Huan
Liu‡
a,
Jie
Fan‡
a,
Peng
Zhang
a,
Youcai
Hu
d,
Xingzhong
Liu
a,
Shu-Ming
Li
*b and
Wen-Bing
Yin
*ac
aState Key Laboratory of Mycology, CAS Key Laboratory of Microbial Physiological and Metabolic Engineering, Institute of Microbiology, Chinese Academy of Sciences, Beijing 100101, P. R. China. E-mail: yinwb@im.ac.cn
bInstitut für Pharmazeutische Biologie und Biotechnologie, Fachbereich Pharmazie, Philipps-Universität Marburg, Robert-Koch-Straße 4, Marburg 35037, Germany. E-mail: shuming.li@staff.uni-marburg.de
cSavaid Medical School, University of Chinese Academy of Sciences, Beijing 100049, P. R. China
dState Key Laboratory of Bioactive Substance and Function of Natural Medicines, Institute of Materia Medica, Chinese Academy of Medical Sciences, Peking Union Medical College, Beijing 100050, P. R. China
First published on 8th February 2021
Epidithiodiketopiperazines (ETPs) are a group of bioactive fungal natural products and structurally feature unique transannular disulfide bridges between α, α or α, β carbons. However, no enzyme has yet been demonstrated to catalyse α, β-disulfide bond formation in these molecules. Through genome mining and gene deletion approaches in Trichoderma hypoxylon, we identified a putative biosynthetic gene cluster of pretrichodermamide A (1), which requires a FAD-dependent oxidoreductase, TdaR, for the irregular α, β-disulfide formation in 1 biosynthesis. In vitro assays of TdaR, together with AclT involved in aspirochlorine and GliT involved in gliotoxin biosynthesis, proved that all three enzymes catalyse not only the conversion of red-pretrichodermamide A (4) to α, β-disulfide-containing 1 but also that of red-gliotoxin (5) to α, α-disulfide-containing gliotoxin (6). These results provide new insights into the thiol-disulfide oxidases responsible for the disulfide bond formation in natural products with significant substrate and catalytic promiscuities.
Previous studies have elucidated many aspects of the molecular mechanism involved in the biosynthesis of gliotoxin from Aspergillus fumigatus, including cyclo-L-Phe-L-Ser generation by an NRPS enzyme GliP, hydroxylation by cytochrome P450, sulfur moiety incorporation by the addition of glutathione (GSH) and its sequential degradation to the dithiol precursor.5,29,30 A flavin adenine dinucleotide (FAD)-dependent oxidoreductase GliT is responsible for the oxidation of the dithiol group yielding the α, α disulfide-containing gliotoxin.31,32 In addition, biosynthesis of several ETPs including sirodesmin PL from Leptosphaeria maculans and acetylaranotin from Aspergillus terreus was extensively studied by multiple genetic and biochemical approaches.12–15 The halogenated aspirochlorine was also investigated with respect to its crucial roles in the formation of a cyclo-L-Phe-Gly containing ETP from cyclo-L-Phe-L-Phe.24 Although the interesting α, β-disulfide bridge formation was speculated to be catalysed by a GliT homolog, AclT, no experimental evidence was provided.
To identify the enzyme involved in α, β-disulfide bridge formation, we focused on 1 reported in Trichoderma sp. BCC5926, which exhibits antibacterial activity towards Mycobacterium tuberculosis H37Ra.19 The typical α, β-disulfide bridge and 1,2-oxazadecaline moiety of 1 as well as its derivatives have shown to be essential for the observed biological activities.21,22 A biosynthetic route was proposed for gliovirin, a derivative of 1, in Trichoderma virens, based on genome mining and core gene deletion experiments.21 It was speculated that the FAD-dependent oxidoreductase Glv4 or Glv16 was possibly involved in the α, β-disulfide bridge formation. However, data for neither its detailed biosynthesis nor for the α, β-disulfide formation was reported. In this study, we identified 1 from a fungicolous fungus Trichoderma hypoxylon and proposed a possible biosynthetic pathway by genome mining and deletion of two targeted genes. Biochemical investigations proved that a FAD-dependent oxidoreductase, TdaR, was responsible for the α, β-disulfide formation in the biosynthesis of 1. Further in vitro assays demonstrated that three oxidases TdaR, AclT and GliT from different pathways catalysed both α, α- and α, β-disulfide bond formation in the biosynthesis of fungal ETP alkaloids.
