Hybridorubrins A–D: Azaphilone Heterodimers from Stromata of Hypoxylon fragiforme and Insights into the Biosynthetic Machinery for Azaphilone Diversification

Abstract The diversity of azaphilones in stromatal extracts of the fungus Hypoxylon fragiforme was investigated and linked to their biosynthetic machineries by using bioinformatics. Nineteen azaphilone‐type compounds were isolated and characterized by NMR spectroscopy and mass spectrometry, and their absolute stereoconfigurations were assigned by using Mosher ester analysis and electronic circular dichroism spectroscopy. Four unprecedented bis‐azaphilones, named hybridorubrins A–D, were elucidated, in addition to new fragirubrins F and G and various known mitorubrin derivatives. Only the hybridorubrins, which are composed of mitorubrin and fragirubrin moieties, exhibited strong inhibition of Staphylococcus aureus biofilm formation. Analysis of the genome of H. fragiforme revealed the presence of two separate biosynthetic gene clusters (BGCs) hfaza1 and hfaza2 responsible for azaphilone formation. While the hfaza1 BGC likely encodes the assembly of the backbone and addition of fatty acid moieties to yield the (R)‐configured series of fragirubrins, the hfaza2 BGC contains the necessary genes to synthesise the widely distributed (S)‐mitorubrins. This study is the first example of two distant cross‐acting fungal BGCs collaborating to produce two families of azaphilones and bis‐azaphilones derived therefrom.


Introduction
The Hypoxylaceae, which were recently resurrected in the course of am ajor phylogenetic study,a re the second largest family of the ascomycete order Xylariales, [1] and they are known for ap articularlyd iverses econdary metabolism. [2] In contrastt oo ther familieso ft he order,b otht heir mycelial cul-tures and their stromata (a mass of fungal tissuet hat has embedded spore-bearings tructures such as ascomata) have been shownt oc ontain diversep igments and other secondary metabolites. The first of these pigments were reported in 1974 by Steglich et al. from Hypoxylonf ragiforme,t he type specieso f the largest genus of the Hypoxylaceae, and shown to belong to the mitorubrin-azaphilonec lass of metabolites. [3] Several years later,t he same speciesw as subjected to an intensive study,a nd various cytochalasans and other unknown compounds were detected and isolated from the young,g rowing stromata. [4] In the same study,i tw as found that the composition of secondary metabolite profilesd iffers drastically during the vegetativeg rowth period,a nd this points toward differential activationo fs econdary metabolite biosynthesis genes. From cultureso ft he fungus,s everald ifferentm etabolites such as dihydroisocoumarins, [5] ad ibenzoxanthenone, [6] various cytochalasans, [7] and small polyketidesh ave been reported. [8] Some of these metabolites were found to have prominenta ctivitiesi nb iologicals ystems, while others, like the complex azaphilones that were recently detected in fossil stromata of H. fragiforme and isolated from freshly collected material, are unprecedentedc ompounds. [1b] We have recently started to further evaluate the diversity of secondary metabolites in twelve selected species of the Hypoxylaceae for which we generated high-quality genome sequences with the aim of establishing correlationsb etween the biological and chemical diversity in these organisms at the genomic level. [9] The ex-epitype strain of H. fragiforme,t he type specieso ft he genus Hypoxylon and the most frequently encountered species in the Northern hemisphere, was selected for genome sequencing. As expected from the various reports on the chemical diversity of secondary metabolites, the genome harbours ag reat many biosynthetic gene clusters (BGCs) that putativelye ncode the biosynthesis of variousp olyketide and polyketide-peptide hybrids. We have recently reported on the identity of the cytochalasin gene clustero ft his fungus and its partial heterologous expression in Magnaporthe grisea. [10] Furthermore, we reported the occurrence of the novel azaphilones fragirubrins A-E and the bis-azaphilones rutilins Ca nd Di ns tromata of H. fragiforme in addition to the known mitorubrins. [1b] The present study deals with the isolation and identification of azaphilone heterodimers with interesting structurala nd biological features, asw ell as the assignment of their biosynthesis genes.

