Characterisation of the biosynthetic pathway to agnestins A and B reveals the reductive route to chrysophanol in fungi

Identification of a reductase (AgnL4) confirms that in vivo anthraquinone to anthrol conversion is an essential first step in aromatic deoxygenation of anthraquinones catalysed by AgnL6 (reductase) and AgnL8 (dehydratase).


1.0
Experimental Details

Preparative LCMS
Compounds were generally purified using a Waters mass or time directed autopurification system

Cephalone F 10 3
Close inspection of the 1 H NMR spectrum of monodictyphenone 7 (Fig S4) revealed the presence of its co-eluting structural isomer, cephalanone F 10. 3 Figure S4.   were mixed and run on the LCMS, a single peak was observed for peak at 17.9 min (Figure S9), supporting the idea that the compound was present in both samples. At first glance, all seemed to point to monodictyxanthone 12 being the compound eluting at tR 17.9 min in crude extracts, which could be formed from both agnestin A 11 and B 15 and monodictyphenone 7. However, there were no reports that monodictyphenone 7 can undergo spontaneous 4a, 10a -ring-closure to give monodictyxanthone 12. However formation of lactone 32 (which we name monodictylactone) seems feasible. Monodictylactone 32 and monodictyxanthone 12 are structural isomers that could very well co-elute and were also expected to give practically the same NMR spectra (splitting pattern and HMBC correlations). Analysis of NMR spectra of this type of compound is further complicated by keto-enol equilibrium and concentration-dependent chemical shifts. This ambiguity prompted us to reinvestigate compounds formed in the dehydration process occurring in pure samples of monodictyphenone 7 and agnestin B 15 separately.
Aged samples of purified monodictyphenone 7 and agnestin A 11 were examined by LCMS.
Both samples contained a degradation peak eluting at 17.9 minutes ( Figure S12). however, showed two sets of peaks (with ~ 1:1 intensity ratio) with the same splitting pattern as monodictyxanthone 12, none of which seemed to match the first spectrum. The two samples were then mixed and re-submitted for 1 H NMR analysis. The resulting 1 H NMR spectrum ( Figure S13) showed only two sets of peaks, confirming that monodictyxanthone was indeed one of the degradation components of monodictyphenone 7. The second component was consequently deduced to be a co-eluting structural isomer -monodictylactone 32. Figure S13. Aromatic region of the 1 H NMR spectrum of a mixture of degradation products obtained from aged samples of 7 and 11 .
As a result of this detailed analysis, we have established that monodictyphenone 7 does spontaneously close to both 12 and 32 (in ~1:1 ratio based on the relative intensities of signals in the 1 H NMR spectrum, Figure S11), although the equilibrium favours the ring-open form.        We also attempted to obtain 13 C, COSY, HSQC and HMBC spectra, and although not much structural information could be extracted from the data due to advancing degradation, some important resonances were observed. One of the compounds appeared to be a benzophenone derivative: there was a doublet of 2H (based on signal integration) showing COSY correlation to one of the two aromatic triplets at δH 7.26 ( Figure S22), indicating two equivalent protons coupled to the same proton, as in the structure of monodictyphenone (5-H and 7-H appearing as one doublet coupled to 6-H triplet). This indicates that monodictyphenone is the likely precursor for oxidation (Scheme S1).
The advancing decomposition was captured by 1 H NMR over a period of 15h ( Figure S23).

