Background & Summary

The fungal pathogen Agroathelia rolfsii poses a significant threat to global agriculture, particularly in tropical and subtropical regions1. It affects over 500 plant species, including economically important crops such as peanut, tomato, and cotton2,3,4. Its asexual stage, Sclerotium rolfsii, is characterized by the production of highly resilient sclerotia, which can survive under adverse environmental conditions5. Once infected, plants exhibit basal stem and root rot, leading to wilting and, in severe cases, total crop failure. Yield losses can reach to 30% ~50%, making A. rolfsii a critical target for plant disease management research6,7.

While chemical fungicides, such as tebuconazole and flutolanil, have been used to control A. rolfsii, their prolonged application has raised concerns over environmental pollution and the development of fungicide resistance8,9. Biological control strategies, including the use of antagonistic microorganisms like Trichoderma spp., have shown potential but may lack consistency and universal applicability in field conditions10. Furthermore, the pathogen’s broad host range and strong ecological adaptability complicate efforts to develop effective and sustainable control measures.

In recent years, the rapid advancement of high-throughput sequencing technologies has provided powerful tools for studying fungal pathogens. Genomic studies have revealed virulence-related gene families in many plant-pathogenic fungi, including glycoside hydrolases, peptidoglycanases, and secondary metabolite biosynthetic gene clusters, which play key roles in host infection and tissue colonization11,12. Comparative genomic analyses have also provided insights into the genetic basis of pathogenicity, host adaptation, and environmental resilience, helping to identify key genes involved in fungal survival and infection strategies13. The emergence of genome-editing technologies, such as CRISPR-Cas9, has opened new possibilities for enhancing plant resistance by modifying key susceptibility genes14,15. Additionally, genomic information has facilitated the identification of novel targets for disease control, including genes essential for fungal virulence and survival16.

Studies on closely related pathogens, such as Sclerotium delphinii, which causes southern blight in Dendrobium officinale, provide additional context17. Genomic analyses of S. delphinii have revealed a significantly higher abundance of glycoside hydrolases (GHs) compared to A. rolfsii, which may explain its enhanced virulence. Furthermore, the identification of secondary metabolite gene clusters and pathogen-host interaction (PHI) factors in S. delphinii highlights the diverse strategies employed by fungal pathogens during infection. These findings emphasize the importance of comparative studies in understanding the molecular basis of fungal pathogenicity.

Here, we assembled a high-quality genome of A. rolfsii using an integrated approach that combined PacBio Single Molecule Real-Time (SMRT) long-read sequencing and Illumina short-read sequencing technologies. The quality of the genome assembly and annotation was evaluated based on the contiguity of assembled sequences (N50) and the completeness of conserved protein-coding genes. Protein functional annotations were performed by aligning the sequences against multiple databases, including Nr, SwissProt, KEGG, and KOG, ensuring comprehensive coverage. Furthermore, evolutionary insights were gained through comparative analyses with genomes of other Agroathelia species. This study provides a valuable genomic resource for advancing research on the pathogenicity, genetic diversity, and evolutionary dynamics of the Agroathelia genus.

Methods

Fungal strains

The A. rolfsii strain LC1 was isolated from the leaf of Coptis chinensis, and is stored at the Institute of Chinese Herbal Medicines, Hubei Academy of Agricultural Sciences, China. For fugal growth, the strain was incubated on solid potato dextrose agar (PDA) at 28 °C in the dark (Fig. 1).

Fig. 1
Fig. 1
Full size image

Disease symptoms caused by Agroathelia rolfsii strain LC1 on field and potted Coptis chinensis, and colony morphology of LC1 on PDA. (a) Southern blight symptoms on field-grown C. chinensis infected by strain LC1. Dense white mycelia and sclerotia are observed on the left, accompanied by plant death, while healthy leaves appear on the right. (b, c) Pathogenicity assay on potted C. chinensis. (b) Mock-inoculated control plants at 10 days post-inoculation (dpi), showing no symptoms. (c) Inoculated plants at 10 dpi showing severe wilting and death. (d–f) Colony morphology of A. rolfsii strain LC1 on PDA at different time points: (d) 4 dpi, (e) 8 dpi, and (f) 10 dpi, with visible sclerotia formation.

DNA extraction and genome sequencing

Total genomic DNA was extracted using the GP1 method (Novogene, Beijing, China) following the manufacturer’s instructions. The purity of the DNA was assessed using a NanoPhotometer® spectrophotometer (IMPLEN, CA, USA), and DNA concentration was quantified using the Qubit® DNA Assay Kit with a Qubit® 2.0 Fluorometer (Life Technologies, CA, USA). For Illumina short-read sequencing, a DNA library with a 350 bp insert size was prepared using the TruSeq Nano DNA HT Sample Preparation Kit (Illumina, USA) in accordance with the manufacturer’s recommendations. Sequencing was performed on the Illumina HiSeq X Ten platform with a paired-end read length of 2 × 150 bp, generating 2.3 Gb of Illumina data, corresponding to approximately 53 × genome coverage. For PacBio long-read sequencing, a sequencing library was constructed and sequenced on the Sequel II System platform. High-quality long reads were obtained using the Guppy basecalling software with default parameters, yielding 8.06 Gb of PacBio sequencing data, which represents approximately 186 × genome coverage.

