In order to identify the genes that are involved in T. atroviride (Ta), T. virens (Tv) and T. reesei (Tr) interaction with R. solani, we performed dual confrontation assays. R. solani, a plant pathogenic fungus was chosen as a model prey fungus because several biocontrol formulations based on Trichoderma target various plant diseases caused by this fungus 14 and as it makes severe agricultural damages on a wide range of crops worldwide 15 . The three Trichoderma spp. revealed essential differences in their response to confrontation with R. solani (Figure 1): Tr was at first unable to stop the growth of R. solani, and the growth of the latter one was even stimulated for 21% more than when confronted to itself. However, after 10 days at the borderline between the two cultures there was a clear barrage zone that was later consequently overgrown by Tr (Figure 1). Tv inhibited the growth of R. solani by 9%, but it was able to fully overgrow R. solani colony and finally killed it. In contrast, at 10th day Ta did not stop the growth of R. solani, but it straightforwardly overgrew it almost completely (86%, Figure 1). Thus, in terms of this experiment, Tv and Ta expressed interactions best described as predation (immediate kill and consumption) and parasitism respectively. Tr showed neutral reaction with signs of weak mycoparasitism.

The tiling microarray data of Ta, Tv and Tr were analyzed at three stages of the interaction: (1) before contact with alien hyphae (BC), (2) during the initial contact with the hyphae (C), and (3) when the interaction has been established (after contact; AC) (Figure 2). After data normalization we applied the linear modeling approach and the Bayes statistics implemented in the LIMMA package (http://bioinf.wehi.edu.au/limma/, 16 ) to the biological replicates as described in Materials and Methods. In total more than 60% of the genes present in the three genomes were found to be transcribed. Among them 651, 303 and 424 genes were found to be differentially regulated (larger/smaller than log₂ (ratio) = 1.5) in Ta, Tv and Tr, respectively (Table 1; Additional file 1: Tables S1-S3) and thus involved in the interaction. All of up- and down-regulated genes were annotated and categorized as described in the Materials and Methods section. In addition, Functional Catalogue (FunCat, http://mips.helmholtz-muenchen.de/proj/funcatDB/) protein families were assigned to those genes for which a function could be predicted (Table 2). 66, 70 and 62% of the significantly expressed genes of Ta, Tv and Tr, respectively, encoded genes with significant similarity to other fungal proteins (blastp E < e^-100; Table 1). The remaining genes were annotated as unknown (when orthologues were found in other fungal genera but no function was assigned) or orphan (unique; present only in Trichoderma spp.) proteins (see Materials and Methods for definition). The latter comprised only 0.5-0.7% of all transcripts for the three Trichoderma species (Table 1). Statistical analysis revealed that the transcriptomic profiles of three investigated species were significantly different (ANOVA, p < 0.001). The majority of involved genes in Tr and Tv (63 and 84% respectively) were down-regulated, whereas the up- and down-regulated genes in Ta occurred in approximately equal proportions (52% up- and 48% down-regulated). However, in all species the pattern of up- and down-regulation depended on the confrontation stages (Figure 3).

*the percentages of up- and down regulated genes are calculated based on the total number of transcripts found within each group of genes; bold fond highlights larger values.

*ratio values cannot be given when denominator is zero; n.a. means not applicable.

Interestingly, the majority of significantly up-regulated genes were represented by genes for which orthologues in other Trichoderma species are present (Figure 4). In addition, 88 of these genes were shared only between Tv and Ta. Thus 93.2, 74.4 and 84.8% of all genes that were significantly expressed in either Tr, Ta or Tv, respectively, were orthologues or at least conserved in Ta and Tv. In contrast, among those most of genes involved in mycoparasitic response were only found in one species. Only nine orthologuous genes were up-regulated in all three species (Figure 4) when confronted with R. solani, and 29 genes were significantly expressed only in the two mycoparasitic species Ta and Tv; they comprised 16 unknown, three putative MFS transporters, two orphan and two unknown transcriptional activators genes.

The most significant differences were evident at the pre-contact stage (ANOVA, p < 0.001): in Tr, the majority of differentially regulated genes were up-regulated, whereas in Tv and in Ta genes in this stage were mostly down-regulated and showed only a small response (Figure 3). At contact with hyphal tips of R. solani, Tr and Ta displayed only a low transcriptomic response, whereas Tv showed significant down-regulation of most of its genes (ANOVA, p < 0.001). During overgrowth of the prey, expression of most of genes increased in Ta and Tv, but were down-regulated in Tr. However, in general transcriptional response in this stage was significantly different between Ta and the other two species (ANOVA, p < 0.001).

