De novo transcriptomic assembly and mRNA expression patterns of Botryosphaeria dothidea infection with mycoviruses chrysovirus 1 (BdCV1) and partitivirus 1 (BdPV1)

Three B. dothidea strains, LW-1, LW-C, and LW-P, infected with hypovirulent and non-hypovirulent mycovirus were generated [5, 16]. LW-1 (or designated as LW-CP) was infected with Botryosphaeria dothidea chrysovirus 1 (BdCV1) and Botryosphaeria dothidea partitivirus 1 (BdPV1), which was isolated and identified from the trunks of sand pear‘Jinshuiyihao’ cv., collected in Wuhan city, Hubei province of China [5]. The LW-P strain infection with BdPV1 and LW-C with BdCV1 were derived from LW-1 by single hyphae isolation, respectively, as previously described [5, 16]. Meanwhile, mock derived from LW-C strain by protoplast isolation was virus-free as negative control. The fungal strains were grown at 25 °C in darkness on potato dextrose agar (PDA) medium and stored on PDA plate at − 4 °C or in glycerin at − 80 °C.

The mycelia of the mycovirus-infected three strains and the virus-free strain of B. dothidea were harvested on 5 day of PDA culture. The mycelia were frozen and ground using liquid nitrogen for nucleic acid extraction. Total RNAs were isolated and obtained with TRIzol Reagent (Invitrogen, USA), phenol and chloroform to remove proteins and ethanol precipitation according to manufacturer’s instructions. Total RNAs digested by DNase I (Takara, Dalian, China) were precipitated with ethanol and dissolved in DEPC treated water to be used as template for Next Genomics Sequencing (NGS). The RNA integrity Number (RIN > 7) and concentration of obtained tRNAs were further assessed using the Bioanalyzer Agilent 2100 and NanoDrop spectrophotometer (Agilent, USA). High quality samples were used to construct the four de novo libraries.

For each RNA-seq library, total RNA from three biological repetitions were pooled. Following the total RNA extraction and treatment with DNase I, mRNA was isolated and enriched by Oligo (dT). mRNA was cut into small fragments and cDNA synthesis was performed with random hexamer primers. PCR amplification was then prepared for library construction according to the Illumina RNA library protocol for transcriptome sequencing. Then, the quantification of the four cDNA libraries was verified by the Agilent 2100 Bioanalyzer and ABI StepOnePlus Real-Time PCR System. Finally, the libraries were sequenced using the HiSeq 4000 instrument (Illumina, USA) at the Beijing Genomics Institute (BGI) (Shenzhen, Guangdong province, China).

Internal software was used to filter raw reads to generate clean reads data via the following processes: 1) The removal of adaptor-polluted reads, 2) removal of reads containing a high content (> 5%) of unknown bases (N), 3) removal of low-quality reads. Following filtering, the remaining reads were used for downstream analyses. Trinity was used to perform de novo assembly with clean reads which PCR duplication removed, and Tgicl was used to cluster transcripts to Unigenes. The sequence by Trinity was called transcripts. Then, gene family clustering with Tgicl for each sample’s Unigene to obtain final Unigenes was performed for downstream analyses. The Unigenes were divided into two classes, including clusters and singletons. The prefix used for clusters was CL with the cluster id behind it (in one cluster, there are several Unigenes with more than 70% similarity between them). The prefix used for singletons was Unigene. For unigene functional annotation analysis, NT (ftp://​ftp.​ncbi.​nlm.​nih.​gov/​blast/​db), NR (ftp://​ftp.​ncbi.​nlm.​nih.​gov/​blast/​db), GO (http://​geneontology.​org), COG (http://​www.​ncbi.​nlm.​nih.​gov/​COG), KEGG (http://​www.​genome.​jp/​kegg), SwissProt (http://​ftp.​ebi.​ac.​uk/​pub/​databases/​swissprot), and InterPro (http://​www.​ebi.​ac.​uk/​interpro) were used as functional databases. Blasts aligned unigenes to NT (NCBI non-redundant nucleotide sequences), NR (NCBI non-redundant protein sequences), COG (Cluster of Orthologous Groups), KEGG (Kyoto Encyclopedia of Genes and Genome), and SwissProt to obtain the annotation. Blast2GO with NR annotation to obtain the GO (Gene Ontology) annotation, and InterProScan5 was used to get the InterPro annotation [28‐30]. The functional databases with a priority order of NR, SwissProt, KEGG, and COG were selected to determine the sequence annotation of unigenes. Coding sequences (CDS) were extracted from sequences of unigenes that best mapped to functional databases. ESTScan was used to predict coding regions of unigenes if no hits were obtained in any database mentioned above [31].

