Background

The Next-Generation Sequencing (NGS) technologies are being extensively used for genome-wide genetic marker development through RNA-seq, reduced-representation sequencing, restriction-site-associated DNA sequencing (RAD-seq) and low-coverage genotyping [1]. Availability of abundant markers will facilitate association mapping, marker aided selection (MAS), and fine mapping of regions of interest for circumventing the problem of linkage drag during introgressions and for map based cloning.

NGS technologies have also contributed to completion of reference genome sequences of many important crops [2, 3]. Availability of reference genomes will facilitate characterization of variability within a crop and its wild relatives by high throughput re-sequencing. In the family Brassicaceae, model species Arabidopsis thaliana was sequenced by the Sanger method using aligned overlapping BACs [4]. Using NGS technologies, a large number of ecotypes have been sequenced in a much shorter span of time [5]. Eventually 1001 ecotypes will be sequenced.

The first crop species sequenced from Brassicaceae is Brassica rapa (2n = 20, AA genome) [6]. The assembled sequence of 283.8 Mb covers more than 98% of the gene space. Sequencing was carried out using Illumina GA II technology. Sequence data was integrated with BAC-end sequences obtained through the Sanger sequencing method. Sequencing work was carried out on line Chiifu, a leafy vegetable type of B. rapa belonging to ssp. pekinensis. The genome sequence of Chiifu is available on BRAD, a genomic database created for B. rapa and other Brassica species [7, 8].

Genus Brassica contains some of the most important vegetable and oleiferous crops of the world. The relationship of the six crop species namely, B. rapa (AA, 2n = 20), B. nigra (BB, 2n = 16), B. oleracea (CC, 2n = 18), B. juncea (AABB, 2n = 36), B. napus (AACC, 2n = 38), B. carinata (BBCC, 2n = 34) was first described by U [9] and later confirmed by others using molecular markers [10]. Crop Brassicas display a range of morphotypes, which include vegetable types where root, leaves, stems and inflorescence have been modified for human consumption, oilseed types and condiment types, all selected under domestication [11]. Three of the species namely B. juncea, B. napus and B. carinata are recent allopolyploids with the full chromosome compliment of the two parental genomes. The three diploids are paleohexaploids with extensive chromosomal rearrangements, gene subfunctionalization and loss [6, 12, 13]. Comparative genomic studies have shown that the gene blocks identified in A. lyrata and A. thaliana are represented at least three times in B. rapa, although every gene in a block is not necessarily represented by three paralogs [12–15].

B. rapa as a species shows enormous morphological variability, containing both vegetable types and oilseed types, and has extensive geographical distribution [11, 16]. Different morphotypes have been classified under subspecies (ssp.). There is however, no consensus on the relationship of different types [17]. A recent classification has recognised 10 ssp. in B. rapa[18]. The most extensive study to date on variability within B. rapa, conducted on 161 accessions with AFLP markers, has shown that oleiferous types of a region are closer to the vegetable types of that region rather than to the oleiferous types of the other regions, thereby implying independent domestication of the oleiferous types in many regions [16].

In the present study we have carried out RNA-seq of three different oleiferous lines of B. rapa namely, YSPB-24 and Tetra (both belonging to the Yellow sarson group, ssp. trilocularis), Candle (turnip rape, ssp. oleifera) and a vegetable type line Chiifu using Illumina GA II technology to find out if sufficient numbers of SNPs are available for genome-wide mapping and for fine mapping in specific regions of the genome. YSPB-24 has a typical bilocular ovary and Tetra is an interesting variant in the Yellow sarson group as it has a tetralocular ovary. The most probable region of origin of the Yellow sarson lines is Eastern India. These lines are extensively grown in this region. Candle is an oilseed line of European origin. Therefore, the study includes two closely related oleiferous lines (YSPB-24 and Tetra), which are distant from the oleiferous line Candle. All the three lines are divergent from the leafy vegetable type line Chiifu that has been sequenced recently.

