Introduction

Studies published so far on plant–pathogen interactions have been mainly focused on the molecular changes related to pathogen attack and/or plant response (Baker et al. 1997; Dangl and Jones 2001; Hammond-Kosack and Parker 2003; Martin et al. 1993; Ronald 1997; Suzuki et al. 2004). Several signal molecules and defence factors have been identified in plant (Cánovas et al. 2004; Ramonell and Somerville 2002; Rep et al. 2002), as well as virulence and avirulence factors in microbes (Kazemi-Pour et al. 2004; Smolka et al. 2003). Nevertheless, the molecular bases of multiple-player systems that may produce beneficial effects on plant health are largely unknown. Moreover, the influence that a biocontrol agent may have on the interactions between a plant and a pathogen has not yet been investigated by using proteomics, while this technique clearly represents an effective tool to analyze such biological processes (Woo et al. 2006).

Since biocontrol fungi of the genus Trichoderma have developed the ability to interact simultaneously both with plants and fungal pathogens, they can be used as model microorganisms to study complex and multiple-player plant–microbe interactions. Antagonistic Trichoderma spp. use numerous mechanisms against pathogens, including production of antifungal compounds, direct parasitism or inhibition of pathogen growth and, as determined more recently, induction of plant systemic and localized resistance (ISR and LAR) (Benítez et al. 2004; Brunner et al. 2005; Chet 1987; Handelsman and Stabb 1996; Harman and Kubicek 1998; Harman et al. 2004a; Lorito and Woo 1998; Sivasithamparam and Ghisalberti 1998). The plant interaction with Trichoderma, which often involves an ISR effect, normally increases fitness and ability to withstand both biotic and abiotic stresses (Bigirimana et al. 1997; De Meyer et al. 1998; Howell 2003; Miller and Jastrow 1990). De facto, the activity of these fungal biocontrol agents commonly determines the level of plant susceptibility or resistance to pathogens (Bigirimana et al. 1997; De Meyer et al. 1998; Lo et al. 2000; Lu et al. 2004).

Structural genomic studies have provided vast information on the identity and structure also of genes expressed in plant–microbe interactions (Baker et al. 1997; Singh et al. 2004; Talbot 2003). Functional genomics combined with bioinformatics provides an overall picture of the metabolic status of an organism in a given moment or condition (Pandey and Mann 2000; Zhu et al. 2003). In particular, proteomics permits a large-scale analysis of protein production, and in recent years has been widely used to investigate protein profiles produced from diverse interaction conditions and physiological states of cells and tissues (Pandey and Mann 2000; Zhu et al. 2003), especially in pathology-related research (Lim and Elenitoba-Johnson 2004).

By using a proteomic approach, we studied the concurrent interactions of the biocontrol agent T. atroviride strain P1 with a host plant and different fungal pathogens, in order to identify and analyze the proteins differentially produced by the three players. 2-D maps of protein extracts were obtained from plant and fungi singly and in any possible combination. Differential proteins in the gels were confidently identified and characterized by using tryptic digestion, mass spectrometry (MS) and in silico analysis. Proteins putatively important for plant–pathogen–antagonist interaction were analyzed in order to determine their accumulation pattern, gather hints on their role and eventually improve methods for disease control.

Materials and methods

Growth and interaction conditions

Bean plants (Phaseolus vulgaris L. cv. Cannellino) were grown in sterile soil. After 3 weeks bean leaves were collected, washed and placed on the top of 1.5% water agar (WA) plates 15 cm in diameter (Fig. 1). For root interaction, bean seeds were sterilized for 1 min with a 1% hypochlorite solution, rinsed with sterile water and placed in sterilized magenta filled with a sterile inert support (perlite). After seed germination, the magenta were opened and the seedlings were rinsed with water two times per week. Three-week-old plants were collected, their roots washed extensively, and entirely transferred to WA plates by laying them on the medium (Fig. 1).

Fig. 1
figure 1

Scheme of the plate system used to arrange three- and two-way interactions between plant, pathogens and the antagonist Trichoderma atroviride strain P1. Plant material (bean leaves or roots) was placed on the water agar (WA) plate of a large Petri dish covered with a sterile cellophane membrane (CM). Fungal mycelia of the antagonistic or pathogenic microbes, previously grown on PDA plates on a sterile CM, were layered on the top of the plant tissues

Full size image

The fungal pathogens Botrytis cinerea (strain 309, isolated from tobacco) and Rhizoctonia solani (strain 1556, isolated from tomato) were maintained on malt extract (SIGMA, St. Louis, MO, USA) and potato dextrose agar (PDA) (SIGMA) plates, respectively. One hundred microliters of a 5 × 105 spores/ml suspension of B. cinerea, and 1 cm2 of a R. solani colony were used to separately inoculate PDA plates covered with a sterile cellophane membrane (CM).

