Genome-wide association mapping of growth dynamics detects time-specific and general quantitative trait loci

Abstract

Growth is a complex trait determined by the interplay between many genes, some of which play a role at a specific moment during development whereas others play a more general role. To identify the genetic basis of growth, natural variation in Arabidopsis rosette growth was followed in 324 accessions by a combination of top-view imaging, high-throughput image analysis, modelling of growth dynamics, and end-point fresh weight determination. Genome-wide association (GWA) mapping of the temporal growth data resulted in the detection of time-specific quantitative trait loci (QTLs), whereas mapping of model parameters resulted in another set of QTLs related to the whole growth curve. The positive correlation between projected leaf area (PLA) at different time points during the course of the experiment suggested the existence of general growth factors with a function in multiple developmental stages or with prolonged downstream effects. Many QTLs could not be identified when growth was evaluated only at a single time point. Eleven candidate genes were identified, which were annotated to be involved in the determination of cell number and size, seed germination, embryo development, developmental phase transition, or senescence. For eight of these, a mutant or overexpression phenotype related to growth has been reported, supporting the identification of true positives. In addition, the detection of QTLs without obvious candidate genes implies the annotation of novel functions for underlying genes.

Keywords: Arabidopsis thaliana; GWAS; PLA; genome-wide association mapping; growth dynamics; natural variation; plant phenotyping; projected leaf area; rosette growth..

Fig. 1.

(A) Images of one of the replicates of CS28014 (Amel-1), a representative accession, at all time points included in the analyses. (B) Pictures processed by ImageJ to determine the projected leaf area (PLA). Pictures were segmented based on colour, saturation, and brightness, and thereafter made binary. Particles which were too small (<120 pixels) were excluded from the analysis. In the images of days 8, 11, and 14, more than one plant is present, but only the remaining one (days 16 and onwards) is taken into account for PLA determinations.

Fig. 2.

Pearson correlations between fresh weight of rosettes (FW) at the end of the experiment (day 28), projected leaf area (PLA) over time (day 8 till 28), and parameters ‘r’ and ‘A ₀’ of the growth model Expo2. r ²-values are given in the left lower part of the figure, whereas corresponding P-values are given in the right upper part of the figure. Blue and red indicate positive and negative correlations, respectively. The stronger the intensity of the colour, the stronger the correlation.

Fig. 3.

Modelling of growth using three mathematical functions, Expo1, Expo2, and Gom (see Table 2 for details). (A) S-curve as a model for determinate growth consisting of three phases: the acceleration phase, the linear phase, and the saturation phase. The S-curve has three characteristic points: the inflection point ‘K’ where the growth changes from increasing to decreasing, the point of maximal acceleration ‘s1’, and the point of maximal deceleration ‘s2’. (B) Data and curve fits for line CS76226 (Se-0), representative for the 11% of the plants for which the growth curve reached ‘K’. The data were fitted to the three models. It is shown that even if the inflection point ‘K’ is reached, model Expo1 and Expo2, which both assume indeterminate growth, resulted in good fits. (C) Data and curve fits for line CS76308 (ZdrI2-25), representative for 56% of the plants for which the growth curve did not reach ‘s1’. The data were fitted to the three models. (D) Magnification of (C) allowing better comparison of the various models for the time window of the experiment.

Fig. 4.

Comparison of the goodness-of-fit for the three growth models used: exponential function with one (Expo1) or two (Expo2) parameters, and Gompertz function (Gom). Plot of the measured PLA on days 8, 11, 14, 16, 18, 20, 22, 24, 25, 26, 27, and 28 against the predicted PLA on the same days. The black line represents y=x (PLA measured=PLA predicted).

Fig. 5.

Genome-wide association (GWA) mapping of FW, PLA, and parameters of growth model Expo2. Univariate GWA analyses were performed for all traits; in addition, the model parameters ‘r’ and ‘A ₀’ were also analysed together in an MTMM-GWA approach. A Manhattan plot of the –log(P) marker–trait association for FW, PLA, and model parameters of Expo2 is shown. PLAs on the different days are represented by one value; for each SNP, only the –log(P) value of the day with the highest association is plotted. Univariate analyses of ‘r’ and ‘A ₀’, and the MTMM analyses of ‘A ₀’ and ‘r’ jointly are also represented by one value; for each SNP, only the –log(P) value of the analysis with the highest association is plotted. The total number of tested SNP markers was 214 000, but only the ~10 000 SNPs with –log10(P)>2 are plotted. The dotted line indicates the arbitrary threshold of –log(P)=5.

Fig. 6.

Association profiles of SNPs that were identified by GWA mapping to be highly associated with the traits FW, PLA over time, and Expo2 model parameters. Each number in the columns with heading ‘FW’ or ‘PLA (mm²)’ represents the association found by univariate GWA mapping of growth trait by time point as indicated at the top of the column (FW at day 28 or PLA on one of the indicated days) and the SNP at the position indicated in the first two columns. In the last three columns, with the heading ‘Expo2’, the numbers indicate the association found between SNPs and parameters of model Expo2. Columns with the heading ‘Expo2: A₀’ and ‘Expo2: r’ refer to univariate GWA mapping of model parameters ‘A ₀’ and ‘r’ respectively, and the column with heading ‘Expo2: full’ refers to multivariate GWA mapping of both growth model parameters. All SNPs with –log(P)>5 for at least one trait are shown. The intensity of the grey colour corresponds to the strength of the association. Curly brackets indicate that associated SNPs are located within 10kb and are considered as one QTL.