Complex phenotypes, such as strain tolerance to toxic compounds, are difficult to be rationally engineered due to the limited knowledge about molecular mechanism. Adaptation is a frequent method to gain insight into strain response to a specific stress condition [18]. In our work, to figure out the potential factors relevant to strain tolerance to combined inhibitors FAP, an adaptation experiment was carried out and thoroughly investigated by the metabolomic analysis. As shown in Fig. 2a, yeast cells were first cultivated in FAP-containing medium till stationary phase, and then an aliquot of the culture was transferred to fresh FAP-containing medium twice for additional rounds of growth. As the results described previously [19], in FAP-free medium, the cells in G0 entered the exponential phase after a transient lag phase. After the addition of FAP, the lag phase of cells in G1 was extended to 39 h, the fermentation time was delayed to 74 h from 40 h, and the final OD600 was reduced to 4.37 from 9.05 compared to cells of G0 in FAP-free medium. After transferring the cultures of G1 to next round, the cells of G2 rapidly adapted to combined inhibitors and started to grow only after 4 h in the lag phase [19]. The growth of cells in G3 was further slightly improved [19]. The glucose consumption was in consistence with cell growth and the final ethanol production was almost the same in all cultures. The duration of the lag phase could be interpreted as a measurement of varied tolerance to inhibitors [20, 21]. Meanwhile, the prolonged lag phase in adverse condition reflected a physiological shift of cells to adapt environmental stress. Therefore, samples for comparatively metabolomic study were collected at the lag phase of G0, G1, G2 and G3 (Fig. 2a). As shown in Additional file 1: Table S1, 70 putative intracellular metabolites were identified and quantified.
The discovery of biomarkers associated with yeast tolerance to FAP by PLS-DA, mRMR and metabolic pathway analysis
For understanding and interpreting the adaptation model, the discovery of biomarkers is critical in the process of metabolomic analysis. Thus, appropriate statistical tools are essential for mining the huge data information. In this work, PLS-DA, mRMR and metabolic pathway analysis were carried out together to analyze our metabolite datasets. In the PLS-DA score plot (Fig. 2b), the first two predictive components explained 84.49 % of the total variance, while the first one accounted for almost 58.41 % alone. The clustering results revealed three major groups. The samples from G2 and G3 were clustered together, which were separated from G0 and G1 clearly, indicating the adjustment of intracellular metabolism of the yeast strain in response to FAP stimuli. In the PLS-DA loading plot (Fig. 2b), the intermediates of central carbon metabolism (glycerol and citrate), amino acids (glutamate, lysine, ornithine, 5-oxo-proline, alanine, aspartate, proline, glycine and GABA) and myo-inositol were identified as the most contributive metabolites in the separation of three groups, which were postulated to be important for strain FAP tolerance. In the results of mRMR analysis, the top 12 metabolites included the intermediates of central carbon metabolism (citrate and glycerol), amino acids (glutamate, lysine, ornithine, 5-oxo-proline, alanine, aspartate, proline, glycine and GABA) and adenine, which were all markedly affected in yeast cells during the adaptation to FAP (Fig. 2c). Among them, 11 of 12 biomarkers in mRMR were overlapped with those found in PLS-DA.
To further study the potential metabolic pathways involved in the yeast tolerance to FAP, the metabolic pathway analysis was performed using MetPA. The detected metabolites were correlated to 42 metabolic pathways, for instance, including pyrimidine metabolism, glutathione metabolism, and arginine and proline metabolism. It is well established that changes in the important positions of a network could trigger a more severe impact on the pathway than that in the marginal or relatively isolated positions [22]. Thus, only the out-degree for node importance measurements was considered here. The impact-value threshold was set to 0.40, above which the pathway was considered as the potential target pathway. Hence, we screened out three unique pathways potentially responsible for the improved inhibitor tolerance, including alanine, aspartate and glutamate metabolism, arginine and proline metabolism and glycerolipid metabolism (Fig. 2d). Meanwhile, the alanine, aspartate and glutamate metabolism and the arginine and proline metabolism were also significant in pathway enrichment analysis (Fig. 2d). Interestingly, the biomarkers alanine, aspartate, glutamate, GABA, and citrate located in the key positions in the alanine, aspartate and glutamate metabolic pathways, while the biomarkers proline, glutamate and ornithine were in the arginine and proline metabolic pathway. Taken together, we determined that alanine, aspartate and glutamate metabolism, arginine and proline metabolism, glycine, lysine, citrate, glycerol and myo-inositol might be strongly associated with yeast tolerance to FAP.
