Isolate Dependency of Brassica rapa Resistance QTLs to Botrytis cinerea

Generalist necrotrophic pathogens including Botrytis cinerea cause significant yield and financial losses on Brassica crops. However, there is little knowledge about the mechanisms underlying the complex interactions encoded by both host and pathogen genomes in this interaction. This potentially includes multiple layers of plant defense and pathogen virulence mechanisms that could complicate in breeding broad spectrum resistance within Brassica species. Glucosinolates (GSLs) are a diverse group of defense metabolites that play a key role in interaction between Brassica and biotic attackers. In this study, we utilized a collection of diverse B. cinerea isolates to investigate resistance within the Brassica rapa R500 × IMB211 recombinant inbred line population. We tested variation on lesion development and glucosinolate accumulation in parental lines and all population lines. We then mapped quantitative trait loci (QTL) for both resistances to B. cinerea and defense metabolites in this population. Phenotypic analysis and QTL mapping demonstrate that the genetic basis of resistance to B. cinerea in B. rapa is isolate specific and polygenic with transgressive segregation that both parents contribute resistance alleles. QTLs controlling defensive GSLs are highly dependent on pathogen infection. An overlap of two QTLs identified between resistance to B. cinerea and defense metabolites also showed isolate specific effects. This work suggests that directly searching for resistance loci may not be the best approach at improving resistance in B. rapa to necrotrophic pathogen.

Generalist necrotrophic pathogens including Botrytis cinerea cause significant yield and financial losses on Brassica crops. However, there is little knowledge about the mechanisms underlying the complex interactions encoded by both host and pathogen genomes in this interaction. This potentially includes multiple layers of plant defense and pathogen virulence mechanisms that could complicate in breeding broad spectrum resistance within Brassica species. Glucosinolates (GSLs) are a diverse group of defense metabolites that play a key role in interaction between Brassica and biotic attackers. In this study, we utilized a collection of diverse B. cinerea isolates to investigate resistance within the Brassica rapa R500 × IMB211 recombinant inbred line population. We tested variation on lesion development and glucosinolate accumulation in parental lines and all population lines. We then mapped quantitative trait loci (QTL) for both resistances to B. cinerea and defense metabolites in this population. Phenotypic analysis and QTL mapping demonstrate that the genetic basis of resistance to B. cinerea in B. rapa is isolate specific and polygenic with transgressive segregation that both parents contribute resistance alleles. QTLs controlling defensive GSLs are highly dependent on pathogen infection. An overlap of two QTLs identified between resistance to B. cinerea and defense metabolites also showed isolate specific effects. This work suggests that directly searching for resistance loci may not be the best approach at improving resistance in B. rapa to necrotrophic pathogen.

INTRODUCTION
Generalist necrotrophic pathogens including Botrytis cinerea are significant economic challenges on Brassica crops because of its ability to cause lesions on nearly all harvestable components of the plant including leaves, stems, florets, and pods (Bolton et al., 2006;Williamson et al., 2007). Thus, there is interest in breeding for resistance to B. cinerea within Brassica crop species. This is, however, complicated because resistance to B. cinerea is a highly quantitative trait with little evidence of major effect resistance loci (Denby et al., 2004;Finkers et al., 2007aFinkers et al., ,b, 2008Rowe and Kliebenstein, 2008;Staal et al., 2008;Davis et al., 2009;Anuradha et al., 2011). The effort if further complicated by the desire to breed for broad spectrum resistance which runs into the complication that B. cinerea is a highly polymorphic pathogen with this genetic variation altering the virulence mechanisms by which the pathogen interacts with the plant (ten Have et al., 1998;Denby et al., 2004;Kliebenstein, 2007, 2008;Amselem et al., 2011;Staats and van Kan, 2012). However, most genetic or molecular studies looking at how plants resist B. cinerea typically utilize individual isolates hindering the effort to find loci that provide potential resistance to a broad spectrum of B. cinerea isolates.
To test the influence of genetic variation in B. cinerea upon quantitative resistance in Brassica rapa, we utilized a collection of 14 genetically variable B. cinerea isolates to test for variation in lesion development on the IMB211 and R500 parents of a B. rapa RIL population (Iniguez-Luy et al., 2009). This identified a significant interaction of the host and pathogen genotypes on the quantitative resistance outcome of the interaction. We then proceeded to map resistance to five B. cinerea isolates in the IMB211 × R500 RIL population. Simultaneously, we measured glucosinolate accumulation in all the RILs in the presence and absence of the pathogen to map defense metabolite quantitative trait loci (QTLs). This showed that the detached leaf assay approach identified isolate specific resistance QTLs in B. rapa and that the defense metabolite QTLs were highly dependent upon the interaction with the pathogen. There was also an overlap of two QTLs between defense metabolites and resistance to B. cinerea but both loci showed isolate specific effects. This work suggests that a potential avenue going forward to breed for quantitative resistance to B. cinerea in B. rapa would be to focus on breeding for the proper defense metabolite blend. Or alternatively to stack isolate specific resistance loci to create the appearance of broad spectrum resistance. Further work is required to assess if these detached leaf identified loci will work in whole plant field based assays.

