Xiong et al.

Microbiome (2020) 8:27

https://doi.org/10.1186/s40168-020-00799-9

RESEARCH Open Access

Rhizosphere protists are key determinants

of plant health
Wu Xiong1,2†, Yuqi Song1†, Keming Yang1, Yian Gu1, Zhong Wei1*, George A. Kowalchuk2, Yangchun Xu1,
Alexandre Jousset1,2, Qirong Shen1 and Stefan Geisen1,3,4

Abstract
Background: Plant health is intimately influenced by the rhizosphere microbiome, a complex assembly of
organisms that changes markedly across plant growth. However, most rhizosphere microbiome research has
focused on fractions of this microbiome, particularly bacteria and fungi. It remains unknown how other microbial
components, especially key microbiome predators—protists—are linked to plant health. Here, we investigated the
holistic rhizosphere microbiome including bacteria, microbial eukaryotes (fungi and protists), as well as functional
microbial metabolism genes. We investigated these communities and functional genes throughout the growth of
tomato plants that either developed disease symptoms or remained healthy under field conditions.
Results: We found that pathogen dynamics across plant growth is best predicted by protists. More specifically,
communities of microbial-feeding phagotrophic protists differed between later healthy and diseased plants at plant
establishment. The relative abundance of these phagotrophs negatively correlated with pathogen abundance
across plant growth, suggesting that predator-prey interactions influence pathogen performance. Furthermore,
phagotrophic protists likely shifted bacterial functioning by enhancing pathogen-suppressing secondary metabolite
genes involved in mitigating pathogen success.
Conclusions: We illustrate the importance of protists as top-down controllers of microbiome functioning linked to
plant health. We propose that a holistic microbiome perspective, including bacteria and protists, provides the
optimal next step in predicting plant performance.
Keywords: Rhizosphere, Pathogen of Ralstonia solanacearum, Protists, Predator-prey interactions, Secondary
metabolite genes, Plant health

Background assembly of diverse microorganisms, including bacteria,

Plant pathogens can colonize the rhizosphere and have a fungi, and protists that together influence plant health
severe influence on plant health [1, 2]. However, patho- [5–8]. Despite the fact that the microbiome consists of
gen success and plant health are ultimately controlled by diverse groups, most research aiming to understand the
other biota, particularly the rhizosphere microbiome [3, role of the microbiome in plant health or disease sup-
4]. The plant rhizosphere microbiome is a complex pression has focused on bacteria [9–11] and fungi [12,
13]. A whole-microbiome view to decipher the main mi-
* Correspondence: [email protected] crobial determinants and their potential interactions that
†
Wu Xiong and Yuqi Song contributed equally to this work.
1
Jiangsu Provincial Key Lab of Solid Organic Waste Utilization, Key Lab of
determine plant performance is currently missing [14].
Plant Immunity, Jiangsu Collaborative Innovation Center of Solid Organic As such, a more complete microbiome analysis is needed
Wastes, Educational Ministry Engineering Center of Resource-saving fertilizers, to identify the microbial groups and potential interac-
Nanjing Agricultural University, Nanjing 210095, Jiangsu, People’s Republic of
China
tions that help predicting plant health.
Full list of author information is available at the end of the article

© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Xiong et al. Microbiome (2020) 8:27 Page 2 of 9

In particular, protists that steer the taxonomic and plant health, as for instance, protist communities were
functional composition of the rhizosphere microbiome shown to respond more strongly to environmental inputs
through trophic predator-prey interactions have so far and vary more in their composition between seasons than
rarely been included in microbiome analyses linked to bacteria and fungi [27].
plant performance [8]. Protists, especially microbial- To investigate potential key microbiome groups that
feeding phagotrophs [15, 16], have various functions might predict plant health, we here used a rhizobox sys-
within the rhizosphere [6, 17, 18]. For instance, some of tem in an agricultural system under field conditions, in
these phagotrophs can directly prey on plant pathogens which we grew tomato plants. Soils were naturally infested
[19]. Studies using model protists have shown that pro- with pathogenic Ralstonia solanacearum bacteria, one of
tists control microbiome diversity and structure leading the most devastating and globally distributed soil-borne
to plant growth promotion [17, 18, 20]. These changes plant pathogens that can infect a range of important crops
are at least partly explained by the fact that protists feed [28, 29]. In the rhizosphere of plants that either later de-
selectively on microbial prey taxa, which differs between veloped disease symptoms or remained healthy, we tem-
protistan species [21, 22]. Through this selective preda- porally investigated the microbiome composition,
tion, protists can, for instance, increase those bacteria including bacteria, fungi, and protists, as well as potential
that produce pathogen-suppressive secondary metabo- microbial functions using metagenomics. We tested the
lites [23, 24]. Yet, all these studies investigating potential hypothesis that protists rather than other microbial com-
links of protists with plant performance were carried out munities in the rhizosphere microbiome best predict
under artificial laboratory or greenhouse conditions fo- pathogen dynamics and plant health.
cusing on one or few protistan species. As such, we have
yet to identify the links between a complex diversity of Results and discussion
protists, the microbiome and plant performance, espe- Here, we show that the community structure of protists
cially in agricultural systems under field conditions. could best predict the density of the R. solanacearum patho-
Protists and their interactions with other microorganisms gen across plant growth in healthy and diseased datasets
are also subject to change throughout plant growth [14, (Fig. 1a). In healthy plants, the diversity and community
25]. Yet, the composition of the microbiome is often inves- structure of bacteria could significantly predict pathogen
tigated only once during plant growth, usually at the time density (Fig. 1b), which is in line with previous findings that
of plant maturity or after disease has already developed. soil bacterial composition can predetermine future plant
Such approaches make it difficult to disentangle causality health [26]. In diseased plants, the community structure of
between plant health and potentially underlying character- protists was the best predictor for pathogen density (Fig.
istics in microbial communities, especially for diseased 1c). At plant establishment, the community structure of bac-
plants that host high amounts of pathogens. Recently, it teria differed (ANOSIM, P < 0.001; Table S2) between
was shown that bacterial communities at plant establish- healthy and diseased plants as shown before [26] but not
ment can predict plant health at maturity [26]. Yet, other that of fungi and protists (ANOSIM, P > 0.05; Table S2).
microbial groups might be even better indicators to predict However, we found that the community structure of

