SEARCH

SEARCH BY CITATION

Keywords:

  • field experiment;
  • flowering;
  • FRIGIDA (FRI);
  • genotype × environment interaction;
  • life-history trait;
  • natural variation;
  • QSTFST comparison;
  • vernalization

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • The study of the evolutionary and population genetics of quantitative traits requires the assessment of within- and among-population patterns of variation.
  • We carried out experiments including eight Iberian Arabidopsis thaliana populations (10 individuals per population) in glasshouse and field conditions. We quantified among- and within-population variation for flowering time and for several field life-history traits. Individuals were genotyped with microsatellites, single nucleotide polymorphisms and four well-known flowering genes (FRI, FLC, CRY2 and PHYC). Phenotypic and genotypic data were used to conduct QSTFST comparisons.
  • Life-history traits varied significantly among- and within-populations. Flowering time also showed substantial within- and among-population variation as well as significant genotype × environment interactions among the various conditions. Individuals bearing FRI truncations exhibited reduced recruitment in field conditions and differential flowering time behavior across experimental conditions, suggesting that FRI contributes to the observed significant genotype × environment interactions. Flowering time estimated in field conditions was the only trait showing significantly higher quantitative genetic differentiation than neutral genetic differentiation values.
  • Overall, our results show that these A. thaliana populations are genetically more differentiated for flowering time than for neutral markers, suggesting that flowering time is likely to be under divergent selection.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The regulation of flowering time has substantial biological significance because it defines the vegetative-to-reproductive transition and determines the length of the post-embryonic life history of annual plants. Variation in flowering time may affect different plant fitness components, so the switch from vegetative to reproductive development is expected to be under strong selective pressure (Coupland, 1995; Ausín et al., 2005; Roux et al., 2006; Anderson et al., 2011). For this reason, unraveling the genetic basis and environmental control of flowering time has represented a major study issue for the Arabidopsis thaliana research community for many years (Coupland, 1995; Koornneef et al., 1998, 2004; Simpson & Dean, 2002; He et al., 2003; Ausín et al., 2005; Flowers et al., 2009; Brachi et al., 2010).

Given the evolutionary relevance of flowering time, research efforts have focused on the detection and assessment of natural selection on flowering time as well as its underlying genetic mechanisms in controlled conditions (Le Corre, 2005; Stenøien et al., 2005; Li et al., 2006; Scarcelli et al., 2007; Kover et al., 2009) and/or natural field settings (Weinig et al., 2002; Caicedo et al., 2004; Stinchcombe et al., 2004; Korves et al., 2007; Brock et al., 2009; Wilczek et al., 2009; Ågren & Schemske, 2012). These studies have shown that the genetic basis of flowering time variation depends strongly on the environment in which flowering time is estimated (Weinig et al., 2002; Olsen et al., 2004; Malmberg et al., 2005; Li et al., 2006; Korves et al., 2007; Brachi et al., 2010). Therefore, significant genotype × environment (G × E) interactions may reflect the wide array of mechanisms accounting for flowering time plasticity across different environments (Pigliucci & Schlichting, 1998; Stratton, 1998; Pigliucci, 2003). As the genetic architecture of a trait depends on the population and environment where it has evolved (Remington & Purugganan, 2003; Atwell et al., 2010), systematic comparisons of experiments performed in different environmental settings are essential if we aim to investigate properly the evolutionary genetics of flowering time in A. thaliana.

In the Iberian Peninsula, adaptive variation in flowering time among natural accessions of A. thaliana seems to be largely determined by variation in winter temperature: accessions from cold environments exhibit late-flowering behaviors and stronger vernalization requirements (Méndez-Vigo et al., 2011). Furthermore, different polymorphisms in flowering genes of the vernalization pathway, such as FRI and FLC, show significant associations with flowering traits and winter temperatures, revealing some of the genetic mechanisms underlying flowering time variation in Iberian A. thaliana accessions (Méndez-Vigo et al., 2011; Sánchez-Bermejo et al., 2012). As these findings were obtained from experiments conducted in glasshouse conditions, we ought to conduct experiments in realistic field settings also and compare results with those obtained in controlled conditions. These comparisons may enable us to evaluate the extent of shared genetic bases for flowering time variation among controlled and natural conditions (Weinig et al., 2002; Brachi et al., 2010).

