Reagents used in this work were commercially sourced and were of analytical grade or better. Industrially-sourced waste paper fines (JK Paper Ltd, India) were kindly supplied by Dr Binod Parameswaran (CSIR NIIST, India) (Suvachan et al., 2021). Culture media pre-blended preparations were obtained from Formedium. The standard E. coli DH5α culture medium used for cloning was Luria-Bertani (LB), which is composed of 10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl. For linalool production assays the culture medium was Terrific Broth (TB), containing 24 g/L yeast extract, 12 g/L tryptone, 9.4 g/L KH₂PO₄ and 2.2 g/L K₂HPO₄. When antibiotic selection was required, the culture medium was supplemented with carbenicillin (50 μg/ml), hygromycin (150 μg/ml), kanamycin (50 μg/ml for plasmids or 30 μg/ml for integrated constructs) or chloramphenicol (34 μg/ml for plasmids or 10 μg/ml for integrated constructs). Agar plates were composed of LB containing 1.5% (w/v) agarose and the required antibiotic (Table 1). Growth of E. coli DH5α on mineral based medium with cellulose-based carbon sources was performed using a thiamine supplemented medium described previously (Li and Sha, 2017). For E. coli BL21 strain, growth on paper as a carbon source was performed using M9 minimal salts with modified trace element supplementation (3 g/L KH₂PO₄, 0.5 g/L NaCl, 6.78 g/L Na₂HPO₄, 1 g/L H₄Cl, 1 mM MgSO₄, 0.3 mM CaCl₂, 1 mg/L thiamine, 1 mg/L biotin, 17 µM FeCl₃, 84 mg/L EDTA, 150 mg/L MnCl₂.4H₂O, 105 mg/L ZnSO₄.7H₂O, 30 mg/L H₃BO₃, 25 mg/L Na₂MoO₄.2H₂O, 15 mg/L CuCl₂.2H₂O) (Blank et al., 2008).

E. coli genomic DNA was extracted and purified using the DNeasy Blood & Tissue Kits (Qiagen). CloneAmp HiFi PCR Premix (Takara Bio) was used for the construction of plasmids, while diagnostic PCR and sample preparation for sequencing was performed using GoTaq Green Master Mix (Promega). Each PCR product was analyzed by agarose gel electrophoresis, followed by purification with NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel). Plasmid purification was performed using the ISOLATE II Plasmid Mini Kit (Bioline). DNA concentration was measured using a NanoDrop 2000 (Cole-Palmer). The Q5 Site-Directed Mutagenesis Kit was obtained from New England BioLabs. Restriction enzymes for plasmid construction were purchased from New England BioLabs. All other plasmid recircularization and ligations were performed using the In-Fusion cloning kit (Takara Bio).

Gene sequencing and oligonucleotide synthesis were performed by Eurofins MWG (Ebersberg, Germany). To facilitate plasmid construction, DNA primers were designed to contain either 5’ and 3’ restriction endonuclease sites or 15–30 bp overlaps homologous to the target construct for ligation by In-Fusion cloning. The list of primers used for PCR reactions are shown in Supplementary Table S1. Details of the biocatalytic plasmid and chromosomal constructs used in this study are described in Table 1. Plasmids acting as templates for the assembly of linalool production pathway constructs and to facilitate genomic integration of DNA are described in Supplementary Table S2.

After construct assembly, each plasmid was transformed into E. coli strain DH5α and incubated overnight on antibiotic-selective LB agar plates. Individual colonies were picked, and small liquid cultures (5 ml) were cultivated overnight at 37°C using antibiotic-selective LB medium (Table 1). Following plasmid recovery and purification, each plasmid was sequenced to confirm the correct construct had been assembled. At each stage of cloning the protocols specified by the individual kit manufacturers were followed.

Plasmid pMVA is a derivative of the limonene-producing construct pJBEI6410 (Alonso-Gutierrez et al., 2013) lacking the genes geranyl pyrophosphate synthase (GPPS) and limonene synthase (Leferink et al., 2016). It contains operons for the upper and lower MVA pathway (MevT and MevB, respectively), controlled by the lacUV5 and trc promoters, respectively (Figure 1). Construct pGLinS contains the Abies grandis geranyl pyrophosphate synthase (AgGPPS) and Streptomyces clavuligerus linalool synthase (LinS) controlled by a tet promoter (Karuppiah et al., 2017). Plasmid pMVA-GLinS is a combination of pMVA and pGLinS (Figure 1), with the latter operon controlled by a trc promoter (Ferraz et al., 2021).

