WO2022101290A1 - Regulation of flowering time - Google Patents

Abstract

The present invention relates to methods for determining likely vernalisation requirements for plants, and to methods for selecting plants for breeding to produce a desired vernalisation requirement.

Description

REGULATION OF FLOWERING TIME FIELD OF THE INVENTION The present invention relates to methods for determining likely vernalisation requirements for plants, and to methods for selecting plants for breeding to produce a desired vernalisation requirement. BACKGROUND TO THE INVENTION Flowering time is a key adaptive trait, being governed by plant responses to different environmental conditions, including temperature. Vernalisation is the induction of a plant's flowering process by exposure to cold temperatures (either naturally during winter, or artificially). Many important crop plants require vernalisation and must experience a period of low winter temperature to initiate or accelerate the flowering process and lead to generation of seeds or fruit. The needed cold is often expressed in chill hours. Typical vernalisation temperatures may be between two and ten degrees Celsius (C). Different cultivars or ecotypes of the same species may be selected for different vernalisation requirements to suit different geographical locations, growing conditions or intended uses. The genetic and biochemical basis of flowering time has been extensively investigated, but not all details are known. In many plants with a vernalisation requirement (especially members of the Brassicaceae), expression of the gene FLOWERING LOCUS C (FLC) renders the plant incompetent to flower until its expression is repressed via environmental cold. Studies in the reference plant Arabidopsis thaliana have revealed a link between differences in flowering time and variation in expression levels and the epigenetic silencing of FLC. However, many important crop plants are polyploid, and have multiple copies and variants of each gene, such that a simple correlation between gene expression and phenotype is not possible. The polyploid (amphidiploid) crop Brassica napus has many different cultivars which show different vernalisation requirements. Broadly, these may be classed as spring cultivars, requiring little or no vernalisation, semi-winter cultivars, requiring limited vernalisation, and winter or biennial cultivars, requiring extensive vernalisation. In B. napus, flowering time paralogues are over-represented relative to other genes. The floral regulator FLC has nine copies. Previous studies in B. napus have identified individual FLC paralogues as determinants of flowering behaviour, associated with spring, semi-winter and winter cultivars, but how these FLC paralogues determine flowering behaviour has until now remained unclear. It might be expected that pre- vernalisation expression levels of these characteristic FLC copies relate to their vernalisation requirement or response. However, pre-vernalisation FLC expression levels do not map well to cold requirement, and the reason for the relative importance of different FLC copies is not clear.

Gaining a greater understanding of those features of FLC expression in B. napus and other polyploid crops which contribute to specific ecotypes and vernalisation requirements would be a significant advance in understanding of this feature of crops. Further, being able to predict which FLC genomic backgrounds are associated with specific ecotypes would enable more targeted selection of cultivars for crossbreeding programmes in order to obtain a desired ecotype.

SUMMARY OF THE INVENTION

Here, we show that it is total FLC expression dynamics rather than that of any specific FLC paralogue which best explains differences in cold requirement between B. napus cultivars. We find that the response to cold varies between FLC paralogues and between ecotypes, and that both these factors contribute to total FLC expression dynamics. The range of vernalisation requirements of B. napus crops and the combinatorial manner with which each flowering behaviour can be achieved from differentially expressed FLC paralogues represents an instance of gene-dosage balance and drift compensation in action and provides an example of how phenotypic plasticity may arise in polyploids. The principles and approaches behind the techniques described herein may also be applied to other genes in other polyploid species.

According to a first aspect of the present invention, there is provided a method for predicting vernalisation requirements of a plant cultivar, the method comprising the steps of: a) determining total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the decay rate of total expression levels of FLC in said plant cultivar under vernalisation growth conditions; and c) calculating the duration of vernalisation growth conditions required for the total expression levels of FLC in said plant cultivar to reduce to below a predetermined threshold, based on the results of steps a) and b).

Previous work has considered specifically only some of the expressed FLC paralogues, and in particular, those believed to contribute most significantly to vernalisation requirements. However, the current invention has determined that reduction of expression of these most significant paralogues can be compensated for or overridden by expression of other paralogues, such that it is necessary to consider total expression of all paralogues.

The predetermined threshold of FLC expression levels is that expression level determined to permit flowering of the plant. This may be estimated and/or experimentally determined based on a spring cultivar of the plant, being a cultivar which does not require vernalisation. For example, where the plant is B. napus, the threshold level may be determined based on the spring cultivar Westar. Alternatively, the predetermined threshold may be determined based on the same cultivar of the plant, by for example experimental determination of the levels at which flowering is permitted. In some embodiments, the predetermined threshold may be determined from previously published information, rather than calculated anew.

Preferably the plant genome includes multiple paralogues of FLC; for example, the plant may be a polyploid plant, preferably an allopolyploid plant. In some embodiments, the plant is selected from among the Brassicaceae, and preferably the plant is a Brassica spp. For example, the plant may be selected from Brassica carinata, Brassica juncea, Brassica napus, Brassica nigra, Brassica oleracea, Brassica rapa, Brassica rupestris. More preferably, the plant is selected from B. napus, B. oleracea, B. rapa. In preferred embodiments, the plant is B. napus, and most preferably B. napus subsp. napus. In some embodiments, B. carinata and B. juncea may be used.

The step of determining the total expression levels of FLC under normal growth conditions may comprise determining the expression levels of each individual FLC paralogue and combining the totals. Expression levels may be determined by any suitable method; for example, detecting expressed protein, or more conveniently by determining amounts of mRNA species for each paralogue. Expression levels may be determined in any convenient or suitable plant organ or tissue; for example, leaves, stems, roots; or for example ground tissue, dermal tissue, vascular tissue, and meristem tissue.

“Normal growth conditions” means suitable non-winter growth conditions for the particular plant cultivar; for example, 18-24 degrees C, 12 hour light/dark cycles, and sufficient water. “Vernalisation growth conditions” means temperatures below 10 degrees C, preferably below 7 degrees C, but above 0 degrees C and preferably above 2 degrees C.

The step of determining the decay rate of total expression levels of FLC may comprise determining the decay rate of each FLC paralogue. The decay rate may be determined experimentally in that plant cultivar, but preferably the decay rate is determined from a model for each FLC paralogue; such a model may be based on a training set of data, for example, data recording decay rates for a given FLC paralogue in one or more representative plant cultivars of the same species as being investigated. Although - as described herein - there may be some variation in decay rates of the same paralogue in different cultivars, we believe this training set approach will nonetheless provide robust predictions.

While a precise decay rate may be used for calculations, in some embodiments FLC paralogues may be classified as cold-responsive or non-cold-responsive (or may include a range of cold-responsive classifications providing finer discrimination between paralogues, for example, whether expression levels will decay after 2, 4, 6, or 8 weeks of vernalisation) depending on their response to vernalisation conditions - that is, whether expression levels do decay in response to cold exposure or do not. The decay rate of total expression levels of FLC may then be determined by calculating the relative contribution to total FLC expression levels of cold-responsive and non-cold- responsive paralogues and applying a weighted decay value to expression levels of cold-responsive and non-cold-responsive paralogues accordingly to determine decay of total FLC expression.

In some embodiments FLC paralogues may be classified as cold-responsive or non- cold-responsive experimentally. In other embodiments, the classification may be predicted based on, for example, gene or mRNA sequence data. For example, one aspect of vernalisation is the epigenetic silencing of FLC expression by the PHD-PRC2 complex, which interacts with the RY motif of FLC. Paralogues which include mutations in the RY motif are less able to interact with the PHD-PRC2 complex, and so are less cold-responsive or even non-cold-responsive. Hence the method may include the step of determining the presence or absence of an RY motif in an FLC paralogue and classifying the paralogue as cold-responsive or non-cold-responsive accordingly. The method may further comprise the step of classifying the plant cultivar as spring, semi-winter, or winter, based on the calculated duration of vernalisation growth conditions of step c). The method may further or alternatively comprise the step of repeating steps a) to c) for multiple plant cultivars, and ranking each of said multiple plant cultivars in order of calculated duration of vernalisation growth conditions of step c), to select a desired plant cultivar for a given environment. The method may yet further comprise the step of selecting said plant cultivar for use in a plant breeding programme in the event that the plant cultivar has a desired calculated duration of vernalisation growth conditions, or has a desired classification as spring, semi-winter, or winter. In some embodiments, the method may further comprise the step of repeating steps b) and c) for different vernalisation growth conditions, for example, for different vernalisation temperatures, to determine a range of durations of vernalisation growth conditions. This may be useful in determining whether vernalisation requirements may change at different temperatures. The foregoing features of certain embodiments may also apply to the following additional aspects of the invention. In a further aspect of the invention, there is provided a method for predicting vernalisation requirements of a plant cultivar, the method comprising the steps of: a) determining the relative contributions of each of a plurality of paralogues of FLC to the total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the proportion of said FLC paralogues which are cold- responsive; and c) determining the vernalisation growth conditions for said plant cultivar based on the relative contributions of cold-responsive FLC paralogues and non-cold- responsive paralogues to total FLC expression levels of step a).

