Branching and converging pathways in fungal natural product biosynthesis

In nature, organic molecules with great structural diversity and complexity are synthesized by utilizing a relatively small number of starting materials. A synthetic strategy adopted by nature is pathway branching, in which a common biosynthetic intermediate is transformed into different end products. A natural product can also be synthesized by the fusion of two or more precursors generated from separate metabolic pathways. This review article summarizes several representative branching and converging pathways in fungal natural product biosynthesis to illuminate how fungi are capable of synthesizing a diverse array of natural products.


Introduction
Structural diversity is a characteristic feature of naturally occurring organic compounds (natural products). Natural products possess a wide range of biological activities that can be attributed to their diverse and complicated molecular architectures. In nature, a great array of natural products are synthesized from a relatively small number of starting materials derived from primary metabolism. One reason behind the structural diversity of natural products is the presence of a significant number of core synth(et)ases, which are involved in the backbone synthesis of natural products. Core synth(et) ases include polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs), and terpene cyclases and are often responsible for the first committed step of a natural product biosynthesis, providing branching points between primary and secondary metabolic pathways. For example, PKSs accept limited starter and extender units, as represented by acetyl-coenzyme A (CoA) and malonyl-CoA, and perform repeated condensation of C 2 units to generate polyketide products with diverse chain lengths, oxidation levels, or cyclization patterns [1,2]. Although polyketide biosynthesis resembles that of fatty acids, an incomparable number of molecules can be produced by PKSs [3]. Likewise, terpene cyclases can cyclize achiral and linear substrates, such as polyprenyl pyrophosphates and (oxido)squalene, in a diverse manner, generating great structural diversity in terpene natural products [4].
In addition to structural diversification by core synth(et)ases, nature also uses other synthetic strategies to afford a variety of natural products [5]. Pathway branching is one such example, in which a molecule is synthesized as a common biosynthetic precursor and then undergoes distinct tailoring reactions to generate several different natural products (Fig. 1A). In another biosynthetic process, two or more molecules synthesized in separate pathways are combined together to provide a single natural product (Fig. 1B). In other words, convergent biosynthesis is nature's alternate approach to expand natural product diversity. Recent advances in genome sequencing technology and the development of molecular biological tools have allowed the elucidation of biosynthetic pathways and the synthesis of many natural products, including those synthesized in branching and converging pathways. In this review article, we summarize representative examples of branching and converging pathways in fungal natural product biosynthesis, with Page 2 of 24 Wei et al. Fungal Biology and Biotechnology (2022) 9:6 the aim to highlight nature's sophisticated strategies for synthesizing diverse molecules.

Branching pathways
As mentioned above, core synth(et)ases utilize substrates also used by the enzymes involved in primary metabolism but synthesize diverse molecules, thus greatly contributing to pathway divergence in natural product biosynthesis. As the functions and diversity of core synth(et) ases are well summarized elsewhere [1,2,4], this review focuses on the biosynthetic processes in which pathway branching occurs at the mid-or late-biosynthesis stage, especially those in which a single molecule serves as a common precursor of more than two different pathways.

