The efficient production of enantiomerically pure chemicals from renewable resources has gained considerable attention especially in the fine chemical/pharmaceutical industry. Stereo-selective chemical processes generally employ expensive chiral catalysts, require harsh physical conditions and solvents, and suffer from extensive byproduct formation. In contrast, enzyme-catalyzed reactions are highly stereo-selective and can be performed in aqueous solutions under mild conditions [1]. As a result, replacing chemical processes by biological ones for the synthesis of chiral compounds has been extensively investigated not only due to superior stereo-selectivity of enzymatic reactions but also due to sustainability as an implementation of green chemistry [2–5]. One example is the production of hydroxyacids, a family of versatile chiral molecules containing one hydroxyl group and one carboxyl group [6]. These molecules have the potential to serve as useful chiral building blocks for a diverse range of products, including polyhydroxyalkanoates (PHAs) (biodegradable polymers) and optically-active fine chemicals, such as pharmaceuticals, vitamins, antibiotics, and flavor compounds [7–10]. Naturally, hydroxyacids are primarily found to be polymerized as PHAs where they serve as intracellular storage materials for numerous microbes. Those PHAs consist mostly of monomers with 3-hydroxy, 4-hydroxy, and 5-hydroxy groups with different lengths of main and side chains [11].

Among the hydroxyacid monomers, 3-hydroxybutyrate (3HB) is the most prolific, with several reports on engineering E. coli for its production from renewable feedstocks [5, 12–15]. Biosynthesis of 3HB begins with the condensation of two acetyl-CoA molecules, a commonly found cellular metabolite regardless of carbon source (Figure 1). However, economically-feasible production of longer-chain hydroxyacids is complicated by issues such as low yields and high prices of feedstocks due to the need to supplement a second carbon source. One example of such hydroxyacids is 3-hydroxyvalerate (3HV). The production of 3HV has been realized by the hydroxylation of valeric acid through fermentation of Candida rugosa [16]. It has also been reported that 3-hydroxyvaleronitrile can be converted into 3HV using the nitrilase activity of Comamonas testosteroni [17]. More recently, direct biological production of 3HV was demonstrated using recombinant P. putida KT2440 and levulinic acid as substrate, although the levulinic acid metabolism pathway in P. putida KT2440 has not yet been fully elucidated [18]. In the aforementioned cases, valeric acid, 3-hydroxyvaleronitrile, and levulinic acid were supplied as secondary carbon sources (in addition to glucose). Additionally, the chirality and/or enantiopurity of the 3HV produced in the above-mentioned studies is unclear as they did not report whether the synthesized 3HV was in the R, S, or racemic form. Alternatively, 3HV can be obtained through either the in vivo or in vitro enzymatic depolymerization of synthesized poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV), a well known biodegradable polymer marketed as Biopol™ which is produced by the natural PHA accumulating bacterium Ralstonia eutropha when grown on glucose and propionate [19]. The production of PHBV has also been reported in recombinant E. coli upon introduction of the PHA biosynthesis genes of R. eutropha and when grown in glucose medium supplemented with valine or threonine [20]. Regardless of whether the end product is 3HV or PHBV, it can be generally concluded that supplementation of a second carbon source, such as valeric acid, 3-hydroxyvaleronitrile, levulinic acid, propionate, valine, or threonine in addition to glucose, is necessary to provide the 5-carbon unit precursor of 3HV. Unfortunately, the high price and/or toxicity of the added second carbon sources could limit industrial production of 3HV [21]. Therefore, synthesis of 3HV from a single carbon source has been proposed as an efficient and sustainable avenue in contrast to the above-mentioned systems.

A novel pathway for the production of PHBV solely from glycerol has been established in recombinant Salmonella enterica Serovar Typhimurium, containing a heterologous pathway that converts succinyl-CoA to propionyl-CoA, the essential precursor molecule of 3HV-CoA in PHBV synthesis [22]. However, expensive cyanocobalamin (CN-B₁₂) was supplemented to the medium to provide the precursor of coenzyme B₁₂ required for the activity of one of the enzymes in the B₁₂-dependent biosynthetic pathway. It should also be noted that the pathway only functioned in S. enterica, a pathogen, but not E. coli, thus limiting its applicability to other industrially-relevant host organisms. In this study, we proposed an alternative biosynthetic pathway that does not require coenzyme B₁₂ for its functionality to synthesize 3HV from glucose or glycerol. Specifically, we metabolically engineered E. coli to exploit its native metabolism for endogenous supply of propionyl-CoA via the threonine biosynthesis pathway, and introduced a heterologous pathway for chiral 3HV biosynthesis using acetyl-CoA and propionyl-CoA as precursor molecules. As stated above, several previous methods for producing 3HV did not focus on enantiopure synthesis. Similarly, due to the stereospecific constraints of PHBV synthesis, in which polymers are composed exclusively of (R)-3HB and (R)-3HV monomer units, the synthesis of (S)-3HV from PHBV remains effectively impossible. On the contrary, our proposed pathway makes possible the direct synthesis of both enantiomerically pure (R)-3HV and (S)-3HV.

