Improvement of acetoin production by deleting ppc/pyc to reduce succinate accumulation
To analyze the carbon distribution in strain CGR6, batch fermentation in flasks was carried out. As shown in Fig. 2A, CGR6 was able to produce 11.30 g/L acetoin when glucose was almost depleted at 24 h, corresponding to a yield of 0.302 g/g glucose. As expected, no 2,3-butandiol or lactate was detected, since their synthesis pathways were blocked. Although 0.52 g/L α-ketoglutarate was detected when fermenting this strain in CGXIIY medium , it was not detected in CGXIIP medium in this study. The major by-product was succinate, with a titer of 2.63 g/L, followed by acetate (1.12 g/L) and glycerin (0.23 g/L). In our previous work, two key genes of succinate synthesis, ppc encoding phosphoenolpyruvate (PEP) carboxylase and pyc encoding pyruvate carboxylase, were respectively deleted via one-step homologous recombination (single crossover) with a tetracycline resistance marker . The ppc/pyc-deficient strain showed an improved acetoin production in batch fermentation in flasks. However, the use of the tetracycline resistance marker impeded further genetic manipulation and was not acceptable for an industrial producer. In this study, we deleted the ppc and pyc genes in CGR6 using a markerless two-step recombination method (double crossover) , resulting in the strains CGS1 and CGS2, respectively.
As shown in Fig. 2B, succinate production was almost completely abolished in the ppc knockout strain CGS1, with a titer of only 0.03 g/L. The glucose consumption rate of CGS1 was decreased compared with that of CGR6, and 11.96 g/L of acetoin was obtained when glucose was exhausted at 29 h. This corresponds to a yield of 0.328 g/g glucose, which was 9.3% higher than that of CGR6. By contrast, the succinate production in the pyc knockout strain CGS2 was almost unchanged (Fig. 2C), but its acetoin production was enhanced with a titer of 11.75 g/L when glucose was depleted at 29 h, and a yield of 0.323 g/g glucose. The glucose consumption rate of CGS2 was also decreased compared with that of CGR6, but it was somewhat higher than that of CGS1 at 24 h.
Mutants deficient in ppc rather than pyc could efficiently reduce succinate accumulation, indicating that PEP rather than pyruvate is the key precursor of succinate under aerobic conditions. In agreement with previous reports [36, 37], the growth of CGS1 and CGS2 was only mildly decreased, but their acetate titers (respectively 0.06 and 0.44 g/L) were unexpectedly significantly decreased compared with that of CGR6 (1.12 g/L) (Table 2). We suspected that the reduced acetate accumulation in CGS1 and CGS2 was resulted from the decreased glucose consumption rates (Table 2), which can reduce the flux in overflow metabolism. The only another detected by-product was glycerin, with a titer of about 0.30 g/L in both CGS1 and CGS2. The two strains both showed improved acetoin yields, which was consistent with our previous results . However, strain CGS3 with ppc and pyc-deficient was unable to grow on glucose in mineral medium and still grew poorly with 10 g/L yeast extract addition (data not shown), which was consistent with previous reports [37, 38]. After considering the acetoin titer, yield and by-products accumulation, strain CGS1 was chosen for further manipulation.
Improvement of acetoin production by blocking anaplerotic pathways and introducing isocitrate dehydrogenase mutants
Most studies on improving precursor pyruvate (Fig. 1) availability focused on complete inactivation or attenuation of the pyruvate dehydrogenase complex (PDHC) by deleting the aceE gene or reducing its promoter activity [39, 40]. In our previous work, the aceE gene was also deleted to conserve pyruvate and improve acetoin production. However, additional acetate was required for cell growth and the best strain CGL3 (C. glutamicum ∆aceE∆ldh∆butA; pEC-XK99E-alsSD) could only accumulate 8.33 g/L acetoin from about 33 g/L glucose and 10 g/L acetate under optimal conditions . In cases of attenuating PDHC, although the growth of the resulting strain was independent of acetate addition, the cell growth and glucose consumption rate were dramatically reduced . Therefore, deletion or attenuation of PDHC might not be optimal for manufacturing optically pure (3R)-acetoin.