Results and Discussion
Isolation and structure elucidation Freshlyc ollected stromata of Hypoxylon fragiforme were extracted with acetone. In the crude extract the new compounds 1-6 ( Figure 1) wered etected by HR-ESI-MS analysis and subsequently purified by preparative chromatography. Proton and carbon NMRd ata of the pure 1-6 are given in Tables 1a nd 2. Hybridorubrin A( 1)w as shownt oh ave the molecular formula C 52 H 62 O 15 by HR-ESI-MS. The IR spectrumo f1 showed characteristic absorptions at ñ max = 1717 and 1621 cm À1 ,r epresenting ester and conjugatedd ouble bonds, respectively (Figure S42 in the Supporting Information). In the 1 Ha nd 1 H/ 13 C HSQC NMR spectra, the presence of six methyl groups,t wo methylene groups, and an uncertain number of methylene groups in an alkyl chain, as well as one two aliphatic and nine aromatic/olefinic methine groups, was observed. The 13 Ca nd 1 H/ 13 CH MBC spectra showed the presence of four conjugated keto groups, three carboxylic ester groups, as well as eleven sp 2 -a nd two sp 3 -hybridized carbon atoms. 1 H/ 1 HC OSY signals ( Figure 2) revealed 12-H 2 ,1 3-H, and 14-H 3 to be contiguous. For the propyl chain protons 12a-H 2 , 13a-H,a nd 14a-H 3 ,asimilar link was established. The first azaphilone core was identified by 1 H/ 13 CH MBC correlations ( Figure 2) from 4-H to C-3, C-6, andC -10,f rom 1-H 2 to C-3, C-5, C-9, and C-10, as well as from 11-H 3 to C-7, C-8, and C-9. Mutual correlationso f4 -H and 12-H 2 linked C-12 to C-3. An acetate moiety was connected to C-13 by correlations from 13-Ht oC -15 and from 16-H 3 to C-12. The second azaphilone unit was established analogously.C orrelations from 13 a -H to C-6 as well as 14a-H to C-5, C-6, and C-7 linked the two azaphilone units. The Z configuration of the D 6,14a alkene was deduced from the presence of as trong 1 H/ 1 HR OESY correlation between 14a-H and 4-H (Figure 2a nd Figure S10 in the Supporting Information), while that of the D 12a,13a alkene was determined as E from the couplingc onstant of the respective protons ( 3 J = 15.3 Hz).
For the fatty acid moiety, the carboxylict erminus was established by 1 H/ 1 HC OSY correlations linking 2'-H 2 ,3 '-H 2 ,a nd 4'-H 2 as well as by 1 H/ 13 CH MBCc orrelations from 3'-H 2 to C-1',C -2', C-4',a nd C-5'.T he methyl terminus 16'-H 3 showed correlations to C-15' and C-14',w hich were supported by 1 H/ 1 HC OSY data. The hydroxyl group 13'-OH showed 1 H/ 13 CH MBC correlations to C-12',C -13',a nd C-14'.T he methylene 12'-H 2 had ac orrelation to C-11',w hich wasa ccordingly placed in the alkyl chain. The missing signals forC -5' to C-10' overlap and could not be assigned unambiguously.C onsequently,t he length of the fatty acid chain was deduced from the molecular formula of 1.B y using Mosher's method, [11] the stereochemistry of C-13' was assigneda s( R)( Figure S38 in the SupportingI nformation). Ultimately, the fatty acid moiety of 1 was deducedt ob e( R)-13'hydroxypalmitic acid.T he fatty acid was linked to C-8 by 4 J 1 H/ 13 CH MBC correlations from 11-H 3 to C'-1 and from 2'-H 2 to C-8. Lastly,t he orsellinic acid moiety was established by 1 H/ 13 C HMBC correlationso f7 a '-H 3 to C-1a',C -5a',C -6a',a nd C-8a', from 4a'-OH to C-3a',C -4a',a nd C-5a',a sw ell as from 2a'-OH to C-1a',C -2a',a nd C-3a'.C orrelationsf rom 11a-H 3 to C-8a' linked the orsellinic acid to C-8a. The stereochemistry of C-8(R) and C-8a(S)w as deduced from their respective building blocks mitorubrin and fragirubrin (see Stereochemistry section below for details). Due to occurrence of af ragirubrin building block in 1,t he stereochemistry of C-13(S)w as deduced from the fragirubrins. [1b] Eventually,t he same C-13(S) configuration was also found in lenormandin F. [12] Analysis of hybridorubrin B( 2)r evealed its molecular formula to be C 54 H 62 O 15 ,i ndicating two additional carbon atoms and two additional degrees of unsaturation compared with 1.I nstead of (R)-13'-hydroxypalmitic acid, it bears (R)-16'-hydroxylinoleic acid, as shown by its NMR data. The stereochemistry of C-16' was assigned by Mosher's method ( Figure S39 in the Supporting Information). The chemical shifts of C-8' and C-14' were characteristic for a cis (Z)/cis (Z)1 ,4-diene configuration of D 9',10' and D 12',13' . [13] HybridorubrinC (3)h as am olecular formula of C 54 H 64 O 15 ,a s shown by HR-ESI-MS data. This implied af ormal loss of hydrogen relative to 1,r epresenting one additional degree of unsaturation, while having af atty acid moiety extended by two carbon atoms. Accordingly,t wo olefinic protons were observed in the 1 H/ 13 CH SQC spectruma nd placed in the fatty acid chain of 3.T he exact position of the alkene was deduced to be D 9',10' due to occurrence of two diagnostic MS/MS fragments (m/z 155.1123 and 171.1066) after epoxidation of the double bond (see Experimental Section and Figure S45 in the Supporting Information). [14] The stereochemistry of this alkene was determined as cis (Z)f rom comparison of chemical shifts of the allylic carbon atoms C-8' and C-11' (both d C = 27.4). [15] By applying Mosher'smethod, the stereochemistry of C-17'(R)w as deduced ( Figure S40 in the Supporting Information).