Here, we report on the comparison between the genetic variation for flowering time in A. thaliana in controlled and natural conditions. Natural conditions were represented by a field experiment in southwest Spain where the species occurs naturally (Picó et al., 2008; Méndez-Vigo et al., 2011). Under controlled conditions, we tested the flowering inductive effect of low temperature – so-called vernalization – by comparing experiments with and without vernalization treatment. In both conditions, we conducted population-based experiments using a well-known collection of A. thaliana populations from the Iberian Peninsula (Picó et al., 2008; Méndez-Vigo et al., 2011; Kronholm et al., 2012). The evolutionary relevance of flowering time was studied by comparing quantitative genetic differentiation (QST) in the various conditions with neutral genetic differentiation (FST) estimated with two sets of neutral molecular markers. Although comparisons of QST vs FST can be constrained and biased for multiple reasons (O'Hara & Merilä, 2005; Leinonen et al., 2008; Miller et al., 2008; Holsinger & Weir, 2009; Edelaar et al., 2011), they are still valid to identify promising traits that may be under selection (Whitlock, 2008). We also analysed nucleotide variation in four well-known flowering genes (FRI, FLC, CRY2 and PHYC; Méndez-Vigo et al., 2011) from two different developmental pathways (vernalization and photoperiod pathways) to assess the association of these genes with among- and within-population patterns of variation in flowering time in different conditions.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study species and source populations

The annual plant Arabidopsis thaliana (L.) Heyhn. (Brassicaceae) is widely distributed across the Iberian Peninsula (Picó et al., 2008; Méndez-Vigo et al., 2011). In Iberian populations, seed germination generally peaks in autumn, and to a lesser extent in spring, and flowering occurs in late winter and spring (Montesinos et al., 2009; Picó, 2012).

In this study, we included eight natural populations that cover the genetic and environmental diversity of A. thaliana in the Iberian Peninsula. Seven populations (codes: AGU, CDC, LEO, MAR, PRA, QUI and SAN; Fig. 1) were previously genotyped to assess the extent of local genetic differentiation in the Iberian Peninsula (Picó et al., 2008). The eighth population of study was found in spring 2006 (code: GRA; 36.77°N, 5.39°W, 1284 m asl; Fig. 1). In this population, individuals clumped into openings of a Mediterranean scrubland. GRA is a relatively small population of hundreds of individuals in comparison with thousands of individuals in the other populations (F. X. Picó & C. Alonso-Blanco, pers. obs.). GRA is located 9.2 km far from the botanical garden where field experiments were conducted so it can be considered as the local population in this study. It must be emphasized that despite the size differences of study populations, sampling individuals were collected from a similar area within each population, which was estimated to be c. 200 m2. All sampling individuals were separated by a few meters from each other.

image

Figure 1. Geographical location of Arabidopsis thaliana populations (red dots) and El Castillejo Botanical Garden (blue dot), Spain, where the field experiment was conducted.

Download figure to PowerPoint

In 2008, field-collected seeds from 10 individuals per population were multiplied by the single-seed descent method in a glasshouse from the Centro Nacional de Biotecnología (CNB-CSIC) of Madrid to be used in all experiments reported in this study. All individuals included in this study are available through the Nottingham Arabidopsis Stock Centre (NASC; http://arabidopsis.info).

