Introduction

Plant essential oils (PEOs) constitute a substantial and important part of commercial odorants1. Their ingredients have been extensively studied, and many of them are comprised of sesquiterpenes2, e.g., sandalwood oil. In addition to being one of the world’s most highly prized natural perfumes1,2, sandalwood oil has been in clinical trials for the treatment of skin disorders3, and also possesses other bioactivities, such as anticancer4,5, antihyperglycemic6, antioxidant6, and neuroprotection activities7. The major components of sandalwood oil are santalenes (mainly including α-santalene, β-santalene, epi-β-santalene, and exo-α-bergamotene) and santalols (mainly including Z-α-santalol, Z-β-santalol, Z-exo-α-bergamotol, and Z-epi-β-santalol) (Supplementary Fig. 1)8. Among these compounds, Z-α-santalol and Z-β-santalol are the most critical components contributing to the bioactivities and fragrance of sandalwood oil5,7,9,10. The International Organization for Standardization has issued the standard (ISO 3518:2002) for Santalum album L. oil in which the contents of Z-α-santalol and Z-β-santalol must fall in the 41−55% and 16−24% range, respectively11. Although the oil can be extracted from several Santalum species, the most renowned is Indian S. album oil. Due to overexploitation, Indian S. album trees are highly endangered and have been listed as Vulnerable by the International Union for the Conservation of Nature. Heterologous biosynthesis in engineered microbial hosts is a promising alternative for manufacturing plant natural products on a large scale1,12. In S. album, the biosynthetic pathway of santalols has been decoded, harboring a santalene synthase (SaSSy) and ten cytochrome P450 enzymes (CYPs) responsible for oxidizing santalenes to santalols (Supplementary Fig. 1)13,14,15. Besides, SauSSy from S. austrocaledonicum13, SspiSSy from S. spicatum13 and CiCaSSy from Cinnamomum camphora16 were characterized to possess the product distributions similar to that of SaSSy, and SanSyn from Clausena lansium was found to be a product-specific enzyme, with α-santalene as its predominant product (Fig. 1)17. The metabolic engineering efforts have led to construction of α-santalene-producing microbial platforms18,19,20,21. Recently, we built an engineered Saccharomyces cerevisiae (yeast) which can produce both santalenes and santalols, with 35.7% Z-α-santalol and 17.8% Z-β-santalol which is very close to the standard of ISO 3518:200222.

Fig. 1: The product-promiscuous and product-specific santalene synthases.
figure 1

SaSSy, SauSSy, SspiSSy, and CiCaSSy produce the multiple santalene products, and SanSyn predominantly produce α-santalene.

Full size image

As mentioned above, the component ratios of PEOs determine their qualities. Properly controlling the component ratios is challenging in construction of the heterologous biosynthetic systems of PEOs. Enzyme engineering has been proven to be a powerful approach to optimize or alternate enzymatic properties such as efficiency, stability, substrate selectivity and product profile23. Understanding catalytic mechanisms of enzymes is commonly believed to be the fundamental basis for redesigning the desired function. In the previous studies, we explored the mechanism which leads to catalytic promiscuity and fidelity of two sesquiterpene synthases TEAS (Nicotiana tabacum 5-epi-aristolochene synthase) and ATAS (Aspergillus terreus aristolochene synthase) by the combination of quantum mechanics/molecular mechanics (QM/MM) and molecular dynamics (MD)24,25,26,27. In this study, owing to such approach, the reaction energy profiles of both SaSSy and SanSyn are mapped out, and the interactions between the enzymes and the carbocation intermediates are explored, which lead to identification of a residue F441 in SanSyn imbuing a unique structural plasticity. And the subsequent site-saturation mutagenesis leads to generation of a mutant SanSynF441V which produces 57.2% α-santalene, 28.6% β-santalene, 6.7% epi-β-santalene and 7.6% exo-α-bergamotene, representing a desirable product profile. Meanwhile, a yeast chassis strain SZ16 (Supplementary Data 1) is obtained by optimization of the mevalonate (MVA) pathway and acetyl-CoA synthesis, knockout of the genes encoding the enzymes involved in both farnesyl diphosphate (FPP) consumption and Z-α-santalol transformation, and deletion of the yeast transcriptional repressor ROX1. Then, SanSynF441V is used to construct the santalene/santalol-producing yeast from SZ16 (Supplementary Data 1). The resulting SZ24 (Supplementary Data 1) totally produces 704.2 mg/L santalenes and santalols, with 43.4% Z-α-santalol, 22% Z-β-santalol and only 6.7% Z-exo-α-bergamotol which well matches the standard of ISO 3518:2002.