Fig. 1 The relative importance of the main microbial parameters in predicting pathogenic Ralstonia solanacearum density across plant growth
with the combined datasets including healthy and diseased plants (a), the healthy plant dataset (b), and the diseased plant dataset (c). Diversity
(Shannon index) and structure (PCoA2) of bacterial, fungal, and protistan communities were selected as the six main microbial predictors (Fig. S2).
Asterisk means P < 0.05, two asterisks mean P < 0.01, and three asterisks mean P < 0.001 (statistical significance was calculated by multiple
regression using linear models between the microbial predictors and R. solanacearum)
Xiong et al. Microbiome (2020) 8:27 Page 3 of 9

phagotrophic protists at plant establishment was indicative than in diseased plants at plant establishment (Fig. 2d and
for later plant health, as indicated by the differences (ANO- Table S4). Particularly phagotrophs (a taxon within Trine-
SIM, P = 0.013) observed between plants developing disease matidae, Flectomonas ekelundi, Proleptomonas faecicola,
and those remaining healthy (Fig. 2a, b). The community and two Eocercomonas spp., all mostly bacterivorous Cerco-
composition of other protistan functional and taxonomic zoa) but also a phototrophic Chloroidium saccharophila
groups did not differ between later healthy or diseased were negatively linked with the pathogen at plant establish-
plants at plant establishment (Fig. 2a). Indicator analysis re- ment (Fig. 2d). Although those protistan OTUs were also
vealed 13 protistan OTUs in healthy plants at plant estab- present in diseased plants, they did not correlate with the
lishment (with only 3 in diseased plants) that indicate later pathogen in the network analysis (Table S4). Thus, we con-
plant health (Fig. 2c and Table S3). Seven protistan OTUs clude that phagotrophic protists in general as well as specific
indicative for healthy plants were identified as phagotrophs, taxa at plant establishment can predict pathogen density
including one amoebozoan and six cercozoan taxa, that and plant health at plant maturation, as supported by the
likely prey entirely or as part of their diet on bacteria [30]. community structure of phagotrophs, phagotrophic indica-
Of these, the protistan Pro_OTU8 (Cercozoa, Trinematidae) tor taxa, and negative links between phagotrophic protistan
was the most abundant at plant establishment (Table S3) OTUs and the pathogen in co-occurrence networks. This
and across plant growth accounting for around 11% of all supports the perspective that functional units rather than
protistan reads (Table S5). This taxon likely represents an taxonomic units underlie microbial functioning and as such
omnivorous protist that mostly feeds on bacteria [30]. Co- should be considered as better indicators [31–33], even
occurrence network analysis revealed more negative links across different trophic levels in the microbiome. In
between R. solanacearum and protistan OTUs in healthy addition, we found that the relative abundance of total

Fig. 2 Community structure of protistan taxonomic and functional groups explaining differences between diseased and healthy plants at plant
establishment (week 0) (a). Community structure of phagotrophic protists (b) and indicator protistan OTUs (c) in diseased and healthy plants at
plant establishment, and networks of the functional groups of protistan OTUs directly associated with the R. solanacearum pathogen in healthy
and diseased plants at plant establishment (d). Correlations between the relative abundance of phagotrophic protists and R. solanacearum in
diseased and healthy plants across plant growth (e). In panel a, only abundant taxonomic and functional groups of protists were selected
(average relative abundance over 1%). In panel a and b, asterisk means P < 0.05. In panel c, protistan OTUs with LDA score > 2.0 are indicators
for healthy plants, while protistan OTUs with LDA score < − 2.0 are belonging to diseased plants. In panel d, blue lines indicate positive, and red
lines indicate negative correlations. In panel e, the solid line shows a significant (P < 0.05) correlation, and the dashed line shows a non-
significant (P > 0.05) correlation
Xiong et al. Microbiome (2020) 8:27 Page 4 of 9

