The transcriptomic fingerprint of glucoamylase over-expression in Aspergillus niger

Background Filamentous fungi such as Aspergillus niger are well known for their exceptionally high capacity for secretion of proteins, organic acids, and secondary metabolites and they are therefore used in biotechnology as versatile microbial production platforms. However, system-wide insights into their metabolic and secretory capacities are sparse and rational strain improvement approaches are therefore limited. In order to gain a genome-wide view on the transcriptional regulation of the protein secretory pathway of A. niger, we investigated the transcriptome of A. niger when it was forced to overexpression the glaA gene (encoding glucoamylase, GlaA) and secrete GlaA to high level. Results An A. niger wild-type strain and a GlaA over-expressing strain, containing multiple copies of the glaA gene, were cultivated under maltose-limited chemostat conditions (specific growth rate 0.1 h-1). Elevated glaA mRNA and extracellular GlaA levels in the over-expressing strain were accompanied by elevated transcript levels from 772 genes and lowered transcript levels from 815 genes when compared to the wild-type strain. Using GO term enrichment analysis, four higher-order categories were identified in the up-regulated gene set: i) endoplasmic reticulum (ER) membrane translocation, ii) protein glycosylation, iii) vesicle transport, and iv) ion homeostasis. Among these, about 130 genes had predicted functions for the passage of proteins through the ER and those genes included target genes of the HacA transcription factor that mediates the unfolded protein response (UPR), e.g. bipA, clxA, prpA, tigA and pdiA. In order to identify those genes that are important for high-level secretion of proteins by A. niger, we compared the transcriptome of the GlaA overexpression strain of A. niger with six other relevant transcriptomes of A. niger. Overall, 40 genes were found to have either elevated (from 36 genes) or lowered (from 4 genes) transcript levels under all conditions that were examined, thus defining the core set of genes important for ensuring high protein traffic through the secretory pathway. Conclusion We have defined the A. niger genes that respond to elevated secretion of GlaA and, furthermore, we have defined a core set of genes that appear to be involved more generally in the intensified traffic of proteins through the secretory pathway of A. niger. The consistent up-regulation of a gene encoding the acetyl-coenzyme A transporter suggests a possible role for transient acetylation to ensure correct folding of secreted proteins.


