Transcriptome analysis of sulfur depletion in Arabidopsis thaliana: interlacing of biosynthetic pathways provides response specificity

Authors

  • Victoria Nikiforova,

    1. Department 1 of L. Willmitzer, Max Planck Institute of Molecular Plant Physiology, Am Mühlenberg 1, 14476 Golm, Germany, and
    Search for more papers by this author
  • Jens Freitag,

    1. Department 1 of L. Willmitzer, Max Planck Institute of Molecular Plant Physiology, Am Mühlenberg 1, 14476 Golm, Germany, and
    Search for more papers by this author
  • Stefan Kempa,

    1. Department 1 of L. Willmitzer, Max Planck Institute of Molecular Plant Physiology, Am Mühlenberg 1, 14476 Golm, Germany, and
    Search for more papers by this author
  • Monika Adamik,

    1. Department 1 of L. Willmitzer, Max Planck Institute of Molecular Plant Physiology, Am Mühlenberg 1, 14476 Golm, Germany, and
    Search for more papers by this author
  • Holger Hesse,

    1. Freie Universität Berlin, Institut für Biologie, Angewandte Genetik, Albrecht-Thaer-Weg 6, 14195 Berlin, Germany
    Search for more papers by this author
  • Rainer Hoefgen

    Corresponding author
    1. Department 1 of L. Willmitzer, Max Planck Institute of Molecular Plant Physiology, Am Mühlenberg 1, 14476 Golm, Germany, and
      For correspondence (fax +49 331 5678201; e-mail hoefgen@mpimp-golm.mpg.de).
    Search for more papers by this author

For correspondence (fax +49 331 5678201; e-mail hoefgen@mpimp-golm.mpg.de).

Summary

Higher plants assimilate inorganic sulfate into cysteine, which is subsequently converted to methionine, and into a variety of other sulfur-containing organic compounds. To resist sulfur deficiency, plants must demonstrate physiological flexibility: the expression of an extensive set of genes and gene regulators that act in the affected pathways or signalling cascades must be delicately tuned in response to environmental challenges. To elucidate this network of interactions, we have applied an array hybridisation/transcript profiling method to Arabidopsis plants subjected to 6, 10 and 13 days of constitutive and induced sulfur starvation. The temporal expression behaviour of approximately 7200 non-redundant genes was analysed simultaneously. The experiment was designed in a way to identify statistically significant changes of gene expression based on sufficient numbers of repeated hybridisations performed with five uniform pools of plant material. The expression profiles were processed to select differentially expressed genes. Among the 1507 sulfur-responsive clones implicated in this way, 632 genes responded specifically to sulfur deficiency by significant over-expression. The sulfur-responsive genes were grouped according to functional categories or biosynthetic pathways. As expected, genes of the sulfur assimilation pathway were altered in expression. Furthermore, genes involved in flavonoid, auxin, and jasmonate biosynthesis pathways were upregulated in conditions of sulfur deficiency. Based on the correlative analysis of gene expression patterns, we suggest that a complex co-ordination of systematic responses to sulfur depletion is provided via integration of flavonoid, auxin and jasmonate pathway elements. Plait concept for transduction of specificity via the main non-specific signalling stream is proposed.

Introduction

Plant growth and crop yield are dependent on a variety of abiotic and biotic factors. Nutrient acquisition from soil is a major challenge for plants because nutrient distribution and availability is not uniform in time or space. This even holds true under agricultural conditions when artificial or organic fertilisers are supplied. As plants are immobile, they have evolved a variety of molecular and physiological mechanisms with which to attain the nutrients they need.

Under optimal conditions, nutrient uptake is achieved via specialised transport systems at the root–soil interface, such as sulfate transporters (Takahashi et al., 2000), nitrate or ammonium transporters (reviewed by Howitt and Udvardi, 2000; Williams and Miller, 2001) and phosphate translocators (reviewed by Flügge, 1999). Nutrient paucity, especially of the macronutrients S, N and P, triggers the induction of high affinity transport systems that allow uptake of low concentrations of ions against concentration gradients (Fischer et al., 1995; Frommer et al., 1995). In parallel, increased root growth is a physiological adaptation that allows plants to actively explore new resources to acquire needed nutrients (Forde and Lorenzo, 2001).

However, if these programmed adaptation mechanisms do not provide sufficient nutrient supply, the immobile plant is forced to respond with further adaptive metabolic strategies to safeguard survival or termination of the life cycle (e.g. through premature ripening, or reduced seed setting). These metabolic changes in response to nutrient deprivation involve alterations in expression levels of a number of genes (Ohlrogge and Benning, 2000).

Sulfur is one of the essential macroelements for plant life with numerous biological functions (Leustek et al., 2000). Plants, together with bacteria and fungi, are able to assimilate inorganic sulfur in form of sulfate from the soil or sulfur dioxide and hydrogen sulfide gases from the atmosphere. In the initial step of the biological sulfur cycle, plants assimilate and reduce inorganic sulfate to sulfide and then incorporate it into cysteine, the first organic molecule carrying reduced sulfur. Cysteine, besides being an important proteinogenic amino acid, is subsequently converted to a huge number of derived compounds such as methionine, glutathione, S-adenosylmethionine, S-methylmethionine and biotine.

Plant response to sulfur-deficiency stress demands physiological flexibility involving sensing mechanisms, transduction of perceived signal(s) to the control point(s) of assimilatory regulation, and triggering and silencing of responsive gene systems. The delicate tuning of these complex tasks requires the involvement of an extensive set of genes and their regulators and/or regulating cascades. The analysis of the behaviour of individual genes in response to sulfur depletion is helpful for gathering knowledge of single regulatory units, but to understand how components of the network interact, it is necessary to analyse the temporal and spatial transcription behaviour of all (or most) of the genes in the genome (the transcriptome) simultaneously.

In the past 2 years, transcription profiling data suggesting a general pattern of plant stress response have accumulated (Bohnert et al., 2001; Kawasaki et al., 2001– salt stress; Desikan et al., 2001– oxidative stress; Negishi et al., 2002; Thimm et al., 2001; Wang et al., 2000, 2001– nutrient deficiency; Reymond et al., 2000– mechanical wounding and insect feeding; Reymond, 2001; Rowland and Jones, 2001; Scheideler et al., 2002; Schenk et al., 2000; Stintzi et al., 2001– plant defence; Seki et al., 2001– drought and cold stress). Different stresses are first perceived and transduced through specific mechanisms, while later converging into more general stress behaviour. To achieve higher effectiveness, the final stress response pathways diverge to effect particular responses. This concept poses the question of how the specificity is transduced from specific sensing through non-specific stream to obtain particular responses. Comparative analysis of transcription profiling data reflecting changes induced by various stresses may confirm the viability of this strategy, help us to locate points of convergence–divergence within networks, and clarify how specific responses can be elicited via a non-specific signalling stream.

In order to investigate the complex network of adaptive gene regulation in response to sulfur depletion and its relationship to general stress response pathways, we profiled time-dependent gene expression of Arabidopsis thaliana seedlings grown on sulfur-depleted medium at four time points using array hybridisation. This system's approach has enabled us to describe the adaptation of the transcriptome to sulfur depletion. Known contributors to the gene response system (sulfate transporters for example) served as landmarks to calibrate data and led us towards new regulatory circuits.

