Niche partioning and partially host and growth stage explain rhizosphere bacterial microbiota variation
The variation in the bacterial microbiota considering all plant hosts and growth stages was mainly affected by spatial differentiation in the rhizosphere as indicated by 21% of variance explained (R2) by contrasting close and distant rhizosphere (PERMANOVA). Additionally, both rhizosphere compartments were significantly different from the initial bulk soil (Table 3). Crop species and growth stage accounted for 10% and 6% respectively, with all factors being statistically significant (P < 0.001) (Table. 4).
Table 3
Divergence of rhizosphere from initial planting soil bacterial microbiota composition*.
Factor
|
Df
|
SumOfSqs
|
R2
|
F
|
P*
|
Rhizosphere vs bulk soil
|
2
|
0.645
|
0.304
|
23.611
|
< .001
|
Residual
|
108
|
1.476
|
0.706
|
|
|
Close rhizosphere vs. bulk soil
|
55
|
0.172
|
10.204
|
114.11
|
< .001
|
Residual
|
56
|
0.669
|
0.796
|
|
|
distant rhizosphere vs. bulk soil
|
56
|
0.268
|
0.245
|
18.157
|
< .001
|
Residual
|
57
|
0.826
|
0.755
|
|
|
* Permutational analysis of variance and weighted UniFrac distance |
**bold font: factors considered significant |
Table 4
Influence of rhizosphere compartment, crop species, and growth stage on entire and core bacterial microbiota1,2.
Factor
|
Df
|
SumOfSqs
|
R2
|
F
|
P*
|
Full dataset (close & distinct rhizosphere)
|
Growth stage
|
2
|
0.108
|
0.057
|
4.969
|
< .001
|
Crop species
|
3
|
0.188
|
0.100
|
5.838
|
< .001
|
Rhizosphere (close&distant)
|
1
|
0.425
|
0.225
|
39.53
|
< .001
|
Growth stage:crop species
|
6
|
0.156
|
0.083
|
2.421
|
< .001
|
Residual
|
94
|
1.01
|
0.535
|
|
|
Core bacterial microbiota distant rhizosphere
|
Growth stage
|
2
|
0.081
|
0.020
|
8.398
|
< .001
|
Crop species
|
3
|
3.694
|
0.918
|
253.680
|
< .001
|
Growth stage:crop species
|
6
|
0.031
|
0.008
|
1.076
|
0.394
|
Residual
|
0.218
|
0.054
|
|
|
|
Core bacterial microbiota close rhizosphere
|
Growth stage
|
2
|
0.119
|
0.030
|
8.842
|
< .001
|
Crop species
|
3
|
3.384
|
0.851
|
166.968
|
< .001
|
Growth stage:crop species
|
6
|
0.195
|
0.049
|
4.807
|
< .001
|
Residual
|
41
|
0.277
|
0.070
|
|
|
Bacterial microbiota distant rhizosphere
|
Growth stage
|
3
|
0.380
|
0.347
|
14.685
|
< .001
|
Crop species
|
3
|
0.199
|
0.182
|
7.698
|
< .001
|
Growth stage:crop species
|
6
|
0.127
|
0.116
|
2.454
|
< .002
|
Residual
|
45
|
0.388
|
0.355
|
|
|
Bacterial microbiota close rhizosphere
|
Growth stage
|
2
|
0.075
|
0.115
|
4.209
|
< .001
|
Crop species
|
3
|
0.084
|
0.129
|
3.150
|
< .001
|
Growth stage:crop species
|
6
|
0.126
|
0.195
|
2.370
|
< .001
|
Residual
|
41
|
0.365
|
|
|
|
1based on permutational analysis of variance and weighted UniFrac distance. |
2Core bacterial microbiota of wheat, barley, rye and oilseed rape includes ASVs present within each replicate per crop species and at every growth stage (booting, flowering, ripening) |
*bold font: factors considered significant. |
Looking at the individual compartment, we observed a stronger growth stage effect in the distant (R2 = 34%) than in close rhizosphere (R2 = 11%). Crop species captured a significant proportion of microbiome variation, which was around 18% and 12 % in the distant and close rhizosphere, respectively. We found a significant interaction between plant growth stage and crop species, which explained an additional 12% of variation in both rhizosphere compartments. PCoA on weighted UniFrac distance resembled the separation of the close from the distant rhizosphere as primary effect (Fig. 1; A-B). Bacterial microbiota assembly of cereal species and oilseed rape diverged at booting, but tended to associate at mid and late growth stages. Contrastingly, barley and oilseed rape were temporally invariant within the distant rhizosphere and formed separate clusters. Rye and wheat associated with barley at booting and diverged between flowering and late growth stages. Hence, the distance to oilseed rape decreased implying a temporal shift in microbiome structure.