The proposed targets relevant to FAP tolerance
The alanine, aspartate and glutamate metabolism and glycerol metabolism have been verified to be important for yeast cells to resist FAP stress in the previous study [23]. The metabolite citrate is one of TCA cycle intermediates. The significant variance of citrate level might be a suggestion that TCA cycle were markedly affected during the adaptation to FAP (Additional file 2: Figure S1). The metabolic flux analysis and comparative proteomics analysis showed that the addition of furfural could affect the activity of TCA cycle, which are involved in energy metabolism, as well as NADH production for the reduction of furfural [24, 25].
The detected intermediates (glutamate, proline, 5-oxo-proline and ornithine) in arginine and proline metabolic pathway were mainly involved in proline synthesis. Their significant variance reflected the marked effect of FAP on this pathway. As shown in Fig. 2e, the comparison of G0 and G1 showed that the intermediates of proline synthetic pathway including glutamate, 5-oxo-proline and ornithine were all significantly decreased when cells suffered sudden exposure to FAP, while the pathway end product of proline was increased by two times in cells of G1 as a response to FAP (Fig. 2e). It was consistent with the conclusion from the previous studies that proline was accumulated as a stress protectant in plants in response to various stress conditions [16]. With the adaptive evolution to FAP, the levels of proline and other intermediates in proline synthetic pathway were all reduced in cells of G2 and G3 (Fig. 2e). The contents of glycine and lysine in cells were also increased as a response to FAP stress, and then returned to low level with the adaptation to FAP (Fig. 2e). In our results, myo-inositol was also significantly affected by FAP. As shown in Fig. 2e, the cells of G1 were characterized by lower levels of myo-inositol compared to cells of G0 in FAP-free medium. In S. cerevisiae, the gene PIS1 encodes the phosphatidylinositol synthase to catalyze the de novo synthesis of PI from myo-inositol and CDP-diacylglycerol. When yeast cells were treated by FAP, the transcriptional level of gene PIS1 was distinctly elevated [26]. Thus, more myo-inositol would be applied for PI synthesis under FAP stress, which might reduce the intracellular accumulation of myo-inositol as observed in cells of G1. With the adaptation to FAP, myo-inositol content returned to high level.
According to the above results, we predicted that the synthesis of proline, myo-inositol, glycine and lysine might be important for strain tolerance to FAP. The growth phenotypes of the mutants deficient in the synthesis of proline (ΔPRO1 and ΔPRO2), glycine (ΔGLY1 and ΔSHM1), lysine (ΔLYS1) and myo-inositol (ΔINO1 and ΔINM2) were investigated to identify whether genetically disturbing the synthesis of these metabolites could affect cell tolerance to FAP. As shown in Additional file 3: Figure S2, the genetic perturbation of glycine or lysine biosynthesis through disrupting relevant gene GLY1, SHM1 or LYS1 does not affect cell growth no matter with or without combined inhibitors. However, the four mutants involved in proline synthesis and myo-inositol synthesis all exhibited increased sensitivity to FAP with enlarged lag phase and fermentation time compared to the parental strain (Fig. 3b). In the absence of FAP, the disruption of gene PRO1, PRO2 and INM2 had no effect on cell growth (Fig. 3a). The deletion of gene INO1 triggered a lower final biomass but the fermentation time was slightly affected (Fig. 3a). These results proved the important role of proline and myo-inositol biosynthesis in maintaining strain growth ability under FAP stress. Accordingly, a schematic view of the targets which might be responsible for strain tolerance to FAP was proposed (Additional file 4: Figure S3).
