As noted, parent line A was less inbred than the other parent lines. This may explain why four out of the five BoFLC2 heterozygotes found in parent lines were parent A (one in the S2
Chapter III The Contribution of BoFLC2 to Flowering Time in Cauliflower
67
found in the other late-flowering line, parent C (S6). While finding heterozygous lines after
six generations of inbreeding is unusual, a low frequency of +/- may be expected. As noted previously, all heterozygous parent lines were excluded from statistical analysis, but the fact that a potentially heterozygous line was used to make the crosses would be expected to affect the segregation of the BoFLC2 alleles (and potentially other genes) in subsequent generations – especially in the AB F2 population, where the relative frequency of heterozygous lines was
higher in the parent lines. This may partially explain the greater phenotypic variation that was observed in the AB F2 plants when compared to CD F2 plants.
The genetic and phenotypic analysis of segregating BoFLC2 F2 populations derived from
crosses between the late-flowering (+/+) and early-flowering (-/-) cauliflowers (as opposed to a group of unrelated cauliflower lines) enabled a more „controlled‟ method of assessing
BoFLC2 influence on flowering time. Another benefit of this experiment was that it enabled the incomplete dominance and quantitative nature of the BoFLC2 gene to be studied through analysis of heterozygotes; Okazaki et al. (2007) state that the annual habit in B. oleracea is dominant over the biennial habit, but the conditions used in this study exposed dosage effects and incomplete dominance of the BoFLC2 gene. Furthermore, the use of these F2
populations enabled the collection of flowering time data at a level of detail not possible for the parent lines used in this study, which were classified into different flowering classes based on in-house trials and inaccessible data.
The correlation of the BoFLC2 genotype with curd appearance date in the AB F2 population
is plain; plants with more functional copies of BoFLC2 produced curds later (Figure 9). The fact that this correlation remains clear despite parent lines not being „thoroughly‟ inbred (which would potentially result in greater heterozygosity of other genes and more phenotypic „noise‟), emphasises the strong influence of BoFLC2 on the transition to reproductive development. Compared to the differences in curd initiation between +/+ and -/- genotypes in the F2 population, the difference in curd initiation between +/+ and -/- in parent lines is
even more extreme, primarily due to the very late appearance of A (S2) curds. The
differences between genotypes in the F2 population may be smaller because none of the
parent lines used was isogenic. This means that in addition to BoFLC2,many other genes that influence flowering time would be segregating and influencing the phenotype of the F2
population, resulting in less extreme phenotypes than in parent lines. It was also noted that the average curd appearance date for parent line A plants inbred for an additional generation
Chapter III The Contribution of BoFLC2 to Flowering Time in Cauliflower
68
(S3) was slightly earlier than those inbred for just two generations (S2). Without specific
knowledge of which characteristics formed the basis for selection in the inbreeding programme, it is difficult to account for this finding.
The CD cross produced an F2 population that was greatly preferred for analysis of BoFLC2
influence. Both C/D parents and F2 population were less affected by disease than the AB
plants, meaning that, as well as curd initiation, subsequent curd development and flowering time could also be investigated. Furthermore, the C/D parents were more highly inbred (six generations as opposed to two and four for the A and B parents, respectively). A significant departure from the expected theoretical BoFLC2 distribution ratio was observed in the CD F2
population; specifically, the +/+ fraction of the offspring was unusually large, and the -/- fraction was unusually small. This may possibly be due to preferential transmission of the functional allele by gametic selection.
Despite the somewhat distorted segregation ratio, the importance of the BoFLC2 gene to the flowering phenotype in the CD F2 population is also clear. In the segregating population, the
BoFLC2 gene appears to function in a dosage-dependent manner with curd formation delayed in an additive manner by additional copies of the functional allele (Figure 10). Similarly, as the number of functional alleles increased, the developmental stage at a fixed time point was observed to decline (Figure 11). The non-normality of the data in this method of classifying flowering time was expected due to the non-uniform development of curds, combined with the fact that the index used to classify developmental stage is not a perfect reflection of actual development; for example, the length of time required for a curd to pass through stages 2-5 may be brief relative to the amount of time spent as a curd (stage 1) or in full flower (stage 6). This means that observations of developmental stage at a fixed point in time are expected to result in positively or negatively skewed distribution. BoFLC2 also appeared to delay developmental stage in an additive manner in F1, C and D parent lines, but this was not the
case for measurements of curd appearance, with F1 curds appearing at a similar stage to curds
in parent D (Figure 10). It is possible that compared to the CD F2 population, the effect of
BoFLC2 heterozygosity on this phenotypic trait is modified or masked in the CD F1
population by its presumed near-complete background heterozygosity predicted by the higher degree of inbreeding of C and D parent lines. This could only be confirmed by performing additional crosses.
