Biosynthesis of Xylariolide D in Penicillium crustosum Implies a Chain Branching Reaction Catalyzed by a Highly Reducing Polyketide Synthase

Fungi are important sources for the discovery of natural products. During the last decades, technological progress and the increasing number of sequenced genomes facilitated the exploration of new secondary metabolites. Among those, polyketides represent a structurally diverse group with manifold biological activities. In this study, we successfully used genome mining and genetic manipulation for functional proof of a polyketide biosynthetic gene cluster from the filamentous fungus Penicillium crustosum. Gene activation in the native host and heterologous expression in Aspergillus nidulans led to the identification of the xil cluster, being responsible for the formation of the 6-methyl-2-pyrone derivative xylariolide D. Feeding with 13C-labeled precursors supported the hypothesis of chain branching during the backbone formation catalyzed by a highly reducing fungal polyketide synthase. A cytochrome P450-catalyzed hydroxylation converts the PKS product to the final metabolite. This proved that just two enzymes are required for the biosynthesis of xylariolide D.


Introduction
Filamentous fungi, including Penicillium and Aspergillus species, are important sources for natural product (NP) discovery and drug development [1,2]. Polyketide NPs are structurally diverse secondary metabolites. Their backbones are assembled by polyketide synthases (PKSs) and often modified by different tailoring enzymes [3]. Type I fungal PKSs are iterative multifunctional enzymes that mainly use simple acyl building blocks to form a linear carbon chain [4]. In the first step, the PKS is primed with a starter unit, like acetyl-CoA, which is then condensed with an extender unit, e.g., malonyl-CoA. The C-C bond formation is catalyzed by the keto synthase (KS) domain via a decarboxylative Claisen condensation. To allow further chain elongation, the growing polyketide chain bound to the acyl carrier protein (ACP) is relocated to the KS domain. The acyltransferase (AT) domain is responsible for loading a new extender unit onto the ACP domain. Highly reducing (HR) PKSs consist of all reductive domains, i.e., dehydratase (DH), enoyl reductase (ER), and keto reductase (KR), which allow the optional full reduction of the β-keto group after Claisen condensation [4,5].
α-Carbon branches can be introduced into the backbone by methyltransferase (MT) domains using S-adenosyl methionine as the methyl donor or by extender units with α-substituents like methylmalonyl-CoA [6]. Several enzymes that catalyze ß-branching by condensation at the ß-position to the carbonyl group of the ACP-bound polyketide chain have been reported, including RhiE with a B domain for the branching reaction [7,8].
Penicillium crustosum (P. crustosum) PRB-2 is a marine fungus isolated from deep-sea sediment [9]. Mining the genome of PRB-2 and gene prediction with antiSMASH [10] indicated the presence of at least 58 putative biosynthetic gene clusters for secondary metabolites. Previous work with P. crustosum PRB-2 revealed that only a few of these gene clusters, such as those of peniphenones, penilactones, and their precursors, are expressed under laboratory conditions [11,12]. For uncovering NPs encoded by these silent or cryptic genes, gene activation in the producer and heterologous expression in a well-studied host have proven reliable genetic approaches [13,14], which were also successfully used for gene identification in PRB-2 [15]. To enhance genetic manipulation, a pyrG deficient strain FK15 was created in a previous study [16]. This strain was further modified by deletion of pcr4870, coding for the DNA repair protein subunit Ku70, to improve gene targeting efficiency. The resulting P. crustosum ∆pcr4870∆pyrG strain JZ02 can be conveniently used for gene deletion and activation experiments [17].
In this study, combined genetic approaches, including activation of genes in the native host as well as heterologous expression in Aspergillus nidulans (A. nidulans), were used to elucidate the function of a silent gene cluster from P. crustosum (Table S1). This led to the isolation of xylariolide D (1) and its congener prexylariolide D (2). Incorporation of 13 C-labeled precursors proved that the methyl residue of xylariolide D is derived from acetate, indicating that the highly reducing PKS also catalyzed a branching reaction during the biosynthesis.