De novo sequencing from four constructed cDNA libraries designated as LW-CP, LW-C, LW-P, and Mock was performed. To analyze the differentially expressed genes (DEGs) between two samples in response to mycovirus, clean reads to assembled unigenes using Bowtie2 were mapped [32]. Then, gene expression value was calculated with RSEM [33]. DEGs were identified with PossionDis based on the Poisson distribution, which was performed as described by Audic et al. [34]. The parameters of PossionDis was set as |Fold Change | > 2.0-fold (|log₂FC| > 1) with a false discovery rate (FDR) < 0.001.

With DEGs, pheatmaps of hierarchical clustering were performed using the heatmap.2 package in R software [35]. For clustering more than two groups, we used the intersecting DEGs.

RT-qPCR was used to validate DEG expression levels obtained from sequencing and to assess the expression levels of cp and RdRp from BdCV1 and BdPV1. Primers used are listed in Additional file 8. Total RNA was extracted from the four groups: Mock, LW-CP, LW-C, and LW-P cultured for 5 d, 10 d and 15 d using TRIzol (Invitrogen, USA) according to manufacturer’s instructions. Total RNA was digested by DNase I and used as template (Takara, Dalian, China). The cDNA was obtained to perform reverse transcription using PrimeScript™ RT reagent Kit with genomic DNA Eraser (Takara, Dalian, China). According to the manufacturer’s instruction, qPCR was performed using 2.5 μl diluted cDNA, 10 μl of the SYBR Premix Ex Taq II PCR mixture (Tli RNaseH Plus) (Takara, Dalian, China), 1 μl of each 5 mM primer, and deionized water to a final volume 20 μl. All reactions were run in triplicate. The qPCR was performed on the Bio-Rad iQTM5 Real-time System machine (BIO-RAD, USA). The products were verified by melt curve analysis. The 18 s gene from B.dothidea strains was used as the internal reference for standardization of cDNA expression levels from samples. mRNA relative expression level changes were quantified by a comparative C_T method (ΔΔC_T) using the formula 2^-ΔΔCt [36].

To determine the culture time point to analyze the B. dothidea transcriptome in response to mycovirus infection, the relative expression levels of cp and RdRp mRNA from BdCV1 and BdPV1 in LW-1, LW-C, and LW-P cultured for 5 d, 10 d, and 15 d were quantified by RT-qPCR analysis. The results demonstrate that the relative expression levels of cp and RdRp mRNA from BdCV1 and BdPV1 are expressed differentially in B. dothidea strains LW-C, LW-P, and LW-1 infected with single and mixed mycovirus. In general, BdCV1 viral accumulation increased, while BdPV1 viral accumulation decreased slowly in LW-1 strain cultured periods from 5 to 15 d (Fig. 1).

The mycelia of the mycovirus-infected three strains and the virus-free strain of B. dothidea at 5 d grown on PDA were collected for RNA extraction to be further used for De novo sequencing. The B. dothidea transcriptome was sequenced and assembled de novo. This is a valuable resource since there are no current genomic data available for B. dothidea except two B.dothidea strains reported [26, 27]. Four libraries designated as LW-CP, LW-C, LW-P, and Mock were constructed to conduct the comprehensive genome-wide transcriptome analysis following Illumina Hiseq 4000 platform sequencing. In total, clean reads [47.65 Mb (98.4% of raw reads)] corresponding of 4.77 Gb bases from LW-C, 4.67 Gb bases from LW-CP [46.71 Mb (98.38%)], 4.75 Gb bases from LW-P [47.53 Mb (98.24%)], and 4.68 Gb bases from Mock [46.84 Mb (98.16%)] were obtained after trimming, and then assembled using de novo assembly program of Trinity (Additional file 1: Table S1). The assembly of the clean reads resulted in 26,461 transcripts for LW-C, 30,707 for LW-CP, 25,808 for LW-P, and 25,298 for Mock. Tgicl was used to cluster transcripts to unigenes. The clustering quality metrics are shown, including of 23,056 unigenes for LW-C, 27,554 for LW-CP, 22,403 for LW-P, and 22,068 for Mock, which reveals the effect of mycovirus infection on a number of gene expression at the transcriptome level. In total, 30,058 unigenes with an average length of 2128 bp and N50 length of 3338 bp were obtained from 4 libraries of B. dothidea strains (Additional file 2: Table S2). The size distribution of these unigenes ranged from 300 bp with 3889 unigenes to higher than 3000 bp with 7593 unigenes as shown in Fig. 2. To validate the unigene sequence by de novo assembly, RT-PCR and sequencing were used to measure seven unigenes selected randomly. Results demonstrated that they were more than 98% identity comparable to de novo sequencing (Additional file 3: Table S3). This is the first report of B. dothidea transcriptome.