A double deletion of ppc and pyc is generally lethal for C. glutamicum due to a lack of oxaloacetate . However, a novel strategy for improving the biosynthesis of pyruvate-derived metabolites by introducing newly identified isocitrate dehydrogenase mutants (A94D, G407S or R453S) was recently proposed . These icd mutations both lowered the ICD activity and activated the glyoxylate shunt, and were consequently able to recover oxaloacetate re-supply and cell growth of C. glutamicum ∆ppc∆pyc . Thus, three ICD mutants A94D, G407S and R453S were introduced into CGS3 to test their effects on acetoin production, resulting in strains CGS4, CGS5 and CGS6, respectively. As shown in Fig. S1, strains CGS4 and CGS6 showed almost the same growth inhibition as strain CGS3. This is probably because strain CGS3 is an acetoin producing host, in which the major carbon fluxes have already been redirected toward acetoin synthesis, and are therefore insufficient to support an activated glyoxylate shunt to re-supply oxaloacetate. Nevertheless, CGS5 with the mutant ICDG407S successfully recovered cell growth to some extent. However, acetoin production in CGS5 was dramatically decreased to a titer of only 11.37 g/L when glucose was exhausted at 34 h, and a yield of 0.280 g/g glucose, which was 14.6% lower than that of CGS1 (Fig. S2). This indicated that the carbon fluxes from acetoin synthesis rather than its competing pathways were redirected to regain biomass synthesis, but the reason for this is still unknown.
Improvement of acetoin production by reducing citrate synthase activity
Since directly blocking acetyl-CoA or oxaloacetate synthesis severely affected the growth performance and failed to improve acetoin production, the focus of engineering was moved to citrate synthase (CS, encode by gltA), which condenses acetyl-CoA and oxaloacetate to citrate and is one of the most important sinks for the flux of these two precursors. Therefore, reduction of CS activity was viewed as a promising strategy to lower the synthesis of acetyl-CoA and oxaloacetate, and thereby conserve pyruvate for improved acetoin production. C. glutamicum mutants devoid of CS were unable to grow on glucose . Therefore, a reduction rather than inactivation of CS activity was preferable. Moreover, regulation of the gltA gene, which is responsible for 95% of CS activity in C. glutamicum , was widely applied to improve the biosynthesis of pyruvate-derived products [34, 43-45]. However, the impact of reducing CS activity on acetoin production is still unclear. Therefore, we used a weak promoter P1 with approximately 1% activity of the Ptac  to replace the native promoter of gltA gene in CGR6 and CGS1, yielding strains CGS7 and CGS8 respectively.
The strains’ CS activity was measured to confirm its successful downregulation. As shown in Fig. S3A, the CS activity of CGS7 at 12 h in the middle exponential phase was 75.8% lower than that of CGR6, and 60.0% lower at 22 h in the late exponential phase. As shown in Table 2, the succinate yield of CGS7 at 24 h was 20.0% lower than that of CGR6 (0.056 versus 0.070 g succinate/g glucose). Moreover, the cell growth and glucose consumption were also decreased (Fig. 3A), suggesting that the flux into the TCA cycle was indeed reduced. Unexpectedly, the acetate yield decreased by 46.7% (0.016 versus 0.030 g acetate/g glucose). We speculated that the decreased glucose consumption rate in CGS7 would reduce the fluxes to acetate synthesis caused by overflow metabolism, and the newly available carbon fluxes flowed to acetoin effectively, leading to an increase of the acetoin yield at 24 h in contrast to CGR6 (0.337 versus 0.302 g acetoin/g glucose). Then, the yield further increased to 0.350 g/g glucose with a titer of 13.61 g/L when glucose was depleted at 29 h (Fig. 3A). However, the results indicated that neither weakening gltA nor the deletion of ppc could effectively improve the intracellular pyruvate pool, but a combination of both eventually led to a 31.1% increase of intracellular pyruvate (Fig. S3B). At the same time, the acetoin production was significantly enhanced to 14.56 g/L, with a yield of 0.389 g/g glucose when glucose was exhausted at 29 h, which was 18.6% higher than that of CGS1 (Fig. 3B). While deletion of ppc obviously inhibited glucose consumption of CGS8 compared to CGS7 (Table 2), however, the acetoin productivity was 0.52 g/L/h at 24 h, even higher than that of CGS7 (0.50 g/L/h). Moreover, the by-products acetate (0.15 g/L), glycerin (0.17 g/L) and succinate (0.03 g/L) remained at low concentrations. After considering acetoin production and by-product accumulation, CGS8 was selected for further engineering.
