Genome-wide search and functional identification of transcription factors in the mycorrhizal fungus Tuber melanosporum


Author for correspondence:
Simone Ottonello
Tel: +39 0521 905646


  • Developmental transitions associated with the life cycle of plant-symbiotic fungi, such as the ascomycete Tuber melanosporum, are likely to require an extensive reprogramming of gene expression brought about by transcription factors (TFs). To date, little is known about the transcriptome alterations that accompany developmental shifts associated with symbiosis or fruiting body formation.
  • Taking advantage of the black truffle genome sequence, we used a bioinformatic approach, coupled with functional analysis in yeast and transcriptome profiling, to identify and catalogue T. melanosporum TFs, the so-called ‘regulome’.
  • The T. melanosporum regulome contains 102 homologs of previously characterized TFs, 57 homologs of hypothetical TFs, and 42 putative TFs apparently unique to Tuber. The yeast screen allowed the functional discovery of four TFs and the validation of about one-fifth of the in silico predicted TFs. Truffle proteins apparently unrelated to transcription were also identified as potential transcriptional regulators, together with a number of plant TFs.
  • Twenty-nine TFs, some of which associated with particular developmental stages, were found to be up-regulated in ECMs or fruiting bodies. About one-quarter of these up-regulated TFs are expressed at surprisingly high levels, thus pointing to a striking functional specialization of the different stages of the Tuber life cycle.


The life cycle of the symbiotic ascomycete, and prized truffle producer, Tuber melanosporum comprises three main stages: free-living mycelium (FLM), ectomycorrhiza (ECM), and fruiting body (FB; ‘truffle’). Little is known about the molecular mechanisms underlying morphogenetic transitions between the above life cycle stages, and most of the available information relies on transcript profiling. As revealed by pregenomic cDNA array analyses conducted on basidiomycete ectomycorrhizal associations, mycorrhizal development is indeed accompanied by changes in the abundance of particular groups of transcripts. Genes involved in fungal cell division and proliferation, differentiation and signaling, synthesis of cell wall and extracellular matrix, plant defense or stress response, and primary metabolism appeared to be specifically up-regulated in mycorrhizas (Martin et al., 2007). Similar large-scale gene expression analyses have been conducted in the ectomycorrhizal ascomycetes Terfezia boudieri (Zaretsky et al., 2006) and in the whitish truffle Tuber borchii (Lacourt et al., 2002; Gabella et al., 2005; Montanini et al., 2006a). In the case of T. borchii, nitrogen starvation, an abiotic stress that accompanies (and perhaps contributes to) symbiosis establishment, was used to identify changes in gene expression in stages immediately before mycorrhiza formation. Genes directly involved in nitrogen metabolism as well as genes involved in cell wall and hyphal morphology modification emerged as candidates from the latter analysis, while unknown genes were the main outcome of the study in T. boudieri (Zaretsky et al., 2006). By comparison, most genes identified as differentially expressed during FB formation in T. borchii were related to cell wall biosynthesis, antioxidant defense, and nitrogen/carbon metabolism, especially gluconeogenesis and the glyoxylate cycle (Lacourt et al., 2002; Gabella et al., 2005). A more comprehensive picture of gene expression pattern variations associated with specific life cycle stages of ectomycorrhizal fungi has been obtained recently, following completion of the genome sequences of the ectomycorrhizal basidiomycete Laccaria bicolor and T. melanosporum, the first sequenced plant-symbiotic ascomycete (Martin et al., 2008, 2010). Transcriptome analyses carried out in both studies confirmed, and further detailed, gene expression pattern variations accompanying symbiosis establishment, but also highlighted profound differences in the so-called ‘symbiosis toolbox’ utilized by these fungi to implement the mycorrhizal program. Furthermore, despite genome-wide expression studies of increasing resolution, transcription factors potentially responsible for the observed changes in gene expression have not been systematically examined in a plant-symbiotic fungus so far.

Ascomycetes are the eukaryotic organisms with the highest genome sequence coverage to date. This, together with the recently released genome sequence of T. melanosporum (Martin et al., 2010), allowed access, for the first time, to the entire repertoire of transcription factors (TFs; the so-called ‘regulome’) in this organism. TFs are regulatory proteins that control gene expression and are involved in a variety of cellular functions. Their basic domain architecture includes at least one DNA-binding domain (DBD) that binds to sequence-specific DNA elements in the promoter region of target genes, and an activation domain (AD) that interacts with (and recruits) the transcription machinery. Additional domains can mediate homodimerization, interaction with other TFs, or the binding of co-activators/co-repressors or small molecule ligands. TFs are classified into superclasses and classes according to their DBDs, which usually display fairly conserved folds and structures (Stegmaier et al., 2004). ADs, instead, are generally more idiosyncratic, lack common structural elements and are usually less easily recognizable. In silico TF predictions thus need to be corroborated and extended by functional analyses.

One of such analyses takes advantage of the distinct and independent functions played by TF domains and relies on a modified version of the yeast two-hybrid system, named ‘transcriptional activator trap’ (TAT) assay, in which selected open reading frames or an entire cDNA library from the organism (and tissue/stage) of interest are cloned downstream to a vector-provided DBD (yGAL4-DBD) (Ye et al., 2004; Titz et al., 2006). When the tested sequence codes for a transcriptional activator, the expression of three reporter genes is activated. Despite the high-throughput capabilities and apparent simplicity of the TAT assay, both false-negative and false-positive results can be expected. The latter, which are perhaps more relevant for the present study, may be caused by forced nuclear internalization of otherwise cytoplasmic proteins as a result of the presence of a vector-provided nuclear localization signal (NLS) associated with the yGAL4-DBD. It should be noted, however, that many instances of apparently cytoplasmic proteins capable of entering the nucleus and activating transcription have been reported recently. These include, for example, actin-related proteins and proteins involved in endocytosis (Pilecka et al., 2007; Zheng et al., 2009), but also a number of metabolic enzymes and other proteins (Hall et al., 2004; Shi & Shi, 2004; Hu et al., 2009), which despite the lack of a recognizable DBD have been found to be capable of DNA binding and transcriptional activation.

