A genome-wide transcriptional analysis using Arabidopsis thaliana Affymetrix gene chips determined plant responses to phosphate deprivation

Abstract

Phosphorus, one of the essential elements for plants, is often a limiting nutrient because of its low availability and mobility in soils. Significant changes in plant morphology and biochemical processes are associated with phosphate (Pi) deficiency. However, the molecular bases of these responses to Pi deficiency are not thoroughly elucidated. Therefore, a comprehensive survey of global gene expression in response to Pi deprivation was done by using Arabidopsis thaliana whole genome Affymetrix gene chip (ATH1) to quantify the spatio-temporal variations in transcript abundance of 22,810 genes. The analysis revealed a coordinated induction and suppression of 612 and 254 Pi-responsive genes, respectively. The functional classification of some of these genes indicated their involvement in various metabolic pathways, ion transport, signal transduction, transcriptional regulation, and other processes related to growth and development. This study is a detailed analysis of Pi starvation-induced changes in gene expression of the entire genome of Arabidopsis correlated with biochemical processes. The results not only enhance our knowledge about molecular processes associated with Pi deficiency, but also facilitate the identification of key molecular determinants for improving Pi use by crop species.

Full text

A genome-wide transcriptional analysis using
Arabidopsis thaliana
Affymetrix gene chips
determined plant responses to phosphate deprivation
Julie Misson*, Kashchandra G. Raghothama
†
, Ajay Jain
†
, Juliette Jouhet
‡
, Maryse A. Block
‡
, Richard Bligny
‡
,
Philippe Ortet
§
, Audrey Creff*, Shauna Somerville
¶
, Norbert Rolland
‡
, Patrick Doumas
储
, Philippe Nacry
储
,
Luis Herrerra-Estrella**, Laurent Nussaume*, and Marie-Christine Thibaud*
††
*Laboratoire de Biologie du De´ veloppement des Plantes, Unite Mixte de Recherche 6191, Centre National de la Recherche Scientifique–Commissariat a`
l’Energie Atomique, Aix-Marseille II, 13108 Saint-Paul-lez-Durance, France;
†
Department of Horticulture and Landscape Architecture, Purdue University,
West Lafayette, IN 47907-1165;
‡
Laboratoire de Physiologie Cellulaire Ve´ge´ tale, Unite Mixte de Recherche 5168, Centre National de la Recherche
Scientifique兾Commissariat a´ l’Energie Atomique兾Universite´ Joseph Fourier兾Institut National de la Recherche Agronomique, 38054 Grenoble, France;
§
De´ partement d’Ecophysiologie Ve´ge´ tale et de Microbiologie, 13108 Saint-Paul-lez-Durance, France;
¶
Department of Plant Biology, Carnegie Institute,
Stanford, CA 94305;
储
Laboratoire de Biochimie et Physiologie Mole´ culaires des Plantes, Unite Mixte de Recherche 5004, Institut National de la Recherche
Argonomique兾Centre National de la Recherche Scientifique兾E
´
cole Nationale Supe´ rieure d’Arts et Me´ tiers兾Universite´ Montpellier II, 2 Place Viala, 34060
Montpellier, France; and **Departamento de Ingenieria Genetica de Plantas, Centro de Investigacio´ n de Estudios Avanzados del Instituto Politecnico
Nacional, Unidad Irapuato, Apartado Postal 629, 36500 Irapuato, Guanajuato, Mexico
Contributed by Luis Herrera-Estrella, June 23, 2005
Phosphorus, one of the essential elements for plants, is often a
limiting nutrient because of its low availability and mobility in
soils. Significant changes in plant morphology and biochemical
processes are associated with phosphate (Pi) deficiency. However,
the molecular bases of these responses to Pi deficiency are not
thoroughly elucidated. Therefore, a comprehensive survey of
global gene expression in response to Pi deprivation was done by
using Arabidopsis thaliana whole genome Affymetrix gene chip
(ATH1) to quantify the spatio-temporal variations in transcript
abundance of 22,810 genes. The analysis revealed a coordinated
induction and suppression of 612 and 254 Pi-responsive genes,
respectively. The functional classification of some of these genes
indicated their involvement in various metabolic pathways, ion
transport, signal transduction, transcriptional regulation, and
other processes related to growth and development. This study is
a detailed analysis of Pi starvation-induced changes in gene ex-
pression of the entire genome of Arabidopsis correlated with
biochemical processes. The results not only enhance our knowl-
edge about molecular processes associated with Pi deficiency, but
also facilitate the identification of key molecular determinants for
improving Pi use by crop species.
P
hosphate (Pi) is an essential macronutriment required for plant
growth and development. Its low availability to plants in many
soils re sults not only from limiting amounts, but also from its
association with cations and organic compounds creating insoluble
complexe s. Thus, Pi has become one of the major plant nutrition
problems limiting growth in both acidic and calcareous soils (1).
Applications of large quantities of fertilizers to correct this problem
are not economically sustainable and also lead to environmental
pollution. Therefore, efforts have been directed to understanding
the molecular basis of plants responses to Pi deficiency and to
identifying Pi-responsive genes whose expression can be manipu-
lated to enable plant growth in low-Pi environments. Although a
host of genes representing Pi transporters, phosphatase s, RNase s,
and others (2–5) have been identified and characterized by tradi-
tional expression studies, knowledge of global changes in the
expre ssion of Pi-responsive genes is still lacking. This technical
limitation has now been circumvented by recently developed mi-
croarray technology, which has been used successfully to study the
effects of different abiotic stresse s to large number of plant genes
in parallel (6). To date, relatively small microarrays containing
probes for less than one-third of all genes have been used in
Arabidopsis, rice, and white lupin to monitor molecular response s to
Pi deprivation (7–10). In Arabidopsis, the expression of ⬇29% of
6,172 genes examined changed in response to Pi starvation for up
to 3 days (10). In another study with 8,100 genes of Arabidopsis,
differential expression of a group of 61 genes was observed after
100 h of Pi starvation (7). These studies represented the evaluation
of only a part of the Arabidopsis genome for Pi-responsive genes.
