Multi-environment QTL studies suggest a role for cysteine-rich protein kinase genes in quantitative resistance to blackleg disease in Brassica napus

Resistance to the blackleg disease of Brassica napus (canola/oilseed rape), caused by the hemibiotrophic fungal pathogen Leptosphaeria maculans, is determined by both race-specific resistance (R) genes and quantitative resistance loci (QTL), or adult-plant resistance (APR). While the introgression of R genes into breeding material is relatively simple, QTL are often detected sporadically, making them harder to capture in breeding programs. For the effective deployment of APR in crop varieties, resistance QTL need to have a reliable influence on phenotype in multiple environments and be well defined genetically to enable marker-assisted selection (MAS). Doubled-haploid populations produced from the susceptible B. napus variety Topas and APR varieties AG-Castle and AV-Sapphire were analysed for resistance to blackleg in two locations over 3 and 4 years, respectively. Three stable QTL were detected in each population, with two loci appearing to be common to both APR varieties. Physical delineation of three QTL regions was sufficient to identify candidate defense-related genes, including a cluster of cysteine-rich receptor-like kinases contained within a 49 gene QTL interval on chromosome A01. Individual L. maculans isolates were used to define the physical intervals for the race-specific R genes Rlm3 and Rlm4 and to identify QTL common to both field studies and the cotyledon resistance response. Through multi-environment QTL analysis we have identified and delineated four significant and stable QTL suitable for MAS of quantitative blackleg resistance in B. napus, and identified candidate genes which potentially play a role in quantitative defense responses to L. maculans.


Background
Prevention of catastrophic crop loss to plant pathogens is most often achieved through the incorporation of resistance genetics into commercial cultivars. Host responses to plant pathogens are broadly divided into two categories; basal defense responses induced by generic pathogen signals or elicitors called "pathogen-associated molecular patterns" (PAMPs), resulting in mild defense responses collectively known as 'PAMP triggered immunity' (PTI) and R gene mediated 'effector triggered immunity' (ETI) in which race-specific pathogen avirulence (Avr) proteins trigger robust defense mechanisms including hypersensitive response (HR) leading to host cell death at the site of infection [1]. When studying foliar plant pathogens, the HR response of race-specific R genes often provides a visual phenotype, indicating an incompatible interaction and allowing for the determination of pathogen virulence. This distinction is used to separate specific R gene interactions from quantitative resistance which can provide effective 'adult plant resistance' (APR) within a crop variety through the cumulative action of multiple resistance loci. APR is usually measured at the end of the growing season in field trials.
APR is particularly important for combating diseases of Brassica napus L. (canola/oilseed rape) in which R gene mediated resistance is lacking, such as Sclerotinia Stem Rot (Sclerotinia sclerotiorum) [2][3][4] and Verticillium Wilt (Verticillium longisporum) [5][6][7] or for diseases where pathogen populations often display a rapid adaptation towards R gene mediated resistance, such as in the case of blackleg disease, caused by the hemibiotrophic fungal pathogen Leptosphaeria maculans [8,9].
Avoidance of R gene mediated resistance by L. maculans can occur both rapidly and in a geographically localised fashion when a pathogen population is under heavy selection pressure. A rapid decline in the efficiency of the blackleg R gene Rlm1 in controlling the disease in Europe highlighted the evolutionary potential of the pathogen [10]. A high frequency of mutation and deletion of the L. maculans avirulence gene AvrLm4-7 was reported to occur within a small plot area sown continually to B. napus harbouring Rlm7, while virulent pathotypes remained undetectable in samples from the surrounding local pathogen population [11]. High rates of infection were observed in some areas of Australia in canola varieties carrying the R gene LepR3 only three years after first commercial release of the material [12], though this rapid loss of effective resistance may have been aided by pre-exposure to Rlm1 varieties, as avirulence towards LepR3 and Rlm1 is conferred by the same L. maculans avirulence gene; AvrLm1 [13].
B. napus cultivars containing only APR usually show no difference in the development of leaf lesions when compared with susceptible cultivars, yet they restrict the development of internal stem infection by the pathogen, resulting in lower levels of crown canker formation [14]. This is in contrast to R gene mediated resistance which leads to arrest of L. maculans growth at the site of infection on cotyledons and leaves. When major R gene mediated resistance is avoided by virulent strains within the mixed pathogen population, APR reduces the selection and proliferation of virulent pathotypes in crop residues and the potential for catastrophic crop loss in following seasons [15][16][17].
While R gene mediated resistance can often be detected efficiently and rapidly by observing hypersensitive response after inoculation of B. napus cotyledons with well-characterised L. maculans isolates, assessment of APR is much more difficult. Resistance needs to be measured either through field-based studies, or under controlled conditions through infection with single spore-derived L. maculans isolates and assessment of stem infection in plants grown for several months [18,19]. Assessment of APR in field-based studies can be difficult considering the complexity of plant-pathogenenvironment interactions. Populations of L. maculans in most disease nurseries are genetically heterogeneous mixtures arising from sexual recombination and variation of pathotypes should be expected both within a trial site and between trial years. Also, variation of host response due to heterozygosity of B. napus lines may be confused for polygenic control of resistance [20]. There has also been a widely-held view that blackleg APR is race non-specific [17], based largely on experience of the French variety Jet Neuf, which provided durable resistance to blackleg disease over many years in Europe and was also utilised in early efforts to improve blackleg resistance in Australian germplasm [21,22]. However, more recent studies utilising single L. maculans isolates have questioned the "race non-specific" nature of blackleg APR [19,23].
