Physiological evaluation of the filamentous fungus Trichoderma reesei in production processes by marker gene expression analysis

Background Biologically relevant molecular markers can be used in evaluation of the physiological state of an organism in biotechnical processes. We monitored at high frequency the expression of 34 marker genes in batch, fed-batch and continuous cultures of the filamentous fungus Trichoderma reesei by the transcriptional analysis method TRAC (TRanscript analysis with the aid of Affinity Capture). Expression of specific genes was normalised either with respect to biomass or to overall polyA RNA concentration. Expressional variation of the genes involved in various process relevant cellular functions, such as protein production, growth and stress responses, was related to process parameters such as specific growth and production rates and substrate and dissolved oxygen concentrations. Results Gene expression of secreted cellulases and recombinant Melanocarpus albomyces laccase predicted the trends in the corresponding extracellular enzyme production rates and was highest in a narrow "physiological window" in the specific growth rate (μ) range of 0.03 – 0.05 h-1. Expression of ribosomal protein mRNAs was consistent with the changes in μ. Nine starvation-related genes were found as potential markers for detection of insufficient substrate feed for maintaining optimal protein production. For two genes induced in anaerobic conditions, increasing transcript levels were measured as dissolved oxygen decreased. Conclusion The data obtained by TRAC supported the usefulness of focused and intensive transcriptional analysis in monitoring of biotechnical processes providing thus tools for process optimisation purposes.


Background
Microorganisms used in various types of biotechnical processes encounter constantly changing environmental conditions, to which they adapt by changing their cellular physiology. The performance of the microorganism has a major impact on the performance of the process and as a consequence, bioprocess monitoring and control strategies based on the physiological status of the culture have become more popular [1,2]. In bioprocesses the physiological status of the culture is generally measured indirectly by analysis of extracellular variables such as enzyme activities, pH and exhaust gas concentrations. Recent developments in analytical methods allow direct evaluation of cellular physiology by analysis of intracellular variables such as proteins, RNAs or metabolites, either in focused or systems-wide manner [1,3].
By combination of transcriptomic, proteomic and/or metabolic profiles with advanced bioinformatic analysis it is possible to predict from the massive amount of data biologically meaningful analytes, i.e. biomarkers which can predict certain physiological events. In biotechnical processes such biomarkers have been used e.g. for detection of osmotic and oxidative stress conditions or trace element deficiencies in yeasts [2,4] and bacteria [3]. Simple and robust assays are required in order to use these biomarkers in monitoring or control of bioprocesses. Transcriptional analysis of selected marker genes provides one potential way for robust monitoring of physiological events, since adaptation to changes in the environment takes place rapidly at the level of gene expression [5,6]. In addition it is possible to measure multiple targets in a single assay with RNA which is more difficult to achieve with enzymes or metabolites.
Few microarray studies have been performed specifically to identify gene markers to be used in monitoring of industrial fermentations [4,7]. Conventional Northern blot analysis is still abundantly used for the actual monitoring of the selected markers in process conditions [2,4,7]. New tools for more robust RNA analysis are emerging, such as quantitative RT-PCR [8], and combination of sandwich hybridisation with fluorescent [9] or electric detection [10,11]. Depending on the organism and length of the process these methods may be feasible for at-line mRNA expression monitoring, but would generally be applicable for only a limited number of targets and/or samples.
We have recently developed a method for transcriptional analysis method called TRAC (Transcript analysis with aid of affinity capture) [12,13]. The TRAC method is suitable for rapid simultaneous analysis of 96 samples with multiplex target detection, allowing high-throughput, focused expression analysis [12] or alternatively it can be performed with amplifiable DNA probes and PCR [13] to reach high sensitivity. TRAC has been applied for gene expression analysis of microbial cultures [12], multiplex quantification of bacterial populations [14,15] and to monitor gene expression stability in continuous cultures [16]. In this study we have used TRAC for expression monitoring of more than 30 marker genes in different types of protein production processes performed with the industrially important filamentous fungus Trichoderma reesei. Expression levels of the marker genes, which are involved in different process relevant pathways, were compared with extracellular process parameters from the cultures, such as specific growth rates, production rates of secreted enzymes and availability of nutrients and oxygen. Identification of relationships between gene expression fingerprints and process performance would offer possibilities for process optimisation and control.