The advent of new microarray technique of Affymetrix ATH1
GeneChip, containing 22,810 Arabidopsis probe sets, has now made
it feasible to evaluate the expression of almost all of the genes in the
Arabidopsis genome with a sensitivity of one transcript per cell (6).
Here, we used the ATH1 GeneChip for a global evaluation of
genes that are spatio-temporally regulated in response to short-,
medium-, and long-term Pi deprivation. The sensitivity of this
technique was corroborated by expre ssion analysis. Identification of
differentially expre ssed genes revealed the coordinated activation
and repression of genes involved in many biochemical pathways that
are closely associated with plant responses to Pi deficiency. The se
genes could serve as potential candidates to decipher the compo-
nents of Pi-sensing mechanisms and developing strategies to im-
prove P efficiency in crops.
Materials and Methods
Supporting Information. For further details, see Figs. 5 and 6 and
Tables 1–9, which are published as supporting information on the
PNAS web site.
Plant Material and Growth Conditions. Arabidopsis thaliana (L.)
plants were raised in liquid culture and transferred in a medium
with or without Pi as described (11) for evaluating the short-term
(3, 6, and 12 h pooled) and medium-term (1 and 2 days pooled)
effects of Pi deficiency on the gene expression. For studying
long-term effects of Pi deficiency, surface-sterilized seeds were
sown in square (12 ⫻ 12 cm) Petri dishes on Murashige and Skoog
(MS)兾10 medium, 0.5% sucrose, 0.8% agar, and supplemented with
either 500
␮
M(P⫹)or5
␮
M(P⫺) Pi; plants were grown vertically
(12). Because some Arabidopsis genes are regulated by diurnal
rhythm and circadian clocks (13), the roots and the leaves were
harvested separately at the beginning and at the end of the
photoperiod and pooled. Samples were rinsed with distilled water,
blot-dried, and frozen in liquid nitrogen.
RNA Extraction and cRNA Preparation. Total RNA from shoot and
root (long-term) and from the whole plant (short- and medium-
Abbreviations: Pi, phosphate; qRT-PCR, quantitative RT-PCR.
††
To whom correspondence should be addressed. E-mail: mcthibaud.cea.fr.
© 2005 by The National Academy of Sciences of the USA
11934–11939
兩
PNAS
兩
August 16, 2005
兩
vol. 102
兩
no. 33 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0505266102

term) was extracted as described (11, 12). cRNA was prepared
following the manufacturer’s instructions (www.affymetrix.com兾
support兾technical兾manual兾expression㛭manual.affx).
Microarray Hybridization and Data Analysis. The Affymetrix micro-
arrays (Arabidopsis ATH1 genome array) contain 22,810 probe
sets representing ⬇80% of the gene sequence s on a single
array. Labeling and hybridization on the ATH1 microarrays (one
sample per chip) were performed according to the manufacturer’s
instr uctions (www.af fy metrix.com兾support兾technical兾manual兾
expre ssion㛭manual.affx). The probe arrays were scanned and fur-
ther analyzed with
GENESPRING software (version 5.0; Silicon
Genetics). Normalization per gene and per chip of the log
2
values
was performed to allow the comparison of the three independent
replicates performed for each set of experiment. In addition,
normalization was performed separately for each experiment and
plant tissue for all measurements using the flags (‘‘present,’’ ‘‘mar-
ginal,’’ or ‘‘absent’’) assigned by Affymetrix treatment of the arrays.
However, only those transcripts that were declared present or
marginal in at least three of six chips were taken into account. Such
a procedure facilitated the elimination of transcripts with very low
signals in both treatments (declared ‘‘absent’’). This elimination was
achieved by selecting the genes that were absent in at least four
microarrays (four, five, or six arrays) and on this basis, comple-
mentary lists of genes were generated. Data were then analyzed for
each of the experiments comprising plants deprived of Pi for
different lengths of time and results were compared with corre-
sponding P(⫹) plants (control). The genes that revealed significant
changes in their expression in P(⫺) plants were selected by applying
a t test (one-way ANOVA Welch t te st, P ⫽ 0.05). Moreover, a
cutoff value of 2-fold change, which is commonly used for microar-
ray analysis (10), was adopted to discriminate expression of gene s
that were differentially altered in re sponse to Pi deficiency. Anno-
tation of the genes represented on the microarray to genomic ORFs
was done with ‘‘gene description’’ and ‘‘gene ontology’’ programs of
GENESPRING (based on information from the international Arabi-
dopsis Genome Initiative sequencing project in collaboration with
The Institute for Genome Research). To test the hybridization
quality, ‘‘Arabidopsis control genes’’ coding for GAPDH, actin,
tubulin, ubiquitin, and several ribosomal RNAs (25S, 5S), spotted
by the manufacturer, were verified. The expression ratios [P(⫺)兾
P(⫹)] of the control genes were consistently in the range of
0.81–1.29.
Real-Time Quantitative RT-PCR (qRT-PCR) and Northern Analysis. A
few differentially regulated Pi-responsive genes identified from the
microarray analysis were selected for validation of the results by
qRT-PCR and Northern analysis. cDNA was used for performing
qRT-PCR (iCycler Real-Time PCR Detection System, Bio-Rad).