RNA-seq can provide the most informative SSRs and SNPs for gene synteny based comparative genomics [19–22] and association mapping. A number of programs have been developed for SNP identification from the NGS data [23, 24]. Marking SNPs in the single-copy genes has been worked out reasonably well. However, in paleoploid species like B. rapa we require SNPs not only for marking allelic variation but also for marking the paralogs as has been the case for more recent allopolyploid species like wheat or B. napus[25–28].

We report that a sufficient number of SNPs are available in the B. rapa single-copy genes for genome-wide mapping and in the paralogs to mark both the allelic differences and paralog specific differences to saturate a specific region with unique marker probes. This strategy for genome-wide mapping and specific-region fine mapping has been tested using a F7-RIL population of a Chiifu x Tetra cross using KASPar oligo technology [29]. A total of 542 SNPs have been mapped on the 10 linkage groups of B. rapa and 52 SNPs have been mapped in a region of LG A4, which we have recently shown to contain the tet-o locus that encodes the trait - tetralocular ovary [unpublished].

Results

Plant material and sequencing

Four Brassica rapa lines – YSPB-24 and Tetralocular (Yellow sarson types, ssp. trilocularis), Candle (ssp. oleifera) and Chiifu (ssp. pekinensis) were used for RNA sequencing. Inflorescence, with all the unopened flower buds, along with a few small leaves was used for RNA extraction. This amalgam of organs represents most of the aerial tissues of a Brassica plant. Paired end cDNA libraries, used for sequencing, were prepared from the poly-A containing RNA and sequenced as 2x101 nt reads on the Illumina GAIIX sequencer.

Sequencing samples of Tetra, YSPB-24 and Candle were run in two lanes of the flow cell for sequencing whereas Chiifu sample was run in only one lane. As information on more than 98% of the gene space is available for Chiifu and has been organized as CDS in the BRAD database [8], the limited Chiifu transcriptome sequencing was carried out to check the overall quality of the assembly of the raw sequencing data obtained in this study.

Data filtering and de-novoassembly of the transcriptome

Paired end sequencing of transcriptome generated 84,458,126, 117,128,230, 145,049,468 and 154,228,832 sequence reads for Chiifu, Tetralocular, YSPB-24 and Candle, respectively (Table 1), the lowest being for Chiifu for which the data was obtained from single lane. After filtering low-quality and single-end reads, assembly of the cleaned reads was carried out using the Velvet de-novo assembly program [30] with default settings except that the minimum contig length was set at 100 bp as this length would be useful for designing oligos for SNP analysis. Reads were assembled for different K-mer values (K21 to K57) and the obtained data were analyzed for the total number of contigs, percentage of reads assembled, N50 values and the average contig length. Best assembly was found at K-mer value of 47 in case of Chiifu, Tetra and Candle and at K-mer value of 51 in case of YSPB-24 (Figure 1). The number of contigs obtained for the four B. rapa lines ranged from 38,220 for Tetra to 69,636 for Candle (Table 1). Approximately 61–78 percent of the reads could be assembled into contigs with N50 values ranging from 515 to 1094 bp. The maximum contig size obtained for Chiifu was 8956 bp whereas for the other three lines it was more than 10 kb. The mean depth of the line specific assemblies was found to be in the range 76.3 to 181 (Table 1).

Table 1 Sequencing and assembly statistics of four different lines of B. rapa
Full size table
Figure 1
figure 1

Comparison of total contig number, average contig length and N50-length, obtained after the Velvet assembly. Figures A, B, C and D represent the contig assembly results of B. rapa line Chiifu, Tetra, YSPB-24 and Candle, respectively. The bars indicate the total number of contigs assembled (primary axis). The green line represents the N50 contig length while the red line indicates the average contig length.

Full size image

Sequence comparison of YSPB-24, Tetra and Candle with Chiifu

We compared the Chiifu transcriptome sequence obtained in this study with the Chiifu gene models described in the BRAD database [8]. A total of 43,110 out of 44,972 contigs obtained in this study could be aligned with the CDS sequences reported in the BRAD database using the parameter of minimum 96% identity, indicating quality assembly of the raw reads in this study. However, the assembled contigs of Chiifu and the three oleiferous type lines were predominantly partial sequences (ESTs). Therefore, in the rest of the study the contigs obtained from the three oleiferous lines – YSBP-24, Tetra and Candle were compared with the more detailed full-length CDS (predicted gene models) available for the line Chiifu in the BRAD database.