Introduction
Due to its well annotated genome sequence, newly established gene transfer systems, and the availability of high-quality tools for obtaining and evaluating transcriptomic and proteomic data, Aspergillus niger has become a model fungus for industrially exploited filamentous fungi (Meyer 2008;Meyer et al. 2007a;Meyer et al. 2011b;Pel et al. 2007). Its impressive natural capacity to secrete high amounts of hydrolytic proteins into the environment combined with its ability to synthesize and secrete various organic acids makes it highly suitable for the production of various food ingredients, pharmaceuticals and industrial enzymes (Fleissner and Dersch 2010;Meyer 2008;Meyer et al. 2011b). As it is also capable of efficiently degrading plant-derived polysaccharides such as starch, cellulose, hemicellulose, pectin and inulin, the biotechnological importance of A. niger will probably rise even more in the near future. For example, A. niger-derived (hemi)cellulases might be used to improve the efficiency of the saccharification process of second-generation feedstock used for bioethanol production (de Souza et al. 2011;Pel et al. 2007).
To analyze and eventually control the secretory capabilities of A. niger, several attempts have been undertaken to identify the key players and regulatory mechanisms involved in protein secretion. For example, galacturonic acid, xylose and maltose were shown to induce expression of secretory proteins, including pectinolytic, (hemi)cellulolytic and glucan-hydrolyzing enzymes, respectively, whereas sorbitol acts as a repressing carbon source (Adav et al. 2010;Braaksma et al. 2010;de Oliveira et al. 2011;Lu et al. 2010). Xylose is the main inducer of XlnR, the master transcription factor that regulates expression of all major enzymes involved in the degradation of (hemi)cellulose (de Oliveira et al. 2011;Lu et al. 2010;Mach-Aigner et al. 2012). Among those, endoxylanase (XynB) and ferulic acid esterase (FaeA) are the most abundant, secreted proteins of A. niger (Lu et al. 2010). When starch or maltose are used as carbon source, the synthesis of amylolytic enzymes is induced, a step that is mediated by the transcription factor AmyR. The most abundant enzyme secreted under these conditions is a glucan 1,4-α-glucosidase (glucoamylase, GlaA), which is an exo-enzyme that releases glucose from the non-reducing end of starch or maltose and accounts for more than 50% of the extracellular proteome (Lu et al. 2010). Induction of extracellular hydrolytic enzymes is thus mainly regulated at the transcriptional level in A. niger, and either repressed by carbon catabolite repression via CreA (Yuan et al. 2006) or activated by AmyR or XlnR depending on the presence of maltose or xylose, respectively (Tsukagoshi et al. 2001).
A. niger proteins and enzymes destined for secretion into the culture medium follow the secretory pathway. The journey of secretory proteins starts with protein translocation into the lumen of the endoplasmic reticulum (ER) via a translocon that forms a channel through the ER membrane (Romisch 1999). In the ER lumen, several ER-resident chaperones and foldases, including the binding protein (BipA), protein disulfide isomerase (PdiA) and calnexin (ClxA), assist secretory proteins in proper folding (Maattanen et al. 2010). Most secretory proteins become glycosylated in the ER, both through the attachment of a conserved, pre-assembled oligosaccharide to specific asparagine residues (Nglycosylation) and by the initiation of O-glycosylation of serine and threonine residues. After proper folding and glycosylation, secretory proteins are packed into COPII-coated vesicles and transported to the Golgi complex, where protein glycosylation is completed, and subsequently through another vesicle-mediated process delivered to the cell surface where the vesicles release their cargo into the periplasmic region. High protein flux through the ER or expression of heterologous proteins can result in the accumulation of misfolded proteins. This misfolding, however, is recognized by a quality control system known as ERassociated degradation (ERAD). The aim of ERAD is to direct misfolded proteins to the cytosol where they become degraded by the proteasome. In addition, another cellular protein quality system, the unfolded protein response (UPR), is induced, which aims at proper refolding of misfolded proteins via induced expression of chaperones and foldases (Fleissner and Dersch 2010;Geysens et al. 2009). For this purpose, the UPR transcription factor HacA becomes activated via splicing of an unconventional 20-nt intron out of the hacA mRNA. This subsequently facilitates translation of hacA mRNA and formation of HacA, which in turn induces transcription of a number of UPR target genes including bipA, which encodes the major ER chaperone protein and pdiA, which encodes for protein disulfide isomerase (Guillemette et al. 2007;Mulder et al. 2004). Hence, both ERAD and UPR are crucial for effective functioning of the secretory pathway -not only in filamentous fungi but also in yeasts and mammals (Kaufman 1999;Saloheimo and Pakula 2012;Saloheimo et al. 2003;Yoshida et al. 2001).
Interestingly, the UPR can be viewed as a general response of A. niger, which becomes activated when the secretion machinery becomes challenged by metabolic changes or ER stress conditions. For example, transcript levels of UPR genes become strongly enhanced, when A. niger is exposed to the reducing compound dithiothreitol (DTT), which blocks the formation of disulfide bridges, or to tunicamycin, which inhibits N-glycosylation, especially when forced to express heterologous proteins or when cultivated in carbon sources, which differentially induce expression and secretion of homologous proteins (Carvalho et al. 2011a;Guillemette et al. 2007;Jørgensen et al. 2009). These observations suggest that the UPR functions act as a homeostatic control mechanism that allows A. niger to flexibly adapt its protein secretion capacity to change in environmental conditions.
In this study, we investigated the transcriptomic fingerprint of A. niger when forced to overexpress and secrete a specific hydrolytic enzyme. We chose glucoamylase (GlaA) as a model enzyme, because it is a naturally highly abundant and secreted enzyme used in the food industry (Kumar and Satyanarayana 2009). For comparison, we used two strains, a wild-type strain expressing a single copy of the glaA gene, and a mutant strain expressing multiple copies of glaA. In order to reduce the number of conflicting variables, such as changes in growth rates and fluctuations in environmental conditions, maltose-limited chemostat cultures were used. Maltose was selected as carbon source to transcriptionally induce expression of the glaA gene. Physiological and transcriptomic data were collected from both wild-type and overexpressing strains and analyzed to identify the glaA-specific overexpression transcriptome. Finally, we compared these global transcriptional changes with previously published transcriptomic data related to secretion stress in A. niger. This analysis allowed us to distinguish condition-specific responses from general transcriptomic responses of A. niger that are important to overcome different triggers of secretion stress. We could identify a core set of 40 genes whose expression is key to ensure high protein fluxes through the secretory route of A. niger, independently of the cause of the secretion stress.