A spore suspension of T. atroviride strain P1 (ATCC 74058) obtained from a colony grown on PDA was used to inoculate CM-covered PDA plates (100 μl of a 5 × 105 spores/ml suspension). After 3–4 days of incubation at 25°C, the CM with the pathogen/antagonist mycelia was transferred on top of the WA plate containing the plant tissue (bean leaves or roots), as shown in Fig.  1. The CM used to separate the pathogens and the antagonist between themselves and from the plant allowed separation and transfer of fungal mycelia, but still permitted micro- and macro-molecules diffusion (Kullnig et al. 2000).

The plates hosting the three-player (plant–pathogen–antagonist) and the two-player (plant–pathogen, plant-antagonist) interactions and the relative controls (plant/fungi alone) were maintained at room temperature for 3 days and then the plant/fungi samples were separately collected for protein extraction.

Protein extraction

The protein extraction protocol described by Jacobs et al. (2001) was applied with some modifications. Approximately 1 g of fungal mycelium (wet weight) from the pathogen, the antagonist, or of a plant tissue (leaves or roots) was suspended in 10 ml of a cold (−20°C) acetone solution [20% trichloroacetic acid (TCA) and 0.2% dithiothreitol (DTT)] and ground with an ultraturrax (T25 basic, IKA Labortechnik, Germany) by keeping the tube in an ice bath. Samples were maintained at −20°C for at least 3 h to allow protein precipitation, then centrifuged (20 min, 30,000g at 4°C). The pellet was washed three times with cold (−20°C) acetone solution containing 0.2% DTT, then resuspended in a rehydration buffer [9 M urea, 2% 3-[(3-cholamidopropyl)-dimethyl-ammonio]-1-propane sulfonate (CHAPS), 1% DTT, 10 mM phenylmethylsulfonyl fluoride (PMSF)], vortexed and kept on an orbital shaker for 2 h to obtain complete protein solubilization. The samples were centrifuged (60 min, 30,000g at 20°C) and the supernatants were recovered. Protein concentration was determined by a Bradford Dc protein assay (Bio-Rad, Richmond, CA, USA) and samples were stored at −40°C until use.

Two-dimensional electrophoresis (2-DE)

Isoelectric focusing (IEF) was conducted by using 7 cm immobilized-pH-gradient (IPG) strips (Bio-Rad) with a pH range from 3 to 10, rehydrated in a solution of 9 M urea, 2% CHAPS, 1% DTT, 2% carrier ampholyte and 10 mM PMSF proteinase inhibitor (SIGMA). Two hundred microliters of the total protein solution (equivalent to 200 μg) were loaded in the focusing tray and absorbed into the gel strip (1 h passively at room temperature and 12 h actively with a 50 V current applied). IEF was carried out with a PROTEAN IEF Cell system (Bio-Rad). IPG strips were focused up to a total of 14 kVh by using a three-step program (250 V for 1 h, 4 kV for 3 h and until 10 kVh were reached). The strips were equilibrated by placing them in a solution of 6 M urea, 0.05 M Tris/HCl pH 8.8, 20% glycerol, 2% SDS, 2% DTT for 10 min, and then in 6 M urea, 0.05 M Tris/HCl pH 8.8, 20% glycerol, 2% SDS, 2.5% iodocetamide for 10 min more. IPG strips were finally loaded on a 15% polyacrylamide gel in a Mini-Protean 3 Cell (Bio-Rad), and run with a constant current of 150 V for 75 min in 1X tris-glycine-SDS (TGS) buffer (Bio-Rad). The same rehydration protocol was used for the 17 cm IPG strips. The IEF program was 300 V for 2 h, 10 kV for 4 h and until 40 kVh were reached. After equilibration, strips were loaded onto 8–16% polyacrylamide gradient gels for SDS-PAGE in a Protean plus Dodeca Cell (Bio-Rad) which was run at 10°C, with a constant current of 200 V for about 8 h. Gels were stained for at least 3 h with SimplyBlue SafeStain G-250 (Invitrogen, California, USA) according to the manufacturer instructions. Each protein extract was run on triplicate or duplicate gels for the 7 cm and 17 cm IPG strips, respectively.

Gel images were acquired by a GS-800 Imaging Densitometer (Bio-Rad) and analyzed with the PD-Quest software. Image files were recorded by using a red filter (wavelength 595–750 nm) and a resolution of 36.3 × 36.3 μm. The signal intensity of each spot was determined in pixel units (Optical Density, OD) and normalized to the sum of the intensities of all the spots included in the standard gel. Protein spots were considered to be differentially produced if at least a twofold intensity variation was detected when responses to different interaction conditions were compared.

