From Bugs to Bioplastics: Total (+)‐Dihydrocarvide Biosynthesis by Engineered Escherichia coli

Abstract The monoterpenoid lactone derivative (+)‐dihydrocarvide ((+)‐DHCD) can be polymerised to form shape‐memory polymers. Synthetic biology routes from simple, inexpensive carbon sources are an attractive, alternative route over chemical synthesis from (R)‐carvone. We have demonstrated a proof‐of‐principle in vivo approach for the complete biosynthesis of (+)‐DHCD from glucose in Escherichia coli (6.6 mg L−1). The pathway is based on the Mentha spicata route to (R)‐carvone, with the addition of an ′ene′‐reductase and Baeyer–Villiger cyclohexanone monooxygenase. Co‐expression with a limonene synthesis pathway enzyme enables complete biocatalytic production within one microbial chassis. (+)‐DHCD was successfully produced by screening multiple homologues of the pathway genes, combined with expression optimisation by selective promoter and/or ribosomal binding‐site screening. This study demonstrates the potential application of synthetic biology approaches in the development of truly sustainable and renewable bioplastic monomers.

We propose am ore direct route, in which a M. spicata-like pathway to (R)-carvone production is combined with specific alcohol dehydrogenase (ADH) and CHMO enzymes within one recombinant strain of E. coli (Scheme 1). Limitations in C5 isoprenoid precursor production would be minimised by incorporating as econd construct containing ae ukaryotic mevalonate pathway to enablel actone production from simple carbon sources. [15] The latter pathway wass hown previously to substantially increase the in vivo production of the limonene derivativep erillyl alcohol in E. coli. [15a] Homologues and modifications of key enzymes were screenedi ni nv ivo reactions to develop an optimised pathway to (+ +)-DHCD. Functional pathway constructs underwent further modifications of the controlling The monoterpenoid lactone derivative (+ +)-dihydrocarvide ((+ +)-DHCD) can be polymerised to form shape-memory polymers. Synthetic biology routes from simple, inexpensive carbon sourcesa re an attractive, alternative route over chemical synthesis from (R)-carvone. We have demonstrated ap roofof-principle in vivo approach for the complete biosynthesis of (+ +)-DHCD from glucose in Escherichia coli (6.6 mg L À1 ). The pathway is based on the Mentha spicata route to (R)-carvone, with the addition of an 'ene'-reductase and Baeyer-Villiger cy-clohexanone monooxygenase. Co-expression with al imonene synthesis pathway enzymee nables complete biocatalytic productionwithin one microbial chassis. (+ +)-DHCDwas successfully produced by screening multiple homologues of the pathway genes, combined with expression optimisation by selective promoter and/orr ibosomal binding-site screening. Thiss tudy demonstrates the potentiala pplication of synthetic biology approaches in the development of truly sustainable and renewable bioplastic monomers.
elements (e.g.,p romoters) to enable significant levels of the terminal lactone product to be generated.

Limonene hydroxylation
The entry step into the M. spicata biosynthesis of (R)-carvone is the hydroxylationo f( S)-limonene to (1S,5R)-carveol (Scheme 1) catalysed by the cytochrome P450 enzyme limonene-6-hydroxylase (L6H) with its electron-transfer partner cytochrome P450 reductase( CPR). [16] Based on earlier studies, we generated an N-terminally truncated and modified form of L6H [17] (L6H m )t o eliminate the signal sequence and increases oluble expression in E. coli. Unfortunately only ap artial sequence was available for mint CPR (205 aa;G enBank:A W255332)f rom studies with expressed sequence tags (EST) from mint glandular trichomes. [18] However the CPR from Salvia miltiorrhiza (Chinese sage;S mCPR) has high amino acid sequence homology( 92 %) to the EST CPR sequence from mint.
Additionally,e arly studies with native L6H showed that hydroxylation occurs in the presence of aC PR from Arabidopsis thaliana (AtCPR). [17] Therefore, we generated C-terminally His 6tagged versions of both SmCPR andA tCPR to determinet he best electron-transfer partner for L6H m .
