Genome-scale analyses and characteristics of putative pathogenicity genes of Stagonosporopsis cucurbitacearum, a pumpkin gummy stem blight fungus

Outbreaks of gummy stem blight (GSB), an emerging seed pumpkin disease, have increased in number and have become more widespread in recent years. Previously we reported that Stagonosporopsis cucurbitacearum (Sc.) is the dominant fungal cause of pumpkin seedling GSB in Northeast China, where it has greatly reduced crop yields in that region. Here, high-throughput whole-genome sequencing and assembly of the Sc. genome were conducted toward revealing pathogenic molecular regulatory mechanisms involved in fungal growth and development. Zq-1 as representative Sc. strain, DNA of Zq-1was prepared for genomic sequencing, we obtained 5.24 Gb of high-quality genomic sequence data via PacBio RS II sequencing. After sequence data was processed to filter out low quality reads, a hierarchical genome-assembly process was employed that generated a genome sequence of 35.28 Mb in size. A total of 9844 genes were predicted, including 237 non-coding RNAs, 1024 genes encoding proteins with signal peptides, 2066 transmembrane proteins and 756 secretory proteins.Transcriptional identification revealed 54 differentially expressed secretory proteins. Concurrently, 605, 130 and 2869 proteins were matched in the proprietary databases Carbohydrate-Active EnZymes database (CAZyme), Transporter Classification Database (TCDB) and Pathogen–Host Interactions database (PHI), respectively. And 96 and 36 DEGs were identified form PHI database and CAZyme database, respectively. In addition, contig00011.93 was an up-regulated DEG involving ATP-binding cassette metabolism in the procession of infection. In order to test relevance of gene predictions to GSB, DEGs with potential pathogenic relevance were revealed through transcriptome data analysis of Sc. strains pre- and post-infection of pumpkin. Interestingly, Sc. and Leptosphaeria maculans (Lm.) exhibited relatively similar with genome lengths, numbers of protein-coding genes and other characteristics. This work provides a foundation for future exploration of additional Sc. gene functions toward the development of more effective GSB control strategies.


Scientific Reports
| (2020) 10:18065 | https://doi.org/10.1038/s41598-020-75235-x www.nature.com/scientificreports/ quality of cucurbit crops 5 . Therefore, new strategies are urgently needed to combat this threat, with genomics data sought as a powerful information resource toward achieving this goal. However, since no complete Sc. genome sequence has yet been reported, most available genetic information with regard to this plant pathogen is limited to data obtained from past studies of spatial genetic structure and population dynamics 6 , genetic structure and evolutionary analysis 7 , diversity analysis [8][9][10] , rapid diagnosis 2 , host range 11,12 , host resistance genes 13,14 and development of efficient fungicides 15,16 . While this information has been helpful, a genome-scale level of information would dramatically expand our ability to understand pathogenic Sc. mechanisms. Whole-genome sequencing and genome-wide analysis have already been completed for thousands of fungi 17 and are particularly important for understanding fungal pathogenic mechanisms. In 2005, the draft genome sequence of Magnaporthe grisea was annotated and released, revealing a variety of secreted proteins, including new virus-related genes and enzymes involved in secondary metabolism. Such information has greatly enhanced our understanding of fungal life processes and our ability to predict fungal pathogenic mechanisms 18 . As a separate project involving the sequencing and annotation of the genome of a strain of Fusarium graminearum, a major pathogen of cultivated grain, Cuomo (2007) discovered a highly polymorphic region containing a cluster of genes related to plant-fungus interactions 19 . Subsequently, Kronstad (2008) reviewed partial genome sequence results for Ustilago maydis and demonstrated the power of comparative genomics to determine genomic similarities and differences between species as a means of harnessing genomic information from diverse fungi to enhance understanding of fungal processes 20 . More recently (2017), systematic bioinformatics analysis of bZIP genes of rice smut pathogenic fungus (Ustilaginoidea virens) for detection of bZIP regions in predicted genes discovered 11 UvbZIP genes associated with fungal infection as an example of how bioinformatics tools can reveal genetic structures, motifs and phylogenetic relationships 21 . In 2018, six keratinase coding genes were identified within the Rhizoctonia cerealis genome (designated RcCUT1-RcCUT6). Subsequent analysis of fungal gene expression patterns during infection led to the discovery of RcCUT1, a toxic factor produced by fungi 22 . Taken together, these studies illustrate the power of whole-genome sequencing and genome-wide analysis to greatly facilitate the discovery of fungal pathogenic molecular regulatory networks and disease-causing genes. Indeed, such methodologies have already greatly expanded our understanding of fungal pathogenic processes, prompting us to use these techniques to study S. cucurbitacearum in this work.
