A synthetic biology and green bioprocess approach to recreate agarwood sesquiterpenoid mixtures

Certain endangered Thymelaeaceous trees are major sources of the fragrant and highly valued resinous agarwood, comprised of hundreds of oxygenated sesquiterpenoids (STPs). Despite growing pressure on natural agarwood sources, the chemical complexity of STPs severely limits synthetic production. Here, we catalogued the chemical diversity in 58 agarwood samples by two-dimensional gas chromatography–mass spectrometry and partially recreated complex STP mixtures through synthetic biology. We improved STP yields in the unicellular alga Chlamydomonas reinhardtii by combinatorial engineering to biosynthesise nine macrocyclic STP backbones found in agarwood. A bioprocess following green-chemistry principles was developed that exploits ‘milking’ of STPs without cell lysis, solvent–solvent STP extraction, solvent–STP nanofiltration, and bulk STP oxy-functionalisation to obtain terpene mixtures like those of agarwood. This process occurs with total solvent recycling and enables continuous production. Our synthetic-biology approach offers a sustainable alternative to harvesting agarwood trees to obtain mixtures of complex, fragrant, oxygenated STPs.

Methods
Agarwood sample collection and processing
Agarwood samples were procured from the old-town market "Al Balad" in Jeddah, Saudi Arabia (21.481° N, 39.187° E) in Winter 2023. The collection includes 36 different samples of agarwood chips, dried Aquilaria spp. wood (also known as "bahkour"), and 22 agarwood steam-distillate oils, known as "oudh". The ages of the plants at harvest and origins of these samples cannot be accurately traced; however, many were labelled with their country of origin. The samples varied in price at the time of purchase, reflecting their purported rarity, the density of fragrant compounds, and complexity of aromatic notes (Supplementary Table 1). It was determined that the appropriate organic solvent for extraction was acetone based on preliminary analysis with different solvents, and each agarwood sample was then diluted to obtain chromatograms with clear peaks for product detection and identification. All samples were processed within 16 weeks of being procured. The wood samples were weighed (1 g) and ground into a fine powder using a combination of freeze–thaw cycles with liquid nitrogen and mechanical grinding. The homogenised samples were immersed in 5 mL of 1:1 hexane:acetone and subjected to ultrasonic agitation at 40°C for 8 h to facilitate terpenoid extraction. The samples were then passed through a 0.2 µm filter to obtain clear solvent extracts that were evaporated under a nitrogen stream for 20 min to concentrate the terpenoids in the samples. Concentrated samples were resuspended in 500 µL acetone and stored at –20°C until gas chromatography–flame-ionisation detector and mass spectrometry (GC–MS/FID) and two-dimensional gas chromatography with time-of-flight mass spectrometry (GCxGC–TOF/MS) analyses. All photographs were captured with a Canon EOS RP camera using a Canon RF 24–105 mm f/4-7.1 IS STM lens (Canon, Tokyo, Japan) and ColorChecker Passport (CCPP2, Calibrite LLC, DE, USA) used for colour calibration.

Algal strain cultivation
Chlamydomonas reinhardtii strain UPN22 was used for all experiments. C. reinhardtii UPN22 is a derivative of the UVM4 strain. It has been genetically enhanced to use phosphite and nitrate as a sole source of phosphorous and nitrogen, respectively, to minimise contamination and maximise cell densities in cultivation. Strains were cultured in Tris-acetate phosphite nitrate (TAPhi-NO3) liquid medium with shaking at 120 rpm or on solid agar and 150 µmol m–2 s–1 light intensity from a combination of cool- and warm white LED tubes (light spectra as reported in). 500 mL algal cultures were agitated with stirring in Erlenmeyer flasks under the same growth conditions.

