Identification of ferredoxin II as a major calcium binding protein in the nitrogen-fixing symbiotic bacterium Mesorhizobium loti

Abstract

Background

Legumes establish with rhizobial bacteria a nitrogen-fixing symbiosis which is of the utmost importance for both plant nutrition and a sustainable agriculture. Calcium is known to act as a key intracellular messenger in the perception of symbiotic signals by both the host plant and the microbial partner. Regulation of intracellular free Ca²⁺ concentration, which is a fundamental prerequisite for any Ca²⁺-based signalling system, is accomplished by complex mechanisms including Ca²⁺ binding proteins acting as Ca²⁺ buffers. In this work we investigated the occurrence of Ca²⁺ binding proteins in Mesorhizobium loti, the specific symbiotic partner of the model legume Lotus japonicus.

Results

A soluble, low molecular weight protein was found to share several biochemical features with the eukaryotic Ca²⁺-binding proteins calsequestrin and calreticulin, such as Stains-all blue staining on SDS-PAGE, an acidic isoelectric point and a Ca²⁺-dependent shift of electrophoretic mobility. The protein was purified to homogeneity by an ammonium sulfate precipitation procedure followed by anion-exchange chromatography on DEAE-Cellulose and electroendosmotic preparative electrophoresis. The Ca²⁺ binding ability of the M. loti protein was demonstrated by ⁴⁵Ca²⁺-overlay assays. ESI-Q-TOF MS/MS analyses of the peptides generated after digestion with either trypsin or endoproteinase AspN identified the rhizobial protein as ferredoxin II and confirmed the presence of Ca²⁺ adducts.

Conclusions

The present data indicate that ferredoxin II is a major Ca²⁺ binding protein in M. loti that may participate in Ca²⁺ homeostasis and suggest an evolutionarily ancient origin for protein-based Ca²⁺ regulatory systems.

Background

Rhizobia are Gram-negative soil bacteria, which establish in the rhizosphere an intimate mutualistic interaction with leguminous plants. During this symbiosis, in which new organs called nodules are de novo generated on plant roots, the microsymbiont provides the plant with fixed nitrogen, receiving in turn fixed carbon from the host [1]. Nitrogen-fixing symbiosis is one of the most important beneficial plant-microbe interactions, supplying dozens of million tons of reduced nitrogen every year to agricultural systems [2]. The rhizobial bacterium Mesorhizobium loti and the leguminous plant Lotus japonicus, whose genomes have been sequenced [3,4], are widely used as model organisms in the study of rhizobium-legume symbiosis [5]. In the last decade a great deal of scientific attention has been focused on the biochemical dialogue that is progressively established in the rhizosphere, based on the reciprocal exchange of diffusible signals between the two symbiotic partners in the soil. In particular, the role of calcium as critical intracellular messenger involved in the perception and transduction of symbiotic signalling molecules has become increasingly apparent. A Ca²⁺-mediated symbiotic signalling pathway has been shown to be activated by symbiosis-related signals not only in the host plant [6], but also in the rhizobial partner [7-10]. Measurements of intracellular Ca²⁺ concentration by means of recombinant aequorin expression demonstrated a tight regulation of cytosolic Ca²⁺ level in M. loti and the occurrence of Ca²⁺-based mechanisms to detect changes in the soil environment [7].

A fine-tuned Ca²⁺ homeostatic machinery underlies the use of Ca²⁺ as intracellular messenger for signal transduction in any biological systems, either eukaryotic or bacterial [11,12]. The complex Ca²⁺ homeostasis toolkit of eukaryotic cells is made of several membrane-located active Ca²⁺ transporters (Ca²⁺ pumps and/or Ca²⁺ exchangers), that mediate the efflux of the ion into the extracellular milieau or its sequestration into different intracellular Ca²⁺ stores [13]. Several intracellular compartments are equipped with a variety of soluble Ca²⁺ buffering proteins, able to bind Ca²⁺ with different affinities and capacities [14]. Recent evidence indicates that Ca²⁺ buffers are likely to play a critical role in Ca²⁺ homeostasis also in bacteria. In the cyanobacterium Anabaena sp., the Ca²⁺-binding protein CcbP has been shown to be involved in the regulation of intracellular free Ca²⁺ concentration during heterocyst differentiation [15,16]. Based on nuclear magnetic resonance spectroscopy analysis of the structures of Ca²⁺-free and Ca²⁺-bound CcbP forms, it has been proposed that CcbP may represent a novel class of Ca²⁺-binding proteins, involved in the cytosolic sequestration of the ion in Anabaena sp. [17].