Maintenance of strong APR in canola varieties can most efficiently be achieved through marker-assisted breeding based on the molecular characterisation of quantitative trait loci (QTL) associated with resistance [17]. The French variety Darmor, derived from Jet Neuf, is the most extensively studied B. napus variety harbouring quantitative resistance to L. maculans. A doubled-haploid (DH) population produced from a cross between Darmor-bzh and the susceptible Korean cultivar Yudal (DY) was utilised to map 10 QTL contributing to blackleg resistance, with four of the QTL detected stably across two years of field testing [24]. The resistance was further analysed in Darmor x Samouraï (DS) DH and F 2 populations, revealing four QTL that were common to both the DY and DS populations [25]. Near-isogenic lines (NILs) were also produced for four Darmor QTL; LmA2, LmA9, LmC2 and LmC4, though only LmA2 was fully validated as having a significant effect on reducing disease severity [26].
Blackleg APR has also been assessed in several Australian varieties, revealing multiple QTL that are potentially common to several Australian and French cultivars [9,19,27].
Little is known about the molecular basis of APR to L. maculans infection in Brassica species. While two racespecific genes responsible for ETI-mediated blackleg resistance, LepR3 and Rlm2, have been cloned from B. napus and shown to encode extracellular leucine-rich repeat (eLRR) receptor-like proteins recognising the L. maculans effectors AvrLm1 and AvrLm2, respectively [13,28,29], no genes underpinning blackleg resistance QTL have been identified. Infection of B. napus by L. maculans results in attempted physical restriction of the pathogen by the host, via callose deposition, while an increased lignification response has also been reported for APR varieties [30,31]. L. maculans infection triggers induction of the salicylic acid (SA) signalling pathway [31,32] which plays a critical role in plant defense [33]. SA signalling can be triggered in B. napus by purified L. maculans cell wall components [34] and is greatly induced during ETI, along with the ethylene signalling pathway and H 2 O 2 accumulation [31,32,35]. However, these studies have all focused on early infection events in the cotyledons of B. napus seedlings; nothing is known about which defense mechanisms may be active against the invading hyphae as they grow asymptomatically through the petiole [18] and stem [36].
In this study we identified several stable blackleg resistance QTL, with resistance alleles derived from AG-Castle and AV-Sapphire, two blackleg-resistant Australian B. napus varieties released in 2002 and 2003, respectively. We used Topas/AG-Castle (TC) and Topas/AV-Sapphire (TS) DH populations to assess the APR of the varieties over multiple years at two locations, performed both single-and multi-environment QTL mapping and defined the physical locations of the QTL relative to the recently released B. napus Darmor-bzh reference genome [37], allowing for the identification of candidate defense-related genes.

Population data
Field tests were conducted in south eastern Australia in disease nurseries located near Horsham, Victoria and Wagga Wagga, New South Wales (Fig. 1 Figure S1). Heritability was calculated based on total entries in each environment for each scoring metric; survival (S) and internal infection (II) and was generally high, producing similar ranges for each metric (S: 0.75-0.9, II: 0.73-0.89) ( Table 1).

Linkage analysis
While the lack of heterozygosity in the populations showed the lines produced to be true doubled haploids, marker distortion was detected in many regions. 40 % of TC marker bins and 20.8 % of TS markers showed mild distortion (failed to conform to a 1:1 ratio; Chi-square test, p = 0.05 to 0.001), while 21 % of TC markers and 6 % of TS markers showed severely distorted segregation (p < 0.001). The severely affected regions were generally towards the ends of chromosomes; in the TC population they were the upper portions of chromosomes A02, C01, C03 and C09, and the lower portions of A05, C01 and C06. Additionally, the entire A07 chromosome was uniformly distorted, with all markers favoring the Topas parent allele in an approximately 3:1 ratio. In the TS population, upper C03, lower A08 and all of C02 were severely distorted. Two draft maps were initially produced for each population; one containing all markers, and a second in which all severely affected marker loci were removed. Whole-genome QTL analysis was performed using both draft map versions. No changes to QTL positions were observed between draft maps, nor were significant differences in QTL LOD and variance scores, with the exception of the Rlm3 locus, positioned on A07 on the 'distorted' TC map and absent from the 'non-distorted' map. To accommodate mapping of the A07 Rlm3 resistance locus and its associated QTL in the final analysis, TC A07 was resolved independently and added to final TC map. All other distorted markers were removed prior to final map construction. The final TC map consisted of 307 marker bins (collections of cosegregating markers) spanning 2182.3 cM in 21 linkage groups (LGs) representing all 19 B. napus chromosomes. Linkage mapping for the TS population produced 23 LGs (199 bins, 1714.97 cM), which were assigned to 18 of the 19 chromosomes, with no representation for chromosome C02.

