Efficient Biosynthesis of R-(-)-linalool through Adjusting Expression Strategy and Increasing GPP Supply in Escherichia coli

Background: R-(-)-linalool is a versatile acyclic monoterpene alcohol with applications in the flavor and fragrance, pharmaceutical, and agrochemical industries. However, plant extraction furnishes only limited and unstable R-(-)-linalool yields that do not satisfy market demand. Therefore, a sustainable yet efficient and productive method of R-(-)-linalool synthesis is urgently needed. Results: To induce the R-(-)-linalool biosynthesis pathway in E. coli , we expressed heterologous (3R)-linalool synthase (LIS) from Lavandula angustifolia (laLIS). We then enhanced R-(-)-linalool production in the cells using a suitable LIS from Streptomyces clavuligerus (bLIS). The bLIS expression was markedly elevated by using optimized ribosomal binding sites (RBSs) and protein fusion tags. R-(-)-linalool output rose from 4.8 mg L -1 to 33.4 mg L -1 . To increase the geranyl diphosphate (GPP) content in E. coli , we tested various alterations in geranyl diphosphate synthases (GPPSs) and their mutants. The final E. coli strain harboring GPPS from Abies grandis ( Ag GPPS) accumulated ≤ 100.1 mg L -1 R-(-)-linalool after 72 h shake-flask fermentation. This yield gain constitutes a 60.7-fold improvement in R-(-)-linalool biosynthesis over the parent strain. Fed-batch cultivation of the engineered strain in a 1.3-L fermenter yielded 1,027.3 mg L -1 R-(-)-linalool. Conclusions: In this study, an efficient R-(-)-linalool production pathway was induced in E. coli via the heterologous MVA pathway, AgGPPS, and (3R)-linalool synthase (bLIS). By overexpressing the key enzyme in the engineered E. coli strain, R-(-)-linalool production reached 100.1 mg L -1 and 1,027.3 in the , , pETDuet- NusA*bLIS , pETDuet- CmR29*bLIS , and pETDuet- GST*bLIS . The optimized RBS sequence was ligated into the 5’-end of fused CmR29*bLIS and generated pETDuet- CmR29*bLIS (maximum). The gene encoding GPP synthase from Abies grandis (Genbank No. AAN01134.1) was truncated by 86 aa from its N -terminus, codon-optimized for E. coli , and inserted behind the second tac promoter of pETDuet- CmR29*bLIS (max). The plasmid product was pETDuet- CmR29*bLIS (max)- AgGPPS . The plasmids pETDuet-CmR29*bLIS (max)- IspA and pETDuet- CmR29*bLIS (max)- Erg20 were constructed based on pETDuet-Cm29R*bLIS (max)- AgGPPS by replacing AgGPPS with IspA from E. coli and Erg20 from Saccharomyces cerevisiae . The mutants of IspA (S80F), Erg20 (F96W), Erg20 (N127W), and Erg20 (F96W/N127W) were obtained by inverse PCR and construction of pETDuet- CmR29*bLIS (max)- IspA Three independent biological / experimental replicates were prepared. The minimum inhibitory concentration of R-(-)-linalool was defined as that which reduced colony formation by ≤ 2× on inhibitory LB plates.