The core rhizosphere bacterial microbiota encompassed growth stage-invariant Bacteria enriched within individual crop species
Host-species effect signatures within the rhizosphere bacterial microbiota were inferred from eight core bacterial microbiota for each host species separated into close and distant rhizosphere (Additional file 1). The core bacterial microbiota gathered from joint datasets of all four crop species represented more than half of summed ASV counts (RZP 58%, RZS 50%), but were limited to a low fraction of ASVs (8%) (Table 4).
Table 4
Number and relative abundance of ASVs included in the host crop core microbiota.
rhizosphere compartment
|
wheat
|
barley
|
rye
|
oilseed rape
|
agglomerated
|
distant
|
ASVs
|
283
|
345
|
154
|
122
|
453
|
ASVs (% of total)*
|
4.6
|
5.6
|
2.5
|
2
|
8.1
|
ASV count (% of total)*
|
11.6
|
12.8
|
8.3
|
8.4
|
50.2
|
close
|
ASVs
|
446
|
296
|
203
|
137
|
512
|
ASVs (% of total)
|
7.0
|
4.8
|
3.3
|
2.2
|
8.3
|
sequence count (% of total)
|
21.0
|
13.0
|
9.2
|
15.0
|
58.0
|
*proportion of ASVs and ASV counts in relation to the unfiltered microbiota |
This implies, that a low number of core Bacteria were the dominant microbiota associated with the four crops studied. Moreover, PCoA analysis demonstrated an uniform clustering by crop species within the core bacterial microbiota independent of growth stage (Fig. 1; C-D), with the host plant effect explaining almost the entire variation (PERMANOVA, R2 = 92 RZS, R2 = 82 RZP). Thus, the core bacterial microbiota derived in our study encompassed plant host-specific Bacteria. These originate from a typical arable soil and can be considered as non-transient host traits, since they persisted over growth stages.
For taxonomic description, the core bacterial microbiota were aggregated at the genus level. Generally, more than two genera unique to each crop species (ratio RZS:RZP) occurred in wheat (15:26) and barley (12:5), i.e. with a total abundance below 1.5% of total sequences count within the core microbiota (Fig. 2). Thus, unique genera were scarce and might belong to the rare species pool. Contrastingly, the 52 (distant rhizosphere) and 44 (close rhizosphere) genera that occurred in all core bacterial microbiota were highly abundant (about 80% of total sequences count), meaning that this comparable small subset were dominant traits of the core microbiota. Additionally, more than ten genera were shared between the three cereals or between barley, wheat an oilseed rape thereby contributing to more than 5% of the total sequences counts within the aggregated core microbiota. As a consequence, enriched or depleted genera were the most abundant core microbiota members for each of the four crop species, while unique bacterial genera were scarce.
The aggregated core microbiota were predominantly composed of the genera of the classes Alpha- and Gammaproteobacteria, Actinobacteria, Thermoleophilia, Bacilli, Verrucomicrobia, unclassified Chloroflexi of the group ‘KD4-96’, which together accounted for more than 50% of relative abundance. Proteobacteria were higher abundant within the close rhizosphere, whereas genera of the Thermoleophilia and ‘TK10’ were lower abundant than in distant rhizosphere. Genera of the class Bacilli were substantially enriched (25%) in oilseed rape close rhizosphere. Exclusive bacterial genera of barley belonged mainly to the Alpha- and Gammaproteobacteria (close rhizosphere) or Actinobacteria, Acidobacteria, Planctomycetes (distant rhizosphere). In contrast to barley, genera of the Bacilli and Polyangia were uniquely present in the distant rhizosphere of wheat.
Genera exclusively shared between barley, wheat, and oilseed rape constituted belonged mainly to the Alpha- and Gammaproteobacteria and were characterized by the presence of Rubrobacteria and Clostridia. Remarkably, genera exclusively found in cereals were mainly Gammaproteobacteria but no Alphaproteobacteria.