Elimination of the repeated regions R1-R3 (see below) in pMVA-GLinS was performed in two stages. The initial phase involved amplifying the whole plasmid by PCR, which selectively deleted region R2 and the majority of region R1 (Figure 2A). This was followed by plasmid recirculation and ligation by In-Fusion cloning to generate pMVA-GLinS NR1. A second round of whole plasmid PCR amplification was performed that eliminated R2 and the residual 3’ end of R1. In addition, the MevB pathway trc promoter was substituted for the constitutive promoter P _J23116 . This PCR product was ligated by In-Fusion cloning to generate pMVA-GLinS NR2.

A variety of modifications of the base pMVA-GLinS NR2 plasmid were performed to increase the expression and/or functioning of the terminal pathway genes gpps and/or linS. A modified version of pMVA-GLinS NR2 contains a N-terminal 29 amino acid solubility tag from the chloramphenicol resistance marker on LinS (pMVA-G-CmR29*LinS NR2). This plasmid was kindly supplied by Dr Robin Hoeven (University of Manchester, United Kingdom). Whole plasmid PCR amplification was performed on pMVA-GLinS NR2 to substitute wild-type LinS with either a single (L72M) or double (L72M/V214I) variant from pET-bLinS_L72M or pET-bLinS_L72M/V214I plasmids, respectively (Ferraz et al., 2021). New pMVA-GLinS NR2 constructs incorporating variant LinS genes were assembled by In-Fusion cloning (Table 1). The pMVA-GLinS NR2 plasmid also underwent modification of the LinS ribosomal binding site (aat---GGA​GCT​TTT​TAG​AAG​GAG​GTA​TAG), which also included the removal of the upstream stop codon to generate a GPPS-LinS fusion protein. Finally, pMVA-GLinS NR2 underwent the insertion of neryl pyrophosphate synthase (NPPS) from Solanum lycopersicum (GenBank: NM_001,247,704.1) with or without gpps gene deletion (Table 1). Further details of the latter two plasmid modification strategies can be found in the Supplementary Methods document.

Commercially available chemically competent E. coli strains DH5α or BL21 (New England BioLabs) were used for routine cloning of plasmids. The E. coli PIR2 strain (Invitrogen) was used for cloning and maintenance of pKIKO plasmids (Supplementary Table S1). This strain harbors the λpir gene that encodes the π protein to support the R6K replication origin in pKIKO plasmids (Sabri et al., 2013). DNA transfer into E. coli was performed by transformation of chemically competent cells (plasmid construction) or electroporation (construction of genome integrated strains), the latter of which is described in Supplementary Methods. Chromosomal integration pKIKO plasmids were obtained from AddGene (Sabri et al., 2013). These plasmids contain homology arms that allow for integration at the lacZ, arsB, and rbsAR genes in E. coli (Sabri et al., 2013). An additional set of plasmids (pKIKOgltACm and pKIKOgltAKm) were generated incorporating a fourth loci (gltA) for chromosomal insertion, as described in the Supplementary Methods ( Supplementary Figure S1). Other plasmids utilized for pKIKO-based chromosomal integration of DNA constructs were pSIM18 (Chan et al., 2007) and pCP20 (Cherepanov and Wackernagel, 1995) (Supplementary Table S1). pSIM18 is a heat-sensitive plasmid that harbors the λ-RED recombination gene (Datta et al., 2006). The recombinase flippase system for antibiotic resistance excision from the genome is encoded on the heat sensitive plasmid pCP20 (Cherepanov and Wackernagel, 1995; Datsenko and Wanner, 2000).

The methods used for chromosomal integration are based on the λ-RED recombination system (Thomason et al., 2014) using the pKIKO vector series of plasmids (Supplementary Table S1) (Sabri et al., 2013; Jervis et al., 2021) as described previously. Further details are provided in the Supplementary Methods. In this general protocol, DNA cassettes for integration were inserted upstream of the kanamycin or chloramphenicol resistance gene and between the specific chromosomal loci homology arms (HA1 and HA2). The antibiotic resistance gene was enclosed on each side by the flippase-mediated recombination sequence (FRT) (Sabri et al., 2013). The linear DNA cassettes used for each genomic integration (HA1-pathway-HA2-FRT-antibiotic gene-FRT) was generated by PCR amplification and electroporated into pSIM18 plasmid-containing E. coli DH5α. Confirmation of the presence of the integrated DNA cassette was performed by PCR, using primers specific to regions outside of the chromosome HA1 and HA2 sites, or with one primer specific for the inserted DNA. Following curing of the heat-sensitive pSIM18 plasmid, elimination of the integrated antibiotic resistance gene flanked by FRT sites was performed using the recombinase flippase system encoded on the heat sensitive plasmid pCP20 (Cherepanov and Wackernagel, 1995; Datsenko and Wanner, 2000). Final confirmation of successful gene cassette integration and antibiotic removal was performed by sequencing the genomic DNA in the region of interest.

