The strains and plasmids used in this study are listed in Supplementary Table S1. Escherichia coli strains DH5α and JM109 were routinely cultivated at 30°C or 37°C in a lysogeny broth (LB) medium composed of 1% (w/v) tryptone (Nacalai Tesque, Kyoto, Japan), 0.5% (w/v) yeast extract (Becton Dickinson, Franklin Lakes, NJ, United States), and 1% (w/v) NaCl. Carbenicillin (100 μg/ml), kanamycin (100 μg/ml), gentamycin (30 μg/ml), and/or chloramphenicol (30 μg/ml) were added into the medium when necessary.

DNA manipulations were carried out according to standard procedures, and PCR reactions were performed with KOD-Plus ver.2 DNA polymerase (Toyobo, Osaka, Japan). The sequences of oligonucleotide primers used in this study are shown in Supplementary Table S2. The transformation of E. coli strains were performed according to conventional chemical competent or electroporation procedures.

pBKS-PCJAB was constructed by inserting a KpnI-XbaI restriction fragment of pBBR1phaP(D4N)CJ _Ac AB _Re (Ushimaru et al., 2015) containing P _Ac -phaP _D4N C _NSDG J _Ac and P _Re -phaAB1 _Re into pBluescript II KS(+) at the corresponding sites, where P _Ac and P _Re are native promoter regions of phaPCJ from A. caviae and phaCAB1 from R. eutropha, respectively. pBKS-PCJA was obtained by deleting the phaB1 _Re region from pBKS-PCJAB by inverse PCR followed by self-ligation. pBtac-CJ_ReE was constructed by replacing the phaP promoter region in pBPP-ccr_MeJ4a-emd (Insomphun et al., 2015) by the tac promoter region amplified from pBBRtac (Fukui et al., 2011). pBtac-CJ_ReEB and pBtac-CJ4_PaE were derivatives of pBtac-CJ_ReE obtained by insertion of bktB _Re at downstream of emd _Mm and replacement of phaJ4a _Re by phaJ4 _Pa , respectively. pSTV-HC was constructed by replacement of the lac promoter-lacZα region in pSTV28 by a tandem of had _Re and crt2 _Re fused with a tac promoter. Further insertion of bktB _Re into pSTV-HC at the downstream of crt2 _Re gave pSTV-HCB. pSTV-PCB was constructed by replacement of had _Re in pSTV-HCB by paaH1 _Re . pMW-Gm-pxtD_EAARABC for phosphite dehydrogenation was constructed by inserting an EcoRI-SacI restriction fragment containing pxtD _EAAR ABC excised from pBBR1MCS2::pxtD _EAAR ABC (Miyahara et al., 2018) into pMW218-Gm which is a pMW218 derivative harboring Gm^r instead of Km^r.

E. coli single-gene knockout mutants and the parent strain BW25113 in the Keio collection (Yamamoto et al., 2009) were obtained from the National BioResource Project (National Institute of Genetics, Japan): E. coli. The Km ^r gene inserted within the target gene was removed by using temperature-sensitive pCP20 harboring flp as described previously (Cherepanov and Wackernagel, 1995). Double-gene knockout strains JW∆cra∆rng, JW∆pgi∆rng, and JW∆pta∆poxB were constructed by additional deletion of the second gene (rng or poxB) from the corresponding single deletion strains by using λ-Red recombinase-mediated recombination (Datsenko and Wanner, 2000). In the case of JW∆pta∆poxB for an example, the FLP recombinase target (FRT)-Km^r-FRT region flanked to 50-bp extensions in poxB was amplified from the genomic DNA of JW∆poxB by PCR. JW∆pta/pKD46 grown in the presence of 0.1% l-arabinose for the expression of λ-Red recombinase was transformed with the corresponding PCR product by electroporation. The cells were incubated in an SOC medium at 37°C for 1 h and inoculated onto an LB plate medium containing 50 μg/ml kanamycin. The colonies formed on the Km-LB plate medium were picked, and ampicillin sensitivity of the clones was confirmed in the LB medium with 50 μg/ml ampicillin. Insertion of the Km^r cassette at the target locus in the obtained clones was confirmed by PCR, and the cassette was then removed by using pCP20.