An interesting alternative is the metabolic engineering of microorganisms to produce R-(-)-linalool. This compound is biosynthesized by (3R)-linalool synthase (LIS) from the universal monoterpene precursor geranyl diphosphate (GPP). Two different metabolic pathways have evolved to generate GPP. One is the mevalonate (MVA) pathway in eukaryotes and the archaea. The second is the deoxyxylulose 5-phosphate (DXP) pathway in most prokaryotes and plants ( Fig. 1) [9]. Previous studies showed that overexpression of certain native DXP pathway enzymes in E. coli combined with the complete exogenous MVA pathway may efficiently increase the levels of isopentenyl diphosphate (IPP) and its isomer dimethylallyl diphosphate (DMAPP). These molecules are building blocks for GPP [10,11]. In practice, however, the integration of these strategies limits monoterpene production [12].
To date, the yield of biosynthetic R-(-)-linalool has remained very low. We hypothesized that insufficient precursor GPP, low catalytic activity, and inadequate heterologous LIS expression explain these low R-(-)-linalool production rates.
In the present study, we established two-step R-(-)-linalool biosynthesis in recombinant E. coli and improved yield by metabolically engineering LIS and geranyl diphosphate synthase (GPPS). The LIS portion was optimized by screening LISs from various sources and upregulating them. Expression of the LIS portion was modulated by changing the ribosomal binding sites (RBSs) and adding fusion tags to tune R-(-)-linalool production. The GPPS portion was optimized by comparing various GPPSs and their mutants. Fed-batch fermentation with an efficient recombinant E. coli strain also improved R-(-)-4 linalool production. The successful R-(-)-linalool production by E. coli demonstrated here may lay theoretical and empirical foundations for engineering terpenoid pathways and optimizing other metabolic pathways.

Evaluation of the performance of different LISs on R-(-)-linalool production
In a previous study, the oleaginous yeast Yarrowia lipolytica was constructed. In shake flask cultures, it yielded a maximum linalool titer of 7.0 ± 0.3 mg L -1 [13]. However, linalool biosynthesis has only had limited success. Its titers have been lower than those for other terpenoids such as geraniol, βfarnesene, and amorphadiene [14][15][16][17][18][19]. We propose that the factors contributing to this deficiency include insufficient precursor supply, linalool toxicity, and inefficient LIS. To screen for effective LIS, we compared the performances of those extracted from Lavandula angustifolia (laLIS), Osmanthus fragrans var. thunbergii (ofLIS), and Streptomyces clavuligerus (bLIS) [20]. Truncated laLIS exhibits activity towards its natural substrate GPP (K m = 42.7 ± 4.6 and K cat = 3.9 × 10 -2 ). Therefore, laLIS may have relatively superior catalytic properties [21]. ofLIS is derived from ornamental and medicinal plants and has a strong fragrance. The only known bacterial terpene synthase is bLIS. It produces acyclic compounds and in purified form it uses GPP as a substrate to form R-(-)-linalool (K m = 12.9 ± 1.3 and K cat = 8.4 × 10 -2 ). Here, the aforementioned genes were codon-optimized and synthesized to avoid any impediments to protein expression. We examined the R-(-)-linalool production profile in E.
coli strains overexpressing the three optimized LISs. For 72 h, all strains were cultured in 250-mL shake flasks each containing 50 mL TB medium + 5 g L -1 glycerol. As shown in Fig. 2a, bLIS expression generated 4.8 mg L -1 R-(-)-linalool, which was a higher concentration than those achieved with laLIS (1.7 mg L -1 ) or ofLIS (ND). The strain containing bLIS produced certain amounts of nerolidol, as bLIS is both a monoterpene and sesquiterpene synthase. It converts farnesyl diphosphate (FPP) produced by native host-encoded enzymes [22]. For all strains, trace amounts of geraniol were detected in the organic culture overlays but diminished with incubation time. Previous studies showed 5 that E. coli bears an unidentified endogenous pathway that converts GPP into geraniol and then dehydrogenates or isomerizes it into other geranoids. Alkaline phosphatase (PhoA) may convert GPP to geraniol [23,24]. Thus, the biosynthetic pathway for R-(-)-linalool production was elucidated via exogenous LIS expression. The catalytic activity of bLIS was higher than those of laLIS and ofLIS.
Hence, the bLIS enzyme was selected to enhance R-(-)-linalool production in the subsequent experiments.
R-(-)-linalool formation was accompanied by the abundant generation of indole, which is an intercellular signaling molecule in E. coli (Additional file 1: Fig. S1). We speculate that certain overlooked effects such as an imbalanced metabolic pathway, the accumulation of toxic intermediate metabolites, or certain environmental factors may have stimulated E. coli to secrete indole to coordinate their behavior in order to adapt and survive in environmental niches [25][26][27]. Mevalonate kinase (MK) is a key enzyme that connects the upstream and downstream portions of the MVA pathway. It was regarded as a potential bottleneck [28]. Addition of exogenous MK lowered the indole titer by 50% (Fig. 2b) and reduced R-(-)-linalool production by 17% (Fig. 2c). Thus, indole formation was associated with the toxic intermediate metabolite HMG-CoA. Moreover, the carbon flux from the upstream to downstream portions of the MVA pathway were not limited by MK overexpression.
However, the R-(-)-linalool yield did not increase. Therefore, certain bottlenecks remained in the R-(-)linalool synthesis pathway. As bLIS catalyzed the final step in R-(-)-linalool synthesis, sufficient bLIS expression might be required to obtain higher R-(-)-linalool yields.

