Pak choy (B. rapa subsp. chinensis) cv. “Flower Nabana” (FN), a hybrid developed by Evergreen Y.H. Enterprises, Anaheim, CA, USA, is highly resistant to pathotype 3 of P. brassicae. Rcr1 was identified and genetically mapped in 1587 plants from a F₁ population that segregated for resistance (R) and susceptibility (S) [11]. This population was derived from a cross between FN, which is heterozygous at Rcr1 locus, and canola line ACDC, which is a clubroot-susceptible B. rapa doubled haploid developed at the Saskatoon Research Centre of Agriculture and Agri-Food Canada. An additional 200 plants in the F₁ population were self-pollinated, but most of the plants did not produce seed due to self-incompatibility in B. rapa. Seed was obtained from 38 plants, resulting in 38 F₂ lines that could be evaluated for resistance to multiple pathotypes of P. brassicae. These lines were tested for resistance to pathotypes 2, 3, 5 and 6 under controlled conditions at the University of Guelph, ON, Canada. Plant seedlings were grown in soil-less mix in tall narrow plastic pots (164 mL conetainers, Stuewe & Sons Inc., Corvallis, OR, USA) and inoculated with 5 ml of resting spore suspension (1×10⁶ spores ml⁻¹) at 10 days after seeding. Roots of 7-14 plants per line were assessed for clubroot severity at 6 weeks after inoculation using a standard 0 to 3 scale where: 0 = no clubbing; 1 = small clubs only; 2 = moderate clubs; and 3 = severe clubbing. A disease severity index (DSI) [63, 64] was calculated using the following formula: DSI(%)=∑(ratingclass)×(#plantsinratingclass)total#plantsintreatment×3×100

The highly susceptible Shanghai pak choy cv. “Mei Qing Choi” (Stokes Seeds, ON, Canada) was included as a susceptible control for each group to ensure that the inoculation was highly effective. Evaluation of the F₂ lines and controls for resistance to the pathotypes of P. brassicae was replicated two times.

At 15-days post-inoculation, root tissue from 9 R plants was combined to form the R bulk and from 9 S plants to form the S bulk; together, the two bulks comprised one biological replicate as described by Chu et al. 2014 [11]. In total, three replicates were assessed. The total RNA from each replicate was isolated using an RNeasy Plant Mini Kit (Qiagen; Toronto, ON, Canada) with oncolumn deoxyribonuclease (DNase) digestion using a Qiagen RNase-Free DNase set following the manufacturer’s instruction. The cDNA library was prepared using TruSeq RNA Sample Preparation Kits v2 (Illumina; San Diego, CA, USA). RNA-seq was carried out using a Illumina Hiseq 2500 platform (Plant Biotechnology Institute, National Research Council, Saskatoon, SK, Canada) and global gene expression by the RNA-seq was conducted using the B. rapa reference genome v1.2 (B. rapa subsp. pekinensis cv. “Chiifu”) [11]. In the current study, the short reads from the RNA-seq were aligned to the B. rapa reference genome v1.5, which was downloaded at http://brassicadb.org/brad/downloadOverview.php. The reference genome consists of 10 chromosomes and 40,357 scaffolds. The total lengths of chromosomes and scaffolds are about 258 million bases (Mb) and 27 Mb, equivalent to about 90% and 10% of the reference genome, respectively. To simplify downstream data analysis, only the 258 Mb chromosome sequences were used in the current study, so the short reads from the RNA-seq project were assembled into 10 chromosomes A01 to A10 of B. rapa. The program SeqMan NGen 11 (DNASTAR, Madison, WI, USA) was used for short read assembly. Two methods were used: 1) short reads from three biological replicates each of the R and S bulks were assembled into the reference genome, resulting in six assembly files—hereafter, this method is called the single sample assembly (SSA); 2) short reads from a pool of the three R bulks and a pool of the three S bulks were assembled into the reference genome, resulting in two assembly files—hereafter called the pooled sample assembly (PSA). Standard assembling and filtering parameters were used.

