Anaerobic ethyl acetate production in E. coli and reduction of by-product formation
To enable ethyl acetate production in E. coli BW25113 (DE3), we introduced the K. marxianus eat1 (Kma eat1) under the control of the IPTG-inducible LacI/T7 promoter. Under anaerobic conditions the strain produced 2.7 ± 0.1 mM ethyl acetate, representing a yield of 0.03 ± 0.00 C-molethyl acetate/C-molglucose (Fig. 3a, c). Due to the formation of by-products, particularly lactate and acetate, the ethyl acetate titre was low. To maximise the metabolic flux towards ethyl acetate, we disrupted the acetate kinase (ackA) and lactate dehydrogenase (ldhA) genes to reduce acetate and lactate formation, respectively. This increased the ethyl acetate titre to 9.1 ± 0.3 mM (Fig. 3b). The final ethyl acetate yield increased to 0.13 ± 0.00 C-molethyl acetate/C-molglucose, or 21.4% of the maximum pathway yield (Fig. 3c). Lactate production was almost completely abolished. Acetate yields did not decrease significantly despite the ackA disruption (Fig. 3c). A possible explanation is that acetate is produced via the hydrolysis of ethyl acetate or acetyl-CoA by the esterase and thioesterase side activity of eat1, respectively.
Since lactate production can no longer act as sink of NADH, ethanol synthesis should fulfil this role. The conversion of ethanol together with acetyl-CoA to ethyl acetate would basically consume all available NADH and make the entire process redox neutral. However, the accumulation of ethanol and also pyruvate suggests that synthesis of ethyl acetate is limited and that Eat1 activity is the bottleneck of the process (Fig. 3b, c). We therefore focused on optimising the activity of the AAT step.
Selection of ethyl acetate-producing AAT and gene expression optimisation
We compared the ethyl acetate-production capacity of S. cerevisiae atf1 (Sce atf1), Kma eat1 and W. anomalus eat1 (Wan eat1) genes in E. coli BW25113 ΔackAΔldhA (DE3) cultivated in anaerobic serum bottles. The genes were placed under the control of the inducible LacI/T7 (Fig. 4a–c) or XylS/Pm promoter (Fig. 4d–f) to allow modulation of their expression levels. To induce gene expression, IPTG or m-toluate was added at various concentrations.
Increased AAT activities will reduce the accumulation of pyruvate, increase the production of ethyl acetate when it is active as an AAT, and increase the production of acetate when it is active as either an esterase or thioesterase.
Strains expressing Wan eat1 (Fig. 4b, e) showed the highest ethyl acetate yields compared to the other AAT genes controlled by the same promoter. High yields of ethyl acetate were also reached by strains expressing Kma eat1 under control of the lac-T7 promoter (Fig. 4a). Surprisingly, strains expressing Kma eat1 under the XylS/Pm promoter produced only traces of ethyl acetate under all induction levels (Fig. 4d). Sce atf1 also evoked ethyl acetate production, but the yields were significantly lower compared to the two eat1 genes (Fig. 4c, f). Pyruvate accumulation decreased significantly, indicating that Atf1 was active, but primarily as an esterase/thioesterase as the acetate yields increased. These results show that Eat1 homologues are better catalysts than Sce Atf1 for in vivo ethyl acetate production in E. coli BW25113 ΔackAΔldhA (DE3) under the tested conditions.
The expression of Wan and Kma eat1 under the control of the LacI/T7 promoter resulted in 0.2 C-molethyl acetate/C-molglucose or higher. However, Wan eat1 required 10-fold less IPTG to reach the same or higher ethyl acetate yields than Kma eat1 (Fig. 4ab). Moreover, the strains expressing Wan eat1 under the control of the XylS/Pm promotor produced up to 0.16 ± 0.01 C-molethyl acetate/C-molglucose (Fig. 4e), whilst Kma eat1 produced almost no ethyl acetate (Fig. 4d). This difference may be explained by the fact that the XylS/Pm promoter is weaker compared to the LacI/T7 promoter [2]. The higher yield obtained with lower gene expression levels indicates that Wan Eat1 was more active than its K. marxianus homologue under these cultivation conditions.
