Homologous NRPS-like Gene Clusters Mediate Redundant Small-Molecule Biosynthesis in Aspergillus flavus

Fungi are among the most prolific sources of pharmacologically relevant natural products.¹ This large diversity of fungal small molecules serves important functions in fungal ecology, for example as virulence factors and as chemical defense agents. However, only a fraction of the biosynthetic capabilities suggested by genomic analyses has been observed under laboratory conditions, because expression of many, perhaps even most biosynthetic pathways depends strongly on environmental conditions.² Here we demonstrate the use of comparative metabolomics³ for the analysis of knock-out, overexpression, and knock-down strains to identify metabolites derived from two nonribosomal peptide synthetase (NRPS) gene clusters in the aflatoxin-producing ascomycete Aspergillus flavus, a crop contaminant⁴ and opportunistic pathogen that causes aspergillosis in immunocompromised humans.⁵

Although the A. flavus genome encodes at least 25 polyketide synthase (PKS), 18 NRPS, and two hybrid NRPS-PKS gene clusters,⁶ assignments have been made for only four metabolites.⁷ Many A. flavus secondary-metabolite pathways are under the control of the nuclear protein LaeA, a global regulator of morphogenesis and virulence factor in A. flavus and other pathogenic fungi.⁸ Two LaeA-regulated clusters, which were named lna and lnb, exhibit a striking level of genetic similarity (Figure 1a, Table S1). The lna and lnb clusters contain two noncanonical NRPS genes with high sequence homology (58% identical at the amino acid level), lnaA and lnbA, respectively, which are accompanied by matching sets of genes likely coding for tailoring enzymes. lna and lnb are orphan clusters that have no known associated metabolites and belong to a family of noncanonical NRPS genes that consist of an adenylation (A) domain, a peptidyl carrier protein (PCP) domain, and a thioester reductase (R) domain, but lack a canonical condensation (C) domain (Figure 1a). The functions of this unusual family of NRPSs, which share homology to reductases participating in the biosynthesis of l-lysine in fungi,⁹ have not been explored. We here demonstrate that the lna and lnb clusters encode sets of enzymes that produce overlapping sets of previously undescribed metabolites, and show that one primary function of the noncanonical NRPSs LnaA and LnbA likely consists in the reduction of L-tyrosine. Furthermore, the lna and lnb biosynthetic pathways appear to be part of a signaling network that controls the formation of sclerotia, a resilient overwintering structure.

Figure 1.

a) The lna and lnb gene clusters in A. flavus.^2c Both lna and lnb include noncanonical NRPS genes encoding adenylation (A), carrier protein (PCP), and a short-chain dehydrogenase/reductase domain (R-domain). There is no lnaD orthologue in the lnb gene cluster. See also Table S1 in the Supporting Information. b) Metabolites identified in this study and the known piperazinomycin (9). Compounds marked with an asterisk have been previously described.

To identify lna-associated metabolites by means of comparative metabolomics, we created deletion and overexpres-sion¹⁰ mutants of the NRPS gene lnaA (ΔlnaA and OE∷lnaA, respectively) as well as double-mutant strains ΔlnaA, KD∷lnbA and OE∷lnaA, KD∷lnaB, where KD indicates a knock-down of gene expression using RNAi (strains are listed in Table S3; for methods, see the Supporting Information, Section 1). Unexpectedly, we noted that in the ΔlnaA, KD∷lnbA double-mutant strain, in which both the lna and lnb pathways are disrupted, formation of sclerotia was strongly suppressed relative to the wild-type(WT; see the Supporting Information, Figure S1).

The WT, ΔlnaA, and OE∷lnaA metabolomes were compared using differential analysis by 2D NMR spectrosco-py (DANS) as described previously.^3b–d Comparison of the OE∷lnaA spectra with either WT or ΔlnaA revealed a large number of spin systems present only in the OE∷lnaA spectra (see Figure S2). Available structure databases indicated that most of the OE∷lnaA-specific signals did not correspond to any known metabolites from Aspergilli or other fungi. Detailed NMR spectroscopic and high-resolution (HR) MS analysis of purified fractions containing the OE∷lnaA-specific components suggested a pair of diastereomeric piperazines, 1 and 2, as major lnaA-dependent compounds (Figure 1b; see Tables S4–S10 for spectroscopic data). Comparison with synthetic samples prepared from cyclo(L-Tyr-L-Tyr) and cyclo(D-Tyr-L-Tyr) derivatives confirmed these structural assignments (Supporting Information, Section 8). The structures of 1 and 2 are reminiscent of piperazinomycin (9),¹¹ a bacterial metabolite first identified from Streptomyces olivoreticuli, which may be derived from oxidative macrocyclization of 1.

