We investigated the molecular processes involved in the maintenance of intracellular NADPH homeostasis, by studying the fluxomic, metabolic and transcriptional responses to increases in NADPH oxidation in S. cerevisiae. The biological system used, which was based on the overproduction of an NADPH-dependent Bdh1p enzyme and the use of acetoin as an electron acceptor, made it possible to impose unprecedented levels of redox disturbance on yeast cells, corresponding to 22 times the anabolic demand for NADPH .
We previously showed using constraint based metabolic flux analysis (dMBA for dynamic Mass Balance Analysis, ) that (i) yeast responds to increases in NADPH demand by increasing the flux through the two main NADPH-producing pathways (the PP and acetate pathways) (ii) for NADPH demands of more than 22 times the anabolic demand, the PP pathway was saturated and additional NADPH was produced through the glycerol-DHAP cycle, involving exchanges of the NADH and NADPH redox cofactors. In this study, we characterized the mechanisms involved in flux rerouting for two moderate (100 and 200 mM acetoin) and one large (300 mM) increase in NADPH, corresponding to 8, 12 and 22 times the anabolic NADPH demand, respectively. In these conditions, little effect on growth was observed only for the two highest levels of NADPH demand, consistent with a decrease in biomass synthesis due to a lack of NADPH.
Regulatory mechanisms involved in the rerouting of carbon to NADPH-producing pathways
We combined 13 C-based metabolic flux, intracellular metabolite and global gene expression analyses, to investigate the way in which yeast adjusts its metabolism to respond to a high NADPH demand. We first confirmed experimentally that carbon was rerouted principally to the acetate and the PP pathways for the generation of NADPH. Flux through the PP pathway, which was estimated by 13 C-flux analysis, was increased by factors of 1.6, 2.4 and 2.9 in the presence of 100, 200 and 300 mM acetoin, respectively, consistent with the predictions of the DynamoYeast model .
We showed that the response of yeast to an increase in NADPH demand to eight times the anabolic demand was mediated mostly by metabolic control, as no transcriptional regulation of the genes involved in the acetate and PP pathways was observed. However, further increases in NADPH demand resulted in the induction of several PP pathway genes: SOL3 and GND1 in the presence of 200 mM and 300 mM acetoin and, in the presence of 300 mM acetoin, another two genes, TAL1 and TKL1. These genes were probably upregulated to increase the flux capacity of the PP pathway, which was limited in these conditions. Indeed, the accumulation of 6-phosphogluconate and of all intermediates of the nonoxidative part of the PP pathway (Figure6) suggests that the activities of 6-phosphogluconolactonase or 6-phosphogluconate dehydrogenase, catalyzed by Sol3p and Gnd1p, and of enzymes of the nonoxidative part of the pathway become limiting.
Both metabolic and genetic controls are therefore required to cover a demand for NADPH more than eight times greater than the anabolic demand for this compound.
GND1 upregulation has been observed in various situations in which the flux through the PP pathway increases [20–22]. This is not surprising, because this gene encodes a protein that catalyzes an irreversible reaction. By contrast, ZWF1, which encodes a protein that catalyzes the first irreversible NADPH-producing reaction of this pathway, was not transcriptionally regulated. These findings may be accounted for by differences in the cost of the proteins encoded by these genes. Indeed, a recent study in Escherichia coli suggested that the production of enzymes with a high cost (estimated from their abundance and molecular weight) is more tightly controlled than that of enzymes with a lower cost . If the same is true in yeast, it would account for the transcriptional regulation of GND1 rather than ZWF1, because the cost of Gnd1p is 10 times that of Zwf1p .
We initially hypothesized that the transcription factor Stb5p might be involved in the maintenance of redox homeostasis under anaerobic conditions. However, only eight of the 69 Stb5p target genes  (including 4 PP pathway genes) displayed differential regulation (Table 1, Additional file 1), consistent with Stb5p playing no role in the redirection of flux observed in response to increases in NADPH oxidation. This low proportion of Stb5p target genes responding to NADPH/NADP+ imbalance suggests that Stb5p does not sense NADPH changes directly during oxidative stress.
The repression of genes involved in the synthesis of reserve carbohydrates under conditions of the greatest NADPH demand might optimize NADPH synthesis by rerouting carbons towards the PP pathway. The same may also be true for the repression of genes involved in glycolysis and alcoholic fermentation. However, as pyruvate decarboxylase is also used to generate acetoin from acetaldehyde, the downregulation of PDC1 and PDC5 may result from disturbance of the acetaldehyde node. By contrast, PDC5 was upregulated in response to altered NADH demand. Whether this may have resulted from the accumulation of acetadehyde in this condition (Table 3) remains to be explored.
Indirect effects of changes in NADPH demand on amino-acid synthesis
The other main changes in transcription concerned the synthesis of amino acids, particularly the NADPH-consuming methionine and lysine pathways and serine synthesis, which is connected to the methionine pathway.