Fungal xanthones
Xanthone is a tricyclic organic molecule, and xanthone derivatives are a class of natural products that exhibit a wide range of biological activities such as anti-bacterial activities and cytotoxicity [6,7]. Fungi are prolific producers of natural products with a xanthone (including di-and tetrahydroxanthone [THX]) scaffold (Fig. 2). In recent years, biosynthetic gene clusters of several fungal xanthones have been identified and characterized, revealing how fungi produce diverse xanthone compounds [8][9][10][11][12][13][14][15][16]. Although the xanthone skeleton can be synthesized through several different pathways in nature, all fungal xanthones are considered to be derived from the anthraquinone chrysophanol (1), which is generated by seven enzymes from one acetyl-CoA molecule and seven malonyl-CoA molecules [12].
Blennolides A-C (2-4) are representative fungal THXs and the precursors of many THX dimers [17], such as secalonic acids and neosartorin [18][19][20]. In the biosynthesis of neosartorin, chrysophanol (1) is transformed into monodictyphenone (5) and isomonodictyphenone (6), providing a branching point (Fig. 3A) [12]. Previously, it was thought that NsrF serves as a Baeyer-Villiger monooxygenase to insert an oxygen atom at two different positions in 1 [12,15]. However, the recent in-depth characterization of GedF, an NsrF homologue in geodin biosynthesis, indicated that chrysophanol is not a direct NsrF substrate [21]. It is currently proposed that 1 is first reduced to chrysophanol hydroquinone (7) by the shortchain dehydrogenase/reductase (SDR) NsrR, which is also required for chrysophanol formation. NsrF then accepts 7 to deprotonate from the C-10 position, followed by single electron transfer (SET) from the carbanion to molecular oxygen to generate the peroxy anion 8. The peroxy anion 8 subsequently attacks either the C-4a or C-10a position, followed by bond migration to yield 5 or 6, respectively (Fig. 3B). Interestingly, AacuH, the NsrF homologue engaged in secalonic acid biosynthesis, selectively generates 5 [16], indicating that this enzyme controls Michael addition by the peroxy anion.
NsrQ subsequently inverts the stereochemistry at the C-6 position to give 14. NsrO then reduces the C-5 carbonyl of 14 to the (5S)-hydroxy group to yield 15, followed by spontaneous heterocyclization to yield 2. This cyclization appears to occur in a stereoselective manner to avoid the 1,3-axial-pseudoaxial interaction between the C-6 methyl group and the C-10a methyl ester. Intriguingly, CPUR_05418, a NsrO homologue in Claviceps purpurea, accepts the first isomerized product of NsrQ, 13, and serves as the C-5S reductase to generate 11 with a (5S,6R,10aS) configuration via 16. Furthermore, the Aspergillus aculeatus secalonic acid pathway adopts two SDRs with distinct stereoselectivity [14]. One SDR, AacuD, is homologous to NsrO and responsible for blennolide A formation, whereas the other SDR AacuF conducts the C-5R reduction, leading to the production of 3 possessing the (5R,6S,10aR) configuration via 17 [16]. In addition, 10 undergoes similar reactions catalyzed by NsrK, NsrQ, and NsrO to yield blennolide C (4) and its C-10a epimer epi-blennolide C (18) [15]; cyclization occurs in a nonselective manner, probably due to lack of the C-6 methyl group. Chrysophanol (1) appears to be the precursor of many other fungal xanthones whose biosynthetic pathways have not been fully elucidated. On the basis of the dicerandrol structure [23,24], their biosynthesis should involve a precursor that is analogous to blennolide A (2) but contains a hydroxymethyl group instead of a methyl ester, 19 (Fig. 3A), which is derived from a monodictyphenone analogue 20 with a hydroxymethyl group. Compound 20 could also be utilized as a common biosynthetic intermediate for other fungal xanthones, such as cryptosporioptide (via 21) [14] and tajixanthone/shamixanthone (via 22) [8], although the mechanism to provide the hydroxymethyl group has yet to be elucidated.

Ergot alkaloids
Ergot alkaloids are a diverse group of fungal indole alkaloids with pharmaceutical and agricultural value [25,26]. They are synthesized by filamentous fungi of several different genera and provide another example in which pathway branching significantly contributes to structural diversification. All ergot alkaloids are believed to be biosynthesized from chanoclavine-I aldehyde (23) as a common precursor (Fig. 4A), which is generated by five dedicated enzymes using l-tryptophan and dimethylallyl pyrophosphate (DMAPP) as starting materials.
Lysergic acid (24) undergoes further modification by several NRPSs to yield ergopeptines and lysergic acid amides (Fig. 4B). In C. purpurea, 24 is activated by and loaded onto a single-module NRPS, LPSB, with adenylation (A)-thiolation (T)-condensation (C) domain organization [33][34][35]. The activated 24 is then transferred to the assembly line of LPSA1 or LPSA2, two other NRPSs homologous to each other [35]. After the incorporation of three amino acids, the elongated peptide chain is released by the terminal condensation-like (C T ) domain to furnish the diketopiperazine structure. The resultant ergopeptams are then transformed into ergopeptines, namely ergotamine (34) and ergocryptine (35), by the αKG-dependent dioxygenase Cp_EasH [36]. LPSA1 and LPSA2 have the same domain architecture but utilize different combinations of amino acids. It has been proposed that the three LPSA1 A domains activate l-alanine, l-phenylalanine, and l-proline, and that those of LPSA2 recognize l-valine, l-leucine, and l-proline [37]. Furthermore, LPSB-loaded 24 can be utilized by another single module NRPS with A-T-C-reductase (R) domain organization, LPSC, thus incorporating l-alanine and forming ergometrine (36) by reductive release of the peptide chain [34]. In some fungi, such as Claviceps paspali, 36 can be further transformed into lysergic acid α-hydroxyethylamide (37; LAH), which is thought to be produced by the FMO EasO and α/β hydrolase-fold enzyme EasP [38].