HybridorubrinD (4)s howed am olecular formula of C 50 H 60 O 13 ,i mplying the formal loss of aC 2 H 2 O 2 fragment compared to 1.T he NMR spectra of 4 were highly similart ot hose of 1,w ith the key differences being the lack of an acetyl group attachedt oO -13 as well as ad ifferent fatty acid moiety,w hich was identified as palmitic acid.
The molecular formula of fragirubrin F( 5)w as determined from its HR-ESI-MSd ata as C 31 H 46 O 8 .I ts 1 Ha nd 13 CNMR data showed high similarity to those of fragirubrin A ( 15). [1b] Com-pound 5 contains four methyl groups, three olefinic and two aliphatic methine groups, asw ella s1 4m ethylene groups. Additionally, signals for two conjugatedk eto groups, two ester carbonyl groups,o ne oxygenated sp 3 carbon atom andt hree sp 2 carbon atomsw ere observed in the 13 CNMR spectra.T he main differencet o15 was the replacement of the palmitoyl moiety by (R)-14'-hydroxypalmitic acid. The absolute stereochemistry of C-14' was determined by using Mosher's method ( Figure S41 in the Supporting Information).
HR-ESI-MS data determined the molecular formula of fragirubrin G( 6)a sC 31 H 44 O 7 ,i mplying one additional degree of unsaturationr elative to fragirubrin A ( 15). The 1 Ha nd 1 H/ 13 CH SQC spectra located an additional olefin in the fatty acid moiety (D 9',10' ). The positiona nd stereochemistry of this double bond was determined by degradation of the compound to its fatty acid methyl ester (6-FAME) and subsequent comparison of GC retention times with those of authentic standards, which resulted in the identification of 9-cis (Z)-hexadecenoic acid (palmitoleic acid;s ee Figure S46 in the Supporting Information and Experimental Section).

Stereochemistry of azaphilones occurringi nH. fragiforme and revision of rutilinsCand D
The stereochemistry of the azaphilones, particularly of C-8 in the backbone,i sa ni mportant aspecto fs tructural complexity. While the first occurrence of (À)-mitorubrins was described in 1965 by Büchi et al. from cultures of Penicillium rubrum [16] (current name: Talaromyces ruber), Stegliche tal. later described (+ +)-stereoisomers of mitorubrins from the stromata of H. fragiforme. [3] Curiously,t he genus Talaromyces was reported to contain either (+ +)-or (À)-mitorubrins depending on the species. [17] We utilized electronic circulard ichroism (ECD) spectroscopy as am eanst oa ssess the stereochemistry of the monomeric azaphilones. As tudy by Clark et al. on chemical synthesis of the azaphilone backbone [18] allowed for relatively simple assignment:m itorubrinol (8)f rom H. fragiforme showed ap osi-tive cotton effect (CE) at 245 nm and negative CE at 226 and 272 nm ( Figure S44 in the Supporting Information). These ECD data indicate an (S)-(+ +)configuration, [18] which we conferred to all mitorubrin-type azaphilones from H. fragiforme due to their common biosynthetic origin (see Biosynthesis section ford etails). However,a ll fragirubrins [1b] showedi nverted ECD spectra with positive CE at about2 30 and 274 nm and negative CE around2 50 nm, accordingly renderingt hem (R)-(À)i somers ( Figure S44 in the Supporting Information).
