Arabidopsis thaliana cDNAs of 1-DEOXY-D-XYLULOSE-5-PHOSPHATE SYNTHASE (DXS, AT4G15560.1), 4-HYDROXY-3-METHYLBUT-2-ENYL DIPHOSPHATE REDUCTASE (HDR, AT4G34350.1, Phillips et al., 2008), and GERANYLGERANYL PYROPHOSPHATE SYNTHASE 11 (GGPPS11, AT4G36810.1, Beck et al., 2013) coding for plastidial enzymes referenced in TAIR (The Arabidopsis Information Resource) and cDNA of CASBENE SYNTHASE (CAS) from Jatropha curcas (King et al., 2014) were cloned into pEAQ-HT (Sainsbury et al., 2009) vector as described in Forestier et al. (2021). A codon optimized cDNA of NaGLS based on the sequence accession number XM_019410085 was synthesized by gBlock IDT with extensions allowing to clone it directly with In-Fusion® in AgeI/StuI linearized pEAQ-HT vector. Transient expression in wild-type N. benthamiana was performed as described in Forestier et al. (2021).

We detected both casbene and 16-OH-casbene in transiently expressed plants by extracting around 200mg of dry material with 5ml of hexane containing 100μg/ml of β-caryophyllene, then sonicating for 15min. We quantified the compounds by GC–MS as detailed in Forestier et al. (2021). For geranyllinalool and its derivatives, we extracted around 150mg of dry weigh (DW) of infiltrated tobacco with 1ml of ethyl acetate containing 100mg/L of β-caryophyllene. The samples were shaken overnight at 2000rpm on a IKA Vibrax VXR basic shaker and then centrifuged, and 100μl of the supernatant was used directly for GC–MS.

We isolated GGPP and 16-OH-GGPP by adapting the protocol described by Nagel et al. (2014). Approximately 750mg of ground dry material was extracted with 15ml of methanol/H₂O (7:3, v/v) and sonicated for 30min. We then added 5ml of water to the mixture, centrifuged for 3min at 2000g, and filtered through Whatman filter paper grade 1 and cotton. The cleared extracts were passed through Chromabond HX RA columns and pre-conditioned with 5ml of methanol and 5ml of water, and compounds were eluted with 3ml of ammonium formate 1M in methanol. Each eluate was dried under a stream of nitrogen and re-dissolved in 250 μl of water/methanol (1:1). We transferred 100 μl into glass HPLC vials, and 2 μl aliquots were analyzed by LC–MS as described by Catania et al. (2018). Additional high-resolution mass spectral data were obtained on a parallel LC interfaced to a Thermo Orbitrap Fusion mass spectrometer, operating in ESI mode at 500000 (FWHM) resolution for MS1 data, with MS2 data collected at 120000 resolution using stepped collision energies between 20 and 60units in both HCD and CID modes.

To identify 16-OH-casbene, we ground 4.9g of dry material obtained from 10 full-grown plants infiltrated with DXS, GGPPS, and CAS, and extracted with 100ml of hexane. After 1h of sonication and 2days shaking, the extract was centrifuged for 3min at 2000g, filtered through Whatman paper grade 1 and cotton, and evaporated to obtain 350mg of oily residue. The residue was re-suspended in 10ml of hexane/ethyl acetate (70:30, v/v) and purified through a 40g Buchi silica column on a PuriFlash® 4,250 system (Interchim). We used the same method of flash chromatography as described in King et al. (2014) to fractionate the extract into 80 samples. GC–MS was used to identify the fraction containing our compound of interest, and 2.6mg of this was obtained after evaporation, at sufficient purity for direct ¹H NMR analysis on a Bruker AVIII 700MHz instrument equipped with a cryoprobe.

For 16-OH-geranyllinalool, we infiltrated 40 young plants with DXS, GGPPS, and NaGLS, which provided 10g of dry material after freeze-drying and grinding. We extracted this with 150ml of ethyl acetate and left to shake for 5days on a rotary shaker. After centrifugation and filtration as detailed above, we reduced the volume down to 1ml before re-suspending in 9ml of hexane/ethyl acetate and purifying with the same column and method as described above. The fractions of interest were combined and dried to obtain 16.4mg of extract that was further purified with a reverse phase column [C18-HQ 5μm 250 mm×10 mm (Interchim)] to remove the pigment content. The reverse phase column was first equilibrated with solvent A – mix of water/acetonitrile (95:5, v/v) – for eight column volumes (CV), before injecting the extract, diluted in 2.5ml of the same solvent, into a 5ml injection loop. The separation method consisted of one CV of solvent A, followed by a gradient of nine CV, to reach 100% acetonitrile (solvent 0B). This solvent was maintained for a further 10 CV, and the entire run was carried out at a flow rate of 3ml/min. We used an in-line connected Advion Expression compact mass spectrometer (CMS), which enabled product isolation guided by mass spectra. To evaluate the fragmentation of 16-OH-geranyllinalool, we additionally ran the extracts on UPLC-MS, allowing us to determine two main ions at m/z 271 and m/z 289, which we used to select our compound of interest on the Puriflash-CMS. We collected one fraction, which after evaporation contained 4.6mg of sufficiently pure metabolite for NMR identification.

