Root microbiota drive direct integration of phosphate stress and
immunity
Gabriel Castrillo
*,1,2
, Paulo José Pereira Lima Teixeira
*,1,2
, Sur Herrera Paredes
*,1,2,3
,
Theresa F. Law
1,2
, Laura de Lorenzo
4
, Meghan E. Feltcher
1,2
, Omri M. Finkel
1,2
, Natalie W.
Breakfield
1,2
, Piotr Mieczkowski
5,6,7
, Corbin D. Jones
1,3,5,6,7,8
, Javier Paz-Ares
4
, and Jeffery
L. Dangl
1,2,3,7,8,9
1
Department of Biology, University of North Carolina, Chapel Hill, NC 27599-3280, USA
2
Howard Hughes Medical Institute, University of North Carolina, Chapel Hill, NC 27599-3280,
USA
3
Curriculum in Bioinformatics and Computational Biology, University of North Carolina, Chapel
Hill, NC 27599-3280, USA
4
Department of Plant Molecular Genetics, Centro Nacional de Biotecnología, CNB-CSIC, Darwin
3, 28049 Madrid, Spain
5
Department of Genetics, University of North Carolina, Chapel Hill, NC, USA
6
Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC
27599-3280, USA
7
Carolina Center for Genome Sciences, University of North Carolina, Chapel Hill, NC
27599-3280, USA
Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research,
subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms
Correspondence and requests for materials should be addressed to J.L.D. (dangl@email.unc.edu).
*
Indicates co-first author
Current addresses:
NB – NewLeaf Symbiotics, St. Louis, MO 63132 USA
LdL -- Department of Plant and Soil Sciences, University of Kentucky, Lexington, KY 40546 USA.
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature
Author contributions
G.C., P.J.P.L.T., S.H.P and J.L.D. designed the project, G.C., S.H.P, T.F.L and M.E.F. set up the experiments, harvested samples and
organized construction of 16S sequencing libraries. G.C and T.F.L performed control experiments related with PSR induced by the
SynCom. G.C, N.W.B, M.E.F and T.F.L set up the experiments, harvested samples and isolated RNA. P.J.P.L.T. organized, performed
construction of RNA-seq libraries and analysed RNA-seq data. S.H.P. analysed 16S sequencing data. S.H.P and P.J.P.L.T oversaw data
deposition. G.C, T.F.L. and P.J.P.L.T performed pathology experiments. G.C., P.J.P.L.T, S.H.P, T.F.L, O.M.F and J.L.D. analysed data
and created figures. L.d.L. performed the ChIP-seq experiment. C.D.J and P.M. advised on sequencing process and statistical methods.
G.C., P.J.P.L.T., S.H.P., and J.L.D. wrote the manuscript with input from O.M.F., C.D.J., and J.P.-A.
Author Information
All data generated from this project is publicly available. Raw sequences from soil census and SynCom colonization are available at
the EBI Sequence Read Archive accession PRJEB15671. Count tables, metadata, taxonomic annotations and OTU representative
sequences from the Mason Farm census and Syncom experiments are available as Supplementary Datasets 1 and Supplementary
Datasets 2 respectively. Custom scripts for analysis and visualization are at https://github.com/surh/pbi. Raw sequences and counts
from RNA-seq experiments are available at the NCBI Gene Expression Omnibus under accession number GSE87339. The
corresponding metadata information is provided in Supplementary Table 15. All code is available upon request. Reprint and
permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interest. Readers are
welcome to comment on the online version of this article at www.nature.com/nature.
Published as:
Nature
. 2017 March 23; 543(7646): 513–518.
HHMI Author Manuscript HHMI Author Manuscript HHMI Author Manuscript
8
Curriculum in Genetics and Molecular Biology, University of North Carolina, Chapel Hill, NC
27599-3280, USA
9
Department of Microbiology and Immunology, University of North Carolina, Chapel Hill, NC
27599-3280, USA
Abstract
Plants live in biogeochemically diverse soils that harbor extraordinarily diverse microbiota. Plant
organs associate intimately with a subset of these microbes; this community’s structure can be
altered by soil nutrient content. Plant-associated microbes can compete with the plant and with
each other for nutrients; they can also provide traits that increase plant productivity. It is unknown
how the plant immune system coordinates microbial recognition with nutritional cues during
microbiome assembly. We establish that a genetic network controlling phosphate stress response
influences root microbiome community structure, even under non-stress phosphate conditions. We
define a molecular mechanism regulating coordination between nutrition and defense in the
presence of a synthetic bacterial community. We demonstrate that the master transcriptional
regulators of phosphate stress response in Arabidopsis also directly repress defense, consistent
with plant prioritization of nutritional stress over defense. Our work will impact efforts to define
and deploy useful microbes to enhance plant performance.
