Diketopiperazine Formation in Fungi requires Dedicated Cyclization and Thiolation Domains

Abstract

Cyclization of linear dipeptidyl precursors derived from nonribosomal peptide synthetases (NRPSs) into 2,5-diketopiperazines (DKPs) is a crucial step in the biosynthesis of a large number of bioactive natural products. However, the mechanism of DKP formation in fungi has remained unclear, despite extensive studies of their biosyntheses. Here we show that DKP formation en route to the fungal virulence factor gliotoxin requires a seemingly extraneous couplet of condensation (C) and thiolation (T) domains in the NRPS GliP. In vivo truncation of GliP to remove the CT couplet or just the T domain abrogated production of gliotoxin and all other gli pathway metabolites. Point mutation of conserved active sites in the C and T domains diminished cyclization activity of GliP in vitro and abolished gliotoxin biosynthesis in vivo. Verified NRPSs of other fungal DKPs terminate with similar CT domain couplets, suggesting a conserved strategy for DKP biosynthesis by fungal NRPSs.

Graphical Abstract

Seemingly extraneous: Diketopiperazines derived from non-ribosomal peptide synthetases, an important family of fungal virulence factors, do not form spontaneously, as presumed; instead cyclization relies on previously unannotated domains of the synthetase.

NRPS-derived DKPs form a large class of natural products with diverse biological activities.^[1–5] Gliotoxin (1) is the best-known member of the epipolythiodiketopiperazines (ETPs), a family of toxic DKPs produced by a variety of filamentous fungi.^[1–5] The biosynthesis of 1 has been extensively studied, due to its significant contribution to the virulence of the devastating human pathogen, A. fumigatus^[6–12] and as a model system for fungal NRPS pathways. Production of 1 is accomplished via the gli biosynthetic gene cluster (BGC) in A. fumigatus and related fungi (Figure 1a). The 13-gene gli BGC encodes the transcriptional regulator, GliZ, one transporter (GliA), several backbone tailoring enzymes, and the core NRPS, GliP (Figure 1a).^[13] Whereas, the majority of gli-cluster tailoring enzymes have previously been characterized extensively in vitro and in vivo (blue arrows, Figure 1a), functional analysis of the core enzyme, GliP, is incomplete (Figure 1b).^[14]

Figure 1.

Gliotoxin biosynthesis in A. fumigatus. (a) gli-cluster gene annotations, including characterized tailoring enzymes (blue), tailoring enzymes with inferred functions (grey), and genes without predicted function (black). gliZ: transcription factor; gliI: pyridoxal phosphate dependent desulfurase; gliC, gliF: cytochrome P450 oxidases; gliM: O-methyltransferase; gliG: glutathione-S-transferase; gliK: glutamate cyclase; gliA: transporter; gliN: N-methyltransferase; gliT: oxidase; (b) Putative function of GliP and abbreviated biosynthesis of 1 showing the most abundant intermediates or shunt metabolites 2 and 3, as well as the detoxification product, 4.^{[6,9,16–19]}

Homology-based annotation of GliP uncovers two adenylation (A), two condensation (C), and three thiolation (T) domains, referred to as A₁-T₁-C₁-A₂-T₂-C₂-T₃.^[14] Previous work showed that recombinant GliP converts L-Phe and L-Ser into the DKP cyclo(L-Phe-L-Ser) (2), which was speculated to result from spontaneous cyclization of a GliP-T₂-tethered-L-Phe-L-Ser intermediate (Figure 1b).^[7,14] Although 2 could plausibly be derived from non-enzymatic cyclization of T₂-tethered L-Phe-L-Ser, the presence of the C₂ and T₃ domains, which in this model have no function, may suggest an alternative mechanism. In 2012, Gao et al. reported that macrocyclization of linear peptidyl precursors produced by a variety of fungal multi-module NRPSs is catalyzed by conserved terminal condensation-like (C_T) domains,^[15] and that C_T domain activity was dependent on a conserved histidine within the amino acid sequence SHXXXDXXS/T. We noted that this C_T-conserved sequence exists in the GliP-C₂ domain (Figure S1), suggesting that the C₂ domain in GliP may be involved in cyclization of a tethered L-Phe-L-Ser dipeptide (Figure 1b). We further hypothesized that the seemingly extraneous T₃ domain also plays a role in DKP formation.