Results

Sulfur-deficient status of plants: physiological and metabolic phenotype

The first morphological differences – slight growth retardation and chlorosis – between plants on normal (+S) and sulfur-deficient medium (–S) were visible at day 11 of sulfur depletion in experiment 1 (constitutive sulfur starvation) and at day 7 of sulfur depletion in experiment 2 (induced sulfur starvation after pre-germination on sulfur-sufficient medium). The time course of morphological changes on sulfur-deficient medium showed high reproducibility in repeated experiments. Thus, days 10 and 6 were chosen to be the first time points for experiments 1 and 2, respectively (referred further as experiments 1.1 and 2.1) because at these time points the plants did not show visible differences to control plants. In the subsequent 3 days, the following symptoms were observed in plants growing on sulfur-deficient medium: plants were retarded in growth, leaves became first pale green, then pale yellow, then reddish; later whole plants starting from petioles became purple (Figure 1a). Additionally, at the second time point (referred further as experiments 1.2 and 2.2) the plants had a more extensive root system in comparison to control plants. This observation was confirmed in further experiments when Arabidopsis plants were grown on sulfur-deficient medium in hydroponic culture. Here, after 4 weeks at –S, the root-to-shoot mass ratio was significantly higher in treated plants (0.67) in comparison to the control (0.40).

Figure 1.

Experimental design.

(a) Arabidopsis seedlings, grown on sulfur-sufficient (left) and sulfur-deficient medium for 14 and 22 days.

(b) Experimental scheme of sulfur starvation showing number of days at sulfur deficient conditions for four experimental points: 1.1, 1.2, 2.1 and 2.2. In experiment 1 (Exp 1.1 and 1.2), the seeds were sown directly in –S medium, while in experiment 2 (Exp 2.1 and 2.2) the seeds were first sown in normal medium (+S), then the seedlings were transferred to –S medium (Exp, experiment).

Under conditions of induced sulfur starvation after pre-germination on sulfur-sufficient medium (experiment 2), the symptoms developed earlier than in experiment 1, where the plants were grown from seeds germinated on sulfur-deficient medium.

To confirm the metabolic status of the plants starving from sulfur deficiency, the total content of elemental sulfur in these plants was measured. After 6 days on sulfur-deficient medium, plants still contained 43% of the sulfur level measured in control plants (experiment 2.1). By day 10 at –S, sulfur went further down to 30% (experiment 1.1 and 2.2). At the latest time point, day 13 of sulfur depletion (experiment 1.2), plants contained only 27.2% residual sulfur (Figure 2a).

Figure 2.

Levels of sulfur and sulfur-related metabolites in Arabidopsis seedlings grown on sulfur-sufficient and sulfur-deficient medium. Values ±SD characterise the average of five independent repetitions (Exp, experiment, DWT, dry weight, FWT, fresh weight, RR, relative response).

Thiol content was determined as a further indicator for sulfur nutrient status (as was shown earlier for glutathione in Brassica napus L. by Blake-Kalff et al. (1998) and Hordeum vulgare by Vidmar et al. (1999)). In Arabidopsis, under our experimental conditions of sulfur depletion, glutathione dropped 7 to 15-fold, and cysteine, another major thiol group-containing compound, was reduced to non-measurable levels (Figure 2b,c).

Under conditions of sulfur depletion, internal levels of serine (Figure 2d), O-acetylserine (Figure 2e) and tryptophan (Figure 2f) were significantly increased. The relative response for serine was 2–3-fold higher in comparison with internal serine levels in control plants grown under conditions of sulfur sufficiency. For O-acetylserine this difference was even higher and accounted for 5–37-fold. The increase of O-acetylserine content was also shown by Awazuhara et al. (2000) in rosette leaves of Arabidopsis plants subjected to sulfur starvation for 9 weeks. Levels of tryptophan in sulfur-starved plants increased 6–28-fold compared to control plants.

Among the compounds measured after 48 h of sulfur depletion (Hirai et al., 2003), levels of sulfate, O-acetylserine, cysteine and glutathione were altered correspondingly: sulfate and glutathione decreased, O-acetylserine increased.

Sulfur-deficient status of plants: transcript profile

Using array hybridisation experiments, we examined transcript levels of 16 128 Arabidopsis EST clones representing approximately 7200 individual genes, corresponding to about 30% of the total Arabidopsis genome. For 15 051 clones, the absolute transcript levels in five repetitions of independently sampled material were determined. The rest of the clones were excluded from further considerations due to the normalisation procedure, which was performed with the mathematical tools incorporated in the Haruspex database (http://www.mpimp-golm.mpg.de/haruspex). The changes between transcript levels of plants grown on normal medium as a control and on sulfur-deficient medium were analysed by calculating the ratio (R) of the average transcript level from five repetitions on sulfur-deficient medium to the average transcript level from five repetitions on control sulfur-sufficient medium for each clone. Statistical significance of the differences in expression levels was analysed with t-test. In all analyses the difference was considered significant with a probability of P < 0.05. For all EST clones, R more than 2.5 or less than 0.4 with P < 0.05 at least in one of the four experimental points were the criteria used to assign statistically significant differential expression. These parameters were used to identify genes responding significantly to the experimental conditions of sulfur depletion, which we termed as sulfur-responsive genes. Expression levels and ratios for genes encoding transcriptional factors were generally lower than those of other functional groups. Therefore, we applied slightly less strict parameters to this group of genes and included genes with less marked changes in expression. Thus, these genes were considered to be sulfur responsive, if R was more than 2 or less than 0.5, with P < 0.05 at least in one of the four experimental points. When applying described selection criteria to the whole dataset containing expression levels of 15 051 clones at all four experimental points, 1507 sulfur-responsive clones were revealed. Among them 826 EST clones totally and 80 clones referring to transcriptional factors exhibited over-expression in sulfur-deficient conditions. Respectively, 681 clones and four transcriptional factors were downregulated during sulfur starvation.

The transcription profiles of the four experimental points were analysed using scatterplots in which all differentially expressed clones (P < 0.05) were plotted; the results are summarised in Table 1. Changes in the numbers of sulfur-responsive clones with increasing duration of sulfur starvation indicate a shift from controlled regulation of gene expression to a general loss of gene activities. After 6 days on sulfur-deficient medium, the differentially expressed genes were distributed more or less equally between the groups of over-expressed and downregulated clones (153 and 141 clones, respectively, with differences higher than 2.5-fold; experiment 2.1). After 10-day growth on sulfur-deficient medium, the majority of the differentially expressed genes were over-expressed, indicating a maximum of gene activation in response to the challenge of sulfur deficiency (experiments 2.2 and 1.1). By day 13 of sulfur depletion, the highest number of clones (2749, or 18%) were differentially expressed, however, presumably in a disorganised manner. Twenty percent of these clones were more than 2.5-fold decreased in expression, and only 7% were still over-expressed with R > 2.5. Such an abrupt re-distribution between up- and downregulated clones to this time point reflects drastic changes in the control of gene expression (experiment 1.2). These changes correspond to the observed severity of the phenotype under continued nutrient stress, resulting finally in plant death.

Table 1.  Complex analysis of normalised signal intensities obtained from array hybridisations (R, ratio of average intensity at –S to average intensity at normal S)
Experiment IDDifferentially expressed clones, P-value < 0.05 (% to total number of clones analysed)Contributions to a total number of differentially expressed clones for:
Upregulated clones (%), R > 2.5Downregulated clones (%), R < 0.4
Experiment 2.1, 6 days at –S2473 (16.4)153 (6.2)141 (5.7)
Experiment 2.2, 10 days at –S2381 (15.8)282 (11.8)93 (3.9)
Experiment 1.1, 10 days at –S2031 (13.5)346 (17)167 (8.2)
Experiment 1.2, 13 days at –S2749 (18.3)197 (7.2)543 (19.8)

Functional categorisation of sulfur-responsive genes

Functional analysis was first achieved by grouping sulfur deficiency-responsive genes according to the predicted functions of their proteins. Individual genes were assigned to ten different functional categories, which were automatically derived using MIPS A. thaliana database (MATDB) at the server of Munich Information Center for Protein Sequences (MIPS, http://mips.gsf.de/proj/thal/db). In nine of the ten functional categories, the number of clones upregulated at –S was higher than the number of downregulated clones (Table 2). However, in one functional category, comprised primarily of the genes encoding accessory proteins of electron transport and membrane-associated energy conservation (super-category Energy), the ratio of over-expressed to downregulated genes exhibited the opposite tendency. Total number of upregulated to downregulated clones from this super-category was 75 to 183.

