Variation in lodging-related traits across a panel of 79 B. napus accessions
We first assessed lodging resistance across the diversity panel. Using ANOVA, the trait data obtained from the largest trial (J2012) was assessed for normality. In all cases, approximately normally distributed residuals were observed. Through ANOVA, for all traits included in the study, we observed high levels of phenotypic variation between genotypes (P = < 0.01), as might be expected for an association panel of diverse accessions.
Two smaller trials were conducted, K2010 and K2011. In the majority of cases, accessions were replicated across both trials. However, a number of accessions were represented in just a single year (see Supplementary Data File 1). Due to the unbalance nature of these data sets, the data obtained from each field trial were analysed independently. Through ANOVA of the 2010 and 2011 trial results, with the exception of stem hollow area in the K2010 analysis (P = 0.618), significant differences were seen between accessions for all of the traits assessed (P < 0.05). Data summaries and all ANOVA outputs for each of these field trials can be found in Supplementary Data File 9. Given the more extensive variation represented by the J2012 trial, this data set was used for the majority of analyses discussed in the following sections. However, the data obtained from K2010 and K2011 provide validation of trends observed in the J2012 data set and will, for this reason, be discussed where appropriate.
Relationships between stem mechanical strength and stem morphological and structural traits
To assess the importance of stem morphological and structural traits on stem mechanical strength, regression analysis was conducted. A summary of the results obtained from this analysis is given in Table 1. Many significant correlations were observed. Firstly, the absolute strength traits Fmax and F/V correlate positively with an R2 of 0.88 (P < 0.001). This suggests that these traits are highly related, i.e. stems resistant to bending are also more resistant to breaking. These absolute strength traits also correlate significantly with a number of stem morphological and structural traits. The structural strength trait, second moment of area, correlates positively with both Fmax and F/V with respective R2 values of 0.63 (P < 0.001) and 0.66 (P < 0.001). Stem diameter was found to explain a greater level of variation than second moment of area in these absolute strength traits with an R2 value of 0.70 (P < 0.001) for both Fmax and F/V. Stem outer cortex thickness and parenchyma area were also found to correlate positively with Fmax and F/V. These traits are also highly related to stem diameter and second moment of area. It is therefore difficult to assess whether these tissues themselves are contributing to strength, or whether the correlations are due to their relationship with a more general increase in stem thickness.
A strong relationship was detected between the absolute strength traits and stem weight where R2 values of 0.62 (P < 0.001) and 0.64 (P < 0.001) were seen for Fmax and F/V respectively. When assessing the relationships between this biomass trait and the stem structural and morphological traits, we observed that the correlations with Fmax and F/V were stronger. Similar correlations were also identified between stem weight and stem diameter/second moment of area. It is therefore likely that the relationship observed between stem weight and absolute strength reflects this, but may also suggest that stem density is important. Stem absolute strength was also seen to correlate positively with plant height, with R2 values of 0.35 and 0.39 obtained for F/V and Fmax respectively.
Interestingly, when looking at the relationship between stem absolute and stem material strength traits, a negative relationship can be seen. Stem material strength describes the mechanical strength contributed by the material composition of the stem and is considered independent of stem structural strength which is described by the stem geometry. Of course, these strength components are not truly mutually exclusive, but it is useful, when aiming to understand the underlying genetics, to break down complex traits into component traits—in this case material strength, structural strength and absolute strength (material and structural strength combined). Fmax and F/V are used in calculating MOR and MOE respectively. Based on the equations used, it would be expected that an increase in these absolute strength traits would result in increased material strength values. However, when looking more closely at the raw data, it can be seen that in the majority of cases, high absolute strength values can be largely explained by high structural strength (large diameter/high second moment of area) and in these accessions, low material strength is often observed. While this suggests that stem structural strength is the most important contributor to absolute stem strength in these accessions, it does not mean that the material strength properties do not have a role to play. When looking at the mean trait values, there are several examples which illustrate a certain importance of material strength in determining overall stem strength. An example can be seen when comparing the trait data for the varieties, Dwarf Essex and Taisetu. The absolute strength values for Dwarf Essex are 24.2 and 19.9 kg/s for Fmax and F/V respectively. The mean values for stem diameter and second moment of area for this accession are 9.2 and 362 mm4 respectively. When comparing these data to those obtained for Taisetu, it can be seen that despite displaying very similar values for stem diameter and second moment of area (9.288 and 365.8 mm4 respectively), Taisetu shows much lower absolute strength values of 6.259 and 8.79 kg/s for Fmax and F/V respectively. This may be explained by the much lower material strength seen for this accession in comparison to that of Dwarf Essex. For Taisetu, MOR and MOE values of 2.52 and 173.4 N/mm−2 were seen. In comparison, Dwarf Essex was found to have a mean MOR of 11.03 N/mm−2 and a mean MOE of 39.1 N/mm−2. This clearly shows the effect that material strength can have in the absence of variation in stem structural strength. As seen in the regression analysis results carried out for the 2012 JIC trial material, in the analyses carried out on material from the 2010 and 2011 KWS trials, the absolute strength traits Fmax and F/V were found to correlate most significantly with stem diameter, second moment of area and stem weight (data not shown).