Thus, where a greater proportion of FLC paralogues are highly cold-responsive, then a reduction in total FLC expression will be relatively swift and the plant will have a lesser vernalisation requirement than a cultivar with a greater proportion of non-cold- responsive paralogues.

The invention further provides a method for selecting a plant cultivar for use in a breeding programme to obtain a desired ecotype, the method comprising: predicting vernalisation requirements of the plant cultivar in accordance with the first or second aspects of the present invention; and selecting the plant cultivar for use in said breeding programme in the event that the predicted vernalisation requirements are consistent with the desired ecotype.

Still further provided is a method of breeding a plant cultivar having a desired ecotype, the method comprising selecting a plurality of plant cultivars for use in a breeding programme as described above, and crossing said plant cultivars in a breeding programme to generate progeny.

The method may further comprise the step of predicting the vernalisation requirements of said generated progeny, as described herein, to determine whether the progeny has the desired ecotype.

A further aspect of the invention provides a method for selecting a plant cultivar having a desired ecotype, the method comprising the steps of: a) determining the relative contributions of each of a plurality of paralogues of FLC to the total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the proportion of said FLC paralogues which are cold- responsive; and c) selecting a plant cultivar having a greater contribution to total FLC levels from cold-responsive FLC paralogues in the event that a spring ecotype is desired, or selecting a plant cultivar having a greater proportion to total FLC levels from non-cold- responsive FLC paralogues in the event that a winter ecotype is desired.

The plant cultivar may be selected for cultivation; or may be selected for participation in a breeding programme.

A still further aspect of the present invention provides a method for generating a model for determining the decay rate of an FLC paralogue under vernalisation conditions, the method comprising recording decay rates for said FLC paralogue in a plurality of plant cultivars under vernalisation conditions, and combining said recorded decay rates to provide a modelled decay rate.

The recorded decay rates may be combined using any suitable statistical method; for example, in some circumstances a simple mean or median value may be used.

Said plurality of plant cultivars may consist of plant cultivars having the same haplotype in respect of the FLC paralogue. This may assist in reducing variation arising from FLC expression against different genetic backgrounds.

In some embodiments, the plurality of plant cultivars may comprise plant cultivars having multiple FLC haplotypes; the model may comprise multiple decay rates determined for each separate haplotype.

Also provided is a model generated by such a method.

In some embodiments, the model may include determined decay rates for a plurality of FLC paralogues.

BRIEF DESCRIPTION OF FIGURES

The invention is further described in the following non-limiting figures:

Figure 1 : Total FLC expression corresponds well to vernalisation requirement when differences in expression response to cold are accounted for. A) Inconsistencies between pre-vernalisation FLC expression, and vernalisation requirement are resolved by differences in response of FLC expression to cold. Time at which total FLC expression drops below a threshold value in each accession is consistent with cold requirement for competence to flower. Measured total FLC (summed expression of all 9 FLC genes) is plotted against days of vernalisation at 5°C (points). Mean and 95% confidence interval for mean plotted for each accession assuming an exponential decay model (line and ribbon). Horizontal line indicates approximate proposed floral competence threshold based on pre-vernalisation FLC expression in Westar. B) The fitted FLC response to cold is statistically different between accessions. 95% confidence interval range for the estimated exponential decay rate (cold response, b) of total FLC expression over vernalisation time. C) Below the threshold floral competence level, floral development rate corresponds well to remaining post-vernalisation total FLC. Predicted total FLC expression is derived from models fit to FLC expression over vernalisation for periods of 12 weeks (Ragged Jack), 6 weeks (Tapidor_JIC, Westar), or 3 weeks (Zhongshuang 11) as shown in A. The number of days post vernalisation to reach developmental stage BBCH51 (buds visible, Meier et al., 2009) were measured in separate experiments after 6- or 12-weeks of vernalisation. Mean values and 95% confidence intervals for mean values are plotted. The association between predicted FLC and time taken is surprisingly clear, considering that different individuals are considered, and for some cases the extrapolation of vernalisation times from the data used to fit the models. This is most extreme for Zhongshuang 11 (ZS11), in which final measurements at 3 weeks are used to make predictions at 12 weeks, which may explain its relatively poor agreement with the consensus pattern.

Figure 2: Differences in response of total FLC expression during vernalisation are a consequence of pre-vernalisation composition of FLC paralogues, and of differences in the cold response of the same loci between accessions. A) Experimental and fitted models for expression of individual FLC paralogues over vernalisation. Y-axis is log scale, meaning that the gradient of the lines is equal to the exponential decay rate b. Differences in this rate are clearly seen between the major expressed paralogues. B) Confidence intervals show that differences between paralogues, and between accessions at a single locus can be statistically significant. 95% confidence interval range for the estimated decay rates in the different FLC paralogues among the six-accession panel. C) Predicted expression levels of individual FLC paralogues, stacked to show contribution to total FLC level in different accessions. Variation in the relative expression of cold responsive and unresponsive paralogues, as well as cold response at each locus contribute to the quantitatively different behaviour of total FLC over cold. Figure 3: Divergent expression dynamics are associated with relaxed selection on coding sequences. A) BnaFLC genes, clustered by similarity in expression during vernalisation. Dendrogram shows clustering by Euclidean distance between expression profiles during vernalisation in the six core accessions. Three main expression types are observed (A, B, and C) corresponding to cold stable expression (A), moderate decay rate through vernalisation (B) and rapid decay or low expression levels (C). B) A phylogenetic tree based on codon aligned coding sequences with Arabidopsis FLC {AtFLC) as an outgroup. Bootstrap confidence levels (percentage of 1000 replicates) are indicated at each node. Colour scale indicates the ω ratio calculated along each branch. Asterisks indicate the significance levels of the RELAX test for relaxed selection (* p < 0.05, ** p < 0.01).

Figure 4: Pre-vernalisation FLC levels do not distinguish well between ecotypes.

Each point corresponds to FLC expression in a 3-week old un-vernalised leaf sample. Although total FLC expression is associated with ecotype (as there is a difference in the mean expression level between spring and winter types), it does not distinguish between ecotypes (as the distributions of expression levels in spring and winter ecotypes overlap), and so it is not sufficient to explain why an accession has a spring type or a winter type vernalisation requirement. Similarly, expression of some individual paralogues {Bna.FLC.A03b [BNAA03G13630D], Bna.FLC.A10, Bna.FLC.C02) are statistically associated with the spring, winter split (Schiessl et al., 2019), but are not sufficient to separate them, either considered individually or in combination {Bna.FLC.A03b+ Bna.FLC.A10+ Bna.FLC.C02).

Figure 5: Different FLC composition strategies exist among winter ecotype accessions

Pre-vernalisation FLC composition of winter ecotype accessions, clustered by similarity in FLC paralogue expression. Dendrogram shows clustering by Euclidean distance between pre-vernalisation FLC expression in each accession. Horizontal bars show cluster membership if cut at height of 40 (blue), or 20 (red). Bar-plot shows pre- vernalisation FLC expression, FLC paralogues are ordered by approximate cold stability from most stable {BNAC09G46540D) to least {BNAA03G02820D). There are three broad FLC compositions (A, B, C groups) within winter type accessions. The B group has low total expression, but proportionately more stable paralogue expression. The A & C groups have higher total FLC expression, the A group has high expression of the rapidly decaying FLC.A3a (BNAA03G02820D) copy. Within these groups, further subtypes exist (see e.g. red clusters). For example: the A group is further split by whether Bna.FLC.A 10 is expressed.

Figure 6: Difference in decay rate between accessions is required to achieve realistic vernalisation predictions.

The predicted FLC expression levels over vernalisation for RIPR panel accessions. A threshold FLC expression for floral competence is shown at 75 TPM. Bold lines show modelled FLC expression over vernalisation for a six-accession panel, based on measured expression (also shown in fig 1A). The time at which each accession crosses the floral competence threshold is shown with vertical dashes. A threshold of 75 TPM is used as this is the measured pre-vernalisation expression level of Westar, which has no vernalisation requirement, and is taken as an approximation of the required level to be able to flower upon cessation of cold treatment.

