Systems Analysis of Lactose Metabolism in Trichoderma reesei Identifies a Lactose Permease That Is Essential for Cellulase Induction

Trichoderma reesei colonizes predecayed wood in nature and metabolizes cellulose and hemicellulose from the plant biomass. The respective enzymes are industrially produced for application in the biofuel and biorefinery industry. However, these enzymes are also induced in the presence of lactose (1,4-0-ß-d-galactopyranosyl-d-glucose), a waste product from cheese manufacture or whey processing industries. In fact, lactose is the only soluble carbon source that induces these enzymes in T. reesei on an industrial level but the reason for this unique phenomenon is not understood. To answer this question, we used systems analysis of the T. reesei transcriptome during utilization of lactose. We found that the respective CAZome encoded all glycosyl hydrolases necessary for cellulose degradation and particularly for the attack of monocotyledon xyloglucan, from which ß-galactosides could be released that may act as the inducers of T. reesei’s cellulases and hemicellulases. In addition, lactose also induces a high number of putative transporters of the major facilitator superfamily. Deletion of fourteen of them identified one gene that is essential for lactose utilization and lactose uptake, and for cellulase induction by lactose (but not sophorose) in pregrown mycelia of T. reesei. These data shed new light on the mechanism by which T. reesei metabolizes lactose and offers strategies for its improvement. They also illuminate the key role of ß-D-galactosides in habitat specificity of this fungus.


Introduction
The b-(1,4)-linked glucose polymer cellulose and hemicellulose polysaccharides of varying composition make up 60-80% of the plant cell wall and arise from the utilization of solar energy and carbon dioxide by plants at an annual production rate of about 7.2 and 6610 10 tons, respectively [1]. These polymers therefore represent the global reservoir of renewable carbon for the production of biofuels and platform chemicals used in biorefineries.
The fungus Trichoderma reesei (the anamorph of the tropical ascomycete Hypocrea jecorina) serves as the major producer organism for the enzymes needed to degrade the above mentioned polymers to soluble monosaccharides [2]. Since most of these enzymes are formed only adaptively, the cultivations must be performed in the presence of an inducer which is, in most cases, a cellulose and hemicellulose containing waste material. Interestingly, lactose (1,4-0-ß-D-galactopyranosyl-D-glucose), which is produced to around 1.2 million tons per annum worldwide as a by-product from cheese manufacture or from whey processing industries, also induces cellulase formation in T. reesei (but not in other fungi) and is the only economic soluble carbon source for this purpose [3]. On the other side, lactose metabolism is slow and leads to lower cellulase yields than on cellulose, thus warranting a deeper understanding of lactose metabolism towards its targeted improvement [4,5].
In this study, we exploit functional genomics resources to perform a systems analysis of the T. reesei transcriptome during utilization of lactose and formation of cellulases and hemicellulases. We have produced strains containing deletions in genes encoding proteins identified from the transcriptome datasets and evaluated them for their ability to use lactose and produce cellulases. From this analysis, we identified a protein whose deletion completely impairs lactose utilization and cellulase gene expression in T. reesei. These data shed new light on the mechanism by which lactose induces cellulase formation in T. reesei, and illuminate the key role of ß-D-galactosides in habitat specificity of this fungus.