Thirty thousand fifty eight non-redundant sequences were submitted to a BLASTx search and annotated to classify function from seven public functional databases including NR, NT, GO, COG, KEGG, Swissprot, and Interpro. In total, the number of unigenes annotated to least one functional database was 24,836 (82.63%). 24,024 (NR: 79.93%), 13,501 (NT: 44.92%), 18,859 (Swissprot: 62.74%), 15,159 (COG: 50.43%), 19,722 (KEGG: 65.61%), 4052 (GO: 13.48%), and 16,057 (Interpro: 53.42%) unigenes were annotated (Table 1). With functional annotation results, 24,097 CDS were detected (Additional file 4: Table S4). Based on NR annotation, the distribution of annotated species was 60.79% B. dothidea unigenes (14,603 unigenes) with a high match to Macrophomina phaseolina MS6 species genome [37], 20.09% (4826 Unigenes) with Neofusicocum parvum UCRNP2, and 1.63% (391) with Coniosporium apollinis CBS 100,218 (Table 2).

The unigenes functional annotation in COG, GO, and KEGG were determined and classified, and are summarized in Fig. 3. In the COG classification, 15,159 unigenes (50.43%) were divided into 25 functional groups: ‘General function prediction only’ (4835 genes, 31.89%), representing the largest group, followed by ‘transcription’ (2834 genes, 18.7%), ‘function unknown’ (2399 genes, 15.8%), ‘replication, recombination, and repair’ (1660 genes, 10.95%), ‘signal transduction mechanism’ (1338 genes, 8.83%), and ‘defense mechanism’ (164 genes, 1.08%). These are summarized in Fig. 3a. The GO classification separated the 4052 unigenes (13.48%) into 43 functional groups representing biological processes, cellular component, and molecular function ontologies. In the biological processes group, ‘metabolic processes’ (2099 genes) and ‘cellular processes’ (1677 genes) were the top two GO terms. In the molecular function group, the top three GO terms were related to ‘catalytic activity’ (2367 genes), ‘binding’ (1614 genes), and ‘transporters activity’ (285 genes). A detailed analysis of the cellular component group showed ‘cell part’ (772 genes), ‘cell’ (772 genes), ‘membrane’ (672 genes), ‘organelle’ (496 genes), ‘membrane part’ (481 genes), and ‘macromolecular region part’ (353 genes) were highly represented (Fig. 3b). Searching the KEGG database revealed that 19,722 unigenes (65.61%) matched to KEGG pathways. The assembled unigenes were assigned to five specific functional categories, including cellular processes, environmental information processing, genetic information processing, metabolism, and organism systems. The most represented KEGG function category was the “global an overview maps” (7138 unigenes) and the ‘carbohydrate metabolism’ (3927 unigenes), belonging to the metabolism cluster. ‘Signal transduction’ (776 unigenes) and ‘membrane transport’ (122 unigenes) were annotated. ‘Drug resistance’ (5 unigenes) was the smallest group (Fig. 3c). These results reveal that B. dothidea infection with mycovirus mainly involve the metabolism pathway.