polyA RNA-and biomass-based normalisation of gene expression data
Expression of a set of 34 marker genes, representing various physiological states, was monitored in different phases of two batch and fed-batch cultures and one continuous culture of an M. albomyces laccase producing strain of the filamentous fungus Trichoderma reesei. Potentially useful molecular marker genes were selected on the basis of their responses in conditions of interest as reported in publicly available transcriptional analysis data for various species of filamentous fungi or Saccharomyces cerevisiae. The genes used in this study are listed in Table 1 with references to available data on these genes or corresponding homologues in other organisms.
The TRAC expression analysis for the selected genes was performed directly from cell lysates, and the gene expression levels were related to the amount of biomass used in the hybridisation or to overall polyA RNA concentration of the lysate (see Methods). The ratio of marker gene mRNA to total polyA RNA shows the change in the expression of a specific gene relative to overall gene expression, whereas the ratio of marker gene mRNA to biomass shows the changing level of a transcript relative to the overall culture (viable/non-viable).
PolyA RNA concentration in the lysed biomass in different growth phases of the two batch cultures, which differed considerably in growth and substrate consumption rate are presented in Figure 1. In batch I the polyA RNA concentration varied between 2 and 2.5 mg g -1 dry weight (DW) for the first 35 h, while μ was above 0.06 h -1 and lactose concentration above 10 g l -1 . After 35 h, when the culture conditions had become growth limiting, the polyA RNA content relative to biomass decreased to approx. 0.4 mg g -1 dry weight. However in batch culture II the specific growth rate and uptake of lactose were slower than in batch I after 20 h and the polyA RNA per g biomass decreased at this time already, although the polyA RNA content in the biomass subsequently increased between 30 and 45 h. After 45 h the cultures were similar. To verify that the decrease in the polyA RNA content after 20 h was not a result of experimental error (e.g. variation in cell lysis efficiency), total RNA was extracted from the samples of batch culture II and the polyA RNA to total RNA ratio was determined. This confirmed that a decrease in overall gene expression had occurred after 20 h ( Figure 1B).
The specific growth rates (μ) were compared to the expression of two ribosomal protein mRNAs (rpl16a, rps16b) as growth-related marker genes in these batch cultures. The changes in the transcription of these genes relative to biomass were similar to μ ( Figure 1C). The relatively low specific growth rate between 20 and 30 h for batch II (25% lower than in batch I), corresponded to a 1.7-fold lower expression of the two ribosomal mRNAs during this period. However, in both fermentations, an approximately 2-fold increase in the ribosomal protein mRNA to biomass ratio was observed 3 to 5 h before the decrease in specific growth rate which occurred after all lactose had been consumed.

Monitoring carbohydrate metabolism markers
A number of genes coding for enzymes involved in central carbohydrate metabolism (glycolysis, TCA cycle) are known to be transcriptionally regulated according to the availability of glucose in T. reesei [17]. We measured the relative expression of three of these genes involved in central carbohydrate metabolism in batch and fed-batch fermentations with lactose as carbon source ( Figure 2). Lack of glucose has been shown to have an increasing (acetyl-CoA-synthethase, acs1), decreasing (enolase, eno1) or indifferent effect (glyceraldehyde-3-P-dehydrogenase, gpd1) on their expression [17]. Both gpd1 and acs1 increased from 25 to 45 h in lactose-rich conditions (>10 g l -1 ) in batch cultures ( Figure 2A). After the lactose concentration decreased below 10 to 6 g l -1 , the abundance of gpd1 mRNA relative to polyA RNA decreased 30 -65% compared to its maximal expression around 50 h, whereas acs1 mRNA continued to increase at low lactose concentration (below 5 g l -1 ) and during starvation. About 20 h after complete exhaustion of lactose, acs1 was expressed at a 2.5 to 3 -fold higher level than in high lactose concentra-PolyA RNA and ribosomal protein mRNA levels in twobatch cultures of T. reesei Figure 1 PolyA RNA and ribosomal protein mRNA levels in twobatch cultures of T. reesei. A. Lactose, biomass and CO 2 concentrations in batch cultures I and II. B. PolyA RNA concentration relative to biomass in cultures I and II and polyA RNA concentration relative to total RNA in culture II. C. Specific growth rate (μ) and expression of two ribosomal protein mRNAs (rpl16a, rps16b) relative to biomass. Cultures were maintained at 28°C and pH 5.5 -6.
tions. The eno1 mRNA to polyA RNA ratio was constant when lactose concentration was above 5 g l -1 , whereas during starvation eno1 mRNA was undetectable.
