Targeted discovery of polyketides with antioxidant activity through integrated omics and cocultivation strategies

ABSTRACT

Fungi generate a diverse array of bioactive compounds with significant pharmaceutical applications. However, the chemical diversity of natural products in fungi remains largely unexplored. Here, we present a paradigm for specifically discovering diverse and bioactive compounds from fungi by integrating genome mining with building block molecular network and coculture analysis. Through pangenome and sequence similarity network analysis, we identified a rare type I polyketide enzyme from Penicillium sp. ZJUT-34. Subsequent building block molecular network and coculture strategy led to the identification and isolation of a pair of novel polyketides, (±)-peniphenone E [(±)−1], three known polyketides (2–4), and three precursor compounds (5–7) from a combined culture of Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23. Their structures were established through extensive spectroscopic analysis, including NMR and HRESIMS. Chiral HPLC separation of compound 1 yielded a pair of enantiomers (+)−1 and (−)−1, with their absolute configurations determined using calculated ECD methods. Compound (±)−1 is notable for its unprecedented structure, featuring a unique 2-methyl-hexenyl-3-one moiety fused with a polyketide clavatol core. We proposed a hypothetical biosynthetic pathway for (±)−1. Furthermore, compounds 2, 5, and 6 exhibited strong antioxidant activity, whereas (−)−1, (+)−1, 3, and four exhibited moderate antioxidant activity compared to the positive control, ascorbic acid. Our research demonstrates a pioneering strategy for uncovering novel polyketides by merging genome mining, metabolomics, and cocultivation methods. This approach addresses the challenge of discovering natural compounds produced by rare biosynthetic enzymes that are often silent under conventional conditions due to gene regulation.

IMPORTANCE

Polyketides, particularly those with complex structures, are crucial in drug development and synthesis. This study introduces a novel approach to discover new polyketides by integrating genomics, metabolomics, and cocultivation strategies. By combining genome mining, building block molecular networks, and coculturing techniques, we identified and isolated a unique polyketide, (±)-peniphenone E, along with three known polyketides and three precursor compounds from Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23. This approach highlights the potential of using combined strategies to explore fungal chemical diversity and discover novel bioactive compounds. The successful identification of (±)-peniphenone E, with its distinctive structure, demonstrates the effectiveness of this integrated method in enhancing natural product discovery and underscores the value of innovative approaches in natural product research.

INTRODUCTION

Structurally complex and diverse natural products remain challenging targets for total synthesis and the leading compounds in new drug development (1, 2). Polyketides are predominantly synthesized by enzymes known as polyketide synthases (PKS) and show a broad spectrum of biological activities including antioxidative, anticancer, immunosuppressive, antibacterial, anti-inflammatory, and antiparasitic activities (3–5). PKS are classified into three distinct types based on structural and functional differences: Type I PKS (T1 PKS), Type II (T2 PKS), and Type III PKS (T3 PKS) (6–8). Considering the estimated over 2 million fungal species in the earth, the chemical space of polyketides in fungi remains largely unexplored (9). Traditional strategies have proven to be effective, yet high rediscovery rates of known compounds and the restriction to the more abundant compounds increasingly frustrate the identification of novel polyketides. Therefore, it is important and necessary to develop target search methods for discovering unidentified polyketides from the vast number of fungal species.

The development of genomics effectively explores unknown biosynthetic gene clusters (BGCs), significantly enhancing the efficiency of discovering novel compounds (10). With more and more fungal genomes being sequenced, many unannotated or misannotated PKS remain to be identified in fungi. Ultrahigh-performance liquid chromatography-mass spectrometry (UPLC-MS)-based methods can rapidly separate and sensitively detect and characterize natural products present at trace levels from small amounts of materials and have been developed to identify new compounds by analyzing and visualizing their tandem MS (MS²) fragments as a relational spectral network, such as a molecular network (MN), feature-based MN, or building-block-based MN (11–13). Nevertheless, the presence of silent genes still limits the exploration of fungal chemical diversity by genomics and metabolomics. The coculturing strategy simulates natural ecological environments to harness interactions between microorganisms, thereby activating silent genes and facilitating the characterization and exploration of increasingly diverse active natural products (14). Considering the advantages and shortcomings of the above methods, a combined strategy of genomic and metabolomic techniques alongside cocultivation strategies might be a promising alternative to explore polyketide compounds from fungi.

In this study, we introduced a novel strategy combining gene mining and building block molecular network analysis with coculture techniques to facilitate the efficient discovery of new polyketides (Fig. 1). This approach led to the discovery and isolation of a pair of novel polyketides, (±)-peniphenone E [(±)−1], three known polyketides (2–4), and three precursor compounds (5–7). The compound (±)−1 features a unique 2-methyl-hexenyl-3-one moiety attached to a polyketide clavatol core. A hypothetical biosynthetic pathway for (±)−1 was proposed. Additionally, (−)−1, (+)−1, 3, and 4 exhibited moderate antioxidant activity, while 2, 5, and 6 exhibited strong antioxidant activity when compared to the positive control, ascorbic acid.

Fig 1.

Diagrammatic workflow for genome mining and building block molecular network with coculture techniques for the discovery of polyketides.

RESULTS

Biosynthetic potential of Ascomycota

To explore the potential of Ascomycota fungi in synthesizing secondary metabolites, we conducted a comprehensive study utilizing pangenome analysis. We accessed BGCs statistics for Ascomycota fungi from the Joint Genome Institute (JGI) database and filtered the data set to exclude genera with fewer than five samples, ensuring robust statistical analysis. Our calculations revealed that Ascomycota fungi possess a diverse array of BGCs.