Initial co-expressionc onstructs of L6H m with either SmCPR or AtCPR wereg enerated in plasmidp CWoriu nder the control of a tac promoter. [17] In vitro biotransformations of cell lysates with limonene showed only poor (1S,5R)-carveolp roduction by L6H m with either SmCPR or AtCPR over 24 h( e.g.,( 95.2 AE 3.3) mm with SmCPR;1 .9 %y ield;F igure S6 in the Supporting Information). Therefore new L6H m /CPR constructs were generated withoutH is 6 tags and controlled by either araBAD (arabinose) or tet (tetracycline) promoters on different plasmid backbones (pBbB8k and pBbE2k,r espectively). We performedi n vivo reactions for the detection of functional L6H m -CPR pairs insteado fu sing purified proteins or cell lysates. Thisi sd ue to difficulties in obtaining sufficient quantities of soluble, active membrane-associated L6H m (resultsn ot shown). This method involved the co-expression of L6H m -CPR constructs with alimonene production plasmid pJBEI6410, [15a] thereby eliminating the need to supplementt he culture with limonene. Cultures were grown in the presence or absence of an onane overlay, which efficientlys equestered the monoterpenoids away from the aqueousp hase to minimise cytotoxicity.
(S)-Limonenep roduction was detected in all cultures, with a range of titres of 137-220 mg L À1 /OD 600 dependento nt he L6H m -CPR construct ( Figure 1). These differences likely reflect the efficiency of production versus the rate of utilisation by the expressedL 6H m /CPR;h owever,t he nature of the L6H m -CPR plasmidb ackbone appeared to have an impact on limonene titres. The best (1S,5R)-carveol-producing construct was L6H m -SmCPR in pBbB8k ((6.7 AE 4.3) mg L À1 /OD 600 ), with the equivalent AtCPR-containing plasmid showing a2 0-foldr eduction in yield ( Figure 1). The higher than expectedv ariability in (1S,5R)-carveol yields within replicates is areflectiononthe nonoptimised growth and induction conditions;h owever,aclear preference for the sage CPR was seen. No detectable levels of (1S,5R)-carveol were found with the constructsi nt he tetracycline-inducible pBbE2k plasmid. This could be indirectly related to the higher copy number and promoters trength, leading to changes in soluble recombinantp rotein expression levels and/ or ahigher metabolic burden on E. coli. This wasseen by an increase in the relative proportion of insoluble protein expressed in these constructs (results not shown).
Optimisation trials were performed in vivo with the bestperforming construct L6H m -SmCPR in pBbB8kc o-expressed with the limonene production plasmid pJBEI6410. The presence/ absence of a n-nonaneb ilayer,c ulture density at induction, inducerc oncentration (isopropyl-b-d-thiogalactopyranoside  (Table S10). In contrast to studies with in vivo production of limonene and other monoterpenoids, [20] the presence of a n-nonanec o-solvent reduced the levelso f( 1 S,5R)-carveol productiona tl east sevenfold ((1.7 AE 0.9) vs. (12.8 AE 4.4) mg L À1 / OD 600 ). This is likely due to the sequestering of the (S)-limonene generatedb yt he pJBEI6410p lasmid into the co-solvent, thereby reducingt he intracellular concentrations and availability for the hydroxylation enzyme. Increasingt he kanamycin concentration (selective for L6H m -SmCPR) from 15 to 60 mgmL À1 led to at hreefold increasei n( 1 S,5R)-carveol. The conditions leading to the highest yields of (1S,5R)-carveol with the highest reproducibility were found to be induction at a mid-logp hase, with 25 mm IPTG and 25 mm arabinose ((33.8 AE 5.0) mg L À1 /OD 600 ).

(S)-Limonene to (R)-carvone operons
The next stage involved combining the highest performing (1S,5R)-carveol-producing construct (L6H m -SmCPRi np BbB8k) with the four ADH enzymest of ind the optimal set of biocatalysts. Each operonw as constructed by inserting the ADH gene downstream from SmCPR, separated by one of two ribosome binding sequences. RBS1 (GAATA ACTAT TTAAG AGGGAG ATTA ATAAC A) has ap redicted translation rate of 13 969, [27] whereas RBS2 (TAAGGAGGT) was chosen as it successfully increased the production of p-coumaryl alcohol in E. coli when using at ricistronic operon. [28] Each construct was co-transformed with plasmidp JBEI6410 into E. coli strain NEB10b to screen for the in vivo production of (R)-carvone from glucose.