Chapter III The Contribution of BoFLC2 to Flowering Time in Cauliflower
69 Applying the formula σ2
phenotype = σ2genotype + σ2environment to the observed variance under a
model of normal distribution would roughly equate to the BoFLC2 gene explaining 40% of total phenotypic variance in flowering time, and 65% of the genetic variance in flowering time, where flowering time is based on the developmental stage of development 117 days after transplanting, and assuming environmental variance in the F2 population is the same as
the average of the variance in the homozygous parent lines (parental lines were used as environmental controls). As mentioned earlier, it appears likely that the remaining variation may be explained by other flowering time genes. For example, as this thesis was going to press, data was published suggesting that allelic variation at the BoFRIa locus may be responsible for variation in B. oleracea flowering time (Irwin et al. 2012).
A very recent QTL study by Uptmoor et al. (2011) using a B. oleracea population derived from a cross between late/early flowering types indicated that flowering time variability was primarily due to differences in vernalisation response. However, neither parent used to generate that population (the same as that used by Razi et al. (2008)) possessed functional alleles of BoFLC2 and no other BoFLC genes were found to co-segregate with flowering time. The authors raise the possibility that FLC-independent pathways are responsible for the vernalisation response seen in their population, and this may also be the case in the populations generated in this study. However, it is also possible that other members of the
BoFLC family may be contributing to the flowering response. In the previously mentioned study by Pires et al (2004), segregation of the BnFLC3 explained 29% of the phenotypic variation in flowering time, with plants containing B. oleracea FLC3 alleles flowering significantly earlier than those with B. rapa FLC3 alleles. In the (likely) absence of fully- functional FLC2 in that segregating B. napus population, it seems as though FLC3 is a significant determinant of flowering time. In the same way, it is possible that variation in genes such as BoFLC3 could contribute to the remaining variability in the AB and CD populations. However, this cannot be established due to the fact that no polymorphisms were detected in BoFLC3 using the primers employed here, meaning that it is unknown whether this gene was segregating. Further examination of the promoter and introns of BoFLC3
would enable the contribution of the BoFLC3 gene to flowering time in these populations to be evaluated.
In both F2 populations, it would have been preferable to commence observations earlier in
Chapter III The Contribution of BoFLC2 to Flowering Time in Cauliflower
70
curds by the first week of observation. This is particularly so in the AB population, where 84% of -/- plants had already formed curds by the first week of observation (followed by 58% of +/- and 48% of +/+ plants). In the CD population, the overall curd initiation was later (64% of the -/-, 29% of the +/- and 12% of the +/+); nonetheless, in both populations, the percentage of initiated plants decreased as the number of functional BoFLC2 alleles increased, meaning that it is likely that the findings reported here would be equally, if not more pronounced had observations begun earlier.
Differences in average curd size at the beginning of the observation period are likely to be directly related to the differences in initiation date. It was not unexpected to observe that in the later-initiating genotypes, the average curd sizes were smaller in the initial observation stages (Figure 12). Even though significant differences were observed in average maximum observed curd size, the differences in curd initiation date make it somewhat misleading to compare these values: a more accurate measure of maximum curd size was taken to be the predicted maximum size, based on the quadratic trend line. This predicted maximum size was similar for all three genotypes, indicating that BoFLC2 functionality does not have any bearing on the final curd size. Indeed, it was observed that within each of the three genotypes, some curds begin to bolt and flower when the curd size is very small, and others do not bolt and flower until the curd is quite large. As well as being influenced by other genes, it is expected that this characteristic is very much determined by environmental factors.
It is worth noting that there is a significant delay in reaching the maximum predicted curd size in +/- and +/+ genotypes. Again, this may be partially explained by the delayed initiation of +/- and +/+ curds compared to -/- lines. For example, Figure 12 shows that +/- lines were, on average, six days slower to reach this maximum predicted size than the -/- lines, which is likely to be because they also appeared an average of six days later than the -/- lines. However, the +/+ curds, which appeared an average of twelve days later than the -/- lines did not reach this maximum predicted size until 32 days later than the -/- lines. This is consistent with the idea that the rate of curd growth, as well as curd development, is slower in genotypes with more functional copies of BoFLC2. Although the statistical calculations of curd growth rate are based on comparisons between older curds in the -/- genotype and younger curds in the +/- and +/+ genotypes, the quadratic nature of the curves used to generate these predicted maximum averages means that even if the curds in all three
Chapter III The Contribution of BoFLC2 to Flowering Time in Cauliflower
71
genotypes were able to be compared at the same age, the differences in slope (and therefore rate of growth) would actually be even more pronounced.