Biofuels produced by microbial fermentations of renewable resources represent an important replacement for fossil fuels. Today, bioethanol is the most important renewable gasoline substitute. However, higher alcohols like butanol are regarded as more suitable fuel substitutes, due to their higher energy density, lower hygroscopicity, and therefore lower corrosiveness [1].

n-Butanol is produced by various Clostridium species via the acetone butanol ethanol (ABE) fermentation process with a ratio of 1:3:6 and titers up to 13 g/l [2, 3]. However, clostridial fermentations are associated with several problems, including sporulation, slow growth rates, bacteriophage infections, and strict anaerobic cultivations [4]. For this reason, other industrially more relevant organisms have been engineered for n-butanol production by introducing variants of the clostridial pathway [5] or using ketoacid degradation pathways [6]. Although n-butanol titers up to 15 g/l [7] were achieved in E. coli, large-scale industrial fermentations with this bacterium also suffer from the high risk of phage infections and bacterial contaminations, for instance [8, 9]. Compared to E. coli, S. cerevisiae holds significant advantages in terms of scalable industrial fermentations, due to long-lasting experiences with this yeast in fermentation processes, high robustness, and tolerance against inhibitory compounds [10].

S. cerevisiae is able to produce small amounts of n-butanol on rich medium via endogenous pathways dependent on threonine or glycine catabolism, the first one stimulated by deletion of ADH1 [11, 12]. By improving these pathways, n-butanol titers up to 242.8 and 92 mg/L, respectively, could be achieved. Combining the amino acid-dependent endogenous pathways with a synthetic ABE pathway and a mutant version of translation initiation factor eIF2B in an adh1 mutant strain resulted in up to 300 mg/L n-butanol, but only in YEPD medium after 15–20 days [13]. Nevertheless, as biofuel production from amino acids is economically not viable, for industrial use threonine or glycine overproduction strains would be required which, however, exhibit lower theoretical maximum yields [1].

Therefore, it is more promising to enhance n-butanol production by improvements of variants of the n-butanol pathway expressed in S. cerevisiae. In this pathway, two molecules of acetyl-CoA are condensed to acetoacetyl-CoA, which is reduced to 3-hydroxybutyryl-CoA and then dehydrated to crotonyl-CoA. This is further reduced to butyryl-CoA, butyraldehyde, and finally n-butanol (Fig. 1). Extending the pathway at the level of butyryl-CoA in a reversal of the β-oxidation cycle would enable even the production of longer molecules [14, 15]. Besides alcohols, also the intermediates of the pathway are valuable compounds or can be used to produce the corresponding aldehydes, fatty acids, and alkanes with different chain lengths. First attempts to establish n-butanol production in yeast using this acetoacetyl-CoA-derived pathway resulted in n-butanol titer as low as only 2.5 mg/L [16]. Further pathway optimization and engineering of substrate supply could enhance n-butanol titers up to 120 mg/L [9, 17]. However, these titers were obtained only in synthetic complete media in high-cell density fermentations and were critically reviewed in [1].

As n-butanol yield in yeast is still much lower than in E. coli, it was hypothesized that cytosolic acetyl-CoA as the initial substrate of the n-butanol pathway is a limiting factor [9]. In yeast, acetyl-CoA is present in four different compartments: cytosol, mitochondria, peroxisomes, and the nucleus. Cytosolic acetyl-CoA is produced via the decarboxylation of pyruvate into acetaldehyde, which is oxidized to acetate by aldehyde dehydrogenases and finally converted to acetyl-CoA, at the expense of two ATPs. This pathway is known as pyruvate dehydrogenase (Pdh) bypass. Aldehyde dehydrogenases compete with alcohol dehydrogenases (Adh) which due to the crabtree effect convert most of the sugar-derived acetaldehyde into ethanol [18]. On the other hand, mitochondrial acetyl-CoA obtained via pyruvate dehydrogenase obviously cannot be transported to the cytosol [19]. Different strategies were tested to increase the level of cytosolic acetyl-CoA. In one approach, alcohol dehydrogenases were deleted, the reactions of the Pdh bypass were overexpressed, and the withdrawal of acetyl-CoA into the glyoxylate cycle was prevented [18]. In another approach, an ATP independent pyruvate dehydrogenase complex from Enterococcus faecalis was expressed in the yeast cytosol replacing the Pdh bypass, but being dependent on supplementation with lipoic acid [20]. Another possibility of ATP independent cytosolic acetyl-CoA production is the use of acetylating acetaldehyde dehydrogenases or pyruvate formate lyases [21, 22].