MET genes expression is normally activated by the Met4p transcription factor when the levels of these sulfur-containing compounds are low [25, 26]. Sulfate assimilation requires considerable amounts of NADPH (7 moles of NADPH are oxidized to NADP+ to generate 1 mole of methionine from 1 mole of SO4
2−). Our results therefore suggest that decreases in NADPH availability limit the flux through this pathway, resulting in low levels of sulfur-containing compound production and a global derepression of the sulfate assimilation pathway. These findings are consistent with the coordinated regulation of this pathway, which is controlled principally through feedback repression exerted primarily by cysteine . The relationship between the PP phosphate pathway and the sulfur pathway has long been known  and is illustrated by the sensitivity to osmotic stress of a zwf1 mutant. The connection between these two pathways does not seem to be based on transcriptional control. Indeed, no common regulator of the genes of these two pathways has been identified except Met4p, the main activator of the sulfur pathway, which can also bind GND2, the minor isoform of the 6-phosphogluconate dehydrogenase .
The upregulation of the LYS12
LYS9 and LYS2 genes suggests that the same mechanism may operate when lysine production decreases. On the other hand, SER3 and SER33 encoding the first enzyme of the serine synthesis pathway, were also upregulated despite the description of this pathway as NADH-dependent . However, as the biosynthetic reactions involving 3-phosphoglycerate, including those contributing to serine synthesis, constitute a highly branched network connecting the purine, thiamine, histidine and methionine biosynthesis pathways, the upregulation of these genes may be an indirect consequence of a decrease in the intracellular concentration of cysteine.
In addition, the repression of several genes linked to the sulfate assimilation pathway (SAM3, TPO1, QDR2) may reflect a need for the cells to conserve methionine, as the synthesis of this compound is probably compromised by the higher demand of NADPH.
Effects of redox alteration on the synthesis of aroma compounds
We observed distinct effects of redox alteration on the production of fermentative aroma compounds, depending on the cofactor (NADH or NADPH) and on the class of aroma compounds. Indeed, increases in the oxidation of NADH and NADPH decreased the production of ethyl esters but did not alter the production of acetate esters. Changes in esters profile could reflect a decrease in the availability of acetyl-CoA , a precursor of these molecules, as suggested by the lower level of flux through acetyl-CoA synthase (Table 2). This suggests that the alcohol-O-acetyltransferase responsible for catalyzing this reaction may be less sensitive to acetyl-CoA concentration than the esterases involved in ethyl ester synthesis.
On the other hand, although the synthesis of phenyl ethanol increased in response to altered NADPH demand, consistent with increased flux through the PP pathway, the synthesis of other fusel alcohols (isobutanol, propanol) was specifically increased in response to NADH alterations. Isobutanol is synthesized from valine, which is produced from α-acetolactate, a compound also involved in acetoin synthesis. We suggest that the synthesis of acetoin by the PDC route is favored in conditions of acetaldehyde accumulation (as described by Heux et al. ), and that α-acetolactate is rerouted towards the production of valine, isobutanol and isobutyl acetate. A key mechanism underlying this remodeling of flux may be the increase in PDC5 gene expression observed in response to NADH oxidation (Table 1). PDC5 encodes a protein involved in the decarboxylation step of the Erhlich pathway. The greater production of butanol and propanol may therefore be explained by increased PDC5 expression.
The energy problem and the glycerol-DHA cycle
Analysis of the metabolic response to an NADPH demand 22 times greater than anabolic demand revealed a metabolic shift that we attributed to saturation of the PP pathway . This conclusion is consistent with the intracellular metabolite profiles obtained in this study. Moreover, a previous constraint-based metabolic flux analysis suggested that a glycerol-DHA cycle exchanging NADH for NADPH was activated in these conditions . The intracellular metabolite profile revealed energetic inconsistencies (Figure6), consistent with the hypothetical operation of this cycle. Indeed, the ATP and UTP pools were markedly smaller in the presence of 300 mM acetoin than in the presence of the other two concentrations of this compound. AXP concentration is one of the major factors controlling the regulation of the glycolysis and carbohydrate pathways [14, 33]. The accumulation of glycolytic intermediates may result from a decrease in phosphofructokinase activity due to a decrease in ATP availability or from an increase in glycolytic flux. An increase in glycolytic flux was actually previously observed  in the presence of 300 mM acetoin (25.7 mmol/gDW/h instead of 19.5 mmol/gDW/h for lower −100 and 200 mM- acetoin concentrations).
Flux toward acetaldehyde, via pyruvate decarboxylase, increased, at the expense of the reactions supplying the TCA pathway. The NADPH imbalance results in the rerouting of the carbon atoms of acetaldehyde for the production of acetate via Ald6p. The marked accumulation of acetate when the cells were incubated in the presence of 300 mM acetoin (Figure2B) probably resulted from a lower level of flux through acetyl-CoA synthase, resulting from both ATP limitation and a smaller acetyl-CoA requirement for lipid biosynthesis. Consistent with the limited production of acetyl-CoA, the production of acetate esters and fatty acids from this precursor decreased in response to increases in NADPH demand.