Generation, annotation and analysis of ESTs from Trichoderma harzianum CECT 2413

EST sequence determination and analysis

ESTs were generated from different libraries with the expectation that more unique genes would be identified from moderate levels of sequencing from diverse libraries rather than deep sequencing of a single library. The combined ESTs for all libraries clustered into 889 multisequence contigs, with 2,589 singlets yielding an estimate of 3,478 unique sequences. These unique sequences are listed in the additional file 1. The ESTs that are contained in each contig are listed in the additional file 2.

The distribution of ESTs across multiple libraries was assessed as a measure of uniqueness of gene concurrence across libraries. No contig contained sequences that were from seven or eight libraries, and only nine unique sequences were present in six libraries. Most of the contigs contained sequences found in one to three libraries, with the largest category being contigs represented by clones from two libraries (393 clones). This pattern is mainly due to the fact that many of the contigs contain only a small number of ESTs. Considering only those contigs representing 20 or more ESTs, it is found that most of the genes are usually represented in multiple libraries.

Functional annotation and analysis

The gene distribution in the main ontology categories was studied and the percentages of unique sequences with assigned GO terms that fell into these categories were calculated. For this purpose, we considered 100% as the total number of unique sequences from each of the libraries that possessed an assigned GO term in each of the three organizing principles of GO (Biological Process, Molecular Function and Cellular Component) [22]. It must be taken into account that these percentages do not add up to 100% because many deduced proteins can have more than one GO assigned function. Gene distribution was very similar across the libraries that came from Trichoderma grown in liquid media (L02, L03 and L10). However, differences were found in L06, made from Trichoderma grown in solid media (solid minimal medium containing 0.1% glucose or PPG medium). For example, in L06 the percentage of GO terms in the following categories was significantly higher than in the other libraries: "metabolism" (92.9%), "biosynthesis" (56.4%), "cellular metabolism" (85.8%), "macromolecule metabolism" (64.4%), "primary metabolism" (80.6%), "nucleic acid binding" (22.8%), "structural molecule activity" (27.6%) and "intracellular" (93.0%). Additionally, in L06 a lower percentage was found in other categories like "catalytic activity" (46.5%), "hydrolase activity" (14.5%), "extracellular region" (0.6%) or "membrane" (15.4%).

Identification of putative secreted proteins

T. harzianum is the source of a number of secreted proteins produced for various industrial applications. To identify potential secreted proteins we used SignalP 3.0 [25] in order to search for predicted proteins with a signal peptide. We found that 800 of these predicted proteins (23.0%) possessed this putative signal peptide (see additional file 1).

Exploration of more abundantly expressed genes

We also found genes probably involved in the synthesis of thiamine (vitamin B1): T34C494 (46 ESTs) and T34C829 (24) were similar to the NMT1 protein and T34C811 (28) was similar to the THI4 protein, a thiazole biosynthetic enzyme. Additionally, we also detected sequences similar to an ADP/ATP carrier protein (T34C470), the ATP synthase protein 9 (T34C216) [28], a peptide involved in stress response (T34C759) and a cyclophilin (T34C418).

Comparison to the nr database and InterProScan annotation

Sequence comparison using the BLASTX algorithm against the NCBI non redundant (nr) database allowed the identification of 2,832 unique sequences (81.4%). Thus, 646 sequences did not exhibit significant similarity (E-value < 10^-5) to genes in the nr database.

Additional information on the ESTs was obtained by protein-signature scanning. InterProScan was used for sequence comparison to the InterPro database [29]. This database contains signature information from Hidden Markov models, regular expressions, fingerprints and profiles for protein families, and domains from public domain database projects including Pfam [30], PROSITE [31], ProDom [32], PRINTS [33], TIGRFAMS [34] and SMART [35]. The submission facilitated the annotation of 3,331 (95.8%) unique sequences with significant protein signatures. Within them, 2,006 (57.7%) sequences contained associated InterPro (IPR) numbers. Of these 2,006 sequences, 76 had not been annotated during the sequence comparison against the NCBI nr database. These data are included in the additional file 1.