In the fed-batch fermentations a 2 to 3 -fold increase of acs1 mRNA level occurred when lactose concentration decreased from 40 to below 5 g l -1 , before the start of the lactose feed ( Figure 2B) as in the deceleration to stationary phases of the batch cultures. Lactose was fed to the fermentor after 45 h at a rate that maintained the rate of base (NH 4 OH) consumption at around 0.012-0.014 ml l -1 min 1 . As long as the feed rate of lactose was below 30 mg h -1 g -1 DW, acs1 mRNA level continued to increase, but it decreased rapidly after the feed rate increased above this value ( Figure 2B), even though the residual lactose con-centration in the medium remained zero. Termination of the lactose feed caused another increase in acs1 expression. gpd1 and eno1 mRNA levels showed only minor changes during the different phases of the fed-batch cultures, and eno1 was again undetectable under starvation conditions.

Monitoring marker genes responding to oxygen
Expression of the heat shock protein gene hsp70 and the heme biosynthesis gene hem6 has been shown to be regulated by oxygen availability in T. reesei [16]. The trx2 gene encoding thioredoxin was chosen as a potential marker for oxidative stress based on S. cereviciae transcriptional data [18]. In batch cultures ( Figure 3) and in the batch phase of fed-batch cultures the expression of these three genes increased 1.6 to 2-fold between 18 to 45 h, while the biomass increased from approximately 3 to 20 g l -1 and pO 2 decreased from 90 to 30 %. A 20-35% decrease in the expression level relative to polyA RNA was observed for these 3 genes after the maximal expression was measured ( Figure 3A).
The maximal expression levels of these genes in batch cultures were compared to the corresponding levels measured in steady state of chemostat cultures with high and low cell density, and in anaerobic conditions (TRAC data of chemostat and anaerobic cultures for this comparison from Rautio et al. [16]) as well as in shake flask precultures ( Figure 3B). In anaerobic conditions the expression levels of hsp70 and hem6 were 3.5 and 2.8 -fold higher than the maximal expression levels measured in batch or fed-batch cultures. Comparable hsp70 and hem6 expression levels were measured between high cell density (16-20 g l -1 ) batch or fed-batch cultures and chemostat cultures with 30% pO 2 , whereas at low cell density (4 g l -1 and a pO 2 of 80%) the expression level of hsp70 was 2.7-fold lower and that of hem6 was appr. 2-fold lower than in the high cell density aerated cultures. In the shake flask cultures (~2 g l -1 ) hem6 expression levels were comparable to those observed in high density cultures, whereas the hsp70 transcript level was 2-fold lower. This indicates that other stress factors besides oxygen limitation resulted in up-regulation of the hsp70 mRNA level in the high density cultures. Trx2 expression was similar in all aerobic bioreactor cultures, but was on average 1.5-fold higher in aerobic than in anaerobic or shake flask culture conditions.

Comparison of secreted enzyme production rate and mRNA expression
The transformant strain LLK13/295 used in these studies produced M. albomyces laccase under the cellobiohydrolase (cbh1) promoter, as well as all the native cellulases, including cellobiohydrolase I (CBHI). Cbh1 promoter is induced by cellulose and by oligosaccharides and disaccharides derived from cellulose, such as cellubiose or sophorose. Also several other disaccharides such as lactose induce cbh1 expression [19]. Transcript levels of the genes (cbh1, lac1) expressing these secreted enzymes and two folding factors (protein disulfide isomerase pdi1 and protein chaperon bip1) were monitored during batch and fedbatch cultures and were compared with specific extracellular production rates of laccase and CBHI ( Figure 4). Both biomass (lac1, cbh1) and polyA RNA (bip1, pdi1, cbh1) concentrations were used to normalise gene expression so that lac1 and cbh1 could be readily compared with bio-mass specific production rates and to allow physiological interpretation of the data.
