Genomic insights into the biosynthetic capacity of the sponge-associated fungus Aspergillus puulaauensis Hmp-F48

Abstract

Background

Marine-derived fungi are prolific producers of structurally diverse secondary metabolites with significant pharmaceutical potential. The discovery of natural products has been unprecedentedly accelerated by the prediction of biosynthetic gene clusters (BGCs) within the whole-genome context. This study presents the whole-genome sequencing, comprehensive annotation, and biosynthetic potential predictions of the sponge-associated fungus Aspergillus puulaauensis Hmp-F48.

Results

Genome sequencing of A. puulaauensis Hmp-F48 generated a high-quality draft assembly of 35.86 Mb. Structural annotation revealed a complex genomic architecture, comprising 10,611 protein-coding genes, 210 non-coding RNAs, and 155 tRNAs. Functional annotation using NR, Swiss-Prot, GO, KEGG, and eggNOG databases highlighted significant enrichment in biosynthetic, metabolic, and transport processes. AntiSMASH analysis identified 78 putative BGCs, including 16 type I polyketide synthase (T1PKS), 27 nonribosomal peptide synthetase (NRPS), 7 hybrid PKS-NRPS, 9 terpene-related clusters, 6 RiPP-related clusters and 13 clusters associated with other secondary metabolites. Several clusters exhibited high homology to known BGCs responsible for bioactive secondary metabolites, including asperthecin, sterigmatocystin, calbistrins, F-9775 A/B, nidulanin A, aspercryptins, fellutamide B, acetylaranotin, burnettramic acid A, equisetin, and pyranonigrin E. Experimental isolation confirmed the presence of PKS-derived metabolites, including sterigmatocystin, averantin, and decumbenones — one of which represents a previously undescribed congener.

Conclusions

This study highlights the extensive biosynthetic potential of A. puulaauensis Hmp-F48, offering valuable insights into its capacity for secondary metabolite production.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12569-2.

Background

Marine ecosystems, characterized by unique physicochemical conditions, drive the evolution of distinctive metabolic pathways in marine microorganisms, resulting in structurally diverse secondary metabolites with pharmaceutical potential [1–4]. Marine fungi, particularly those associated with marine sponges are recognized as prolific producers of bioactive secondary metabolites [5]. Among these, marine-derived Aspergillus species have attracted significant interest due to their ability to synthesize diverse bioactive compounds, including polyketides, peptides, alkaloids, and terpenoids with potent antimicrobial, anticancer, anti-inflammatory, and antioxidant activities [6–8]. A notable example is plinabulin, a synthetic derivative of halimide from marine-derived Aspergillus ustus, is currently undergoing Phase III clinical trials as a second-line chemotherapy agent in combination with paclitaxel for the treatment of advanced or metastatic non-small cell lung cancer (NSCLC) and for the prevention of chemotherapy-induced neutropenia (CIN) caused by docetaxel, underscoring the therapeutic promise of marine Aspergillus metabolites [9].

Recent advancements in genome sequencing technologies have established genome-based strategies as powerful tools for natural product discovery [10, 11]. Meanwhile, bioinformatic tools for the prediction of biosynthetic gene clusters (BGCs) efficiently bridge genomic dataset to chemically diversity, offering unprecedented opportunities for targeted natural product discovery [12, 13].

In this study, Aspergillus puulaauensis Hmp-F48 was prioritized for whole-genome sequencing from the Hymeniacidon perleve-associated fungal community due to its distinctively rich metabolic profile and significant bioactivity observed in preliminary screenings [14]. Elucidating its genomic architecture is therefore essential to decipher the biosynthetic machinery driving these traits and to facilitate the discovery of novel natural products. The whole-genome of Hmp-F48 was sequenced using Illumina HiSeq and PacBio platforms. Comprehensive genome annotation, including prediction of protein-coding genes and non-coding RNAs, was performed. For functional annotation, we utilized databases such as Nr, Swiss-Prot, GO, KEGG, eggNOG, and CAZy. BGCs were predicted and analyzed using antiSMASH. Finally, we cultivated A. puulaauensis Hmp-F48, isolated its secondary metabolites, and established direct links between its predicted BGCs and chemical profiles through structural characterization.