Trichothecenes
Trichothecenes are a large group of sesquiterpenoid mycotoxins produced by a variety of fungi of different genera and represented by T-2 toxin (38) and satratoxin H (39) (Fig. 5) [39,40]. The trichothecene core structure features an epoxide ring at the C-12/C-13 positions, which is crucial for their biological activities. The first committed biosynthesis step of all trichothecenes is the cyclization of farnesyl pyrophosphate (FPP) into the sesquiterpene hydrocarbon trichodiene (40), which is catalyzed by the terpene cyclase Tri5 (Fig. 5A) [41,42]. The sesquiterpene 40 then undergoes multiple oxidations catalyzed by the cytochrome P450 monooxygenase Tri4.

Dioxafenestrane sesquiterpenoids
The dioxafenestrane sesquiterpenoid pathways provide another example in which fungal sesquiterpenoids are generated in branching pathways ( Fig. 6) [63,64]. The biosyntheses of penifulvins and asperaculins both begin with the cyclization of FPP into the sesquiterpene silphinene (59), which is synthesized by the terpene cyclase PeniA/AspeG. Subsequently, the P450 PeniB/AspeF performs a series of oxidative reactions to yield the tetracyclic sesquiterpenoid 60. Compound 60 then undergoes Baeyer-Villiger oxidation in both pathways, but an oxygen atom is inserted at two distinct positions by the FMO, PeniC or AspeB, to provide penifulvin A (61) and 9-deoxyasperaculin A (62), respectively. Interestingly, although PeniC and AspeB, the key enzymes for the pathway divergence in the dioxafenestrane sesquiterpenoid biosynthesis, are both FMOs, they are phylogenetically distantly related to each other. In the late-stage of biosynthesis, 61 is converted into penifulvin G (63) by two αKG-dependent dioxygenases PeniD and PeniF and the acetyltransferase PeniE, whereas AspeC and AspeD, which are PeniF and PeniD homologues, respectively, oxidize 62 to form asperaculin C (64).