The relationship between stem lodging risk and stem mechanical strength
Using the apparatus illustrated in Fig. 1a, the J2012 material was assessed for stem lodging risk (SLR) under field conditions. Assessment of the data obtained through ANOVA revealed significant variation for this lodging risk trait between accessions (P = 0.043). Through a further regression analysis, the relationship between this field-based scoring and the mechanical strength traits obtained from the three-point bend test were explored. Some of the key relationships observed are plotted in Fig. 2. Significant positive correlations were observed between the absolute strength traits and the SLR scores measured in the field with R2 values of 0.42 (P < 0.001) and 0.33 (P < 0.001) for Fmax (Fig. 2a) and F/V respectively. These lodging risk scores were however found to correlate negatively with the material strength measures with R2 values of 0.16 (P < 0.001) for MOR and 0.21 (P < 0.001) for MOE (Fig. 2e) respectively. This again illustrates the importance of stem structural strength in determining stem strength in B. napus. Indeed, very similar correlations were observed between the stem structural traits stem diameter and second moment of area and the SLR score as those seen for absolute strength. For both of these structural traits, positive correlations were observed with R2 values of 0.42 (P < 0.001) and 0.37 (P = 0.001) for stem diameter (Fig. 2b) and second moment of area respectively. Stem height and plant weight were also found to correlate positively with the SLR scores (Fig. 2c, d) with respective R2 values of 0.25 (P < 0.001) and 0.26 (P < 0.001). However, given that these correlations are weaker than those seen between SLR and the absolute strength measures, it seems that the variation in these morphological phenotypes has not had a confounding effect on the SLR measurements. The detection of these significant correlations is very promising and suggests that the mechanical strength measures obtained through the three-point bend test are capturing important variation relevant to stem mechanical strength under field conditions. Furthermore, as with the results obtained from the three-point bend test assay, the results presented here suggest that understanding the genetic control of stem absolute strength and stem structural strength (with contributions made by stem diameter and stem second moment of area being of highest value) would be of great interest for the improvement of lodging resistance in B. napus. The work presented also confirms the widely accepted link between plant height and lodging susceptibility.
Associative Transcriptomics for stem absolute and structural strength traits, Fmax, structural and morphological traits
In order to identify molecular markers that might be useful to support breeding for traits of relevance for lodging resistance, we conducted association analysis by Associative Transcriptomics.
Through regression analysis, we showed that stem lodging risk is most highly related to the absolute stem strength traits Fmax and F/V. Given the very high correlation observed between Fmax and F/V, we limited the consequent Associative Transcriptomics analysis to Fmax only. We also explored marker-trait associations for several stem structural traits that showed high correlation with absolute stem strength. The structural traits included were stem diameter, stem second moment of area, stem parenchyma area, outer cortex thickness and stem weight. Finally, given the known role that height has to play in contributing to lodging risk (which was confirmed in the present study), variation in plant height was also included in the Associative Transcriptomics analysis. An overview of the most highly significant SNP and GEM marker associations identified for each trait is provided in Supplementary Data File 10. Full Manhattan plots for all traits analysed can be found in Supplementary Figures 1–9. With the exception of one marker (JCVI_19156:300), the highly associated markers occur at similar frequencies in the sub-populations defined by STRUCTURE, as shown in Supplementary Figure 10, confirming their association with trait variation is not simply due to population structure within the panel.