Plant organs create distinct physical and chemical environments that are colonized by
specific microbial taxa
1
. These can be modulated by the plant immune system
2
and by soil
nutrient composition
3
. Phosphorus is present in the biosphere at high concentrations, but
plants can only absorb orthophosphate (Pi), a form also essential for microbial
proliferation
4, 5
and scarce in soil
6
. Thus, plants possess adaptive phosphate starvation
responses (PSR) to manage low Pi availability that typically occurs in the presence of plant-
associated microbes. Common strategies for increasing Pi uptake capacity include rapid
extension of lateral roots foraging into topsoil where Pi accumulates
7
and establishment of
beneficial relationships with some soil microorganisms
8, 9
. For example, the capacity of a
specific mutualistic fungus to colonize Arabidopsis roots is modulated by plant phosphate
status implying coordination between the PSR and the immune system
8, 10
. Descriptions of
pathogen exploitation of PSR-immune system coordination are emerging
11, 12
.
We demonstrate that Arabidopsis mutants with altered phosphate starvation responses (PSR)
assemble atypical microbiomes, either in phosphate-replete wild soil, or during
in vitro
colonization with a synthetic bacterial community (SynCom). This SynCom competes for
phosphate with the plant and induces PSR in limiting phosphate. PSR in these conditions
requires the master transcriptional regulator PHR1 and its weakly redundant paralog, PHL1.
The severely reduced PSR observed in
phr1 phl1
mutants is accompanied by transcriptional
changes in plant defense leading to enhanced immune function. Negative regulation of
immune system components by PHR1 is direct, as measured by target gene promoter
occupancy, and functional, as validated by pathology phenotypes. Thus, PHR1 directly
activates microbiome-enhanced response to phosphate limitation while repressing
microbially-driven plant immune system outputs.
Castrillo et al. Page 2
Nature
. Author manuscript; available in PMC 2017 September 15.
HHMI Author Manuscript HHMI Author Manuscript HHMI Author Manuscript
The root microbiome in plants with altered phosphate stress response
We linked PSR to the root microbiome by contrasting the root bacterial community of wild-
type Arabidopsis Col-0 with three types of PSR mutants (Fig. 1a, b, Supplementary Text 1;
Extended Data Fig. 1; Supplementary Table 1). PSR, historically defined in axenic seedlings
and measured by Pi concentration in the plant shoot, is variable across these mutants. In
replete Pi and axenic conditions,
phr1
plants accumulate less free Pi than wild-type
13
;
pht1;1, pht1;1 pht1;4
and
phf1
accumulate very low Pi levels and express constitutive
PSR
14, 15
; and
pho2
,
nla
and
spx1 spx2
express diverse magnitudes of Pi hyper-
accumulation
16– 18
. We grew plants in a previously characterized wild soil
19
that is not
overtly phosphate deficient (Extended Data Fig. 2). Generally, the Pi concentration of PSR
mutants grown in this wild soil recapitulated those defined in axenic conditions, except for
phf1
and
nla
which displayed the opposite phenotype to that observed in axenic agar, and
phr1
which accumulated the same Pi concentration as Col-0 (Fig. 1b). These results suggest
that complex chemical conditions, soil microbes, or a combination of these can alter Pi
metabolism in these mutants.
Bacterial root endophytic (EC) community profiles were consistent with previous
studies
2, 19
. Constrained ordination revealed significant differences between bacterial
communities across the Pi accumulation gradient represented by these PSR mutants [5.3 %
constrained variance, canonical analysis of principal coordinates (CAP)] (Fig. 1c).
Additionally, CAP confirmed that
phr1
and
spx1 spx2
carried different communities, as
evidenced by their separation on the first three ordination axes, and that
phf1
was the most
affected of Pi-transport mutant (Fig. 1c). Specific bacterial taxa had differential abundances
inside the roots of mutant plants compared to wild-type. Mutants from the same PSR type
had a similar effect on the root microbiome at a low taxonomic level [97 % identity
Operational Taxonomic Unit (OTU)] (Fig. 1d), while they had no overlapping effect at a
higher taxonomic level (Family, Extended Data Fig. 1g). This suggests that closely related
groups of bacteria have differential colonization patterns on the same host genotypes.