The ethyl acetate yields increased with rising inducer concentrations (Fig. 4e, f), reached a plateau (Fig. 4a) and even began to decline at higher inducer concentration (Fig. 4b, c). Determining the optimal inducer concentrations thus resulted in significantly improved ethyl acetate yields. For example, optimised IPTG concentrations used for gene induction in E. coli BW25113 ΔackAΔldhA (DE3) (pET26b:hKma Eat1) led to an increase of the ethyl acetate yield from 0.13 ± 0.00 (Fig. 3c) to 0.19 ± 0.00 C-molethyl acetate/C-molglucose (Fig. 4a). The highest ethyl acetate yield was achieved in E. coli BW25113 ΔackAΔldhA (DE3) (pET26b:hWan Eat1) that was induced with 0.01 mM IPTG. It produced 0.27 ± 0.01 C-molethyl acetate/C-molglucose or 40.7% of the theoretical pathway maximum (Fig. 4b).
Selecting the best AAT gene and optimising, its expression level diminished the metabolic bottleneck present in ethyl acetate production, but pyruvate still accumulated (Fig. 4a–f). This indicated that the conversion efficiency of Eat1 was still insufficient to handle the EMP metabolic pathway flux.
Using truncated Eat1 variants
Removal of the mitochondrial pre-sequences of K. marxianus and W. anomalus Eat1 resulted in a higher stability of the enzyme when expressed in E. coli [20]. We tested if this elevated stability also led to more ethyl acetate production. Optimisation of gene expression levels for Kma trEat1 F-26 and K-30 resulted in a lower accumulation of pyruvate compared to the unprocessed version (Fig. 5a–c), suggesting a higher efficiency of the truncated Eat1 variants. However, this did not result in a higher ethyl acetate yield; only the acetate yield increased.
Nevertheless, the optimum inducer concentration shifted to 0.05 mM IPTG, which was 50% lower compared to the native Eat1. Ethyl acetate production was also higher at 0.01 and 0.02 mM IPTG, indicating that the in vivo production capacity of ethyl acetate improved. At the same time, the acetate yields increased for induction levels above 0.05 mM IPTG, whilst the pyruvate yields decreased (Fig. 5b, c).
At the lowest IPTG concentration, the strains producing the truncated Wan Eat1 (Wan trEat1 N-13) reached a 3.5-fold higher ethyl acetate yield on glucose than the unprocessed Wan Eat1 (Wan Eat1) (Fig. 5e, f). However, at higher IPTG concentrations these differences were absent. The acetate yield in the strain producing Wan trEat1 N-13 also increased relative to the strain producing Wan Eat1. (Figure 5d, f). The increase in acetate production was not as pronounced as with the Kma trEat1 F-26 and K-30 (Fig. 5a–c). No difference was found between the Wan trEat1 V-11 and the unprocessed Wan Eat1 (Fig. 5d, e).
As discussed above, a limiting Eat1 efficiency resulted in the accumulation of ethanol. Pyruvate was produced by E. coli BW25113 ΔackAΔldhA (DE3) to counter the redox imbalance caused by ethanol accumulation. The higher stability of the truncated Eat1 versions did indeed result in a decreased pyruvate yield (Fig. 5b, c, e, f) but the ethyl acetate yield did not increase accordingly. Instead, the acetate yield increased, likely due to the esterase and thioesterase side activities of Eat1.
Improving ethyl acetate production with H2 co-production in controlled bioreactors
In all serum bottle experiments described above, glucose consumption was incomplete, most likely caused by the accumulation of organic acids, especially formate, and the associated pH decreased due to a limited buffering capacity of the medium. To avoid limitations caused by medium acidification, additional cultivations were performed in pH-controlled reactors under anaerobic conditions. To limit the accumulation of formate even further, Na2SeO3 was added to stimulate the conversion of formate into H2 and CO2 by Fhl. A constant flow of nitrogen gas was applied to keep the culture conditions anoxic. This resulted in stripping of ethyl acetate, H2 and CO2 from the broth and the concentrations of these compounds in the exhaust gas were therefore analysed.