Specific primers (T
m
, 58–63°C) were designed to generate PCR
products between 150 and 350 bp (Table 1). qRT-PCR of GAPDH
C (At3g04120) was performed for standardization. Platinium
Quantitative PCR SuperMix-UDG (Invitrogen) was used for the
PCRs according to the manufacturer’s protocol with a minor
modification (0.33
␮
M of each primer in a final volume of 15
␮
l).
Northern analysis was performed as described (11).
S6K2 Promoter Fusion. The gene AtS6K2 (At3g08720), coding for a
ribosomal protein S6 kinase is induced during Pi deficiency (Table
7). A 520-bp fragment of its promoter located upstream of the
transcription start was PCR amplified and cloned into the modified
binary vector pBIN-
35
S-mgfp4 at HindIII兾XbaI replacing the
35
S
promoter (kindly provided by J. Haseloff, University of Cambridge,
Cambridge, U.K.). Transgenic plants were generated by vacuum
infiltration of inflorescence with Agrobacterium as described (14).
Quantification of Lipids, Anthocyanins, Macro- and Microelements,
and Pi. Lipids, fatty acids, and anthocyanins were quantified as
described (15–17). Soluble inorganic Pi was measured in plants
from short- and medium-term Pi deficiency treatments (12),
whereas samples from long-term treatment were mineralized in
14% HNO
3
in a microwave system (MarsX, CEM) for the deter-
mination of macro- and microelements by ICP (ICP OES Vista
MPX, Varian).
31
P- and
13
C-NMR Spectroscopy. Metabolite analysis was performed
by NMR as described (18) on leaf extracts from the plants grown
in Petri dishes in a medium without sucrose for 20 days. The plants
were enriched with
13
C in a growth chamber that allowed accurate
regulation of the atmospheric gas composition and environmental
parameters. The growth chamber conditions were: 10-h photope-
riod with 100
␮
mol䡠m
⫺2
䡠s
⫺1
light, 22兾18°C day兾night temperature,
and 90% humidity. CO
2
(containing 10% of
13
CO
2
purchased from
Euriso-Top, Saint-Aubin, France) concentration in the chamber
was maintained at 350
␮
l䡠liter
⫺1
during the light period by auto-
matic injections to compensate for photosynthetic assimilation.
Results and Discussion
Phenotypic, Physiological, and Biochemical Modifications in Pi-
Starved Plants. The plants were grown on low Pi (5
␮
M) to reduce
the nonspecific effects of complete nutrient deficiency on growth.
The effects of Pi deficiency treatments on some of the morpho-
logical, biochemical, and physiological traits were evaluated. A
significant decline of 55% and 68% in soluble Pi content after 12
and 48 h of Pi deficiency, respectively (Fig. 1A) indicated a rapid
effect of Pi withdrawal from the growth medium. This response to
Pi deprivation could be due to small seed size of Arabidopsis with
low Pi re serves. Generally, the amount of metabolic reserve in a
seed is correlated with its size, which could be critical for seedling
survival during various environmental stresse s (19). However, a
decline in Pi content did not result in a significant accumulation of
anthocyanins (data not shown), which is one of the traits associated
with Pi deficiency (1). Although many of the well characterized Pi
starvation-induced genes including Pht1;4 (10, 11), are induced
during short- and medium-term Pi deficiency, the degree of Pi stress
may not be sufficient enough to elicit anthocyanin accumulation.
Accumulation of anthocyanins was observed in the leaves only
when the plants were subjected to long-term Pi deficiency (Fig. 1B).
In addition, there was a reduction in leaf size and a modification in
the root architecture (i.e., higher densities of lateral roots and root
hairs), typical responses of plants to Pi deficiency (20). Pi-deficient
plants also exhibited an early arrest of the primary root growth (0.3
cm䡠day
⫺1
and2cm䡠day
⫺1
in low and high Pi, respectively), whereas
secondary roots continued to grow (Fig. 1C). Long-term Pi defi-
ciency-induced modifications in morphological and biochemical
traits could be attributed to a decline in the total Pi content in the
leaves and the roots (Fig. 1D). A significant decline in the concen-
tration of K in leaves, and an appreciable increase in the concen-
tration of S in both the roots and the leaves, was also observed (Fig.
1D). Likewise, higher accumulation of some micronutrients (i.e.,
Fe, Zn) was observed during long-term Pi deprivation (Fig. 1E).
This observation suggests that, during Pi deficiency, the activity
and兾or utilization of other nutrients was altered. This finding is not
surprising considering the importance of Pi in numerous energy-
requiring metabolic and transport processes (21). Effects of long-
term Pi deficiency on various water-soluble metabolite s involved in
P and C metabolism were also examined.
13
C-NMR spectroscopy
analysis showed a reduction in the concentration of fumarate,
whereas glutamine and arginine increased in Pi-deficient plants
(Fig. 1F). The level of fumarate, a storage form of C, is known to
be affected during Pi deficiency (22). Accumulation of polyami-
nated glutamine and arginine in Pi-deficient plants could be an
adaptive response toward meeting the demand for N for protein
synthesis. The
31
P-NMR analysis also revealed a decrease in the
Misson et al. PNAS
兩
August 16, 2005
兩
vol. 102
兩
no. 33
兩
11935
PLANT BIOLOGY

concentrations of inorganic Pi and phosphorylcholine (results not
shown). Reductions in both inorganic Pi and soluble phosphory-
lated compounds in plants grown under Pi-deficiency have been
correlated with an inhibition of growth (23).
Microarray Analysis of the Spatio-Temporally Regulated Pi-Respon-
sive Genes.