Table 2.  Distribution of differentially expressed clones between the groups of up- and downregulated in the main functional categories at sulfur-deficient conditions (R, ratio of average intensity at –S to average intensity at normal S)
CategoryUpregulated, R > 2Downregulated, R < 0.5
Cellular organisation222119
Cellular communication and signalling11856
Cell rescue, defence, cell death and ageing10545
Energy75183
Metabolism772413
Protein destination23387
Protein synthesis13753
Transcription225123
Transport facilitation10066
Not identified (unknown)1317992

Tables 3 and 4 contain selected examples of differentially expressed clones. For each clone, four ratios corresponding to the four experimental points are given. Ratios printed in bold are statistically significant with P-values less then 0.05. Table 3 shows clones belonging to the above-mentioned functional categories for each category in two groups: significantly upregulated and significantly downregulated. Temporal analysis of gene activation from the first (6 days at –S) to the last (13 days at –S) experimental point revealed clusters of genes correlated in the patterns of gene behaviour, as well as mutual activation of these gene clusters. The activated gene clusters were further correlated with biosynthetic pathways. The interaction of pathways in conditions of sulfur starvation was analysed. The highest response was found for sulfur assimilation, jasmonate biosynthesis, and auxin-related pathway genes. The expression data related to these three pathways are summarised in Table 4.

Table 3.  Sulfur-responsive genes sorted according to functional categories, selected up- and downregulated clones (R, ratio of average intensity at –S to average intensity at normal S)
#Clone IDR 1.1R 1.2R 2.1R 2.2Gene IDFunctionFunctional subcategory
  1. Ratios printed in bold are statistically significant with P-values less than 0.05.

Cellular organisation
 194A11T7-04.06.10.80.3AT3G23890Topoisomerase IIDNA synthesis and replication
 2122P4T7-04.63.81.41.7AT5G20490Myosin-like protein my5Cell polarity and filament formation
 380A10T7-02.22.32.32.1AT5G01600Ferritin 1 precursorChloroplast organisation
 4E6B11T7-00.40.30.50.4AT2G38530Lipid transfer protein 2Organisation of plasma membrane
 580F1T7-00.20.20.60.4AT3G23610Dual-specificity protein phosphataseMeiosis
 6143K8T7-00.40.30.40.4AT3G01500Carbonic anhydraseExtracellular transport (secretion)
Cellular communication and signalling
 785A9T7-01.82.22.32.5AT1G33600BLAST match to receptor protein kinaseReceptor proteins
 8230O3T7-12.32.32.02.8AT1G21270Cell wall-associated receptor kinase 2Receptor proteins
 9F4D12T7-02.81.21.92.4AT1G28440Similar to receptor kinase, Arabidopsis thalianaReceptor proteins
 10147A15T7-00.40.50.50.6AT3G23750Receptor-like kinase, putativeReceptor proteins
 11144C15T7-00.50.40.60.4AT2G27190Purple acid phosphatase precursorKey phosphatases
 12162E9T7-00.30.20.30.4AT5G34850Secretory acid phosphatase precursorKey phosphatases
Cell rescue, defence, cell death and ageing
 13192C2T7-03.13.82.52.6AT1G08830Cu,Zn superoxide dismutaseDetoxificaton
 14117J6T7-02.22.42.32.8AT4G14030Selenium-binding protein likeDetoxificaton
 15224N7T7-02.41.52.22.3AT5G64100Peroxidase ATP3aDisease, virulence and defense
 16114J4T7-00.50.80.60.4AT2G43350Putative glutathione peroxidaseStress response
 17206N24T7-00.40.40.70.7AT4G28050Senescence-associated protein -likeStress response
 1863A2T7-00.20.20.60.6AT4G16950Disease resistance RPP5 like proteinStress response
Energy
 1991H16T7-02.63.41.21.2AT3G16480Mitochondrial processing peptidaseRespiration
 20125O15T7-01.73.61.42.4AT3G05930Germin-like protein (GLP8)Respiration
 21240K13T7-02.11.92.12.2AT4G16190Cysteine proteinase like proteinOther energy generation activities
 22183O15T7-00.20.10.40.5AT2G34430Lhcb1B1 chlorophyll a/b bindingAccessory proteins of electron transport
and membrane-associated energy
conservation
 23E8F3T7-00.30.20.40.4AT5G54270Lhcb3 chlorophyll a/b binding 
 24231L1T7-10.30.10.50.4AT2G34420Lhcb1B2 chlorophyll a/b binding 
Metabolism
 25122B24T7-028.712.211.07.8AT1G75280Isoflavonoid reductase homologueCarbohydrate utilisation
 26216D20T7-012.030.85.511.9AT3G05400Sugar transporter, putativeCarbohydrate transport
 27205G7T7-02.31.62.32.0AT4G15390HSR201 like proteinLipid and fatty-acid biosynthesis
 28111D3T7-00.30.10.30.4AT4G35770sen1, senescence-associated proteinNitrogen and sulfur utilisation
 29173O24T7-00.40.40.40.8AT5G01220SQD2Lipid and fatty-acid biosynthesis
 30128N22T7-00.20.10.40.4AT3G53260Phenylalanine ammonia-lyasePhenylpropanoids biosynthesis
Protein destination
 31197L7T7-03.23.92.51.9AT5G17060ADP-ribosylation factor -like proteinAssembly of protein complexes
 3294P10T7-03.13.12.11.0AT2G40880Cystatin B, putativeProteolysis
 33231K13T7-02.02.12.62.2AT4G01620Cathepsin B-like cysteine proteaseProteolysis
 34146C23T7-00.30.30.70.7AT2G22990Serine carboxypeptidase I, putativeProteolysis
 3583H12T7-00.40.40.70.5AT2G37410Protein translocase, putativeProtein targeting, sorting and translocation
 36189E4T7-00.50.60.50.3AT1G72750Inner mitochondrial membrane protein 
Protein synthesis
 37222P11T7-04.20.92.74.0rrn2626S ribosomal RNA proteinRibosomal proteins
 38F7F4T7-02.31.13.32.2AT1G07940Elongation factor 1-alpha (EF-1)Translation
 39132G8T7-04.01.80.71.8AT1G0836060S ribosomal protein L10A, putativeRibosomal proteins
 40187O9T7-00.40.60.70.5AT5G22800Alanyl-tRNA synthetasetRNA-synthetases
 41E1D9T7-00.50.30.60.4AT3G2783050S ribosomal protein L12-ARibosomal proteins
 42146L8T7-00.40.40.40.4AT5G2244060S ribosomal protein L10ARibosomal proteins
Transcription
 43137F4T7-02.11.91.91.5AT5G62190RNA helicaserRNA processing
 44186O19T7-02.82.40.70.7AT5G15310AtMYB16Transcriptional control
 45170K1T7-11.31.71.71.9AT3G48920AtMYB45Transcriptional control
 4696N14T7-03.34.41.80.3AT5G59780AtMYB59Transcriptional control
 4799P9T7-02.11.71.41.2AT1G56650AtMYB75Transcriptional control
 48193M15T7-01.72.31.21.1AT1G66390AtMYB90Transcriptional control
 49157G9T7-02.21.8#N/A#N/AAT5G10280AtMYB92Transcriptional control
 50111E18T7-01.41.41.92.1AT3G47600AtMYB94Transcriptional control
 51115B13T7-00.40.40.70.2AT3G27360Histone H3, putativeTranscriptional control
 52187D4T7-00.40.40.50.6AT3G60220RING-H2 zinc finger protein ATL4Transcriptional control
 53177M6T7-00.40.30.40.4AT3G56510Similarity to TATA-binding proteinmRNA synthesis
Transport facilitation
 54165N2T7-06.810.63.813.7AT5G26220Similarity to cation transport proteinCation transporters
 55246H14T7-03.07.01.22.1AT5G59310Lipid-transfer protein precursor - likeLipid transporters
 56204C3T7-01.31.82.62.4AT5G09220Amino acid transport protein AAP2Amino-acid transporters
 5775C11T7-00.60.5#N/A#N/AAT5G33320Phosphate/PEP translocatorPhosphotransferase systems
 58E6H9T7-10.40.30.50.6AT3G47960Putative peptide transporterAmino-acid transporters
 59210L8T7-00.40.20.50.5AT5G19980Lipophosphoglycan biosynthesisC-compound transporters
Table 4.  Sulfur-responsive genes sorted according to biosynthetic pathways (R, ratio of average intensity at –S to average intensity at normal S)
#Clone IDR 1.1R 1.2R 2.1R 2.2+Herb*–Fe*Gene IDMIPS function
  • *

    Results of comparative analysis of clone expression behaviour during treatment with herbicides (+Herb, unpublished data) and iron starvation (−Fe, Thimm et al., 2001); up, upregulation and down, downregulation in respective experiments.