Modulation of bLIS expression by modification of RBS sequences and application of fusion tags
Here, bLIS expression was independently regulated at the translational level. This process has been widely used in whole-cell biocatalysis as it is convenient and highly efficient. Bacterial translation processing comprises initiation, elongation, termination, and ribosome turnover. Of these, initiation is the rate-limiting step [29]. Translation initiation can be modulated by modifying RBS sequences [30].
Computational algorithms conveniently design RBS for specific translation initiation rates (TIRs) [29,31]. In this study, we used the online program "Salis Lab's RBS Calculator" 6 (https://salis.psu.edu/software/) to redesign the bLIS RBS. To control bLIS expression, we designed a set of RBS sequences with TIRs of 100, 200, 500, 5,000, 50,000, and a maximum of 176,000 arbitrary units (au) (Additional file 1: Table S2). The RBS used in pETDuet-bLIS (original plasmid without RBS sequence modification) was predicted to have TIR = 21,711 au. The bLIS had an expected molecular weight (MW) of 37 kDa. However, the SDS-PAGE analysis disclosed no evidence of target protein overaccumulation in any of the transformants (data not shown). We hypothesized that this outcome was the result of low heterologous protein expression in E. coli. Moreover, there was no indication that bLIS expression increased with TIR strength. To elucidate the effect of bLIS expression on R-(-)-linalool production, six different RBS strains were made by transformation with pAGES and pHGFH (Additional file 1: Table S1). The R-(-)-linalool titer increased with TIR strength (Fig. 3a). The RBS with TIR = 50 k au and 176 k au for bLIS contributed to higher R-(-)-linalool production levels than the unmodified RBS. WX6600 had the maximum bLIS TIR strength and produced the greatest R-(-)-linalool titer (13.5 mg L -1 ) after 72 h cultivation. This rate was 2.8-fold higher than that determined for the original WX6000. Thus, RBS-based modulation of protein expression level is feasible. Low bLIS TIR strengths (100, 200, and 500 au) significantly decreased R-(-)-linalool titers and corresponded with poor cell growth (Fig. 3b). The latter, along with low R-(-)-linalool production levels, may have been caused by the abnormal accumulation of toxic IPP and DMAPP when there was insufficient bLIS protein [9].
R-(-)-linalool yield could be improved by modifying the bLIS RBS sequence. However, the R-(-)-linalool titer was lower than expected and there was a theoretical maximum TIR (Additional file 1: Fig. S2). As R-(-)-linalool is a secondary metabolite, its actual quantity is very low relative to overall plant biomass. The selection of bLIS from bacteria is conducive to protein solubility and expression in prokaryotic E. coli. Nevertheless, heterologous bLIS expression in E. coli was still very low (Fig. 3c).
Therefore, it does not support high R-(-)-linalool production rates and yields. To overcome the bLIS expression barrier, we used fusion tags in the attempt to enhance solubility. Maltose-binding protein (MBP), N-utilization substrate A (NusA), thioredoxin (Trx), and glutathione-S-transferase (GST) are efficient solubility enhancers in E. coli [32,33]. Moreover, a peptide ≤ 29 aa in length derived from the N-terminal of chloramphenicol acetyltransferase (CmR29) is an effective solubility enhancer in Synechocystis. Thus, we constructed the fused bLIS genes MBP*bLIS, NusA*bLIS, GST*bLIS, and coli. These tools, then, could be applied towards the regulation of heterologous protein production pathways to improve the yield of target compounds.

