Wurfbainia villosa is from Yangchun City, Guangdong Province, China. The leaves, rhizome, flowers, and fruits from healthy plants were collected and frozen at −80°C. The fruits at different developmental stages: 30-days after flowering (DAF), 45-DAF, 60-DAF, 75-DAF, and 90-DAF were separated into pericarp and seeds.

To identify putative BAHD genes in W. villosa, we used the Protein Basic Logical Alignment Search Tool (BLASTP) (E-value cutoff of 1e-5) to compare Arabidopsis thaliana and Malus domestica BAHD protein sequences (Liu et al., 2020) with the W. villosa genomic data (unpublished). The amino-acid sequences of A. thaliana and M. domestica were obtained from phytozome.¹ Then, based on the hidden Markov model (HMM) of the BAHD domain (PF02458) obtained from the Pfam database,² we also identified candidate WvBAHDs using Hmmsearch (3.2.1) (Potter et al., 2018). Then, we used the Web CD-Search Tool³ and Web SMART⁴ to confirm the domains of the candidate WvBAHDs identified by the two methods mentioned above, and then their HXXXD and DFGWG motifs were inspected. Furthermore, PF07247 (alcohol acetyltransferase, AATase) was used to screen AAT as the supplement of BAHD genes. Collinear blocks of the WvBAHD were identified using MCSanX and the results of circular plots were generated using Circos (Wang et al., 2012). The amino acid sequences alignments and the maximum-likelihood (ML) phylogenetic tree with bootstraps of 1,000 were obtained by using the MUSCLE and IQ-TREE of TBtools (Chen et al., 2020a), and the clade of clustering adapted from Liu et al. (2020).

The WvAAT candidate genes were selected directly by their clade clustered into and their transcripts per million (TPM) expression value (from unpublished RNA-seq data) combined with the upstream gene WvBPPS in different tissues and the fruits at different developmental stages. To predict the substrate preference of WvAAT, the multi-sequence alignments and phylogenetic analysis with the reported AATs were performed using the ML method with the tool MEGA X (Kumar et al., 2018). The information of the AATs from other plants used for the phylogenetic analysis was shown in Supplementary Table 2. Candidates were further analyzed for a multiple sequence comparison analysis of the nucleotide and amino acid sequences using the software Jalview (Waterhouse et al., 2009). The conserved and active motifs predictions were used by the online tool MEME⁵ (Bailey and Elkan, 1994); the protein structure, signal peptide, and transit peptide of the WvAAT candidates were predicted and analyzed by online platform TMHMM 2.0,⁶ SignalP,⁷ and ChloroP 1.1 Server⁸ (Emanuelsson et al., 1999; Krogh et al., 2001; Petersen et al., 2011). According to the results of the sequence alignment phylogenetic tree, the evolutionary tree was beautified by the online tool iTOL⁹ (Letunic and Bork, 2021).

The extraction method of total RNA and complementary DNA (cDNA) was the same as the previous reports (Wang et al., 2018). The coding regions of AAT candidate genes were amplified from W. villosa cDNA using the Primer STAR Max DNA Polymerase (Takara, China) with appropriate primers (Supplementary Table 3). The PCR conditions used were the following: 98°C, 1 min; 98°C 10 s, 50–60°C, 15 s, 72°C, 15 s, 30 cycles; 72°C, 5 min. The purified PCR products were then ligated into pLB (Tiangen, China) or 007 vs. (TSINGKE, China) cloning vector, which were consequently transformed into E. coli DH5α cells and sequencing.

The full-length ORFs of the candidate WvAATs were ligated into pET32a (+) expression vector, using the In-Fusion Cloning Kit (Takara, China), and then transformed into the Escherichia coli Rosetta (DE3). The primers were described in Supplementary Table 3. Cells were grown at 37°C until the OD₆₀₀ reaches.4–0.6, and induced at 16°C with isopropyl-β-D-thiogalactopyranoside (IPTG) at a final concentration of 10 μM for 16–20 h in Luria–Bertani (LB) media supplemented with 50 μg/μL carbenicillin and 25 μg/μL chloramphenicol. The recombinant protein was purified using NI-NTA resin (Qiagen, Hilden, Germany). The purified protein was dialyzed on a PD-10 desalting column (GE Healthcare).