Bulking of Brassica rapa Germplasm
Seeds of the two parental lines of the B. rapa BraIRRi population, the annual yellow sarson R500 (male) and the rapid cycling IMB211 (female) (Williams and Hill, 1986) as well as the resulting recombinant inbred lines (RILs) population, were obtained (Iniguez-Luy et al., 2009). Both parental lines and 133 lines of RILs in this study were bulked during 2013 in the University of California, Davis greenhouses under a controlled environment. The plants were grown under a 12-h photoperiod under metalhalide lamps using a day/night temperature set at 25/18 • C and relative humidity of 70. Plants were grown in 15-cm-diameter plastic round pot, filled with standard potting soil (Sunshine Mix #1; Sun Gro Horticulture) and ∼4.5 g of a slow-release fertilizer (14-14-14 Osmocote, Scotts). For bulking, all plants were bagged with mesh pollination bag during flowering to prevent crosspollination. The plants were staked upright to produce larger fruits and reduce pathogen and herbivore attack. We watered the plants once a day and then reduced the watering times for about 2 weeks when the fruits began to mature. The seeds were harvested once the majority of the seedpods were dry and then stored separately in paper bags in a cool, dry, dark place until further use.

Growth of the R500 X IMB211 RIL Mapping Populations for QTL Analysis
For measuring the resistance to B. cinerea and GSL metabolite accumulation, the B. rapa parental lines and 120 RILs that had sufficient seed were raised in a controlled environment chamber at University of California, Davis. Three seeds of each genotype were sown in the center of a separate well of a 6 × 12 well tray filled with standard potting soil matrix and the tray was placed in a large planting flat (280 mm × 540 mm × 58 mm). All genotypes were randomized in a randomized complete block design. Two liters of nutrient-enriched water (0.5% N-P-K fertilizer in a 2-1-2 ratio; Grow More 4-18-38) were added into the flat to ensure the compost around seeds was moist. The flat was covered with a transparent plastic hood to maintain humidity during germination and placed into a chamber at 5 • C to complete vernalization. After 3 days of chilling, the transparent hood was removed and the flat was transferred into a climate-controlled chamber with the temperature at 22 • C and a photoperiod 10 h light: 14 h dark photoperiod. All plants were watered twice a week using nutrient-enriched water. At 21 days after sowing, 4-5 true leaves were harvested from each plant for analysis of lesion size and GSL content. The entire experiment was repeated a second independent time.