In this work, we investigated the possible occurrence of Ca²⁺ binding proteins as a component of the Ca²⁺ homeostatic machinery of the nitrogen-fixing bacterium M. loti. Herewith we provide evidence for the presence of a small soluble acidic protein, identified by Q-TOF MS/MS analyses as ferredoxin II that, on the basis of its biochemical features, may act as a major Ca²⁺ buffer in this rhizobial species.

Results

Isolation and biochemical characterization of Ca²⁺ buffering proteins in Mesorhizobium loti

As a first step in the search for a proteinaceous Ca²⁺-buffering system in M. loti, we considered the possibility that rhizobia might express proteins sharing a certain degree of amino acid sequence homology with the Ca²⁺-binding protein CcbP of the cyanobacterium Anabaena, which has been shown to regulate the bacterial intracellular free Ca²⁺ concentration during heterocyst differentiation [15-17]. Bioinformatic searches into the complete genome of M. loti [3] did not highlight any gene sharing a significant degree of nucleotide sequence identity with the Anabaena CcbP-encoding gene (GenBank accession no. AY919604). As a consequence, a biochemical approach was undertaken to search Ca²⁺-binding proteins potentially acting as Ca²⁺ buffers in M. loti.

Crude protein extracts were prepared from mid-exponential phase cultures of M. loti strain 3147^T and fractionated into a soluble fraction and a membrane fraction. To isolate soluble Ca²⁺ buffering proteins, a selective ammonium sulfate precipitation procedure, followed by anion exchange chromatography on DEAE-Cellulose, was applied. This experimental procedure has been traditionally used for the purification of acidic Ca²⁺-binding proteins, such as calsequestrin [18] and calreticulin from both animal [19] and plant [20] tissues. The same procedure has been later adopted for the isolation of the acidic Ca²⁺-binding protein CcbP from of the cyanobacterium Anabaena [15]. After separation on SDS-PAGE of protein samples, the slab gels were incubated with the cationic carbocyanine dye Stains-all, that is commonly used for the preliminary detection of acidic Ca²⁺-binding proteins [21]. As shown in Figure 1, a M. loti protein with an apparent molecular mass of about 20 kDa stained blue with Stains-all, whereas most other proteins stained pink. As a positive control, calsequestrin, purified from rabbit skeletal muscle, was used (Figure 1). The soluble M. loti Stains-all blue-staining protein was eluted from the DEAE-Cellulose column at 0.45-0.55 M NaCl, showing a several-fold level of enrichment (Additional file 1).

Figure 1

Isolation of acidic Ca ²⁺ buffering proteins from M . loti . Stains-all staining of SDS-PAGE (10-12.5% polyacrylamide linear gradient gel). Acidic proteins which bind Ca²⁺, such as calsequestrin and calreticulin, stain blue with Stains-all, whereas other proteins stain pink [21]. Lane 1, M. loti crude extracts (50 μg); lane 2, soluble fraction (50 μg); lane 3, protein sample after selective precipitation with ammonium sulphate (50 μg); lane 4, calsequestrin purified from rabbit skeletal muscle, used as a positive control (1 μg). Arrow, M. loti Stains-all blue-staining band with an apparent molecular mass of 20 kDa; arrowhead, rabbit calsequestrin.