To confirm the microarray results, quantitative Real-Time-PCR (qPCR) was performed on a subset of the genes from Tv and Ta. They were chosen from gene families that showed abundant up-regulation during confrontation with R. solani, such as GH16 glucanases, proteases, PKS and SSCPs (see below). Expression of these genes correlated well with the data of the microarray (Additional file 1: Table S4). We therefore conclude that the results of tiling microarray analysis indeed reflect differences in the function of genes involved in mycoparasitism.

The enrichment for the major functional categories (FunCat) was assessed for each category (Table 2) separately. To reveal the common (shared by all three species) and species-specific mechanisms acting at every stage of interaction between Trichoderma and R. solani we performed the statistical analysis of the respective FunCats. In general, all three species showed significantly different FunCat profiles indicating the presence of alternative interaction strategies (MANOVA; p < 0.05). Shifts in gene expression between BC, C and AC stages of the interaction were also strongly species-specific (Figure 5): only a minor portion of genes became down-regulated in C compared to BC and at AC compared to C (ANOVA < p < 0.001). However the massive differences were observed in Ta when a large group of genes became up-regulated at AC compared to BC and C (Figure 5A).

Statistical analysis of the global comparison of FunCat categories for the three stages for Tr, Tv and Ta, revealed that most categories broadly scattered both in the up- and down-regulated areas (Table 2). However, some specific features were observed: in Tr, genes involved in metabolism and transport were all up-regulated and the latter also in Tv; expression of proteins with binding function was increased in Ta during the contact phase and proteins involved in self-defense decreased in Tr during overgrowth. It is also apparent that during overgrowth regulation of all genes was much more comparable (as indicated by less scattering of values), and only few categories also included up-regulated genes.

A comparison of the protein families [Pfam, http://pfam.sanger.ac.uk/] assigned for up-regulated genes in the above analysis revealed very little common responses among the three Trichoderma spp., and also between the two opportunistic and strong mycoparasitic species. In fact only a few gene families shared a common trend in all three Trichoderma spp. (proteases, heat shock proteins, cytochrome C peroxidase, proline oxidase, ER-bound glutathione-S-transferases, ABC efflux transporters, the pleiotropic drug resistance (PDR) transporters, multidrug resistance MDR-type transporters and nitrilases), indicating that stress response connected with detoxification of potentially hazardous metabolites is a general reaction of Trichoderma during antagonism (see the Discussion below).

In addition to the genes named above, all three Trichoderma spp. had distinct specific responses. Thus, in addition to expression of the highest number of several protease families (dominated by subtilisin-like and aspartyl proteases), Ta increased the transcription of oligopeptide transporters, C-type lectins, small secreted cysteine-rich proteins (SSCPs), PTH11-receptors and ß-glucanases of the GH16 family (Table 2; Additional file 1: Tables S1-S3). In contrast, Tv mainly increased the expression of genes for gliotoxin biosynthesis including the genes for its precursors, and genes encoding heat shock proteins. The minor additional responses in Tv included increased up-regulation of chaperone proteins, and several dehydrogenases and monooxygenases and significant down-regulation of AAA + −type ATPases and transcriptional regulators. However, it was conspicuous that both strongly opportunistic and mycoparasitic species expressed all these genes already at a stage before hyphal contact with R. solani, and were not expressed anymore after the contact with the prey (Figure 5B).

Ta also largely relied on antibiosis: one polyketide synthase (PKS) of the reducing (lovastatin/citrinin) clade I (Triat2:134224; 17 ) was expressed at all stages of interaction, and peaked at C. A second PKS gene (Triat2:85006; 17 ) from the reducing clade I was expressed only at contact but not at BC and AC. Interestingly, Ta also expressed genes encoding two KP4-like killer-toxins that have so far not been described for filamentous fungi. In addition, the up-regulation of genes encoding macrophomate synthases, isoflavone reductases and pyoverdin dioxygenases was notable. These genes may be involved in the formation of yet unknown secondary metabolites of Trichoderma. Finally, a gene encoding a lipoxygenase (Triat2:33350), which is absent in Tr and Tv, was up-regulated at C stage. It may be involved in the biosynthesis of γ-pentylpyrone (see details in the Discussion).

In contrast, the transcriptome of Tr before contact revealed unique response to the presence of R. solani. It showed a massive up-regulation of cellulolytic and hemicellulolytic CAZymes, and ribosomal proteins (Table 2). Furthermore, several transporter proteins (MSF permeases, inorganic phosphate, oligopeptide and amino acid transporters) as well one PKS (Trire2:60118) were up-regulated at this stage. Interestingly, Tr also up-regulated subtilisin proteases but none of the aspartyl proteases that were active in Ta. At C stage Tr significantly reduced the total gene expression, and only genes encoding a multidrug transporter, a stress-responsive protein (RDS1) and another PKS were induced and may indicate defense reaction rather than an attack. The AC stage was characterized by an increased activity of the genes encoding ribosomal proteins, while genes for cellulolytic and hemicellulolytic CAZymes were not expressed anymore.