Plasmids and Strains
Plasmids and strains used in this study are listed in Tables S2 and S3. All plasmids were constructed by using an assembly approach based on homologous recombination in Saccharomyces cerevisiae HOD114-2B [18]. For heterologous expression in A. nidulans LO8030 [19], an expression vector was created by using pJN17 as a template for the amplification of Amp/URA3, wA flanking regions, and gpdA(p) [20]. Instead of AfriboB, AfpyrG was amplified from the plasmid pYH-wA-pyrG [21] and used as a selection marker for all strains. The expression vector was sequenced to confirm correct assembly and linearized with SfoI. The genes for expression were amplified from genomic DNA (see Supplementary Materials for detail).
Plasmids for the activation of genes in the native host P. crustosum were assembled likewise: Amp/URA3, flanking regions, AfpyrG, gpdA(p), and genes of interest were amplified by PCR and assembled via yeast homologous recombination. E. coli was used to amplify plasmids after construction in S. cerevisiae (see Supplementary Materials for more detail). Genetic manipulation in E. coli was carried out as described before [22].

Transformation of Fungal Strains and Confirmation of Integration
A. nidulans LO8030 and P. crustosum JZ02 were used for PEG-mediated protoplast transformation according to protocols described previously [12]. For selection of recombinant strains, media lacking uridine and uracil were used. Correct integration of the constructs into the genome was verified by PCR amplification with genomic DNA as template (see Supplementary Materials for more details). Afterward, transformants were cultivated, and the extracts were submitted to LC-MS analysis.

Cultivation of Fungal Strains, Extraction and Isolation of Secondary Metabolites
For LC-MS analysis, 50 mL of liquid glucose minimal medium [3] (GMM, see Supplementary Materials for more details) and Czapek-Dox (CD) medium were inoculated with spores and cultivated at 25 • C in the dark under static conditions. After 14 days, 1 mL of liquid culture was extracted with the same volume of EtOAc twice. After evaporation of the solvent, the residue was dissolved in 95% MeOH and submitted to LC-MS analysis (see Supplementary Materials for details).
Large-scale cultivation for the isolation of metabolites was carried out in 3 L GMM or Czapek-Dox (CD) medium. Compounds 1 and 2 were isolated from culture broths of P. crustosum SSt12, and compounds 2 and 3 were isolated from A. nidulans SSt04 and A. nidulans SSt26. After extraction with the same volume of EtOAc twice, the compounds were isolated by silica gel chromatography and semi-preparative HPLC (see Supplementary Materials for details).

Feeding with Prexylariolide D (2)
Cultivation was performed using 10 mL liquid GMM in 50 mL flasks at 25 • C in the dark under static conditions. Prexylariolide D (2) was dissolved in MeOH and added in a final concentration of 0.5 mM to 3-day-old cultures of A. nidulans SSt28 and A. nidulans LO8030. For LC-MS analysis, 0.5 mL samples were taken after 10 days of cultivation and extracted with the same volume of EtOAc.

Bioinformatic Analysis
One of the putative silent clusters from P. crustosum PRB-2, termed the xil cluster in this study (Figure 1), attracted our attention. The cluster of interest comprises three genes coding for an HR-PKS XilA, a fungal-specific transcription factor XilB, and a cytochrome P450 enzyme XilC. BLAST search revealed that xilA and xilC share identical nucleotide sequences with those coding for KAF7517450.1 and KAF7517449.1 from Penicillium crustosum G10, respectively. An ortholog for xilB was not reported for strain G10. BLAST search also indicated the presence of a similar gene cluster in Penicillium rubens Wisconsin 54-1255 (=DSM 1075). XilA, XilB, and XilC show high sequence identities of 89.1, 85.8, and 93.0% to Pc16g04890, Pc16g04880, and Pc16g04870 from P. rubens, respectively ( Figure 1) [23]. The polyketide synthases XilA and Pc16g04890 share the same domain structure comprising KS-AT-DH-MT-ER-KR-ACP with sequence identities from 82% to 96% on the amino acid level between the respective domains. In addition, XilA shows a sequence identity of 52.2% to the HR-PKS Sol1 (D7UQ44.1) from Alternaria solani with the same domain structure. This PKS assembles a 3-methyl-2-pyrone derivative as the PK scaffold [24][25][26].