The primary result of both core inference and venn diagram partitions was that shared genera between all crop species incorporate most of the of the total sequences counts of the aggregated core microbiota and comprise also the host-distinctive bacterial classes. Triplicate comparisons of oilseed rape, wheat and barley or of the three cereals revealed shared taxa with elevated abundance, that additionally promote differences of bacterial rhizosphere bacterial microbiota structures.
LEfSe biomarker analysis
Most genera were present across core bacterial microbiota of all crop species. Hence, we identified biomarker taxa from genus to phylum rank among them using LEfSe analysis. These biomarker were the dominant bacterial phylotypes associated with a specific crop species and were the main taxa that explained differences between the core bacterial microbiota of the four crops species (Fig. 3 and Table. 5).
Table 5
Global co-occurrence network metrics for each core microbiota of the crop species and rhizosphere compartment.
Niche
|
Crop species
|
Nodes
|
Clustering coefficient
|
Modularity
|
Positive edges (%)
|
Average degree
|
Average betweenness
|
Natural connectivity
|
close
|
Barley
|
107
|
0.083
|
0.563
|
50
|
0.035
|
0.028
|
0.011
|
Wheat
|
139
|
0.098
|
0.483
|
56
|
0.036
|
0.017
|
0.009
|
Rye
|
77
|
0.119
|
0.608
|
52
|
0.036
|
0.048
|
0.015
|
Oilseed rape
|
71
|
0.156
|
0.604
|
51
|
0.041
|
0.043
|
0.016
|
distant
|
Barley
|
99
|
0.091
|
0.512
|
52
|
0.040
|
0.027
|
0.012
|
Wheat
|
102
|
0.087
|
0.530
|
55
|
0.039
|
0.028
|
0.012
|
Rye
|
61
|
0.133
|
0.645
|
51
|
0.040
|
0.054
|
0.018
|
Oilseed rape
|
58
|
0.089
|
0.677
|
47
|
0.038
|
0.078
|
0.019
|
Additionally, we determined bacterial genera characteristic of spatial differentiation between the close and distant rhizosphere based on the entire dataset irrespective of the crop species (Supplementary Fig. 2).
One third of the biomarker taxa occurred in both rhizosphere compartments but were indicative of two different crop species (Fig. 3C1). Exemplarily, several lower taxonomic groups of the Gammaproteobacteria were indicative of oilseed rape within the distant rhizosphere, whereas these were a biomarker taxon of rye or barley in the close rhizosphere. The families Blastocatellia and Dokdonella were biomarker taxa of rye in the close and belonged to oilseed rape in the distant rhizosphere. Accordingly, several biomarker that were assigned to oilseed rape in the distant rhizosphere, were assigned to wheat within the close rhizosphere, e.g. Paenbacillus or Bradyrhizobium Additionally, several biomarker taxa of a crop species bacterial microbiome identified in the close rhizosphere became non-discriminative in the distant rhizosphere. This was most apparent for biomarker taxa belonging to the order Rhizobiales. Interestingly, Bacillus and Streptomyces were biomarker of oilseed rape in botch compartments. Further, the bacterial microbiota of barley tended to show identical biomarkers in both rhizosphere compartments, such as the family Gemmataceae or the genus Candidatus Alysiosphaera
Biomarker of rye belonged uniformly to the close rhizosphere, e.g. Optitutaceae. The majority of biomarker taxa exclusively derived from the close rhizosphere belonged to the bacterial microbiota of oilseed rape such as Sphingomonas and Pseudnarcodia (Fig. 3B;C3). Verrucomicrobiales were a unique biomarker of wheat within the close rhizosphere. Biomarker exclusively detected in the distant rhizosphere that belonged to barley were among other the order Frankiales and Nocardioides.
Thus, the rhizosphere compartment essentially determined biomarker taxa assignment. Thereby, we observed ambiguous assignments of biomarker taxa to different crop species within different rhizosphere compartments. This required further investigations of the identified individual biomarker genera and the underlying differences between hosts core microbiota, they were characteristic of. Thus, we examined their impact on microbiome structure and associations, i.e. their co-occurrence patterns.
Network hubs of the crop core bacterial microbiota
We performed co-occurrence network analyses for each core bacterial microbiota visualized as chord diagrams to identify structurally important interactions within the individual crop bacterial microbiota in separated data sets of the close and the distant rhizosphere (Supplementary Fig. 3; Supplementary Fig. 4)
The size of the largest connected component was 1.5-fold larger in wheat and barley, compared to rye and oilseed rape (Table. 5). Higher average degree, betweenness and natural connectivity were observed in all networks of (a) the close compared to the distant rhizosphere or (b) rye and oilseed rape compared to barley and wheat (Table. 5). Accordingly, the network structure of wheat and barley was more dependent on individual nodes compared to the other host plant species.
