We used NCBI BLAST to screen the DfFPS gene sequences from the transcriptome data of D fragrans. The primers designed to amplify the genes and the PCR results are shown in Supplementary Figure S1 . Sequence analysis revealed that the sequences of the amplified genes were consistent with the transcriptome data, and the open reading frames (ORF) of DfFPS1 and DfFPS2 were 1041 and 1065 bp in length, respectively. The DfFPS1 gene encoded a protein with 346 amino acids (39.87 kDa) and a predicted isoelectric point of 5.56; the DfFPS2 gene encoded a protein with 354 amino acids (40.67 kDa) and a predicted isoelectric point of 4.94. A BLAST search of the NCBI protein database revealed that the deduced amino acid sequences of DfFPS1 and DfFPS2 were closely related to the FPS protein of other terrestrial non-seed plants, such as Ceratopteris richardii, Huperzia serrata, Ceratodon purpureus, and Physcomitrium patens. Conserved between two proteins, the nucleotide sequence identity was 84.23% and the amino acid identity was 86.72%. Next, the sequences were compared with the amino acid sequence of A. thaliana FPS (AtFPS), which revealed that the DfFPS proteins had the same five conserved domains as the AtFPS proteins, which are numbered I to V (Chen et al., 1994) ( Supplementary Figure S2 ). The highly conserved aspartate-rich motif in region II with the amino acid sequence DDXX(XX)D is called FARM (first Asp-rich motif), which is highly conserved in all known prenyltransferases. Region V with the amino acid sequence DDXXD is called SARM (second Asp-rich motif). These regions (marked with lines) are characteristic of prenyltransferases, which can be used to synthesize isoprenoid diphosphates (Zhao et al., 2015). To explore the genetic relationship between DfFPS1 and DfFPS2 and FPS from other plant species, 36 FPS amino acid sequences of different plants were selected ( Supplementary Table S1 ) and a phylogenetic evolutionary tree was constructed using MEGA6. The results revealed that the FPS proteins of dicotyledons, monocotyledons, gymnosperms, and ferns were clustered in different evolutionary clades. The two DfFPS proteins clustered in a small branch and were closely related to FPS from Huperzia serrata, Phlegmariurus carinatus, Physcomitrium patens, and Ceratodon purpureus ( Figure 1 ).

The 3D structures of DfFPS1 and DfFPS2 were predicted using the online tool SWISS-MODEL, and the figures were prepared using PyMOL. All prediction models were built using ProMod Version 3.70 with x-ray at 2.20 Å. The template (SMTL ID: 4kk2.1.A) used for the 3D structure modeling of DfFPS1 had 66.37 and 50% sequence identity and similarity, respectively, with DfFPS1, and the coverage was 99%. The template (SMTL ID: 7bux.1.A) used for the 3D structure modeling of DfFPS2 had 68.13 and 51% sequence similarity and identity, respectively with DfFPS2, and the coverage was 99%. The two predicted proteins sequence were described as FPS, which were consistent with the results obtained for the cloned gene. Next, we compared the amino acid sequence of the two proteins with that of 1ubx.1.A, obtained from the SWISS-MODEL database, which is described as FPS with ligands FPP. We compared the structures of the two predicted proteins with that of 1ubx.1.A and found that the FPP binding site and amino acid residues of the two predicted proteins were the same of those of 1ubX.1.A ( Figures 2A, B ). We also analyzed the binding sites of the proteins, which revealed that DfFPS1, DfFPS2, and lubx.1.A had similar binding sites for FPP ( Figures 2B–D ).

To explore the expression pattern of the two DfFPS genes, qPCR was performed to analyze their relative expression levels in different growth-stages and parts of D. fragrans. The analyses revealed that both DfFPS genes were expressed in the roots, stems, and leaves, and their expression levels in the roots and petioles in different stages varied. The expression levels of the two genes in the leaves were significantly different based on growth-stages in the following order: sporangium mature and developed > sporangium shedding > sporangium not developed ( Figures 3A, B ).

However, based on the similarity in the expression patterns of the two genes in different tissues, we investigated whether this pattern is observed in other species, such as A. thaliana, which has two FPS genes that show a complementary pattern in seeds (Keim et al., 2012). AtFPS1 is the main gene in the process of plant growth and development (Closa et al., 2010). Therefore, we speculated whether one of the two genes in D. fragrans was acting as the main effect gene. We performed a comparative analysis of the expression of the two genes in the leaves, petioles, and roots of mature sporophytes and young gametophytes of D. fragrans. We found that the expression level of DfFPS1 was higher than that of DfFPS2 in these tissues ( Figure 3C ).