In-gel digestion, mass spectrometry and in silico analysis

Protein spots were excised from gels and digested with trypsin (SIGMA), as described by Ha et al. (2002). Tryptic peptides were resuspended in 10 μl of a 1% acetic acid solution. The samples were mixed 1:1 with a matrix of a saturated α-cyano-4-hydroxycinnaminic acid solution [10 mg/ml acetonitrile (ACN)/0.2% trifluoroacetic acid (TFA), 70/30] (SIGMA), and 1 μl aliquots were applied to the MALDI (matrix-assisted laser desorption/ionization) sample plate and dried. Peptide mass spectra were obtained on a Voyager-DE Pro MALDI-TOF (time of flight) mass spectrometer (Applied Biosystem, Foster City, CA, U.S.A.) equipped with a 337 nm laser and delay extraction, operated in positive-ion reflector mode for the mass range between 890 and 3,500 Da. Mass calibration was performed with the ions from human adrenocorticotropic hormone—ACTH (fragments 18–39) (SIGMA) at 2,465.1989 Da, and Angiotensin III human (MP Biomedicals, Irvine, CA, USA) at 931.5154 Da as internal standards.

Peptide mass fingerprint (PMF) data were matched to the National Centre Biotechnology Information non-redundant (NCBInr) database entries against proteins from fungal, plant or all species, using the Mascot software (Matrix Science, London, UK). The following search parameters were applied. One incomplete cleavage was allowed and alkylation of cysteine by carbamidomethylation was set as possible modification. The Mascot program compares theoretical and experimental peptide values derived by virtual hydrolysis of proteins present in the database with a specific proteolytic agent, then supplies a list of hypothetical candidates with the probability that the peptides found belong to that entries. A modified Mascot analysis was also performed by using a Trichoderma Expressed Sequence Tags (EST) database, built as described by Suárez et al. (2005) and supported by TRICHOEST European Union project (http://www.trichoderma.org; Rey et al. 2004). The database contains more than 14,000 cDNA clones obtained from libraries of mRNAs isolated from the mycelia of different Trichoderma species, including T. atroviride strain P1, grown also in the presence of plant and/or pathogens. Similarities between the peptide fragmentation of some Trichoderma spp. clones and known proteins were determined.

Moreover, the Motif program (http://www.motif.genome.jp) was applied to Pfam and Pfam_fs databases to determine if the known proteins share a significant degree of homology with the analyzed protein spots or contained conserved domains in the areas that matched the peptide deduced sequences.

Some differential spots, such as those produced by B. cinerea in the interaction with bean leaves and Trichoderma or by bean leaves in the interaction with Trichoderma and B. cinerea, were subjected to MALDI-TOF/TOF analysis by using a Proteomics Analyzer 4700 (Applied Biosystems) that was calibrated immediately prior to each experiment. The samples were desalted and concentrated using ZipTips C18 (Millipore, Bedford, MA, USA), washed with 0.1% TFA solution (10 μl × 2), and eluted with 0.1% TFA in a 1:1 water : ACN solution (10 μl). They were applied onto the MALDI sample plate and treated with a matrix solution of saturated α-cyano-4-hydroxycinamic acid in a 1:1 water : ACN solution + 0.1% TFA. The analyte and matrix were air-dried at room temperature. The acquired MS/MS data were then submitted to the NCBInr database for protein identification by using the GPS Explorer Software with the integrated Mascot search engine.

Results

Multi-player interactions and proteome separations

The use of large water agar Petri dishes to support the interaction and cellophane membranes to separate the three players, permitted extensive contact and selective recovery of the different fungal and plant tissues (Fig. 1). Instead, co-cultures in liquid medium by separating the three partners with dialysis membranes did not allow a good interaction between the plant and the two microbes (data not shown).

Experimental procedures have been optimized by changing the length of the interaction time, the quantity of the microbial inoculum, the age of the plantlets transferred in the Petri plates, etc. in order to achieve the best separation and recovery of proteins from the live material. Generally, good yield and quality of proteins from both leaves and mycelia were obtained. For instance, 5–10 mg of total proteins were obtained from 1 g of fresh material (both from plant and fungal mycelia). On the other hand, the proteins from root samples were not well resolved on 2-D gels, even when a specific protocol for root protein extraction was applied (Saravanan and Rose 2004). Although the experiments were performed with different plants (i.e., tobacco, potato and tomato), only the results obtained on bean are included in the present report.

