Quantification of Causal Relationships Between Heading Date and other Traits Using Structural Equation Modeling: Application to Bi-Parental and Association Mapping Populations in Rice

In crop plants, the timing of heading and flowering influences the phenotypes of other agronomic traits sometimes largely. Thus it is important to take into account their causal relationships in breeding and genetic researches. Here we quantified the influence of timing of heading (days to heading: DH) on culm length (CL) and panicle length (PL) using structural equation models (SEMs) in two rice (Oryza sativa L.) populations: a bi-parental mapping population consisting of 100 backcross inbred lines (BILs) derived from the cross between Nipponbare and Koshihikari, and an association mapping population consisting of 101 Japanese cultivars. The former population was tested at three locations (A, B and C) in Japan in 2009 and the latter was tested at one location (D) over four years (2006-2009). Parameters of SEMs were estimated separately for each location via Markov chain Monte Carlo. The posterior means of the causal influence (path coefficient) from DH to CL were 0.021, 0.105, 0.304 and 0.683 (cm/day) at locations A, B, C, and D, respectively, and those from DH to PL were -0.052, 0.048, 0.051 and 0 (cm/day). While the estimated coefficients varied among the three locations for the bi-parental population, those were similar between locations C and D. Considering that the locations C and D have similar climates especially in temperature during cultivation season, these results suggest that the influences of DH on CL and PL are determined mainly by the environmental conditions of field-testing rather than the genetic composition of experimental populations.