Results

QM/MM simulations of SaSSy and SanSyn catalysis

CYP736A167 exhibits the approximately equal substrate preference to α-santalene, β-santalene, epi-β-santalene and exo-α-bergamotene, and hence the component profile of sandalwood oil is shaped mainly by SaSSy22. As mentioned above, Z-α-santalol and Z-β-santalol are the most important components which contribute to the odor and bioactivity of sandalwood oil, and used for quality control in ISO 3518:2002. However, SaSSy produces much exo-α-bergamotene, resulting in the large ratios of exo-α-bergamotene and Z-exo-α-bergamotol in the fermentation oil, which are close to those of α-santalene and Z-α-santalol and even larger than those of β-santalene and Z-β-santalol22. In this study, we sought to elevate the ratios of Z-α-santalol and Z-β-santalol through engineering santalene synthase to attenuate the production of exo-α-bergamotene and Z-exo-α-bergamotol. To this end, we first explored the catalytic mechanisms of SaSSy and SanSyn by QM/MM calculation.

SaSSy produces multiple bridged ring products, as shown in Fig. 1. Among these compounds, α-santalene and β-santalene could be directly derived from (6S)-bisabolyl cation (A state) through cyclization, alkyl transfer and deprotonation after initial 1,6-closure of farnesyl diphosphate, while deprotonation from B state leads to exo-α-bergamotene (Fig. 2a). To date, the structures of both holo-SaSSy and holo-SanSyn are still absent, although the apo-SaSSy structure (PDB ID: 5ZZJ) has been reported28. Thus, the SaSSy and SanSyn models were predicted using Alphafold2 herein (Supplementary Fig. 2)29, and A state was considered in the active sites (see Figs. 2 and 3 and Supplementary Method 1). As shown in Fig. 2b, the conformation of A state is preorganized for further cyclization with the homoprenyl group surrounded by the aromatic residues (e.g., F545 and F424) and the cyclohexene ring on the other side of the pocket in SaSSy. The 7,2(3)-cyclization of A state conquers 4.0 kcal/mol barrier to yield a non-classical carbocation, namely B state (Fig. 2a, c). The subsequent alkyl transfer leads to C state with notable heat release (Fig. 2a, c). As shown in Fig. 2d, B state is stabilized by the aromatic residues (F424 and F545), and the subsequent alkyl transfer from B to C state has to overcome a barrier of 8.3 kcal/mol (Fig. 2c), indicating that B state has a lifetime to a certain extent and the intrinsic conformational dynamics allow it to produce exo-α-bergamotene. After F424 and F545 were respectively replaced by alanine, the catalytic efficiency of two mutants SaSSyF545A and SaSSyF424A dramatically decreased (Supplementary Figs. 3a, b and 4a, b), supporting their substantial contribution to configuring of the intermediate conformation. And we found that T318 in SaSSy might appropriately serve as the major general base for deprotonation at C-4 (Fig. 2d) to yield α-santalene across a barrier of 9.3 kcal/mol (Fig. 2c) which indicates that the deep energy well of C state confers a considerable lifetime to it, relatively longer than B state. Besides, the unoccupied space of the active pocket is enlarged as the bridged ring formation at C state (Fig. 2d), further increasing the conformational dynamics of the carbocation intermediate at C state. As a result, deprotonation from C-13 to yield β-santalene is also feasible by the help of T318 (Fig. 2a, d). This was supported by the T318A mutation which resulted in almost complete loss of the ability to produce α-santalene and β-santalene (Supplementary Fig. 3c and 4c). Regarding to the minor product epi-β-santalene, it could be directly derived only from (6R)-bisabolyl cation but not A state, i.e., (6S)-bisabolyl cation, if going through the reaction pathway similar to synthesis of β-santalene (Fig. 2a and Supplementary Fig. 5). However, it was found that the (6R)-bisabolyl cation conformation is not likely existed in SaSSy, since none of the corresponding reasonable conformations could be well maintained in QM/MM MD simulations. In addition, the other compounds (such as 7-epi-α-santalene) possibly derived from the R-configurational pathway are absent in the product mixture of SaSSy (Supplementary Fig. 5), which further suggests the absence of (6R)-bisabolyl cation in the reaction pathway catalyzed by SaSSy. Alternatively, it was reported that epi-β-santalene could derive from cleavage of the cyclopropane ring of α-santalene (Supplementary Fig. 6)30. Our calculation proved that the cyclopropane ring opening of α-santalene is feasible (only 4.3 kcal/mol barrier) (Fig. 2c), which could be triggered through attracting a proton from the T318. This leads to generation of D state, and the subsequent deprotonation from D state results in the production of epi-α-santalene (Supplementary Figs. 6 and 7).

Fig. 2: QM/MM MD simulation of the reaction pathway catalyzed by SaSSy.
figure 2

a The reaction pathway from (6S)-bisabolyl cation (A state) to the multiple products. b A state in the active site of SaSSy model. c Relative energy profiles from A state to α-santalene. d Representative structures of QM/MM simulations for A, B, and C states are shown.