First, proteins were separated on 7 cm long IPG strips across a gradient of pH from 3 to 10, and results were successively confirmed by using a larger gel format (17 cm strip size). The gel analysis performed with the PD-Quest software (Bio-Rad) was found to be useful to generate the 2-D maps from the various conditions of interaction, and perform quantitative and qualitative analysis of the differential protein spots. In fact, the gels obtained from the interaction plates allowed the separation of hundreds of proteins and provided a representative picture of each proteome. For some spots, the homology found was confirmed by the presence in the known proteins of conserved domains matching the deduced sequences of the peptides obtained from 2-D gels. An example is provided by the matching of spot 3703 with a Brassica oleracea SGT1-like protein (gi|40974917) containing an SGS conserved domain (Fig. 2).

Fig. 2
figure 2

Peptide mass fingerprint of spot 3703 from bean leaves, in the presence of T. atroviride and B. cinerea, and identification of conserved domains. a Sequence of the Brassica oleracea SGT1-like protein (gi|40974917) found to be homologous to spot 3703. Matched peptides are shown in bold. The sequence of the SGS conserved domain is underlined. b Sequences of the digested peptides obtained. Start–End = amino acid position indicating the portion of the known protein matching the peptide deduced sequence, Observed = experimental m/z value, Mr  (expt) = experimental m/z value transformed to a relative molecular mass, Mr (calc) = calculated relative molecular mass of the matched peptide, Delta = difference (error) between the experimental and calculated masses, Miss = number of missed enzyme cleavage sites, Deduced sequence = sequence of the peptide in one-letter code

Full size image

Preliminary experiments with protein extracts from liquid cultures of T. harzianum grown in the presence of various fungal pathogen sources (i.e., cell walls or heat killed mycelia of Pythium ultimum, R. solani or B. cinerea) or plant tissues, produced more differential spots (present, absent, increased, decreased) than cultures supplemented with different simple sugars as the only carbon source (data not shown). As expected, the presence of the fungal host or plant extensively modified the proteome of the antagonist and activated a variety of interaction-related genes in comparison to a non-induced phase. These results were confirmed by using the WA plate-based method, when Trichoderma mycelia were left to interact either with the plant, the pathogen or both.

Differential proteins from the interaction between T. atroviride, bean leaves and B. cinerea

The proteome of T. atroviride grown alone was used as a control for comparisons with the two-way (Trichoderma–bean leaves) and the three-way interactions (Trichoderma–bean leaves–Botrytis). More than 220 differential spots were noted as ex novo, absent, increased or decreased if Trichoderma was exposed to the plant or to the plant and the pathogen together (Table 1). When the Trichoderma–plant–pathogen profile was compared to the Trichoderma–plant interaction, 57 spots appeared to be produced ex novo, 93 were absent, 25 were up-regulated and 62 down-regulated, indicating that the presence of B. cinerea induces major changes in the proteome of the antagonist interacting with the plant (Table 1). About 50 of the most strongly modified T. atroviride protein spots (examples shown in Fig. 3) were further characterized by MALDI-TOF MS followed by in silico analysis, but only a few identification cases are reported here (Table 2). These include a kinase containing a DnaJ conserved domain, a cyclophilin A-like protein and a chitin synthase.

Table 1 Changes occurring in the proteome of plant (Pln), pathogens (Bot = Botrytis cinerea; Rhi = Rhizoctonia solani) and Trichoderma atroviride strain P1 (Trc) when different interaction conditions (three-way, two-way and no interaction) were compared
Full size table
Fig. 3
figure 3

Differential intensity levels of two-dimensional (2-D) gel spots produced by the three-way, two-way and no-interaction conditions of bean plants, B. cinerea, R. solani and T. atroviride. Spot intensity changes are shown by the enlarged gel regions (picture insets) placed over the corresponding relative intensity (histogram). 2-D gels of intracellular proteins were separated in the first dimension on IPG strip (7 cm, pH 3–10) and in the second dimension on 15% polyacrylamide SDS gel. Proteins were visualized by SimplyBlue SafeStain G-250 (Invitrogen). Spot intensity was quantified by using the PD-QUEST software (Bio-Rad). The relative intensity of a spot is the sum of the signal intensities (expressed as optical density units) of all the pixels that make up the object

Full size image
Table 2 Homologies of differential proteins in the T. atroviride strain P1 proteome during interaction with bean leaves and B. cinerea. MW (kDa)/pIexp/th: molecular weight and isoelectric point of differential proteins determined experimentally (from gels) and theoretically (from in silico analysis)
Full size table