Single-environment QTL
Permutation tests performed for each scoring metric, survival (S) and internal infection (II) percentage, in each environment determined significant LOD thresholds between 2.76 and 3.07 for the TC population and 2.66-2.95 for the TS population. Analysis of the singleenvironment data produced multiple QTL exceeding their respective LOD thresholds (LOD 3.37-41.47), which were localised to seven chromosomes for the TC population (A01, A08, C03, C04, C05, C06 and C07) and five chromosomes of the TS population (A01, A03, A09, C01 and C06), and accounting for between 2.48 and 31.77 % of the phenotypic variance (Additional file 2: Table S1). To identify chromosomal regions harbouring 'stable' and significant QTL regions, QTL identified from individual environments were only considered significant if they exceeded both the LOD threshold for each analysis, based on permutation test (1000 permutations, 0.05 error) and accounted for >5 % of the variance. After applying these criteria three significant QTL, each with favorable alleles derived from the respective resistance donor parent (AG-Castle or AV-Sapphire), were identified in each population. These regions consisted of clustered QTL located on chromosomes A01, A08 and C06 of AG-Castle, and A01, A09 and C06 of AV-Sapphire, with the A01 and C06 QTL regions appearing to be common to both resistance donor parent lines ( Table 2, Fig. 2). For the TC larger population (242 lines), the three clustered QTL regions were represented by QTL

Multi-environment QTL
Multi-Environment Trait (MET) analysis performed using S and II data for all environments produced MET QTL corresponding to each of the seven chromosomes previously identified from the single-environment analysis, except for TC A10 and C07. In addition, multiple lowvariance MET QTL were detected for the TS population.
Estimates of the additive genetic (A) and environmental (E) components of the multi-environment variance (σ 2 ) were used to calculate the narrow-sense heritability (h 2 ) of each MET QTL. As doubled-haploid populations were used for the study there was no heterozygosity and thus no dominance component for the variance. h 2 values between 0.59 and 0.88 were determined for the TC QTL, while slightly higher h 2 values (0.69-0.94) were determined for QTL in the smaller TS population (Additional file 2: Table S1). Multi-Environment Traits (MET) were considered significant, and the associated chromosomal regions harbouring MET QTL were considered to make a stable contribution to resistance, if they accounted for greater than 5 % of the total variance and were attributed a heritability score of 0.5 or higher. After applying these criteria only MET QTL associated with the previously-identified single-environment QTL clusters were retained. The total multi-environment survival (MET S) and internal infection (MET II) variance explained by the major QTL regions was 30.19 and 33.64 %, respectively, for the TC population, and 28.59 and 24.35 %, respectively, for the TS population ( Table 2).

Single-isolate characterisation
Each DH line in the TC and TS populations were characterised for the presence or absence of hypersensitive response via cotyledon infection tests, as differential phenotypic reactions were initially observed to the L. maculans isolates WA30 (avirulent on AG-Castle, virulent on AV-Sapphire) and v23.1.3 (virulent on AG-Castle, avirulent on AV-Sapphire). After the phenotypic data was converted to Mendelised resistant (+) and susceptible (−) scores and incorporated into the genetic maps for each population, the cotyledon resistance loci were determined to localise to chromosome A07, co-segregating with the simple sequence repeat (SSR) marker sR12173 in both cases (Fig. 2). As chromosome A07 is known to harbour the race-specific blackleg resistance genes Rlm1, Rlm3, Rlm4, Rlm7 and Rlm9 [19,[38][39][40] further characterisation of the parental lines with differential L. maculans isolates varying in their reactions to the A07 R genes was performed. Only isolates avirulent towards Rlm3 or Rlm4 produced resistant reactions on AG-Castle or AV-Sapphire, respectively (Table 3). Additional evidence for the presence of Rlm4 in AV-Sapphire was produced using the transgenic isolate 3R11: AvrLm4-7, which demonstrated the AvrLm4-7 gene conveys avirulence towards AV-Sapphire (Table 3, Additional file 3: Figure S2). The Rlm3 locus of the TC population co-segregated with a group of three SSR alleles; sR12294a, sR12173 and sR2834a (Fig. 2), which span a region of 80 genes of the B. napus Darmor-bzh reference genome [37] on chromosome A07 (BnaA07g20270D to BnaA07g21070D). Unfortunately the closest flanking markers to the Rlm3 cluster were the SSR markers sNRA59 and sR12829, both of which match to portions of the genomic sequence which are not currently incorporated into the B. napus chromosome A07 model (matches to 'chrUn_random' and 'chrA07_random' , respectively). The next-closest flanking markers, sR12119a (closest gene = BnaA07g17900D) and    [37].
The mapping of Rlm4 with the TS population placed the gene between the SSR markers sN2555Ra and sN2834a, a region spanning approximately 3.9 Mb and 704 genes on chromosome A07 (BnaA07g14030D-BnaA07g21070D) which overlapped the Rlm3 interval defined in the TC population.
