Differential expression analysis
A false discovery rate of 0.05 was set as the threshold for significantly different expression. In order to understand the response of C. sativus to B. cinerea infection, GO analysis was implemented to the above DEGs, and enrichment analysis was applied based on the hypergeometric distribution, using a false discovery rate (FDR) of < 0.05 as the cutoff.
In C. sativus, more DEGs were divided into terms in the molecular function and biological process domains than to cellular component terms. The dominant terms in each domain were “phosphotransferase activity”, “oxidation–reduction process”, and “integral to membrane”, respectively (Additional file 7: Figure S1A). The most significantly enriched GO terms in the molecular function domain included “phosphotransferase activity—alcohol group as acceptor” (GO:0016021), “GTP binding” (GO:0015979), “ATP binding” (GO:0005576), “protein tyrosine/serine/threonine phosphatase activity” (GO:0005506), and “heme binding” (GO:0009765). The most significantly enriched GO terms in the biological process domain included “oxidation–reduction process” (GO:0055114), “negative regulation of transcription”, “DNA-dependent” (GO:0009734), “protein phosphorylation” (GO:0004601), “oxidation–reduction process” (GO:0009522), and “carbohydrate transport” (GO:0004497) (Additional file 1: Table S1).
In B. cinerea, more DEGs were divided into terms in the biological process and cellular component domains than to molecular function terms. The dominant terms in each domain were “transport”, “cytosol”, and “hydrolase activity”, respectively (Additional file 7: Figure S1B). The most significantly enriched GO terms in the molecular function domain included “hydrolase activity” (GO:0005975), “oxidoreductase activity” (GO:0004553), “TBP-class protein binding” (GO:0003868), “purine nucleobase transmembrane transporter activity” (GO:0070884; GO:0046355), and “RNA polymerase I activity” (GO:0045461). The most significantly enriched GO terms in the biological process domain included “transport” (GO:0030248), “oxidation–reduction process” (GO:0055114; GO:0016812), “metabolic process” (GO:0016491; GO:0007346), “mitochondrial transport” (GO:0008864), “vesicle-mediated transport” (GO:0030245), and “methylation” (GO:0004076) (Additional file 2: Table S2).
To further clarify the functions of DEGs, they were mapped to KEGG terms to identify genes involved in significantly enriched biosynthetic or signal transduction pathways in C. sativus and B. cinerea. 277 DEGs were assigned to 19 KEGG pathways in C. sativus (Additional file 3: Table S3). The top five significantly enriched biosynthetic pathways included “phenylpropanoid biosynthesis”, “photosynthesis”, “biosynthesis of antibiotics”, “fatty acid elongation”, and “valine, leucine, and isoleucine degradation” (Additional file 8: Figure S2A). The pathway involving the highest number of DEGs was “biosynthesis of antibiotics” (86; 31.05%), followed by “phenylpropanoid biosynthesis” (53; 19.13%), “starch and sucrose metabolism” (35; 12.64%), “pentose phosphate pathway” (22; 7.94%), and “glycine, serine, and threonine metabolism” (14; 5.05%). Therefore, we considered the DEGs involved in these pathways as candidates associated with C. sativus susceptibility to B. cinerea.
In B. cinerea, 150 DEGs were assigned to 26 KEGG pathways (Additional file 4: Table S4). Among these, “starch and sucrose metabolism”, “pentose and glucuronate interconversions”, “cyanoamino acid metabolism”, “biosynthesis of antibiotics”, and “phenylpropanoid biosynthesis” were the top five most significantly enriched pathways (Additional file 8: Figure S2B). The pathway involving the highest number of DEGs was “biosynthesis of antibiotics” (27; 18.00%), followed by “starch and sucrose metabolism” (16; 10.67%), “pentose phosphate pathway” (11; 7.33%), “phenylpropanoid biosynthesis” (9; 6.00%), “cyanoamino acid metabolism” (7; 4.67%), and “glyoxylate and dicarboxylate metabolism” (7; 4.67%).