Results
The Lactose-regulated Transcriptome of T. reesei In order to identify genes that are induced or repressed by lactose, we determined the transcriptional profiles of T. reesei QM 9414 during the initial growth phase (when 25-30% of the carbon source have been consumed) on lactose, glucose and glycerol, and monitored those that were .2-fold different at a p,0.05 between lactose and either of the two other sugars. In order to remove those genes that would be specific to either glucose or glycerol, but not lactose, we then selected those genes that were present in both comparisons. These comprised a total of 1540 genes ( = 16.8% of the total genome), of which 899 (58.4%) and 641 (41.6%) were upand downregulated, respectively. Of these 1540 genes, 523 upand 367 downregulated genes could either be identified or at least attributed to a function (Table 1; Table S1). It was found that 376 and 274 genes, respectively, encoded unknown or orphan, lineagespecific proteins. Also, 127 (14.1%) and 59 (9.2%) of the up-and downregulated genes encoded genes containing an N-terminal signal sequence and therefore putatively comprise secreted and/or membrane proteins.
We employed two strategies to identify genes that are overrepresented in the lactose-specific transcriptome (see Materials and Methods): after grouping genes according to FunCat categories, we first calculated the ratio of the number of up-vs downregulated genes belonging to each category and tested which of them was beyond the average (p,0.05). In addition, we expressed the number of transcripts that belong to individual Funcat groups as the percentage of all genes in the .2-fold transcriptome, and divided this value by the percentage of the total number of genes of this Funcat group in the 9143 genes of the T. reesei genome. These ''enrichment'' values were then again tested for those lying beyond the mean variation (p,0.05). Genes involved in solute transport (of the major facilitator superfamily; ''MFS permeases'' and of amino acids), and ''CAZymes'' ( = enzymes acting on extracellular polysaccharides) were enriched in lactose-grown cultures, and genes required for proteasome and mitochondrial function were strongly downregulated ( Table 2). In addition, genes that only passed the first criterion were genes encoding GH-auxiliary proteins, G-protein coupled receptors and transcription factors (which all contained more up-than downregulated members), and genes for ribosomal functions (which were mostly downregulated).

Metabolic Characteristics of the Lactose Transcriptome
When the CAZyme encoding genes were not considered, 156 and 120 genes of the up-and downregulated transcriptome, respectively, comprised genes encoding enzymes known to be involved in metabolism. Of these genes, 106 and 92 could be attributed to specific metabolic pathways. Genes encoding enzymes for oxidative reactions and amino acid metabolism accounted for the highest number, but no significant differences were noted between the up-and downregulated genes.
Growth on lactose proceeds slower than on glucose and glycerol, and we therefore tested whether some of the genes specifically expressed on lactose could be subject to regulation by the relationship between growth rate and repression by the carbon catabolite repressor CRE1 as described by Portnoy et al. [9]. We have therefore investigated whether genes found by these authors would overlap with those upregulated on lactose. Indeed, we found 93 genes to be shared between these two conditions (Table S2). However, they were distributed between all nine groups found by Portnoy et al. [9], and also at about equal numbers. Consequently, none of these groups is particularly enriched and we therefore conclude that lactose does not create conditions for carbon catabolite repression or growth rate-dependent derepression.
Most of the transcription factors that were upregulated on lactose comprised members of the fungal-specific Zn(2)Cys(6) cluster family (58 genes), almost all of which have not yet been characterized in fungi. However, the genes encoding the general cellulase regulator XYR1 (Trire2:122208) and the recently described N. crassa cellulase-regulator CLR-2 (Trire2:60282; [10]) were found to be significantly upregulated on lactose (22-and 7-fold vs. glucose and 10 and 5-fold vs. glycerol, respectively).
Interestingly, eight genes that were significantly upregulated during growth on lactose comprised genes involved in sexual differentiation, including the mating type gene mat1-2-1, the a-type pheromone precursor hpp1, two pheromone receptors and genes involved in synthesis and processing of the pheromone precursor ( Table 4).

