AUTHOR=DuVal Ashley , Gezan Salvador A. , Mustiga Guiliana , Stack Conrad , Marelli Jean-Philippe , Chaparro José , Livingstone Donald , Royaert Stefan , Motamayor Juan C. 

TITLE=Genetic Parameters and the Impact of Off-Types for Theobroma cacao L. in a Breeding Program in Brazil

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 8 - 2017

YEAR=2017

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2017.02059

DOI=10.3389/fpls.2017.02059

ISSN=1664-462X

ABSTRACT=Breeding programs of cacao (Theobroma cacao L.) trees share the many challenges of breeding long-living perennial crops, and genetic progress is further constrained by both the limited understanding of the inheritance of complex traits and the prevalence of technical issues, such as mislabeled individuals (off-types). To better understand the genetic architecture of cacao, in this study, 13 years of phenotypic data collected from four progeny trials in Bahia, Brazil were analyzed jointly in a multisite analysis. Three separate analyses (multisite, single site with and without off-types) were performed to estimate genetic parameters from statistical models fitted on 10 important agronomic traits (yield, seed index, pod index, % healthy pods, % pods infected with phytophthora, % pods infected with witches broom, % of pods other loss, vegetative brooms, diameter and tree height). Genetic parameters were estimated along with variance components and heritabilities from the multisite analysis, and a trial was fingerprinted with low-density SNP markers to determine the impact of off-types on estimations.  Heritabilities ranged from 0.37-0.64 for yield and its components and from 0.02-0.16 for disease resistance traits. A weighted index was used to make selections for clonal evaluation. The impact of off-types to breeding progress in cacao was assessed for the first time.  Even when present at less than 5% of the total population, off-types resulted in a 41% difference in estimated genetic gains for yield, altered selections by 48%, and impacted heritability estimations for nine of the 10 traits analyzed. These results show that in a mixed model analysis, even a low level of pedigree error can significantly alter estimations of genetic parameters and selections in a breeding program.