phagotrophs correlated negatively (regression analysis, P < functional genes (2 links) in healthy plants than in dis-
0.05) with the abundance of R. solanacearum in diseased eased plants (0 links) (Fig. 3f). Especially, Pro_OTU105
plants or in healthy and diseased combined datasets across (Cercozoa; Eocercomonas sp.), which also negatively cor-
plant growth (Fig. 2e). Interestingly, the relative abundance related with the pathogen at plant establishment, showed
of total phagotrophs significantly decreased (regression ana- negative (Spearman’s correlation coefficient (ρ) < − 0.8
lysis, P < 0.05) with plant growth time in diseased plants with P < 0.01) correlations with six bacterial OTUs
(Fig. S4). Phagotrophic protists may control pathogen devel- across plant growth. Among those was one bacterial
opment throughout plant growth, as a decreased relative OTU (Bac_OTU17: Bacteroidetes; Terrimonas) that
abundance of phagotrophs in diseased plants coincided with positively linked with non-ribosomal peptide synthetase
pathogen outbreak. Although the pathogen was present in gene (COG1020), one of the key genes involved in the
healthy plants, a stable relative abundance of phagotrophic suppression of R. solanacearum pathogen [26, 36] (Fig.
protists throughout plant development might have helped 3f and Table S5). Future targeted experiments using iso-
to keep the pathogen in check. Together, these findings sug- lated phagotrophic protists and bacterial strains are
gest that direct trophic interactions between phagotrophic needed to evaluate such a role. Healthy plants showed
protists and the pathogen at plant establishment and higher numbers of phagotrophic protistan OTUs, bacter-
through plant growth steer later plant performance. In con- ial OTUs, and metabolism [Q] genes, resulting in a more
trast, R. solanacearum in diseased plants at plant establish- complex network (55 nodes with 90 links) than diseased
ment was positively linked with two oomycete species plants (41 nodes with 59 links) (Fig. 3f and Table S6).
(OTU), including one likely plant-pathogenic Pythium spe- Specific linkages within co-occurrence networks only
cies (Fig. 2d). This suggests that a pathobiome forms in dis- provide information about potential interactions, but
eased plants [34, 35], here consisting of a simultaneous further mechanistic proof for the interaction needs spe-
infection with different pathogens. While, a dominance of cific co-culture experiments. In addition to individual
predator-prey interactions might mitigate negative pathogen links, network structure and composition can provide in-
effects and thereby stimulating plant health. sights about system’s stability and increased potential for
We also found that protists might determine pathogen providing ecosystem services [37–40], suggesting that
development and plant health through functional healthy plants benefit from the presence of a more com-
changes in the bacterial microbiome. Healthy plants plex network, among them higher numbers of phago-
showed significantly (student’s t test, P < 0.05) higher trophs (higher-trophic level organisms in general).
relative abundances of metabolism genes related with Our findings bridge evidence from laboratory or green-
carbohydrate and coenzyme functions at plant establish- house studies focusing on single protist model species [18,
ment (Fig. S5). Strikingly, most metabolism genes had 20, 41, 42] to the community level in agricultural systems
significantly (student’s t test, P < 0.05) higher relative under field conditions, showing that protists affect bacter-
abundances in healthy than in diseased plants at week 5 ial communities and their functioning through predation,
(Fig. S5). Among the eight metabolism gene categories, leading to changes in plant performance. Compared with
secondary metabolite biosynthesis [Q] genes were most diseased plants, healthy plants hosted higher relative
strongly linked (lineal model, P < 0.001) with R. solana- abundances of phagotrophic protists, potentially plant-
cearum density (Fig. 3a). Furthermore, the relative abun- beneficial bacteria, and secondary metabolite genes likely
dance of metabolism [Q] genes increased over time in implicated in pathogen suppression 5 weeks after plant es-
healthy plants, showing significantly (student’s t test, P < tablishment—the time point when pathogen symptoms
0.05) higher relative abundance in healthy than in dis- first developed in diseased plants (Fig. S1). Moreover, pha-
eased plants at week 5 (Fig. 3b). Metabolism [Q] genes gotrophic protists negatively correlated with bacteria that
did not differ between healthy and diseased plants at positively linked with a pathogen-suppressing gene coding
weeks 0, 3, and 4 (Fig. S5). Heathy plants with a higher for non-ribosomal peptides across plant growth. This find-
(student’s t test, P < 0.05) relative abundance of phago- ing might also contribute to pathogen suppression. How-
trophic protists (Fig. 3c) had a higher (student’s t test, P ever, the interaction between plants and the rhizosphere
< 0.05) relative abundance of metabolism [Q] genes (Fig. microbiome is a complex and dynamic process [43]. Fu-
3b), a higher (student’s t test, P < 0.05) relative abun- ture experiments are needed to further examine how
dance of Bacillus OTUs (Fig. 3d), and a lower (student’s plants affect bacterial, fungal, and protistan communi-
t test, P < 0.01) level of pathogen density than diseased ties and their interactions and how those changes in
plants at week 5 (Fig. 3e and Fig. S1). In addition, co- the soil microbiome in turn affects plant performance.
occurrence networks encompassing phagotrophic pro- Together, we propose that predation-induced shifts in
tistan OTUs, bacterial OTUs, and metabolism [Q] genes microbiome composition and functioning are likely in-
across plant growth showed that phagotrophs had more volved in controlling pathogen development and there-
correlations with bacteria (9 links with 7 negative) and fore plant health.
Xiong et al. Microbiome (2020) 8:27 Page 5 of 9

Fig. 3 Relative importance of the eight metabolism gene categories in predicting R. solanacearum density across plant growth in the combined
datasets including healthy and diseased plants (a). Changes in relative abundance of metabolism Q genes (secondary metabolite biosynthesis,
transport, and catabolism genes) in diseased and healthy plants at week 0 and week 5 (b). Relative abundance of phagotrophic protists in
diseased and healthy plants at week 0 and week 5 (c). Relative abundance of Bacillus OTUs in diseased and healthy plants at week 0 and week 5
(d). Abundance of R. solanacearum in diseased and healthy plants at week 0 and week 5 (e). Co-occurrence networks between abundant
phagotrophic protistan OTUs, bacterial OTUs, and metabolism Q genes for healthy and diseased plants across plant growth (f). In panel a, asterisk
means P < 0.05 and three asterisks mean P < 0.001 (statistical significance was calculated by multiple regression using linear models between
metabolism genes and R. solanacearum pathogen). In panel b, c, d, and e, “ns” means non-significant, asterisk means P < 0.05 and two asterisks
mean P < 0.01 under student’s t test (n = 4 for metabolism Q genes, n = 8 for phagotrophic protists, Bacillus and R. solanacearum). In panel d,
relative abundance of Bacillus OTUs combines the two Bacillus OTUs from the bacterial OTU table. In panel f, blue lines indicate positive, and red
lines indicate negative correlations; detailed annotation of bacterial OTUs and metabolism Q genes are provided in Table S5