Genome assembly

The genome size of A. rolfsii was first inferred from a 15-mer distribution using jellyfish software (v2.3.0) based on Illumina paired-end reads18, and the genome size was estimated to be 41.18 Mb (Fig. 2a; Table 1). For de novo genome assembly, PacBio long reads corrected with Falcon (v1.8.7)19 were assembled to generate an initial assembly by Wtdbg2 with the following parameters: -x sq and -g 44 m. The PacBio sequencing data provides significant advantages for genome assembly due to its long-read lengths, which simplify the complexity of assembly algorithms by enabling overlap-based assembly. To further enhance the accuracy and completeness of the assembly, Pilon (v1.23)20 was utilized to perform three iterative rounds of base correction using high-quality Illumina short reads with default settings. To the end, the assembled genome of A. rolfsii was found to be 43.3 Mb in length, with a Contig N50 of 3.3 Mb and a GC content of 46.3% (Fig. 2b; Table 1). The assembled genome consists of 50 Contigs (Table 1).

Fig. 2
Fig. 2
Full size image

Genomic features of the A. Agroathelia isolate. (a) Genome survey based on 21-mer frequency distribution of base error-corrected reads. (b) Circos plot of the high-quality genome assembly. Tracks from outer to inner circle represent: contigs (n = 50), gene density (20-kb window), transposon density, simple repeat density, GC content (red above and green below the genome average), and syntenic blocks among contigs.

Table 1 Statistics of the Agroathelia rolfsii genome assembly.
Full size table

Genome annotation

Using RepeatModeler v1.0.8 (http://www.repeatmasker.org/RepeatModeler/), a de novo repeat library was generated and subsequently combined with the Repbase library v23.09 (http://www.girinst.org/repbase/). The merged library was then used for repeat identification with RepeatMasker v4.1.2-p1. This analysis identified 13.38 Mb of repetitive sequences, which accounted for 30.93% of the genome. Among these, long terminal repeats (LTRs; 27.56%) and DNA elements (1.6%) were the most abundant repetitive components, with unclassified repeats also making a significant contribution.

The repeat-masked A. rolfsii genome was annotated using Funannotate (http://github.com/nextgenusfs/funannotate), an automated pipeline designed for fungi but also applicable to other eukaryotes. Protein evidence was incorporated from curated UniProtKB/SwissProt datasets included in Funannotate. Gene prediction employed multiple tools integrated within the pipeline, including including Augustus v3.3.321, GeneMark-ES v4.3222, SNAP v2006-07-2823, and GlimmerHMM v3.0.424 (all ab initio predictors), as well as CodingQuarry v2.025, which utilizes transcriptome data for gene model inference. The combined results yielded a consensus gene model comprising 8,826 protein-coding genes, with an average length of 2,504 bp and a maximum length of 15,609 bp. A total of 7,907 (89.6%) of which were successfully annotated in at least one database (Table 2). Similarly, we utilized several complementary software tools to predict non-coding RNAs in the A. rolfsii genome, including RNAmmer for rRNA prediction, tRNAscan-SE for identifying tRNA regions and secondary structures, and cmscan (with default parameters) for detecting sRNAs and miRNAs by aligning sequences to the Rfam database. A total of 209 ncRNAs were identified in the A. rolfsii genome, including 160 tRNAs, 41 rRNAs (13 5S, 12 28S, and 16 18S), 1 sRNA, and 7 miRNAs.

Table 2 Functional annotation of Agroathelia rolfsii proteins.
Full size table

Identification of CAZymes and biosynthetic gene cluster

The putative carbohydrate-active enzymes (CAZymes) in A. rolfsii were annotated using the dbCAN server (dbCAN HMMs 5.0)26, applying thresholds of an e-value lower than 1e-5 and at least 70% sequence coverage. CAZymes were categorized based on the modules outlined in the CAZyme database classification (http://www.cazy.org). The A. rolfsii genome encodes a total of 1,260 genes for putative CAZymes, which are distributed across six functional categories. Among these, glycoside hydrolases (475 genes, 37.7%) were the most prevalent, followed by glycosyltransferases (331 genes, 26.3%) (Table 3). Polysaccharide lyases were the least represented, with only one gene identified (Table 3). To explore secondary metabolism, the antiSMASH 4.0 tool was employed for identifying gene clusters within the A. rolfsii genome. A total of 22 clusters associated with secondary metabolites were predicted, including four terpene clusters, three type I polyketide synthase clusters, and one non-ribosomal peptide synthase cluster (Fig. 3a).