B. cinerea Isolates, Preparation of Conidia and Inoculation
Information for all B. cinerea isolates used in this study is described in previous reports (Denby et al., 2004;Kliebenstein et al., 2005). To collect spores for plant innoculation, all isolates were maintained as conidial suspensions in 30% glycerol at −80 • C at our lab for long time storage. Conidia suspensions were swabbed on freshly prepared potato dextrose agar (PDA, Gibco/Invitrogen, Carlsbad, CA, USA) medium in Petri dishes and cultured at room temperature. Spores used for infection on B. rapa leaves were obtained as described (Rowe and Kliebenstein, 2008). The detached leaf assay has been utilized in numerous settings to identify causal loci controlling resistance to necrotrophic fungi. While this assay will miss loci controlling resistance in a whole plant context like pedicel transmission barriers, it is a useful approximation (Sharma et al., 2005;Mulema and Denby, 2012;Cowley et al., 2014;Boydom, 2015). For lesion assays and trypan blue staining, the fully developed detached leaves placed on 1% phytoagar in large plastic trays. Detached leaves were inoculated with 4 µL droplets of B. cinerea spore suspensions (10 spores/uL) in 50% filtered grape juice (Santa Cruz Organics, CA) at room temperature with light illumination. Control leaves (mock) were inoculated with of the 4 µL droplet of 50% filtered grape juice without spores. An abiotic GSL elicitor, AgNO3 (5 mM), was inoculated in the same way. Six independent infections were conducted per isolate/genotype pair across the two independent experiments. Digital photographs were taken every 8-12 h to examine the lesion development on leaves.
To test for differences in the B. rapa parental lines, we screened fourteen B. cinerea isolates for differential virulence against the two parental lines using six independent biological replicates per genotype/isolate combination. Susceptibility of B. rapa parental lines, R500 and IMB211, to diverse B. cinerea isolates was digitally measured by measuring the size of the developing fungal lesion after 72 h post inoculation. The lesion size for each isolate was compared between the two genotypes using ANOVA to test the statistical significance of influence of each experimental factor, or a specific interaction between experimental factors on the lesion size. The ANOVA model was lesion = plant genotype + fungal isolate + experiment replicate + plant genotype × fungal isolate + plant genotype × experiment replicate + fungal isolate × experiment replicate + error.
For QTL mapping, three isolates that showed significantly differential virulence between the two B. rapa parents, Ausubel, Davis navel, and Pepper, as well as the most virulent isolate, Katie tomato, and least virulent isolate UK Razz were used to measure lesion size on each RIL in threefold replication per experiment in two independent experiments for a total of six biological replicates. Seeds were sown in replicate and plants were measured for resistance to the different isolates as described above. After FIGURE 1 | Differential Botrytis cinerea virulence on the R500 and IMB211 Brassica rapa RIL parents. B. rapa, R500 (light, left) and IMB211 (dark, right) were treated with various B. cinerea isolates at 72 h after inoculation and lesions quantified as described. Means of lesion size (n = 6) and standard error are presented. Asterisks identify isolates that have significant difference in virulence on the two B. rapa parents as determined using ANVOA and post hoc Tukey's t-test to compare virulence within each isolate: * P < 0.05, * * P < 0.01, * * * P < 0.001. planting, there were 114 RILs left for the final lesion size analysis due to the failure of 6 RILs to grow. Least square means for all lesion data was then obtained using the ANOVA as described above for the parents.

Trypan Blue Staining
Trypan blue staining was used to visualize the growth and structure of mycelium and accompanying plant cell death caused by different B. cinerea isolates on leaves of both B. rapa parental lines. Staining of B. rapa leaves was performed at 12 h postinoculation as previously described (van Wees, 2008). Briefly, the infected leaf tissues were transferred into a 50 mL plastic tube with lid and covered with 2.5 mg/mL trypan blue-lactophenol solution diluted in ethanol (96%; 1 :2 v/v). The plastic tube (lid slightly unscrewed) was heated in a boiling water bath for 1 min and the leaf tissues were left in staining solution at RT for about 12 h. Leaf tissues were destained by removing the staining solution and covering the tissues in chloral hydrate solution for 6 h and the distaining solution changed several times until the leaf tissues were clear. The cleared leaf tissues were placed into 50 mL plastic tubes with 70% glycerol. For analysis, stained leaf tissue was spread on a transparent plastic Petri dish and examined by taking high-resolution digital photos of the entire leaf tissue and each lesion

GSL Analysis
To measure the plants response to infection, GSLs were extracted, identified and quantified using a high-throughput analytical system from all of the above leaf tissue treated with B. cinerea isolates, Silver nitrate or grape juice after 72 h post inoculation (Kliebenstein et al., 2001a,b,c). Briefly, GSLs were identified by comparing the retention time of HPLC peak and UV absorption spectrum with standards (Reichelt et al., 2002). For the RILs, we were only able to obtain GSL values for the control and B. cinerea isolate Pepper infected samples due to a technical failure during sample storage. Each GSL was analyzed using the same statistical models as for the respective lesion size analysis described above.