We have identified a pathway which combines elements of our previously developed chiral 3HB biosynthesis pathway together with the natural threonine biosynthesis pathway of E. coli for direct biosynthesis of chiral 3HV (Figure 1). In the proposed pathway, chiral 3HV is produced from direct hydrolysis of 3HV-CoA catalyzed by a thioesterase II (encoded by tesB) where 3HV-CoA is obtained from condensation of one acetyl-CoA and one propionyl-CoA to form 3-ketovaleryl-CoA catalyzed by a thiolase (encoded by bktB), followed by a reduction of the 3-ketovaleryl-CoA to 3HV-CoA catalyzed by a 3-hydroxybutyryl-CoA dehydrogenase. Here, two enantio-selective 3-hydroxybutyryl-CoA dehydrogenases were utilized to control the chirality of 3HV-CoA produced. The NADPH-dependent dehydrogenase encoded by phaB produces (R)-3HV-CoA while the NADH-dependent dehydrogenase encoded by hbd produces (S)-3HV-CoA. It should be noted that in order to yield the highest 3HV titers and 3HV/3HB ratios, BktB was used as the thiolase in this study as opposed to other thiolases such as PhaA from R. eutropha H16 or Thil from C. acetobutylicum ATCC 824 because BktB has been shown to have highest in vitro enzyme activity towards the C₅ substrate while PhaA and Thil were specific towards the C₄ substrate [19]. Next, a pathway allowing for endogenous propionyl-CoA synthesis from glucose or glycerol, through the threonine metabolic pathway intermediate 2-ketobutyrate, was introduced to circumvent the need for feeding propionate. To examine the upstream pathway for endogenous supply of propionyl-CoA, we used a bottom-up approach where 2-ketobutyrate and threonine were, at first, fed to provide propionyl-CoA, in addition to glucose, to support 3HV production. In the final stage, a single carbon source of glucose or glycerol was used to provide both acetyl-CoA and propionyl-CoA to support 3HV biosynthesis in our metabolically engineered E. coli.

Overall, in this study we successfully demonstrated the direct biological production of enantiomerically pure (R)-3HV and (S)-3HV from a single carbon source. Improvements of the biosynthetic pathway and E. coli host strains have also been carried out to elevate 3HV titers and 3HV/3HB ratios.

In general, two approaches can be taken to engineer E. coli for direct 3HV production via the threonine biosynthesis pathway. The first is to utilize an existing threonine producer, such as E. coli ATCC 21277 [33], followed by further engineering to introduce our constructed 3HV pathway. However, this and other available threonine producing strains have typically been developed through multiple rounds of random mutagenesis and selection due to the difficulty of engineering this highly regulated and complex metabolic network. Although there are several successful cases in developing industrial threonine producers by such approaches, resultant strains usually also suffer from undesired phenotypes including, for example, growth retardation, low transformation efficiency, and by-product formation as a result of random mutations [34]. In addition, other uncharacterized mutations may hinder further strain development as often needed. Fortunately, recent advances in computational genomics have allowed for rational development of production strains [34]. Therefore, as a second approach, a genetically-defined threonine producing strain was established and introduced with the 3HV pathway to achieve direct microbial production of chiral 3HV from glucose or glycerol.

As seen in Figure 5 acetate is the major byproduct to the production of hydroxyacids (3HB and 3HV). In an effort to decrease acetate and increase 3HV production, a mutant strain HCT 20 with deletions on atoDA, poxB, and ackA-pta genes was developed. Counter-intuitively, the recombinant acetate pathway knockout strains of HCT 20 and HCT 21 produced slightly more acetate and less 3HB than recombinant HCT 10. We suspected that the enzymatic activity responsible for acetate production was restored by Ptb-Buk and TesB in the recombinant HCT 20 and HCT 21. In fact, in a separate experiment, both enzymes were found to have CoA-removing activities on acetyl-CoA and propionyl-CoA (data not shown), so an introduction of TesB and/or Ptb-Buk to strains HCT 20 or HCT 21 would likely restore the ability to produce acetate.

Apparently, knocking out enzymes responsible for acetate production failed to reduce acetate synthesis. Alternatively, to alleviate the substrate promiscuity of TesB and Ptb-Buk on acetyl-CoA, and thus reduce acetate production, one approach called enzyme co-localization could be implemented to allow substrate channeling between enzymes [35]. For example, pathway enzymes of Hbd and TesB, catalyzing successive reactions, can be co-localized in an attempt to reduce the amount of freely floating TesB that may hydrolyze acetyl-CoA as well as to increase accessibility of 3HV-CoA by TesB. The spatial organization of the enzymes can be achieved using either the leucine zipper, a dimer resulting from interaction between leucine residues [36], or the synthetic scaffolds, constructed from protein-protein interaction domains [37]. Furthermore, expressing enzymes that would assimilate produced acetate is another way to reduce acetate accumulation. For example, acetyl-CoA synthetase (encoded by acs) from E. coli can be over-expressed to convert acetate to acetyl-CoA with the use of one ATP. While successfully demonstrated in one work [38], in our case, over-expression of acs was found to have essentially no effect on acetate reduction (data not shown). Additionally, to overcome the hurdle of acetate reduction, approaches like protein engineering of TesB and/or Ptb-Buk to alleviate their substrate promiscuity, or utilization of better isozymes with more stringent substrate specificity could also mitigate the carbon loss in the form of acetate.