Conclusions systems to predict plant performance based on protistan

Using a holistic microbiome investigation of bacteria, communities before a crop plant is grown. Furthermore,
fungi, and protists in the rhizosphere across plant our findings suggest a potential for targeted microbiome
growth, we show that in addition to bacteria, protists engineering to promote plant performance through the
serve as key indicators that predict plant health. Particu- application of key microbiome predators: protists. This
larly, the community composition of phagotrophic pro- would bring us closer to the holy grail in reaching a
tists during plant establishment can predict later plant more sustainable, pesticide-reduced agriculture.
performance in the presence of pathogens. These pro-
tists might indeed protect plants by directly feeding on Methods
the pathogen and through predation-induced shifts in Experiment description and soil samples collection
the taxonomic and functional composition of bacteria. We used a semi-open mesocosm system (rhizobox) as de-
These results hold promise in creating tailor-made scribed previously [26], which allowed repeated collection
Xiong et al. Microbiome (2020) 8:27 Page 6 of 9

of rhizosphere soil from each individual plant without Bioinformatic analyses of bacteria, fungi, and protist
damaging the root system under field condition. Briefly, communities
each individual tomato plant was grown in a rhizobox 16S rRNA gene sequence data was processed with the
filled with the original local soil. Triplicate soil samples UPARSE pipeline as described previously [26]. After re-
were collected from the inner and outer sides of the mid- moving the reads assigned as chloroplast, mitochondria,
dle sampling layer, which were thoroughly homogenized and unknown taxa, we obtained 9108 prokaryotic OTUs
and pooled. These soil samples were regarded as the initial (9051 bacteria OTUs and 57 archaea OTUs). We further
bulk soil samples (week 0). In order to track the imprint removed archaeal OTUs (accounting for less than 0.05%
of the tomato rhizosphere, three nylon bags from each of total prokaryotic reads) to generate a bacterial OTU
rhizobox were collected 3, 4, 5, and 6 weeks after trans- table. We selected 8 replicates from the 12 replicates for
plantation. We focused our analyses on the two cat- bacterial community profiles which matched the 80
egories that clearly differed between plants, i.e., plants eukaryotic datasets (2 symptoms × 5 time points × 8
with wilt symptoms and detectable pathogen (R. sola- replicates). Each sample from the bacterial OTU table
nacearum) levels and no wilt symptoms with no detect- was rarefied to 26,014 reads resulting in 8656 bacterial
able pathogen levels. Ten other plants that did not OTUs. We extracted bacterial OTUs from Bacillus and
show wilt symptoms, but did have detectable pathogen Pseudomonas (Fig. 3d and Fig. S6), both well-known po-
levels at a later stage (latently infected plants), were not tentially biocontrol agents against various soil-borne
included in further analyses [26]. Soils from the three pathogens including R. solanacearum [46–48].
nylon bags (4 g soil per bag) for each tomato plant at Eukaryotic sequences were processed according to previ-
each time point were separately homogenized with ster- ously established protocols [49, 50] with some modifications.
ilized forceps and stored at – 80 °C for further use. Soil In short, sequences with expected errors > 1.0 or a length
DNA was extracted from 0.5 g soil using the MoBioPo- shorter than 350 bp were removed. After discarding single-
werSoil™ DNA Isolation Kit (Mo Bio Laboratories Inc., tons, the remaining reads were assigned to operational taxo-
Carlsbad, CA, USA) according to the manufacturer’s in- nomic units (OTUs) with a 97% similarity threshold,
structions. We used the DNA samples to determine followed by a removal of chimeras using UCHIME [51]. Fi-
rhizosphere bacterial and eukaryotic communities as nally, eukaryotic OTUs were matched against the PR2 data-
well as functional genes in both healthy and diseased base [52]. In order to obtain the protistan OTU table, we
tomato plants across plant growth. removed sequences belonging to Rhodophyta, Streptophyta,
Metazoa, and Fungi, resulting in 1,475,483 reads for the 80
samples (average 18,444 reads per sample). In order to ob-
Illumina MiSeq sequencing of the 16S rRNA gene and the tain an equivalent sequencing depth for later analyses, all
18S rRNA gene samples were rarefied to 4537 sequences in 1926 protistan
The V4 region of the 16S rRNA gene was PCR-amplified OTUs. We further assigned the protistan OTUs into differ-
to investigate bacterial communities using the primer set ent functional groups according to their nutrient-uptake
563F and 802R [44] as described previously [26]. In mode based on literature [49, 50], including parasites, pha-
addition, we selected 80 DNA samples (2 symptoms × 5 gotrophs, phototrophs, plant pathogens, and saprotrophs
time points × 8 replicates) for eukaryotic community (Table S1). From the eukaryotic OTU table, we extracted
profiling. For that, the V4 region of the 18S rRNA gene OTUs assigned as fungi resulting in 525,927 reads for the
was broadly targeted to investigate eukaryotes using the 80 samples (average 6574 reads per sample). Each sample
primer set V4_1f (CCAGCASCYGCGGTAATWCC) from the fungal OTU table was rarefied to 1085 reads in
and TAReukREV3 (ACTTTCGTTCTTGATYRA) [45]. 234 fungal OTUs.
PCR was performed in a 20 μl volume consisting of 4 μl
of 5× reaction buffer, 2 μl dNTPs (2.5 mM), 0.8 μl of Functional genes from meta-genomic sequencing
each primer (10 uM), FastPfu Polymerase 0.4 μl, 10 ng We had 12 replicates (each time point) for both diseased
of DNA template, and the rest being ddH2O. Amplifica- and healthy plants. We selected 4 of those replicates (40
tion was performed with the following temperature re- samples in total: 2 symptoms × 5 time points × 4 repli-
gime: 5 min of initial denaturation at 95 °C, followed by cates) for metagenome analyses. Meta-genomic analysis
30 cycles of denaturation (95 °C for 30 s), annealing (55 and functional annotation were performed previously [26].
°C for 30 s), extension (72 °C for 45 s), and a final exten- In short, all reads were trimmed by the Sickle software
sion at 72 °C for 10 min. PCR products were pooled in that removing reads quality below 20 and shorter than 50
equimolar concentrations of 10 ng μl− 1. Paired-end se- bp. Filtered reads were assembled with SOAPdenovo. As-
quencing was performed on an Illumina MiSeq sequen- sembled contigs were then predicted using MetaGene [53]
cer at Shanghai Biozeron Biological Technology Co. Ltd and clustered with a 0.95 similarity threshold using CD-
(Shanghai, China). HIT to generate non-redundant gene catalog. The quality
Xiong et al. Microbiome (2020) 8:27 Page 7 of 9