In the bean proteome, the presence of Trichoderma or Botrytis determined a similar accumulation of differential proteins (about 140), as observed by comparing the three- versus two-way interactions (Table 1). However, the comparison between the two two-way interactions (plant–Trichoderma vs. plant–Botrytis) indicated a different proteomic response in the plant to the antagonist and the pathogen. Bean accumulated 43 novel, 34 absent, 29 increased and 19 decreased spots in the presence of the antagonist instead of the pathogen. Four of the about 30 selected spots for PMF analysis are presented here (Fig. 3). All corresponded to proteins involved in defence response against pathogens, and in particular the MS data matched with specific domains related to disease resistance (i.e., leucine rich repeats, SGS and Barwin domains, thaumatin family sequences) (Table 3).

Table 3 Homologies of differential proteins in the bean leaves proteome during interaction with T. atroviride strain P1 and B. cinerea. MW (kDa)/pIexp/th: molecular weight and isoelectric point of differential proteins determined experimentally (from gels) and theoretically (from in silico analysis)
Full size table

Differential proteins were also produced by the pathogen B. cinerea in the presence of the plant alone or in combination with Trichoderma (i.e., 157 and 204 differential spots, respectively, in the interaction with plant vs. Botrytis alone and in the three-player vs. the Botrytis–plant interaction) (Table 1). These results also indicated that the presence of Trichoderma induces major changes in the Botrytis proteome while the fungus is interacting with bean plant. Of the spots further analyzed by 4700 MALDI-TOF/TOF, four gave the homologies reported in Table 4. The proteins involved in the infection process of Botrytis–plant and Botrytis–plant–Trichoderma interactions were significantly higher than in the control, as indicated by the increased spot intensities (Fig. 3).

Table 4 Homologies of differential proteins in the B. cinerea proteome during interaction with bean leaves and T. atroviride strain P1. Data obtained by using the 4700 MALDI-TOF/TOF apparatus. MW (kDa)/pIexp/th: molecular weight and isoelectric point of differential proteins determined experimentally (from gels) and theoretically (from in silico analysis)
Full size table

Differential proteins from the interaction between T. atroviride, bean plants and R. solani

When the Trichoderma–bean roots–Rhizoctonia was compared to the Trichoderma–plant interaction, more than 230 differential spots were accumulated in the antagonist proteome. Sixty-three spots appeared ex novo, 116 were absent, 27 up-regulated and 29 down-regulated (Table 1), indicating that the presence of R. solani induces major changes in the proteome of Trichoderma interacting with the plant. In addition, about 200 differential spots present ex novo, absent, increased or decreased were produced by Trichoderma during the interaction with the plant alone or in combination with Rhizoctonia, when compared to the control (Trichoderma grown alone). Figure 3 shows the selected spots that were further subjected to in silico analysis and the main results are presented in Table 5. Spot 4301 revealed a strong similarity to cyclophilin A, a protein containing a peptidyl-prolyl-cis–trans isomerase (PPIase) domain involved in protein folding and possibly in other cellular functions (Arévalo-Rodriguez et al. 2000; Gothel and Marahiel 1999). Spot 6301 was confidently identified as a fungal hydrophobin, while different members of the ABC transporter family were found to be differentially accumulated in the antagonist proteome (spots 6502 and 7501). Finally, spot 5208 was found to be a homologue of T. reesei Hex1 protein by searching the Trichoderma EST database.

Table 5 Homologies of differential proteins in the T. atroviride strain P1 proteome during interaction with bean roots and R. solani. MW (kDa)/pIexp/th: molecular weight and isoelectric point of differential proteins determined experimentally (from gels) and theoretically (from in silico analysis)
Full size table

Proteins were extracted from bean leaves and analyzed in order to characterize the systemic defence response in plants whose roots were in contact with pathogen and/or antagonist mycelia. The protein profiles from the bean roots-Trichoderma–Rhizoctonia (three-way) as compared to the bean roots-Trichoderma (two-way) interaction indicated 127 differential spots caused by the presence of the pathogen, of which 40 were novel, 29 absent, 28 of increased and 30 of decreased intensity (Table 1). By comparison with the bean roots–Rhizoctonia treatment, the three-way interaction revealed more than 300 differential spots: 28 novel, 168 absent, 23 of increased and 93 of decreased intensity. Interestingly, the presence of the antagonist resulted in a strong reduction in the number and the level of plant proteins produced as compared to the interaction of the plant with the pathogen, that produced the greatest number of novel and increased differential spots in comparison to the plant alone. In the case of plant–Rhizoctonia interaction compared to the control, 300 differential spots were obtained, of which 151 novel, 23 absent, 107 of increased and 19 of decreased intensity. Of the approximately 30 differential spots selected for PMF analysis, 5 are presented in Fig. 3. Spots 1206, 6101, 6201 and 6305 showed a particularly high increase in intensity when the plant was exposed to the pathogen and less in the three-way interaction with the fungal antagonist. These proteins were homologues to disease resistance or pathogenesis-related (PR) proteins (Table 6), and found to contain conserved motifs (NB-ARC, Bet v I family, NBS-LRR) known to be involved in pathogen- or stress-related responses. Spot 7501, which was similar to an Arabidopsis thaliana resistance protein, was absent in the control (plant grown alone) and was induced by the presence of R. solani (either alone or combined with Trichoderma).