Assembled sequences of each of the three oleiferous lines showing ≥ 80% sequence identity in ≥ 100 bp sequence stretch with any of the predicted gene models of Chiifu were considered to be homologous sequences. All the contigs showing identity with a Chiifu gene model as per the criteria described above were considered to be a part of the gene model and collectively referred to as a homolog. Around 90% contigs of Tetra, 88.2% contigs of YSPB-24 and 89.2% contigs of Candle matched with one or the other gene model of Chiifu listed in the BRAD database. The BRAD database lists 17,562 single-copy gene models for Chiifu in the syntenic paralog data (http://brassicadb.org/brad/searchSynteny.php). We could identify 13,808, 9,081 and 8,143 homologs in the lines Tetra, YSPB-24 and Candle, respectively (Table 2). Homologs have been grouped-‘Chiifu vs YSPB-24’, ‘Chiifu vs Tetra’ and ‘Chiifu vs. Candle’ and these have been described in the Additional file 1. Hitherto, gene models and their RNA-seq based homologs will be referred to as genes.

Table 2 Homology based grouping of contigs of three oleiferous B. rapa lines with Chiifu
Full size table

SSR identification

The MISA program [31] was used to identify SSRs in the Chiifu gene models available in the BRAD database and in the homolog-contigs of the three oleiferous lines. The minimum cut off for the identification of mono-, di-, tri-, tetra-, penta- and hexanucleotide SSRs was set at 10, 6, 5, 5, 5 and 5 repeats, respectively. SSRs were found in 4327, 8308, 6160 and 5296 genes of Chiifu and in their homologs in Tetra, YSPB-24 and Candle, respectively. Mono-, bi- and tri-nucleotide motifs were found to be the most abundant in the three different lines (Table 3). We have recorded more mono- and di- SSRs in the contigs of Candle, YSPB-24 and Tetra as compared to the SSRs present in the gene models described in the BRAD database for Chiifu. This could be due to the presence of UTR sequences in the RNA-seq data while the CDS in BRAD database contains only those sequences that are translated into a protein. When different lines were compared in silico for the identification of polymorphic SSRs using the stringent criteria of complete identity in 50 bp flanks on either side of the repeat motif, most of the SSRs were found to be monomorphic. The maximum number of polymorphic SSRs (238) were found between Candle and YSPB-24 and the minimum number of polymorphic SSRs (92) were identified between Tetra and YSPB-24. The number of polymorphic SSRs identified between all the four lines in various combinations is shown in Figure 2.

Table 3 Number of repeat motifs identified in the gene models of Chiifu and transcriptome sequences of the three lines of B. rapa
Full size table
Figure 2
figure 2

Polymorphic SSRs between different Brassica rapa lines. Venn diagram showing the number of polymorphic SSR markers available between four different lines of B. rapa.

Full size image

Identification of SNPs between different lines of B. rapa

Two different programs Maq [32] and MUMmer [33] were used separately to identify the single nucleotide variations between different lines of B. rapa. For the Maq based SNP identification between Chiifu and the other three lines, the Chiifu CDS sequence was taken as the reference and the short reads of each of the three oleiferous lines (Additional file 1) were assembled independently on the reference. The SNPs thus obtained were filtered using SNPfilter script and SNPs with less than read depth of 7 and a quality score of less than 40 were discarded. This yielded 240,424, 277,237 and 346,189 SNPs for ‘Chiifu vs. Tetra’, ‘Chiifu vs. YSPB-24’ and ‘Chiifu vs. Candle’, respectively (Table 4a). The identified SNPs were sorted based on whether these were present in single-copy genes, or in genes with two or three paralogs.