During the phenotypic screening of the mapping populations with single isolates, some intermediate phenotypes (scores between 4.1 and 6.9 on the 0-9 scale) were observed. Therefore, to test for QTL affecting the expression of the cotyledon resistance phenotype when  Table 3 Determination of A07 blackleg R genes in AG-Castle and AV-Sapphire a L. maculans isolates used in this study; name followed by avirulence genes corresponding to Brassica A07 R genes carried by each isolate. 3R11: AvrLm1 and 3R11: AvrLm4-7 are transgenic isolates of 3R11 carrying additions of AvrLm1 and AvrLm4-7, respectively. Isolates in bold were used for TC and TS population cotyledon assays b B. napus lines used in the study; name followed by blackleg R gene content of line c Interaction of isolate and B. napus line; avr = virulence, Avr = avirulence, followed by mean cotyledon rating (in brackets) on 0-9 scale. Entries in red indicate a virulent interaction challenged with single isolates of L. maculans, the phenotypic data was again analysed, this time as continuous data (0 to 9) rather than the discrete "Mendelised" data ('+' or '−') analysed previously. As expected, large portions of the variance were associated with each major resistance gene locus (Rlm3 and Rlm4). We also detected a second significant QTL for the TC population which accounted for 9.26 % of the cotyledon phenotypic variance and co-localised with the TC C06 QTL cluster (Table 2, Fig. 2).

Delineation of QTL loci and identification of candidate genes
The 0.5 cM map interval containing the TC A01 MET S and MET II loci was flanked by the markers sN12790 and sN4638 (Fig. 2), which corresponds to a span of 49 B. napus genes (BnaA01g12170D-BnaA01g12660D). The peak MET LOD scores for each metric were flanked by the markers sR8420 and sN4638 which corresponded to a 10 gene interval of the Darmor-bzh reference B. napus genome (BnaA01g12560D-BnaA01g12660D). The TS A01 QTL locus was flanked by the markers sR9228a and sN12176, representing a span of 296 genes (BnaA01g09950D to BnaA01g12910D), with peak LODs for all QTL contained with the 4.2 cM marker interval sR6202b-sN12176 (BnaA01g10980D to BnaA01g 12910D), representing a span of 193 B. napus genes which also encapsulates the physical interval defined for the TC A01 QTL locus.
The TC population produced a cluster of single-and multi-environment QTL on chromosome A08, with peak values for all QTL positioned between 39 and 46 cM ( Table 2). The multi-environment analysis defined overlapping MET S and MET II QTL contained within a marker interval of 13.7 cM flanked by the markers sN4513Fa and sN12352a. This interval corresponds to a region of 396 genes in the Darmor-bzh reference genome (BnaA08g18290D-BnaA08g22250D).
The QTL cluster detected for the TS population on chromosome A09 was positioned within the marker interval sS2212 to sR9373, which corresponds to a physical interval of 1314 genes in the B. napus reference genome (BnaA09g22470D-BnaA09g35610D). While the peak LOD for many of the QTL, including both the MET S and MET II QTL, was positioned within the smaller sR6410-sR9373 marker interval, the physical region could not be refined any further using the B. napus reference genome, as sR6410 was assigned to the 'chrA09_random' molecule which is not incorporated into the main A09 chromosome build. However, by using the B. rapa Chiifu A genome reference sequence [41], we identified homology between sR6410 and the B. rapa A09 gene Bra006927 and used a neighbouring gene (Bra006925) to determine an approximate B. napus physical position for this gene as equivalent to BnaA09g32910D of B. napus. This produced a physical delineation for the A09 QTL peak interval of approximately 270 genes (BnaA09g32910D-BnaA09g35610D).
QTL clusters were also detected on chromosome C06 in each population, though poor resolution of these linkage groups hampered precise delineation of the support intervals. The TS C06 QTL support interval was delimited to a 2998 gene span of the B. napus genome (BnaC06g05190D-BnaC06g35170D) between the markers sS2486 and sN5088F while in the TC population, a B. napus physical interval could not be defined as the upper flanking marker for the TC C06 support interval (DArT marker brPb-841355) provide a match to the unincorporated 'chrUn_random' molecule. We were, however, able to define a syntenic physical interval in the B. oleracea reference genome [42] of approximately 9 Mb on C06 (1,797,307..10,864,498), which contains 1336 predicted genes.
Delineation of the TC A01, TC A08 and TS A09 QTL loci produced physical intervals sufficiently small to warrant identification of gene candidates that had potential roles in underpinning pathogen resistance QTL. The TC A01 QTL peak LOD interval spanned only ten genes, which straddled a cluster of eight genes (BnaA01g12580D-12590D, BnaA01g12610D-12630D, BnaA01g12650D-12670D) with homology to the Cysteine-rich Receptor-Like Protein Kinase (CRK) genes of A. thaliana chromosome 4 (At4g23190, At4g23300 and At4g04570, respectively). CRKs are one of the largest super-families of receptor kinases in Arabidopsis with 44 members [43], several of which have been implicated in plant defense responses [44][45][46][47][48][49]. The TC A08 and TS A09 peak QTL intervals span 396 and 270 genes within the reference B. napus genome, respectively. Within these spans are several potential resistance-related genes including receptorlike proteins, a receptor-like kinase and TIR-NB-LRR homologues.