Effects of proline or myo-inositol supplementation on tolerance of S. cerevisiae to FAP
Considering the above results, we speculated that strain tolerance to FAP could be improved through increasing the availability of proline or myo-inositol either by external addition or enhancing their intracellular synthesis. In the following study, the effect of proline or myo-inositol supplementation on the tolerance of strain BY4742/pRS426 to FAP was first tested. As shown in Fig. 4a, the addition of proline exerted a protective effect on cell growth under FAP stress. When extra 500 or 1000 mg/L proline was added, the moderate growth advantage was observed under FAP stress (Fig. 4a). Supplementation of 1500 mg/L proline further shortened the lag phase and fermentation time, and thus improved strain tolerance against FAP stress (Fig. 4a). For the effect of myo-inositol, the addition of extra 500 mg/L myo-inositol slightly increased strain growth ability under FAP stress, while an obvious growth advantage was observed when 1000 mg/L myo-inositol was added (Fig. 4b). In the FAP-free medium, neither proline nor myo-inositol addition has the positive effect on cell growth phenotypes (Fig. 4), indicating that the growth advantage appeared under FAP stress is not due to the minimal dose requirement for proline and myo-inositol. In addition, we found that the combination of proline and myo-inositol supplementation could also induce a moderate growth improvement under FAP stress compared to the single addition of proline or myo-inositol (Additional file 5: Figure S4). These results demonstrated that increasing proline and myo-inositol by exogenous addition contributed to the enhanced FAP tolerance in S. cerevisiae.
Effects of enhancing proline synthesis on tolerance of S. cerevisiae to FAP
Although exogenously added proline or myo-inositol could alter yeast tolerance against combined inhibitors, this exogenous addition is not desirable for large-scale fermentations due to the additional costs. Therefore, to facilitate the enhanced tolerance without exogenous addition, we sought to genetically modify the strain by altering expression levels of the enzymes involved in proline or myo-inositol biosynthesis pathway to increase their intracellular synthesis. In S. cerevisiae, the main proline synthetic pathway from glutamate consisted of three enzymes: γ-glutamyl kinase (GK, the PRO1 gene product), γ-glutamyl phosphate reductase (encoded by PRO2), and Δ1-pyrroline-5-carboxylate reductase (encoded by PRO3). The PRO1 gene was first overexpressed to test its effect on strain tolerance against FAP stress (Fig. 5a). As shown in Fig. 5b, the overexpression of gene PRO1 triggered approximate twofold higher intracellular proline in the recombinant strain (BY4742/PRO1) than that in the control strain (BY4742/pRS426). In FAP-free medium, the recombinant strain BY4742/PRO1 exhibited the similar fermentation pattern with the control strain (Additional file 6: Figure S5). When exposed to FAP condition, a great growth advantage was observed in strain BY4742/PRO1 compared to strain BY4742/pRS426 (Fig. 5c). The lag phase was significantly shortened due to the overexpression of PRO1 gene (Fig. 5c). At about 54 h, the glucose could be depleted totally by the strain BY4742/PRO1, while the control strain was still in lag phase (Fig. 5c, d). The ethanol production rate was in parallel with the glucose consumption rate and ethanol yield was slightly affected (Fig. 5d). As mentioned above, the duration of the lag phase has been well established to evaluate the strain tolerance to inhibitors [20, 21]. The significantly shortened lag phase and increased glucose consumption rate of strain BY4742/PRO1 indicated that the enhancement of proline synthesis through overexpressing gene PRO1 could successfully improve strain tolerance against FAP stress. However, the recombinant strain with the overexpression of gene PRO2 did not exhibit the enhanced tolerance to FAP (Additional file 7: Figure S6). In the proline synthesis of S. cerevisiae, GK activity (encoded by PRO1) is sensitive to feedback inhibition by proline, and GK has been proved to be the rate-limiting and key regulatory enzyme that controls intracellular proline biosynthesis [27, 28]. Therefore, the gene PRO1 was generally modified to regulate the proline level in S. cerevisiae. Through overexpressing the wild PRO1 gene or expressing the mutant PRO1 gene encoding the D154N mutant GK, which is less sensitive to proline feedback inhibition, increased proline accumulation in S. cerevisiae can be implemented and the enhanced tolerance was observed under freezing, air-drying, ethanol and high-sucrose stresses [29–32]. In our study, the overexpression of gene PRO1 could also enhance proline accumulation and successfully improved strain tolerance to FAP. Correspondingly, the overexpression of PRO2 gene might not play an active role in regulating intracellular proline synthesis to affect strain tolerance against FAP due to the strictly regulated role of GK on proline synthesis. The PRO1 gene would be an alternative target for further improving strain tolerance against lignocellulose-derived inhibitors.