In the present study, high-throughput sequencing technology was used for the first time to sequence the genome of S. cucurbitacearum isolate zq-1, an Sc. isolate obtained from a cold region of China. A high quality whole-genome fine map was obtained after genome assembly with a total size of 35.28 Mb. Based on the complete Sc. genome sequence, differentially expressed Sc. genes at the transcriptional level before and after Sc. infection of pumpkin were analyzed, resulting in identification of genes of numerous secreted proteins, plant cell wall degradation enzymes and secondary metabolic enzymes that may play important roles in GSB pathology. Ultimately, the Sc. genome sequence will likely provide additional insights for development of new strategies to control GSB and other fungal diseases of crops. Characteristic of genome. The genome of Sc. was assembled from data generated using the Single Molecule Real-Time (SMRT) sequencing strategy, an effective method introduced by Pacific Biosciences to decode important sequence regions that can facilitate the generation of gapless eukaryotic genome sequences 23 . After obtaining 527,073 subreads, a total of 5.24 Gb of genome sequence was obtained. High quality of the 10-kb library was confirmed using subreads distribution analysis ( Supplementary Fig. S2) that provided a subreads N50 number of 12,559 sequences and a mean sequence length of 9946 bp (Supplementary Table S1). Subreads were assembled into 33 scaffolds (N50, 1.57 Mb) to generate a total genome length of 35.28 Mb, a genome size smaller than that of Lm. (45.12 Mb; genome assembly # ASM23037v1) ( Table 1). We have performed BUSCO to evaluate genome completeness from the annotation ( Supplementary Fig. S3).A Circos-plot, used to display genome assembly information (Fig. 1), shows thirty-three scaffolds arranged in the outermost circle and displays the predicted 9844 protein-coding genes, for a gene density of 279 genes per 1 Mb that represents a total genome coverage rate of approximately 48.22%. Sequence analysis predicted 176 secreted proteins and 14 pseudogenes from the genome sequence (Table 1). BLAST comparisons were conducted between Sc. gene sequences and multiple functional databases to obtain functional annotation results for 9562 genes (Supplementary Table S2). It is noteworthy that 8913 putative protein-coding genes were detected within the RNA-seq data.

Results
Repetitive DNA sequence and DNA methylation. Repetitive DNA sequences play important roles in genome structure, gene evolution and fungal functions 24 . Here, we identified 3,387,350 bp of repeat sequences in the Sc. genome that accounted for 9.6% of the genome, including LTR retrotransposons, DNA transposons, tandem repeat sequences and other unclassified transposons ( Fig. 2A). DNA methylation plays an important role in biological processes, such as in genomic imprinting and regulation of gene transcription 25 . In recent years, more and more reports on DNA methylation have been reported for fungi, while numerous reports exist describing this phenomenon in higher plants and animals 26 . In order to detect 4-methyl-cytosine (m4C) and 6-methyl-adenosine (m6A) methylation events within the Sc. genome, SMRT technology was used. Subsequently, 1,228,069 m4C and 30,883 m6A sites in the Sc. genome were identified (Fig. 2B), with most methylation sites (23,845,215) uncategorized. Ultimately, 99.71% of categorized DNA methylation events were categorized as m4C, whereas m6A only accounted for 0.29% of events. However, compared with m6A, m4C, DNA methylations occur with low frequency in the regions of repetitive elements ( Fig. 3A-C).  ). The outermost circle is a marker of genome size with each demarcation representing 5 kb that shows the entire genome size of Sc. zq-1 of 35.28 Mb. The second and third circles from the outside show positions of genes coded by plus and minus strands of the genome, respectively. The function and gene number annotated by COG are listed on the right of the figure whereby each function corresponds to a color and each color corresponds with a different COG functional classification. The fourth circle shows repeating sequences. The fifth circle shows genes coding for tRNA and rRNA, where blue and purple represent tRNA and rRNA, respectively. The sixth circle shows the GC content, the light yellow part shows GC content is higher than the mean GC content of genome, the higher the peak value is, the bigger the difference is, and the blue part shows GC content is lower than the mean GC content of genome; The innermost circle shows the GC-skew, with dark gray representing regions where G is greater than C and red representing regions where C is greater than G. Circos (https ://circo s.ca/) was used for drawing program.
Scientific Reports | (2020) 10:18065 | https://doi.org/10.1038/s41598-020-75235-x www.nature.com/scientificreports/  www.nature.com/scientificreports/ Comparative genome. The evolutionary relationship between Sc. and analogous species was studied using BLAST searches against the Nr database ( Supplementary Fig. S4). As shown in Supplementary Fig. S4, Sc. had higher genomic similarity with Phaeosphaeria nodorum (18.56%) and Leptosphaeria maculans (18.39%), while only Leptosphaeria maculans (Lm.) had genome datas (FA format) and genome annotation files (GFF format) in the public database. Considering the results of the analysis of the public database information reported so far, although the similarity of only 18% between Sc. and Lm., the similar sequences still provide data support for genome annotation and further functional gene research of Sc. In terms of their relatedness based on homology (similarities of nucleotide composition and gene sequences) and their evolutionary relatedness (based on distances between genes and gene positioning patterns of biological systems), the results of analysis of both types of comparisons between the two plant pathogenic fungi are shown in Table 1.