Since the production of cytosolic acetyl-CoA proceeds from two substrates, acetate and coenzyme A (CoA), the level of CoA might limit acetyl-CoA production as well. Moreover, nearly all the intermediates of the n-butanol pathway exist as CoA esters which might additionally decrease CoA availability. Therefore, we hypothesized that engineering of CoA biosynthesis could further increase the concentration of acetyl-CoA in yeast, resulting in elevated n-butanol production. CoA biosynthesis starts from pantothenate. In yeast, pantothenate can be produced via an endogenous pathway starting from 2-ketoisovalerate, derived from the valine metabolic pathway, and spermine, derived from l-ornithine and methionine [23]. Besides, it can be taken up from the medium by the transporter Fen2 [24]. Pantothenate is then converted into 4-phosphopantothenate by pantothenate kinase Cab1, which is postulated to catalyze the rate-limiting step in CoA biosynthesis. Cysteine is incorporated by phosphopantothenate cysteine ligase Cab2, which is weakly glucose repressed like Cab1 [23]. The resulting product is finally converted into CoA in three further steps [25].

In the present study, we first established and improved n-butanol production in S. cerevisiae by testing different variants of the acetoacetyl-CoA-derived pathway. We then improved synthesis of CoA by overexpression of pantothenate kinase coaA from E. coli, which is not inhibited by acetyl-CoA [23], or by overexpression of the yeast Fen2 pantothenate transporter. The level of CoA could further be increased by adding pantothenate to the medium. By deletion of alcohol dehydrogenases and glycerol-3-phosphate dehydrogenase genes, we increased the availability of acetaldehyde and NADH as driving forces for the n-butanol pathway. Overexpression of a mutant form of an acetylating acetaldehyde dehydrogenase could further improve n-butanol production, resulting in n-butanol titers and yields of up to 130 ± 20 mg/L and 0.012 g/g glucose, respectively. These are the highest values ever reported for S. cerevisiae in synthetic medium via the acetoacetyl-CoA-derived pathway.

The next step was to redirect the metabolic flux away from the predominant fermentation product ethanol to n-butanol, and to increase the redox power in the form of NADH. Therefore, alcohol dehydrogenase genes ADH1-5 were deleted in order to reduce NADH-dependent acetaldehyde reduction and the concomitant formation of ethanol. As expected, we observed less ethanol production, but therefore an increased glycerol formation as a result of an enhanced alternative NADH oxidation route of yeast (Additional file 1: Figure S6). Fermentations of VSY0 (Δadh1-5) transformed with pVS6 (n-butanol pathway) resulted in n-butanol production up to 30 mg/L (Additional file 1: Figure S2), which is a significant increase compared to the reference strain (13 mg/L) (Fig. 5). The effect of elevated CoA supply was also tested in this strain by introducing pVS5_4 (coaA) together with pVS6. n-Butanol production was significantly higher in the adh1-5 deletion strain with coaA overexpression (52 versus 30 mg/L) (Additional file 1: Figure S2, Fig. 5). After the addition of 25 µM pantothenate, the final n-butanol titer even increased up to 71 mg/L in the Δadh1-5 (pVS6, pVS5_4) strain (Additional file 1: Figure S8, Fig. 5). These results show that providing more substrate in the form of acetaldehyde and more reducing power in the form of NADH can increase n-butanol production.

In yeast, cytosolic acetyl-CoA is mainly produced by acetyl-CoA synthetases Acs1 and 2 [18]. It has been shown before that acetyl-CoA-dependent product formation can be increased by increasing acetyl-CoA formation [8, 32]. However, acetyl-CoA synthetases use hydrolysis of ATP to AMP which is not favorable for anaerobic n-butanol production, since only glycolysis and not respiration can provide ATP. Therefore, we tested whether the expression of an ATP independent acetylating acetaldehyde dehydrogenase from E. coli, adhE, can further increase n-butanol production. Normally, the enzyme adhE catalyzes the reduction of acetyl-CoA into ethanol. Membrillo-Hernandez and coworkers [33] succeeded in engineering the enzyme to mainly perform the reverse reaction of converting acetaldehyde into acetyl-CoA. The specific acetaldehyde dehydrogenase activity could be increased by changing two amino acids: A267T and E568 K. For this reason, we chose to use adhE^A267T/E568K to increase the cytosolic acetyl-CoA pool. The mutant adhE ^A267T/E568K allele could complement the growth defect of an acs2 mutant strain on glucose medium, indicating that it is functional in yeast (data not shown) [21, 22]. We integrated this allele, expressed under control of the strong glycolytic PFK1 promoter, into the SFA1 locus resulting in strain VSY8. As SFA1 encodes an additional but minor alcohol dehydrogenase [34], we chose this locus for integration to stabilize the strain and to prevent the occurrence of ethanol-producing suppressor mutants. Surprisingly, fermentations of VSY8 expressing the n-butanol pathway (pVS6) showed almost no n-butanol production, but increased ethanol production instead (Fig. 5, Additional file 1: Figure S6). One reason could be that adhE^A267T/E568K mainly converts acetaldehyde or acetyl-CoA into ethanol if there is not enough free CoA, the second substrate for the acetylating acetaldehyde dehydrogenase activity, available. Indeed, concomitant overexpression of coaA reduced ethanol and increased n-butanol formation (Additional file 1: Figure S9, Additional file 1: Figure S6). On the other hand, it might be possible that the additional alcohol dehydrogenase gene ADH6 is upregulated under those conditions. Indeed, when ADH6 was deleted by integration of coaA, a further decrease in ethanol formation could be observed [calculated yield for ethanol: 0.1 g/g glucose in VSY10 and 0.13 g/g in VSY8 (pVS6, pVS5_4)] (Additional file 1: Figure S6). In fermentations of the Δadh1-5, adhE ^A267T/E568K-overexpressing strain VSY8 (pVS6, pVS5_4), n-butanol production was increased up to 75 mg/L, which is a significant improvement compared to the strain without adhE ^A267T/E568K overexpression (52 mg/L) (Fig. 5). To further increase CoA availability, we added 25 µM pantothenate which resulted in up to 95 mg/L n-butanol production (Fig. 5, Additional file 1: Figure S9). These results show that increasing the cytosolic acetyl-CoA synthesis by adhE^A267T/E568K does improve n-butanol production, especially when combined with elevated CoA synthesis. It should be noted that although we used butyraldehyde dehydrogenase eutE in pVS6, which was also shown to convert acetaldehyde into acetyl-CoA [22], we could not observe a clear improvement in n-butanol production due to this activity.