The aim of this work was to use state-of-the-art prediction tools in combination with the TAT assay (outlined in Fig. 1) to obtain the largest possible (and less biased) repertoire of DBD-containing proteins and bona fide transcriptional regulators encoded by the T. melanosporum genome, especially TFs potentially involved in ECM and FB formation. Gene expression programs underlying mycorrhiza formation and symbiosis establishment are still largely unknown. Fruiting body development has been studied extensively in Neurospora and Aspergillus, but at variance with these model ascomycetes, T. melanosporum is a macrofungus producing large-sized underground ascomata, whose formation strictly depends on prior interaction with a host plant. There is thus much potential for the discovery of new TFs, or homologs of previously known TFs reallocated to a new function in T. melanosporum. Therefore, in our analysis, special attention was paid to orphan genes, to duplicated genes that may reflect a novel (nonredundant) function in truffles, and to TF genes preferentially expressed in mycorrhiza or fruiting bodies.

Figure 1.

 Outline of the combined, bioinformatic and functional transcription factor (TF) identification strategy utilized in this study. In silico analysis, focused on the conserved DNA-binding domain (DBD) region, allowed the assignment of putative Tuber melanosporum TFs to different categories, based on sequence similarity with characterized TFs from other fungi (see text and Supporting Information, Table S1 for further details). Heterologous functional analysis in yeast (‘transcriptional activator trap’, TAT; see text and Fig. S1 for details) was carried out in parallel and led to the validation of 44 in silico identified TFs and to the identification of 63 DBD-lacking unconventional activators (UAs and putative UAs (PUAs); see Tables 2, S3). Forty-one TAT-positive proteins of plant origin were retrieved from the screening of the ectomycorrhiza (ECM) library (see Table S2). These analyses, coupled with transcriptome data, led to a comprehensive list of positively acting Tuber regulators (‘regulome’). pTFs, putative transcription factors; CH-TFs, conserved hypothetical transcription factors; HP-DBDs, DBD-containing hypothetical proteins; FB, fruiting body; FLM, free-living mycelium; AD, activation domain (see text for further details on these definitions).

Materials and Methods

Bioinformatic and transcriptome analyses

Sequences corresponding to functionally characterized or predicted TFs from Saccharomyces cerevisiae, Neurospora crassa and Aspergillus nidulans (Borkovich et al., 2004; Beskow & Wright, 2006; Colot et al., 2006; Wilson et al., 2008) were employed as queries for a BLASTP search against the predicted T. melanosporum proteome, which was conducted at the Génoscope and INRA Tuber sites (, using a threshold e-value of 10−5. This initial set of putative TFs was further scrutinized by a BLASTP search run against the GenBank database in order to remove TF-unrelated as well as mispredicted sequences. All putative proteins containing a recognizable DBD (pfam) were also included in the analysis. Gene products sharing a significant similarity with TFs, but lacking a typical DBD, were individually verified (and corrected if necessary) in order to uncover a potentially mispredicted DBD. If a DBD could not be detected, no putative transcription factor function was assigned to the protein.

This set of Tuber TFs was then subjected to BLASTP analysis against the Swiss-Prot database, masking PROSITE profiles or Pfam hidden Markov models matching regions in order to distinguish proteins with homology restricted to the DBD from those sharing homology with bona fide TFs along the entire sequence. Based on these results, putative Tuber TFs were classified as ‘DBD-containing hypothetical proteins’ (HP-DBD) when sequence homology was restricted to the DBD, or as ‘conserved hypothetical TFs’ (CH-TF) in the case of proteins for which sequence similarity was also significant outside of the DBD. CH-TFs similar to functionally identified TFs from other fungi (i.e. Tuber proteins that are best reciprocal hits of characterized proteins) were separately grouped and designated as ‘putative TFs’ (pTFs). Phylogenetic analysis (not shown) was used to define paralogous relationships between Tuber genes.

Whole-genome exon oligoarray and RNA-Seq analyses were conducted as described (Martin et al., 2010).

Yeast-expressible cDNA library construction

Total RNA (500 μg per sample) was prepared from T. melanosporum fruiting bodies and from free-living mycelia (strain Mel28) grown in liquid malt culture for 3 months (kindly provided by Gérard Chevalier, INRA-Clermont-Ferrand) as described elsewhere (Balestrini et al., 2000). The RNeasy Plant Mini Kit (Qiagen) was used for RNA cleaning and DNase I treatment, followed by mRNA purification on Oligotex columns (Qiagen). FB and FLM cDNA libraries were constructed with the CloneMiner cDNA Library Construction Kit (Invitrogen), starting from 2 μg of purified mRNA for each library. For the ECM library, total RNA was extracted from T. melanosporum-inoculated hazelnut (Corylus avellana) roots (kindly provided by Raffaella Balestrini, University of Torino), with the RNeasy Plant Mini Kit (Qiagen), using RLC buffer containing 20 mg ml−1 PEG 8000 and applying DNaseI on column. mRNA was then retrotranscribed, amplified and cloned, followed by Gateway®-mediated transfer of the initial libraries to the yeast-expressible pDEST™32 vector (supplied with the ProQuest™ Two-Hybrid System kit, Invitrogen).

Functional analysis in yeast

For the ‘transcriptional activator trap’ assay, yeast strain MaV103 (kindly provided by Marian Walhout, University of Massachusetts) harboring Gal4-dependent LacZ, HIS3 and URA3 reporter genes was transformed with the three pDEST™32 cDNA libraries and plated on selective medium (SD-Leu-His) containing 25 mM 3-amino-1,2,4-triazole (3AT). At least 500 000 colonies were screened for each library. 3AT-resistant colonies were individually transferred to 384-well SD-Leu plates, handled with a 384-multipinner device (V&P) and analyzed for reporter gene expression. Briefly, for the HIS3 reporter assay, freshly grown colonies were replica-plated on SD-Leu (positive control), SD-Leu-His + 25 mM 3AT and SD-Leu-His + 100 mM 3AT (test); growth was monitored by visual inspection after 2 and 4 d at 30°C for 25 and 100 mM 3AT, respectively, and colony sizes were compared with those of the corresponding colonies grown for 2 d on positive control plates. Clones yielding large or small colonies on 100 mM 3AT were classified as strong (+++) or medium-strength (++) activators; clones that failed to grow on 100 mM 3AT but which grew on 25 mM 3AT were classified as weak activators (+), whereas clones that did not grow on either medium were considered as false positives (−). An example of this assay is shown in Supporting Information, Fig. S1. For the URA3 reporter, freshly grown colonies were replica-plated on SD-Leu (positive control) and SD-Leu -Ura (test) for 5 d. After culture, activator strength was classified as none (−), weak (+), medium (++) or strong (+++), based on visual inspection of colony size using the positive control as a reference. For the LacZ (β-Gal) reporter assay, candidate clones were pinned onto 150 mm-YPD (1% yeast extract, 2% peptone, 2% dextrose, 2% agar) plates overlaid by a nitrocellulose or nylon membrane. The β-Gal assay was perform as described by Walhout & Vidal (2001). After culture incubation at 37°C for 24 h, clones were classified based on color development as no activation (−, white), weak (+, green), medium (++, light-blue), and strong (+++, dark-blue) (Fig. S1). Cumulative scores from the three assays were used to evaluate activation strength: clones with score values ≥ ++ in at least two assays and ≥ + in the remaining assay were considered as putative strong activators. Despite this fairly stringent criterion, c. 70% of the clones that passed initial selection were retained and subjected to sequence analysis.