Although genome similarity between Sc. and Lm. was relatively high, some differences were observed: the genome of Sc. was smaller (35.28 Mb) than that of Lm. (45.12 Mb) and had lower GC content (44.1%) than did that of Lm. (50.2%); The numbers of genes encoding proteins of Sc. and Lm. were 9,844 and 12,469, respectively; the respective average gene lengths of Sc. and Lm. were 1728 bp and 1323 bp, with N50 values of 1.80 Mb and 1.57 Mb, respectively. Notably, the most obvious difference between Sc. and Lm. was due to the larger number of scaffolds and greater proportion of repeated sequences for Lm. which were 2.3 times and 3.56 times higher than respective Sc. values.
BLAST comparisons of protein sequences of Sc. and Lm. genomes revealed a collinear relationship at the nucleic acid level that reflected differences between species in relative positions of analogous genes within their genomes (Fig. 4). In general, comparative genetic studies show that genes and sequences of many fungal genomes are highly conserved at the genetic map level, but the level of collinearity is less conserved at the micro mapping level (with the exception of very closely related species that maintain good collinearity). In addition, MCScanXbased analysis was also used to conduct collinearity analysis of amino acid sequences of both genomes and demonstrated a similar linear relationship between Sc. and Lm. genes. Here, the collinearity analysis was performed on the genome, for strains Lm. and Sc. homologous proteins, and the collinearity fragments were obtained with the help of the relative position of genes in the chromosome. All these fragments provide support for the detection of genome-wide gaps in the coding genes for unknown functional protein, which will be one of the continuing research focuses of this research.
Differentially expressed gene analysis. After sequencing and quality control, a total of 44.62 Gb clean data were obtained, with clean data per library approaching 7.15 Gb, and the Q30 base percentage were greater than 92.38% (Supplementary Table S3). During the detection of differentially expressed genes (DEGs), Fold Change ≥ 2 and FDR < 0.01 were used as the screening criteria (Table 2), 319 DEGs between Sc.1 (the average value of three replications of T01, T02 and T03) and Sc.2 (the average value of three replications of T04, T05 and T06) were obtained, in which 190 and 129 genes were down-and up-regulated ( Supplementary Fig. S5). 58 DEGs were identified as pathogenicity genes involved in multiple molecular functions within biological processes, such as hydrolytic enzyme, protein kinase, phosphatase, cellulase, etc. Among them, 22 DEGs were associated with transmembrane transport (GO:0055085, GO:0003333 and GO:0005215) and 16 DEGs were associated with hydrolytic enzyme (GO:0016787, GO:0004553 and GO:0016810). In addition, 20 genes may be directly or indirectly involved in growth and pathogenicity of Sc. (Supplementary Table S4), since they belong to biological processes of cellulase activity (GO:0008810), phosphorelay sensor kinase activity (GO:0000155), proteolysis (GO:0006508) and interaction with host via protein secreted by type II secretion system (GO:0052051). Hierarchical cluster analysis of the abovementioned 58 DEGs based on log10 (RPKM) values showed that the overall gene expression pattern could be divided into several clusters based on DEGs expression levels for T01, T02, T04, and T05; this result demonstrates obvious differences in DEGs between Sc.1 and Sc.2, a result consistent with Pearson correlation coefficient r results and comparison diagram analysis of the FPKM density distribution ( Supplementary Fig. S6). The most significant KEGG pathways associated with DEGs are shown in Supplementary Fig. S7; DEGs between Sc.1 and Sc.2 were assigned to 7 KEGG pathways.

Secreted proteins and effector factors.
Annotation of the Sc. genome yielded a total of 7238 proteins based on Pfam database matches. Domain calling analysis revealed that more than 2738 domain types exist in the Sc. proteome, which include effectors secreted by pathogens that may play critical roles in plant suppression of pathogen proliferation 27 . Among these effectors, 1024 signaling peptide proteins were identified. After removal of transmembrane helix proteins from this list, 759 secreted proteins of the Sc. Genome were identified. Of these proteins, secreted protein functional domains with the highest frequency of representation were counted to reveal that the largest number of structural domains belonged to glycosyl hydrolase family 61 (PF03443.9), followed by zinc carboxypeptidase (PF00246. 19), the WSC domain (PF01822.14), the subtilase family (PF00082.17) and the kelch motif (PF13854.1). Functional domains with lowest frequency of representation included chitin recognition protein (PF00187.14) and fungal cellulose-binding domain (PF00734.13) (Supplementary Table S5). Ultimately, 54 predicted secretory proteins were subsequently shown to be differentially expressed ( Fig. 5 and Supplementary Table S6).