Transcriptional changes in fungal genes following mycovirus infection have been little reported [17, 18, 22]. Here, we examined genome-wide transcriptional differences in B. dothidea expression in response to mycovirus infection. The number of up-regulated genes ranged from 2590 to 5598 genes while the number of down-regulated genes ranged from 2325 to 4291, at a cut-off of FDR < 0.001 and |log 2 (fold change)| > 1 (Fig. 4a). The number of DEGs was the highest in LW-CP (In total of 8896 genes, 5598 up-regulated and 3298 down-regulated), LW-C (8,759 genes, 4468 up-regulated and 4291 down-regulated), and lowest in LW-P (4,915 genes, 2590 up-regulated and 2325 down-regulated), indicating that many B. dothidea genes are regulated by mycovirus infection. Representative B. dothidea genes that show significant differential gene expression with functional annotation are summarized in Table 3. In addition, we detected the expression of genes involved in gene silencing pathways. These were up-regulated in response to mycovirus infection, such as argonaut-like genes (Ago), dicer-like genes (Dicer), and RNA-directed RNA polymerase (RdRp) genes (Additional file 5: Table S5). Genes annotated heat shock protein (Hsp)-related proteins involved in ‘response to stimulus’ pathway were also up-regulated (Additional file 6: Table S6). The above results reveal that the numbers of up-regulated transcripts were more prevalent than down-regulated transcripts. Notably, more genes were differentially expressed under BdCV1 compared to BdPV1. A hierarchical clustering of the mycovirus-responsive transcriptome demonstrates the significant similarities and differences in gene expression profiles among the three groups (LW-CP, LW-C, and LW-P; Fig. 4b). We next compared DEGs to determine the numbers of mycovirus-specific or commonly expressed genes in the three groups (Fig. 4c). A total of 2483 genes were identified in the three groups, revealing a potentially critical biological process. The number of specifically expressed genes in each mycovirus-infected sample was 2203 (1876 up-regulated and 602 down-regulated) for LW-CP, 2057 (881 up-regulated and 1501 down-regulated) for LW-C, and 1177 (1059 up-regulated and 746 down-regulated) for LW-P. Although many DEGs were mycovirus-specific, 2483 DEGs were common to LW-CP, LW-C, and LW-P, suggesting that they could be involved in stress response. The B. dothidea transcriptome analysis revealed that many host genes responsive to viral infection were common and mycovirus-specific expression [17, 38]. Increasing the stringency of differentially expressed genes progressively up to FDR < 0.001 and log₂FC ≥ 2 (FC ≥ 4) also reveals a high number of significantly expressed transcripts (Additional file 7: Figure S7). DEGs were the highest in LW-CP (5313 genes) and the lowest in LW-P (2,738 genes). Furthermore, the hierarchical clustering and Venn diagram analyses demonstrate a similar trend with a cut-off of FDR < 0.001 and log₂FC > 1(FC > 2) in the gene expression profiles among the three groups.

To confirm RNA-Seq results, we selected nine genes and purified total RNAs from different biological samples. The results of RNA-Seq were highly consistent with those of RT-qPCR using specific primers (Additional file 8: Table S8). For example, the expression of CL1217.Contig1 and CL1346.Contig4_All were strongly down-regulated, and CL2349Contig4_All, CL51.Contig10_All, CL1218.Contig1_All, Unigene3107_All, CL4042.Contig3_All, Unigene4082_All, and CL5019.Contig2_All were up-regulated in response to mycoviruses, which were all confirmed by RT-qPCR. This reveals that the transcript patterns by RT-qPCR are consistent with the RNA-Seq data analysis; however, the extent of differential expression differed (Table 4).

Because many fungal genes are unknown, only a limited number of transcription factors (TFs) were annotated [17, 18, 22]. In this study, 1083 TFs belonging to 18 families, including Zn2Cys6 (684 TFs) and C2H2 zinc finger (165 TFs), were predicated from de novo transcriptome data, many of which were differentially expressed in response to chrysovirus 1 BdCV1 and partitivirus 1BdPV1 infection. In total, 11 putative TF families were differentially expressed in response to different mycovirus infections. One hundred twenty eight TFs were commonly identified in all three mycovirus-infected groups by Venn diagrams. Up-regulated TFs were more numerous than down-regulated TFs in the LW-CP and LW-P groups, while more numerous TFs were down-regulated in the LW-C group. We further examined the portion of TF families enriched in DEGs. The Zn2Cys6 family was the dominant TF family followed by C2H2 zinc finger, MYB, and bHLH (Table 5; Additional file 9: Table S9 and Additional file 10: Figure S10).

