Introduction

Prenylation modifications, which are ubiquitously present across diverse life forms, play important roles in both primary and secondary metabolisms1,2. Natural products (NPs) that undergo prenylation modifications generally exhibit enhanced biological and pharmacological activities compared to their non-prenylated counterparts3,4. Additionally, The prenylation of diverse NPs (e.g., flavonoids, alkaloids, coumarins, quinones, xanthones) leads to the creation of innovative structures with functions beyond those already known, thereby garnering intensive attention from the scientific community5,6,7. The prenylation reaction is mainly catalyzed by several families of enzymes known as prenyltransferases (PTs) in living organisms, including (i) membrane-bound UbiA-type PTs, (ii) dimethylallyl tryptophan synthase (DMATS)-type PTs, (iii) NphB/CloQ-type PTs, and (iv) peptide, protein, and tRNA PTs8. Besides, some class I terpene cyclases can be switched to aromatic prenyltransferases, such as AaTPS and FgGS9.

DMATS PTs belong to the soluble ABBA-PT superfamily, characterized by a distinctive αββα (ABBA) fold comprising a 10-stranded antiparallel β-barrel surrounded by α-helices10. This family of PTs primarily utilizes dimethylallyl diphosphate (DMAPP) as the prenyl donor for prenylation. And they have demonstrated the ability to catalyze both regular prenylation (addition of the primary carbon of the prenyl donor to the acceptor) and reverse prenylation (addition of the tertiary carbon of the prenyl donor to the acceptor) of diverse compounds. Common substrates include tryptophan, indole-containing diketopiperazines (DKPs), tyrosine, and other aromatic compounds (e.g., flavonoids, hydroxynaphthalenes, xanthones, benzophenones) (Fig. 1a and Supplementary Figs. 1 and 2)8,11. Most notably, DMATS PTs demonstrate remarkably broad substrate scopes, rendering them promising biocatalysts for applications in chemoenzymatic synthesis and medicinal chemistry8.

Fig. 1: Common prenylated products.
figure 1

a The reported representative structures of various substrate types that can be catalyzed by DMATS PTs include tryptophan, tyrosine, indole-containing DKPs, flavonoids, hydroxynaphthalenes, xanthones, benzophenones, indole diterpenes, polyketides, and so on. The background color for the reaction sites corresponding to regular prenylated products is pink, while the background color for reverse prenylated products is purple. Additionally, the background color for the reaction sites associated with regular and reverse products is green. b Reported DKPs with imidazole-prenylated moieties.

Full size image

In principle, imidazole, being an aromatic and electron-rich heterocycle, serves as a potential receptor for prenylation reactions. However, only a limited number of enzymes have been reported to catalyze imidazole prenylation. LimF, a peptide PT from cyanobacteria, has been identified to catalyze the geranylation of imidazole-C2 in histidine-containing ribosomally synthesized and post-translationally modified peptides (RiPPs)12,13. Furthermore, the recently characterized FunA from Fusarium tricinctum catalyzes the formation of imidazole C5-regular dimethylallylation products from free L-histidine14. But PTs capable of imidazole moiety prenylation on DKPs have yet to be discovered, despite the existence of diverse prenylated imidazole fragments in DKPs, such as aurantiamine (1), roquefortine E, (-)-phenylahistin and gartryprostatin C, featuring C5-reverse dimethylallylation, and viridamine (5), which displays a rare C2-regular dimethylallylation (Fig. 1b)15,16,17,18,19,20. These prenylated imidazole-containing NPs demonstrate significant bioactivities, with particular emphasis on (-)-phenylahistin, the leading compound of the potent microtubule-targeting agent ‘plinabulin’, showcasing cytotoxic activity against a wide array of tumor cell lines18,21,22. Moreover, the relationship between the structure and biological activity of phenylahistin derivatives underscores the essential role of the prenylated imidazole moiety in cytotoxicity23,24. Hence, the study of PTs capable of imidazole prenylation can provide valuable post-modification enzymatic tools for the production of bioactive natural products.

In this work, we report the discovery of the DMATS family PT AuraA from the fungus Penicillium solitum HDN11-131, which is involved in the biosynthesis of aurantiamine (1), a histidine-containing DKP with C5-reverse dimethylallylation. In vitro assays demonstrate that AuraA catalyzes the reverse dimethylallylation of the dehydrogenated substrate cyclo-(L-Val-DH-His) (3) at the C5 position of imidazole. Furthermore, AuraA can accept the non-dehydrogenated cyclo-(L-Val-L-His) (2) as a prenyl acceptor and produces the C2-regular prenylated dihydroviridamine (4). Structures of AuraA, both in the absence and presence of ligands, determined by Cryo-electron microscopy (cryo-EM), reveal the molecular basis of imidazole dimethylallylation and the factors governing chemoselectivity, encompassing regioselectivity and regular or reverse dimethylallylation selectivity. Finally, the mutants of AuraA result in a shift in its chemoselectivity, although they consistently produce imidazole-C-dimethylallyl products. Therefore, the identification and characterization of AuraA enhance our understanding of controlling positional selectivity and cis-trans selectivity of products in DMATS PTs.

Results

Identification of the P. solitum HDN11-131 aura gene cluster

In our continued exploration of bioactive secondary metabolites from fungi, the known DKP alkaloid aurantiamine (1) was isolated from the ethyl acetate extract of the mangrove-derived fungus P. solitum HDN11-131, and the structure was confirmed by nuclear magnetic resonance (NMR) (Supplementary Table 6 and Supplementary Figs. 24-31)15. Compound 1 features C5-reverse dimethylallylation on the imidazole moiety, proven to be a typical pharmacophore of (-)-phenylahistin24. In the analysis of cytotoxic activity, aurantiamine (1) exhibited significant cytotoxicity against multiple cells (Supplementary Table 1). However, the enzymes catalyzing the prenylation of histidine-containing DKPs on imidazole moiety have not been discovered. Thus, the biosynthetic origin of 1 was investigated to determine the unique PT.

