Medicago truncatula Root Nodule Proteome Analysis Reveals Differential Plant and Bacteroid Responses to Drought Stress 1, 2

Drought is one of the environmental factors most affecting crop production. Under drought symbiotic nitrogen fixation is one of the physiological processes to first show stress responses in nodulated legumes. This inhibition process involves a number of factors whose interactions are not yet understood. The present work aims to further understand changes occurring in nodules under drought stress from a proteomic perspective. Drought was imposed on Medicago truncatula cv. Jemalong A17 plants grown in symbiosis with Sinorhizobium meliloti 2011. Changes at the protein level were analysed using a non-gel approach based on liquid chromatography coupled to tandem mass spectrometry. Due to the complexity of nodule tissue, the separation of plant and bacteroid fractions in M. truncatula root nodules was first checked with the aim of minimising cross-contamination between the fractions. Secondly, the protein plant fraction of M. truncatula nodules was profiled, leading to the identification of 377 plant proteins, the largest description of the plant nodule proteome so far. Thirdly, both symbiotic partners were independently analysed for quantitative differences at the protein level during drought stress. Multivariate data mining allowed for the classification of proteins sets which were involved in drought stress responses. The isolation of the nodule plant and bacteroid protein fractions enabled the independent analysis of the response of both counterparts, gaining further understanding of how each symbiotic member is distinctly affected at the protein level under a water-deficit situation.


INTRODUCTION
One of the best studied N 2 -fixing symbioses is the one established between certain members of the Leguminosae family and soil bacteria, collectively termed rhizobia. This symbiotic interaction results in the formation of a unique plant organ, the root nodule, to which the plant supplies reduced carbon for the bacteroids (differentiated form of bacteria) in exchange for fixed nitrogen.
In the last decade Medicago truncatula, which establishes a symbiosis with Sinorhizobium meliloti, has emerged as a useful model legume for molecular and genetic studies (Barker et al., 1990;Cook et al., 1997) and has been the focus of extensive research at both transcript and protein level. The proteomic characterisation of different plant organs, as well as cell cultures, has been carried out (Mathesius et al., 2001;Gallardo et al., 2003;Watson et al., 2003;Catalano et al., 2004;Valot et al., 2004;Lei et al., 2005). Detailed proteomic characterisation of S. meliloti both as free-living cells and bacteroid has been described elsewhere (for a comprehensive review, see Djordjevic, 2004). However, little is known about the plant proteome in the M. truncatula root nodule. One of the closest approaches was carried out by Bestel-Corre et al. (2002) in which they compared the root proteome when inoculated with symbiotic bacteria and arbuscular mycorrhizal fungi. In a similar manner, Natera et al. (2000) analysed the symbiotic interaction between Medicago alba and S. meliloti, but the microbial counterpart was the main focus of the study. Therefore, in the first part of the present work the soluble plant proteome in nodules of M. truncatula in symbiosis with S. meliloti was characterised. The approach chosen was a proteomic technique based on two dimensional liquid chromatography separation of peptides in complex mixtures. Proteomic studies are traditionally carried out using two dimensional gel electrophoresis techniques (2D-PAGE). However, despite the fact that it was pioneered nearly 37 years ago (Kaltschmidt and Wittmann, 1970), 2D-PAGE has some restrictions: automation, reproducibility and quantification are still major fields of development. To address some of the technical limitations inherent to 2D-PAGE, alternative gel-independent separation methods have been developed. One of them is the direct identification of proteins from complex protein mixtures based stresses, with drought being one of the major environmental factors affecting plant productivity (Boyer, 1982;Zahran, 1999). Under this stress, symbiotic nitrogen fixation (SNF) is one of the physiological processes to first show stress responses in nodulated legumes, occurring before the decrease in CO 2 -photosynthetic assimilation rates (Durand et al., 1987). Several hypotheses have been proposed by different research groups to explain the decline in nitrogen fixation during drought: i) regulation through the control of carbon flux within nodules, mainly due to down regulation of sucrose synthase (González et al., 1995;Gordon et al., 1997;Arrese-Igor et al., 1999); ii) regulation based on the internal oxygen levels within nodules (Durand et al., 1987;Diaz del Castillo et al., 1994;Serraj and Sinclair, 1996) and iii) nitrogen feedback regulation (Serraj et al., 2001;King and Purcell, 2005). However, to date, the molecular mechanisms responsible for these physiological responses are not yet understood. In order to further characterise this complex regulation process, the variation in the protein profiles of control and water-stressed nodules was analysed. Symbiotic root nodules are complex structures, where plant and bacteria cells are in close proximity. Therefore, special care was taken to ensure minimal cross-contamination during the separation of the plant and bacteroid fractions. This enabled the independent analysis of the response of both symbiotic partners, gaining further understanding of how each symbiotic member is distinctly affected by drought.