Full size image
Fig. 3: Comparative analysis of the active cavities of SaSSy, SanSyn, and SanSynF441V.
figure 3

a Comparisons of the A state conformations in SaSSy (yellow) and SanSyn (green). b The C state conformations in SanSyn and SanSynF441V. The space released by F441V mutation is indicated by the red dashed line.

Full size image

Compared with SaSSy, SanSyn dominantly produces α-santalene, with only a trace of the other relevant products (Fig. 4), suggesting that the conversion from B state to C state and the followed H4 leaving to the C3-C4 bond formation is dominant. As shown in Fig. 3a, the conformation of A state in SanSyn is more compact than that in SaSSy, which means the unoccupied space for the intermediate conformational dynamics is limited in SanSyn. The homoprenyl group of A state is well oriented for hyperconjugation as an aligned conformer to stabilize the carbocation in SanSyn, while more flexible as an extended conformer in SaSSy. The effect of hyperconjugation donation on the relative stability of B state has been previously discussed by the extensive QM calculations31. Herein the QM/MM calculations from A to C states in SanSyn (Supplementary Fig. 8) validated such hyperconjugation effect that the potential energy well of B state in SanSyn (depth of 2.2 kcal/mol) is shallower than that in SaSSy (depth of 4.3 kcal/mol), indicating a shorter lifetime of B state in SanSyn. Meanwhile, the limited unoccupied space configures the compact conformation of B state preferable to the conversion to C state, thus largely avoiding the production of exo-α-bergamotene in SanSyn. Subsequently, T298 corresponding to T318 in SaSSy (Supplementary Fig. 9) could serve as the potential major general base to achieve deprotonation at C-4 for the production of α-santalene. This was supported by the mutagenesis experiment in which the replacement of T298 with alanine resulted in a significant decrease in the catalytic efficiency (Supplementary Figs. 3i and 4e). Comparing Sansyn and SaSSy, we found that the lack of unoccupied space in Sansyn would largely restrict the intermediate conformational dynamics, which is also a key factor for its high fidelity. We, therefore, surmised that the replacement of the residue containing a large side chain with a proper small one in the cavity to vacate more space for the intermediate (especially C state) conformational dynamics might achieve our end. Based on the structure comparisons between SanSyn and SaSSy and our understanding of the reaction pathways, a single amino acid variant F441/S459 (SanSyn in front) attracted our attention (Fig. 3a). In SanSyn, F441 appears to restrict the conformational space of the intermediates (Fig. 3b), thereby likely playing a critical role in guiding the intermediate conformation.

Fig. 4: GC-MS analysis of the products of SaSSy, SanSyn and SanSynF441V.
figure 4

SanSynF441V can produce both α-santalene and β-santalene.

Full size image

Creation of SanSynF441V with a desirable product profile

Given that F441 of SanSyn corresponds to S459 of SaSSy, we conducted site-directed mutagenesis to obtain SanSynF441S. The yeast strain WL0722 expressing SanSynF441S produced both α-santalene and β-santalene (Supplementary Fig. 3d), which encouraged us to conduct site-saturation mutagenesis of F441. The results showed that the replacement of F441 with the residues (including threonine, valine, leucine and isoleucine) similar to serine resulted in the production of both α-santalene and β-santalene (Fig. 4 and Supplementary Figs. 3e−g and 4d). Among these mutants, SanSynF441V exhibited the highest efficiency and a desirable product profile, 57.2% α-santalene, 28.6% β-santalene, 6.7% epi-β-santalene and 7.6% exo-α-bergamotene (Fig. 4). QM/MM MD simulations of C state in SanSyn and SanSynF441V were performed. As shown in Fig. 3b, C state adopts a compact conformation in SanSyn. While in SanSynF441V, the homoprenyl group occupies the additional space arising from F441V mutation and adopts a more extended conformation. As observed in QM/MM MD and traditional MD simulations, the conformational flexibility of C state in SanSynF441V is more notable with larger fluctuations of the dihedral C6–C7–C8–C9 (Supplementary Figs. 10 and 11) compared to that in the wildtype SanSyn, which brings about the rotatability of the bridged ring moiety and thus deprotonation could occur at more than one position (i.e., C-4 and C-13). Besides, the long lifetime of C state would further enhance the conformational dynamics of the carbocation intermediate at this state. As a result, T298 could eliminate both H4 and H13 of C state for the production of α-santalene and β-santalene, respectively, as we surmised above (Fig. 3b and Supplementary Fig. 12). Indeed, the replacement of T298 with alanine in SanSynF441V led to substantial loss of the ability to produce α-santalene and β-santalene (Supplementary Figs. 3h and 4f). These results document that the structural features, especially the cavity space, substantially contribute to configure preorganized substrate folding modes and give rise to various intermediate conformations with different dynamic behavior25,32.