Targeted functional analysis of individual putative TFs was conducted by cloning each gene in a modified version of the pDEST™32 plasmid, in which the Gateway® recombination cassette is replaced by a CpoI restriction site. CpoI restriction sites flanking the coding sequence of the gene of interest and in frame with the yGAL4-DBD were introduced by PCR. Amplifications were carried out using as template 200 ng of the library corresponding to the life cycle stage in which the TF of interest was most expressed. 3′-rapid amplification of cDNA ends (3′-RACE) was used to identify the C-terminal portion of the protein encoded by GSTUMT00012451001.

Transcriptional activator trap-positive clones were then sequenced, and the resulting sequences were first trimmed to remove low-quality and vector-derived regions. Polished sequences were subsequently assembled into contigs using the EGAssembler (Masoudi-Nejad et al., 2006). The resulting unisequences were used as queries for a BLASTX search against the Tuber proteome and, when available, a gene model was assigned to each sequence (identity > 95%). Sequences overlapping the same gene model were checked manually and included in contig sequences. Nucleotide and amino acid sequences were then subjected to BLASTX and BLASTP searches at the National Center for Biotechnology Information. Based on the outcome of these searches, sequences were classified as belonging to either plant or fungi, and to specific structural/functional categories (see the Results section). Sequences without a significant protein match (e-value < 10−4) were analyzed with BLASTN. No sequence similar to noncoding RNAs (or vector DNA) was retrieved from the TAT assay. Putative NLSs were searched with the PSORT II prediction tool (Horton & Nakai, 1997).


Predicted T. melanosporum regulome

Putative transcription factors bearing a recognizable DBD were identified by searching the T. melanosporum genome sequence with BLASTP, using functionally characterized fungal TFs and predicted TFs from S. cerevisiae, N. crassa and A. nidulans as queries (Borkovich et al., 2004; Beskow & Wright, 2006; Colot et al., 2006; Wilson et al., 2008). The resulting candidate sequences were classified according to the TF classification system of Stegmaier et al. (2004), taking into account the major classes of fungal TFs: basic domain leucine zipper (bZIP), basic helix-loop-helix (bHLH), and the zinc-coordinating GATA Zn finger and C2H2 Zn finger, and the fungal Zn2-Cys6 binuclear cluster domain (Zn cluster). As shown in Fig. 2, Zn cluster, followed by C2H2 Zn fingers, are the prevalent DBDs in T. melanosporum, yet the total number of predicted TFs as well as their distribution among different classes do not differ significantly from those of other ascomycetes. The sole exceptions are the apparently higher number of TFs in T. melanosporum compared with Schizosaccharomyces pombe and the nearly twofold lower number of TFs compared with A. nidulans, which displays a strong prevalence of Zn cluster transcription factors.

Figure 2.

 Distribution of Tuber melanosporum (T. mel) transcription factors (TFs) among different structural categories (Stegmaier et al., 2004) and comparison with the TF repertoire of unicellular and filamentous ascomycetes: Saccharomyces cerevisiae (S. cer) and Schizosaccharomyces pombe (S. pombe) (Beskow & Wright, 2006), Aspergillus nidulans (A. nid) (Wilson et al., 2008) and Neurospora crassa (N. cra) (Borkovich et al., 2004). The following types of DNA-binding domains (DBDs) were considered: Zn cluster; C2H2 Zn finger; GATA Zn finger; basic helix-loop-helix (bHLH); basic leucine zipper (bZIP); others, less represented types of DNA binding motifs (‘Others’).

Functional assignment simply based on sequence similarity may lead to ambiguous results. This is especially true in the case of TFs, for which sequence similarity can be limited to (and biased by) highly conserved DBDs. We addressed this potential problem by performing a separate set of BLAST searches leaving out the DBDs. Based on these additional BLASTP results, putative Tuber TFs were subdivided into three classes: proteins apparently unique to Tuber with homology restricted to the DBD were classified as HP-DBDs (42 proteins); proteins for which sequence similarity is also significant outside of the DBD were classified as CH-TFs (57 proteins), or as pTFs if a functionally characterized ortholog could be identified (102 proteins). Predicted Tuber TFs belonging to each of the above classes, a total of 201 proteins, along with information on the particular type of DBD, best-hit sequence are listed in Table S1. All of the TF genes thus predicted gave above-background hybridization signals in oligoarray and/or RNAseq transcriptome analyses in at least one of the three life cycle stages that were investigated (FLM, ECM, FB); expression data for each candidate sequence are also reported in Table S1.

In addition to conserved basal transcriptional regulators (e.g. the RNA polymerase III transcription factors TFIIIA and TFIIIB; see Table S1), a specific transcriptional control function could be assigned to 91 of the predicted T. melanosporum proteins, which are similar to known TFs from other fungi, mainly yeast, N. crassa and A. nidulans. As shown in Table 1, putative Tuber TFs were assigned to five groups based on the different cellular functions of their potential target genes: cell wall modification, development, cell cycle, metabolism, and response to stress and stimuli.

Table 1.   DNA-binding domain (DBD)-containing Tuber melanosporum transcription factors
Functional classes and gene namesaPutative gene product functionbPreferential expressionc
  1. aT. melanosporum DNA binding domain-containing transcription factors were grouped into major functional classes based on homology with functionally characterized TFs from other fungi; gene names were derived from those of the corresponding homologs (see Supporting Information Table S1 for further sequence information).

  2. bSpecific putative function of Tuber TFs as deduced from the known function of their characterized homologs.

  3. cArrows indicate expression ratios, derived from oligoarray and RNA-seq data, ≥ 2 (↑) or ≤ 0.5 (↓) in one of the investigated lifecycle stages with respect the other two (ECM, ECMs; FB: fruiting bodies; FLM: free-living mycelium); TAT-positive TFs are in bold.