Growth physiology of maltose-limited chemostat cultures of A. niger
In order to identify the transcriptomic adaptations of A. niger to forced overproduction of GlaA, we compared the transcriptomes of chemostat-grown cultures of strain B36 (overproducing strain) and wild-type strain N402 (reference strain). Strain B36 was selected as GlaA overproducer, because it is reported to contain multiple copies of the glaA gene at chromosome V (Verdoes et al. 1993;Verdoes et al. 1994b).
Maltose-limited chemostat cultures were used to induce expression of the glaA gene, to control the specific growth rate and to obtain highly reproducible data due to well-defined steady state conditions. Initial chemostat experiments were conducted using three different dilution rates (D = 0.05, 0.1 and 0.15 h -1 ); however, steady state conditions for both strains were only reached at D = 0.1 h -1 . At a higher dilution rate (D = 0.15 h -1 ), N402 reached a steady state, but the B36 strain was washed out before reaching a steady state. Although the morphology of the N402 strain at the lowest dilution rate (D = 0.05 h -1 ) was similar to the higher dilution rates, and no sign of mycelial aggregation was apparent (data not shown), N402 was not able to reach a steady state at this dilution rate. Maltose-limited chemostat cultures were therefore run in triplicate at D = 0.1 h -1 for both strains. The cultures were highly reproducible and gave rise to homogenous cultures of dispersed mycelial morphologies ( Fig. 1, Table 1 and data not shown). The initial batch cultivation (duration about 30 h) was followed by approximately 50 h of continuous cultivation. After four volume changes (4 x D -1 ), the cultures reached a steady state as reflected by a constant alkali addition rate and constant CO 2 , O 2 and biomass concentrations ( Fig. 1 and data not shown). Biomass concentrations for both strains stabilized at about 4 g kg -1 with a small relative standard deviation (RSD) (approx. 0.002 for N402 and 0.02 for B36). In both strains, the respiratory quotient (RQ) was lower than 1, probably due to high production of organic acids in parallel to GlaA secretion. Notably, the RQ value calculated for B36 cultures was slightly but significantly lower than the RQ value obtained for N402 cultures ( Table 1), suggesting that protein and acid production is somewhat higher in B36 compared to N402. In agreement, the carbon concentration in culture filtrates obtained from steady state samples was also higher in B36 compared to N402 (data not shown) and the specific productivity of extracellular protein (q protein-EC ) was five-to six-fold higher in B36 than N402, demonstrating that B36 indeed secretes more protein than the wild-type strain.

Protein secretion and glucoamylase production during steady state
The GlaA overproducer strain B36 was previously estimated to contain about 80 copies of the glaA gene as inferred from Southern analysis (Verdoes et al. 1993). As multiple gene copies can cause frequent recombination in A. niger resulting in genetic instability and loss of glaA copies (Verdoes et al. 1994b), we decided to re-determine the number of glaA copies present in B36 using quantitative real time PCR (qPCR). As summarized in Table 2,  B36 contains about 32 glaA gene copies based on the fact that N402 contains only a single copy of glaA (Pel et al. 2007). We also determined glaA mRNA levels by qPCR and observed that glaA transcript levels in B36 were about seven times higher compared to N402. Consistent with this observation, under steady state conditions extracellular GlaA production in B36 was about seven-to eight-fold higher than in N402 as estimated by Western analysis and measuring glucoamylase activity (Table 2). These data confirm previous observations that the amount of GlaA produced correlates well with the amount of glaA mRNA but is not proportional to the number of glaA gene copies, a well-known phenomenon in filamentous fungi, where protein overproduction is often limited at the transcriptional level (Tsukagoshi et al. 2001;Verdoes et al. 1994b). It should be noted, however, that the increased glucoamylase production in the B36 strain is still significantly lower compared to industrial strains which produce glucoamylase up to 30 gram/liter (Finkelstein et al. 1989  changes in gene expression <2, these values are considered to be significant in view of the identical culture conditions used, the identical specific growth rates of both strains during steady state conditions, and the stringent statistical analysis of the data. A comprehensive list of all differentially expressed genes is depicted in Additional file 1. An enrichment analysis was performed to identify gene ontology (GO) terms which were over-represented in the differentially expressed gene set. We used the recently published improved GO annotation tool for A. niger (Fisher's exact test Gene Ontology annotation tool, FetGOat (Nitsche et al. 2011)), an open source tool accessible at http://www.broadinstitute.org/fetgoat/index.html. GO terms (up-or down-regulated) with FDR values < 0.05 were defined as over-represented. Overall, 129 enriched GO terms were identified among the differentially expressed gene set, 54 of which belonged to 'biological processes' (BP), 63 to 'cellular components' (CC) and 12 to 'molecular functions' (MF).
The corresponding network maps and gene lists are depicted in the Additional file 2 and 3.