GPD2 encodes glycerol-3-phosphate dehydrogenase II which is the major enzyme for glycerol production in yeast under anaerobic conditions [35]. Glycerol production is used to re-oxidize a surplus of NADH. Deletion of GPD2 would further increase NADH levels in the adh deletion mutants [36] and could therefore push n-butanol production. Moreover, in order to overcome the negative effects of using different plasmids we decided to integrate coaA into the ADH6 locus of VSY8 (resulting in strain VSY9), and the n-butanol pathway genes of pVS6 into the GPD2 locus of strain VSY9, finally resulting in strain VSY10. Fermentations of strain VSY10 resulted in n-butanol titers up to 108 mg/L, which is significantly higher compared to its plasmid-based parental strain VSY8 (pVS6, pVS5_4; 75 mg/L) (Fig. 5). The increase might be due to both the stable integration of n-butanol pathway genes into the genome as well as reduced glycerol production [calculated yield for glycerol: 0.17 g/g glucose in VSY10 and 0.27 g/g in VSY8 (pVS6, pVS5_4)] (Additional file 1: Figure S6). Nevertheless, VSY10 did not consume all of the glucose (Additional file 1: Figure S5). This might be due to deletions of adh1-6, sfa1, and gpd2, resulting in inefficient NADH oxidation. The capacity of the introduced n-butanol pathway is obviously not strong enough to compensate this deficiency. This can also be seen by measured OD₆₀₀ values. Without deletions, the wildtype is able to reach an OD₆₀₀ of about 6, but all deletion strains reached not more than an OD₆₀₀ of about 2, except for VSY8 (pVS6), that produced ethanol instead of n-butanol (Additional file 1: Figure S7). As a result of incomplete glucose consumption the n-butanol yield of VSY10 (0.01 g/g glucose) is much higher than for all fermentations with VSY8 (up to 0.0057 g/g) (Additional file 1: Figure S4). The addition of 25 µM pantothenate increased the n-butanol yield of strain VSY10 up to 0.012 g/g (Additional file 1: Figure S4), resulting in a titer of 130 ± 20 mg/L (Fig. 5, Additional file 1: Figure S10). This is to our knowledge the highest titer reported so far for n-butanol production in synthetic media in yeast via an acetoacetyl-CoA-derived n-butanol pathway.

Compared to bacterial systems, n-butanol production in S. cerevisiae via heterologous acetoacetyl-CoA-derived pathways is very low. The reason for this is not fully understood. Our results confirm that several reasons contribute to the low efficiency of n-butanol production in yeast. By stepwise engineering various aspects of n-butanol production, we could gradually increase n-butanol titers and yields. Of course, a very important aspect was the composition of the heterologous pathway. By testing various isoenzymes for the different reactions, we could establish a functional pathway. n-Butanol production was then increased by stable expression of the heterologous genes. Expression from centromeric plasmids was clearly superior to expression from multicopy plasmids (Table 2). Other important aspects were the elevation of acetaldehyde levels and NADH by reduction of alcohol dehydrogenase activities and glycerol synthesis. Moreover, increasing acetyl-CoA availability by means of a mutant form of an acetylating acetaldehyde dehydrogenase from E. coli further improved n-butanol production. However, our results indicate that obviously free CoA levels in S. cerevisiae are limiting n-butanol production. By increasing synthesis of CoA via overexpression of pantothenate kinase coaA of E. coli and by feeding pantothenate to the medium, n-butanol production levels could be increased up to 130 ± 20 mg/L, the highest level ever reached in yeast via an acetoacetyl-CoA-derived pathway in synthetic medium.