QTL Analysis
To detect QTL for the lesion size and GSL content in the R500 × IMB211 RIL population, we used the least-square means for each trait for each RIL across all experiments. A highresolution genetic map was obtained for the R500 × IMB211 RIL population from previously published resources (Devisetty et al., 2014). This was used in conjunction with the Composite interval mapping (CIM) algorithm as implemented by the cim function in the R/qtl analysis package to map QTL (Broman et al., 2003). The imputation method was selected  and forward regression was used to identify three markers as covariates, with window size of 10 cM, an error of 0.0001, and 0 cM steps: cim(cross, method = "imp, " n.marcovar = 3, window = 10).
The LOD thresholds to call significant QTLs were estimated using 1000 permutation for each phenotype with a genomewide significance level of p = 0.05 (Churchill and Doerge, 1994;Doerge and Churchill, 1996). Results obtained by CIM were analyzed and the define.peak function in R/eqtl analysis package was used to define the QTL with support LOD interval for each phenotypic trait (Broman et al., 2003). QTL were named with respect to their phenotypic traits and the cM position on the chromosome number. The additive effects of the loci along all chromosomes were estimated using the effectscan function in R/qtl package (Broman et al., 2003).

Testing of QTL Interactions
To identify QTL × Isolate or QTL x QTL interactions using the detected QTL, we conducted an ANOVA using all of the RILs. In the ANOVA model, the markers that most closely associated with each QTL were used as factors. Furthermore, the different isolates as well as the untreated data were all used within the model to allow B. cinerea isolates and the treatments to be used as factors in the model. We tested all the QTL main effects as well as all possible pairwise interactions, including the QTL × isolate, QTL x infection or QTL × QTL interactions where appropriate.
FIGURE 5 | Botrytis cinerea isolate dependent effects of resistance QTLs. The estimated phenotypic effect of the alleles from R500 and IMB211 at markers associated with isolate dependent QTL are shown. Error bars indicate standard error. Isolates that were significant affected by the QTL were tested by ANOVA with post hoc Tukey's HSD and is indicated by asterisks above bars: * P < 0.05, * * P < 0.01. Light shows the value for the R500 allele and dark shows the value for the IMB211 allele at the given QTL. QTLs are as labeled in Figure 4

Variable Resistance of B. rapa Genotypes to Diverse B. cinerea Isolates
To investigate resistance to B. cinerea, we tested two B. rapa genotypes, R500 and IMB211, for resistance to 14 isolates of this necrotrophic pathogen using a previously published foliar resistance assay (Table 1). This detached lesion assay has been widely used to identify necrotrophic resistance genes in a number of different systems (Sharma et al., 2005;Mulema and Denby, 2012;Cowley et al., 2014;Boydom, 2015). These isolates show extensive genomic variation . Leaves of B. rapa R500 and IMB211 were inoculated with B. cinerea spore suspension from each of the 14 isolates and visible expansion of necrotic lesions appeared between 12 and 24 h post inoculation (HPI), indicating outgrowth of hyphae and the initial establishment of primary lesion. Most of the lesions induced by isolates were observed to spread beyond the inoculation droplets at 24 HPI, with lesions expanding in general more rapidly on IMB211 leaves. Chlorotic zones adjacent to the developing lesion were observed for all isolates on both plant genotypes with a tendency to extend to the distal regions of leaves, plants.
Quantifying lesion diameter for all infections showed that there was a statistically significant effect of the B. rapa and B. cinerea genotypes and an interaction between the plant and pathogen genotype in controlling resistance (Table 1 and Figure 1). Comparison of the mean lesion diameters showed that in all instances where there was a significant effect, B. rapa R500 plants had smaller lesion sizes than IMB211 (Figure 1). In addition to the plant genotype, there were significant differences across the B. cinerea isolates for lesion size on the B. rapa genotypes ranging from the low virulence Fresa SD to higher virulence Apple 517 (Table 1 and Figure 1). While most isolates showed equal virulence on the two B. rapa genotypes, four B. cinerea isolates (Ausubel, Davis Navel, Pepper and Supersteak) showed significant differences in virulence across the B. rapa genotypes suggesting that there are Host × pathogen genetic interactions underlying the quantitative resistance of B. rapa to B. cinerea.