To investigate the relationship between BGC count and phylogeny, we obtained its internal transcribed spacer (ITS) sequences for Ascomycota fungi from the National Center for Biotechnology Information (NCBI) database. We extracted the ITS regions using the ITSx software and performed multiple sequence alignment with MAFFT. Based on these alignment results, we constructed a phylogenetic tree using IQ-TREE 2. Integrating BGC count data with the phylogenetic tree (Fig. 2), we observed a notable trend: the top 10 genera with the highest average BGC counts, each exceeding 50 BGCs, were significantly above the average for other genera. Furthermore, these BGC-rich genera were not randomly distributed but exhibited clustering on the phylogenetic tree. This pattern suggests that these genera may have experienced BGC expansion or horizontal gene transfer, enhancing their ability to produce secondary metabolite synthesis. This finding underscores the significant natural product synthesis potential of Ascomycota fungi and highlights the crucial role of BGCs in their evolutionary adaptation.

Fig 2.

The distribution of BGCs across different groups of Ascomycota fungi. The dendrogram on the left illustrates the phylogenetic relationships among the fungi. The heatmap on the right depicts the number of various types of secondary metabolite gene clusters present in different fungal groups. DMAT: dimethylallyl diphosphate synthase; NRPS: non-ribosomal peptide synthetase; PKS: polyketide synthase; Hybrid: PKS-NRPS hybrid.

An overview of Penicillium BGCs

In our study of the genus Penicillium, we employed comparative genomics and bioinformatics to systematically analyze its pangenome. We categorized the core biosynthetic genes by their chemical classes for two main reasons: (i) different classes of BGCs use distinct building blocks to produce a wide variety of natural products, except for RiPPs (ribosomally synthesized and post-translationally modified peptides) and NRPS BGCs, which use the same blocks; (ii) the diverse chemical compounds produced by these BGCs may contribute uniquely to the adaptive evolution of Penicillium species.

Although PKS and NRPS BGCs are widely distributed in Penicillium, their distribution patterns vary significantly. PKS BGCs are primarily found in the accessory and singleton regions, suggesting their role in producing species-specific secondary metabolites. These metabolites likely contribute to interactions with hosts, environmental adaptation, and other specialized functions, driving the evolutionary diversity of Penicillium species. Conversely, NRPS BGCs are more prevalent in core and accessory regions, indicating that their metabolic pathways are relatively conserved and fundamental for core functions like basic metabolism and cell growth, which are crucial for species survival.

Our functional analysis revealed 6,959 known clusters and 25,062 unknown clusters in Protein Homology Groups 2020 (COG20). This indicates a substantial presence of proteins with unknown function within Penicillium. Pathway analysis further identified 2,463 known and 29,558 unknown clusters, with over 90% of genes related to metabolic pathway remaining unannotated (Fig. 3). These findings suggest that Penicillium possesses significant metabolic potential and is high degree of novelty in biosynthesis.

Fig 3.

Pangenomic analysis and BGC overview of nine Penicillium genomes by anvi'o. The central diagram represents a hierarchical clustering tree based on gene presence/absence. In the circular interface, each layer (gray) represents all genes in a single genome (black/green) and the distribution of BGCs (red). For the bin names, the core region (green) contains genes present in all nine Penicillium sp. genomes, while the accessory region (blue) includes genes shared by some Penicillium sp. genomes. The distribution of BGCs in the strains is depicted by bar graphs under the species phylogeny. The maximum number and classification of each BGC are indicated on the right side of the bar graphs. The double-layer BGC circular diagram provides an overview of the proportion of each BGC category (outer layer) and unknown/known BGCs (inner layer) in Penicillium sp. The core biosynthetic genes circular diagram indicates the proportion of BGC classes to which the core biosynthetic genes belong in different pangenome regions. The BGC core biosynthetic genes circular diagram indicates the total number of BGCs and the number of unknown BGCs.

Unique Penicillium BGCs

In our assessment of biosynthetic novelty and capabilities in Penicillium species, we employed antiSMASH analysis and BiG-SCAPE (cutoff value of 0.6) to map the diversity of BGCs. This approach, combined with data from the Minimum Information about a Biosynthetic Gene cluster (MIBiG) database, allowed us to construct a comprehensive map of Penicillium BGCs, revealing significant insights into their novelty (Fig. 4). We identified a total of 155 NRPS BGCs, 121 PKS BGCs, 34 PKS-NRPS hybrid BGCs, and 13 additional BGCs potentially classified as PKS under the category of “Others” (Table S1).

Fig 4.

Sequence similarity network of BGCs from nine Penicillium genomes. Each node represents a BGC, with different colors indicating various types of secondary metabolites.

Our analysis encompassed 408 BGCs from nine Penicillium strains. Out of 2,497 reference BGCs in the MIBiG database, 201 were identified in our gene cluster network. This network comprised 609 BGCs organized into 278 gene cluster families (GCFs), including 204 singletons. Notably, only 44 of GCFs contained characterized or known BGCs, suggesting that 84% of Penicillium’s secondary metabolism potential remains unexplored. Specifically, 28% of T1 PKS BGCs, 75% of NRPS BGCs, 100% all of RiPPs-Likes, 51% of Others, and 57% of Terpene BGCs have not been identified. Our findings indicate that the majority of T1 PKS and PKS-NRP hybrids are concentrated in the main network, with T1 PKS largely located at the network periphery. This suggests that these BGCs are distantly related to known ones, potentially reflecting ongoing innovation in Penicillium that supports its adaptation to diverse environments.