Analysis of PHI related genes and cell degrading enzymes. PHI is an important database repository
of pathogen-host interaction factors and contains a large number of important pathogenicity genes that are useful for analyzing the pathogenic potential of pathogen virulence factors 28 . After comparing the entire genome of Sc. to the PHI database, 2869 genes with high homology were detected, accounting for 29.14% of the total number of genes in the Sc. genome, of which 96 genes were differentially expressed (Fig. 5  The assembled genome is shown on the left and the reference genome on the right. Analogous genes between species are joined with a line. Thus, the homology of coding sequence and structure between modular genomes is established, whereby mapping information of known genomes is used to locate genes within other genomes that, in turn, reveals potential functions of genes, evolutionary relationships between species and internal structures of genomes. MCScanX (https ://chibb a.pgml.uga.edu/mcsca n2/MCSca nX.zip) was used for drawing program. www.nature.com/scientificreports/ Notably, four classes of DEG annotations were identified in PHI that included secretome, putative carbohydrate-active enzymes (CAZymes) and secondary metabolic processes, all of which indicate that successful phytopathogenic fungi can break down and utilize plant cell wall polysaccharides by secreting CAZymes. Such plant cell wall-degrading enzymes are especially relevant to fungal pathogenicity, due to their important roles in penetration and successful infection of plant hosts. In this study, 36 DEGs of the total 605 genes encoding carbohydrate enzymes were predicted to have pathogenic relevance based on matches with CAZyme database proteins ( Fig. 5 and Supplementary Table S8).
In the Sc. genome, approximately 37.69% of potential secreted proteins (278/605) were predicted to be CAZymes, demonstrating that Sc. may employ a large group of CAZymes to digest host cell walls during invasion. Interestingly, one family of CAZymes, the glycoside hydrolase GH109 family, was highly enriched (Supplementary Table S9); one member of this family is known to biochemically function as an α-N-acetylgalactosaminidase (α-NAGAL), which can cleave the bond linking terminal alpha-linked N-acetylgalactosamine group to the A blood group antigen protein 29 . Notably, we found that Sc. encodes 10 putative GH109 genes, a higher number than observed for most fungi, including Magnaporthe oryzae, Botrytis cinerea and Neurospora crassa 30 . Thus, these data suggest that expansion of the GH109 family in Sc. may have endowed the pathogen with the ability to overcome pumpkin lectin-mediated resistance to infection.
The GH18 family, the third largest group within the glycoside hydrolyase (GH) family, contains many fungal chitosan enzymes that function along with other cell wall-degrading enzymes to ultimately transform and recycle the fungal cell wall 31 . Ten potential GH18 family coding genes from the Sc. genome were identified and appear to code for secreted proteins, due to their homology with pathogenicity or virulence genes PHI 144 and PHI 1178 (reduced virulence) within the PHI database (Supplementary Table S10). As carbohydrate enzymes have been found among many important enzyme families, analysis of protein families may provide insights regarding the regulation of expression of carbohydrate enzyme-coding genes of pathogenic fungi.
Of annotated carbohydrate esterase genes described in this work, we found a total of 74 genes belonging to two major subfamilies, CE1 and CE10. Both families encode proteins exhibit activities characteristic of carboxylesterases and endo-1,4-β-xylanases. Meanwhile, although 47 potential carbohydrate-binding module (CBM) proteins that digest complex carbohydrates extracellularly 32 have been identified in fungal genomes, unexpectedly only one CBM protein, CBM21, was found in the Sc. genome. Nevertheless, this observation may be relevant to pathogenicity, since many phytopathogenic fungi generate abundant quantities of CBM21 32 .
Secondary metabolic enzyme. During interactions between fungi and host plants, many secondary metabolites, such as fungal toxins, have been shown to participate in and enhance the virulence of pathogenic fungi. Biosynthesis of fungal secondary metabolites mainly relies on five key enzymes: polyketone synthase (PKS), non-ribosomal peptide synthase (NRPS), non-ribosomal peptide synthase-polyketide synthase hybrid (hybrid NRPS-PKS), dimethyl tryptophan pyrophosphate synthetase (DMATS) and terpene cyclase (TC). Here a total of 3327 Sc. genes related to functional pathways within metabolic process (GO: 0008152) were found. Of these genes, 86 metabolically relevant genes were found to be differentially expressed (40 up-regulated and 46  Tables S11, S12). This finding is relevant to pathogenicity, since pigments are an important group of secondary metabolites that protect pathogens from host oxidative stress damage during invasion. Notably, the Sc. genome encodes multiple putative PKSs that likely participate in pigment production ( Fig. 6 and Supplementary Table S12).
Transporters. Amino acid sequence BLAST comparisons conducted between Sc. sequences and the TCDB database predicted a total of 130 transport-related genes, of which the largest number (45 genes) were related to transporters (one-way transporter, cotransporter, reverse transporter) and 44 genes were related to P-P-bondhydrolysis-driven transmembrane transporters ( Supplementary Fig. S8). These genes together constitute the main transporter system of Sc., of which eight genes exhibit differential expression before versus after infection and were designated as contig00003.94, contig00009.78, contig00009.120, contig00009.376, contig00013.104, contig00013.149, contig00015.279 and contig00016.184.