Activation of Transcription Factor xilB and PKS xilA in P. crustosum
To identify the product of the xil cluster, we activated the transcription factor xilB by cloning the gpdA(p) promoter from A. nidulans [16] in front of its coding region. The resulting linearized construct pSSt02 was introduced into P. crustosum JZ02 via PEG-mediated protoplast transformation, as mentioned in Section 2.2. The obtained transformant P. crustosum SSt02 (AfpyrG::gpdA(p)::xilB) was verified via PCR ( Figure S6) and cultivated in Figure 1. Comparison of xil cluster from P. crustosum PRB-2 with its ortholog from P. rubens Wisconsin 54-1255. Accession numbers for genes from P. rubens are given in parentheses.

Activation of Transcription Factor xilB and PKS xilA in P. crustosum
To identify the product of the xil cluster, we activated the transcription factor xilB by cloning the gpdA(p) promoter from A. nidulans [16] in front of its coding region. The resulting linearized construct pSSt02 was introduced into P. crustosum JZ02 via PEG-mediated protoplast transformation, as mentioned in Section 2.2. The obtained transformant P. crustosum SSt02 (AfpyrG::gpdA(p)::xilB) was verified via PCR ( Figure S6) and cultivated in GMM for 14 days. In comparison to the control strain JZ02, LC-MS analysis revealed the presence of a new compound 1 with a [M+H] + ion at m/z 183.1019 in the extract of SSt02 (Figure 2A). However, the low quantity prohibited us from structure elucidation. Therefore, we used the same strategy as for xilB to activate the PKS gene xilA, resulting in strain P. crustosum SSt12 (AfpyrG::gpdA(p)::xilA). LC-MS analysis indicated an at least 17-fold higher accumulation of 1 in the xilA overexpression strain SSt12 than in the strain SSt02 after xilB activation ( Figure 2A). Furthermore, another new compound 2 was detected with a [M+H] + ion at m/z 167.1067. Product yields of 27.5 and 12.9 mg/L were calculated for 1 and 2 in SSt12, respectively. can be speculated that 2 is the PKS (XilA) product and hydroxylated to 1 by the cytochrome P450 enzyme XilC in P. crustosum (Figure 4). Comparison of the NMR data of 2 and 3 revealed that the signals at δH 2.22 (s, 3H) and δC 17.2 ppm for the methyl group at the 2-pyrone ring of 2 disappeared in the spectra of 3. Instead, signals for a methylene group were observed at δH 4.45 (s, 2H) and δC 59.0 ppm (Figures S12, S13, S16, and S17), indicating the conversion of the methyl to a hydroxymethyl group in 3. This was also confirmed by HSQC and HMBC analyses. Since 3 (named xylariolide G) is not produced by the native host P. crustosum, even after activation of xilA or xilB, we assume that the conversion of 2 to 3 in A. nidulans SSt04 is catalyzed by an enzyme from the host strain, as observed for a methylated isocoumarine after heterologous expression of a PKS gene [22].

Heterologous Expression of xilC and Feeding of Prexylariolide D
To provide evidence for the conversion of 2 to 1 in P. crustosum by the cytochrome P450 XilC and 2 to 3 in A. nidulans by a host enzyme, we created a xilC overexpression strain A. nidulans SSt28. Compound 2 (0.5 mM) was added to the three-day-old cultures of SSt28 and the control strain LO8030. As shown in Figures 2C and S24, 2 was almost completely converted to 1 in SSt28 and about 30% to 3 in the control strain. These results proved that 2 was hydroxylated by XilC at the side chain and by an unknown enzyme from A. nidulans at the methyl group of the 2-pyrone ring (Figure 4).  To confirm that the PKS Pc16g04890 from P. rubens Wisconsin 54-1255 produces the same compounds as XilA, its genomic sequence was also expressed in A. nidulans. Cultivation of the obtained transformant SSt26 led indeed to the accumulation of 2 and 3 ( Figures 2B and S4).