Additionally, our GCF clustering revealed unique BGCs that do not match any reference BGCs in the MIBiG database. These unique GCFs, each containing fewer than nine BGCs and originating from various Penicillium species, likely reside in accessory genomic regions. This pattern suggests that they represent species-specific capabilities for secondary metabolite synthesis, potentially driven by selective pressures in specific ecological niches. For instance, these unique GCFs might be involved in synthesizing compounds for host interactions or environmental adaptations. In contrast, GCFs with nine or more BGCs, as well as those containing MIBiG reference BGCs, are generally located in core genomic regions. These core BGCs are critical for fundamental metabolic functions across Penicillium species and reflect essential life-supporting capabilities. Overall, Penicillium exhibits a differentiated distribution of secondary metabolite synthesis capabilities, with core regions maintaining basic functions and accessory regions contributing to specialized adaptations. This differentiation underscores Penicillium’s evolutionary strategy to balance fundamental life functions with competitive advantages in varied environments.

Homology-based genome mining for fungal polyketides using T1 PKS as a probe

Our analysis indicates that ClaF and SorB are located on the same major branch of the phylogenetic tree (Fig. 5A), suggesting that T1 PKS enzymes on this branch are capable of generating a 2,4-dimethyl-1,3-benzenediol core structure. In contrast, TraA and SorA are positioned on separate major branches, which implies that the synthesis of polyketide compounds involving these T1 PKS enzymes may involve a collaborative process. Specifically, one T1 PKS might be responsible for creating the core skeleton, while the other could be involved in modifying this skeleton to produce a range of polyketide compounds.

Fig 5.

Genome mining of T1 PKS BGCs from Penicillium sp. (A) The phylogenetic analysis of T1 PKS from Penicillium sp. The asterisk indicates that the catalytic functions of T1 PKS were verified. (B) Clusters containing T1 PKS from diverse Penicillium sp. analyzed by antiSMASH database and MIBiG hits.

Further analysis of sequence similarities among different BGCs revealed that in Penicillium sp. ZJUT-34, scaffold 30.1 contains a T1 PKS gene with 93% similarity to ClaF and 40% similarity to SorB (Fig. 5B). Based on these observations, we propose that a T1 PKS gene from this cluster, when paired with a distinct T1 PKS from a different branch than TraA, could potentially yield a novel secondary metabolite with a 2,4-dimethyl-1,3-benzenediol core structure. Notably, no T1 PKS gene from Penicillium sp. ZJUT-34 was found within the clade containing TraA. Considering the 42% sequence similarity between ClaF and SorB, we also speculate that SorA’s ortholog may interact with ClaF to synthesize polyketide compounds. In targeting the clade with SorA, we identified a T1 PKS gene on scaffold 13.1 from Penicillium sp. ZJUT-34 with 32% similarity to SorA. This gene appears relatively novel compared to SorA and TraA. We hypothesize that this novel T1 PKS, in combination with the ClaF ortholog from scaffold 30.1, could produce a unique polyketide compound. Consequently, we propose that any strain possessing an ortholog of ClaF alongside either an ortholog of TraA or SorA, with sufficient novelty, might generate new polyketide compounds featuring a 2,4-dimethyl-1,3-benzenediol core structure.

Targeted discovery of diverse fungal polyketide guided by molecular networking and coculture

In our quest to discover novel polyketide compounds, we constructed a molecular network based on biosynthetic building blocks (Fig. 6A). We collected HPLC-MS/MS data from monocultures of Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23, as well as from their coculture. Using the open-source software MZmine v2.53, we processed these data sets through deconvolution and filtering, which yielded 8,231 high-intensity features (>1.0E⁴; Fig. S1). The refined data set was further analyzed using the Global Natural Products Social Molecular Networking (GNPS) platform and SIRIUS software to assess confidence and annotate the compounds. Features with SIRIUS confidence scores below 0.5 were prioritized for their potential to represent novel compounds. The analysis results were visualized with Cytoscape v3.8.1.

Fig 6.

Cocultivation and secondary metabolite analysis. (A) Flowchart for mining polyketides compounds using co-cultivation and metabolomics. (B) Molecular networking analysis of extracts is depicted, illustrating the relationships and differences in metabolite profiles. (C) Fragmentation patterns of compounds 1–3, as analyzed by MS² collision-induced dissociation, are provided, offering detailed information on their structural characteristics. (D) The chromatograms for secondary metabolites from Penicillium sp. ZJUT-34, Penicillium sp. ZJUT23, and their coculture are compared, highlighting the differences and similarities in metabolite production.

Our phylogenetic analysis of T1 PKS enzymes from Penicillium species suggested that the secondary metabolites should contain a fragment with an m/z of 179.07, as indicated by our molecular network. We identified 154 features with this fragment (Fig. 6B). Subsequent investigation revealed that these metabolites were produced in the coculture of Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23. Notably, compounds 1–3, which were exclusively produced in coculture, likely share a common core structure, as evidenced by signals at m/z 161.060 and 179.070 (Fig. 6C and D). From this, we successfully isolated and identified a novel polyketide, (±)-peniphenone E [(±)−1], along with three known polyketides (2–4) and three precursor compounds (5–7) (Fig. 7).

Fig 7.

Selected molecular networking analysis of extracts from coculture of Penicillium sp. ZJUT-34 with Penicillium sp. ZJUT23.