Improvement of R-(-)-linalool production by enhancing GPP accessibility
For strain WX6600 with a modified RBS sequence and strain WX6030 with the fusion protein CmR29*bLIS, the R-(-)-linalool titers were 13.5 mg L -1 and 16.4 mg L -1 , respectively. Nevertheless, strain WX6630 with both the modified RBS sequence and the fusion protein CmR29*bLIS produced only 33.4 mg L -1 R-(-)-linalool. This rate was, in fact, lower than the theoretical 49.4 mg L -1 yield gain.
A low GPP level may have accounted for this unsatisfactory output. The DXP and heterogeneous MVA pathways had adequate IPP and DMAPP. However, the GPP pool was very low and a substantial amount of nerolidol by-product was generated. These results were obtained because of the absence of specific, efficient GPPS in the recombinant strains. In E. coli, GPP is produced by the enzyme IspA, 8 which also synthesizes FPP. Therefore, IspA both supplies and competes for the bLIS substrate GPP.
To ensure adequate GPP, we selected truncated Abies grandis GPPS (AgGPPS), E. coli IspA and its mutant IspA (S80F), Saccharomyces cerevisiae Erg20, and the Erg20 mutants F96W, N127W, and F96W/N127W and used them to generate R-(-)-linalool [10,11,[34][35][36]. Upon co-expression of bLIS, GPPSs, and their mutants, R-(-)-linalool production was altered to varying degrees (Fig. 4a). The coexpression of bLIS and AgGPPS provided the highest R-(-)-linalool yield (100.1 mg L -1 ). Thus, the major limiting step in the supply of GPP is its conversion from DMAPP/IPP. Erg20 and its mutants have been widely used for terpenoid production by S. cerevisiae. However, few studies have reported on Erg20 application for terpenoid synthesis in prokaryotic E. coli. Wild type Erg20 had very low catalytic activity for GPP and FPP generation in E. coli. By changing Phe96 to Trp in Erg20 (F96W), FPP synthase activity was significantly enhanced but the GPP content did not significantly change.
Replacing Asn127 with Trp (N127W) increased both GPP and FPP and generated nearly equal amounts of R-(-)-linalool and nerolidol. To obtain relatively higher titers of R-(-)-linalool, the strain WX6636 was induced to express double-mutant Erg20 (F96W/N127W). Its R-(-)-linalool and nerolidol yields were 47.0 mg L -1 and 22.4 mg L -1 , respectively. The three Erg20 variants showed comparatively higher catalytic activity in E. coli. Nevertheless, the mutant Erg20 (F96W) had the highest FPP synthase activity while the double mutant Erg20 (F96W/N127W) was relatively more conducive to GPP generation (Fig. 4b). IspA (S80F) and AgGPPS had similar performances as precursor suppliers for pinene production in E. coli [37]. In the present study, however, their relative efficacies for R-(-)linalool production were markedly different. Analysis of the fermentation products revealed that 32.3 mg L -1 and 10.2 mg L -1 nerolidol were obtained from strains WX6632 and WX6637, respectively, after 72 h shake flask culture (Fig. 4c). Strains WX6632 and WX6637 accumulated total C 5 unit concentrations of 0.75 mM and 0.82 mM, respectively, after 72 h shake flask culture. AgGPPS can convert IPP/DMAPP into GPP without generating any FPP by-product. IspA (S80F) can co-synthesize FPP with GPP and bLIS can convert the former into nerolidol. For these reasons, the precursor suppliers IspA (S80F) and AgGPPS achieved different results in pinene and R-(-)-linalool production.
Here, we increased R-(-)-linalool production to 100.1 mg L -1 and generated 10.2 mg L -1 nerolidol by using strain WX6637 as its GPP access was relatively greater. Some of the GPP was converted to FPP by the native IspA in E. coli. Further study is required to determine how to eliminate this competing native FPP biosynthesis pathway or enhance bLIS substrate selectivity.