Discovery of variants (SNPs and InDels) in comparison with the DNA sequences in the B. rapa “Chiifu” [65] were performed using SeqMan Pro 11 (DNASTAR, Madison, WI, USA) with Q call ≥ 15 and depth ≥ 5. Analysis of variance and t-tests (LSD) were used to assess the variation in SNPs and InDels [66].

Selected SNPs identified in the target region were confirmed using the Kompetitive Allele Specific PCR (KASP) method (http://www.lgcgroup.com/) following the manufacture’s instruction. PCR reactions were performed using a StepOne Plus Real Time PCR System (Applied Biosystem, Mississauga, ON, Canada). FN8, a line derived from the B. rapa cv. FN and homozygous at the Rcr1 locus, was included for this study.

To validate the SNP markers for MAS selection with Rcr1, DNA samples from 38 F₁ plants from the cross between FN and the susceptible B. rapa line ACDC and 26 recombinants from the B. rapa population that was originally used to identify Rcr1 [11] were assessed. The accuracy of the SNP markers for use in MAS was determined by testing the recombinants from a BC₁ population derived from B. napus introgressed with Rcr1 described by Chu et al, 2014 [11].

Rcr1 was initially identified based on resistance to pathotype 3 (Williams’ differentials) [41] of P. brassica, the most prevalent pathotype identified in western Canada [11]. The resistant donor FN was found to be resistant to multiple pathotyes of P. brassica [43]. To determine if resistance to other pathotypes in the R donor cv. FN was also associated with Rcr1, we tested 38 F₂ lines derived from the F₁ segregating population for resistance to pathotypes 2, 3, 5 and 6. As expected, FN conferred resistance to pathotype 3 (0 DSI), but ACDC (95 DSI) and “Mei Qing Choi” (100 DSI) were highly susceptible (Fig 1). Eight of the 38 lines (F₂-6, 18, 19, 24, 29, 30, 34 and 41) were highly susceptible to pathotype 3 (severity > 80 DSI) so their parental F₁ plants likely did not carry Rcr1. Severity was low to moderate in the other 30 lines (0-52 DSI), which indicates that the parental F₁ plants for these lines likely carried the Rcr1 gene. To further confirm the F₁ phenotypes determined by F₂ lines, DNA samples from the original 38 F₁ plants were assessed using two microsatellite markers, MS7-9 and sN8591, that flanked Rcr1 [11]. The 30 F₁ parental plants of the F₂ lines with low to moderate DSI carried the allele associated with resistance; the 8 F₁ plants with high DSIs in their F₂ lines carried the allele associated with susceptibility (Fig 1), confirming the phenotypes in the 38 F₁ plants determined by DSIs in the F₂ lines.

The clubroot reaction of the parental lines and the F₂ lines to pathotypes 2, 5 and 6 was also assessed. The parental cv. FN was highly resistance to pathotypes 2, 5 and 6, but ACDC and “Mei Qing Choi” were highly susceptible (Fig 1). The eight F₂ lines that were susceptible to pathotype 3 were also highly susceptible to pathotypes 2, 5 and 6 (> 80 DSI). In contrast, clubroot severity was low to moderate (0-55 DSI) in the other 30 lines (Fig 1). Therefore, we conclude that resistance to pathotypes 2, 5 and 6 of P. brassicae was associated with resistance to pathotype 3 that was used for identification of Rcr1 in the 38 lines.

The average sequence counts and average accumulated sequence length were 125.9 million (M) and 11,093.8 Mb per R bulk, and 114.0 M and 10,005.8 Mb per S bulk using the SSA method (Table 1). This provided an average depth of coverage of the reference genome of 43 fold in R bulks and 39 fold in S bulks. The average sequence length assembled into the reference genome was 88 bases. More sequences were assembled into the longer chromosomes A03 and A09, while fewer sequences were assembled into the shorter chromosomes A04 and A10 (Table 1). The sequence counts assembled into the genome for each chromosome were highly correlated to chromosome length (r = 0.90) for both R and S bulks.