Structure Elucidation
The obtained analytically pure compounds 1-3 were subjected to NMR analysis, including 1 H-NMR, 13 C-NMR, HMBC, and HSQC ( Figures S7-S20). Interpretation of the NMR spectra and comparison with data in the literature proved 1 and 2 to be xylariolide D [27] and its non-hydroxylated precursor 5-butyl-6-methyl-2H-pyran-2-one (named prexylariolide D in this study) [28], respectively (see Figures 2 and 3 for structures). ECD spectrum ( Figure S23) confirmed the S-configuration of 1, as reported previously [27]. It can be speculated that 2 is the PKS (XilA) product and hydroxylated to 1 by the cytochrome P450 enzyme XilC in P. crustosum (Figure 4). Comparison of the NMR data of 2 and 3 revealed that the signals at δ H 2.22 (s, 3H) and δ C 17.2 ppm for the methyl group at the 2-pyrone ring of 2 disappeared in the spectra of 3. Instead, signals for a methylene group were observed at δ H 4.45 (s, 2H) and δ C 59.0 ppm (Figures S12, S13, S16, and S17), indicating the conversion of the methyl to a hydroxymethyl group in 3. This was also confirmed by HSQC and HMBC analyses. Since 3 (named xylariolide G) is not produced by the native host P. crustosum, even after activation of xilA or xilB, we assume that the conversion of 2 to 3 in A. nidulans SSt04 is catalyzed by an enzyme from the host strain, as observed for a methylated isocoumarine after heterologous expression of a PKS gene [22].

Feeding with 13 C-Labeled Precursors
To clarify whether the methyl residue at the 2-pyrone ring of 1 is derived from Lmethionine or acetate, feeding experiments with [2-13 C] and [1,2-13 C] acetate were conducted in P. crustosum SSt12. Both 1 and 2 were isolated and subjected to 13 C NMR analysis. As shown in Figures 3 and S21, incorporation of 13 C with enrichments between 5.0 and 9.0% (Table S5) was observed for C3, C5, C8, C10, and C11 in the spectra of both 1 and 2 isolated after [2-13 C] acetate feeding. These results proved unequivocally the incorporation of five acetate units in their structures and acetate as the origin of both methyl groups. This conclusion was also supported by the observed 13 C-13 C coupling pattern in the 13 C NMR spectra of 1 and 2 (Figures 3 and S22, Table S5) after feeding with [1,2-13 C] acetate.

Discussion
Xylariolide D was isolated from different fungal strains, e.g., Xylaria sp. NCY2, Neodidymelliopsis sp., and Dictyosporium digitatum [29][30][31]. However, the gene cluster responsible for its biosynthesis has not been described previously. Based on the results from feeding and incorporation experiments, we postulate a mechanism for the formation of prexylariolide D (2) catalyzed by the type I iterative HR-PKS XilA (Figure 4). During the biosynthesis, the linear polyketide chain is branched by the addition of an acetyl unit. Product release is achieved via 2-pyrone ring formation accompanied by water elimination. As shown in Figure 1, a putative MT domain was predicted for XilA and Pc16g04890. However, activation of xilA in P. crustosum and heterologous expression of both genes in A. nidulans, as well as feeding experiments, proved the accumulation of metabolites without the involvement of a methyl transfer reaction. It seems that the MT domain in both PKSs is inactive or has another function.
As mentioned in Section 3.1, bioinformatic analysis indicates that xilB encodes a fungal-specific transcription factor. By xilB activation, we were able to slightly increase xylariolide D (1) production in P. crustosum. However, activation of the PKS gene xilA not only resulted in the formation of 2, but also in a significantly higher yield of product 1 compared to that after xilB activation. Obviously, XilB had less influence on the regulation of xylariolide D biosynthesis. It is possible that the entire xil cluster is under the control of other systems like the global regulation of secondary metabolism. Furthermore, the production of both compounds 1 and 2 in P. crustosum SSt12 after overexpression of xilA indicates that the formation of the polyketide scaffold, and not the hydroxylation, is the limiting step in the biosynthesis of xylariolide D.
As shown in Figure 2, feeding of prexylariolide D (2) led to the accumulation of xylariolide D (1) in the xilC overexpression strain A. nidulans SSt28 and xylariolide G (3) in the control A. nidulans SSt01. Dihydroxylated derivative was not detected in SSt28, proving the higher substrate specificity of XilC toward its natural substrate 2 than the unknown hydroxylase from A. nidulans. Only in the absence of XilC, prexylariolide D (2) was converted to 3 by the hydroxylase from A. nidulans.