Isolation and structure elucidation

Compound 1 [(±)−1] was obtained as a white amorphous powder. The molecular formula of 1 is C₁₆H₂₀O₄, with ion peaks at m/z 277.1434 [M + H]⁺ (Fig. S2), corresponding to 7° of unsaturation. The ¹H NMR spectrum (Fig. S3) displayed signal for two reactive hydrogens (δ_H 9.13, s; 12.94, s), an aromatic proton (δ_H 7.34, s), two olefinic protons (δ_H 7.00, dq, J = 13.9, 6.8 Hz; 6.19, d, J = 15.6 Hz), three aliphatic protons (δ_H 2.85, dd, J = 14.2, 10.7 Hz; 2.74, dd, J = 14.2, 2.3 Hz; 3.27, m), and four methyls (δ_H 1.89, d, J = 6.8 Hz; 2.18, s; 2.52, s; 1.31, d, J = 7.4 Hz). The ¹³C NMR data and HSQC spectrum (Fig. S4 and S5) revealed that 1 has a skeleton based on 16 carbons, including two carbonyls (δ_C 207.2 and 202.8), six aromatic carbons (δ_C 112.9–161.9), two olefinic carbons (δ_C 130.1 and 145.7), two oxygenated methines (δ_C 67.7 and 65.3), two aliphatic carbons (δ_C 25.4 and 44.6), and four methyls (δ_C 16.2–26.3). The correlations observed were H-7 to C-1, C-5, and C-6, Me-13 to C-1, C-2, and C-3, and the aromatic proton H-3 to C-1 and C-5 in the HMBC spectrum (Fig. S6), indicating a pentasubstituted benzene ring. Additional HMBC cross-peaks from Me-15 to C-4 and C-14 and from H-3 to C-14 supported the location of the acetyl at C-4. The above 1D NMR data (Table 1) of a pentasubstituted benzene ring were similar to those of clavatol (6), a known polyketide derivative also obtained in the present study. The HMBC correlations of H-7 to C-9, H-10 to C-9, and H-11 to C-9, along with the correlated spectroscopy (¹H-¹H COSY, Fig. S7) cross-peaks of H-7/H-8, H-8/H-16, H-10/H-11, and H-11/H-12 revealed the presence of a 2-methyl-hexenyl-3-one group. Furthermore, the HMBC correlations from H-7 to C-1, C-5, and C-6 facilitated the linkage of 2-methyl-hexenyl-3-one to clavatol. Therefore, the planar structure of 1 was assigned as a polyketide. However, compound 1 was found to be optically inactive based on the optical rotation [(α)20D = 0 (c = 0.1, MeOH)], suggesting that 1 was a racemic mixture. Subsequently, chiral resolution of 1 by HPLC afforded the anticipated enantiomers, (−)−1 and (+)−1, respectively. The absolute configurations of 1 were established by comparing experimental and calculated ECD spectra. As shown in Fig. 8, the calculated ECD spectrum for (8R)−1 matched the experimental ECD spectrum of (+)−1, whereas the ECD spectrum calculated for (8S)−1 agreed well with the measured one for (−)−1. Thus, the absolute configurations of (+)−1 and (−)−1 were determined and named (+)-peniphenone E and (−)-peniphenone E, respectively.

TABLE 1.

¹H (600 MHz) and ¹³C NMR (150 MHz) spectroscopic data of 1 (in Chloroform-d)

Position	C, type	H (J in Hz)
1	160.6, C	–
2	118.0, C	–
3	130.6, CH	7.34 s
4	112.9, C	–
5	161.9, C	–
6	113.7, C	–
7	25.4, CH2	2.85 dd (14.2, 10.7), 2.74 dd (14.2, 2.3)
8	44.6, CH	3.27 (1H, m)
9	207.2, C	–
10	130.1, CH	6.19 d (15.6)
11	145.7, CH	7.00 dq (13.9, 6.8)
12	18.5, CH3	1.89 d (6.8)
13	16.2, CH3	2.18 s
14	202.8, C	–
15	26.3, CH3	2.52 s
16	19.4, CH3	1.31 d (7.4)
1-OH	–	9.13 s
5-OH	–	12.94 s

^a –, hydroxyl group at different positions.

Fig 8.

Isolation and chemical structures of compounds (±)−1. (A) Key ¹H-¹H COSY and HMBC correlations of 1. (B) Chiral HPLC separation profile of (±)−1. (C) The experimental CD spectra of (−)−1 and (+)−1 in MeOH, along with the calculated ECD spectra of (8R)−1 and (8S)−1. (D) MS/MS collision-induced dissociation fragmentation for 1.

Hypothetical biosynthetic pathway for peniphenone E (1)

The UPLC-MS² data indicated that peniphenone E could only be detected in the coculture broth of Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23. This result suggests that the upregulation of biosynthetic pathways for peniphenone E under coculture conditions facilitated its detection in the HPLC-DAD profile. Peniphenone E features a unique 2-methyl-hexenyl-3-one moiety attached to a polyketide clavatol core. Compounds 5–7 were speculated to be precursors in the synthesis process of peniphenone E. Biosynthetic pathways for peniphenone E were proposed based on a comprehensive analysis of their biosynthetic gene clusters and structural features (Fig. 9).

Fig 9.

Proposed biosynthesis pathway of (±)−1.