ATH1 microarrays were used to evaluate the spatio-
temporal regulation of genes in response to Pi deficiency. During
short-term Pi deficiency (Table 2), 72 genes were induced, whereas
only four genes were suppressed (Fig. 2). These numbers increased
significantly (291 genes induced, 34 genes suppressed) during
medium-term Pi starvation (Table 3). At these two time points,
16% of the induced genes had overlapping expression, whereas only
1 gene was suppressed (Fig. 2). Furthermore, the induction (91
genes) or suppression (22 genes) of some genes was only transient.
This pattern of gene expression points to a very rapid but transient
change occurring even during short periods of Pi deficiency.
Modulation in the expression of the Pi-responsive genes correlated
with a decline of soluble Pi content during early stages of Pi
deficiency treatments (Fig. 1 A) as reported earlier (10). Long-term
Pi deprivation resulted in the differential regulation of 732 genes of
which 501 were induced [228 in the roots (Table 4) and 404 in the
leaves (Table 5)] and 231 were suppressed, 74 in the roots (Table
4) and 169 in the leaves (Table 5). Expression of 26.1% of the
induced and 4.8% of the suppressed genes overlapped in both leaves
and roots. Nevertheless, most of the genes were specific for either
roots or leaves, suggesting that different plant organs respond to Pi
deficiency by activating distinct sets of gene s. Comparison of the
microarray data from all of the three time points showed the
common induction of 48 genes and suppression of only one gene
(Fig. 2). These results are in agreement with results from the smaller
microarrays with Arabidopsis, rice, and white lupin showing similar
patterns of gene expre ssion (7–10). The differential expre ssion of
Pi-responsive genes is considered an adaptive response by plants to
Pi deficiency, which facilitate acquisition of sparingly available Pi
and concurrent attenuation of some of the energy-requiring met-
abolic pathways (21). Furthermore, differential regulation of the
closely related members of the purple acid phosphatase gene family
was found in the roots and leaves of long-term Pi-starved plants.
Likewise, the microarray discriminated among closely related mem-
bers of Pht1 family (Pht1;2 and Pht1;3), which share ⬎95% se-
quence homology (2). This finding clearly demonstrates one ad-
vantage conferred by the use of Affymetrix ATH1 microarray (6).
Validation of the ATH1 Microarray Data. The expression levels
determined from ATH1 arrays were confirmed by a combination
of qRT-PCR, Northern analysis, and promoter-reporter gene fu-
sion studies. Although qRT-PCR results were in general agreement
with the microarray data, quantitative difference s in the modifica-
tion of expression level of some of the genes (At5g05340,
At2g02990, and At5g56870) were observed (Fig. 3A). Extreme
expre ssion ratios of some of the genes that were also barely
expre ssed in one of the conditions (declared absent by the Af-
fymetrix analysis) have a poor quantitative significance (raw values
of microarray re sults are provided in Tables 2–5). Earlier microar-
ray analysis had also shown value s derived from qPCR generally
exceeded those from the array (24). Affymetrix technique was
found to be more sensitive than Northern analysis (e.g., At5g56870
transcripts in the leaves, Fig. 3B). In addition, GFP fused to the
promoter of AtS6K2 [induced in roots of P(⫺) plants] revealed in
Fig. 1. Effects of Pi deficiency on
some of the traits of Arabidopsis.
(A) Soluble Pi in whole plant after
transfer to low (open bars) or high
(filled bars) Pi. (B) Anthocyanin ac-
cumulation in the leaves in re-
sponse to Pi deficiency. (C) second-
ary兾primary root length ratio in
low-Pi (open bars) or high-Pi (filled
bars). Macroelements (D) and mi-
croelements (E) in the high Pi (black
bars) and low Pi (white bars) leaves
and high Pi (dark gray bars) and low
Pi (light gray bars) roots. (F)
13
C-
NMR analysis. fum, fumarate; Glu,
glutamate; Gln, glutamine; Arg, ar-
ginine; ref, reference (500
␮
mol
maleate). Upper traces correspond
to the enlargement of the corre-
sponding spectrum in the rectangle
in the lower trace.
Fig. 2. Number of genes induced (A) or repressed (B) in low Pi. Comparison of
short-term (diamonds), medium-term (squares), and long-term (circles) experi-
ments. The whole plant was analyzed to monitor the regulation of Pi-responsive
genes during short- and medium-term experiments. Results from leaves and roots
analyzed separately were mixed as long term experiment.
11936
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0505266102 Misson et al.

vivo gene expre ssion restricted only to the stele and emerging lateral
roots of Pi-deprived plants (Fig. 3C). This result illustrates the high
sensitivity of the Affymetrix technique, which allowed the evalua-
tion of Pi-responsive genes that are either expre ssed at low levels or
restricted to specific cell type s and substantiates the robustness of
transcript profiling with Affymetrix arrays (6).
Functional Classification of Pi-Responsive Genes with Altered Expres-
sion Patterns.
Pi-responsive genes were categorized into different
functional groups (Table 7). Approximately 84% and 75% of the
induced and suppressed genes, re spectively, had known functions.