To explore the biosynthetic pathway of 1 and identify the PT catalyzing the dimethylallylation of the imidazole ring, we sequenced the producing strain P. solitum HDN11-131 and unveiled a potential biosynthetic gene cluster (BGC) named aura (Fig. 2a and Supplementary Table 2, accession number PP622746). The aura gene cluster comprises four adjacent genes encoding the di-modular NRPS (auraC), the DMATS-type PT (auraA), the major facilitator superfamily (MFS) transporter (auraB), and the P450 monooxygenase (auraD). To validate that the aura gene cluster is accountable for the synthesis of 1, we performed heterologous expression in the widely used host Aspergillus nidulans A1145 due to the low yield in the native producer25. We initially expressed the three functional genes auraACD in A. nidulans (AN). However, the liquid chromatography (LC) analysis indicated the absence of target compound 1 or new products in the extracts of transformant AN-auraACD, which may be due to the lack of amino acid precursors (Supplementary Fig. 3). Consequently, we attempted to enhance the fermentation medium by adding L-His and L-Val. As a result, the target compound 1 was successfully produced, as confirmed by comparison to the standard (Fig. 2b, trace ii). This finding suggests that the aura gene cluster is responsible for the biosynthesis of 1.

Fig. 2: Confirmation of the aura cluster.
figure 2

a Depiction of the organization and proposed function of auraA-D. b LC-MS analyses of extracts from A. nidulans transformants, featuring the extracted ion chromatograms (EIC) under positive ionization. Source data are provided as a Source Data file. c Illustration of the biosynthetic pathway of aurantiamine (1), dihydroviridamine (4), and viridamine (5).

Full size image

The function of these three genes was further investigated by reconstituting the biosynthetic pathway in A. nidulans. As depicted in Fig. 2b, trace iii-v, the expression of the core NRPS gene auraC alone resulted in the successful isolation of 2, identified as cyclo-(L-Val-L-His) through comprehensive NMR analysis (Supplementary Table 7 and Supplementary Figs. 32-38). The P450 monooxygenase aruaD exhibits a high identity and similarity with roqR (42% identity and 60% similarity), a protein known for catalyzing the α,β-dehydrogenation of the DKP scaffold26,27, and the co-expression of auraCD led to the production of 3 as expected (Fig. 2b, trace iv). Compound 3 was isolated and the structure was fully characterized, based on NMR data (Supplementary Table 8 and Supplementary Figs. 39-46). Surprisingly, we detected a small amount of 4 with a molecular weight 68 Da greater than that of 2 from the transformant harboring auraA and auraC (Fig. 2b, trace iii). Isolation and subsequent NMR analyses elucidated that 4 is the direct C2-regular prenylated product of 2 (Supplementary Table 9 and Supplementary Figs. 47-53). The in vivo results above indicate that AuraA, as a prenyltransferase, can prenylate the imidazole moiety of 2 and 3 but with distinct regioselectivity and dimethylallylation manners (Fig. 2c). Thus, we performed further investigations on the function of AuraA.

Heterologous expression and in vitro activity of AuraA

AuraA shares 27.5% and 26.7% sequence identity with AnaPT and NotF, which catalyze the prenylation of tryptophan-containing alkaloids, respectively28,29. Phylogenetic tree analysis indicated that AuraA belongs to the DMATS superfamily (Fig. 3a). To further confirm the function of AuraA and reveal its enzymatic reaction characteristics, intron-free auraA was cloned from the cDNA of A. nidulans expressing the aura gene cluster and ligated to the expression vector pET-28a(+). AuraA was heterologously expressed in Escherichia coli BL21 (DE3) and purified as a His12-tagged fusion protein by Ni-chelate affinity chromatography (Fig. 3b and Supplementary Fig. 4). Following the incubation of AuraA with 2 or 3 and DMAPP, we detected the corresponding prenylated products 4 and 1, consistent with the in vivo results (Fig. 3c). The results provide reliable evidence that AuraA catalyzes the C2-regular dimethylallylation of 2 and C5-reverse dimethylallylation of 3 at the imidazole ring. However, the incomplete prenylation of substrate 2 in the in vitro reaction indicates that the fundamental function of AuraA is to catalyze the reverse dimethylallylation of 3 at the imidazole C5 position. Subsequently, the comparison of kcat/KM values demonstrated that AuraA prefers to use 3 as a substrate, as the catalytic efficiency of AuraA towards 3 was nearly 8.2-fold greater than that towards 2 (Supplementary Fig. 5b, c). Up to now, AuraA is the unusual DMATS PT to achieve the prenylation of imidazole systems.

Fig. 3: Characterization of AuraA.
figure 3

a Phylogenetic tree analysis of AuraA. Representative DMATS, UbiA, NphB/CloQ, and peptide PT form four clades, indicated by blue, orange, purple, and green background, respectively. AuraA is highlighted in red. b Purification of AuraA as a His12-tagged fusion protein post gene expression. Lane 1: purified His12-AuraA; lane 2: molecular mass standard. This experiment was repeated independently three times. c LC-MS traces from in vitro assays with AuraA using substrate 2 or 3. X5 represents that the trace values of m/z 237 have been amplified fivefold. Source data are provided as a Source Data file.

Full size image

DMATS-type aromatic PTs do not contain the conserved metal ion-binding (N/D)DXXD motif29,30,31,32. They are generally considered metal-independent enzymes, although, in several cases, the addition of metal ions such as Ca2+ and Mg2+ enhances their activities33,34. To examine the influence of metal ions on enzymatic reactions of AuraA, the enzymatic activity was assayed in the presence of different concentrations of metal ions (Ca2+, Mg2+, Mn2+, Ni2+, and Zn2+) or the chelating agent EDTA using the preferred substrate 3 and DMAPP. The LC-MS results revealed that the presence of metal ions and EDTA, with the exception of Zn2+, did not affect the occurrence of the AuraA reaction. (Supplementary Fig. 5d). These results clearly indicate that AuraA is not dependent on divalent metal cations, which corresponds well to those reported for most members of the DMATS superfamily11,30.