To integrate the GLinS cassette into the arsB loci in E. coli DH5α, we constructed the plasmid pKIKOarsBCm:GLinS using PCR and restriction enzyme cloning (Table 1). The GPPS and LinS genes and regulatory elements were PCR amplified from pGLinS and digested with enzymes SacI and SalI. This DNA cassette was ligated to pKIKOarsBCm plasmid previously digested with the same restriction enzymes. Following integration into the chromosome of E. coli DH5α, the new strain (GL) was transformed with the pMVA plasmid to supply the geranyl pyrophosphate precursor for linalool production.

To incorporate the MVA pathway into E. coli DH5α or the GL strain, the pathway was divided into the upper MevT and lower MevB operons (Figure 1) (Alonso-Gutierrez et al., 2013) for integration at loci rbsAR and lacZ, respectively. The upper pathway integration plasmid pKIKOrbsARCm:MevT was assembled using the PCR and restriction enzyme cloning protocol described above. For the lower MevB pathway, the integration plasmid pKIKOlacZCm:MevB was assembled by PCR amplification of the three-gene operon with its regulatory elements, PCR linearization of pKIKOlacZCm and ligation by In-Fusion cloning. Following successful genomic integration, the MVA pathway integrated strain (M1) was transformed with pGLinS to allow linalool production to occur. The strain containing the MevT, MevB, and GLinS cassettes (MGL) did not require any plasmid addition.

The general protocol for in vivo linalool production in E. coli was performed using individual colonies from freshly transformed cultures (plasmid constructs only) or overnight liquid cultures as the starting inoculum. Cultures (3 ml) were incubated in TB medium containing 0.4% glucose in sealed glass bottles and incubated at 37°C and 200 rpm until growth became visible (up to 7 h). Recombinant protein expression was induced with 50 µM isopropyl β-D-1-thiogalactopyranoside (trc and lacUV5 promoters) and in some cases with 25 nM anhydrotetracycline (tet promoter; Table 1). Cultures were incubated for a further 68–72 h at 30°C. In some assays, 300–600 μL of nonane or isopropyl myristate (organic overlay) was added to the cultures after induction to sequester linalool from the culture medium. Linalool production from cellulose was performed as above (no overlay), except growth was performed in M9 medium containing 1.5% carbon source (glucose or cellulose) with DH5α or 0.5% carbon source (glucose or cellulose) with BL21 (DE3) and 1% Cellic cTec2 cellulase enzyme blend (Sigma-Aldrich).

Linalool titers were determined by extracting 1 ml of culture or organic overlay with an equal volume of ethyl acetate containing 0.01% (v/v) sec-butyl benzene (internal standard). The organic layer was dried with anhydrous MgSO₄ and analyzed for linalool content by GCMS. To assess the retention of the recombinant plasmids (where applicable), minipreps were performed on 1–2 ml of each culture. Each purified plasmid underwent a diagnostic restriction digest with EcoRI and XhoI, and the fragments were analyzed by electrophoresis using a 0.5% agarose gel. The predicted fragment DNA band sizes and plasmid maps for pMVA-GLinS and potential recombined constructs are shown in Supplementary Figure S2.

Linalool toxicity studies were performed by monitoring the optical density of cultures exposed to linalool. The starter culture of E. coli DH5α was incubated overnight in LB medium (5 ml) with agitation. Replicate aliquots of E. coli DH5α (200 μL; 2% inoculum) were set up in LB medium containing 0–2 g/L linalool in a 96-well microtiter plate (Costar). Cultures were covered with a Moisture Barrier Seal (4titude) and incubated at 30°C with 500 rpm agitation within a FLUOstar Omega Microplate reader (BMG Labtech). Optical density readings (OD_{600 nm}) were performed every 15 min for 24 h.

Adaptive laboratory evolution (ALE) of E. coli DH5α for increased linalool tolerance was performed by exposure of cultures to increasing concentrations of linalool. The initial starter culture was composed of E. coli DH5α in 5 ml LB containing 0.5 g/L linalool, which was cultivated overnight at 30°C and 200 rpm. This culture served as the inoculum for subsequent cultures of increasing linalool concentrations (2–20 g/L). Growth was monitored in a microtiter plate reader, as described above. This was performed iteratively until growth was observed up to 20 g/L linalool, with the new strains designated as E. coli ALE.