We cultivated E. coli BW25113 ΔackAΔldhA (DE3) producing several Eat1 variants. Gene expression was induced with the optimal IPTG concentration of each strain based on the findings of previous experiments (Figs. 4, 5). In contrast to the shake-flask experiments, glucose was fully consumed at the end of the batch fermentations and ethyl acetate production proceeded until glucose was depleted (Fig. 6a, b). Formate was converted into CO2 and H2 by E. coli BW25113 ΔackAΔldhA (DE3), but conversion percentages were inconsistent and the conversion was incomplete (Fig. 6 c, d). Strains’ conversion was between 6% and 27% of the formate into CO2 and H2 whilst in most fermentations, the conversion averaged around 10%. There was no correlation between the conversion efficiency, the strain, or reactor vessel. Between 93.0 and 103.7% of the carbon was recovered in all runs performed when biomass formation and the main fermentation products such as ethyl acetate, ethanol, acetate, pyruvate, formate, succinate and CO2 were included (Fig. 7a, b).
Consistently, all strains cultivated in pH-controlled bioreactors showed improved performance compared to the serum bottle cultivations. Once the unprocessed Kma Eat1 was induced with an optimal 0.1 mM IPTG, a beneficial effect on ethyl acetate yield was apparent. A yield of 0.35 ± 0.01 C-molethyl acetate/C-molglucose corresponded to a 1.8-fold increase when compared to a serum bottle yield of 0.19 ± 0.01 C-molethyl acetate/C-molglucose, reaching about 50% of the maximum pathway yield. A similar yield was obtained in strains producing Kma trEat1 K-30 in the presence of 0.05 mM IPTG (Fig. 7a).
The best producers tested in pH-controlled bioreactors were E. coli BW25113 ΔackAΔldhA (DE3) producing Kma trEat1 K-30 and Wan trEat N-13. They formed 27.6 ± 3.7 mM (2.4 ± 0.3 g/L) and 42.8 ± 3.3 mM (3.8 ± 0.3 g/L) ethyl acetate from 55.6 ± 2.5 mM (10.0 ± 0.5 g/L) glucose, respectively (Fig. 6, Additional file 1). Generally, ethyl acetate yields were between 1.6- and 2.8-fold higher in bioreactors compared to serum bottles. The highest yield was obtained by the strain producing Wan trEat1 N-13, reaching 0.48 ± 0.03 C-molethyl acetate/C-molglucose, or 72.3% of the maximum pathway yield.
The yields of ethanol and pyruvate decreased with increasing ethyl acetate yields (Fig. 7a, b). Strains producing the unprocessed Kma Eat1 and Kma trEat1 K-30 in the presence of optimal IPTG concentrations accumulated 66% less pyruvate (Fig. 7a) compared to cultivations in serum bottles (Fig. 5a, c). For the strains producing unprocessed Wan Eat1 and Wan trEat1 N-13 pyruvate accumulation was almost entirely abolished (Figs. 6b and 7b). The ethyl acetate yield for Wan Eat1 was consequently higher compared to the strains producing the Kma Eat1 variants. It should be noted that a statistically significant difference in ethyl acetate yields (p = 0.03) was only found for the strain producing Wan trEat1 N-13 (Fig. 7a, b). This strain converted approximately 72% of glucose to ethyl acetate based on the maximum pathway yield.
Not only did the trEat1 variants require lower induction levels and accumulated less by-products, glucose was also depleted faster. As a result, the volumetric productivity of ethyl acetate (QEA) was higher. The QEA of the strain producing Kma trEat K-30 (0.05 mM IPTG) was 35% higher (p = 0.013) compared to the strain producing unprocessed Kma Eat1 (0.1 mM IPTG) (Fig. 7c). A similar trend was present in E. coli BW25113 ΔackAΔldhA (DE3) producing unprocessed Wan Eat1 and trEat1 N-13 in the presence of 0.01 mM IPTG. The QEA of the latter strain was 26% higher (p-value = 0.042) compared to the strain producing the unprocessed Wan Eat1 (Fig. 7d).
The hydrolysis of ethyl acetate by the side activity of Eat1 might be restricted by efficiently removing all ethyl acetate by gas stripping. But due to low gas flow rates, ethyl acetate still accumulated in the liquid during the fermentation. At times of maximum productivities, liquid ethyl acetate concentrations ranged from 2.87 ± 0.1 mM for Kma Eat1 with 0.05 mM IPTG induction to up to 14.7 ± 0.4 mM for Wan trEat1 N-13 with 0.01 mM IPTG induction (data not shown).