Comparison with other Trichoderma collections of ESTs

We used the tBLASTX algorithm [36] in order to study the presence of similar sequences in the collections of ESTs from Trichoderma publicly available. At the E-value < 10^-10 level, the highest percentages of similar sequences were found in three collections from T. reesei: 44.7% [18], 40.5% [17] and 25.1% [15]. Unexpectedly, only 21.6% of similar sequences was found in the related to biocontrol collection obtained from T. harzianum [19]. Finally, a very low percentage (4.5%) of similar sequences was found with the ESTs recently obtained from T. reesei following a subtractive strategy [20].

BLAST analysis to species sequence datasets

We used the BLASTX algorithm [36] to identify ESTs with sequence similarity to the protein sequences derived from the genomic sequence datasets of 15 eukaryotic species (two animals, two plants and eleven fungi). We used an E-value < 10^-5 as indicative of sequence similarity significance. In these comparisons it is important to note that the T. harzianum sequences do not represent the complete genome.

Further analysis of the T. harzianum unique sequences grouped these sequences based upon the eukaryotic taxa in which sequence similarity was found. The classification of unique sequences conserved across all eukaryotes was defined by at least one similar sequence in each of the fungi, plants and animals. Under these criteria, 981 of the 2,793 unique sequences (35.1%) that showed sequence similarity (E-value < 10^-5) with at least one organism, were present in all eukaryotic taxa (Figure 1). These sequences are listed in the additional file 3.

Moreover, 1,138 unique sequences were found in genomes from fungi and animals (not present in plants) and a slightly higher number of genes (1,159) was shared between fungi and plants (not present in animal genomes). Furthermore, the number of unique sequences with similarity to the fungal and plant genomes but not the animal genomes was 178. Meanwhile, the number of unique sequences with similar sequences in the fungal plus in either of the two animal, but not in the plant genomes was 157 (Figure 1). The T. harzianum unique sequences found to have similar sequences only in fungal species totalled 1,474 (42.4%). These sequences are also listed in the additional file 3.

The number of orphan unique sequences (those showing no similarity with any other organism or the nr database) was 487 (14.0%). However, the number of unique sequences lacking a similarity to sequences in the nr database was 646 (18.8%). Thus, after our cross-species identification of similar sequences, we reduced the number of orphan genes by another 159 unique sequences.

EST abundance

The analysis of the relative abundance of individual ESTs that make up unique sequences (contigs) from different libraries can be used as a first indicator of transcript abundance. We identified a number of unique sequences generated from 20 or more ESTs (Table 4). Apparently, no particular gene family was predominant. This agrees with some previous works [10, 18], but it is in contrast with other analysis where particular classes of genes (e.g., encoding ribosomal proteins) dominated these frequency tables [8, 14]. However, numerous housekeeping genes were detected as expected.

A unique sequence similar to the hydrophobin II from H. jecorina was the most represented gene. A similar sequence (contig1201) was also found as one of the most expressed genes in the library LT002 from T. reesei [18]. However, we have not found another possible hydrophobin in a similar frequency table made in any other fungus including T. harzianum [19]. In a recent study [52], it was found that the T. reesei hydrophobins I and II had a role in hyphal development and sporulation, respectively. Among the most highly expressed genes we also detected sequences that could be involved in thiamine biosynthesis, similar to the NMT1 and THI4 proteins respectively. THI4 is a thiazole biosynthetic enzyme that had been previously found (contig437) as highly expressed in T. harzianum [19]. However, both genes could also be involved in other processes [53]. In S. cerevisiae THI4 appears to be a dual function protein involved in thiazole biosynthesis and tolerance to mitochondrial DNA damage [54].

A putative cyclophilin was also identified as one of the most abundant transcripts. Cyclophilins include the binding proteins of the cyclic peptide cyclosporin A, they possess peptidyl-prolyl cis-trans isomerase activity in vitro and can play roles in protein folding and transport, RNA splicing and the regulation of multi-protein complexes in cells [55]. One similar sequence was also found as one of the most expressed genes (contig429) in the EST collection from T. harzianum [19]. So far, cyclophilins have been more studied in yeasts [56, 57] than in filamentous fungi [58, 59].