Materials and methods

Fungal strain

The fungal strain A. puulaauensis Hmp-F48 was isolated from a marine sponge (Hymeniacidon perleve) collected from the Bohai Sea at Heishi Reef Bay (38°51′52.4′′N121°33′05.0′′E) in Dalian, China and is preserved at the China Center for Type Culture Collection (CCTCC NO: AF 2025006). The Hmp-F48 strain was maintained on Czapek–Dox Medium (0.2% NaNO₃, 3% Sucrose, 0.1% K₂HPO₄, 0.05% KCl, 0.05% MgSO₄·7H₂O, 0.001% FeSO₄·7H₂O) at 28 °C with shaking at 180 rpm for 5 days. Large-scale fermentation (20 L) for metabolite isolation was performed in PDB Medium (200 g/L Potato, 2% Sucrose) at 28 °C with shaking at 180 rpm for 7 days. Spore morphology was observed after 4–5 days of cultivation on PDA Medium (200 g/L potato, 2% sucrose, 2% agar). The morphology of Hmp-F48 was examined using scanning electron microscopy (SEM, Phenom™ ProX, Phenom World BV, Netherlands).

Genome sequencing and assembly

Genomic DNA was extracted from mycelia using the OMEGA Fungal Genomic DNA Isolation Kit following the manufacturer’s protocol. DNA was assessed via agarose gel electrophoresis. The genome of the Hmp-F48 strain was sequenced using the Whole Genome Shotgun (WGS) strategy combining Illumina and PacBio platforms. For PacBio sequencing, SMRTbell libraries with an average insert size of ~ 10 kb were prepared and sequenced on the PacBio Sequel II platform, producing approximately 8.37 Gb of raw long-read data, corresponding to an estimated ~ 233× genome coverage. For Illumina sequencing, paired‑end libraries (2 × 151 bp) with an average insert size of ~ 400 bp were prepared and sequenced on an Illumina HiSeq instrument, yielding approximately 5.41 Gb of clean reads, corresponding to an estimated ~ 147× genome coverage. Raw Illumina reads were quality‑checked using FastQC v0.11.9, adapter contamination was removed with AdapterRemoval v1.5.4, and base quality correction was performed with SOAPec v2.0 (K‑mer = 17) before being used for polishing. PacBio long reads were first assembled de novo using Canu v2.2 to generate draft contigs and scaffolds [15]. The corrected Illumina short reads were then aligned to these PacBio scaffolds using BWA‑MEM, and two rounds of error correction were performed using Pilon v1.23 to correct base substitutions, small insertions and deletions, and local misassemblies [16]. The completeness of the genome assembly was assessed using BUSCO (v3.0.2, https://busco.ezlab.org) [17].

Genome annotation

Gene model prediction was carried out using three de novo prediction tools: Augustus (v3.03), GlimmerHMM (v3.0.1), and SNAP (v2006-07–28). Homology-based gene prediction was conducted using Exonerate (v2.2.0) with protein sequences from closely related species as references. The results from de novo predictions and homology-based approaches were integrated using EVidenceModeler (vr2012-06–25), producing a comprehensive gene annotation. Non-coding RNA (ncRNA) annotation included identifying tRNAs with tRNAscan-SE (v1.3.1) and rRNAs with RNAmmer (v1.2). Additional ncRNA elements were identified through comparative analysis with the Rfam database (http://rfam.xfam.org/), allowing for the identification of conserved RNA families and regulatory RNAs. Functional gene annotation was performed using five databases, including Nr (https://ftp.ncbi.nlm.nih.gov), eggNOG (https://www.ncbi.nlm.nih.gov/COG/), Swiss-Prot (https://www.expasy.org/resources/uniprotkb-swiss-prot), KEGG (https://www.kegg.jp/kegg/) and GO (http://geneontology.org/), to gain insights into gene functions and associated metabolic pathways. Additionally, carbohydrate-active enzyme (CAZy) genes (http://www.cazy.org) within the genome were predicted using hmmscan (v3.1b2). The analysis focused on open reading frames (ORFs) exceeding 80 amino acids in length, with an E-value threshold of 1e-5 to ensure high-confidence alignments.