A total of 62 genes annotated in PHI database were found to be associated with a total of 44 PHI-base accessions (Supplementary Table S13) and 3 ABC transporter gene proteins (contig00001.338, contig00004.179 and contig00011.394). ABC transporters play an important role in the processing of antifungal toxins and in cytotoxin transport, which are processes that are needed to overcome the toxic effects of anti-cytotoxins produced by the host during fungal invasion 33 . Notably, transcriptome analysis showed that contig00011.93 contained an up-regulated DEG that was involved in ATP-binding cassette metabolism in infected leaves. Thus, transporter genes may be important pathogenic factors involved in Sc.-plant interactions.
Expression of predicted pathogenic genes. Eight genes were randomly selected from the total 136 predicted pathogenic genes for assessment of gene expression levels (Supplementary Tables S6, S7, S8). Primers used in these experiments are listed in Supplementary Table S14. In order to study the functions of predicted genes, we transiently expressed the eight genes in Nicotiana tabacum then screened plants to detect potential effectors that suppressed BAX-triggered PCD (BT-PCD). Our results showed that the eight genes suppressed BT-PCD. As for the cracks after Contig00006.409 and Contig00007.120 produced pathogenicity, this uncontrollable factor may have a slight influence on the results, but it will not affect the overall evaluation (Fig. 7A). This result aligns with qRT-PCR results showing that expression levels of these seven genes increased and one decreased after 5 days of infection of pumpkin leaves (Fig. 7B). Therefore, the results obtained in the field support our predicted results.

Discussion
Whole genome sequencing and analysis. Genomics belongs to the scientific branch of applied bioinformatics and requires high quality nucleotide sequences for successful results. Currently, fungal DNA isolation methods mainly include mycelial and spore DNA extraction methods 34 . Due to the large amount of GSB fungal mycelial material obtained from PDA culture, mycelial enrichment was carried out by scraping material directly from the surface of the solid PDA medium followed by grinding of the liquid nitrogen-frozen fungal material and associated agar into a powder. The powder was then subjected to DNA extraction, resulting in isolation of DNA of high purity and integrity for this study. www.nature.com/scientificreports/ Genome-wide analysis of plant pathogenic fungi is helpful for understanding evolutionary relationships between pathogens, for detecting disease-related genes and for determining mechanisms that underlie observed variations in virulence, all of which are relevant to controlling fungal diseases of plants 35 . Here, comparisons of sequences within the Sc. genome to known genes within multiple databases resulted in prediction of 9844 protein-coding genes, including 2869 pathogen-host interaction-related genes, 605 carbohydrate synthase-coding genes and 130 transporter-coding genes. By studying variations of such genes and how they correspond to variable levels of pathogenicity, clues to interactions between pathogens and hosts may be found. For example, if the virulence of GSB is low only toward a single specific variety of seed pumpkin (but not to other varieties), identification of gene variations in that pumpkin variety could reveal an unknown but important participating host factor that could serve as a gene target for use in development of GSB-resistant seed pumpkin varieties.
In another example, proteins secreted by pathogens often must interact with key plant effectors in order to cause pathogenic effects in plants; therefore, variations in plant effector genes may sometimes undermine plant www.nature.com/scientificreports/ resistance to pathogens 36,37 . Indeed, most transporter protein families that participate in pathogen-host interactions are likely clues to understanding pathogenic mechanisms. However, based on this dialectical view, the use of bioinformatics tools to deduce functions of predicted genes may also have limitations. For example, although pathogenic genes for Sc. interaction with host were identified and classified based on results of BLASTp searches of the PHI database, the expression of several analogous genes varied among different pathogenic organisms; thus these genes require further verification to confirm disease specificity or selectivity. In addition, many identified Sc. gene-encoded proteins, such as protein kinases, ABC transporters and major facilitator superfamily (MFS) transporters, are frequently directly or indirectly involved in these processes. Therefore, although some pathogenically relevant protein gene functions have been identified using bioinformatics methods, others genes require further functional verification to confirm their functions. DNA methylation is an important field of epigenetics in eukaryotes and prokaryotes. Of the three most studied DNA methylation types observed in fungi, m5C, m4C and m6A 38 , only two types of DNA methylation (30,833 sites for 6-mA and 1,228,069 sites for 4-mC DNA methylation) were predicted in the Sc. genome using SMRT software. Generally, m6A is considered to be an epigenetic signal involved in DNA-protein interactions, while m4C is considered to be a modification present at lower frequency within regions of repeating elements. In this study, 100 times more m4C sites were observed than m6A sites, indicating that m4C may be generated during transposition of Sc. transposons. Nevertheless, SMRT can directly detect DNA methylation through information provided by polymerase kinetics, analysis of GSB DNA methylation confirms that SMRT is a powerful tool for studying epigenetic modifications of fungal genomes.
At present, the absence of genomic studies for Sc. has hindered development of reference materials for the study of GSB in real time. However, high-throughput sequencing platforms have made it possible to obtain and analyze whole-genome sequences, especially since third-generation sequencing technologies that are now available are very powerful tools for genomics research and sequence analysis 39 . In this work, the whole-genome sequence was obtained for a cold-adapted strain of Sc. using second-and third-generation high-throughput sequencing methods. After data quality control, genome assembly, genome component analysis and DNA methylation analysis were completed, functional annotation ultimately identified numerous possible pathogenic genes that code for cell wall degradation enzymes, protein transport system proteins and pathogen-host interactions. These Sc. genome sequence results will facilitate future research efforts to understand GSB pathogenic mechanisms.
