We constructed a Δcre1 recombinant strain of T. reesei QM 9414 by replacing its ORF and part of its 5'-and 3'-nt regions by a hygromycin B resistance gene 19 under constitutive expression signals. In accordance with Nakari-Setälä et al. 18 , the corresponding knock-out strains exhibited a reduced radial growth rate on plates, and formed smaller colonies, fewer aerial hyphae and less spores. In addition, the Δcre1 strain displayed shorter but more robust hyphae, that contained a considerably thickened cell wall and less septa (Additional File 1: Figure S1 and Table S1). All these mutant phenotypes were eliminated by retransforming the Δcre1 strain with the cre1 gene, thus proving that they are cre1 specific (data not shown). We used phenotype microarrays to analyze whether the above noted reduction in the growth rate is general or specific for some carbon sources only. The results showed that - in contrast to the radial growth on plates - the Δcre1 strain grew significantly slower on only 5 of 95 carbon sources (Glycogen -37 [±4], p = 0.004; arbutin -27 [±4], p = 0.0006; adenosine -42 [±5], p = 0.022; salicin -34 [±4], p = 0.007; and amygdalin -42 [±6], p = 0.013), but on the other hand was unaffected on the majority of them (within ± 25% of variation, Chi square test p > 0.05; see Additional File 1, Figure S2). However, increased growth of > 30% of the control, which would be expected if CRE1 represses growth on a given carbon source, was observed for 9 carbon sources shown in Additional File 1, Figure S3 (D-galactose +38 [±3], p = 0.0008; L-sorbose +34 [±3], p = 0.027; D-xylose +48 [±5], p = 0.002; palatinose +31 [±6], p = 0.033; maltose +58 [±6], p = 0.002; stachyose +45 [±4], p = 0.03; xylitol +37 [±4], p = 0.017; adonitol +46 [±6], p = 0.0008); and glucuronamide +73 [±3], p = 0.044).

We have used whole-genome DNA microarrays with the goal of identifying the main genes in T. reesei that are controlled by CRE1. Since carbon catabolite repression is known to be dependent on the growth rate, the microarray experiments were therefore performed in chemostat cultures at two constant growth rates that were earlier shown 16 to be carbon repressing (0.07 h^-1) and potentially carbon derepressing (0.025 h^-1), respectively (Figure 1). After data pretreatment and normalization, we applied the linear modeling approach and the Bayes statistics implemented in the limma R package 20 to our biological replicates as described in the Methods part. Using these criteria we retrieved a list of 251 genes whose expression was regulated in at least one of the four experimental conditions. One of them was cre1 itself (in the parent strain), which is trivial and was omitted from all further investigations. The (in part overlapping) occurrence of the truly regulated 250 genes under the four different experimental conditions is shown in Additional File 1, Figure S4. We used clustering algorithms (see Methods) to divide them into 9 different clusters (Figure 2). The effect of CRE1 and its interplay with the growth rate is summarized in Figure 3: 47.3% of the identified genes are in fact repressed by CRE1, but 29.0% are CRE1 induced and 17.2% are CRE1 independent. In addition, 62.2% of the genes are influenced by the high growth rate (34.5% induced and 27.7% repressed), 24.9% are derepressed at the low growth rate and 6.4% are growth rate independent. A complete list of all genes is given in Additional File 1, Table S2. To confirm these microarray results, quantitative Real-Time-PCR (qRT-PCR) was performed on a subset of the genes belonging to different clusters. The gene for isocitrate lyase was included in this set, as its regulation by CRE1-mediated carbon catabolite repression is known ( 16 , and references therein) and its appearance in our dataset confirms that the growth rates were chosen appropriately. As shown in Additional File 1, Table S3, these genes that showed differential expression between two or more of the used conditions in the microarray study were also differentially expressed in the same direction upon Real Time-PCR analysis. We therefore conclude that the microarray expression ratios indeed reflect differences in the expression of these genes.

Up and down regulated genes were annotated and categorized as described in the Methods section. The enrichment for the major FunCat categories was assessed in each cluster, and a Gene Ontology (GO) enrichment analysis was also done on the whole dataset as further control (see Additional File 1, Tables S4 and S5).