Moreover, qPCR analysis was performed to examine the changes in the expression of the two DfFPS genes under salicylic acid (SA), ABA, MeJA, and ethylene (ETH) treatments. The expression level of DfFPS1 first increased and then decreased under the four hormone treatments. DfFPS1 mainly responded to MeJA treatment, and its expression began to increase at 0.5h after MeJA treatment and reached the maximum at 24h, which was 11.47 times of that observed at 0h. DfFPS1 also responded to SA, ABA, and ETH treatments, and reached the maximum value at 1.5h, 6h, and 12h, with the expression levels being 3.77, 4.28, and 4.06 times of that at 0h, respectively ( Figure 4A ). The expression pattern of DfFPS2 under hormone treatment was similar to that of DfFPS1 ( Figure 4B ). However, the change in the expression level of DfFPS2 was unobvious after treatment. Meanwhile, the gene mainly responded to MeJA treatment, reaching 3.48-folds after 6h of treatment compared with that at 0h.

Abiotic stress can promote the production of secondary metabolites in plants (Wang et al., 2021). FPS is a key enzyme of the terpenoid metabolic pathway; hence, we hypothesized that FPS may be induced by abiotic stress. We investigated the expression patterns of the two genes under four types of abiotic stress. We found that DfFPS1 mainly responded to high temperature stress, followed by drought stress, but its response against salt stress and low temperature stress was negligible ( Figure 4C ). The gene expression level began to increase after 1.5h of high temperature treatment, reached 11.25 times after 6h of treatment compared with that at 0h, and returned to normal at 48h; after 1.5h of 20% PEG solution–simulated drought treatment, the expression level increased to 5.51 times of that at 0h and gradually decreased with time; with 200 mM NaCl treatment, at 12h–24h, the gene expression level increased to twice of that at 0h and decreased to original level at 48h; DfFPS1 expression did not change significantly under low temperature treatment. The expression pattern of DfFPS2 under stress revealed that it mainly responded to high temperature and drought stress but was not sensitive to salt stress and almost does not respond to low temperature stress ( Figure 4D ). The maximum expression of DfFPS2 was observed after 12h of high temperature stress, which was 3.98 times of that at 0h.

Using the PEG-mediated genetic transformation method, empty vector T-Egfp and fusion vectors T-DFFPSs-Egfp and T-Egfp-DFFPSs were transformed into Arabidopsis protoplasts. The protoplasts were cultured in the dark at 22–24°C for 12h–16h and observed using a confocal laser scanning microscope. The results are shown in Figure 5 . We observed a green fluorescence signal in the nucleus, cytoplasm, and membrane of Arabidopsis protoplasts transformed with the empty vector T-Egfp; the cytoplasm of Arabidopsis protoplasts transformed with T-DFFPS1-Egfp, T-DFFPS2-Egfp, T-Egfp-DFFPS1, and T-Egfp-DFFPS2 plasmids also exhibited green fluorescence signals. These results indicated that DfFPS1 and DfFPS2 proteins were localized in the cytosol of the Arabidopsis protoplasmic cells.

Confocal laser scanning micrographs showing the distribution of fluorescence signals in Arabidopsis protoplasts.

The entire reading frame of DfFPS genes was cloned into pET-32a(+) vector and expressed in Escherichia coli BL21 (DE3) cells to obtain DfFPS1 and DfFPS2 proteins for the characterization of FPS activity.

After induction by 0.5, 1.0, and 1.5 mM isopropyl thio-β-galactoside (IPTG) at 28°C and 200 rpm for 6h, the recombinant proteins were expressed at the same time, and we observed that the target protein content in the supernatants of both proteins was higher under 1.5 mM-IPTG treatment. The molecular mass of DfFPS1 ( Supplementary Figure S3A ) fused with a Trx-tag and a His-tag on N-terminal was approximately 57.6 kDa and that of the DfFPS2 ( Supplementary Figure S3B ) fusion protein was approximately 55–70 kDa, as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis. For all further experiments, 1.5 mM IPTG was selected for the induction of protein and the His-tagged proteins were purified using a Ni-column. And obtain purified TrxA-DfFPS1 ( Supplementary Figure S3C ) and TrxA-DfFPS2 ( Supplementary Figure S3D ) proteins.