The first group contained genes that are related. In addition to
genes harboring general metabolic functions, this study highlights
the involvement of genes related to the uptake and transport of Pi
and other inorganic ions and to the Pi salvage systems. Among the
genes coding the Pht1 family of Pi transporters (2), Pht1;4 was
induced during short-, medium-, and long-term treatments, indi-
cating a rapid and sustained induction of this gene in response to
Pi deprivation, which is consistent with its proposed role in both
acquisition and mobilization of Pi (11, 12). Some of the members
of this family were induced only during medium- and兾or long-term
Pi-deficiency (Fig. 5A). Spatio-temporal regulation of the gene s of
Pht1 family indicates an apparent lack of functional redundancy
among family members (2). Induction of Pht3;2, encoding a mito-
chondrial ATP兾ADP antiporter, during medium- and long-term Pi
deprivation, was shown (Fig. 5A). Genes involved in Pi mobilization
from organic compounds such as the gene encoding glucose-6P兾Pi
translocator were induced only upon prolonged Pi deficiency. These
‘‘late’’ genes are thought to play a role in promoting the efficient use
of Pi by the plant (25). It is therefore evident from the microarray
results that there is Pi-deficiency-induced spatio-temporal regula-
tion of genes involved in Pi acquisition and its mobilization; complex
mechanisms necessary for plants to thrive under Pi deficiency. Late
induction of PHO1;H1, which is involved in the loading of Pi into
the xylem vessels (26), supports this view. Furthermore, there was
induction of genes coding sulfate transporters (SULTR 1;3, SULTR
3;4) that may have facilitated the higher uptake of S during Pi
deficiency (Fig. 1D), possibly to meet the demand for increased
sulfolipid synthe sis and for balancing the anion兾cation ratio in the
absence of Pi ions (27). Mechanisms involved in maintenance of Fe
homeostasis (28), such as the differential and coordinated suppres-
sion of the iron transporter IRT1 in the roots and induction of
AtFER1 (encoding a protein involved in iron storage in the chlo-
roplast) in leaves, reflected the plant response to Fe overload
induced by Pi-deficiency. These expression data correlated with the
elevated concentrations of Fe, and other metals observed in
Pi-deficient plants (Fig. 1E) may be linked to increased availability
in the medium in absence of Pi. This theory was confirmed by
running the chemical equilibrium model (
VISUAL MINTEQ version
2.30). In addition, genes for several metal and ATP-binding cassette
transporters were induced in Pi–deprived plants (Tables 6 and 7).
This finding suggests that one adaptive response to Pi deficiency by
the plant is to enhance its capacity to absorb and scavenge Pi-
complexing metals to release sequestered Pi from the medium.
Another group of genes identified by our experiment is involved
in triggering Pi-salvage via the conversion of organic phosphorus
into available Pi (4, 29). Of the 29 genes that encode different
purple acid phosphatases in Arabidopsis, 27 were pre sent on the
ATH1 microarray analysis, and 11 of them were induced by Pi
deficiency (Fig. 4B). In addition, one of the ribonuclease genes
(RNS1) was induced early in response to Pi deprivation, suggesting
that ribonucleases could play a role in the remobilization of P during
Pi deprivation (4). Induction of pyrophosphate-dependent phos-
phofructo-1-kinase and nucleotide pyrophosphatase genes during
medium- and long-term Pi deficiency treatments also repre sent an
important mechanism to reduce Pi demand in plant organs (21).
The distinct induction of the expression of genes for glycerol-3-P
permease s also points toward the complexity of the various adap-
tive modifications that occur in the plant for optimal utilization of
Pi under limiting conditions.
Fig. 3. Validation of ATH1 results. (A) Comparison of
chip results and q-PCR. P(⫺)兾P(⫹) ratio in leaves and
roots for some selected genes are shown. Two differ-
ent scales were used for the genes that were up- or
down-regulated. Q-PCR and ATH1 results are means
and SD of three assays performed on triplicates. (B)
Northern blot analysis and chip results [ratio P(⫺)兾
P(⫹)] of Pi-responsive genes in leaves (L) and roots (R).
Plants were harvested after 5, 10, and 15 days of
transfer in low [P(⫺)] or high Pi [P(⫹)], and 10
␮
gof
total RNA, isolated from the whole plant, was hybrid-
ized with
32
P-labeled cDNA fragments of the genes.
Equivalence of RNA loading in all of the lanes is shown
by
32
P-labeled tubulin hybridization and ethidium bro-
mide-stained rRNA (Lower). (C) Expression of GFP
fused to AtS6K2 (At3g08720) promoter in the stele and
emerging lateral roots of transgenic plants grown un-
der high [P(⫹)] and low Pi [P(⫺)]. (Scale bar, 50
␮
m.)
Misson et al. PNAS
兩
August 16, 2005
兩
vol. 102
兩
no. 33
兩
11937
PLANT BIOLOGY

Detailed analysis of Pi-re sponsive genes also revealed that ⬇7%
(44 genes) are involved in lipid biosynthetic pathways (Table 6). Of
these genes, only two were suppressed. About 50% of the lipid
related genes were induced within 2 days of Pi deprivation. Induced
genes largely represented those coding for enzyme s involved in
phospholipid degradation and galacto- and sulfolipid synthesis (Fig.