Enhancement of the santalene and santalol titers by metabolic engineering

In order to easily quantify the titers, the product-specific enzyme SanSyn was utilized in this part. We put forth the first effort to screen the initial yeast host strains including S. cerevisiae BY4741, CEN.PK2-1C and CEN.PK2-1D (Supplementary Data 1). These three strains were transformed with pSZ1 for expressing tHMG1 (the truncated 3-hydroxy-3-methylglutaryl coenzyme A reductase, a limited enzyme in terpene synthesis)33 and SanSyn and pSZ2 for expressing CYP736A167 and SaCPR2 (a cytochrome P450 reductase from S. album) (Supplementary Data 2 and 3)22. The highest total α-santalene and Z-α-santalol titer (5.4 mg/L) was observed when using S. cerevisiae CEN.PK2-1D (SZ7, Supplementary Data 1), ~1.5-folds relative to those when using S. cerevisiae BY4741 (SZ1, Supplementary Data 1) and CEN.PK2-1C (SZ4, Supplementary Data 1), respectively (Fig. 5a). Because cytochrome P450 reductases play important roles in the functions of cytochrome P450 enzymes and significantly influence their efficiencies34,35, we tested the efficiency of CYP736A167 assisted by two other cytochrome P450 reductases ATR1 (from Arabidopsis thaliana)36 and CrCPR (from Catharanthus roseus)37, respectively (Supplementary Data 1 and 2). However, the better efficiency was not achieved (Fig. 5a). Thus, SaCPR2 was still used in the following modification.

Fig. 5: Metabolic engineering of yeast for the production of santalenes and santalols.
figure 5

a The production of α-santalene and Z-α-santalol by expression of SanSyn and CYP736A167 assisted with the different CPRs in the yeast hosts (SZ1-SZ9, see Supplementary Data 1), and enhanced the production of α-santalene and Z-α-santalol by knockout of OYE2, OYE3, ATF1, and ATF2 (SZ13, see Supplementary Data 1). b GC-MS analysis of α-santalene, Z-α-santalol, dihydro-α-santalol (by-products 1) and acetyl dihydro-α-santalol (by-product 2) before and after knockout of OYE2, OYE3, ATF1, and ATF2 (SZ7 and SZ13, see Supplementary Data 1). c Enhanced the production of α-santalene by overexpression of IDI1, UPC2-1, ADH2, ALD6, and ACSL641P, and knockdown of ERG9 (SZ18−SZ20, see Supplementary Data 1). All data represent the mean of n = 3 biologically independent samples and error bars show standard deviation. Statistical analysis was performed by using two-tailed and two-sample unequal variance t test (*P < 0.05, **P < 0.01). Source data are provided as a Source Data file.

Full size image

In addition to α-santalene and Z-α-santalol, two unknown peaks were detected by GC-MS from the culture of SZ7 (Fig. 5b). The titer of by-product 1 was 4.3 mg/L which was even higher than the Z-α-santalol titer (Fig. 5b). As the MS spectral profiles of these two by-products are very similar to that of Z-α-santalol (Supplementary Fig. 13), we surmised that they derived from Z-α-santalol through the reactions catalyzed by yeast endogenous enzymes. From 12 L SZ7 culture, these two by-products were isolated and their structures were respectively determined as dihydro-α-santalol and acetyl dihydro-α-santalol by NMR analysis (Supplementary Figs. 14–17)38. It has been reported that OYE2 and OYE3 (two yeast old yellow enzymes) can reduce geraniol into citronellol39,40,41 and ATF1 and ATF2 (two yeast alcohol acetyltransferases) can catalyze acetylation of geraniol39,40,41. We assumed that these four endogenous enzymes may also play roles in derivatization of Z-α-santalol. Indeed, knockout of OYE2 and OYE3 by CRISPR/Cas9 (SZ12, Supplementary Data 1) resulted in a 27.9% decrease of the dihydro-α-santalol titer to 3.1 mg/L and a 23.1% increase of the Z-α-santalol titer to 3.8 mg/L compared with those of SZ7 (Fig. 5a and Supplementary Fig. 18). This indicates the presence of the other endogenous enzymes which can catalyze hydrogenation of Z-α-santalol to dihydro-α-santalol. Then, all of OYE2, OYE3, ATF1, and ATF2 in S. cerevisiae CEN.PK2-1D were knocked out to afford SZ11 (Supplementary Data 1). After transforming SZ11 with pSZ1 and pSZ2, the Z-α-santalol titer in the culture of the resulting strain SZ13 was 4.2 mg/L, a 35.5% increase relative to that of SZ7 (Fig. 5a), and the by-product 2 acetyl dihydro-α-santalol was undetectable (Fig. 5b).