GO terms

The percentage of assigned GO-terms (51.1%) was slightly higher than in one similar study in A. niger [46]. As far as we know, this is the first time that the program Blast2GO [23] has been used for this purpose. Clear differences were found in the distribution of the GO terms among the gene library L06 and the other three (L02, L03 and L10). This could be explained considering that L06 was the only library among them made using mRNA from Trichoderma growing in solid medium. However, it must be also considered that these differences could arise more from their different composition than from the solid or liquid nature of the media. The liquid media included different stress-related growth conditions like nitrogen or carbon starvation, chitin or fungal cell walls as sole carbon source whereas the solid medium covered only carbon starvation.

We identified 800 predicted proteins with a putative signal peptide. Gene Ontology annotation categorized only 80 of the predicted proteins as "extracellular", included in the GO terms "extracellular matrix" (GO:0031012) and "extracellular region" (GO:0005576) (see [24]). However, it must be taken into account that we were able to assign a GO term to 51.1% of the unique sequences. The percentage of predicted proteins with signal peptide (23.0%) was almost identical to what was found in A. niger (23.4%) [46] but lower than in T. reesei (33%), using the same algorithm [18]. These differences could be related to the different growth conditions from which the cDNA libraries were made or to the fact that they are different species.

Comparison with other Trichoderma collections of ESTs

We found that only 21.6% of the unique sequences had similar sequences in the related to biocontrol collection obtained from T. harzianum [19]. A higher number of similar sequences was found in the other comparable collections from T. reesei, specifically in the collection of ESTs involved in protein processing and secretion [18]. These results were unexpected because in principle, a highest number of similar sequences should have been found in the EST collection related to biocontrol [19] than in the T. reesei collections. The lack of information on the strain and the growth conditions in which that EST collection for T. harzianum was obtained [19] makes difficult to explain this fact. As for the differences in the presence of similar sequences between the three comparable collections from T. reesei, it is logical that the lowest percentage of similar sequences was found in a collection obtained in very different growth conditions, with glycerol as sole carbon source [15], than the ones used in our study.

BLAST analysis to species sequence datasets

The unique sequences were compared to the genomes of eleven different fungi, two animals (C. elegans and D. melanogaster) and two plants (A. thaliana and O. sativa). Within the total of 2,793 unique sequences (80.3%) that showed sequence similarity with at least one species (at E-value < 10^-5), 2,616 (93.7%) had a similar sequence in the T. reesei complete genome. It must be considered that many gaps are still present in the current publicly available version (v1.0) of this genome [5]. The new version (v2.0) containing very few gaps will be available soon [45] and will allow us to further study the unique sequences that could be found in T. harzianum but not present in T. reesei.

Behind T. reesei, the ascomycete F. graminearum possessed the largest number of similar sequences to the T. harzianum unique sequences. This is not surprising due to the close taxonomic relationship between the Trichoderma and Fusarium genera, because their teleomorphs are located in the close families Nectriaceae and Hypocreaceae, respectively [60].

In the species-by-species comparison, 981 (28.2%) unique sequences were present in all the eukaryotic taxa. A similar percentage (29%) was found in a similar study in U. maydis [13]. These 981 clones constitute a 35.1% of the unique sequences (2,793) that showed sequence similarity (at the level E-value < 10^-5) with at least one organism (animals, plants or fungi). This number is perhaps slightly low because it has been described that about 40% of the total genes in an eukaryotic genome may be shared with other eukaryotes of different kingdoms, although this does not mean that they are all essential genes [61, 62]. This slightly low percentage could be due to the fact that our EST data set constitutes only a portion of the genome.

T. harzianum sequences found to have similar sequences only in fungal species totalled 1,474 (42.4%). This percentage is much higher than what was found in U. maydis (12.3%) [13]. There may be genes among these that have retained phylogenetic signatures dating to the separation of fungi and animals, or genes with signatures representing further changes leading to the current state of T. harzianum.