As shown in Figure 3, genes which were upregulated in the Δcre1 mutant strain could be grouped into four classes: genes that were upregulated in the CRE1 knockout independently of the growth rate (cluster C, 16 genes); genes upregulated in the CRE1 knockout at high growth rate only (cluster E, 50 genes); genes for which CRE1 function counteracted an induction at high growth rate (cluster G, 26 genes); and genes for which CRE1 function counteracted an induction at low growth rate (cluster H, 26 genes).

Figure 4a shows the distribution of functional categories (FunCat) within these 4 gene clusters. Genes which have orthologues in other fungi but for which no function can be predicted were most abundant in clusters E and G. Genes for carbohydrate degradation were most abundant in the cluster that groups together the genes upregulated in the CRE1 knockout at high growth rate (cluster E). In this cluster, we detected a significant enrichment of genes involved in cellular transport (see Table S4). This category was also significantly enriched in cluster C, which comprises the genes upregulated in the CRE1 knockout independently of the growth rate. The nature of cellular transport proteins was diverse in cluster E, but was dominated by permeases for transport of nitrogenous compounds in cluster C.

The cluster of the genes upregulated in the CRE1 knockout at high growth rate (E) also exhibited genes linked to transcription regulation (e.g. the helicase SNF2 involved in chromatin remodeling and the transcriptional regulator MedA involved in sporulation), as well as PhiA and RAS1, both participating in hyphal development.

This latter function was also represented with the bys1 transcript in the cluster of genes for which CRE1 function counteracted an induction at high growth rate (cluster G). This cluster comprised different genes linked to carbohydrate metabolism (e.g. glycoside hydrolases and glycoside transferases), as well as hsp30, a single integral plasma membrane heat shock protein that is induced by several stresses, including glucose limitation 21 .

As found for cluster E, the cluster that groups together genes for which CRE1 function counteracted an induction at low growth rate (cluster H) exhibited a majority of genes involved in cellular transport. A membrane protein linked to sporulation, CMP1, has also been found in this cluster. The significance of the enrichment of these functional categories has been verified using Gene Ontology annotation (Additional File 1, Table S5).

Genes contained in cluster X exhibited a heterogenous pattern of gene expression, but their expression was consistently highest in the Δcre1 strain at low growth rates, and lowest in the parental strain at high growth rates. We therefore conclude that genes of this cluster are also subject to CRE1-dependent CCR. Gene with putative functions did not reveal any common characteristics. However, we detected the orthologue of A. nidulans creD (cre4), which encodes a protein with arrestin and PY motifs known to interact with ubiquitin ligase 22 . Disruption of creD in A. nidulans confers partial relief from CCR 23 .

The expression of a smaller number of genes was decreased in the Δcre1 strain, either only at the low (cluster D, 36 genes) or the high growth rate (cluster F, 36 genes), suggesting that CRE1 acts - either directly or indirectly - positively on these genes. The highest number of genes from the cluster that groups together the genes downregulated in the CRE1 knockout at high growth rate (F) corresponds to genes that encode proteins involved in cellular transport or proteins for which no function has been characterized (Figure 4b). This cluster also comprised 4 genes for carbohydrate metabolism, 2 synthases and 2 oxidoreductases involved in secondary metabolite formation, as well as 3 transcription factors with Zn2Cys6 and bZIP DNA binding domains. Cluster D, which comprise genes downregulated in the CRE1 knockout at low growth rate, exhibited a significant enrichment of genes encoding carbohydrate active enzymes. This cluster also exhibited a broad composition of other various gene categories, those encoding unknown proteins and compounds involved in cellular transport being most abundant.

17.2% of the genes detected in this study displayed an expression pattern that was independent of CRE1 function, i.e. their expression was different only between the two growth rates, but not in any direct comparison between the parent and the Δcre1 mutant strain. 24 of them were activated (cluster B), whereas 19 (cluster A) were actually repressed by a high growth rate. A significant enrichment of genes for carbohydrate metabolism has been detected in the cluster of genes repressed by the high growth rate (cluster A), whereas the group of genes induced at high growth rate (cluster B) comprised the whole set of genes needed to catabolize N-acetylglucosamine to fructose-6-phosphate and exhibited five genes dedicated to fatty acid metabolism (Figure 4c; see also Additional File 1, Tables S4 and S5 for enrichment statistics).