Importantly, we found that the enrichment and depletion profiles were better explained by
PSR mutant signaling type rather than the mutant’s capacity for Pi accumulation: all of the
Pi-transport-related mutants had a similar effect on the root microbiome, and the
antagonistic PSR regulators
phr1
and
spx1 spx2
each exhibited unique patterns (Fig. 1a, d
and Extended Data Fig. 1f, g). Our results indicate that PSR components influence root
microbiome composition in plants grown in a phosphate-replete wild soil, leading to
alteration of the abundance of specific microbes across diverse levels of Pi accumulation
representing diverse magnitudes of PSR.
Phosphate starvation response in a microcosm reconstitution
Our observations in a wild soil suggested complex interplay between PSR and the presence
of a microbial community. Thus, we deployed a tractable but complex bacterial synthetic
community (SynCom) of 35 taxonomically diverse, genome-sequenced bacteria isolated
from the roots of Brassicaceae (nearly all from Arabidopsis) and two wild soils. This
SynCom approximates the phylum level distribution observed in wild-type root endophytic
compartments (Extended Data Fig. 3, Supplementary Table 1, Supplementary Table 2). We
Castrillo et al. Page 3
Nature
. Author manuscript; available in PMC 2017 September 15.
HHMI Author Manuscript HHMI Author Manuscript HHMI Author Manuscript
inoculated seedlings of Col-0,
phf1
and the double mutant
phr1 phl1
(a redundant paralogue
of
phr1
13
) grown on agar plates in low or high Pi (Supplementary Text 2). Twelve days later,
we noted that the SynCom had a negative effect on shoot Pi accumulation of Col-0 plants
grown on low Pi, but not on plants grown on replete phosphate (Fig. 2a). As expected, both
PSR mutants accumulated less Pi than Col-0; the SynCom did not rescue this defect. Thus,
in this microcosm, plant-associated microbes drive a context-dependent competition with the
plant for Pi.
We sought to establish whether PSR was activated by the SynCom. We generated a
literature-based core set of 193 PSR transcriptional markers and explored their expression in
transcriptomic experiments (Extended Data Fig. 4a, b, Supplementary Table 3). In axenic
low Pi conditions, only the constitutive Pi-stressed mutant
phf1
exhibited induction of these
PSR markers. By contrast, Col-0 plants expressed only a marginal induction of PSR markers
compared to those plants grown at high Pi (Fig. 2b). This is explained by the purposeful
absence of sucrose, a key component for the PSR induction
in vitro
20
(Supplementary Text
2; Extended Data Fig. 5) that cannot be used in combination with bacterial SynCom
colonization protocols. Remarkably, the SynCom greatly enhanced the canonical
transcriptional response to Pi starvation in Col-0 (Fig. 2b); this was dependent on PHR1 and
PHL1 (Fig. 2b; Extended Data Fig. 4b). Various controls validated these conclusions
(Supplementary Text 2; Extended Data Fig. 4–Fig. 6). Importantly, shoots of plants pre-
colonized with SynCom on 0 or 50 µM Pi, but not on 650 µM Pi, accumulated 20–40 times
more Pi than shoots from similarly treated non-colonized plants when subsequently
transferred to full Pi conditions in the absence of additional bacteria (Fig. 2c and
Supplementary Table 4). This demonstrates functional PSR activation by the SynCom. We
thus propose that the transcriptional response to low Pi induced by our SynCom reflects an
integral microbial element of normal PSR in complex biotic environments.
We evaluated agar- and root-associated microbiomes of plants grown with the SynCom
(Supplementary Text 3; Fig. 2d, e; Extended Data Fig. 7e, f; Supplementary Table 5). In line
with results from plants grown in wild soil, we found that PSR mutants failed to assemble a
wild-type SynCom microbiome (Fig. 2f). Some strains were differentially abundant across
PSR mutants
phf1
and
phr1 phl1
(Fig. 2e, f; Extended Data Fig. 7c), Pi concentration (Fig.
2g; Extended Data Fig. 7d), or sample fraction (Extended Data Fig. 7b, e, f). These results
established a microcosm reconstitution system to study plant PSR under chronic competition
with plant-associated microbes and allowed us to confirm that the tested PSR mutants
influence root microbiome membership.