Field experiment

In summer 2008, for each of the 80 A. thaliana individuals, 16 batches of 60 filled seeds each were prepared and stored in 1.5 ml plastic tubes at room temperature in darkness until the sowing day. On October 9, 2008, we sowed the 60 seeds batches in square plastic pots (12 × 12 × 12 cm3) filled with standard soil mixture (Abonos Naturales Cejudo Baena S.L., Utrera, Spain) at El Castillejo Botanical Garden of Sierra de Grazalema Natural Park in SW Spain (El Bosque, Cádiz, 36.46°N, 5.30°W, 329 m above sea level (asl); Fig. 1). The experiment included four complete blocks to account for environmental variation within the experimental garden, and each individual was randomly replicated four times within each block (see the Supporting Information, Fig. S1). In total, this experiment included 1280 pots (8 populations × 10 individuals × 4 blocks × 4 replicates) and 76 800 seeds (1280 pots × 60 seeds). Each block was covered by 2-cm wire mesh to protect seedlings from bird and rodent depredation. We observed several invertebrates in the pots (e.g. grasshoppers, beetles, spiders, ants) but never detected any significant plant damage.

We counted the number of plants per pot every 2 wk after seed sowing. The experiment was surveyed between 23 October 2008 and 23 February 2009 to track changes in rosette density over time. We stopped counting rosettes when some individuals started flowering. A flowering date was given at the pot level when the majority of the plants in the pot switched from vegetative to reproductive stage and had their first flower open. Flowering time in field conditions was estimated as the number of days between the first census, 15 d after seed sowing when all individuals had already germinated, and flowering date. In general, all plants within the same pot exhibited a rather homogeneous behavior so assigning flowering time at the pot level in field conditions was relatively straightforward. After flowering, we let all surviving plants complete the life cycle and set fruits. We counted the number of fruiting plants per pot and the total number of fruits per plant when all plants in the pot finished flowering (i.e. no more flowering buds) and fruiting (i.e. all fruits developed). Flower abortion rate was low in this experiment (< 1%) and was not taken into account. We surveyed the experiment every few days during flowering and fruiting between late February and late April 2009 to obtain accurate data. Overall, we conducted 21 surveys to complete this experiment. After counting the number of fruits per plant, plants were taken away and incinerated to minimize uncontrolled seed dispersal at the experimental facility.

Weather conditions during the field experiment were recorded daily at the meteorological station of the El Castillejo Botanical Garden (Fig. S2). On average, mean monthly minimum temperatures ranged between 5.2°C in January and 13.8°C in October, and mean monthly maximum temperatures ranged between 12.2°C in January and 22.2°C in October. There were only 3 d below 0°C in January varying between −1°C and −2°C. Maximum daily temperatures of 28°C were recorded in late April. Total monthly precipitation was regular throughout the experiment, varying from a low of 76 mm in April to a high of 166 mm in January. During the whole study period, precipitation totaled 856 mm.

Glasshouse experiments

In 2010, the 80 A. thaliana individuals were grown in the glasshouse facilities at CNB-CSIC in two environmental conditions, with and without 6 wk of vernalization, as previously described (i.e. 4°C with a short-day photoperiod during 6 wk; Méndez-Vigo et al., 2011). Several A. thaliana Iberian genotypes require obligate vernalization to flower (Méndez-Vigo et al., 2011) so any attempt to study flowering time variation in this set of populations must include a vernalization treatment. For each individual and vernalization treatment, we estimated leaf number and flowering time. Leaf number was estimated at flowering as the total number of rosette and caulinar leaves along the main inflorescence developed by a plant. Flowering time was estimated as the number of days from the planting date until the opening of the first flower for the samples without vernalization, or from the end of the vernalization treatment until the opening of the first flower for the vernalized samples (Méndez-Vigo et al., 2011).