For the absolute strength trait, Fmax, three main SNP association peaks were identified based on the results obtained from J2012 trial (a full Manhattan plot for these data can be found in Supplementary Figure 1). The first and most significant of these was found to be on chromosome A2/C2 (the A2 association can be seen marked with *1 in Fig. 3a). The most significant marker within this peak was JCVI_31359:1657 (P = 4.45E-05, trait effect = 30.9% (a percentage based on trait range across accessions)). This region was also detected (more weakly) in the SNP association analysis carried out for the K2010 material for absolute stem strength (Supplementary Figure 2). Identification of a common signal across two field trials is very promising. These markers were found to be in close proximity to orthologs of Arabidopsis genes expected to have pectin methylesterase (PME)/pectin methylesterase inhibitor (PMEI) activity, AT5G50030 and AT5G50060. Previous studies have found that the methylesterification state of cell wall pectin has a key role to play in determining stem mechanical strength (Hongo et al. 2012). Pectin methylesterification is also known to effect cell expansion (Wolf et al. 2009), a process which likely contributes to variation in stem thickness. As we have seen, stem diameter explains a high level of variation in stem absolute strength in B. napus, making these genes very good candidates for controlling the trait variation observed.
The second SNP association for this absolute strength trait was detected on chromosome A3/C3 (A3 association can be seen marked with an *3 in Fig. 3c) with the most significant marker being JCVI_38663:382 (P = 4.16E-05). No clear candidate genes were identified in close proximity to this marker.
A final SNP association for Fmax was seen on chromosomes A5/C5, with the top marker being JCVI_39914:1128 (the A5 association is shown with *2 on Fig. 3b). This marker was assigned a P value of 4.07E-04 and a trait effect of 26.9%. Within close proximity to these markers is again an orthologue of an Arabidopsis gene described as having PMEI/PME activity (AT3G12880). Weak signals in this region were also identified for stem weight in the Associative Transcriptomics analysis carried out on both the J2012 and K2011 trials (Supplementary Figures 3 and 4).
In addition to these very promising SNP associations, highly significant GEM associations were also detected for Fmax for the J2012 trial. A summary of these associations is given in Supplementary Data File 10. One of these GEMs, JCVI_28286, detected on chromosome C7, corresponds to an orthologue of an Arabidopsis CYTOCHROME P450 (CYP450) (AT4G27710). The expression of this gene correlates negatively with Fmax (P = 5.41E-08, R2 = 0.34) (Fig. 3d). In Arabidopsis, the expression of this gene has been linked with localised cell wall deposition (McCurdy et al. 2008). To further explore the role of this gene in contributing to stem mechanical strength in B. napus, the respective transcript abundance values for this marker were mapped as a trait against the SNP data. This analysis revealed SNP associations. The clearest of these associations was identified on chromosome A9/C8 (the A9 association can be seen marked as *4 in Fig. 3d). A second clear peak was identified on A6/C6. The most significant marker detected for the A6/C6 association was JCVI_11271:1070 (P = 1.09E-05). The second association, seen on A9/C8, had a much stronger signal, with the most significant marker being JCVI_7691:268 (P = 5.44E-07). In close proximity to the SNP association on A6/C6, there are several genes involved in cell wall biosynthesis. The first of these is an orthologue of Arabidopsis INCURVATA 4 (AT1G52150). This gene is predicted to act as a transcriptional regulator, important for vascular development and cell differentiation. In addition, two genes known to contribute to cell wall mannan structure were identified. Firstly, there is a gene orthologous to Arabidopsis MANNAN SYNTHESIS-RELATED 2 (AT1G51630). Cell wall mannans are thought to bind to cellulose, providing structural integrity to the cell wall (Rodríguez-Gacio et al. 2012). As a second example, an orthologue of Arabidopsis ALPHA MANNOSIDASE 1 was identified. Mannosidases are a group of enzymes, involved in the cleavage of linkages between cell wall mannose and other cell wall polysaccharides (Mast and Moremen 2006). No clear candidate genes were identified in close proximity to JCVI_7691:268. These associations suggest that there may be an interaction between these loci and the CYP450 gene detected in the GEM analysis. Based on the candidate genes identified in the A6/C6 region and the expected role of this CYP450 gene in contributing to cell wall biogenesis, these findings may be indicative of a cell wall biosynthesis pathway which contributes to stem mechanical strength in B. napus.