Gene cluster A is responsible for the assembly and transport of clavatol (6), with the non-reducing PKS ClaF playing a crucial role in its formation. Clavatol is oxidized to hydroxyclavatol (7) by the non-heme FeII/2-oxoglutarate-dependent oxidase ClaD, which then undergoes spontaneous dehydration to form an O-quinone methide. In cluster B, PheC possesses the domain architecture KS-AT-DH-MeT-ER-KR-ACP and is located on the same major branch as SorA in the sorbicillin biosynthesis pathway. Since their domain structures are identical, we believe that PheC, like SorA, catalyzes two elongation cycles to produce a non-methylated, reduced triketide starter unit. However, unlike SorA, PheC’s KR and DH domains only carry out one round of ketoreduction and dehydration, without invoking the ER domain, resulting in a diketide precursor.

The active hydrogen in this diketide precursor attacks the β-carbon (the terminal carbon of the double bond) of the O-quinone methide, leading to the formation of a new carbon-carbon bond. During this process, the attack by the diketide precursor causes the double bond in the O-quinone methide shifts, resulting in the generation of a new intermediate via Michael addition (15). These intermediates can then be converted into a racemic mixture of peniphenone E (±)−1 via a series of reductions (16). Finally, chiral resolution of compound 1 using HPLC yielded a pair of enantiomers, (−)−1 and (+)−1.

Evaluation of in vitro antioxidant capacity

The antioxidant activity of the seven compounds was evaluated by measuring their ability to scavenge ABTS^•+ capacity (Table 2). Compounds 2, 5, and 6 exhibited strong antioxidant activity and a high capacity to scavenge ABTS^•+. In contrast, compounds (−)−1, (+)−1, 3, and 4 showed moderate antioxidant activity and a moderate ability to scavenge ABTS^•+. The strong antioxidant activity of compounds 2, 5, and 6 may be attributed to their ability to effectively donate electrons or hydrogen atoms to ABTS^•+ capacity, thereby neutralizing their activity and achieving free radical scavenging. Although compounds (−)−1, (+)−1, 3, and 4 also exhibit antioxidant activity, their effects are more moderate. This may be due to the lack of certain key free radical-stabilizing features in their structures or a weaker mechanism for free radical scavenging. Additionally, compounds (−)−1 and (+)−1, as enantiomers, show moderate antioxidant activity, indicating that stereochemistry has some impact on their antioxidant capacity. However, in this study, this impact did not significantly enhance their ability to scavenge free radicals. The strong antioxidant activity of compounds 2, 5, and 6 highlights their potential use in food preservation. Antioxidants play a crucial role in extending the shelf life of food or agricultural products by preventing oxidation, which can lead to rancidity and loss of nutritional value. The effective free radical scavenging ability of these compounds could be utilized to maintain the quality and safety of food or agricultural products over time. Overall, compounds 2, 5, and 6 demonstrated superior performance in antioxidant studies, showing potential as highly effective antioxidants for use in related fields. Compounds (−)−1, (+)−1, 3, and 4 may require further structural modifications or combination with other compounds to improve their antioxidant effects. Studying the antioxidant mechanisms and structure-activity relationships of these compounds will aid in the design and development of more effective antioxidants.

TABLE 2.

Effects of compounds 1–7 on antioxidant activities

Antioxidant activity	ABTS•+ scavenging activity (IC50 µg/mL)
(−)−1	13.92 ± 0.94
(+)−1	12.89 ± 1.05
2	0.23 ± 0.04
3	14.40 ± 0.43
4	6.37 ± 0.12
5	0.39 ± 0.09
6	0.85 ± 0.05
7	>100
Ascorbic acid	7.80 ± 0.17

DISCUSSION

Penicillium, commonly referring to the genus Penicillium, belongs to the phylum Ascomycota. This genus represents a large and complex group of fungi, first discovered in 1809. By 2022, more than 200 species have been identified, demonstrating its extensive diversity. Penicillium species are widely distributed and can be commonly found in various environments, including soil, air, food, decaying organic matter, and wood. Notably, among the top 10 genera with the highest BGC counts in Ascomycota fungi, the Penicillium genus stands out for the remarkable structural diversity of its secondary metabolites. These compounds include terpenes, polyketides, alkaloids, and macrolides, and they exhibit a wide range of bioactivities such as antibacterial, anticancer, hypolipidemic, antioxidant, and insecticidal effects. The most famous example is Penicillium rubens, the primary industrial producer of penicillin (17). Additionally, other species of Penicillium are widely utilized in the production of various pharmaceuticals, enzymes like cellulases, and applications in the food industry (18).

To explore the potential of Penicillium species in synthesizing secondary metabolites, we conducted a comprehensive pangenomic analysis using comparative genomics and bioinformatics. This approach allows us to understand how the pangenome of Penicillium reflects its adaptive evolution under natural selection (19). Core genes, shared by all species, are essential for maintaining basic fungal functions. In contrast, accessory genes and singletons, which are present in only some species, may seem “dispensable” but often contribute to unique biochemical pathways and functions. These non-core genes can provide Penicillium with competitive advantages in specific ecological niches, driving adaptive evolution.

T1 PKS are complex enzyme systems composed of multiple modules, each responsible for a specific reaction. These systems can be highly programed to catalyze multiple interactions, generating compounds with varying chain lengths and reduction states. To investigate the diversity of T1 PKS BGCs within the Penicillium genus, we collected amino sequences. Multiple sequence alignment and phylogenetic analysis revealed that the involvement of S-adenosylmethionine (SAM) in T1 PKS-catalyzed reactions varies, leading to different products. For instance, AdrD, MapC, and MacA enzymes each catalyze the formation of distinct methylated orsellinic acids using different combinations of acetyl-coA, malonyl-coA, and SAM. We also observed that when two T1 PKS are present in the same BGC, they often catalyze distinct reactions. Moreover, T1 PKS from different clusters can collaborate to produce secondary metabolites, such as sorbicillin and penilactones A/B (15, 20).