Table 4 Number of SNPs identified between Chiifu and three lines of B. rapa using Maq and MUMmer tools
Full size table

When homologs of different B. rapa lines (Additional file 1) were analysed using the MUMmer tool, 249,671, 231,259 and 266,349 SNPs were identified between ‘Chiifu and Tetra’, ‘Chiifu and YSPB-24’ and ‘Chiifu and Candle’, respectively (Table 4b). These were further categorized on the basis of their being present in single-copy genes, or in genes with two or three paralogs.

For the identification of SNPs between lines other than Chiifu, homologs were compared in pair wise combinations of ‘Tetra vs. YSPB-24’, ‘Tetra vs. Candle’ and ‘YSPB-24 vs. Candle’ using both Maq (Table 4a) and MUMmer programs (Table 4b). Very high SNP frequencies were found for Candle vs. the two Yellow sarson lines. The least amount of polymorphism (20,310 SNPs with the MUMmer program) was recorded between Tetra and YSPB-24, the two closely related lines.

For the purpose of genome-wide mapping studies, one SNP is enough for marking a gene. We, therefore, carried out pair-wise analysis of SNPs in single-copy genes and in genes with two or three paralogs of all the four lines (Table 4). Except for ‘YSPB-24 vs. Tetra’ all others comparisons showed that 60-70% of the single-copy genes contained at least one SNP (Figure 3). For the two closely related lines YSPB-24 and Tetra, only 2,557 homologs could be identified with at least a single SNP - marking 14% of the single-copy genes, 9.6% of the two-gene paralogs and 7.2% of the three-gene paralogs. This number is sufficient for genome-wide linkage analysis but may turn out to be insufficient for fine mapping in a specific region.

Figure 3
figure 3

Percentage of genes that were marked with SNPs. Percentage of single-copy genes (first column), genes with two paralogs (second column) and genes with three paralogs (third column), which could be tagged with at least one SNP marker in pair-wise comparison of different lines of B. rapa. The least number of SNPs were found in Tetra vs. YSPB-24 comparison.

Full size image

In general, the number of SNPs recoded when the three oleiferous lines were compared with each other was lesser than the number recorded when the assembled contigs were compared with the full-length Chiifu CDS sequences available in the BRAD database (Table 4). As the contigs generated through Velvet assembly in the study were partial sequences, increasing the coverage could provide more SNPs for fine mapping.

Marker development from the identified SNPs

A survey of the BRAD database showed that 42% of the gene models have been reported as single-copy genes and these are well distributed throughout the B. rapa genome (Additional file 1). Developing markers from the single-copy genes for genome-wide linkage analysis was therefore considered to be the ideal strategy.

We used Chiifu vs. Tetra polymorphism data for SNP-marker development as these could be tested on a F7-RIL population of Chiifu x Tetra developed in our laboratory. SNPs were identified using the following criteria

  • SNPs should be surrounded with a conserved flanking region of ≥50 bp on both sides. This length of sequences would allow flexibility in positioning the oligos for amplification.

  • The region of 101 bp, containing the variable base at the middle, should not have any exon-intron junctions.

  • Read-depth of each of the identified SNPs should be ≥7.

A total of 6,451 single-copy genes that showed ≥ 90% sequence identity between Chiifu and Tetra were compared for SNPs. Applying the first criteria of 50 bp conserved sequences around the SNP, the number of useful contigs was reduced to 4,990. The number got reduced to 2,836 when sequences containing the intron/exon junctions were removed. Further application of the read-depth criteria reduced the number of contigs to 2,113, which could be used for marker development. To test our selection criteria for SNP marking, a set of 580 sequences, all representing single-copy genes that are well distributed throughout the genome were selected for developing SNP assays.

For specific-region mapping, the region around the tet-o locus was targeted. This locus has been earlier mapped (unpublished) on LG4 (A4) in a region that contains the S and I blocks of this linkage group. Sixty genes from this area were selected for developing SNP markers of which 42 were single-copy genes and 18 had two or three paralogs present in the genome. For such multi-copy genes, the paralogs were aligned and both the allele-specific SNP and paralog-specific variations (PSVs) were marked. Allele specific SNPs were selected wherein the 50 bp on either side of the SNP had some paralog specific variations (PSVs), so that locus specific primers could be designed for marking the specific paralog of interest.