Ta king the ECD resultsa nd the BGC analysis (see Biosynthesis section below) into account then allows fors tereochemical assignmento ft he heterodimers:r utilins such as 12-13,w hich consist of two (S)-mitorubrin-type building blocks, are hence (S)-configured at both C-8 and C-8a. Hybridorubrins 1-4,i n turn, are (S)-configured at C-8a in their mitorubrin moiety and (R)-configured at C-8 in their fragirubrin part. We hence have to revise data from our prior study with rutilins C( 12)a nd D (13), and the mitorubrins, [1b] to be (S)-configured at C-8 and C-8a.

Bioactivity testing
Compounds 1-2, 4-10,a nd 12-18 were tested for their antimicrobial activity in am inimumi nhibitory concentration (MIC) assay as well as for their cytotoxicity,b ut were found to be devoid of activity against the examined test organisms or cell lines (Table S3 in the Supporting Information). The lack of antimicrobial and cytotoxic activities is largely in accordance with former findings for mitorubrin-type azaphilones. [4] In addition, 1, 3-4, 7-10, 12,a nd 14-15 were tested for their inhibitory effect on biofilm formation of Staphylococcus aureus (Table 3). Due to minor impurities in the samples, the given percentagev alues only allow for ar ough estimationo f bioactivity.Strong activity was observed for the bis-azaphilones hybridorubrins A( 1), C( 3), D( 4), andr utilin C ( 12). These compounds have potency similartothat of the reference microporenic acid A, [19] as wellass clerin,and sclerin diacid from H. fragiforme. [8] The mitorubrin-type azaphilones 7 and 9,a sw ella s the fattya cid-containing 15,s howed weak activity,w hile 8, 10, and 14 showed no inhibition.
These results allow for preliminary structure-activity relationships to be deduced:s ince rutilin C( 12)s howedm uch stronger inhibition of biofilm formation than 7,astrong influence of the fused second azaphilone backbone is suggested. In addition, the differing inhibition of biofilm formation of the mitorubrin-type azaphilones 7-10 indicatesamodest influence of the functional group at C-14:amethyl group or an acetate unit (7, 9)a llowed for weak activity,w hile azaphilones carrying more polar hydroxyl or carboxylic acid moieties at C-14 (8,10) exhibitedn oi nhibition of biofilm formation.
Lenormandin F( 14)a nd fragirubrin A (15), which both carry aC 16 fatty acidm oiety instead of an orsellinic acid residuea t C-8, only differ in the presence of an acetate moiety at C-13 in 15.W hile 14 showed no activity, 15 exhibited weak activity similart o7 and 9.H ence, ap ositive effect of C-13 acetylation can be deduced.B yc omparing 7 and 9 to 15,t he presenceo f an orsellinic acid or af atty acid moiety at C-8 does not seem to be highly relevant for activity against S. aureus biofilm formation.
Ta king thesef indingsi nto account,t he strong bioactivity measured for hybridorubrins A( 1), C(3), and D( 4)a nd rutilin C (12)c an be mainly explained by the fusion of two azaphilone buildingblocks. As acetylation of C-13 was deduced to be beneficial for bioactivity, 4 consequently exhibitedaweaker effect than 1 and 3.

Azaphilone BGC analysis
In order to understand how the wide diversity of azaphilonetype compoundsi nt he stromata of H. fragiforme is genetically encoded, we investigated the genome sequence of the producer organism. Genome sequencing of the fungus had been performed in the contexto faprevious study. [9] To identify al ikely candidate gene cluster,t he previously published sequences of the BGCs encoding azaphilones in Monascusr uber [20] (i.e.,m onascin, ankaflavin,a nd monascorubrin), azanigerones in Aspergillus niger, [21] and mitorubrinol in Talaromyces marneffei [22] were used for homology searches. In M. ruber, A. niger,a nd T. marneffei assembly of the azaphilone core structure is initiated by the action of an on-reducing polyketide synthase (NR-PKS) and finalized by subsequent processing of ak etoreductase (KR) and FAD-dependent monooxygenase (FAD-MO). [20] These three core proteins were initially used as the template for BLASTP searches against a H. fragiforme protein databasec reatedb yu sing the softwareG eneious 9.1.8.