Expression of cbh1, lac1, pdi1, and bip1 increased during the first 40 h of growth in batch cultures. In the fed-batch culture there was a 1.8 and 2.6 -fold increase in mRNAs for lac1 and cbh1 respectively, after the lactose feeding was started, whereas their expression decreased in the batch cultures. There were corresponding increase in the specific extracellular production rates of laccase (3.8 -fold) and CBHI (4.3 -fold), after starting of the lactose feed in the fed-batch culture. The highest amounts of lac1, cbh1, pdi1 and bip1 mRNAs were measured between 7 and 12 h after the start of the lactose feed. 36 h after the lactose feed had been started both the gene expression (mol/biomass) and the specific enzymes production rates (nkat h -1 g -1 DW) had decreased to levels similar to those observed during the batch phase.
To study the variation between the biomass and polyA mRNA based normalisation they were compared in the normalisation of the major cellulase cbh1 expression in the fed-batch culture ( Figure 4B). An expected difference between the profiles of cbh1 to biomass and cbh1 to polyA RNA ratios was observed at the end of the culture. Cbh1 to biomass ratio showed a faster decreasing trend after about 70 h than cbh1 expression relative to polyA RNA, since increasing proportion of the biomass became metabolically inactive and polyA was produced only by the metabolically active part of the culture. It should be noted, however, that e.g. increase in a gene mRNA to polyA RNA ratio is not necessarily a sign of specific up-regulation, but can also indicate slower down-regulation than for the majority of genes. However increase or decrease in this ratio indicates the growing or decreasing physiological importance of the gene product in the particular conditions.
The specific production rate of both CBHI and laccase followed trends similar to the cbh1 and lac1 gene expression profiles, although a delay was observed between the gene expression and the corresponding extracellular enzyme production rate. For example the specific extracellular laccase production rate, which was more frequently measured than CBHI, was highest 6 to 10 h after the highest gene expression level was measured. An increase in laccase production rate was consistently observed at the end of the fed-batch cultures after lactose feed was stopped (Figure 4). This was not observed for CBHI.

Monitoring responses caused by starvation
The highest expression levels for genes coding the cellulases, laccase and folding factors were detected at growth rate 0.05 -0.03 h -1 , at low lactose concentration in batch and lactose limiting in fed-batch cultures ( Figure 4). To Expression of oxygen sensitive hem6, hsp70 and trx2 genes in different cultures of T. reesei determine the responses of the marker genes during the transition from deceleration to stationary phase, samples were taken frequently around the time when lactose was expected to be exhausted in batch cultures ( Figure 5). Gene expression levels are presented relative to overall gene expression (polyA RNA), because the responses of metabolically active cells during starvation was of interest.
The time when the base consumption rate had decreased to zero and the lactose was undetectable was considered to be the time at which starvation began. The expression of genes coding for carbohydrate degrading enzymes was at this time on average 1.5 -fold lower than the maximal expression level measured during these cultures. The time of maximal expression of cbh1, lac1, egl1 (endoglucanase) Expression of the recombinant M. albomyces laccase gene lac1, the cellulase gene cbh1 and protein folding factor genes pdi1 and bip1 in batch and fed-batch cultures of T. reesei Figure 4 Expression of the recombinant M. albomyces laccase gene lac1, the cellulase gene cbh1 and protein folding factor genes pdi1 and bip1 in batch and fed-batch cultures of T. reesei. A. lac1 mRNA level relative to biomass, specific extracellular production rate of laccase r MaL , specific growth rate μ. B. cbh1 mRNA relative to biomass (batch and fed-batch) and relative to polyA RNA (fed-batch) and specific extracellular production rate of CBHI. C. pdi and bip1 mRNA relative to polyA RNA. Dotted, vertical lines in the fed-batch culture represent the start and end of lactose feed.
and bgl2 (β-glucosidase) was 3.6 to 6.4 hours prior to the start of starvation, when μ was 0.03 -0.045 h -1 , the rate of base consumption was between 0.1 and 0.2 μmol min -1 g -1 DW and the lactose concentration was below 6 g l -1 . Along with cellulases and laccases the majority of the genes analysed showed down-regulation during starvation, including the folding factors (pdi1, bip1), oxidative stress genes (thioredoxin reductase trr1 and thioredoxin protein trx2) and glycolysis genes (eno1, gpd1) (data not shown).