Fungal cocultivation, especially with fungi from the same ecological niche, has proven to be an effective strategy for enhancing the production of known compounds and inducing the generation of novel bioactive molecules (14). For instance, a study comparing the chemical composition of monocultures and cocultures of the marine-derived fungus Cosmospora sp. and Magnaporthe oryzae found that monocultures lacked specific isochromanones present in the cocultures, leading to the discovery of two new derivatives (21). Although the cocultivation strategy is widely recognized for enhancing microbial metabolite diversity, the ecological and molecular mechanisms underlying these complex microbial interactions remain unclear. Future research should focus on exploring these mechanisms within cocultivation systems to deepen our understanding and improve the application of microbial cocultivation strategies.

Our approach, which integrates genomics, metabolomics, and coculture strategies, successfully led to the discovery of a pair of novel polyketides, (±)−1, from a coculture of Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT23. (±)−1 Stands out due to its unique structure, featuring a 2-methyl-hexenyl-3-one moiety integrated with a polyketide clavatol core. This combined strategy offers significant advantages over traditional methods that rely solely on genomics, metabolomics, or coculture techniques. Genomics efficiently identifies new polyketide synthases but often requires additional steps to activate previously silent gene clusters. Building block molecular networks further refine this process, reducing the risk of overlooking potentially bioactive molecules, yet without the context of coculture, many bioactive molecules may remain unexpressed. Coculture can activate silent enzymes and uncover novel compounds, but it may lack the precise targeting provided by genomics and metabolomics, making the process less efficient and more time-consuming. By integrating these three approaches, our strategy not only activates otherwise silent gene clusters but also streamlines the discovery process, enhancing both efficiency and precision. This synergistic strategy shows considerable potential for advancing the targeted discovery of bioactive fungal compounds, offering a more comprehensive and effective method.

MATERIALS AND METHODS

General experimental procedures

NMR spectra were recorded on a record Bruker AVANCE III 600 MHz spectrometer (Bruker, Fällande, Switzerland). HREIMS data were obtained using an AB Sciex 5500 spectrometer (AB Sciex Technologies, USA). Optical rotations were determined using an automatic polarimeter (Rudolph Research Autopol III, POLAXL, ATAGO, USA). IR spectra were recorded on a Nicolet-Avatar-iS5 spectrometer with KBr disks. UV spectra were acquired using a DU640 microvolume spectrophotometer (Beckman, USA). CD spectra were obtained on a JASCO J-1700 CD spectrometer. Semi-preparative HPLC was performed using a SHIMADZU LC-2030C HPLC system (SHIMADZU, Japan) with an Agilent Eclipse XDB C18 semi-preparative HPLC column (9.4 × 250 mm, 5 µm). Chiral separation was carried out on a Chiralpak AD-H column (5 µm, 4.6 × 250 mm). MCI gel (75–150 µm), silica gel (200–300 mesh), and ODS C-18 (SP-120–50-ODS) were used for column chromatography. All solvents used were of analytical grade and sourced from commercially available suppliers.

Strains and culture conditions

Penicillium sp. ZJUT-34 was previously isolated and identified as described (22). Penicillium sp. ZJUT23 was identified as Penicillium sp. with ITS sequence showing a 99.01% identity match with Penicillium copticola in homology alignment. Penicillium sp. ZJUT-34 and Penicillium sp. ZJUT-23 were separately and jointly incubated on potato dextrose agar plates at 28°C for 4 days. For secondary metabolite production, cocultivation was conducted on plates containing rice broth medium at 28°C for 14 days.

The biosynthetic potential of ascomycetes

Statistical information on secondary metabolism clusters of Ascomycota was obtained from the JGI database (23). The data were classified by genus, excluding genera with fewer than five occurrences, and averages were calculated to generate secondary metabolites of Ascomycota (Table S2). Subsequently, the ITS sequences of each genus were obtained from the NCBI, and a phylogenetic tree was constructed using MEGA (v 11.0.13) (24).

antiSMASH annotations and BiG-SCAPE classification

Nine Penicillium genomes were randomly selected from NCBI (GCF_000315645.1, GCF_000769745.1, GCF_001723175.1, GCF_028827035.1, GCF_028827245.1, GCF_028827825.1, GCF_028828155.1, GCF_02882945.1, and GCF_028829775). The local version of antiSMASH 7.0.0 was utilized to mine all genome sequences, identifying natural product BGCs with analysis and querying of the MIBiG database using ClusterBlast and Pfam (25). BGC clustering was then performed using BiG-SCAPE 1.1.5 with a cutoff of 0.6 (26). Finally, the results were visualized using Cytoscape (v3.10.2) (27).

Pangenome analysis

Pangenome analysis was conducted using anvi'o v8 (28). The procedure followed the Vibrio jasicida pangenome mini workshop (https://merenlab.org/tutorials/vibrio-jasicida-pangenome/). A filter with a “minimum number of genomes for gene homolog group = 9” was applied. Gene clusters not classified into the core gene bin were appended to the accessory bin. Protein sequences were exported using “anvi-get-sequences-for-gene-clusters” and subsequently analyzed with antiSMASH to classify core biosynthetic genes in each region. The results were visualized as a circular plot using Origin 2024b. Whole-genome antiSMASH annotation results were then imported back into anvi'o using “anvi-import-misc-data” and visualized through the interface.

