As cytotoxic PQs, phaeosphaerins can be photoactivated to generate ROS, which accumulates in the lysosomes of human tumor cells and induces apoptosis of these cells. Phaeosphaerins A-F (1–5) have been isolated from Phaeosphaeria sp. (Supplementary Figure S2; Li et al., 2012).

Calphostin A, B, C, D, and I (6–10; Supplementary Figure S3) have inhibitory activity against protein kinase C (PKC), and Calphostin C is the strongest inhibitor (Kobayashi et al., 1989b). Calphostin C induces cell apoptosis, both via the dual leucine zipper-bearing kinase and c-Jun N-terminal kinase (JNK) signaling pathways (Robitaille et al., 2008) and by inducing endoplasmic reticulum stress (Kaul and Maltese, 2009). Calphostin C also inhibits the expression of Chlamydophila pneumoniae-associated intracellular adhesion molecule-1 and blocks NF-kappaB translocation (Vielma et al., 2003a,b). These potentially useful biological activities have stimulated extensive research aimed at identifying alternative sources of Calphostins and improving their biosynthetic efficiency. For instance, Calphostins were first isolated from Cladosporium cladosporioides culture medium, during screening for protein kinase C (PKC) inhibitors (Iida et al., 1989; Kobayashi et al., 1989a), and other Cladosporium species can also produce calphostin derivatives. In addition, C. cucumwinum from etiolated cucumber seedlings produces Cladochrome A (12; Overeem et al., 1967) and C. cladosporioides can biosynthesize diverse PQs after epigenetic modifier treatment, including Cladochromes A (12), B (13), D (14), E (15), F (16), G (17), and calphostin B (Williams et al., 2008). Phaeosphaeria sp. also produces (+)-calphostin D (Li et al., 2012). The homologous PQs to calphostin, phleichrome (18) and isophleichrome (19) have been isolated and characterized from the plant pathogens C. phlei and C. herbarum (Yoshihara et al., 1975; Robeson and Jalal, 1992). Isophleichrome is also produced by C. cucumwinum in low yield (Arnone et al., 1989). These phleichromes are stronger inhibitors of HIV-1 and PKC than calphostin, and have antimicrobial, and antitumor activities based on their photodynamic properties (So et al., 2018).

Cercosporin (20) is another well-known photoactivated PQ toxin from diverse Cercospora species (Supplementary Figure S3). These phytopathogens cause leaf spot and blight diseases on diverse crop species (Daub, 1981; Daub and Ehrenshaft, 2000). Cercosporin also has antileishmanial activity and cytotoxicity for glioblastoma, breast adenocarcinoma, and human tumor cell lines (Kumarihamy et al., 2012; Grigalavicius et al., 2019; Mastrangelopoulou et al., 2019). Photoactivation of cercosporin has applications in the biosynthesis of medicinal compounds. For instance, the cercosporin-photocatalyzed sp³(C-H) activation reaction can be used to synthesize pyrrolo[3,4-c]quinolones, which are the backbone structures of various bioactive compounds, such as “A Disintegrin and Metalloproteinase with Thrombospondin Motif” (ADAMTS) inhibitors (Li et al., 2019). The cercosporin-photocatalyzed oxidation reaction can be also employed to synthesize different kynurenine derivatives and the relevant peptides, which present the neuroprotective bioactivity (Yuan et al., 2022). The great pharmacological value of cercosporin and its derivatives has stimulated their isolation from diverse microorganisms. Cercospora kikuchii was the first reported cercosporin producer (Kuyama and Tamura, 1957) and at least 24 other Cercospora species can produce cercosporin (Assante et al., 1977). Various derivatives of cercosporin with modified functional groups have been also isolated, such as (+)-isocercosporin (21) from Cercospora kikuchii, and (+)-Isocercosporin and acetylisocercosporin (22) from the plant pathogen Scolecotrichum graminis Fuckel (Tabuchi et al., 1994). Septoria pistaciarum produces (+)-Cercosporin, (+)-14-O-Acetylcercosporin and (+)-di-O-Acetylcercosporin (Kumarihamy et al., 2012). In a study characterizing the cercosporin toxin biosynthesis (CTB) gene cluster in the phytopathogenic genus Cercospora, de Jonge et al. also found that treatment of Colletotrichum fioriniae with trichostatin A stimulates the production of a cercosporin-associated PQ (de Jonge et al., 2018), thus indicating that Cercospora has the potential to produce cercosporin-associated PQs.

