Reconstitution of the early steps of gliotoxin biosynthesis in Aspergillus nidulans reveals the role of the monooxygenase GliC

Abstract

The gliotoxin, a member of the epipolythiodioxopiperazine (ETP), has received considerable attention from the scientific community for its wide range of biological activity. Despite the identification of gliotoxin cluster, however, the sequence of steps in the gliotoxin biosynthesis has remained elusive. As an alternative to the gene knock out and biochemical approaches used so far, here we report using a heterologous expression approach to determine the sequence of the early steps of gliotoxin biosynthesis in A. nidulans. We identified the GliC, a monooxygenases that involved in the second step of gliotoxin biosynthesis pathway through the catalyzing the hydroxylation at the α-position of l-Phe.

Gliotoxin belongs to a class of fungal secondary metabolites produced by nonribosomal peptide synthetase (NRPS) called epipolythiodioxopiperazine (ETP). ETPs are toxins that are characterized by the presence of the transannular disulphide bridge which is the source of their biological activity. Extensive studies have revealed that this unique structure motif primarily imparts various properties of gliotoxin such as antiviral,^{1, 2} antibacterial³ and immunosuppresssive^{4, 5} activities. The gliotoxin biosynthesis gene cluster has been identified in the human pathogen Aspergillus fumigatus, however, the exact sequence of steps in gliotoxin biosynthesis has remained unclear.⁶ Various pathways for gliotoxin biosynthesis have been proposed by different groups using either knock-out approaches in A. fumigatus targeting gliP,⁷ gliT,⁸ gliZ,⁹ gliG,^{10, 11} gliH,¹² gliK¹³ or gliI¹⁴ genes and by biochemical approaches. However to date, the role of other genes in gliotoxin biosynthesis remains unclear. As an alternative to the gene knock out and biochemical approaches used so far, we used a bottom up heterologous expression approach using A. nidulans an organism not known to produce gliotoxin as host to reconstitute the early steps of gliotoxin biosynthesis.

Works from the Walsh group have definitively demonstrated that the first step in gliotoxin biosynthesis is the formation of the cyclo-l-phenylalanyl-l-seryl by the NRPS, GliP.¹⁵ As the first step in our approach, we carried out the heterologous expression of the A. fumigatus gliP in A. nidulans under the control of the alcA promoter that can be induced strongly using cyclopentanone.¹⁶ We amplified the gliP gene and pyroA selective marker by PCR from A. fumigatus strain AF293. The alcA promoter and the two regions flanking the yA gene were amplified from the genomic DNA of A. nidulans strain LO2026.¹⁷ LO2026 is an A. nidulans strain where the FAS gene stcJ in the sterigmatocystin pathway has been deleted, thus will not produce major metabolite sterigmatocystin. This recipient strain also contains an nkuA mutation that greatly increases homologous recombination rate. Due to the large size of the gliP gene we split the gene into two incomplete fragments with 1 kb overlapping region. The first gliP fragment was fused to the upstream yA-flanking sequence-linked with the alcA promoter and the second gliP fragment was fused to the downstream yA-flanking sequence-linked with the pyroA marker. These constructs were generated by fusion PCR (Table S1) and characterized by DNA sequence analysis which is then subsequently used for A. nidulans transformation. The transformants with the correct insertion were further verified with diagnostic PCR (Fig. S1). In each case at least two transformants carrying the correct insertion were used for further study. We cultivated the alcA-gliP A. nidulans strain (CW2129, Table S2) in LMM medium (15 g/l lactose, 6 g/l NaNO₃, 0.52 g/l KCl, 0.52 g/l MgSO₄·7H₂O, 1.52 g/l KH₂PO₄, 1 ml/l trace element) supplemented with uracil (1 mg/ml) and uridine (10 mM) for 18 hours and then induced with cyclopentanone (10 mM) for another 48 hours. After 48 hours of incubation the medium was extracted with ethyl acetate followed by evaporating in vacuo. The residue was resuspended in 0.5 ml of 20% DMSO/MeOH and a portion (10 µl) was examined by HPLC-DAD-MS. We detected two UV active peaks with masses of 235 and 274 (m/z[M+H]⁺) (Fig. 1).

Figure 1.

(A) HPLC-DAD-MS analysis of organic extracts from: recipient (stcJΔ); GliP-expressing (stcJΔ, alcA-gliP); GliP- and GliC-expressing (stcJΔ, alcA-gliP, alcA-gliC); and GliP- and GliF-expressing (stcJΔ, alcA-gliP, alcA-gliF) A. nidulans strains. (B) Structures of compound 1–4.

Characterization of the purified molecules from scaled-up 1 liter culture by NMR revealed that compound 1 is cyclo-(lphenylalanyl-l-seryl) and compound 2 is cyclo-(l-tryptophanyl-l-seryl) (Fig. 1). The diketopiperazine 1 has been identified from a biochemical study using purified recombinant GliP and is an expected product from the pathway.¹⁵ The production of the diketopiperazine 2¹⁸ however was unexpected and suggests that the substrate specificity of the first adenylation (A) domain in GliP is sufficiently relaxed to accommodate both l-Phe and lTrp.