The in vitro enzymatic reactions were performed at 250–350 μL total volume [20 mM sodium hydrogen phosphate buffer (pH 9.0), 10% glycerol] containing 20–80 μg purified protein, 0.2 mM acetyl-CoA, and 0.02 mM terpene alcohol substrate [(+)-borneol, (−)-borneol, isoborneol, geraniol, nerol, α-terpineol, (−)-menthol, linalool and nerolidol]. According to the previous reports on AAT enzymatic assays (Croteau and Hooper, 1978; Sarker and Mahmoud, 2015) and the results of our preliminary experiments, the optimum pH and temperature for WvAATs were determined to be 9.0 and 32°C, respectively. All reaction mixture was incubated at 32°C for 6 h, then immediately overlaid with 250 μL hexane. The mixture was then centrifuged at 12,000 rpm for 5 min to separate the phases. The hexane extraction was analyzed using Agilent 7890B Gas Chromatograph with 5977A inert Mass Selective Detector (Agilent, United States). Helium was used as the carrier gas (1 mL/min) and then separated on the HP-5MS column (30 m × 250 μm × 0.25 μm film thickness). The gas chromatography (GC) oven temperature was programmed at an initial temperature of 35°C for 2 min with an increase of 12°C/min to 300°C. The temperature was then kept at 300°C for 5 min. For chiral compounds, CycloSil-B column (30 m × 0.25 mm id, 0.25 mm film thickness) was used for separation, initially at 50°C for 2 min, and then at 5°C/min from 50 to 180°C, increase to 230°C at 10°C/min, and hold at 230°C for 2 min. NIST14/Wiley275 Mass Spectral Library was used for metabolite identification. Meanwhile, the standards were also utilized for further identification. Each WvAAT was tested in triplicate.

Approximately, 0.1 g of seeds at different developmental stages were ground frizzed in liquid nitrogen and extracted with 1 mL hexane using an ultrasonic cleaner for 30 min, and then incubated at 40°C for 1 h. The samples were then centrifuged at 10,000 rpm for 15 min and the resulting supernatants were pipetted into novel 2 mL tubes. One milliliter of hexane extract was pipetted into 1 mL vial for GC- mass spectrometry (MS) analysis. The extraction was analyzed using Agilent 7890B Gas Chromatograph with 5977A inert Mass Selective Detector (Agilent, United States). Helium was used as the carrier gas (1 mL/min) and then separated on the CycloSil-B column (30 m × 0.25 mm id, 0.25 mm film thickness). The GC oven temperature was programmed at an initial temperature of 35°C for 2 min, and then at 5°C/min from 35to 200°C, increase to 230°C at 10°C/min, and hold at 230°C for 2 min. The temperature was then kept at 240°C for 5 min. NIST14/Wiley275 Mass Spectral Library was used for metabolite identification. The terpene compounds were identified by the mass spectral library. The predominant terpene acetates and their precursor terpene alcohols in this research were further identified using their authentic standards. There were three biological replicates and three technical replicates for each organ.

Based on the correlation analysis between transcriptome and ester data of W. villosa, eight identified WvBATs were selected for quantitative real-time PCR (qRT-PCR). The primers for real-time PCR of all the WvBAT genes were designed manually (Fluorescence-specific primers shown in Supplementary Table 3). The quantitative PCR (qPCR) of WvBATs was performed using 2xTSINGKE^®Master qRT-PCR Mix-SYBR (+ UDG, TSINGKE, China) in the CFX96 real-time PCR detection system (Bio-Rad, United States). The WvBAT transcript levels were monitored using the internal reference gene TUA, and calculated using the 2^–ΔΔCt method. There were three biological replicates and three technical replicates for all experiments.

Preliminary element analysis was performed on the approximate area of the promoter. The 2,000 bp non-coding region upstream of the WvBATs coding region was extracted through TBtools (Chen et al., 2020a) and then analyzed on the online website PlantCare¹⁰ (Lescot et al., 2002) to find out whether the element was related to seed-specific expression. The classification of the cis-acting elements of the promoters of functional genes was referred to Abdullah et al. (2018).