As seen for Fmax, a clear SNP association can be seen on chromosome A2/C2 for stem diameter with the most significant marker being JCVI_31359:1651 (Supplementary Figure 5). This association signal was also identified for stem parenchyma area (Supplementary Figure 7). Although these associations are less clear than the SNP signal identified at this locus for Fmax, the detection of common associations between these traits suggests that there is a genetic component underlying the observed trait correlations. A second SNP association for stem parenchyma was seen on chromosome A1/C1 with the most significant marker, JCVI_36764:445, reaching a P value of 0.00023. No clear candidate genes for this trait were identified in this region.
For stem outer cortex thickness, a SNP association was seen on chromosome C6 with the most significant marker being JCVI_335:451 with a P value of 0.00027. Again, no clear candidate genes for this trait were found in close proximity to this marker (Supplementary Figure 8).
One clear association signal was detected for plant height. This was seen on chromosome A3/C7 where the most significant marker is JCVI_26003:352 (P = 2.86E-04, trait effect 21.7%) (Supplementary Figure 9). This marker is in close proximity to a gene orthologous to Arabidopsis BRI1 (BRASSINOSTEROID-INSENSITIVE 1) SUPPRESSOR 1 (AT4G30610). BRI1 is known to be involved in regulation of plant height in Arabidopsis. Mutants defective for this gene have a very clear dwarfing phenotype (Noguchi et al. 1999). The detection of a suppressor of this gene for plant height in B. napus is therefore very encouraging. A more modest peak was also detected for plant height on chromosome A9/C8. The most significant marker is JCVI_19156:300 (P = 8.82E-04, trait effect of 16.8%). No clear candidate genes were detected in this region.
The GEM analysis identified two highly associating GEM markers for plant height on chromosome A5, both of which correspond to a gene orthologous to Arabidopsis MICROTUBULE ORGANISATION 1 (MOR1) (AT2G35630). Arabidopsis mutants for this gene exhibit a temperature-dependant reduction in organ size. This was found to be due to the inability of the mutant plant to correctly organise cortical microtubules (Whittington et al. 2001). To further explore the potential role of this gene in contributing to the genetic control of plant height in B. napus, transcript abundance levels for the most highly associated of these GEM markers, A_EE440437 (P = 2.98E-07, R2 = 0.26 (correlating negatively with plant height)), were mapped as a trait against the SNP data. In doing this, a very clear SNP association on chromosome A2/C2 was uncovered. The most highly associated marker within this peak was JCVI_20133:246. No clear candidate genes were identified within this region. However, the results presented here suggest that there may be a gene, which resides at this A2/C2 locus, which either directly or indirectly regulates the expression of MOR1, contributing to the observed variation in plant height in B. napus.
Validation of markers for the selection of high stem absolute and structural strength
The discovery of marker-trait associations through methods such as Associative Transcriptomics has the potential to contribute greatly to crop improvement through marker-assisted breeding. The durability of such markers can however depend on many factors including environmental interactions. It is also possible that the associations may be the result of false-positive errors due for example to any unaccounted-for population structure or relatedness between individuals used within the study. It is therefore important that the efficacy of detected markers in selecting for the trait of interest is validated. To do this, a test panel (i.e. a set of unrelated accessions not used for the original association analysis) of B. napus accessions of previously unknown genotype was screened for marker variation detected through Associative Transcriptomics. Following mechanical testing, it was then possible to assess whether the allelic variation segregates with the target trait as would be expected based on the Associative Transcriptomics results.