After the first step of gliotoxin biosynthesis by GliP, the subsequent steps are unclear. Intermediates isolated from a gliG, glutathione synthase knock out mutant and gliZ transcription factor knock out mutant of A. fumigatus revealed a series of intermediates oxidized at the α carbon of l-Phe suggesting that the next step in gliotoxin involves one of the two cytochrome P450 monooxygenases, GliC or GliF.^{9, 10} To identify which of the monooxygenase is responsible for the oxidation at the α-carbon of l-Phe, we created an alcA-gliP/alcA-gliC and an alcA-gliP/alcAgliP/alcA-gliF mutant strains using the same strategy we used to create gliP. We cultured both alcA-gliP/alcA-gliC (CW2142) and alcA-gliP/alcA-gliF (CW2221) A. nidulans mutant strains separately and examined the extracts from the two strains by HPLC-DAD-MS. Only in the alcA-gliP/alcA-gliC (CW2142) strain did we observe the disappearance of compound 1 and the emergence of a new peak (compound 3) with a mass of 247 (m/z[M-H]⁻) (Fig. 1). Compound 2 can still be detected in the alcA-gliP/alcA-gliC (CW2142) strain, suggesting that only cyclo-(l-phenylalanyl-l-seryl) is a substrate for GliC (Fig. 1). The alcA-gliP/alcA-gliF (CW2221) strain has similar metabolite profile as the single alcA-gliP strain (Fig. 1) suggesting that gliF is involved in the latter parts of the gliotoxin biosynthesis pathway if at all.

To characterize the new metabolite 3, we scaled up the culture of the alcA-gliP/alcA-gliC (CW2142) strain and isolated the metabolite by column chromatography followed by preparative HPLC. Spectroscopic analysis of compound 3 (Table 1) revealed that compound 3 isolated from this study is structurally quite similar to the shunt product 4 isolated by Davis et al and Scharf et al from a gliG knockout mutant.^{10, 11} The two molecules differ only in the functional group attached to N1. In our molecule it is a hydroxy group while in the Davis et al molecule a methoxy group is attached to N1. The enzyme responsible for catalyzing the methoxy group is currently unknown although the gliotoxin cluster contains two methyltransferases GliM and GliN which are possible candidates. In our experiments, neither GliM nor GliN have been reconstituted in A. nidulans and could explain the lack of methylation in the substrate we isolated.

Table 1.

NMR data comparison between compound 4 published from Davis et al and scharf et al. (600 and 150 MHZ)^a and compound 3 isolated from this study (400 and 100 MHZ)^b.

	4 (CDCl3)			3(CD3OD)
position	δC		δH	δC		δH
2	157.3		—	160.2		—
3	132.2		—	N.O.c		—
5	161.3		—	162.4		—
6	82.4		—	84.0		—
7	102.6		5.11, 5.57 (each 1H, S)	101.9		5.26, 5.34 (each 1H, brs)
8	47.6		3.18, 3.33 (each 1H, d, J = 13.5 HZ)	47.1		2.94, 3.52 (each 1H, br d, J = 13.5 HZ)
9	132.6		—	135.7		—
10 and 14	130.7		7.19–7.20 (2H, m)	131.8		7.19 (5H, br s)
11 and 13	128.8		7.24–7.27 (3H,m)	129.5
12	128.0			128.3
OCH3	62.0		3.71 (3H,s)	—		—

^apublished data obtained from Davis et al and Scharf et al.^10,15.

^bData obtained from this study.

^cNot observed.

From the structure of compound 3, we propose that the second step in the gliotoxin biosynthesis pathway is the hydroxylation at the α-position of l-Phe and this is catalyzed by GliC (Scheme 1). Our data is consistent with the work by Forseth et al. Where oxidation at the α-position is proposed to occur at an early state in the biosynthesis.⁹ The structure of compound 3 suggests that the molecule we isolated is probably a stable shunt metabolite that is formed because we have not reconstituted additional genes in the pathway.

Scheme 1.

Proposed biosynthetic pathway for gliotoxin. Following the condensation of l-Phe and l-Ser by GliP, the oxidation of the α carbon of l-Phe was catalyzed by GliC. A series of subsequent reactions including a GliT-mediated tailoring reaction result in gliotoxin formation. Compound 3 was isolated as a stable shunt metabolite following co-expression of GliP and GliC. ^aCompound 5 is a hypothetic product which has not been detected in HPLC-DAD-MS.

From our approach, it is not clear the exact nature of the substrate of GliC. It is possible that the free standing diketopiperazine 1 is an intermediate although we can not exclude the other possibility that GliC recognizes the substrate still tethered to GliP. Additional studies will be necessary to clarify the two competing mechanisms.

In summary, using a heterologous expression technique we demonstrated that one of the two cytochrome P450 monooxygenase, GliC, is responsible for the oxidation of the α-carbon of l-Phe, and is the subsequent step after the condensation of l-Phe and l-Ser by GliP in gliotoxin biosynthesis. Further investigation of other components within the gliotoxin cluster will not only provide for intriguing mechanistic studies but also lay a foundation for the utilization of the gliotoxin pathway to generate new ETP molecules.