Plasmid construction, algal transformation, and screening for sesquiterpenoid synthase expression
Heterologous expression of sesquiterpenoid synthases (STPSs) in C. reinhardtii was achieved through synthetic transgene redesign based on the amino-acid sequences of each STPS, following previously established protocols. We selected 21 STPSs from various species encoding isoforms that yield nine different STP skeletons (Table 1). Targeted STPSs for aristolochene, δ-guaiene, santalene, valencene, valerianol, zizaene, τ-cadinol, bisabolol, and patchoulol were designed; all accession numbers are listed in Table 1. Their amino-acid sequences were used to generate algal-adapted nucleotide coding sequences using the Intronserter programme. This programme back-translates amino-acid sequences to frequently used codons, removes unwanted restriction sites, and systematically integrates the first intron of CrRBCS2i1 at a set distance to enable expression from the C. reinhardtii nuclear genome, as previously reported. Algal-adapted coding sequences were synthesised and sub-cloned into pOptimized_3 expression plasmids by Genscript (Piscataway, NJ, USA). Ketocarotenoid biosynthesis in alga was achieved by transformation with the pOpt2_CrBKT_aadA plasmid, and knockdown of C. reinhardtii squalene synthase (Uniprot: A8IE29) was achieved using the previously reported luciferase–artificial microRNA expression plasmid pOpt2_ cCA_gLuc-TAA_i3_ami_Spec. Both plasmids confer resistance to spectinomycin as a selectable marker. Combining both cassettes into a single plasmid was unsuccessful. Co-transformation of both plasmids and selection on spectinomycin was combined with robotics assisted colony picking and plate-level screening of 768 transformants to find those with luciferase activity (indicative of SQS knockdown) and a brown-colony phenotype from CrBKT-mediated ketocarotenoid biosynthesis. All plasmid constructs used are listed in Supplementary Table 6 and the complete annotated plasmid sequences are available in Supplementary File 5.

The glass-bead protocol was used to transform the nuclear genome of C. reinhardtii with a plasmid DNA. Each plasmid was linearised using restriction enzymes (XbaI + KpnI, Thermo Scientific FastDigest), and 10 µg of DNA was used for each transformation. Following an ~8 h recovery period in liquid TAPhi-NO3 medium under low light, algal cells were plated on a selective medium with paromomycin (10 µg mL–1), spectinomycin (200 µg mL–1) or zeocin (15 µg mL—1) antibiotics either individual or combinations of selection agents relative to each target plasmid. Plates were illuminated continuously for ~7 d before colony picking. A PIXL robot (Singer Instruments, Watchet, UK) transferred up to 384 colonies per transformation event to TAPhi-NO3 agar plates. After an additional 3 d, a ROTOR robot (Singer Instruments) was used to replicate colonies onto new medium and plates containing amido black (150 µg mL–1) for fluorescence screening as previously described.

All algal-optimised STPSs were expressed as fusions with mVenus (yellow fluorescent protein) or the monomeric teal (cyan) fluorescent protein 1 (mTFP1). Transgene expression of each STPS was determined by fluorescence imaging at the agar-plate level, as previously described. Chlorophyll fluorescence was observed with 2 sec of 475/20 nm excitation and 640/160 nm emission to show colony presence/absence on amido black-containing plates. Cyan-green fluorescence was captured with 420/20 nm excitation and 480/20 nm emission filters using 2.5 min exposure. Yellow fluorescent signals were captured with 504/10 nm excitation and 530/10 nm emission filters with 30-sec exposures. Transformants displaying strong fluorescent-protein signals were selected and inoculated into 12-well plates containing 2 mL of liquid TAPhi-NO3 medium and grown with shaking at 160 rpm. Predicted molecular masses of the expressed heterologous STPS–FP fusions were verified using sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS–PAGE) in-gel fluorescence against the fusion-protein fluorescent reporter.

Heterologous sesquiterpenoid biosynthesis analyses
For each transformant, biosynthesis of heterologous STPs was assessed by gas chromatography–mass spectrometry (GC–MS/FID). Four individual transformants containing each plasmid with the highest fluorescence signals were selected, and solvent–culture two-phase living extractions were performed using a 10% v/v dodecane–culture overlay in 6-well plates, as previously reported17,53. Cultivations were performed in biological triplicate for 6 d in 4.5 mL TAPhi-NO3 media with 500 µL dodecane overlay. Phases were separated, and culture samples were taken for cell-density analysis by flow cytometry as previously described, and dodecane was spun at 20,000 x g for 3 min,  then 150 µL clarified solvent was transferred into amber GC vials in triplicate prior to analysis.