Phylogenetic analysis

For phylogenomic analysis with whole genome sequence, the final annotated data were processed to retain the protein-coding genes and the longest transcript. OrthoFinder v2.5.5 was used to perform gene family analysis using Hmp‑F48 and 17 other Aspergillus strains, with Penicillium chrysogenum IBT 35,668 included as an outgroup (Table S1 lists all genomes and NCBI accession numbers), resulting in a total of 25,144 clusters of orthologous proteins (COPs) across all genomes analyzed. Multiple sequence alignments of these COPs were generated with ParaAT v1.0, and the resulting sequences were concatenated into a supergene using seqkit v2.7.0. Phylogenomic trees were constructed using RAxML-NG v1.2.1 under the GTR + G + I model with 1000 bootstrap replications. All resulting trees were visualized using FigTree v1.4.4 or ITOL (Interactive Tree of Life), with final layout adjustments made in Adobe Illustrator CC 2019. In addition, Average Nucleotide Identity (ANI) values between Hmp-F48 and related strains were calculated using FastANI v1.33 with default parameters, providing quantitative measures of genome similarity to support phylogenomic relationships.

Prediction of biosynthetic gene clusters

The BGCs responsible for secondary metabolite production in the Hmp-F48 genome were predicted using the antiSMASH (fungal version 8.0.0) web server (http://antismash.secondarymetabolites.org/). Each identified gene cluster were extracted from the genome and subjected to BLAST analysis to identify homologous sequences and predict potential product of the cluster.

Fungal metabolite extraction and separation

The Hmp-F48 strain was cultured in PDB medium at 28 °C with shaking at 200 rpm for 7 days. Following fermentation, the culture broth was extracted three times with ethyl acetate. The combined organic phases were concentrated under reduced pressure to yield a crude extract (2.63 g). Initial fractionation of the crude extract was performed using silica gel column chromatography with dichloromethane-methanol gradient (90:10 to 80:20, v/v), resulting in two subfractions (Fr.1 and Fr.2). Subfraction Fr.2 was further purified using an ODS column with a methanol-water gradient (20:80 to 40:60, v/v), yielding four subfractions (Fr.1-Fr.4). Finally, semi-preparative HPLC was performed on a SHIMAZU LC-20 equipped with a DAD detector and a YMC-ODS-A C18 column (250 × 4.6 mm, 5 μm) to isolate compounds 1–4 from these subfractions. Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker AV-600 spectrometer (Billerica, MA, USA) using tetramethylsilane (TMS) as the internal standard. HR-ESI-MS analysis were performed on a Bruker micrOTOF-Q and a Agilent.

Results

Genome sequencing and assembly

The genome of A. puulaauensis Hmp-F48 consists of 30 contigs with a GC content of 49.72% and a total assembly length of 35.86 Mb. The N50 and N90 values are 2,260,453 bp and 688,255 bp, respectively, indicating high assembly contiguity (Table 1). BUSCO analysis (290 total groups) of the assembly revealed that 99.3% of the genes are complete, confirming the genome’s high completeness and reliability (Table S2). These findings confirm that the Hmp-F48 genome is highly contiguous and nearly complete, supporting further genomic and functional investigations.

Table 1.

Statistics of genome assembly results of Hmp-F48

Property	Contig
Shortest (bp)	15,037
Longest (bp)	5,691,020
Total sequence number	30
N20 (bp)	5,140,981
N50 (bp)	2,260,453
N90 (bp)	688,255
N number	0
N rate	0
Total sequence length (bp)	35,864,449
GC content (%)	49.72%
Sequences greater than 1 kb	30

Total sequence number: The number of contiguous sequences generated during assembly; Total sequence length (bp): The assembly length of the genome; Shortest sequence length (bp): The length of the shortest contig; Longest sequence length (bp): The length of the longest contig; GC content: The percentage of guanine and cytosine bases in the genome; N number: The total number of undetermined nucleotides (N) across all contigs; N rate: The percentage of ambiguous bases relative to the total assembly length; Sequences greater than 1 kb: The number of contigs longer than 1 kb in length; N20 (bp): The weighted median contig length at 20% of the total assembly size, calculated by summing the lengths of contigs from longest to shortest until 20% of the total assembly size is reached; The N50 and N90 algorithm is the same as N20

