Manuscrit en préparation
Station de Génétique Végétale, INRA-CNRS-INAPG-UPS, Ferme du Moulon, 91190 Gif sur Yvette, France. Tel: +0033 (0)1 69 33 23 44; Fax: +0033 (0)1 69 33 23 40
* Unité Éco-Anthropologie et Ethnobiologie, UMR CNRS 5145 - MNHN USM 104, Équipe "Génétique des Populations Humaines", Musée de l'Homme, 17 place du Trocadéro, 75116 Paris, France.
Corresponding author: B. Rhoné
Station de Génétique Végétale, INRA-CNRS-INAPG-UPS, Ferme du moulon, 91190 Gif sur Yvette, France. Tel: +0033 (0)1 69 33 23 44; Fax: +0033 (0)1 69 33 23 40
Abstract
In annual plant species, flowering time is a major adaptive trait that synchronizes the initiation of reproduction with favorable environmental conditions. In order to understand how evolved diversity and spatial structure of genes involved in the control of an adaptive trait submitted to selection, we studied the evolution of flowering time, genes associated with this trait and microsatellite markers supposed to be neutral in three experimental wheat populations grown in contrasting environments (Northern to Southern France). By comparing phenotypic to presumably neutral variation, we showed that flowering time was selected during the 12 generations of evolution. We then tested whether six candidate genes, presumably involved in the trait expression, exhibited signatures of selection that could be identified by spatial or temporal methods adapted for our experimental population design using three different methods of detection. Only genes with a major impact on flowering time were detected as responding to selection. With low-to-moderate phenotypic effect, or when covariance among genes occurs, even combining several spatial methods of outlier loci detection appeared inefficient. In such a case, it seems necessary to consider multilocus allelic combinations between genes, which could be targets of selection.
Keywords: Adaptation, Differentiation, Diversifying selection, Heading date, Triticum
Introduction
Local adaptation resulting from spatial and temporal heterogeneity in selection pressures acting on heritable traits is the main force shaping phenotypic diversity in natural populations (Schlötterer 2002). Understanding the genetic mechanisms involved is therefore one of the greatest challenges of evolutionary biology. Yet the question of how the selection acting on an adaptive trait influences the evolution of genes involved in the structure of that trait is still largely unanswered (Orr 2005, Ehrenreich & Purugganan 2006). To get insight into the genetic mechanisms of adaptation, the identification of genes that influence ecologically important traits is a prerequisite, before detecting those responding to selection during adaptation. Flowering time is a major adaptive trait in plant species because it ensures reproduction to occur in favourable conditions with respect, e.g., to climate, herbivory, pathogen pressures, or the presence of pollinators (Remington & Purugganan 2003; Roux et al. 2006). The genetic architecture of this trait has been extensively studied in model species and crops. For example, in Arabidopsis thaliana, more than 80 genes involved in the floral transition have been identified (Putterill et al. 2004) revealing that flowering time is a complex trait governed by many interacting genes.
In the present work, we studied the evolution of flowering time and genes associated with this trait in three experimental wheat populations grown in contrasting environments, from Southern to Northern France. The three experimental populations were derived from the same initial population, and evolved independently for 12 generations. Wheat (Triticum aestivum L.) is an annual autogamous crop. Under temperate climates, it is usually sown in autumn and requires winter frost (vernalization) and exposure to long photoperiod before flowering. These major environmental signals ensure that the plant flowers in spring under favorable conditions (Loskutov 2001; Goldringer et al. 2006). In a preliminary study, Goldringer et al. (2006) found a latitudinal differentiation for earliness in these experimental wheat populations, after only 10 generations. Populations from the North exhibited both a higher frequency of the winter growth habit genotypes (which need a period of exposure to low temperature to initiate flowering) and a later flowering time than populations from the South, presumably in response to local climatic conditions. These findings led to questions about the genetic changes underlying the phenotypic differentiation among populations on such a small time scale.