Subsequently, the terpene biosynthetic pathways in yeast were engineered. Only SanSyn was expressed in the engineered strains in this part to simplify the experimental process, and thus α-santalene served as the sole reference compound for weighing the effect of each modification. The mevalonate (MVA) pathway was firstly optimized by integration of one copy of IDI1 encoding isopentenyl diphosphate isomerase42 and UPC2-1 encoding a mutant of the transcription factor UPC2 which can enhance expression of the genes of yeast MVA pathway22. Because DPP1 (a yeast diacylglycerol pyrophosphate phosphatase) can dephosphorylate FPP into farnesol18, we integrated one copy of IDI1 and UPC2-1 into DPP1 locus for both optimization of MVA pathway and attenuation of FPP consumption through dephosphorylation. After transformation with pSZ5 harboring SanSyn and tHMG1 (Supplementary Data 2 and 3), the resulting SZ18 produced 3.1 mg/L α-santalene, a 73% increase relative to that of SZ17 generated by introducing pSZ5 into SZ11 (Fig. 5c and Supplementary Data 1). In order to increase the supply of acetyl-CoA (the precursor of MVA pathway) and further decrease FPP consumption through dephosphorylation, we enhanced conversion from ethanol to acetyl-CoA by integrating ADH2 (a yeast alcohol dehydrogenase), ALD6 (a yeast acetaldehyde dehydrogenase) and ACSL641P (the L641P variant of a Salmonella enterica acetyl-CoA synthetase)20 into the locus of LPP1, encoding a yeast lipid phosphate phosphatase which, like DPP1, can also dephosphorylate FPP18. The α-santalene titer of the resulting stain SZ19 (Supplementary Data 1) increased to 4.8 mg/L (Fig. 5c). To further redirect more FPP to α-santalene biosynthesis, ERG9 (yeast squalene synthase) was downregulated to attenuate FPP consumption in triterpenoid/steroid biosynthesis by replacing its native promotor with a glucose-induced promotor PHXT1, which led to an α-santalene titer of 13.9 mg/L in the resulting SZ20 (Fig. 5c). Thus, the corresponding chassis strain SZ16 (Supplementary Data 1) with the optimized MVA and acetyl-CoA synthesis pathways and ERG9 knockdown was used to construct the santalene/santalol-producing strains.

Because expression of biosynthetic genes by the integration mode often give rise to higher and more stable yields of the corresponding compounds43, we integrated one copy of tHMG1, CYP736A167 and SaCPR2 and two copies of SanSynF441V into SZ16 to afford SZ21 (Supplementary Data 1). In SZ21 culture, the titers of santalenes (α-santalene, β-santalene, epi-β-santalene and exo-α-bergamotene) and santalols (Z-α-santalol, Z-β-santalol, Z-exo-α-bergamotol, and Z-epi-β-santalol) were 1.3 mg/L and 4.5 mg/L, respectively (Fig. 6a), when fermented under the optimal condition with 0.2% w/w glucose and 1.8% w/w galactose. The previous studies reported that deletion of ROX1 (a yeast transcriptional repressor) elevated the expression levels of terpenoid and steroid biosynthetic genes43,44. Thus, we integrated an additional copy of tHMG1, SanSynF441V and CYP736A167 into ROX1 locus of SZ21 to obtain SZ22 (Supplementary Data 1). SZ22 produced 6.6 mg/L santalenes and 14.3 mg/L santalols in the optimal condition, respectively corresponding to 5.1- and 3.2-fold increase relative to those of SZ21 (Fig. 6b). The ratios of Z-α-santalol and Z-β-santalol in the santalene and santalol mixture produced by SZ22 were 48.1% and 20.2%, respectively, which well matches the ISO 3518:2002 standard.

Fig. 6: The titers of santalenes and santalols in SZ21 and SZ22.
figure 6

The titers of santalenes and santalols in SZ21 (with two copies of SanSynF441V and one copy of tHMG1 and CYP736A167) (a) and SZ22 (with three copies of SanSynF441V and two copies of tHMG1 and CYP736A167) (b) when fermented with the different ratios of glucose and galactose. All data represent the mean of n = 3 biologically independent samples and error bars show the standard deviation. Statistical analysis was performed by using two-tailed and two-sample unequal variance t test (*P < 0.05, **P < 0.01). Source data are provided as a Source Data file.