Marker genotyping and flowering gene sequencing

We isolated DNA from all A. thaliana individuals using the Bernartzky & Tanksley (1986) protocol without mercaptoethanol. Individuals were genotyped for 14 nuclear microsatellites and 335 genome-wide nuclear single-nucleotide polymorphism (SNP) loci. These markers were genotyped as previously described (Picó et al., 2008; Gomaa et al., 2011; Méndez-Vigo et al., 2011). The SNPs were selected from three sources: frequent polymorphisms in Central Europe (Schmid et al., 2006; Picó et al., 2008), in the Iberian Peninsula (Picó et al., 2008) and in a worldwide collection of accessions (Warthmann et al., 2007). Preliminary analyses indicated low ascertainment bias between sets of polymorphic markers (results not shown), so all SNPs were analysed simultaneously. We used a total of 198 polymorphic SNPs (35, 57 and 106 SNPs that were polymorphic in Central Europe, the Iberian Peninsula and the worldwide collection, respectively) after excluding markers with missing values higher than 20%, non-polymorphic markers and singletons. On average, the 14 microsatellites and the 198 SNPs included in this study exhibited 2.3% and 8.1% of missing data, respectively. The genotyping error of these sets of neutral markers has previously been estimated and varies between 0.0004% and 0.74% (Picó et al., 2008; Gomaa et al., 2011).

Four flowering-time genes previously analysed in a larger collection of Iberian A. thaliana populations (Méndez-Vigo et al., 2011) were also sequenced for the 80 A. thaliana individuals of study: FRI and FLC, which are involved in the vernalization pathway, and CRY2 and PHYC, which are involved in the photoperiod pathway. Based on the nucleotide diversity patterns described elsewhere (Méndez-Vigo et al., 2011), we sequenced the complete FRI gene (3.5 kb), a fragment corresponding to 0.7 kb segment of FLC intron 1, a 0.9 kb segment from the promoter region of PHYC and 1.5 kb from the coding region of CRY2. For sequencing, one to seven overlapping fragments of 0.5–0.7 kb were PCR amplified using the primers described in Méndez-Vigo et al. (2011). The PCR products were sequenced using an ABI PRISM 3700 DNA analyser (Applied Biosystems, Foster City, CA, USA). DNA sequences were aligned using dnastar v.8.0 (Lasergene, Madison, WI, USA). Alignments were inspected and edited by hand with genedoc v.2.7.0 (Nicholas et al., 1997). Nucleotide diversity was estimated with dnasp v.5 (Librado & Rozas, 2009). The GenBank accession numbers of DNA sequences generated in this study are JX291242JX291525.

Experimental data analyses

The effects of block, population and individual nested within population on phenotypic traits estimated in field and glasshouse conditions were analysed with general linear models (GLM). When significant, differences among levels within main factors were analysed with the Student–Newman–Keuls post-hoc test. For glasshouse experiments, we took into account four traits including leaf number and flowering time both with and without vernalization. For the field experiment, we estimated several life-history traits: maximum number of vegetative rosettes, flowering time, total number of fruiting plants and mean number of fruits of the six largest plants within each pot (mean number of fruits per plant hereafter). We used the maximum number of rosettes observed during the experiment as a surrogate of the recruitment potential of each individual. Individual analyses on the number of rosettes from each census gave consistent patterns (results not shown). Certain asymmetric competition operated within some pots as a few plants performed significantly better than the rest of plants in the pot. After different trials with different variables (e.g. mean number of fruits among all plants, the number of fruits of the largest plant), we estimated that the six largest plants captured the winners of the within-pot asymmetric competition process in this experiment. Preliminary analyses on all traits, including the maximum number of plants per pot as a covariate, did not indicate a significant effect of density on any plant trait (results not shown).

We performed an additional GLM to test the effect of environment (i.e. the field experiment and the two glasshouse experiments), population and individual nested within population on flowering time, which was the only trait fully comparable among experiments. Block was excluded from this design because a preliminary analysis indicated that block and the interactions between block and the other factors did not have a significant effect on flowering time (> 0.26 in all cases). Results were totally consistent when including or excluding block as a factor in the design. Population- and individual-level correlations for flowering time estimated in field and glasshouse conditions were analysed with Pearson's correlation tests.

We estimated the amount of quantitative genetic differentiation, QST (Spitze, 1993), for all study traits. Assuming complete selfing, QST was estimated as VB/(VB + VW) (Bonnin et al., 1996; Le Corre, 2005), where VB is the among-population variance and VW is the within-population variance. We also estimated broad sense heritability for each trait as h2 = VG/(VG + VE), where VG is the estimated among-family variance component and VE is the residual variance (Le Corre, 2005). The 95% confidence intervals (95% CI) for all QST and h2 values were computed with the (co)variances method using restricted maximum likelihood (REML) variance components (Lynch & Walsh, 1998).