Day 0 expression is taken from measured expression in the leaf at 21 days. Decay parameters are based on values measured in six accessions, under different assumptions; left) common response; all decay rates are the same for all paralogues in all accessions, middle) paralogue specific response; FLC paralogues have different cold response rates, which is set to the mean measured value or each paralogue, right) paralogue & accession specific response; the same paralogue of FLC may have different decay rates in different accessions. The decay rate for each paralogue in each accession is set to the most extreme maximum-likelihood value estimated for that FLC paralogue among the six-accession panel, in order to maximise the difference between the low vernalisation (spring, & semi-winter) and high vernalisation (winter & swede) groups.

Only by accounting for differences in cold response between paralogues and accessions, is good separation of ecotypes and realistic vernalisation requirements for winter ecotypes predicted (vertical dashes).

Figure 7: Ragged Jack is an extreme winter type relative to the winter oilseed cultivars Tapidor_JIC and Express-617

Days taken after removal of plants from vernalisation to achieve BBCH51 developmental stage (buds visible). Results are plotted for individual plants across multiple trials. That Ragged Jack has a stronger vernalisation requirement that the other winter types (Express-617 and Tapidor_JIC) can be seen in the response to 6 weeks of vernalisation. Values of 144 days indicate that the individual did not flower before the end of the experiment, 144 days after vernalisation finished.

Figure 8: VIN3 is rapidly induced, meaning that epigenetic dependent and independent vernalisation periods are not distinguishable under experimental conditions. Plots show gene expression against days from germination, vernalisation treatment at 5°C is carried out between vertical lines. Total (summed) VIN3 expression is high at the first sampling timepoint, 1 day into vernalisation, indicating that the epigenetic dependent and independent periods of vernalisation are indistinguishable under these experimental conditions.

Figure 9: FLC gene sequences are sufficiently different that RNA-seq can distinguish them. Simulated 150 bp paired-end reads generated from each of the FLC gene models were aligned to the Darmor-bzh reference genomes (Chalhoub et al., 2014), using the same alignment pipeline as for the real data. For each generative template sequence (facets), the number of reads mapping to each of the FLC gene models are plotted (colours). Divergence between the template sequence used to generate the reads, and the reference sequence of the “correct” FLC paralogue in the reference sequence aligned against was also considered (x-axis). For each parameter combination, 10 independent simulations were run, the mean result and estimated 95% confidence limits are plotted. For all FLC paralogues, the true generative paralogue can clearly be distinguished, even for relatively high levels of divergence from the reference sequence. For comparison, the maximum experimentally identified coding sequence divergence from the reference sequence within a single accession in the RIPR panel was 1.68%, in BnaC03g04170D (Table 2).

Figure 10: As Fig 7, but for publicly available 100bp single-end RNA reads. Although mis-mapping rates are higher than for paired-end reads, they still map to the correct generative paralogue assuming moderate to low sequence divergence from the reference sequence. All paralogues can be clearly distinguished with some moderate mis-mapping of reads generated from FLC.A10 to FLC.C09a and FLC.C09b, and from FLC.C09a to FLC.A10 and FLC.C9b.

DETAILED DESCRIPTION OF THE INVENTION The present invention will now be further described. In the following passages, different aspects of the invention are defined in more detail. Each aspect so defined may be combined with any other aspect or aspects unless clearly indicated to the contrary. In particular, any feature indicated as being preferred or advantageous may be combined with any other feature or features indicated as being preferred or advantageous.

The practice of the present invention will employ, unless otherwise indicated, conventional techniques of botany, microbiology, tissue culture, molecular biology, chemistry, biochemistry and recombinant DNA technology, bioinformatics, which are within the skill of the art. Such techniques are explained fully in the literature.

The duration of cold that plants need to be exposed to prior to flowering (vernalisation requirement) is a major determinant of when they flower (reviewed Xu & Chong, 2018). The allotetraploid crop Brassica napus (B. napus, rapeseed, canola) is bred for different environments in order to maximise time available for growth, whilst reducing the exposure to adverse environmental conditions and fitting within the land management cycle (Marjanovic-Jeromela et al., 2007; Canola Council of Canada, 2013). Consequently, B. napus cultivars can be grouped into “ecotypes”. Spring types have essentially no cold requirement and are planted and harvested in the same season, with no over-wintering. Semi-winters have a weak vernalisation requirement and are grown in regions with mild winters. Winter types are often characterised as having a strong vernalisation requirement, and need to experience an extended period of cold prior to flowering (reviewed by Leijten et al., 2018). Within each ecotype, further variation exists in the amount of cold required for flowering, and in the sensitivity of flowering time to vernalisation treatment (Schiessl et al., 2017).

In the diploid Arabidopsis thaliana, natural variation in vernalisation requirement between accessions is largely a consequence of differences in pre-vernalisation expression of FLOWERING LOCUS C (FLC) (Johanson et al., 2000; Lempe et al., 2005; Bloomer & Dean, 2017). Much is known about FLC in Arabidopsis (AtFLC) (Fornara et al., 2010; Hepworth & Dean, 2015; Xu & Chong, 2018). FLC protein transcriptionally represses the central flowering regulators FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1), preventing premature flowering (Lee et al., 2000; Michaels et al., 2005). During winter, AtFLC undergoes cold-induced silencing through reduced transcription and epigenetic modifications at the AtFLC locus by the PHD-PRC2 complex (reviewed by Song et al., 2013). After a sufficient period of cold, AtFLC expression is reduced enough that it can no longer prevent flowering. As repression of AtFLC is partly epigenetic, it provides a “memory” of cold, meaning that upon a return to warm conditions in spring, AtFLC expression remains low, allowing flowering to be triggered by a combination of environmental and internal signalling pathways (Michaels & Amasino, 1999; Searle et al., 2006).

The role of FLC is similar in other Brassicaceae, however polyploids such as B. napus have multiple copies of genes (paralogues), many of which exhibit differences in regulation (Jones et al., 2018). It was recently reported that flowering time genes are over-represented in the B. napus genome relative to other families (Jones et al., 2018), in particular, FLOWERING LOCUS C (FLC) has nine copies in B. napus. Over- retention of paralogues suggests they may be dosage sensitive genes, retained to maintain stoichiometric expression balance, or alternatively, that they may play a role in facilitating adaptation and the acquisition of new functionality (Maere et al., 2005). A persuasive explanation for the retention of paralogues is the gene dosage hypothesis, in which gene loss would lead to negative consequences by perturbing the stoichiometric ratio of gene products (Birchler & Veitia, 2010, 2012). Dosage-balance selection acts at the level of total gene expression, allowing individual paralogues to drift in their expression as long as this effect is compensated for by other paralogues (Thompson et al., 2016). Thus, within the context of a polyploid organism not all paralogous copies of a gene are necessarily equally important for its canonical function and reduced selection pressure arising from multiple copies can allow sub-, or neofunctionalization to a different role (Conant & Wolfe, 2008; Hua et al., 2009; Yu et al., 2018; Jones et al., 2018; Wu et al., 2019). By bringing together two paleo-polyploid genomes with unique evolutionary histories, Brassica napus has assembled a large pool of FLC paralogues with potentially wide variation in expression profiles and/or function to draw upon when adapting to different flowering requirements. How, and whether all FLC paralogues contribute to vernalisation requirement and determine flowering behaviour is therefore not clear.

Indeed, there is much evidence that only certain FLC paralogues contribute to vernalisation. Expression of some BnaFLC paralogues are found to be unresponsive to cold (Schiessl et al., 2019). This has been interpreted to indicate that they may no longer be involved in the cold requirement machinery and may have sub-functionalised to some different role in other processes besides cold response in flowering regulation (Schiessl et al., 2019). Consistent with this, associative genomics studies using different panels of accessions have identified sequence or expression variation in only a subset of paralogues (BnaFLC. A02, BnaFLC.A03b, BnaFLC.AW, and BnaFLC.C02) as associated with cold required for flowering (Hou et al., 2012; Wu et al., 2012, 2019; Raman et al., 2013; Song et al., 2020). See Table 1 for paralogue naming conventions. Further evidence for the roles of specific paralogues was provided by the demonstration that a combination of BnaFLC.AIO with a defined transposable element together with a functional BnaFLC.A02 is required for winter ecotypes (Yin et al., 2020).

Here, we demonstrate that when FLC expression over vernalisation is considered, total BnaFLC expression (rather than individual paralogues) best explains variation in the vernalisation requirements between rapeseed ecotypes, suggesting that all expressed paralogues of BnaFLC are important in determining ecotype cold requirement. This finding is consistent with the gene dosage hypothesis (Conant et al., 2014; Cheng et al., 2018) and drift compensation model (Thompson et al., 2016), which can be extended naturally to expression dynamics during development.

