Plasticity screening using Arabidopsis thalianaaccessions in response to N availability
We screened 190 Arabidopsis accessions (Table S1), grown on soil supplemented with three different concentrations of N (Methods), for plasticity of focal traits (i.e. phenotypes) including early rosette diameter (ERD), final rosette diameter (FRD), flowering time (FT) and yield (YIE) (Table S2). A pilot experiment with a subset of accessions, following an established protocol (Pandey et al., 2019), was used to determine the N conditions considered as limiting, intermediate and optimal for growth (Methods). Accessions with missing data in at least one N condition were removed from further analysis, resulting in 142 (ERD), 109 (FRD), 127 (FT) and 102 (YIE) accessions to characterize plasticity of the four phenotypes. On average, the accessions were larger, flowered earlier and produced more seeds when grown under optimal N in comparison to intermediate and limiting N (Figure 1a).
Clustering of the accessions according to their reactions norms, representing the phenotype mean as a function of different environments, partitioned the accessions into three groups for ERD and two groups for each of the other complex traits (FRD, FT, and YIE) (k-means clustering, k determined by silhouette index analysis, Figure 1b, Table S2). The differences in response of the studied phenotypes between the clusters of accessions indicated that there is a genetic variation in plasticity of these phenotypes to N availability.
To quantify plasticity of the phenotypes, we calculated the coefficient of variation (CV) for each accession using the mean values in the three N conditions (Methods, Table S3). The average CVs for all traits were smaller than 1, implying a generally low degree of plasticity (Figure 2a). ERD showed higher median plasticity (0.46) than the other phenotypes (Figure 2a). In contrast, FT, measured as the number of days from pricking of the seedlings to the first open flower, had the lowest median CV (0.11) in response to the three N conditions (Figure 2a). We also asked if the accessions differed with respect to the pattern of plasticities for the four phenotypes. To this end, we clustered the 77 accessions for which the CV of all four phenotypes were measured, and found that they can be partitioned into three groups (Figure 2b; Table S3). The accessions in cluster 1 showed lower plasticity in the ERD than those in clusters 2 and 3, while the accessions in cluster 2 showed higher plasticity in the ERD and FRD than those in clusters 1 and 3 (Figure 2b, Table S3). These findings demonstrated that accessions differed with respect to the plasticities of the four investigated phenotypes. Next, using the CVs of the 77 accessions, we tested if the plasticities of the four phenotypes correlated. We found that the CV of the ERD showed significant positive Spearman correlation with the CV of the FRD (0.42, p-value = 4.7x10-5) and FT (0.37, p-value = 5.7x10-5), indicating that plasticity of the size in the beginning of vegetative growth is moderately associated with plasticity in flowering time (Figure 2c).
To investigate whether the plasticity in the flowering time was due to change in the developmental timing or the growth pattern, we scored the number of rosette leaves until bolting for six accessions with varying plasticities of FT. We noted that under the same N condition, the total number of leaves differed between the selected accessions, but all of them produced approximately twice the number of leaves under optimal N in comparison to the limiting N (Figure S1). This suggests that the earlier flowering under optimal N was due to a faster developmental transition to the reproductive phase.