Fungal strains

T. harzianum CECT 2413 (Spanish Type Culture Collection, Valencia, Spain) was used in this study. Phytophthora syringae CECT 2351, B. cinerea CECT 2100, Penicillium aurantiogriseum IMI 374515 (International Mycological Institute, Egham, UK) and Rhizoctonia solani CECT 2815 were used as a source to obtain fungal cell walls. The fungal strains were maintained on potato dextrose agar (PDA, Difco Becton Dickinson, Sparks, MD).

cDNA libraries construction

A set of more than eight different conditions, most of them designed in order to simulate biocontrol processes were used to build specific and mixed cDNA libraries. The following libraries were made for the "TrichoEST" project: L02, L03, L05, L06, L08, L10, L11 and L15 (Table 1).

For the L02 library the biomass was obtained following a two-step liquid culture procedure. First, T. harzianum CECT 2413 was grown in a minimal medium [63] (MM: 15 g/l NaH₂PO₄, 5 g/l (NH₄)₂SO₄, 600 mg/l CaCl^.2H₂O, 600 mg/l MgSO₄^.7H₂O, 5 mg/l FeSO₄, 2 mg/l CoCl₂, 1.6 mg/l MnSO₄, 1.4 mg/l ZnSO₄), containing 2% glucose as carbon source, in baffled flasks at 25°C and 160 rpm for two days. Biomass was harvested, rinsed twice with sterile distilled water and transferred to MM [63] containing 0.5% of a 1:1 mixture of fungal cell walls from P. syringae and B. cinerea. It was incubated during 9 h.

For the L03 library, T. harzianum was grown in potato dextrose broth (Difco Becton Dickinson), in baffled flasks at 25°C and 160 rpm for two days. Biomass was treated as indicated above and transferred to MM under the following conditions in separate cultures: (i) 0.1% glucose, (ii) 1.5% chitin (Sigma, St. Louis, MO), (iii) a 100-fold increase in the concentration of metals or (iv) 1% of a 1:1 mixture of fungal cell walls from P. aurantiogriseum and B. cinerea. The cultures were incubated under these conditions for 8 and 12 h.

No pre-culture was performed when L05 library was made. T. harzianum was directly grown in MM containing 5% glucose during 30 h.

For the L10 library, T. harzianum was grown in MM containing 4% glucose, in baffled flasks at 25°C and 160 rpm for 36 h. Biomass was treated as indicated above and transferred to MM under the following conditions in separate cultures: (i) (ii) MM containing 1.5% chitin during 8 and 20 h, respectively (iii) MM buffered at pH 2.5 with HCl containing 1.5% chitin (Sigma, St. Louis, MO) during 8 h, or (iv) MM containing 2% glucose, in nitrogen starvation conditions (50 mg/l ammonium sulphate), during 8 h.

In order to make the L15 library no pre-culture was performed. Trichoderma was directly grown in MM containing 0.1% polygalacturonic acid, 0.1% carboxymethylcellulose, 0.1% pectin, 0.1% xylan and 0.2% glucose, at 28°C during 36 h.

Growth conditions for the remaining libraries involved solid media. For the L06 library, the biomass was obtained after Trichoderma was cultured (1.5 × 10⁶ spores onto cellophane sheets) in solid MM containing 0.1% glucose or PPG medium (2% dextrose, 2% smashed potatoes, 2% agar) plates. For the L08 library, the biomass was obtained after Trichoderma was cultured during three days (1.5 × 10⁶ spores onto cellophane sheets) in solid MM containing 0.02% colloidal chitin as carbon source.