Separation of plant and bacteroid fractions of M. truncatula root nodules
Most of the published proteomic studies include a protein extraction step based on tissue homogenisation using liquid nitrogen. However, when this is applied to nodule tissue, a significant level of cross-contamination among plant and bacteroid protein fractions is observed (Natera et al., 2000;Bestel-Corre et al., 2002). As the separation of plant and bacteroid fractions in nodules was one of the aims of this study, we initially evaluated the degree of protein cross-contamination between the fractions when comparing homogenisation using liquid nitrogen to homogenisation using ice-cold extraction buffer (adapted from Saalbach et al., 2002) followed by centrifugation of the extract to pellet bacteroids. The level of contamination between the fractions was firstly estimated by western blot analysis employing antibodies against components of the bacteroid nitrogenase complex (NifDK) (Figure 1). Secondly, plant fractions were analysed via LC/MSMS and peptide mass spectra were matched against the S. meliloti protein database to detect bacterial proteins present in the putative plant fraction. While maintaining total protein content in nodules (26.2 ± 2.3 mg protein/g fresh weight nodule), the ratio bacteroid:plant fraction protein content was 1:2 when ice-cold buffer was used, whereas the ratio decreased to 1:7 when liquid nitrogen was employed. Furthermore, LC/MSMS analyses showed the presence of bacterial origin proteins in the putative plant fraction after liquid nitrogen homogenisation, representing almost 40% of the total peptides detected in the bacteroid fraction. However, when homogenisation buffer was employed, the presence of contaminating bacterial origin peptides was reduced, representing only 3% of the total bacteroid peptides detected. This reduction was not due to protein degradation, as total nodule protein content was maintained, but to a different distribution of the protein among the fractions.
Taking all these data together, both approaches indicated that liquid nitrogen was the key factor involved in the unwanted breakage of bacteroid cells and, consequently, responsible for the bacterial protein contamination in the plant fraction. Once the method for minimal crosscontamination was established, two types of analyses were performed. A flow diagram of the experimental work is represented in figure 2.