4). Interestingly, only a few of the genes coding for phospholipases
D (At3g05630) and C (At3g03540) were induced during Pi defi-
ciency. These results suggest a role for these genes in the lipid
metabolic pathway during Pi deficiency. Genes involved in the
subsequent utilization of DAG to synthesize galactolipids (MGDG,
DGDG) were strongly up-regulated at early stages of Pi depriva-
tion, which is consistent with previous data (30). Genes coding for
MGDG synthases (MGD2 and MGD3) were induced 4- to 10-fold
during short-term Pi deprivation; whereas expression of DGD1 and
DGD2, coding for DGDG synthases, was enhanced during medi-
um- and long-term Pi deficiency, re spectively. Similarly, the genes
enc oding UDP glucose-4-epimerase and UDP galactose-4-
epimerase, which convert UDP-glucose to UDP-galactose (galac-
tolipid precursor), were induced during medium- and long-term Pi
deficiency. This induction could facilitate the production of galac-
tose required for galactolipid synthesis. Comparatively, the genes
coding for UDP-sulfoquinovose synthase and UDP-sulfoquinovo-
syl:DAG sulfoquinovosyltransferase exhibited early and sustained
induction during Pi deficiency treatments; this was reflected by a
4-fold increase in the level of SQDG in P(⫺) leaves during
long-term Pi deficiency. Although SQDG is not considered essen-
tial for plant development (31), under Pi deficiency it could possibly
replace PG and may allow photosynthesis to continue despite a
reduction in phospholipid content. Arabidopsis mutants defective in
sulfolipid synthase show impaired growth during Pi deprivation
(31). These modulations of lipid biosynthetic pathways indicate a
complex mechanism to replace membrane phospholipids with
nonphosphorus galacto- and sulfolipids that may have evolved to
scavenge and conserve Pi in plants under Pi-limiting conditions
(30). These results correlated with variations in phospholipids,
sulfolipids and glycosylglyceride s (Fig. 4). Alteration of the lipid
content became apparent within 2 days, whereby a decrease of PG
and PC was compensated by an increase of SQDG and DGDG. In
leaves of plants grown in Pi-deficient medium, reduction in the
levels of all phospholipids except DPG was observed. Interestingly
in the P(⫺) roots no significant difference was detected in any of
the phospholipid species including PC, but there was a substantial
increase in the level of DGDG. This finding suggests that lipid
composition is more sensitive to Pi deficiency in leaves than in roots,
as indicated earlier (16). Despite an early induction of MGD2 and
MGD3, there was no significant increase in MGDG level, even
during long-term Pi deficiency; this may be due to rapid conversion
of MGDG into DGDG by DGD1 and DGD2, whose activity
increased during long-term Pi deficiency. Furthermore, the genes
DGD1 and DGD2 exhibited differential regulation in roots and
leaves. Earlier studie s have also shown the preferential biogenesis
of DGD2 and DGD1 outside of plastids and in the chloroplast
membrane, respectively (30, 32). The pre sent microarray analysis
revealed an early, sustained, and coordinated induction of a host of
Pi-responsive genes involved in Pi acquisition and conversion of
organic phosphorus into available Pi. These experiments also
indicated that Pi deprivation is perceived at the molecular level as
soon as Pi is withdrawn from the medium, suggesting that the plant
is able to sense decrease of Pi concentration either in the medium
or in cells.
Because accumulation of anthocyanin is a characteristic response
of plants to long-term Pi deficiency (Fig. 1B), microarray data were
also evaluated for the differential regulation of genes involved in its
biosynthesis (Fig. 6). At each step of the anthocyanin biosynthetic
pathway leading to the synthesis of cyanidin, pelargonidin, and
flavonoids, at least one gene was found to be induced. Although the
majority of these genes were induced only during long-term Pi
deficiency treatment and more specifically in the leaves, a few of
them encoding flavonol 3-O-glucosyltransferase were significantly
up-regulated after 2 days of Pi deficiency. On average, genes were
induced 3- to 4-fold, with the notable exceptions of dihydroflavo-
nol-4-reduct ase, anthoc yanindin synthase, and flavonone-3-
hydroxylase, which showed 5- to 6-fold induction. Moreover, two
genes coding for anthocyanin 5-aromatic acyltransferase and one
for anthocyanin2 were induced (Table 7). Interestingly, in Pi-
deficient leaf samples, there was an induction of a gene coding for
chalcone synthase, which is responsible for the conversion of
4-coumarate to naringenin in flavonoid biosynthetic pathway (Fig.
6). Accumulation of naringenin has been shown to affect the
transport of auxin to roots (33), which could be re sponsible for an
altered root architecture, a hallmark of plant response s to Pi
deficiency (34). In addition, the expression of some of the genes
associated with phytohormone responses was also altered during Pi
deficiency. For instance, genes involved in auxin response were
modulated during medium-and long-term Pi-deprivation (Table 7).
Fig. 4. Transcriptional regulation in
pathway of glycosylglyceride biosyn-
thesis in short-term (squares), medium-
term (diamonds), and long-term
[leaves (circles) and roots (triangles)
analyzed separately] experiments.
Fold change: red, ⬎10; orange, 4–10;
yellow, 2– 4; white, 0.5–2; green, 0.25–
0.5; pale blue, 0.1– 0.25; dark blue,
⬍0.1. *, Significant, one-way ANOVA,
P ⫽ 0.05. (Insets) Metabolite quantifi-
cation [% of total lipids in P(⫹) and
P(⫺) and ratio P(⫺)兾P(⫹)] in short-,
medium- and long- (leaf, root) term
treatments. DPG, diphosphatidylglyc-
erol; PG, phosphatidylglycerol; PI,
phosphatidylinositol; PE, phosphati-
dylethanolamine; PC, phosphatidyl-
choline; SQDG, sulfoquinovosyldiacyl
glycerol; MGDG, galactosyl-1,2-diacyl-
glycerol; DGDG, digalactosyl-diacyl-
glycerol.
11938
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0505266102 Misson et al.

A gene for an auxin response element, which was induced in
Pi-deficient roots, is known to be involved in transcriptional regu-
lation (35). Likewise, ethylene-response genes also showed differ-
ential regulation in response to Pi deprivation. The role of auxin and
ethylene in the Pi-starvation response is well established and thus
it is not surprising to see the changes in the expre ssion of genes
involved in the response pathways for these hormone s.
