In silico analysis of triacylglycerol production on different substrates

With glucose as the sole carbon and energy source, the computed optimum flux distribution is shown in Figure 1. The most common product found in 94% of the modes is cell mass whereas TAG production is found in 25% of the modes. Maximum theoretical yield of the TAG is found to be 0.07 mol.mol⁻¹ on the basis of an assumed TAG molecule size of 51 carbons (3X C₁₆ palmitic acid + glycerol 3-phosphate). Taking the molecular weight of the assumed TAG (C₅₁H₉₂O₆ = 801.27 g.mol⁻¹) into account the resulting maximum yield is 0.30 g.g⁻¹ glucose. For the production of one mol of TAG, 42 moles of NADPH and 21 moles of ATP are required. From the flux distribution map (Figure 1), 63% of the required NADPH is mainly supplied by the pentose phosphate pathway (PP) and partially by the cytosolic malic enzyme (values normalized to glucose uptake rate as shown in Figure 1). This high flux through the PP pathway might lead to carbon loss in form of CO₂, but also for a better precursor availability and growth. In fact, 37.2 mol CO₂ would be released for the synthesis of one mol TAG. Neither the malic enzyme in the cytosol nor the PP pathway alone can compensate for the required NADPH. The flux through the malic enzyme depends on the cytosolic NADH availability as the malic enzyme functions together with the pyruvate-oxaloacetate-malate cycle (POM cycle). About 10% ATP is generated by the respiratory chain reaction using the mitochondrial NADH.

On the other hand, with glycerol as the sole substrate (Figure 2), the NADPH demand is mainly met by the cytosolic malic enzyme whereas there is a reduced PP pathway flux and hence reduced CO₂ formation. Glycerol is channeled into glycolysis via the dihydroxyacetone to glyceraldehydes 3-phosphate. Most of the fluxes pass through the lower part of the glycolysis into the pyruvate dehydrogenase complex. Two more optimal modes (data not shown) were identified where similar maximum TAG yields were observed with a succinate to fumarate interconversion in the mitochondrion and cytoplasm. The required ATP is also generated by both the respiratory chain reactions in the mitochondrion.

Similarly, the optimal flux distributions with xylose and arabinose as substrates are shown in Additional file 1: Figures S2 and S3, respectively in the Supplement. Higher numbers of EMs were observed on arabinose compared to xylose. About 96% of the modes produced cell mass and 23% of the modes showed TAG production on both the pentoses. A very high PP pathway fluxes is observed on both pentoses, obviously due to the requirement of NADPH for uptake of these substrates [24]. Xylose showed slightly higher theoretical yields of TAG than arabinose. Most of the fluxes are channeled through the gluconeogenic glucose 6-phosphate isomerase reaction into the PP pathway on both the pentoses. Similar channeling of fluxes is observed towards the pyruvate dehydrogenase complex on both the substrates. Due to the high demand for NADPH on arabinose, the oxaloacetate transporter flux into the mitochondrion should be inactive whereas on all the other substrates it should be active for maximal TAG synthesis. This accumulation of oxaloacetate may drive a high flux through the malate dehydrogenase (MDH) in the cytosol towards malate which in turn also increases the flux through the cytolsolic malic enzyme on arabinose than on xylose.

Using the same method, optimal flux distributions and EFM analysis with the proposed metabolic network of R. toruloides were performed using blend of substrates. The maximum TAG yield obtained is 0.58 C_mol. C_mol ⁻¹ substrate with a blend of glucose and glycerol (Figure 3), compared to 0.55 C_mol. C_mol ⁻¹ substrate with a blend of xylose and glycerol (Additional file 1: Figure S4). The represented modes are the best modes consuming both the substrates. Comparing the yields from these analyses, similar maximum TAG yields are observed compared to the yields where single sugars are used (Table 1). Adding glycerol uptake did not show any effect on maximum TAG yields in these in silico analyses. CO₂ produced per mole TAG on different carbon sources are compared from the in silico analyses, arabinose as a carbon source has the highest production of CO₂ per TAG molecule (45.6 mol/mol TAG), whereas glycerol as a sole carbon source has the lowest CO₂ production (30.7 mol/mol TAG), indicating that glycerol is a desirable substrate in terms of efficiency of carbon utilization. CO₂ production is proportional to the flux entering into the PP pathway as clearly observed from the in silico predictions.