To assess the performance of these two enzymes in E. coli,a variety of multigene constructs were generated and assessed for both (2R,5R)-dihydrocarvonea nd (+ +)-DHCD production under standard fermentation conditions. Cell extracts of each construct were tested by in vitro biotransformations in the presence of ac ommercially available( 1 S,5S)-and( 1 R,5R)-carveol mix, NAD + (IPDH) and an NADPH cofactor-recycling system (PETNR and CHMO 3M ). These early constructs contained the complete pathway from (S)-limonene to (+ +)-DHCD (L6H m -IPDH-PETNR-CHMO 3M ;L 6H m IPC 3M )e xcept for CPR, as the most suitable CPR (andA DH homologue) had not been determined at the time of pathway construction. However,t he focus of these operon designs was to generate the most suitable PETNR-CHMO 3M gene arrangement to maximise (+ +)-DHCD production from exogenously supplied( R)-carvone, so the absence of CPR and the presence of IPDH insteado fC DH was inconsequential. Full details of the production of thesec onstructs can be found in the Supporting Information (Experi-mentalS ections 1-5, Tables S1-S7 and Figures S1-S4).
The next approach to boost expression was to insert av ariety of promoters upstream of CHMO 3M .T he selected promoters were inducedb yI PTG (trc/lacO, tacII/lacO, lacUV5), rhamnose (rhaBAD) or tetracycline (PtetA), allowing either as ingle (IPTG) control over the expression of all three genes or differential control for CHMO 3M . [34] Biotransformations of cell extracts were performed with three different substrates to determine the most effective expression control system for CHMO 3M (Tables 1, S13 and S14). As expected, in each case, the highest (+ +)-DHCD production was seen in the presence of (2S,5R)-dihydrocarvone (CHMO 3M substrate), with the best yields obtained with CHMO 3M under the control of a trc/lacO promoter ((0.57 AE 0.07) mm;T able 1). When the CHMO 3M promoter was substituted for PtetA and rhaBAD, the yields decreased by 1.7-and 4.4fold, respectively.B iotransformations in the presenceo fc arveol showedasignificant decrease in (+ +)-DHCDp roduction ((0.12 AE 0.01) mm with trc/lacO). In the case of the rhaBAD-containing construct, no (+ +)-DHCD was produced in the presence of carveol. Therefore, the inclusion of the promoters trc/lacO and PtetA upstream of CHMO 3M have successfully led to the productiono f( + +)-DHCDfrom carveol.

Lactone production from glucose
Full pathway assembly wasp erformed by using the most successfulc arvone-producingc onstruct as the backbone (L6H m -SmCPR-CDH RBS1;a rabinose inducible), and inserting PETNRpromoter-CHMO 3M genes downstream of CDH. Constructs L6HIP-trc-C 3M andL 6HIP-tet-C 3M were chosen as the source of PETNR-promoter-CHMO 3M genes due to their ability to produce (+ +)-DHCD in the presenceo fc arveol. Additionally,t he L6HIPrha-C 3M construct was chosen as it generateds ignificant (+ +)-DHCD in the presence of (R)-carvone. The three dual promoterc onstructs (L6HCCP-trc-C 3M ,L 6HCCP-tet-C 3M and L6HCCP-rha-C 3M )w ere co-expressed in E. coli with the limonene synthesis plasmid for total in vivo productiono f (+ +)-DHCD lactone from glucose. Given the length of the number of steps in the pathway to (+ +)-DHCD, CHMO 3M inducer was added either at the same time as the other inducers (IPTG and arabinose) or 6hlater so as to give time for the intermediate monoterpenoid concentrationst ob uild up within the cell. The trc-promoter is IPTG inducible, so the expression of CHMO 3M in construct L6HCCP-trc-C 3M could not be postponed for 6hours, as IPTG is required for the induction of the limonene synthesis genes (pJBEI6410 plasmid).
In vivo studies showedt hat two of the three construct combinations successfully generated (+ +)-DHCD from glucose ( Figure 4). The most successful limonene to lactone-producing construct in E. coli was L6HCCP-rha-C 3M ,w hich showed around 6mgL À1 (+ +)-DHCD, dependent on the inductionc onditions. Interesting, the highest in vitro (+ +)-DHCD-producing construct L6HCCP-trc-C 3M did not show any detectable levels of (+ +)-DHCD under in vivo conditions when co-expressed with the limonene-producing plasmid. This highlights the importance of screening multiple constructs with differentc ontrolling elements, as the addition of an extra IPTG-inducible pathway can sometimes have an (unpredictable) impact on thee xpression of each recombinant gene.