Conclusions
In conclusion, we elucidated the function of a silent gene cluster for the biosynthesis of xylariolide D (1) in two fungal strains by gene activation in the native strain and heterologous expression in a well-studied host. Activation of the PKS gene xilA led to the accumulation of 1 with a much higher yield than the activation of the putative transcription factor xilB. Feeding experiments confirmed that XilC catalyzes the hydroxylation of prexylariolide D (2) to 1. The incorporation of 13 C-labeled precursors provides clear evidence for acetate as the origin of the methyl group at the 2-pyrone ring. Further investigation,

Heterologous Expression of xilC and Feeding of Prexylariolide D
To provide evidence for the conversion of 2 to 1 in P. crustosum by the cytochrome P450 XilC and 2 to 3 in A. nidulans by a host enzyme, we created a xilC overexpression strain A. nidulans SSt28. Compound 2 (0.5 mM) was added to the three-day-old cultures of SSt28 and the control strain LO8030. As shown in Figure 2C, 2 was almost completely converted to 1 in SSt28 and about 30% to 3 in the control strain. These results proved that 2 was hydroxylated by XilC at the side chain and by an unknown enzyme from A. nidulans at the methyl group of the 2-pyrone ring (Figure 4).

Discussion
Xylariolide D was isolated from different fungal strains, e.g., Xylaria sp. NCY2, Neodidymelliopsis sp., and Dictyosporium digitatum [29][30][31]. However, the gene cluster responsible for its biosynthesis has not been described previously. Based on the results from feeding and incorporation experiments, we postulate a mechanism for the formation of prexylariolide D (2) catalyzed by the type I iterative HR-PKS XilA (Figure 4). During the biosynthesis, the linear polyketide chain is branched by the addition of an acetyl unit. Product release is achieved via 2-pyrone ring formation accompanied by water elimination. As shown in Figure 1, a putative MT domain was predicted for XilA and Pc16g04890. However, activation of xilA in P. crustosum and heterologous expression of both genes in A. nidulans, as well as feeding experiments, proved the accumulation of metabolites without the involvement of a methyl transfer reaction. It seems that the MT domain in both PKSs is inactive or has another function.
As mentioned in Section 3.1, bioinformatic analysis indicates that xilB encodes a fungal-specific transcription factor. By xilB activation, we were able to slightly increase xylariolide D (1) production in P. crustosum. However, activation of the PKS gene xilA not only resulted in the formation of 2, but also in a significantly higher yield of product 1 compared to that after xilB activation. Obviously, XilB had less influence on the regulation of xylariolide D biosynthesis. It is possible that the entire xil cluster is under the control of other systems like the global regulation of secondary metabolism. Furthermore, the production of both compounds 1 and 2 in P. crustosum SSt12 after overexpression of xilA indicates that the formation of the polyketide scaffold, and not the hydroxylation, is the limiting step in the biosynthesis of xylariolide D.
As shown in Figure 2, feeding of prexylariolide D (2) led to the accumulation of xylariolide D (1) in the xilC overexpression strain A. nidulans SSt28 and xylariolide G (3) in the control A. nidulans SSt01. Dihydroxylated derivative was not detected in SSt28, proving the higher substrate specificity of XilC toward its natural substrate 2 than the unknown hydroxylase from A. nidulans. Only in the absence of XilC, prexylariolide D (2) was converted to 3 by the hydroxylase from A. nidulans.

Conclusions
In conclusion, we elucidated the function of a silent gene cluster for the biosynthesis of xylariolide D (1) in two fungal strains by gene activation in the native strain and heterologous expression in a well-studied host. Activation of the PKS gene xilA led to the accumulation of 1 with a much higher yield than the activation of the putative transcription factor xilB. Feeding experiments confirmed that XilC catalyzes the hydroxylation of prexylariolide D (2) to 1. The incorporation of 13 C-labeled precursors provides clear evidence for acetate as the origin of the methyl group at the 2-pyrone ring. Further investigation, such as the biochemical characterization or elucidation of the XilA structure, will gain deeper insight into the unusual formation of the 5-alkyl-6-methyl-2-pyrone backbone.