Full size image

Fed-batch fermentation and tuning of the copy numbers of SanSyn F441V and CYP736A167

Because galactose is too expensive for industry-scale fermentation, we sought to use the cheaper carbon source. Given that the route from ethanol to acetyl-CoA in SZ22 was optimized, the mixture of glucose and ethanol was utilized in the fed-batch fermentation. In the batch media, glucose (25 g/L) was the sole carbon source. When glucose was completely consumed, the significant OD increase was observed. And at this point the feeding phase started. To optimize the feeding media, three different ratios of glucose and ethanol were tested, including 350 g/L glucose and 150 g/L ethanol (Supplementary Fig. 19a), 250 g/L glucose and 250 g/L ethanol (Supplementary Fig. 19b), and 150 g/L glucose and 350 g/L ethanol (Fig. 7a). The highest titer (842.7 mg/L santalenes and 1075.7 mg/L santalols) was observed when 150 g/L glucose and 350 g/L ethanol were used (Fig. 7a). The ratios of Z-α-santalol and Z-β-santalol were 29.4% and 15% (Table 1), respectively, indicating that much santalenes were not oxidized into santalols. Therefore, we increased the copy number of CYP736A167 to yield SZ23 and SZ24 (Fig. 7b, Supplementary Fig. 20, and Supplementary Data 1). 43.4% Z-α-santalol and 22.0% Z-β-santalol were observed in the oil produced by SZ24 (Table 1) which well matches the ISO 3518:2002 standard, but the total titer of santalenes and santalols decreased to 704.2 mg/L (Fig. 7b). The titer decrease might result from the biomass decrease, as the OD of SZ24 was approximately 58% of that of SZ22 (Fig. 7). This observation indicated that the high copy number of CYP736A167 could inhibit yeast growth. In the next research, CYP736A167 should be also engineered to increase it catalytic efficiency, and hence the higher conversation of santalenes to santalols would be achieved using the low copy number of the resulting mutant. In addition, intriguing was that the ratios between dihydro-α-santalol and Z-α-santalol in the oils produced by SZ22 and SZ24 were only 1:17 and 1:20, respectively (Supplementary Fig. 21). This indicates that the catalytic capacity of the yeast endogenous enzymes responsible for hydrogenation of Z-α-santalol is limited, and thereby the ratio between dihydro-α-santalol and α-santalol significantly decreased when the Z-α-santalol titer was dramatically elevated.

Fig. 7: Fed-batch fermentation of SZ22 (with three copies of SanSynF441V and two copies of CYP736A167) and SZ24 (with three copies of SanSynF441V and CYP736A167).
figure 7

The santalenes and santalols titers and OD600 of SZ22 (a) and SZ24 (b). All data represent the mean of n = 2 biologically independent samples. Source data are provided as a Source Data file.

Full size image
Table 1 The component ratios of the oils produced by SZ22 and SZ24.
Full size table

Discussion

The qualities of plant natural product mixtures (including PEOs, herbal medicines, etc.) are determined by their component ratios. Accordingly, to accomplish biotechnological production of the high-quality products requires not only enhancement of yields not also optimization of component ratios. In the biosynthetic pathways of these compound mixtures, some product-promiscuous enzymes often play critical roles in shaping their component ratios. Clarification of catalytic mechanism of these enzymes would enable optimization of product distribution by enzymatic engineering45,46. In sandalwood oil biosynthesis, SaSSy produces a large amount of exo-α-bergamotene, which leads to the ratio of Z-exo-α-bergamotol comparable to those of Z-α-santalol and Z-β-santalol (two most valuable components) in the oil produced by the engineered yeast constructed with SaSSy22. In this study, we established the models of SaSSy and SanSyn using Alphafold2, and simulated their catalytic reaction pathways using QM/MM calculation.

In SaSSy model, only (6S)-bisabolyl cation (A state) (Fig. 2) is likely existed, while (6R)-bisabolyl cation isn’t. And the possible products derived from (6R)-bisabolyl cation are absent. These results indicate that the substrate folding mode in the cavity is well preorganized, which determines the fate of the reaction pathway. It has been reported that aromatic amino acid residues in terpene cyclases can stabilize carbocation intermediates through cation−π interactions47. Indeed, mutagenesis of F545 and F424 resulted in a significant loss of catalytic efficiency (Supplementary Figs. 3a, b and 4a, b).