The Lactose-induced MFS Permeases Comprise Members Involved in Transport of Hemicelluloses Monomers
As noted above, MFS permeases comprised one of the largest gene groups upregulated on lactose ( Table 2; Table S1). The function of most of them was unknown. However, BLAST analysis of their amino acid sequences resulted in putative monoor disaccharide transporters of other fungi as next neighbours. One of them (Trire2:104072) has recently been described to enable S. cerevisiae to take up D-xylose [11]. To learn more about their function, we prepared deletion strains of the 14 most upregulated MFS permeases, and screened them for their ability to grow on glucose, lactose, D-galactose, and Trire2:104072 also on L-arabinose and D-xylose ( Figure 1). The most severe effect was observed with Trire2:3405, whose deletion completely blocked the ability of the strain to grow on lactose containing medium, but not on D-glucose or D-galactose ( Figure 2a). A phylogenetic analysis shows that Trire2:3405 is a member of a sister clade of a larger cluster containing the K. lactis and A. nidulans lactose permeases (Figure 3), and other so far uncharacterized orthologues from other Pezizomycota. All the other deletion mutants did not display any clear phenotype, and their identity therefore remains obscure. Interestingly, none of the mutants exhibited decreased growth on cellobiose, not even in the presence of nojirimycin (to inhibit the extracellular ß-glucosidases which could compensate the defect of a putative cellodextrin transporter; Figure S1). The expression of the MFS permeases was verified by qPCR analysis, which confirmed upregulation on lactose after 24 h of growth (Table S3).

The Putative Lactose Permease is Essential for Cellulase Induction by Lactose
Having identified a putative lactose transporter of T. reesei, we were also interested whether this transporter would be relevant for cellulase production on lactose. To this end, we cultivated the deletion strain in Trire2:3405 on glycerol and then transferred it to lactose and recorded the formation of cellulases and the accumulation of cellulase transcripts (Figure 2 b-e). The D3405 strain did not grow after transfer to lactose, and also did not take it up; cellulase activities were reduced to zero and the levels of the cel6A and cel7A transcripts eliminated, as determined with qPCR. Placing Trire2:3405 under the strong and constitutive expression signals of the tef1 promoter [12] resulted in an earlier appearance of the cel6A and cel7A transcripts (after 16 h of growth on 1% lactose) compared to the parental strain, and this finding was also reflected in the appearance of these two enzymes at 16 h in the culture filtrate but not in the parental strain, as shown by SDS-PAGE analysis ( Figure 2e).
We also tested whether induction by sophorose, a strong inducer of cellulase formation in T. reesei (8), would also be affected by the knock out in the lactose transporter. However, pregrowth on glycerol and subsequent transfer to 1 mM sophorose revealed the same abundance of cel7A and cel6A transcripts in QM 9414 and D3405 after 4 and 6 hrs of incubation (data not shown). We therefore conclude that the lactose permease Trire2:3405 is specifically involved in cellulase induction by lactose.  Consistent with recent findings that induction of the cellulase regulator XYR1 does not require metabolism of D-galactose (14), expression of xyr1 remained unaffected in the D3405 strain. Similarily, formation of the extracellular ß-galactosidase BGA1 was unaffected in strain 3405 (Figure 2 f ).

