Introduction

Brassica juncea var. tumida Tsen (Brassicaceae) is a feature vegetable in Fuling, Chongqing and is the raw material for “Fuling mustard” pickles, which is a famous fermented food. This crop’s production is severely affected by clubroot disease caused by Plasmodiophora brassicae (Ludwig-Muller and Schuller 2008), and this has caused significant economic losses. Owing to cell elongation and cell division in infected roots and hypocotyls, the symptoms caused by P. brassicae are typical hypertrophied cells (root galls) (Ludwig-Muller and Schuller 2008; Kobelt et al. 2000). The vasculature of the infected plant was destroyed, making the uptake of nutrition and water difficult and resulting in foliar wilting when under slight soil moisture stress. Galls constitute strong metabolic sinks that redirect assimilates and other nutrients from shoots to roots to meet the pathogen’s own needs (Paesold et al. 2011), leading to yield decreases or plant death.

P. brassicae is an obligate biotrophic pathogen that can exist in soil for over 20 years (Donald and Porter 2009). The disease is difficult to control using only conventional chemical or cultural means (Donald and Porter 2009). Although the breeding of resistant cultivars is an effective measure to control the disease, resistant varieties are very scarce and the variation in the pathogen’s virulence is great (Hirai 2006). Biological control has received more attention recently because it is environmentally friendly, cost-effective, and sustainable (Jaschke et al. 2010). The antagonism mechanisms of biocontrol strains include the competition for nutrients and space, mycoparasitism, the production of antifungal metabolites, plant growth promotion, and the induction of defense responses in plants (Howell 2003).

The lifecycle of P. brassicae consisted of two phases: a primary phase that takes place mainly in root hairs and occasionally in epidermal cells and a secondary phase that occurs predominantly in the root cortex (Ludwig-Muller and Schuller 2008). The primary phase of infection can still occur in resistant hosts, but not the secondary phase (Donald et al. 2008).

Plants combat pathogen attacks through two innate immune techniques: pathogen-associated molecular pattern (PAMP)-triggered immunity (PTI) and effector-triggered immunity (ETI) (Jones and Dangl 2006). When PAMPs are detected by pattern recognition receptor (PRR) proteins on the surface of the host cell, the PTI pathway is triggered (Jones and Dangl 2006). The pathogen can avoid PTI by secreting effectors into host cells. When the effectors are recognized by specific resistance (R) genes in the host, ETI is triggered (Jones and Dangl 2006).

The analysis of global gene expressions is a powerful tool to study the interactions between pathogens or insect and hosts, especially the resistance mechanisms and resistant genes. Wang et al. (2016) study the differential gene expression profiles between the Ectropis oblique damage-induced tea plants and undamaged control using RNA sequencing (RNA-Seq) and indicate that plant secondary metabolites and the signaling pathways may play an important role in defense against insects. So far, some work has been conducted to examine the interaction between host and P. brassicae. Using the Arabidopsis Affymetrix ATH1 full-genome chip, Siemens and his colleagues identified differently expressed genes between infected and control roots, including genes associated with growth and cell cycle, sugar phosphate metabolism, and defense (Siemens et al. 2006). The protein profile changes at the primary infection stage in Arabidopsis thaliana and Canola showed that most of the differentially regulated proteins were involved in metabolism, cell defense, and the detoxification of reactive oxygen species (ROS) (Devos et al. 2006; Cao et al. 2008). In A. thaliana, at 4 days after inoculation (DAI), the expression levels of genes that took part in the pathogen’s recognition and signal transduction were highly induced (Agarwal et al. 2011). A transcriptome analysis of Brassia rapa using two near-isogenic lines carrying clubroot-resistant and clubroot-susceptible alleles showed that the differently expressed genes were mainly associated with signal transduction, defense, transport, and metabolism (Chen et al. 2016).