In seed plants, FPS can condense IPP and DMAPP to form FPP under suitable conditions (Kajiura et al., 2017). To validate the function of DfFPS proteins in vitro, IPP and DMAPP were used as substrates to perform enzyme assays at pH 7.0 and 30°C for 30 min. The enzymatic properties of purified DfFPS1 and DfFPS2 proteins were validated by comparing the activity of the two proteins inactivated at high temperature.

High-performance liquid chromatography (HPLC) analyses of the standards and enzyme-activated products are presented in Figure 6 . In the figure, the red line is the IPP standard, which has a retention time (RT) of 5.058 min; the orange line is the DMAPP standard, which has an RT of 5.365 min; the blue line is the FPP standard, which has an RT of 12.340 min; the green line is the reaction between IPP and DMAPP as substrates at pH 7.0 and 30°C for DfFPS1 protein, and the RT of the product is 12.309 min, same with FPP standard; the purple line is the reaction between IPP and DMAPP as substrate at pH 7.5 and 30°C for DfFPS2 protein, and the RT of the product is 12.370 min, same with FPP standard.

Eight and seven transgenic tobacco plants overexpressing DfFPS1 and DfFPS2 were obtained, respectively, by agrobacterium-mediated genetic transformation ( Supplementary Figure S4 ).

Wild-type seedlings and T2 transgenic tobacco seedlings were grown up to the four-leaf stage in the medium, transferred to nutrient soil for culture, and the growth and development of the plants were observed after 21 days. Wild-type, 35S::DFFPS1-8, and 35S::DFFPS2-4 tobacco plants are presented in Figures 7A, B (from top to bottom). A significant difference was not observed between transgenic and wild-type tobacco plants. The length and dry weight of transgenic plants was measured ( Figures 7C , 8 ), which revealed that the root length of transgenic plants was not significantly different from that of wild-type and empty vector plants; however, the root dry weight of transgenic plants increased significantly compared with that of wild-type and empty vector plants, whereas a significant difference was not observed between the root dry weight of 35S::DfFPS1-8/9 and 35S::DfFPS2-4/6 transgenic tobacco plants.

FPS is the key enzyme of the terpene metabolic pathway. Overexpression of FPS genes may affect the expression of terpene metabolites in plants; therefore, we performed gas chromatography-mass spectrometry (GC-MS) to detect the metabolites in tobacco leaves. The leaves of the same part and growth stage were selected from the transgenic tobacco lines and were mixed in equal mass. GC-MS analyses revealed that the RT of the standard nonyl acetate was 7.433 min. Moreover, the main volatile secondary metabolites in tobacco were (1) nicotine, (2) neophytadiene, (3) n-docosanol, (4) cis-7,10,13,16,19-docosapentaenoic acid methyl ester, (5) manool, (6) geranyl linalool, and (7) eicosapentaenoic acid ethyl ester ( Supplementary Figure S5 ). Qualitative and quantitative analysis of tobacco leaves using nonyl acetate revealed that the nicotine content of 35S::DfFPS1 transgenic tobacco leaves was 3.12 times that of the wild-type leaves and 1.47 times that of neophytadiene content; cis-7,10,13,16,19-docosapentaenoic acid methyl ester, manool, geranyl linalool, and eicosapentaenoic acid ethyl ester contents in 35S::DfFPS2 transgenic tobacco leaves were 1.62, 2.47, 3.18, and 1.87 times those of wild-type tobacco leaves, respectively ( Figure 9 ).

Based on the fact that the DfFPS genes were induced by abiotic stress, we speculated that the overexpression of these two genes might lead to a stress-resistant phenotype in tobacco. Four-week-old transgenic and wild-type tobacco plants were selected and subjected to 150 mM NaCl treatment (salt stress), 20% PEG (drought stress), 42°C (high temperature stress), or 4°C (low temperature stress). Next, the leaves from the same parts of the plants were selected for nitro blue tetrazolium (NBT) and 3, 3’-diaminobenzidine (DAB) staining to preliminarily verify gene function. The leaves of DfFPS1- and DfFPS2-overexpressing tobacco plants under high temperature and drought stress were lighter than those of the wild-type plants, whereas the leaves of transgenic and wild-type tobacco plants were similar in color under low temperature and salt stress. NBT and DAB staining revealed similar patterns for these treatments ( Figure 10 , S6 ). The results revealed that transgenic tobacco plants exhibited a certain degree of resistance to high temperature and drought stress.