Comparative genomics analysis. In this study, BLAST alignments of genome sequences of various related species resulted in identification of Lm. as the closest relative of Sc., with genomic collinearity between the two species aiding discovery of pathogenic and functional genes. Indeed, many fungal genomes contain highly conserved genes and gene sequences that exhibit good collinearity among closely related species (that are separated by small evolutionary distances).  30 . These differences may explain why the two strains, although they are closely related, are different and illustrate why it is important to conduct genetic similarity analysis of scaffolds.
At the start of this study, none of the genomes of S. cucurbitacearum, Stagonosporopsis citrulli or Stagonosporopsis caricae, all of which cause GSB, had been released. However, due to the fact that the genome sequence for Lm. was available and that Lm. is highly similar to Sc., the Lm. genome was selected for comparative genome-wide analysis against the Sc. genome 40 . After the analysis, genomes of Sc. and Lm. were found to be relatively similar in length (35.28 Mb vs. 45.12 Mb, respectively) with similar numbers of protein-coding genes (9844 vs. 12,469), fueling speculation that they are closely related and probably share common past evolutionary rearrangement events 41 . Moreover, subsequent MCScanX-based analysis revealed a highly conserved collinearity relationship between Sc. and Lm. Thus, the large number of conserved sequences and highly conserved genome structures between these organisms suggest that the Lm. genome sequence will be highly useful for completing gap repair of the Sc. genome now that numerous Sc. genes have been mapped and structures and functions of numerous genes have been characterized 42,43 . Furthermore, additional results have confirmed relatively high genomic correlation between Sc. and Lm., further supporting a high degree of relatedness based on both sequence homology and evolutionary history 44 . Here, comparative genomics was conducted to comprehensively analyze the relationship between Sc. and other related species from a whole-genome perspective, groupings of genes with similar functions and characteristics were generated based on homologies to known genes, thus demonstrating that gene functions can be predicted. In addition, these results provide valuable insights for the development of methods to achieve early and rapid diagnosis of GSB and for development of strategies to combat GSB.
Pathogenic gene analysis. About  www.nature.com/scientificreports/ understanding of pathogenic fungal mechanisms from a molecular perspective. In this way, accurate prevention and control of pathogenic fungi may be achieved through development of target drugs that induce disease resistance in the future. The plant cell wall is the main physical barrier between plant and pathogen during the infection by plant pathogenic fungi, which invade host tissues by producing a series of cell wall-degrading enzymes (CWDEs) 47 .
Here, CAZyme was used as a tool to predict approximately 278 putative secreted proteins in the Sc. genome that notably include cell wall-degrading enzymes with relevance to GSB. In the future, prediction of proteins involved in Sc. infection structures involved in fungal invasion of pumpkins will also be conducted.
Using RNA-seq technology, scientists have discovered that many secreted proteins, secondary metabolic enzymes and receptor protein-coding genes play important roles in host-pathogen interactions. Here, we found that the disease-causing genes predicted from the special function database were not corresponding with the transcriptional analysis, this was attributed to the fact that transcriptional expression analysis was only the differential expression of genes before and after the host infection, while the expression of pathogenic genes predicted by the database is not significantly different during the treatment period. In addition, the Venn diagram was drew from the three databases and eight infection-related genes that were not significantly differentially expressed at transcriptome level during this treatment period.
In this study, 22 DEGs were detected using GO analysis that was related to transmembrane transport. Although these genes appear to be related to Sc. toxin secretion and drug resistance, their specific functions require further verification. One group of transmembrane transport proteins that contains the largest number of efflux pumps, the ABC family of transporters, are primary active transport systems that play an important role in antifungal toxin and cytotoxin transport. During plant host invasion, pathogenic fungi rely on these transporters to counter toxic effects of host-generated anti-cytotoxins 48 . Interestingly, one DEG in this group found here (Contig 00011.93), was an ATP-binding cassette transporter (a type of ABC transporter) that was up-regulated in infected leaves. In addition to transmembrane transport proteins, hydrolytic enzymes of pathogenic fungi were also detected in Sc. This is an interesting result, since hydrolytic proteins have important pathogenic roles. Such roles have been demonstrated in studies showing that fungal pathogenicity can be reduced by inhibiting expression of a hydrolytic enzyme gene 48 , possibly because hydrolytic enzymes can participate in the release of fungal defense substances that facilitate pathogen host invasion 49 . In this study, 16 Sc. genes were found to be related to hydrolytic enzymes, which may participate in pathogen infection and colonization of host tissues.