It has to be noted that Stains-all blue-staining cannot be considered as necessarily diagnostic for acidic Ca²⁺-binding proteins, since this dye metachromatically stains also phosphorylated proteins as well as sialoglycoproteins [22,23]. Therefore, the potential ability of the M. loti protein to bind Ca²⁺ was further investigated by a Ca²⁺-dependent electrophoretic mobility shift assay [24]. To this aim, M. loti proteins partially purified by DEAE-Cellulose were electrophoresed on SDS-PAGE containing either EGTA or CaCl₂. As shown in Figure 2, the major M. loti Stains-all blue-staining band of about 20 kDa shifted to a lower apparent M_r in the presence of Ca²⁺, like rabbit calsequestrin did.

Figure 2

Ca ²⁺ - dependent changes of electrophoretic mobility. Proteins were electrophoresed on 12.5% SDS-PAGE either in the presence of 1 mM EGTA (A) or 1 mM CaCl₂ (B) and stained with Stains-all. Lane 1, 20 kDa M. loti protein partially purified by DEAE-Cellulose (2 μg, arrow); lane 2, rabbit calsequestrin, used as positive control (0.5 μg, arrowhead).

Since eukaryotic high-capacity, low-affinity Ca²⁺ binding proteins such as calsequestrin and calreticulin are characterized by an acidic isoelectric point [25], isoelectric focusing (IEF) analyses of the M. loti putative Ca²⁺-binding protein were carried out. The M. loti Stains-all blue-staining protein band was indeed found to migrate in the low pH region of the gel (Figure 3) and, when analysed by liquid phase IEF, exhibited a pI of 4.2 (data not shown).

Figure 3

Isoelectric focusing of the DEAE -Cellulose fraction from M . loti . The slab gel was stained with Stains-all. About 15 μg of protein was loaded. Arrow indicates the major Stains-all blue-staining protein band of 20 kDa.

To further purify the protein to homogeneity, electroendosmotic preparative electrophoresis (EPE) was performed, yielding a fraction containing the virtually pure 20 kDa M. loti protein (fraction 8 in Additional file 2). The Ca²⁺ binding ability of the purified protein was confirmed by ⁴⁵Ca²⁺ overlay assays, in which calsequestrin and cytochrome c were used as positive and negative controls, respectively (Figure 4).

Figure 4

Identification of M . loti Ca ²⁺ binding proteins by ⁴⁵ Ca ²⁺ -ligand overlay. Proteins were dot-blotted onto nitrocellulose and then incubated with ⁴⁵CaCl₂ in a calcium overlay assay [44]. ⁴⁵Ca-labeled proteins were visualized by autoradiography (7-d exposure). 1, equine cytochrome c, used as negative control (5 μg); 2, purified M. loti 20 kDa Stains-all blue-staining protein (5 μg); 3, rabbit calsequestrin, used as positive control (1 μg).

Identification of the M. loti Ca²⁺ binding protein by mass spectrometry

Mass spectrometry (MS) analyses, carried out after digestion with either trypsin or endoproteinase AspN, identified the protein purified from EPE as ferredoxin II. Best coverage and identification results were then obtained from fraction 8, while some ferredoxin II peptides were identified below significance threshold also in fraction 7 and 9 (data not shown). This is probably due to different abundance of the protein in each fraction or to uncovered post-translational modifications of its residues in fraction 7 and 9. Overall sequence coverage of 54% was reached among trypsin and AspN digestion, thus confirming the presence of ferredoxin II in fraction 8 (Figure 5). With reference to the RhizoBase at Kazusa Genome Resources, there are five ferredoxins (msl0793, mll5100, mlr5869, mlr5930, msl8750) plus a ferredoxin-like protein (msl5859) and a probable ferredoxin (msr9193) in Mesorhizobium loti with low sequence homology to the one identified by our analyses (mlr3855). Moreover, we retrieved sequences from at least other three ferredoxins spanning high sequence homology to identified mlr3855 in Rhizobium leguminosarum, Sinorhizobium meliloti and Bradyrhizobium japonicum, thus supporting a relevant role for this protein in nitrogen-fixing bacteria (Figure 6). In silico sequence analysis with ScanProsite tool [26] (http://prosite.expasy.org/scanprosite/) confirmed the presence of two 4Fe-4S ferredoxin-type iron-sulfur binding domain profiles between residues 1-30 and 31-60 in the cysteine-rich region of mlr3855, which we partially covered by MS analysis.