We show that the expression of different BnaFLC paralogues decline at different rates during vernalisation, (suggesting that the paralogues may have sub-functionalised within their roles as quantitative environmental sensors in the vernalisation pathway) and find evidence that reduced selection pressure has potentiated this divergence of response.

In Arabidopsis, flowering time is sensitive to the rate of reduction of AtFLC expression in cold as well as pre-vernalisation expression levels (Bloomer & Dean, 2017; Takada et al., 2019; Song et al., 2020). In B. napus, we also see variation in sensitivity of expression over cold of each BnaFLC paralogue between accessions and find that this variation (in addition to pre-vernalisation expression levels) is important in determining vernalisation requirement.

These findings suggest that total BnaFLC expression can be controlled by many strategies, as expression of one paralogue can be compensated for by a cultivar of combinations of expression of the others. Consistent with this, we see examples of numerous different FLC composition strategies which result in the same ecotype within the well-studied RIPR accession panel (Havlickova et al., 2017; Schiessl et al., 2019).

Results Total FLC expression over vernalisation treatments corresponds to expected duration of cold requirement

We hypothesised that different ecotypes of B. napus will exhibit different total FLC expression dynamics. This was based on previous reports that link FLC expression to vernalisation requirements in Arabidopsis accessions (Johanson et al., 2000; Lempe et al., 2005; Bloomer & Dean, 2017), that flowering time depends on the decline of FLC expression during cold (Bloomer & Dean, 2017; Takada et al., 2019; Song et al., 2020), and that expression levels can be tuned to be optimal for their environment (Dekel & Alon, 2005).

To test this hypothesis, we studied BnaFLC expression over a vernalisation time- course in six exemplar accessions of different ecotypes. Stellar and Westar are spring oil seed rape (OSR) accessions, Zhongshuang 11 is a Chinese semi-winter OSR, Tapidor_JIC and Express-617 are winter OSR accessions, and Ragged Jack is a Siberian kale accession with a stronger vernalisation requirement (Fig 7).

In Arabidopsis, AtFLC expression changes during vernalisation have been described by a biphasic exponential decay model, corresponding to periods of epigenetically independent and dependent downregulation, as diagnosed by low and high expression of the PHD protein VIN3 respectively (Hepworth et al., 2018). However, under experimental vernalizing conditions, VIN3 expression rapidly increased (Fig 8), so monophasic models were sufficient. Exponential decay models were fit to the total BnaFLC expression data with parameters corresponding to pre-vernalisation expression levels and expression decay rates during vernalisation (also referred to here as “cold response”).

As shown in Fig 1A, pre-vernalisation total BnaFLC expression does not correspond well to vernalisation requirement. For example, the winter cultivar Tapidor_JIC has higher pre-vernalisation total BnaFLC expression than Ragged Jack but requires a shorter vernalisation period (Fig 7). Furthermore, the semi-winter Zhongshuang 11 has similar pre-vernalisation BnaFLC expression levels to the winter type accessions Express-617 and Ragged Jack. However, there are clear differences in the rate at which total BnaFLC expression declines under vernalisation between cultivars. For example, the initially high levels of BnaFLC seen in Zhongshuang 11 decay very rapidly under these vernalising conditions. These differences are statistically significant given the exponential decay assumption (Fig 2B). If pre-vernalisation total BnaFLC expression in the spring cultivar Westar is taken as the approximate level which is insufficient to prevent flowering, the order in which the cultivars are predicted to cross this threshold corresponds to their vernalisation requirement based on ecotype. This assumed threshold level is also consistent with the experimental observation that Ragged Jack requires approximately six weeks (42 days) vernalisation at 5°C for floral competence (Fig 7). Below this threshold, predicted total BnaFLC expression levels after 6- and 12-weeks vernalisation correspond well to the time taken to reach BBCH51 developmental stage (buds visible, Meier et al., 2009) after removal from vernalising conditions (Fig 1C).

We conclude that total BnaFLC expression dynamics can explain different cold requirements between B. napus ecotypes.

Differences in total BnaFLC cold response are caused both by differences between paralogues within an accession and differences in homologue behaviour between accessions

We next asked what might cause of the different rates of decline in total BnaFLC expression. We hypothesised that different FLC paralogues may have different expression decay rates. To investigate this, we examined the expression of the nine individual paralogues in each B. napus accession.

As different BnaFLC paralogues are highly similar, being up to 98.48% identical at the sequence level (BnaA03g02820D vs BnaC03g04170D), we first ensured that RNA sequencing data was able to distinguish between the paralogues. Under simulation, paired-end reads are well able to distinguish between paralogues (Fig 9, Table 2). For shorter, single-end RNA seq data (Havlickova et al., 2017; Jones et al., 2018), more mis-mapping can be expected, but we still see that paralogues can be distinguished well (Fig 10).

As previously reported, some paralogues are expressed at different levels in the different accessions prior to vernalisation (Schiessl et al., 2019). Interestingly, Express- 617 (a winter type accession), has spring-type levels of pre-vernalisation BnaFLC. A10 (Fig 2A), which is a commonly identified spring-winter discriminatory BnaFLC paralogue (Hou et al., 2012; Schiessl et al., 2017, 2019; Wu et al., 2019; Song et al., 2020). However, this is compensated for by higher expression of a combination of other BnaFLC paralogues, notably BnaFLC.A02; BnaFLC.C02 BnaFLC.C03a, resulting in a high total level of BnaFLC. This finding emphasises the importance of considering the BnaFLC loci in combination, rather than individually.

In addition to differences in pre-vernalisation expression level, statistically significant differences in cold expression response exist between paralogues (Fig 2A, B). For example, the BnaFLC. A03a (BnaA03g02820D) paralogue is extremely cold responsive: its expression declines rapidly across all tested accessions in the first three weeks of vernalisation. In contrast, expression of the BnaFLC. C09b (BnaC09g46540D) locus is extremely stable during vernalisation: in all accessions it either maintains, or slightly increases expression during cold. Other paralogues have varied cold responses between these two extremes. Interestingly, there are also significant differences in the rate at which expression declines between accessions for equivalent FLC paralogues, including in the BnaFLC. A02, BnaFLC. A10, and BnaFLC. C02 copies, genes which have recently been associated with ecotype differences (Song et al., 2020; Yin et al., 2020).

In order to assess the relative importance of the pre-vernalisation BnaFLC composition of highly cold responsive and unresponsive BnaFLC paralogues, as compared to differences in responsiveness at the same loci between accessions, we plotted how the individual paralogues are predicted to contribute to the total BnaFLC level over cold days (Fig 2C). This analysis suggests that Zhongshuang 11 has a low vernalisation requirement compared to winter ecotypes primarily because the rapid BnaFLC.A03a type makes up a large proportion of its total pre-vernalisation FLC. Although expression of many other paralogues decays rapidly in Zhongshuang 11 relative to other accessions, they are not initially expressed at a high level, so that their large cold response has relatively little effect on total BnaFLC response. In contrast, Ragged Jack appears to require an extreme vernalisation treatment relative to Tapidor_JIC and Express-617, partly because its pre-vernalisation FLC composition is weighted away from BnaFLC. A03a, towards more stable paralogues, but also because many paralogues of FLC appear to be less cold responsive than in Express-617 or Tapidor_JIC. For example, BnaFLC. A10 is highly expressed in both Tapidor_JIC and Ragged Jack, but is more stably expressed in the latter.

Together, these analyses suggest that total FLC expression dynamics, comprising individual FLC paralogues that have diverged in their response to cold, determine vernalisation requirements in the six studied B. napus accessions. Divergent expression dynamics are associated with relaxed selection strength acting on coding sequences

In order to gain insights into the diversity of expression dynamics observed for the multiple BnaFLC paralogues, as well as the origin of this variation, we first clustered the BnaFLC genes based on their expression dynamics in the six accessions (Fig 3A). This clustering identified three main expression types (A, B & C). Cluster A comprised genes with relatively cold stable expression (BnaFLC. C02, BnaFLC. C09b). Cluster B comprised genes with a moderate expression decay rate (BnaFLC. A10, BnaFLC. A02) and cluster C consisted of genes that had either a rapid decay rate or low overall levels of expression (BnaFLC. A03a, BnaFLC. A03b, BnaFLC. C03a, BnaFLC. C03b, BnaFLC. C09a). Notably, this clade included all genes that arose through tandem or segmental duplications (BnaFLC. A03b, BnaFLC. C03b, BnaFLC.C09a), including the presumed pseudogene (BnaFLC. C03b).