The piperazines 1 and 2 are accompanied by a derivative that includes a quaternary carbon (δ = 93.1 ppm), two isolated methylene groups, and an OCH₂-CH(N)-CH₂-phenyl fragment. HRMS indicated a molecular formula of C₁₈H₂₁NO₄, whichinconjunction with 2D NMR spectroscopic data (Table S9) led to the identification of the hemiacetal-containing morpholine 7. ROESY correlations between the hydroxy proton with the axial proton at C-6 defined the relative configuration of 7 as (2R*,5S*). We further identified a fully deaminated “monomer”, 3-(p-hydroxyphenyl)-1,2-propanediol (8), a previously described fungal metabolite,¹² and four pyrazine derivatives. These include the known actinopolymorphol C (3),¹³ an O-sulfonated derivative (4), as well as two other highly polar derivatives. The NMR spectra of one of these suggested N-oxidation which was confirmed by HRMS; the molecular formula of C₁₈H₁₆N₂O₄ found corresponded to an unusual N,N-dioxide (5; for a related synthetic compound, see Ref. [14]). NMR and HRMS spectra of the second derivative of 3 indicated sulfonylation at C-3 in this metabolite (→6). In addition to 1–8, which were consistently produced by OE∷lnaA but not ΔlnaA, we detected occasional production of several other metabolites, whose structures (10–13) appeared unrelated to those of 1–8, suggesting that they are not derived from the lna biosynthetic pathway. Among these, compound 10 is a novel metabolite featuring an intriguing ring system that is likely derived from the known A. flavus terpenoid aflavinine (see the Supporting Information, Section 9).¹⁵

Next we investigated whether production of 1–8 can be elicited under specific conditions in wild-type A. flavus, and whether these metabolites are in fact strictly lna-dependent (see Table 1). Highly sensitive single-ion-monitoring MS (SIMMS) analysis of ΔlnaA, WT, and OE∷lnaA cultures grown using conditions previously shown to support lnaA expression^2c revealed the presence of piperazines 1 and 2 in WT, although at roughly 100-fold lower concentrations than in OE∷lnaA (see the Supporting Information, Section 11). Pyrazine 3 was found only in some, but not all WT extracts, whereas 4–8 were not detected. OE∷lnaA produced compounds 1–3, 7, and 8, whereas 4–6 were absent under these culturing conditions. Notably, SIMMS also detected very small quantities of 1 and 2 in ΔlnaA extracts, at levels roughly 10 times lower than in WT. This suggested that 1 and 2 are perhaps also produced by enzymes encoded by the homologous lnb cluster. To test this hypothesis, extracts derived from two different ΔlnaA, KD∷lnbA double-mutant strains were analyzed by HPLC/ESI⁺-SIMMS, which showed that compounds 1–8 are completely absent in the double mutants, whereas WTand ΔlnaA, grown as positive controls in parallel, produced 1 and 2 as before (Figure S4). These results indicate that the amounts of 1 and 2 found in ΔlnaA are derived from lnbA expression and thus the lna and lnb clusters (Figure 1a) encode partially redundant biosynthetic pathways.

Table 1. A. flavus strains (also see Table S3) and occurrence of lna/lnb-associated metabolites 1–8.^a.

Cmpd	OE∷lnaA (rich medium)	OE∷lnaA	OE∷lnaA, KD∷lnaB	Wild type	ΔlnaA	ΔlnaA, KDlnbA
1	+++	+++	+++	++	+	–
2	+++	+++	+++	++	+	–
3	+++	+++	+++	++*	–	–
4	+++	–	–	–	–	–
5	+++	–	–	–	–	–
6	+++	–	–	–	–	–
7	+++	+++	+++	–	–	–
8	+++	+++	+++	–	–	–

^a“+++”=abundant, detected by DANS and MS, “++”=detected by MS only (“*”=occasionally detected), and “+”=detected by MS, roughly 10-old less abundant than in WT. Strains were cultured on GMM unless indicated otherwise.