Clustering using the Skellam model
The Skellam model was used to cluster RNA genes into distinct groups. Because it incorporates sample size information, we used the Bayesian information criterion (BIC) as the model-selection criterion. First, we clustered 4200 differentially expressed genes (DEGs) in C. sativus into distinct groups. From the plot of the BIC against the group numbers, all the DEGs are categorized into 17 distinct groups (Fig. 1A). We had illustrated the mean expression in each group of C. sativus and these 17 groups displayed differential levels in expression (Fig. 2A and Additional file 5: Table S5). Figure 3A plotted the pattern of the C. sativus gene expression differences before and after fungal infection, which showed that DEGs in 11 groups were up-regulated, whereas those in 6 groups were down-regulated. The gene groups were not parallel and different patterns of gene expression plasticity was exhibited in response to environmental changes from an uninfected state to an infected one. Subsequently, based on the BIC values under different numbers of clusters, 670 DEGs in B. cinerea were clustered into 12 groups (Fig. 1B). The mean expression values in each group of B. cinerea were showed in Fig. 2B and Additional file 6: Table S6. The pattern of pathogen gene expression differences before and after host infection, in which DEGs in 7 groups were up-regulated, whereas those in 5 groups were down-regulated (Fig. 3B).
Plasticity expression pattern
Of these 17 groups in C. sativus, gene expression levels from groups 1, 2, 3, 4, 7, 8, 9, 12, 13, 15, 17 (accounting for nearly 42.5% of genes) were clearly up-regulated after B. cinerea infection. Nearly 50% of genes (groups 6, 10, 11, 14, 16) were clearly down-regulated and gene expression levels from group 5 (about 9.4%) tended to be slightly down-regulated. In group 7, the most significantly enriched GO term responded to “oxidative stress” (GO:0006979), indicating that the plant reacted to pathogen infection. GO term “cell wall” (GO:0005618) was significantly enriched in group 11 and term “photosystem” was significantly enriched in group 14. In B. cinerea, of these 12 groups, only the mean expression values of group 2 (about 18.4%) were clearly down-regulated after infecting C. sativus. Approximately 40.15% of genes from groups 3, 4, 5, 9 were slightly down-regulated. Genes in other groups are up-regulated after infecting C. sativus. Hypothesis tests were performed to examine whether each cluster of genes expressed significantly differently between the two treatments and determined whether a particular pair of gene groups interacted with the environment. Plasticity gene expression was statistically significant (P < 0.05). This indicated that DEGs tended to obvious changes in response to B. cinerea infection. All pairs of gene clusters displayed significant gene-environment interactions (P < 0.05).
Gene regulatory network
The core-periphery structure is a vital feature of many biological networks, including protein-protein interaction networks as well as gene regulatory and metabolic networks (Csermely et al. 2013). In this study, we constructed networks of genes that interacted with each other to screen hub genes based on a directed graphical model known as Bayesian networks. Through a detailed GO analysis, we detected hub genes which were biologically meaningful.
For example,the gene regulatory network of group 7 in C. sativus was shown in Fig. 4A. All the 116 genes were displayed in green (Additional file 5: Table S5), in which two in red were No. 63 (Csa5G285030, Proteinase inhibitor) and 73 (Csa1G265640, Uncharacterized protein). They were two hub genes detected by the Bayesian networks. Csa5G285030 was enriched in “response to wounding” (GO:0009611), which might be involved in response to stress such as wounding and pathogens. Group 12 contained 149 genes in the network (Additional file 5: Table S5), in which No. 29 (Csa2G075440) was screened as a hub gene. Csa2G075440 was annotated as “disease resistance protein RPS2” in KEGG orthology and enriched in the pathway of “plant–pathogen interaction” (Bent et al. 1994). In B. cinerea, the mean expression values of group 10 were clearly up-regulated after infecting C. sativus. There were 15 genes in this group, in which No. 4 (B0510_3699) was identified as one of the hub genes (Fig. 4B). The gene probably encodes 1,4-beta-d-glucan cellobiohydrolase which participates in regulating the hydrolase activity or hydrolyzing O-glycosyl compounds (GO:0004553) for pathogens to invade plant cells or exploit the polysaccharides of plant cell walls (Additional file 6: Table S6) (Kong et al. 2015).