Overall structure of AuraA and its ternary complexes

We were intrigued by the diverse dimethylallylations catalyzed by AuraA on substrates that only differ by one double bond in their structures. To further understand the basis for the catalytic selectivity of AuraA, we solved the overall structures of ligand-free AuraA (2.95 Å resolution, referred to as the apo structure) as well as the ternary complexes with the unreactive diphosphate analog, dimethylallyl S-thiolodiphosphate (DMSPP), and 2 or 3 (2.45 Å and 2.67 Å resolution, respectively, referred to as the holo structure), using cryo-EM techniques (Supplementary Figs. 6-11). AuraA is a tetramer, consistent with the aggregated state measured in solution (Supplementary Fig. 14). The architecture of each AuraA subunit features the typical ABBA-fold, consistent with other DMATS PTs (Fig. 4a)29,31,32,35,36,37.

Fig. 4: Cryo-EM structures of AuraA and two ternary complexes of AuraA (DMSPP-2 and DMSPP-3).
figure 4

a Cartoon representation of AuraA (yellow, PDB ID 8Y9E), AuraA-DMSPP-2 (magenta, PDB ID 8Y9D), and AuraA-DMSPP-3 (cyan, PDB ID 8Y9G) viewed from the top. The salmon shadow highlights differences in the β1-β2 turn region (residues A111GES114) between the two holo-structures and the apo structure in the cartoon representation. The β1-β2 turn region (residues A111GES114) of AuraA (yellow), AuraA-DMSPP-2 (magenta), and AuraA-DMSPP-3 (cyan) is displayed in stick and surface (60% transparency) representation. b Key residues in the catalytic chamber of AuraA are categorized into three regions: the prenyl donor binding pocket (green), the tyrosine shield (purpleblue), and the acceptor binding pocket (gray80). c Structures of prenyl donors and acceptors, along with residues of tyrosine shields (purpleblue), from AuraA-Apo, AuraA-DMSPP-2 (magenta), and AuraA-DMSPP-3 (cyan), are shown in sticks. Residues Y207 and Y280 exhibit significantly different conformations in two ternary complexes, marked separately in salmon. Q337 (gray70) acts as a key residue for substrate binding, while the conformational change of Y280 creates space for its interaction with the substrate. The dashed line, colored with gray60, indicates a hydrogen bond interaction between two atoms.

Full size image

AuraA is structurally homologous to AnaPT (PDB code: 4LD7) and NotF (PDB ID: 6VY9), with Cα-root mean square deviations (RMSDs) of 1.55 and 1.72 Å, respectively. Structural analysis demonstrates strong resemblances between the apo form of AuraA and its holo forms, as shown by RMSDs of 0.437 Å for the comparison with the DMSPP/2 complex and 0.385 Å for the comparison with the DMSPP/3 complex (Fig. 4a). The circular β-barrel cavity provides a typical active site to accommodate the substrates DMSPP and 2 or 3 in DMSPP/2 or /3 complexes (Supplementary Fig. 15).

Structural basis of AuraA and its ternary complexes

Generally, DMATS PTs, require three key regions to complete the prenylated reaction in the substrate-binding pocket29,31,32,35,36,37. The first region is responsible for binding the diphosphate group to constrain the direction of the dimethylallyl group. There are five positively charged residues (R118, K205, R276, K278, and K352) anchoring the negatively charged diphosphate group, and Y419 contributes hydrogen bonds to the β-phosphate group (Fig. 4b and Supplementary Fig. 16). The second region is composed of four tyrosine residues (Y207, Y280, Y354, and Y423), used to shield the reactive carbocation intermediate from solvent by stabilizing π-cation interactions between the aromatic rings of Tyr and the carbocation (Fig. 4b). The third region is a hydrophobic substrate binding pocket composed of multiple hydrophobic residues (F98, L103, V191, Y207, Y264, W408, and Y423) and other residues (E104, T120, H186, C188, P209, and Q337) in both complex structures (Fig. 4b).

Comparing the structures of apo and holo AuraA, we observed a clear conformational change in the β1-β2 turn (residues A111GES114) above the prenyl donor binding pocket. This β turn points inward toward the center of the inner barrel in the apo structure but adopts an outward orientation in both holo structures, which may be beneficial for opening the upper part of the cavity to influence the entry of the prenyl donor (Fig. 4a). Additionally, we observed conformational changes in multiple residues in the substrate-binding pockets of the two holo structures, especially in regard to the binding diphosphate groups and the shield of tyrosine, including K205, Y207, R276, K278, Y280, K352, and Y354 (Supplementary Fig. 17). Similar to AtaPT, Y207, and Y280 exhibit an “open state” conformation in both holo structures, which not only facilitates the formation of the tyrosine shield region but also provides additional space for substrate binding (Fig. 4c)36.