We examined six candidate genes, potentially associated with the variation in flowering time. In this allo-hexaploid species (genome AABBDD), each gene is usually present in three
copies across the A, B, and D genomes. Three genes out of the six have already been identified and described in hexaploid wheat: (i) VRN-1 is the major gene involved in the vernalization pathway. Mutations in the promoter of the genome A copy (VRN-1A) and in the first intron of the two genome B and D copies (VRN1-B and VRN-1D) were found to be strongly associated with the variation for flowering time without vernalization and for growth habit (spring or winter depending on the vernalization requirement ;Yan et al. 2004; Sherman et al. 2004; Fu et al. 2005; Rhoné et al. 2008). (ii) Ppd-1 is a major gene involved in the photoperiod pathway. The three copies have recently been isolated and two mutations (a large deletion upstream of the coding region and a deletion in the last exon) have been identified in the genome D copy of the gene as involved in the variation of photoperiod sensitivity (Beales et al. 2007). (iii) An orthologous gene of FLOWERING LOCUS T (FT) gene of Arabidopsis
thaliana was isolated recently in hexaploid wheat. FT polymorphisms (also referred to as VRN-3 in Yan et al. 2006) are expected to affect both vernalization and photoperiod pathways
as it is considered a key integrator gene (ultimately regulated by internal and external signals). In a global association study, FT mutations in the A and D copies were found to be mostly associated with flowering time without vernalizing treatment (Bonnin et al. 2008). The three other gene we studied were orthologous sequences of A. thaliana genes, known in this species to be involved either in the autonomous pathway as LUMINI-DEPENDENS (LD) (Lee et al. 1994) or in the photoperiod pathway as CONSTANS (CO) and GIGANTEA (GI) (Griffiths et al. 2003; Dunford et al. 2005).
Most of the approaches that have been proposed to detect marker loci potentially targeted by selection aim at detecting some loci exhibiting unusual patterns of variation compared to expectations under the neutral model (see the reviews by Beaumont 2005; Storz 2005; Vasemägi & Primmer 2005). Such approaches rely on the fact that demography affects the genome evenly, while selection for a given trait presumably affects a few genes at a time, in a locus-specific manner. Hence, regions of the genome that are driven by directional selection are expected to show a significant decline in genetic diversity, increased differentiation among populations and increased linkage disequilibrium compared to expectations under the neutral model (Vasemägi & Primmer 2005). Statistical methods based on estimates of population differentiation (as measured, e.g., by the parameter FST) have been developed to identify singular loci that exhibit higher (or lower) population differentiation than expected under neutrality (see Beaumont & Nichols 1996; Vitalis et al. 2001; Beaumont and Balding 2004, reviewed in Beaumont 2005). A similar approach based on the temporal changes in allele frequencies (estimated as the standardized variance in allele frequencies Fc), was
developed in order to identify loci that exhibit unusual temporal variation of allele frequencies (extreme Fc) as compared to the rest of the genome (Goldringer & Bataillon 2004). There are many recent studies in non-model organisms that use one or several of these approaches in order to identify genomic regions targeted by selection (see, e.g., Wilding et al. 2001; Kayser et al. 2003; Rogers & Bernatchez 2005; Bonin et al. 2006; Oetjen & Reusch 2007; Bonin et al. 2007; Rhoné et al. 2007; Stinchcombe & Hoekstra 2008). In the present study, we used some of the approaches cited above to specifically test whether the candidate-gene polymorphisms, known to be associated with the variation of flowering time in the different studied populations, were targeted by selection. Simulations of the distribution of different parameters in the neutral case were obtained, based on the variation of 21 microsatellite loci genotyped in the studied populations. The use of experimental populations may increase the power of such methods, because many of the demographic parameters that are usually considered as nuisance parameters (for example the initial genetic composition, the time since divergence from an ancestral population, the rate of dispersal, and the census size of the populations) are controlled or known.
Our main objective in the present study was to investigate whether selection can be detected at candidate genes, which are potentially involved in the variation of a quantitative trait under diversifying selection. We first investigated whether flowering time was responding to natural selection in the wheat populations studied during the course of the experiment. In that aim, we compared the quantitative genetic variation for this trait (QST) with the genetic variation measured at 21 microsatellite loci presumed to be neutral. Then we evaluated whether the polymorphisms found at six candidate genes were associated with variation in flowering time, in order to better characterize the genetic architecture of that trait. Finally, we tested whether the candidate genes exhibited signatures of selection that could be identified by the spatial or temporal methods discussed above adapted for our experimental population design.