Discussion
We describe here the detection and characterisation of highly-stable quantitative resistance loci to the B. napus fungal pathogen L. maculans. Through multi-environment analysis we were able to define highly-heritable resistance loci effective in some of the harshest testing conditions in the world and to identify putative resistance-related genes that are located within the physically-defined QTL regions.
In performing the QTL tests over multiple environments we were able to produce estimates of heritability (h 2 : the degree to which genetics determines phenotype) for each MET QTL (Table 2). Traditionally heritability is calculated as a function of the variance within the entire population (Table 1). While this provides an estimate of the genetic influence on the over-all phenotypic variance, this does not provide information on the environmental variability of individual QTL loci within the population. By calculating h 2 values for each MET QTL we can offer an estimate of how environmental variability will affect the phenotypic variation, particularly when targeting individual QTL loci in breeding programs. We also observed that in the larger TC population (242 DH lines), the h 2 values were consistently higher when MET QTL were determined using the survival metric compared to the internal infection metric, though this did not hold true for the smaller TS population (109 lines). This may be due to the influence of 'escapes' in the scoring of the field trials. While significant differences were seen in survival between the 'susceptible' Topas and 'resistant' parental cultivars AG-Castle and AV-Sapphire, the difference was not always evident with the internal infection metric (Additional file 4: Table S2, Additional file 1: Figure S1). Under Australian field conditions, infection is driven by sexual ascospores and often results in seedling death for susceptible plants [50]. This means very few susceptible plants will remain standing at the end of the growing season when internal infection is assessed. However, the remaining survivors have been enriched for escapes i.e. plants that did not develop the same level of disease due to a delay in, or lack of, exposure to the pathogen, particularly when seedling germination is not consistent. This would result in an under-estimate of internal infection for individual lines, as only a portion of the standing plants are assessed. The same effect would not be as significant in the survival metric where the entire row is counted.
The genomic location of the TC A01 QTL interval matches the previously reported position of a blackleg QTL from the Australian cultivar AG-Spectrum [19]. A recent report detailing QTL mapping in European winter oilseed rape populations [51] placed QTL from Grizzly/Bristol and Darmor/Bristol populations within a region of A01 that spans the TC A01 QTL locus defined in our study (approx. BnaA01g08200D-BnaA01g20140D). QTL on chromosome A01 have also previously been reported for several DH populations derived from Australian varieties, including AV-Sapphire [27]. Favourable alleles from the AG-Castle A08 QTL locus (BnaA08g 18290D-BnaA08g22250D) are positioned adjacent to the QlmA8_DB QTLs previously identified in the Darmor/ Bristol population, which were positioned between the SSR markers BN53449 and sR3688 (BnaA08g12480D-BnaA08g17050D) on chromosome A08 [51], and to the DY A08 QTL detected in the Darmor-bzh/Yudal population [26,52]. Additionally, the TS A09 QTL locus from AV-Sapphire provides a near-complete overlap with the previously-reported syntenic A. thaliana region At3g25805-At3g58680 (equivalent to approximately BnaA09g19610D to BnaA09g37480D of the B. napus genome), which was identified as syntenic to the LmA9 QTL interval from the Darmor-bzh/Yudal QTL map [26]. Finally, QTLs from Aviso and Darmor/Bristol populations were also placed within a 369 gene interval (BnaC06g31460D-BnaC06g 35150D) on lower C06 [51], which is in agreement with the large C06 QTL regions detected in both the TC and TS populations. The correlation of blackleg QTL in many of the studied B. napus varieties suggests the overall pool of APR genetics utilised in canola varieties world-wide may be rather limited.
Pedigree analysis for AG-Castle suggests nearly half of the variety's genetic contribution (46.9 %) is derived from Japanese material, with European material making up the bulk of the remainder (29.8 %) [53]. A study of B. napus germplasm diversity has shown the variety AG-Spectrum is closely related to Rainbow [54], an Australian polygenic resistant variety which is also featured in the pedigree for AG-Castle [53]. The B. napus variety Major, a progenitor of the well-characterised French APR variety Darmor (Major -> Primor -> Jet Neuf -> Darmor) was grouped into a different clade than Rainbow and AG-Spectrum [54], suggesting a low over-all genomic relationship between sources of APR for the Australian and European QTL studies. However, the correlation of QTL on A01, A08, A09 and C06 between Australian and French varieties demonstrates a high degree of selection and retention of these QTL after the introduction of French APR germplasm into Australian breeding programs [22], suggesting enduring efficacy of these QTL against Australian L. maculans populations. However, a "slow erosion of polygenic resistance" has been observed in Australian breeding programs [53] and suggests that future efforts should be focused on the detection and introgression of novel APR genetics from diverse Brassica germplasm rather than the continued recycling of overused QTL. Interestingly, the B. juncea line BJ168 also makes a small contribution (6.3 %) to the pedigree of AG-Castle and many other Australian varieties, though none of the major R genes from B. juncea have been introgressed into Australian cultivars [53]. These R genes, present in the B genome of B. juncea, are potentially valuable sources of resistance for B. napus breeding if stably introgressed [55][56][57]. However, little is known of the quantitative resistance potential of the B. juncea A genome, which harbours distinct genomic diversity [58], or if B. juncea A genome introgressions have contributed to the pool of blackleg APR utilised in Australian germplasm, and this remains a potentially under-utilised resource for resistance genetics.