Furthermore, reduction of CS exhibited a fantastic synergistic effect on acetoin production with inactivation of PEP carboxylase, which is responsible for 90% of total oxaloacetate synthesis in C. glutamicum [47, 48]. As shown in Table 2, ppc knockout in CGS1 almost had no effect on acetoin yield (0.309 versus 0.302 g/g glucose), and gltA attenuation in CGS7 only led to a 11.6% increase in acetoin yield (0.337 versus 0.302 g/g glucose). However, the combination of the two gene modifications resulted in that the acetoin yield was significantly improved by 37.4% (0.415 versus 0.302 g/g glucose). The variations of intracellular pyruvate further confirmed the synergistic effect of CS and PEP carboxylase on acetoin production (Fig. S3B), which would also be potential for improving other pyruvate-acetolactate-derived metabolites production.
Effect of enhancing the acetoin synthesis pathway on acetoin production
To further pull the carbon flux from pyruvate toward product accumulation, two additional alsSD operon (under the control of the constitutive promoter Ptuf) copies were inserted into the chromosome of CGS8 at the ∆butA and ∆nagD sites to generate the strain CGS9.
The activities of ALS and ALDC were listed in Table 3. The ALS activity of CGS9 was 2.33-fold higher than that of CGR6 at 12 h in the middle log phase, and subsequently increased by 11.6% at 22 h in late log phase. However, the ALDC activity of CGS9 was increased by only 47.0% compared with that of CGR6 at 12 h. Then, it decreased by 15.1% at 22 h, but was still 37.3% higher than that of CGR6. With the increase of alsSD copies, the intracellular pyruvate concentration was slightly decreased by 8.2%. However, when glucose was exhausted at 34 h, 15.70 g/L acetoin was accumulated with a yield of 0.408 g/g glucose (Fig. 4A), which was merely 4.9% higher than that of CGS8. Therefore, the increased ALS and ALDC activity appeared to still be insufficient to increase acetoin production.
To further pull pyruvate into acetoin synthesis pathway, the previously constructed alsSD over-expression plasmid pEC-XK99E-alsSD-ΔlacIq , was introduced into CGS9 to yield CGS11. Both the activities of ALS and ALDC were significantly improved compared with CGS9. Moreover, the intracellular pyruvate pool was further decreased by 6.7% compared with that of CGS9 (Fig. S3B), indicating that more flux might have been re-directed toward acetoin synthesis. However, the acetoin production was still not obviously enhanced. When glucose was depleted at 34 h, the final acetoin titer was 16.10 g/L with a yield of 0.419 g/g glucose (Fig. 4B), which was only 7.7% higher than that of CGS8. Although the intracellular pyruvate pool of CGS11 was still 12.1% higher than that of CGR6, the carbon fluxes from pyruvate to competing pathways were successfully controlled as the titers of the main by-products, acetate (0.29 versus 1.12 g/L), glycerin (0.06 versus 0.23 g/L) and succinate (0.12 versus 2.63 g/L), were much lower than those of CGR6. No 2,3-butandiol or lactate was detected. Moreover, despite the decreases of cell growth and glucose consumption, the acetoin productivity was still comparable to that of strain CGS8. Therefore, strain CGS11 was a promising candidate for further scale-up fermentation.
Notably, both strain CGS9 and CGS11 showed much higher acetoin yields at 24 h (Table 2). The acetoin yield of strain CGS11 was even beyond the maximum theoretical yield of 0.489 g/g glucose and reached 0.498 g/g glucose at 24 h, which was mainly caused by the rich nutrients from the yeast extract in the CGXIIP medium. Moreover, all the detected by-products remained at low concentrations and biomass was relatively lower in CGS11 at 24 h (Table 2). Unexpectedly, the acetoin yield was then noticeably decreased to 0.419 g/g glucose at 34 h. The decreases of ALS and ALDC activities over time (Table 3) were initially suspected to be responsible for the decreased acetoin yield. However, the yields of by-products or biomass were not accordingly increased. Another possibility is that acetoin was re-used as an alternative carbon source when glucose was almost exhausted. Acetoin utilization is mainly catalyzed by the acetoin dehydrogenase (AoDH) complex [20, 49]. However, no candidate gene encoding a putative AoDH was identified in C. glutamicum through homologous sequence alignment (data not shown). Furthermore, acetoin degradation was more clearly observable when glucose was exhausted in fermentations of strains CGR6, CGS1 and CGS2 (Figs. 2A-C), but their biomass instead decreased during acetoin degradation. Thus, it did not appear that acetoin was consumed as a reserve carbon source, which would be expected to further support cell growth.