In order to link the diverse regulatory effects of CRE1 identified in this study to a direct gene target, we looked for the CRE1 binding pattern 5'-SYGGRG-3' 10 11 in sequences 1 kb upstream of the coding region of the genes that were up-or down-regulated in the microarray experiments. For each cluster, an average number of sites per gene was calculated and normalized to the average number of sites per gene found in the whole T. reesei genome. Since binding of CreA/CRE1 has been described to require two adjacent binding sites, or may be clustered in cis-regulatory element enriched regions (CRERs, 24 ), we also looked for the occurrence of direct or inverted repeats (motif pairs). The total number of these "paired motifs" was calculated for each gene, and an average per gene was again measured for each cluster, and then normalized to the average number in the genome. Two spacing distances between the two motifs were tested, which were maximally 20 bp or 50 bp. For a value of 50 bp, the average number of complex motifs per gene for the whole genome was 0.9 while it was 0.5 for a 20 bp distance value. For each motif pair alone, values ranged from 0.1 to 0.5. Significance of the differences in average per cluster compared to the average on the whole genome was assayed with bi-directional Student t-test. The relatively high frequency of the motifs, combined with the possible presence of genuine primary and secondary CreA/CRE1 target, resulted in often high standard deviation, making statistical validation difficult. However, the results (Table 1) show that the CRE1 binding motif, was significantly more abundant in the CRE1 repressed cluster E, and in the CRE1 activated cluster D. This observation was less clear for cluster H (genes for which CRE1 function counteracted an induction at low growth rate), while no enrichment was detected for cluster F and G (genes downregulated in the CRE1 knockout at high growth rate, and genes for which CRE1 function counteracted an induction at high growth rate, respectively). While we were not able to statistically validate this, a trend in enrichment is visible for cluster C, which groups together genes that were upregulated in the CRE1 knockout independently of the growth rate. These results suggest that clusters D, E and maybe C are enriched in direct CRE1 targets while clusters F and H are indirect results of the CRE1 regulation. In support of this conclusion, the values obtained for the CRE1 independent genes contained in clusters A and B were in the range of (or even below) the genomic average.

We also looked for the presence of nucleotide motifs that are shared by genes regulated by CRE1, as described in the Methods section. One of the detected motif using the RSA-Tools software (5'-SAGGGRG-3') was indeed significantly enriched in clusters D and E (i.e. the clusters most enriched in CRE1 motifs). This motif is otherwise found at least once at a 57% frequency in all T. reesei promoters.

H. jecorina QM 9414 and the Δcre1 strain derived from it (see below) were maintained on malt extract agar. Constant-mass, chemostat-type continuous cultivations were performed in a 2.5 l glass bioreactor with a working volume of 2 l, essentially as described earlier 31 . The feeding medium contained 3 g l^-1 glucose, a concentration low enough to make the culture carbon-limited. Steady-state of the cultures was established when no changes in biomass dry weight were observed in three successive samples taken over a period of three residence times (= the reciprocal value of the dilution rate). In D-glucose limited cultures, the steady-state biomass concentration was 1.46 ± 0.21 g l^-1, irrespective of the dilution rate. The residual steady-state concentrations of D-glucose in the medium were 0.08-0.10 mM. The calculated growth yield (grams of biomass formed per gram of carbon source consumed) was between 46 and 49% for all cultures, which correlates well with our previous studies 34 . Two subsequently achieved, independent steady-states were sampled and analysed for each dilution rate and fungal strain.