Plant protein identification in N 2 -fixing M. truncatula root nodules
The plant fraction of M. truncatula root nodules was collected as outlined above and soluble protein was extracted. To identify a large number of nodule plant proteins, a double separation step was performed according to a previous analysis of Arabidopsis leaf proteome (Wienkoop et al., 2004). First, using anion-exchange FPLC the initial plant extract was fractionated into 9 subfractions. Then protein from each of the 9 subfractions was precipitated, enzymatically digested and separated using 2D-LC/MS/MS (Fig. 2). The obtained mass spectra were searched against The Institute for Genomic Research (TIGR) Medicago Gene Index release 8.0 database (now located at http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb=medicago) using stringent criteria as described in materials and methods. As a result, 377 unique plant proteins were identified, representing the largest nodule plant proteomic analysis reported so far. This result confirms the usefulness of novel liquid chromatography-based methods for the large scale identification of proteins in complex tissues such as N 2 -fixing nodules. The complete list of identified proteins, ordered by functional groups, is presented in Table S1 as supplemental material. Some of the most relevant identified proteins are discussed below.
The list of identified proteins was functionally classified using the Mapman program (Fig. 3), which classifies genes/proteins based on gene ontology and functional classification databases (Gene Ontology Consortium and Kyoto Encyclopedia of Genes and Genomes database) (Thimm et al., 2004;Usadel et al., 2005). In contrast to previously reported large-scale proteome analyses, in which a large number of unknown proteins are often reported, surprisingly, the present study contains a reduced portion of proteins with unknown function. Although the reason behind remains unclear, this result may be due to the fact that in nodules most of the high abundant and detectable proteins have been characterised, becoming a suitable source for physiological interpretation.
Following classification, the largest functional group comprises proteins involved in amino acid metabolism and protein synthesis and degradation. Apart from the known enzymes involved in nitrogen assimilation in nodules, several proteins related to the biosynthetic pathway of sulfurcontaining amino acids were detected. This is particularly interesting taking into account the limited information available on sulfur metabolism in legume nodules. Uptake of sulfur from the soil occurs almost exclusively in the form of sulfate and the sole entry step for the metabolism of sulphate is the formation of adenosine 5'-phosphosulfate, by ATP sulfurylase (Saito, 2004). A nodule ATP sulfurylase (TC100144) was identified which had a close similarity to a plastidic enzyme recently reported to be abundantly expressed in soybean roots (Phartiyal et al., 2006). In addition, several enzymes were identified which are involved in the biosynthesis of sulfurcontaining amino acids, such as methionine synthase, cysteine synthase or cysteine desulfurase.
Notably, the present analysis enabled the detection of a considerable number of enzymes involved in the biosynthesis of S-adenosyl-L-methionine (SAM), a primary methyl-group donor and also a precursor of metabolites such as ethylene, polyamines and vitamin B1 (Amir et al., 2002). SAM is synthesized from methionine and ATP by the enzyme SAM synthetase, for which several isoforms have been identified in the current analysis. These findings suggest that N 2 -fixing root nodules play an important, so far overlooked, role in plant sulfur assimilation, especially the biosynthesis of sulfur-containing amino acids and regulatory compounds such as SAM.
For proteins related to energy production, the analysis allowed the almost complete identification of enzymes of the glycolytic pathway and tricarboxylic acid cycle (TCA). In nodules sucrose can be hydrolysed to monosaccharides by sucrose synthase (SuSy) or alkaline invertase (AI) (Morell and Copeland, 1985;Sturm et al., 1999). In the present analysis a SuSy isoform was identified (TC100410), which is known to be the nodule-enhanced isoform in M. truncatula nodules (Hohnjec et al., 2003), as well as a putative beta-fructofuranosidase/alkaline invertase (TC106886) recently reported to be expressed in Lotus japonicus during nodule development (Flemetakis et al., 2006). The next largest functional class of proteins is involved in redox state control and defence against biotic and abiotic stress. Besides the dominating presence of leghemoglobin, a key protein controlling the internal oxygen concentration within the nodule, a set of enzymes involved in antioxidant defence were also found. These included members of the ascorbate/glutathione cycle, such as mono-and dehydroascorbate reductase and glutathione reductase, reflecting the active antioxidant defence which occurs within nodules (reviewed in Matamoros et al., 2003). Several pathogen-related proteins were also detected, including a member of the PR-10 family, PR10-1, which has been reported as constitutively expressed in M. truncatula roots (Mathesius et al., 2001;Watson et al., 2003) and upregulated upon pathogen infection in roots and leaves (Gamas et al., 1998;Colditz et al., 2004;2005).
Furthermore, the present proteomic analysis allowed for the identification of several proteins known to be involved in signalling processes in N 2 -fixing nodules. Two calmodulin-like proteins (CaML) were identified, corresponding to proteins CaML 2 and 6b (for which amino acid sequence information is provided in table S4 as supplemental material). A third protein (CaML 4) was also detected but, as only one single peptide was found, it was not included in the protein identification list. A group of six of these calmodulin-like proteins were considered to be specifically expressed in M. truncatula root nodules in the in silico transcript analysis carried out by Fedorova et al. (2002).
These putative calmodulin-like proteins were further characterised in a recent study by Liu and coworkers (2006) in which, using specific antibodies, the authors identified CaML 2 and 5 based on single amino acid peptide sequences. The detection of 2 of these low abundance calmodulin-like proteins in the present study is further evidence for the potential of non-targeted liquid chromatography-based techniques