In the previous work, we tested the transient co-expression of different MEP pathway genes and GGPPS from A. thaliana with CAS from Jatropha curcas to evaluate the best combination for the highest production of casbene. We determined that the combination of DXS (catalyzing the first step in the MEP pathway), HDR (4-hydroxy-3-methylbut-2-enyl diphosphate reductase, catalyzing the last step), GGPPS, and CAS resulted in an up to 5-fold increase in casbene production, compared to CAS expression alone (Forestier et al., 2021). Omitting HDR from this combination resulted in lower production of casbene (Forestier et al., 2021). Further inspection of the total ion chromatograms of plant extracts from this reduced gene combination identified three additional peaks (Figure 1A), compared to co-expression of DXS, HDR, GGPPS, and CAS which only produced casbene (Figure 1B).

The largest of the three additional peaks was present in sufficient amount to allow its identification as 16-hydroxy-casbene (16-OH-casbene) by NMR spectroscopy (Supplementary Figure S1). 16-OH-casbene was present at approximately 30% of casbene levels when HDR was absent from the gene combination (Supplementary Figure S2).

The fact that 16-OH-casbene is only produced when HDR is omitted from the gene combination used to increase flux into the C₂₀ GGPP precursor for casbene production, led us to investigate the possibility that the hydroxyl group at the 16-position of the precursor is also hydroxylated. GGPP is formed by head-to-tail condensations between one molecule of DMAPP and three molecules of IPP (Ogura and Koyama, 1998), with the chain-starter DMAPP ending up distal to the pyrophosphate group (Figure 2A). In the MEP pathway, the immediate precursor to DMAPP is (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP; Figure 1C), which has the same structure as DMAPP, but with a hydroxyl functionality at position 4 (Figures 2A,B). This hydroxyl group would appear at the 16-position of GGPP after chain-extension by IPP if HMBPP is accepted by GGPPS as an alternative chain starter to DMAPP (Figure 2B). We considered therefore that 16-OH-GGPP could be formed if there were an excess of HMBPP, caused by the over-expression of DXS and GGPPS with insufficient conversion to DMAPP and/or IPP due to this being dependent on an endogenous HDR. 16-OH-GGPP could then be further incorporated into the casbene backbone assuming that CAS accepts 16-OH-GGPP as substrate for production of 16-OH-casbene (Figure 2B).

To establish whether 16-OH-GGPP accumulates depending on the gene combination, either DXS+GGPPS or DXS+HDR+GGPPS were transiently expressed in N. benthamiana and C₂₀ prenyl diphosphate intermediates were extracted as described by Nagel et al. (2014). UPLC-MS/MS negative mode analysis of methanolic extracts was used to detect GGPP by selecting the m/z range 449–450 and by comparison with an authentic GGPP standard (Figures 3A–C). In the absence of a 16-OH-GGPP standard, we predicted that since m/z 449.2 represents the [M-H]- ion for GGPP (Figure 3F), hydroxylated forms should be detectable at an added mass of 16; i.e at 465.2. Hydroxylated forms of prenyl diphosphates would be more hydrophilic and therefore elute at an earlier retention time compared to GGPP in reverse phase chromatographic separation. A clear peak at m/z 465.2, was detected at the earlier retention time of 2.5 min, in the extract of material over-expressing DXS+GGPPS (Figures 3D, F), consistent with 16-OH-GGPP. High-resolution mass spectrometry analysis revealed a m/z of 449.1869 for GGPP ([M-H]⁻ theoretical 449.1864; error 1.11ppm) and m/z of 465.1819 for 16-OH-GGPP ([M-H]⁻ theoretical 465.1813; error 1.29ppm). Both peaks generated a common 158.9252m/z MS2 fragment, identified as the diagnostic pyrophosphate group ion, [P₂O₆H]⁻. We were unable to detect the m/z 465.2 peak in the gene combination of DXS+HDR+GGPPS (Figure 3E), consistent with the hypothesis that OH-GGPP is produced when HMBPP reduction is limiting due to lack of HDR activity.

To further explore whether 16-OH-GGPP could be used by other diterpene synthases, we transiently expressed the Nicotiana attenuata geranyllinalool synthase (NaGLS) in N. benthamiana, alone or in combination with DXS+HDR+GGPPS (Figure 4 and Supplementary Figure S3). This resulted in accumulation of geranyllinalool in both cases. Co-expression of NaGLS+DXS+GGPPS produced geranyllinalool but also two additional peaks with retention times (R_t) of 30.0 and 31.6min (Figure 4A and Supplementary Figure S3E). We used UPLC-MS (Supplementary Figure S4) and flash chromatography to purify the compound giving rise to the larger peak at Rt 30.0 min in sufficient quantities to permit its identification by NMR spectroscopy as 16-OH-geranyllinalool (Supplementary Figure S5) This novel natural product represented up to 25% of the geranyllinalool peak (Figure 4C). This is comparable amount of 16-OH-casbene to casbene (30%) with the same reduced gene combination (Supplementary Figure S2).

Despite obvious parallels with the regio-isomeric 17-OH-geranyllinalool (Supplementary Figure S6), the precursor of the insecticidal diterpene glycosides in many Nicotiana species (Snook et al., 1997; Jassbi et al., 2010; Falara et al., 2014), there is no indication that 16-OH-geranyllinalool could be involved in the biosynthetic pathway to 17-OH-diterpenes. A recent work actually demonstrated that two cytochrome P450s from N. attenuata are responsible for the 17-hydroxylation of geranyllinalool (Li et al., 2021).

We did not detect any other hydroxy geranyllinalool compounds apart from 16-OH-geranyllinalool with the DXS+GGPPS+NaGLS gene combination, providing further evidence that 16-OH-geranyllinalool is derived from a direct conversion of 16-OH-GGPP.