filtered reads from each sample were subsequently OTUs, bacterial OTUs, and metabolism Q genes” for
mapped to the represent genes using SOAPaligner. Func- healthy and diseased plants were visualized via the “igraph”
tional gene annotation was carried out against eggNOG package in R (version 3.4.4).
database [54]. In order to focus on potentially functional The α-diversity of bacterial, fungal, and protistan com-
activities of the microbiome in the rhizosphere across munities across plant growth was estimated using the
plants growth, we extracted microbial metabolism genes non-parametric Shannon index [56]. A principal coordin-
(representing 44.85% of all functional genes, Fig. S5), in- ate analysis (PCoA) based on Bray–Curtis distance metrics
cluding the following eight general categories: [C] energy was performed in R (version 3.4.4) to explore the differ-
production and conversion, [E] amino acid transport and ences in bacterial, fungal, and protist community struc-
metabolism, [F] nucleotide transport and metabolism, [G] tures (Hellinger transformed) across plant growth.
carbohydrate transport and metabolism, [H] coenzyme ANOSIM was applied to investigate significant differences
transport and metabolism, [I] lipid transport and metabol- of microbial community structures between diseased and
ism, [P] inorganic ion transport and metabolism, and [Q] healthy plants at each time point. Abundant protistan
secondary metabolite biosynthesis, transport, and OTUs (average relative abundance > 0.1%) were used to
catabolism. examine indicator species, which were assessed in LEfSe
[57] through the “lefse command” in Mothur [58]. In
Co-occurrence network and statistical analyses addition, we used the “relaimpo” package [59] in R (ver-
First, we used co-occurrence networks to uncover the po- sion 3.4.4) to calculate the relative importance of main mi-
tential interactions between the functional groups of protists crobial parameters in predicting R. solanacearum density
and the pathogen R. solanacearum for diseased and healthy across plant growth in the combined dataset including
plants at each time point. We selected abundant functional healthy and diseased plants, healthy plant dataset, and dis-
groups of protistan OTUs (with average relative abundance eased plant dataset. We selected the diversity (Shannon
> 0.1% across all the samples) and the abundance of R. sola- index) and structure (PCoA2) of bacteria, fungi, and pro-
nacearum for network constructions. Second, we used the tists as the six main microbial predictors (Fig. S2) and
co-occurrence networks to uncover potential interactions used multiple regression by lineal models in R (version
between phagotrophic protists, bacteria, and functional 3.4.4) to calculate the significance of the correlation be-
genes for diseased and healthy plants across plant growth tween microbial predictors and R. solanacearum (all data
(combined all time point samples in diseased or heathy was standardized by “scale” function in R). We also used
plants). As we selected 8 replicates from the 12 replicates the “relaimpo” package to calculate the relative import-
for eukaryotic community profiles with 3 metagenomic rep- ance of the eight metabolism genes for R. solanacearum
licates matching both bacterial and eukaryotic datasets, we density across plant growth in the combined healthy and
used the 30 samples in total (2 symptoms × 5 time points × diseased plant samples. Other linear regression relation-
3 replicates) for the analyses. We further selected abundant ships were examined by the “lm” function in R (version
phagotrophic protistan OTUs (top 30), bacterial OTUs (top 3.4.4). Student’s t test was used to compare the microbial
30), and metabolism Q genes (top 30 genes in metabolism taxon and functional gene differences between diseased
Q category) for network constructions (detailed information and healthy plants at each time point. Normal distribution
provided in Table S5). A pairwise Spearman correlation was tested by the Shapiro-Wilk test; non-normal data
matrix was calculated with the “corr.test” function in the were log or log (x + 1) transformed [60].
package “psych” in R (version 3.4.4). The P values were ad-
justed with the false discovery rate method [55]. Spearman’s Supplementary information
correlation coefficient (ρ) higher than 0.7 (or lower than − Supplementary information accompanies this paper at https://doi.org/10.
0.7) with P values < 0.05 was selected for the networks of 1186/s40168-020-00799-9.

“functional groups of protistan OTUs with the R. solana-

Additional file 1: Table S1. Annotation table for assigning protistan
cearum.” In order to select robust correlations between pha- functional groups.
gotrophic protistan OTUs, bacterial OTUs, and metabolism Additional file 2: Table S2. Differences (based on ANOSIM) in the
Q genes, Spearman’s correlation coefficients (ρ) higher than community structure of bacteria, fungi and protists (including the
0.8 (or lower than − 0.8) with P values < 0.01 were selected abundant taxonomic and functional groups of protists) between diseased
and healthy plants at each time point across plant growth. Table S3.
for the network of “phagotrophic protists, bacteria, and Detailed information of indicator protistan taxa (OTUs) in diseased and
functional genes”. Network properties were characterized healthy plants at plant establishment. Table S4. Detailed information of
via the “igraph” package in R (version 3.4.4). Finally, net- protistan taxa (OTUs) in the networks of “functional groups of protists
directly associated with the R. solanacearum pathogen” in diseased and
works of “functional groups of protists directly associated healthy plants at plant establishment. Table S5. Detailed information of
with the R. solanacearum” at plant establishment for healthy the 30 most abundant phagotrophic protistan OTUs, bacterial OTUs and
and diseased plants were visualized in Cytoscape (v3.5.1), metabolism Q genes in the co-occurrence networks shown in Fig. 3f.
Table S6. Topological properties of networks between the abundant
and co-occurrence networks of “phagotrophic protistan
Xiong et al. Microbiome (2020) 8:27 Page 8 of 9