Elsinochrome A is another well-known photoactivated PQs toxin from the phytopathogenic Elsinoe species (Supplementary Figure S3; Weiss et al., 1957). Similarly to Cercosporin, Elsinochrome A is cytotoxic to the mouse fibroblast cell line L929 and human HeLa cell line KB3.1 (Surup et al., 2018). Elsinochrome A also has antimicrobial activity against Bacillus subtilis, Mucor hiemalis, and Staphylococcus aureus (Surup et al., 2018). Further research on Elsinochrome-producing microorganisms resulted in the isolation of Elsinochromes A, B, C, and D (23–26) from other Elsinoe species (Batterham and Weiss, 1963; Lousberg et al., 1969, 1970). Other microorganisms can also produce Elsinochrome, for example, Elsinochrome C is produced by Parastagonospora nodorum (Chooi et al., 2017). The mycobiont fungal strain of the lichen Graphis elongata biosynthesizes Elsinochrome A (Fazio et al., 2018). Hypomyces sp., a parasitic mushroom fungus, produces hypomycin A (27) and B (28), with structures closely related to the elsinochromes (Liu et al., 2001; Zhang et al., 2001). It is noted that Stagonospora convolvuli LA39 can produce both elsinochrome A and cercosporin (Boss et al., 2007). Shiraia sp. SUPER-H168 produces a PQ complex of elsinochrome and hypocrellin (Cai et al., 2011), thus indicating that diverse microorganisms contain the PQ biosynthetic gene clusters (de Jonge et al., 2018).

As an effective defense compound against disease, hypocrellin has been used for centuries in traditional Chinese medicine, to effectively treat skin diseases, rheumatoid arthritis, and gastric diseases (Diwu and Lown, 1990). Hypocrellin A (29) and B (30) present excellent antimicrobial and antileishmanial activities (Ma et al., 2004) and can induce apoptosis of diverse cancer cells, such as hepatocellular carcinoma (Wang et al., 2019), human lung adenocarcinoma A549 cells (Qi et al., 2019), and squamous carcinoma A431 cells (Niu et al., 2020). Hypocrellin can also protect human ACE2 cells against infection by SARS-CoV-2 by blocking the receptor-binding domain of the SARS-CoV-2S protein, thereby inhibiting the entry of the virus into human cells (Li et al., 2021). Given its potential as a SARS-CoV-2 entry inhibitor, hypocrellin and its derivatives have been isolated from diverse microorganisms (Supplementary Figure S3) to improve the supply of hypocrellin, which was first isolated from the stroma of Hypocrella bambuase, growing on bamboo shoots (Chen et al., 1981). Shiraia bambusicola, another bamboo parasitic fungus, is also able to biosynthesize hypocrellins (Liu, 1985), as are Penicillium chrysogenum and Phaeosphaeria sp. (Meng et al., 2011; Li et al., 2012) and the engineered Aspergillus species (Hu et al., 2019).

Genes for PQ biosynthesis are arranged as biosynthetic gene clusters (BGCs; Blin et al., 2017), comprising three modules: i.e., the polyketide synthase (PKS), which synthesizes the core PQ structure; phenol coupling enzymes, which dimerize the core structure; and tailoring enzymes, which modify the functional groups and generate the diversity of PQ structures (Müller et al., 2019; Figures 1A, 2). This collinear BGC structure has facilitated the discovery of new PQs by homology analysis. For example, the elsinochrome BGC from Elsinoë fawcettii and the phleichrome BGC from C. phlei have been identified through homology analysis of the cercosporin and hypocrellin BGCs, respectively (Ebert et al., 2019; Hu et al., 2019).