Paper fines from a waste stream of JK Paper Ltd paper mill were used as a low-cost cellulose source. The paper was sanitized by vortexing in 100% ethanol, followed by solvent evaporation in a sterile environment. To determine the effective glucose titers from paper fines, saccharification was performed using the Novozymes Cellic cTec2 cellulase enzyme blend (Sigma-Aldrich) using an adaptation of the NREL procedure (Resch et al., 2015). Paper fines (100 mg) were suspended in 10 ml of sterile 0.1 M sodium citrate buffer pH 4.8 containing 80 μg/ml tetracycline hydrochloride, 0.002% sodium azide, and 1:100 Novozymes Cellic cTec2 cellulase enzyme blend. Samples were incubated at 50°C with shaking at 250 rpm. A negative control (no paper) was performed to determine the background glucose content present in the Cellic cTec2 cellulase blend. Suspensions were clarified by centrifugation at 3,200 g for 10 min or 17,000 g for 2 min. The aqueous solution was analyzed for glucose content by HPLC.

Growth of E. coli strains DH5α and BL21 (DE3) on cellulose-based carbon sources was performed by cultivation in mineral-based medium containing 0.5% of either glucose (control), carboxymethylcellulose (CMC), Sigmacell cellulose or pretreated paper fines with a 1:1000 inoculum of an overnight culture in LB medium. Simultaneous saccharification and fermentation was performed where cellulose digestion to glucose occurred by the addition of the Cellic cTec2 cellulase blend (1:100) or expression of the Addgene ampicillin-resistant pCellulose multicellulase plasmid (Bokinsky et al., 2011). Recombinant cellulase production was controlled by the cspD and wrbA promoters that are induced in stationary phase (when E. coli is starved for carbon) (Bokinsky et al., 2011). To minimize the likelihood of microbial contamination in cultures containing cTec2, the E. coli strain contained the pBbS5a:RFP vector and each culture was cultivated for 24 h at 37°C in the presence of 50 μg/ml carbenicillin. Culture optical density measurements were not possible due to the opaqueness of some of the carbon sources. Therefore, after 24 h of growth, culture aliquots were diluted and spread on LB agar plates and incubated overnight at 37°C. Growth is expressed as the number of colony-forming units (CFUs).

Linalool production was analyzed by GCMS using an Agilent Technologies 7890B GC equipped with a 5977A MSD detector. Product(s) (1 μL) were separated on a DB-WAX column (30 m × 0.32 mm i.d., 0.25 µM film thickness, Agilent Technologies) using the running conditions described previously (Leferink et al., 2016). Compound identification was carried out using the reference spectra in the NIST library of MS spectra and fragmentation patterns. Linalool quantitation was performed by comparing peak areas to a standard curve generated from authentic standards run under the same conditions. Linalool concentrations obtained from organic overlay samples have been corrected to express the data as mg/L culture.

Glucose release from cellulose was analyzed by HPLC using an Agilent 1260 Infinity HPLC with a 1260 ALS autosampler, TCC SL column heater, a 1260 refractive index detector (RID). Samples were run on an Agilent Hi-Plex Ca column (300 × 7.7 mm) using 100% HPLC-grade water as the mobile phase (0.6 ml/min) at 85°C for 30 min. Glucose quantitation was performed by comparing peak areas to a standard curve generated from authentic standards (retention time of 13 min). For generalized glucose and organic acid detection from cellulose-containing samples, culture supernatant aliquots (20 μL) were injected into an Agilent Hi-Plex H column (7.7 × 300 mm, 8 μm) using 0.005 M H₂SO₄ as the mobile phase (0.7 ml/min). The run was performed at 60°C for 60 min with the RI detector set to 55°C. Retention times for each component were: 10 min for glucose, 14 min for formate, 16 min for acetate, 17 min for levulinic acid, 33 min for 5-hydroxymethylfurfural, and 49 min for furfural. Analyte concentrations were calculated by comparing the peak areas to a standard curve generated from analytical standards of known concentrations.

Bacterial linalool production in E. coli has been previously demonstrated using a one or two plasmid-based system incorporating the entire recombinant pathway from acetyl-CoA (Leferink et al., 2016; Karuppiah et al., 2017; Wang et al., 2019; Karuppiah et al., 2020; Wu et al., 2021) or via a non-canonical isopentenol utilization pathway for GPP precursor production (Ferraz et al., 2021). A screen of multiple E. coli strains showed that the highest linalool titers were obtained using the K-12 strain DH5α with the two-plasmid recombinant linalool production system (pMVA + pGLinS) (Leferink et al., 2016). This strain is also known to produce higher beta-carotene levels than E. coli BL21, MG1655, DH5α, S17-1 and XL1-Blue strains (Yoon et al., 2009). Therefore, we selected E. coli DH5α as our base strain for linalool production and used the one plasmid system (pMVA-GLinS; Table 1) as the template construct for our optimization studies (Figure 2A).

In our hands, we found that biological replicates (individual colonies) of E. coli DH5α containing either the one or two plasmid system generated linalool titers with a high degree of variability (122.7 ± 48.6 and 4.90 ± 4.5 mg/L, respectively; Figures 2B,C). This suggests the plasmid-based linalool production system is not stably maintained. An apparent correlation was observed where the linalool titer was inversely proportional to the size of the freshly transformed E. coli pMVA-GLinS colony used to inoculate the assay medium. This suggested linalool production from leaky promoters adversely affected culture growth, and small colony size was subsequently used as a criterion for selection of higher linalool producers.