Meroterpenoids
Meroterpenoids are natural hybrid products that are in part derived from terpenoid biosynthesis and ubiquitously distributed both in primary and secondary metabolism [71][72][73]. Fungi are especially prolific producers of meroterpenoids, and fungal meroterpenoids exhibit enormous structural diversity, often with complicated molecular architectures and a broad range of biological activities. Fungal meroterpenoid biosynthesis occurs in a highly branching manner, and branching can be observed at different biosynthesis stages, thus allowing for the biogenesis of diverse metabolites from a small number of starting materials. One of the most intensively studied classes of fungal meroterpenoids comprises those derived from the aromatic polyketide 3,5-dimethylorsellinic acid (72; DMOA), and many DMOA-derived meroterpenoids possess a unique and complicated molecular skeleton (Fig. 8). A key factor generating the structural diversity of DMOA-derived metabolites is the existence of terpene cyclases with different activities. In the biosynthesis of andrastin A, austinol, and terretonin, (R)-epoxyfarnesyl-DMOA methyl ester (73) serves as a common intermediate but is cyclized into different products, generating the biosynthesis branching point (Fig. 9A) [74,75]. Three terpene cyclases, AdrI, AusL, and Trt1, all accept 73 as a substrate and cyclize it into the tetracyclic carbocationic species 74; however, in the final step of the reaction, these enzymes perform deprotonation at different positions to yield andrastin E (75), protoaustinoid A (76), and preterretonin A (77), respectively. Intriguingly, the terpene cyclase involved in novofumigatonin biosynthesis utilizes (S)-epoxyfarnesyl-DMOA (78) with a free carboxy group as a substrate to form asnovolin H (79) [76], although methyl-esterification is essential for the activities of AdrI, AusL, and Trt1 [74,75]. Additionally, based on the cyclized product structures, AdrI, AusL, and Trt1 accept (R)-epoxide, whereas (S)-epoxide is used by NvfI, indicating the presence of epoxidases with different stereoselectivities. In a recent study by Mitsuhashi et. al., synthetic epoxyfarnesyl-DMOA methyl ester molecules, including unnatural isomers, were reacted with a series of meroterpenoid cyclases, resulting in the production of several new meroterpenoid species generated by new cyclization modes [77]. Thus, enzymes that catalyze cyclization reactions in an unprecedented manner could be further discovered from unelucidated DMOA-derived meroterpenoid pathways.
Pathway branching in DMOA-derived meroterpenoid pathways also occurs at a later biosynthesis stage. Preaustinoid A1 (80) is a common biosynthetic precursor of several meroterpenoids (Fig. 9B), such as austinol, paraherquonin, and berkeleyacetals. The αKG-dependent dioxygenases AusE and AusE′, which are involved in the biosynthesis of austinol and acetoxydehydroaustin, respectively, accept 80 to catalyze two successive oxidations to afford preaustinoid A3 (81) with the characteristic spirolactone system via preaustinoid A2 (82) [78,79]. Meanwhile, PrhA, a homologue of AusE/AusE′, utilizes the same substrate but performs a distinct oxidative rearrangement to yield berkeleydione (83) harboring the cycloheptadiene core via berkeleyone B (84) [79]. Intriguingly, AusE′ and PrhA share > 90% sequence identity, and a structural biology study revealed that only three amino acid residues in the enzyme active site are crucial for product selectivity [80]. Andiconin (85) serves as a common biosynthetic intermediate of anditomin and emeridone F (Fig. 9C) [81,82]. In the anditomin pathway, AndJ, which acts as a Baeyer-Villiger monooxygenase, transforms 85 into andilesin D (86) with a seven-membered lactone ring [81]. Emeridone biosynthesis lacks an AndJ homologue, and 85 is accepted by the SDR SptI to be converted to andiconin B (87) [82]. It should be noted that the SptI homologue in the anditomin pathway, AndI, does not accept 85 as a substrate but acts only after the AndJ-catalyzed reaction. Polyketides other than DMOA can also be employed as the non-terpenoid portion of fungal meroterpenoids. Orsellinic acid serves as the precursor of diverse fungal meroterpenoids such as ascofuranone (88), ascochlorin (89), chartarlactam A (90), stachybisbin B (91), and SMTP-0 (92) (Fig. 10). Among these metabolites, the complete biosynthetic pathways of 88 and 89 have recently been elucidated [83], exemplifying pathway branching occurring in a single fungus. In the biosynthesis of these two meroterpenoids, orsellinic acid is converted to ilicicolin A epoxide (93) as a common precursor via ilicicolin B (94), which is also a biosynthetic intermediate of many other orsellinate-derived fungal meroterpenoids [84]. In the ascochlorin pathway, the terpene cyclase AscF cyclizes the farnesyl moiety of 93 to provide ilicicolin C (95) with a cyclohexanone ring, which subsequently undergoes dehydrogenation catalyzed by the P450 AscG to yield 89. Ascofuranone biosynthesis involves three dedicated enzymes encoded by a separate gene cluster from that of 89. In this process, 93 is first hydroxylated by the P450 AscH, followed by heterocyclization catalyzed by AscI to generate ascofuranol (96). Interestingly, AscI does not display sequence similarity with AscF and the terpene cyclases involved in the DMOA-derived metabolite pathways but is homologous to CtvD and AurD, the epoxide hydrolases involved in the biosynthesis of citreoviridin and aurovertin, respectively [85,86]. Finally, the SDR AscJ performs an alcohol dehydrogenation reaction to complete the biosynthesis. Triacetic acid lactone (97; TAL) is another polyketide used as a fungal meroterpenoid building block. In the biosynthetic processes of known TAL-derived meroterpenoids, 97 undergoes geranylgeranylation and epoxidation to provide a substrate for terpene cyclases, 98 (Fig. 11) [87]. Chevalone biosynthesis adopts the terpene cyclase Cle3 to yield the pentacyclic product chevalone E (99), whereas Sre3, which is responsible for the sartorypyrone pathway, synthesizes sartorypyrone D (100), which possesses the monocyclic terpenoid moiety. The terpene cyclase for aszonapyrone biosynthesis appears to have another function in producing the tetracyclic aszonapyrone B (101) [88]; however, this function has yet to be identified.
Another example in which terpene cyclases are responsible for pathway divergence can be found in the biosynthesis of chrodrimanins and talaromyolides (Fig. 12) [89,    (97), 5-methyl TAL (105), is often found in fungal meroterpenoid structures, and in recent years, the complete biosynthetic pathways of several 5-methyl TAL-derived meroterpenoids have been revealed (Fig. 13) [91][92][93]. In the biosynthesis of setosusin (106), 105 undergoes a series of reactions similar to those in TAL-derived meroterpenoid pathways to provide a precyclized intermediate 107, which is accepted by the terpene cyclase SetH to give 108 with a tricyclic terpenoid portion [93]. This cyclized product is also used as the precursor for other fungal metabolites, such as brevione E (109) [94], although brevione E biosynthetic genes have not been reported. The biosynthesis of other 5-methyl TAL-derived meroterpenoids, such as subglutinols, higginsianins, and metarhizins, branches from the setosusin/brevione pathways at an early biosynthesis stage. In the biosynthetic processes of these molecules, epoxidation occurs at the second double bond from the geranylgeranyl chain terminus of 110 to afford 111, which is cyclized into metarhizin C (112) with a bicyclic terpenoid moiety. In a recent study by Tsukada et. al. [92], biosynthetic genes for metarhizin C-derived metabolites from five fungal species were examined and heterologously expressed, leading to the production of several known and new meroterpenoids. The FADdependent oxidoreductase DpxxF (xx = ma, as, or ch) conducts oxidative cyclization to synthesize subglutinols A (113) and B (114) from 112. DpasF or DpchF can also install a carbonyl group onto the dimethylallyl portion of 112 to yield a previously undescribed meroterpenoid 115. The SDRs DpxxG and DpxxH (xx = fg, mp, or ch) invert the stereochemistry of 112 at the C-8 position to give higginsianin B (116). Compound 116 is oxidized by DpasF or DpchF to provide higginsianin A (117), which is the C-8 epimer of 113. Compound 116 also serves as the precursor for three new meroterpenoids 118-120, which are synthesized by catalysis of the methyltransferase DpfgI or DpmpI, the P450 DpfgJ or DpmpJ, and the methyltransferase DpfgK.

