S. cerevisiae strains and plasmids used in this study are summarized in Table 1. Escherichia coli strain DH5α was used for sub-cloning and was grown on LB medium supplemented with 100 mg/L ampicillin. Defined mineral medium was used for S. cerevisiae cultivation and was composed of: xylose or glucose, 60 g/L; mineral salts ((NH₄)₂SO₄, 5 g/L; KH₂PO₄, 3 g/L; MgSO₄·7H₂O, 0.5 g/L); buffer (potassium hydrogen phthalate, 50 mM pH 5.5); Tween 80, 0.4 g/L; ergosterol, 0.01 g/L 28; vitamins and trace elements 29. Identical medium was used for pre-cultures and batch fermentation in instrumented bioreactors with the exception that buffering agent was omitted in the latter case. At the start of each experiment, yeast strains were streaked from 15% (v/v) glycerol stocks and grown two days on Yeast Nitrogen Base (YNB) glucose plates. Pre-cultures were inoculated in baffled shake-flasks (10% liquid volume) at a predetermined cell density, OD_{620 nm} = 0.5/0.025 (xylose/glucose), and grown for 20 hrs to yield cells in late exponential phase (OD_{620 nm}~14). Cultivation of S. cerevisiae was performed at 30°C.

Anaerobic batch cultivation was performed in an instrumented bioreactor (Applikon Biotechnology, AC Schiedam, the Netherlands) with 1.5 L working volume and a starting OD_{620 nm} of 0.2. The medium contained 60 g/L of glucose or xylose and was prepared as described above with antifoam (Dow Corning, Midland, USA) added to the reactor at a final concentration of 0.2 mL/L. Temperature was maintained at 30°C and the pH was controlled at 5.5 through addition of 3 M KOH. Cultures were grown under aerobic conditions until the cell density reached OD_{620 nm} = 1.0, upon which conditions were changed to anaerobiosis for the remainder of the experiment (Figure 1). During the aerobic phase, the culture was sparged with air at a flow rate of 0.4 L/min and the stirring was set to 500 rpm. During the anaerobic phase, oxygen free conditions were maintained by nitrogen (> 99.995%) sparging at a flow rate of 0.2 L/min and the stirring rate was reduced to 200 rpm. Dissolved oxygen (DO) was monitored using a DO probe and CO₂ production was detected online by an INNOVA 1313 fermentation monitor (LumaSense Technologies, Ballerup, Denmark). Cultures were sampled for HPLC, OD_{620 nm} and dry cell weight measurements. Samples were collected for transcriptome analysis during exponential aerobic (OD_{620 nm}~1.0) and anaerobic growth (OD_{620 nm}~4.0). Transcriptome samples (50 mL) were collected from the bioreactor into a pre-chilled (-80°C) glass bottle. Samples were immediately centrifuged (5000*g, 2 min, 4°C) and the biomass pellet was frozen in liquid nitrogen. Experiments were performed in biological duplicate.

S. cerevisiae strain TMB 3415 was constructed from the TMB 3043 (Table 1) parent strain 13. This genetic background encompasses genetic changes previously identified as beneficial for xylose utilization. Genes for the non-oxidative pentose phosphate pathway 12 and xylulokinase XKS1 30 have been over-expressed, and the non-specific aldose reductase gene GRE3 has been deleted 31. Genes encoding Pichia stipitis mutated xylose reductase (XYL1 K270R) and xylitol dehydrogenase (XYL2) were integrated in TMB 3043 by transformation with YIpOB9 (Table 1) linearized with EcoRV, using the Lithium Acetate method 32. The resulting strain, TMB 3662, was rendered prototrophic by integration of the linearized (EcoRV) vector YIplac128 (Table 1), yielding strain TMB 3415. Standard molecular biology techniques were employed 33 and Fermentas GeneJet plasmid miniprep kit (Fermentas, Vilnius, Lithuania) was used for plasmid extraction.

Exponential anaerobic growth was confirmed by linear regression of the natural logarithm of cell concentration against time. Specific rates of product formation and substrate consumption were calculated in Matlab (Matlab R2007b, The MathWorks Inc., MA, USA) using Equation 1 & 2 and measured metabolite and cell concentrations. A pseudo-steady state assumption was validated by observing constant specific production- and consumption rates within 2-3 cell duplications, as well as good agreement of measured values between biological replicates.