Morphology, classification and phylogenetic analysis of strain Hmp-F48

Strain Hmp-F48 was grown on PDA medium at 28 °C for 4–5 days, forming colonies with diameters ranging from 2 to 8 mm. The colonies displayed a yellow-green coloration with a green central region and yellow edges, while the reverse side exhibited a yellow-brown hue without distinct striations. The conidia were predominantly spherical, with a granular texture and yellow-green coloration (Fig. 1A). Scanning electron microscopy (SEM) revealed globose conidia with granular and ridged surface ornamentation, borne in dense radiate clusters on smooth-walled hyphae (Fig. 1B). Phylogenomic analysis based on whole-genome protein sequences positioned Hmp‑F48 within the Aspergillus versicolor clade, showing its closest relationship to Aspergillus puulaauensis MK2 [18]. ANI analysis using FastANI v1.33 supported this placement, with an ANI score of 96% between Hmp‑F48 and A. puulaauensis MK2 (Table S3), providing robust genomic evidence for its taxonomic assignment (Fig. 1C).

Fig. 1.

Morphological and molecular taxonomy of Hmp-F48. A The colony of strain Hmp-F48 on PDA Medium. B Scanning electron microscopy (SEM) images of Hmp-F48 conidia and hyphae. a: 1000× b: 5000× c: 20,000×. Scale bars: shown in panels. Imaging parameters: EHT = 5.00 kV, WD = 8.1–8.2 mm, SE2 detector. C Whole-genome protein sequence-based phylogenetic tree of strain Hmp‑F48 and related Aspergillus species, with P. chrysogenum IBT 35,668 as the outgroup

Prediction of genetic structure

The structural prediction analysis of Hmp-F48 identified 10,611 protein-coding genes, with a total coding sequence (CDS) length of 16,857,201 bp, accounting for 47.00% of the genome. The average gene length was calculated to be 1588.6 bp, with an average of 2.9 exons per gene. With a total length of 15,135,382 bp, exons represent 42.20% of the genome. Introns, in contrast, have an average length of 81.4 bp. Notably, the total length of CDSs matches that of exons, also constituting 42.20% of the genome, with an average CDS length of 1426.3 bp (Table S4). The genome contains 30 copies of 5S rRNA gene, along with 5 copies each of 5.8S and 28S rRNA genes, 7 copies of 18S rRNA gene, 155 tRNA genes, and 210 ncRNA genes (Table S5), indicating a complex genomic structure with diverse functional capacities.

Annotation of gene functions

To obtain a comprehensive overview of the functional repertoire of Hmp-F48, protein-coding genes were annotated against multiple complementary databases, each providing distinct insights into taxonomy, biochemical function, and metabolic potential. The 10,611 protein-coding genes of Hmp-F48 were annotated across multiple databases, including NR, Swiss-Prot, GO [19], KEGG [20], and eggNOG [21]. Annotation success rates were 99.35% (10543 genes) for NR, 73.36% (7784 genes) for Swiss-Prot, 72.52% (7695 genes) for GO, 33.91% (3598 genes) for KEGG, and 93.42% (9913 genes) for eggNOG (Table S6). The NR database analysis revealed that most homologous sequences belonged to Aspergillus species, particularly A. sydowii CBS 593.65 and A. versicolor CBS 583.65 (Fig. 2A). GO annotation categorized the genes into biological processes, cellular components, and molecular functions. A significant proportion of genes associated with biological processes are involved in biosynthesis (1752 genes), nitrogen compound metabolism (1822 genes), transport (1307 genes), and small molecule metabolism (1088 genes). Molecular function annotations highlight roles in catalytic activities like oxidoreductase activity (1657 genes), methyltransferase activity (537 genes), and hydrolase activity targeting non-peptide or glycosidic bonds (402 genes). Additionally, a notable number of genes were involved in binding functions, particularly iron binding (2654 genes). In the cellular components category, genes contribute prevalently to cellular structure (2638 genes), intracellular components (2501 genes), organelles (1980 genes), and cytoplasmic composition (1303 genes) (Fig. 2B). KEGG pathway mapping revealed that 36.9% of genes (2,424 genes) were involved in metabolic pathways, suggesting substantial metabolic potential and capacity for diverse metabolite production. (Fig. 2C). The eggNOG classification showed 36.95% of genes (3,663 genes) with unknown functions, indicating potential for discovering novel chemistry. Among characterized genes, those related to carbohydrate transport and metabolism (956 genes, 9.64%), secondary metabolite biosynthesis, transport, and catabolism (704 genes, 7.10%), and amino acid transport and metabolism (539 genes, 5.44%) were prominent (Fig. 2D). Further functional annotations using the CAZy database identified six categories of carbohydrate-active enzymes [22]: glycosyl transferases (GT), polysaccharide lyases (PL), carbohydrate esterases (CE), auxiliary activities (AA), carbohydrate-binding modules (CBM), and glycoside hydrolases (GH) (Fig. 2E). This classification emphasizes the strain’s enzymatic diversity and potential applications in carbohydrate metabolism and biotechnological.