Controlled bioreactor cultivations of R. toruloides and the utilization efficiency of different substrates

To compare the in silico predictions regarding the use of different substrates an initial set of cultivations was performed using glucose and glycerol as the sole carbon and energy source. The time profiles of the cultivations are shown in Figure 4. Typically, in the medium used, and with a fixed aeration rate of 0.5 vvm, four distinct culture phases were observed for both substrates: a growth phase under excess of both nutrient and oxygen (10 h or 12 h for growth on glucose and glycerol, respectively), a growth phase under oxygen-limited condition (last for about 10 and 20 h with glycerol and glucose respectively), and a phase with sufficient oxygen supply, presumably due to limitation by other nutrient(s) (which continued for almost 30 h or 40 h for glycerol and glucose respectively), and lastly a phase of oxygen limitation again which continued till the end of the experiment.

ATP requirements for growth, lipid production and maintenance are supplied mainly through the TCA cycle in the mitochondria. In fact, our in silico analysis was all done with a fixed P:O ratio of 1.2. The effects of oxidative phosphorylation (with varied P:O ratio) on TAG yields using glucose as a substrate were analyzed using the proposed model. The change in P:O ratios (from 1.2 to 2) neither showed significant effect on the TAG yields (from 0.58 to 0.60 C_mol.C_mol ⁻¹ glucose), nor cell mass yield (0.67 to 0.71 C_mol.C_mol ⁻¹ glucose), respectively. Hence, it can be concluded that ATP generated by oxidative phosphorylation is not a major limiting factor for TAG synthesis. In agreement with the in silico analysis, controlling the oxygen concentration at 50% air saturation in the fermentation broth of R. toruloides grown either on glucose or glycerol throughout the cultivation process did not have any significant effect on the yield (Table 2). Similar to the cultivations without pO₂ control (e.g. O2 limitation in phases 2 and 4, see Figure 4), the maximal growth rate with glucose was 0.23 h⁻¹ compared to 0.13 h⁻¹ with glycerol in pO₂ controlled cultivations. The maximum lipid content on glycerol was 61% vs. 57% in pO₂ uncontrolled culture, whereas it was 48% on glucose vs. 41% in pO₂ uncontrolled culture. The lipid productivity increased significantly in the pO₂ controlled experiments, mainly because of the increased growth rates.

Blends of sugars and glycerol were also tested for the production of lipid by R. toruloides. Interestingly, this strain showed a catabolite repression in the presence of glucose as depicted in Figure 5a. Catabolite repression was not observed before in certain oleaginous yeasts. Strains of Y. lipolytica and R. glutinis were shown previously to metabolize both carbon sources simultaneously [23,25,26].

As shown in Figure 5a, a clear diauxic growth was observed in the mixed substrate cultivations: glycerol was consumed only after glucose consumption in this case. Despite of the observed catabolite repression, the addition of glycerol enhanced the lipid yield on glucose significantly (Figure 5b), confirming the predictions from the in silico analysis. In fact, the mixed substrate cultivation has the dual advantages of efficient cell mass production (normally maximized with glucose) with higher lipid content (obtained normally with glycerol). The lipid yields on the substrate consumed and the specific lipid productivities on cell mass are higher when glucose and glycerol are used in mixture rather than only glucose (Table 2).

In the above blend substrate fermentation experiments, glucose was used at a concentration of 10% of the added carbon source. The effect of different combinations of the substrates on the lipid yield was further investigated in flask experiments. As compared to mono-substrate fermentation with glucose, a 26% increase of lipid yield was observed if glycerol is added (Figure 6). Similarly to the above results from controlled bioreactors, glucose as mono-substrate supported the highest cell mass production (13.7 g.L⁻¹) with the lowest lipid content (32%). Similarly, with a closely related yeast strain R. glutinis, Easterling [23] reported that blend substrate fermentation produced more cell mass than when a single carbon source was provided. They also reported that the lipid content was minimum with glucose and significantly higher with either glycerol or blend of both. However, they did not report catabolite repression for R. glutinis. However, it should be stressed that, on mixtures of carbohydrates and with the sequential utilization of sugars, the sugar is being consumed with more nitrogen available (and hence more cell mass formation), followed by glycerol consumption during a period of nitrogen restricted growth which will have a positive effect on the lipid yield. Still, the mechanism of the positive role of glycerol for the lipid synthesis of cells grown in blends deserves more detailed studies. Glycerol can be utilized as a source of carbon and energy. But under certain conditions (e.g. in pO2-controlled culture with glucose as the sole carbon source), R. toruloides can produce glycerol extracellularly up to 5 g/L (data not shown). The two most important functions of glycerol synthesis in yeast are related to redox balancing and the hyperosmotic stress response [27]. In Saccharomyces cerevisiae, production of glycerol occurs to maintain the redox balance. Interestingly, glycerol production is found to be higher in a minimal medium than in a complex medium [28]. It was concluded that in a minimal medium the de novo synthesis of amino acids from glucose and ammonia leads to an excess of NADH which is re-oxidized by glycerol formation. In osmotolerant yeasts, glycerol plays a crucial role in their survival [29]. Recently, Petelenz-kurdziel et al. [30] reported that S. cerevisiae adapts to hyperosmotic stress by activating the HOG signaling cascade, which controls glycerol accumulation, production and import. Moreover, it was also shown that glycerol transport depends on the strain and the growth conditions. Glycerol production has been never reported in R. toruloides, and hence transcriptomic analysis is being further performed in our laboratory.