phagotrophic protistan OTUs, bacterial OTUs and metabolism Q genes Ecology (NIOO-KNAW), 6708, PB, Wageningen, The Netherlands. 4Laboratory
for healthy and diseased plants across plant growth. Figure S1. Abun- of Nematology, Wageningen University & Research, 6700, ES, Wageningen,
dance of R. solanacearum pathogen in diseased and healthy plants across The Netherlands.
plant growth. Figure S2. Shannon diversity of bacteria (a), fungi (b) and
protists (c) in diseased and healthy plants across plant growth, and over- Received: 21 October 2019 Accepted: 5 February 2020
all community structures of bacteria (d), fungi (e) and protists (f) in dis-
eased and healthy plants across plant growth. Figure S3. Relative
abundances of the most abundant (average relative abundance over 1%
across all samples) taxonomic (a, b, c, d, e and f) and functional (g, h, i, j References
and k) groups of protists in diseased and healthy plants across plant 1. Dean R. Kan J a. LV, Pretorius ZA, Hammond-Kosack KE, Pietro AD, Spanu
growth. Figure S4. Correlation between the relative abundance of pha- PD, et al. The top 10 fungal pathogens in molecular plant pathology. Mol
gotrophic protists and plant growth time in diseased and healthy plants. Plant Pathol. 2012;13:414–30.
Figure S5. Relative abundances of the eight metabolism genes and total 2. Mansfield J, Genin S, Magori S, Citovsky V, Sriariyanum M, Ronald P, et al.
metabolism genes in diseased and healthy plants across plant growth. Top 10 plant pathogenic bacteria in molecular plant pathology. Mol Plant
Figure S6. Relative abundance of Pseudomonas OTUs in diseased and Pathol. 2012;13:614–29.
healthy plants at week 0 and week 5. 3. Berendsen RL, Pieterse CMJ, Bakker PAHM. The rhizosphere microbiome and
plant health. Trends Plant Sci. 2012;17:478–86.
4. Mendes R, Garbeva P, Raaijmakers JM. The rhizosphere microbiome:
significance of plant beneficial, plant pathogenic, and human pathogenic
Acknowledgements microorganisms. FEMS Microbiol Rev. 2013;37:634–63.
We thank Prof. Dr. Marco Bazzicalupo and three anonymous reviewers for 5. Philippot L, Raaijmakers JM, Lemanceau P, van der Putten WH. Going back
their constructive comments on the manuscript. We also thank Prof. Dr. to the roots: the microbial ecology of the rhizosphere. Nat Rev Microbiol.
Rong Li from Nanjing Agricultural University for the helpful discussions. 2013;11:789–99.
6. Sapp M, Ploch S, Fiore-Donno AM, Bonkowski M, Rose LE. Protists are an
Authors’ contributions integral part of the Arabidopsis thaliana microbiome. Environ Microbiol.
ZW, YS, AJ, YX, and QS conceived and designed this study. YS, YG, and KY 2018;20:30–43.
conducted the laboratory work and collected the data. WX and YS were 7. Vorholt JA, Vogel C, Carlström CI, Müller DB. Establishing causality:
responsible for the bioinformatics processing and data analysis. WX, ZW, GK, opportunities of synthetic communities for plant microbiome research. Cell
AJ, and SG contributed to the interpretation of the results. WX, ZW, YS, SG, Host Microbe. 2017;22:142–55.
and GK wrote the manuscript with the inputs of all the authors. All authors 8. Gao Z, Karlsson I, Geisen S, Kowalchuk G, Jousset A. Protists: puppet masters
read and approved the final manuscript. of the rhizosphere microbiome. Trends Plant Sci. 2019;24:165–76.
9. Cha J-Y, Han S, Hong H-J, Cho H, Kim D, Kwon Y, et al. Microbial and
Funding biochemical basis of a Fusarium wilt-suppressive soil. ISME J. 2016;10:119–29.
This research was supported by the National Key Research and Development 10. Mendes R, Kruijt M, de Bruijn I, Dekkers E, van der Voort M, Schneider JHM,
Program of China (2018YFD1000800), the Fundamental Research Funds for et al. Deciphering the rhizosphere microbiome for disease-suppressive
the Central Universities (KY2201719, KYT201802), the National Natural Science bacteria. Science. 2011;332:1097–100.
Foundation of China (41922053, 41671248 and 31972504), the Natural 11. Sanguin H, Sarniguet A, Gazengel K, Moënne-Loccoz Y, Grundmann GL.
Science Foundation of Jiangsu Province (BK20170085), the Key S&T Special Rhizosphere bacterial communities associated with disease suppressiveness
Projects of China National Tobacco Corporation [110201601025 (LS-05)], and stages of take-all decline in wheat monoculture. New Phytol. 2009;184:694–707.
the Innovative Research Team Development Plan of the Ministry of 12. Manici LM, Caputo F. Fungal community diversity and soil health in
Education of China (IRT_17R56). Stefan Geisen was supported by an NWO- intensive potato cropping systems of the east Po valley, Northern Italy. Ann
VENI grant from the Netherlands Organisation for Scientific Research Appl Biol. 2009;155:245–58.
(016.Veni.181.078). Wu Xiong and Alexandre Jousset were supported by 13. Penton CR, Gupta V, Tiedje JM, Neate SM, Ophel-Keller K, Gillings M, et al.
NWO from the Netherlands Organisation for Scientific Research Fungal community structure in disease suppressive soils assessed by 28S
(ALWGR.2017.016). LSU gene sequencing. PLoS One. 2014;9:e93893.
14. Hassani MA, Durán P, Hacquard S. Microbial interactions within the plant
Availability of data and materials holobiont. Microbiome. 2018;6:58.
All raw 16S rRNA gene sequence data is available at the DDBJ Sequence 15. Adl MS, Gupta VS. Protists in soil ecology and forest nutrient cycling. Can J
Read Archive (DRA) under the accession number SRP090147. All raw 18S For Res. 2006;36:1805–17.
rRNA gene sequences are available at the NCBI Sequence Read Archive (SRA) 16. Geisen S, Mitchell EAD, Adl S, Bonkowski M, Dunthorn M, Ekelund F, et al.
under the accession number PRJNA525676. The raw data of metagenomics- Soil protists: a fertile frontier in soil biology research. FEMS Microbiol Rev.
derived gene catalogs are publicly available under the accession number 2018;42:293–323.
PRJNA492172. 17. Henkes GJ, Kandeler E, Marhan S, Scheu S, Bonkowski M. Interactions of
mycorrhiza and protists in the rhizosphere systemically alter microbial
Ethics approval and consent to participate community composition, plant shoot-to-root ratio and within-root system
Not applicable. nitrogen allocation. Front Environ Sci. 2018;6:117.
18. Rosenberg K, Bertaux J, Krome K, Hartmann A, Scheu S, Bonkowski M. Soil
Consent for publication amoebae rapidly change bacterial community composition in the
Not applicable. rhizosphere of Arabidopsis thaliana. ISME J. 2009;3:675–84.
19. Geisen S, Koller R, Hünninghaus M, Dumack K, Urich T, Bonkowski M. The
Competing interests soil food web revisited: diverse and widespread mycophagous soil protists.
The authors declare that they have no competing interests. Soil Biol Biochem. 2016;94:10–8.
20. Bonkowski M. Protozoa and plant growth: the microbial loop in soil
Author details revisited. New Phytol. 2004;162:617–31.
1
Jiangsu Provincial Key Lab of Solid Organic Waste Utilization, Key Lab of 21. Schulz-Bohm K, Geisen S, Wubs ERJ, Song C, de Boer W, Garbeva P. The
Plant Immunity, Jiangsu Collaborative Innovation Center of Solid Organic prey’s scent – volatile organic compound mediated interactions between
Wastes, Educational Ministry Engineering Center of Resource-saving fertilizers, soil bacteria and their protist predators. ISME J. 2017;11:817–20.
Nanjing Agricultural University, Nanjing 210095, Jiangsu, People’s Republic of 22. Glücksman E, Bell T, Griffiths RI, Bass D. Closely related protist strains have
China. 2Ecology and Biodiversity Group, Department of Biology, Institute of different grazing impacts on natural bacterial communities. Environ
Environmental Biology, Utrecht University, Padualaan 8, 3584, CH, Utrecht, Microbiol. 2010;12:3105–13.
The Netherlands. 3Department of Terrestrial Ecology, Netherlands Institute for
Xiong et al. Microbiome (2020) 8:27 Page 9 of 9