Figure 5

One letter code amino acid sequence of ferredoxin II (mlr3855) from Mesorhizobium loti strain MAFF303099. Peptides identified by MS analysis are colored in red.

Figure 6

Alignment of ferredoxin sequences from different rhizobial species. Ferredoxin II (mlr3855) from M. loti MAFF303099 showed 85% amino acid sequence identity with a putative ferredoxin II (RL4616) from Rhizobium leguminosarum bv. viciae 3841, 75% with a putative ferredoxin protein from Sinorhizobium meliloti 1021 (SMc03875) and 83% with ferredoxin (bll0157) from Bradyrhizobium japonicum USDA110. mlr3855 sequence covered by MS analysis is highlighted in yellow.

Interestingly, the doubly charged tryptic fragment YFSAEPGEGD (m/z 536.2123) was also identified in a form modified by an adduct with Ca²⁺ (m/z 555.1855) (Additional file 3), further strengthening the notion that the M. loti protein is indeed a Ca²⁺-binding protein.

The protein has a predicted molecular mass of 12553.13 Da and a theoretical pI of 4.18, in perfect agreement with the value (4.2) identified by IEF analyses. The discrepancy between the actual molecular mass and the apparent molecular weight on SDS-PAGE (Figures 1, 2, 3) is a common feature of acidic Ca²⁺ binding proteins, which often migrate anomalously in Laemmli SDS-PAGE [25].

Discussion

In this work we have provided evidence for the occurrence of a major soluble Ca²⁺ binding protein in the nitrogen-fixing bacterium M. loti USDA 3147^T strain. This small acidic protein, which was isolated by using an experimental procedure commonly adopted for the extraction and purification of calsequestrin and calreticulin from animal tissue homogenates, was identified as ferredoxin II by tandem mass spectrometry analyses with an ESI-Q-TOF instrument. Ferredoxin II shares with eukaryotic Ca²⁺ buffering proteins several biochemical features, i.e. metachromatic staining with Stains-all, an acidic isoelectric point, the Ca²⁺-dependent change of electrophoretic mobility and the Ca²⁺-binding ability in ⁴⁵Ca²⁺ overlay assay. All these characteristics, together with mass spectrometry evidence for a Ca²⁺ adduct in the C-terminal region peptide YFSAEPGEGD, suggest that ferredoxin II may participate in Ca²⁺ homeostasis in M. loti.

Further investigations are needed to unravel the physical basis for ferredoxin II ability to bind Ca²⁺, since the protein lacks classical EF-hand motifs. However, it is possible to speculate that the abundance of negatively charged amino acid residues present in the C-terminal region of the protein, where aspartic and glutamic acid residues reach 40% of the sequence along the last ten residues, may play some role, as hypothesized for CP12, a novel Ca²⁺-binding protein which has recently been identified in the chloroplasts of Arabidopsis thaliana [27]. Moreover, based on the class assignment for solvent accessibility of the tool NetSurfP - Protein Surface Accessibility and Secondary Structure Prediction (Technical University of Denmark, http://www.cbs.dtu.dk/services/NetSurfP/), the last eight residues encompassing two glutamic and one aspartic acid are exposed to solvent in mlr3855, thus they are possibly accessible to Ca²⁺ and prone to formation of adducts, like the one confirmed here by MS/MS data.