It is worth noting that two molecules of acetoin (C4H8O2) can be chemically converted to one molecule of 2,3,5,6-tetramethylpyrazine (TMP, C8H12N2, also called ligustrazine) in the presence of inorganic ammonium salts such as (NH4)2SO4 or diammonium phosphate [6, 30]. As shown in Fig. S4, no TMP was accumulated in the fermentation broth of strain CGS11 at 24 h, but a titer of 1.27 g/L TMP was indeed detected at 34 h, which corresponded to a consumption of 1.64 g/L acetoin. Taking this portion of the generated acetoin into calculation, the yield could even reach a high level of 0.462 g/g glucose, which indicated that 94.4% of carbon fluxes had been directed toward acetoin synthesis.
Fed-batch fermentation of CGS11 for efficient (3R)-acetoin production
After considering the promising shake-flask results, strain CGS11 was chosen for fed-batch fermentation in order to evaluate its potential for further industrial application. Before scale-up production, CGS11 was evaluated in the LBRC medium, which was optimized for acetoin production in C. glutamicum in our previous work . As shown in Fig. 5, the acetoin production was significantly enhanced with a titer of 23.53 g/L (3R)-acetoin and a yield of 0.553 g/g glucose when glucose was exhausted at 29 h. The acetoin production in CGS11 was also much higher than that in our previous optimal strain CGR7 (17.10 g/L (3R)-acetoin with a yield of 0.428 g/g glucose from LBRC medium) . The rich nutrients such as lactate (initial 2.07 g/L) and amino acids from corn steep liquor (CSL) in LBRC medium should be the major reason for the improved acetoin production. And the remarkable effect on acetoin production with CSL addition was consistent with previous report in B. subtilis [50, 51]. Moreover, when using organic nitrogen source to replace inorganic nitrogen source (NH4)2SO4, no TMP was detected (Fig. S4), which might be another reason for the improved acetoin production. Given that CSL is cheap and abundant, the cost caused by addition of CSL could be easily made up by the significant enhancement of acetoin production. The LBRC medium was therefore adopted for further fed-batch fermentation.
During the fed-batch fermentation process, the cell growth, residual glucose, product concentrations and relative dissolved oxygen (DO) were determined. As shown in Fig. 6, a titer of 102.45 g/L acetoin was obtained at 55 h, corresponding to an average productivity of 1.86 g/L/h. The final fermentation volume was 2.45 L and a total of 625.85 g glucose was added in 7 batches, while the remaining glucose amount at 55 h was 11 g/L. Therefore, the acetoin yield was 0.419 g/g glucose, reaching 85.7% of the theoretical yield. The optical purity of the produced (3R)-acetoin surpassed 95% (Fig. S5), which was consistent with our previous results . No TMP was detected during the entire fed-batch fermentation (Fig. S4). The final concentrations of acetate, succinate, glycerin and lactate were 1.94, 0.60, 0.41 and 0.33 g/L, respectively. However, α-ketoglutarate, which was undetectable during batch fermentation in shake flask, started to accumulate after 33 h and reached a final concentration of 1.71 g/L at 55 h. Nevertheless, all the by-products remained at acceptably low concentrations.
We noticed that the glucose consumption and acetoin production were prone to stop when acetoin titer surpassed 100 g/L (from 55 to 62 h, as shown in Fig. 6). With an initial acetoin concentration of 80 g/L in shake fermentation, the specific growth rate of B. amyloliquefaciens was decreased by 99% , while wild-type B. subtilis 168 was even unable to grow on plates containing 50 g/L acetoin . Therefore, the toxicity of the very high acetoin concentrations will no doubt prevent glucose assimilation and acetoin production of CGS11, and even activate its underlying acetoin catabolism. Since acetoin metabolism is complex and still not clearly elucidated, several successful strategies, such as physical/chemical mutagenesis , adaptive evolution [52, 53] or omics focusing on global transcriptional/metabolism level responses to acetoin stress , could be adopted to deeply understand and improve the acetoin tolerance of C. glutamicum in future studies.
To our best knowledge, the value of 102.45 g/L is the highest titer of highly enantiomerically enriched (3R)-acetoin reported to date (Table 4), as well as the best result obtained for acetoin production via microbial fermentation. Moreover, the yield and productivity were also good enough to merit further industrial application. Furthermore, the strains engineered in this work could also be used as platform strains to be further developed as microbial cell factories for production of (3S)-acetoin and chiral 2,3-butanediol.