Genetic basis of plasticity of the growth- and flowering-related phenotypes to N availability
Next, using genome-wide association (GWA) analysis, we investigated the genetic architecture of plasticity of the four studied phenotypes in response to N. To this end, we performed the GWA using the CV over all three N conditions and, in addition, with the fold changes (FC) of the phenotypic means between the optimal and limiting N conditions. GWA analysis for the CVs of the four phenotypes identified only one significant biallelic association for FT at the end of chromosome 1 (p-value = 1.41x10-7; Tables 1 and S4, Methods). Interestingly, the GWA with the FC identified four SNPs on the same locus on chromosome 1 that was associated with the CV of FT (Tables 1 and S4). We found no significant association for the other traits.
The low power of our dataset in detecting significant associations could be due to the low heritability (i.e. low proportion of genotype x environment variance from the total phenotypic variance) of the scored plasticities (Sasaki et al., 2015; Tam et al., 2019). Another explanation is that the low number of accessions that we used for our study did not suffice to identify the loci with small effects. A third explanation for the missing associations could be the incomplete set of SNPs, since the initial GWA was performed with 250,000 SNPs (Horton et al., 2012). To investigate if the lack of SNPs was the reason for the few detected associations, we took advantage of those accessions with fully sequenced genome, including 58 for ERD, 49 for FRD, 54 for FT and 44 for YIE, and performed GWA for the two plasticity scores. We detected a significant association for the ERD on chromosome 1 to two biallelic SNPs (Figure 3a; Tables 1 and S4).
Finally, we asked if the plasticity (scored as CV or FC) of the phenotypes is controlled by the same genes that control the differences in the mean trait values in the accessions. To investigate this question, we performed the GWA with the mean value of the traits in the three N conditions. We found a single significant association for ERD at intermediate N on chromosome 1, but to a different SNP than that found for the plasticity of ERD (Methods, Table S1). This implies that the plasticity of the traits in response to N has a different genetic basis than the one controlling the mean trait value in a single condition.