Structural analysis of two different substrate-binding modes of AuraA

To explain the catalytic selectivity of AuraA, we compared the orientation of the prenyl moiety in both two holo-structures. The β-phosphate remained at the same position, while the oxygen atom connecting the two phosphate groups shifted slightly. Remarkably, the side chain of R118 shifted drastically among the two holo-structures (Fig. 5a). In the DMSPP/2 structure, the orientation of the diphosphate group is mainly restricted by the interactions with K278 through salt bridges (Fig. 5a), while in the DMSPP/3 structure, the donor direction is retained by the interactions with R118. In addition, in each structure, the dimethylallyl moiety extends toward its acceptor’s imidazole, but the orientation is nearly perpendicular (Supplementary Fig. 16). In the DMSPP/2 structure, the dimethylallyl moiety is mainly surrounded by the imidazole of 2, and residues T120, Y207, and Y280 (Fig. 5b). In the DMSPP/3 structure, the dimethylallyl group is mainly surrounded by the imidazole of 3, and residues Y354, W408, and Y423 (Fig. 5b). Notably, the indole of Y354 undergoes obvious rotation to expand the cavity, accommodating the dimethylallyl moiety and maintaining tyrosine shielding in the DMSPP/3 structure (Fig. 5b). Thus, for different prenyl acceptors, a few residues, related to the formation and shielding of dimethyl cations, may undergo conformational changes, leading to the dimethylallyl group being reoriented toward its acceptor and increasing the promiscuity of the enzyme.

Fig. 5: Comparison of the active residues involved in substrate binding in three ternary complexes of AuraA.
figure 5

a The residues for the binding of the diphosphate group in AuraA-DMSPP-2 (green) and AuraA-DMSPP-3 (salmon) complexes. b The residues for maintaining tyrosine shielding in AuraA-DMSPP-2/3 complexes. c The residues for interacting with the valine side chains of acceptors in AuraA-DMSPP-2/3 complexes. d The residues for orienting the imidazole moiety of acceptors in AuraA-DMSPP-2/3 complexes. The structures of prenyl donors and acceptors in AuraA-DMSPP-2 and AuraA-DMSPP-3 are represented by lightmagenta and cyan sticks, respectively. R118 and Y354 exhibit significantly different conformations in the AuraA-DMSPP-2 and AuraA-DMSPP-3 structures, marked separately in purpleblue (a, b). The dashed line, colored with gray40, indicates a strong noncovalent bond interaction between two atoms. e The distance between the prenyl acceptors and donors in the AuraA-DMSPP-2 and AuraA-DMSPP-3 structures. These two substrate-binding modes effectively elucidate the reasons behind the formation of products with distinct chemoselectivity. f The residues responsible for maintaining tyrosine shielding within the AuraA-Y207A-DMSPP Cryo-EM structure. Y280 exhibits a significantly different conformation in the AuraA-Y207A-DMSPP structure, marked separately in purpleblue. g Molecular docking of substrate 3 into AuraA-Y207A-DMSPP structure.

Full size image

Consequently, we compared the binding modes of the prenyl acceptors in two holo-structures. The valine side chains of acceptors 2 and 3 are both stabilized by hydrophobic interactions with six residues (F98, L103, H186, C188, V191, and P209) lining one side of the cavity (Fig. 5c). The specific Q337, distinct from the hydrophobic residues in most PTs (Supplementary Fig. 19), forms a hydrogen bond with the carbonyl group on the DKP ring to constrain the imidazole moiety facing the prenyl donor. Interestingly, in addition to the immobilization of the aforementioned fragments, the residues (E104, T120, Y207, Y264, W408, and Y423) involved in orienting the imidazole moiety of 2 and 3 are also identical, but the orientations of acceptors 2 and 3 are different (Fig. 5d). In the DMSPP/2 structure, residue E104 interacts with 2 via a hydrogen bond to the N1 of imidazole, and W408 forms a noncovalent π-π stacking interaction with imidazole, while others (T120, Y207, Y264, and Y423) only stabilize the imidazole moiety through van der Waals interactions. In the DMSPP/3 structure, residue E104 is also capable of forming a hydrogen bond with the imidazole N1. Residue Y207 participates in π-π T-shaped stacking and T120 shows the hydrogen bond interaction with the imidazole moiety. The remaining residues Y264, W408, and Y423 only generate van der Waals force interactions with the imidazole moiety. Unlike the residues involved in donor-binding, the conformations of the residues used for receptor-binding in the two holo-structures are highly similar. The distinct interactions with the above residues are likely responsible for the tolerance of two different acceptor binding modes.

The two different substrate-binding modes in the reaction chamber directly affect the relative distance between the binding sites of prenyl acceptors and donors. Among the four possibilities for imidazole-C-dimethyl allylation, the distance from C2 of 2 to C1 of DMSPP (3.1 Å) is the shortest in the DMSPP/2 structure. In contrast, the distance from C5 of 3 to C3 of DMSPP (3.0 Å) is the shortest in the DMSPP/3 structure (Fig. 5e). These data explained the distinctions of C2-regular dimethylallylation at imidazole of acceptor 2 and the C5-reverse dimethylallylation of acceptor 3. Therefore, the results revealed that the changes in the dimethyl allylation position and mode of products depend on two different substrate-binding modes.

Alteration of the dimethyl allylation promiscuity via mutagenesis

To investigate the residues impact on dimethylallyl promiscuity, we mutated the residues E104, T120, Y207, Y264, Q337, W408, and Y423, which are involved in substrate 3 binding (Fig. 5c, d), based on the DMSPP/3 structure (resulting in the product 1). Among them, mutations of residue Q337 to either Ala or Glu, which plays a crucial role in immobilizing the diketopiperazine ring, resulted in the production of a small amount of 5 (Supplementary Fig. 20). Compound 5 was isolated and identified as viridamine, featuring regular dimethylallylation at the C2 atom of the imidazole nucleus, through detailed spectral analysis (Fig. 2, Supplementary Table 10 and Supplementary Figs. 54-61)16. Residue T120 interacts with the imidazole N1 of 3 through a hydrogen bonding, and its Ala and Val mutants also produce a small amount of compound 5 compared to the wild-type (Fig. 2 and Supplementary Fig. 20). These findings suggest that residues Q337 and T120 have a certain impact on the chemoselectivity of AuraA.