Where: X = biomass (g/L); Met = metabolite concentration (g/L); μ = specific growth rate (h^-1); r = specific production rate (g/h×gDW).

Microarray analysis was performed on cell samples collected from aerobic and anaerobic batch cultivation on glucose and xylose as described above. RNA from two independent biological cultivations was analyzed. Total RNA was extracted from frozen cell pellets using a bead-beater (Biospecs products, Bartlesville, OK, USA) and Trizol reagent (Invitrogen, CA, USA). All extractions where performed on the same total amount of cells (approximately 10 mg dry weight). RNA was further purified using the RNeasy mini kit (Qiagen, Hilden, Germany). RNA quality and concentration were measured using an Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and Nanodrop ND-1000 (Thermo Fisher Scientific, Waltham, MA, USA) respectively. Total RNA was processed using the GeneChip^® Expression 3'-Amplification Reagents One-cycle cDNA synthesis kit (Affymetrix Inc, Santa Clara, CA, USA) to produce double-stranded cDNA. This was used as a template to generate biotin-targeted cRNA following the manufacturer's specifications. Fifteen μg of the biotin labeled cRNA was fragmented to strands between 35 and 200 bases in length, 10 μg of which was hybridised onto the GeneChip^® genome array overnight in the GeneChip^® Hybridisation oven6400 using standard procedures. The arrays were washed and then stained in a GeneChip^® Fluidics Station 450. Scanning was performed with the GeneChip^® Scanner 3000 and image analysis was performed using GeneChip^® Operating Software. The RMA algorithm 34 was used for normalization and scaling of the raw signal data. Student's t-test was used to identify genes with significantly (p ≤ 0.05) altered gene expression. All array data is presented as fold changes, i.e. the log2 ratio of expression signals.

Metabolite concentrations were determined by HPLC using a Waters HPLC system (Milford, MA, USA). An Aminex HPX-87H ion-exchange column (Bio-Rad, Hercules, CA, USA) and a refractive index detector (RID-6a, Shimadzu, Kyoto, Japan) were used for separation and detection, respectively. The column temperature was 45°C and 5 mM H₂SO₄ was used as a mobile phase at a flow rate of 0.6 mL/min. Cell dry weight was determined by filtering a known sample volume through a dried and pre-weighed 0.45-μm pore membrane (Pall Corporation, New York, USA), washing with distilled water and drying in a microwave oven for 8 min at 350 W.

TMB 3415 was cultivated in synthetic medium containing either glucose or xylose as the sole carbon source, under aerobic and anaerobic conditions. Aerobic growth was maintained for approximately 2 cell divisions, after which anaerobiosis was established by switching the sparging gas from air to nitrogen. The fermentation profiles of glucose and xylose growth in this setup are presented in Figure 1. Balanced exponential cell growth was seen for glucose and xylose under both aerobic and anaerobic conditions, allowing for the assumption of pseudo steady-state conditions 24. Growth rates and metabolic fluxes were calculated for the anaerobic growth phase (Table 2).

In glucose medium, growth rates were similar regardless of oxygenation (0.43 h^-1 vs. 0.33 h^-1) (Figure 1A) and comparable to previous reports 35. During xylose consumption on the other hand, the growth rate decreased from 0.20 h^-1 to 0.025 h^-1 under anaerobic conditions (Figure 1C). Specific rates of substrate uptake and metabolite production were also different during anaerobic glucose and xylose fermentation (Table 2). The substrate uptake rate and the ethanol production rate were approximately an order of magnitude lower during xylose utilization than during glucose utilization (Table 2). The productivity, 0.13 g/gDW×h, and yield, 0.43 g/g, of ethanol from xylose (Table 2) was however significantly higher compared to several recently investigated xylose-utilizing strains 20263637. It is reasonable to assume that anaerobic growth of strain TMB 3415 results from increased ethanol productivity and low xylitol yield, which in turn can be ascribed to the K270R mutation in XR 26. Compared to strains expressing a xylose isomerase based pathway, the ethanol productivity in TMB 3415 was higher than in non-growing strains 101138, whereas it was similar to a strain growing at μ = 0.03 h^-1 39 and lower than a strain growing at μ = 0.09 h^-1 40.