  • Ratios printed in bold are statistically significant with P-values less than 0.05.

Tryptophan-IAA pathway, auxin-related genes
 1G2C8T73.42.12.12.9AT1G488605-Enolpyruvylshikimate-3-phosphate synthase
 2G1E7T76.35.22.45.8DownAT1G18870Putative isochorismate synthase
 3240I16T73.22.62.61.4AT5G05730Anthranilate synthase component I-1
 4208G10T71.82.11.62.3AT5G38530Tryptophan synthase beta chain
 5G1C4T75.05.71.92.5DownAT1G58160Myrosinase binding protein, JA-inducible
 6200L14T71.61.71.61.7AT1G51470Myrosinase precursor, putative
 7133J6T71.00.81.61.8AT2G33070Putative myrosinase-binding protein
 8113L3T70.30.30.60.7AT5G25980Myrosinase TGG2
 9116K19T70.30.20.60.5DownAT5G26000Myrosinase precursor
 10159P22T70.30.20.30.2AT3G14210Myrosinase-associated protein, putative
 11205A9T75.34.14.74.4AT3G44320Nitrilase 3
 12155M24T73.92.82.33.1AT1G60680Auxin-induced protein
 13184F12T72.82.90.80.8AT4G02380Induced by auxin, ethylene
 14F3B7T72.61.73.02.6UpAT1G60710Auxin-induced atb2
 15146E7T71.82.02.71.6AT5G25890IAA28 auxin-induced protein
 16F3B6T72.32.20.81.0AT5G65670IAA9 auxin-induced protein
 17129C6T72.71.41.21.5UpDownAT1G51950IAA18 early auxin-induced protein
 18135N23T71.21.81.92.1AT2G33310Auxin regulated protein (IAA13)
 19E7D3T71.91.21.51.6AT1G04250IAA17/AXR3-1 auxin-induced protein
 20242I18T70.60.20.30.8AT2G33830Putative auxin-repressed protein
 21E8D3T70.50.50.50.4UpAT1G04240Putative auxin-induced protein AUX2-11
Sulfur-related pathways
 22213H17T73.24.02.62.0AT1G75270GSH-dependent dehydroascorbate reductase
 23140B17T71.92.21.81.5AT1G19570GSH-dependent dehydroascorbate reductase
 24121P8T72.02.61.91.1UpAT2G25080Glutathione peroxidase
 25106F1T72.62.7#N/A#N/AUp/downAT3G24170Glutathione reductase, cytosolic
 26190P9T71.92.62.11.2UpUpAT1G02920Glutathione transferase
 27115L24T72.23.62.00.6AT3G43800Glutathione transferase-like protein
 28E2E3T72.01.82.21.4AT3G09270Putative glutathione transferase
 2989K20T72.01.00.91.5AT3G13110Serine acetyltransferase (Sat-1)
 30181H17T71.00.91.22.0AT3G59760Cysteine synthase oasC
 31212D22T70.40.50.70.3AT3G03630O-acetylserine (thiol) lyase; cysteine synthase
 32175C9T76.07.95.09.0AT1G36370Putative serine hydroxymethyltransferase
 33E1H8T73.12.31.62.7AT3G17390S-adenosylmethionine synthetase like
 34118N16T71.51.02.22.0AT4G01850S-adenosylmethionine synthetase 2
 35220C7T72.01.52.41.8AT3G02470S-adenosylmethionine decarboxylase
 36227E17T72.01.91.91.4AT4G13940Adenosylhomocysteinase
 37187L7T70.40.30.60.8AT4G23100Gamma-glutamylcysteine synthetase
 38142F20T72.14.13.13.5AT5G10180Sulfate transporter AST68 (Sultr2;1)
 39251F3T72.42.61.33.1AT5G13550Sulfate transporter Sultr4;1
JA biosynthesis pathway
 40230J8T71.31.91.91.4AT5G42650Allene oxide synthase
 41106H21T73.74.23.42.3AT1G7668012-Oxophytodienoate reductase (OPR1)

In the experiments by Hirai et al. (2003) under conditions of short-term sulfur starvation, the genes of sulfur assimilation pathway, and also of the jasmonate biosynthesis pathway were shown to be activated as well.

Genes responding specifically to sulfur deficiency

Arabidopsis plants subjected to sulfur depletion were under severe nutrient stress leading to death in less than a month. This means that under chosen experimental conditions, sulfur-specific responses are mixed with non-specific, or general stress responses. To distinguish between general stress response and the specific response to sulfur deficiency, the data were compared to data obtained under other stress conditions. We assumed that those genes responsible for general stress reactions should similarly respond to different stresses. Thus, subtraction of these genes from the list of genes differentially expressed in conditions of sulfur deficiency should reveal genes responding specifically to sulfur. For this comparative analysis, we have chosen the following two stress-related experiments, which were performed using identical EST arrays. In the first set of experiments, the plants were subjected to iron deficiency stress (Thimm et al., 2001; the dataset is available at http://www.biologie.hu-berlin.de/botanik/index2.html). In the second set, we applied two herbicides, asulam and methotrexate, in sublethal doses (unpublished results). The criteria to select the differentially expressed clones were the same as for sulfur-responsive clones (R > 2 in condition of P < 0.05 at least in one experimental point for transcriptional factors, and R > 2.5 in condition of P < 0.05 at least in one experimental point for all the other clones). Under these selection criteria, 1300 EST clones totally and 103 transcription factors were significantly over-expressed in iron-deficient conditions. Respectively, 383 and 35 clones showed significant over-expression after treatment with the herbicides. Venn diagrams (Figure 3), in which only upregulated clones are represented, visualise the results of comparative analysis for three tested stress conditions. In the subtracted area of sulfur-specific response, there are 743 EST clones referring to 632 individual genes (due to the redundancy of the spotted library). These genes were considered distinct from genes involved in general stress responses and thus constituted the group of primary interest for further studies. Considering the clones representing transcription factors, the sulfur-specific group contained 67 clones corresponding to 49 individual genes, constituting the group of potential sulfur-signalling elements. The results of this comparative analysis for selected genes are shown in Table 4 in columns +Herb (Herbicides) and –Fe (up means upregulation and down means downregulation in respective experiments).

Figure 3.

Comparative analysis of the patterns of clone expression under different stress conditions. Venn diagrams show the numbers of clones over-expressed at significant levels during sulfur deficiency (–S, this paper), as well as in the experiments on iron starvation (–Fe, Thimm et al., 2001) and on treatment with herbicides (+Herb, unpublished data), for the whole clone set (a) and for the annotated transcriptional factors (b). The selection criteria for the clone calculations are depicted.

Discussion

Array hybridisation confirmed known and revealed new sulfur-responsive genes

Extended sulfur depletion of Arabidopsis seedlings resulted overall in the differential expression of about 10% of genes. As experimental conditions mimicked long-term starvation, we expected to see differential expression of genes involved in sulfur uptake, assimilation and downstream processing of cellular sulfur compounds, and of genes involved in physiological responses and adaptation towards this stress.