The PKS enzymes in different PQ biosynthetic pathways are characterized by high sequence homology and structural conservation. Therefore, using PKS as a screening template is an effective approach to discover potential new PQ BGCs. For example, our group has elucidated and confirmed the hypocrellin pathway in S. bambusicola, based on homology analysis of the cercosporin and elsinochrome PKSs (Deng et al., 2017a). A genome evolutionary method has been developed to mine PQ BGCs through PKS homology analysis, based on the collinearity of BGCs (de Jonge et al., 2018), which has been used to discover that the cercosporin BGC has undergone multiple duplications and horizontal transfers across a wide range of plant pathogenic fungi. Three cercosporin BGCs were found in the Cercospora beticola genome, two of which shared the major cercosporin-associated genes. A “phylogenetic roadmap” of the putative evolutionary history of polyketide synthase (PKS) was also created (de Jonge et al., 2018), which showed that four duplications and three transfers of the cercosporin PKS had occurred from diverse Dothideomycetes and Glomerellales species. The identity of the proposed cercosporin PKS remains to be confirmed by gene knockout and chemical characterization of its substrate selectivity/products.

Knockout of a PQ biosynthesis gene may decrease or eliminate the related enzymatic function (Liu et al., 2022). Phenotypical and metabolic differences between wild-type strains and mutants can aid in the elucidation of gene functions, for example, the pentacyclic core structure of PQs confers a variety of colors (Hu et al., 2019; Liu et al., 2022). Wild-type strains are mostly red, whereas mutants exhibit various other colors, owing to the accumulation of biosynthetic intermediates (Newman and Townsend, 2016). This aspect has greatly facilitated the discovery of the specific functions of the essential PQ biosynthetic genes.

Sometimes, gene disruption mutants can generate no, or ambiguous metabolites, thus hampering the confirmation of their functions. To avoid this situation, the corresponding enzymes can be heterologously expressed in optimized host systems, such as Escherichia coli and Aspergillus species, and their products can be characterized (Figure 1D).

Escherichia coli has several advantages as a host for gene expression, such as rapid growth rate, high protein expression levels, and well-constructed genetic tools, which has made it the expression system of choice for PQ genetic studies. However, polyketide synthases, the first enzyme in PQ biosynthesis, comprise 2,000–3,000 amino acids, with starter unit acyltransferase (SAT), β-ketoacyl synthase (KS), malonyl acyltransferase (MAT), product template (PT), dual-tandem acyl-carrier protein (ACP₂), and thioesterase (TE) domains (Herbst et al., 2018). These large, multi-domain enzymes cannot be expressed efficiently in E. coli, so the Udwary-Merski algorithm (UMA) was designed to predict the locations of individual domains and the linker regions between them (Udwary et al., 2002). This allows the domain structure of the gene to be deconstructed, and individual domains can be expressed in E. coli, then purified (Crawford et al., 2008); followed by combining the various domains with substrates and analyzing the PQ products. When the PKS gene was subjected to this analysis, the tri-domain SAT-KS-MAT was expressed in soluble form, but not KS-MAT (Crawford et al., 2008; Herbst et al., 2018), thus indicating that the tri-domain SAT-KS-MAT can assemble as a substrate loading/condensation complex, yielding a heptaketide. Through the same approach, the PT domain is involved in the C4-C9 and C2-C11 cyclizations by this heptaketide (Newman et al., 2014), and the TE domain has been identified to contribute to pyrone formation (O13-C1 cyclization) and nor-toralactone release (Newman et al., 2012).

Unfortunately, codon usage bias and species differences often result in insoluble, or unstable proteins when the bacterial E. coli expression system is used for fungal enzyme expression, thus restricting the utility of the deconstruction approach. However, these problems can often be avoided by using fungal expression systems, such as Aspergillus (Pahirulzaman et al., 2012; Chiang et al., 2013), which has facilitated the complete elucidation of the PQ biosynthetic pathway. For example, the Cppks1 gene from C. phlei has been expressed in Cryphonectria parasitica, thereby facilitating the elucidation of the biosynthetic pathway of the red phleichrome pigment (So et al., 2015). Engineered filamentous fungi such as Penicillium crustosum and Aspergillus species were successfully transformed and enabled to produce nor-toralactone (Chooi et al., 2017; Hu et al., 2019; Kindinger et al., 2019).