Oligos for SNP genotyping were synthesized by KASPar technology [http://www.kbioscience.co.uk] using FRET quencher oligos competitor allele specific arrays. A total of 640 SNP based markers were developed between Chiifu and Tetra lines. The sequence of the marker oligos are provided in Additional file 2.

Linkage mapping in Chiifu x Tetra F7RIL population

SNP marker assays were tested on a Chiifu x Tetra F7-RIL population of 93 individuals. Out of 640 SNPs selected from well-dispersed genes, successful assays could be obtained with 613 SNPs using KASPar genotyping technology. Twelve of the markers showed significant segregation distortion and four were found to be monomorphic. A total of 733 markers including 594 SNP markers generated in this study, 99 Intron Polymorphism (IP) markers, 39 SSR markers and one morphological marker (tet-o) were mapped onto the 10 linkage groups corresponding to the haploid chromosome number of B. rapa (Table 5). The assignment of names to the linkage groups was based on the earlier mapping work [15]. The map covered a total genetic length of 679.7 centiMorgans (cM). The markers were distributed over 653 intervals and the distance among consecutive markers ranged from 0.8 cM to 1.5 cM with an average distance of 1.1 cM. The use of well-dispersed SNP markers allowed excellent general coverage of the linkage groups. The number of markers varied from 45 (linkage group A8) to a maximum of 108 (linkage group A9) SNP markers. A linkage map based predominantly on the SNP markers is given in Figure 4.

Table 5 Characteristics of 10 linkage groups of B. rapa map constructed with 594 SNP and 138 IP and SSR markers
Full size table
Figure 4
figure 4

Linkage map for B. rapa developed from F 7 -RILs with SNP, IP and SSR markers. A total of 594 SNP, 39 SSR and 99 IP markers were used for the development of the linkage map. Linkage groups are named A1-A10. Markers are shown on the right of the linkage group bar and marker positions (cM) are on the left. SNP markers are with the prefix BC_CT_.

Full size image

We further tested the use of SNP markers for fine mapping of a specific-region containing the locus tet-o. Of the 60 SNP markers developed for fine mapping in the region containing the tet-o locus, 52 could be successfully mapped with a mean marker interval of 0.5 cM. All the 18 markers designed for the genes with paralogs, marking both the SNP and PSV information for allele and paralog discrimination could be mapped without any ambiguity. The detailed map of the region containing the tet-o locus is shown in Figure 5.

Figure 5
figure 5

Fine map of the tet-o locus. SNP markers mapped in the region of tet-o locus of B. rapa CTF7 population. A total of 52 SNP markers were mapped in the region. The map on the left was developed with the SNP markers for genome-wide mapping, the tet-o region specific markers and their map distances in cM are shown in the map on the right side.

Full size image

We have earlier developed a linkage map of three different mapping populations in B. rapa – Chiifu x Tetra RIL-F6, Chiifu x Tetra F2 and Chiifu x YSPB-24 F2 using IP (genic) and SSR makers (mostly non-genic). These maps have now been integrated with the SNP map. A comparison of the features of the four parental maps, i.e. CTF2, CTF6, CYF2 and CTF7 is given in Additional file 3. IP and SSR markers genotyped on CTF7 in this study have been mapped previously in all these three mapping populations (unpublished) and these were used as anchor markers for developing the integrated map.

The integrated map has a total of 1,036 markers (211 IP, 230 SSR and 594 SNP) and a morphological marker ‘tet-o’. The features of the integrated map have been described in Additional file 4. The total genetic length spanned by the 10 linkage groups of the integrated map was 831.0 cM. The new integrated map is shown in Additional file 5.

Discussion

Although NGS technologies can be used in a variety of ways for mapping, we have opted in this study to use them for testing whether RNA-seq could provide adequate genic SSRs and SNPs for both genome-wide linkage analysis and for fine mapping of a specific region. Our results show that the number of polymorphic genic SSRs is rather limited, but abundant SNPs are available between the distantly related (Chiifu and Tetra, Chiifu and YSPB-24 and Chiifu and Candle) lines. The extent of polymorphism is low between the two closely related lines, YSPB-24 and Tetra. In general, the numbers of SNPs available are adequate both for genome-wide mapping and specific-region fine mapping.