The extensive data on gene behaviour in conditions of sulfur deficiency obtained from gene expression profiling were used to construct a response network. We first analysed the dataset for reliability by comparison with results on sulfur starvation experiments made earlier by a number of research groups, where the behaviour of individual genes of the sulfur assimilation pathway was described (summarised by Hawkesford, 2000). Sulfate transporters represent a group of known sulfur-responsive genes (Lappartient et al., 1999; Takahashi et al., 1997, 2000). As a confirmation for the validity of our approach, two genes of this group were significantly over-expressed: Sultr2;1 (At5g10180) and Sultr4;1 (At5g13550) (Table 4, entries 38, 39).

Sultr1;2 (At1g78000) is a sulfate transporter known to be strongly induced by sulfur starvation (Takahashi et al., 2000). Yet, it was not represented by any of the EST clones from the MSU library spotted to the filters, as under sulfur-sufficient conditions class I sulfate transporters are expressed at very low levels.

Cysteine is the first reduced sulfur-containing organic compound synthesised in plants after the nutrient sulfate is reduced to sulfide. In conditions of limited sulfate availability, cysteine contents dropped to non-measurable levels (Figure 2c). The amount of glutathione became significantly lower in conditions of sulfur deficiency as well (Figure 2b), while the genes from the glutathione oxidation/reduction cycle were activated by sulfur starvation (Table 4, entries 22–25). This indicates an attempt to sustain free glutathione levels under conditions of sulfur depletion. Glutathione is involved in various stress-related responses, mainly in scavenging reactive oxygen species via the ascorbate–glutathione cycle (Noctor et al., 2000).

Another sulfur-containing amino acid, methionine, is synthesised from cysteine. Methionine and its direct derivative S-adenosylmethionine (SAM) are essential intermediates for a number of plant processes, prime among them methylation reactions within C1-metabolism (Galili and Höfgen, 2002). The array data revealed significant over-expression of two genes of the SAM methylation cycle at –S: AT3g17390 encoding S-adenosylmethionine synthetase-like protein and At4g01850 for S-adenosylmethionine synthetase 2 (Table 4, entries 33 and 34). Interestingly, the genes of the direct methionine biosynthetic pathway were not induced upon sulfur starvation. Studies on the methionine recycling pathway in wheat have shown that under sulfur starvation methionine is preferentially utilised for methylation reactions via S-adenosylmethionine synthesis, whereas sulfur-containing proteins decrease in relative amounts (Fullington et al., 1987; Moss et al., 1981; Wrigley et al., 1984), which is in perfect agreement with our observations of plant priorities directed to methylation under sulfur starvation.

One of the highest over-expressors (8 to 29-fold) revealed in this study was annotated in MATDB as a gene encoding a putative NADPH oxidoreductase (At1g75280, Table 3, entry 25; Figure 4). This very gene was first isolated as a P3 cDNA in a tolerance screen towards the thiol-oxidising drug diamide, which depletes reduced glutathione in the cell, and was shown to be an Arabidopsis homologue of isoflavonoid reductases (IFR) from different plant species (Babiychuk et al., 1995). The activity of legume isoflavonoid reductases suggested that all IFR family members are oxidoreductases utilizing NAD(P)H as cofactors and isoflavonoids as electron acceptors, although a specific isoflavonoid substrate for the enzyme has not been identified yet. The same isoflavonoid reductase gene was isolated later by mRNA differential display applied to maize seedlings grown under sulfate-deprived conditions, where it was selectively induced in response to sulfur starvation (Petrucco et al., 1996). The expression of this maize gene was negatively correlated with glutathione levels, suggesting its role in thiol-independent response to oxidative stress under glutathione shortage conditions. The data from the present study on reduced glutathione level and highly induced isoflavonoid reductase gene activity referring now to Arabidopsis sulfur-starved plants support the previous findings obtained in maize. Furthermore, it indicates that the plant tries to keep its ability to cope with oxidative stress at a maximum level by keeping glutathione levels as high as possible even at the cost of other sulfur compounds and that a second, specific scavenging system comes into play when glutathione capacity is impaired. Interestingly, this itself implies induction of flavonoid biosynthesis to provide the substrate for the IFR gene product.

Figure 4.

RNA gel blot of EST clone T44052 (isoflavonoid reductase homologue/putative NAD(P)H oxidoreductase gene). The mRNA, labelled as ‘–S’, was isolated from samples transferred to sulfur-deficient medium in hydro-culture (control plants, labelled as +S, were transferred from sulfur-sufficient to the fresh sulfur-sufficient medium).

Sulfur-related energy deficit resulted in flavonoid biosynthesis activation

Besides elucidating the differential expression patterns of genes involved in the various sulfur-related pathways, we were interested in learning about the complex response of the whole network. To this end, we performed a systematic comparative analysis of the genes exhibiting altered expression under conditions of sulfur starvation. The general predominance of clones upregulated at –S (Table 2) might well be interpreted as the result of a maximal mobilisation in the adaptation mechanisms of plants under conditions of severe stress. At the same time, we observed significantly decreased levels of gene expression at –S conditions in only one functional category, which is comprised of genes encoding accessory proteins of electron transport and membrane-associated energy conservation (super-category Energy). Co-ordinated downregulation of just these genes indicates that energy-related processes associated with membranes were significantly depressed in response to sulfur starvation. This effect may result from quantitative alterations of iron–sulfur centres of Photosystem I proteins or/and from changes in thylakoid membrane lipid compositions due to decreased sulfur supply.

The possible involvement of altered sulfolipid content in the sulfur-related energy deficit is supported in our transcript profile by significant downregulation of the SQD2 gene (At5g01220, Table 3, entry 29). This gene was identified recently to encode the second of two sulfolipid biosynthesis enzymes (Yu et al., 2002). Furthermore, it was shown that in the sqd2 mutant lacking sulfolipid, the effective quantum yield of photochemical energy conversion in photosystem II was slightly but significantly reduced (Yu et al., 2002).

An increase in the relative amount of sulfolipid and a concomitant decrease in phosphatidylglycerol were observed in a set of counter-experiments, in which plants were exposed to phosphorous deficiency (Essigmann et al., 1998; Hartel and Benning, 2000) or altered light (Hartel et al., 1998). This raises the assumption that a certain amount of anionic thylakoid lipid needs to be maintained if the photosynthetic apparatus is to function. Under phosphate limiting conditions, this might be achieved by substituting phosphatidylglycerol with sulfolipids. At the same time, thylakoids of the Arabidopsis phospholipid-deficient pho1 mutant showed increased sulfolipid to phospholipid ratios but did not show effects on photosynthetic electron transfer (Hartel et al., 1998). This observation supports the assumption that maintaining an optimal thylakoid lipid milieu is more important for vital processes than maintaining the original phospholipid to sulfolipid ratio. Furthermore, in sulfur-deficient mustard and rapeseed, total lipid content was shown to decrease (Ahmad and Abdin, 2000; Munshi et al., 1990). The main reason for photosynthetic organisms to employ sulfolipids is believed to minimise the amount of phosphate required when synthesising the large amounts of membranes, required in turn to support high rates of light capture (Somerville et al., 2000). If so, then limited sulfolipid biosynthesis by Arabidopsis seedlings grown at –S conditions should lead to limited energy supply and further extinction of photo-energetic processes. As a result, light assimilation capability is reduced, and, hence, normal light is perceived as a high light stress. This assumption is corroborated through the observed accumulation of high light-protective anthocyanins under sulfur-deficient conditions (this study) and through accumulation of anthocyanins in Arabidopsis in response to phosphorous starvation stress, as has been described previously (Trull et al., 1997). High light-induced expression of phenylpropanoid and flavonoid biosynthetic enzymes leading to anthocyanin production in Arabidopsis was shown by Graham (1998) and Shirley (1996).