Finally, for the L11 library, plates of MM containing 0.2% glucose were inoculated with one 0.5-cm mycelial plugs of R. solani. After 3.5 days of incubation, R. solani was covered with a cellophane sheet and then, one plug of T. harzianum was inoculated on the top of the cellophane, in the other side of the plate, 5 cm far each other. After 3.5 days of growth at 25°C, mycelia were recovered from the Trichoderma overgrowth zone

In all cases, mycelia were harvested, rinsed twice with sterile distilled water, freeze-dried and kept at -80°C until RNA extraction. RNA was extracted from the mycelia by grinding them with a mortar and pestle under liquid nitrogen and extracted using TRIZOL^® reagent (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer's instructions. After total RNA extraction, an equal amount of RNA from the different growth conditions was mixed and mRNA was purified using poliA columns (Stratagene, La Jolla, CA) (for the libraries L02, L05, L10 and L11) or Dynabeads (Dynal, Oslo, Norway). The cDNA libraries were constructed using the UNI-ZAP^® XR Vector System (Stratagene, La Jolla, CA) following the manufacturer's instructions, excepting for the libraries L08 and L11, where the kit "Creator Smart cDNA library construction" (Clontech, Mountain View, CA) was used. In the latter case, it was performed a PCR in order to amplify the cDNA, before the cDNA cloning step.

Clone isolation

In vivo excision of pBluescript^® plasmids from Uni-ZAP^® XR vector was performed in SOLR Escherichia coli host cells (Stratagene) following the manufacturer's instructions. Cells were plated on Q-Tray plates containing LB (Luria-Bertani) agar medium containing 100 μg/ml ampicillin and then, they were grown at 37°C overnight. Colonies were picked and then were distributed into known positions into 384-well plates, previously filled with 60 μl freezing medium per well, using a QPix robot (Genetix, New Milton, UK). One liter of freezing medium included: 900 ml of LB glycerol (10 g/l NaCl, 10 g tryptone, 5 g/l yeast extract and 44 ml/l glycerol), 100 ml of a solution composed by 90 ml of a mixture of salts (6.27 g K₂HPO₄, 1.80 g KH₂PO₄, 0.5 g trisodium citrate and 0.90 g ammonium sulphate, per each 90 ml) and 10 ml MgSO₄ (0.1 g per each 10 ml), and 100 mg ampicillin. Well plates were incubated at 37°C overnight and then frozen at -80°C until used.

DNA sequencing

Template DNA was extracted using a modified alkaline lysis protocol. Sequencing reactions were performed following standard Big Dye (Applied Biosystems, Foster City, CA) protocols for a 0.25X reaction. Cycle sequencing was performed over 35 cycles (96°C for 10 s; 50°C for 5 s; 60°C for 4 min) in an Applied Biosystems GenAmp 9700 thermocycler. Multiscreen 96-well plates (Millipore, Billerica, MA) were used for dye-terminator removal. The 5' end of each clone was sequenced using an ABI 3,100 capillary sequencer (Applied Biosystems).

Sequence processing

The data were managed and stored using software specifically developed for the project. EST sequencing was performed and only the sequences containing more than 150 bases, with the program Phred [21] quality values greater than 20, were selected. Then, the EST sequences were cleaned using three programs included in the EMBOSS package [64]: Vectorstrip (for removing vector contamination), Trimseq (for removing the ambiguous ends of the sequences) and Trimest (for removing poly-A tails). Finally, the EST sequences were assembled into contigs using CAP3 [65]. Singlets and multisequence contigs resulting from this curation and assembly process were annotated on MySQL tables to build the TrichoEST database. A contig viewer written in PHP was used to browse the contigs. Several scripts in PHP and Python were used to parse the CAP3 standard output into MySQL tables. This method provided us with a clean table to keep exclusively unique sequences.

Assignment of GO terms

Annotations were based on the Gene Ontology (GO) terms and hierarchical structure [22]. The unigene set of EST contigs and singlets were annotated using the program Blast2GO [23, 66] using the E-value < 10^-5 level. Blast2GO assigns the GO terms based on the BLAST definitions. The GO term annotations were merged and loaded into the AmiGO browser and database [67].

Accession numbers

The nucleotide sequences of the generated ESTs were deposited in the EMBL database and have been assigned accession numbers from [EMBL:AJ893574] to [EMBL:AJ904494] inclusive. They are available as supplementary data in the additional files 1 and 3.