A cre1 deletion vector was constructed by using the double joint PCR technique 44 . Primers used are given in Additional File 1, Table S6. Oligonucleotides cre5'F and Cre5'Rtailhph were used for the promoter region, cre3'R and Cre3'Ftailhph for the terminator region, and M1LHhph and M2LHhph for the hygromycin B expression cassette 45 amplification. The outside primers (cre5'F and cre3'R) were used for the amplification of the whole deletion fragment in a fusion PCR assay. The resulting fragment was subsequently cloned in pGEMT-Easy resulting in pΔcre1hph. The cre1 deletion fragment was released from plasmid pΔcre1hph by a NotI restriction digest. The respective 5.3 kb fragment was eluted with a QIAGEN Gel Extraction Kit (QIAGEN GmbH) and was used to transform protoplasts of the strain QM 9414 as described by Gruber et al. 46 . Transformants were selected on malt extract medium containing hygromycin (50 μg/ml). 40 transformants were selected from the transformation plates and transferred to small malt extract plates containing hygromycin (50 μg/ml). 25 transformants exhibited a stable phenotype and their respective spores were plated on malt extract medium containing hygromycin (50 μg/ml) and Triton X-100 (0.1%) to isolate single spore colonies. The cre1 loci of the six transformants and the parental strain QM 9414 as a control were amplified with the primers cre5'F and cre3'R using the following PCR conditions: 2 min of denaturation (94°C), were followed by 32 cycles of 45 s denaturation (94°C), 45 s of annealing (55°C), and 6 min of elongation (72°C), concluded by 7 min at 72°C. Genomic DNA of strain QM 9414 yielded a fragment of 4.3 kb, while homologous insertion of the deletion vector lead to an increase of the fragment size to 5.3 kb fragment in cre1 deleted strains. For the retransformation of the Δcre1 strain the cre1 gene was amplified with oligonucleotides cre1retrafo1 and cre1retrafo2, which are located about 1500 bp upstream and 700 bp downstream of the coding region, respectively. The fragment was cloned into pGEM-T Easy (Promega). Transformation was performed with the A. nidulans amdS as marker 47 . Successful retransformation of the Δcre1 strain with plasmid pcre1amdS was verified by PCR amplification of the cre1 gene and resulted in colonies, which showed a QM 9414 growth phenotype.

Fungal mycelia were harvested by filtration, washed with distilled cold water, frozen and ground under liquid nitrogen. For extraction of genomic DNA, plasmid DNA and RNA, purification kits (Wizard Genomic DNA Purification Kit, PureYield Plasmid Midiprep System and SV Total RNA Isolation System, respectively, all from Promega) were used according to the manufacturer's protocol. Standard methods were used for electrophoresis, blotting and hybridization of nucleic acids. The microarray data and the related protocols are available at the GEO web site (http://www.ncbi.nlm.nih.gov/geo/) under accession number: GSE21072. Briefly, the fungal RNAs of each experiment were reverse-transcribed and labelled with Cy3 or Cy5 dye using the indirect labelling procedure and dye-switch on two biological replicates. We then hybridized 1 μg of labelled cDNA with the 1 × 244k T. reesei DNA chip manufactured by Agilent and designed using the Teloenn software as described previously 48 . Arrays were read using a GenePix 4000B scanner (Molecular Devices) and signals analysed by the GenePix Pro 6.1 software. Spots when the "align blocks" algorithm was not able to locate features on the slide were flagged "not found". A spot was labelled as "detectable" when the raw mean intensities were above the background. Data pretreatment was applied on each result file to discard GenePix flag and saturating spots. The data were normalized without background subtraction by the global Lowess method performed with the Goulphar software 49 . The background threshold was calculated by adding two standard deviations to the average intensity of all the "not found" features. For each probe the log2 hybridization ratio was linked to genome annotation coming from the JGI website. The final log2 ratio for each transcript was obtained by averaging the "detectable" hybridization values from all probes located inside the coding sequence on the matching strand. Transcripts with no or only one probe marked as "detectable" were discarded from further analysis. For the two biological replicates on each four experiments we applied on the pretreated results the linear modeling approach implemented by lmFit (using each replicate as independent variable) and the empirical Bayes statistics implemented by eBayes both from the limma R package 20 . We selected the list of statistically regulated genes with an adjusted p-value, using the Benjamini-Hochberg multiple test correction, lower than 0.05. Finally we kept as the most highly regulated targets only transcripts with a final log2 hybridization ratio greater than 2 or lower than -2.

DNase treated (DNase I, RNase free; Fermentas) RNA (5 μg), obtained from T. reesei QM 9414 and the delta-cre1 strain grown in chemostat cultures (vide supra), 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 RT-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 S7. All assays were carried out in 96-well plates, which were covered with optical tape. 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 57°C) and elongation (10 sec at 72°C). Determination of the PCR efficiency was performed using triplicate reactions from a dilution series of cDNA (1, 0.1, 10^-2 10^-3). Amplification efficiency was then calculated from the given slopes in the IQ5 Optical system Software v2.0. Expression ratios were calculated using REST^© Software 50 .