Ca²⁺ is acknowledged as a ubiquitous and versatile intracellular messenger, able to transduce a wide variety of signals into proper cellular responses in all eukaryotic cells [13,28,29]. Increasing evidence suggest that the basic components of Ca²⁺ homeostasis and signalling machineries are present in bacteria as well [30-32]. In rhizobium-legume symbiosis, Ca²⁺ has been demonstrated to be involved in the perception of symbiotic signals not only by the plant host [33] but also by the bacterial partner. In particular, several plant-derived diffusible molecules, such as tetronic acid in M. loti [7] and the flavonoid naringenin in R. leguminosarum [8,10] were found to trigger transient intracellular Ca²⁺ changes leading to the expression of bacterial nodulation genes. Although the molecular identity of the Ca²⁺-permeable channels involved in the observed Ca²⁺ fluxes remains elusive, in this work another essential component of the Ca²⁺ homeostatic machinery, i.e. a protein Ca²⁺ buffer, was identified. It has recently been hypothesized that Ca²⁺ buffering proteins, in addition to quickly bind and thus remove free Ca²⁺ ions that can have toxic effect for the cell, can also play a crucial role in shaping the Ca²⁺ signals [34].

In the cyanobacterium Anabaena sp. strain PCC 7120 the Ca²⁺-binding protein CcbP was found to play an important role in the regulation of intracellular free Ca²⁺ concentration during heterocyst differentiation [15]. In particular, the protein was found to be degraded by the serine-type protease HetR, thereby releasing bound Ca²⁺ and significantly contributing to the establishment of higher levels of cytosolic Ca²⁺ in mature heterocysts [16].

In a recent comparative proteomic analysis of M. loti MAFF303099, no change in the expression of ferredoxin II was found between the free-living and symbiotic condition [35], in agreement with previous results obtained by transcriptomic analysis [36]. Similarly, a putative ferredoxin II encoding gene (RL4616) in R. leguminosarum bv. viciae was not among the significantly upregulated genes in bacteroids compared with free-living cells [37]. These data suggest that ferredoxin II may play an essential constitutive function in rhizobia as a part of the bacterial Ca²⁺ homeostat. It remains to be established whether the transition from free-living bacteria to bacteroids may involve changes in intracellular free Ca²⁺ concentration in M. loti, thereby possibly involving a post-translational regulatory mechanism for the level of ferredoxin II, similar to that described for cyanobacterial CcbP.

It cannot be ruled out that M. loti ferredoxin II may play additional roles within the rhizobial cell. Based on the Rhizobase at Kazusa DNA Research Institute, M. loti ferredoxin II falls within the functional category of central intermediary metabolism and nitrogen metabolism [35]. Interestingly, a protein sharing a significantly high homology (83%) with M. loti ferredoxin II is found also in Rhodopseudomonas palustris, a metabolically versatile photosynthetic bacterium able to fix nitrogen. A relevant similarity between genes of rhizobia and Rhodopseudomonas is not unexpected as they share a close terminal branch in 16S rRNA phylogeny [38], which entails recent separation and consequently high colinearity of their genomes.

Indeed, a commonality among eukaryotic organellar Ca²⁺ buffering proteins is to be multifunctional proteins, carrying out additional important functions not necessarily directly related to Ca²⁺ homeostasis [14]. In particular, the presence of Fe-S clusters renders ferredoxin II particularly suited for electron transfer during redox reactions, similarly to ferredoxin. Interestingly, Ca²⁺ binding to a plant ferredoxin has been hypothesized to affect both the buffering capacity of the chloroplast stroma and the formation of complexes with ferredoxin-dependent oxidoreductases [39].

Our findings support the notion that proteins with the ability to bind Ca²⁺ appeared early in the history of life, possibly contributing to the regulation of intracellular free Ca²⁺ levels and to the consequent evolution of Ca²⁺-based signalling mechanisms [12].

Conclusions

In this work we demonstrated that ferredoxin II is a major Ca²⁺ binding protein in M. loti, potentially acting as a primitive system to control intracellular free Ca²⁺ concentration in this rhizobial species. The elucidation of the mechanisms involved in Ca²⁺ homeostasis in rhizobia may help to better understand the Ca²⁺ signalling events underlying the early stages of nitrogen-fixing symbiosis with host legumes.