The conserved residue E104 interacts with the imidazole N1 position through hydrogen bonding (Fig. 5d). The catalytic efficiencies of the E104A and E104Q mutants were significantly reduced by more than 1000-fold (Supplementary Table 11 and Supplementary Fig. 21). This suggests that the negative charge of E104, similar to most reported DMATS, enhances the nucleophilicity of the aromatic heterocyclic ring, thereby affecting catalytic efficiency31,32,37,38. Interestingly, the E104A and E104D mutants exhibited alterations in their product spectrum, with the E104A variant mainly producing compound 5 (Supplementary Fig. 20). This suggests that substrate 3 in AuraA-E104A is likely to undergo a significant alteration in the binding pocket, similar to what occurs with the reported substrate DBu in AtaPT-E91A36.

Served as conserved residues involved in tyrosine shielding, Y207 and Y423, play important role in substrate 3 immobilization (Fig. 5d and Supplementary Fig. 19). To our surprise, the Y207F mutant produced a small amount of 5, whereas the Y207A mutant predominantly generated 5 with a catalytic efficiency of 3.2 min−1 mM−1 (Supplementary Figs. 20 and 21). To further elucidate the cause of the change in products associated with the Y207A mutant, we employed cryo-EM techniques to resolve its complex structure with DMSPP and 3 (Supplementary Figs. 12 and 13). Compared to the pre-mutation structure (Fig. 5b), a significant orientational change was observed in DMSPP, accompanied by a conformational change in the Y280 residue (Fig. 5f). However, a limitation of our study is that we could not determine the binding position of substrate 3 within the pocket. To address this limitation, molecular docking was performed to model the interactions of substrate 3 within the Y207A mutant. As shown in Fig. 5g, the distance from C2 of 3 to C1 of DMSPP (3.8 Å) is the shortest in AuraA-Y207A-DMSPP structure, making it most prone to facilitate a nucleophilic attack of the imidazole C2 position at C1 of DMAPP. As for the mutants Y423A and Y423F, although a small amount of 5 still was generated, their catalytic efficiencies were significantly reduced (Supplementary Table 11 and Supplementary Fig. 21), primarily due to the disruption of tyrosine shielding by the mutants, which affected the stability of carbocations.

As for residues Y264 and W408, all of the mutants (Y264A, Y264F, W408A, and W408F) resulted in the generation of product 5, especially Y264A and W408A, mainly producing 5 (Supplementary Fig. 20). These results indicated that Y264 and W408, as hydrophobic residues related to substrate immobilization, significantly influenced the chemoselectivity of the reaction due to changes in their residue size. Furthermore, we investigated the reactions of all mutants towards the non-native substrate 2 and found that most mutants exhibited decreased activity or even inactivation (Supplementary Fig. 22). Based on all the mutation results, it is evident that AuraA exhibits promiscuity in catalyzing the imidazole-C-modified reactions. Mutations in key residues involved in substrate binding can enhance the diversity of imidazole-C-modified products.

Discussion

In the past 30 years, over 20 PTs that catalyze the prenylation of DKPs have been reported (Supplementary Table 13 and Supplementary Figs. 1 and 2). These enzymes mainly belong to DMATS family. Although these enzymes exhibit some flexibility towards DKP substrates, they require the presence of an indole ring in the substrate structure, with the most prenylated positions typically located within the indole ring8. Despite the observation of at least two prenylation modes targeting the imidazole moiety of DKPs, including C5-reverse or C2-regular dimethylallylation (Fig. 1b), there has still no enzyme been found to accept His-containing DKPs and generate imidazole prenylated products. It is worth noting that the reverse dimethylallylation at C5 position in the imidazole ring is essential for the cytotoxicity of DKPs24, while modification at the C2 position of His’s imidazole moiety is more challenging due to its electron-deficient characteristics compared to its other reaction sites12. In our study, AuraA, a member of the DMATS family PTs, was found to catalyze the generation of both C5-reverse and C2-regular dimethyl allylated products on imidazole-containing DKPs. We further employed AuraA and other reported DMATS PTs as probes to search for additional potential imidazole PTs in the public fungal genome database and constructed a sequence similarity network (SSN) (Supplementary Fig. 23). The SSN results revealed that AuraA and its homologs are clustered together, separated from other types of DMATS PTs. Analysis of the neighboring gene regions revealed that AuraA homologs are incorporated in the BGCs containing a di-modular NRPS, a P450 enzyme, a DMATS PT, and an MFS, exactly the same as aura gene cluster, suggesting that the prenylation of imidazole is likely to occur with similar substrates. These findings offered new approaches for synthesizing imidazole C5-reverse and C2-regular prenylated products.

At present, crystal structures of five DKP DMATS PTs have been reported, among which FtmPT1 and NotF reveal the structural basis for catalyzing Trp-containing DKP through their crystal structures in ternary complexes with brevianamide F and DMSPP8,29,31. In our study, we employed cryo-EM technology to unveil the structural basis for the catalytic imidazole-C-prenylation of His-containing DKPs. Although the characteristics of the entire reaction chamber are similar to those of reported DMATS reaction chambers, the AuraA receptor binding pocket comprises only three conserved residues (E104, Y207, and Y423), with the remaining residues being non-conserved (F98, L103, T120, H186, C188, V191, P209, Y264, and W408). In addition, our study revealed the different substrate-binding modes in which a single enzyme catalyzes similar substrates (with only different double bonds), to form various prenylated products with different positions and modes. The occurrence of different prenylated positions and modes indicates a certain plasticity within the active cavity, and the flexibility of these reaction orientations is attributed to the plasticity of the residues surrounding the hydrophobic pocket and the prenyl donor.

Interestingly, most AuraA mutants increased product diversity. The resulting products were also imidazole-C-dimethylallyl products, indicating that AuraA has promiscuity in catalyzing the reaction of imidazole-C-dimethylallylation. Additionally, although Y207 is a conserved residue involved in tyrosine shielding and the binding of substrate 3, its Y207A mutant resulted in the generation of a non-natural product 5 with catalytic efficiency of 3.2 min−1 mM−1. This result suggests that the conserved residue involved in tyrosine shielding also contributes to product diversity. This finding offers valuable insights for engineering DMATS transformations.