Converging pathways
Heterodimeric compounds, which are formed by the fusion of two metabolites derived from two distinct pathways, are often obtained during natural product isolation.
Most heterodimeric natural products are often spontaneously formed due to reactive structural elements in the monomeric natural products that compose the heterodimer. However, in some biosynthetic processes, two or more compounds synthesized via separate pathways undergo an enzyme-catalyzed fusion reaction, expanding the structural diversity of natural products. This section summarizes representative converging pathways observed in fungal natural product biosynthesis and the enzymes responsible for the hybridization of two distinct natural products.

Addition of a polyketide chain
One representative converging pathway in fungal natural product biosynthesis is the addition of a polyketide chain to a molecule synthesized in a separate biosynthetic route. One example is the biosynthesis of harzianum A (54), as mentioned above (Fig. 5C) [58], and similar reactions to attach a polyketide chain are found in many fungal natural product pathways (Fig. 15). In the last step of lovastatin biosynthesis, the acyltransferase LovD transfers the polyketide chain synthesized by the PKS LovF to introduce a methylbutanoyl moiety to monacolin J acid (134) to yield lovastatin acid (135) (Fig. 15A) [109]. Squalestatin biosynthesis involves the acyltransferase Mfm4 that utilizes the tetraketide CoA 136 as a substrate to yield squalestatin S1 (137); 136 is synthesized by the PKS squalestatin tetraketide synthase (SQTKS) and the CoA ligase Mfm9 (Fig. 15B) [110,111]. In fungal meroterpenoid calidodehydroaustin biosynthesis (Fig. 8), the acyltransferase AusQ performs a transacylation reaction by utilizing the polyketide chain generated by the PKS AusV to give precalidodehydroaustin (138) from 1,2-dihydro-7-hydroxydehydroaustin (139) (Fig. 15C) [112]. AusQ is also known to catalyze acetylation reactions at the same position. The biosynthesis of another fungal meroterpenoid, fumagillin, adopts the acyltransferase Fma-AT to install the dodecapentaenoyl moiety, which is synthesized by the PKS Fma-PKS, onto fumagillol (140) to produce prefumagillin (141) (Fig. 15D) [113].
Calbistrins are structurally similar to lovastatin in that they have a polyketide chain attached to a decalin core; however, the biosynthesis of calbistrins and lovastatin somewhat differs. CalD, the enzyme responsible for the acylation reaction in the calbistrin pathway, performs a similar reaction to LovD to introduce the hexaketide moiety to decumbenone A (142) to yield 143 (Fig. 15E), although CalD does not exhibit sequence similarity with LovD [114]. Interestingly, the calbistrin biosynthetic gene cluster encodes a LovD homologue, CalJ, which is unable to perform acylation reactions but instead catalyzes the hydrolysis of the ester bond formed by the CalD-catalyzed reaction. Another intriguing feature of calbistrin biosynthesis is that both the decalin and polyene portions are generated from the single PKS CalA. The decalin portion is synthesized in a CalA and transacting enoylreductase (trans-ER) CalK collaboration, whereas the polyene hexaketide is formed by CalA and the trans-acting methyltransferase CalH. Thus, the PKS CalA contributes both to pathway branching and convergence in the biogenesis of calbistrins.
Gregatin biosynthesis provides a similar example in which a single PKS is responsible for the synthesis of two distinct polyketide chains that are eventually fused together (Fig. 15F) [115]. GrgA, the PKS involved in the gregatin pathway, synthesizes the longer pentaketide chain in the presence of the trans-ER GrgB, whereas it produces a shorter diketide chain when GrgB is absent. These two polyketide chains are fused by the α/β hydrolase fold enzyme GrgF to generate 144.