Growth of T. reesei QM 9414 and two Δcre1 independent transformed clones from it on 95 carbon sources was investigated using Biolog^® phenotype microarrays essentially as described by Druzhinina et al. 51 . Biomass concentration (OD750) was measured after 96 h of growth. One-way or main-effect analyses of variance (ANOVAs) were used to compare the growth of selected strains on individual carbon sources. Tukey's honest significant difference (Tukey HSD) test as implemented in STATISTICA 6.1 was used for post hoc comparisons to detect the contribution of each variable to the main effect of the F test resulting from the ANOVA. The significance of increased growth on selected carbon sources was tested by the Chi-square test. Only p-values < 0.05 were considered as significant.

Clustering analysis was done using the MultiExperiment Viewer software 52 . An expression matrix was built from all the genes sorted as strongly regulated for the four conditions we hybridized. From this matrix we selected the expression profiles corresponding to the main behaviour categories we expected which are: CRE1 repressed genes at high growth rate (GR), CRE1 induced genes at high GR, genes induced by CRE1 at low GR, genes repressed by CRE1 at low GR, genes repressed or induced by GR alone, CRE1 repressed genes alone and CRE1 induced genes alone. Then each of these gene expression groups was extended using the Pavlidis Template Matching method 53 fixating the absolute R threshold when all gene included in the group used as template are retrieved. Genes not classified in one of the groups where clustered using the Clustering Affinity Search Technique 54 procedure using the Pearson correlation coefficient and a 0.8 threshold. From the clusters obtained, two of them could be associated to already detected groups. Finally a hierarchical clustering was performed on each expression groups using Euclidian distance and the average linkage method. From this clustering each gene expression group was separated in two subclusters using the two more external nodes of each tree, resulting in clusters A to H and X.

Genes were first identified according to their ID number. In the case of genes that were either poorly or not yet annotated, they were subjected to BLASTX at NCBI, and all hits with E < -50 retrieved and aligned. Proteins with >75% identity and over >90% of the amino acid sequence in other fungal taxa were considered to be their potential orthologues, and this hypothesis tested by phylogenetic analysis. Proteins that fell outside these criteria were termed "unknown proteins". Proteins with < 25% of identity and less then 75% of the size of the best hits were considered "unique". Identified proteins were categorized according to the Functional Catalogue (FunCat 55 ) using the T. reesei genome database implemented in the Pedant server of the Munich Information Center for Protein Sequences 56 . For each of the FunCat categories selected we calculated for each gene cluster the enrichment ratio compared to the whole FunCat annotation and we used the hypergeometric distribution to compute the statistical p-value associated to this enrichment.

We performed a whole enrichment analysis for all the 171 annotated genes (among 250) detected as highly regulated targets using the Gene Ontology annotation of T. reesei from the JGI web site. We calculated the significance of the enrichment ratio using the hypergeometric distribution and the p-value where adjusted for multiple test using the Benjamini-Hochberg FDR correction. There are 4,977 genes annotated with at least one Gene Ontology term among the 9,129 genes from the T. reesei genome.

All analyses were performed with the RSAT software suite 24 . Promoter sequences were obtained using the "retrieve sequence" algorithm, with -1000 to -1 coordinates input and the "noORF" option unchecked. Motifs were searched using the "DNA Pattern Matching" algorithm, with the "prevent overlapping matches" parameter checked. For a given set of genes (any cluster, or whole genome), the total number of motifs found was collected and an average number of sites per gene was calculated. Finally this number was normalized to the average number of sites in the whole genome. Significance in mean values between cluster and whole genomes groups were assessed using a Student t-test with Microsoft Excel "ttest" function, using bidirectional and heteroscedastic (unequal variances) options. P-values of 0.1 and 0.05 were considered (and indicated in Table 1). For search of new regulation motifs, the "oligo analysis" and "Dyad Analysis" algorithms were used for each cluster or group of clusters. To ensure the selection of statistically relevant motifs, for each condition tested (Dyad, 5, 6, 7 or 8 words length), three different random gene groups of the same size were assessed in parallel (using the "random gene selection" algorithm). Only motifs with scores higher than the highest score obtained with the corresponding control set +0.5 were considered. This usually resulted in a 1.5 score selection threshold. Selected motifs were validated with enrichment in clusters and statistical tests described above.