Investigation of the genomic region defined by the TC A01 peak QTL interval revealed a cluster of cysteine-rich receptor-like kinase genes (CRKs). CRKs are characterised by one or more extracellular C-X8-C-X2-C motifs (DUF26/GNK2) that likely mediate protein-protein interactions [43]. CRK genes have been shown to be induced in plants during pathogen infection [44][45][46][47][48][49], including the upregulation of several CRKs during infection of B. napus cotyledons by L. maculans [32]. Active early in the plant defence response to pathogens, CRKs are induced independent of other host response genes such as EDS1 or NPR1 [44,45]. While some members of the CRK superfamily have been demonstrated to enhance resistance against pathogens by promoting hypersensitive response (HR) and salicylic acid (SA) accumulation [44,45], the use of knock-out mutants and transient gene silencing has suggested that some other CRKs can negatively affect defense responses, acting in a quantitative fashion [47,48]. CRKs have also been shown to have a role in symbiotic plant-microbe interactions, with the symCRK gene of the legume Medicago truncatula being required for maintenance of functional nodules and the suppression of host defense during the symbiotic process [59]. Recently it was demonstrated that over-expression of some Arabidopsis CRKs enhanced PTI defense responses, such as oxidative burst, stomatal closure and callose deposition, towards the bacterial pathogen Pseudomonas syringae pv. tomato DC3000. Additionally, these proteins were shown to associate with the pattern-recognition receptor FLAGELLIN SENSING2 (FLS2) at the plasma membrane [60], an essential component of the bacterial pathogen PAMP response [61]. A second study, profiling a near-complete set of Arabidopsis CRK mutants, showed some CRKs respond to fungal chitin, a common PTI elicitor [62,63], and play a role in regulating stomatal closure [64]. The roles of CRKs in quantitative plant defense responses suggests the cluster of CRK genes colocalised with the TC A01 QTL interval offer viable candidate genes for blackleg APR. Further study of these genes, including their temporal and spatial regulation in response to L. maculans infection, is currently underway. While the other QTL intervals defined in this study were not small enough to warrant extensive candidate gene studies, other CRK homologues were found with the TS A09 (BnaA09g 23770D) and TS C06 (BnaC06g31680) intervals.
While no stable QTL were detected in which the resistant allele originated from the Topas "susceptible" parent, it should be noted that in most environments Topas had a higher survival percentage than the commonly-used susceptible line Westar (Additional file 4: Table S2). Under the disease nursery conditions in Australia, the Westar rows showed near-complete seedling death, often leaving no plants at all at the end of the growing season, while the Topas rows still produced some adult plants, though most were dead or highly cankered at the time of rating. It is conceivable that Topas contains some small degree of APR common to both AG-Castle and AV-Sapphire, and perhaps many other B. napus cultivars, which would not be detectable as QTL in these populations. While some Topas-derived QTL were detected during the study, these were always specific to a single testing environment, usually accounted for less than 5 % of the variance and were not associated with any significant MET QTL (Additional file 2: Table S1). Under the extremely high disease pressure of the Horsham 2012 trial, both Topas and Westar had 0 % survival (Additional file 4: Table S2, Additional file 1: Figure S1), so any minor APR that may have been expressed by Topas was rendered completely ineffective under those conditions.
Our single-isolate characterisation of the TC and TS populations also delimited the physical positions of the major R genes Rlm3 and Rlm4 to intervals of 3.5 Mb and 3.9 Mb, respectively, on chromosome A07. While these genes have been known to cluster genetically, along with Rlm7 and Rlm9, on A07 for over a decade [38], this is the first report to define a physical location for this cluster in the B. napus genome. Previously, Raman et al. (2012c) positioned the Rlm4 gene (from B. napus variety Skipton), assessed at both the cotyledon and adult plant stages, to a 24.8 cM interval of A07 using SNP markers. While the physical interval relative to B. napus was not defined in this study, several SNPs corresponding to approximately 15.9-16.9 Mb (~BnaA07g20250D-BnaA07g21900D) on B. napus A07 were associated with the Rlm4 locus [65]. This location would be in agreement with the Rlm4 physical interval defined for the TS population in this study (BnaA07g14030D-BnaA07g21070D) and the cosegregation of the locus with the SSR marker sR12173 (BnaA07g20910D).