However, six to eight marker genes out of the 34, in addition to acs1 (Figure 2), showed more than 2 -fold increased mRNA levels during starvation ( Figure 5). The trend of increasing expression of these genes started 0.5 to 5 hours before base consumption stopped. These genes coded for heat shock proteins hsp105 and hsp30, metacas-pase (mca1) involved in apoptosis, neutral trehalase (nth1), membrane fusion factor (nsf1), copper transporter (ctaA) and two conidiation related proteins (ccg9 and con6, measured only in batch I).

Monitoring of a continuous culture
Continuous culture was used to study further the effect of growth conditions on the productivity of extracellular proteins and to optimise their production. Different growth conditions were applied during the culture, i.e. altering the medium feed rate and composition and process temperature. In addition, unexpected disturbances occurred during the process, which was maintained for 1000 h. The TRAC assay was used to monitor expression 20 genes in a single pool, including markers that showed predictive value for cellulase productivity (cbh1), specific growth rate (rpl16a), and starvation (acs1) (Figure 6). A Responses of marker genes to starvation in two batch cultures (I and II) of T. reesei Figure 5 Responses of marker genes to starvation in two batch cultures (I and II) of T. reesei. A. Specific growth rate μ, lactose concentrations, specific base consumption rate r base . B. Expression of cellulase and laccase genes as log 2 ratio to maximal expression measured in the cultures (blue symbols). Expression of genes showing >2-fold increasing mRNA level relative to polyA RNA after exhaustion of lactose as log 2 ratio to the expression level in a sample taken before the observed increase (red symbols).
more detailed description of the process parameters in the continuous culture will be published elsewhere (M. Bailey, unpublished results).
The feed rate of the culture was controlled by the rate of base consumption (see Methods). The medium feed to the fermentor started at 41 h ( Figure 6I) when the rate of base consumption decreased below the DELTABAS criterion value (0.02 ml (5% NH 4 OH) l -1 min -1 ). This criterion kept the dilution rate (D) at approximately 0.04 h -1 and the specific base consumption rate, r base , was 0.165 μmol min -1 g -1 DW. The criterion for base consumption was decreased in two steps: at 309 h by 25% ( Figure 6II) and at 407 h by additional 20% (Figure 6III). The average r base value decreased to 0.14 and 0.12 μmol min -1 g -1 dry weight as a result of these two changes and the average dilution rate decreased to 0.036 and 0.026 h -1 . The cbh1 mRNA level increased 1.6-fold ( Figure 6B) and the specific extracellular protein production rate (r prot ) increased 1.2 -1.5 -fold ( Figure 6A) after the first step (II). After the second reduction in base consumption rate (III) there was no immediate effect on cbh1 mRNA or extracellular protein production rate, but after 48-57 h (1.2-1.5 residence times) these values decreased to the level measured prior to the changes in the base consumption rate (before 309 h).
At 550 h the organic nitrogen concentration was halved in the medium ( Figure 6IV), which decreased D to an average value of 0.017 while the r base increased to an average level of 0.134 μmol min -1 g -1 DW. The lower level of nitrogen in the medium will have been partly compensated for by the increased base (NH 4 OH) consumption. The specific production rate of extracellular proteins was unaffected by this change and the only marker gene which responded to the lowered N concentration was the amino acid permease gap1, the mRNA level of which was increased 1.8-fold 65 h after change in the medium (data not shown).
The expression profile of the ribosomal protein mRNA (rpl16a) followed the changes in the dilution (=growth) rate throughout the culture, consistent with earlier observations from batch cultures (Figure 1). The gene coding for acetyl-CoA-synthase (acs1), which showed its highest expression at a low lactose concentration (Figure 2), was somewhat induced when the dilution rate was reduced ( Figure 6C). Two other genes (hsp30, nsf1) showing increased mRNA level under starvation conditions ( Figure  5) were also transitorily induced after 309 h when the rate of medium feed decreased (data not shown).
Finally at 861 h the temperature of the culture was reduced from 28 to 24 °C over a 50 h period ( Figure 6V). A number of marker genes responded by increasing their expression levels. Some of these genes (cbh1, egl1, lac1 and vacuolar protease vpa1) returned to the expression levels measured before the temperature change in less than 45 h, whereas others (hsp30, nsf1, bgl2, gpd1) remained at the higher expression level, even after the temperature was increased to 28°C again for the remainder of the culture. However, it can not be concluded that these responses were specifically related to the temperature change, since a temporary increase in the dilution rate occurred simultaneously because of a process control problem when growth and base consumption were reduced at the lower temperature (see Figure 6V).