Samples were analysed as previously described using an Agilent 7890A gas chromatograph equipped with a mass spectrometer and a flame-ionisation detector (GC–MS/FID). The gas chromatograph comprises a 5975C inert MSD with a triple-axis detector and a DB-5MS column (30 m × 0.25 mm i.d., 0.25 μm film thickness). The injector, interface, and ion-source temperature profiles were set to 250°C, 250°C, and 220°C, respectively. In splitless mode, 1 μL of the sample was injected using an autosampler (G4513A, Agilent). The column flow was constant at 1 mL min−1, with helium as carrier gas. The initial GC oven temperature was set to 80°C for 1 min, increased to 120°C at 10°C min−1, raised to 160°C at 3°C min−1, and to 240°C at 10°C min−1, holding for 3 min. After a 12-min solvent delay, mass spectra were recorded using a scanning range of 50–750 m/z at 20 scans per second. Chromatograms were analysed with MassHunter Workstation software version B.08.00 (Agilent), and STPs were identified using the National Institute of Standards and Technology (NIST) library (Gaithersburg, MD, USA). Further identification was conducted with purified standard calibration curves ranging from concentrations of 1–1,200 μM in dodecane of δ-guaiene (CAT#B942760), patchoulol (CAT#P206200), santalene (CAT#S15065), valerianol (CAT#V914000, Toronto Research Chemicals, ON, Canada), bisabolol (CAT#95426), valencene (CAT#06808), cedrene (CAT#22133, Sigma-Aldrich, MO, USA) (Supplementary Fig. 6). For compound identification, retention-time acquisition, internal digital-library calibration, and method development, we used a set of 12 microampules containing a standard terpene mixture, which covered 98 terpenes at 1 mM in methanol (CAT# MSITPN101, MetaSci, ON, Canada, Supplementary Table 4).

Two-dimensional Gas chromatography time-of-flight mass spectrometry (GCxGC–TOF/MS) analysis
Comprehensive two-dimensional gas chromatography time-of-flight mass spectrometry (GCxGC–TOF/MS) analysis of agarwood and distillate extracts in acetone was performed using an Agilent 7890B gas-chromatography system equipped with a Zoex ZX1 cryogenic thermal modulator and a JEOL TOF MS (AccuTOF GCx-plus, JEOL, Japan). The GCxGC system featured a normal (non-polar x mid-polar) two-dimensional column configuration, comprising a first-dimension column with a 30 m non-polar HP-5MS UI capillary column (5%-phenyl-methylpolysiloxane) and a second-dimension column with a 2 m mid-polar BPX-50 capillary column (50% phenyl polysilphenylene-siloxane). We used helium (99.999%) as the carrier gas at a constant flow rate of 0.8 mL min−1. The GCxGC–TOF/MS injector temperature was maintained at 300°C with a 10:1 split ratio. The oven temperature was initially held at 80°C for 1 min, then increased to 325°C at 2°C min−1. The modulation period was set at 6 s with a pulse time of 0.35 ms. The mass spectrometer operated in electron ionisation (EI+) mode at 70 eV. Both the transfer-line and ion-source temperatures were maintained at 250°C. The detector voltage of TOF was set to 2,500 Volts, and data were acquired at a rate of 50 Hz. Mass spectra were obtained within a mass-to-charge ratio (m/z) from 50 to 600.

Gas chromatography data analysis
The analysis of terpenoid extracts followed a procedure similar to previously reported methods, and qualitative analyses primarily relied on the retention index and match factor. The GC–MS/FID data were processed using the MassHunter Workstation software version B.08.00 (Agilent Technologies, USA). The identification of compounds was assisted by the NIST Mass Spectral Library Version 2.3 (National Institute of Standards and Technology, Gaithersburg, MD, USA). The mass-spectral data, derived from the GCxGC–TOF/MS analysis, were evaluated with GC ImageTM Version 2.9 software (Lincoln, NE, USA) and referenced against the NIST2020/EPA/NIH EI Mass Spectral Library. The spectral data were cross-referenced with library spectra to identify potential chemical structures, facilitated by calculating the match factor and probability, thereby generating a list of probable compound matches. The match factor directly compares the unknown mass spectrum peaks with those in the library spectra, indicating their similarity. In contrast, the probability determines the relative likelihood of the list of hits being accurate, assuming that the unknown spectrum is present in the library. Using these metrics provides the relative assuredness of matching chemical structures within the library spectra.

Sesquiterpenoid production and concentration
Transformants of best-performing STPS isoforms that accumulated the most of each heterologous sesquiterpenoid in screening conditions, were subjected to solvent–culture two-phase cultivations at 300 mL scale in TAPhi-NO3 medium using FC-3283 as a solvent underlay phase, as previously described. FC-3283 is a perfluorinated amine that is inert and denser than water, forming an underlay to the culture and accumulating heterologous terpene products from the algae. After 6 days of cultivation, gravity and gentle centrifugation separated the fluorocarbon and cultures. After centrifugation to clarify, fluorocarbons were subjected to liquid–liquid extraction to partition accumulated STPs into an equal volume of 96% ethanol. The mixture was shaken for 16 h at room temperature at 200 rpm. Next, samples were again centrifuged gently to further separate the phases. FC-3283, after ethanol extraction, can be reused on algal cultures and is effectively recycled in this process. 500 μL aliquots from each phase were sampled and stored in separate GC vials at –20°C for subsequent analysis and separation performance quantification.