The effects of different growth substrates and culture conditions on the fatty acid composition were also studied (Figure 7). In oxygen controlled conditions at pO₂ of 50% of air saturation, and on all the substrates and blends used, palmitic, oleic and linoleic acid constitute more than 90% of the total fatty acids. In oxygen un-controlled cultivation (constant aeration and agitation, pO₂ not controlled), about 65% of the fatty acid measured is unsaturated (oleic acid and linoleic acid) if glycerol or blend are used, compared to 57% of saturated fatty acids (lauric acid, stearic acid and palmitic acid) if glucose is the sole carbon source. Oxygen limitation has significantly influenced the fraction of saturated and un-saturated fatty acid composition in the lipid produced by R.toruloides especially on glycerol or mixed substrates. Oxygen controlled conditions derived fatty acid compositions that are suitable for biodiesel conversion [9] whereas oxygen un-controlled conditions produced significant amounts of saturated fatty acids which are close to be substitutes of certain important products like cocoa butter.

R. toruloides is a known lipogenic yeast and the results presented here confirm its ability to accumulate over 60% of its cell mass as lipid. However, lipid accumulation was strongly affected by the nature of the carbon source provided. The recently sequenced and annotated genome of R. toruloides, its proteomic analysis and the available protein annotation have opened doors to understand the underlying mechanism of this organism for lipid production [31-33]. In this work, elementary modes were used to decompose the complex metabolic network into its basic functioning units and to identify the potentials of triacyglyceride production on different substrates in R. toruloides. Maximum theoretical TAG yields on hexoses (glucose), pentose (xylose and arabionse), glycerol and on mixture of hexoses & glycerol, pentose & glycerol have been estimated (Table 1) and were in general agreement with previously reported theoretical maximum yields [34,35]. As shown from the computed solution span in Additional file 1: Figure S1, the production is still far from the theoretical optimum and the distance between the actual production state on glucose or glycerol and the theoretical optimum suggests an enormous potential for future optimization. In silico optimization, identification of limiting steps and finally suggestions for future genetic modification are discussed below.

One of the key processes for fatty acid biosynthesis is the provision of reducing power (e.g. NADPH) to reduce the acetyl group and channel it into the acyl chain of fatty acid. In Y. lipolytica, the PP pathway is assumed to be the sole supplier of NADPH for lipid synthesis, since it does not possess a cytosolic malic enzyme [36] and the deletion of mitochondrial malic enzyme did not show any effect on lipid production [37]. In R. toruloides, an alternative NADPH supplying reaction would be the cytosolic NADP-dependent isocitrate dehydrogenase (ICDH) which is inactive in the predicted optimal flux distribution of glucose (Figure 1). A single gene deletion study of our model was performed to remove the cytosolic malic enzyme from the reactions. A reduction in maximum theoretical yield of TAG was observed (from 0.30 g.g⁻¹ to 0.27 g.g⁻¹) on glucose and glycerol (0.32 g.g⁻¹ to 0.27 g.g⁻¹). However, the theoretical maximum yield retained to 0.30 g.g⁻¹ on glucose and 0.32 g.g⁻¹ on glycerol if a NADP-dependent acetaldehyde dehydrogenase (ALD6) is incorporated. This enzyme constitutively catalyzes the conversion of acetaldehyde to acetate, generating an NADPH in the cytoplasm. Another alternate route for supplying NADPH as proposed by Ratledge [34] can be a mitochondrial NAD-dependent ICDH acting in the reverse direction. EFM analysis and optimal flux distribution were done (Additional file 1: Figure S5) and the maximum theoretical yield of TAG on glucose is retained to 0.30 g.g⁻¹ glucose. A third possible target could be a heterologous NADPH generating enzyme or an endogenous engineered NADPH generating enzyme, both in the glycolytic pathway [38] which could be a potential source for NADPH. By changing GAPDH, which is a NAD-dependent glyceraldehyde dehydrogenase in the glycolytic pathway, to an NADP-dependent enzyme can increase the theoretical TAG yields by 7% on glucose and 9% on glycerol.