Cell wall
 TmelXlnRActivator of plant cell wall-degrading enzymes  
 TmelAce1Repressor of plant cell wall-degrading enzymes  
 TmelCtf1Transcriptional activator of cutinase genes  
 TmelMrgARegulation of cell wall integrity   
 TmelRlmAMaintenance of cell wall integrity   
 Sexual development and fruiting body formation
  TmelStuARegulator of fruiting body and asexual development   
  TmelSteARegulator of fruiting body development  
  TmelMat 1-2-1MAT1-2-1 mating-type protein   
  TmelMcm1Regulator of pheromone response   
  TmelSte11Regulator of pheromone response   
  TmelAsl1Ascospore lethal 1   
  TmelFf7Female fertility  
  TmelSub2Submerged protoperithecia   
  TmelNosANumber of sexual spores, regulator of sexual development   
  TmelNdt80Meiosis-specific transcription factor   
 Asexual development and basal hyphal growth
  TmelLah3Long aerial hyphae, regulator of asexual development  
  TmelTah1Tall aerial hyphae, regulator of asexual development   
  TmelCol27Colonial, regulator of hyphal growth   
  TmelCol21Colonial, regulator of hyphal growth   
  TmelVad5Vegetative asexual development, regulator of hyphal growth and asexual development   
  TmelVad3-1Vegetative asexual development, regulator of hyphal growth and asexual development  
  TmelVad3-2Vegetative asexual development, regulator of hyphal growth and asexual development  
  TmelKal1KALeidoscope-1, regulator of hyphal growth and asexual development   
  TmelAbaARegulator of conidiation  
  TmelDevRRequired for conidiophore development  
  TmelFlbBFluffy Low BrlA expression, regulator of asexual and conidiophore development   
  TmelFlbCFluffy Low BrlA expression, regulator of asexual and conidiophore development   
  TmelFlbDFluffy Low BrlA expression, regulator of asexual and conidiophore development  
  TmelOefCOver expressed fluffy  
  TmelHms1Regulator of cell morphology   
  TmelCon7Cell morphology regulator   
  TmelAda1All development altered, regulator of basal hyphal growth and asexual and sexual development   
  TmelAda3All development altered, regulator of basal hyphal growth and asexual and sexual development  
Cell Cycle
 TmelSwi6MBF complex, regulator of cell cycle   
 TmelMbp1MBF complex, regulator of cell cycle  
 TmelFkh2Regulator of cell cycle   
 TmelSep1Regulator of cell cycle   
 TmelSwi5Regulator of cell cycle   
 TmelYox1Regulator of cell cycle   
 TmelUme6Key transcriptional regulator of early meiotic genes   
 TmelRme1Regulator of cell cycle: prevents meiosis, promotes mitosis  
  TmelCreAMajor carbon catabolite repression protein   
  TmelNrg1Carbon catabolite repression   
  TmelAzf1Glucose responsive transcription factor  
  TmelRgt1Glucose responsive transcription factor   
  TmelFacBActivates transcription of genes required for acetate utilization   
  TmelAcuMPositive regulator of gluconeogenesis   
  TmelAcuKPositive regulator of gluconeogenesis  
  TmelAreAMajor, positively acting, nitrogen regulatory protein  
  TmelTamAPositively acting nitrogen regulatory protein  
  TmelAreBMajor nitrogen catabolite repression protein  
  TmelMeaBNitrogen catabolite repression   
  TmelGlcDRegulation of nitrogen metabolism   
  TmelCpcACross pathway control protein, amino acid metabolism; regulator of fruiting body development   
  TmelNirA-1Pathway specific, positively acting nitrate regulatory protein   
  TmelNirA-2Pathway specific, positively acting nitrate regulatory protein   
  TmelNirA-3Pathway specific, positively acting nitrate regulatory protein  
  TmelAro80Pathway specific, positively acting aromatic amino acid regulatory protein   
  TmelUay1Pathway specific, positively acting purine regulatory protein  
  TmelAmdXPathway specific, positively acting acetamidase regulatory protein   
  TmelArcAPathway specific, positively acting arginine regulatory protein  
  TmelPpr1Pathway specific, positively acting pyrimidine regulatory protein   
  TmelLys14Activator of lysine biosynthesis pathway   
  TmelMet32Activator of sulfur metabolism  
  TmelMetRActivator of sulfur metabolism   
  TmelCbf1Activator of sulfur metabolism; centromere binding protein  
  TmelOaf1Activator of fatty acid metabolism   
  TmelFarAActivator of short- and long-chain fatty acid metabolism  
  TmelFarBActivator of short-chain fatty acid metabolism  
  TmelUpc2Sterol regulatory element binding protein   
  TmelPalcATranscription factor, involved in regulation of phosphate acquisition  
  TmelHapBCCAAT binding complex, subunit B  
  TmelHapCCCAAT binding complex, subunit C   
  TmelHapECCAAT binding complex, subunit E   
  TmelHapXCCAAT binding complex, subunit X; iron-responsive factor   
Stress and stimuli response
 TmelSfl1Activator of stress responsive genes   
 TmelFcr3FluConazole resistance protein 3  
 TmelSfp1Cell cycle regulator in response to nutrients and stress   
 TmelAsg1Regulator of stress response and drug resistance   
 TmelRpn4Regulator of response to stress  
 TmelHsf1Heat shock transcription factor   
 TmelAp1Regulator of oxidative stress tolerance   
 TmelSkn7Response to osmotic and oxidative stress   
 TmelPacCActivator of alkaline-induced genes; repressor of acid-induced genes   
 TmelHaa1Weak acid stress response regulator   
 TmelCrz1Activator of genes involved in stress response   
 TmelZap1Activator of zinc responsive genes  
 TmelHacAUnfolded protein response   
 TmelWc1Light response and circadian rhythm regulator   
 TmelWc2Light response and circadian rhythm regulator   
 TmalSltAResponse to abiotic stress regulator  

All TFs, which, based on information available in other ascomycetes (both saprotrophic and phytopathogenic) are known to be involved in the regulation of genes encoding plant cell wall-degrading enzymes, are present in the Tuber genome (Table 1); for example, the carbon catabolite repressor CreA, the repressor Ace1, the pH regulator PacC, the Hap complex, the nitrogen regulator AreA, and the activator of genes encoding cellulose/hemicellulose degrading enzymes XlnR (Aro et al., 2005) (Fig. 3a). One exception is the transcription factor Ace2, which at a variance with the situation in Trichoderma reesei and N. crassa, appears to be absent in Tuber. However, the lack of Ace2 was previously reported also in A. nidulans and an Ace2 homolog appears to be missing in A. niger and Magnaporthe grisea as well (Aro et al., 2005), suggesting that different transcriptional regulators are likely to substitute for Ace2 function in these fungi. Of note, XlnR and the cutinase enzyme regulator Ctf1 are both preferentially expressed in mycorrhiza (ECM : FLM ratio c. 6 and c. 3, respectively; see Fig. 3a and Table S1).