Some gene families were consistently down-regulated in all three species during confrontation with R. solani, particularly already at BC stage. These comprised short-chain dehydrogenases/reductases, transcriptional regulators of the fungal-specific Zn(2)Cys(6) family, non-ribosomal peptide synthases, and AAA + −domain proteases.

The only species-specific response was detected for Ta, which showed the opposite expression pattern of genes involved in polysaccharide and lipid hydrolysis and solute uptake to Tr, as it strongly down-regulated all these genes already prior the contact (BC).

We also investigated whether the genes expressed during interaction with R. solani would be clustered in the genome. Our approach was based on the consideration that under random distribution of transcribed genes, we should (on the average) find them once in batches of 18, 42 and 22 subsequent genes in Ta, Tv and Tr, respectively. These values arise by division of the total number of genes in the Ta, Tv and Tr (11865, 12518, and 9143, respectively) by the number of genes significantly expressed by them in this study (651, 303 and 424, respectively). To detect significant deviations from these expectations, we manually screened the three genomes for genes that occurred within the vicinity of each other at an at least 3-fold lower value than the above calculated average distributions, as was also done by Seiboth, Aghcheh et al. 18 . We detected that the distribution of about one third of the genes (36.9, 34.7 and 37.2% in Ta, Tv and Tr, respectively) was not random (Pearson coefficient >0.25) (Table 3). About a third of them were located at the ends of the respective scaffolds implying that they are situated either at the ends of chromosomes or nearby repetitive regions.

* indicates the area on the scaffold in which the cluster was identified, given by the numbers of the first and the last genes on the scaffold. ** the ratio of significantly up- or down-regulated genes to the total number of genes in the cluster.

For mycoparasitism confrontation assays T. virens Gv29-8, T. atroviride IMI 206040 and T. reesei QM 6a were grown on potato dextrose agar plates (BD Dicfo, Franklin Lakes, NJ, USA), covered with cellophane, at 25°C and 12 hours cyclic illumination and harvested when the mycelia were ca. 5 mm apart, at contact of the mycelia and after Trichoderma had overgrown the host fungus Rhizoctonia solani by ca. 5 mm. As control, the respective strain of Trichoderma was confronted with itself and harvested at contact. Peripheral hyphal zones from each confrontation stage were sampled and shock frozen in liquid nitrogen. Mycelia were ground to a fine powder under liquid nitrogen and total RNA was isolated using the guanidinium thiocyanate method 53 . For cDNA synthesis, RNA was treated with DNase I (Fermentas, Burlington, Canada) and purified with the RNeasy MiniElute Cleanup Kit (Qiagen, Valencia, CA, USA). 5 μg RNA/reaction were reverse transcribed using the SuperScript™ III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and a mixture of provided random hexamer primer and oligo(dT) primer.

We designed T. virens, T. atroviride and T. reesei tiling array with 60mer oligonucleotides (oligo length) each 93 bp apart (oligo distance), using the unmasked fasta file of the genomes of T. atroviride [http://genome.jgi.doe.gov/Triat2/Triat2.info.html], T. virens [http://genome.jgi-psf.org/Trive1/Trive1.home.html] and T. reesei [http://genome.jgi-psf.org/Trire2/Trire2.home.html]. The design was done with Teolenn 54 using a maximum prefix length match set to 30 for the uniqueness calculation made by genome tools. Complexity is evaluated using the masked genome by counting the number of masked bases for each probe. T_m values are calculated using the nearest neighbor thermodynamic model. No filters were applied after probe parameter calculations. In order to obtain the probe quality score, the calculated parameters were weighted as follows: 0.4 for T_m, 0.3 for uniqueness, 0.2 for GC content, and 0.1 for complexity. To get final oligonucleotide scores, a weighting of 0.75 was assigned to quality scores, and a weighting of 0.25 was assigned to position scores. Teolenn software designed 415373, 387105 and 359089 final probes respectively for T. atroviride, T. virens, and T. reesei. Oligonucleotides were loaded on Agilent eArray software and 2 × 400 k microarrays were obtained from Agilent. The microarray data and related protocols are available at the GEO web site [http://www.ncbi.nlm.nih.gov/geo/] under accession number GSE23438. Briefly, the RNAs of two independent biological replicates for each condition (which in turn consisted of pooled RNA preparations from at least three separate cultivations) were reverse-transcribed and labeled with Cy3 or Cy5 dye using the indirect labeling procedure. Dye bias was eliminated using dye switch labeling protocol. We then hybridized 1.5 μg of labeled cDNA with the 2x400k DNA chip (Agilent). The array was read using an Agilent G2505C DNA microarray scanner and the TIFF images extracted with the Agilent Feature Extraction software (version 10.5.1.1) using the 20bit coding ability. Data pre-treatment was applied on each result file to discard flagged spots by Feature Extraction software. The data were normalized without background subtraction by the global Lowess method performed with the Goulphar software 55 . For each experimental condition, the two file results were merged together. For each probe the hybridization ratio was linked to genome annotation coming from the DOE JGI website for a respective genome. The final ratio for each transcript was obtained by averaging the detected hybridization values from all probes located inside the coding sequence on the matching strand. Transcripts with no or only one probe marked as detected were discarded from further analysis. Finally we kept only transcript with a final hybridization ratio greater than log₂ (ratio) = 1.5 or lower than log₂ (ratio) = −1.5. The list of statistically significant differentially expressed genes was obtained by applying the linear modeling approach implemented in lmFit package for Python and the empirical Bayes statistics implemented by eBayes from the limma R package (http://bioinf.wehi.edu.au/limma/; 16 ).