E. coli strains were cultivated at 30°C in 100 ml of LB medium supplemented with 20 g/L glucose. Appropriate antibiotics were added at the final concentration as described previously, and 1.0 mM IPTG was added into the medium for induction of gene expression prior to inoculation. After the cultivation for 48 h with reciprocal shaking (115 strokes/min), the cells were harvested, washed once with cold deionized water, and then lyophilized. The cellular PHA content and composition were determined by gas chromatography (GC) after direct methanolysis of the dried cells in the presence of 15% sulfuric acid as described previously (Kato et al., 1996).

PHA polymers were extracted from the dried cells by stirring in chloroform for 72 h at room temperature and recovered by precipitation with methanol. The molecular weight and polydispersity were determined by gel permeation chromatography (GPC) using a Shimadzu 10A GPC system and a 10A refractive index detector equipped with a serial of Shodex K-806M and K-802 columns at 40°C. Chloroform was used as the eluent at a flow rate of 0.8 ml/min. The calibration curve was generated using polystyrene standards with a low polydispersity (STANDARD SM-105, Shodex, Tokyo, Japan).

Glucose, acetate, and pyruvate in the culture supernatant were measured using a flow injection analyzer BF-9 (Oji Scientific Instruments, Hyogo, Japan) at 25°C. Glucose concentration was determined using a glucose oxidase-based biosensor with a mobile phase composed of 100 mM Na₃PO₄, 50 mM KCl, and 1 mM NaN₃ (pH 7). Acetate and pyruvate were simultaneously measured by using a serial of pyruvate kinase- and acetokinase/pyruvate kinase/pyruvate oxidase-based biosensors, respectively, and 50 μM FAD, 50 μM thiamine pyrophosphate, 0.2 mM ATP, and 0.2 mM phosphoenolpyruvate were added to the mobile phase.

Phosphite and phosphate concentrations were determined by a capillary electrophoresis P/ACE system MDQ (Beckman Coulter, CA, United States) and anion analysis kit at 25°C. The analytes were separated with 20 kV voltage detected by the indirect detection mode. The capillary was washed with 0.1 N NaOH and 0.1 N HCl before analysis, and the samples were injected by vacuum (0.5 psi for 8.0 s) (Miyahara et al., 2018).

The pathway for the biosynthesis of P(3HB-co-3HHx) from glucose introduced into E. coli is shown in Figure 1. The C₄ monomer (R)-3HB-CoA is conventionally formed through two reaction steps: condensation of two molecules of acetyl-CoA and subsequent reduction with (R)-stereospecificity to (R)-3HB-CoA. Two kinds of β-ketothiolase [PhaA _Re and BktB _Re (Slater et al., 1998)] and NADPH-acetoacetyl-CoA reductase (PhaB1 _Re ) derived from R. eutropha are applied for these two reactions in this study. Two pathways for crotonyl-CoA generation from acetoacetyl-CoA are dehydration of (R)-3HB-CoA catalyzed by (R)-specific enoyl-CoA hydratase from A. caviae (PhaJ _Ac ) (Fukui et al., 1998; Mifune et al., 2010) and (S)-specific pathway mediated by NAD⁺-(S)-3-hydroxyacyl-CoA dehydrogenase and crotonase [(S)-specific enoyl-CoA hydratase] from R. eutropha (Had _Re and Crt2 _Re , respectively) (Segawa et al., 2019). Crotonyl-CoA is then converted to butyryl-CoA or ethylmalonyl-CoA by bifunctional Ccr _Me catalyzing NADPH-dependent reduction or reductive carboxylation, respectively (Erb et al., 2007). The latter is decarboxylated to butyryl-CoA by Emd _Mm (Linster et al., 2011), thus consequently converting crotonyl-CoA to butyryl-CoA by the combination of Ccr _Me and Emd _Mm . Because the aforementioned BktB _Re , Had _Re , and Crt2 _Re showed rather broad substrate specificity accepting C₄–C₆ intermediates, the three enzymes are expected to establish reverse β-oxidation for butyryl-CoA to 2-hexenoyl-CoA. A medium-chain-length-specific (R)-enoyl-CoA hydratase from R. eutropha (PhaJ4a _Re ) (Kawashima et al., 2012) then generates (R)-3HHx-CoA from 2-hexenoyl-CoA. (R)-3-Hydroxyacyl-CoAs of C₄ and C₆ are copolymerized by PhaC_NSDG (Asn149Ser/Asp171Gly double mutant of PHA synthase from A. caviae) (Fukui and Doi, 1997; Tsuge et al., 2007), which has been shown to synthesize P(3HB-co-3HHx) with a higher 3HHx composition than the wild-type enzyme.