Two technical disturbances occurred during the process.
The first occurred at 68 h when the vessel weight controller malfunctioned and the culture volume decreased to about 8 litres. The volume was increased to 10 l by medium addition, which caused increased expression of e.g. cbh1 and ribosomal protein mRNA (rpl16a) genes. Both reached stable expression levels within 70-90 hours (2.8 -3.8 residence times) of the disturbance (Figure 6*).
At 504 h, oscillation in the feed control mechanism was observed. Samples were collected at 1 h interval after this disturbance and temporarily increased expression in many of the marker genes (e.g. cbh1, rpl16b, pdi1, gpd1) was observed ( Figure 6O).

Discussion
In this work we monitored the expression of 34 marker genes during changing conditions in batch, fed-batch and continuous fermentations of the filamentous fungus T. reesei using the transcriptional analysis method TRAC. In production processes the environmental conditions in the fermentor may be variable, causing corresponding variability in physiological parameters. This makes conventional gene expression data normalisation methods, such as the use of house-keeping genes, unreliable. In this study we related the expression of specific genes either to the amount of biomass or to the total polyA RNA content used in the hybridisation. Relating marker gene expression to biomass can be more useful in prediction of culture performance parameters such as specific growth and production rates, whereas normalisation to polyA RNA predicts the physiological responses in metabolically active cells more accurately.
Both polyA RNA to biomass and polyA RNA to total RNA ratios were shown to be growth phase dependent, and batch-to-batch variation in growth and protein production was also evident at the level of overall mRNA expression ( Figure 1). Quantitative determination of total RNA per g biomass from mycelial biomass is not reliable, however comparing the polyA RNA per g biomass with polyA RNA per total RNA from the same culture (Figure 1) indicated that the total RNA to biomass ratio was not constant in changing environmental conditions of the batch cultures. It is thus beneficial to use a quantitatively measured factor to normalise expression levels between samples collected from different phases of a cultivation, rather than basing the normalisation on the assumption that a chosen factor is constant.
The normalisation based on polyA RNA is similar to the use of the expression levels of all genes for a normalisation in microarray experiments. However, for normalisation of genome-wide transcription data the distributions of the expression signal intensities between samples are equalised, i.e. it is assumed that global shifts in the mRNA population do not occur. This is a reasonable assumption e.g. in steady state conditions that can be achieved in continuous cultures, however disregarding global changes in mRNA levels during non-steady conditions (e.g. batch and fed-batch cultures) can result in the masking of true gene expression differences between compared samples [20]. On the other hand, variability in oligo dT-based mRNA capture can be caused by decreasing polyA tail length, that has been observed in stationary phase yeast cells [21]. However, since deadenylation leads to degradation of the mRNA body [22], and the polyA tail is required for translation initiation [23], the physiological relevance of mRNA species with short or no polyA tail is presumably insignificant, when the mRNA synthesis rate is low.
Expression of the native cellulase gene cbh1 and the recombinant laccase gene lac1 showed correlation with the specific extracellular production rates of the corresponding enzymes. There was a delay of 6 to 10 h between increased level of lac1 transcripts and laccase production ( Figure 4). M. albomyces laccase is processed at its C-terminus and is presumably activated by the processing [24], and a delay in this processing might contribute to the difference between the lac1 expression and laccase production profiles. C-terminal processing of laccase could also explain the strong increase of its activity in the fed-batch cultures after termination of the substrate feed, which was not observed for CBHI (Figure 4). Gene expression of the secreted enzymes was highest when the growth rate was decreasing from 0.05 to 0.03 h -1 in batch and fed-batch cultures (Figures 4 and 5) or was constant between 0.026 and 0.036 h -1 in the continuous culture ( Figure 6). This is in accordance with the optimal growth rate (0.031 h -1 ) for production of extracellular proteins observed in chemostat cultures of T.reesei [25]. When the growth rate was below 0.03 h -1 decreased gene expression of cellulases and recombinant laccase was observed in addition to the responses associated with starvation ( Figure 5). This data indicates that the "physiological window" in which the productivity of these extracellular proteins is optimal is narrow. In general the relation between specific product formation rate of a secreted protein and specific growth rate seems to be more dependent on the protein and its transcriptional promoter than on the organism. The specific product formation rate of a secreted protein can increase consistently with increasing growth rate, as shown e.g. for Fab fragment production in Pichia pastoris [26] and β-galactosidase production in Escherichia coli [27]. Whereas for other secreted proteins, such as α-galactosidase of P. pastoris, the specific product formation rate is at highest at low growth rates [28].