To determine whether the mycovirus-responsive genes are involved in specific pathways, a functional classification of the DEGs was performed by GO enrichment analysis. By comparing LW-CP versus Mock, 4866 DEGs were functionally assigned to the relevant terms in three categories including biological processes, cellular component, and molecular function. The GO terms ‘metabolic process’ (GO: 0008152, 690 genes) in biological process, ‘catalytic activity’ (GO: 0003824, 817 genes) in molecular function, and ‘membrane’ (GO: 0016020, 235 genes) in the cellular component category were highly enriched (Fig. 5a).

In the LW-C group, 4879 DEGs were functionally assigned to: ‘metabolic process’ (GO: 0008152, 730 genes), followed by ‘cellular process’ (GO: 0009987, 541) in biological process. The ‘catalytic activity’ (GO: 0003824, 851 genes) in molecular function, were highly enriched. In the cellular component category, the ‘membrane’ (GO: 00016020, 238 genes), ‘integral component of membrane’ (GO: 00016021, 159 genes), and ‘intrinsic component of membrane’ (GO: 0031224, 238,159 genes) were significantly enriched (p-value is set as less than 0.05). In addition, other terms including ‘response to ‘stimulus’ annotated genes such as heat shock protein (Hsp)-related proteins, ‘Signaling’ and ‘transporter activity’ related to pathways involved in mycovirus-associated response were also enriched (Fig. 5b; Additional file 6: Table S6).

In the LW-P group, 2636 DEGs were functionally assigned. It is similar to the above two groups, such as ‘cell process’ (GO: 0009987, 271) in biological process, ‘catalytic activity’ (GO: 0003824, 453 genes) in molecular function, and ‘membrane’ (GO: 0016020, 131) in the cellular component category were highly enriched, respectively (Fig. 5c; Additional file 11: Figure S11).

To elucidate the biological processes, DEGs from B.dothidea strains regulated by mycovirus infection, and KEGG pathway were analyzed. The top 20 enriched pathways are summarized from the 3 groups. In the LW-CP group, 2554 DEGs were significantly enriched (corrected p-value < 0.05) in ‘metabolic pathways’ (PATH: ko01100), and 830 DEGs were related to ‘biosynthesis of secondary metabolites’ (Fig. 6a). However, this is approximately 36 and 12% of the total genes involved in ‘metabolic pathways’ and ‘biosynthesis of secondary metabolites’, respectively (Additional file 12: Table S12A). In the LW-C group, similar to the results of LW-CP, genes associated with ‘metabolic pathways’ (PATH: ko01100) and ‘biosynthesis of secondary metabolites’ (PATH: ko01110) (Fig. 6b), approximately 39 and 13% of the DEGs, respectively, were significantly enriched (Additional file 12: Table S12B). ‘Stach and sucrose metabolism’ (PATH: ko00500) related to carbohydrate metabolism, ‘Glycerophospholipid metabolism’ (PATH: ko00564) related to lipid metabolism, and ‘Indole diterpene alkaloid biosynthesis’ (PATH: ko00403) related to biosynthesis of other secondary metabolites, were all significantly enriched. Lastly, MAPK signaling pathway of signal transduction (PATH: ko04011) and ABC transporters of membrane transport (PATH: ko02010) were enriched. In the LW-P group, more DEGs were enriched in the ‘metabolic pathways’ and ‘Starch and sucrose metabolism’ (PATH: ko00500), with approximately 39 and 11% of the DEGs significantly enriched among the DEG, respectively (Fig. 6c). In addition, ‘peroxisome’ (PATH: ko04146), with functional classes related to transport and catabolism; ‘ABC transporters’ (PATH: ko02010), involved in membrane transport; and ‘ribosome’ (PATH: ko03010), involved in translation, were also enriched with approximately 2.7, 1, and 2.6% of the DEGs significantly enriched, respectively (Additional file 12: Table S12C).

From the above functional analysis, the DEGs associated with metabolic pathways, transport and catabolism, membrane transport, transcription, and signal transduction are presumed to play a role in the interaction of fungal defense and mycovirus counter defense strategies.