Figure 3.

 Transcription factors (TFs) differentially expressed, and potentially involved in specific stages of the Tuber life cycle. (a) TFs regulating the expression of cell wall-degrading enzymes are preferentially expressed in ectomycorrhizas (ECMs). Also shown in this scheme is the Tuber-unique DBD-containing hypothetical protein (HP-DBD) encoded by GSTUMT00012451001, a gene that is strongly up-regulated in ECM and that was designated as XlnR2 because of its similarity to the cellulase/hemicellulase gene activator XlnR. (b) A subset of TFs preferentially expressed in fruiting bodies (FBs) positively regulate the expression of metabolic enzymes involved in fatty acid metabolism, including the glyoxylate cycle and gluconeogenesis, two pathways known to be particularly sustained in truffles (Abbàet al., 2007). By comparison, only one regulator of sulfur metabolism (Cbf1) and no TF potentially involved in sexual development were found to be differentially up-regulated by more than twofold in FBs. TFs are represented with a false-color code (see the color scale inset) indicating their relative expression levels in ECMs or FBs compared with free-living mycelium (FLM, see text and Supporting Information, Table S1 for more information on these TFs).

Several homologs of Neurospora TF genes involved in development (e.g. hyphal growth and asexual/sexual development; Colot et al., 2006) are present in the Tuber genome, along with regulators of FB development and pheromone response (e.g. SteA, the ortholog of the yeast regulator Ste12; StuA; NosA; Ste11; and the mating type protein Mat1-2-1 (Martin et al., 2010) (see Table 1)). Curiously, Mat1-2-1 and SteA are both weakly expressed in FBs, while an ortholog of the Aspergillus NsdD gene (encoding a GATA-type transcription factor required for sexual development) could not be found in the Tuber genome. However, the homologs of other TFs that are essential for FB or conidiophore formation in N. crassa (e.g. the White Collar proteins WC-1 and WC-2) and A. nidulans (e.g. the general amino acid control activator CpcA, the G-protein modulator FlbA and the TFs FlbB, FlbC, and FlbD) are present in T. melanosporum (Poeggeler et al., 2006).

Homologs of the main regulators of nitrogen and carbon metabolism are encoded by the T. melanosporum genome. These are accompanied by the homologs of general activators for the utilization of alternative nitrogen sources and pathway-specific activators for the metabolism of particular nitrogen sources. Of note, the activator of nitrate metabolism NirA is present in three copies in the Tuber genome. As shown in Table 1, regulators of aromatic amino acid biosynthesis were also identified, along with all known regulators of sulfur assimilation, which may be important for truffle flavor formation. In accordance with the sustained sulfur metabolism operating in fruiting bodies (Martin et al., 2010), the mRNAs for the latter components, which include the Cbf1 activator, exhibit a markedly preferential expression in fruiting bodies. The Met30 repressor is also present in the Tuber genome, but it is not listed in Table 1 because it does not bind to DNA directly.

Finally, regulators of ‘stress and stimuli’ responses, including the ECM-up-regulated activator of zinc homeostasis genes (Zap1), were identified. Genes coding for the major regulator of the unfolded protein response (HacA) are apparently amplified, with as many as 18 copies in the Tuber genome. However, careful inspection of the corresponding gene models showed that 17 putative HacA genes are severely truncated, not expressed, and/or very close to transposable elements, suggesting that all but one of them (TmelHacA) are putative pseudogenes (Table S1).

Functional screening and validation of truffle transcriptional activators in yeast

To validate functionally some of the TFs predicted earlier in the paper and to discover new transcriptional activators, bioinformatic analysis was coupled to the TAT assay, a heterologous gene transactivation screen in yeast (Fig. 1). To this end, cDNA libraries prepared from the three life cycle stages of T. melanosporum were transferred to the pDEST32 yeast expression vector, and transformants bearing an in-frame fusion of the vector-provided yGAL4-DBD with a Tuber transcriptional activator were positively selected and monitored via reporter gene transactivation (see Fig. S1 for representative examples). At least 500 000 cDNA clones from each library were screened in this way and the inserts of positive selected clones were sequenced. A total of 438 sequences (160 from the FLM, 126 from the FB, and 152 from the ECM library) gave good-quality results and were retained for subsequent analysis. They were organized into 53 contigs and 88 singletons for a total of 141 unisequences (see Fig. 4). Most of these sequences (100) were of fungal origin, and 37 of them contained a recognizable DBD. Thirty-three of the latter sequences matched in silico predicted TFs, while the remaining four escaped detection because the DBD was missing as a result of misprediction of the first exon, or because only a partial DBD was present (see Table S1). As expected, a number of positive clones retrieved from the TAT screen of the ECM library (a total of 41 sequences) resembled plant polypeptides: 22 sequences bore a recognizable DBD and 12 of them shared a significant sequence similarity with plant transcriptional activators involved in plant–pathogen interaction (Fig. 5, Table S2). The remaining 82 TAT-positive sequences lacked a recognizable DBD and were collectively designated as ‘unconventional activators’ (see the paragraph ‘Unconventional activators’).

Figure 4.

 Venn diagram representation of the number of independent clones (left) and corresponding unisequences (right) isolated from the transcriptional activator trap (TAT) screening of the free-living mycelium (FLM), fruiting body (FB) and ectomycorrhiza (ECM) libraries. The number of DNA-binding domain (DBD)-containing clones are shown between brackets. Sequences of plant origin retrieved from the screening of the ECM library are shown on a green background.

Figure 5.

 DNA-binding domain (DBD)-containing transcription factors (TFs) of the plant (hazelnut) mycorrhizal partner retrieved from the transcriptional activator trap (TAT) screening of the ectomycorrhiza (ECM) library. TFs involved in microbial pathogen defense (Stracke et al., 2001; Shi & Shi, 2004; Olsen et al., 2005) are emphasized.