Prior studies have suggested that instability in titers of terpenoids and other natural products may be due to plasmid recombination events, which in turn reduces in vivo production of toxic intermediates and products generated by full length constructs (Rinaldi et al., 2021). This may be especially relevant for pathways where either large or multiple plasmids need to be maintained. To check for the presence of plasmid recombination events in the linalool production plasmid, diagnostic restriction digestion and agarose gel electrophoresis of replicate cultures were therefore performed. Cultures displaying significant linalool production showed the “full-length” pMVA-GLinS plasmid restriction map (10.9 + 3.7 kb fragments), whereas cultures displaying low, or no titers had different banding patterns (Figure 2C and Supplementary Figure S2). This suggests that pMVA-GLinS is undergoing recombination, which decreases or eliminates the ability of E. coli to make linalool.

Studies have shown that repeated sequences as short as 8 bp can recombine both within and between plasmids (Sleight et al., 2010). As DH5α is a ΔrecA E. coli strain, any recombinant events must be RecA-independent. This can frequently occur in the presence of pairs of 25 bp homology regions, with recombination frequency increasing up to 411 bp (Lovett et al., 2002). A closer examination of pMVA-GLinS revealed the presence of four sets of identical repeating sequences longer than 15 bp, which were a consequence of maintaining multiple, identical regulatory elements for gene expression (Figure 2A and Supplementary Figure S2). Repeated region 1 (R1; 234 bp) contained the trc promoter upstream of the lower MVA pathway and GLinS. The second region (R2; 138 bp) contained the rrnBT1 and T7Te transcriptional terminators downstream of GLinS. Repeated region 3 (R3; 46 bp) consisted of the common sequence between the trc and lacUV5 promoters upstream of each part of the pathway (embedded in R1), thus appearing three times in the plasmid. The final repeating region (R4; 26 bp) corresponded to an RBS upstream of GPPS and LinS. Simulations of potential recombination events are shown in Supplementary Figure S2, which predicts the diagnostic banding patterns seen in some of the non-linalool producing colonies (Figure 2C).

To reduce the likelihood of plasmid recombination, we eliminated regions R1-R3 in two stages to generate pMVA-GLinS variants NR1 and NR2, respectively (Figure 2A). The final pMVA-GLinS NR2 plasmid contained a P _J23116 constitutive promoter in place of P _trc upstream of the lower MevB pathway. This new promoter was selected due to a prediction by SelProm (Jervis et al., 2019) that its subsequent expression levels would be similar to that of the original trc promoter. Comparative assays showed that both NR1 and NR2 constructs showed minor (1.7-fold) increases in linalool production compared to the original pMVA-GLinS in E. coli (162.8 ± 37.1 for NR2; Figure 2D). While some variability in titres still exists, there was a decrease in the number of non-linalool producing replicates. Other factors that could account for increased titers include the loss of some of the redundant terminator regions and/or the switch of the lower MevB pathway promoter.

It is interesting to note that these repeated regions are derived from the pJBEI-6410 or pJBEI-6409 MVA pathway backbone plasmid (Alonso-Gutierrez et al., 2013) due to the presence of multiple transcriptional control elements with sequence similarities. Variants of these plasmids have been used for the production of multiple terpenoids (Leferink et al., 2016), such as pinene (Bao et al., 2019), geraniol/geranyl acetate (Chacón et al., 2019; Wang et al., 2021), limonene (Rolf et al., 2020), perillyl alcohol (Alonso-Gutierrez et al., 2013) and other linalool production studies (Wu et al., 2021). Other natural product biosynthesis plasmids are known to contain significant repeating regions, such as pBbA5c-MevT (CO)-T1-MBIS(CO, ispA). This plasmid was designed to produce amorphadiene, a precursor of the antimalarial drug artemisinin (Redding-Johanson et al., 2011), and the sesquiterpene α-humulene (Alemdar et al., 2017). Terpenoid production has also been described using constructs based on the pACYCDuet backbone, which has a 51 bp repeated T7 promoter and lac operator (Meng et al., 2011; Liu et al., 2019). In spite of this, only one of these reports describes variable production titers (pinene) due to recombination of a pJBEI-6409-derived plasmid (Bao et al., 2019). In this case, pinene titers were improved by the elimination of repeated regions within the construct.