Fig. 2.

Gene Function Annotation. A Distribution of species based on matches from the non-redundant protein sequence (Nr) database. B Classification statistics based on Gene Ontology (GO) annotations. C Functional classification of pathways based on the Kyoto Encyclopedia of Genes and Genomes (KEGG). D Classification statistics based on Cluster of Orthologous Groups (COG) annotations. E Functional classification of Carbohydrate-Active Enzymes (CAZy)

Annotation of secondary metabolites biosynthetic gene clusters

The biosynthetic potential of Hmp-F48 for secondary metabolites was predicted using antiSMASH [23], identifying 78 putative biosynthetic gene clusters (BGCs). These BGCs mainly include 16 type I polyketide synthase (T1PKS) clusters, 27 NRPS clusters, and 7 hybrid PKS-NRPS clusters. Additionally, 9 terpene-related clusters, 6 RiPP-related clusters and 13 clusters encoding other classes of secondary metabolites were identified (Fig. 3A, Table S7). Based on their completeness and homology to known pathways, four PKS, four NRPS, and three PKS-NRPS clusters were prioritized for further functional annotation and product prediction.

Fig. 3.

A Types and abundance of biosynthetic gene clusters (BGCs) in strain Hmp-F48. B Chemical structures of products from homologous gene clusters corresponding to BGCs 1–11 in strain Hmp-F48

PKS gene clusters

Cluster 1 (T1PKS)

Shares 66–100% sequence identity with the asperthecin BGC in A. nidulans FGSC A4 [24]. Asperthecin, a polyphenolic compound, facilitates ascus formation and confers UV resistance in fungi [25]. Notably, its inhibition of tau protein aggregation highlights therapeutic potential in neurodegenerative diseases [26]. The cluster harbors homologs of aptA (PKS), aptB (thioesterase), and aptC (monooxygenase), suggesting a conserved role in asperthecin-like polyketide biosynthesis (Figs. 3B and 4A).

Fig. 4.

Synteny plots of BGCs 1–11 and their homologous clusters, generated using Easyfig (BLASTn). A Syntenic comparison between the predicted asperthecin cluster in Hmp-F48 and the identified candidate cluster in A. nidulans FGSC A4. B Syntenic comparison between the predicted sterigmatocystin cluster in Hmp-F48 and the identified candidate cluster in A. ochraceoroseus. C Syntenic comparison between the predicted calbistrins cluster in Hmp-F48 and the identified candidate cluster in A. stellatus. D Syntenic comparison between the predicted F-9775 A/B cluster in Hmp-F48 and the identified candidate cluster in A. nidulans FGSC A4. E Syntenic comparison between the predicted nidulanin A cluster in Hmp-F48 and the identified candidate cluster in A. nidulans FGSC A4. F Syntenic comparison between the predicted aspercryptins cluster in Hmp-F48 and the identified candidate cluster in A. nidulans FGSC A4. G Syntenic comparison between the predicted fellutamide B cluster in Hmp-F48 and the identified candidate cluster in A. nidulans FGSC A4. H Syntenic comparison between the predicted acetylaranotin cluster in Hmp-F48 and the identified candidate cluster in A. terreus NIH2624. I Syntenic comparison between the predicted burnettramic acid A cluster in Hmp-F48 and the identified candidate cluster in A. burnettii. J Syntenic comparison between the predicted equisetin cluster in Hmp-F48 and the identified candidate cluster in F. heterosporum. K Syntenic comparison between the predicted pyranonigrin E cluster in Hmp-F48 and the identified candidate cluster in A. niger ATCC 1015