As an important means used by pathogenic fungi for absorption of foreign macromolecular substances and nutrients, intracellular endocytosis is a highly conserved process that is instrumental for fungal growth and pathogenic processes 50 , as demonstrated previously in M. oryzae 51 and U. maydis 52 . In this study, it was predicted that genes within Contig00011.26 may play an important role in the process of endocytosis. Therefore, the study of this pathway will further our understanding of pathogenic mechanisms of Sc.
Protease is considered to be an important virulence factor of pathogenic fungi involving in the process of adhesion, colonization, dissemination and escape of the host during the immune response. In this study, two upregulated DEGs related to proteases were predicted, namely Contig00012.91 and Contig00022.5. Contig00012.91 was coded Alkaline protease 1 in Swiss_Prot_database with multiple GO function terms, including serine-type endopeptidase activity (GO:0004252), proteolysis (GO:0006508) and identical protein binding (GO:0042802). Contig00022.5 was coded Alpha-lytic protease prodomain with in the process of negative regulation of catalytic activity (GO:0043086), serine-type endopeptidase activity (GO:0004252), extracellular region (GO:0005576) and proteolysis (GO:0006508).

Conclusions
GSB, a worldwide disease, is easy to infect melon crops. In recent years, GSB has seriously affected pumpkin production in northeast China, where the seed used pumpkin was one of the most important industrial crops, causing huge losses to agricultural producers. Although appropriate cultivation measures can alleviate the disease to some extent, they cannot cure the disease accurately. This is the first report to carry out high-throughput whole-genome sequencing and genome assembly of Stagonosporopsis cucurbitacearum (Sc.), which was the main pathogenic fungus of pumpkin seedling GSB in northeast China, revealing the pathogenic molecular regulatory mechanism involved in the growth and development of this fungus. Meanwhile, transcriptional analysis of Sc. before and after infecting with host was conducted, and differentially expressed disease-related genes and metabolic pathway were explored. This study laid a foundation for the further development of more effective application of pathogenic genes in the development of green prevention and control of new fungicides.

Materials and methods
Sampling and isolation of fungi. Sc. strain zq-1, obtained from a pumpkin field in northeast China in August 2016, was isolated by Zhao and colleagues as previously described 53 . Briefly, a white colony with limited mycelial expansion was cultured to produce a large colony with irregular edges containing numerous mycelial groupings with mycelial diameters ranging from 15 ~ 20 µm × 5 ~ 6.3 µm, as consistent with known colony morphology of Sc. Further confirmation of the identity of the fungus as Stagonosporopsis cucurbitacearum was obtained using ITS sequence comparison analysis 53 . PDA solid culture medium was used for strain cultivation. After incubation for 72 h at 28 °C, thalli were scraped from surfaces of agar surfaces and placed in EP tubes without medium and frozen at − 80 °C. Approximately 1 g of Sc. thalli (maintained in the frozen state using liquid nitrogen) was ground into a powder that was then subjected to total DNA extraction using a modified CTAB method, as previously described 54 . The ultimate yield of thallus nucleic acid was approximately 7 μg.  55 . DNA libraries with 270-bp or 10-kb inserts were constructed. The 270-bp library was constructed following Illumina's standard protocol, including fragmentation of genomic DNA, end repair, adaptor ligation and PCR amplification. The 270-bp library was quantified using 2100 Bioanalyzer (Agilent, USA) and subjected to paired-ended 150-bp sequencing using Illumina HiSeq 4000. Sequencing data (filtered reads: 3.21G, sequencing depth: 91×) were used to estimate genome size, repeat content and heterozygosity. Next, the 10-kb library was constructed following PacBio's standard methods, including fragmentation of genomic DNA, end repair, adaptor ligation and template purification. The 10-kb library was quantified using a 2100 Bioanalyzer (Agilent, USA) and sequenced using SMRT followed by filtering of sequence data (filtered reads: 5.24G, sequencing depth: 148×). The resulting set of high-quality sequences was assembled using Canu v.1.2 using default parameter settings 56 . Finally, Illumina reads were subjected to error correction and gap filling using SOAPdenovo2 with GAPCLOSER v1.12 57 .
Genome annotation. Protein-encoding genes were annotated using a combination of three independent ab initio gene-finding tools: Augustus, GlimmerHMM and SNAP. Next, GeMoMa 58 was used to make gene predictions based on gene sequences of analogous species (Ascochyta rabiei and Lm. Genome comparison. Gene prediction and annotation results for the Sc. zq-1 genome were used to determine the distribution of identified fungal genes in selected eukaryotic microbial species. Next, matching genes were counted and searched against the Nr database to generate a distribution map of sequences among genomes of analyzed species. According to the gene distribution map, a large number of Sc. genes aligning to Lm. sequences were observed, prompting comparison of the whole genomes of both organisms using BLAST. In addition, MCScanX 64 software was used to conduct collinearity analysis of amino acid sequences of both genomes (using default program parameters) for prediction of genome sequence differences and structural changes that have occurred during evolutionary processes to further clarify the degree of genomic similarity between these two eukaryotic microorganisms.