Predicted up-and down-regulated biological processes inferred from the GlaA-overexpression transcriptome
In order to deduce biological information out of the GO term enrichment analysis, we focused on the BP gene list and removed redundant and less detailed annotations. Among the remaining GO terms in the up-regulated gene set, the following four higher-order categories were identified: i) translocation, ii) protein glycosylation, iii) vesicle transport and iv) ion homeostasis (see Additional file 4). The translocation category included GO terms such as 'posttranslational protein targeting to membrane', 'SRP-dependent cotranslational protein targeting to membrane', 'translocation' and 'protein targeting to ER'. The protein glycosylation category included the GO terms 'related to glycosylation', 'protein N-linked glycosylation', 'oligosaccharide biosynthetic process', 'dolichol-linked oligosaccharide biosynthetic process' and 'oligosaccharide-lipid intermediate biosynthetic process'. The vesicle transport category included the GO terms 'vesicle-mediated transport', 'COPII-coated vesicle budding', 'membrane budding', 'vesicle organization', 'vesicle coating', 'vesicle targeting (rough ER to cis-Golgi)', 'COPII vesicle coating' and 'retrograde vesicle-mediated transport (Golgi to ER)'. The last category, ion homeostasis, contained GO terms involved in iron, calcium and zinc homeostasis (e.g. 'iron homeostasis', 'inorganic cation homeostasis', 'cellular response to iron starvation' and 'ion transport', see Additional file 4).
As three out of the four major categories were related to the secretory pathway, the corresponding genes lists were examined in more detail. The expression of at least 130 predicted secretory pathway genes were changed in the GlaA-overexpressing strain B36 (Table 3). Importantly, this set of genes is causatively linked to GlaA overexpression and does not depend on the growth rate or the carbon source, because both strains were in steady state at the same specific growth rate in maltose-limited chemostat cultures. Only 16 of the 130 genes were down-regulated, indicating that the capacity of the protein secretion machinery is increased compared to the wild-type situation. Among the secretory pathwayrelated genes, the majority of the induced genes belonged to ER-related processes, including translocation into the ER, protein folding, glycosylation, ERAD, UPR, COPI-and COPII-mediated transport processes (Table 3). Notably, the gene with the most significantly increased transcript level in B36 was the bipA gene (FDR, 1.1 x 10 -7 ), which is under transcriptional control of HacA (Mulder et al. 2004), and encodes the main chaperone in the ER and is thus important for protein overproduction in A. niger (van Gemeren et al. 1997). The increased expression of bipA gene in the B36 strain was previously reported by Northern blot analysis in a shake flask culture (Punt et al. 1998). In agreement, other important HacA-dependent ER chaperones and foldases such as clxA, prpA, tigA and pdiA (Table 3) were also significantly higher expressed in B36 (Mulder et al. 2004).
As up-regulation of these genes pointed towards slow or aberrant folding of GlaA and thus induction of UPR in strain B36, we tested whether the central activation mechanism of appropriate and sufficient for A. niger, because strain B36 is better adapted to grow on starch compared to the wild-type strain. Interestingly, starch-and maltose-responsive genes such as the transcription factor AmyR and AmyR-dependent hydrolase genes (Yuan et al. 2008b) are down regulated in B36 except glaA (Table 4). As previously suggested, a possible explanation for the reduced expression could be the titration of the AmyR      Fig. 2A), we propose that down-regulation of AmyR and its target genes reflects a negative feedback mechanism similar to RESS in T. reesei.
Notably, enriched GO terms of the 'ion homeostasis' category indicated an increased demand for iron, calcium and zinc. In the context of increased fluxes of GlaA through the secretory pathway in strain B36, several explanations are conceivable. First, the mature GlaA protein contains nine cysteine residues, eight of which are involved in disulfide bridge formation (Lee and Paetzel 2011;Sorimachi et al. 1996). Hence, in B36 there is an increased demand for disulfide bond formation, which requires increased activity of protein disulfide isomerases. Increased amounts of PdiA, however, might result in sequestration of  including heme or iron-sulfur-cluster (Fe-S) proteins (Lill 2009;Stojilkovic 2005;Waldron et al. 2009). In this context, it is interesting to note that protein secretion in A. fumigatus has most recently been shown to require controlled uptake of iron. Basically, the transcription factor PrtT not only positively regulates expression of protease genes but also strengthens expression of iron uptake genes (Hagag et al. 2012). We thus compared the published expression data of iron uptake genes in A. fumigatus (wt versus ∆prtT) with the expression data of their predicted orthologs in A. niger (B36 versus N402). Twelve out of 15 ironuptake genes showed similar expression profiles (Additional file 5), i.e. their up-regulation in wt versus ∆prtT was mirrored in B36 versus N402. In agreement, expression profiles of the main iron transcription factors also matched, i.e. up-regulation/down-regulation of the activator HapX/repressor SreA in wt versus ∆prtT were comparably up-/down-regulated in B36 versus N402. Hence, both independent observations from two Aspergilli strongly indicate that proper function of the protein secretion machinery and high fluxes through the secretory pathway mandate optimal iron supply and assuring proper ion homeostasis.
GO enrichment analysis of the down-regulated gene set in B36 uncovered three major categories: i) 'carbon catabolism', ii) 'amino acid catabolism' and iii) 'response to oxidative stress' (Additional file 3). The first two categories might be causatively linked to the RESS phenomenon discussed above, i.e. increased GlaA secretion is only possible at the cost of other secreted proteins. As AmyR targets related to starch degradation are downregulated in B36 (see Table 4), processes related to polysaccharide degradation have consequently to be down-regulated as well. In a recent study, 19 proteases of A. niger were identified in the extracellular medium, when the strain was cultivated under sorbitol-, galacturonic acid-or carbon-starvation conditions (Braaksma et al. 2010). Expression of 12 of them was reduced in B36 compared to N402, two of the genes were up-regulated and expression of five genes was unaltered (Additional file 6). Hence, overall reduction in protease activity in B36 will in turn decelerate amino acid catabolic processes. An additional explanation for reduced expression of carbon and amino acid catabolic processes is that many enzymes involved in these processes are iron-dependent. Iron deficiency has been shown to trigger a metabolic response in Saccharomyces cerevisiae or Schizosaccharomyces pombe, which includes down-regulation of enzymes involved in carbon and amino acid metabolism and respiration (Lill 2009;Mercier and Labbe 2010;Philpott et al. 2012;Philpott and Protchenko 2008). Assuming that this also holds true for A. niger, reduction in catabolic processes would lower the fluxes into the citric acid cycle and into the respiratory chain, which in turn would also lower the amount of radical oxygen species produced. Hence, oxidative stress would be diminished, which in turn would result in down-regulation of oxidative stress genes as observed in B36. A slight but significantly reduced RQ in B36 compared to N402 (Table 1) supports this hypothesis. Taken together, the enriched GO term set of down-regulated genes in B36 can hypothetically be viewed as a consequence of the RESS phenomenon and reduced iron availability due to increased GlaA secretion.
Two A. niger proteins lacking an N-terminal signal peptide for entering the secretory route were recently determined to be abundantly present in the secretome of A. niger (Braaksma et al. 2010). It was thus proposed that proteins can also be exported out of the cell by the so-called non-classical secretion pathway (Braaksma et al. 2010). Interestingly, transcript levels of both protein-encoding genes were lowered in B36 versus N402 (An01g09980, fold-change B36/N402 = 0.15, FDR = 8.65 x10 -9 ; An01g00370, fold-change = 0.54, FDR = 0.0001). This could mean that an accelerated activity of the classical secretory machinery is only possible at the expense of the non-classical secretion pathway -which could be another potential example for the existence of the RESS regulatory feedback loop in A. niger.