Cluster 2 (T1PKS)

Displays 66–100% homology to the sterigmatocystin cluster in A. ochraceoroseus [27]. Sterigmatocystin is a potent carcinogen, mutagen and teratogen produced by several species of Aspergillus [28]. With a xanthone nucleus attached to a bifuran structure, sterigmatocystin is a biosynthetic precursor to aflatoxin. Notably, this BGC harbors additional transporter genes compared with the known reference cluster (Figs. 3B and 4B).

Cluster 3 (T1PKS)

Shows 71–100% homology to the calbistrin/decumbenone biosynthetic cluster cal from A. stellatus [29, 30]. Calbistrins are antifungal and cytotoxic natural products structurally analogous to lovastatins due to their decalin core and polyketide side chain [31, 32] (Figs. 3B and 4C).

Cluster 4 (T1PKS)

Shares 68–100% identity with the cysteine protease cathepsin K inhibitor F-9775 A/B biosynthetic cluster in A. nidulans FGSC A4 [33]. In addition to the homologous genes AN7909-AN7914, this cluster contains several other genes with unknown functions (Figs. 3B and 4D).

NRPS gene clusters

Cluster 5

Homologous (66–100%) to the nidulanin A biosynthetic cluster in A. nidulans FGSC A4 [34, 35] which is known for producing tetracyclopeptide/isoprene hybrid metabolites (Figs. 3B and 4E).

Cluster 6

Shares 69–100% identity with the atn gene cluster producing aspercryptin A1 in the A. nidulans FGSC A4 [36]. The core biosynthetic genes for aspercryptin A1 include an NRPS gene (atnA) and two fatty acid synthase genes (atnF and atnM). Notably, antiSMASH predicts the absence of aminotransferase genes critical for α-amino fatty acid synthesis in this gene cluster, suggesting either truncation of the cluster or that these genes lie outside its boundaries (Figs. 3B and 4F). Therefore, this cluster likely encodes a noncanonical or an incomplete aspercryptin biosynthetic machinery.

Cluster 7

Shows significant alignment (69–100% similarity) with the fellutamide B BGC in A. nidulans FGSC A4. The proteasome subunit InpE in this cluster [37] is consistent with fellutamide B’s role as a proteasome inhibitor with antitumor potential [38] (Figs. 3B and 4G).

Cluster 8

Exhibiting 69–100% homology to the acetylaranotin cluster in A. terreus [39], associated with producing cytotoxic diketopiperazine natural products, epipolythiodioxopiperazines [40] (Figs. 3B and 4H).

PKS-NRPS hybrid clusters

Cluster 9

Highly homologous with the burnettramic acid A BGC in A. burnettii, comprising a PKS-NRPS core gene buaA and tailoring genes buaB-G [41]. In previously studies, burnettramic acid A exhibited potent antimicrobial activity, particularly against Candida albicans [42] (Figs. 3B and 4I).

Cluster 10

With 64–100% sequence similarity to the equisetin cluster from Fusarium heterosporum [43], this BGC is propose to synthesize decalin-containing compounds structurally related to lovastatins, known for their lipid-lowering, antimicrobial, and anti-inflammatory activities [44, 45] (Figs. 3B and 4J).

Cluster 11

Exhibits partial homology to the pyranonigrin E cluster from A. niger [46], suggesting a potential for producing pyranonigrin-like metabolites. As is typical for all in silico predicted BGCs within this study, its associated product profiles require future experimental validation (Figs. 3B and 4K).