In our laboratory, we screened an antagonistic bacteria Zhihengliuella aestuarii B18 against P. brassicae, which was obtained from the rhizospheric soil of B. juncea var. tumida grown in Fuling, Chongqing. The biocontrol effects of the antagonistic strain against clubroot on B. juncea var. tumida reached 63.4 and 49.7% in the greenhouse experiment and field trial, respectively (Luo et al. 2017). To better understand the biocontrol mechanism, a transcriptome analysis of B. juncea var. tumida Tsen infected by P. brassicae was conducted after the biocontrol agent Z. aestuarii was applied. We constructed three cDNA libraries using Illumina HiSeq 4000, one from B. juncea only inoculated with P. brassicae (P), one inoculated with P. brassica and the biocontrol agent Z. aestuarii (deposit number CICC11045s in the China Center of Industrial Culture Collection) at the same time (P + B), and the other was the control (H), in which P. brassicae was replaced by sterile water. A total of 107,617 unigenes were obtained and these unigenes’ functional annotations were determined. A differential expression analysis was performed among the biocontrol-treated (P + B), the infected (P), and the control (H) samples, and resistance-related genes were identified among the differentially expressed genes. The transcriptome results were validated by reverse transcription quantitative PCR (qRT-PCR). The data will serve as an important public information platform to study B. juncea var. tumida Tsen and will be useful for breeding mustard plants resistant to P. brassicae.

Materials and methods

Plant material

Seeds of B. juncea var. tumida were sown in a stainless steel tray (35 cm × 20 cm) that was almost full of perlite (Thermal Insulation Material Factory, Chongqing, China) supplemented with 600 mL of Murashige and Skoog nutrient solution (Murashige and Skoog 1962). The mustard seedlings were maintained in an illuminated incubator at 25 °C under a 11-h/13-h (light/dark) photoperiod with light supplied by GXZ-300B lamps at an intensity of 12,000 lx.

Preparation of resting spore suspensions

Naturally infected diseased plant roots of B. juncea var. tumida were derived from fields in the Fuling District of Chongqing. The diseased roots were stored at − 20 °C until required. The resting spores were extracted by homogenizing mature clubroot galls of B. juncea var. tumida in sterile water before filtering the homogenate through gauze (25-μm pore) and clarifying the filtrate by two centrifugation steps at 1475×g (Xiao and Guo 2002). The final precipitate of resting spores was re-suspended in sterile water and adjusted to a concentration of 1.5 × 108 spores mL−1. Three independent experiments were conducted to produce the dataset.

Liquid cultures of bacteria

The bacterial isolates were cultured in LB liquid medium and grown on rotary shakers at 180 rpm at 30 °C for 2 days. The bacterial liquid was collected and adjusted to 108 colony forming units mL−1 using LB medium.

P. brassicae inoculation and root tissue sampling

Seedlings with four leaves were inoculated by adding 400 mL of spore suspension and 200 mL of bacterial culture to each tray (P + B), while the other group was inoculated with 400 mL of spore suspension and 200 mL of sterile water (P) as a positive control. The seedlings inoculated with only 600 mL sterile water served as negative controls (H). Two independent experiments were conducted to produce the dataset. On the 6th and 13th day after inoculation (DAI), seedlings were inoculated with 200 mL of bacterial culture per tray, while the positive and the negative control groups were inoculated with 200 mL of sterile water. The root tissues were harvested at 15 days after the first inoculation and washed with water to remove the perlite. The dried roots were immediately frozen in liquid nitrogen and stored at − 80 °C before further analysis. Every treatment was repeated three times, and every sample consisted of 20 plants per treatment. The disease symptoms were recorded 28 DAI.

RNA extraction

Total RNA was isolated using TRIzol according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA; Cat. no. 15596-026). The extracted RNA was treated with DNase I (Fermentas, Hanover, MD, USA; Cat. no. EN0531) to remove potential genomic DNA contamination. The purity of the RNA was confirmed using a 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) and a Nanodrop spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).

cDNA library construction for transcriptome analysis

The cDNA library construction and sequencing were performed by Beijing Genomics Institute (Wuhan, China) using NEBNext® Ultra™ RNA library Prep kit for Illumina® (NEB, USA; Cat. no. E7530L). The mRNA was isolated from total RNA using beads with Oligo(dT) primers. Under an elevated temperature, mRNA was fragmented into small pieces after the fragmentation buffer was added. Using the mRNA fragments as templates, the first-strand cDNA was synthesized with random primers. Then, the second-strand cDNA was obtained using DNA polymerase I and RNase H. The synthetic cDNAs were end-repaired by polymerase and ligated with Solexa adapters. Suitable fragments were chosen as templates for the AMPure™ XP system (Beckman Coulter, Beverly, MA, USA; Cat. no. A63882), and PCR amplification was performed to enrich the product. The three libraries (P + B, P, and H) were sequenced using an Illumina HiSeq 4000 (Table 1).