In the present study, DfFPS genes were found to respond to MeJA treatment and drought and high temperature stresses, and the transgenic plants exhibited a certain degree of resistance to high temperature and drought. However, the underlying mechanisms responsible for the stress response exhibited by the transgenic plants are unclear; therefore, we cloned the promoter sequences of the two DfFPS genes to analyze their transcriptional activity.

Since the genome of D. fragrans is not sequenced yet, we used fusion primer and nested integrated PCR (FPNI-PCR) for genome walking to amplify the promoter sequence of the gene from the predicted coding sequence (CDS) (Wang et al., 2011). The genome walking results of DfFPS1 revealed that none of the FP primers could successfully amplify the gene ( Supplementary Figure S7A ). The genome walking results of DfFPS2 revealed that the FP2/3/4/5/7 primers (of the nine random primers) could successfully amplify the promoter ( Supplementary Figure S7B ). Sequencing analyses revealed that the amplicon obtained using the FP2 primer could coincide with the sequence of the 5’-UTR of DfFPS2, and the length of the promoter of DfFPS2 was 1554 bp. Since the promoter sequence DfFPS1 could not be amplified using the FP1-9 primers, seven new FP primers (FP10-16) were designed according to the primer design principle described in the materials and methods section. FPNI-PCR using the FP11/14/15 primers successfully amplified the DfFPS1 promoter ( Supplementary Figure S7C ). Sequencing analysis revealed that three sequences were the same and could be compared with the 5’-UTR sequence of DfFPS1. Since the amplicon obtained using the FP15 primer was long (943 bp), this sequence was selected for subsequent experiments.

To analyze the transcriptional activity of the full-length promoters of the two genes, the promoter sequences of the two genes were cloned into pBI121 upstream of the β-glucuronidase (GUS) reporter. GUS expression driven by the two promoters was high in the stomata of tobacco leaves, which has not been reported earlier ( Figure 11 ). Moreover, the transcriptional activity of proDfFPS1 was higher than that of proDfFPS2, which is similar to qPCR results ( Figure 3C ).

To determine the active site of the promoters, according to the PlantCare-based predictions, DfFPS1 and DfFPS2 promoters were truncated into different lengths according to cis elements to control transient expression of GUS in tobacco leaves. proFPS1-Full exhibited high transcriptional activity, whereas proFPS1-Δ1 and proFPS1-Δ2 exhibited negligible transcriptional activity, indicating that the sequence between proFPS1-Full and proFPS1-Δ1 could be the transcriptionally active region of DfFPS1 ( Figure 12A ). We also found that the transcriptional activity of DfFPS2 promoter with different lengths was low, with almost no transcriptional activity ( Figure 12B ).

Next, we analyzed the response of the promoters of different lengths to stress, and we selected the leaves of the same parts of each plant for transformation. To ensure the accuracy of the experiment, different truncated fragments of the same promoter were injected into the same leaf, and the plants were treated with MeJA, 150 mM NaCl, 20% PEG, 42°C, or 4°C for 6h after 3 days of transformation. The results revealed that the transcriptional activity of the truncated DfFPS1 promoter was changed to some extent under MeJA treatment and abiotic stress, and the main transcriptional active site of DfFPS1 promoter was the fragment between proFPS1-Full and proFPS1-Δ1. Similarly, under different treatments, the transcriptional activity of DfFPS2 promoters of different lengths under MeJA and abiotic stress was low ( Figure 13 ).

D. fragrans spores were collected from lava rocks near Bagua Lake, Wudalianchi, Heilongjiang, China (48.733334 N, 126.167966E) within a 30-m radius during July 1–6, 2018, after obtaining necessary permission from the government. To ensure uniformity in samples, we used D. fragrans that were grown at 25°C under a 16h/8h photoperiod for 2 years.