Coordination between phosphate stress response and immune system
output
We noted that
phr1 phl1
and
phf1
differentially activated transcriptional PSR in the presence
of our SynCom (Fig. 2b). Therefore, we investigated the transcriptomes of plants growing in
the SynCom to understand how these microbes activate PHR1-dependent PSR. We identified
differentially expressed genes (DEGs) that responded to either low Pi, presence of the
SynCom, or the combination of both (hereafter PSR-SynCom DEGs) (Supplementary Text
Castrillo et al. Page 4
Nature
. Author manuscript; available in PMC 2017 September 15.
HHMI Author Manuscript HHMI Author Manuscript HHMI Author Manuscript
4; Extended Data Fig. 8a, b; Supplementary Table 6). Hierarchical clustering (Fig. 3a,
Supplementary Table 7) revealed gene sets (c1, c2, c7 and c10) that were more strongly
activated in Col-0 than in
phr1
or
phr1 phl1
. These clusters contained most of the core PSR
markers regulated by PHR1 (Fig. 3b). They were also enriched in PHR1 direct targets
identified in an independent ChIP-seq experiment (Fig. 3c, Supplementary Table 8), PHR1
promoter binding motifs (Extended Data Fig. 4c), and genes involved in biological processes
related to PSR (Fig. 3d and Supplementary Table 9). PHR1 surprisingly contributed to
transcriptional regulation of plant immunity. Five of the twelve clusters (Fig. 3a; c3, c6, c7,
c8 and c11) were enriched in genes related to plant immune system output; four of these
were over-represented for jasmonic acid (JA) and/or salicylic acid (SA) pathway markers
(Fig. 3d, c3, c7, c8, and c11; Supplementary Table 9) and three of these four were enriched
for PHR1 direct targets (Fig. 3c). SA and JA are plant hormone regulators of immunity and
at least SA modulates Arabidopsis root microbiome composition
2
.
To further explore PHR1 function in the regulation of plant immunity, we generated
transcriptomic time course data for treatment-matched Col-0 seedlings following application
of Methyl Jasmonate (MeJA) or the SA analogue Benzothiadiazole (BTH; Supplementary
Table 10). We found a considerable over-representation of SA- and JA-activated genes
among the PSR-SynCom DEGs (468 versus 251 predicted for SA and 165 vs. 80 predicted
for JA; p<0.0001, hypergeometric test) (Extended Data Fig. 8c–h, Supplementary Table 7).
A large proportion of SA-responsive genes were more strongly expressed in
phr1
and
phr1
phl1
than in Col-0; these were strongly enriched for classical SA-dependent defense genes
(Extended Data Fig. 8d, e). A second group of SA-responsive genes that were less expressed
in
phr1
and
phr1 phl1
than in Col-0 lacked classical SA-dependent defense genes and were
weakly enriched for genes likely contributing to PSR (Extended Data Fig. 8d). By contrast,
most JA-responsive genes exhibited lower expression in
phr1
and
phr1 phl1
(Extended Data
Fig. 8g, h), including a subset of 18 of 46 genes known or predicted to mediate biosynthesis
of defense-related glucosinolates (Extended Data Fig. 8i)
21
. This agrees with the recent
observation that
phr1
exhibited decreased glucosinolate levels during Pi starvation
22
.
Analyses of SA- and JA- up-regulated genes revealed enrichment of direct PHR1 targets
(Fig. 3e), consistent with the converse observation that some PHR1-regulated clusters
enriched in direct targets were also enriched in defense genes (Fig. 3c, d). Many of the SA-
and JA- responsive genes were PSR-SynCom DEGs (Fig. 3f; Extended Data Fig. 8c–h,
Supplementary Table 7). Thus, PHR1 directly regulates an unexpected proportion of the
plant immune system during PSR triggered by our SynCom.
PHR1 integrates plant immune system output and phosphate stress
response
We tested whether PHR1 also controls the expression of plant defense genes under
conditions typically used to study PSR (axenic growth, sucrose present, no microbiota
involved). We performed RNA-seq in response to low Pi in these conditions and identified
1482 DEGs in Col-0 and 1161 DEGs in
phr1 phl1
(Fig. 4a, b; Extended Data Fig. 9,
Supplementary Table 11). A significant number of our BTH/SA-activated genes were also
up-regulated in
phr1 phl1
, but not in Col-0 in response to low Pi (Fig. 4a, b; Supplementary
Castrillo et al. Page 5
Nature
. Author manuscript; available in PMC 2017 September 15.
HHMI Author Manuscript HHMI Author Manuscript HHMI Author Manuscript