A number of techniques are available for marking SNPs [34]. These have been broadly classified as allele specific hybridization, allele-specific single-base extension and allele specific enzymatic cleavage. Different methods have been developed for detection of allele specific products and a number of technology platforms have been developed for allele based sequence determination. All the technologies have been used in one or the other study of SNP based mapping in plants. We chose KASPar technology as it seems to be the most appropriate technology for the most frequently encountered mapping situations in crop genetics – (a) relatively small populations and a reasonable number of markers for genome-wide linkage mapping and (b) large populations and a small number of markers for specific-region fine mapping. Markers for 640 SNPs were developed and 594 could be successfully mapped. The technology also allows development of SNP markers that can differentiate paralogs and allelic differences in one reaction.

Considerable difficulty has been encountered in marking allelic SNPs in allotetraploid species like Triticum aestivum (wheat) [28, 35, 36], Gossypium hirsutum (cotton) [37] and B. napus (rapeseed) [27] due to the presence of homeologous chromosomes [26]. B. rapa, in contrast to the three species mentioned above, is an ancient paleoploid with three genomes that have gone through extensive gene fractionation and chromosomal rearrangements. At the genomic level there has been extensive gene loss, leading to many genes being present as single-copy genes besides some having two or three paralogs [6].

The data available for Chiifu in the BRAD CDS database and NGS based RNA-seq carried out in this study show that a very large proportion of genes present in B. rapa (42% in Chiifu) exist as single-copy genes and the nucleotide polymorphism that exists between the single-copy genes of various lines is sufficient for genome-wide as well as specific-region fine mapping. Sufficient nucleotide level polymorphism also exists between the paralogs to mark these through PSVs. In our study, > 92% of identified SNPs could be converted to successful assays using the KASPar technology. Our success with KASPar markers is higher than what has been reported in wheat using this technology [36]. The possible reasons for this could be – (a) a reference genome is available in B. rapa, (b) the frequency of single-copy genes in the genome is very high and (c) more stringent criteria were used in this study for developing SNP detection assays.

Use of molecular markers for genetic mapping in B. rapa began with an extensive use of RFLP markers [38]. Use of AFLP and SRAP markers provided more extensive marker densities [39, 40]. These markers, though abundant, are anonymous and do not provide any information on genomic synteny and therefore, are difficult to use for fine-mapping. Markers obtained from the gene space of a species, are most informative. EST-cDNA probes [14, 41], genic SSRs [42, 43], intron polymorphism (IP) markers [15, 43] and InDels (Sequence Tagged Sites, STS) [44], have all been successfully used for comparative mapping and studying genomic synteny amongst the Brassica species belonging to the U’s triangle [9].

The large number of SNPs available in B. rapa will allow more involved genome-wide linkage mapping and also association studies. A core set of 168 B. rapa lines has been identified and used in a genome wide association study using predominately AFLP markers [45]. It should be possible to use SNPs in future genome-wide association studies. However, we expect that major emphasis in Brassica species will be on mapping specific traits like disease [46–49] and pest resistance and QTL for yield [50–52] through the use of populations derived from biparental crosses. This is evident from extensive work that has been carried out on mapping of such traits using biparental crosses. Precise introgressions while avoiding linkage drag, particularly when unadapted germplasm is used, is going to be the key to improvement of a large number of vegetable and oleiferous crops available within the Brassica species belonging to the U’s triangle. The ready availability of SNPs for both background selection (general markers) and precise introgression (specific-area markers) will help in introduction from unadapted to adapted germplasm and from the diploid species to allotetraploid species. Specific-area marking and mapping of genes, as has been shown in this study for the region containing the tet-o locus, will also be useful for map based cloning.