In total, seven NR-PKS-containing BGCs were found. However,o nly one included the required KR andF AD-MO encoding genes. This BGC (designated hfaza1,G enBank Acc. No. MN736721) is composed of seven genes, the majority of which show high homology with the biosynthetic genes of the M. ruber azaphilone mrPig and the T. marneffei mitorubrinol BGCs ( Figure 3). In addition to theN R-PKS (hfaza1A), the KR (hfaza1F), and the FAD-MO (hfaza1D), genes encoding an NADPH-dependent dehydrogenase (hfaza1B), an ac(et)yltransferase (hfaza1E), at ransporter (hfaza1C), and at ranscription factor (hfaza1G)a re present. Ap revious investigation of the mitorubrinolg ene cluster in T. marneffei showedt hat two PKS genes are involved in the biosynthesis of 8 and 10. [22] Thes econd PKS encodes the biosynthesis of orsellinic acid. We therefore searched for ah omologue of the putative T. marneffei orsellinic acid synthase (OSAS) pks12 in H. fragiforme. Accordingly,w ef ound ag ene cluster encodingahighly similar NR-PKS together with as et of genes of which the majority also appeared in the M. ruber ( Figure 3) and A. niger azaphilone BGCs. This second gene cluster is designated hfaza2 (GenBank Acc. No. MN736720). The re-spectiveN R-PKS (Hfaza2A) hasa nS AT-KS-AT-PT-ACP domain structure, and thus lacks at ypical C-terminal release domain. Additional genes in the hfaza2 BGCe ncodea nF AD-MO (hfaza2D)w ith high homology to hfaza1D and mrPigN,a n ac(et)yltransferase (hfaza2E)h omologous to hfaza1E and mrPigD,aP450 monooxygenase (hfaza2F), anda nN ADPH-dependentd ehydrogenase (hfaza2L)w ith homology to hfaza1B and mrPigH. Furthermore, two similarg enes were also found in the BGC( hfaza2B and hfaza2C)t hat did not produce any hits in BLASTPs earches against the Swiss-Prot database, but showeds trong homology with the azaphiloneb iosynthesis genes mrPigM and mrPigO from M. ruber. On the basis of knockout experiments of the latter two,i tw as deduced that MrPigM is an acetyltransferase, whereas MrPigO performs deacetylation. [20] In addition to these genes, two FAD-dependent oxidoreductases (hfaza2J and hfaza2M)w ere found, which are very similart oazaG and azaL,b oth of which are part of the azanigerone biosynthetic pathway. [21] Finally,t wo putative transcription factors (hfaza2G and hfaza2H)a nd two putative transporters (hfaza2I and hfaza2K) were also assigned to the cluster.Adetailed comparison of the hfaza1 and hfaza2 clusters with the uncharacterized mitorubrinol BGC reported from T. marneffei [22] showed the presenceo f almosta ll genes from the latter in the H. fragiforme BGC (Figure 3). Homologues of Hfaza1A, Hfaza2A, Hfaza1B, Hfaza2B, Hfaza2C,H faza1E, Hfaza2E, Hfaza1F,a nd Hfaza2L, were found. The T. marneffei cluster is expanded by two hydrolase enzymes, but no FAD-dependent monooxygenase, P450 monooxygenases, FAD-dependento xidoreductases, and transcription factors are present. Therefore, we propose, according to the homology analyses, that two unlinked BGCs (hfaza1 and hfaza2)a ct together to assemble and diversify azaphilones in H. fragiforme.
H. fragiforme does not readily produce azaphilones in laboratory culture, so it is not yet possible to investigate the biosynthesis experimentally.H owever,t here is now sufficient detailed knowledge concerningt he biosynthesis of related compounds in other organismstoa llow the development of ad etailed biosynthetic hypothesis based on the combination of structure information and analytical HPLC-MS data (Scheme 1).