The genomes of T. reesei QM6a, T. atroviride IMI 206040 and T. virens Gv29-8 have been annotated 4 and were used as a source for identification of the genes found in this study. A gene was considered to encode a homologue of an already identified protein if it had the Expect value (E) in blastp of < e^-100 with the next Sordariomycetes member. All unknown and orphan (unique) proteins were subjected to blastp search (May, 2012) to check whether any of them has been identified for another fungus since the Trichoderma annotations. Orphan proteins were defined as such that did not occur in any other Pezizomycotina with an E-value > e^-20. All other proteins, for which no function could be predicted, were termed as unknown.

Statistical analyses of microarray data were performed using Statistica 6.1 (StatSoft, Inc., Tulsa, OK, USA). For this purpose the cumulative matrix has been assembled (Additional file 1: Table S1-S3). In this matrix every gene was considered as a case and the following grouping variable (predictors) were integrated: protein ID, gene annotation, species, protein family (Pfam), gene regulation (up-regulated/down-regulated), fold gene regulation prior, at and after the contact, and functional category (FunCat). Data were explored by means of descriptive statistics (mean values and standard deviations were calculated for each predictor). Hypotheses were tested using one-way main effects or factorial ANOVA analyses with post-hoc comparisons with Tukey honest significant difference tests (Tukey HSD).

The antagonistic potential was evaluated for each species based on dual confrontation plate assay carried out as described for the RNA extraction, except the plate were not covered with cellophane. The evaluation of the plates was carried out after 10 days of incubation (see Figure 1) and the growth of Trichoderma against itself was set up as a zero inhibition rate. The growth inhibition of R. solani was calculated implying that the distance between the plugs of both fungi is 100%. The inhibition of growth was calculated as a percentage of Trichoderma growth subtracted for a correction of the growth against itself. Overgrowth of the prey was calculated similarly using the distance between the R. solani plug and the confrontation zone with Trichoderma as 100%, and the growth of Trichoderma over the R. solani mycelium was then calculated as a percentage of overgrowth.

T. atroviride and T. virens were cultivated and harvested as described for the microarrays. DNase treated (DNase I, RNase free; Fermentas) RNA (5 μg) was reverse transcribed with the RevertAid™ First Strand cDNA Kit (Fermentas) according to the manufacturer’s protocol with a combination of the provided oligo(dT) and random hexamer primers.

All real-time PCR experiments were performed on a Bio-Rad (Hercules, CA) iCycler IQ. For the reaction the IQ SYBR Green Supermix (Bio-Rad, Hercules, CA) was prepared for 25 μl assays with standard MgCl₂ concentration (3 mM) and a final primer concentration of 100 nM each. Primers used are given in Additional file 1: Table S4. The amplification protocol consisted of an initial denaturation step (3 min at 95°C) followed by 40 cycles of denaturation (15 sec at 95°C), annealing (20 sec at 52 and 52.5°C for T. virens and T. atroviride respectively) and extension (72°C for 10 sec for both species). Tef1 gene which maintained constant over all conditions (± 20 relative %) was used as the normalizer. Determination of the PCR efficiency was performed using triplicate reactions from a dilution series of cDNA (1, 0.1, 10^-2 and 10^-3). Amplification efficiency was then calculated from the given slopes in the IQ5 Optical system Softwarev2.0. The qPCR were performed with the cDNA of 5 pooled biological replicates for each species and condition. Expression ratios were calculated using Livak test model 56 and are given in the Additional file 1: Table S4. Zero values for qPCR indicate the expression levels between |log₂ (ratio)| > 1.5 as these were not treated as significantly regulated in the microarray analysis.