Among microbes, NADH and NADPH play a central role in energy metabolism by providing the cell with the reducing power for a variety of cellular redox reactions. The availability of such cofactors could impose a huge impact on the functionality of introduced biosynthetic pathways. In fact, we have previously shown that the NADPH/NADP⁺ ratio was two- to three-fold higher than the NADH/NAD⁺ ratio under the culture conditions examined, presumably affecting in vivo activities of PhaB and Hbd and resulting in greater production of (R)-3HB than (S)-3HB [15]. Given that our proposed 3HV pathway was based on the previously established 3HB pathway, it was also expected to see the same trend of greater production of (R)-3HV than (S)-3HV, even though the cofactor dependency of 3HV synthesis may be complicated by the energetically expensive threonine biosynthesis pathway with utilization of both ATP and NADPH. In an effort to perturb the cofactor balance within the cells, thereby tuning the production of (R)-3HV and (S)-3HV, we attempted to used glycerol, a promising, abundant, and highly-reduced carbon source, to support 3HV production. Based on our calculation of reducing equivalents (e^-) of glucose and glycerol, on the same basis of 2 moles of phosphoenolpyruvate synthesized, glucose and glycerol possess, respectively, 24 and 28 reducing equivalents. Potentially, the additional four reducing equivalents can be utilized to generate two NADPH or equivalent amount of ATP. In fact, it has been experimentally confirmed that a higher intracellular NADPH/NADP⁺ ratio was observed when glycerol was used as a carbon source than glucose, and this higher ratio was also reflected in boosted production of thymidine as its biosynthesis requires NADPH as a cofactor [32]. Given that both NADPH and ATP play a central role in threonine biosynthesis, we hypothesized that the use of glycerol, which could generate more NADPH and ATP (Figure 1) relative to glucose, may favor threonine biosynthesis by directing more carbon flux towards production of propionyl-CoA, thus favoring the formation of 3HV relative to 3HB. In agreement with our hypothesis, a higher 3HV/3HB ratio was obtained in the (R)-3HV production when glycerol was used as the carbon source (Figure 6). In addition, much larger ratios of the total (R)-hydroxyacids (summation of (R)-3HB and (R)-3HV titers) to the total (S)-hydroxyacids (summation of (S)-3HB and (S)-3HV titers) were observed in glycerol-fed cultures (Figure 6) compared to glucose-fed cultures (Figure 5). We hypothesize that the higher intracellular NADPH/NADP+ ratio as a result of the use of glycerol would favor (R)-hydroxyacid biosynthesis compared to the use of glucose, thus yielding the larger ratios of total (R)-hydroxyacids to total (S)-hydroxyacids.

As mentioned previously, BktB was chosen as the primary thiolase due to its high enzymatic specificity towards the C₅ substrate. Given that E. coli has an endogenous thiolase (encoded by atoB), a deletion of atoB was expected to increase the ratio of 3HV/3HB as AtoB has been shown to prefer to condense two molecules of acetyl-CoA instead of one propionyl-CoA and one acetyl-CoA [19]. However, our preliminary result showed that the recombinant HCT 11 with an atoB deletion behaved exactly as the recombinant HCT 10, and the deletion in atoB had essentially no effect on 3HV production (data not shown), implying that atoB may not be a constitutively expressed gene. In addition, it is noteworthy that increased 3HV production in the recombinant HCT 20 relative to the recombinant HCT 10 was observed only with 2-ketobutyrate supplementation (Figure 3) but not with the threonine supplementation, solely glucose, or solely glycerol experiments (Figure 4, 5 and 6); similarly, an accumulation of propionate only occurred in the 2-ketobutyrate supplementation experiment (Figure 3), altogether, indicating that 3HV biosynthesis from glucose or glycerol is most likely limited by the precursor propionyl-CoA. Therefore, approaches to increase the availability of propionyl-CoA could enhance the 3HV production.

Carbon skeletons with even-chain number are naturally found in fatty acid metabolism, but those with odd-chain number are pretty novel. As a result, there is a good deal of interest in making odd-carbon chain molecules such as 3HV (C5) and propionate (C3) because they are so much harder to get to than even-carbon chain ones such as acetate (C2) and butyrate/butanol (C4). This paper opens the way for biosynthesis of the odd-carbon chain molecules from renewable feedstocks. Taking together, our work represents the first report of direct microbial production of enantiomerically pure 3HV from a single carbon source. In addition, we have explored the production of each stereoisomer of 3HV across different genetically altered E. coli strains, along with various enzyme homologs, for enhanced chiral 3HV production. Further engineering of host strains and pathway enzymes should lead to higher 3HV titers and a more economical bioprocess for the production of chiral 3HV.