To determine the biosynthetic pathway of 1, the genome of T. hypoxylon was sequenced and the draft genome sequence was used for the prediction of putative gene clusters by using AntiSMASH.34 Based on its ETP structural character, 1 should be derived from cyclo-L-Phe-L-Phe (3), which is expected to be assembled by an NRPS.35 One of the eleven NRPS genes T_hypo_11188, termed tdaA, within a large 49.6 kbp cluster (Fig. 2A and Table S4†), contains a T–C–A–T–C domain structure, being similar to the core enzymes AtaP15 and AclP24 in phenylalanine-containing ETP biosynthesis (abbreviations for NRPS domains are as given before35). TdaA shares a sequence similarity of 31% with GliP9 from A. fumigatus and 88% with Glv2121 from T. virens on the amino acid level (Table S4†). To prove its function, tdaA was deleted from the genome by PEG-mediated protoplast transformation using hygromycin B as a selection marker.36 LC-MS analysis revealed complete disappearance of ETP products 1 and 2, as well as the NRPS product 3, indicating that TdaA acts as the core enzyme for the biosynthesis of 1 and 2 (Fig. 2B). To find the candidate gene for the α, β-disulfide bridge formation in 1, further blast analysis was performed using the disulfide bond formation enzyme GliT in the gliotoxin biosynthetic pathway as a probe.32 Two putative FAD-dependent oxidoreductases TdaE (T_hypo_11193) and TdaR (T_hypo_11206) are located in the pretrichodermamide A (tda) cluster. Alignments with known oxidases for thiol-disulfide formation showed the existence of a conserved active binding site CLFC (CXXC) box in TdaR but not in TdaE (Fig. S4†). Therefore, we performed tdaR deletion according to the aforementioned method for tdaA, leading to complete disappearance of 1 and 2, but accumulation of 3 as indicated by the analysis of extracted ion chromatograms (EICs) (Fig. 2B). This proved the involvement of TdaR in the conversion of 3 to 1 and 2. Inspection of the EICs of the culture extracts and structure analysis implied more conversion steps between 3 and 1. Thus, a biosynthetic pathway of 1 and 2 was proposed as shown in Fig. 2C. Briefly, the DKP core is assembled by NRPS TdaA (i), followed by hydroxylations at α and β positions by a putative cytochrome P450 TdaS (ii). After sulfur incorporation and sequential degradation (iii), TdaR catalyses the α, β-disulfide bond formation (iv). Subsequently, further hydroxylations catalysed by multiple cytochrome P450s and methylations catalysed by different methyltransferases (MeTs) lead to the formation of 1 with a 1,2-oxazadecaline moiety and methoxy groups (v). Lastly, spontaneous degradation results in the transformation of 1 to 2 (vi).
Fig. 2 Pretrichodermamide A (1) biosynthetic gene cluster (A), LC-MS analysis of T. hypoxylon strains (B) and the proposed biosynthetic pathway for 1 (C). |
To investigate its function, the 933 bp long coding sequence of tdaR comprising three exons was cloned from cDNA of T. hypoxylon and overexpressed in E. coli BL21(DE3) cells. The recombinant His6-tagged protein was purified with the aid of Ni-NTA agarose resin to near homogeneity which was confirmed on SDS-PAGE (Fig. 4A), yielding 5.1 mg of purified TdaR per liter of bacterial culture. Due to the instability of the unidentified substrate of TdaR for the α, β-disulfide bond formation, we proposed that this enzyme could also catalyse the conversion of reduced pretrichodermamide A (red-pretrichodermamide A, 4) to 1. Therefore, 1 was chemically reduced to 4 using dithiothreitol (DTT) to provide an appropriate substrate for the enzyme assays (Fig. S6†). HPLC analysis revealed a complete conversion of 0.5 mM 1 to 4 by 1 mM DTT at 30 °C for 30 min (Fig. 5A and S6†). Subsequently, this mixture was directly incubated with 0.1 μM purified TdaR at 37 °C for 30 min. As expected, the enzyme assay led to the formation of 1 which was verified by comparing the retention time with that of an authentic standard (Fig. 5A(i, ii and viiii)). This proved that the FAD-dependent oxidoreductase TdaR was responsible for the α, β-disulfide bond formation in the biosynthesis of 1. Furthermore, the time course of this reaction was examined and monitored by HPLC, indicating that 17% of 0.5 mM 4 was consumed by 0.1 μM TdaR in 5 min (Fig. S7†). However, no conversion of 4 could be observed with heat-denatured TdaR or without O2 in the glove box (Fig. 5A(i and iii)). This indicated a mechanism of α, β-disulfide bond formation similar to GliT-catalysed α, α-disulfide bond formation, which required O2. Accordingly, we proposed that initial transformation occurred in TdaR from the oxidized form (disulfide) to reduced form (dithiol) by the putative CLFC redox system. Subsequently, electron pairs were transferred to FADox from the active site of TdaR. FADred is then reoxidized by O2 to form H2O2, leading to the concomitant production of the disulfide-containing product (Fig. 4B and 6).