Sixty-one putative BAHD genes with complete ORF and two conserved motifs were identified from the genome database of W. villosa using the BAHD HMM configuration file (PF02458, transferase). In addition, three genes (WvBAHD3, 52, and 53) were screened out using PF07247 (AATase) as HMM. AATase (PF07247) comes from CL0149, a protein clan named CoA-acyltrans with a characteristic HXXXD motif, same as the transferase family (PF02458). The AATase family contains a number of AATs from bacteria and metazoa, catalyzing the esterification of isoamyl alcohol by acetyl-CoA, similar to the AATs from the plant (Minetoki et al., 1993; Zhu et al., 2015; Lin et al., 2016; Reyes-Sánchez et al., 2019). Accordingly, we speculated that these three genes screened by PF07247 (AATase) should be members of the BAHD superfamily. In total, 64 WvBAHDs have been identified as BAHD super acyltransferase family members (Supplementary Table 4), which are distributed on 14 chromosomes and nine unanchored contigs (Figure 2A). The genome synteny analysis showed that 15 segmental duplications (23.4%) and four tandem duplication (6.3%) events occurred, suggesting that segmental duplication events have an important contribution to the expansion of the WvBAHD gene family (Figure 2B). The prediction results of the trans-membrane function indicated that most WvBAHDs were located in the cytoplasm (Supplementary Table 5).

To determine the classification of the WvBAHD gene family, we constructed a phylogenetic tree using amino acid sequences encoded by 235 BAHD genes from W. villosa, A. thaliana, and M. domestica, and 14 identified AATs from other plants. The 64 putative WvBAHDs were classified into five clades: I-a (twenty-two genes), I-b (fifteen genes), II-a (nine genes), II-b (fourteen genes), III-a (one gene), and VI (three genes) (Figure 3A). The genes belonging to clade I-a are involved in modifying aromatic and terpenoid alcohols in A. thaliana and M. domestica, therefore, we speculated that the WvBAHDs clustered into clade I-a in W. villosa might encode proteins with similar functions. However, the identified AAT were also clustered into clade I-b and clade IV, which is inconsistent with the reports that the members of clade I-b had functions related to the biosynthesis of lignin monomeric intermediates (Hoffmann et al., 2003; Petrik et al., 2014). This might be due to the phylogenetic tree algorithm and these enzymes have not been comprehensively identified to date. Furthermore, three AATase-WvBAHDs are clustered separately as a novel clade and named clade VI. Therefore, the range of WvAAT candidate genes we inferred here was expanded from clade I-a to clade I-b and clade VI. To confirm the correlation between the above clustering and conserved motifs, we detected eight conserved motifs in the WvBAHDsof W. villosa by MEME. Motif1 and Motif2 correspond to conserved motifs HXXXD and DFGWG, respectively (Figure 3B and Supplementary Figures 1, 2). The similarity in the type and distribution of conserved motifs in the same clade further supports the classification of evolutionary trees. Among the members of clade VI, the histidine residues from the HXXXD motif were substituted by threonine or serine, and four residues from the DFGWG motif were substituted, causing the clade VI members distant relation with other members of BAHD and other identified AATs (Figures 3A, 4).

Then, we compared the expression profiles of the genes from clade I and VI with WvBPPS, the key gene for borneol biosynthesis, and identified four genes (WvBAHD14, WvBAHD40, WvBAHD39, WvBAHD50; TPM >10) co-expressed with WvBPPS in 60-DAF seeds, suggesting that these genes might be the key genes involved in the borneol acetylation (Figure 3C). Furthermore, in clade I-a, WvBAHD8, WvBAHD38, and WvBAHD31 share high identity (>45%) with the reported AATs (CaAT20, LiAAT3, VpAAT1, and FaAAT2); WvBAHD9, WvBAHD14, WvBAHD39, WvBAHD40, and WvBAHD50 are clustered with MsBanAAT into a close branch; in clade I-b, WvBAHD18 and WvBAHD28 share approximately 40% identity with ClLAAT1, an AAT from Citrus limon (Figure 4). These reported AATs to show enzymatic affinity to geraniol and nerol in vitro (Aharoni et al., 2000; Supplementary Table 2), therefore, we speculated that these WvBAHDs might have AAT activity similar to these reported AATs. Thence, ten genes (WvBAHD8, WvBAHD9, WvBAHD14, WvBAHD18, WvBAHD28, WvBAHD31, WvBAHD38, WvBAHD39, WvBAHD40, and WvBAHD50) from clade I and two genes (WvBAHD53 and WvBAHD3) from clade VI, which are all expressed in the seeds (Supplementary Table 6), were selected as candidate WvAATs for cloning and functional characterization.