In summary, our study has identified AuraA, a DMATS family PT, capable of catalyzing the imidazole-C-prenylation of His-containing DKPs with promiscuity. This discovery opens up new avenues for engineering the biosynthesis of imidazole-C-modified unnatural products.

Methods

Cytotoxicity assay

Cytotoxic activities of compounds 1-5 were evaluated against K562 by the MTT method, and against L-02, ASPC-1, MDA-MB-231, NCI-H446, and NCI-H446/EP by the SRB method. Adriamycin was used as a positive control, and the IC50 values are shown in Supplementary Table 1. The detailed methodologies for biological testing based on previous reports39.

Strains and culture conditions

The fungus P. solitum HDN11-131 was isolated from the rhizosphere soil of a mangrove plant (Rhizophora stylosa) in Yingluo Bay, Guangxi Province, People’s Republic of China, and identified through the ITS region (MW261857). It was cultivated at 28 °C for 5 days on PDA media (2.6% Potato Dextrose Broth, 2% agar) to induce sporulation. A. nidulans A1145 served as the host for heterologous expression of the aura cluster. It was cultured at 37 °C on CD media (1 litre: 10 g glucose, 50 mL 20 × nitrate salts, 1 mL trace elements, pH 6.5) for sporulation and at 28 °C on CD-ST media (1 litre: 20 g starch, 20 g casamino acids, 50 mL 20 × nitrate salts, 1 mL trace elements) for fermentation. Escherichia coli strain XL1 was employed for cloning, while E. coli BL21 (DE3) was utilized for protein expression. All E. coli strains were cultured at 37 °C on LB media, with or without antibiotics.

Extraction of gDNA and synthesis of cDNA

P. solitum HDN11-131 was cultured in PDB media (potato dextrose water) at 28 °C for 4 days to facilitate genomic DNA extraction using the DNA Quick Plant System Kit (TIANGEN) following the provided instructions. The genome sequencing data of P. solitum HDN11-131 was obtained by BGI-Qingdao.

A. nidulans harboring the auraACD genes was cultured in liquid CDST production media at 28 °C and 220 rpm for 3 days. Fungal biomass was harvested and subjected to RNA extraction using TRLZOL® Reagent (Ambion). The RNA sample was subsequently treated with DNase, followed by cDNA reverse transcription using the Transcriptor First Strand cDNA Synthesis Kit (Roche).

Construction of expression plasmids

The primers utilized in this study are detailed in Supplementary Table 3, while the plasmids are outlined in Supplementary Table 4.

To construct A. nidulans expression plasimids, individual genes (auraA, auraC, and auraD) along with their terminators (~500 bp) and two homologous arms were amplified from P. solitum HDN11-131 gDNA using Q5® High-Fidelity DNA Polymerase (NEB) through PCR. Subsequently, the PCR products were purified and co-transformed with NotI-digested pANU, BamHI-digested pANR, or BamHI-digested pANP plasmids into Saccharomyces cerevisiae strain BJ5464 for in vivo recombination. The resulting circular plasmids were extracted from yeast and subsequently transformed into E. coli XL1 strain to acquire purified plasmids for further transformation.

To assemble E. coli expression plasimids, intron-free DNAs of auraA were amplified from the cDNA of the A. nidulans-auraACD expression strain and inserted into NdeI-digested pET-28a(+)plasmids using the Seamless Assembly Cloning Kit (Clone Smarter). Mutated fragments were generated by PCR using pET-28a(+)-auraA plasmids as templates and subsequent mutants were created using the Seamless Assembly Cloning Kit (Clone Smarter). The resulting circular plasmids were transformed into E. coli DH10B strain to acquire cloning plasmids. All mutated plasmids have been confirmed through gene sequencing.

Construction of A. nidulans heterologous expression strains

The A. nidulans A1145 strain was cultivated in solid CD media supplemented with 10 mM uridine, 5 mM uracil, 0.5 μg mL−1 pyridoxine HCl, and 0.125 μg mL−1 riboflavin at 37 °C for 5 days. The spores were collected and inoculated into 50 mL liquid CD media in a 250 mL Erlenmeyer flask at 37 °C, 220 rpm for ~9 h. The culture fluid was centrifuged at 4 °C, 3059 × g for 10 min to harvest the mycelia. The mycelia were washed twice with 15 mL osmotic buffer (1.2 M MgSO4· 7H2O, 10 mM sodium phosphate, pH 5.8). Subsequently, the mycelia were resuspended in 10 mL osmotic buffer containing 30 mg lysing enzymes from Trichoderma and 20 mg Yatalase in a 100 mL flask at 28 °C, 80 rpm for 6 h. The culture fluid was poured directly into a sterile 50 mL centrifugal tube and gently overlaid with 10 mL trapping buffer (0.6 M D-sorbitol, 0.1 M Tris-HCI, pH 7.0). Subsequently, the protoplasts were harvested by centrifugation at 3059 × g for 18 min at 4 °C and transferred into a sterile 15 mL centrifuge tube. The protoplasts were rinsed with 10 mL of STC buffer (1.2 M D-sorbitol, 10 mM CaCl2, 10 mM Tris-HCI, pH 7.5) and then resuspended in 1 mL of STC buffer for transformation. 2 μL of recombinant plasmids was added to 100 μL protoplast suspension. After incubating on ice for 45 min, 600 μL of PEG solution (60% PEG, 50 mM CaCl2, and 50 mM Tris-HCl, pH 7.5) was added to the mixture and incubated at room temperature for 30 minutes. The mixture was spread evenly onto the regeneration dropout solid CDS media (CD solid media with 1.2 M D-sorbitol and appropriate supplements) and cultured at 37 °C for 2 days. Finally, the transformants were cultured on CD media at 37 °C for rejuvenation. The transformants of A. nidulans strains were grown on CD-ST media supplemented with L-His and L-Val with a final concentration of 5 mM at 28 °C for 4 days. Subsequently, the fermentation broth was extracted with ethyl acetate (EtOAc) and the organic phase was dried by speed vacuum and dissolved in methanol for analysis or isolation of compounds.