Crystal Growth, Solution and Deposition
The natural tendency for crystallization of Agnestin A 11 was utilized to obtain a crystal structure by X-ray analysis, which confirmed the structure elucidated by NMR. Tiny crystals left in a vial after  Crystallographic data hasve been deposited with CCDC: acession numbers1839028-1839029

Interconversion of Agnestin A 11 and Agnestin B 15
Agnestin A 11 and B 15 were found to interconvert. Agnestin B 15 was observed to undergo rearrangement to agnestin A 11 and the dehydration product -monodictyxanthone 12. This process occurred spontaneously with time and was catalyzed by acid ( Figure S26). A sample of freshly

Gene cluster details + Bioinformatics
A database consisting of translated proteins from the P. variotii genome was BLAST searched using the MdpG PKS responsible for monodictyphenone biosynthesis in Aspergillus nidulans. This analysis identified AgnPKS with 67% identity. The genomic region surrounding agnPKS was compared to the mdp cluster using the Artemis Comparison Tool (ACT, Figure S28). Approximately 50 kb either side of agnPKS were analysed for potential ORFs using Softberry FGENESH, along with manual intron/exon analysis. Putative coding sequences were analysed by BLAST against the NCBI ascomycete database, submitted to InterPro to predict protein families, domains and important sites, and also compared to A. nidulans monodictyphenone cluster (ANmdp cluster). Artemis Comparison Tool was utilised to compare the A. nidulans mdp cluster to the putative agnestin cluster (PVs6c30agn) which allowed identification of homologous genes, and those exclusive to the PVs6c30agn cluster.
The candidate gene cluster was more closely analyzed by pBLAST and the putative proteins encoded by given genes annotated, resulting in a composition of PVs6cl30agn cluster to be proposed as shown on Figure S3.1A and pBLAST annotations listed in Table S6. When the PVs6cl30agn gene cluster was compared with the ANmdp cluster (Table), we were able to differentiate genes that were common for both clusters (homologous genes) and those that are specific for P. variotii cluster (extra genes).
The annotated Agn BGC has been uploaded to genBank with accession number MH898872.

Transformation and KO procedure
Spores from one plate were inoculated into 100 ml of PDB medium and cultured overnight at 25 ˚C with shaking (200 rpm). The culture was then transferred into 50 ml sterile tubes and spun at 9000

DAgnPKS
Disruption of the agnPKS gene led to non-production of agnestins, monodictyphenone and related metabolites by agn∆PKS mutant (Figure S29), confirming that all those compounds shared polyketide precursor and that we have identified the correct gene cluster.

DAgnL4
The agnL4 gene was targeted for disruption. When cultured for metabolite extraction and analysed by the LCMS, the PCR-confirmed mutant was found to accumulate emodin but not chrysophanol ( Figure S30).

DAgnL3
One DAgnL3 transformant was confirmed by PCR analysis to carry the correct mutation and was selected for further testing. LCMS analysis of the secondary metabolites extracted from the mutant showed that neither agnestins nor monodictyphenone were produced. confirming their xanthone affiliation. Preparative TLC yielded two colourful compounds ( Figure   S32).

DAgnR1
time / min 9 6 More than 50 DAgnR1 transformants were isolated after the use of a standard bipartite KO procedure ( Figure S34). The agnR1 gene knockout transformants were analysed by various PCR reactions using genomic DNAs from the transformants and P. variotii WT.  The first PCR is to confirm presence and correct recombination of the two fragments of hygromycin cassette using primer hyg 7.2F and hyg 6.2R (Fig S35). Results show Pv-2, Pv-3, Pv-26, Pv-44 and Pv-47 are negative. All the other transformants tested positive. Fig S35. PCR results for hygromycin cassette using primers hyg7.2F and hyg 6.2R.
PCR for intact agnR1 was carried out using agnM-KO-P1F and agnM-KO-P4R. PCR reaction for agnM1F+agnM4R was set to amplify only intact agnR1 while the agnR1 gene with hygromycin inserted would be too large to be amplified (6 kb). The results show Pv-8 and Pv-11 have the expected insertion ( Figure S36).    Finally strain Pv-11, Pv-8 and Pv7 were grown under producing conditions, extracted and examined by LCMS ( Figure S40 KO strains Pv-11 and Pv-8 still produce WT compounds, while Pv-7 with incorrect integration produces nothing, possibly indicating disruption of agnPKS.