Incorporation of non-proteinogenic amino acids into NRPS assembly lines
One key feature of nonribosomal peptide biosynthesis, distinct from that of ribosomal synthesis, is that nonproteinogenic amino acids and non-amino acid molecules can be directly used as building blocks for peptide chain synthesis [116]. Non-standard amino acids found in nonribosomal peptides are often synthesized by dedicated enzymes prior to acceptance by the NRPS assembly line, thus exemplifying converging pathways in natural product biosynthesis.
Pneumocandin A 0 biosynthesis is one of the most complicated examples in which the dedicated syntheses of unusual amino acids are required (Fig. 16). Pneumocandin A 0 (145) is an antifungal agent in the echinocandin class, obtained from the fungus Glarea lozoyensis [117]. Pneumocandin A 0 biosynthesis is initiated with the synthesis of dimethylmyristoyl-CoA (146) by the PKS GloL (GLPKS4) and the CoA ligase GloD (GLligase), and 146 is loaded onto the NRPS GloA (GLNRPS4) as a starter unit for peptide chain synthesis [118,119]. GloA consists of six modules, five of which accept a non-proteinogenic amino acid. Among these five non-proteinogenic amino acids, ornithine, accepted by the first A domain, is an important primary metabolite sufficiently present in cells. However, the remaining four amino acids need to be prepared by utilizing dedicated enzymes encoded by the pneumocandin biosynthetic gene cluster. The third A domain activates 4R-hydroxy-l-proline (147), which is synthesized by l-proline hydroxylation catalyzed by the αKG-dependent dioxygenase GloF (GLOXY2) [120]. The fourth A domain accepts 3S-hydroxy-l-homotyrosine (148), which is synthesized from 4-hydroxyphenylpyruvic acid (149; 4-HPPA) by five enzymes, GloH, -J, -I, -G, and -M (GLHtyA, -HtyD, -HtyC, -HtyB, and -OXY1) [119,121,122]. The αKG-dependent dioxygenase GloE (GLOXY3) hydroxylates l-glutamine to provide 3R-hydroxy-l-glutamine (150) [123], which is loaded onto the fifth module. The sixth A domain substrate, 3S-hydroxy-4R-methyl-l-proline (151), is generated from l-leucine. In this transformation, l-leucine is oxidized by the αKG-dependent dioxygenase GloC (GLOXY4), and the resultant cyclic imine 152 is reduced to 4R-methyl-l-proline (153) by an uncharacterized enzyme [119,122]. GloF, which is also used in 4R-hydroxy-l-proline synthesis, hydroxylates 153 to yield 151 [120]. The linear peptide chain synthesized by GloA undergoes C T domain-catalyzed cyclization to be released from the NRPS enzyme to provide the cyclic peptide 154. The cyclic peptide undergoes further oxidative tailoring reactions to finally provide 145. Similar biosynthetic strategies incorporating unusual amino acids or starter units are also found in other fungal NRPS pathways, such as the biosynthesis of echinocandins [121,122] and leucinostatins [124]. Some fungal PKS-NRPS hybrids also accept an unusual amino acid synthesized by dedicated enzymes encoded by a gene cluster containing the PKS-NRPS gene (Fig. 17). In the biosynthesis of Sch 210972 (155), γ-hydroxymethyl-l-glutamic acid (156) is accepted by the PKS-NRPS CghG A domain (Fig. 17A) [125]. The biosynthesis of this non-proteinogenic amino acid involves the condensation of two pyruvic acid molecules catalyzed by the aldolase CghB, followed by a stereoselective transamination reaction. It has been proposed that transamination is performed by a transaminase encoded outside the biosynthetic gene cluster. Another similar example is found in the biosynthesis of oxaleimides, in which the PKS-NRPS PoxE utilizes (S,E)-2-aminodec-4-enoic acid (157) as a substrate of the A domain to yield 158 (Fig. 17B) [126]. The amino acid backbone structure is generated by the PKS PoxF, which is modified by the P450 PoxM and the aminotransferase PoxL.