Fig. 4 Analysis of TdaR, AclT and GliT on SDS-PAGE (A) and enzyme-mediated disulfide bond formation in the biosynthesis of 1 and 6 (B). |
Fig. 5 HPLC analysis of enzyme assays of TdaR, AclT and GliT with 4 (A) or 5 (B). UV absorption at 254 nm was illustrated. |
To directly confirm the oxidation of TdaR on 4, 4 was purified from the mixture of 1 with tris(2-carboxyethyl)phosphine hydrochloride (TCEP) as a reducing agent on a semi-preparative HPLC (acetonitrile/H2O, 35:65) (Fig. S9†). Incubation of TdaR with the purified 4 revealed a complete conversion of 4 to 1 after incubation at 37 °C for 30 min (Fig. S10A†). However, the formation of 1 was also observed in the negative control containing heat-denatured TdaR and 4, demonstrating the spontaneous disulfide formation. In comparison, disulfide formation in 1 catalysed by TdaR was more efficient than the spontaneous oxidation.
Notably, bioinformatics study showed that TdaR and AclT involved in the α, β-disulfide bond formation were located in clade IV with several known α, α-disulfide bond-forming enzymes including GliT,9,32 AtaTC15 and SirT12 (Fig. 3 and S5†). The close phylogenetic relationship triggered our interest to test if TdaR and AclT could also catalyse α, α-disulfide bond formation. Thus, reduced gliotoxin (red-gliotoxin, 5) was acquired by co-incubation of gliotoxin (6) with DTT at 30 °C for 30 min or purified from a mixture containing 6 and TCEP (Fig. S6 and S9†). Interestingly, we observed that 5 (0.5 mM) was readily converted into 6 in the presence of TdaR (0.1 μM) or AclT (0.1 μM) after 5 min (Fig. S8†). 5 was completely transformed to 6 in 30 min by both TdaR and AclT (Fig. 5B, S8 and S10B†). These results supported our hypothesis that TdaR and AclT catalyse not only α, β- but also α, α-disulfide bond formation in the biosynthesis of ETP alkaloids. In addition, TdaR and AclT-catalysed α, α-disulfide bond formation with 5 as the substrate was more efficient than the α, β-disulfide bond formation.
In analogy, the coding sequence of GliT, which was proved to catalyse the α, α-disulfide bond formation in the biosynthesis of 6, was cloned into pET28a(+)SUMO and overexpressed in E. coli BL21(DE3). 11.2 mg purified GliT was obtained in a yellow color from 1 liter of bacterial culture and subsequently incubated with 4 and 5 in a similar way to TdaR and AclT. In accordance with the previous study, GliT acts as a FAD-dependent oxidase for the α, α-disulfide bond formation to generate 6 from 5 after 5 min incubation (Fig. 5B and S8†).32 Unprecedentedly, incubation of 4 with GliT also led to the formation of 1, as demonstrated by HPLC analysis (Fig. 5A(vi and vii)). This confirmed that the known α, α-disulfide bond related enzyme GliT also catalysed the α, β-disulfide bond formation. The time course of GliT-catalysed oxidation of 4 was monitored by HPLC, revealing that an increasing amount of 4 ranging from 28–82% was consumed by 0.1 μM GliT from 5 min to 30 min (Fig. S7†).
