Pichia stipitis possesses the highest native ability of any yeast to metabolize xylose , and is therefore a key candidate for ethanol production from biomass, as well as for engineering xylose metabolism in Saccharomyces cerevisiae [2–4]. Xylose consumption requires two additional reactions that are catalyzed by xylose reductase (XYL1, Xyl1p) and xylitol dehydrogenase (XYL2, Xyl2p). Cofactor requirements for these reactions affect the oxygen demand of cells . Xyl1p has a higher affinity for NADPH whereas Xyl2p prefers NAD+, hence the formed NADH cannot be properly recycled at oxygen limited conditions . Therefore, the efficient conversion of xylose to ethanol or biomass occurs under defined aerobic conditions. P. stipitis produces ethanol at yields close to the maximum at low oxygen transfer rates (~2 mM h-1), but xylose uptake rate is only half of the maximum attained under fully respiratory conditions . Attempts to increase the rate of xylose consumption under oxygen-limited conditions have been only partially successful, since the engineered cells were unable to use xylose alone or had reduced ethanol production [6, 7]. Due to its advantages in the production of ethanol under anaerobic conditions, while resisting very high concentrations of ethanol, there has been much focus on converting S. cerevisiae into a xylose fermenter. However, the expression of target genes from P. stipitis into S. cerevisiae has not been quite successful due to the difficulties in regulating redox balances during xylose consumption and ethanol conversion [3, 4]. Furthermore, the production of ethanol is differently regulated in the two yeasts. Whereas S. cerevisiae can produce ethanol via aerobic fermentation (the Crabtree effect) or anaerobically, especially in the presence of high concentrations of glucose , P. stipitis does it mainly in response to oxygen limitations (Pasteur effect) [5, 8, 9].
Pichia pastoris belongs to a small group of microorganisms capable of catabolizing methanol and fatty acids (e.g. oleic acid) as the sole carbon and energy source [10, 11]. It can up-regulate the expression of crucial genes (e.g. alcohol oxidase, AOX; and the multifunctional beta-oxidation protein, FOX2), and multiply peroxisomes when it is growing on such carbon sources [11, 12]. P. pastoris has therefore been used extensively for the production of recombinant proteins using the AOX gene promoter, as well as a model for studying peroxisome proliferation [11, 13]. Consequently, a number of genetic tools and cultivation methods have been developed . Cultivation techniques include fed-batch culture with glycerol, followed by induction with methanol alone, or in combination with glycerol or sorbitol which has shown to be useful for increasing the production of recombinant proteins [14–16]. The success of engineering the glycosylation pathway of P. pastoris to produce sialylated glycoproteins has increased the expectations for its use to produce pharmaceutically relevant proteins, as well as to transfer the glycosylation technology into S. cerevisiae .
With the availability of the complete genome sequences for P. stipitis and P. pastoris [19, 20], there is the opportunity to study, at the system level, the native potentials of these yeasts. Genome-scale metabolic models (GEMs) can be used to support such a task since they can be used to assess metabolic capabilities of cells, to analyze metabolites connectivity and pathway redundancy, or for comparing metabolic capabilities between closely related species . GEMs can also be used to predict genotypic-phenotypic relationships , and for the identification of metabolic engineering targets . Moreover, by incorporating 'omics' data and in silico methods, GEMs can act as scaffolds for the design of optimal metabolic fluxes [24–26], or to evaluate the correlation between gene expression and metabolic changes in response to environmental perturbations [22, 27, 28].
Thus far, two GEMs of P. pastoris have been published [29, 30]. However, neither one incorporates and evaluates methanol metabolism and protein production together, which are among the most important features of this yeast. We also show that our model better predicts the growth phenotype on methanol alone or in combination with glycerol. In this manuscript we report an extensive GEM for P. pastoris (iLC915), together with the first GEM for P. stipitis (iSS884). Both GEMs correctly represented available experimental phenotypes. The iSS884 model, for example, can predict xylose consumption and distribution into biomass and ethanol using different oxygen up-take rates. The iLC915 model can closely simulate the physiological differences of P. pastoris growing at different carbon sources alone (including oleic acid and methanol), as well as methanol mixed with glycerol or sorbitol. iLC915 also includes reactions for the production of a model recombinant protein, for which the production under different oxygen uptake rates compared well with experimental results. The use of such networks on the design of fermentation technologies is also discussed.