Table 1 Summary of sequencing results
Full size table

Sequencing reads filtering, assembly, and unigene functional annotations

To yield clean reads, raw reads of low-quality were removed using the Soap software package (http://soap.genomics.org.cn/), including reads with adaptors, reads in which unknown bases were more than 1%, and reads having a percentage of bases of quality lower than 10 were greater than 20%. The clean reads were stored in FASTQ format. Then, the clean reads were de novo assembled using the Trinity program (https://github.com/trinityrnaseq/trinityrnaseq/wiki) to determine the unigenes (Grabherr et al. 2011). To form longer fragments without unknown sequences, Trinity was used to combine the reads overlapping in certain sequences to obtain contigs. These contigs were processed to form longer sequences (unigenes) using Trinity. Then, the unigenes from different samples were spliced, and redundancy was removed by the clustering software Tgicl (http://www.jcvi.org/cms/research/projects/tdb/overview/) to produce the longest possible non-redundant unigenes (all-unigenes) (Pertea et al. 2003). The generated unigenes were compared with the non-redundant protein (NR, ftp://ftp.ncbi.nlm.nih.gov/blast/db), Swiss-prot (http://ftp.ebi.ac.uk/pub/databases/swissprot), the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg), and the cluster of Orthologous Groups (COG) databases (http://www.ncbi.nlm.nih.gov/COG) using the BLASTX algorithm (E-value < 10−5, http://blast.ncbi.nlm.nih.gov/Blast.cgi), and the unigenes’ direction was determined based on the best aligning result. Using the NR annotation, the Blast2GO program (https://www.blast2go.com) acquired GO annotations for the unigenes. Then, WEGO (http://wego.genomics.org.cn) and Top GO software (http://bioconductor.uib.no/2.7/bioc/html/topGO.html) programs were used to carry out GO functional classifications and enrichment analyses to understand the distributions of the gene functions. The KEGG pathway annotations and enrichment analyses were performed using BLASTALL software against KEGG (http://www.genome.jp/kegg/) databases.

Differentially expressed gene analysis

The genes’ expression levels were evaluated by counting the read density as reads per kilo base per million mapped reads (Mortazavi et al. 2008). The correlation of the detected sequence in two parallel libraries was assessed statistically by Pearson’s correlation coefficient (P value ≤ 0.001) using DEGseq2 software (http://bioinfo.au.tsinghua.edu.cn/software/degseq). P values were corrected using the Benjamini–Hochberg program to determine the false discovery rate (Thissen et al. 2002). Genes were regarded as significantly differentially expressed if the log2 fold change of the reads per kilo base per million mapped reads was > 2 and the false discovery rate was ≤ 0.001, based on published literature (Chen et al. 2016).

qRT-PCR

To validate the transcriptome results, qRT-PCR was performed using the iCyclerTM Thermal Cycler system (Bio-RAD, Hercules, CA, USA) with the specific primers for the selected 22 genes that were related to resistance (Table 2) according to our bioinformatics analysis and previous reports (Siemens et al. 2006; Devos et al. 2006; Cao et al. 2008; Agarwal et al. 2011; Chen et al. 2016; Chu et al. 2014). Total RNA was extracted three times from (B + P, P) and H roots at 15 days post-infection and treated with DNase I (Fermentas) prior to reverse transcription. Three RNA sets were used for cDNA synthesis using both oligo (dT)18 primers and random hexamer primers provided as part of the RevertAidTM First Strand cDNA Synthesis kit (Fermentas, Pittsburg, PA, USA; Cat. no.K1622) following the manufacturer’s instructions. The PCR was performed with three cDNA sets at each point as templates using the SYBR® Premix Ex TaqTM II (Takara, Dalian, Liaoning, China; Cat. no. RR820A). Conditions used for amplification were 95 °C for 2 min, followed by 40 cycles of 95 °C for 10 s, 61 °C (57 °C for reference genes) for 20 s, and 72 °C for 30 s. The ACT and UBE cDNAs (encoding the actin and ubiquitin-conjugating enzymes, respectively) were used as internal controls to standardize transcript levels. The qPCR results were analyzed using ANOVA (analysis of variance).