Conclusions
In vivo production of fine chemicals is one possible solution to the increasing demandf or sustainable and renewable manufacturing. The cost-effectiveness of biological manufacturing strategies is dependento nt he constructiono fr ecombinant microorganisms that express the correct "assemblyl ine" of enzymes at sufficient levels. We have achieved ap roof-of-principle demonstration of in vivo production of the bioplastics precursor (+ +)-DHCD in E. coli,g rown on as imple, inexpensive carbon source.T his overcomes the severe limitations in the existing partial pathwaya pproach( limonene to lactone) caused by the addition of ac ytotoxic precursor (limonene) supply to the microorganism. [13] The in vivo production of limonenei n  Further studies are requiredt oi ncreaset he productivity and cost-effectiveness of this bio-manufacturing approach to bioplastics production.T his is necessary to increase the production titres, concomitant with the elimination of selection agents (antibiotics)a nd expensive chemical induction (e.g., IPTG and rhamnose). For example, host selection and (chromosomal)m odification could be applied to reduce the cytotoxicity and recovery of the monoterpenoids and increasec ellular export.Ahigh-throughput combinatorial approach could be appliedt os creen for the best combination of enzymeh omologues/variants, vector backbone, promoter combinationa nd gene order.H owever our demonstration of the complete in vivo production of (+ +)-DHCD is al eap forwardi nt he development of truly sustainable and renewable bioplastic monomers.
The gene encoding an N-terminally modified mature (4S)-limonene-6-hydroxylase from M. spicata (L6H m ;U NIPROT:Q 9XHE8) [17] was synthesised and subcloned without codon optimisation into pCWori (+ +)b yG eneart. The Nterminus was modified by removing the chloroplast signal sequence in addition to other modifications designed to increase its soluble expression in E. coli,a sd escribed previously. [17] TwoC -terminally His 6 -tagged cytochrome P450 reductases from A. thaliana (AtCPR;U NIPROT:Q 9SB48) and S. miltiorrhiza (SmCPR;U NIPROT:S 4URU2) were synthesised and subcloned into pET21b by incorporating codon-optimisation techniques of rare codon removal. Each gene was transformed into competent cells of E. coli strain BL21(DE3) for functional overexpression according to the manufacturers' protocols.
Limonene hydroxylation construct assembly: Functional limonene hydroxylation constructs were generated by In-Fusion cloning (Takara) [38] between PCR linearised L6H m (3'-end) in pCWori (+ +) and either AtCPR or SmCPR, with the inclusion of aShine-Dalgarno sequence between the genes (L6H m -AtCPR and L6H m -SmCPR, respectively). The constructs were transformed into E. coli strain JM109 for functional expression. The two constructs were further subcloned into vectors pBbB8k-RFP and pBbE2k-RFP (Addgene) [34] under the control of tetracycline and pBAD promoters, respectively. This was performed by using In-Fusion cloning between PCR linearised vector (RFP eliminated) and an L6H-CPR insert. Following each PCR reaction, the template was removed by DpnI digestion, and PCR product size was determined by 0.6 %agarose gel electrophoresis. The oligonucleotide sequences encoding all the PCR primers can be found in Ta ble S1. The correct assembly of each construct was confirmed by DNA sequencing. Each construct was cotransformed with plasmid pJBEI6410 into competent cells of E. coli strain NEB10b for functional overexpression according to the manufacturers' protocols.
Generation of the L6H-CPR-ADH constructs: Eight constructs were generated in which each ADH was inserted downstream of the CPR gene of L6H m -SmCPR in pBbB8k preceded by one of two different ribosome binding sequences (rbs1-2). This was performed by using In-Fusion cloning between PCR linearised L6H m -SmCPR and amplified rbs-ADH insert (L6H m -SmCPR-IPDH, L6H m -SmCPR-CDH, L6H m -SmCPR-LkADH and L6H m -SmCPR-RRADH versions 1a nd 2, respectively). Following each PCR, template removal and DNA clean up were performed as above. The oligonucleotide sequences encoding the PCR primers can be found in Ta ble S2. The correct assembly of each construct was confirmed by DNA sequencing. Each construct was cotransformed with plasmid pJBEI6410 into competent cells of E. coli strain NEB10b for functional overexpression according to the manufacturers' protocols.
Construction of multienzyme-cascade constructs containing PETNR and CHMO 3M : As eries of multigene constructs containing L6H m ,I PDH, PETNR and CHMO WT or CHMO 3M were generated to maximise the production of (+ +)-DHCD lactone from (1S,5R)-carveol. The optimisation parameters varied were the plasmid backbone (pBbE1c or pBbE5c), RBS sequences and the presence of four different promoters upstream of CHMO 3M .F ull details of the assembly techniques and biotransformation data performed for each construct can be found in the Supporting Information.