Discussion
In their natural habitat, fungi are unlikely to encounter the disaccharide lactose frequently as a carbon source. However, T. reesei is able to grow on it and simultaneously secretes cellulases. The present data offer some explanations for this enigma: besides a complete cellulase system, the upregulated CAZome also includes the cellulase accessory proteins CIP1, swollenin and two of the newly detected GH61 cellulose monooxygenases. Further, it comprises about half of the hemicellulases (xylanases, aand ß-D-galactosidases, a-Larabinofuranosidases and cellulose/hemicellulose deacetylases), three of four a-L-fucosidases and all enzymes acting on D-glucuronoyl-side chains (GH30 glucuronoyl xylanases, and GH67/GH79 a-Dglucuronidases). An indirect stimulation of expression of these genes (e.g. by a lower growth rate and/or relieve from carbon catabolite repression) appears unlikely when the respective transcriptomes displayed under these conditions are compared (vide supra).A plausible explanation therefore is that lactose mimics substance(s) present in the natural habitat of T. reesei (i.e. predegraded and decaying wood) that signal the availability of plant biomass tothe fungus. This physiological signal is likely another ß-D-galactoside, as some of them have been shown to be able to induce cellulases [13]. Such ß-D-galactooligosaccharides would typically occur in xyloglucans of the XXGG type [14], in which xylose residues can be substituted with a-1,2-Lfucopyranose-ß-1,2-D-galactopyranose and a-1,2-L-galactopyranoseß-1,2-D-galactopyranose disaccharides and O-linked acetyl groups, side chains for which the respective hydrolytic enzymes are all induced in T. reesei by lactose. Further, the xyloglucans of dicotyledons are partially replaced by glucuronoarabinoxylan, which has a linear ß-1,4linked D-xylopyranosyl backbone with glucuronosyl or 4-O-methyl glucuronosyl side chains, which are again structures for which T. reesei enzymes are induced. Xyloglucans are the hemicelluloses in the primary cell wall of dicotyledons and are strongly associated with cellulose by crosslinking cellulose microfibrils [15]. Our data leads to speculate that the xyloglucans are the initial target for T. reesei when feeding on plant biomass.
The Zn(2)Cys (6) transcriptional regulator XYR1 has been demonstrated to be responsible for the induction of cellulases, xylanases and also some of the a-L-arabinofuranosidases in T. reesei   [12,16], but it is not known whether this extends to all of the CAZyme genes upregulated on lactose. In this context, it is interesting that the orthologue of the recently described N. crassa cellulase regulator CLR-2 [10] was found to be upregulated by lactose. Since in vivo and in vitro studies have shown that only the binding sites for XYR1 and the HAP2/3/5 complex are in the cel6A promoter are occupied and are thus necessary for transcription of the cellulases in T. reesei [17], and a xyr1 knock out is unable to induce cellulase gene transcription [16], it is unclear how CLR-2 could contribute to cellulase induction. It will be interesting to investigate whether it is responsible for the expression of some of the hemicellulase genes for which the regulator is as yet unknown. Lactose also induces the genes for a number of oxidative enzymes such as a GMC oxidoreductase, multicopper oxidase, tyrosinase, and FAD-dependent oxidases and monoxygenases. The majority of these proteins were predicted to possess a signal peptide (by SignalP; p,0.05; [18]) and are thus putative secreted proteins. Their induction by lactose is enigmatic, because no oxidation products of lactose or D-galactose are present in the medium. On the other hand, these enzymes are considered to be involved in lignin degradation by white rot fungi [19], and aid in the attack of brown rot fungi on cellulose by Fenton chemistry [20]. We consider it possible that lactose, as a trigger for formation of the plant biomass decomposing machinery of T. reesei, also induces enzymes that can aid the action of the CAZymes by Fenton-type like reactions. Although the operation of Fenton-type chemistry in cellulose degradation has so far only been shown in brown-rot fungi, recent data in N. crassa also pointed in this direction [21]. The demonstration of Fenton-type chemistry in T. reesei would be an important step in further understanding the superiority of its secretome in cellulose hydrolysis.
The number of genes -1540 -that are significantly regulated during growth on lactose is higher than that recorded in previous studies on carbon catabolite repression or conidiation [9,22], illustrating that growth on lactose involves a major physiological shift in T. reesei. While one would expect changes in genes associated with metabolism of lactose, our data show that other areas of metabolism (such as amino acid and lipid metabolism or metal ion uptake), mitochondrial functions and components of the proteasome are also altered. At this time, the physiological consequences of these changes are unknown; however they do not seem to be related to the different growth rate on lactose in comparison to glucose and glycerol, as only a few genes were shared between the present study and that by Portnoy et al. who used chemostat cultures at different growth rates [9]. Interestingly, secondary metabolism -which was reported to correlate with cellulase and hemicellulase gene upregulation [23] -was almost not affected by lactose: only two nonribosomal peptide synthases (Trire2:60751 and Trire2:67189) and none of the polyketide synthases at all were upregulated.
The lactose upregulated transcriptome also included a vast number of putative transporters of the major facilitator superfamily. The fact that gene deletion could only identify the function of one of them is likely due to a redundancy of their function. However, one of them turned out to be essential for lactose utilization. Interestingly, this transporter is not a member of the major fungal lactose permease clade identified recently ( [24]; cf. Figure 3). The fact that its knock-out impairs triggering of cellulase gene expression by lactose also in resting cells, which do not require lactose for growth, illustrates that either the uptake or the intracellular presence and/or metabolism of lactose is essential for cellulase induction. These findings reject our previous assumption that lactose metabolism in T. reesei proceeds only via extracellular hydrolysis and subsequent uptake and metabolism of the monomers D-glucose and D-galactose [25,26]. This theory was based on the findings that the genome inventory of T. reesei appears not to contain a gene encoding an intracellular ß-galactosidase. Intracellular ß-galactosidases belong to GH family 2. While the genome of T. reesei contains seven members of this group [6], and four have also been found in this study to be induced on lactose, a Blastp search identified all of them as ß-mannosidases (five), exo-ß-D-glucosaminidase (one) and ß-glucuronidase (one) and none as a ß-galactosidase. While we cannot rule out that one of them (e.g. the ß-glucuronidase) has also a ß-galactosidase activity, it is also possible that T. reesei has an intracellular ß-galactosidase that does not belonging to family 2. Identification of the respective enzyme would provide essential information for genetic manipulation of lactose metabolism and thus cellulase formation, and for the understanding of the utilization of this intriguing disaccharide in T. reesei.