The protein expression and purification of AuraA in E. coli

Recombinant plasmids pET-28a(+)-auraA and its mutants were individually transformed into the E. coli BL21 (DE3) strain using the heat shock transformation method for protein expression. The strain was cultured to OD600 = 0.4-0.6 in 1000 mL LB media containing 50 μg/mL kanamycin at 37 °C with 220 rpm. The temperature of the culture was lowered to below 16 °C. Subsequently, the cells were cultured at 16 °C for 18 h with 0.2 mM isopropylthio-β-D-galactoside (IPTG). Following that, the cells were collected at 4 °C, 3059 × g for 5 min. The cells containing AuraA were resuspended in 20 mL of bufferA (50 mM Tris-Cl, 500 mM NaCl, 10% glycerol, pH 7.5) and subjected to sonication on ice for 20 min. Cellular debris was collected at 4 °C, 15,133 × g for 40 min, and subsequently purified by Ni-NTA agarose resin. The protein was then eluted using buffer A containing 20 mM, 50 mM, 100 mM, 200 mM, and 500 mM imidazole. The protein concentration (from the 200 mM imidazole eluent) was obtained by using a 10 kDa ultrafiltration centrifugal tube (Millipore Amicon® Ultra-15 mL) at 4 °C, 3059 × g. The concentrated enzyme solution was resuspended in a small amount of buffer C (50 mM Tris-Cl, 50 mM NaCl, 5% glycerol, pH 7.5), dispensed into 1.5 mL centrifuge tubes, and subsequently stored at −80 °C after being quick-frozen in liquid nitrogen. The purified enzyme was analyzed using 10% SDS-PAGE (Omni-Easy™One-Step PAGE Gel Fast Preparation Kit). Protein concentration was calculated by measuring ultraviolet absorption at A280.

The in vitro biochemical assay of AuraA and its mutants

Reactions were preheated to 28 °C without DMAPP and were initiated by the addition of DMAPP. To assess the in vitro activity of AuraA-catalyzed prenylations involving substrates 2 and 3 (Fig. 3), the reaction system contained 1 μM AuraA, 200 μM DMAPP, and 100 μM substrates 2 or 3, incubated at 28 °C for 10 min. To examine the impact of reaction time on AuraA-catalyzed prenylations of the native substrate 3 (Supplementary Fig. 5a), the reaction was comprised of 0.2 μM AuraA, 400 μM DMAPP, and 200 μM 3, incubated at 28 °C for 0, 3, 10, 20, 40, 60, 80, 100, 120, 180, 240, 360, 480, and 600 min, respectively. To investigate the influence of metals on AuraA-catalyzed prenylations of the native substrate 3, reactions were composed of 0.2 μM AuraA, 400 μM DMAPP, 200 μM substrate 3, pH 7.5, 1.2 mM EDTA, and 2 mM (or varying concentrations of) various metals, incubated at 28 °C for 30 min. To evaluate the enzyme kinetics of AuraA and its mutants concerning substrates 2 and 3 (Supplementary Table 11), the reactions included 1 μM AuraA for 2; 0.2 μM AuraA for 3; and various concentrations of AuraA mutants for 3 (1 μM AuraA-Q337A, 5 μM AuraA-Q337E, 5 μM AuraA-T120A, 5 μM AuraA-T120V, 5 μM AuraA-E104D, 10 μM AuraA-E104A, 5 μM AuraA-Y207A, 5 μM AuraA-Y207F, 20 μM AuraA-Y423A, 5 μM AuraA-Y423F, 7 μM AuraA-Y264A, 5 μM AuraA-Y264F, 10 μM AuraA-W408A, and 5 μM AuraA-W408F), 4 mM DMAPP, and varying concentrations of 2 or 3, incubated at 28 °C for 30 min. To assess the reactivity of AuraA and its mutation proteins (Supplementary Fig. 20), the reactions included 5 μM protein, 400 μM DMAPP, and 200 μM 2 or 3, incubated at 28 °C for 2 h. Subsequently, the reaction was quenched with 100 μL MeOH and analyzed by LC-MS. The relative conversion rates of the enzyme reactions were calculated from ion peak areas of prenylated products and substrates as analyzed by LC-MS. Each experiment was replicated three times.

Chemical analysis of samples

The LC-MS analyses were performed using a Waters ACQUITY H-Class UPLC-MS system equipped with a PDA detector and SQD2 mass spectrometer (MS) detector, employing a reversed-phase C18 column (ACQUITY UPLC® BEH, 1.7 μm, 2.1 × 50 mm, Waters). The general analysis methods in this work involved an isocratic elution of 5% MeCN-H2O with 0.02% formic acid for 1 min, followed by a linear gradient of 5-50% MeCN-H2O with 0.02% formic acid over 9 min, using a flow rate of 0.5 mL min−1.

The HPLC analyses were carried out using a Hitachi High-Tech Science HPLC system with a reversed-phase C18 column (5.0 μm, 4.6 × 250 mm, Silgreen GH20100044). The general analysis methods involved an isocratic elution of 5% MeOH-H2O with 0.05% formic acid for 5 min, followed by a linear gradient of 5-100% MeOH-H2O with 0.05% formic acid over 35 min, using a flow rate of 1 mL min−1.