Isolation of secondary metabolites

To functionally annotate the predicted BGCs in Hmp-F48, we systematically isolated and structural elucidated its secondary metabolites. Notably, averantin and sterigmatocystin, previously isolated from this strain, are likely products of the PKS gene clusters BGC1 and BGC2, respectively [14]. Our recent chemical investigations revealed a series of decumbenone-type derivatives, including the known compounds decumbenone B (1), craterellone F (2), versiol (3) and an undescribed congener designated as 1R, 3S, 5S, 8S, 9S, 10R−9,10-dihydroversiol (4) (Fig. 5). The NMR spectroscopic data for these compounds are provided in Figures S1-S12 and Tables S8-S9. While compound 4 shares an identical planar structure to isoversiol A, it differs in the stereochemistry of C-8 position [47]. In the NOESY spectrum of compound 4, H‑13a showed a cross‑peak with 14‑CH₃, and 14‑CH₃ correlated with 15‑CH₃. In addition, 5‑H exhibited NOE interactions with both 15‑CH₃ and H‑3. These correlations suggest that H‑13a, 14‑CH₃, 15‑CH₃, 5‑H, and H‑3 are α‑oriented. Conversely, NOE cross‑peaks between 10‑H and H‑1, H‑2a, and H‑4a indicate that these protons are β‑oriented (Fig. S1). When the modified Mosher’s method was applied to the absolute configuration at position 1, the results were inconclusive, likely due to steric hindrance from the pyran ring that limited the applicability of this method. Based on the typical configuration observed in compounds isolated from A. versicolor is 1R, 3S, 5S, 9S, 10R, we propose that compound 4 has an absolute configuration of 1R, 3S, 5S, 8S, 9S, 10R [48].

Fig. 5.

Chemical structures of the compounds isolated from strain Hmp-F48

Although in silico analyses suggest diverse biosynthetic machinery including NRPS, terpene, RiPP, and hybrid clusters, all experimentally characterized metabolites from Hmp-F48 thus far are derived from PKS pathways. This is likely due to the limitation in current chemical profiling approaches. On the other hand, many gene clusters are silent under regular culture conditions. To fully understand their products, culture condition optimization, gene regulation and heterologous expression may be required to activate these cryptic BGCs.

Discussion

The comprehensive genomic analysis of A. puulaauensis Hmp-F48 revealed a high-quality assembly (35.86 Mb, N50 2.26 Mb), comparable to those of the reference strains such as A. versicolor CBS 583.65 (33.1 Mb, N50 2.5 Mb) and A. sydowii CBS 593.65 (34.4 Mb, N50 2.3 Mb). Notably, Hmp-F48 contained no ambiguous bases and was assembled into only 30 scaffolds, surpassing the completeness of CBS 583.65 (51 scaffolds) and CBS 593.65 (97 scaffolds). This enhanced continuity provides a robust foundation for functional annotation.

AntiSMASH analysis of the A. puulaauensis Hmp-F48 genome revealed 78 biosynthetic gene clusters (BGCs). This number is comparable to, and slightly higher than, those found in closely related Aspergillus species, such as A. puulaauensis MK2 (69 BGCs) [49], A. nidulans FGSC A4 (61 BGCs) [50], A. sydowii CBS 593.65 (59 BGCs) [51], and A. versicolor CBS 583.65 (66 BGCs) [52]. These observations indicate that Hmp‑F48 has a predicted BGC repertoire at the upper end of the range typically observed in Aspergillus species. The diversity of BGC types further underscores the strain’s metabolic versatility, encompassing 27 NRPS, 16 T1PKS, 7 hybrid clusters, 9 terpene-related clusters, 6 RiPP-related clusters and 13 clusters associated with other secondary metabolites. Particularly noteworthy is the presence of 6 RiPP-related BGCs, while NRPS and PKS type BGCs are commonly observed in Aspergillus species [53]. The relatively high number of RiPP clusters in Hmp-F48 suggests a significant capacity for RiPP biosynthesis and merits further investigation. Several of the most homologous BGCs (1–11) also show partial loss of synteny with their closest references and display gene content variation, such as additional or missing genes (e.g., Cluster 2 and 4). Meanwhile, a considerable number of BGCs with low homology to known clusters incomplete clusters, or clusters lacking core genes remain uncharacterized. This suggests the possibility of unique, novel metabolites and pathways awaiting discovery. Future research should therefore not only characterize these cryptic clusters but also assess the biological activities of selected metabolites to clarify their ecological and biotechnological relevance.