Table 2 The primers designed for the qRT-PCR analysis
Full size table

Results

Disease symptoms

The disease indices of the biocontrol-treated and infected samples were 35.4 and 96.8, respectively, 28 DAI (Luo et al. 2017). The morphology of the biocontrol-treated and infected samplesis shown in Fig. 1.

Fig. 1
figure 1

Root development in biocontrol-treated and the infected samples: a roots from the biocontrol-treated group, b roots from the infected group. 10 and 15d means 10 and 15 days post-infection

Full size image

Sequencing reads filtering and assembly

To evaluate the molecular mechanisms of the potential biocontrol strain against P. brassicae, three cDNA libraries were constructed. The transcriptome samples were P + B, P, and H. A total of 19.94 Gb was generated after Illumina HiSeq sequencing. The ratio of the Q20 was greater than 96%, which indicated that the sequencing contained most of the expressed genes. The details of the sequencing are listed in Table 1. The total lengths and the average lengths of the transcripts from the B + P sample were 105,901,898 and 756 bp, respectively, while those of the transcript from the P sample were 102,641,447 and 742 bp, respectively. The total lengths and the average lengths of the transcripts from the H sample were 113,116,391 and 759 bp, respectively.

The sequence data were de novo assembled, and 107,617 unigenes were obtained. The total lengths and the average lengths of the unigenes were 104,094,386 and 967 bp, respectively. The GC percentage of the unigenes was 44.54%, and the N50 was 1488 bp.

Differentially expressed genes among three transcriptome sample

A total of 13,226 differentially expressed genes were identified between the P + B and H samples. Among these, 5708 genes were up-regulated, while 7518 genes were down-regulated. There were only 2524 genes up-regulated and 13,137 genes down-regulated in P compared with H. In P + B, there were 8161 genes up-regulated and 1643 genes down-regulated when compared with P (Fig. 2).

Fig. 2
figure 2

Venn diagram of differentially expressed genes among biocontrol-treated (P + B), infected (P), the control (H) samples

Full size image

Unigene functional annotation

The generated unigenes were compared with the database NT, NR, Swissprot, KEGG, COG, GO databases, and the unigenes’ functional annotations were determined based on the best aligning result. In the NT database, 93,536 (86.92%) sequences had similarities to known sequences. In the NR and Swissprot databases, 91,646 (85.16%) and 63,035 (58.57%) sequences, respectively, were functionally annotated.

The unigenes were aligned with the COG database, and 33,432 sequences (31.07%) were distributed into 25 COG categories. Among the 25 COG categories, “General function prediction only” (11,842, 16.14%) was the largest group, followed by “Transcription” (7299, 9.95%), “Translation, ribosomal structure and biogenesis” (5711, 7.785%), and “Replication, recombination and repair” (5601, 7.63%). The smallest groups were “Nuclear structure” (10, 0.0136%) and “Extracellular structures” (35, 0.0477%). The functional distribution of the COG annotation is shown in Supplementary Fig. 1.

The unigenes were then compared with the GO database, and 75,926 sequences were grouped to the three categories: biological process, cellular component, and molecular function. The GO functional analysis was also applied to the differentially expressed genes between biocontrol-treated (P + B) and infected (P) samples. A total of 6797 sequences were divided into 54 groups. Among the 54 groups, “cell” (6440, 5.2%) and “cell part” (6440, 5.2%) were the largest groups, followed by “organelle” (5106, 4.14%), “cellular process” (4790, 3.9%), and “metabolic process” (4670, 3.8%) (Supplementary Fig. 2).