Among the functionally validated Tuber transcription factors homologous to TFs previously characterized in other fungi (pTFs; 22 sequences), we found two regulators of cell wall integrity (TmelAce1 and TmelMrgA), seven regulators of development and cell cycle genes, six stress response TFs, and seven regulators of metabolism (Table 1).

Five cDNAs, all belonging to the pTF class, were independently isolated > 20 times in the TAT assay and accounted for > 30% of the entire set of sequenced clones: TmelCpcA (49 clones), TmelMetR (25 clones), TmelKal1 (22 clones), TmelMeaB (27 clones), plus the ethylene-responsive plant TF RAP2.12 (24 clones) (see Tables S1, S2). Analysis of the TmelMeaB contig revealed an expressed sequence tag (EST)/mRNA population which differed from that associated with the predicted gene model (GSTUMT00003085001) at the level of the 5′-splice site of the last intron. In fact, the initial gene model was predicted on the basis of the EST population produced by the downstream splice site (i.e. with a retained partial intron of 14 nt), while eight sequences mapping to this region, picked with the TAT screening, all contained the upstream splice site. Analysis of all the available ESTs, coupled with RNAseq data from FLM and FB, showed that one-third of the MeaB cDNAs had a retained partial intron, while two-thirds of them belonged to the same population identified with the TAT screening. Alternative splicing at these two sites thus produces two forms of MeaB with different C-terminal sequences. The form encoded by the mRNA lacking the retained partial intron is 61 amino acids longer, is more similar to MeaB fungal proteins, and is thus probably the only one that is transcriptionally active.

Fifteen of the TAT-positive clones were either Tuber-unique HP-DBDs (six) or CH-TFs (nine). Since they all met the basic criteria of our analysis (i.e. the presence of a DBD and the ability to activate reporter gene expression in yeast, along with above-background expression levels in at least one stage of the Tuber life cycle), we consider the corresponding proteins as newly identified, Tuber-unique (HP-DBDs) or general filamentous ascomycete (CH-TFs) TFs (see Table S1).

The TAT screen thus allowed the functional validation of about one-fifth of in silico predicted TFs, the adjustment of three gene model(s), and the de novo identification of four DBD-containing transcriptional activators. This was accompanied by the identification of at least 22 bona fide plant TFs potentially bearing on the response of the host plant to T. melanosporum infection and by the discovery of 82 unconventional putative transcriptional activators (see the paragraph ‘Unconventional activators’).

Unconventional activators

DNA-binding domain-lacking sequences similar to previously described nuclear or mixed localization proteins as well as to proteins apparently functioning outside of the nucleus were also found among the TAT-positive clones. These included 20 proteins, eight nuclear and 12 with a mixed (nuclear/cytoplasmic) localization, called ‘unconventional activators’ (UAs), for which a direct (or indirect) role in transcriptional regulation has previously been demonstrated (Table 2), and 43 proteins, called ‘putative unconventional activators’ (PUAs) with no previously documented involvement in transcription (Table S3).

Table 2.   DNA-binding domain (DBD)-lacking, unconventional activators retrieved from the transcriptional activator trap (TAT) screena
Subcellular localization and descriptionbGene modelSaccharomyces cerevisiae homolog
  1. aTAT-positive unconventional activators retrieved from functional screening of the Tuber melanosporum free-living mycelium (FLM), fruiting body (FB) and ectomycorrhiza (ECM) libraries in yeast (see Supporting Information, Table S3 for further information).

  2. bThe subcellular (nuclear or mixed) localization of homologous proteins from other organisms along with the gene name of the yeast homolog, when available, are indicated.

  3. cUnconventional activator differentially expressed in FB, with FB : FLM and FB : ECM expression ratios of 40.0-2.1 and 747.7-5.7 (oligoarrays-RNAseq), respectively.

Nuclear proteins
 Mitogen-activated protein kinaseGSTUMT00005852001Hog1
 DNA-directed RNA polymerase II subunit  RPB1GSTUMT00009762001Rpo21
 Cryptic loci regulator 2GSTUMT00004358001 
 Mitogen-activated protein kinase MAF1GSTUMT00005390001Maf1
 Casein kinase II beta 2 subunitGSTUMT00011807001 
 Serine/threonine-protein kinase KSP1GSTUMT00006520001Ksp1
 DASH complex subunit ask1GSTUMT00005863001Ask1
 Ubiquitin thiolesterase (OtuB1)GSTUMT00000718001 
Mixed localization proteins
 E3 ubiquitin-protein ligase NEDD4GSTUMT00011744001Rsp5
 EH Domain and endocytosis-related proteinGSTUMT00009307001Ede1
 Charged multivesicular body protein 5GSTUMT00012039001Vps60
 Low-temperature viability protein ltv1GSTUMT00001950001Ltv1
 Peptide methionine sulfoxide reductase  msrB/msrAcGSTUMT00004791001Mxr1
 Phospholipase D1 (PLD1), putativeGSTUMT00003661001Spo14
 Mitosis inhibitor protein kinase SWE1GSTUMT00005490001Swe1
 vts1 proteinGSTUMT00007890001Vts1
 Ankyrin repeat proteinGSTUMT00003500001Nas6

Among the UAs we found various kinases, ubiquitinating enzymes, endocytotic and actin-related proteins, some of which are known to shuttle between the cytoplasm and the nucleus, where they act as scaffolds for the recruitment of the basal transcription apparatus or play a role in chromatin modification (Pilecka et al., 2007; Zheng et al., 2009). Various nuclear actions have been reported for these proteins, which are often found associated with actively transcribed mammalian genes. These include transcription stimulation, chromatin remodeling, RNA/protein export from the nucleus in the case of actin, and enhanced transactivation by the nuclear receptor activator SRC-2 in the case of alpha-actinin (Zheng et al., 2009). Two endocytotic proteins with a previous record of nuclear actions were also found among the TAT-positive clones: TmelEde1, which is homologous to human Eps15p (Vecchi et al., 2001), and TmelVps60, the Tuber homolog of charged multivesicular body protein 5 (CHMP5), which has been shown to interact with chromatin-remodeling enzymes as well with proteins involved in nuclear sumoylation and with the RNA polymerase III transcription factor IIIA in human cells (Tsang et al., 2006).