The gene balance hypothesis (Birchler & Veitia, 2010) predicts that duplicates of dosage sensitive genes are less likely to be selectively maintained following segmental duplication than following whole genome duplication (Conant et al., 2014; Cheng et al., 2018). We wondered whether the divergent expression types seen for type C genes might therefore be associated with altered signatures of selection. To investigate this further we determined the ω ratio (dN/dS) along branches of an FLC phylogenetic tree. All genes belonging to expression types A and B showed ω values below 0.5 consistent with purifying selection (Fig 3B). In contrast, four out of five type C expression genes had ω values at or above 1, consistent with relaxed or positive selection (Fig 3B). We subsequently used RELAX, a test based on random effects branch-site evolutionary models (Wertheim et al., 2014) to formally compare the strength of selection acting on expression type C genes, to that acting on type A and B genes. Importantly RELAX can efficiently distinguish between increases in ω due to relaxed selection and positive selection. The RELAX test demonstrated that coding sequences of type C genes have experienced a significant reduction in selection strength compared to those of expression type A and B (p = 0.0017). We re-ran the analyses for individual genes or pairs of genes with similar expression profiles in clade C. Relaxed selection was confirmed for BnaFLC. A03a I BnaFLC. C03a (p = 0.0012) and for the pseudogene BnaFLC. C03b (p = 0.045) but not for BnaFLC. A03b (p = 0.641) suggesting that this last gene is still experiencing purifying selection, consistent with its lower ω value (Fig 3B). BnaFLC.C09a is a very recent duplicate and had therefore accumulated insufficient substitutions (a single non-synonymous substitution) for meaningful analysis. These findings are consistent with the hypothesis that individual paralogues may drift in their expression and that reduced expression may reduce selection pressure and promote neo- and sub-functionalisation (Thompson et al., 2016). In the case of BnaFLC.A03a, and BnaFLC.C03a this sub-functionalisation apparently takes the form of an increased cold response sensitivity, diversifying the range of cold response rates available to the polyploid within the FLC gene group.

Different strategies for pre-vernalisation BnaFLC composition exist within the winter ecotype

The finding that total FLC dynamics is determined by combining paralogues with different cold responses, suggested to us that multiple different such combinations could be employed to achieve a similar outcome. In order to assess this inference we considered expression of BnaFLC paralogues in a subset of the RIPR panel comprising 37 spring, 43 winter, 6 semi-winter ecotypes, and 9 swede types (Havlickova et al., 2017). The RIPR panel data was collected pre-vernalisation (Havlickova et al., 2017). This panel includes the six cultivars measured in time-course over vernalisation.

Consistently among the semi-winter cultivars included in the RIPR panel, a comparatively large proportion of the total BnaFLC is made up of the highly responsive BnaFLC.A03a copy (Fig 4). This fits with a simple model, in which prior to vernalisation the BnaFLC. A03a copy inhibits flowering, but rapidly decays with cold, leaving the total other paralogues insufficiently expressed to prevent flowering. Conversely the swede types, which have an extreme, long vernalisation requirement, commonly have a low BnaFLC. A03a to BnaFLC. A10 expression ratio relative to other accessions.

As previously reported, although total pre-vernalisation BnaFLC expression is associated with ecotype, it is a poor discriminant between spring and winter types in this panel (Fig 4, total FLC subplot), and pre-vernalisation expression levels of BnaFLC.AW, BnaFLC. A03b, and BnaFLC. C02 provide the best discriminants between spring and winter ecotypes in this data (Schiessl et al., 2019). However, pre-vernalisation FLC expression in the RIPR panel makes it clear that Express-617 is not unusual in having atypical expression of one of these key paralogues for its ecotype. Fig 4 shows that expression of these paralogues is insufficient to explain the spring-winter ecotype split: the range of expression levels of every individual paralogue of BnaFLC shows a large overlap between spring and winter ecotypes. Of the 37 spring and 43 winter accessions, 8 spring types and 4 winter types have non-canonical expression of BnaFLC. A03b. 3 spring accessions, and 11 winter accessions have non-canonical BnaFLC. A10 expression levels, and the majority of spring types have greater BnaFLC.C02 expression than the lowest winter type.

It is therefore insufficient to consider each FLC paralogue in isolation to explain the cold requirements of these accessions. Moreover, a simple combination of pre- vernalisation expression of the paralogues which have previously been identified as best distinguishing spring from winter are still not able to separate them. Considering BnaFLC. A03b+BnaFLC.A10+BnaFLC.C02', 34 spring lines express more of these than the lowest winter line, and 12 winter lines express less than the highest spring line.

Among the winter-type accessions, several different “strategies” for the composition of pre-vernalisation FLC have arisen (Fig 5). Broadly, there are two different high pre- vernalisation total FLC strategies (groups A and C), and one lower total FLC group (group B). Members of group B express proportionately more of the cold-stable FLC paralogues identified in the previous section, whereas for example in group A, approximately half of the total FLC is composed of the rapidly cold responsive BnaFLC.A03a copy. Further nested subgroups are evident within each of these broad strategies. For example; within both groups A and B, there are examples of spring and winter type BnaFLC. A10 expression, which are compensated for by expression levels of various other paralogues.

These results confirm that high pre-vernalisation total FLC does not imply a strong vernalisation requirement, and importantly show that different combinations of paralogue expression are employed to achieve the same (winter) ecotype.

Total BnaFLC explains cold requirement in the RIPR diversity panel if differences in the response of FLC expression to cold are considered Pre-vernalisation FLC expression is only weakly associated with cold requirement. To assess whether differences in the cold response of BnaFLC expression are sufficient to explain variation in cold requirement between ecotypes, we modelled the predicted level of BnaFLC expressed in the RIPR lines during vernalisation treatment (Fig 6). We used the measured pre-vernalisation BnaFLC expression levels as a starting condition for each accession. The decay rate parameter was based on the values measured in the six accessions previously under three different assumptions. It was either set to the mean measured cold response value for all paralogues of BnaFLC, ignoring differences between paralogues (Fig 6, left), set to the mean value across accessions for each individual paralogue, meaning for example that BnaFLC. A03a decays relatively rapidly (Fig 6, middle), or set to the most extreme measured response for each paralogue among the six accession panel, so as to best separate spring and semi-winter (rapid cold response) from winter and swede ecotypes (slow cold response) (Fig 6, right). In each panel, bold lines show the predicted values using the accession models fit to measured BnaFLC levels during vernalisation as previously discussed (see Fig 1A). Vertical dashes indicate when each accession crosses below the Westar derived total FLC threshold level and are consequently predicted to be competent to flower.

If differences in cold response between paralogues are not taken account (Fig 6, left), then separation in predictions for required cold treatment between ecotypes is poor, reflecting the low association of pre-vernalisation total BnaFLC with ecotype and its poor predictive power as a classifying criterion.

If different BnaFLC paralogues are allowed different decay rates (Fig 6, middle), the results are slightly better in terms of ecotype separation, reflecting the fact that differences in pre-vernalisation BnaFLC composition between accessions contribute to differences in their cold requirement. However, this simple model predicts unrealistically short cold requirements in winter and swede cultivars, which often require more than 12 weeks vernalisation (Schiessl et al., 2017).

To separate spring and winter ecotypes, and predict realistic cold period requirements for winter accessions, differences in the cold response of individual paralogues between accessions has to be allowed for (Fig 6, right). Under this assumption, three winter types (Falcon, Resyn-H048, Zenith) are predicted to require less cold than the most cold-dependent spring type. This is also the only tested assumption under which measured expression in the winter accessions Express-617 and Tapidor_JIC is not misrepresentative of the predicted BnaFLC expression for winter accessions in the RIPR panel.

This approach successfully demonstrates that differences in cold response of BnaFLC expression are sufficient to largely explain ecotype differences without needing to invoke additional factors such as differences in the potency of BnaFLC paralogues as repressors of flowering. Taken together, this data, analysis and modelling support the view that total BnaFLC expression dynamics rather than individual paralogues are the key factor determining vernalisation requirements.

Discussion

The allotetraploid crop Brassica napus has nine copies of FLC, a key gene that determines a plant’s vernalisation requirement and associates with adaptation to different agricultural environments by adjusting flowering time. Individual FLC paralogues have been linked to different B. napus ecotypes (Hou et al., 2012; Wu et al., 2012, 2019; Raman et al., 2013; Schiessl et al., 2019; Song et al., 2020; Yin et al., 2020); yet there are discrepancies between observed and expected FLC expression levels based on an ecotype classification of accessions; and transferring seemingly specific ecotype-defining paralogues into a different ecotype has not resulted in the expected changes (Yin et al., 2020).

Here, we investigated how multiple paralogues of the flowering gene FLC may function as a system to determine vernalisation requirements in B. napus. We demonstrated that when BnaFLC expression response to cold is considered in addition to pre- vernalisation expression levels, total BnaFLC expression can explain cold requirement in B. napus better than individual paralogues.