The unigenes were also aligned with the KEGG database, and 52,352 sequences were distributed into 5 categories and 20 functional groups. The five categories were as follows: “cellular processes,” “environmental information processing,” “genetic information processing,” “metabolism,” and “organismal systems.” Among the 20 functional groups, “global map” was the largest group, followed by “translation,” “carbohydrate metabolism,” “folding, sorting and degradation,” “signal transduction,” “environmental adaptation,” and “signal transduction.” The KEGG functional analysis was also applied to the differentially expressed genes between P + B and P sample. In total, 5629 differentially expressed genes were assigned to 126 KEGG pathways. Of these, 20 pathways were significantly enriched. The most represented pathways were as follows: metabolic pathways (1614 members, 28.67%), biosynthesis of secondary metabolites (807 members, 14.34%), plant hormone signal transduction (422 members, 7.5%), endocytosis (342 members, 6.08%), RNA transport (332 members, 5.9%), glycerophospholipid metabolism (331 members, 5.88%), and plant–pathogen interaction (311 members, 5.52%) (Supplementary Fig. 3). The transcriptome data were deposited in NCBI’s Sequence Read Archive under the project ID PRJNA339019 with the accession number SRR4034946.

Identification of genes involved in resistance to P. brassicae

To determine the molecular mechanisms of the potential biocontrol strain against P. brassicae, the resistance-related genes were obtained by combining those in the literature with those expressed in enriched KEGG pathways (Fig. 3). These genes include the following: PR proteins, PRRs involved in PTI, R proteins involved in ETI, calcium influx genes, hormone regulation genes, respiratory burst oxidase and mitogen-activated protein kinase (MAPK) cascades genes, cell wall modification genes, and cytokinin synthase gene, which were all up-regulated in BP compared with P, except cytokinin synthase gene (Table 3).

Fig. 3
figure 3

The enriched KEGG pathways of the resistance mechanism of the B. juncea var. tumida Tsen de novo transcriptome affected by biocontrol strain Z. aestuarii. a Plant hormone signal transduction pathways. b Plant–pathogen interaction pathways. The red and green frames represent genes and enriched functions that were up- and down- regulated in biocontrol-treated and the infected samples

Full size image
Table 3 Functional classification of B. juncea var. tumida Tsen transcripts that were differentially regulated between biocontrol-treated (BP) and the infected (P) root tissues at 15 days post-inoculation
Full size table

qRT-PCR validation

To validate the transcriptome results, 22 genes related to resistance were analyzed using qRT-PCR, including lipid transfer proteins (CL4183.contig1), thaumatin (CL7115.contig3), BAK1 (Unigene23458), CERK1 (CL3604.contig1), RPS4 (CL3351.contig1), RPS5 (Unigene41313), calmodulin-binding protein (CL8316.Contig8), calmodulin-like protein (CL6475.contig2), CDPK (CL8426.contig1), PR1 (CL12690.Contig2), NPR1 (Unigene9371), SAUR (Unigene19142), GH3 family protein (CL2550.Contig2), JAZ (CL5009.Contig1), MAPK7 (CL1005.Contig8), MEKK1 (CL6418.Contig7), MKK4 (Unigene45908), WRKY29 (CL3.Contig1), WRKY22 (CL4233.Contig2), glutathione S-transferase (CL6386.Contig3), heat shock protein (CL11002.Contig2), and arabinogalactan protein (Unigene40439). For all 22 genes, the expression tendencies indicated by the qRT-PCR were almost the same as in the transcriptome analyses (Fig. 4, supplementary Fig. 4), except CL6475.contig2, CL11002.Contig2, and CL3.Contig1.

Fig. 4
figure 4

qRT-PCR validation. The relative expression was analyzed at 15 days post-inoculation with ACT and UBE as standards. BP, P, and H represent biocontrol-treated, infected, and control samples, respectively. Asterisks indicate significant differences (p value < 0.05) compared with H. Error bars represent ±SD of three independent PCR amplifications and quantifications.

Full size image

The expression level of CL6475.contig2 and CL3.Contig1 in BP was slightly less than H, and CL11002.Contig2 in H was slightly higher than P, but the difference between two samples was much smaller. Thus, the results indicated that our transcriptome data was reliable.