Strains, Cultivations and Measurement of Growth, Lactose Consumption and Cellulase Formation
T. reesei QM9414 (ATCC 26921), an early cellulase producing mutant was used throughout this work and kept on potato dextrose agar (Sigma, St. Louis, MO) at 28uC. T. reesei Dtku70 strain [27], an uridine auxotrophic strain lacking the tku70 gene required for non-homologous end joining DNA-repair, was maintained on malt extract medium supplemented with uridine (10 mM).
Growth on solid medium was tested by inoculating agar plates with a small agar piece (5 mm diameter). The biomass in submerged culture was determined by filtering portions of culture onto Whatman no. 1 preweighted filter papers. The harvested biomass was dried at 80uC for 3 days, and then weighted.
SDS-PAGE was performed as described by Ausubel et al. [31] using 10% polyacrylamide gels. To this end, the proteins from the culture supernatant were precipitated by the addition of 2 vol EtOH andthen and dissolved in one tenth of the original volume of SDS sample buffer. Gels were stained with Coomassie Brilliant Blue G250.

Vector Construction and Generation of Gene Deletion Mutants
Deletion cassettes consisting of 1.-to 1.5 kb fragments of the genespecific flanking regions interrupted by the T. reesei pyr4 marker gene (encoding orotidine-59-monophosphate decarboxylase) were assembled by yeast recombinational cloning [32]. Oligonucleotides (10 mM) 5F +5R and 3F +3R were used for amplification of the individual flanking regions from genomic T. reesei DNA using Taq Polymerase (Promega). By PCR approximately 19 bp were introduced at each flanking end that overlap with the pRS426 (URA +) yeast shuttle vector or the pyr4 gene to allow homologous recombination. A 3.2 kb fragment of the T. reesei pyr4 marker gene was amplified with oligonucleotides Pyr4 fw and Pyr4 rev. A PCR touchdown program ranging from 62uC to 58uC for annealing was used for amplification. Oligonucleotide sequences are shown in Table S4. Yeast transformation was performed as described [33] using the lithium chloride/polyethylene glycol procedure. Yeast strain WW-YH10 (ATCC 208405) was transformed with both flanking regions, the pyr4 marker gene and an EcoRI/XbaI digested plasmid pRS426 [34]. Transformants were selected on synthetic complete dropout media (SC-URA with uracil dropped out). Following total DNA isolation from liquid SC-URA media [35], plasmids were introduced into chemically competent JM109 E. coli cells (Promega). The outside primers 5F +3R were used to synthesize the complete deletion cassette from S. cerevisiae.
For construction of a strain overexpressing Trire2:3405 under the constitutive tef1 promoter, the coding region of Trire2:3405 plus approximately 500 bp downstream of it were amplified using primer st_infusion_1 and st_infusion_2 (Table S4) and inserted into the SalI/HindIII linearized plasmid pLHhph1-tef1 [12] using the In-FusionH HD Cloning Kit (Clontech). Transformation of T. reesei protoplasts was performed as described by Gruber et al. [36]. Integration of the ptef1::3405 fragment was verified using primer Ptef F-ch and St R-ch (Table S4).