Table 6 Homologies of differential proteins in the proteome of bean leaves whose roots were in contact with T. atroviride strain P1 and R. solani. MW (kDa)/pIexp/th: molecular weight and isoelectric point of differential proteins determined experimentally (from gels) and theoretically (from in silico analysis)
Full size table

The differential proteome of the soilborne pathogen R. solani was also studied. In comparison to the control, a total of about 200 spots, present ex novo, absent, increased or decreased, were obtained by exposing Rhizoctonia to plant roots, or to plant roots and the antagonist together (Table 1). In the interaction with the bean roots as compared to the pathogen alone, 37 novel spots, 69 absent, 39 of increased and 48 of decreased intensity were found. The differential spots of the three-player versus the Rhizoctonia–plant interaction were 81 novel, 12 absent, 51 of increased and 56 of decreased intensity. These data indicated that the presence of Trichoderma induced important variations in the pathogen proteome while the fungus is interacting with bean plant. Unfortunately, the MS results for the about 20 spots excised from gels and digested showed no significant homologies with proteins present in the NCBInr database.

Discussion

In this work we used bean plants, fungal pathogens and the antagonistic fungus T. atroviride strain P1 to analyze the changes in the proteome of the three organisms caused by multiple-player interactions. The use of cellophane membranes permitted both the exchange of compounds and thus a chemical interaction in situ, as well as the separated extraction and recover of the individual proteomes. Placing of such type of membranes between fungal spores and seed surface did not affect the improvement of seed germination or crop yield caused by treatment with Trichoderma (Benítez et al. 1998). Kullnig et al. (2000) reported that the cellophane membrane allowed the diffusion of proteins up to 90 kDa, as noted in confrontation assays between T. atroviride strain P1 and R. solani, whereas the dialysis membrane (12 kDa cut-off size) prevented macromolecule diffusion.

Comparison of results from different extraction methods demonstrated that the TCA/acetone-based precipitation was the best protocol in terms of quality and quantity of the proteins obtained, minimizing protein degradation and the presence of interfering compounds (i.e., polysaccharides, salts, polyphenols, etc.) (Görg et al. 2004).

During the three-way interaction, major changes in the proteome of T. atroviride, as compared to single and double player (antagonist–plant) conditions, were observed. In particular, the presence of fungal pathogens (B. cinerea and R. solani) strongly modified the protein pattern of the antagonist during plant interaction. Many spots absent in the three-way (i.e., 93 spots with B. cinerea and 116 with R. solani, respectively) were present in the Trichoderma–plant condition (Table 1). Possibly, the activity of compounds released by either of the two pathogens may somehow interfere with the expression of Trichoderma genes used to interact with the plant. Alternatively, the increased extracellular protein production (i.e., cell wall degrading enzymes, antibiotics, etc.) that follows the activation of the antagonistic/mycoparasitic mechanisms may be associated with a reduced number of intracellular protein species.

In the bean proteome, the interaction with each pathogen induced more protein spots than the interaction with the antagonist alone or the combination of both fungi (Table 1). As expected, Trichoderma induced different sets of plant proteins than the pathogens, and the presence of the beneficial fungus clearly changed the expression pattern of plant genes responding to pathogen attack, which may be related to increased pathogen control. In fact, some of the spots analyzed by PMF corresponded to PR-proteins and were less up-regulated than by the pathogen alone when both Trichoderma and the pathogen were interacting with the plant (Fig. 3; Table 6). Both B. cinerea and R. solani produced the greatest number of novel and increased differential spots in comparison with the plant alone, thus confirming that the presence of the antagonist extensively modifies the proteome of the plant affected by a pathogen. These results underline the importance of conducting comparative analysis of the multiple interactions involved in biocontrol and pathogenesis processes.

