Microbial Hub Taxa Link Host and Abiotic Factors to Plant Microbiome Variation

Abstract

Plant-associated microorganisms have been shown to critically affect host physiology and performance, suggesting that evolution and ecology of plants and animals can only be understood in a holobiont (host and its associated organisms) context. Host-associated microbial community structures are affected by abiotic and host factors, and increased attention is given to the role of the microbiome in interactions such as pathogen inhibition. However, little is known about how these factors act on the microbial community, and especially what role microbe-microbe interaction dynamics play. We have begun to address this knowledge gap for phyllosphere microbiomes of plants by simultaneously studying three major groups of Arabidopsis thaliana symbionts (bacteria, fungi and oomycetes) using a systems biology approach. We evaluated multiple potential factors of microbial community control: we sampled various wild A. thaliana populations at different times, performed field plantings with different host genotypes, and implemented successive host colonization experiments under lab conditions where abiotic factors, host genotype, and pathogen colonization was manipulated. Our results indicate that both abiotic factors and host genotype interact to affect plant colonization by all three groups of microbes. Considering microbe-microbe interactions, however, uncovered a network of interkingdom interactions with significant contributions to community structure. As in other scale-free networks, a small number of taxa, which we call microbial "hubs," are strongly interconnected and have a severe effect on communities. By documenting these microbe-microbe interactions, we uncover an important mechanism explaining how abiotic factors and host genotypic signatures control microbial communities. In short, they act directly on "hub" microbes, which, via microbe-microbe interactions, transmit the effects to the microbial community. We analyzed two "hub" microbes (the obligate biotrophic oomycete pathogen Albugo and the basidiomycete yeast fungus Dioszegia) more closely. Albugo had strong effects on epiphytic and endophytic bacterial colonization. Specifically, alpha diversity decreased and beta diversity stabilized in the presence of Albugo infection, whereas they otherwise varied between plants. Dioszegia, on the other hand, provided evidence for direct hub interaction with phyllosphere bacteria. The identification of microbial "hubs" and their importance in phyllosphere microbiome structuring has crucial implications for plant-pathogen and microbe-microbe research and opens new entry points for ecosystem management and future targeted biocontrol. The revelation that effects can cascade through communities via "hub" microbes is important to understand community structure perturbations in parallel fields including human microbiomes and bioprocesses. In particular, parallels to human microbiome "keystone" pathogens and microbes open new avenues of interdisciplinary research that promise to better our understanding of functions of host-associated microbiomes.

Fig 1. Experiment 1 and 2: Ecological and host factors are important in shaping phyllosphere microbial communities.

A. Experiment 1: Sampling location and sampling time correlated to microbial community structure variation observed between Tübingen wild sites. Circles and triangles are samples collected in fall and spring, respectively. Colors of points illustrate the location where the samples were collected. Dot plots are unconstrained endophytic communities, while barcharts show factor correlations to endophytic (endo) and epiphytic (epi) variation. Overlap of bars represents factors correlated to the same variation. B. Experiment 2: The host A. thaliana accession correlated to microbial community structure variation observed in the Cologne garden experiment. Colors of points represent the host accession. For A and B, figures are based on genus-level data from bacterial 16S V3/V4 region, fungal ITS1 region and oomycete ITS1 region amplicons. For A and B, a star indicates that the measured correlation is statistically significant (p < 0.05) based on random permutations of sample classes. (S1_Data.xlsx)

Fig 2. Computational Experiment 3: Inter- and intra-kingdom microbe–microbe interactions affect phyllosphere microbiome structure.

A. A correlation network demonstrates that correlations between microbes within kingdoms tend to be positive (orange solid), while correlations between kingdoms tend to be negative (black dashed). Boldness of lines is related to the strength of the correlation. Correlations were made using samples from both Experiment 1 and Experiment 2. Additional care was taken to ensure correlations were robust (see S1 Text). The network structure was typical of a scale-free network since only a few nodes were highly connected (a power-law fit to the node degree distribution has alpha = −1.072 and r² = 0.846). B. “Hub” microbes were identified as those which were significantly more central based on all three measurements of centrality. For the network shown in A (based on one of several cutoffs for “good” correlations, see S1 Text), three microbes, Albugo sp., Dioszegia sp., and a genus of Comamonadaceae were identified as “hubs” (yellow line: p = 0.1 based on a log-normal distribution fit). Other genus-level hub microbes indicated in Fig 2A were identified by combining the results of several other correlation cutoffs (see S11 Fig and S7 Table). (S1_Data.xlsx)

Fig 3. Computational Experiment 3: Hub microorganisms are critical determinants of the microbiome interaction network structure.