Discussion

RNA-sequencing technology is a powerful tool for studying gene expression variation and can be used in non-modal organisms without known genome sequences. RNA-sequencing is more comprehensive and accurate than expressed sequence tag libraries and is more widely used to obtain differently expressed genes and their annotations under various conditions. Using Illumina paired-end RNA-sequencing, Tang et al. (2017) investigated the transcriptome changes during barley grain development from two barley landraces with the differential seed starch synthesis traits. Janská et al. (2011) reported gene expression in the leaf and crown of the winter barley cultivar Luxor, following the exposure of young plants to various periods of low (above and below zero) temperatures. Szeszko et al. (2016) identified differentially expressed genes of porcine luteal ovarian cells in response to adiponectin treatment. In our study, 19.94 Gb was generated by Illumina HiSeq sequencing, and 107,617 unigenes were obtained with a mean length of 967 bp. Compared with H, there were more up-regulated genes and less down-regulated genes in P + B than in P. Compared with H, the genes with opposite expression patterns in B + P and P were most likely related to resistance against P. brassicae (Fig. 5).

Fig. 5
figure 5

Schematic diagram illustrating the defense mechanisms induced by P. brassicae. Red and green represent up- and down-regulated genes, respectively

Full size image

Up-regulated genes expressed in the P + B sample but not in the P sample

PR proteins

Plants defend against pathogen invasion in various ways, including the production of PR proteins that are induced specifically under pathological or other related conditions (Sels et al. 2008). Among these PR proteins, thaumatin family proteins and lipid transfer proteins belonged to PR-5 and PR-14, respectively (Liu et al. 2010; De Oliveira Carvalho and Gomes 2007). A transcriptome analysis of B. rapa CR BJN3-2 carrying clubroot-resistance genes showed that five thaumatin family proteins and six lipid transfer proteins were up-regulated following inoculation with P. brassicae compared with the susceptible allele BJN3-2 (Chen et al. 2016). In our experiment, compared with H, six PR proteins (five thaumatin family proteins and one lipid transfer proteins) were up-regulated in P + B, while they were down-regulated in P. This means that the expression levels of thaumatin family proteins and lipid transfer proteins were suppressed by the pathogen and induced by the biocontrol strain Z. aestuarii. Thus, the PR proteins may be involved in resistance to P. brassicae in B juncea var. tumida Tsen induced by Z. aestuarii.

PRRs involved in PTI

Plants first combat pathogens through PTI pathways, which occur when PAMPs are detected by PRR proteins on the surface of the host cell (Jones and Dangl 2006). Some important PRR genes triggered by PAMPs, such as brassinosteroid insensitive 1-associated kinase 1 (BAK1), flagellin sensing 2 (FLS2), chitin elicitor receptor kinase (CERK), and chitin elicitor-binding protein (CEBiP), can transduce signals that trigger PTI (Dodds and Rathjen 2010). In our study, three CERKs were induced in P + B and suppressed in P compared with H. The expression levels of four BAK1 genes were greatest in H and lowest in P. The BAK1 genes were marginally induced by the biocontrol agent. In previous reports, BAK1, CEBiP, and CERK1 were up-regulated in a resistant banana line infected with Fusarium oxysporum (Li et al. 2012). Our results were in agreement with theirs. Thus, we hypothesized that the PTI pathways play roles in the resistance to P. brassicae.

R protein involved in ETI

Pathogens secrete effectors into host cells to avoid PTI. When the effectors are recognized by specific resistance (R) genes in the host, the ETI pathways are triggered to combat the pathogens. RPS4, containing the avirulence gene AvrRPS4, was specifically resistant to P. syringae pv. tomato (Wirthmueller et al. 2007), and RPS5, carrying the avirulence gene AvrPphB, was specifically resistant to P. syringae strains (Ade et al. 2007). RPS4 and RPS5 were up-regulated in clubroot-resistant B. rapa “CR BJN3-2” compared with in clubroot-susceptible allele lines at 0, 12, 72, and 96 h after inoculation (Chen et al. 2016). Consistent with these reports, two RPS4 and two RPS5 proteins were up-regulated after the root of Brassica juncea var. tumida Tsen inoculated with P. brassicae was drenched with the biocontrol strain. In our experiment, not only PTI, but also ETI, pathways were triggered after the addition of biocontrol strain and took part in the resistance to P. brassicae.