Pi deficiency is known to cause significant stress to the plant and
this was reflected in modulation of an array of stress-related genes
(Tables 6 and 7). A large number of them are related to disease or
pathogen resistance (e.g., chitinases, PR-proteins) and toxin catab-
olism (GSTs). General features of oxidative stress [superoxide
dismutase (SOD), peroxidase, GST, cytochrome P450] are also
strongly induced in Pi-deficient samples mainly in long-term ex-
periments (Table 7). This finding suggests interactions among
several stre ss-related pathways that may have functional implica-
tions in plant survival under Pi deficiency. Pi starvation also
resulted in spatio-temporal expre ssion of the genes (10 induced and
two suppressed) involved in controlling the level of reactive oxygen
species (ROS) (Table 7). They encode different ROS-scavenging
enz y mes such as SOD, monodehydroasc orbate reduct ase
(MDAR), glutathione peroxidase (GPX). One of the genes coding
for NADPH oxidase-like enzyme that is responsible for generating
ROS showed a transient suppression during medium-term Pi
deprivation. Recent microarray analysis of Arabidopsis, subjected to
different abiotic stre sses, had also demonstrated differential regu-
lation of 152 genes coding for different enzymes involved in
scavenging and generation of ROS (36).
In Pi-limiting conditions, genes coding for enzyme s involved in
protein degradation (six genes) and protein biosynthesis (21 genes)
are induced and suppressed, respectively (Table 7), sugge sting that
initiation of Pi recycling processe s. Furthermore, gene s coding for
protein phosphatase s and kinase s were up-regulated during Pi
deficiency. Some of the protein kinase genes were also found to be
down-regulated in leaves and roots of Pi-deficient plants (Table 7).
During medium-and long-term Pi deficiency, modulation in the
expre ssion of the genes encoding various enzymes involved in cell
wall metabolism was observed, and this finding is in conformity with
earlier microarray analysis (9). Majority of these induced genes (2-
to 10-fold) encoded enzymes like xyloglucan endo-1,4-
␤
-D-
glucanase, polygalacturonase inhibiting protein-1, put ative
pectinesterase, (1,4)-
␤
-mannan endohydrolase precursor, and po-
lygalacturonase inhibiting protein. A few genes for xylosidase or
cellulose synthase-like protein were found to be down-regulated.
Aside from the production of galactose to support increased
synthesis of MGDG and DGDG, increased galactose may also have
a critical role in cell wall biosynthesis of the modified root archi-
tecture, as shown by transcript analysis of lateral root induction (37).
Analysis of long-term Pi-deprived plants revealed that genes for
these enzymes were mainly modulated in roots.
Because Pi deficiency responses are known to be regulated at the
transcription level (1), microarray data were further analyzed for
the Pi-deficiency-induced genes encoding transcription regulatory
elements. The expre ssion of a total of 80 genes, presumed to be
associated with transcriptional regulation of gene expression, was
altered during Pi deficiency (Table 6). A few of them were
up-regulated during short-term (five gene s) and medium-term (10
genes) Pi deficiency. However, their induction was more pro-
nounced (47 genes) during long-term Pi deprivation. Interestingly,
the induction of only a small number of transcription factor genes
overlapped during different stage s of Pi deficiency. This finding
suggests that specific sets of transcription factors are involved in
regulating early and late response s of plants to Pi deficiency. To gain
further insight into the mode of regulation of Pi-responsive genes,
the conserved sequences located upstream (⫺1to⫺2,000 bp) of the
ATG start codon were analyzed (Tables 8 and 9). Promoters of the
Pi-responsive genes coding for Pi transporters, phosphatase s, and
those involved in protein synthesis were found to be significantly
enriched with the PHR1 binding sequence, which is recognized by
a MYB-domain-containing transcription factor (5). Furthermore,
deduced protein sequence of some of the Pi-responsive gene s
harbored an SPX domain, which has been identified in proteins
involved in either transport or sensing of Pi (26).
Induction of phosphatase s and kinase s further suggests the
involvement of numerous posttranslational modifications, which
remain to be identified. This study thus presents a global analysis of
plant transcriptomic responses to Pi deficiency and physiological
and biochemical correlations with observed phenotypes. The list of
putative targets established will significantly add to our knowledge
about the complex molecular processes associated with Pi nutrition.
We thank P. Auroy and P. Richaud for the analysis of ion ic content; M.
Pean, S. Boiry, and the GRAP team for plant growth and
13
C enrichment
experiments; D. Varadarajan for Arabidopsis culture; K. Ramonell and
L. Rose for the chip hybridization and initiation to Genespring; M. H.
Montane and A. Nublat for their help for the qPCR technique; and A. M.
Boisson and E. Gout for the NMR analysis. This project was supported
in part by a grant from Commissariat a` l’Energie Atomique and region
Provence-Alpes-Coˆte d’Azur (to J.M.) and U.S. Department of Agri-
culture (to K.G.R.).
1. Raghothama, K. G. (1999) Annu. Rev. Plant Physiol. Plant Mol. Biol. 50, 665–693.
2. Mudge, S. R., Rae, A. L., Diatlof f, E. & Smith, F. W. (2002) Plant J. 31, 341–353.
3. Baldwin, J. C., Karthikeyan, A. S. & Raghothama, K. G. (2002) Plant Physiol. 125, 728–737.
4. Bariola, P. A., Howard, C. J., Taylor, C. B., Verburg, M. T., Jaglan, V. D. & Green, P. J.
(1994) Plant J. 6, 673–685.
5. Rubio, V., Francisco, L., Roberto, S., Ana, C., Martin, J. I., Antonio, L. & Paz-Ares, J. (2001)
Gene Dev. 15, 2122–2133.