We hoped to build on the success of titer increase and plasmid stability improvements by implementing additional strategies to increase plasmid retention. One approach was to apply environmental stress by co-expressing with a second plasmid (pBbB2k-RFP) containing a compatible ori and a second antibiotic resistance gene (Gama et al., 2020) (Supplementary Figure S3). An alternative approach is to decrease the cytotoxicity of linalool by sequestering it into a microbial compatible organic co-solvent overlay (nonane or isopropyl myristate; Supplementary Figure S4). Monoterpenoid sequestering using this approach has been shown previously (Leferink et al., 2016). We also targeted key bottleneck enzymes GPPS and LinS for modification by a) incorporation of the more active LinS_L72M/V214I variant (Ferraz et al., 2021); b) increasing the solubility of LinS by incorporating an N-terminal 29 amino acid tag from the chloramphenicol acetyltransferase enzyme (Wang et al., 2019) c) generation of a GPPS-LinS fusion gene with a small linker region and d) substitution or inclusion of the N-terminally truncated neryl pyrophosphate synthase (NPPS) from tomato (Solanum lycopersicum) (Wu et al., 2019) (Supplementary Figure S5). Overall, these approaches showed little to no additional positive effects on linalool production titers and/or plasmid stability. This highlights that strategies shown to impart positive outcomes with one recombinant pathway will not necessarily translate into another. Further discussion on the environmental stress and GPPS/LinS modification approaches related to linalool titres can be found in the Supplementary Material document.

Linalool is known to be cytotoxic to microorganisms (Liu et al., 2020). We conjectured this may contribute towards linalool-producing plasmid instability and subsequent titer variability in E. coli. We investigated the cytotoxicity of linalool towards E. coli DH5α by monitoring its growth profile in the presence of externally supplied linalool. The results indicated that linalool significantly affects cell growth rate at levels < 1 g/L, with high concentrations (20 g/L) leading to a rapid loss of cell density after the addition of linalool (Supplementary Figure S6). Therefore, we performed Adaptive Laboratory Evolution (ALE) to increase the tolerance of E. coli DH5α towards linalool by performing continuous cultivation in increasing concentrations of linalool (Rinaldi et al., 2021).

The growth profile of a subset of iterative ALE strains in the presence of 0–20 g/L linalool was compared to see if linalool tolerance had improved. Strains ALE-2.5 to 20 showed significant increases in culture growth in the presence of linalool compared to the wild-type strain (Supplementary Figure S6 and Figure 3A). For example, ALE-20 in 5 g/L linalool achieved around 70% of the optical density after 24 h as the same strain without any linalool present. For wild-type DH5α, only ∼ 10% of the growth was achieved in 5 g/L linalool, with a gradual loss of optical density over time. In addition, significant growth was detected in the presence of 10 g/L linalool, a concentration that completely inhibited growth in the wild-type strain (Supplementary Figure S6).

Each of the ALE strains was transformed with the pMVA-GLinS NR2 plasmid to assess whether increasing the linalool tolerance leads to an increase in linalool production titers. Unfortunately, linalool titers of each of the ALE strains were found to be lower than the wild-type strain (Figure 3B). It is not clear why the more tolerant strains do not generate higher titers of linalool, as the mechanism(s) of increased linalool tolerance in the evolved strains has not been determined. A possible explanation is that ALE strains have adapted to decreased uptake of exogenously supplied linalool from the culture medium (lower permeability across the membrane), which could also translate to decreased secretion of intracellular linalool into the culture medium. This could in turn lead to higher accumulation of in vivo-generated linalool intracellularly, leading to product feedback inhibition, cytotoxicity or genetic instability of the plasmid. These findings illustrate how trialing an approach shown previously to successfully improve product titers does not necessarily translate into a positive outcome on a different system, even though the same problem (cytoxicity of the product) is present.

The chromosomal integration of recombinant pathways into microbial hosts is generally considered preferable to maintaining genes on a plasmid. This is because it eliminates the need for toxic and/or expensive selection agents (antibiotics) for plasmid maintenance (Friehs, 2004), which could prolong fermentation time and maximize product yields (Tyo et al., 2009; Amer et al., 2020a; Amer et al., 2020b). To this end, we performed partial and full integration of the more genetically stable linalool biosynthetic pathway into E. coli DH5α (Figure 4A). The simplest genome integrated system tested contained only integrated GPPS and LinS under control of a tetracycline inducible P _tet promoter (strain GL). Precursor geranyl pyrophosphate (GPP) was supplied by the presence of the MVA pathway on a plasmid. Almost no linalool was visible with the genome integrated GL strain compared to the plasmid-borne system (Figure 4A). A dramatic decrease in linalool titers with genome integrated strains is not surprising, given the genes are now present as a single copy in the genome instead of multiple copies on a plasmid (Amer et al., 2020a). However, such a dramatic titer reduction may have been in part due to the additional metabolic burden placed upon E. coli by the necessity to provide two antibiotics for linalool production (tetracycline for GLinS induction and kanamycin for pMVA maintenance). Therefore, additional constructs were incorporated that did not utilize a P _tet promoter.