Morphological Analysis of the Interaction of B. rapa Genotypes with Diverse B. cinerea Isolates
To investigate if the quantitative variation in resistance between these B. cinerea isolates and B. rapa genotypes is apparent at the microscopic level, we stained the infected B. rapa leaves at 24 HPI (Figure 2). This allowed us to visualized plant vasculature, dead plant cells and fungal hyphae. In all cases, the fungal hyphae developed in the primary lesion, including those isolates that had extremely low virulence UK Razz, Rose or Fresa SD suggesting that the plant was preventing hyphal growth. There no consistent relationship between lesion outgrowth and hyphal production at 24 h with the strongly virulent Apple 517 and Katie Tomato showing similar hyphal distribution as the low virulent UK Razz, Rose or Fresa SD (Figure 2). Thus, the quantitative resistance that we are measuring in the B. rapa/B. cinerea system is not preventing hyphal germination or establishment (Figure 2). Instead, the quantitative resistance is altering the relative rate of outgrowth of the hyphae in each interaction. This is similar to what had previously been found in Arabidopsis thaliana .

QTLs for Quantitative Resistance to B. cinerea
To begin identifying the loci that may control the quantitative interaction of B. rapa with B. cinerea, we measured lesion  ( Figures 1 and 3). Lesion development induced by the five fungal isolates was significantly influenced by B. rapa genotypic variation (p < 0.001), B. cinerea isolate variation (p < 0.001) and the interaction of the two as determined using analysis of variance (ANOVA ; Table 1). This further supports that there are genotypic dependent interactions between B. cinerea and B. rapa. The B. rapa RIL population displayed a range of variation for lesion size trait that was different for all five B. cinerea isolates (Figure 3). Using the Ausubel and Pepper isolates showed a distribution that skewed toward the sensitive IMB211 parent. In contrast, the Davis Navel isolate highlighted a distribution that was more evenly spread between the parental values. Interestingly, for both isolates that had no difference between the IMB211 and R500 parents, we were able to identify underlying variation in the resistance traits. For example, the Katie Tomato isolate that was equally virulent on the IMB211 and R500 parents identified a large number of RILs that transgressively segregated for increased resistance (Figure 3). Thus, the equal resistances of IMB211 and R500 to these two isolates are caused by opposing resistance alleles in the two B. rapa RIL parents.
Using the lesion measurements, we mapped QTL controlling phenotypic variation in lesion size within the RILs populations derived from R500 × IMB211. This identified five QTL governing lesion size induced by B. cinerea isolates (Table 3 and Figure 4). Two of these QTL, L7.26.6 and L9.0.1, were detected using B. cinerea isolate Ausubel (Table 3 and Figure 4). The QTL L1.51.7 detected by B. cinerea isolate Katie tomato was located on chromosome I. The L3.2.5 and L9.72.3 QTL on chromosomes III and IX were detected with the B. cinerea isolate UK Razz isolate. Surprisingly, no QTL were detected using the B. cinerea isolates Davis navel and Pepper and no QTL were found consistently for all isolates. This is in contrast to the effect size plots that are similar for most isolates. This is probably due to the limited number of RILs available for this population that may generate a potentially elevated false negative rate with QTL mapping (Joseph et al., 2013(Joseph et al., , 2014. Thus, we proceeded to use a linear modeling approach to directly test if the detected QTLs were actually isolate specific. This analysis showed the QTLs were all isolate specific (Figure 5 and Supplementary Table 1). Interestingly, The QTL L1.51.7 found only for the Katie Tomato isolate also significantly altered resistance to Davis Navel. Thus, we were able to identify QTL that appear to control isolate specific aspects of quantitative resistance in the interaction of B. rapa with B. cinerea.