Calcium influx

Calcium is an important secondary messenger involved in signal transduction pathways that is regulated in many plant cell processes, such as responses to abiotic and biotic stress (Zipfel 2009). Calcium cannot only reduce root-hair infections but also inhibits the production of differentiated and dehisced sporangia of P. brassicae in infected Chinese cabbage root hairs (Donald and Porter 2009). In the field, calcium nitrate has been widely used to control clubroot (Donald and Porter 2009). Calcium-mediated signaling processes appeared necessary for the increase in phenylalanine ammonia-lyase activity induced by P. brassicae (Takahashi et al. 2002). In our study, some genes involved in calcium influx were up-regulated in the biocontrol-treated sample compared with P, including CDPK, calmodulin-like genes, and the calmodulin-binding protein, while the calmodulin-like and CDPK genes were expressed less in the P + B sample than the H. Due to the P. brassicae infection, the expression level of calmodulin-like and CDPK genes were down-regulated and could be compensated partly by the biocontrol strain. In susceptible Arabidopsis infected by P. brassicae, the CDPK gene was up-regulated at 4 DAI when pathogen recognition and signal transduction occurred to trigger the early stages of plant defense responses (Agarwal et al. 2011). At 4 DAI, but not 7 or 10 DAI, some resistance genes, including CDPK, which also existed in other resistant host-pathogen interactions, were induced in susceptible Arabidopsis (Agarwal et al. 2011; Kasukabe et al. 2004; Kuznetsov et al. 2006). Compared with susceptible BJN 3-2, calmodulin-like genes were induced at 0 h after inoculation with P. brassicae in resistant Chinese cabbage “CR BJN 3-2” (Chen et al. 2016). Our data was aligned with these genes, and the calcium-influx signaling pathway may be involved in the resistance to P. brassicae.

Hormone regulation

In the current study, the genes related to auxin regulation, such as the SAUR and GH3 family proteins (Wu et al. 2012; Kant et al. 2009; Ding et al. 2008; Fu et al. 2011; Zhang et al. 2007; Wang et al. 2012; Glawischnig 2007), were induced in the B + P sample compared with the P sample. The expression level of GH3 was much higher in the B + P sample than the other two samples, while the SAUR protein was comparable with H. Thus, the suppression effect of the pathogen may be counteracted by the biocontrol strain. The GH3 family protein is linked to intensive immunity in rice by suppressing auxin accumulation (Ding et al. 2008; Fu et al. 2011), and the over-expression of GH3.5 in Arabidopsis resulted in the enhanced biosynthesis of camalexin, which is a major phytoalexin resistant to pathogens (Zhang et al. 2007; Wang et al. 2012). The biosynthesis of camalexin may lead to the suppression of auxin synthesis owing to the common precursor, tryptophan (Glawischnig 2007). The SAUR protein takes part in auxin biosynthesis and signaling pathway (Wu et al. 2012) and acts as negative regulator of auxin synthesis and transport in rice (Kant et al. 2009). In previous reports, the SAUR and GH3 family proteins were induced in resistant canola compared with sensitive plants (Chu et al. 2014). Our results were in accordance with these studies. In clubroot disease, auxin is involved in pathogenicity (Ludwig-Muller and Schuller 2008). The up-regulation of the SAUR and GH3 family proteins results in a reduction of auxin and is related with the resistance to P. brassicae.

In our study, the PR1 and NPR1 genes, which are involved in salicylic acid (SA) signaling, were also up-regulated in the B + P sample (Spoel et al. 2003). As shown in Fig. 4, the expression level of PR1 and NPR1 in the B + P sample was much higher than the P and H samples. The NPR1 and PR1 genes were localized downstream of the SA pathway and were linked with disease resistance (Fig. 3). Plant resistance to biotrophic pathogens is mainly controlled by the SA pathway, and host resistance to necrotrophic pathogen is regulated by the ethylene and jasmonic acid signaling pathway (Glazebrook 2005). Lovelock et al. (2013) demonstrated that exogenous applications of SA can suppress clubroot in broccoli. The SA signaling pathway was down-regulated in susceptible A. thaliana (Agarwal et al. 2011), and the PR1 and NPR1 genes involved in the SA pathway were up-regulated in clubroot-resistant near-isogenic lines of Chinese cabbage (Chen et al. 2016). Thus, the SA-signaling pathway was related with resistance to P. brassicae, which was induced by the biocontrol strain. In our experiment, the JAZ protein, which was related to JA biosynthesis and signaling (Fig. 3), was also up-regulated in the BP sample, and the expression level was much higher than the P and H samples (Fig. 4). In other reports, the SA pathway was induced, and the JA pathway was suppressed in resistant plants (Chen et al. 2016; Jubault et al. 2013). Our result conflicted with those of previous studies. Lemarie et al. (2015) demonstrated that both the SA and JA pathways contribute to resistance against P. brassicae in Arabidopsis. Consistent with these results, in our experiment, both the SA and JA pathways were involved in resistance to P. brassicae after the biocontrol strain was applied.