Table 3 Identification of CAZymes in the Agroathelia rolfsii genome.
Full size table
Fig. 3
Fig. 3
Full size image

Identification of carbohydrate-active enzymes (CAZymes) and biosynthetic gene clusters (BGCs), and phylogenetic analysis. (a) Distribution of CAZyme and BGC families. (b) Phylogenetic analysis of nine species based on single-copy orthologous genes.

Data Record

All sequencing data, including the raw PacBio long reads and Illumina short reads, have been deposited in the Genome Sequence Archive (GSA27) in National Genomics Data Center28 under BioProject PRJCA042508. The PacBio raw data are available under accession CRA02743029, and the Illumina raw data under CRA02741730. The assembled genome has been made available in GenBank with JBQWDA00000000031. Additionally, the annotation data has been deposited at the Figshare repository32.

Data Overview

Protein sequences of 9 genomes were used for comparative genomics and phylogenomic analyses: A. rolfsii (this study), A. delphinii SD strain (GCA_041475395.1), A. rolfsii GP3 isolate (GCA_018343915.1), A. rolfsii Sr-RCBB strain (GCA_002940785.1), A. rolfsii CCF1 strain (GCA_036172465.1), A. rolfsii ZY isolate (GCA_018343895.1), A. rolfsii MR10 strain (GCA_000961905.2), Fusarium oxysporum Fo5176 strain (GCA_030345115.2), Fusarium solani F421 strain (GCA_040436825.1). The OrthoFinder (v2.5.1)33 with default settings was used to identify orthologous and paralogous genes. All protein sequences were aligned with MAFFT (v7.123b)34 and trimmed by trimAl (v1.4.rev22)35. The sequences were concatenated into a gene matrix and used for phylogenetic analysis (Fig. 3b). A phylogenetic tree was constructed by maximum likelihood (ML) method using IQ-TREE (v2.1.2)36, with 1,000 ultrafast bootstrap replicates. The best model (JTT + F + R6) was determined by ModelFinder37 and two Fusarium species were used as outgroups. A total of 758 single-copy orthogroups were identified and extracted for phylogenetic analysis (Fig. 3b). Among the analyzed strains, F. oxysporum and F. solani exhibit the highest gene counts, with 11,419 and 10,495 genes, respectively, forming an outgroup relative to the Agroathelia species. The Agroathelia rolfsii strain analyzed in this study contains 6,921 genes and clusters closely with A. delphinii (SD strain, 9,877 genes). Within A. rolfsii, notable gene count variations are observed, ranging from 4,051 genes in the GP3 isolate to 10,816 genes in the ZY isolate.These results highlight substantial genomic diversity within A. rolfsii, which may be attributed to differences in ecological adaptation, host specificity, or pathogenicity potential.

Technical Validation

Quality control of PacBio long reads was performed using Chopperv1.5.0 (http://gitcode.com/gh_mirrors/cho/chopper), resulting in 606,486 reads with an N50 length of 26 kb. Illumina short reads were processed using Fastp38 to remove low-quality reads, generating a total of 15,321,002 high-quality reads for subsequent analysis. The completeness of the genome was assessed using BUSCO (Benchmarking Universal Single-Copy Orthologs) v5.5.0 with the basidiomycota_odb10 database (genome mode)39. The evaluation yielded the following BUSCO statistics: 97.7% complete, 96.8% complete single-copy, 0.3% duplicated, 0.6% fragmented, and 2.3% missing (Table 4). The annotated and integrated proteins were further evaluated using BUSCO v5.5.0 with the basidiomycota_odb10 database (protein mode), which showed a complete BUSCO score of 88.6% (Table 5). Additionally, 89.6% of the genes were functionally annotated across four major databases (Nr, SwissProt, KEGG, and KOG) (Table 2), and future improvements will involve refining the genome annotation with full-length transcriptome data using the cDNA_Cupcake toolkit (https://github.com/Magdoll/cDNA_Cupcake). The genome assembly quality value (QV value) was assessed using Merqury (http://github.com/marbl/merqury), a k-mer-based tool designed for evaluating genome assembly accuracy and completeness without a reference genome. The process involved comparing the k-mers (specifically 21-mers) from short-read sequencing data with those from the assembled genome. The assessment yielded a high QV value of 44, reflecting the high base accuracy of the assembly.

Table 4 Busco of the Agroathelia rolfsii genome.
Full size table
Table 5 Busco of the Agroathelia rolfsii protein-coding genes.
Full size table