Comparison of the GlaA overexpression transcriptome with the maltose transcriptome
In order to identify genes whose transcript levels are generally important for high-level secretion in A. niger, i.e. independent of the conditional trigger, we first compared the GlaA transcriptomic fingerprint with two other recently determined fingerprints from our laboratory. First, we compared the GlaA overexpression transcriptome with the maltose transcriptome of A. niger. The maltose transcriptome was recently obtained by comparing carbon-limited chemostat cultures supplied with different carbon sources (maltose and xylose) but with the same specific growth rate (D = 0.16 h -1 ) (Jørgensen et al. 2009). This comparison showed that the production rate of extracellular proteins was about three-fold higher in maltose-grown cultures compared to xylose (Jørgensen et al. 2009). As shown in the Venn diagram depicted in Fig. 3, 150 genes were commonly induced in both B36 versus N402 and maltose versus xylose and thus represent the most interesting genes with respect to high-level secretion in A. niger. GO enrichment analysis of this set of genes confirmed that genes related to ER import, translocation, N-glycosylation and COPII transport are of utmost importance for ensuring high protein traffic through the secretory pathway (Additional file 7).
Genes that are induced in B36 versus N402, but not in maltose versus xylose, are likely to be genes that are mainly related to GlaA overexpression per se. There are 640 such genes and GO enrichment analysis uncovered two main categories (Additional file 7), one of which is the 'ion homeostasis' category, as expected due to the high iron, calcium and zinc demand for GlaA overexpression (see former section). The second category involves genes involved in 'DNA repair and DNA replication' and could hypothetically also be linked to the iron deficiency in B36 versus N402, as many proteins involved in nucleotide excision repair and DNA replication are iron-dependent (Lill 2009). Genes that induced in maltose versus xylose but not in B36 versus N402 are likely to be genes, which are important for starch/maltose degradation but not for xylose catabolism. Indeed, the enriched GO categories of this gene set (484 genes) contained 'starch metabolic processes' and other catabolic process related to 'carbon source oxidation', 'amino acid metabolism' and 'respiration and oxidative stress' (Additional file 7). These are processes important for (or related to) high ATP generation during oxidative phosphorylation in the respiratory chain and do likely form the energetic basis for the observed higher secretory flux in maltose cultures compared to xylose cultures despite similar growth rates. There were 146 commonly repressed genes in B36 versus N402 and in maltose versus xylose, but only a small fraction of these genes could be assigned to enriched GO terms related to 'saccharide catabolic processes' (Additional file 7).