Increased biosynthesis of anthocyanins provides an explanation for the observed over-expression of genes from aromatic amino acids synthesis pathway (Table 4, entries 1–2). Over-expression of other sets of genes, such as glutathione S-transferases, may also be partially explained by increased flavonoid biosynthesis. After inductive light treatment, flavonoids accumulate in vacuoles. Deposition of anthocyanins in vacuoles is facilitated by conjugation of anthocyanins with glutathione through specific glutathione S-transferases, as shown for maize (Marrs et al., 1995) and petunia (Mueller et al., 2000). Three genes from the Arabidopsis genome encoding glutathione S-transferases exhibited significant over-expression at –S: At1g02920, At3g43800 (glutathione transferase-like protein) and At3g09270 (putative glutathione transferase) (Table 4, entries 26–28). The exact functions of the proteins encoded by these genes need to be determined.

The expression of phenylpropanoid and flavonoid pathway genes needs to be triggered and controlled through transcriptional regulators. A definite number of transcription factors have been identified as specifically expressed under –S conditions. Genes of the R2R3-MYB family of transcriptional factors are especially over-expressed. R2R3-MYB genes were shown earlier to regulate phenylpropanoid metabolism in Arabidopsis (Stracke et al., 2001; Weisshaar and Jenkins, 1998). In our experiments, as many as seven R2R3-MYB transcription factors demonstrated over-expression under sulfur starvation: AtMYB16, AtMYB45, AtMYB59, AtMYB75, AtMYB90, AtMYB92 and AtMYB94 (Table 3, entries 44–50). For two of them, AtMYB75 and AtMYB90, a correlation to accumulation of anthocyanins was shown earlier as a result of T-DNA activation-tagged over-expression (Borevitz et al., 2000), again indicating the accuracy of our profiling dataset. In an extensive study of expression patterns of R2R3-MYB genes in plants grown under more than 20 different growth conditions (Kranz et al., 1998), both genes were expressed after exposure to UV light stress. Significant over-expression of these two genes in response to sulfur starvation, as revealed in the present study, supports their involvement in light assimilation regulation, on one hand, and confirms the development of high radiation stress via decreased thylakoid capacity at –S, on the other. Interestingly, for five of the seven R2R3-MYB genes (except AtMYB45 and AtMYB94), over-expression was higher in experiment 1 in which the plants suffered from sulfur deficiency from the time of germination, having no chance to amass sulfolipids for thylakoid formation before sulfur depletion was applied.

Among these MYB genes, AtMYB59 showed the strongest positive response to sulfur depletion (Table 3, entry 46). The functional analysis performed by Kranz et al. (1998) showed that this gene, like AtMYB90, was strongly induced by white light, and, surprisingly, by ethylene application, again together with AtMYB90. The likely involvement of these two MYB transcriptional factors in ethylene pathway regulation could explain their induction by sulfur starvation in the present study, where low sulfur caused a drop down in reduced sulfur-containing compounds and an induction of genes involved in SAM recycling. In turn, ethylene is synthesised from methionine through SAM. This would necessitate the tight regulation of this process during low sulfur conditions, a process in which the AtMYB59 and AtMYB90 genes are likely to participate.

The finding that pigment content was not altered in the sqd2 mutant lacking sulfolipid (Yu et al., 2002) suggests conditional roles for sulfolipid in photosynthesis. Sulfur starvation may be one such condition in which decreased levels of glutathione (Figure 2b) exacerbate the energy deficit the plants have to face. In order to counteract this deficit, the glutathione recycling genes are induced. Such an explanation might also imply the first point of connection with jasmonic acid, which has been shown to activate glutathione synthesis genes in Arabidopsis (Xiang and Oliver, 1998). With these points in mind, the array data revealed strong and consistent, among all experimental points, over-expression of the OPR1 gene (At1g76680) (Table 4, entry 41) during sulfur depletion. OPR encodes 12-oxophytodienoate reductase, one of the main enzymes in jasmonic acid biosynthesis (Figure 5c). Although OPR 3 was shown to be acting on jasmonate biosynthesis (Schaller et al., 2000), we rather detected over-expression of OPR 1, which indicates a function in sulfur-specific response for this isoform. The other gene of the same pathway, namely AT5G42650 (allene oxide synthase), showed synchronous over-expression as well (Table 4, entry 40), corroborating the assumption of the involvement of jasmonic acid with respect to regulation of plant response to –S.

Figure 5.

Selected sulfur-induced biosynthetic pathways and potential cross-talk between them.

Biochemical pathway connections are shown with dotted arrows. The up and down expression responses of the corresponding genes are marked with white-filled arrows and are summarised in Table 4, the altered increased (up) and decreased (down) levels of some metabolites are marked with white-filled arrows and are summarised in Figure 2.

(a) Auxin biosynthesis pathway: TRP-S beta, tryptophan synthase β-subunit; TRP-D, tryptophan decarboxylase; S-GT, UDPG:thiohydroximate glucosyltransferase.

(b) Sulfur assimilation pathway: SAT, serine acetyltransferase; CgS, cystathionine γ-synthase; CbL, cystathionine β-lyase; MS, methionine synthase; AHC, adenosylhomocysteinase; SAHC, S-adenosylhomocysteine; MMT, AdoMet:Met S-methyltransferase; HMT, homocysteine S-methyltransferase; SAM-S, S-adenosyl-l-methionine synthetase; SMM, S-methylmethionine; SAM, S-adenosyl-l-methionine; ACC, aminocyclopropane-carboxylic acid; SAM-DC, S-adenosylmethionine decarboxylase; DMSP, dimethyl sulfoniopropionate; DC-SAM, decarboxylated SAM; CMT, Met2-type cytosine DNA-methyltransferase-like protein.

(c) Jasmonate biosynthesis pathway: AOS, allene oxide synthase; AOC, allene oxide cyclase; OPR, 12-oxophytodienoate reductase; SAM:JA CMT, SAM:JA carboxyl methyltransferase.

Sulfur-deficient metabolic state caused surplus flux via auxin

Assignment of the identified sulfur-responsive genes to their respective biochemical pathways fostered further considerations regarding the main routes of co-ordinated response to sulfur deficiency. Besides sulfur assimilation and jasmonic acid biosynthetic pathways, genes involved in aromatic amino acid synthesis and downstream genes (leading via tryptophan to auxin and its derivatives) exhibited transcriptional activation (Table 4, entries 4, 6, 11). Significantly increased levels of transcription products for several genes of this pathway point to information flux to and/or via indole-3-acetic acid (IAA) biosynthesis. Furthermore, genes encoding auxin-induced proteins also showed over-expression (Table 4, entries 12–19) indicating that the build-up of auxin is part of the sulfur starvation response. Under sulfur-deficient conditions, accumulation of auxin might trigger induced root growth to provide additional access to exogenous sulfur. Elevated auxin levels are known to increase root elongation and branching, among other processes of plant growth and development (Ploshchinskaya et al., 2002).

The clear evidence of co-ordinate gene expression between and among biochemical pathways indicates the presence of a sophisticated response network. A picture starts to emerge about how the plant reacts to sulfur starvation by pathway integration. The involvement of auxin pathway genes can be traced directly from the metabolic state of sulfur-deficient plants as there are two direct links with the sulfur assimilation pathway (Figure 5). The first cross-point is tryptophan biosynthesis. Tryptophan is produced in plants from indole and serine through the activity of the tryptophan synthase β-subunit (Coruzzi and Last, 2000). The gene encoding this enzyme (At5g38530, Table 4, entry 4) was significantly induced by sulfur deficiency. In conditions of limited sulfate, cysteine production from O-acetylserine drops to non-measurable levels of cysteine (as we could confirm by cysteine measurements, Figure 2c), leading to accumulation of O-acetylserine and its precursor serine (Figure 2d,e). The surplus serine might be exploited for increased tryptophan production (as we measured under sulfur deficiency conditions, Figure 2f) and increased flow via downstream products including IAA as a result. This assumption is supported by an experiment in which the addition of exogenous O-acetylserine led to a strong induction of promoter activity of the gene encoding nitrilase, the final enzyme in auxin biosynthesis (Kutz et al., 2002).