More than ten hubs where identified for wheat and barley, which interconnected almost the entire network (Fig. 4; Fig. 5), while rye and oilseed rape comprised less than four hubs that were directly associated with less than 16 cohort nodes (Fig. 4; Fig. 5). The networks of wheat, rye and oilseed rape within the distant rhizosphere comprised less than three hubs, which directly connected only a small subset of cohort nodes. In contrast, the seven hubs of barley were similarly complex interconnected than observed within the close rhizosphere. As a result, the co-occurrence networks of (a) rye and oilseed rape as well as of (b) wheat and barley are structurally more similar to each other, respectively.
Hubs shared between at least two cereal crop species in the close rhizosphere were Massilia, Rubrobacter, and Phycicococcus. Thereby, more than three hubs of barley and wheat belonged to the class Alphaproteobacteria. Hubs of wheat, barley and to a lower extend of rye associated with similar nodes. Individual hubs of oilseed rape were unclassified Chloroflexi group ‘KD496’, Phenylobacterium, and Rubrobacter. Exclusive hubs of wheat were among others Steroidobacter, Glycomyces, Vampirovibrionales, Clostridium ss.13 and Solirubrobacter. The only hubs exclusively found in rye were Kitasatospora and unclassified Pyrinomonadaceae ‘RB41’. Remarkably, Nitrospira was an exclusive hub of barley, which was not associated with cohort nodes in the other crop species (Fig. 4).
Differentially associated nodes of the bacterial rhizosphere core microbiota
We examined pairwise significant differences in associations of shared genera between each crop core bacterial microbiota to demonstrate that the core taxa across crop species distinctively interfere with bacterial microbiota structure and assembly. In most comparisons (Fig. 6; Fig. 7), half of the differentially associated nodes belonged to the Alphaproteobacteria and Actinobacteria. The most frequent and differentially associated genera (Table 6) were unclassified Tepidisphaeraceae ‘WD2101 soil group’, Hyphomicrobium, Terrabacter, uncultured Beijerinckiaceae ‘alphaI cluster’, Nocardioides, Massilia, and Bradyrhizobium.
Table 6
Most frequent* significant differential associated nodes within pairwise comparisons of the crop species** core bacterial microbiota co-association networks
|
Number of differential associations
|
rhizosphere
|
close
|
distant
|
WD2101_soil_group
|
32
|
26
|
Hyphomicrobium
|
31
|
23
|
Terrabacter
|
28
|
21
|
alphaI_cluster
|
27
|
19
|
Skermanella
|
NA
|
17
|
Nocardioides
|
25
|
NA
|
Nitrospira
|
21
|
NA
|
Massilia
|
21
|
17
|
TK10
|
NA
|
15
|
Bradyrhizobium
|
21
|
14
|
Devosia
|
NA
|
14
|
*(0.9 percentile) |
**barley, wheat, rye and oilseed rape |
While Bradyrhizobium was identified as a hub of oilseed rape, the association of Bradyrhizobium and Hyphomicrobium was specific to the three cereals. Another specific feature of cereals was the association of Nocardioides and uncultured Beijerinckiaceae ‘alphaI cluster’ (only wheat and barley) with unclassified Tepidisphaeraceae of the group‘WD2101 soil group’, while in oilseed rape Nocardioides formed associations with other genera of the Actinobacteria, such as Streptomyces. Differential associations contrasting the individual cereal crop species were formed by Cellulomonas and Skermanella as well as unclassified Pyrinomonadaceae ‘RB41’ for rye. Significant different association specific to wheat and barley were guided by Massilia, Sphingomonas, Bosea, Devosia and Candidatus Xiphinematobacter as well as Optitutus in rye. Thus, a limited set of genera promoted different microbiome structures between the crop species. These shared genera were indicative of differences between cereal crops and oilseed rape and between individual crop species. Moreover, cereal hosts tended to associate with similar nodes compared to oilseed rape.
As a result, co-occurrence network analyses explained differences in rhizosphere microbiome assembly between plant families as well as individual crop species. Thus, cohort subnetworks and differential associations elucidated a host species effect that was not evident from differential abundance of specific bacterial genera enriched or unique to a specific plant host.