Quantitative proteomic analysis of M. truncatula root nodule
To evaluate the physiological effect of the imposed water deficit, two parameters were measured: nodule water potential (Ψ W ) and apparent nitrogenase activity (ANA), which was evaluated as H 2 evolution ( Fig. 4A and 4B). In drought-stressed plants, Ψ W showed a slight decline on day 3 (-1.07 ± 0.08 MPa), but the decrease was only statistically significant on day 6 (-2.42 ± 0.29 MPa). Values for control plants remained constant over the course of the experiment (-0.81 ± 0.05 MPa). In parallel, drought-stress caused a 30% and 70% reduction in ANA on day 3 and 6 respectively when compared to control plants. These results are in agreement with several previous studies (reviewed in Zahran, 1999).
Following physiological characterisation, nodules were analysed to obtain an overview of changes at the protein level. Extracts were separated into plant and bacteroid fractions and independent analyses were carried out. The obtained mass spectra of the plant and bacteroid fractions were searched against the TIGR M. truncatula Gene Index release 8.0 database (http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb=medicago) and the S. meliloti strain selection of the relatively most abundant proteins found in every fraction is shown in Tables I and   II, together with graphs showing their relative abundance during the drought treatment. The complete list of proteins together with their spectral count values is included as supplemental material (Tables S2 and S3 for plant and bacteroid analysis, respectively). Also provided is the amino acid sequence information obtained from the mass spectra, including a selection of the different peptides found per protein and their Sequest-identification scores (Tables S4 and S5).
One of the main problems when dealing with large complex datasets is extracting the important variables in an unsupervised way which allows for the interpretation of the results. In our case, just considering the plant fraction, we are dealing with around one hundred proteins measured, two treatments, at two time points, and five biological replicates per sample. For analysing these multifactorial and multivariate data sets statistical techniques are needed. In the present study, MetaGeneAlyse was used, a web-based service which provides the integrated analysis of data sets containing genetic, proteomic or metabolite information (Daub et al., 2003;Scholz et al., 2004).