Complete integration of the linalool production pathway was also explored. Due to the length of the MVA pathway, it was incorporated in two parts (MevT and MevB) into the high expression genome loci rbsAR and lacZ, respectively (strain M1) (Figure 4A). Incorporation of the pGLinS plasmid resulted in linalool titres of 4.51 ± 3.68 mg/L; (Figure 4B). Titres were reduced to 2.32 ± 0.84 mg/L when the full linalool pathway was incorporated within the genome (strain MGL; Figure 4A and Table 2). However, given the variability among biological replicates, these integrated or partially integrated strains overall showed similar linalool titers, with about a 100-fold reduction compared to using the pMVA-GLinS NR2 plasmid. This is not surprising given that the MVA pathway exists as a single copy on the chromosome compared to the high copy number plasmid-borne construct (Martin et al., 2003; Tyo et al., 2009). Similar terpenoid pathway titer reductions have also been reported by others in genome integration studies (Zhao et al., 2013; Alonso-Gutierrez et al., 2018; Wei et al., 2018).

To obtain higher titers of linalool it is preferable that multiple copies of key bottleneck genes be integrated stably within the genome. This approach was taken with lycopene production from chromosomally-integrated pathways, where titers matched and exceeded those of plasmid-borne sequences by increasing chromosomal copy number (Tyo et al., 2009). A rapid approach of achieving this is to use the CreloxP system for high-copy number integration of gene cassettes using growth under increasing antibiotic resistance as the stimulus for multi-copy insertion, as has been applied for the integration of the phaCAB genes at the asnB loci (50 copies) in Halomonas TD01 (Yin et al., 2015). This may prove to be a viable approach for significantly improving linalool titres in a genomic encoded system. In addition, other approaches to increasing linalool titers could include modifying regulatory elements such as promoters and ribosomal binding sites in native and genomic-integrated sequences. This proved successful in some cases to increase titers to match or even surpass those obtained with plasmids (Table 2) (Tyo et al., 2009; Lemuth et al., 2011; Chen et al., 2013; Ghanegaonkar et al., 2013; Li Y. et al., 2015; Ye et al., 2016; Wei et al., 2018). Given that single copy genomic integration of the entire linalool biosynthetic pathway led to a significant decrease in titers, we decided to focus subsequent optimization studies using the enhanced plasmid-based system.

We have demonstrated linalool production in E. coli DH5α via both plasmid-borne and genome integrated pathway incorporation. A key constraint in transitioning this “proof-in-principle” bacterial linalool production into a commercially viable venture is the reduction in costs associated with bacterial cultivation. Standard laboratory E. coli rich growth medium (TB) is prohibitively expensive on a large scale, and existing pilot scale bacterial cultivations typically rely on a mineral-based medium supplemented with glucose as a carbon source (Ruiz et al., 2013; Li J. et al., 2015). An abundant supply of glucose can be sourced via the enzymatic saccharification of waste lignocellulose biomass (de Souza, 2013; Beig et al., 2021). Paper mill wastewater contains secondary or rejected paper fiber, generated by the depuration of bleached pulp (Suvachan et al., 2021). This fiber has a high cellulose content, and enzymatic saccharification of this waste to glucose has been shown to be a suitable low cost carbon source for the growth of lysine-producing Corynebacterium glutamicum (Suvachan et al., 2021). Therefore, we investigated the potential of hydrolyzed paper as a feedstock for E. coli linalool production.

We tested the ability of a commercial cellulase blend to release glucose from Indian paper mill secondary fiber samples. This includes acid and base pre-treated fibers (Suvachan et al., 2021), designed to improve the access of saccharolytic enzymes to the cellulose (Elliston et al., 2015). Saccharification studies yielded maximal glucose titers [∼ 70% (w/w)] between 24 and 48 h (Supplementary Figure S7). These titers are consistent with compositional analysis of paper fibers showing a more than 50% cellulose content (Aguayo et al., 2018; Suvachan et al., 2021). The acid or base pre-treatment did not appear to have influenced the rate of glucose release or final glucose titers significantly. This is surprising given that previous studies showed acid pre-treatment removes the hemicellulose fraction in the secondary fiber, thereby making enzymatic saccharification more efficient (Suvachan et al., 2021). However, given that excess levels of cellulase blend were added to the saccharification reaction, any significant differences in the glucose release rate within the first 24 h would not be apparent.

An additional consideration for the use of secondary fiber as a bacterial carbon source is the potential for the presence of microbial growth inhibitors generated during the paper making process. However, analytical HPLC of aqueous secondary fiber extracts showed no evidence of significant levels of known growth inhibitors furfural, 5-hydroxymethylfurfural, acetate, formate or levulinic acid (Elliston et al., 2015). Therefore, glucose release from secondary fiber has potential to be a low-cost carbon source for E. coli growth.