The leaves, petioles, and roots of D. fragrans were selected from 2-year-old plants at different stages, that is, undeveloped, developed, mature, and shedding sporangia. Two-year-old D. fragrans sporophytes were subjected to the following hormone treatments: 100 μM SA, 10 μM ABA, 120 μM MEJA, and 500 μM ETH, and water was used as the control; they were also subjected to the following four types of abiotic stress: 200 mM NaCl (salt stress), 20% PEG irrigation (drought stress), 42°C (high temperature stress), 4°C (low- temperature stress), and 24°C and water irrigation (control). The samples were collected, immediately frozen in liquid nitrogen, and stored at −80°C until further use.

A. thaliana ecotype Columbia-0 (Col-0) was grown in a growth chamber under a photoperiod of 16 h/8h (light/dark) at 22°C.

Nicotiana tabacum L. (Shan Xi) seeds were sterilized and cultured in 1/8 MS medium (M519; Phytotech ,Lenexa, USA). After 4 weeks, the leaves were cut into 1-cm square pieces and placed on M1 medium (MS agar medium containing 0.1 mg/liter NAA and 1 mg/liter 6-BA) for 2–3 days. The prepared agrobacterium culture was adjusted to an OD₆₀₀ of 0.8 using sterile water and transferred to the leaves for 8–10 min; next, the culture was drained and returned to the original medium. After 4 days of culture, the leaves were washed with sterile water and transferred to M2 medium (M1 medium containing 50 mg/liter Kan and 50–400 mg/liter Cef) until resistant buds developed. The resistant buds were grown up to a height of 2 cm and then transferred to 1/2 MS medium for rooting.

Four-week-old N. tabacum L. plants were used to investigate instantaneous expression, and the plants were watered a day before transformation. The prepared agrobacterium culture was adjusted to an OD₆₀₀ of 0.6 by injecting a buffer (10 mM MgCl₂, 10 mM MES, and 150–200 μM AS). The bacterial solution (1 ml) was injected using a medical syringe into the tobacco leaves from the lower epidermis through the stomata. After injection, the plants were kept at 22–24°C for 1 day in dark and then in low light for 1 day, followed by growth under normal conditions for 2 days.

Wild-type and T2 transgenic tobacco seedlings were grown to the four-leaf stage in the MS medium, then transferred to nutrient soil for further growth, and their growth and development were observed after 21 days. Simultaneously, the plants were subjected to the following four types of abiotic stress: 200 mM NaCl (salt stress), 20% PEG irrigation (drought stress), 42°C (high temperature stress), 4°C (low-temperature stress) for 12h; then, the leaves of the same part were stained using NBT and DAB.

Transient transformed tobacco plants were subjected to 120 μM MeJA treatment and the abovementioned abiotic 12h before sampling, and the leaves were stained using X-gluc solution.

RNA of D. fragrans was extracted from the sporophyte leaves using a Trelief TM RNAprep Pure Plant Plus Kit (Polysaccharides & Polyphenolics-rich; Tsingke, China), and reverse transcription was performed using a HiScript III 1st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, China) to obtain cDNA. The putative DfFPS genes were amplified using 2×Phanta^® Max Master Mix (Vazyme, Nanjing, China) with the following primer sets: DfFPS1 (F: 5′-ATGTCCAAGGATGCATCAGG-3′; R: 5′-TCATTTTTGTCTTTTGTAAATTTTTCCCAAG-3′); DfFPS2 (F: 5′ATGGCTCCTACTGTGGAACT-3′; R: 5′-TCATTTTTGTCGTTTGTAGATTTTTCCTAAG-3′). The obtained amplicons were inserted in a vector (pClone007 Versatile Simple Vector Kit), and the plasmid was sequenced.

The nucleotide and protein sequences were compared using the NCBI tools (http://www.ncbi.NLM.NIH.gov). The ORF of the two genes was identified using ORF Finder (www.ncbi.NLM.NIH.gov/Gorf/Gorf.html). Multiple sequence alignments were performed using the DNAMAN software. Phylogenetic analysis was performed using the MEGA 6.0 software, and the sequences were aligned using ClustalW. A phylogenetic tree was constructed using Evolview (https://www.evolgenius.info/evolview/#login) and was based on the P-distance model of the neighbor-joining method, with partial deletion and interior-branch test replication being 1,000 times (Tamura et al., 2013) to construct the evolutionary tree. The 3D structures of DfFPS1 and DfFPS2 were predicted using the SWISS-MODEL online tool (https://swissmodel.expasy.org/) (Bienert et al., 2017), and figures were prepared using PyMOL.