Short reads from three R bulks and three S bulks were further assembled using the PSA method. As observed previously with the SSA method, more sequences were aligned into the longer chromosomes A03 and A09 and fewer sequences were aligned into the short chromosome A04 and A10 (Table 1). A total of 351.8 M sequences, 30,836.5 Mb in length, with coverage of 120 fold of the reference genome were assembled into B. rapa chromosomes from the pool of three R bulks, and 322.9 M sequences, 28,216.9 Mb in length, with 109 fold coverage were assembled from the pool of three S bulks (Table 1). About 50% of the aligned sequences were aligned to the top strand of the reference DNA and the other 50% to the bottom strand using the either the SSA or PSA methods (S1 Table).

Variants in both R and S samples were very frequent, with means of about 605.9 K SNPs and 82.2 K InDels per R bulk and 579.4 K SNPs and 76.3 K InDels per S bulk by using the SSA method. There was a strong positive correlation between the numbers of SNPs compared to InDels (r = 0.99 in both R and S bulks). The numbers of SNPs and InDels varied among the chromosomes (Fig 2), indicating that the longer chromosomes (A03, A09) carried more variants than the shorter chromosomes (A04, A10). However, the proportion of SNPs and InDels was similar among chromosomes, with about 88% and 12%, respectively, in both R and S bulks (S1 Fig).

The numbers of SNPs and InDels identified using the PSA method were higher than with the SSA method, with 776.2 K SNPs and 122.2 K InDels in the R bulk, and 762.8 K SNPs and 118.7 K InDels in the S bulk. As with the SSA method, there was a strong positive correlation between the numbers of SNPs and InDels with the correlation coefficient (r = 0.99) in either R or S bulks. The proportion of SNPs and InDels on each chromosome was 87% and 13% in both the R and S bulks (S1 Fig), which is a slightly lower proportion of SNP and higher proportion of InDel identified using the PSA method than the SSA method. As observed with the SSA method, the numbers of SNPs and InDels identified using PSA method (Fig 2) were higher on the longer chromosomes.

Variants identified in the R and S samples could be the same (monomorphic, mono) or different (polymorphic, poly) (Fig 3). Mono variants comprised 73% (range 63.7-75.6%) of the variants identified across the B. rapa genome using the SSA method, and poly variants 27% (range 24.4%-36.3%). There were no differences among chromosomes in the proportion of mono and poly variants except for chromosome A03, where the clubroot resistance gene Rcr1 is located, which carried more ploy variants and fewer mono variants (64%, P ≤ 0.05) than the other chromosomes (Fig 4).

With the PSA method, the proportion of mono variants was higher than for the SSA method for each chromosome (Fig 4). Overall, approximately 5% of the poly variants identified by SSA method were identified as mono using the PSA method. On chromosome A03, there were 66% mono variants and 34% poly variants, very similar to the results from SSA (Fig 4).

A previous study reported that Rcr1 was mapped in an interval of 0.28 centi-Morgan on B. rapa chromosome A03 and located in the 24.26-24.54 Mb of A03 [11]. In the current study, the target region was first analyzed using the SSA method. There were 35 genes annotated in the region, based on the reference genome v1.5 (Table 2). Three genes (Bra038750, Bra019396 and Bra019394) did not show any expression and no short reads were assembled into the reference genome, so no variants could be identified. Short reads with sequence counts < 20 and number of variants < 1 was associated with 11 genes (Bra038751, Bra038747, Bra019411, Bra019408, Bra019404, Bra019402, Bra019401, Bra019399, Bra019397, Bra019393 and Bra019392). More than 20 sequence counts with > 1 variants were associated with 21 genes in both R and S bulks (Table 2). Information on BLASTX (best hit) to A. thaliana and gene ontology annotation for the genes was obtained from http://brassicadb.org/brad/index.php (S2 Table). Four genes (Bra019409, Bra019410, Bra019412 and Bra019413) encode TNL-class disease resistance proteins.

The numbers of poly variants between the R and S bulks among the genes in the coding regions was assessed because the poly variants represent differences in the DNA sequences between the R and S bulks. The numbers of poly variants differed significantly among the 21 genes (Table 3).Three TNL genes (Bra019409, Bra019410, and Bra019413) had the most poly variants, with means of 81.7, 75.0 and 34.7 poly variants per gene, respectively, followed by four genes (Bra019406, Bra038754, Bra019390 and Bra038757) with means of 33.3, 31.3, 29.0 and 24.0. These seven genes exhibited more poly variants than the other 14 genes (P ≤ 0.05). The fourth TNL-class gene (Bra019412), with mean of 4.0 poly variants, was one of the 14 genes that showed few polymorphic variants (Table 3).