Standard mineral-based growth medium with a single carbon source is not suitable for scaled cultivation of E. coli DH5α due to it being a thiamine auxotroph (thi ⁻) (Durfee et al., 2008) and the presence of a purB mutation (Jung et al., 2010). In contrast, E. coli BL21-based strains grow more efficiently in a mineral-based medium without the addition of thiamine. Therefore, we investigated the ability of both DH5α and BL21 strains to generate linalool in cellulose-based carbon sources expressing the pMVA-GLinS NR2 plasmid. We initially looked at growth of E. coli BL21 on cellulose-based carbon sources in the absence of pMVA-GLinS NR2 plasmid by indirectly monitoring colony forming units. This was due to the opaqueness of the paper-containing growth medium that prevented growth monitoring by optical density (see Supplementary Results for further details). The direct addition of glucose derived from saccharification of secondary fibers to culture medium was not performed due to the addition of biocides to prevent contamination during cellulolytic degradation. Instead, simultaneous saccharification and fermentation was performed whereby the carbon source was added direct to the minimal medium in addition to the cTec2 cellulase blend or via co-expression of a pCellulose plasmid (Bokinsky et al., 2011). Growth was higher when the cTec2 enzyme blend was added to the culture as the means of releasing glucose from the cellulolytic carbon sources (Supplementary Figure S8). We therefore used this latter approach to screen both E. coli strains for linalool production on a variety of cellulose-based substrates.

Overall, E. coli strain DH5α generated higher linalool titres compared to BL21 for most carbon sources, although the errors were higher with the former strain (Figure 5). This is consistent with previous observations that linalool titers tend to be the highest for the DH5α strain (Yoon et al., 2009; Han et al., 2016; Leferink et al., 2016). Minor linalool production was detected in control cultures without an added carbon source due to the presence of the commercial cellulase, whose preparations often contain sugar. The highest linalool titres for DH5α were seen with Sigmacell cellulose and glucose as the carbon sources (143 ± 111 and 75 ± 43 mg/L, respectively). In contrast, BL21 cultures generated similar titres of linalool with both glucose (32.03 ± 5.71 mg/L) or wastepaper (36.90 ± 3.79 mg/L). These titers are lower than those obtained in the presence of rich medium (83.3 ± 38.1 mg/L), but higher than for CMC and Sigmacell cellulose carbon sources. Additionally, terpenoid titers tend to be lower when E. coli is cultivated in minimal media compared to in rich media (Vickers and Sabri, 2015). This may be due in part to the observation that plasmid loss tended to be higher in minimal medium [data not shown; (Friehs, 2004)], possibly due to rich media relieving metabolic burden and alleviating stress (Singha et al., 2017).

Surprisingly, DH5α cultures showed poor linalool titres on wastepaper fibers compared to other cellulose-based substrates and the BL21 strain cultivated on paper (4.07 ± 4.08 mg/L; Figure 5). In addition, the variability of titres amongst biological replicates was higher for DH5α cultures compared to BL21 when cultivated on minimal medium. The low linalool titres of DH5α cultures grown on paper can be explained by the drastic reduction seen in plasmid levels compared to other carbon sources (Supplementary Figure S9). However, the relative stability of the pMVA-GLinS NR2 plasmid can be seen in non-paper cultures as a significant reduction in the appearance of apparent recombination events, although a few colonies appeared to have lost the plasmid during cultivation. This suggests other factors may be impacting on the high variability in linalool titres beyond plasmid effects, such as differences in the level of fitness of E. coli DH5α vs. the BL21 strain in minimal medium. In addition, the volatility of linalool results in losses into the headspace of the assay cultures, which can significantly impact on the recovery after a prolonged incubation.

Despite the complications discussed above, to our knowledge, this is the first demonstration of linalool production from lignocellulose waste, and is encouraging given similar titers are obtained with comparative studies with glucose as the sole carbon source. Previous studies have shown that careful optimization studies from lab scale to 1.3 L bioreactors can lead to significant increases of linalool titers from 100–270 mg/L to 1–1.5 g/L (Wang et al., 2019; Wu et al., 2021). Scaled fermentations under controlled process conditions could impact on the reproducibility of linalool titres by the careful control over carbon source feeding and the use of exhaust gas condensers to improve the retention of linalool in the liquid phase. Therefore, further studies are required beyond this “proof-of-principle” demonstration to improve the scaling potential of the process by optimizing the culture conditions to increase linalool titers and switching from adding commercially sourced cellulases to the efficient co-expression of secretable cellulolytic enzymes (Li et al., 2021) and cytoplasmic monoterpenoid production pathways.