Comparison of the GlaA overexpression transcriptome with other highsecretion transcriptomes
In order to determine so far unknown HacA targets and their involvement in UPR, we have recently determined the HacA transcriptome by expressing a constitutively active form of HacA (hacA CA ) in A. niger (Carvalho et al. 2012). Using bioreactor-controlled batch cultures, we compared the genome-wide expression profiles of a strain expressing hacA CA with a wild-type strain (hacA WT ) at three different time points during the exponential growth phase. At each of the three time points the up-regulated gene set of the hacA CA cultures contained genes related to protein traffic through the secretory pathway, and to the UPR and ERAD response (Carvalho et al. 2012). When comparing these data with the GlaA-overexpression transcriptome, one might expect a considerably overlap due to the fact that overexpression of GlaA induced splicing of the hacA mRNA (Fig. 2) et al. 2007). This analysis uncovered 40 genes whose transcript levels were commonly modified under all seven secretion stress conditions: 36 genes were up-regulated, and 4 genes were down-regulated ( Table 5). The genes from this set thus probably represent genes that are crucial for coping with stress conditions that target the secretion machinery. This gene set includes ER chaperones and foldases (prpA, clxA, lhs1, pdiA, bipA, tigA), genes important for translocation of secretory proteins into the ER (sec63, sec11, sss1, spc3, sec71) and genes important for protein glycosylation and COPII-based vesicle trafficking (Table 5). Fascinatingly, a gene encoding the predicted acetyl-coenzyme A transporter (An02g13410) was consistently up-regulated under all seven conditions. In higher eukaryotes, this protein was shown to be involved in translocation of membrane-impermeable coenzyme A from the cytosol into the ER lumen, where it is used for transient acetylation of ER-based proteins, thus improving the folding efficiency of nascent secretory proteins (Jonas et al. 2010). It becomes therefore important to examine whether the predicted A. niger CoA transporter fulfills the same function. Similarly, the set of core genes included six hypothetical proteins, whose precise relevance for the secretory machinery remains to be elucidated in future studies.