Methods

Strain and growth conditions

M. loti USDA strain 3147^T, kindly provided by P. Van Berkum (USDA, Beltsville MD), was cultured in minimal BIII medium [40] at 28°C, as previously described [7]. For protein sample preparations, cells were harvested at OD_600nm = 0.6 (mid-growth phase).

Isolation and purification of acidic Ca²⁺ binding proteins

To prepare protein crude extracts from M. loti suspension cell cultures, cells were harvested by centrifugation at 3000 g for 20 min at 4°C, washed twice with fresh medium, centrifuged again and resuspended in 4 volumes of phosphate-buffered saline (PBS) supplemented with 0.5 mM benzamidine and 0.5 mM phenylmethylsulfonyl fluoride. Bacteria were lysed by 3 cycles of 30 s each of sonication at 35 Hz (Fisher Sonic, Artek Farmingdale, NY, USA), each followed by 30 s on ice. After sonication, samples were centrifuged at 1600 g for 15 min at 4°C to pellet down and discard non lysed bacteria.

Total protein extracts (from about 25 g cells, fresh weight) were further fractionated in a soluble and membrane fractions by centrifugation at 150.000 g for 45 min at 4°C: the supernatant, containing soluble proteins, was subjected to a selective ammonium sulfate precipitation procedure followed by DEAE-Cellulose column chromatography, as described by Slupsky et al. [18]. Protein fractions from the DEAE-Cellulose column (DE52, Whatman, Maidstone, UK) collected between 0.4 and 0.5 M NaCl were further purified by using electroendosmotic preparative electrophoresis (EPE) [41], which was carried out with an ELFE apparatus (Genenco Life Science, M-Medical srl, Firenze, Italy). Protein concentration was determined by using the Bio-Rad Protein Assay (Bio-Rad Laboratories, Hercules, CA, USA), according to manufacturer’s instructions.

In some experiments calsequestrin, purified from rabbit skeletal muscle as previously described [42] was used as positive control.

Gel electrophoresis analyses

One-dimensional SDS-PAGE was performed according to Laemmli [43], using 10 to 12.5% polyacrylamide linear gradient gels or 12.5% gels. Ca²⁺-dependent electrophoretic mobility shift assays were performed by separating protein samples on SDS-PAGE slab gels containing either 1 mM EGTA or 1 mM CaCl₂, both in the stacking and in the separating gels [24].

For in gel-isoelectric focusing (IEF), purified M. loti proteins were thoroughly desalted and resuspended in 0.2 ml of 9 M urea, 2 M thiourea, 4% (w/v) 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS) and 2 mM tris(2-carboxyethyl)phosphine (TCEP) using Micro Bio-Spin columns (Bio-Rad). Ampholytes, 3-10 pH range (Bio-Rad), and bromophenol blue were then added to the sample at 0.2% and 0.01% (w/v) final concentration, respectively. Isoelectric focusing (IEF) was performed on 11 cm IEF strips, 3-10 pH range (Bio-Rad). When 7 cm IEF strips, pH range 4-7 (Bio-Rad), were used, the separation was performed on 0.125 ml sample, in the presence of ampholytes, 4-7 pH range (Bio-Rad), (0.2%, w/v), for 10,000 V/h. Before the second dimension separation on a 10-12.5% gradient SDS-PAGE, the strips were equilibrated in 10 ml of 0.375 M tris-HCl, pH 8.8 containing 2% (w/v) SDS, 6 M urea, 20% (v/v) glycerol and 2% DTT for 10 minutes and then in the same buffer with 2.5% (w/v) iodoacetamide.

Liquid-phase IEF was instead performed by using a MicroRotofor Cell (Bio-Rad). In this case M. loti proteins were thoroughly desalted on NAP 5 columns (GE Healthcare, Little Chalfont, Buckinghamshire, UK) and resuspended in 2.5 ml of focusing buffer: 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 2 mM TCEP and 3% (w/v) ampholytes, 3-10 pH range. Proteins were separated at constant power (1 W) for 5 h using 0.1 M H₃PO₄ at the anode and 0.1 M NaOH at the catode. At the end of the run ten protein fractions of almost the same volume were collected and used for further studies.