Identification of QTL Controlling Defense Responses
Previous work has linked genetic variation in defense compounds to variation in biotic interactions including for B. cinerea (Denby et al., 2004;Rowe and Kliebenstein, 2008). Thus, we measured GSL content in all of the RILs in both control leaves and B. cinerea isolate Pepper infected leaves. This showed that all detected GSLs had genetic variation and were affected by the infection with the B. cinerea Pepper isolate ( Table 2). There were no presence or absence polymorphisms affecting GSL abundance in this population allowing us to focus on quantitative variation controlling their relative abundance ( Table 2; Kliebenstein et al., 2001b;Wentzell et al., 2007;Chan et al., 2010). All five GSLs identified significant QTLs with the majority of them appearing to be dependent on the presence or absence of B. cinerea (Figure 6 FIGURE 7 | Glucosinolates QTL that are show pathogen responsiveness. Shown is the estimated phenotypic effect of the alleles from R500 and IMB211 at GSL accumulation QTL that are treatment dependent. Error bars indicate standard error. Significance differences between the treatment × genotype groupings were determined by ANOVA with a post hoc Tukey's HSD test and are indicated with different letters. The top of each graph indicates the GSL and QTL being shown with abbreviations as listed in Table 2. Light shows the accumulation for RILs with the R500 allele and dark shows the value for the IMB211 allele at the given QTL. and Table 3). The aliphatic and benzylic GSLs identified more QTLs under the control treatment whereas the indolic GSLs identified more QTLs under the B. cinerea infected tissue. This agreed with the observation that the treatments in general lead to higher indolic GSLs and lower 4MSB and Benzylic GSLs. This allowed QTL effects to be seen for indolic GSL following treatment and Benzylic and 4MSB prior to treatment (Figure 7). Most of the indolic QTLs were such that the IMB211 allele leads to lower pathogen induced indolic GSLs in contrast to the R500 allele (Figure 7). One QTL, G9.5.0, on chromosome IX shared by both aliphatic and benzylic GSLs, was also detected as affecting lesion development when the RILs were infected with the B. cinerea isolate Ausubel isolate. This suggested that there might be a link between these GSL and resistance to at least this B. cinerea isolate (Figures 4 and 6).

Epistasis Analysis
Previous work on quantitative resistance to B. cinerea has shown that the identified QTLs are typically epistatic to each other (Finkers et al., 2007a(Finkers et al., ,b, 2008Rowe and Kliebenstein, 2008). To investigate the epistatic architecture underlying isolatespecific resistance to B. cinerea and chemical defense within the R500 × IMB211 population, we conducted ANOVA models for QTL controlling lesion development and individual GSL using the genetic markers closest to the QTL peak as terms in the linear model. In contrast to previous studies, we only found a single epistatic interaction with any evidence of significance in altering Botrytis resistance. This interaction was between the L3.2.5 and L9.72.3 QTLs that were unique to the UK Razz isolate (Figure 8). In comparison to B. cinerea resistance, more epistatic interactions were detected for GSL accumulation, one for the accumulation of aliphatic GSLs and three for the accumulation of the I3M GSL (Figure 8, Supplementary Tables 2-6). Interestingly, all of the epistatic interactions between GSL loci were also found to interact with the presence or absence of the pathogen suggesting that they may be linked to the regulation of the defense compounds.