Conclusions
The evidence is accumulating that the secretory machinery of A. niger is equiped with a high inherent flexibility, which enables the fungus to dynamically respond to changes in the secretory protein load. Irrespective of whether different amounts of a homologous protein or a heterologous protein have to be accommodated and escorted, or a specific step in the secretory pathway becomes blocked, A. niger follows general and specific strategies to adapt to these challenges. On the one hand, the burden of high protein loads is in general dealt with increased transcript levels of genes involved in stabilizing the secretory pathway, including chaperone-encoding and foldase-encoding genes, transport genes and UPR-and ERAD-related genes. In addition, it seems likely that for the efficient secretion of individual proteins (either overexpressed or a heterologous proteins) an additional, particular set of genes becomes up-regulated, whose expression is important to deal with  the specific requirements of the respective protein. On the other hand, A. niger attempts to avoid exceeding the maximum capacity of the secretory machinery. In doing so, the RESS regulatory system offers a flexible way to lower transcript levels of those genes whose function is less important for growth and survival under the given circumstances.

Strains and inoculums
The laboratory strain Aspergillus niger N402 (Bos et al. 1988) was used as a reference strain and B36 (Verdoes et al. 1993), which contains multiple copies of the glucoamylase gene, was used as an overproducer strain. Strains were grown on solidified Complete Medium (CM) containing 1% (w v -1 ) glucose, 0.1% (w v -1 ) casamino acids and 0.5% (w v -1 ) yeast extract in addition to Minimal Medium (MM) (Alic et al. 1991). Spore plates were incubated for four or five days at 30 o C and stored for no more than three months at 4 o C. Conidia were harvested from CM agar plates with a sterile detergent solution containing 0.05% (w/v) Tween 80 and 0.9% (w/v) NaCl.

Bioreactor cultivation conditions
Maltose-limited batch and chemostat cultivations were performed as described previously for A. niger (Jørgensen et al. 2009) with slight modifications. The N402 and B36 strains were both grown in triplicate in maltose-limited chemostat cultures.

(i) Batch cultures
Batch cultivation was initiated by inoculation of 5 kg ammonium-based minimal medium with a conidial suspension to give 10 9 conidia L -1 . Maltose was heat sterilized separately from the MM and the final concentration was 0.8% (w/v). Germination was induced by addition of 0.003% (w/w) yeast extract. During the first six hours of cultivation the culture was aerated (air flow = 1 L min -1 ) through the headspace of the reactor and the stirrer speed was kept low at 250 rpm to avoid loss of the hydrophobic conidia. After six hours when most conidia had germinated, air was sparged into the culture broth, mixing was intensified (750 rpm) for more efficient oxygen transfer, and 0.01% (v/v) of polypropylene glycol (PPG) was added as an antifoaming agent. The temperature was 30°C and the pH was kept constant at pH = 3 by computer-controlled addition of 2 M NaOH or 1 M HCl. Acidification of the culture broth was used as an indirect growth measurement (Iversen et al. 1994). Submerged cultivation was performed with 6.6 L BioFlo3000 bioreactors (New Brunswick Scientific, NJ, USA).

(ii) Chemostat cultures
Continuous cultivation was started in the late-exponential growth phase, when 90 mmol of NaOH had been added to the batch culture (75% of maltose had been consumed, at a biomass concentration of about 3.5 g dry weight per kg of culture). The dilution rate (D) was set at 0.1 h -1 . Steady-states, where the specific growth rate (µ) is equal to the dilution rate, were defined by constant alkali addition rate and constant CO 2 , O 2 and biomass concentrations after four residence times (~ 40 h). Samples were taken regularly to monitor growth and to determine if a steady-state had been reached. All samples were quickly frozen in liquid nitrogen. Mycelium harvested during steady-state conditions was used for micro-array analysis.

Analysis of culture broth
Dry weight biomass concentration was determined by weighing lyophilized mycelium separated from a known mass of culture broth. Culture broth was filtered through GF/C glass microfiber filters (Whatman). The filtrate was collected and frozen for use in solute analyses. The mycelium was washed with demineralised water, rapidly frozen in liquid nitrogen, and stored at -80°C until lyophilization. Extracellular protein concentration was determined using the Quick Start Bradford Protein Assay (Bio-Rad) with BSA as a standard. The total organic carbon in the culture filtrate was measured with a Total Organic Carbon Analyzer (TOC-Vcsn; Shimadzu, Japan), using glucose as a standard.