The availability of the PQ precursor nor-toralactone has enabled the verification of other PQ pathway genes (Figure 2); for example, the O-methyltransferase domain of the di-domain enzyme CTB3 O-methylates nor-toralactone (Newman and Townsend, 2016). The monooxygenase domain of CTB3 catalyzes the opening cycle, decarboxylation, and hydroxylation (Newman and Townsend, 2016); similarly, the di-domain protein in the elsinochrome pathway has similar functions (Hu et al., 2019). Engineered A. nidulans with elcA and elcB-MT transformed nor-toralactone to toralactone, suggesting the O-methylation function of elcB-MT. Co-expression of elcA and elcB can also convert toralactone to three different products, including one with a molecular mass consistent with cercoquinone A and D (Newman and Townsend, 2016). This finding indicates that elcB is involved in O-methylation, the opening cycle, decarboxylation, and hydroxylation.

It appears that an FAD-linked oxidase is involved in dimerizing the phenolic precursor to the corresponding PQ. However, a laccase-like multicopper oxidase (ElcG) can also catalyze dimerization by combining with a berberine bridge enzyme-like oxidase (ElcE; Hu et al., 2019). The FAD-dependent CTB5 from Cercospora nicotianae and Cz_CTB12 from Cercospora zeae-maydis appear to synergistically catalyze the dimerization reaction. In addition, a multicopper oxidase has been identified, which can biosynthesize the atropisomer (P)-viriditoxin by dimerizing two regionally-selective semiviriditoxin molecules (Hu et al., 2019). A fungal laccase was determined to regionally dimerize the phenol coupling precursor to the (R)-semi-vioxanthin, combined with an auxiliary protein (Furtges et al., 2019). Overall, oxidases, particularly multicopper oxidase, perform a crucial function in PQ biosynthesis by dimerizing two phenol precursors.

The subsequent functional group modification of these dimer intermediates results in the wide diversity of PQs, for example, a flavin-dependent monooxygenase (ElcH) can transform hypocrellin into elsinochrome (Hu et al., 2019). In the absence of this flavin-dependent monooxygenase, an engineered A. nidulans generated hypocrellin through an intramolecular aldol reaction. A FAD-dependent monooxygenase, CTB7, has been proposed to catalyze the formation of the dioxepine ring of cercosporin (Chung and Liao, 2008); when the CTB7 gene was transferred into Cercospora zeina, it enabled cercosporin production (Swart et al., 2017), confirming the involvement of CTB7 in dioxepine ring formation. However, recent reports suggest that CTB9 and CTB10 also contribute to methylenedioxy bridge formation (de Jonge et al., 2018), which was confirmed by the crystal structures of CTB9 with various substrates and the subsequent site-directed mutagenesis study (Liu et al., 2022).

The regulation system of PQs biosynthesis is extremely complex and employs different modulation levels, including pathway-associated regulators, epigenetic regulators, and global regulators (Macheleidt et al., 2016). This section describes the corresponding regulatory mechanisms and strategies for PQs biosynthesis on basis of these regulatory systems.

The initial PQ precursor, acetyl-CoA is obtained from pyruvate through three enzymic steps (Figure 2). Acetyl-CoA carboxylase (ACC) transforms acetyl-CoA to malonyl-CoA, the extension unit for PQ biosynthesis; then PQ is biosynthesized by the central polyketide pathway (Figure 2). Therefore, modulating precursor supply gene expression and the main polyketide pathway increases PQ production (Figures 2, 3C). For example, ACC overexpression increases the acetyl-CoA substrate pool and stimulates polyketide production in various strains, such as Streptomyces albus (Lu et al., 2016), Saccharomyces cerevisiae (Wattanachaisaereekul et al., 2008), and Aspergillus species (Hasan et al., 2018). This suggests that increasing the polyketide precursor pool may be a feasible and generally applicable way to up-regulate production of targeted metabolites. Similarly, overexpressing two amylases can up-regulate the expression levels of carbon catabolic genes (such as acc and pdc) and hypocrellin pathway genes, resulting in three-fold higher hypocrellin yields in the engineered strains (Gao et al., 2018b). Regulating the genes from the central PQ pathway is another strategy to increase metabolite production. For instance, overexpressing a di-domain protein and a hydroxylase increases hypocrellin production three-and two-fold, respectively (Li et al., 2019). Overexpressing multicopper oxidase promotes the corresponding dimerization reaction and enhances the expression of hypocrellin-associated genes; hypocrellin production in the engineered transformants increases 5-fold compared with wild-type S. bambusicola (Deng et al., 2016b). The CRISPR-mediated transcriptional activation system can stimulate multigene activation in A. nidulans and should facilitate PQ production by improving related multi-gene expression (Roux et al., 2020).