The changes in the proteomes of each player during the complex three-way cross-talk were investigated and the most interesting differential spots were analyzed via PMF. These were selected by comparing proteomes obtained from three-way, two-way and no interaction conditions and selecting those spots showing the most evident and reproducible changes, either qualitatively or quantitatively. In addition to simple sequence matching, we also determined if the peptide fragments obtained by MS matched the sequence of conserved domains of known proteins. Both Trichoderma EST and all-species-entries databases were used, also to perform cross species identification (CSI) (Grinyer et al. 2004a, b, 2005; Wilkins and Williams 1997).

Several interesting proteins, among those differentially expressed, were found in the T. atroviride proteome during the three-way interaction with bean leaves and the foliar pathogen B. cinerea (Table 2) or with bean roots and the soilborne pathogen R. solani (Table 5). We confidently identified a homolog of a 40 kDa heat shock protein (HSP) (spot 203; Table 2) whose intensity increased in the presence of both the plant and the pathogen, including a conserved DnaJ domain typically related to environmental stress (Lindquist 1986; Morimoto et al. 1994) and possibly associated to a Trichoderma defence response. Spot 8101, whose intensity strongly increased in the presence of the plant (with or without the pathogen) (Fig. 3), was found to be a homologue of an enzyme (chitin synthase) involved in the synthesis of fungal cell wall components after chemical, physical or osmotic stress. This could also represent a stress-related protein, since increased expression of glucan and chitin synthase genes may be required to repair cell wall damages caused by the pathogen or plant enzymes (Valdivia et al. 2003). Similarly, spot 5208 (Table 5), which corresponded to a T. reesei Hex1 protein of comparable MW and pI (Lim et al. 2001), doubled in intensity when the pathogen was present (three-way vs. two-way interaction) (Fig. 3). Hex1 is one of the most abundant proteins in fungal cell walls, mainly contained in the Woronin body, and associated to repairing of damaged hyphae (Jedd and Chua 2000).

We found proteins similar to different members of the cyclophilin family in the proteome of the antagonist during a three-player condition, with either B. cinerea or R. solani (Tables 2, 5). Cyclophilins have an enzymatic PPIase activity that has been demonstrated to have a role in protein folding (Marks 1996) and possibly in intracellular tracking, signal transduction, cell cycle regulation, differentiation and maintenance of multi-protein complex stability (Arévalo-Rodriguez et al. 2000; Gothel and Marahiel 1999). A protein with a predicted PPIase activity was previously identified by LC MS/MS in the proteome of T. harzianum grown in liquid culture (Grinyer et al. 2004b). These proteins could be used by Trichoderma in a wide range of processes, that may include the interaction with the plant since the relative spot intensity increased remarkably in the presence of bean leaves or roots (Fig. 3).

Both spots 6502 and 7501 were identified as membrane pumps of the ABC transporter family (Table 5), which could be related to the well-known resistance of Trichoderma spp. to natural toxins, antimicrobial compounds, synthetic pesticides and chemical pollutants (Harman et al. 2004a, b). Recently, a few ABC transporter genes have been cloned from T. atroviride strain P1 (Lanzuise et al. 2002), and the culture filtrates of different pathogens (B. cinerea, R. solani, P. ultimum) were found to strongly induce their expression (Woo et al. 2006). Interestingly, the promoters of these genes have many putative regulation factor-binding sites corresponding to those of an endochitinase involved in biocontrol activity (Lorito et al. 1996). These results, including the fact that the relative intensity of spot 7501 increased remarkably in the presence of either the pathogen R. solani or the plant (Fig. 3), suggesting that several ABC transporters may support the antagonistic activity of Trichoderma and its ability to colonize the roots. Spot 3214 matched a protein that may confer resistance to the fungal toxin brefeldin A (Table 5) capable of interfering with protein secretion and transport (Fujiwara et al. 1988), and may be possibly involved in cell detoxification. The presence in T. harzianum of the RNA from a different brefeldin A resistance protein (accession no. P41820) has been recently reported by Liu and Yang (2005) who used an EST-based approach. These data support the hypothesis that Trichoderma possesses a variety of mechanisms to protect itself from microbial and plant toxins.

A fungal hydrophobin matched to spot 6301 (Table 5). This kind of protein helps fungi to penetrate barriers, has in some cases a structural function (Linder et al. 2005; Wösten 2001), and mediates the attachment of pathogenic fungi to the host plant surface (Kershaw and Talbot 1998). Several hydrophobins have been found in the Trichoderma genome, but their role has not been determined yet (Linder et al. 2005). Benítez et al. (2004) have recently suggested that hydrophobins are specifically up-regulated during colonization by Trichoderma of tomato roots, and similarly we have found in the culture filtrates of T. harzianum strain T22 a hydrophobin containing a chitin binding domain able to induce the hypersensitive response (HR) and systemic resistance in plant (Ruocco et al., unpublished data).