Methods

Plant material, RNA extraction and library preparation

The four Brassica rapa lines - YSPB-24, Tetralocular (Yellow sarson types, ssp. trilocularis, seeds procured from Indian Agriculture Research Institute, India) Candle (ssp. oleifera, seeds procured from Gerhard Rakow, Agri-Food, Canada), Chiifu (ssp. pekinensis, seeds procured from Lim Yong Pyo, Chungnam National University, South Korea) used for transcriptome analysis were grown in the field during the mustard growing season (October – March). Tissues for RNA isolation were taken from the field-grown plants at the time of flowering.

Inflorescence with unopened flower buds along with a few small leaves, was used for RNA extraction. Harvested tissues were immediately frozen in liquid nitrogen. Total RNA was isolated using Total RNA Spectrum Kit (Sigma), following the manufacturer’s instructions. Contaminating DNA was removed by DNase treatment (DNaseA Kit, Ambion). RNA was further purified by treatment (thrice) with acidic phenol: chloroform (1:1). RNA was quantified using Nanodrop ND1000 spectrophotometer (Nanodrop Technologies). Integrity of the obtained RNA samples was checked on Agilent 2100 Bio analyzer. RNA samples with RIN value ≥ 7 were used for further experiments.

Paired end cDNA libraries, used for sequencing, were prepared from 20 μg of total RNA using predominantly the reagents available in the mRNA-seq Sample Preparation Kit (Illumina). mRNA was isolated from the total RNA with magnetic oligo (dT) beads. Purified mRNA was fragmented by treatment with divalent cations for 5 min (solution is provided in the kit). The obtained mRNA fragments were transcribed into first strand cDNA using Superscript II reverse transcriptase (Invitrogen), followed by second-strand cDNA synthesis using DNA polymerase and RNaseH. Double stranded cDNA molecules were purified by a QIAquick PCR purification kit (Qiagen). End repair of the double-stranded cDNA was carried out using T4 DNA polymerase, the Klenow DNA polymerase and T4 polynucleotide kinase. DNA was purified again using a QIAquick PCR purification kit. This was followed by a single ‘A’ base addition at the 3’ end of the double stranded cDNA molecules using Klenow 3’ to 5’ exo-polymerase followed by purification of modified cDNA molecules using a MinElute PCR purification kit (Qiagen). Sequencing adaptors were ligated to the ‘A’ tailed fragments using T4 ligase. Adaptor ligated cDNA fragments were separated on 2% agarose gel and fragments ranging in size from 200–250 bp were excised from the gel and purified using a QIAquick Gel Extraction Kit. PCR was performed for 15 cycles from the purified DNA molecules using the adaptor specific primers (available in the Illumina kit) and the amplified fragments were purified using a QIAquick PCR Purification Kit. The size and quantity of the obtained enriched cDNA libraries was checked on an Agilent 2100 Bio analyzer. Libraries with single discrete band of 200–250 bp were selected for sequencing reactions. The libraries were sequenced as 2 x 101 nt paired end reads on the Genome Analyzer IIx instrument (GAIIx, Illumina).

Data filtering and de-novoassembly of the transcriptome

Data obtained from the sequencer were processed for image deconvolution and quality value calculation of each base using the CASAVA package [Version 1.6, Illumina]. Quality check was done using Fastx-toolkit [53]. Reads with more than 30% bases having a Phred quality score < 20 were removed from the analysis using fastq_quality_filter (−q 20, -p 70). Thirty-one bases of the tail region, which were found to have low Phred score, were removed from the obtained sequences using Fastx-quality_trimmer (−t 31).

The assembly of the filtered reads was done using Velvet de-novo assembly program with the velvetg main parameters: -ins_length_sd 20 -ins_length (variable, depending on the insert length) -read_trkg yes -min_contig_lgth 100 -scaffolding yes -alignments yes. Obtained contig sequences are available on request.