For each of the nine STPs biosynthesised by the algae, 20 mL 96% ethanol-containing STOs were generated. The ethanol fractions were pooled and subjected to organic solvent nanofiltration (OSN) in a dead-end cell to concentrate the terpenes in ethanol without evaporative losses. A Duramem solvent-resistant membrane (Evonik, Germany) with a nominal molecular weight cut-off value of 300 g mol–1 suitable for STP retention and chemical compatibility with ethanol was selected. The 200 mL ethanol–sesquiterpenoid mixture was loaded into the OSN chamber containing a 16 mm membrane disc and subjected to 20 bar pressure delivered by CO2 as an inert gas, to drive the ethanol phase through the nanofiltration membrane at 2.84 L m–2 h–1 flux. While ethanol permeated the membrane, STPs were kept in the retentate due to their molecular weight and consequently concentrated in the ethanol. The permeate ethanol is suitable for recycling and can be reused in subsequent liquid–liquid extraction processes.

Bulk hydroxylation of sesquiterpenoid backbones
To selectively introduce hydroxyl groups at the double bonds present in the terpenes, hydroboration-oxidation reactions were performed using two distinct organoboron reagents: borane–tetrahydrofuran complex (BH3·THF) (CAT#176192, Sigma-Aldrich) and 9-borabicyclo[3.3.1]nonane (151076, Sigma-Aldrich). Three different stoichiometric ratios (0.5, 1.0, and 2.0) were explored for each reagent to investigate different degrees of hydroxylation within the STP mixture. The sterically hindered 9-BBN was anticipated to selectively react with the least sterically demanding sites of the terpenes, while the borane–tetrahydrofuran complex was expected to facilitate hydroxylation even at the more challenging endocyclic positions. To remove residual ethanol and water content from the concentrated terpene mixture, a rotary evaporator was used under reduced pressure at 40°C, ensuring effective removal without compromising the integrity of the terpenes. To ensure the complete removal of water, molecular sieves (CAT#105734, Sigma-Aldrich) were added to the terpene–THF solvent mixture. The reactions were carried out in anhydrous tetrahydrofuran (THF, Sigma-Aldrich) to enable removal by evaporation after completion. Hydrogen peroxide (H2O2, 30%, VWR Chemicals) and sodium hydroxide aqueous solution (NaOH, Sigma-Aldrich) were used to convert the organoboron intermediates into non-toxic, water-soluble boric acid after the reaction. All subsequent terpene derivatives were extracted from the aqueous mixture with ethyl acetate.

Data analysis
To evaluate the impact of engineered modifications on STP biosynthesis and growth characteristics in C. reinhardtii strains, one-way analysis of variance (ANOVA) was performed to compare mean STP (patchoulol) titres and growth rates among different transformants and the parental control strain. This statistical approach allowed for the simultaneous analysis of differences between multiple groups, to provide a robust assessment of the experimental manipulations. ANOVAs were performed separately for STP titres and growth rates under mixotrophic and phototrophic conditions. Post-hoc pairwise comparisons using Tukey's HSD were conducted to identify which groups were statistically significantly different. A one-way ANOVA was used to assess the effects of differences in STP biosynthesis among tested strategies (single or double transformation), followed by a post hoc Tukey's HSD test for specific pairwise data-set comparisons. Mean production values (in fg cell–1 and mg L–1) for each STP were compared, considering standard deviations to evaluate data variability. Mean values were considered statistically significantly different at a level of p < 0.05. For data analyses, JMP v.16 (SAS Institute Inc, NC, USA) and Rstudio 3.6.2 (Posit Software, Boston, USA) were used. Data visualisation was done using JMP v.16 and GraphPad Prism v.10 (GraphPad Software, MA, USA). Image adjustments involved ColourChecker calibration (Calibrite LLC, DE, USA) paired with Adobe Lightroom (Adobe Inc., CA, USA) for colour accuracy. The Gardner Colour Scale was used for agarwood sample assessment. Images were cropped and organised in Affinity Photo v.1.10.6 (Serif Ltd., WB, UK). Analysis of images and assignment of colour values according to the Gardner scale was conducted using ImageJ software (NIH, USA). Diagrams and illustrations were made using Affinity Designer v.1.10.6 (Serif Ltd., WB, UK), chemical structures with ChemDraw v.20.1 (PerkinElmer, MA, USA) and all visual elements were harmonised in Affinity Publisher v.1.10.6 (Serif Ltd., WB, UK).