The present study aimed at highlighting regulatory differences between anaerobic glucose and xylose growth by xylose-utilizing S. cerevisiae. Specifically, differences in growth rate, substrate consumption and ethanol productivity were analyzed on a transcriptional level. Transcriptional characterization have not previously been performed under anaerobic conditions due to the inability of recombinant S. cerevisiae strains to grow on xylose in the absence of oxygen 17202122. A few strains expressing Piromyces XI are able to grow anaerobically on xylose 3940, however to the best of our knowledge transcription analysis of these strains has not been reported.

The overall structure of the microarray data was examined using Principle Component Analysis (PCA) 41. PCA has been used to reduce the dimensionality of microarray data and to identify features of experimental conditions that best explain the observed variance in gene expression 42. The projection of the two principle components, oxygen availability and carbon source, segregated the individual samples in a two dimensional space (Figure 2A). The four conditions, glucose aerobic (GA), xylose aerobic (XA), glucose anaerobic (GAnA) and xylose anaerobic (XAnA), were separated along the axis of the first and second principle component. The first principal component, which was responsible for the most variance in the data set, separated samples according to oxygen availability (aerobic or anaerobic) (Figure 2A). The second principal component separated samples according to carbon source (glucose or xylose). The PCA projection of the transcription data singled out anaerobic xylose growth as the most unique group in the data set (Figure 2A).

Next, the microarray data was organized into four relevant pairwise comparisons: GA vs. GAnA (Comparison 1 = C1); GA vs. XA (Comparison 2 = C2); XA vs. XAnA (Comparison 3 = C3) and GAnA vs. XAnA (Comparison 4 = C4) (Figure 2B). Comparisons C1 and C2 have previously been reported in slightly different experimental setups 172023. However, comparisons C3 and C4 allowed for the first time analysis of transcription during anaerobic growth on xylose as a sole carbon source. Globally, the transition from aerobic to anaerobic growth changed the expression level of more genes on xylose (C3, 809 genes) than on glucose (C1, 546 genes) (Figure 2B). This difference is presumably due to that aerobic metabolism on xylose is fully respiratory, while it is respiro-fermentative on glucose. Likewise a higher number of genes displayed different expression levels between glucose and xylose utilization under aerobic conditions (C2, 1059 genes) than under anaerobic conditions (C4, 641 genes) (Figure 2B).

Finally, subsets of genes were isolated according to the following criteria: (i) genes that changed expression levels between aerobic and anaerobic conditions regardless of carbon source (C1&C3, Figure 3A) and (ii) genes that changed expression levels on glucose and xylose regardless of oxygenation level (C2&C4, Figure 3B). Relatively few genes, 113, changed expression on both glucose and xylose during transition from aerobic to anaerobic conditions (Figure 3A). Likewise, only 130 genes changed expression level on xylose compared to glucose irrespective of oxygenation level (Figure 3B).

Within a group of genes, up- and down-regulated pathways and processes can be identified by searching for over-represented gene ontology (GO) terms http://www.geneontology.org/. Each annotated gene in the S. cerevisiae genome is associated with one or several GO terms that describe the corresponding biological process, e.g. amino acid synthesis. Significantly enriched gene ontology terms (p < 0.01) were identified within the previously described groups (C1, C2, C3, C4, C1&C3 and C2&C4) (Table 3). If more than one GO term in the same "family" were identified, only the most significant term was listed.

Comparing aerobic and anaerobic metabolism on glucose (C1, Figure 2B), unilateral down-regulation of respiratory processes was identified (Table 3). Down-regulation of respiratory genes in response to anaerobiosis has previously been reported for glucose grown cells in batch culture 17 and C- and N-limited chemostat culture 4243. In contrast, respiratory pathways were not down-regulated between aerobic and anaerobic xylose metabolism (C3, Figure 2B), whereas processes related to protein synthesis were significantly repressed (Table 3). Down-regulation of ribosome biogenesis and amino acid synthesis 4244 reflects the ten fold reduction of growth rate between aerobic and anaerobic conditions on xylose (Figure 1B). In addition, the GO terms "alcohol metabolism" and "hexose catabolism" were up-regulated under anaerobic conditions on xylose (Table 3). In the group of genes that changed expression in response to anaerobiosis irrespective of carbon source (C1&C3, Figure 3A), repression of cellular respiration was identified (Table 3).