Bioinformatics analysis of T1 PKS and phylogenetic tree construction

We collected all T1 PKS amino acid sequences from the antiSMASH database, the MIBiG database, and the NCBI database and combined them with the antiSMASH analysis results of two strains from our laboratory (29, 30). All T1 PKS amino acid sequences were extracted and subjected to multiple sequence alignment using MAFFT (31). Subsequently, a phylogenetic tree was constructed using IQ-TREE 2 (32). Gene cluster comparisons were conducted using the CompArative GEne Cluster Analysis Toolbox, (CAGECAT, https://cagecat.bioinformatics.nl/) (33).

Molecular networking and coculture analysis

The extracts underwent repeated extraction with ethyl acetate and subsequent evaporation under reduced pressure. They were dissolved in 1 mL of methanol and analyzed using HPLC-DAD and LC-MS/MS. MS/MS data files were converted to mzXML format using MSConvert software (www.proteowizard.sourceforge.net). The data were processed using MZmine v2.53 (34), resulting in a peak list comprising 8,231 individual features. This list was exported as .mgf files and a .csv quantification table, which were submitted to the GNPS platform and the SIRIUS platform, respectively (35, 36).

Fermentation, purification, and structure characterization

Spores were washed with 10 mL of sterile water and inoculated into Erlenmeyer flasks (30 × 1 L) containing rice medium (100 g of rice in 150 mL of distilled water, sterilized at 121°C for 20 min). The flasks were subsequently incubated at 28°C in static conditions for 30 days. The fermented rice was extracted three times with MeOH/EtOAc (10:90) at room temperature, and the resulting extracts were concentrated under reduced pressure to yield a total extract (12 g). The extracts were separated using MCI column with a gradient elution of MeOH/H₂O (from 20:80 to 100:0), which produced nine fractions (fractions 1–9). Based on UPLC-MS data, fraction 6 was selected for further purification. Fraction 6 underwent additional separation by ODS C-18 using a gradient of CH₃OH/H₂O [40:60 to 100:0, (vol/vol)], resulting in three sub-fractions. Fractions 6–1 underwent semi-preparative HPLC with CH₃CN/H₂O (65:35) as the eluent, yielding compounds 4 (3.0 mg, t_R = 15.5 min) and 3 (4.5 mg, t_R = 18.0 min). Compound 2 (2.2 mg, t_R = 13.0 min) was obtained from Fractions 6–2 under the semi-preparative HPLC conditions (CH₃CN/H₂O, 70:30). Fractions 6–3 were fractionated by semi-preparative HPLC using a gradient elution of CH₃CN/H₂O (65:35), resulting in compound 1 (2.5 mg, t_R = 6.0 min). Fraction 1 was subjected to ODS C-18 with elution using CH₃OH/H₂O [10:90 to 60:40, (vol/vol)], yielding two fractions (Fraction 1–1 to Fraction 1–2) and 6 (20.0 mg). Fractions 1–3 underwent further purification by Sephadex LH-20 with methanol elution, followed by silica gel with CH₂Cl₂-CH₃OH [100:1 to 10:1, (vol/vol)], yielding 5 (12.0 mg) and 7 (80.0 mg). One (3.0 mg) was subjected to chiral HPLC with elution using isopropanol/n-hexane [5:95, (vol/vol)] at a flow rate of 1 mL/min, which led to the isolation of compounds 1a (0.8 mg, t_R = 6.9 min) and 1b (0.6 mg, t_R = 7.8 min; Fig. 10).

Fig 10.

Chemical structures of 1–7.

peniphenone E (1). Colorless oil. [α]20D: ± 0 (c 0.1, MeOH), UV λ_max (MeOH) nm (log ɛ): 219 (3.18), 282 (2.91), and 328 (2.56; Fig. S8); IR (KBr): 3,232, 2,924, 1,622, 1,371, 1,333, 1,286, 1,266, and 1,184 cm⁻¹ (Fig. S9). HR-ESI-MS m/z: 277.1434 [M + H]⁺ (calcd. for C₁₆H₂₀O₄H⁺, 277.1434). For ¹H- and ¹³C-NMR data, see Table 1. (−)-peniphenone E (1a): Colorless oil, [α]20D: − 44 (c 0.05, MeOH), ECD (MeOH) 219 (Δε −2.68), 282 (Δε + 1.45), and 328 (Δε −0.65). For ¹H-NMR data, see Fig. S10. (+)-peniphenone E (1b): Colorless oil, [α]20D: +32 (c 0.1, MeOH), ECD (MeOH) 219 (Δε + 1.75), 282 (Δε + 0.83), and 328 (Δε + 0.36). For ¹H-NMR data, see Fig. S11.

penicophenones A (2). Colorless oil. [α]20D: +8.0 (c 0.1, MeOH). ¹H NMR (600 MHz, Chloroform-d) δ_H 7.28 (1H, s), 4.00 (1H, m), 3.61 (1H, dqd, J = 12.5, 6.3, 2.1 Hz), 2.69 (1H, dd, J = 16.8, 6.4 Hz), 2.33 (3H, s), 2.20 (1H, dd, J = 16.8, 2.8 Hz), 2.13 (1H, ddd, J = 12.6, 4.8, 1.7 Hz), 1.94 (3H, s), 1.88 (1H, m), 1.81 (1H, m), 1.09 (1H, dd, J = 12.6, 11.3 Hz), 0.99 (1H, dt, J = 12.1, 11.6 Hz), 0.87 (3H, d, J = 6.3 Hz), and 0.73 (3H, d, J = 7.1 Hz); ¹³C NMR (150 MHz, Chloroform-d) δ_C 156.2, 117.0, 129.5, 113.1, 161.2, 109.4, 23.7, 33.3, 101.8, 39.9, 65.1, 42.5, 66.4, 15.2, 202.9, 26.4, 15.6, and 21.6 (37).