Enzymatic assays were conducted in vitro using recombinant proteins extracted and purified from E. coli expression strains (Supplementary Figure 3). The enzymatic assays showed that eight WvAAT recombinant proteins (WvBAHD8, WvBAHD14, WvBAHD28, WvBAHD39, WvBAHD40, WvBAHD50, WvBAHD3, and WvBAHD53) can acetylate geraniol and nerol, which are primary alcohols easily to be catalyzed, to form geranyl acetate and neryl acetate, respectively, even with the small amount of soluble proteins, verifying that these eight candidates have AAT enzymatic activity, and this method is effective for the AAT characterization in vitro, according to previous research (Shalit et al., 2003; Sarker and Mahmoud, 2015; Supplementary Figure 4). In the case of the other four candidate WvAATs, no geranyl acetate or neryl acetate was detected from the enzymatic reaction, and they could not catalyze bornyl-type substrate either.

To determine the BAT activity of these eight WvAAT recombinant proteins mentioned above, the catalytic products using borneol-type substrates, including (−)-borneol, (+)-borneol, and isoborneol, were analyzed. The results reveal that all these WvAATs can catalyze the substrate (−)-borneol to produce (−)-BA, however, (+)-BA was only detected from the catalytic products of five WvAATs (WvBAHD8, WvBAHD14, WvBAHD39, WvBAHD40, and WvBAHD3) when (+)-borneol was used as substrate (Figures 5A,B); we thus concluded that these five WvAATs capable to catalyze both (−)-borneol and (+)-borneol have no strict substrate specificity for different optical isomers of borneol, while WvBAHD28, WvBAHD50, and WvBAHD53 have substrate specificity for (−)-borneol. Among these five WvAATs capable to catalyze both (−)-borneol and (+)-borneol, WvBAHD8, WvBAHD39, WvBAHD40, and WvBAHD3 can also catalyze isoborneol to produce IBA (Figure 5C). In addition to borneol-type substrate, more diverse substrates were used to perform the enzyme assay, and none of these WvAATs could react with tertiary alcohols, such as (−)-menthol, linalool, or nerolidol, to form relative acetate products, similar to that of LiAAT3 and LiAAT4 (Sarker and Mahmoud, 2015). Since all these WvAATs can convert at least one type of borneol to BA, demonstrating that they have the BAT activity, therefore they are renamed as WvBAT1–8, respectively (Table 1 and Supplementary Table 7). To compare the catalytic activity for different optical isomers of borneol substrate, the product peak areas of the five WvBATs which can catalyze both (−)-borneol and (+)-borneol, reacting with the same amount of each substrate, were analyzed. The results indicated that these WvBATs had the better catalytic ability for (−)-borneol than for (+)-borneol; notably, WvBAT3 (WvBAHD39) exhibited the best catalytic efficiency for (+)-borneol than the other WvBATs, followed by WvBAT4 (WvBAHD40) (Figure 5D). In addition, WvBAT3 is the only enzyme capable to acetylate the α-terpineol to synthesize α-terpinyl acetate (Table 1 and Supplementary Figure 5).

As we mentioned above, most of the WvBATs present a better catalytic ability for (−)-borneol in vitro. (−)-Borneol and (+)-borneol are optical isomers and both of them are natural metabolites in plants, however, the optical configuration of endogenous borneol in the seeds of W. villosa and the WvBPPS-catalyzed product hasn’t been identified with the chiral column to date (Wang et al., 2018). Therefore, we used GC-MS with CycloSil-B column to identify the chiral configuration of borneol in the seeds of W. villosa, and that of the product in the enzymatic reaction of WvBPPS. The results showed that no (−)-borneol was detected but a significant amount of (+)-borneol and a small amount of isoborneol were detected in the mature (90-DAF) seeds of W. villosa (Figure 6A); this result was also observed in the volatile extract from seeds at different developmental stages (30-DAF to 75-DAF), and consistent with the observation that (+)-borneol was the WvBPPS-catalyzed product with phosphatase treatment (Figure 6B). Furthermore, the content of (+)-BA rises with the increasing of (+)-borneol throughout the seed development, peaking at 90 DAF; (+)-borneol has a substantially larger content than isoborneol, which matches the data that (+)-BA has a significantly higher amount than IBA (Figure 6C). These results suggest that (+)-borneol is the major endogenous borneol-type substrate in the seeds of W. villosa.