Perhaps more problematic are the ‘putative unconventional activators’. In fact, only half of these proteins – six of which are conserved hypothetical proteins and two of which are Tuber-unique proteins – bear a recognizable NLS and can thus be hypothesized to be capable of localizing to the nucleus autonomously. Indirect nuclear localization mechanisms (e.g. ‘piggy back’ translocation) along with some false positives (i.e. cytoplasmic proteins that function as transcriptional activators when artificially forced into the nucleus by the vector-borne NLS) can instead be hypothesized for the other 35 TAT-positive polypeptides. As previously noted, however, there is an increasing number of apparently cytoplasmic proteins for which a genuine nuclear function has been documented (Shi & Shi, 2004; Pilecka et al., 2007). In keeping with recent findings in yeast (Hall et al., 2004), among Tuber PUAs we found metabolic enzymes, but also proteins involved in vesicular transport. Plant-derived UAs (e.g. the nuclear signaling decoder calmodulin, involved in the response to pathogen invasion) and PUAs were also retrieved from TAT analysis of the ECM library and are listed in Table S2. A detailed analysis of the ability of TAT-positive PUAs to enter the nucleus autonomously is under way.

Transcription factors preferentially expressed in ECMs and FBs

Preferential expression of all predicted TFs in the ectomycorrhizal and/or FB stage of the Tuber life cycle was investigated by microarray hybridization and RNA sequencing (see the ‘Materials and Methods’ section and (Martin et al., 2010) for details). Twenty-five TFs were found to be up-regulated more than fivefold in ECM (16) or FB (nine) relative to FLM, with up-regulation levels as high as 49-fold and 4500-fold for the ECM and the FB stage, respectively (Table 3). A homolog of the stress response regulator Rpn4, two conserved hypothetical TFs and a Tuber-specific Zn cluster TF were up-regulated more than fivefold in both life cycle stages. Five of the above TFs were independently retrieved from both in silico analysis and TAT screening, while seven of them were individually subjected, and scored positive to the TAT assay following prioritarization by expression profiling.

Table 3.   Transcription factors (TFs) differentially up-regulated in ectomycorrhizas (ECMs) or fruiting bodies (FBs)aThumbnail image of

Transcription factors with no similarity to functionally characterized proteins were prevalent among transcriptional activators specifically up-regulated in ECMs. Thus, only indirect inferences regarding the function and potential targets of these TFs can be made at present. For example, the hypothetical Zn cluster DBD-containing protein encoded by GSTUMT00012451001 – one of the top up-regulated TFs in ECM – is distantly related to XlnR, an activator of plant cell wall-degrading enzymes (cellulases/hemicellulases) in other fungi. An ortholog of the XlnR activator, named TmelXlnR, is present in the Tuber genome and is up-regulated by more than fivefold in ECM compared with FLM. However, its expression levels are 10-fold lower than those for the transcript encoded by GSTUMT00012451001, a novel putative activator that we designated as XlnR2 (Fig. 3a, Table S1), whose polysaccharide substrates and physiological role in Tuber are presently unknown.

Among ECM-preferential TFs with a functionally characterized homolog in other fungi, we found the regulator of cell morphology TmelHms1, the regulators of asexual development TmelFlbD and TmelDevR, and the regulator of sexual development TmelFf7. Interestingly, TmelFf7 as well as its Neurospora homolog contain the maltose acetyltransferase domain typical of the Rhizobium nodulation factor NodL. The stress response regulators TmelSFP1 and TmelSFL1 were also found to be overexpressed in ECM, suggesting that control of stress responsiveness and morphological differentiation may be jointly regulated processes during ECM formation. Of note, the homologs of TmelSfp1 and of the conserved C2H2-Zn finger TF GSTUMT00007914001 are also up-regulated in ECMs from the basidiomycete Laccaria bicolor.

Nine predicted TFs, six containing a Zn cluster DBD, are preferentially expressed in FBs. Five of these TFs have a functionally characterized homolog in other fungi: TmelAbaA, a regulator of asexual development; the positive regulators of gluconeogenesis TmelAcuM and TmelAcuK; the sulfur metabolite activator TmelCbf1; and the activator of short-chain fatty acid metabolism TmelFarB. FB-preferential expression of the latter four regulators fits with the notion that gluconeogenesis, especially via fatty acid utilization through the glyoxylate cycle (Fig. 3b), and sulfur assimilation are highly sustained metabolic pathways in Tuber FBs. The remaining four FB-preferential activators are either ‘conserved hypothetical TFs’ (two) or Tuber-unique HP-DBDs (two). One of the latter proteins, bearing a Zn cluster DBD, is encoded by the second most up-regulated gene (GSTUMT00009814001) in FBs.


This work describes the repertoire of TFs, the so-called ‘regulome’, of a plant-symbiotic ascomycete producing highly prized subterranean FBs. A total of 201 sequences coding for putative and/or functionally validated TFs has been identified by the combined use of in silico prediction and a TAT assay conducted in the model ascomycete S. cerevisiae. The TAT assay allowed the validation of functionally about one-fifth of the in silico predicted Tuber TFs; a hit rate that is only twofold lower than that reported for the homologous validation of yeast TFs, where the lack of reporter gene activation for 83 out of 138 predicted yeast transcriptional activators was ascribed to the requirement for specific promoter elements, cofactors and/or growth conditions (Titz et al., 2006). On the other hand, the totally unbiased nature of the TAT assay led to the de novo identification of four TFs along with a number of ‘unconventional activators’ that would have been impossible to predict simply based on sequence similarity. The same screening also allowed the assignment of a putative ‘transcriptional activation function’ to 12 hypothetical, functionally uncharacterized proteins (CH-TFs) with a widespread occurrence in fungi. Additionally, we verified that all the candidate TFs we identified are expressed in at least one stage of the Tuber life cycle, and 14% of them were found to be transcriptionally up-regulated in a stage-dependent manner in ECMs or FBs.

Most TFs are encoded by single-copy genes in the T. melanosporum genome. The sole exceptions (leaving out a few instances of TF pseudogenes) are the homolog of the N. crassa regulator of hyphal growth and asexual development VAD3, which is present in two copies, and the nitrate metabolism activator NIRA, which is present in three copies. The occurrence of extra copies of NIRA is potentially interesting in view of the fungal nitrogen–plant carbon trade taking place in ECMs and because nitrogen metabolism is likely to be altered during symbiosis establishment and maintenance. In fact, the expression of nitrate assimilation enzymes has previously been shown to be modified in Tuber borchii ECMs and, unlike the situation in most fungi, nitrate assimilation enzymes appear not to be subjected to nitrogen catabolite repression in Tuber (Montanini et al., 2006b; Guescini et al., 2007).

