Identifying genes that determine SOC plasticity
To uncover the genetic basis of SOC plasticity, we conducted genome-wide
association studies (GWASs) using two reaction norm parameters as
phenotypic input, the intercept and the slope, obtained from joint
regression analysis based on the environmental mean for SOC across all
environments (Fig. S7). When using the intercept as the phenotype, we
detected genomic regions containing the PROBABLE METHYLTRANSFERASE
PMT6 (PMT6 ) and DIOXYGENASE FOR AUXIN OXIDATION1(DAO1 ) genes, with PMT6 previously reported to influence
SOC variation in B. napus (Tang et al., 2021). Using the slope as
phenotype, we identified two genomic regions containing genes homologous
to Arabidopsis (Arabidopsis thaliana ) MYB
DOMAIN PROTEIN 106 (MYB106 ) and DIGALACTOSYL
DIACYLGLYCEROL DEFICIENT1 (DGD1 ). MYB106 encodes a
MIXTA-like transcription factor involved in cuticle development and
lipid transport, while DGD1 encodes a galactosyltransferase-like
protein involved in lipid trafficking (Dormann et al., 1999, Oshima et
al., 2013).
To uncover the genomic regions that respond to changes in the identified
environmental indices (DTR183–192,
PR166–195 and UVB144–186), we
calculated the slope from a regression between SOC and each index,
representing the specific response of each plant to each index. We used
the resulting slopes as phenotypes for the GWAS (Figs. 5a, S8). We
identified DGD1 for DTR183-192 slopes andMYB106 for UVB144-186 slopes. We also mapped
three additional genomic regions associated with
PR166-195 slopes. We prioritized the candidate genes
within these regions with the Arabidopsis homologs3-HYDROXYACYL-[ACYL-CARRIER-PROTEIN] DEHYDRATASE(HAD ), MYO-INOSITOL-1-PHOSTPATE SYNTHASE3 (MIPS3 ),
and PHOSPHATIDYLSERINE DECARBOXYLASE1 (PSD1 ). HAD is a
3-hydroxyacyl-ACP dehydratase involved in fatty acid biosynthesis. MIPS3
catalyzes the rate-limiting step in myo-inositol biosynthesis. PSD1 is a
mitochondrion-localized protein involved in phosphatidylethanolamine
biosynthesis (Fig. 5a–c). The involvement of these genes in lipid
metabolism in Arabidopsis is consistent with their putative role
in SOC variation in B. napus .
To delineate potential polymorphisms in the above candidate genes, we
aligned their genomic sequences from eight publicly available
whole-genome assemblies of B. napus lines (Fig S9; Supplementary
Table S16). We discovered structural variations in PMT6 andDGD1 , in the form of transposon insertions in their promoters,
defining two haplotypes among these eight B. napus inbred lines.PMT6 in ‘Gangan’, ‘QuintaA’ and ‘Shengli’ contains an insertion
of a Mutator retrotransposon; Helitron retrotransposon was
inserted in the DGD1 promoter in ‘Shengli’. PMT6 andDGD1 may therefore affect SOC through changes in their expression
levels. We also investigated potential coding variants affecting protein
function. The most significant amino acid polymorphisms are C106G, H32N
and S108C in DAO1, MYB106 and HAD, respectively. These variants provide
relevant evidence for follow-up studies of functional polymorphisms.
To understand gene–environment interactions, we examined genetic effect
dynamics along the identified environmental indices. We calculated the
genetic effects within each environment for the most significant markers
within identified candidate genes and used regression analysis to obtain
fitted lines representing the genetic effects along the environmental
gradient (Fig. 5d–f). This helps interpret the underlying mechanism of
plant perceptron interacting with diverse environmental and development
cues. Differential sensitivity (DS; magnitude change), conditional
neutrality (CN, effect limiting to specific environment), antagonistic
pleiotropy (AP; sigh change), and no G×E (no change) (Des Marais et al.,
2013) can be viewed as emergent properties in systems. When DGD1interacts with DTR183-192, 57.14% of effect pairs show
AP, with 35.71% showing DS and 7.14% indicating no G×E (Fig. 5d). TheMYB106 and UVB144–186 interaction is
predominantly AP (42.86%), followed by 25.00% CN, 21.43% DS, and
10.71% no G×E (Fig. 5f). In contrast, the HAD and
PR166–195 interaction is primarily CN (39.29%),
followed by 32.14% AP, 21.43% DS, and 7.14% no G×E (Fig. 5e). We
asked if the two alleles at each SOC plasticity gene showed differences
in their expression levels; to this end, we used RNA sequencing data
from 289 lines of developing seeds at 20 days after flowering in WH2016
(Tang et al., 2021). Indeed, the two possible alleles differed
significantly in their expression, supporting a role in SOC plasticity
(Figs. 5g–i, S10).