Comparative analysis of SaSSy and SanSyn showed they respectively harbor a bigger and a smaller cavities (Fig. 3b). Accordingly, the intermediates in SaSSy adopt the flexible conformations with different dynamic behavior, which allow a versatile reaction pathway for the production of multiple products (Fig. 1). Oppositely, the limited cavity space in SanSyn confers the compact conformations to the intermediates, which makes it a product-specific enzyme. Therefore, the residues highly related to the cavity space may substantially influence the conformation dynamic of the intermediates, and thereby mutagenesis of these residues could alter the product distribution. F441 in SanSyn was considered to be the residue of this type base on QM/MM MD simulations (Fig. 3). The additional space released by F441V mutation was observed (Fig. 3b), and the resulting SanSynF441V exhibited a desirable product profile, 57.2% α-santalene, 28.6% β-santalene, 6.7% epi-β-santalene and 7.6% exo-α-bergamotene. Meanwhile, metabolic engineering was performed, which led to the establishment of the chassis strain SZ16 by optimization of MVA pathway and acetyl-CoA synthesis, knockout of the genes encoding the enzymes involved in both FPP consumption and Z-α-santalol transformation, and deletion of the yeast transcriptional repressor ROX1. Next, SanSynF441V was used for construction of the santalene/santalol-producing yeast strains (Supplementary Fig. 22). After tuning of the copy numbers of SanSynF441V and CYP736A167, SZ24 totally produced 704.2 mg/L santalenes and santalols by a fed-batch fermentation, with 43.4% Z-α-santalol and 22.0% Z-β-santalol which well matches the ISO 3518:2002 standard (Table 1).

The active side chains of the amino acids locating at the cavity of terpene cyclases often serve as general bases for deprotonation to produce olefins27. Herein, QM/MM MD simulations and site-directed mutagenesis suggested that T318 in SaSSy and T298 in SanSyn are the major bases for deprotonation at C-4 and C-13 of C state (Figs. 2d, 3b and Supplementary Figs. 3c, i and 4c, e). This indicates that the residues serving as the general bases in SaSSy and SanSyn are highly conserved although these two enzymes only share a 33% sequence identity (Supplementary Fig. 9).

This study documents that multiscale simulations substantially facilitate the identification of key amino acid residues associated with conformational dynamics of intermediates in key enzymes, and the resulting findings can lead to redesign of the product distribution. Such enzymatic engineering approach, in association with metabolic engineering, represents a general strategy for construction of microbial platforms of not only PEOs but also other valuable natural product mixtures.

Methods

Computational details

All the computational details, including system setup, classical MD simulations and QM/MM MD simulations, are included in Supplementary Method 1. The PDB files of SaSSy and SanSyn computational models including predicted models and the important states of QM/MM simulations are provided in Supplementary Data 4.

Strains and media

The S. cerevisiae strains BY4741, CEN.PK2-1D, and CEN.PK2-1C were maintained on YPD plates at 30 °C. Engineered yeast strains were grown in SD media lacking corresponding amino acids. Escherichia coli DH10B was used for genes cloning and plasmids construction, and cultivated at 37 °C in NZY media; E. coli BL21 (DE3) was used for protein expression, and cultivated at 16 °C in LB media.

Mutagenesis of SaSSy and SanSyn, and construction of gene expression cassettes and plasmids

The genomic DNA of S. cerevisiae CEN.PK2-1D was obtained using the DNAiso reagent (Takara Biomedical Technology Co., Ltd., Beijing, China.). Polymerase Chain Reactions (PCR) were performed using Phanta Max Super-Fidelity DNA Polymerase (Vazyme Biotech Co., Ltd.). The yeast endogenous genes, promotors, terminators, and homology arms were amplified from CEN.PK2-1D genomic DNA. The gene expression cassettes were prepared by overlap PCR (Supplementary Fig. 23)48. The plasmids were constructed using ClonExpress II One Step Cloning Kit. The resulting plasmids were confirmed by gene sequencing. To obtain the mutants of SaSSy and SanSyn, SaSSy and SanSyn were respectively ligated into the pEASY Blunt Cloning Vector (TransGen Biotech Co., Ltd.), and the resulting plasmids were used as the templates for mutagenesis by a PCR-based method49. Expression cassette fragments were amplified from the corresponding plasmids. All the plasmids and primers are listed in Supplementary Data 2 and 3.

Strain construction

The expression cassette fragments were amplified from the corresponding plasmids, and subsequently integrated into the yeast chromosome. Yeast transformation was performed by Frozen-EZ Yeast Transformation II Kit (Zymo Research Co., Ltd.). All the strains are listed in Supplementary Data 1.

For CRISPR/Cas9-mediated knockout50, the complete crRNA arrays were synthesized by GenScript Co., Ltd (Nanjing, China), and ligated into the vector pCRCT. The resulting plasmids were introduced into the yeast strains. After growing on plates for 4 days, the colonies were inoculated into 10-mL test tube containing 5 mL SD-URA media. 100 μL of each cell culture was transferred into 5 mL fresh SD-URA media. After incubation for 6 days, the diluted cell culture was plated onto SD-URA media plate, and incubated at 30 °C for 3 days. Then, the single colonies were picked and tested by colony PCR. Finally, the pCRCT plasmids were removed by successive subculturing.

Flask cultivation of yeast strains

The single colonies of each engineered strain were inoculated into 10-mL test tubes containing 5 mL SD media without corresponding nutrition at 30 °C and 230 rpm for 24 h. Then, 1 mL starter culture was inoculated into a 250-mL flask containing 50 mL SD media and grown at 30 °C and 230 rpm for 4 days.