In this study, we identified an α, β-disulfide-containing metabolite 1, together with its spontaneous desulfurized product 2, in T. hypoxylon. Genome mining led to the identification of a tda cluster encoding analogues of all aforementioned gliotoxin (gli) cluster enzymes and five additional cytochrome P450s as well as three methyltransferases (Fig. 2A and Table S4†).5 Deletion of the NRPS coding gene tdaA and the FAD-dependent oxidase gene tdaR proved their involvement in the biosynthesis of 1 and 2 from the CDP 3 (Fig. 2B and C). Plausibly, dihydroxylation of 3, α, β-disulfide formation and subsequent hydroxylations as well as methylations will yield the ETP alkaloid 1. Further biochemical investigation of TdaR provided evidence for its involvement in the metabolism of 3 to 1 and proved its crucial role in the formation of the α, β-disulfide bridge (Fig. 4B). Thus, our results first represent an example for understanding the biosynthesis of α, β-disulfide-containing ETP alkaloids, especially the α, β-disulfide formation. Other pretrichodermamide A-like compounds derived from cyclo-L-Phe-L-Phe, including pretrichodermamides,22 gliovirin,21 FA2097,47 aspergillazines,48 peniciadametizines49 and aspirochlorine,24 are very likely also formed in a similar way regarding the installation of the typical α, β-disulfide and a 1,2-oxazadecaline moiety. Therefore, further attempts to mine α, β-disulfide related enzymes would provide evidence for our hypothesis.
Previous study on aspirochlorine biosynthesis has already implied a putative α, β-disulfide related oxidase AclT.24 Herein, we proved that AclT is an analogous oxidase of TdaR catalysing the α, β-disulfide formation. Further biochemical investigations on TdaR and AclT revealed, interestingly, their dual function of catalysing not only the α, β-disulfide formation in 1 but also the α, α-disulfide formation in 6 (Fig. 4B and 5). These results indicated that α, β-disulfide related enzymes also catalysed α, α-disulfide formation, and vice versa. Indeed, enzyme assays confirmed that the α, α-disulfide related GliT was also able to catalyse α, β-disulfide formation in 1. It can be deduced that hydroxylations on the CDP core catalysed by P450 enzymes determine the positions of the two hydroxyl groups and finally the type of the disulfide bond (α, α or α, β).
Taken together, TdaR, AclT and GliT are dual oxidases independently forming the α, α- or α, β-disulfide bridge in fungal ETP biosynthesis. It might well explain why most fungal related FAD-dependent oxidases were located in the same clade IV (Fig. 3), no matter they naturally catalyse α, α- or α, β-disulfide formation. In the phylogenetic analysis of TdaR homologs, the unique clade III, exemplified by TrxR from E. coli, catalysed the reduction of a disulfide bond in thioredoxin.41 In contrast to DepH (clade I) from C. violaceum and HlmI (clade II) from S. clavuligerus involved in intra- or interchenar disulfide bonds, TdaR, AclT and GliT (clade IV) catalysed transannular disulfide bond formation.39,40 They shared a catalytically active CLFC box for redox activation, but different FAD binding sites, i.e. His139 of TdaR and AclT, but Asp139 in GliT (Fig. S4†). Plausibly in analogy to GliT-catalysed disulfide bond formation, fungal oxidases utilized a similar mechanism to install disulfide bonds across α, α- or α, β-positions (Fig. 6).32 A sulfhydryl group in the DKP-containing substrate attacks a disulfide bond in the thioredoxin motif (CXXC) within the oxidoreductase, creating a transient mixed-disulfide bond between the substrate and oxidoreductase (Fig. 6A and B). In this step, FADox receives the electrons at the C4a position, leading to a FAD-4a-thiol adduct. Subsequently, it undergoes attack at the other sulfhydryl group with disulfide formation in the ETP alkaloid and concomitant reduction to FADred (Fig. 6C and D). Eventually, FADred is reoxidized by molecular oxygen instead of NADP+ with the elimination of H2O2. We speculated that the large pocket across the active site and FAD-binding motif of GliT as well as TdaR and AclT50 might result in the broad substrate specificity, making it attractive for further enzyme structure analysis and protein engineering.37
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0sc06647h |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2021 |