RNA was extracted as described above; for cDNA synthesis, 1000 ng of total RNA was mixed with the HiScript^® II Q RT SuperMix for qPCR (+gDNA wiper) (Vazyme, China). We performed qPCR using ChamQ Universal SYBR qPCR Master Mix (Vazyme, China), as per the manufacturer’s instructions. The primers were designed using Primer Premier 6 ( Supplementary Table S2 ). The transcript levels were normalized to that of the 18S rRNA gene of D. fragrans. The relative expression levels were calculated using a previously reported formula (Livak and Schmittgen, 2001). All biological replicates were measured in triplicate, and a plot was constructed using GraphPad Prism 7 and TBtools.

To verify the subcellular localization of DfFPS proteins, fusion plasmids T-DFFPS1-Egfp, T-DFFPS2-Egfp, T-Egfp-DFFPS1, and T-Egfp-DFFPS2 were constructed using the Gibson assembly. All plasmids used in the transient transfection assays were endotoxin free. The plasmids were extracted using the GoldHi EndoFree Plasmid Maxi Kit (CWBIO), and their concentration was adjusted before transfection to approximately 4000 ng·µl⁻¹ using isopropanol and 5 M NaCl solution. The transient Arabidopsis protoplast transfection assays were performed as previously described (Yoo et al., 2007). Briefly, 3- to 4-week-old rosette leaves of wild-type Arabidopsis plants were harvested and immediately cut into strips using a razor. The strips were incubated in an enzyme solution containing 1% (w/v) cellulase R10 (Yakult Pharmaceutical Ind. Co., Ltd, Japan) and 0.25% (w/v) macerozyme R10 (Yakult Pharmaceutical) at 22–24°C for 3h. The protoplasts were purified, and the fusion plasmids were transfected into the protoplasts using PEG. After transformation, the protoplasts were cultured at 22–24°C for 12–16h under dark conditions, and they were observed using a confocal laser microscope.

To verify the enzyme activity of the two proteins, the genes encoding the proteins were cloned in a prokaryotic expression vector [pET32a(+)-DfFPS1 and pET32a(+)-DfFPS2] using the Gibson assembly. E. coli BL21 (DE3) cells were used for recombinant protein expression; the protein was induced using 1 mM IPTG at 16°C for 20h. After the cells were harvested, they were resuspended in 2 ml of lysis buffer A [20 mM tris-HCl (pH 8.0), 500 mM NaCl, 1 mM PMSF, and bacterial protease inhibitor cocktail], lysed by sonication (sonication for 5 s, followed by pause for 5 s), centrifuged for 30 min at 13,000 × g, and the protein was harvested from the supernatant.

The supernatant was loaded on a Ni-NTA column, which was pre-equilibrated using lysis buffer A. The column was washed with 10 column volumes of lysis buffer A containing 20 mM imidazole. The target proteins were eluted using buffer B [20 mM tris-HCl (pH 8.0) containing 500 mM NaCl and 50 mM imidazole] and buffer C [20 mM tris-HCl (pH 8.0) containing 150 mM NaCl and 250 mM imidazole). The purified proteins were subjected to SDS-PAGE. The concentration of the proteins was determined using the Bradford method.

Enzyme activity analysis and HPLC were performed as previously described (Ferriols et al., 2015), with minor modifications. The enzyme assays were performed in 1.5-ml Eppendorf tubes in a total volume of 100 μl containing 35 mM tris-HCl (pH 7.5), 10 mM MgCl2, 4 mM dithiothreitol (DTT), 50 μM isopentenyl pyrophosphate (IPP; Echelon Biosciences, Inc) and 150 μM dimethylallyl pyrophosphate (DMAPP; Echelon Biosciences, Inc) as the allylic substrates, and 100 ng of the purified enzyme. The assay mixtures were prepared in bulk without the substrates, and the reactions were initiated by substrate addition after sufficient equilibration of the system temperature. The reactions were performed at 30°C for 10 min and stopped by snap freezing the assay tubes in liquid nitrogen. To denature and precipitate the proteins before HPLC analysis, 100 μl of acetonitrile was added to the reaction mixture, and the mixture was vortexed for 10 s and centrifuged at 19,000 × g for 5 min at 4°C.