Slab gels were stained with Coomassie Blue, completely destained and then restained with Stains-all, as described by Campbell et al. [21].

⁴⁵Ca²⁺ overlay assay

⁴⁵Ca²⁺ overlay assay was carried out by the technique described by Maruyama et al. [44]. Proteins were spotted directly onto nitrocellulose membranes. Rabbit calsequestrin was used as positive control, equine cytochrome c (Sigma-Aldrich) was used as negative control. ⁴⁵CaCl₂ was purchased from Perkin-Elmer (Boston, MA, USA). ⁴⁵Ca²⁺-labeled proteins were visualized by autoradiography on Amersham Hyperfilm MP (GE Healthcare).

Q-TOF analyses

For Q-TOF analyses, proteins purified by the EPE technique (fractions 7-8-9 separately) were used, after buffer exchange with Vivaspin 6 concentrators with a 5000 MW cut-off (Sartorius Stedim Biotech GmbH, Goettingen, Germany) against 40 mM NH₄HCO₃, 10% (v/v) acetonitrile pH 8.0. Each fraction was then digested either with trypsin and endoproteinase AspN based on obtainable coverage predicted in silico [45] (http://web.expasy.org/peptide_mass/). 100 ng of each protease were used for overnight digestion at 37°C and reaction was stopped by acidification with formic acid (FA) 1% (v/v) final concentration. Digests were dried by vacuum centrifugation and resuspendend into a minimal volume of 0.1% FA for subsequent MS analysis. 1 to 10 μl of each peptide mixture, obtained from the digestion of protein samples, were analyzed by means of reversed-phase chromatography on a nano-fluidic HPLC-Chip apparatus coupled with a quadrupole ion trap and time of flight mass spectrometer, using the 6520 Accurate-Mass Q-TOF LC/MS system (Agilent Technologies) equipped with MassHunter Workstation Software Qualitative Analysis B.02.00 as graphical interface for data handling. A 1200 Rapid Resolution system (Agilent Technologies, Santa Clara, CA, USA) containing a binary pump and degasser and a well-plate autosampler with thermostat were associated to the HPLC-Chip interface directly connected to a nanoESI ionization source. Loaded samples were thus enriched on a 160 nl enrichment column and separated with an acetonitrile gradient on a 75 μm x 150 mm separation column packed with Zorbax 300SB-C18 5 μm material. Data dependent MS/MS analysis were carried on the 3th most intense peaks from each MS scan using collision induced dissociation fragmentation (CID) with dynamic exclusion in order to enhance sequencing of less abundant peptides.

Data analysis

MS/MS data were extracted for identification purpose from raw data by means of MassHunter Workstation Software Qualitative Analysis and converted to Mascot generic format datafile (.mgf). Thus, data were analysed with Mascot server (ver. 2.3, Matrix Science Ltd.) being classified by a probability based implementation of the Mowse algorithm: experimental mass spectra produced were correlated to peptide sequences obtained by comparison with the theoretical mass spectra in the RhizoBase protein database downloaded from the Kazusa DNA Research Institute website (http://genome.microbedb.jp/RhizoBase). As the confidence of protein identification a precursor mass tolerance of no more than 10 ppm and MS/MS product mass tolerance of no more than 0.05 Da were adopted. Methionine oxidation, asparagine/glutamine deamidation and cysteine disulfide or acrylamide adducts formation were considered as routine possible post-translational modification (PTM). Sequence alignments were performed by means of ClustalW2 alignment tool (http://www.ebi.ac.uk/Tools/msa/clustalw2/) using Gonnet protein weight matrix.

Accession numbers

All MS/MS data were searched with Mascot server 2.3 against Mesorhizobium loti protein sequences downloaded from the RhizoBase database (http://genome.microbedb.jp/rhizobase/Mesorhizobium/genes.faa). Accession numbers of identified proteins refers to the ones contained in this database.