Transcriptome analysis. DESeq was used to obtain two differentially expressed gene sets, Sc.1 (Sc. zq-1 grown on PDA medium) and Sc.2 (Sc. zq-1 grown on pumpkin leaves). Sc.1 was stored in a refrigerator at -20 °C with PE tube slant culture medium. After activation at room temperature, mycelia were transferred to a new plate (20 mL PDA medium) with inoculation ring and continued to grow in a 28 °C light incubator. While Sc.2 inoculated plant host material "Yinhui 2", is a typical seed pumpkin variety. All samples were immediately frozen and stored in liquid nitrogen until RNA extraction. Fungal samples grown on either PDA or leaves were collected in triplicate to generate a total of six cDNA libraries. To make each library, 3 μg RNA was used to make cDNA using an NEBNextUltraRNA Library Prep Kit for Illumina (NEB, MA, USA) following the manufacturer's recommendations. Index codes were added to link sequences to their corresponding starting samples. Clustering of index-coded samples was performed using a cBot Cluster Generation System employing a TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer's instructions. After cluster generation, library preparations were sequenced on an Illumina Hiseq 2000 platform and paired-end reads were generated. Raw data (raw reads) in FASTQ format were first processed using in-house Perl scripts and Q20, Q30, GC content, and the level of sequence duplication within the clean data were calculated. R package (DESeq 1.10.1, https ://www.bioco nduct or.org/packa ges/relea se/bioc/html/DESeq .html) was used for differential expression analysis using the threshold of P < 0.05 65 .
Analysis of the genome characteristic. Due to the relatively low conservation of repeat sequences among species compared here, a pipeline combining four software tools, LTR FINDER 66 , MITE-Hunter 67 , RepeatScout 68 and PILER-DF 69 , was used to construct a repeat sequence database of genomic sequence data based on principles of structure prediction and ab initio prediction. The database was classified using PASTEClassifier 70 then was merged with the Repbase database 71 to construct the final repeat sequence database. Next, RepeatMasker 72 software was used to predict repeat sequences from sequencing data. Non-coding RNAs, including various RNAs with known functions (such as rRNA and tRNA), were predicted based on structural characteristics, with dif-Scientific Reports | (2020) 10:18065 | https://doi.org/10.1038/s41598-020-75235-x www.nature.com/scientificreports/ ferent strategies adopted to predict different types of non-coding RNAs, including genome-wide comparisons based on BLASTn search results 73 , identification of rRNA based on matches to Rfam database sequences 74 and identification of t-RNA sequences using tRNAscan-SE 75 software. Meanwhile, pseudogenes (non-functional genes) were predicted using gene sequence comparison tool GeneWise 76 to detect immature termination codons and code shift mutations within gene sequences. SignalP v4.0 77 and TMHMM 78 were used to analyze the protein sequences of all the predicted genes and identify the signal peptide proteins and transmembrane proteins, respectively. And secreted proteins were predicted by remove the transmembrane proteins from signal peptide proteins.
Functional study of predicted pathogenic genes. All products generated via polymerase chain reaction (PCR) amplification using high-fidelity enzyme were digested using suitable restriction enzymes then were ligated to corresponding restriction enzyme-digested vector cloning sites. Recombinant plasmids were then transformed into Escherichia coli for plasmid micropropagation. Bacterial colonies containing plasmids with suitable inserts were verified via PCR then were isolated and transformed into Agrobacterium tumefaciens. The apoptosis inducer BAX (which can induce a hypersensitive necrosis reaction in tobacco plants) and a characteristic pathogen-associated molecular pattern (PAMP) protein of Phytophthora oocytes (infestin-1 or INF1) served as positive controls and GFP as negative control. An agrobacterium-mediated viral expression system was used to instantaneously express effector factors in tobacco plants 30 . Effector factors directly causing plant responses were screened out to allow for detection of candidate fungal effector functions.
qRT-PCR validation. Differentially expressed genes related to fungal pathogenesis were randomly selected for qRT-PCR analysis. For RNA preparation, Sc. samples were collected 5 days post-inoculation from fungal cultures growing on PDA medium or on pumpkin leaves then total RNA was prepared using a PrimeScript RT reagent Kit with gDNA Eraser (Perfect Real Time) (Takara, Otsu, Japan). 1 μg of total RNA as template was reverse transcribed to generate cDNA using instructions provided with the kit. Next, qRT-PCR was performed for biological samples in triplicate using an ABI PRISM 7500 fast real-time PCR system (Applied Biosystems, Foster City, CA, USA) and thermal cycling conditions as follows: 1 cycle of initial denaturation at 95 °C for 5 min followed by 35 cycles of: denaturation at 95 °C for 15 s, annealing at 60 °C for 15 s and extension at 72 °C for 15 s. Finally, the t-test was used to conduct difference analysis of data to assess if expression level differences were statistically significant based on the threshold p < 0.05. Primers used in qRT-PCR for DEGs validation are shown in Supplementary Table S13 .
Ethical approval. The authors note that this research was performed and reported in accordance with ethical standards of scientific conduct.