Transformation of T. reesei and Analysis of Transformants
All deletion strains were generated in the Dtku70 strain [27] and the QM 9414 strain (ATCC 26921) was used as a control in all experiments. Protoplast preparation and transformation were performed as previously described [36]. The deletion cassettes were purified from agarose gels (QIAquick Gel Extraction kit, QIAGEN) and concentrations were determined (Nanodrop Spectrophotometer, Peqlab). After transformation protoplasts were stabilized and regen-erated on minimal mediumsupplemented with 1 M D-Sorbitol. In the case of the QM9414 tef1::3405 strain the selection media additionally contained 100 mg/mL hygromycin B (Roth). For sporulation, transformants were transferred to small plates and purified by plating conidiospores onto plates with 0.1% Triton X-100 as colony restrictor. Single colonies were transferred to selective media and screened for correct integration of the deletion cassettes or of the tef1::3405 fragment, respectively. Genomic DNA of the transformants was extracted [37] and transformants were screened for the presence of the deletion cassettes by amplifying a fragment by PCR with one oligonucleotide specific for the upstream region outside of the deletion cassette (primer abbreviated with ''ch'' in Table S4) and the second specific for the pyr4 marker gene (ch_pyr4 neu).

Transcriptome Analysis
Mycelia were harvested from cultures growing on lactose, glucose and glycerol, respectively, for 24 hrs. Total RNAs were extracted using TRIzolH reagent (Invitrogen Life Technologies, Carlsbad, CA, USA), according to the manufacturer's instructions, and then purified. cDNA synthesis, labelling and hybridization was performed by Roche NimbleGen (Roche-NimbleGen, Inc., Madison, WI, USA) with a high density oligonucleotide (HDO) microarray using 60-mer probes representing the 9.129 genes of T. reesei. Microarray scanning, data acquisition and identification of probe sets showing a significant difference (p = 0.05) in expression level between the different conditions were performed essentially as described by Metz et al. [22]. Gene accession numbers were annotated according to version 2 of the T. reesei genome assembly (http://genome.jgi-psf.org/Trire2/Trire2.home.html), and ambiguous cases annotated manually.
Genes were then classified according to their major annotation in the MIPS Functional Catalogue [38]. To determine whether there were differences in the functional categories in each cluster, the distribution within each cluster was compared to the total distribution of all the annotated genes using independent chisquare tests. The microarray data and the related protocols are available at the GEO web site (www.ncbi.nlm.nih.gov/geo/) under accession number GSE39276.

Real Time PCR
DNase treated (DNase I, RNase free; Fermentas) RNA (5 mg) was reverse transcribed with the RevertAid TM First Strand cDNA Kit (Fermentas) according to the manufacturer's protocol with a combination (1:1) of the provided oligo-dT and random hexamer primers. All assays were carried out in 96-well plates which were covered with optical tape, as described by Metz et al. [22]. Primers, amplification efficiency and R-square values are given in Table S5. Determination of the PCR efficiency was performed using triplicate reactions from a dilution series of cDNA, and the amplification efficiency was then calculated from the given slopes in the IQ5 Optical system Software v2.0. Expression ratios were calculated using RESTß Software. All samples were analyzed in two independent experiments with three replicates in each run.

Phylogenetic Analysis
Phylogenomic relationship between Trire2:3405 and its closest neighbors in BLAST (blastP) was studied by performing a randomized bootstrap maximum-likelihood analysis using RAxML software [39], setting the bootstrap analysis to 1000 runs and the bootstrap random seed value to 12,311. The Dayhoff mutation data matrix was used for the analysis of the alignment.