Our experimental isolation efforts primarily led to structure determination of polyketides, such as averantin, sterigmatocystin, and decumbenone derivatives. These findings likely result from the inherent bias of conventional organic solvent extraction and column chromatography, which favors moderately polar PKS products while potentially overlooking hydrophilic peptides or volatile terpenoids. Expanding the range of extraction techniques — including advanced liquid-liquid partitioning, solid-phase extraction, and headspace trapping — may enhance the recovery of diverse metabolites. Additionally, optimizing culture conditions (e.g., altering nutrient composition, co-cultivation, or stress induction) and exploring heterologous expression or epigenetic manipulation could help activate cryptic BGCs [54].

Although Hmp-F48 exhibits remarkable potential for secondary metabolite production, it should be noted that the current isolation and identification efforts were restricted to major metabolites under specific culture conditions. Consequently, these findings cannot reflect the strain’s comprehensive biosynthetic repertoire. Future studies employing an integrated multi-omics approach—encompassing transcriptomics, proteomics, and metabolomics—across diverse culture conditions are expected to address this limitation. This strategy will enable us to link predicted gene clusters with detected metabolites and elucidate the strain’s unique biosynthetic machinery [55, 56], ultimately facilitating the discovery of novel chemical scaffolds and biocatalysts.

Conclusion

This study presents a comprehensive genomic analysis of the sponge-associated fungus A. puulaauensis Hmp-F48, offering valuable insights into its genetic architecture and biosynthetic capacities. The high-quality genome assembly (35.86 Mb) and the annotation of 10,611 protein-coding genes illustrate the strain’s extensive metabolic potential. Functional annotation across multiple databases (Nr, Swiss-Prot, GO, KEGG, eggNOG, and CAZy) revealed a broad spectrum of metabolic pathways and a substantial proportion of uncharacterized genes, indicating the possibility of discovering novel bioactive compounds. AntiSMASH predictions identified 78 BGCs, predominantly involved in the biosynthesis of polyketides and nonribosomal peptides. The detailed analysis of conserved clusters confirmed Hmp-F48’s ability to produce notable bioactive metabolites, including the proteasome inhibitor fellutamide B and the antimicrobial burnettramic acid A. The isolation of a potentially novel decumbenone derivative (compound 4) further supports the strain’s ability to generate unique secondary metabolites. Despite the promising genomic data, the limited diversity of isolated metabolites highlights the need for deeper exploration of the strain’s secondary metabolism. Targeted activation of cryptic BGCs, and heterologous expression techniques should be employed to maximize the recovery of structurally diverse and biologically significant metabolites. Ultimately, this research underscores the potential of Hmp-F48 as a promising resource for natural product discovery, with implications for biotechnology and drug development.

Abbreviations

NSCLC

Non-small cell lung cancer

CIN

Chemotherapy-induced neutropenia

BGCs

Biosynthetic gene clusters

PDA

Potato dextrose agar

PDB

Potato dextrose broth

ITS

Internal transcribed spacer region

SEM

Scanning electron microscopy

CDS

Protein coding sequence

Nr

Non-redundant protein sequence

COG

Cluster of orthologous group of proteins

KEGG

Kyoto encyclopedia of gene and genome

GO

Gene ontology

CAZy

Carbohydrate-active enzymes

GT

Glycosyl transferases

PL

Polysaccharide lyases

CE

Carbohydrate esterases

AA

Auxiliary activities

CBM

Carbohydrate-binding modules

GH

Glycoside hydrolases

PKS

Polyketide synthase

NRPS

Nonribosomal peptide synthetase

NOESY

Nuclear overhauser effect spectroscopy

MTPA

α-Methoxy-α-(trifluoromethyl) phenylacetic acid

Authors’ contributions

N.W. and N.L. conceived the project; N.W., N.L., Y.Y., and X.W. designed the experiments; Y.Y. performed bioinformatics, X.W. and Y.Y. performed natural product isolation and structure determination; N.W., Y.Y., X.W., N.L., Q.M. Y.L. and Z.C. analyzed the results; Y.Y. wrote the first draft with input from all co-authors; N.W. and Y.Y. revised and finalized the manuscript. All authors reviewed the manuscript.

Funding

This work was financially supported by the Key Research and Development Program of Guangxi (AB24010108) and the National Natural Science Foundation of China (22077113).

Data availability

Sequence data that support the findings of this study have been deposited in the NCBI database under the BioProject PRJNA1249370, with the accession number JBNBLO010000000.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.