The second cross-point of auxin and sulfur-assimilation pathways is the biosynthesis of indole glucosinolates, sulfur-containing glucosides derived from tryptophan via thiohydroximate production (reviewed in Halkier, 1999). Although the sulfur donor for the thiol sulfur in thiohydroximate is not known, cysteine was the most efficient sulfur source incorporated into glucosinolates in vivo (Wetter and Chisholm, 1968; cited by Halkier, 1999). Enzymatic degradation of indole glucosinolates by myrosinases releases sulfate and indole-acetonitrile, a direct IAA precursor. In Brassicaceae, glucosinolates are one of the major sulfur-containing compounds: as much as 6% of total sulfur is stored in glucosinolates, as shown for the youngest leaves of Brassica napus (Blake-Kalff et al., 1998). Under sulfur-deficient conditions, de novo synthesis of glucosinolates is limited by the sulfur donor for the thiohydroximate. At the same time, hydrolysis of the glucosinolate stores by myrosinase can provide additional sulfate, as suggested by Visvalingam et al. (1998) and confirmed experimentally by Blake-Kalff et al. (1998). In Brassica napus, these authors demonstrated the reduction of indolyl glucosinolates down to zero by day 13 of sulfur starvation.

Conversion of indole-acetonitrile, the product of indole glucosinolate degradation, to IAA is catalysed by nitrilases (Figure 5a). The gene encoding nitrilase 3 is one of the strongest over-expressors in our experiments (Table 4, entry 11). In sulfur starvation experiments made by Hirai et al. (2003), the gene encoding nitrilase was also over-expressed at –S. Auxin biosynthesis is believed to proceed via different pathways. Our experiments indicate that the nitrilase 3-driven pathway is the most prominent one under sulfur-deprived conditions. Kutz et al. (2002) have also demonstrated recently that nitrilase 3 is induced during sulfur depletion. Moreover, it has been proposed that the main route for auxin biosynthesis in Brassicaceae is via indole glucosinolates (reviewed in Grsic et al., 1999), which thus may function as a regulatory sink for IAA (Bak et al., 2001).

In sulfur-sufficient conditions, three enzymes involved in the glucosinolate portion of auxin biosynthesis, the tryptophan oxidizing enzyme, myrosinase and nitrilase, were enhanced after treatment with jasmonic acid (Grsic et al., 1999). Additionally, the IAR3 gene, which encodes IAA-Ala hydrolase, was induced by jasmonate (Sasaki et al., 2001). The regulation by jasmonate was also demonstrated at the metabolic level: indole glucosinolates accumulate systematically in plants treated with jasmonate (reviewed in Chen and Andreasson, 2001). On the contrary, auxin inhibits the expression of some jasmonic acid-induced genes (DeWald et al., 1994; Rojo et al., 1998). It can be assumed from these data that a circuit of jasmonic acid and glucosinolate/auxin-triggered changes provide a regulatory control network in response to sulfur starvation.

Thus, sulfur deficiency may lead to the excess of free IAA via: (i) serine accumulation channelled into tryptophan synthesis and (ii) indole acetonitrile production by indole glucosinolate degradation. In our experiments, this possibility is confirmed by activation of the corresponding pathway genes and by the auxin-rich phenotype (increased rooting) of sulfur-starved seedlings. Circumstantially, a number of auxin-inducible genes exhibited over-expression in the arrays. Involvement of IAA as a plant hormone in delicate regulation of a wide range of developmental processes and the toxicity of excess free IAA (mentioned in Normanly and Barlet, 1999) necessitate its transport, turnover, or degradation. Increase of IAA degradation was shown in transgenic lemna in response to the greater flux of metabolites through the tryptophan biosynthetic pathway (Tam et al., 1995). Flavonoids were shown to inhibit auxin polar transport in white clover (Mathesius et al., 1998) and Arabidopsis (Murphy et al., 2000). Moreover, sulfonation of flavonoids decreases their inhibition ability (Faulkner and Rubery, 1992). Hence, under conditions of sulfur deficiency, the activation of specific genes of flavonoid biosynthesis, which we indeed observed in the transcription profiles, can be interpreted as a measure to prevent auxin accumulation in extra-bioactive concentrations. We suggest from these data that involvement of auxin in the sulfur starvation response may be provided via the direct cross-talk of the auxin biosynthesis pathway with the sulfur assimilation pathway. Further studies on metabolite levels of free IAA and its conjugates or derivatives that we are pursuing now will help to elucidate the regulation of the response via IAA turnover.

Integration of the pathways provided specific response to sulfur deficiency

We are aware that the EST collection used for this study comprises only about one-third of the full Arabidopsis transcriptome, though presumably the majority of non-specialised gene functions are represented, we can deduce from the complex expression pattern a sufficiently corroborated and reasonable network. Yet, it has to be kept in mind that 36.5% of the spotted EST clones are of unknown function. As their functions are determined, our understanding of this network will improve.

The central question of plant response to sulfur starvation is transduction of the information flux to provide the most effective reaction. To be effective, the response must use a general stress-reactive information stream and finally provoke specific effects. Genes over-expressed or biochemical pathways activated in conditions of sulfur starvation may as well be, fully or partially, common to other information flux systems providing responses to other nutrient or environmental conditions. As cases in point, induction of tryptophan pathway enzymes (Zhao et al., 1998), flavonoids (Winkel-Shirley, 2002) and jasmonate biosynthesis genes (de Bruxelles and Roberts, 2001) as a response to different stresses was shown. However, specificity of the reaction to –S is caused by the uniqueness of the situation, when in order to deal with low sulfur conditions, plants must (i) channel excess auxin via inactivation of auxin transport inhibiting flavonoids, and (ii) produce more anthocyanins via the flavonoid biosynthesis pathway to neutralise high-light radiation in conditions of decreased thylakoid membrane capacity. This specificity can be provided by activation of specific genes of flavonoid biosynthesis, which are able to redirect the corresponding flows of metabolites through the pathway. An isoflavonoid reductase homologue gene, one of the highest over-expressors in our experiments, may be responsible for this re-direction.

Additional response regulation may be provided via jasmonic acid, which activates the genes for glutathione synthesis, causes indole glucosinolate accumulation and enhances auxin biosynthesis enzymes. In this study, the genes of jasmonic acid biosynthesis, especially the OPR1 gene, showed strong activation during sulfur starvation. Specificity of the response is likely achieved here by the combinatorial activation of these genes together with other sulfur-responsive genes revealed by this study. We suggest that integration of auxin, jasmonate and flavonoid pathway elements into a complex regulatory unit provides the basis of a co-ordinated, systematic response to sulfur deficiency. This process – convergence of informational fluxes from particular types of specific inputs into a general transduction path that subsequently diverges to induce specific responses – can be imagined as a daughter's plait or braid. Specific sensing and signal transduction mechanisms are separate hairs before the plait. Then, to effectively pilot information, fluxes converge into the main stream of non-specific general stress-induced signalling response pathways (together with signals derived from other stress factors), as strands converge into a plait. In this way, transduction of specificity via the main non-specific stream would be provided here by a characteristic interlacing of the affected informational flows, like the hairs are interwoven in a plait. To obtain finally specific stress response, pathways of informational fluxes should diverge again to provide effectiveness in chosen direction – like the unravelling at the end of a plait. Under sulfur depletion, the integration of auxin, jasmonate and flavonoid pathway elements with the sulfur assimilation pathway may be a point of convergence, and possible re-direction of flavonoid biosynthesis by putative isoflavonoid reductase enzyme may then represent a point of divergence. To see if this plait concept is a realistic explanation for the processes of nutrient stress-induced signal transduction, we are now integrating transcriptome data with metabolic networks revealed by metabolite profiling.