The formationo fa zaphilones likely starts in as imilar fashion as provedf or azanigerones and Monascus pigments with the NR-PKS Hfaza1A producing ah exaketide chain, whichi ss ubsequently cyclised by the product template (PT) domain and released by the reductiver elease( R) domain of the PKS to yield the reactive benzaldehyde intermediate 20.C hen et al. reported that in M. ruber ketoreductiona tC -13 is required prior to hydroxylation at C-8 to afford the pyran ring. [20] In the crude stromatal extracts of H. fragiforme we could not find any evidence for the existence of such ab icyclic pyranoquinonei ntermediate 22.I nstead, we found ac onspicuousp eak with m/z = 249 [M+ +H] + ,w hich showedf ragmentation patterns, aU V/Vis spectrum,a nd molecular formula consistentw itht he putative keto intermediate 21 (Tables S4, S5 in the Supporting Information). We therefore concludet hat 21 might be produced by hydroxylation of 20 at C-8 by the FAD-MO Hfaza1D and subsequent spontaneousp yran ring formation.A sH faza2D also encodes ah omologous enzyme, it is possible that it can perform the same reaction. In ap revious study based on crystal structure data and quantum mechanical/molecularm echanical calculations of the homologous FAD-MO TropB, it has been demonstrated that such enzymes govern ah ighlye nantioselective transformation. [23] The occurrence of the homologous pair Hfaza1D/Hfaza2D would therefore be consistentw ith the observation of different stereoconfigurations at C-8 between mitorubrin-type and fragirubrin-type azaphilones. Compound 21 can then be furtherp rocessed by the ketoreductase Hfaza1F to yield 22.A sp reviously stated, we could not detect 22, which can possibly be explained by differences in metabolic rates due to differences in enzyme reactionr ates or expression levels of hfaza1F and subsequent processing enzymes.
In the next step, the pathway branches into two directions depending on the attached side chain. In order to yield lenormandin-type azaphilones such as 14,t he backbone 22 can undergo acylation at the C-8 alcohol mediated by the acyltransferase Hfaza1E. Subsequent acetylation at the C-13 alcohol by the putative acetyltransferase Hfaza2B will lead to the highly diverseg roup of fragirubrins (5)(6)(15)(16)(17)(18)(19), which differ among each other in the chain length, desaturation level, and hydroxylation pattern of the side chain. This side chain very likely originates from different free or CoA-bound fatty acids of the primary metabolism implying ab road substrate tolerance of Hfaza1E.A cyltransferases accepting aw ide range of enzyme-free acyl substrates are also involved in the biosynthesis of squalestatin. [24] As only lenormandin F ( 14)h as been isolated as ar epresentative of this type of compounds from H. fragiforme,w ea ssume that the majority of lenormandins are transformed into the respective fragirubrins. This hypothesis is consistent with observations made in H. lenormandii,w hich only Figure 3. Gene cluster comparison of the BGCse ncodingf or azaphilone production visualized with the clinker tool: A,the known mitorubrin BGC of T. marneffei; B, hfaza1 and hfaza2 from H. fragiforme; C,the known mrPig BGC from M. ruber. In accordance with the original publication, [22] no furtherl abels were assigned to the T. marneffei genes.
Scheme1.Biosynthetic hypothesis for the production and diversification of azaphilone-type compounds in H. fragiforme. Putativeintermediates that could not be isolateda nd were only detectable in traces by HPLC-MS or not detectable at all are shown in square brackets. Ri nt he free fatty acids (FFA) and respectives ide chain indicates variations in chain length, hydroxylationand unsaturation pattern depending on the final product.
produces the azaphilones nameda fter this fungus. [12] Thus, it can be speculated that H. lenormandii lacks homologueso f Hfaza2B. The diversity of mitorubrin-type azaphilones likely starts by the attachment of orsellinic acid (23)t ot he hydroxyl group at C-8 catalysed by Hfaza2E leadingt ot he intermediate 24.D ue to the structural differences between 23 and fattya cids, it seems unlikely that transfer reactions are conducted by the very same enzyme. Hence, we expect the acyltransferases from both clusters to be specific for different types of substrates. In addition, we assume that these enzymes are also highly stereoselectivec oncerning the substrate 22,a so nly as ingle enantiomer for each compound can be detected.
