Multi-trait multi-environment quantitative trait loci mapping for a sugarcane commercial cross provides insights on the inheritance of important traits.

G R A Margarido,A P Souza,A A F Garcia,M M Pastina

doi:10.1007/s11032-015-0366-6

Abstract

Breeding trials typically consist of phenotypic observations for various traits evaluated in multiple environments. For sugarcane in particular, repeated measures are obtained for plant crop and one or more ratoons, such that joint analysis through mixed models for modeling heterogeneous genetic (co)variances between traits, locations and harvests is appropriate. This modeling approach also enables us to include molecular marker information, aiding in understanding the genetic architecture of quantitative traits. Our work aims at detecting QTL and QTL by environment interactions by fitting mixed models with multiple QTLs, with appropriate modeling of multi-trait multi-environment data for outcrossing species. We evaluated 100 individuals from a biparental cross at two locations and three years for fiber content, sugar content (POL) and tonnes of cane per hectare (TCH). We detected 13 QTLs exhibiting QTL by location, QTL by harvest or the three-way interaction. Overall, 11 of the 13 effects presented some degree of pleiotropy, affecting at least two traits. Furthermore, these QTLs always affected fiber and TCH in the same direction, whereas POL was affected in the opposite way. There was no evidence in favor of the linked QTL over the pleiotropic QTL hypothesis for any detected genome position. These results provide valuable insights into the genetic basis of quantitative variation in sugarcane and the genetic relation between traits.Electronic supplementary materialThe online version of this article (doi:10.1007/s11032-015-0366-6) contains supplementary material, which is available to authorized users.

Highlights

Plant breeding is an essential activity to agriculture, affording short-term new elite cultivars and long-term potential for increased yields and response to adversities (Moose and Mumm 2008)
Given that our ultimate goal was to map QTLs, we decided to use the model selected by BIC because a simpler model results in smaller runtime for fitting and is more amenable to QTL mapping, which is important in the multiple interval mapping (MIM) context
Genetic and residualvariances are expected to be different across traits and environments, which in turn makes this type of data suited for mixed model analysis

Summary

Introduction

Plant breeding is an essential activity to agriculture, affording short-term new elite cultivars and long-term potential for increased yields and response to adversities (Moose and Mumm 2008) To select both broadly stable genotypes and those adapted to specific environmental conditions, it is paramount to have. Appropriate modeling of genetic and residual (co)variance matrices alleviates the need to make unrealistic assumptions about the distribution of errors and affords unbiased estimates of model effects, such as QTL effects (Balzarini 2001) In this context of modeling G 9 E interaction and correlation between traits, mixed models are a natural statistical approach to use (Smith et al 2005)

Objectives

Methods

Results

Conclusion