Neither Rlm3 nor Rlm4 were found to be associated with QTL detected in the field studies, suggesting a high degree of virulence towards both R genes in the populations of L. maculans infesting the disease nurseries. This result was not unexpected, as high rates of virulence towards Rlm3 [66] and Rlm4 [23] have been previously reported for Australian L. maculans isolates. Analysing the cotyledon reactions to single L. maculans isolates avirulent towards the respective R genes produced high LOD, high variance 'CotQTL' loci associated with each gene, as expected. An additional cotyledon-responsive interaction with the isolate WA30 was also demonstrated in the TC population which accounted for 9.26 % of the cotyledon phenotypic variation and localised to the C06 QTL interval (Fig. 2, Table 2). No additional significant cotyledon-responsive QTL were detected in the TS population, which also contains a C06 QTL locus, using isolate v23.1.3. Previous studies have observed a significant difference in L. maculans cotyledon lesion development between APR and non-APR varieties [18]. Cotyledon-expressed QTL similar to the one detected in our study have also previously been defined in the SASDH (Skipton/ AG-Spectrum) population and localised on chromosomes A01 and A10, though only the A01 QTL was associated with field resistance in the adult plant. These cotyledon QTL interactions also appeared to be race-specific, having been detected with the L. maculans isolate 06MGPP041 but not isolate 04MGPS021 [19]. Blackleg resistance at seedling and adult stages of the host have often been viewed as separate mechanisms; the former governed by racespecific R gene interaction in cotyledons and leaves, and the latter by non-specific QTL active in the stem. The detection of common QTL loci in both seedling (cotyledon) and adult plant (field) tests, and the differential interaction of individual L. maculans isolates to those QTL, suggests that such distinctions are not always valid. The associated resistance response is likely triggered early in the interaction between host and pathogen, activated long before entry of the pathogen into the stem tissues of the plant. The action of these QTL may be to impair the initial infection process, rather than to actively resist invasion of adult stem tissues, leading to delayed development of stem canker symptoms in the adult plant. If the quantitative blackleg resistance found in B. napus is truly "nonrace specific" then the entirety of the environmental variance observed in field trials can be attributed to differences in the physical environmental factors (rainfall, temperature etc.) affecting each trial. However, if L. maculans isolates react differentially to QTL, then a large component of the environmental variance could be explained by variation in the pathogen populations used in the trials, both between sites and years. Alternatively, as it has been previously shown that B. napus blackleg R genes require other partner proteins in order to successful trigger the resistance response to L. maculans [29,67], minor QTL detected during cotyledon tests may represent either allelic variation or altered transcriptional patterns for R gene interacting proteins. Further dissection of the temporal and spatial effects of quantitative resistance responses during L. maculans infection, as well as the effectiveness of these responses against a number of diverse isolates, will aid in bettering our overall understanding of the blackleg resistance response and how best to utilise the available genetics in crop protection.

Conclusions
As pathogen populations become enriched for virulence against commonly-deployed R genes, alternative resistance genetics need to be incorporated into crop breeding programs in order to maintain global agricultural production. The identification and mapping of QTL is required for the efficient introgression of stable resistance genetics in modern MAS breeding programs, enabling the effective utilisation of quantitative genetics in plant protection. In this study we have delineated four stable blackleg resistance QTL through multi-environment QTL analysis. We have identified a cluster of cysteine-rich receptor-like kinase genes associated with a significant field-derived QTL, suggesting a role for these genes in quantitative defense responses to L. maculans. The genomic delineation of effective blackleg resistance QTL described in this work should benefit resistance breeding efforts, and further the investigation into the identity of genes controlling APR in Brassica germplasm.

Methods
Plant populations F 1 seedlings, produced by crossing the Swedish springtype, blackleg-susceptible B. napus variety Topas with the Australian blackleg-resistant varieties AG-Castle and AV-Sapphire, were used to produce DH populations via microspore culture by the method of Ferrie et al. [68] using modified Lichter medium (NLN) with 13 % sucrose. The Topas/AG-Castle DH population (TC) utilised for this study consisted of a total of 242 DH lines from two sub-populations of 139 and 103 DH lines, with each sub-population produced from a single F 1 plant. The Topas/AV-Sapphire DH population (TS) contained a total of 109 lines, also consisting of two sub-populations of 66 and 43 DH lines. Seed of parental and DH lines used in the QTL trials was produced in tented field plots in Saskatoon, Canada during the 2007 (TS) and 2008 (TC) summer growing seasons, packaged into replicates of 100 seeds per line and sent to Australia for testing field resistance to blackleg. The Topas B. napus line was sourced from the AAFC Saskatoon collection. AG-Castle and AV-Sapphire lines were provided by DEDJTR Victoria (formerly Department of Primary Industries).