Due to the absence of another obviousO SAS encoded in the genome, the involvement of Hfaza2A in the synthesis of 23 seems most likely.T his is also supported by the strong homologyo fH faza2A to PKS12 of the mitorubrinol BGC in T. marneffei. The latter enzyme was shown to be crucial formitorubrinol 8 [25] and mitorubrinic acid 10 [25] biosynthesis by knockdown experimentsi nt he producing fungus, but the actualf unction could not be deduced from the data. Hence, it was speculated that PKS12 might be responsible for orsellinica cid biosynthesis. Unfortunately,t he authors only looked specifically for the absence of 8 and 10,b ut did not search for additional products in the extracts of their transformantst oc onfirmt his idea. [22] The lack of ar elease domain in the proposed OSAS Hfaza2A could be compensated by the acyltransferase Hfaza2E, which might directly load the ACP-bound 23A.Such areaction has already been suggested for the acyltransferase MrPigD,w hich presumably accepts ACP-bound fattya cids in Monascus pigment biosynthesis [20] and hasb een well characterised for the acyltransferase LovD involved in lovastatin biosynthesis. [26] Because 23B can also be detected as free acid in the stromatal crude extracts, we expect ah ydrolytic self-release mechanism analogously to truncated formso ft he methylorcinaldehyde synthase. [27] Intermediate 24 is then acetylated by the putative acetyl transferase Hfaza2B to give 25.M ass searches for compounds 24 and 25 revealed the presence of respective traces in the stromatal crude extracts (Table S5 in the Supporting Information).B ecause of the very small amount of compound the corresponding relationships can, however,n ot be verified. Thef ollowing step mighti nvolve deacetylationc arried outb y Hfaza2Ct oy ield mitorubrin (7), whichi nr eturni sh ydroxylated at C-14 putatively by theP 450m onooxygenase Hfaza2Ft o afford oneo ft he majors tromatal metabolites, mitorubrinol (8).
Mitorubrinol ( 8)t hen acts as the startingm aterial for the biosynthesis of 9 throught he acetylation of the C-14 alcohol performed either by Hfaza2B or ac luster-independenta cetyltransferase. In addition, 8 is also likelyt ob ean intermediate towardsm itorubrinic acid (10)v ia formation of the aldehyde mitorubrinal (11). We found ac orresponding peak in stromatal crude extracts exhibiting the expected mass spectra ( Figure S2, Ta bles S4, S5 in the Supporting Information). We wereu nable to isolate this compound, but standards of 11 obtained semisynthetically by oxidation of 8 with manganese oxide proved that the observed peak indeed corresponds to 11.
The respective biosynthetic steps to 10 might be carriedo ut by the action of the FAD-dependento xidoreductases Hfaza2J and Hfaza2M. As the T. marneffei BGC also encodes the production of 10,b ut lacks oxidoreductase genes, ad ifferent mechanism is also possible. Interestingly,t he mitorubrinol clustero f T. marneffei only leaves limited options to explain the conversion of 11 to 10.T he function of the highly conserved NADPHdependentd ehydrogenase still remainso bscure in all azaphilone biosynthetic pathways.H ence, it could theoretically also be involved in such oxidationsteps.
Finally,w ep ropose that the aldehyde functionality of 11 acts as an electrophile for the nucleophilic C-6 in all H. fragiforme monomeric azaphilones to afford dimers of the rutilin-(12-13)a nd hybridorubrin-type (1)(2)(3)(4), as already postulated by Quang et al. for rutilins Aa nd B. [28] The presence of rutilins in Hypoxylon rutilum as major stromatal metabolites could also indicate that condensation is enzyme-catalysed. [28] However,t his phenomenon could also be explained by the lack of an FADdependento xidoreductase to prevent the biosynthesis of a carboxylic acid functionality and leave the reactive aldehyde as the final enzymatic product. Thisisa lso consistentwith the observation that no carboxylated azaphilones have been detected in H. rutilum. [28] On the other hand, the mechanism could also involve radicals. The structures of the known bis-azaphilone diazaphilonic acid [29] or the azaphilone-derived nitrogencontaining chaetoglobins [30] ( Figure S3 in the Supporting Information) mightp ossibly be formed by recombination of radicals establishing the carbon-carbon bond connecting the substructures.
When comparing the biosynthetic machinery of mitorubrins in H. fragiforme and T. marneffei variousq uestions remain. The lack of monooxygenase genes in the cluster of the latter would prevent backbonea ssembly.F urthermore, monooxygenases are also very likely required to obtain mitorubrinol (8). Therefore, it seems likely that enzymes encoded outside of the BGC participate in the azaphilone formationo fT. marneffei. Based on our biosynthetic hypothesis, we propose that the production of lenormandin-type azaphilones requires only genes from hfaza1 and thus is likelye volvede arlier in these fungi. Consequently, hfaza2 might be acquired later,f or example, by horizontal gene transfer from T. marneffei or related fungi andh as proved for the fungus to be compatible with hfaza1.