6. Redman, J. C., Haas, B. J., Tan imoto, G. & Town, C. D. (2004) Plant J. 38, 545–561.
7. Hammond, J. P., Bennett, M. J., Bowen, H. C., Broadley, M. R., Eastwood, D. C., May, S. T.,
Rahn, C., Swarup, R., Woolaway, K. E. & White, P. J. (2003) Plant Physiol. 132, 578–596.
8. Uhde-Stone, C., Zinn, K. E., Ramirez-Yanez, M., Li, A., Vance, C. P. & Allan, D. L. (2003)
Plant Physiol. 131, 1064–1079.
9. Wasaki, J., Yonetani, R., Kuroda, S., Shinano, T., Yazaki, J., Fujii, F., Shimbo, K.,
Yamamoto, K., Sakata, K., Sasaki, T., et al . (2003) Plant Cell Environ. 26, 1515–1523.
10. Wu, P., M, a L., Hou, X., Wang, M., Wu, Y., Liu, F. & Deng, X. W. (2003) Plant Physiol.
132, 1260–1271.
11. Karthikeyan, A. S., Varadarajan, D. K., Mukatira, U. T., D’Urzo, M. P., Damsz, B. &
Raghothama, K. G. (2002) Plant Physiol. 130, 221–233.
12. Misson, J., Thibaud, M. C., Bechtold, N., Raghothama, K. G. & Nussaume, L. (2004) Plant
Mol. Biol. 55, 727–741.
13. Schaffer, R., Landg raf, J., Accerbi, M., Simon, V., Larson, M. & Wisman, E. (2001) Plant
Cell 13, 113–123.
14. Bechthold, N., Elis, J. & Pelletier, G. (1993) C. R. Acad . Sci. Paris 316, 1194–1199.
15. Folch, J., Lees, M. & Stanley, G. H. S. (1957) J. Biol. Chem. 226, 497–509.
16. Jouhet, J., Marechal, E., Bligny, R., Joyard, J. & Block, M. A. (2003) FEBS Lett. 544, 63– 68.
17. Lange, H., Shropshire, W., Jr., & Mohr, H. (1971) Plant Physiol. 47, 649–655.
18. Aubert, S., Bligny, R. & Douce, R. (1996) in Current Topics in Plant Physiology, eds.
Shashar-Hill, Y. & Pfeffer, P. (Am. Soc. Plant Physiol., Rockv ille, MD), pp. 109–154.
19. Westoby, M., Leishman, M. & L ord, J. (1997) in Plant Life Histor ies: Ecology, Phylogeny, and
Evolution, eds. Silvertown, J. W., Franco, M. & Harper, J. L. (Cambridge Univ. Press,
Cambridge, U.K.), pp. 143–162.
20. Ma, Z., Bask in, T. I., Brown, K. M. & Lynch, J. P. (2003) Plant Physiol. 131, 1381–1390.
21. Plaxton, W. C. & Carswell, M. C. (1999) in Plant Responses to Environmental Stresses: From
Phytohormones to Genome Organization, ed. Lerner, H. R. (Dekker, New York), pp. 349–372.
22. Chia, D. W., Yoder, T., Reiter, W.-D. & Gibson, S. (2000) Planta 211, 743–751.
23. Gout, E., Boisson, A. M., Aubert, S., Douce, R. & Bligny, R. (2001) Plant Physiol. 125, 912–925.
24. Wang, R., Okamoto, M., Xing, X. & Crawford, N. M. (2003) Plant Physiol. 132, 556–567.
25. Hammond J. P., Broadley, M. R. & White, P. J. (2004) Ann. Bot. 94, 323–332.
26. Wang, Y., Ribot, C., Rezzonico, E. & Yves Poirier, Y. (2004) Plant Physiol . 135, 400 –411.
27. Essigmann, B., Guler, S., Narang, R. A., Linke, D. & Benning, C. (1998) Proc. Natl . Acad.
Sci. USA 95, 1950–1955.
28. Petit, J. M., Briat, J.-F. & L obre´aux, S. (2001) Biochem J. 359, 575–582.
29. Duff, S. M. G., Sarath, G. & Plaxton, W. C. (1994) Physiol. Plant. 90, 791–800.
30. Benning, C. & Otha, H. (2005) J. Biol. Chem. 280, 2397–2400.
31. Yu, B., Xu, C. & Benning, C. (2002) Proc. Natl. Acad. Sci. USA 99, 5732–5737.
32. Jouhet, J., Mare´chal, E., Baldan, B., Bligny, R., Joyard, J. & Block, M. A. (2004) J. Cell Biol.
167, 863–874.
33. Brown, D. E., Rashotte, A. M., Murphy, A. S., Normanly, J., Tague, B. W., Peer, W. A., Taiz,
L. & Muday, G. K. (2001) Plant Physiol. 126, 524–535.
34. Lo´pez-Bucio, J., Herna´ndez-Abreu, E., Sa´nchez-Caldero´n, L., Nieto-Jacobo, M. R., Simp-
son, J. & Herrera-Estrella L. (2002) Plant Physiol. 129, 244–256.
35. Ulmasov, T., Hagen, G. & Guilfoyle, T. J. (1999) Plant J. 19, 309 –319.
36. Mittler, R., Vanderauwera, S., Gollery, M. & Van Breusegem, F. (2004) Trends Plant Sci. 9, 490– 498.
37. Brinker, M., van Zyl, L., Liu, W., Craig, D., Sederoff, R. R., Clapham, D. H. & von Arnold,
S. (2004) Plant Physiol. 135, 1526–1539.
Misson et al. PNAS
兩
August 16, 2005
兩
vol. 102
兩
no. 33
兩
11939
PLANT BIOLOGY