General down-regulation of nodule plant metabolism under water-deficit
Five nodule plant proteins showed a statistically relevant variation during drought stress: methionine synthase, sucrose synthase, asparagine synthetase, leghemoglobin and the transcriptional elongation factor eEF-2, with the relative content of all of them decreasing in the drought samples (see the upper section of Table I). This occurred at day 3 for most of the proteins, when nitrogen fixation had only declined by 30% compared to control plants (Fig. 4B).
The plant protein showing the strongest response to drought stress was identified as methionine synthase (TC106598), which is involved in both the de novo synthesis of methionine and in the regeneration of the methyl group SAM. In plants it has been estimated that about 20% of the methionine is incorporated into proteins while 80% is converted to SAM (Giovanelli et al., 1985).
SAM functions as a primary methyl-group donor and, most interestingly, as a precursor for metabolites such as ethylene (Adams and Yang, 1977). Thus, methionine occupies a central position in cellular metabolism: as a protein constituent, in the initiation of mRNA translation and as a component of the regulatory molecule SAM. Although it is assumed that methionine synthesis, accumulation and consumption are under stringent regulatory control (Hesse and Hoefgen, 2003;Hesse et al., 2004), the molecular and biochemical characterization of plant methionine synthase is still limited. Very little is known about the role of methionine synthase in N 2 -fixing nodules, and to our knowledge this is the first report in which an abiotic stress is linked to a decrease on the relative content of this protein. Furthermore, it may be speculated that a decline in the levels of methionine synthase may cause a reduction in SAM content in drought-stressed nodules.
The second highest impact on the separation of control and drought samples is attributed to TC100410, which corresponds to nodule-enhanced SuSy (Hohnjec et al., 2003). Root nodules are highly specialised sink tissues, in which at least one SuSy isoform is strongly induced (Morell and Copeland, 1985) and reduction in SuSy activity is considered to be one of the key factors responsible for the inhibition of SNF during drought (González et al., 1995;Gordon et al., 1997;Gálvez et al., 2005). The fact that the present study also identified a decline in this protein as part of the plant response to water-deficit highlights the usefulness of proteomics as an unbiased approach to the detection of stress markers.
The next proteins found to be involved in the plant response to water-deficit were identified as two isoforms of asparagine synthetase (AS; TC100391, TC100393). In temperate-climate legumes, fixed nitrogen is mainly exported from nodules to the rest of the plant as asparagine, which is synthesized by the concerted action of two enzymes, aspartate aminotransferase (AAT) and AS (Vance et al., 1994). In alfalfa and soybean nodules AS has been described in detail (Huber and Streeter, 1985;Shi et al., 1997), whilst information about this enzyme in nodules from M. truncatula is mostly restricted to transcriptomic data (Györgyey et al., 2000;Carvalho et al., 2003;Colebatch et al., 2004;Tesfaye et al., 2006). In the present work AS showed a decrease in level while other enzymes involved in the nitrogen assimilation pathway in nodules, such as AAT or glutamine synthetase (GS), showed no response. Interestingly, a possible role of phosphorylation as a post-translational regulatory mechanism of GS activity in M. truncatula nodules has recently been reported (Lima et al., 2006). Whether this type of regulation takes place during drought stress is a question for further exploration.
Leghemoglobins (Lb) are symbiotic oxygen carrying proteins found at millimolar concentrations in the cytoplasm of bacteroid containing nodule cells. They are essential for the control of low internal oxygen while maintaining an adequate supply for bacteroid respiration (Appleby, 1984;Ott et al., 2005). In the present proteomic analysis, the relative content of Lb was by far the highest compared to other nodule plant proteins and up to eight Lb isoforms, some of them sharing a high sequence similarity, were detected and quantified. Drought caused a significant decrease in one leghemoglobin isoform (TC106593). Lb has been suggested to play a role during abrupt water stress in nodules, whereas gradual or moderate water deficit did not affect Lb content (González et al., 2001;Marino et al., 2006). In this study, although the drought treatment applied can be considered as gradual, the LC/MS/MS technique allowed for the quantification of small variations in protein content in an isoform-specific manner.
Finally, a protein with high similarity to eukaryotic elongation factor 2 (eEF-2; TC93936) was highlighted by the statistical analysis. eEF-2, formerly known as aminoacyltransferase II, catalyses the translocation of tRNAs, facilitating the movement of the ribosome relative to the mRNA during protein synthesis (Proud, 1994). In plants homolog genes have been cloned in sugar beet and Arabidopsis thaliana (Vogel et al., 1999;Guo et al., 2002) but little is known about the role of this elongation factor in legume nodules. Guo and collaborators (2002) reported that Arabidopsis mutants in this gene failed to induce the transcription of cold-responsive genes. Furthermore, they showed that this elongation factor was involved in new protein synthesis during cold stress. The decline in the relative content of the eEF-2-like protein in nodules subjected to drought stress may be related to a down regulation of de novo protein synthesis under water-limiting conditions. Water-deficit stress causes a well-documented decline in SNF in N 2 -fixing legumes (Arrese-Igor et al., 1999;Serraj et al., 2001;Zahran, 1999) which is also shown for nodulated M. truncatula plants in the present study (Fig. 4B). It can be inferred from the proteomic analysis that the plant response in nodules involves a global reduction of protein biosynthesis and a down-shift of cellular carbon, nitrogen and sulfur metabolism, thus reducing the energy-demanding process of nitrogen fixation. Altogether, this study shows a clear integration of both physiology and proteome analyses, with the added value of identifying which specific proteins, up to the level of isoforms, are first affected by water deficit.

Bacteroid upregulation of protein biosynthesis as an adaptation mechanism to drought stress
Although proteomic characterisation of S. meliloti bacteroid has been previously described (Djordjevic, 2004), some of the 97 proteins identified in the present comparative analysis (supplementary material, Table S3) deserve special mention. Several chaperonin proteins, including GroEL, GroES and heat shock proteins, essential to the establishment of the symbiosis (Yeh et al., 2002), were detected. Indeed, the largest spectral count number corresponded to a member of the GroEL family, assigned to GroEL1 (Djordjevic et al., 2003). The second most abundant category was comprised of proteins related to SNF, predominantly components of the nitrogenase complex, such as NifD, NifH and NifK. It is generally accepted that bacteroids assimilate very little fixed ammonia, which is mainly exported to the host plant (Brown and Dilworth, 1975;Vance et al., 1994). However, the present study identified several nitrogen assimilation enzymes which have been described as absent or present at very low levels in bacteroids, such as glutamine synthetase (GlnA), nitrogen regulatory proteins PII (GlnB) and PIIA (PtsN) and urease accessory protein (UreE). This is consistent with the recent hypothesis that an active amino acid cycling between both partners occurs in nodules rather than a simple export of NH 4 + from bacteroids to the host plant cell (Lodwig et al., 2003).
In a similar manner to the plant analysis, bacterial fractions from control and water-stressed nodules were separated, and changes at the protein level were analysed. Table II  The other two identified bacteroid proteins which responded to drought stress are related to protein biosynthesis. One of these is a putative glutamyl-tRNA amidotransferase (GatB; SMc01350) which catalyses the biosynthesis of glutaminyl-tRNA, an intermediate in RNA translation (Gagnon et al., 1996). 50S ribosomal protein L3 (RplC; SMc01309) was also found to be more abundant in drought-stressed bacteroids. Under stress conditions, protein biosynthesis and ribosomal genes are reported to be repressed due to slower cell growth. However, the observed increases of glutamyl-tRNA amidotransferase and ribosomal protein RplC suggest that bacteroids from nodules experiencing water deficit may temporarily activate protein biosynthesis as a stress adaptation mechanism.