23. Jousset A, Lara E, Wall LG, Valverde C. Secondary metabolites help 45. Bass D, Silberman JD, Brown MW, Pearce RA, Tice AK, Jousset A, et al.
biocontrol strain pseudomonas fluorescens CHA0 to escape protozoan Coprophilic amoebae and flagellates, including Guttulinopsis, Rosculus and
grazing. Appl Environ Microbiol. 2006;72:7083–90. Helkesimastix, characterise a divergent and diverse rhizarian radiation and
24. Mazzola M, de Bruijn I, Cohen MF, Raaijmakers JM. Protozoan-induced contribute to a large diversity of faecal-associated protists. Environ
regulation of cyclic lipopeptide biosynthesis is an effective predation Microbiol. 2016;18:1604–19.
defense mechanism for Pseudomonas fluorescens. Appl Environ Microbiol. 46. Raaijmakers JM, De Bruijn I, Nybroe O, Ongena M. Natural functions of
2009;75:6804–11. lipopeptides from Bacillus and Pseudomonas: more than surfactants and
25. Hünninghaus M, Dibbern D, Kramer S, Koller R, Pausch J, Schloter-Hai B, antibiotics. FEMS Microbiol Rev. 2010;34:1037–62.
et al. Disentangling carbon flow across microbial kingdoms in the 47. Raza W, Ling N, Liu D, Wei Z, Huang Q, Shen Q. Volatile organic compounds
rhizosphere of maize. Soil Biol Biochem. 2019;134:122–30. produced by Pseudomonas fluorescens WR-1 restrict the growth and
26. Wei Z, Gu Y, Friman V-P, Kowalchuk GA, Xu Y, Shen Q, et al. Initial soil virulence traits of Ralstonia solanacearum. Microbiol Res. 2016;192:103–13.
microbiome composition and functioning predetermine future plant health. 48. Chen MC, Wang JP, Zhu YJ, Liu B, Yang WJ, Ruan CQ. Antibacterial activity
Sci Adv. 2019;5:eaaw0759. against Ralstonia solanacearum of the lipopeptides secreted from the Bacillus
27. Zhao Z-B, He J-Z, Geisen S, Han L-L, Wang J-T, Shen J-P, et al. Protist amyloliquefaciens strain FJAT-2349. J Appl Microbiol. 2019;126:1519–29.
communities are more sensitive to nitrogen fertilization than other 49. Xiong W, Jousset A, Guo S, Karlsson I, Zhao Q, Wu H, et al. Soil protist
microorganisms in diverse agricultural soils. Microbiome. 2019;7:33. communities form a dynamic hub in the soil microbiome. ISME J. 2018;12:
28. Jiang G, Wei Z, Xu J, Chen H, Zhang Y, She X, et al. Bacterial wilt in China: 634–8.
history, current status, and future perspectives. Front Plant Sci. 2017;8:1549. 50. Xiong W, Li R, Guo S, Karlsson I, Jiao Z, Xun W, et al. Microbial amendments
29. Salanoubat M, Genin S, Artiguenave F, Gouzy J, Mangenot S, Arlat M, et al. alter protist communities within the soil microbiome. Soil Biol Biochem.
Genome sequence of the plant pathogen Ralstonia solanacearum. Nature. 2019;135:379–82.
2002;415:497–502. 51. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. UCHIME improves
30. Dumack K, Fiore-Donno AM, Bass D, Bonkowski M. Making sense of sensitivity and speed of chimera detection. Bioinformatics. 2011;27:2194–200.
environmental sequencing data: ecologically important functional traits of 52. Guillou L, Bachar D, Audic S, Bass D, Berney C, Bittner L, et al. The Protist
the protistan groups Cercozoa and Endomyxa (Rhizaria). Mol Ecol Resour. Ribosomal Reference database (PR2): a catalog of unicellular eukaryote small
2019; Available from: https://onlinelibrary.wiley.com/doi/abs/10.1111/1755- sub-unit rRNA sequences with curated taxonomy. Nucleic Acids Res. 2013;
0998.13112. 41:D597–604.
31. Burke C, Steinberg P, Rusch D, Kjelleberg S, Thomas T. Bacterial community 53. Noguchi H, Park J, Takagi T. MetaGene: prokaryotic gene finding from
assembly based on functional genes rather than species. Proc Natl Acad Sci. environmental genome shotgun sequences. Nucleic Acids Res. 2006;34:
2011;108:14288–93. 5623–30.
32. Louca S, Parfrey LW, Doebeli M. Decoupling function and taxonomy in the 54. Huerta-Cepas J, Szklarczyk D, Heller D, Hernández-Plaza A, Forslund SK, Cook