The existence of intertwining secondary metabolite gene clusters has already been reported for the production of the structurally distantc ompounds fumagillin and pseurotin Ai n Aspergillus fumigatus. [25] However,t hese clusters werep hysically linked and consequently translocation of genes into neighbouring BGC can be explained by simple inversion of certain genomic regions within such as upercluster.R ecently,i ndependentg ene clusters have been demonstrated to be responsible for the formation of the azaphilone azasperpyranone Ai n Aspergillus terreus. While one BGC produced the azaphilone backbone, the other BGC afforded and processed 5-methyl orsellinic acid . The respective 5-MOA PKS contained a methyltransferase and thiolesterase domain and shared only little homology with Hfaza2A. Chem. Eur.J.2021, 27,1438-1450 www.chemeurj.org 2020 The Authors.C hemistry-AEuropeanJournal published by Wiley-VCH GmbH In addition to the elucidation of the biosynthetic pathway of azasperpyranon A, the regulatory network of the participating BGCs wasd eciphered by gene knockout of the encoded transcriptionf actors (TF) and gene expression analysis. It was shown that each BGC is upregulated by ac luster-specific TF, which in return are regulated by at hird TF located in one of the BGCs. [31] The regulatory network for azaphilone production in H. fragiforme could be highly similar,a st hree TFs have been identified across hfaza1 and hfaza2. We thus tried to experimentally link hfaza1 and hfaza2 with the knowna zaphilones by ectopic overexpression of the individual TF genes using previously described methods. [32] However,t his provedu nsuccessful. Knockout strategies are not viable in the Hypoxylaceae, as azaphilones are exclusively formedd uring stromatal development, which cannot be induced under laboratory conditions.
We could also find highly similar homologues of the two clusters in the taxonomically related fungus H. rickii and the more distantly related H. rubiginosum (data not shown), which are known to produce mitorubrins and/or the closely-related rubiginosins. [33] This observation further supports our theory about azaphilone biosynthesis in H. fragiforme and enables further options to study the pathways in detail.H owever,i tw ill be as pecial challenge to obtain final proof of the biosynthetic mechanism, because the stromatac an presently not be grown in the laboratory,a nd hence the only path forward would be heterologous expression.

Conclusion
We used ac ombination of classical natural product chemistry and state-of-the-art genome sequencing to deduce the biosynthesis of azaphilone pigments in H. fragiforme,d emonstrating the powerful combination of those two methods. We showed that both possible C-8 stereoisomerso fa zaphilones are produced in the stromata, which allowsf or assignment to subgroups:1 )the C-8(R)-configured azaphilones consist of the acyl-carrying lenormandinsa nd fragirubrins;2 )the group of C-8(S)-configured azaphilones carry an orsellinic acid moiety and belong to the family of mitorubrins and their fusion products, rutilins;a nd 3) the novel hybridorubrins, which are of mixed stereochemistry,a st heir buildingb locks originate from groups 1a nd 2. Furthermore, the hybridorubrins A( 1), C( 3), and D( 4)e xhibited high bioactivity against formation of S. aureus biofilms.
Examinationo ft he H. fragiforme genome revealed two BGCs to be most likely responsible for biosynthesis of azaphilone polyketides. The hfaza1 BGC is likely responsible for biosynthesis of the azaphilone backbone and addition of fatty acid moieties to yield group 1c ompounds.I np arallel, the hfaza2 BGC synthesizes orsellinic acid, whichi se sterified to the backbone to yield group 2a zaphilones and tailors the gained mitorubrins to obtain ah igh diversity of derivatives. We suggest that a spontaneous aldol condensation reactioni sr esponsible for the formation of hybridorubrina nd rutilin bis-azaphilones from reactive aldehyde intermediates in H. fragiforme;h owever,t his needs experimental verification. These results revealt he first example of two distant, cross-acting BGCs that enablealarge diversity of azaphilone products through natural mix-andmatch strategies.

Fungal material and extraction
To generate crude extract 1a ir-dried stromata (fruiting bodies, ca. 65 g) of Hypoxylon fragiforme were collected in 2017 from Fagus sylvatica in the vicinity of Braunschweig, Germany,b yL ucile Wendt. Extraction was performed by adding 500 mL of acetone, followed by ultrasonication at 40 8Cf or 1h.T his procedure was repeated twice. The extracts were combined and dried in vacuo, which led to approximately 6g of crude extract 1. For crude extract 2, about 25 go fa ir-dried stromata of Hypoxylon fragiforme were collected in 2016 from Fagus sylvatica in the vicinity of Lake Starnberg, Germany,b yL ucile Wendt. Extraction was performed as described above. This yielded approximately 3gof crude extract 2.