Symbiotic partners respond simultaneously, but differently, to drought stress
The relative influence of each symbiotic partner on the regulation of nitrogen fixation in the legume-Rhizobium symbiosis is subject of debate. It has been suggested that the massive amplification of bacterial numbers in the nodule leads to the proteome being dominated by bacterial proteins (Djordjevic et al., 2003). In order to further investigate the relative influence of both microand macrosymbiont proteome on the nodule response to drought, both protein data sets were combined and an integrative ICA was carried out (Fig. 5C). After data normalisation, the integrative statistical analysis suggested that both plant and bacteroid fractions respond simultaneously to water-deficit and with similar statistical significance at the protein level. However, whilst there was a general down regulation of proteins related to plant metabolism, bacteroid cells were upregulating protein biosynthesis, probably as an adaptation to the water deficit imposed. The simultaneous proteomic analysis of both symbiotic partners has been shown to be a useful tool which, together with transcriptomic developments such as the dual-genome chip M. truncatula-S.
meliloti (Barnett et al., 2004), will help to further understanding of this intriguing symbiotic interaction.

CONCLUSIONS
The identification of 377 plant proteins using 2D-LC/MS/MS has confirmed the efficacy of liquid chromatography-based methods for proteomic analysis in complex tissues such as legume nodules. This profiling can now serve as a nodule proteome database for future targeted protein studies. The independent analysis of the nodule plant and bacteroid fractions confirmed the role of SuSy as a key enzyme involved in drought stress and also identified new marker enzymes such as plant methionine synthase and bacteroid serine hydroxymethyltransferase. This study has highlighted several novel avenues for further investigation to better understand the complexity of the SNF response to drought stress.  Witty and Minchin (1998) using an electrochemical H 2 sensor (Qubit System Inc., Canada). Following this, root nodules were collected, frozen in liquid nitrogen and stored at -80 ºC for further analysis. Four types of nodule samples were analysed: control plants at day 3 (C3), drought-stressed plants at day 3 (D3), control plants at day 6 (C6) and drought-stressed plants at day 6 (D6).
Pellets were washed twice with extraction buffer to avoid plant protein contamination and subsequently sonicated for bacteroid disruption. Sonicated fractions were further centrifuged at 10,000g at 4°C for 15 min and supernatants were collected as nodule bacteroid fractions. Soluble proteins were precipitated overnight at -20°C after adding 20 volumes of acetone containing 0.07% (v/v) 2-mercaptoethanol. Pellets recovered by centrifugation were rinsed with cold acetone solution, air dried and resuspended in 0.6 mL of solubilisation buffer (8 M urea, 0.1 M NH 4 HCO 3 , pH 8.5).

Immunoblot Analysis
Plant and bacteroid protein extracts (20 µ g) were separated by SDS-PAGE on 10% (w/v) polyacrylamide gels. Gels were transferred onto nitrocellulose membranes and blocked overnight with 5% (w/v) non-fat milk powder in Tris Buffer Saline (TBS). Polyclonal antibodies raised against nitrogenase complex components NifDK were incubated for 1h at 1:10,000 dilution in TBS.