Since the structures of 1–8 appear to be tyrosine (Tyr) derived, the two putative NRPS proteins LnaA and LnbA were tested for their ability to specifically activate Tyr by assaying amino acid dependent ATP-[³²P]-pyrophosphate exchange activity.¹⁶ Both LnaA and LnbA specifically activated L-Tyr, although LnaA was somewhat D/L-unspecific and also activated D-Tyr (Figure 2a). However, neither LnaA nor LnbA include a condensation domain, and correspondingly, none of the lna- or lnb-associated metabolites identified in the WT nor in the OE∷lnaA background feature a peptide bond. In addition, neither protein activated the L-Tyr-L-Tyr dipeptide (Figure S5). However, both LnaA and LnbA include C-terminal R- (putative short-chain reductase and/or epimerase) domains, with amino acid sequence similarity to other microbial reductase domains.¹⁷ Based on the chemical structures of the identified lna- and lnb-dependent metabolites, it appeared likely that the R-domains in LnaA and LnbA are involved in the reduction of L-Tyr derivatives. In a biosynthetic model, LnaA- and/or LnbA-derived L-Tyr aldehyde (15) may form a dimer such as 17, perhaps via a tethered intermediate, for example, 16 (Figure 2b). Formation of the most abundant lna- or lnb-associated metabolites (1, 2, 7, 8) requires additional reduction from aldehydes or imines to the corresponding amines and alcohols. These reductive steps seemed likely to involve putative NmrA-like proteins encoded by lnaB and lnbB (Figure 1a), as NmrA-like proteins are believed to serve enzymatic functions as epimerases or reductases,¹⁸ in addition to possible roles in gene regulation.¹⁹

Figure 2.

a) Substrate specificity of LnaA (black bars) and LnbA (gray bars). Recombinant LnaA and LnbA were tested for their ability to adenylate various amino acids. Shown are relative activity values normalized to L-Tyr activity. Error bars: standard deviation. b) Simplified model for the biosynthesis of lnaA-dependent metabolites by lna and lnb- cluster genes (see also Figure S6). Tethered L-Tyr (14) is reduced to 15 by the R-domains of LnaA or LnbA, resulting in the formation of a dimeric imine (17) potentially via a tethered intermediate (16). Cyclization and reduction by LnaB (and perhaps LnbB) leads to formation of 1 or 2, whereas oxidation yields 3. In the absence of LnaB, isomerization of 17 (or a related intermediate) to the enamine followed by loss of ammonia leads to increased formation of 7. c) Single-ion chromatograms from HPLC/ESI⁺-SIMMS analysis corresponding to 1, 2, and 7 for the OE∷lnaA and OE∷lnaA, KD∷lnaB strains. The OE∷lnaA, KD∷lnaB strain shows a dramatic increase in production of 7 relative to 1 and 2, compared to the OE∷lnaA strain.

To examine the putative role of LnaB in the biosynthesis of 1 and 2, the metabolite profile of an OE∷lnaA, KD∷lnaB double-mutant strain was compared to that of OE∷lnaA by 2D NMR and HPLC/ESI⁺-SIMMS studies. These analyses showed that the morpholine 7 is much more abundant in OE∷lnaA, KD∷lnaB than in the OE∷lnaA background, relative to the piperazines 1 and 2 (Figure 2c). This shift towards a relatively greater production of morpholine 7, whose biosynthesis requires one reduction step less than production of 1 and 2, suggests that LnaB participates in the reduction of LnaA-derived intermediates. Knock-down of lnaB would result in increased accumulation of 17, which may partially undergo hydrolysis leading to increased production of 7 (Figure 2b, Figure S6). Notably, production of piperazines 1 and 2 is not fully abolished in the OE∷lnaA, KD∷lnaB strain, and formation of morpholine 7 from L-Tyr aldehyde (15) or 17 still requires one reductive step. The residual reductase activity in the OE∷lnaA, KD∷lnaB double mutant could result from expression of the second nmrA-like gene lnbB (or another related gene proximal to lnbB, Table S1) or incomplete silencing of lnaB. Lastly, whereas cyclization of 17 to a corresponding diimine followed by LnaB- (or LnbB-) mediated reduction leads to piperazines 1 and 2, oxidation, non-enzymatically or catalyzed by putative oxidases LnaC, LnaD, or LnbC, would explain formation of pyrazine 3.