Respiratory burst oxidase and MAPK cascades

In resistant plants, early defense responses occur within minutes of infection, including the activation of MAPK cascades and burst of reactive oxygen intermediates (Torres and Dangl 2005; Zhao et al. 2005; Jones and Dangl 2006). In our study, 2 MAPK7, 20 heat shock proteins, 9 glutathione S-transferase, 2 MEKK1, 2 MKK4, and 11 WRKY (22, 29) genes were up-regulated in the B + P sample compared with the P sample and H, except the MEKK1 and WRKY 29, which was lower in the B + P sample than H. MEKK1 is required for basal defenses (Zhang et al. 2012) and was suppressed by P. brassicae. After the addition of Z. aestuarii, the expression level of MEKK1 was only slightly up-regulated. The MEKK1 gene may not be involved in resistance to P. brassicae. The heat shock proteins and glutathione S-transferase are involved in the oxidative burst, which functions as antimicrobial compounds and signaling molecules in the plant defense reaction (Wan et al. 2002; Sagi and Fluhr 2001; Agarwal et al. 2011). The oxidative burst was repressed at 4 days after infection in the compatible A. thaliana infected by P. brassicae (Agarwal et al. 2011) and the ROS level was higher in resistant B. rapa inoculated with P. brassicae (Chen et al. 2016). Our result contradicted that of a compatible phenotype but was in accordance with the resistant phenotype. The WRKY proteins belong to a superfamily of transcription factors, which respond to wounding and other stresses and act in a complex defense-response network in plant immunity (Dellagi et al. 2000; Pandey and Somssich 2009). In our experiment, WRKYs 22 was highly up-regulated in the B + P sample compared with P and H, while the expression level of WRKY 29 in the B + P was slightly lower than H. The WRKY 22 and WRKY 29 gene may be involved in resistance to P. brassicae. Thus, the MAPK cascades and reactive oxygen intermediates may work together to combat P. brassicae.

Cell wall modification

In our study, arabinogalactan protein, which is related to cell wall modification, was largely induced by the addition of the biocontrol strain and was much higher than in the P and H samples. The fasciclin-like arabinogalactan protein is an important component of plant cell walls (Estevez et al. 2006) and is involved in cell adhesion and the plant’s detection of invading pathogens (Mellersh and Heath 2001; Hardham et al. 2007). In susceptible Arabidopsis infected by P. brassicae at 4 DAI, the arabinogalactan protein was down-regulated (Agarwal et al. 2011). In contrast, the arabinogalactan protein was up-regulated in the B + P sample and may strengthen cell walls and inhibit the invasion of P. brassicae.

Down-regulated genes

The gene encoding cytokinin synthase (Unigene13014) was down-regulated in the B + P sample compared with the P sample. Compared with H, the cytokinin synthase gene was up-regulated in the P sample, while it was down-regulated in the B + P sample. The cytokinin synthase gene plays a role at late time points during root-gall formation (Ludwig-Muller and Schuller 2008). The down-regulated expression of the cytokinin synthase gene results in the alleviation of the clubroot disease symptoms by biocontrol strain.

Conclusions

In our study, the genes and pathways involved in plant resistance were induced by the biocontrol strain. The transcriptome data explained the molecular mechanism of the potential biocontrol strain against P. brassicae. The data will also serve as an important public information platform to study B. juncea var. tumida Tsen and will be useful for breeding mustard plants resistant to P. brassicae.