Microorganism rapidly responds to the dynamic environmental signals for the continuous PQs formations through real-time signal pathways (Keller, 2019). Therefore, comprehending the transient signal translations and the relevant PQs regulatory mechanisms can facilitate PQs biosynthesis. Namely, PQs production might be improved by rationally engineering these signal pathways.

Genes from the PQ pathway are highly homologous, and other non-conserved genes may be useful for modifying PQ functional groups and generating new combinations thereof (de Jonge et al., 2018; Ebert et al., 2019). Modifying genes for functional group modification should be an effective strategy to construct a diverse PQ platform. For example, Cryphonectria parasitica can generate various complex pigments through the polyketide pathway, and transformants with Cppks1 also produced a non-natural phleichrome (So et al., 2015). Recently, our group has constructed an efficient CRISPR system (Deng et al., 2017a), which can be used to integrate or replace functional group modification genes to generate novel PQs.

Reconstructing PQ BGCs in tractable filamentous fungi is another feasible approach to generate novel PQs (Figure 3E) because these strains have been engineered as auxotroph. Thus, transforming plasmids (containing the corresponding prototrophic genes) into these mutants can facilitate the growth of positive strains on selective media (Chiang et al., 2013). In addition, PQ biosynthesis genes are colinear in the PQ BGCs (Blin et al., 2017), containing nucleotides larger than 30 kb (Hu et al., 2019). These longer sequences require efficient methods to express partial or whole pathways; therefore, a fungal artificial chromosome (FAC) approach has been designed to randomly transfer the relevant genomic regions (up to 300 kb), thereby facilitating the transfer of the intact PQ BGCs into the expression plasmids (Tsunematsu et al., 2013; Bok et al., 2015). These plasmids contain bacterial and filamentous fungal replication origins, which ensure successful pathway construction in E. coli and target metabolite expression in Aspergillus species. After expression in Aspergillus species, high throughput detection tools are essential to confirm the presence of the target products. Liquid chromatography/mass spectrometry (LC/MS) was previously the method of choice to analyze Aspergillus metabolomics. Molecular networking and in-silico MS/MS fragmentation tools have further emerged as efficient methods to exploit complex metabolite profiles and the relevant PQ derivatives in filamentous fungi (Allard et al., 2016). The recently developed Global Natural Products Social Molecular Networking (GNPS; Wang et al., 2016) is an open-access data analysis tool, offering diverse raw, or assigned MS/MS libraries, which facilitates continuous dereplication and reliable product confirmation. In addition, a metabolomics-scoring strategy has been established to determine the FAC-encoded metabolites of different Aspergillus species, based on GNPS (Clevenger et al., 2017). For instance, the elsinochrome pathway in Aspergillus nidulans has been deconstructed and the target metabolites have been detected; engineering Aspergillus nidulans with a flavin-linked monooxygenase generates elsinochrome, whereas the transformants without this monooxygenase produce hypocrellin (Hu et al., 2019). Given the substantial homology between the elsinochrome and cercosporin BGCs, ElcE and ElcG have been used as reference genes, and two oxidases, i.e., FAD-dependent CTB5 from Cercospora nicotianae and Cz_CTB12 from Cercospora zeae-maydis, have been identified. Through the gene swap principle, ElcE and ElcG of A. nidulans have been replaced with CTB5 and Cz-CTB12, and the modified transformants generates the PQ core ring structure (Hu et al., 2019). These findings indicate the possibility of generating diverse PQ structures by reconstructing the pathways in Aspergillus species.

A droplet microfluidic platform has been developed in Aspergillus species via complex digital microfluidic/channel-based droplet chips (Gach et al., 2016). Similarly, liquid-handling robots can largely eliminate repetitive sample handling and facilitate very large-scale automated biochemical experiments; for example, liquid-handling robots have achieved 100% genomic editing frequency in Aspergillus (Kuivanen et al., 2019). These two systems offer the possibility to reconstruct the PQ pathway in Aspergillus species in a fully automated manner.

Except for endogenously genetic regulation, other methods, such as physical and chemical strategies can also stimulate PQs biosynthesis by increasing expressions of genes that are involved in hypocrellin biosynthesis, secretion, and auto-resistance (Sage and Shikazono, 2017; Sun et al., 2017).