Fast Protein Liquid Chromatography (FPLC)
For FPLC analyses 0.6 g fresh weight nodule was homogenized as described above and 5 mM PMSF proteinase inhibitor was added to the nodule homogenate. The plant protein fraction was sterile-filtered using a 0.45 mm filter (Schleicher and Schuell, Dassel, Germany) and the FPLC run was performed as previously described (Wienkoop et al., 2004). Briefly, a total of 25 mg nodule plant protein was loaded onto a 1-mL Resource Q column (Amersham-Pharmacia Biotech, Freiburg, Germany) equilibrated with 10-column bed volumes of buffer A (50 mM Tris-HCl, pH 8.0). The column was washed with buffer A until absorbance at 280 nm reached baseline level.
Bound proteins were eluted with a 50 mL-linear gradient from 0 to 500 mM NaCl in buffer A at a flow rate of 3 mL min -1 . Nine protein fractions of 2 mL were collected and protein was precipitated as described above.

In-solution Protein Digestion
Digestion was carried out according to Washburn et al. (2001). Aliquots containing 500 µ g of protein were digested for 5 h at 37°C with sequencing-grade endoproteinase Lys-C (1:100, v/v, Roche, Mannheim, Germany) in solubilisation buffer. Samples were then diluted in buffer containing 2 M urea, 10% (v/v) acetonitrile, 0.1 M NH 4 HCO 3 and 1 mM CaCl 2 (pH 8.5). Proteins were further digested overnight at 37°C with Porosyzme immobilized trypsin beads (1:10, v/v, Applied Biosystems, Darmstadt, Germany). After centrifugation for beads removal, the obtained peptide mixtures were desalted using SPEC C18 columns according to the manufacturer's instructions (Varian). Finally, desalted digest solutions were dried and pellets stored at -20°C until use.

Liquid Chromatography/Mass Spectrometric Analysis
Prior to the mass spectrometric measurement, protein digest pellets were dissolved in 5% ( Jose, CA) operated in a data-dependent mode. Each full MS scan was followed by three MS/MS scans, in which the three most abundant peptide molecular ions were dynamically selected for collision-induced dissociation (CID) using a normalized collision energy of 35%. The temperature of the heated capillary and electro spray voltage were 150°C and 1.8 kV, respectively.

Database Search
After mass spectrometric analyses, DTA files were created from raw files and were then  Liu et al. (2004). Spectral count is a semi-quantitative measure for tracking changes in protein abundance in complex samples, based on the cumulative sum of recorded peptide spectra that can be matched to a given protein. The validity of this label-free quantification method, compared to others such as peak integration, has been widely demonstrated (Cox et al., 2005;Old et al., 2005;Wienkoop et al., 2006, among others).

Statistical Analysis
Statistical independent component analysis (ICA) was carried out using MetaGeneAlyse (http://metagenealyse.mpimp-golm.mpg.de). A distance matrix was written and subsequently uploaded to the MetaGeneAlyse server. Firstly, principal component analysis (PCA) was applied as a pre-processing step for dimensionality reduction to a set of three principal components (PCs) and visualisation of relevant variances. ICA was then applied to this reduced data set for the evaluation of those covariant protein sets that were involved in the response of the plant and bacteroid fractions to water deficit and could be used to differentiate between the samples. The extracted independent components were then ranked by kurtosis as a measure of the influence of a specific protein on the separation of control and drought samples. For more details see "User manual" downloadable at the website and also Scholz and Selbig (2007).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. List of identified plant proteins in M. truncatula root nodules.
Supplemental Table S2. Quantitative proteomic analysis of the plant fraction of M. truncatula root nodules during drought stress.
Supplemental Table S3. Quantitative proteomic analysis of the bacteroid fraction of M.
truncatula root nodules during drought stress.
Supplemental Table S4. Summary of the amino acid sequence information and Sequest identification scores of the plant fractions of M. truncatula root nodules.
Supplemental Table S5. Summary of the amino acid sequence information and Sequest identification scores of the bacteroid fractions of M. truncatula root nodules.       First and third columns represent the average peptide ion count values of control nodules at day 3 (C3) and day 6 (C6), respectively. Second and fourth columns represent average values of droughtstressed nodules at day 3 (D3) and day 6 (D6), respectively. Bars represent standard error (n=3 biological replicates). For the complete set of proteins, see supplementary material (table S3).