To investigate the function of the putative GliP-C_T and the GliP-T₃ domains in vivo, we constructed two truncation mutants in A. fumigatus, GliP-ΔT₃ and GliP-ΔC_TT₃, and assessed the impact of these truncations on the biosynthesis of 1 and gli-pathway metabolites (1–4, Figure 2a–d). Comparison of whole-metabolome extracts from WT (Af293), GliP-ΔT₃, and GliP-ΔC_TT₃ revealed a complete loss of 1–4 in the GliP-ΔC_TT₃ strain (Figure 2a–d). In extracts from the GliP-ΔT₃ strain we were unable to detect 1 and the most abundant gli-pathway shunt metabolites 2 and 3, however, we observed trace quantities of 4 (99 % less than Af293), the detoxification product of 1 (Figure 2a–d).^[20] These data suggested that normal biosynthesis of 1 requires the T₃ domain of GliP.

Figure 2.

LC-HRMS analysis of whole-metabolome extracts from WT and mutant A. fumigatus strains. (a-d) Representative overlaid extracted-ion-chromatograms (EICs) for 1–4 in Af293 (black), GliP-ΔT₃ (green), and GliP-ΔC_TT₃ (purple). Darker lines representative overlaid EICs for 1–4 in the indicated strains with added exogenous 1. “10 x” indicates scaling of weak signals applied for comparison. “*” = unrelated peaks.

Gliotoxin (1) serves as a positive feedback loop for its own production through regulation of gli-cluster expression.^[21] Accordingly, the inability of the GliP-ΔC_TT₃ mutant strain to produce 1 resulted in almost complete loss of gli-cluster expression (Figure S2), which could have affected our results. Therefore, we repeated the experiment by growing cultures supplemented with exogenous 1, which largely rescued gli-cluster expression (Figure S3).^[21] LC-HRMS comparison of extracts from gliotoxin-supplemented cultures showed that rescue of cluster gene expression did not recover production of gli-pathway metabolites. As in the case without supplementation of 1, the GliP-ΔC_TT₃ mutant did not produce any 1–3 (Figure 2a–d), whereas the GliP-ΔT₃ strain produced small quantities of 2 (96 % less than Af293) and the tailored metabolite, 3 (96 % less than Af293) (Figure 2a–d). These results indicate that normal biosynthesis of 1 requires the T₃ domain of GliP, whereas in the absence of T₃, only small amounts of 1 and other gli-pathway metabolites are produced, possibly via cyclization of a T₂-tethered intermediate.

We further considered the possibility that the GliP-ΔC_TT₃ and GliP-ΔT₃ variants may have reduced adenylation activity and do not load L-Phe and L-Ser as efficiently as the WT GliP. To address this concern, we heterologously produced and purified GliP-WT, GliP-ΔC_TT₃ and GliP-ΔT₃ proteins from E. coli (Figure S4) and conducted an ATP-[³²P]pyrophosphate exchange assay. We found that adenylation activity of the A₁- and A₂ –domains is not reduced in the truncated proteins, GliP-ΔC_TT₃ and GliP-ΔT₃, relative to GliP-WT (Figure S5).^[14,22] Notably, the heterologously expressed GliP truncations remained largely functional. Production of cyclo-Phe-Ser (2) was only slightly reduced with the GliP-ΔT₃ mutant and still about one third of WT with the GliP-ΔC_TT₃ mutant (Figure S6). Residual production of 2 by these mutants likely results from spontaneous or C_T-catalyzed cyclization of the T₂-tethered L-Phe-L-Ser (Figure 1b), which does not appear to occur in vivo in the corresponding A. fumigatus mutants (compare Figures 2a and S6).

Taken together, these results indicate that the GliP-T₃ domain is required for DKP formation in vivo, since in the absence of T₃, only small amounts of 1 and other gli-pathway metabolites are produced, possibly via cyclization of a T₂-tethered intermediate. Therefore, we hypothesized that in full length GliP the L-Phe-L-Ser dipeptide is transferred from the T₂ domain to T₃ prior to cyclization via C_T.

To assess whether the T₃ domain is functional, we tested for appropriate post-translational decoration of the predicted active site residue, serine 2095, with a phosphopantetheinyl-(ppant-) moiety. LC-HRMS/MS analysis of tryptic digests of GliP-WT showed diagnostic ppant-fragmentation of the peptide containing S2095, confirming ppant attachment to the active site serine in T₃ (Figure S7).^[24]

To further probe the roles of the T₃ and C_T domains for DKP formation, we constructed A. fumigatus strains carrying point mutations at the T₃ and C_T active sites. For this purpose, we first created a new ΔGliP strain, then reintroduced either GliP(WT), a point mutant of the T₃ active site, GliP-S2095A, or a point mutant lacking the putative catalytic histidine of the C_T domain, GliP-H1754A (see Figure S2 and Supporting Methods for details). The ΔGliP+GliP(WT) strain produced a distribution of pathway metabolites (Figure 3a) very similar to that of WT A. fumigatus, proving that the reintroduced GliP is fully functional. In contrast, production of gli pathway metabolites was almost completely abolished in the S2095A and H1754A mutant strains (Figure 3b,c). To rescue potentially suppressed expression of gli-cluster genes in the absence of gliotoxin (1), we repeated the experiment with exogenously added 1, which resulted in a production of a small amount (<2% of WT) of cyclo-Phe-Ser in the S2095A mutant while cyclo-Phe-Ser remained undetectable in the H1754A mutant (Figure 3b,c).