Experimental procedures

Plant material and growth conditions

Arabidopsis thaliana ecotype Columbia G1 plants were grown in a growth chamber (16-h light/8-h dark cycle) on sterile agarose medium in 12.5 cm square Petri dishes. In each Petri dish, plants were arranged in five rows of 100–120 seedlings each, on a solidified sulfur-sufficient medium (0.5N Murashige–Skoog salts), containing 915.01 µm sulfate. This sulfate concentration resulted from summarising sulfate in macro salts (750 µm), micro salts (65.01 µm) and Fe-EDTA, which was prepared by co-dissolving iron sulfate with Titriplex III (Merck), and in this way contributed 100 µm sulfate to the medium. To allow easy transfer to the new medium and sampling, sterilised seeds were put on sterile fine nylon filter laying on the surface of the medium (Figure 1a).

Sulfur starvation experiments

To subject seedlings to sulfur starvation, sulfur-sufficient medium was changed to a medium in which the macro element MgSO4 was replaced with an equimolar amount of Mg(NO3)2, and among micro salts CuCl2, MnCl2 and ZnCl2 were used instead of CuSO4, MnSO4 and ZnSO4, correspondingly. Thus, the whole sulfate from macro and micro salts was replaced, giving zero sulfate concentration in them. The only sulfate left in sulfur-deficient medium originated from iron sulfate, which was used to prepare Fe-EDTA. The resulting medium (modified 0.5 Murashige–Skoog medium, referred as ‘–S medium’, 100 µm sulfate concentration) contained 89% less sulfur, 8% more nitrogen and 4% more chlorine. Two types of sulfur starvation experiments were conducted. To obtain constitutive sulfur starvation, in the experiment 1 the seeds were sown directly to –S medium. To induce the genes involved in sulfur-starvation response in time-controlled manner, in the experiment 2 the seeds were first sown to normal medium, then after growth for 8 days the seedlings were transferred to –S medium. In both experiments, the samples were taken twice (Figure 1b): before (experiment 1.1 and 2.1) and after (experiment 1.2 and 2.2) appearance of visible changes caused by the sulfur depletion (for the exact number of days at –S before sampling, see Figure 1b). Five Petri dishes per treatment were used. To minimise biological and physiological variability, the sampling procedure followed a Latin square method, when seedlings from row 1 of Petri dish 1, row 2 of Petri dish 2, row 3 of Petri dish 3, row 4 of Petri dish 4 and row 5 of Petri dish 5 were all pooled together to represent sample 1 for RNA isolation 1. In this way, five pooled samples were obtained from five Petri dishes for each experimental point. By applying this method to each treatment, we got totally 40 independent pools for RNA preparation containing a minimum of 500 seedlings each.

Biochemical measurements of sulfur and sulfur-related metabolites

The levels of total sulfur were determined by inductively coupled plasma-atomic emission spectroscopy (Applied Research Laboratories, Accuris, Ecublens, Switzerland). The measurements were done by Laura Hopkins in IACR-Rothamsted, Harpenden, UK.

The levels of glutathione and cysteine were determined by the high performance liquid chromatography (HPLC)-based method after subsequent extraction, reduction and derivatisation, as described by Williams et al. (2002). The internal levels of serine, O-acetylserine and tryptophan in seedling samples were determined by GC–MS. Tissue extraction and derivatisation were performed as described by Szopa et al. (2001), with slight modifications. Serine and O-acetylserine (Sigma) were used as standards for spiking. For the quantification (MassLab software-ThermoQuest, Manchester, UK) of relative response, the following ions were used: ribitol m/z 319, serine m/z 116; 204, O-acetylserine m/z 132, tryptophan m/z 202. The levels of serine, O-acetylserine and tryptophan were determined as relative response ratios of peak areas of these compounds to peak area of internal standard (ribitol), normalised with respect to the fresh weight of the sample.

RNA isolation and radiolabelling

Total RNA was extracted from the pools of whole seedlings (RNeasy Plant Kit, Qiagen GmbH, Germany). Concentration and quality of isolated RNA was monitored on electropherograms by 2100 Bioanalyser (Agilent Technologies). For reverse transcription and radiolabelling, 10 µg total RNA was used. Radiolabelling was performed as described by Thimm et al. (2001).

Array hybridisation and data analysis

Filter design and production, as well as reference and complex hybridisations were performed as described by Thimm et al. (2001). For data analysis, the signal intensities of the reference and experimental hybridisations were quantified using the software arrayvision (Imaging Research Inc., Haverhill, UK). Normalisation, statistical analysis and response evaluation were done using the mathematical tools incorporated into the Haruspex database (http://www.mpimp-golm.mpg.de/haruspex/index-e.html), as described by Thimm et al. (2001). In total, eight filters were hybridised up to five times each with newly synthesised and labelled cDNA probe of the corresponding RNA pool. To control the influence of the individual filters to the hybridisation quality, two filters were cross-hybridised with the same RNA pools.

Validation of data on transcription profiling

The reliability of the gene expression profiling results was confirmed by:

  • 1Pooling of plant material, as designed by the experiment set-up (each experimental point was represented by repetitive pools, with 500–600 seedlings in each pool).
  • 2Statistically sufficient number of repetitions (five independently isolated RNAs per experimental point were hybridised separately to produce five profile repeats).
  • 3Redundancy of EST library (normally differential expression of several EST clones representing one and the same gene was confirmed).
  • 4t-test and P-value analysis.
  • 5Cross-check of the filters.
  • 6Positive controls (expected expressional behaviour of genes known to be induced by sulfur deficiency).
  • 7Northern hybridisation of the first 12 over-expressors with newly isolated RNA pools obtained in sulfur starvation experiments performed in hydro-culture.

The sequences of the first 25 over-expressors were checked and showed 100% correspondence with their annotation, confirming the quality of the Michigan State University EST library spotted to the macroarray filters.

Acknowledgements

We thank Dr Laura Hopkins and Dr Malcolm Hawkesford of IACR-Rothamsted, Harpenden, UK, for total sulfur measurements in Arabidopsis samples. We thank Professor Thomas Altmann and Dr Bernd Essigmann for provision of macroarray filters and technical advice. We also thank Dr Karin Koehl for advice concerning experimental design and statistical evaluation, Dr Sebastian Kloska and Peter Krueger for the Haruspex database creation and maintenance, Astrid Basner for constructive criticism, and Megan McKenzie for manuscript editing. We thank Professor Lothar Willmitzer for constant support and lively discussions promoting progress of this work. This research was supported by the EU commission through funding of FP5 project QLRT-2000–00103 and by the Max-Planck Society.

Footnotes

  1. The paper of Hirai et al., 2003 in this issue also describes the results of sulfur starvation experiments. Both papers (Hirai et al., 2003 and this paper) give a complementary view on the responses of gene expression and, partially represented, metabolite contents upon sulfur depletion. Especially, as the Hirai et al. paper discusses short-term starvation and induction with the putative effector molecule, OAS, while this paper discusses late phases in plant adaptation to nutrient stress. Apart from the generally lower ratios detected in the Hirai et al. paper, which might be due to differences in technical procedures, a number of key genes of sulfur assimilation and jasmonic acid biosynthesis pathways responded similarly in both studies, as well as a number of individual genes (as IFR, At1g75280) can be detected in both studies. Also, the metabolite levels, which were measured in both studies, were altered correspondingly: sulfate and glutathione decreased, O-acetylserine increased. Both EST sets, though comparable in size, do not fully overlap, providing on one hand more information, but also preventing a 1 : 1 comparison on the other. For the publication, selected ESTs were chosen. Yet, both groups intend to directly compare their data sets in future and make the whole data set available to the scientific community.

Ancillary