R-(-)-linalool production from an engineered strain in a 1.3-L bioreactor
Fed-batch fermentation was conducted in a 1.3-L bioreactor to evaluate the overall performance of strain WX6637 which accumulated the highest R-(-)-linalool levels. As illustrated in Fig. 5a,  It was proposed that the bacterial toxicity of organic solvents is inversely correlated with their log P ow . Organic solvents with log P ow = 2-5 are considered toxic to most bacteria. The log P ow for R-(-)linalool is 2.97 [38]. A minimum inhibitory concentration assay was run on R-(-)-linalool to quantify its antimicrobial property. E. coli CIBTS1758 was grown in LB medium with 100 mg L -1 , 200 mg L -1 , 500 mg L -1 , 1,000 mg L -1 , or 2,000 mg L -1 exogenous R-(-)-linalool. As shown in Additional file 1: Fig. S3, the culture reached OD 600 = 3.1 after 12 h incubation in the absence of R-(-)-linalool. R-(-)-linalool concentrations < 1,000 mg L -1 did not significantly inhibit CIBTS1758 growth. However, at R-(-)linalool concentrations ≥ 1,000 mg L -1 , CIBTS1758 growth was drastically inhibited. After 12 h, only scant E. coli were detected in the culture. The number of colonies on the plates significantly decreased when the R-(-)-linalool titer increased to 1.3 g L -1 (Fig. 5b). At ~1.4 g L -1 , R-(-)-linalool had reduced the number of surviving bacterial colonies by 50% relative to the untreated (Fig. 5c). Twophase cultures may partially alleviate R-(-)-linalool inhibition. However, substantial improvements in the tolerance of E. coli to organic solvents and microbial terpenoid production will require considerable metabolic and protein engineering and fermentation optimization.

Conclusions
Here, R-(-)-linalool production via recombinant E. coli strains was optimized by systematic and rational approaches (Fig. 6)

Chemicals and materials
Restriction enzymes, T4 DNA ligase, T4 polynucleotide kinase, PrimeSTAR Max DNA polymerase, and PCR reagents were purchased from Takara (Dalian, China). The primers were synthesized by Sangon (Shanghai, China). The corresponding codon-optimized genes were chemically synthesized by Generay (Shanghai, China). The gel extraction, PCR purification, and plasmid mini kits were purchased from Axygen Scientific (Suzhou, China). All other chemicals used were of the highest available purity.

Product analysis by GC-MS and GC-FID
One milliliter of uniformly mixed culture broth with the isopropyl myristate phase was sampled and  Fig. S4).
coli BL21 to yield the recombinant strains WX6009, WX6019, WX6029, WX6039, and WX6049, respectively. Heterologous expression of the fused bLISs in E. coli BL21 was compared by 12% SDS-PAGE analysis of the crude protein extracts.

Fed-batch fermentation in a 1.3-L bioreactor
The process used for R-(-)-linalool production was adapted from a previously published method [18]. A single colony of strain WX6637 was incubated overnight in 5 mL LB medium at 37 °C and rotated at 200 rpm. Then 40 mL LB medium (10% v/v) was inoculated with this pre-culture.
Two hundred-microliter samples were transferred to 96-well plates and incubated with continuous shaking at 37 °C in a Biotek Cytation 3 imaging plate reader (Biotek Instruments Inc., Winooski, VT, USA). OD 600 was automatically measured every 0.5 h. To determine the minimum inhibitory concentrations of R-(-)-linalool, equal amounts of cells at OD 600 = 1 were subjected to serial 10-fold dilutions with fresh LB medium. Then 100 μL of a 10 -5 dilution was spread onto inhibitory LB plates containing 0, 1, 1.1, 1.2, 1.3, 1.4, or 1.5 g L -1 R-(-)-linalool and incubated for 12 h at 37 °C. The viable cells on each plate were then counted. Three independent biological / experimental replicates were prepared. The minimum inhibitory concentration of R-(-)-linalool was defined as that which reduced colony formation by ≤ 2× on inhibitory LB plates.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.