We suggest that differences between accessions in the response of total BnaFLC expression to vernalisation are controlled both by varying the pre-vernalisation composition of expressed BnaFLC paralogues (with different cold responses), as well as through differences in the cold responses of BnaFLC orthologues between accessions.

How then does such wide variation in the expression dynamics of different paralogues arise? Once example is provided by the drift compensation model (Thompson et al., 2016) which demonstrates how the expression of individual paralogues can move freely so long as other paralogues compensate for any introduced expression changes. One prediction of this model is that over time the expression level of some genes will reduce sufficiently that selection is relaxed, allowing the evolution of new function, new expression dynamics, or pseudogenisation. Consistent with this model, BnaFLC paralogues with low levels of expression such as BnaFLC. C03b or novel dynamics, such as the highly cold responsive paralog BnaFLC. A03a show evidence of relaxed selection.

It is also possible for variation to arise in the context of purifying selection. Of the paralogues that have ω < 0.5, those belonging to the C genome form one expression type (cold-stable), while those belonging to the A genome form another (cold responsive). This is despite each A genome paralogue being more closely related phylogenetically to its direct C genome homoeologue (Fig 3). This suggests that differential selection in the ~2.5-4.5 MY since divergence of the A and C genomes (Liu et al., 2014; Arias et al., 2014) but prior to the formation of Brassica napus may also have contributed to variation in expression dynamics. Thus, compensatory drift, relaxation of selection and differential selection all have the potential to generate diversity in expression dynamics which, when brought together in a single species provides a pool of variation that may be more rapidly evolvable than a single gene.

An example of how altered expression levels contribute to the rapid evolution of cold sensitivity within the context of FLC can be seen in the Chinese semi-winters (Fig 4). Since the introduction of winter accessions to China in the 1940s, selective breeding for a milder cold requirement has apparently led to an increase in the relative pre- vernalisation expression of the highly cold responsive BnaFLC. A03a paralogues, and decrease in the expression of cold stable paralogues, resulting in a dramatically different expression of total BnaFLC over vernalisation, and consequently a novel vernalisation response phenotype, distinct from both spring and winter ecotypes.

Based on the hypothesis that total FLC is important, we might expect to find different ways of implementing the same total BnaFLC behaviour through different combinations of paralogue dynamics. Consistent with this idea, within the winter ecotype, we observe a number of different strategies for managing total BnaFLC through variation in the pre- vernalisation BnaFLC composition of differently cold responsive paralogues. A consequence of this diversity of viable strategies is that association studies are more likely to be sensitive to the composition of the accessions studied, than for phenotypes where the strategy space is smaller. When paralogues can compensate for each other and produce the same phenotype, it is not clear whether a statistical association between a trait and a particular paralogue means that it is truly contributing more to producing a phenotype within an individual, or is instead associated with a more common strategy within the panel. Here, cold requirement provides a case study for the additional difficulties of gene association studies with polygenic traits.

Although we find that separation between the cold requirement of spring and winter ecotypes is much better after consideration of differences in the cold response of gene expression, the model is not perfect. For example, three winter accessions (out of 43) are not predicted to require more vernalisation than the most cold-requiring spring type. This is likely to be caused by a combination of several factors (1-4). 1) Although the ASSYST ecotype labels (spring, semi-winter, winter, swede) are a useful shorthand for vernalisation phenotype, they are not perfect; there is variation in cold requirement with each ecotype group, and the labels can be misleading. For example the winter cultivar Mansholt has been reported to flower without vernalisation, whereas the spring cultivars Giant-xr707 and Daichousen (fuku) do require vernalisation (Schiessl et al., 2017). 2) The hard FLC expression threshold for floral competence considered here is an approximation and a simplification. The level used was based on the pre- vernalisation BnaFLC expression of the spring cultivar Westar, and thus was considered to provide a lower bound on the true threshold level. However, any “threshold level” is also likely to vary between accessions, partly as a consequence of 3). 3) Variation at genes downstream, or independent of FLC expression in the vernalisation pathways can also affect cold requirement. For example, the low vernalisation requirement in Mansholt has been linked to sequence variation in the promoter of BnaFT.C02 (Schiessl et al., 2017). BnaFT.A02, BnaFT.C06a, and BnaFT.C06b have also been associated with two major QTL clusters for flowering time (Wang et al., 2009; Tudor et al., 2020), and may modify the effects of FLC. 4) The range of decay parameters allowed for each FLC paralogue were derived from measurements of only six accessions, and it is likely that more extreme FLC paralogue responses to cold exist within the RIPR panel.

Differences in the cold sensitivity of expression between paralogues suggest that, as in Arabidopsis (Li et al., 2014; Bloomer & Dean, 2017), cold sensitivity is at least partially a consequence of cis sequence differences at the BnaFLC loci. However, we also see common patterns across paralogues within accessions, (for example many paralogues have relatively stable expression in Ragged Jack). This suggests that functionally important differences in responsiveness are also caused by trans factors, which regulate multiple BnaFLC loci during vernalisation. Consistent with this, genes involved in chromatin modification at FLC have experienced selection during ecotype improvement (Lu et al., 2019). More thoroughly studying the range of cold responses in a broad panel of accessions is therefore an important next step in order to understand this system.

In summary, these findings support the idea that total BnaFLC expression, rather than individual BnaFLC paralogues, determine vernalisation requirement in B. napus. Central to this proposition is consideration of the dynamic cold response of BnaFLC expression, which we modelled using an exponential decay function. The proposed hypothesis provides a simple, mechanistically grounded explanation of an accession’s cold requirement, and thus is a useful framework from which to further study vernalisation in this polyploid system.

The importance of combined BnaFLC levels suggests that B. napus may have retained so many FLC paralogues, because (in the absence of significant feedback regulation) loss of a duplicate results in a quantitative difference in total BnaFLC expression level, which may be sufficient to lead to detrimental changes in phenotype. Notably, this is the case even for paralogues which are themselves unresponsive to vernalisation. Such considerations are consistent with gene-dosage selection (Conant et al., 2014) and drift compensation (Thompson et al., 2016) applied to expression dynamics, and suggest a means by which paralogue combinatorics could be exploited to potentiate phenotypic diversity in polyploids.

Methods

Plant growth conditions, RNA extraction, sample preparation and sequencing

Brassica napus cv. Stellar, Zhongshuang 11, Tapidor_JIC, Express-617, and Ragged Jack plants were sown in cereals mix (40% medium grade peat, 40% sterilised soil, 20% horticultural grit, 1.3 kg m^-3 PG mix 14-16-18 + Te base fertiliser, 1 kg m^-3 Osmocote Mini 16-8-11 2 mg + Te 0.02% B, wetting agent, 3 kg m^-3 maglime, 300 g m ³ Exemptor). Material was grown in a Conviron MTPS 144 controlled environment room with Valoya NS1 LED lighting (250µmol.m^¯2.s^¯1) 18°C day/15°C night,70 % relative humidity with a 16hr day. At day 21, plants were put into vernalisation at 5°C (8h day). The Tapidor accession used to generate this data is from a different seed lineage to that used in other studies (Havlickova et al., 2017), and during this analysis it became clear that differences exist in its FLC complement. Here this accession is referred to as Tapidor_JIC to differentiate it from the canonical Tapidor accession. At each sampling timepoint 3 replicates samples containing either three seedings or three first true leaves were collected. For Express-617 only one leaf sample was collected at each timepoint. Samples were ground in LN2 to a fine powder before RNA extraction and DNase treatment were performed following the method provided with the E.Z.N.A^® Plant RNA Kit (Omega Bio-tek Inc., http://omegabiotek.com/store/). RNA samples were processed at Novogene (Beijing); cDNA libraries were constructed using NEB next ultra-directional library kit (New England Biolabs Inc), sequencing was performed using Illumina HiSeq X, resulting in 150bp paired end reads. Bioinformatics Publicly available single-end fastq files of gene expression over vernalisation were downloaded from NCBI SRA, project ID PRJNA398789 (Jones et al., 2018). Publicly available single-end RNA seq data for pre-vernalisation gene expression data in the RIPR panel were downloaded from NCBI SRA, project ID PRJNA309367 (Havlickova et al., 2017). Gene expression quantification was carried out using HISAT v2.0.4 (Kim et al., 2015), & StringTie v1.2.2 (Pertea et al., 2015). Reads were aligned to the Darmor-bzh reference genome (Chalhoub et al., 2014), downloaded from http://www.genoscope.cns.fr/brassicanapus/data/. Phylogenetic analyses Coding sequences and corresponding protein sequences were curated based on aligned RNAseq reads, apart from BnaFLC.C03b which was not expressed in any sample. A codon aligned nucleotide sequence was then generated as described (Suyama et al., 2006) manually edited in Geneious Prime (v2020.0.3). Protein sequence alignment and Neighbour-Joining tree were generated using Geneious Prime with default settings. ω ratios (dN/dS) were calculated using the Ka/Ks webtool from CBU (http://services.cbu.uib.no/tools/kaks). The phylogenetic tree was plotted using the phytools R package (Revell, 2012). Simulation of expected RNA-seq mis-mapping rate between FLC paralogues Using the FLC gene models in the Darmor-bzh reference sequence (Chalhoub et al., 2014, genoscope.cns.fr/brassicanapus/data/), ArtificialFastqGenerator (Frampton & Houlston, 2012), was used to generate simulated reads, with read-length, and sequencing error rates sampled from real fastq files. Simulated reads were put through the same alignment pipeline as the real data. FLC expression model fitting FLC expression over vernalisation was assumed to follow an exponential function. For each paralogue of FLC and total FLC in each accession an ANOVA test was used to select between