Separation and structural elucidation of by-products

In all, 12 L SZ7 culture was extracted with ethyl acetate (EtOAc). The crude extract was preliminarily purified by silica gel column chromatography, with gradually increasing EtOAc in petroleum ether (PE) as the mobile phase. Two by-products were detected by GC-MS in the fractions eluted by 30:1 and 20:1 PE/EtOAc solutions, respectively. These two fractions were further subjected to high-performance liquid chromatography (HPLC) performed on an Ultimate 3000 instrument (Thermo Scientific, USA) with a YMC-Pack Pro C18 (5 μm, 4.6 × 250 mm) column (the mobile phase: 50–100% methanol in water within 0.00–35.00 min). The structures of the by-products were determined by NMR on TopSpin 2.1 (Bruker).

Fed-batch cultivation of yeast strains

Fed-batch fermentation was performed in a 5-liter fermentor. The agitation ranged from 250 to 800 rpm, the airflow rate ranged from 2 to 3 L/min to keep the dissolved oxygen (DO) above 40%, and the flow rate of feeding media ranged 2 to 14 mL/h. 1 L batch medium contained 15 g (NH4)2SO4, 8 g KH2PO4, 3 g MgSO4, 0.72 g ZnSO4.7H2O, 12 mL vitamin solution, 10 mL trace metals and 25 g glucose. The trace metal solution (per L) contained 15 g EDTA, 10.2 g ZnSO4.7H2O, 0.5 g MnCl2.4H2O, 0.5 g anhydrous CuSO4, 0.86 g CoCl2.6H2O, 0.56 g Na2MoO4.2H2O, 3.84 g CaCl2.2H2O, and 5.12 g FeSO4.7H2O. The vitamin solution contained (per L) contained 0.05 g biotin, 1 g calcium pantothenate, 1 g nicotinic acid, 25 g myoinositol, 1 g thiamine hydrochloride, 1 g pyridoxal hydrochloride, and 0.2 g p-aminobenzoic acid. The feed media (per L) contained 10 mL trace metal solution, 12 mL vitamin solution, 9 g KH2PO4, 2.5 g anhydrous MgSO4, 3.5 g K2SO4, 0.28 g Na2SO4, and total 500 g glucose and ethanol (with various ratios).

Protein expression and purification

pET28a and pCold TF were respectively used to construct the plasmids for expression of the SaSSy and SanSyn mutants in E. coli BL21 (DE3) (Supplementary Data 2)51. Each transformant was grown in LB medium at 37 °C, shaking at 220 rpm, when the OD600 reached 0.4−0.6, 0.4 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) was used to induce the protein expression at 16 °C for 16 h. The cells were harvested by centrifugation at 4000×g for 5 min at 4 °C, and resuspended in buffer D (50 mM Tris-HCl, 5 mM MgCl2, 250 mM NaCl, 10% glycerol, pH 7.5). The cells were lysed and centrifuged at 13,000×g for 10 min to obtain the supernatant crude protein.

SaSSy mutant protein was subjected to a Ni-NTA affinity column successively eluted with buffer A (50 mM Tris-HCl, 250 mM NaCl, 5 mM MgCl2, 10% glycerol, pH 8.4), buffer B (50 mM Tris-HCl, 250 mM NaCl, 5 mM MgCl2, 10% glycerol, 100 mM imidazole, pH 8.4), and buffer C (50 mM Tris-HCl, 250 mM NaCl, 5 mM MgCl2, 10% glycerol, 300 mM imidazole, pH 8.4). After the removal of imidazole by dialysis, the resulting enzyme solution was concentrated by centrifugation.

In vitro enzymatic reactions

Enzymatic reactions were performed in 500 μL buffer D (50 mM Tris-HCl, 5 mM MgCl2, 250 mM NaCl, 10% glycerol, pH 7.5) containing 5 μM enzyme and 3 mM FPP (Sigma). The reaction mixtures were incubated at 30 °C for 3 h, followed by extraction with n-hexane for GC-MS analysis. For the SanSyn mutants, the reaction mixtures were established with the crude proteins which were prepared by centrifugation of E. coli (expressing the SanSyn mutants) cell lysates at 13,000 × g for 10 min.

Extraction and analysis of santalols and santalenes

The cultures were extracted twice with an equal volume of EtOAc. The EtOAc layers were combined and dried in vacuum, and then the residues were dissolved in hexane for GC-MS analysis.

Fermentation products were analyzed on an Agilent GC-MS instrument (7890B/5977B) equipped with an HP-5MS column. The correlation co-efficients of α-santalene and Z-α-santalol are shown in Supplementary Fig. 24.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.