The DELTABAS feed control system, uses the rate of base consumption to try to maintain a culture in the optimal production phase by limiting the growth of the culture but avoiding total carbon source depletion [29]. During the optimal production phase the base (NH 4 OH) was consumed at a rate of 0.1 -0.2 μmol min -1 g -1 DW (Figures 5 and 6). However, in the fed-batch and continuous cultures in which this feeding strategy was applied, the optimal production was maintained for rather short periods of time. In the continuous culture most of the time the growth (and base consumption) rates were either too high or too low ( Figure 6). During the feeding phase in the fed-batch cultures the growth rate was mainly below 0.025 h -1 (Figure 4), being too low for optimal production. Induced expression level of some of the starvation markers during the lactose (fed-batch, data not shown) or medium feed (acs1, continuous culture, Figure 6C) further indicated that too little substrate was available for the culture at this low dilution rate. However, at the end of the fed-batch cultures it was observed that the metabolically active part of the culture was still expressing cellulase genes (cbh1, Figure 4B) at a relatively high level, suggesting that improved productivity would be attained if culture viability could be maintained for a longer time.
Altogether 9 out of 34 marker genes (Figures 3 and 6) were observed to have more than 2-fold increased mRNA level under starvation and slow growth conditions, indicating their increasing importance under these conditions. Of the induced genes, nth1 codes for neutral trehalase, which is involved in degradation of storage carbohydrates [30].
CtaA codes for a copper transporter [31]. and in these fermentations the copper requirement may have been increased because of production of the copper-containing laccase . The acs1 gene has been shown to be strongly induced in glucose starvation in T. reesei [17]. The induction of these genes elucidates some of the physiological events during starvation and slow growth: mobilisation of storage carbohydrates, acquisition of nutrients from the medium, stress responses, onset of conidia formation and triggering of apoptosis. The consistent induction of these genes preceding the complete exhaustion of the carbon source can make them useful markers in optimisation of substrate feeding strategies, when their expression is monitored continuously.
The heme biosynthesis gene HEM13 of S. cerevisiae has been shown to be induced under anaerobic conditions [39] and the Ssa type Hsp70 proteins have been shown to be part of a complex that regulates transcription of aerobic genes [40]. The homologues of these genes are induced by anaerobic conditions in T. reesei (Figure 3). Based on the expression of these genes during the batch, fed-batch and continuous processes, 30% dissolved oxygen (DO) concentration even at relatively high cell density (DW ~ 20 g l -1 ) is apparently sufficient to maintain aerobic culture conditions, although at lower density (4 g l -1 ) and higher DO (80%) the expression of these genes was more than 2fold lower. Strong induction of these anaerobic marker genes and decreased levels of the cellulase gene mRNAs were observed only when DO was lower than 10% (data not shown). In large scale industrial fermentations with thick filamentous slurry, inhomogeneity of the system can cause local limitations in oxygen supply [41]. Thus monitoring of the expression of these markers in such fermentations could be useful in devising intervention strategies which would maintain better aeration and productivity.
Simultaneously with the increase in hsp70 and hem6 expression, moderately increased expression of the thioredoxin gene (trx2) was observed (Figure 3), which indicated a response to accumulation of reactive oxygen species presumably because of culture ageing [42].
The TRAC method was shown to be an efficient tool for monitoring sets of marker genes in different bioprocess conditions. The possibility to analyse simultaneously in 96-well format large number of samples, allows frequent sampling during the cultivations. Crude lysed cell material can be directly used in the hybridisations, thus RNA extraction and cDNA conversions are avoided. Detection of multiple mRNA targets from a single sample and low hands-on time of the protocol makes the assay cost-effective. For off-line monitoring, the advantages of the system are most clearly demonstrated in the analysis of the con-tinuous culture. The protocol starting from the lysis of all the 48 mycelial samples to analysis of expression of 18 genes in duplicate (1728 expression levels) was carried out in 7 h with 1 to 2 h hands-on time. For at line monitoring (less than 10 samples at once) the entire protocol (from sampling to results) can be carried out in approximately 2 h. Increasing the degree of automation by combining some or all of the 4 steps (cell lysis, hybridisation, sample treatment, CE analysis) would make the overall protocol time even shorter and frequent at-line monitoring more practical. Such an analysis system would have a wide application potential.