Construction of the complete lactone-producing pathway from limonene: (+ +)-DHCD-producing constructs from (S)-limonene (Figure S5) were generated by In-Fusion cloning between the PCR linearised L6H M -SmCPR-CDH construct (contains rbs1)i np BbB8k and one of three PETNR-promoter-CHMO 3M inserts amplified from L6HIP-tet-C 3M ,L 6HIP-rha-C 3M and L6HIP-trc-C 3M .T hese inserts differ by the type of promoter located upstream of the CHMO 3M gene, that is tetracycline-, rhamnose-or IPTG-inducible, respectively.P CR linearisation of L6H M -SmCPR-CDH was performed between the 3'-end of CDH and the terminator region,; amplification of the PETNR-promoter-CHMO 3M inserts included rbs2 upstream of PETNR. Following each PCR reaction, template removal and DNA clean up were performed as above. The oligonucleotide sequences encoding the PCR primers can be found in Ta ble S8. The correct assembly of each construct was confirmed by DNA sequencing (L6HCCP-tet-C 3M ,L 6HCCP-rha-C 3M and L6HCCP-trc-C 3M ). Each construct was cotransformed with plasmid pJBEI6410 into competent cells of E. coli strain NEB10b for functional overexpression according to the manufacturers' protocols. As ummary of all the gene constructs is found in Ta ble S9.
Analytical techniques: Monoterpenoid content was quantified by using an Agilent Te chnologies 7890A GC system with af lame ionization detector (FID). Biotransformation extracts (1 mL) were analysed on aD B-WAX column (30 m; 0.32 mm;0 .25 mmf ilm thickness;J WS cientific). In this method, the injector temperature was 220 8Cw ith as plit ratio of 20:1. The carrier gas was helium with a flow rate of 1mLmin À1 and ap ressure of 5.1 psi. The program began at 40 8Cw ith ah old for 2min, then the temperature was increased to 210 8Ca tar ate of 15 8Cmin À1 ,w ith af inal hold at 210 8Cf or 3min. The FID was maintained at at emperature of 250 8Cw ith af low of hydrogen at 30 mL min À1 .P roduct was quantitated by comparing the peak areas to those of authenticated standards of known concentration. Where authentic standards were not commercially available (by-products only), the concentrations were estimated by using an average concentration per peak area value based on 11 related monoterpenoid standards.
Monoterpenoids were identified on an Agilent Te chnologies 7890B GC system with a5 977A MSD extractor EI source detector by using the same DB-WAX column. In this method, the injector temperature was 240 8Cw ith as plit ratio of 50:1. The carrier gas was helium with af low rate of 3mLmin À1 and ap ressure of 8.3 psi. The program began at 50 8Cw ith ah old for 1min, then the temperature was increased to 68 8Ca tarate of 5 8Cmin À1 ,w ith ah old at 68 8Cf or 2min. As econd temperature gradient was applied at 25 8Cmin À1 until 230 8Cw ith afinal hold of 2min. The mass spectra fragmentation patterns were entered into the NIST/EPA/NIH 11 mass spectral library to identify any potential match.
Upscaled in vitro biotransformations and analysis: Reactions (10 mL) were performed in buffer (50 mm Tris, pH 7.0) containing (R)-carvone/(+ +)-dihydrocarvone starting substrate (5 mm; % 30 mg), NADP + (10 mm), glucose (15 mm), GDH (10 U) and the enzyme(s) (10 mm). The samples were incubated for 24 ha t3 0 8Ca nd 180 rpm, then cooled in ice, and the organic compound(s) were extracted with petroleum ether (PET;1 :2, v/v). Twof urther PET extractions were performed, and the pooled organic phase was dried over anhydrous MgSO 4 .T he product(s) were recovered following solvent removal with ar otor evaporator with the water bath set to 30 8C, at 20-30 To rr.P roduct(s) was/were purified chromatographically on silica gel (pore size 60, 220-240 mesh size, particle size 35-75 mm), which was equilibrated with 100 %P ET.T he com- pounds were eluted with am ix of PET and ether (5-40 %), and each elution fraction was analysed by thin-layer chromatography (TLC) using am obile phase composed of aP ET/ether (70:30). The TLC plate was stained with phosphomolybdic acid stain (PMA; 12 gi n2 50 mL ethanol) and exposed to UV light. The fractions containing the desirable metabolites were pooled, and the solvent was removed as before. 1 Ha nd 13 CNMR spectra of the scaled-up purified product(s) (10 mg mL À1 )i nd euterated chloroform were recorded on aB ruker Avance4 00 MHz NMR spectrometer at 298 Kw ithout the addition of an internal standard. Chemical shifts were calibrated against the residual solvent signal. 1 Ha nd 13 CNMR spectra were analysed by using MestreNova.