DISCUSSION
Using 14 diverse B. cinerea isolates to measure quantitative resistance in two B. rapa lines that are the parent of a common RIL population showed that the two parents differed in their resistance to some but not all of the isolates (Figure 1). Further QTL mapping with a subset of these isolates showed that the genetic basis of this is likely polygenic with transgressive segregation showing that both parents can contribute resistance alleles. This was even the case when both B. rapa parents had identical resistance to an isolate (Figures 3 and 4). These loci need to be assessed in whole plant assays under field growth conditions to assess their agronomic utility. This quantitative, isolate specific and transgressive basis of quantitative resistance to B. cinerea is similar to what has been seen in A. thaliana (Denby et al., 2004;Rowe and Kliebenstein, 2008). This suggests that directly searching for QTL that provide broad spectrum resistance to B. cinerea within Brassica's may not be the best approach to identifying successful avenues to achieve this goal. An alternative approach to utilizing these genetic loci may be to stack isolate specific loci with overlapping specificity to create a quilt of loci that can provide a protective blanket against an array of diverse isolates.
We were also able to identify QTLs that control the accumulation of the defensive glucosinolate metabolites in response to infection (Figure 6). The lower number of overlapped QTLs between GSLs and resistance may be due to the significant level of false negative error even in a large size of RIL population during QTL mapping (Chan et al., 2011;Joseph et al., 2013). While a few of these QTLs overlapped with one B. cinerea resistance QTL, GLS have frequently been linked to altered biotic interactions within the Brassica's (Mithen et al., 1986(Mithen et al., , 1987Mithen, 1992;Mithen and Magrath, 1992;Kroymann and Mitchell-Olds, 2005;Pfalz et al., 2007Pfalz et al., , 2009Fan et al., 2011;Stotz et al., 2011;Buxdorf et al., 2013;Cargnel et al., 2014;Weis et al., 2014;Calmes et al., 2015;Kerwin et al., 2015). However, the specific mechanism by which the GSL can alter biotic interactions is not yet well understood. Some studies have provided evidence of direct toxicity to the biotic attacker that can be compensated by resistance mechanisms in the pathogen (Bednarek et al., 2009;Fan et al., 2011;Stotz et al., 2011). In contrast other studies have begun to illuminate a different possibility, specifically that the GLS alter the defense signaling pathways by which the plant responds to pathogens or jasmonic acid (Clay et al., 2009;Kerwin et al., 2011;Burow et al., 2015). Using the QTL results, it is not possible to discriminate between these two possibilities and further work will be required to test if the link between GSL accumulation and B. cinerea resistance in B. rapa is due to direct toxicity, altered defense responses or a blend of both. Partitioning between these two possibilities will be key to develop a maximally efficient effort at improving resistance to a broad array of B. cinerea isolates in B. rapa.

CONCLUSION
This work begins to highlight the underlying genetic complexity of breeding for improved resistance to B. cinerea within B. rapa. The directly identified resistance loci are highly isolate specific but it may be possible to improve the breeding efficiency by breeding for an optimal defense compound blend. This, however, needs to be balanced by the fact that these same defense compounds also influence the flavor and nutritive value of the resulting Brassica crop. As such any effort at resistance breeding will by fact of this link also alter the quality of the resulting crop. By combining quantitative resistance loci that target overlapping ranges of Botrytis isolates together may be a more feasible breeding strategy to confer a broad-spectrum and durable resistance to crops against this pathogen. It is also important to note that this level of isolate specific resistance loci has significant impact on the mechanistic analysis of quantitative resistance (Broekgaarden et al., 2015). This indicates that the use of individual isolates will only provide mechanistic insight into how that individual isolate is resisted and that a broad range of pathogen genetics needs to be incorporated to assess the broader mechanistic influences on quantitative resistance. The low number of identified QTLs given the high heritability of the resistance trait suggests that there is a need to increase the available RIL population sizes to decrease the false negative error rate and to obtain a more precise picture of the genetic architecture underlying the resistance to B. cinerea in B. rapa.

AUTHOR CONTRIBUTIONS
S-TK, DK conceived and designed the experiments. WZ, S-TK performed the experiments. WZ, S-TK, DK analysis the data. WZ, DK wrote the paper.