These genes of the heterologous enzymes were introduced into E. coli by three compatible plasmids. pBKS-PCJAB harbored phaP _D4N-phaC _NSDG-phaJ _Ac and phaA _Re -phaB1 _Re , both with the native promoter regions derived from A. caviae and R. eutropha, respectively. phaP _D4N is a gene of Asp4Asn (D4N) mutant of PhaP (granule-associated protein, phasin), in which the nucleotide change corresponding to the D4N mutation has been reported to elevate expression levels of the phaP-C-J gene cluster in E. coli (Ushimaru et al., 2015). The lac promoter in the plasmid was also expected to contribute to the high expression of phaP-C-J potentially important for the efficient synthesis of PHAs. pSTV-HCB harbors a tandem of had _Re , crt2 _Re , and bktB _Re located downstream of P _tac . pBtac-CJ_ReE was constructed by insertion of a tandem of ccr _Me , phaJ4a _Re , and codon-optimized emd _Mm into pBBRtac, a broad-host range expression plasmid containing P _tac (Fukui et al., 2011).

E. coli JM109 strains were transformed by the constructed plasmids and cultivated for PHA production at 30°C in the LB medium containing 2% (w/v) glucose with IPTG induction. The strain harboring phaP _D4N C _NSDG J _Ac and phaAB _Re on a high-copy number plasmid (pBKS-PCJAB) accumulated a large amount (5.6 g/L) of P(3HB) homopolymer with 57 wt% of the cellular dry weight (entry 1, Figure 2 and Supplementary Table S3). As this PHA production was higher than the previous case using a medium-copy number plasmid [pBBR1phaPCJ _Ac AB_Re (Ushimaru et al., 2015)], pBKS-PCJAB was used as the module for (R)-3HB-CoA formation and polymerization for further engineering. Unlike in R. eutropha (Insomphun et al., 2015; Zhang et al., 2019), introduction of genes for butyryl-CoA formation (ccr _Me -phaJ4a _Re -emd _Mm ) did not enable the P(3HB)-producing E. coli to synthesize P(3HB-co-3HHx) from glucose, as the transformant having pBKS-PCJAB and pBtac-CJ_ReE still produced P(3HB) homopolymer with a low content of 38 wt% (entry 2). P(3HB-co-3HHx) biosynthesis from glucose was finally achieved by further introduction of the third plasmid pSTV-HCB harboring the genes for reverse β-oxidation (had _Re -crt2 _Re -bktB _Re ). This triple transformant produced 2.8 g/L of P(3HB-co-14 mol% 3HHx) with 41 wt% content after 48 h cultivation (entry 3). The incorporation of the 3HHx unit into PHA was not observed when the cultivation temperature was elevated up to 37°C, probably due to the unstable property of any of the heterologous enzymes for formation of (R)-3HHx-CoA monomer. All the cultivation was thus carried out at 30°C in this study.

The time-course of growth and PHA synthesis by the strain JM109/pBKS-PCJAB/pBtac-CJ_ReE/pSTV-HCB are shown in Figure 3. The cell growth (residual cell mass, RCM) was saturated at 24 h (Figure 3A), and PHA accumulation was started from 12 h and continued to 48 h (Figures 3A,B). Although the PHA accumulation was looked to be decreased after 48 h, the difference of PHA amounts between 48 and 72 h was not significant (p-value > 0.25). The 3HHx composition was constant ranging within 12–14 mol% regardless of the cultivation time.