peniphenone A (3). pale yellow crystals. [α]20D: +80.0 (c 0.1, MeOH). ¹H NMR (600 MHz, Chloroform-d) δ_H 12.81 (1H, br s), 7.32 (1H, s), 2.53 (1H, dd, J = 16.6, 3.1 Hz), 2.77 (1H, dd, J = 16.6, 5.6 Hz), 2.34 (1H, ddd, J = 13.8, 11.3, 0.8 Hz), 2.50 (1H, dd, J = 13.8, 3.1 Hz), 2.09 (1H, m), 2.87 (1H, d, J = 6.7 Hz), 3.93 (1H, m), 2.04 (3H, s), 2.54 (3H, s), 1.23 (3H, d, J = 6.8 Hz), 1.14 (3H, d, J = 6.7 Hz), and 1.19 (3H, d, J = 6.2 Hz); ¹³C NMR (150 MHz, Chloroform-d) δ_C 206.4, 202.9, 160.0, 155.6, 129.5, 117.0, 113.2, 112.2, 105.0, 67.5, 49.2, 48.6, 30.8, 26.2, 23.5, 21.7, 15.2, 15.1, and 7.5 (16).

6-acetyl-2α,5-dihydroxy-2-(2-hydroxypropyl)−3α,8-dimethylchroman (4). Pale yellow powder. [α]20D: −22.0 (c 0.1, MeOH). ¹H NMR (600 MHz, Chloroform-d) δ_H 7.33 (1H, s), 4.78 (1H, m), 2.74 (1H, m), 2.53 (3H, s), 2.48 (1H, m), 2.18 (1H, m), 2.13 (3H, s), 2.04 (1H, d, J = 14.2 Hz), 1.81 (1H, d, J = 14.2 Hz), 1.34 (3H, d, J = 6.3 Hz), and 1.15 (3H, d, J = 6.8 Hz). ¹³C NMR (150 MHz, Chloroform-d) δ_C 202.8, 160.5, 157.2, 130.7, 117.3, 112.8, 110.6, 100.4, 65.6, 44.3, 34.9, 26.3, 25.1, 23.7, 15.9, and 15.8 (38).

1-(2,4-dihydroxy-5-methylphenyl)ethan-1-one (5). White powder. ¹H NMR (600 MHz, Chloroform-d) δ_H 7.44 (1H, s), 6.35 (1H, s), 2.54 (3H, s), and 2.17 (3H, s); ¹³C NMR (150 MHz, Chloroform-d) δ_C 202.7, 163.4, 161.7, 133.1, 116.4, 113.9, 103.0, 26.3, and 15.3 (16).

Clavatol (6). Colorless oil. ¹H NMR (600 MHz, Methanol-d₄) δ_H 7.42 (1H, s), 2.50 (3H, s), 2.16 (3H, s), and 2.05 (3H, s); ¹³C NMR (150 MHz, Methanol-d₄) δ_C 204.4, 162.3, 162.1, 131.1, 117.2, 113.8, 111.8, 26.2, 16.3, and 7.9 (16).

hydroxyclavatol methyl ether (7). Colorless oil. ¹H NMR (600 MHz, Chloroform-d) δ_H 7.40 (1H, s), 4.83 (2H, s), 3.49 (3H, s), 2.53 (3H, s), and 2.15 (3H, s); ¹³C NMR (150 MHz, Chloroform-d) δ_C 202.8, 162.3, 160.2, 131.8, 117.1, 112.6, 108.2, 68.4, 58.9, 26.3, and 15.4 (15).

Quantum chemistry calculations

The low-energy conformers of different configurations were generated using Sybyl-X 2.1.1 software with the MMFF94S force field (Table S3) (39). Conformations with a Boltzmann population above 1.0% were selected for further optimization using Gaussian 09 Software (Table S4) (40). The optimization was conducted with density functional theory. Subsequently, the optimized conformers were used to calculate the ECD spectra (41). The ECD spectra of 8R/8S were obtained by a sigma/gamma value of 0.3 eV to process the calculated ECD (Table S5) (42).

Evaluation of in vitro antioxidant capacity

The ABTS radical cation (ABTS^•+) was generated by mixing an aqueous solution of 2,2’-azino-bis-3-ethylbenzthiazoline-6-sulfonic acid (ABTS; 7 mM) with an aqueous solution of potassium persulfate (2.45 mM) in equal quantities and letting the reaction proceed for 12–16 hours. Subsequently, 190 µL of ABTS^•+ solution was combined with 10 µL of either the isolates or ascorbic acid at 12.5–200 µg/mL. The mixture was then incubated in the dark for 10 minutes. A control was prepared by mixing 190 µL of ABTS^•+ solution with 10 µL of methanol solution. The scavenging activity percentage was determined as % inhibition using the following formula:

$${\displaystyle \text{ Inhibition }(\mathrm{\%})=\left(\frac{{\mathrm{O}\mathrm{D}}_{734\text{ control }}-{\mathrm{O}\mathrm{D}}_{734}}{{\mathrm{O}\mathrm{D}}_{734\text{ control }}}\right)\times 100\mathrm{\%}.}$$

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.01603-24.

Supplemental material. aem.01603-24-s0001.pdf.

Tables S1 to S5; Figures S1 to S11.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.