To analyze the temporal and spatial expression patterns of WvBATs, qRT-PCR was performed using the leaf, rhizome, flower, and 60-DAF fruit (separated into pericarp and seeds), and the seeds at five developmental stages. The result demonstrated that the relative expression levels of WvBATs were basically consistent with the TMP (transcripts per million) expression value from the RNA-sequencing (RNA-seq) data, except for WvBAT5; WvBAT5 was actually expressed higher in rhizome than in the seeds. The expressional patterns of WvBAT3, WvBAT4, and WvBAT6 are seed-specific, similar to that of WvBPPS. Although WvBAT6 was expressed specifically in the seeds, the enzyme it encodes cannot catalyze (+)-borneol and isoborneol, indicating that it might participate in the biosynthesis of other acetates in the seeds. WvBAT1 was expressed specifically in the flower, suggesting that it might be involved in the acetates biosynthesis in the flower. Furthermore, WvBAT7 was expressed in all the organs with the highest level in the rhizome, indicating that it might be involved in the BA synthesis in the rhizome, which contains low levels of BA (Wang et al., 2018; Figure 7A).

As we mentioned above, (+)-borneol is the major endogenous substrate in the seeds, and BA accumulates in the seeds with an obvious occurrence at 45-DAF (Figure 6C); only four WvBATs (WvBAT2, WvBAT3, WvBAT4, and WvBAT7) are capable to catalyze (+)-borneol and expressed in the seeds as well. Therefore, to determine the major BAT(s) responsible for the BA biosynthesis, we investigated their expression patterns in the seeds at different developmental stages and compared their expression levels in 45-DAF and 60-DAF seeds. Of which, the expression levels of WvBAT2 and WvBAT7 peaked at 45 DAF, while their expression levels were much lower than WvBAT4 (Figures 7B,C). Notably, both WvBAT3 and WvBAT4, the seed-specific expressed genes, were not expressed at 30-DAF, but started to be highly expressed at the 45-DAF stage, similar to that the highest expression level of WvBPPS presented in 45-DAF, in accord with the initial accumulation levels of (+)-borneol and (+)-BA in the seeds (Wang et al., 2018; Figure 7B). In addition, WvBAT4 was the highest-expressed gene in the seeds at 45-DAF and 60-DAF, and the expression level of WvBAT3 in 60-DAF seeds is also higher than the other two WvBATs (Figure 7C). Considering WvBAT3 (WvBAHD39) is the enzyme presenting the best catalytic efficiency on (+)-borneol (Figure 5D), we speculated that both WvBAT3 and WvBAT4 are the key enzymes responsible for the synthesis of BA. In addition, IBA initial accumulation in the seeds also increased from 45-DAF (Figure 6C), implying that the WvBATs (WvBAT3, WvBAT4, and WvBAT7) enable to catalyze (+)-isoborneol, were also involved in the IBA synthesis in the seeds of W. villosa.

To ascertain the potential biological roles of WvBATs in W. villosa, the element analysis was performed on the approximate area of the promoter of WvBAT1–8 genes. In total, 308 cis-acting elements were identified, and they were grouped into three categories, phytohormone responsiveness, stress responsiveness, and plant growth and development (Figures 8A,B). The greatest proportion of cis-acting regulatory elements related to phytohormone response was ABRE elements that were associated with abscisic acid (ABA, 22%), followed by MYC (21%), TGACG-motif (10%), and CGTCA-motif (10%); the latter three elements were related to methyl jasmonate (MeJA). Nearly half of all the cis-acting elements were associated with stress responsiveness (161/304), of which the most common 4 cis-acting elements were the box 4 (14%), G-box (11%), GT1-motif (9%), and MYB (9%), the first three were associated with responsiveness to light, and the last one was related to water stress (Figure 8C). These results suggested that WvBATs might be induced or suppressed by MeJA and ABA, and involved in plant responses to a variety of abiotic stressors.

We previously observed that promoters of WvBPPS featured a seed-specific expression element GCN4-motif, which was consistent with the main accumulation of borneol in the seeds (Zhao et al., 2021a). The seed-expression cis-acting elements include the GCN4 motif involved in the endosperm cis-elements (Onodera et al., 2001), RY-element, and AT-rich element involved in seed-specific regulation (Bäumlein et al., 1992; Ellerström et al., 1996), and G-box (Izawa et al., 1993, 1994). Expectedly, excluding WvBAT5, these WvBAT promoters all contained cis-acting elements related to seed expression; for example, the WvBAT7 promoter contained two GCN4 motifs, and the promoters of WvBAT1, WvBAT2, WvBAT4, WvBAT6, and WvBAT7 contained G-box.

There are relatively few studies devoted to the HMM of the BAHD super acyltransferase family for screening BAHD members in plants. PF02458, the HMM of the transferase family, which belong to CL0149 with a characteristic HXXXD motif, was obtained by comparing the HMM in the Pfam database of reported members of the BAHD family, which was consistent with the results of previous studies (Liu et al., 2020; Kumar et al., 2021). We identified 61 WvBAHDs by PF02458 from W. villosa genomic data, featured “HXXXD” and “DFGWG,” which is consistent with the characteristic of the BAHD family.