global ocean microbiome. Science. 2016;353:1272–7. H, et al. eggNOG 5.0: a hierarchical, functionally and phylogenetically
33. Ma X, Zhang Q, Zheng M, Gao Y, Yuan T, Hale L, et al. Microbial functional annotated orthology resource based on 5090 organisms and 2502 viruses.
traits are sensitive indicators of mild disturbance by lamb grazing. ISME J. Nucleic Acids Res. 2019;47:D309–14.
2019;13:1370–3. 55. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical
34. Bass D, Stentiford GD, Wang H-C, Koskella B. Tyler CR. Trends Ecol Evol: The and powerful approach to multiple testing. J R Stat Soc Ser B Methodol.
pathobiome in animal and plant diseases; 2019. Available from: http://www. 1995;57:289–300.
sciencedirect.com/science/article/pii/S0169534719302241. 56. Chao A, Shen T-J. Nonparametric estimation of Shannon’s index of diversity
35. Vayssier-Taussat M, Albina E, Citti C, Cosson JF, Jacques M-A, Lebrun M-H, when there are unseen species in sample. Environ Ecol Stat. 2003;10:429–43.
et al. Shifting the paradigm from pathogens to pathobiome: new concepts 57. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al.
in the light of meta-omics. Front Cell Infect Microbiol. 2014;4:29. Metagenomic biomarker discovery and explanation. Genome Biol. 2011;12:
36. Michelsen CF, Watrous J, Glaring MA, Kersten R, Koyama N, Dorrestein PC, R60.
et al. Nonribosomal peptides, key biocontrol components for Pseudomonas 58. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, et al.
fluorescens In5, Isolated from a Greenlandic Suppressive Soil. mBio. 2015;6: Introducing mothur: open-source, platform-independent, community-
e00079–15. supported software for describing and comparing microbial communities.
Appl Environ Microbiol. 2009;75:7537–41.
37. Morriën E, Hannula SE, Snoek LB, Helmsing NR, Zweers H, de Hollander M,
59. Groemping U. Relative importance for linear regression in R: the package
et al. Soil networks become more connected and take up more carbon as
relaimpo. J Stat Softw. 2006;17:1–27.
nature restoration progresses. Nat Commun. 2017;8:14349.
60. McDonald JH. Handbook of biological statistics. Baltimore: sparky house
38. de Araujo ASF, Mendes LW, Lemos LN, Antunes JEL, Beserra JEA. Lyra M do
publishing; 2009.
CCP de, et al. Protist species richness and soil microbiome complexity
increase towards climax vegetation in the Brazilian Cerrado. Commun Biol.
2018;1:1–8. Publisher’s Note
39. Montoya JM, Rodríguez MA, Hawkins BA. Food web complexity and higher- Springer Nature remains neutral with regard to jurisdictional claims in
level ecosystem services. Ecol Lett. 2003;6:587–93. published maps and institutional affiliations.
40. Soliveres S, van der Plas F, Manning P, Prati D, Gossner MM, Renner SC,
et al. Biodiversity at multiple trophic levels is needed for ecosystem
multifunctionality. Nature. 2016;536:456–9.
41. Flues S, Bass D, Bonkowski M. Grazing of leaf-associated Cercomonads
(Protists: Rhizaria: Cercozoa) structures bacterial community composition
and function. Environ Microbiol. 2017;19:3297–309.
42. Kreuzer K, Adamczyk J, Iijima M, Wagner M, Scheu S, Bonkowski M. Grazing
of a common species of soil protozoa (Acanthamoeba castellanii) affects
rhizosphere bacterial community composition and root architecture of rice
(Oryza sativa L.). Soil Biol Biochem. 2006;38:1665–72.
43. Shi S, Nuccio EE, Shi ZJ, He Z, Zhou J, Firestone MK. The interconnected
rhizosphere: high network complexity dominates rhizosphere assemblages.
Ecol Lett. 2016;19:926–36.
44. Cardenas E, Wu W-M, Leigh MB, Carley J, Carroll S, Gentry T, et al. Significant
association between sulfate-reducing bacteria and uranium-reducing
microbial communities as revealed by a combined massively parallel
sequencing-indicator species approach. Appl Environ Microbiol. 2010;76:
6778–86.