The number average-molecular weight (M _n ) and polydispersity (PDI) of P(3HB-co-14 mol%) synthesized by JM109/pBKS-PCJAB/pBtac-CJ_ReE/pSTV-HCB were determined to be 5.80 ×10⁴ and 1.55, respectively, by GPC (Table 1). The molecular weight values were roughly half of P(3HB) homopolymer synthesized by the strain solely having pBKS-PCJAB, which would be due to some effect of incorporation of the C₆ units on the enzymatic polymerization. Both the PHAs synthesized by the recombinant E. coli in this study showed a lower molecular weight but narrow distribution than P(3HB-co-12 mol% 3HHx) synthesized by the previously engineered R. eutropha strain from glucose (Zhang et al., 2019) (M _n and PDI were 29.5 ×10⁴ and 3.59, respectively). It has been reported that introduction of an additional copy of phaC resulted in the decrease in PHA molecular weight in R. eutropha (Kawashima et al., 2015). This agreed with the lower molecular weight in E. coli harboring phaC on the high-copy number plasmid than the previous R. eutropha strain having the single phaC gene, probably attributed to more active centers in the catalytic polymerization.

The roles of PhaB1 and reverse β-oxidation on the formation of (R)-3HA-CoAs of C₄ and C₆ in the recombinant E. coli were investigated by gene deletion analyses. The introduction of pBKS-PCJA lacking phaB1 _Re resulted in a significant decrease in the accumulated polyester (entry 4). Although the 3HHx composition of 22 mol% was higher than that by the strain having phaB1 _Re , the amounts of 3HB and 3HHx units in the polymer fraction were remarkably reduced. This indicated the critical role of PhaB1 _Re in the provision of both the monomers within the cells. When pSTV-HC-lacking bktB _Re was used, the strain synthesized P(3HB) homopolymer with low content (entry 5). The strain lacking had _Re and crt2 _Re but harboring bktB _Re was constructed by double transformation with pBKS-PCJAB and pBtac-CJ_ReEB, in which the latter plasmid contained bktB _Re at downstream of emd _Mm . The resulting strain synthesized PHA containing only a trace fraction of the 3HHx unit although the accumulation was as much as that by the full triple transformant (entry 6), indicating the essential roles of the (S)-specific enzymes for formation of the C₆-monomer.

It has been clarified that two kinds of (S)-3HB-CoA dehydrogenases Had and PaaH1 were active in R. eutropha grown on fructose or soybean oil (Segawa et al., 2019). Both the dehydrogenases have broad substrate specificity toward 3-oxoacyl-CoAs of C₄–C₈, in which PaaH1 showed slightly higher catalytic efficiency to the C₆ substrate than Had, while one of the (R)-enoyl-CoA hydratases from Pseudomonas aeruginosa (PhaJ4 _Pa ) was reported to show activity to trans-2-enoyl-CoAs of C₆ and longer but no activity to crotonyl-CoA (Tsuge et al., 2003), different from PhaJ4a _Re showing low activity to crotonyl-CoA (Kawashima et al., 2012). We here replaced had _Re and phaJ4a _Re in the plasmids by paaH1 _Re and phaJ4 _Pa , respectively. As shown in Supplementary Table S4, the strains having PaaH1 and/or PhaJ4 _Pa produced P(3HB-co-3HHx) with similar cellular content and composition when compared to the parent strain having Had _Re and PhaJ4a _Re . The catalytic properties of (S)-3HB-dehydrogenase and (R)-enoyl-CoA hydratase were not significantly affected by the PHA copolymer biosynthesis in E. coli.

