3.2. Applying the systemic approach for improving ethanol production of S. cerevisiae
To find candidate reactions for S. cerevisiae , the correlated reactions with ethanol production at pH=5 were identified using the Pearson correlation coefficient (between fluxes of each reaction and ethanol production) in the range of proton exchange rate between -2 and 23 mmolgDCW-1h-1. This range was determined based on robustness analysis (Figure 3b), considering that the metabolic model of pH=5 was robust for growth compared to the two other models and maximum ethanol is produced at optimal growth in this range (Figure S8). So, flux distributions at optimal growth in the selected range were determined and correlated reactions with ethanol production were determined (Supplementary File 2). These 252 reactions are key reactions that indicate the manner of using metabolism for maximal ethanol production during optimal growth. 67 reactions with positive coefficient are candidates for up-regulation and 185 with negative coefficient are predicted for down-regulation. The Pearson correlation coefficients between these reactions and growth indicate that the correlated reactions with ethanol production are non-growth associated and vice versa (Supplementary File 2). So, the correlation coefficients predict that up and down-regulation of candidate reactions for overproduction of ethanol results in growth reduction(Naghshbandi et al., 2019; Pagliardini et al., 2013).
For more screening of the key reactions and determining the important genes affected by the pH change, PCA was performed using fluxes of the key reactions at pH levels of 5, 6 and 7. Figure 4a illustrated a clear distinction between pH=5, and the other three pH models were achieved based on the first PC. Only the first two components have been considered as they demonstrate the high percentage of variation between pH=5 and other pH levels: 90.18% for PC1 and PC2 (Figure 4b). The decomposition of data by PCA indicates the contribution of each correlated metabolic reaction to the differentiation of models at various pH values. Among the 252 key reactions, 12 reactions including ILETA, ME2m, PDHm, ICDHyr, CSm, ACONTm, DESAT16, NDPK1, THRD_L, ICL, AGTi, and MDH were essential for the discrimination of pH models based on the PCA results presented in Figure 4c. These reactions were selected to evaluate the effects of regulators on their enzymes on ethanol production. Table S6 provides information for these 12 reactions in detail.
It can be seen that various approaches have been suggested by the systemic approach for ethanol overproduction. Figure 5 shows the comprehensive connection of each proposed reaction, which is suitable for up-regulation or down-regulation. For instance, through observing the model reactions, mitochondrial pyruvate is not capable of converting to acetaldehyde and must enter the cytosol for conversion to ethanol. Thus, according to the PDHm reaction (Table S7), coenzyme A reacts with the mitochondrial pyruvate to produce acetyl coenzyme A. The product of this reaction is then converted to citrate by the CSm reaction and the produced citrate is responsible for providing isocitrate via ACONTm reaction. Cytosolic isocitrate is generated by mitochondrial isocitrate from the CITtcm and then converted to 2-oxoglutarate by the ICDHyr reaction. This metabolite is further produced to cytosolic pyruvate by the ALATA_L reaction, which leads to acetaldehyde by PYRDC. Eventually, acetaldehyde is reduced to ethanol by the ALCD2ir.
The predicted redirecting flux to the TCA cycle under acidic conditions is expected. Isocitrate dehydrogenase (ICDHyr), citrate synthase (CSm), Aconitate hydratase (ACONTm), which were ethanol-associated (or non-growth associated) reactions and Malic enzyme NADP mitochondrial (ME2m), with growth correlated coefficient (or non-ethanol associated), were the identified candidate reactions in TCA. The predictions of the systemic approach were in compliance with fluxes determined by experimental data, where Blank et al.(Blank & Sauer, 2004) showed that malic enzyme and mitochondria were more active under acidic conditions, as was clear from metabolic reactions located at crucial branch points in central metabolism. Furthermore, previous studies have supported the predicted effects of some reactions. These were the bottleneck reactions for both growth and ethanol production at pH=5. Isocitrate dehydrogenase (ICDHyr) reaction, which is encoded by IDP2, converts isocitrate to α-ketoglutarate, Scalcinati et al.(Scalcinati et al., 2012) showed that overexpression of IDP2 has resulted in an increased NADPH production to supply the required energy for biomass and yeast growth. The alanine glyoxalate aminotransferase (AGTi) reaction, which encoded by AGX1, converts glyoxalate and alanine to glycine and pyruvate. Chidi et al.(Chidi, Rossouw, & Bauer, 2016) experimentally confirmed that the deletion of AGX1 increased pyruvic acid, which led to improved ethanol production.