RNA isolation and quality control
Total RNA was extracted by modified Trizol extraction. Frozen ground mycelium (~200 mg) was directly suspended in 800 µl Trizol reagent (Invitrogen) and vortexed vigorously for 1 min. After centrifugation for 5 min at 13000 rpm, 450 µl of the supernatant was transferred to a new tube. Chloroform (150 µl) was added and after 3 min incubation at room temperature, samples were spun down for 15 min at maximum speed. The upper aqueous phase was transferred to a new tube to which 400 µl of isopropanol was added, followed by 10 min incubation at room temperature and centrifugation for 10 min at 13000 rpm. Pellet was washed with 75% ethanol and finally precipitated in 100 µl H 2 O. RNA samples for micro-array analysis were additionally purified on NucleoSpin RNA II columns (Machery-Nagel) according to manufacturer's instructions. RNA quantity and quality was determined on Nanodrop spectrophotometer.

Microarray analysis
Probe synthesis and fragmentation were performed at ServiceXS (Leiden, The Netherlands) according to the GeneChip Expression Analysis Technical Manual (Affymetrix inc., 2002). DSM (Delft, The Netherlands) proprietary A. niger gene chips were hybridized, washed, stained and scanned as described in the GeneChip Expression Analysis Technical Manual (Affymetrix inc., 2002).

Normalization, filtering, statistical significance and comparisons
Transcriptomic analysis was basically performed using Bioconductor and statistical programming language R (Nitsche et al. 2012b). Two experimental conditions (N402 vs B36) were compared to each other; each condition was represented by independent triplicate cultures. Using the robust multi-array analysis (RMA) package (Irizarry et al. 2003), background correction, normalization and probe summarization steps were performed according to the default setting of the RMA package. Differential gene expression was evaluated by moderated t-statistics using the Limma package (Smyth 2004) with a threshold of the Benjamini and Hochberg (BH) False Discovery Rate (FDR) (Benjamini and Hochberg 1995) at 0.005. A minimal fold-change criterion was not applied for the identification of differentially expressed genes, as fold-changes are not necessarily related to biological relevance (van den Berg et al. 2010;van den Berg et al. 2006). Foldchanges of gene expression from N402 to B36 (B36/N402) were calculated from normalized expression values. Means of the expression values for B36 and N402 as well as classifiers for the moderated t-statistics are summarized in Additional file 1.

Gene Ontology (GO) and enrichment analysis
Controlling the FDR at q<0.05, over-represented GO terms in sets of differentially expressed genes were determined by Fisher's exact test (Fisher 1922). An improved GO annotation for the A. niger CBS513.88 was applied that is based on ontology mappings from A.nidulans FGSCA4 (http://www.broadinstitute.org/fetgoat/index.html) (Nitsche et al. 2011).

Quantitative real-time PCR (qPCR)
Quantitative real-time PCR was performed as described (Nowrousian et al. 2005) with slight modifications. After a DNAse treatment (DNA-freeTM Kit Applied Biosystems) of 10 µg of total RNA, 600 ng were used for cDNA synthesis using the iScriptTM cDNA Synthesis Kit from BioRad according to the manufactures instructions. Real time PCR was performed in PTC-200 Peltier Thermal Cycler with a Chromo4 Continuous Fluorescence detection system from Bio-Rad with a SYBRgreen mix (iQ™ SYBR® Green supermix) in a volume of 25 µl. Each reaction was carried out in triplicate with ~200 ng of cDNA and each oligonucleotide primer at 0.3 µM. Oligonucleotide primers used for real time PCR are listed in Additional file 8. Two reference genes, H 2 B and Cox5 were used for normalization. The PCR program was as follows: 95 o C for 3 min, followed by 39 cycles of 95 o C for 15 sec and 60 o C for 1 min, followed by a melting curve analysis. The efficiency of each primer pair and mean Ct (threshold cycles) values were calculated and used for determination of glucoamylase RNA transcript levels (Pfaffl 2001). Estimation of glucoamylase gene copy number was also performed by qPCR as well but with genomic DNA as a template.

Analysis of glucoamylase activity
One unit of glucoamylase activity was expressed as the amount of enzyme that liberates one micromole of glucose per minute from starch. Filtrates from steady-state cultures were mixed with 1% soluble starch (Sigma, S9765) in 50 mM NaAc buffer at pH 4.5 and incubated at 37 o C for 3.75 min. Liberated glucose was determined with the glucose kit from ABX (Pentra Glucose HK CP).