SNP marker development

For marker development SNPs were mined from the EST contigs of Tetra generated in this study and the reference gene model sequence data of Chiifu available in the BRAD database. For identifying SNPs in the single-copy genes, the reference sequence dataset of 17,572 single-copy gene sequences was compared against corresponding Tetra contigs identified as single-copy genes. Sequences showing ≥ 90% homology were compared for SNPs using the MUMmer software. After removing sequences with insertions/deletions, sequences with SNPs and 50 bp conserved flanking sequence on both sides of the SNP were excised from the contigs using custom perl scripts. For identifying allele specific SNPs in genes with two/three paralogs, the Tetra contigs were compared against the reference sequence of Chiifu by conducting a BLAST search [54]. The paralogs were separated based on the homology and paralog specific variations (PSVs). The allelic variations were marked using custom PERL scripts and SNPs with 50 bp on each side of SNP were excised as before. PSV containing sequences were selected to design paralog specific primers. As the SNPs were identified using EST sequences, BLAST search was conducted with all the excised 101 bp sequences containing SNPs against the B. rapa genome sequences of Chiifu available in the BRAD database (Brapla_genome_data_v2.1) to identify intron/exon junctions. Selected SNPs showing the presence of such boundaries in the flanking regions were discarded, as the oligo designing at such positions would result in assay failure.

The cut-off read depth for the identified SNP was set at ≥ 7. For this, a Velvet assembly generated .afg file was converted to .ace file and further pileup file was generated using Samtools [55]. The depth of each of the base (A, T, G and C) was calculated using custom perl scripts.

SNP genotyping and construction of linkage map

For validation and mapping of SNPs, KBioscience KASPar assay [29] was used. The primer design and assay development was undertaken by KBiosciences [29]. The ID and features of identified putative SNPs along with 100 bp sequences are provided in the Additional file 2.

For genotyping the SNPs markers, a Recombinant Inbred Line (RIL) population of 93 individuals in the F7 generation derived from a cross between the leafy vegetable type line Chiifu and the oleiferous line Tetra was used. From the already developed map of B. rapa using F6 -RIL population of Chiifu and Tetra in our lab (unpublished), randomly selected 99 IP and 39 SSR markers, well dispersed on all the 10 linkage groups were also genotyped on F7-RILS to develop a skeleton map. PCR reactions and product separation for IP and SSR markers were performed as described earlier [15]. Linkage groups were established at a LOD > 6.0 with Join Map 4.0 [56] following the mapping criteria of Pradhan et al. [57]. The recombination fractions were transformed to map distances with the Kosambi function [58]. The graphical representation of the linkage groups was generated by Map Chart 2.2 [59].

For the construction of an integrated map, marker information from three mapping populations CTF2 (Chiifu x Tetra F2), CTF6 (Chiifu x Tetra F6) and CYF2 (Chiifu x YSPB-24 F2) [unpublished] along with the genotyped SNP markers was used. Recombination fraction data from the CTF2, CTF6 and CYF2 were merged with the recombination data of CTF7 map and heterogeneity tests were performed for pairs of markers common to the four mapping populations. A set of markers from each linkage group was identified on the basis of the order among the component maps (CTF2, CTF6, CTF7 and CYF2) and was used to define a fixed order for the construction of an integrated map. Common marker pairs that differed significantly (p < 0.01) in the recombination frequencies were excluded from mapping. An integrated map was generated by Join Map 4.0 with the combine groups for a map integration function using the regression mapping algorithm.

Conclusion

RNA-seq of three agronomically interesting oleiferous lines of B. rapa using paired end sequencing provided a large number of SNPs for both genome-wide mapping and for fine mapping in specific areas of the genome. High confidence SNPs in homologs were selected for KASPar based genetic marker development by three-step selection criteria. High frequency (~96%) of markers could be mapped successfully in the B. rapa genome. KASPar technology can be effectively used for marking allelic SNPs and PSVs for marking paralogs in B. rapa lines, and can be used for marking genes with multiple paralogs in any genome. Also, availability of large number of SNPs will allow fine mapping of regions of interest.

Author’s contributions

KP carried out all the sequencing reactions and the bioinformatics work. VG helped with the development of the SNP markers. SKY, PP-M and AKP carried out the mapping work. YSS developed and maintained the mapping population. DP initiated the study and wrote the manuscript with inputs from all the authors. All authors read and approved the final manuscript.