which models FLC decay to zero given sufficient time and

which models FLC decay to some non-zero value c. where y is FLC expression measured in TPM, x is days of cold, and a, b, & c are fitted parameters. The best model was carried forward for analysis. For the prediction of RIPR panel FLC expression levels over vernalisation, model 1 was assumed. The measured FLC expression level at 21 days was used as α , and b was set according to the assumptions described in Fig 6. Table 1

as % of

Only minor divergence from the reference sequence is observed in the FLC coding sequence. The number of SNPs in FLC paralogues identified in the RIPR panel from mRNA data (Havlickova et al., 2017). The apparent true divergence rate is well within the simulated range. Arias T, Beilstein MA, Tang M, McKain MR, Pires JC. 2014. Diversification times among Brassica (Brassicaceae) crops suggest hybrid formation after 20 million years of divergence. American Journal of Botany 101: 86–91. Birchler JA, Veitia RA. 2010. The gene balance hypothesis: implications for gene regulation, quantitative traits and evolution. New Phytol 186: 54–62. Birchler JA, Veitia RA.2012. Gene balance hypothesis: Connecting issues of dosage sensitivity across biological disciplines. Proceedings of the National Academy of Sciences 109: 14746–14753. Bloomer RH, Dean C. 2017. Fine-tuning timing: Natural variation informs the mechanistic basis of the switch to flowering in Arabidopsis thaliana. Journal of Experimental Botany 68: 5439–5452. Canola Council of Canada.2013. Canola Encyclopedia: time-of-seeding. Chalhoub B, Denoeud F, Liu S, P Parkin IA, Tang H, Wang X, Chiquet J, Belcram H, Tong C, Samans B, et al. 2014. Early allopolyploid evolution in the post-Neolithic Brassica napus oilseed genome. Science 345: 950–953. Cheng F, Wu J, Cai X, Liang J, Freeling M, Wang X. 2018. Gene retention, fractionation and subgenome differences in polyploid plants. Nature Plants 4: 258–268. Conant GC, Birchler JA, Pires JC. 2014. Dosage, duplication, and diploidization: clarifying the interplay of multiple models for duplicate gene evolution over time. Curr Opin Plant Biol 19: 91–98. Conant GC, Wolfe KH. 2008. Turning a hobby into a job: How duplicated genes find new functions. Nature Reviews Genetics 9: 938–950. Dekel E, Alon U.2005. Optimality and evolutionary tuning of the expression level of a protein. Nature 436: 588–592. Fornara F, De Montaigu A, Coupland G. 2010. SnapShot: Control of Flowering in Arabidopsis. Cell 141: 550. Frampton M, Houlston R. 2012. Generation of Artificial FASTQ Files to Evaluate the Performance of Next-Generation Sequencing Pipelines (JH Badger, Ed.). PLoS ONE 7: e49110. Havlickova L, He Z, Wang L, Langer S, Harper AL, Kaur H, Broadley MR, Gegas V, Bancroft I.2017. Validation of an updated Associative Transcriptomics platform for the polyploid crop species Brassica napus by dissection of the genetic architecture of erucic acid and tocopherol isoform variation in seeds. The Plant Journal. Hepworth J, Antoniou-Kourounioti RL, Bloomer RH, Selga C, Berggren K, Cox D, Collier Harris BR, Irwin JA, Holm S, Säll T, et al.2018. Absence of warmth permits epigenetic memory of winter in Arabidopsis. Nature Communications 9: 639. Hepworth J, Dean C. 2015. Update on Flowering Locus C Flowering Locus C’s Lessons: Conserved Chromatin Switches Underpinning Developmental Timing and Adaptation. Plant Physiology 168: 1237–1245. Hou J, Long Y, Raman H, Zou X, Wang J, Dai S, Xiao Q, Li C, Fan L, Liu B, et al. 2012. A Tourist-like MITE insertion in the upstream region of the BnFLC.A10 gene is associated with vernalization requirement in rapeseed (Brassica napus L.). BMC Plant Biology 12: 238. Hua S, Shamsi IH, Guo Y, Pak H, Chen M, Shi C, Meng H, Jiang L. 2009. Sequence, expression divergence, and complementation of homologous ALCATRAZ loci in Brassica napus. Planta 230: 493–503. Johanson U, West J, Lister C, Michaels S, Amasino R, Dean C. 2000. Molecular analysis of FRIGIDA, a major determinant of natural variation in Arabidopsis flowering time. Science 290: 344–347. Jones DM, Wells R, Pullen N, Trick M, Irwin JA, Morris RJ. 2018. Spatio-temporal expression dynamics differ between homologues of flowering time genes in the allopolyploid Brassica napus. The Plant Journal 96: 103–118. Kim D, Langmead B, Salzberg SL. 2015. HISAT: a fast spliced aligner with low memory requirements. Nature Methods 12: 357–60. Lee H, Suh SS, Park E, Cho E, Ahn JH, Kim SG, Lee JS, Kwon YM, Lee I. 2000. The AGAMOUS-lIKE 20 MADS domain protein integrates floral inductive pathways in Arabidopsis. Genes and Development 14: 2366–2376. Leijten W, Koes R, Roobeek I, Frugis G, Leijten W, Koes R, Roobeek I, Frugis G. 2018. Translating Flowering Time From Arabidopsis thaliana to Brassicaceae and Asteraceae Crop Species. Plants 7: 111. Lempe J, Balasubramanian S, Sureshkumar S, Singh A, Schmid M, Weigel D. 2005. Diversity of Flowering Responses in Wild Arabidopsis thaliana Strains. PLOS Genetics 1: e6. Li P, Filiault D, Box MS, Kerdaffrec E, van Oosterhout C, Wilczek AM, Schmitt J, McMullan M, Bergelson J, Nordborg M, et al.2014. Multiple FLC haplotypes defined by independent cisregulatory variation underpin life history diversity in Arabidopsis thaliana. Genes and Development 28: 1635–1640. Liu S, Liu Y, Yang X, Tong C, Edwards D, Parkin IAP, Zhao M, Ma J, Yu J, Huang S, et al. 2014. The Brassica oleracea genome reveals the asymmetrical evolution of polyploid genomes. Nature Communications 5: 3930. Lu K, Wei L, Li X, Wang Y, Wu J, Liu M, Zhang C, Chen Z, Xiao Z, Jian H, et al. 2019. Whole-genome resequencing reveals Brassica napus origin and genetic loci involved in its improvement. Nature Communications 10: 1154. Maere S, De Bodt S, Raes J, Casneuf T, Montagu M Van, Kuiper M, Van De Peer Y. 2005. Modeling gene and genome duplications in eukaryotes. Proceedings of the National Academy of Sciences 102: 5454–5459. Marjanović-Jeromela A, Marinković R, Mijić A, Jankulovska M, Zdunić Z. 2007. Interrelationship between oil yield and other quantitative traits in rapeseed (Brassica napus L.). Journal of Central European Agriculture 8: 165–170. Meier U, Bleiholder H, Buhr L, Feller C, Hack H, Heß M, Lancashire P., Schnock U, Stauß R, Van Den Boom T, et al.2009. The BBCH system to coding the phenological growth stages of plants–history and publications. Journal für Kulturpflanzen 61: 41–52. Michaels SD, Amasino RM. 1999. FLOWERING LOCUS C encodes a novel MADS domain protein that acts as a repressor of flowering. Plant Cell 11: 949–956. Michaels SD, Himelblau E, Sang YK, Schomburg FM, Amasino RM. 2005. Integration of flowering signals in Winter-annual Arabidopsis. Plant Physiology 137: 149–156. Pertea M, Pertea GM, Antonescu CM, Chang T-C, Mendell JT, Salzberg SL.2015. StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. Nature Biotechnology 33: 290–295. Raman H, Raman R, Eckermann P, Coombes N, Manoli S, Zou X, Edwards D, Meng J, Prangnell R, Stiller J, et al.2013. Genetic and physical mapping of flowering time loci in canola (Brassica napus L.). Theoretical and Applied Genetics 126: 119– 132. Revell LJ. 2012. phytools: an R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution 3: 217–223. Schiessl S, Huettel B, Kuehn D, Reinhardt R, Snowdon R. 2017. Post- polyploidisation morphotype diversification associates with gene copy number variation. Scientific Reports 7. Schiessl S V., Quezada-Martinez D, Tebartz E, Snowdon RJ, Qian L. 2019. The vernalisation regulator FLOWERING LOCUS C is differentially expressed in biennial and annual Brassica napus. Scientific Reports 9. Searle I, He Y, Turck F, Vincent C, Fornara F, Kröber S, Amasino RA, Coupland G. 2006. The transcription factor FLC confers a flowering response to vernalization by repressing meristem competence and systemic signaling in Arabidopsis. Genes & Development 20: 898–912. Song JM, Guan Z, Hu J, Guo C, Yang Z, Wang S, Liu D, Wang B, Lu S, Zhou R, et al. 2020. Eight high-quality genomes reveal pan-genome architecture and ecotype differentiation of Brassica napus. Nature Plants 6: 34–45. Song J, Irwin J, Dean C. 2013. Remembering the prolonged cold of winter. Current Biology 23: R807–R811. Suyama M, Torrents D, Bork P. 2006. PAL2NAL: robust conversion of protein sequence alignments into the corresponding codon alignments. Nucleic Acids Research 34: W609–W612. Takada S, Akter A, Itabashi E, Nishida N, Shea DJ, Miyaji N, Mehraj H, Osabe K, Shimizu M, Takasaki-Yasuda T, et al.2019. The role of FRIGIDA and FLOWERING LOCUS C genes in flowering time of Brassica rapa leafy vegetables. Scientific Reports 9. Thompson A, Zakon HH, Kirkpatrick M. 2016. Compensatory Drift and the Evolutionary Dynamics of Dosage-Sensitive Duplicate Genes. Genetics 202: 765. Tudor EH, Jones DM, He Z, Bancroft I, Trick M, Wells R, Irwin JA, Dean C. 2020. QTL-seq identifies BnaFT.A02 and BnaFLC.A02 as candidates for variation in vernalisation requirement and response in winter oilseed rape (Brassica napus). Plant Biotechnology Journal n/a. Wang J, Long Y, Wu B, Liu J, Jiang C, Shi L, Zhao J, King GJ, Meng J.2009. The evolution of Brassica napus FLOWERING LOCUS T paralogues in the context of inverted chromosomal duplication blocks. BMC Evolutionary Biology 9. Wertheim JO, Murrell B, Smith MD, Kosakovsky Pond SL, Scheffler K. 2014. RELAX: Detecting Relaxed Selection in a Phylogenetic Framework. Molecular Biology and Evolution 32: 820–832. Wu D, Liang Z, Yan T, Xu Y, Xuan L, Tang J, Zhou G, Lohwasser U, Hua S, Wang H, et al. 2019. Whole-Genome Resequencing of a Worldwide Collection of Rapeseed Accessions Reveals the Genetic Basis of Ecotype Divergence. Molecular Plant 12: 30– 43. Wu J, Wei K, Cheng F, Li S, Wang Q, Zhao J, Bonnema G, Wang X. 2012. A naturally occurring InDel variation in BraA.FLC.b (BrFLC2) associated with flowering time variation in Brassica rapa. BMC Plant Biology 12: 1–9. Xu S, Chong K. 2018. Remembering winter through vernalisation. Nature Plants 4: 997–1009. Yin S, Wan M, Guo C, Wang B, Li H, Li G, Tian Y, Ge X, King GJ, Liu K, et al.2020. Transposon Insertions within Alleles of BnaFLC.A10 and BnaFLC.A2 are Associated with Rapeseed (Brassica napus L.) Seasonal Crop-type. J Exp Bot. Yu H, Jiao B, Lu L, Wang P, Chen S, Liang C, Liu W. 2018. NetMiner-an ensemble pipeline for building genome-wide and high-quality gene co-expression network using massive-scale RNA-seq samples. PLOS ONE 13: e0192613.