Isolation and structural characterization of compounds

Compound 1 was obtained through the fermentation of P. solitum HDN11-131 under static conditions at 28 °C in 1000 mL Erlenmeyer flasks containing 300 mL of liquid media with soluble starch (4.00%), sucrose (4.00%), maltose (3.00%), peptone (0.20%), sodium glutamate (0.20%), yeast extract (0.10%), soybean meal (0.05%), KH2PO4 (0.05%), MgSO4⋅7H2O (0.03%) dissolved in naturally collected seawater (Huiquan Bay, Yellow Sea, Qiangdao, China) over a 4-week period. Compounds 2-5 were generated by the respective A. nidulans heterologous expression strains (Supplementary Table 5) in solid CD-ST media, supplemented with L-His and L-Val to a final concentration of 5 mM at 28 °C for 4 days. The fermentation media was extracted with EtOAc three times, and the EtOAc phase was evaporated under reduced pressure to obtain EtOAc extracts. The extract was fractionated by VLC of ODS using a step gradient elution of MeOH-H2O (5:95-60:0). The fractions containing the target compounds were combined for additional purification using a Sephadex LH-20 column with MeOH, followed by semi-preparative HPLC (Hitachi High-Tech Science Semi-preparative HPLC) employing a YMC-Pack ODS-A column (5 μm, 10 × 250 mm) with a flow rate of 3.0 mL min−1. The structures of compounds 1-5 were established through extensive spectroscopic analysis. NMR spectra were recorded on Bruker Avance NEO 400 (400 MHz), Agilent DD2 500 (500 HMz), and JNM-ECZ600R/S1 (600 MHz) spectrometers. 1H and 13C spectra were referenced to the residual deuterated solvent peaks at δH 2.50 and δC 39.52 (DMSO-d6).

Commund 1, 8.2 mg, 45% MeOH/H2O system, tR = 15 min, 1H and 13C NMR see Supplementary Table 6, ESIMS m/z [M + H]+ 303.24.

Commund 2, 5.8 mg, 5% MeOH/H2O system, tR = 8 min, 1H and 13C NMR see Supplementary Table 7, ESIMS m/z [M + H]+ 237.11.

Commund 3, 10.1 mg, 40% MeOH/H2O system, tR = 12 min, 1H and 13C NMR see Supplementary Table 8, ESIMS m/z [M + H]+ 235.12.

Commund 4, 9.5 mg, 35% MeOH/H2O system, tR = 11 min, 1H and 13C NMR see Supplementary Table 9, HRESIMS m/z [M + H]+ 305.1969 (calcd for C16H25N4O2, 305.1972).

Commund 5, 1.8 mg, 45% MeOH/H2O system, tR = 11 min, 1H and 13C NMR see Supplementary Table 10, ESIMS m/z [M + H]+ 303.28.

Cryo-EM data collection

A 4.0 μL aliquot of AuraA (or AuraA-Y207) and its complexes (containing DMSPP and 2 or 3) was applied to a freshly glow-discharged holey carbon grid (Quantifoil Au R2/1, 200 mesh) with continuous carbon support. The grids were plunge-frozen into liquid ethane using an FEI Vitrobot Mark IV (ThermoFisher Scientific) at 4 °C and >90% humidity with 2 s blot time and −1 blot force. The grids were placed into a 300 kV Titan Krios G3i microscope (Thermo Fisher) that was equipped with a K3 BioQuantum direct electron detector (Gatan, USA) or a 200 kV Glacios 2 microscope (Thermo Fisher) that was equipped with a Falcon 4 direct electron detector (Thermo Fisher) for data acquisition. A total of 5062 movies for AuraA, 6620 movies for the DMSPP/2 complex, and 5503 movies for the DMSPP/3 were collected at a total dose for a stack of ~50 e− Å−2 in a defocus range of −1.2 to −2.2 μm. Super-resolution mode was used at a nominal magnification of ×81,000 corresponding to a pixel size of 0.53 Å, with the energy filter slit set to 20 eV. In contrast, a total of 4723 movies for AuraA-Y207A-DMSPP was collected at a total dose for a stack of ~40 e− Å−2 with a pixel size of 0.89 Å.

Data processing of Cryo-EM data

Data processing was conducted using cryoSPARC Software40. After patch-motion and CTF correction, particles were picked using the blob-picking algorithm in cryoSPARC. The particles underwent multiple rounds of 2D classification. These runs produced particles for analysis, which underwent Ab-initio reconstruction. After homogeneous refinement, 3D nonuniform refinement, sharpening, global (per-group) CTF refinement, and local (per-particle) CTF refinement, were conducted. The overall resolutions of the maps for AuraA, the DMSPP/2 complex, and the DMSPP/3 complex were 2.95, 2.45, and 2.67 Å, respectively. The overall resolution of the map for AuraA-Y207A-DMSPP was 2.86 Å. Details of the cryo-EM data processing and associated parameters can be found in Supplementary Table 12 and Supplementary Figs. 6-13.

Model building and refinement of structures

The AnaPT structure (PDB code: 4LD7) was initially docked into the cryo-EM map of AuraA with resolution using UCSF Chimera41. The amino acid sequences were then mutated to match those of AuraA. The AuraA model was manually rebuilt based on the cryo-EM density using COOT42 and then subjected to real-space refinement using Phenix 1.16.354943. The atomic models of the DMSPP/2 complex, DMSPP/3 complex, and AuraA-Y207A-DMSPP were built using a method similar to that for AuraA but utilizing AuraA as a reference starting model. Geometrical restraints for substrates were generated using the Grade Web Server. MolProbity 4 was employed to assess the geometries of the structures44. Images were rendered using Chimera and ChimeraX45.

Molecular docking

Molecular docking studies for compound 3 and AuraA-Y207A-DMSPP were performed using MOE (version 2014.09) software46. Compound 3 was drawn using ChemDraw (version 15.0), and its lowest energy conformation was optimized using Chem3D. The receptor utilized for docking was derived from AuraA-Y207A-DMSPP obtained through cryo-EM technology. PyMOL (version 4.6.0) was employed for structural visualization and figure preparation.

Reporting summary

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