Aerobic glucose metabolism was respiro-fermentative, while aerobic xylose metabolism was completely respiratory with absent ethanol production (Figure 1). On a transcription level, this difference was visible in higher expression of several respiration related GO terms on xylose compared to glucose during aerobic conditions (C2, Table 3) 1720. During anaerobic conditions, protein synthesis was down-regulated on xylose compared to glucose, while expression of respiratory processes and the GO term "hexose metabolism" was up-regulated (C4, Table 3). The lower expression of amino acid biosynthesis 4244 is related to the lower growth rate on xylose compared to glucose under anaerobic conditions. Oxidative phosphorylation was up-regulated on xylose in the group of genes that were differently expressed on glucose and xylose under both aerobic and anaerobic conditions (C4&C2, Table 3).

Respiration was not repressed between aerobic and anaerobic growth on xylose (C3, Table 3), and higher expression of respiratory processes was observed during both aerobic and anaerobic conditions compared to glucose (C4&C2, Table3). Certain processes in cellular respiration were however down-regulated both on glucose and xylose in response to anaerobiosis (C1&C3, Table 3). This indicates that although oxidative metabolism was not unilaterally down-regulated under anaerobic conditions, some processes were still repressed.

Expression of the GO term "hexose metabolism" increased specifically under anaerobiosis on xylose, both in comparison to anaerobic glucose utilization (C4) and to xylose under aerobic conditions (C3) (Table 3). This observation was dependent on anaerobic xylose growth and has thus not previously been reported. The GO term "hexose metabolism" encompasses several aspects of the central carbon metabolism (glycolysis, pentose phosphate pathway, gluconeogenesis etc.), which is responsible for the cell's energy metabolism. As such, regulation of central carbon metabolism is linked to growth rate, protein synthesis and ethanol production rate 4445. Several reaction steps in the central carbon metabolism are catalyzed by more than one isozyme, which enable the cell to regulate the direction/flux of the pathway in response to the energetic state of the cell. Therefore, expression of genes in the central carbon metabolism was investigated in greater detail to pinpoint differences between glucose and xylose anaerobic growth (Figure 4).

Increased reversed flux in glycolysis during anaerobic xylose utilization was indicated by expression of several isozymes specific for gluconeogenesis. The hexokinase gene HXK1 was expressed higher on xylose than on glucose irrespective of aeration (Figure 4). Previously, high HXK1 expression has been reported under aerobic xylose growth 1720 and during metabolism of non-fermentable carbon sources 46. Further down the pathway, the exclusively gluconeogenetic enzyme fructose-1,6-bisphosphatase FBP1 was up-regulated specifically during anaerobic xylose utilization (Figure 4). Expression of glyceraldehyde-3-phosphate dehydrogenase isozyme TDH1 is linked to stationary growth and gluconeogenesis 47 and was similarly up-regulated on xylose during anaerobic conditions (Figure 4). Also the minor isoform of phosphoglycerate mutase GPM2 was up-regulated on xylose during anaerobic conditions (Figure 4) but the function of this enzyme is largely unknown 48.

While reversed flux in glycolysis was indicated during anaerobic xylose growth, increased activity of the oxidative pentose phosphate pathway was observed through expression of 6-phosphogluconate dehydrogenase(GND2) and glucose-6-phosphate dehydrogenase (ZWF1) (Figure 4). Both GND2 and ZWF1 catalyze the reduction of NADP⁺ and were specifically up-regulated under anaerobic conditions on xylose (Figure 4). During anaerobic conditions, NADP⁺ reduction in the oxidative pentose phosphate pathway controls the rate of xylose utilization as a consequence of the cofactor imbalance in the XR-XDH reaction 49. Increased flux in the oxidative pentose phosphate pathway (PPP) must however be supported by proportionally increased reversed flux in glycolysis, which indeed was seen in the current study.