A. Most high-degree bacteria (including the genus of Comamonadaceae designated as a hub) are first neighbors (i.e., direct and negative correlates) of the hub microbial genera Albugo sp. and Dioszegia sp., and many group into an intercorrelated cluster. First neighbors of the three “hub” microbes are shown in color and the rest of the network is shown in greyscale. The depiction is a spring-loaded visualization of the network in Fig 2 where tightly correlated nodes cluster together. B. The hub microbes were partly independent, since about half of the nodes to which they correlated were unique and half were shared. They together directly reach over half (100/191) of all nodes in the network. C. Hub microbes (high degree organisms with high centrality) can be considered as reasonable keystone species, since the magnitude of their effects in the network extend over more edges than nonkeystone nodes (high abundance organisms with low degree and low centrality) but over fewer than keystone nodes (high degree organisms with low centrality). An edge was considered dependent if it was not observed in a network built using partial correlations controlling for abundance of the test microbes. Error bars show standard deviation, and significance was tested with a one-sided Welch’s t test where (*): p < 0.1, (**): p < 0.05 and (***): p < 0.01. Hub nodes: Albugo sp., Dioszegia sp. and a genus of Comamonadaceae. Keystone nodes: Mycobacterium sp., Rhodoplanes sp., and Rhizobiales (other). Nonkeystone nodes: Pseudomonas sp., Oxalobacteriaceae (other), and Sphingomonas sp. (S1_Data.xlsx)

Fig 4. Experiment 4: Species of the obligate biotrophic pathogen and hub genus Albugo can affect colonization of microbes in the phyllosphere, linking abiotic or host genotype factors to a mechanism for observed microbial community variation.

A. When Alb. laibachii Nc14 or Alb. candida Nc2 are absent due to abiotic (physical spore removal) or host (resistance) factors, the pathogen-associated microbial community increases in alpha diversity (also see S12 Fig) is less replicable (A. laibachii only) and shifts significantly (A. laibachii only). B. Several genera of bacteria were observed to more efficiently colonize the endophytic compartment of the phyllsophere in plants infected with Albugo sp. than in controls. For A and B: Ws-0: A. thaliana Ws-0, Col-0: A. thaliana Col-0, Ksk-1: A. thaliana Ksk-1. Green: Susceptible hosts, Red: Resistant hosts, Yellow: Filter removal of Albugo on all hosts. (S1_Data.xlsx)

Fig 5. Experiment 1 and 3: Effects on bacterial colonization of eukaryotic “hub” microbes overlap with effects of location and sampling time.

Each bubble represents the amount of microbial community variation between samples at Tübingen wild sites (Experiment 1) that could be correlated to the factor’s location and sampling time (blue), Albugo abundance (green), and Dioszegia abundance (red) using constrained ordination analysis. About 40% of observed variation in both epiphytic or endophytic bacterial colonization could be attributed to the external factors location and sampling time and about 50% when considering Albugo and Dioszegia in addition (total model). For epiphytes and endophytes, respectively, about 20%–35% and 15%–20% of variation (the overlap percent) linked to location and sampling time could also be correlated to either Albugo sp. or Dioszegia sp. (The “overlap percent” is the “factor overlap” divided by location/sampling time-correlated variation, where “factor overlap” is the percent of total community variation shared by Albugo/Dioszegia and location/sampling time). Black lines show the percent variation correlated to pairs of factors, and stars indicate that the two factors connected by the black line were significantly (p < 0.05) independent of one another.

Fig 6. Direct targeting of a hub by a biotic or abiotic factor results in a cascade of abundance shifts throughout the community.

A. The magnitude of effects of host or abiotic factors on microbial community structures is dependent on the connectivity of microbes targeted by the factor. For example, host “A” directly limits the colonization of an “edge” microbe (with low degree in the community network), and a relatively small shift in beta diversity is observed compared to the inoculating microbiome. On the other hand, when host “B” affects the colonization of a “hub” microbe, a drastic shift of many members of the inoculating microbiome is observed. B. A microbial community without a main “hub” microbe shows high fluctuations in beta diversity upon perturbation by different factors that might act on different “edge” microbes differently and is likely to be more subject to stochastic variation. A microbial community that is structured by a main “hub” microbe shows lower levels of fluctuation in beta diversity but is highly sensitive to perturbations of the “hub” microbe.