Communesin B
Communesins are dimeric indole alkaloids isolated from Penicillium fungi [127][128][129][130]. Some of them, such as communesin B (159), harbor the hexadienoyl moiety, which originates from the polyketide pathway. The communesin pathway involves both pathway branching and convergence to furnish the unique molecular architecture [131,132]. The communesin core structure is derived from two l-tryptophan molecules, each undergoing different pathways to provide monomeric units of communesins (Fig. 18) [131]. In one branch, l-tryptophan is converted to tryptamine (160) by the tryptophan decarboxylase CnsB. l-Tryptophan is also transformed into aurantioclavine (161) via 4-dimethylallyl-l-tryptophan (162; 4-l-DMAT) by catalysis with the DMATS CnsF, the FMO CnsA, and the catalase CnsD. In the subsequent biosynthetic step, 160 and 161 undergo an oxidative coupling reaction catalyzed by the P450 CnsC to afford the communesin backbone structure 163 [132], which is further converted to communesin I (164) by the methyltransferase CnsE and the αKG-dependent dioxygenase CnsJ. Finally, the acyltransferase CnsC performs N-acylation by utilizing the hexadienoyl chain synthesized by the PKS CnsI to yield 159.

Flavunoidine
Flavunoidine (165) is obtained by genome mining in the fungus Aspergillus flavus and features a sesquiterpenoid core modified with dimethylcadaverine (166) and 5,5-dimethyl-l-pipecolic acid (167) [133]. It is a rare example of terpenoid-amino acid conjugated natural products. Compound 165 is generated by the fusion of three separately synthesized building blocks (Fig. 19). In one of the three pathways, l-lysine is converted into 166 by the N-methyltransferase FlvH and the decarboxylase FlvG. Meanwhile, the terpene cyclase FlvE cyclizes FPP into (1R,4R,5S)-( +)-acoradiene (168). The P450 FlvD accepts the sesquiterpene hydrocarbon 168 as a substrate to transform it into 169 with a tetracyclic cage. The sesquiterpenoid is further oxidized to the carbocationic species 170, which then reacts with 166; this nucleophilic addition is mediated by the terpene cyclase-like enzyme FlvF. This hybrid molecule 171 is hydroxylated by the P450 FlvC to generate 172, which is further combined with 167 to yield the end product 165. The unusual amino acid 167 is synthesized from O-acetyl-l-homoserine (173) and α-ketoisovaleric acid (174) by FlvA, the chimeric protein of PLP-dependent lyase and αKGdependent dioxygenase, and the SDR FlvB. Compound 167 is then loaded onto the single-module NRPS FlvI, which performs the esterification reaction to afford 165.