The fact that > 50% of the Tuber TFs are homologous to previously characterized transcriptional regulators from other fungi served as a ‘positive control’ for our analysis and provided important clues regarding the gene targets and potential function of these TFs. This is especially true for some TFs preferentially expressed in ECMs (e.g. TmelXlnR, TmelCtf1, TmelAce1), whose homologs in other filamentous ascomycetes have been implicated in biochemical processes related to plant cell wall degradation and to the establishment of biotrophic (plant–pathogen) interactions (Fig. 3a); for example, genes with a role in the deconstruction of plant cell walls, host cell adhesion, detoxification of plant defense metabolites, and nutrient exchange, many of which are differentially up-regulated in T. melanosporum ECMs (Martin et al., 2010). Similar considerations hold for a subset of TFs preferentially expressed in FBs (e.g. TmelACUM, TmelFARB, TmelACUK and TmelCBF1), which pinpointed potential target genes involved in metabolic pathways (e.g. gluconeogenesis, glyoxylate cycle and sulfur assimilation) that are particularly active in FBs (Fig. 3b) (Lacourt et al., 2002; Gabella et al., 2005; Abbàet al., 2007; Martin et al., 2010).

Identification of putative homologs of TFs acting on specific genes and pathways in other fungi may also prompt investigations aimed at verifying the occurrence of functionally related target genes in Tuber. This is the case, for example, of the two Tuber homologs of XlnR, a major regulator of sugar hydrolytic enzyme genes (endoglucanases, cellobiohydrolases and xylanases) in saprotrophic filamentous ascomycetes (Aro et al., 2005). Three beta-1,4-endoglucanases are encoded by the Tuber genome and one of them (TmelCMC3) is strongly up-regulated (> 6000-fold) in mycorrhiza, similar to its L. bicolor ortholog (Martin et al., 2008). In contrast, no cellobiohydrolase and one, but ECM-down-regulated, xylanase (endo-1,4-beta-xylanase precursor; TmelXylA) are present in T. melanosporum.

Similar considerations hold for the Tuber homolog of the cutinase enzyme regulator (Ctf1), for which no apparent cutinase gene could be identified in the T. melanosporum genome. It should be noted, however, that the physiological targets of mycorrhizal fungi are host plant roots, that do not contain cutin, and that the Ctf transcription factors initially characterized in Nectria haematococca are the orthologs of the lipid metabolism regulators FarA and FarB, with which they share a common DNA binding site. Indeed, comparative genomic analyses in Aspergillus spp. highlighted the presence of conserved FarA and FarB DNA binding sites upstream of cutinase, but also of lipid metabolism and peroxisomal enzyme genes (Hynes et al., 2006). Interestingly, one of the genes most strongly up-regulated in T. melanosporum mycorrhiza codes for a secreted lipase/esterase (Martin et al., 2010), which might be involved in host membrane remodeling.

Of note, about one-quarter of the FB- or ECM-up-regulated TFs (see Table 3) are among the top 10% most highly expressed genes in FBs or ECMs. This is a rather peculiar situation, considering that TF-encoding genes are usually expressed at very low levels. This might point to the high degree of functional specialization associated with different stages of the Tuber life cycle and warrants future investigations aimed at discovering the actual target genes and validating the biological processes associated with these TFs. A good case in point is represented by a few TFs whose homologs in other ascomycetes are implicated in asexual development and somatic spore (e.g. conidia) formation, a process whose existence in Tuber spp. is still very controversial (Urban et al., 2004). Similar in-depth investigations, including the experimental identification of the DNA binding sites, will specifically address Tuber-unique TFs, such as the TAT-validated activator encoded by GSTUMT00009814001 – the second most expressed gene in FBs – as well as newly identified UAs. The latter components, especially those without a prior record of nuclear localization and activity (PUAs), represent another significant outcome of this analysis. Although a relatively high rate of false positives may be expected as a result of the presence of a vector-borne NLS, the occurrence of moonlighting proteins capable of functioning both in the cytoplasm and in the nucleus is in line with recent findings in yeast (and other organisms) as to the gene transactivation ability of otherwise cytoplasmic proteins, especially metabolic enzymes (Hall et al., 2004; Pilecka et al., 2007; Hu et al., 2009). This points to the existence of an as yet largely unexplored set of hard-to-predict transcriptional regulators and raises the exciting possibility of a direct link between metabolism and transcription.

A final note regards the plant TFs retrieved from the TAT screen of the ECM library, 12 of which belong to specific classes of transcriptional regulators (e.g. NAC, WRKY, MYB and ER) that in plants attacked by phytopathogenic fungi have been shown to control various kinds of defense-related processes (Stracke et al., 2001; Singh et al., 2002; Olsen et al., 2005). Another case in point is the presence among plant UAs of the hazelnut homolog of Arabidopsis calmodulin, a multifunctional protein that, despite its predominant cytoplasmic localization, is capable of entering the nucleus, where it has recently been shown to negatively regulate defense responses to pathogen invasion in a Ca2+-dependent manner (Kim et al., 2009). Plant TFs functionally identified in T. melanosporum/hazelnut mycorrhiza may thus open a new window on the mutualistic interaction between symbiotic fungi and their plant partners, and provide important clues as to its mechanistic peculiarities compared with better characterized host and nonhost plant–pathogen interactions.


We thank Patrick Wincker and all the Génoscope team (Evry, France) for their invaluable contribution to the sequencing and assembly of the T. melanosporum genome. We are grateful to Gérard Chevalier (INRA-Universite′ Blaise Pascal, Clermont-Theix, France) for T. melanosporum mycelia and fruiting bodies; Raffaella Balestrini (Istituto per la Protezione delle Piante, CNR, Torino, Italy) for mycorrhizal root tips; and Marian Walhout (University of Massachusetts Medical School, Worcester, USA) for yeast strain MaV103. Financial support from the Fondazione Cariparma (Parma, Italy) and from the Ministry of Education, University and Research (Rome, Italy; MIUR-PRIN 2007 program) to S.O. is gratefully acknowledged. B.M. was funded by a postdoctoral fellowship from the Fondazione Cariparma; E.L. is a graduate student in ‘Biochemistry and Molecular Biology’ at the University of Parma. Work in F.M.’s laboratory was supported by grants from the European Network of Excellence EVOLTREE and the Région Lorraine Council (Project FORBOIS). E.T. was funded by a joint PhD scholarship from the European project ENERGYPOPLAR and Région Lorraine.