Further modifications were introduced based on previous knowledge for central metabolisms in E. coli. Several E. coli mutant strains were obtained from the Keio collection (Yamamoto et al., 2009) and used after the excision of kanamycin-resistant marker genes. The mutant strains transformed with pBKS-PCJAB/pBtac-CJ_ReE/pSTV-HCB were subjected to PHA synthesis from glucose. E. coli BW25113 (a parent of Keio collection strains) harboring the three plasmids produced P(3HB-co-16 mol% 3HHx) with the cellular content of 32 wt% (entry 7, Figure 4 and Supplementary Table S5), whose property was similar to that of the JM109-based strain. Cra (FruR) is a global transcriptional regulator modulating central metabolisms and respiration (Ramseier et al., 1995; Saier and Ramseier, 1996), and PdhR is a pyruvate-sensing transcriptional repressor for genes of pyruvate dehydrogenase complex and respiration (Gohler et al., 2011; Maeda et al., 2017). Although we assumed that acetyl-CoA formation following PHA synthesis may be enhanced by deletion of cra (entry 8) or pdhR (entry 9), these modifications resulted in a marked decrease in both cell growth and PHA production. Acetate was formed by the JW∆cra-based strain in higher levels (35 mM) than the parent strain (Supplementary Table S5). RNase G, encoded by rng, is an endonuclease-cleaving rRNA precursor and glycolysis gene-derived mRNA. It was reported that pyruvate was overproduced by ∆rng∆cra double mutant of E. coli MG1655 possibly owing to enhanced glycolysis (Sakai et al., 2007). However, we here observed that the strain lacking rng produced PHA with a high cellular content (43 wt%) whereas only a trace amount of the 3HHx unit was detected in the polymer fraction (entry 10). The strain doubly lacking rng and cra (entry 11) showed the PHA biosynthesis property similar to that of the single ∆rng strain. Inactivation of phosphoglucose isomerase Pgi has been demonstrated to shunt glucose metabolisms from the Embden–Meyerhof pathway to NADPH-forming Entner–Doudroff and pentose-phosphate pathways, which was expected to be favorable for PHA synthesis since PhaB1 and Ccr _Me utilize NADPH as a reduced cofactor. The strain lacking pgi produced P(3HB-co-3HHx) with a high 3HHx composition (22 mol%), while the cellular PHA content was decreased to 19 wt% (entry 12). It was considered that double deletion of pgi and rng might compensate the respective disadvantages to each other; however unexpectedly, the JW∆pgi∆rng-based transformant produced P(3HB-co-6.6 mol% 3HHx) with a low cellular content of 19 wt% (entry 13).

We further investigated the effects of block of acetate formation (Dittrich et al., 2005) on the copolyester biosynthesis. The JW∆pta-based triple transformant lacking phosphate acetyltransferase Pta produced P(3HB-co-3HHx) with a high cellular content but low 3HHx composition (41 wt% and 1.9 mol% 3HHx, respectively) (entry 14) and still secreted acetate (Supplementary Table S5). The acetate formation was completely lost by double deletion of pta and poxB encoding pyruvate oxidase (entry 15). However, pyruvate was secreted instead of acetate, and both the cell growth and PHA production were severely impaired.

Miyahara et al (2018) reported that NADPH supply coupled with phosphite oxidation led to a 3.2-fold increase in P(3HB) production in recombinant E. coli under non-growth conditions. We thus investigated the effects of the phosphite oxidation system on P(3HB-co-3HHx) biosynthesis by E. coli. The fourth plasmid pMW-Gm-ptxD_EAARABC was constructed, in which ptxD _EAAR and ptxABC are genes encoding the Glu175Ala/Ala176Arg double mutant of phosphite dehydrogenase and phosphite transporters, respectively, both derived from Pseudomonas stutzeri. It has been reported that PtxD was NAD⁺-specific, but the mutant enzyme could accept NADP⁺ as well as NAD⁺ as the electron acceptor (Woodyer et al., 2003). The quadruple transformant of E. coli JM109 was subjected to two-stage cultivation on glucose in a MOPS-based mineral medium. The transformant having ptxD _EAAR ABC produced P(3HB-co-11 mol% 3HHx) with a cellular content of 31 wt% from glucose in the absence of phosphite, while the cellular content and 3HHx composition were significantly increased up to 48 wt% and 24 mol%, respectively, by addition of 5 g/l phosphite into the medium (Table 2).