Given the importance of the absolute strength trait, Fmax, in contributing to stem lodging resistance, we focused this marker validation study on the marker association detected on chromosome A2/C2 with the most significant marker being JCVI_31359:1657. This marker was not only found to show association with Fmax across multiple field trials (across different years and environments), but was also seen for additional traits which were found to be related to stem absolute strength through regression analysis. Due to low sequence read-depth in the region surrounding the most significant marker in this association peak, an alternative, tightly linked marker, JCVI_31359:1723, was used (P value 9.32E-05; trait effect of 26.5%). Following the assessment of the efficacy of the developed genome-specific marker assay in screening for the target variation (using a subset of Associative Transcriptomics accessions of known genotype), the marker assay was used to explore variation across the 86-accession test panel. High levels of variation in Fmax are seen across accessions and there is no discernible relationship with the sub-populations identified through STRUCTURE analysis, as shown in Fig. 4a. The alleles scored, as summarised in Fig. 4c, had approximately equal frequencies in the sub-populations, as shown in Fig. 4b. There is a significant (P < 0.05) increase in Fmax associated with the increasing (A/G) allele, as shown in Fig. 4d. Together, these results showed that the allelic variation does not simply reflect population structure and that this marker can be used for selection of genotypes with greater stem strength (Supplementary Data File 7 summarises the results obtained).
PMEI positively regulates stem strength in Arabidopsis T-DNA mutants
In addition to molecular markers to assist breeding, the Associative Transcriptomics approach is also remarkably effective at identifying candidate genes for the control of trait variation. We therefore assessed the plausibility of a candidate gene for the control of Fmax based on functional analysis of orthologues in the related species A. thaliana.
Given that two independent SNP associations identified for Fmax (and several related structural traits) were in close proximity to genes implicated in pectin methylesterification, we decided to explore the potential causality of these genes through screening Arabidopsis T-DNA mutants. Although no T-DNA insertion mutants were available for the PME/PMEI candidate genes identified on chromosome A2/C2, several mutants were available for the PME/PMEI candidate identified on chromosomes A5/C5. These mutants each carried an insertion affecting the coding regions of Arabidopsis gene model AT3G12880. Supplementary Data File 6 summarises the mutants screened and provides information regarding the primers used in determining their genotype. For simplicity, these mutant lines have been renamed as pme(i)1 to pme(i)3.
Following genotyping, homozygous mutants were grown alongside wild-type (WT) control plants and the mature stem tissue mechanically tested using a three-point bend test method. Figure 4e summarises the mean Fmax values obtained for the three T-DNA lines included relative to WT plants. Two of the three T-DNA lines assessed for altered stem strength showed a significant difference in Fmax relative to WT plants. In both cases, a decrease in stem strength was observed, suggesting that this PME(I) acts to positively regulate stem strength in WT plants.
Based on information available from the Arabidopsis Information Resource (TAIR), it is not clear whether AT3G12880 has a role in promoting or inhibiting pectin methylesterification. Studies have reported high similarities in the sequence of proteins performing these seemingly antagonistic functions (Pelloux et al. 2007; Wang et al. 2013). Given this, based on sequence information alone, it is difficult to propose the mode of gene function here, i.e. whether it has a role in cleaving methyl ester pectin side groups (making it a PME), or whether it is an inhibitor of this process (making it a PMEI). To further assess the role of this gene in contributing to stem mechanical strength, FTIR analysis was carried out on the stem material used in mechanical testing for homozygous pme(i)-2 homozygous and WT plants. Figure 4f summarises the results obtained. Reports comparing the FTIR spectra of high and low methylesterified pectin suggest that it is a shift in the ratio of two spectral peaks which is the important discriminating factor. The first of these peaks is found at 1740 cm−1. The second is a peak at between 1600 and 1630 cm−1, which would show a higher or lower absorbance relative to the 1740 cm−1 in low methylesterified and high methylesterified pectins respectively (Szymanska-Chargot and Zdunek 2013). As can be seen from Fig. 4f, the T-DNA mutant stems show enrichment relative to WT at 1740 cm−1 and depletion at 1624 cm−1. This suggests that this pectin-related gene is functioning as a pectin methylesterase and that a lack of demethylesterification in these mutant lines is contributing negatively to absolute stem strength.