We have demonstrated here that by marker genes analysis it is possible to evaluate various physiological factors of the culture, such as nutrient and oxygen limitation, growth and extracellular protein production rates in the changing environmental conditions. Transcriptional analysis can provide a more complete picture of the physiological state of an organism than can be achieved by the external parameters that are measured from production processes. In addition to increasing the knowledge of gene regulation, gene expression data collected at high frequency may suggest strategies for optimising process parameters, such as medium composition and feeding strategy. Accumulation of microarray data from T. reesei and other industrially relevant microorganisms grown under various conditions will further help in the selection process of marker genes, the expression of which has a predictive value in the evaluation of the physiological state and performance of the cultures.

Conclusion
We have monitored by the TRAC method the expression of a set of 34 marker genes, involved in different process relevant pathways, in batch, fed-batch cultures and continuous cultures of a filamentous fungus T. reesei transformant strain producing M. albomyces laccase. Many of the marker gene expression levels measured at frequent intervals showed to have value in prediction of consecutive physiological effects and process performance. mRNA levels of genes coding for industrially relevant secreted cellulases and recombinant laccase followed the trends in the corresponding extracellular enzyme production rates and was at highest in narrow specific growth rate range of 0.03 -0.05 h -1 . The specific growth rate of the fungal cultures was possible to evaluate on the basis of ribosomal protein mRNA expression. Increasing expression of altogether nine starvation related genes preceded the complete exhaustion of carbon source at least by some hours, indicating their usefulness in prediction of insufficient substrate feed for optimal production. Deficiency in oxygen supply was manifested by increased level of two oxygen sensitive genes hem6 and hsp70. The TRAC method was shown to be an effective tool in focused transcrip-tional monitoring of biotechnical processes and the data produced by this method supported the usefulness of intensive gene expression analysis for process optimisation work.

Transcriptional analysis by TRAC
Biomass was harvested from fermentations for transcriptional analysis by anaerobically withdrawing medium containing 50-150 mg fresh biomass. Biomass was separated from medium by quick filtration with glass-fibre filter disks (Whatman GF/B 47 mm ∅, Kent, UK). The biomass was immediately washed with RNAse-free (dimethyl pyrocarbonite (DMPC)-treated) water, after which the biomass was transferred in tarred screw-cap tubes to liquid nitrogen and stored at -80°C. This sampling procedure took <5 min.
The steps following hybridization, including affinity capture, washing and elution, were automated with a King-Fisher 96 magnetic bead particle processor (Thermo Electron, Vantaa, Finland) in 96-well plates as follows: 1) affinity capture of hybridized RNA targets to 50 μg of streptavidin-coated MyOne DynaBeads (Dynal, Oslo, Norway) at room temperature (RT) for 30 min, 2) washing of the beads twice for 1. The eluents were analyzed by capillary electrophoresis with ABI PRISM 310 or 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA). In order to compare individual samples and to calibrate the separation of the detection probes by size, GeneScan-120LIZ size standard (Applied Biosystems) was added to each sample. The identity of the probes was determined by the migration and the quantity by the peak area. In order to convert the measured peak area to the molar amount of a probe, the fluorescence signal intensity relative to molar amount was determined for each 6-FAM labelled probe.
Total polyA RNA quantification from prepared lysates was performed with the above TRAC protocol without addition of detection probes. The final elution of polyA RNA was performed in 50 μl DMPC treated water. RNA concentration in the eluent was quantified with a RiboGreen RNA quantitation kit (Molecular Probes, Leiden, the Netherlands).
The expression levels of the marker genes during the cultivations are presented here either as changes relative to a control sample or as molar amounts of the target specific probe detected. The expression levels presented as molar amounts can be compared between conditions for a given gene, but not between different genes, since the hybridisation efficiency between probe-target pairs can vary.

Probe selection
The detection probe oligonucleotides, labelled at the 3' and 5' ends with 6-FAM, were synthesized by Thermo Electron (Ulm, Germany). The biotinylated Oligo(dT) capture probe was from Promega. The HPLC-purified oligonucleotide detection probes (Table 1)