The PSA method identified more poly variants among the 21 genes in the target region than the SSA method (Table 3). For both methods, three TNL genes (Bra019409, Bra019410, and Bra019413) consistently carried the highest number of poly variants among the 21 genes.

The poly variants that uniquely occurred in the R bulks but not in the S bulks (Fig 3) with depth > 5 in both samples were further assessed using the PSA method. A total of 108 poly variants were identified from the coding sequences in the four TNL genes (S3 Table). SeqMan Pro software was used to sort the 108 variants that affect amino acid sequences into four groups: non-synonymous, nonsense, frameshift and synonymous variants. Non-synonymous variants occurred in each of the TNL genes, but Bra019409 and Bra019410 carried many more non-synonymous variants than the other two TNL genes (Fig 5). Three nonsense variants were identified in Bra019409 and one in Bra019410, but none were identified in the other two genes. Frameshift variants caused by InDels were found in Bra019409 and Bra019413, and synonymous variants occurred in each of the genes except Bra019412 (Fig 5). Overall, Bra019409 and Bra019410 carried higher number of poly variants that uniquely occurred in the R samples and could affect the amino acid sequence in their proteins.

The DNA sequences flanking 11 SNPs in the four TNL genes (S3 Table) and three SNPs (SNP_A03_18, 19, and 32) in the gene Bra019406 that was one of genes harboring more poly variants (Table 3) were obtained from the B. rapa reference genome (S4 Table). These SNP markers, which spanned about 42 Kb in the target region (Table 4), were genotyped using the KASP method. As expected, each SNP marker was polymorphic between the R parental line FN8 and the S parental line ACDC (Fig 6), which confirmed that a SNP was present at each locus. In an allelic discrimination plot, PCR products from the FN8 (R) were usually close to the right bottom and ACDC (S) to the left top (Fig 6). This indicates that they carried homozygous alleles that might be associated with R and S, respectively. In contrast, when DNA samples from the 38 F₁ plants (Fig 1) were analyzed using the 14 SNP markers, the PCR products from 30 R plants were in the middle (Fig 6). This indicates that these plants were heterozygous at this locus (Table 4), with one allele from the R parent and the other one from the S parent. PCR products from the eight S plants were close to the right top (Fig 6), where the PCR product from the susceptible line ACDC was located, indicating that they carried homozygous alleles from the S parent (Table 4). These results confirm that the 14 SNP markers identified by RNA-seq were completely associated with Rcr1 on chromosome A03 in this population.

To determine if the SNP markers were located in the target region, we analyzed 26 recombinants, including LP9-1, LP5-12, LP5-71 and LP5-183 that were identified by the closest flanking markers, MS7-9 and sN8591 from 1587 F₁ plants with five SNP markers (SNP_A03_08, 09, 12, 16, and 32) that reside in the genes Bra019413, Bra019412, Bra019410, Bra019409 and Bra019406 respectively (Fig 6). Each SNP marker co-segregated with phenotypes and no recombination event was found in these plants (Fig 7). This indicates that the SNP markers were completely associated with the clubroot resistance gene Rcr1.

To assess the potential use of the SNP markers in MAS for introgression of Rcr1 into canola, DNA samples were assessed from seven recombinants in the BC₁ populations consisting of 176 plants, which were previously used for validating Rcr1 linked to microsatellite markers for MAS [11], with the 14 SNP markers (Table 4). Five SNP markers (SNP_A03_16, 18, 19, 32, 103) were polymorphic in the B. rapa population; however, they were monomorphic in the B. napus population. Polymorphism was identified in the rest of 9 SNP markers (Table 5). The SNP alleles from the R parent occurred in 3 R plants (BN-4, 27, and 58) and the alleles from the S parent occurred in 4 S plants (BN-1, 86, 103, and 113) (Fig 6; Table 5). The accuracy for MAS in the BC₁ population using the 9 KASP SNP markers was 100%, and the SNP markers completely co-segregated with the phenotypes.