Field characterisation of DH populations
The parental lines and their respective DH populations, along with several B. napus variety checks (Additional file 4: Table S2) were grown May through November in blackleg disease nurseries situated at Wagga Wagga (New South Wales) and two sites (Wonwondah and Green Lake) south of Horsham (Victoria) between 2008 and 2012 (Fig. 1). Rainfall differed by an average of 103.2 mm/year over the three trial years at each site, with the Wagga Wagga site receiving an average 38 % more precipitation (Additional file 5: Table S3). Each trial was planted in a randomised block design with 3 to 4 replicates per line. The TC population was tested in three separate environments; both Wagga Wagga and Horsham (Wonwondah site) in 2009 (3 replicates per trial) and once more at Wagga Wagga in 2010 (4 replicates). The TS population was tested in four environments; Horsham (Wonwondah site), 2008 and Wagga Wagga, 2009 as 3-replicate trials and Wagga Wagga, 2011 and Horsham (Green Lake site), 2012 as 4replicate trials. Emergence counts were made 4-5 weeks after sowing. For trials at the Horsham sites, local diseased canola stubble, collected in the previous year, was spread as the source of L. maculans inoculum, while at Wagga Wagga the trials were sown into in situ canola stubble from the preceding season as previously described [19]. At the end of the growing season trials were assessed for blackleg disease using both survival percent (S) and internal infection (II) metrics. S was calculated for each row as (Standing plants at end of season/Emergence count)*100, II determined by examining 5-20 random plants per row for internal infection after cutting the stem at the crown and rating fungal discoloration on a 0-5 scale (0 = no infection, 1 = 1-25 % infection, 2 = 26-50 % infection, 3 = 51-75 % infection and 4 = 76-99 % infection and 5 = 100 % infection/missing plant) and median scores were calculated for each entry and metric using Microsoft Excel. Statistical analysis of populations and plotting of distribution histograms were performed using R software [69].
Selected lines representative of the 3 major QTL identified from the TC DH population in this study, both singularly and in combination, were also entered into Saskatchewan disease nurseries at Melfort (2012) and Saskatoon (2013) in attempt to test the efficacy of the QTL resistance against Canadian populations of L. maculans. However, neither site produced sufficient disease pressure to assess the resistance, with Westar control entries presenting average internal infections of less than 25 % at the end of the growing season.

Single-isolate characterisation
Each DH line in the TC and TS populations were characterised for the presence or absence of hypersensitive response during infection by the single spore-derived L. maculans isolates WA30 (Rimmer Collection, AAFC Saskatoon) or v23.1.3 (INRA-Bioger collection, France), respectively, via cotyledon infection tests. Four plants for each line were scored at 14 days post-infection on a 0-9 scale as described by Larkan et al. [13]. Further differential pathology testing, using isolates of L. maculans varying in their avirulence gene profiles (00-100, 98-15, 99-53, 89-12, 2354-Rimmer Collection, AAFC Saskatoon; 3R11-Howlett Lab, University of Melbourne, Australia, 3R11: AvrLm1 [13]) was performed with the parental lines and the R gene control lines Topas DH16516 (a DH line of Topas with no effective blackleg resistance), Columbus (Rlm1, Rlm3), Quantum (Rlm3), Jet Neuf (Rlm4), Roxet (Rlm7) and Goeland (Rlm9) in order to produce a hypothesis for the likely R gene content of each population (Table 3). Inoculum for the tests was produced as described previously [70]. Spores were diluted to 2x10 7 spores/mL and 10 μL was used to inoculate each of four small wounds on each plant (one wound at the centre of each cotyledon lobe).

Map construction and QTL identification
The TC and TS populations were characterised using 225 and 212 simple sequence repeat (SSR) markers, respectively, as described previously [73]. A subset of the TC population had previously been characterised using 279 DArT markers as part of a consensus map of B. napus [74], and this data was also incorporated into the TC map to provide additional resolution and to allow orientation to other maps (503 markers total, grouped into 360 bins). Linkage map construction and inclusive composite interval mapping (ICIM) [75] was performed using QTL IciMapping version 3.3 (http://www.isbreeding. net). Prior to final map construction severely distorted markers (Chi-square Test, p < 0.001) were excluded. A threshold of LOD 6 was applied for initial grouping of markers.
LGs consisting of > 3 markers were anchored to chromosomes based on BLAST-like alignment tool (BLAT) analysis of SSR marker sequences to the published B. napus genome [37] before recalculating LGs at LOD 3. Marker ordering was performed using nearest neighbour algorithm and rippling was performed once using the sum of adjacent recombination fractions (SARF) criterion. For QTL identification, the 'ICIM-ADD' and 'ICIM-EPI' functions of the software were utilised to investigate both additive and digenic epistatic QTL [76,77] expressed in each environment and for each scoring metric (S and II). Permutation tests (1000 permutations, 95 % confidence level, 1 cM interval) were performed for each scoring metric in each environment to determine significant LOD thresholds. The precision of QTL positions were then improved by reanalysing the data using the determined LOD thresholds and a smaller (0.5 cM) scan interval. The 'Multi-Environment Trials' (MET) function of the software was also utilised to determine the consensus positions for the major QTL and to assess the genotype by environment (GxE) interactions. QTL identified from individual environments were considered significant if they exceeded the LOD significance threshold and accounted for > 5 % of the variance. MET QTL were considered significant if they accounted for > 5 % of the variance and had a heritability (h 2 = σ 2 (Additive)/σ 2 (Total)) > 0.5. QTLs positions were defined by both support and peak marker intervals, with