The differentially expressed proteins found in the proteome of the pathogen B. cinerea (Table 4) included the cutinase encoded by the cutA gene particularly important for pathogen infection (van der Vlugt-Bergmans et al. 1997), different isoforms of cyclophilin 1 has been suggested to act as virulence factors in pathogen penetration or in planta growth (Viaud et al. 2003), and a superoxide dismutase (SOD) that may support pathogenicity by removing reactive oxygen species (ROS) produced by the plant (Gil-ad et al. 2000). The over-expression of SOD in B. cinerea could be related to the induction of plant resistance mechanisms by Trichoderma, since the intensity of the relative spot increased considerably when the antagonist was added to the B. cinerea–bean interaction (Fig. 3).

A variety of differentially expressed proteins were identified in the plant proteome during the interaction with the pathogens, the antagonist and both fungi (Tables 3, 6). The in silico analysis of data from plant–Botrytis and plant–Trichoderma interactions revealed many homologues to PR-proteins. Conserved domains, such as leucine rich repeats (LRR), nucleotide binding sites (NBS) and SGS domains, as well as conserved sequences of Barwin and Bet v I PR-protein families, were found. For instance, a protein of the tobacco PR-4 family with a Barwin domain (spot 6306; Table 3) and a thaumatin-like protein (spot 6302; Table 3) involved in the defence response of rice to Magnaporthe grisea (Kim et al. 2004) were differentially accumulated in presence of Trichoderma, either alone or in combination with B. cinerea (Fig. 3). Some differential spots in the bean leaves proteome obtained from plants which had root interactions with R. solani and T. atroviride showed a high level of similarity to proteins associated with plant disease resistance and pathogen recognition (Table 6). Homologies with both PR-proteins or proteins involved in defence activation mechanisms were found, including a NBS-LRR type (spot 6101) (Table 6) that may recognize pathogen products and induce defence responses such as apoptosis and HR (Moffett et al. 2002). These proteins accumulated in the plant proteome particularly in the presence of R. solani (Fig. 3), which was expected considering the activation of plant defence mechanisms after a pathogen attack. Regardless, we noticed that in many cases spots corresponding to defence-related compounds showed a decreased intensity when Trichoderma was present as compared to plant–pathogen condition (Fig. 3), which is in agreement with the differences found in terms of spot number (see above). These results indicated that specific resistance genes may regulate the plant–Trichoderma–R. solani interaction, and that the presence of the antagonist may reduce quantitatively and qualitatively the protein-based response of the plant to the pathogens. However, the addition of Trichoderma, either alone or in combination with B. cinerea, induced an increased expression in bean leaves of at least two PR-proteins (spots 6302 and 6306) (Fig. 3), suggesting the activation of a specific response to the biocontrol agent (Yedidia et al. 2000, 2003). In fact, the simple comparison between the plant–Trichoderma and the plant alone conditions indicated up to 191 differential spots, of which 39 appeared ex novo, 87 were absent, 27 up-regulated and 38 down-regulated (Table 1). In addition, we found, both by proteome matching and band hybridization, possible homologues in T. harzianum and T. atroviride of known avr proteins (avr4, avrE, NIP1) (Harman et al. 2004a), which is not surprising considering the “avirulent” nature of these plant-colonizing fungi. In particular, the homolog of NIP1 showed similarity to a T. harzianum β-1,3-exoglucanase with a conserved cysteine pattern typical of hydrophobins (Rohe et al. 1995). These findings, together with the absence of such homologies in T. reesei genome (data not shown), suggest a significant involvement of hydrophobins in the Trichoderma avirulence behaviour. Finally, we consider that there is a general similarity in the plant molecular interaction with pathogenic and beneficial saprophytic microbes like Trichoderma spp., including common gene-for-gene and avr-R gene mechanisms.

In conclusion, a proteomic approach allowed us to identify numerous differential proteins involved in multiple-player cross-talk normally occurring in nature between plant, pathogens and biocontrol agents. The majority of the studies reported so far have been focused on two-partner conditions (Baker et al. 1997; Dangl and Jones 2001; Hammond-Kosack and Parker 2003; Harman et al. 2004a; Suzuki et al. 2004), thus providing a relatively incomplete view of a pathogenicity/resistance processes as mediated by both beneficial and pathogenic microbes. Proteomic analysis can be very useful to provide both general and specific information on the “interaction proteomes” used by plants and microbes. However, the complexity of the system, which requires more than one player to act at the same time, indicated that a more integrated approach is necessary to deeply understand the biology of biocontrol agents such as Trichoderma spp.