In our investigation, besides PF02458, the other Pfam HMM PF07247 of the AATase family, which belongs to CL0149 as well, was also used to identify the BAHD family members of W. villosa. Accordingly, three WvBAHDs (WvBAHD3, WvBAHD52, and WvBAHD53) annotated as AATase had been screened out. However, there was no reported HMM profile of BAHD members similar to PF07247. Actually, “AATase” usually describes the AAT of bacteria, metazoan, and fungi, such as yeast Saccharomyces and Kluyveromyces (Zhu et al., 2015; Reyes-Sánchez et al., 2019). To further confirm whether these three AATase-WvBAHDs belong to the BAHD acyltransferase family, we constructed another phylogenetic tree using 64 WvBAHDs and the identified BAHDs from other species (Supplementary Table 8) based on the cluster system from D’Auria, which clusters BAHD acyltransferases into six major groups (D’Auria, 2006; Sarker and Mahmoud, 2015). Unexpectedly, these three AATase-WvBAHDs were all clustered into the clade V, which consists mostly of AATs involved in volatile ester biosynthesis (Supplementary Figure 6); however, they are not clustered closely with the other identified AATs from plants, which might be due to amino acid residue substitutions in their DFGWG motif, one of the conserved motifs of BAHD acyltransferase family. Notably, WvBAHD3 and WvBAHD53 do have the activity of AAT, and both can acetylate borneol, geraniol, and nerol (Table 1). Therefore, this work has verified “PF07247/AATase” is feasible for screening BAHD superfamily members in the plant, thus, the BAHD family would be enriched by adding new AAT members.

Bornyl acetate is the aromatic monoterpene ester with a wide range of pharmacological and biological activities, but only a few plants accumulate a significant amount of BA. BPPS, the enzyme catalyzing the synthesis of borneol precursor, the first step on BA downstream biosynthesis pathway of BA, has been identified from other plants, including Salvia officinalis (SoBPPS), Lippia dulcis (LdBPPS), Lavandula angustifolia (LaBPPS), and Cinnamomum burmannii (CbBPPS) (Wise et al., 1998; Despinasse et al., 2017; Hurd et al., 2017; Ma et al., 2021); except for Lippia dulcis, these plants all contain BA (Al-Dhubiab, 2012; Arceusz et al., 2013), however, the gene encoding BAT, catalyzing the acetylation of borneol, has not yet been reported. In this study, based on the identification of the BAHD family, we cloned and characterized eight WvBATs from W. villosa. This is the first report on the BAT responsible for the last biosynthetic step of BA. These WvBATs can acetylate at least one type of borneol to synthesize BA, and they are substrate-promiscuous enzymes, similar to other AATs. AAT is the substrate-promiscuous enzyme, which can acetylate different types of alcohols and acyl-CoA to synthesize aromatic volatile ester compounds in plants (Beekwilder et al., 2004; Souleyre et al., 2005). However, the type and content of acylation products depend on the presence and content of endogenous substrates in plants. For example, the type of acetate ester synthesized in mint depends not on the specificity of AATs, but on the availability of the terpene alcohol in mint oil (Croteau and Hooper, 1978). With regard to W. villosa, the volatile oil of the seeds contains high levels of (+)-borneol, much lower levels of isoborneol, and only trace levels of α-terpineol, accordingly, (+)-borneol is the predominant endogenous substrate for WvBATs in the seeds. Two WvBATs, WvBAT3 and WvBAT4, capable of acetylating (+)-borneol with higher efficiency, and expressed in the seeds with positive correlation to the BA accumulation, might be the key enzymes collaborate to synthesize BA in the seeds of W. villosa. This speculation is consistent with the previous reports that the plants contain several AATs which might collaborate in the synthesis of a single ester (Cumplido-Laso et al., 2012; Sarker and Mahmoud, 2015).

In summary, based on the biochemical characterization of WvBATs in this study and previous literature (Wise et al., 1998; Despinasse et al., 2017; Wang et al., 2018), we illuminated the downstream biosynthetic pathway of BA in W. villosa (Figure 9). This work will contribute to a better understanding of the biosynthesis of BA and other acetates in W. villosa, and lay the foundation for further studies on BA biosynthesis and metabolism engineering of BA in other plants or heterologous hosts.