HPLC was performed using an Agilent Technologies 1200 Series HPLC system. The separation of the reaction product FPP was achieved on an Agilent ZORBAX SB-C18 column (particle size: 5 μm, 4.6 mm × 150 mm) using a mobile phase consisting of 5 mM NH₄HCO₃ and 100% acetonitrile as solvents A and B, respectively. A gradient system was used for the analyses, wherein 2% solvent B was increased to 5% in 5 min, followed by an increase of solvent B to 50% in 10 min and to 100% in 5 min, and this was maintained for 5 min. The flow rate was 0.8 ml·min⁻¹; the column temperature was 30°C; the injection volume was 10 μl; and the detection wavelength was 214 nm for all samples and standards.

Secondary metabolites from transgenic plants were detected as per a previously described method (Song et al., 2022). Briefly, we added 0.1 g of tobacco leaves (three plants of each transgenic line were selected at the four-leaf stage; the samples from both lines were mixed for the experiment) and 1 ml of ethyl acetate standard solution to a 1.5-ml Eppendorf tube, which contained 8.64 µg·ml⁻¹ nonyl acetate as the internal standard. The Eppendorf tube was centrifuged using an Ultra Centrifugal Mill (700 rpm, 5 min) and ultrasonicated using a small ultrasonic cleaner (60 kHz, 40°C, 30 min). After sonication, the tube was centrifuged at low temperature (5000 × g, 5 min), and the supernatant was aspirated and filtered through a 0.22-µm microporous membrane for GC-MS.

Terpenoids were analyzed using an Agilent 7890A gas chromatograph coupled to an Agilent 5975C Network Mass Selective Detector (MS, insert XL MSD with triple-axis detector) and an HP-5MS column (30 m × 0.25 mm × 0.25 µm; J&W Scientific, Folsom, CA, USA). The GC-MS conditions were as follows: Carrier gas: He (purity ≥ 99.999%); injection port temperature: 280°C; injection method: splitless injection; injection volume: 1 µl. The temperature conditions were as follows: the initial temperature was 60°C for 2 min, then increased to 220°C for 1 min at 20°C·min⁻¹, followed by an increase to 250°C for 1 min at 5°C·min⁻¹, and final increase to 290°C for 7.5 min at 20°C·min⁻¹. The MS conditions for the analyses were as follows: ion source: EI; ion source temperature: 230°C; ion energy mode: tuning setting; ion energy (eV): 70; detector setting: gain factor; solvent delay: 5 min; mass scan range: 30–500 m/z.

The promoter was cloned using FPNI-PCR (Wang et al., 2011), and 2×Rapid Taq Master Mix (Vazyme, China) were used to obtain the promoter.

The 1^st PCR was a thermal asymmetric interlaced (TAIL)-PCR performed using D. fragrans genomic DNA as the template with FP1-9 primers ( Supplementary Table S3 ) and gene specific primers DfFPS1-SP1and DfFPS2-SP1 ( Supplementary Table S4 ). The amplicons from the 1^st PCR were used as the template for the 2^nd PCR, and nested PCR-specific primers FSP1 ( Supplementary Table S3 ) and secondary gene specific primers DfFPS1-SP2 and DfFPS2-SP2 ( Supplementary Table S4 ) were used. In the 3^rd PCR, the amplicons from the 2^nd round were diluted 100 times and used as the template, and specific primer FSP2 ( Supplementary Table S3 ) and gene specific primers DfFPS1-SP3 and DfFPSP2-SP3 ( Supplementary Table S4 ) were used for common PCR. The specific programs that were used are listed in Supplementary Table S5 .

Since the promoter of DfPS1 could not be amplified using the FP1-9 primers, we designed seven new FP primers. The sequence of primers FP10-16 is presented in Table 1 . The same PCR procedure and gene-specific primers were used to amplify the promoter of DfFPS1.

We used the plant cis-acting regulatory element (plantCARE) website (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) to predict potential cis promoter regions of the two genes (Lescot et al., 2002). The transcriptional activity of the full-length promoters of the two genes was verified by investigating the transient expression of DfFPS genes in tobacco leaves. The promoters for both genes were truncated according to the homeopathic element cis on the promoters, and truncated DfFPS1 (938, 740, and 327 bp) and DfFPS2 promoters (1554, 1223, 1051, 632, and 345 bp) with different lengths were obtained, using the primers listed in Supplementary Table S6 .