Figure 3.

LC-HRMS analysis of whole-metabolome extracts from mutant A. fumigatus strains. Representative EICs for compounds 1, 2 and 4 in (a) A. fumigatus ΔGliP +GliP-WT, (b) A. fumigatus ΔGliP +GliP-S2095A, and (c) A. fumigatus ΔGliP +GliP-S2095A. Dashed chromatograms representative EICs for 1, 2, and 4 in the indicated strains with added exogenous gliotoxin, 1. Note that the extent of conversion of gliotoxin (1) to its detoxification product 4 is highly variable between experiments, e.g. in the example chromatogram shown in (a) no 1 remains following addition of 1.

To confirm the essential role of the GliP-T₃ domain, we heterologously expressed a GliP mutant protein in which the first and second T-domains were inactivated by substitution of serines 555 and 1582 for alanine (GliP-ΔT₁T₂). We then used synthetic L-Phe-L-Ser-N-acetylcysteamine (L-Phe-L-Ser-SNAC, 5) to effect attachment of L-Phe-L-Ser to T₃ via transthiolation and monitored production of 2 (Figures 4a and S8a). Since thioester 5 could cyclize non-enzymatically or via catalysis by the C_T domain without first getting attached to T₃ (Figure S8b), we included control assays without GliP, without Sfp, or without coenzyme A (= no ppant functionalization on T₃), and using a GliP mutant protein in which all three T-domains were inactivated (GliP-ΔT₁T₂T₃). Incubation of 5 under these conditions (Figure 4a) demonstrated that the presence of a functional T₃ domain significantly increases the cyclization activity of GliP.

Figure 4.

(a) In vitro production of cyclo-Phe-Ser (2) using different GliP variants. Relative yields of 2 were measured via integration of the corresponding peak in LC-MS ion-chromatograms (n = 4). **p < 0.01, ***p < 0.001. (b) Model for C_TT_C-catalyzed DKP formation in gliotoxin (1) biosynthesis.

Our results demonstrate that cyclization of tethered Phe-Ser dipeptide en route to 1 is not spontaneous, as proposed,^[7–11] and rather requires two additional GliP domains, the second condensation-like domain (C_T domain) and the enigmatic terminal thiolation domain T₃ (“T_C” domain) to which the nascent dipeptide appears to be transferred prior to cyclization via C_T (Figure 4b). Significantly, mutation of the catalytic histidine in the C_T domain or the ppant attachment site within the T₃ domain is sufficient to almost completely abrogate gliotoxin biosynthesis. These observations highlight the importance of dedicated cyclization domains in fungal NRPSs.^[25,26] Gliotoxin (1) is a representative member of a large class of NRPS-derived DKPs that seem likely to be produced via similar mechanisms. Analysis of available fungal genomes revealed 56 putative NRPSs that feature terminal domains homologous to the C_TT₃ tandem in GliP (see Supporting Methods, Table S1, and Figure S9), indicating a conserved biosynthetic strategy for DKP formation.^[13,27]

Whereas the GliP C_T domain is homologous to the recently described C_T domains that catalyze cyclization of fungal tripeptides;^[15] the requirement of an additional T_C domain for DKP formation is perhaps unexpected, given that dipeptide thioesters often cyclize non-enzymatically, as we showed in our in vitro studies. We note that T_C domains in DKP-producing NRPSs could serve as a tether for linear dipeptides during tailoring by other cluster enzymes. Although there is substantial evidence that, in the case of gliotoxin, many of the later steps of its biosynthesis proceed via untethered, cyclic intermediates, the structures of shunt metabolites in related DKP biosynthesis pathways may suggest tailoring of tethered dipeptides. For example, in the case of hexadehydroastechrome, which is derived from cyclo-Trp-Ala derivatives, abundant production of prenylated tryptophan in mutants defective in late-stage tailoring enzymes could be due to recycling of a tethered prenylated Trp-Ala dipeptide (Figure S10).^[28]

In conclusion, our study suggests a general framework for fungal DKP biosynthesis, wherein the T_C domain serves as a tether for dipeptide cyclization by an adjacent C_T domain and potentially for tailoring by other cluster enzymes. Furthermore, our characterization of the biosynthetic roles of the C_T and T_C domains in DKP formation extends the functional repertoire of NRPS domains.