HPLC/ESI⁺-SIMMS analysis further showed that the ratio of the two diastereomers 1 and 2 is significantly greater in WT than in ΔlnaA (Figure 3a and Figure S7), suggesting that the lnb pathway produces relatively larger amounts of the (2R*,5S*) isomer, 2, than the lna pathway. Given that LnbA activates L-Tyr with high selectivity, it seems unlikely that this greater relative abundance of 2 results from incorporation of both D-Tyr and L-Tyr. Instead, these stereochemical differences probably originate from different degrees of epimerization at the aldehyde or imine stage. Next, we asked whether the ratio of diastereomers is actively regulated at the gene expression level. Northern analysis of WT cultures incubated with synthetic 1 or 2 showed that treatment with 2 did not lead to any significant changes in lna gene expression, whereas addition of 1 to WT cultures greatly increased expression of lnaB (Figure 3b). This result strongly suggests metabolite-mediated crosstalk between the lna and lnb pathways.

Figure 3.

a) Comparison of the diastereomeric ratios of compounds 1 and 2 in WT and ΔlnaA metabolite extracts. Error bars: standard deviation. b) Effect of supplementation with synthetic 1 or 2 on lna gene cluster expression. Northern analysis showed that addition of 1, but not 2 increased lnaB expression. rRNA served as loading control.

In conclusion, comparative metabolomics revealed eight lna-associated metabolites, 1–8, of which 1–3 could be detected in WT A. flavus. The two most consistently produced lna metabolites, 1 and 2, are also produced under participation of lnb genes, providing a first example for partially redundant biosynthesis involving an NRPS-like pathway. LnaA and LnbA deviate from canonical NRPS domain structures in that they lack condensation domains, and thus, as the structures of identified metabolites suggest, do not act as peptide synthetases. Instead, LnaA and LnbA follow the domain layout of Lys2 and related fungal enzymes that serve as α-aminoadipate aldehyde reductases in fungal l-lysine biosynthesis (24% and 25% identity of LnaA and LnaB to S. cerevisiae Lys2, respectively⁹). Examination of fungal genomes showed that LnaA-like proteins are primarily confined to the Ascomycete taxon Plectomycetes, with between one and five different LnaA-like proteins in different Aspergilli. Recently published genomes (e.g., of Serpula lacrymans and Heterobasidion annosum²⁰) show that LnaA/LnbA-type enzymes are also encoded in basidiomycete genomes. The piperazinomycin-producing bacteria (Streptomyces sp.) have not been sequenced; however, available Streptomyces genomes contain a number of putative genes encoding proteins with high (up to 38%) amino acid sequence similarity to LnaA. The close similarity of some bacterial NRPS-like proteins to LnaA/LnaB despite bacteria not synthesizing lysine by the Lys2 pathway suggests a complex evolution of these proteins.²¹

Our work provides the first evidence that Lys2-type enzymes also function in secondary metabolism and additionally suggests that NRPS-like proteins may serve as amino acid reductases, indicating that a reaction hitherto considered as an NRP-offloading mode can stand by itself. This study also demonstrates the consequences of lnaA/lnbA gene disruption/silencing and the concomitant loss of piperazine/pyrazine/morpholine secondary products on fungal development, linking sclerotia formation in A. flavus and the lna/lnb metabolic pathways. This observation adds a unique role to previously known functions of NRPS-like genes in fungi. The occurrence of a functionally duplicated biosynthetic pathway can be interpreted as a safeguard to ensure timely sclerotial production and hence, persistence during unfavorable environmental conditions. Our results are suggestive of a complex signaling network regulating the biosynthesis of the lna- and lnb-pathway metabolites, which may include cross-pathway interactions mediated by sensing of 1 and 2 or shared biosynthetic intermediates. Detailed analysis of the metabolomes and associated phenotypes of lna/lnb single and double knock-out strains will be required to clarify the extent of interactions between the lna/lnb clusters and determine the roles of the identified metabolites for sclerotia formation and other aspects of A. flavus biology.