Other converging pathways
In addition to the biosynthetic pathways described above, fungi accomplish pathway convergence in diverse ways. Fungal THX dimer biosynthesis is one such example. As mentioned above, fungal THX biosynthesis occurs in a highly branching way (Fig. 3); however, these monomeric THXs often undergo (hetero)dimerization reactions to provide THX (hetero)dimers, and the biosynthesis of THX heterodimers can be regarded as converging pathways. The P450 NsrP performs oxidative coupling of 5-acetylblennolide A (175) and blennolide C (4) to complete neosartorin (176) biosynthesis (Fig. 20A) [12]. Similarly, AacuE, an NsrP homologue responsible for secalonic acid biosynthesis, can accept blennolides A (2) and B (3) as substrates to generate secalonic acid F (177) (Fig. 20B) [16]. It should be noted that these P450s appear to utilize several different substrates to perform both homo-and heterodimerization reactions. For  example, AacuE can synthesize secalonic acid D [16], a homodimer of 3 and a major product of the A. aculeatus secalonic acid pathway [134]. Fumonisins are mycotoxins mainly derived from Fusarium species and possess characteristic tricarballylic ester moieties [135]. The tricaballylate portion is derived from aconitic acid (178), which is activated by the standalone A domain Fum10 and subsequently loaded onto the NRPS-like enzyme Fum14 with T-C domain organization (Fig. 21) [136]. It has been proposed that the tricaballylate bound to Fum14 undergoes enoylreduction catalyzed by the iron-containing dehydrogenase/reductase Fum7 [137]. The Fum14 C domain would then utilize hydrolyzed fumonisin B 3   Maleidrides are carbocyclic molecules harboring one or two maleic anhydride groups [138], and their biosynthesis involves an enzyme-catalyzed heterodimerization reaction to synthesize the core structure. In the biosynthesis of byssochlamic acid (183), butenyl carboxymethyl maleic anhydride (184) and its decarboxylated form 185 are enzymatically hybridized (Fig. 22A) [139]; the maleic anhydride derivative 184 originates from hexenoate synthesized by a PKS and oxaloacetate. The reaction to generate 183 is catalyzed by the ketosteroid isomerase (KSI)-like proteins BfL6 and BfL10, and the product yield is significantly increased in the presence of phosphatidylethanolamine-binding proteins (PEBPs) BfL5 and BfL9. Deoxyscytalidin (186) biosynthesis occurs in a similar manner to 183, except that octanoate is used as a building block instead of hexenoate (Fig. 22A) [140]. The fusion reaction in the scytalidin pathway is performed by the KSI ScyR6 and the PEBPs ScyL1 and ScyR12. Rubratoxin A biosynthesis constructs the maleidride scaffold in a similar manner with the aid of the KSI RbtR and the PEBPs RbtM and RbtO by utilizing two carboxymethyl maleic anhydride molecules with slightly different side chains, 187 and 188, to afford prerubratoxin A1 (189) (Fig. 22B) [141].
Xenovulene A and eupenifeldin (190) are polyketidederived fungal meroterpenoids synthesized in a distinct manner from those described above. The biosynthesis of these molecules employs a class I terpene cyclase, and therefore, the resultant terpene products need to be hybridized with a polyketide molecule at a later biosynthesis stage. In the biosynthesis of xenovulene A, the sesquiterpene hydrocarbon humulene (191) and the tropolone quinomethide (192) undergo a hetero [4 + 2] cycloaddition reaction catalyzed by AsR5 to give 193 with the meroterpenoid scaffold (Fig. 23A) [142]. The eupenifeldin pathway also adopts a homologous enzyme EupF or EupfF, which catalyzes hetero [4 + 2] cycloaddition reactions at two distinct positions of the sesquiterpene alcohol humulenol (194) using two tropolone orthoquinone molecules (195) (Fig. 23B) [143,144].
Azasperpyranone A (196) is an azaphilone family natural product biosynthesized by the fusion of preasperpyranone (197) and a polyketide-derived aromatic aldehyde 198, which, unlike the other cases covered in this review, are generated by two separate gene clusters (Fig. 24) [145]. However, a gene deletion experiment indicated that ATEG_03636 is responsible for this heterodimerization reaction. ATEG_03636 does not exhibit sequence similarity to any characterized proteins, and the detailed mechanism by which the enzyme performs the hybridization reaction has yet to be elucidated.

Concluding remarks
In this review, we summarize several representative examples of branching and converging pathways found in fungal natural product biosynthesis. Core synth(et)ases, such as PKSs, NRPSs, and terpene cyclases, undoubtfully contribute to the structural diversity of natural products by creating a large number of branching pathways, which is achieved by the diverse programming rules adopted by core synth(et)ases. However, pathway branching and converging at mid-and late-stage biosynthesis, which are performed by a variety of tailoring enzymes, are other major factors that generate the natural product diversity and complexity. It should be noted that only slight genetic changes can sometimes cause a pathway divergence; for example, the divergence of paraherquonin and acetoxydehydroaustin pathways in Penicillium brasilianum stems from the change of a few amino acid residues in the key dioxygenases [79]. The presence of diverse branching and converging pathways might be attributed to nature's evolutionary efforts to synthesize diverse molecules from a relatively small number of available starting materials. Given the recent rapid accumulation and advances in genome editing and heterologous expression technologies [146,147], an increasing number of branching and converging biosynthesis pathways will be identified and characterized in the near future, which could lead to the discovery of novel biosynthetic mechanisms for pathway branching or convergence. The elucidation and understanding of such biosynthetic pathways will also facilitate the construction of artificial branching and converging metabolic pathways, as exemplified by the expansion of meroditerpenoid pyrone pathways [92], the evolution of the acyltransferase LovD into a simvastatin synthase [148], and the side-chain engineering of pneumocandins [149]. Future studies might discover more branching and converging pathways and rationally engineer biosynthesis to further expand natural product diversity and provide useful molecules in an efficient and selective manner.