Proline involves in many important intracellular functions, such as maintaining protein and membrane stabilization, lowering the Tm of DNA, and scavenging of reactive oxygen species (ROS) [33]. Allen demonstrated that furfural induced ROS accumulation in S. cerevisiae, resulting in damage to mitochondria and vacuole membranes, the actin cytoskeleton and nuclear chromatin [34]. Acetic acid could induce a programmed cell death process with an apoptotic phenotype, which was related to the intracellular ROS level [35, 36]. As shown in Fig. 6, ROS accumulation was detected both in the parental strain and recombinant strains after the addition of FAP, reflecting that the multiple inhibitors might cause oxidative stress in yeast strain. As an antioxidant, proline has the ability to scavenge intracellular ROS and thereby suppresses ROS-mediated apoptosis [37]. Under FAP-free condition, the ROS levels in PRO1Δ and PRO2Δ were similar with their control strain BY4742. However, when cells were challenged by FAP, the interruption of proline biosynthesis by deleting gene PRO1 or PRO2 resulted in much higher ROS level accumulated in cells compared to the control strain BY4742 (Fig. 6a). On the contrary, the enhancement of proline biosynthesis by overexpressing gene PRO1 largely mitigated the ROS accumulation in vivo (Fig. 6b). The ROS level in strain BY4742/PRO1 is only one-third of that in the control strain BY4742/pRS426 under FAP stress. Therefore, we predicted that one potential role of proline under FAP stress is serving as an ROS scavenger to protect cells from oxidative damage.
Effects of enhancing myo-inositol synthesis on tolerance of S. cerevisiae to FAP
Myo-inositol is a precursor for many inositol-containing compounds and implicated in various physiological and biochemical processes such as membrane phospholipid synthesis, nuclear processes and alcohol stress [17, 38, 39]. In S. cerevisiae, myo-inositol is produced from glucose-6-phosphate via the reactions catalyzed by the INO1-encoded inositol-3-phosphate synthase and INM1-encoded inositol monophosphatase, and then participated in the biosynthesis of phosphatidylinositol (PI). The reaction catalyzed by INO1-encoded enzyme is known to be the key step in the synthesis of inositol-containing phospholipids. To engineer the strain with enhanced myo-inositol synthesis, the key gene INO1 was overexpressed (Fig. 7a). The intracellular myo-inositol level of recombinant strain BY4742/INO1 was eightfold higher than that of the strain BY4742/pRS426 (Fig. 7b). The slight difference was observed between these two strains when the fermentation was performed in SC-Ura medium without FAP (Additional file 8: Figure S7). However, in the presence of FAP, the overexpression of gene INO1 significantly reduced the lag phase and the strain BY4742/INO1 was more tolerant to inhibitors than the control strain (Fig. 7c). As illustrated in Fig. 7c, d, the strain BY4742/INO1 could grow into the stationary phase and exhaust the glucose totally at about 60 h, while the strain BY4742/pRS26 was still in the lag phase. The ethanol production rate was also in parallel with the glucose utilization, and the final ethanol production was slightly affected (Fig. 7d). Meanwhile, the overexpression of gene INM2 could also improve strain ability to resist FAP stress (Additional file 7: Figure S6). These results further indicated that myo-inositol metabolism was one of the targets facilitating strain tolerance to FAP. Increasing myo-inositol by the overexpression of genes involved in myo-inositol biosynthesis could improve strain tolerance against multiple inhibitors. However, the role of myo-inositol in protecting cells against FAP was still unclear. As a direct precursor of PI synthesis, intracellular myo-inositol level may affect the lipid composition and membrane integrity [39]. Yang et al. [26] has reported that combined inhibitors interrupted the membrane integrity and permeability, and the increase of PCs and PIs with long fatty acyl chains might be an important compensatory mechanism for the increase of plasma membrane permeability and fluidity when subjected to combined inhibitors. Therefore, it was most likely that myo-inositol acted on cell membrane to play its protective role in resisting tolerance.