Claims

CLAIMS: 1. A method for predicting vernalisation requirements of a plant cultivar which expresses multiple FLC paralogues, the method comprising the steps of: a) determining total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the decay rate of total expression levels of FLC in said plant cultivar under vernalisation growth conditions; and c) calculating the duration of vernalisation growth conditions required for the total expression levels of FLC in said plant cultivar to reduce to below a predetermined threshold, based on the results of steps a) and b). 2. The method of claim 1, wherein the predetermined threshold of FLC expression levels is that expression level determined to permit flowering of the plant and which is experimentally determined based on a cultivar of the plant which does not require vernalisation. 3. The method of claim 1 or claim 2, wherein the plant is a polyploid plant. 4. The method of any preceding claim wherein the plant is selected from among the Brassicaceae, and preferably the plant is a Brassica spp. 5. The method of claim 4, wherein the plant is selected from B. napus, B. oleracea, B. rapa, and is preferably B. napus subsp. napus. 6. The method of any preceding claim wherein the step of determining the decay rate of total expression levels of FLC comprises determining the decay rate of each FLC paralogue. 7. The method of claim 6 wherein the decay rate is determined from a model for each FLC paralogue, wherein the model is preferably based on a training set of data including decay rates for a given FLC paralogue in one or more representative plant cultivars of the same species as the plant. 8. The method of any preceding claim wherein the decay rate is determined by calculating the relative contribution to total FLC expression levels of cold-responsive and non-cold-responsive paralogues and applying a weighted decay value to expression levels of cold-responsive and non-cold-responsive paralogues accordingly to determine decay of total FLC expression. 9. The method of claim 8 comprising the step of determining the presence or absence of an RY motif in an FLC paralogue, and classifying the paralogue as cold- responsive or non-cold-responsive accordingly. 10. The method of any preceding claim further comprising the step of classifying the plant cultivar as spring, semi-winter, or winter, based on the calculated duration of vernalisation growth conditions of step c). 11. The method of any preceding claim further comprising the step of selecting said plant cultivar for use in a plant breeding programme in the event that the plant cultivar has a desired calculated duration of vernalisation growth conditions, or has a desired classification as spring, semi-winter, or winter. 12. The method of any preceding claim further comprising the step of repeating steps b) and c) for different vernalisation growth conditions to determine a range of durations of vernalisation growth conditions. 13. A method for predicting vernalisation requirements of a plant cultivar, the method comprising the steps of: a) determining the relative contributions of each of a plurality of paralogues of FLC to the total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the proportion of said FLC paralogues which are cold- responsive; and c) determining the vernalisation growth conditions for said plant cultivar based on the relative contributions of cold-responsive FLC paralogues and non-cold- responsive paralogues to total FLC expression levels of step a). 14. A method for selecting a plant cultivar for use in a breeding programme to obtain a desired ecotype, the method comprising: predicting vernalisation requirements of the plant cultivar in accordance with any preceding claim; and selecting the plant cultivar for use in said breeding programme in the event that the predicted vernalisation requirements are consistent with the desired ecotype. 15. A method of breeding a plant cultivar having a desired ecotype, the method comprising selecting a plurality of plant cultivars for use in a breeding programme in accordance with claim 14, and crossing said plant cultivars in a breeding programme to generate progeny. 16. The method of claim 15 further comprising the step of predicting the vernalisation requirements of said generated progeny in accordance with any of claims 1 to 13, to determine whether the progeny has the desired ecotype. 17. A method for selecting a plant cultivar having a desired ecotype, the method comprising the steps of: a) determining the relative contributions of each of a plurality of paralogues of FLC to the total expression levels of FLC in said plant cultivar under normal growth conditions; b) determining the proportion of said FLC paralogues which are cold- responsive; and c) selecting a plant cultivar having a greater contribution to total FLC levels from cold-responsive FLC paralogues in the event that a spring ecotype is desired, or selecting a plant cultivar having a greater proportion to total FLC levels from non-cold- responsive FLC paralogues in the event that a winter ecotype is desired. 18. A method for generating a model for determining the decay rate of an FLC paralogue under vernalisation conditions, the method comprising recording decay rates for said FLC paralogue in a plurality of plant cultivars under vernalisation conditions, and combining said recorded decay rates to provide a modelled decay rate.