General linear models, Pearson's correlations and REML variance components were performed with SPSS v.17 statistical software (IBM, Chicago, IL, USA). We did not transform any trait because inspection of residuals showed that the assumptions of all analyses were met.

Genetic data analyses

For each population, we estimated the mean number of observed alleles per locus (na) and mean gene diversity (HS) using fstat v.2.9.3 (Goudet, 1995). We also calculated the percentage of polymorphic loci (PL), the total number of nonredundant multilocus genotypes (NG), and the percentage number of markers that differed among all pairs of nonredundant multilocus genotypes for the set of 14 nuclear microsatellites, the set of 198 polymorphic SNPs and the set of 57 polymorphisms found in the four flowering genes (see the 'Results' section). We also computed the extent of genetic differentiation, given by the FST index (Weir & Cockerham, 1984) and its 95% confidence intervals with fstat for the 14 microsatellites and the 198 polymorphic SNPs.

The population genetic structure based on nonredundant multilocus genotypes (NG = 54) was assessed using the model-based clustering algorithm implemented in structure v. 2.2 (Pritchard et al., 2000; Falush et al., 2003) following the protocols described elsewhere (Méndez-Vigo et al., 2011). The relationship between genetic distance, based on FST/[1 – FST] (Rousset, 1997), and Euclidean geographical distance among all population pairs was determined by Mantel correlation test using ibdws v. 3.22 (Jensen et al., 2005). The relationship between genetic distance, based on the proportion of allelic differences over the total number of alleles using the set of 198 SNPs and Euclidean phenotypic distance for each phenotypic trait among all nonredundant multilocus genotype pairs was also computed by Mantel correlation tests using passage v. 2 (Rosenberg & Anderson, 2011). We also conducted Mantel tests to analyse the relationship between Euclidean geographical distance and Euclidean QST distance for each phenotypic trait, again using passage. Genetic, phenotypic, QST and geographical distances were log-transformed before analysis and the significance of correlations was calculated with 1000 randomizations.

Phenotype–genotype associations were tested using the mixed-model approach for structured populations as implemented in tassel v.2.1 (Bradbury et al., 2007) and as described in Méndez-Vigo et al. (2011). Two levels of genetic relatedness, the population structure (Q matrix), and the relative kinship (K matrix) were included in the model (Yu et al., 2006; Zhao et al., 2007; Kang et al., 2008). Population structure was estimated with structure as the Q matrix containing the membership proportions of all genotypes to the K ancestral populations. To ensure independence, only K − 1 clusters of the Q matrix were used as covariates in the model. The kinship matrix was estimated as twice the proportion of shared alleles from the 198 polymorphic SNPs genotyped (Bradbury et al., 2007; Zhao et al., 2007). Associations were tested for nonsingleton gene polymorphisms that did not show within-gene linkage disequilibrium (LD) using the mixed linear model (MLM) and the EM algorithm.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Demographic surveys in field conditions

We recorded the maximum number of rosettes 1 month and a half after seed sowing for all populations (Fig. 2). In late January, the number of rosettes peaked again for all populations (Fig. 2). Although we could not monitor the fate of each plant individually, most of these January-germinated plants died within the next 15 d (Fig. 2). Visual inspection of plant sizes before flowering also indicated that most of the plants belonged to autumn-germinated cohorts that survived the winter. After the first density peak, the number of plants decreased gradually until reproduction (Fig. 2). The final number of flowering and fruiting plants remained fairly similar (Fig. 2). Overall, we recorded a maximum number of 24 375 vegetative plants growing together in a single survey (20 November 2008), 9302 flowering plants, and 8252 fruiting plants at the end of the experiment, which produced a total of 24 806 fruits (grand mean ± SE = 2.53 ± 0.10 fruits per plant).

image

Figure 2. Total number of vegetative rosettes and reproductive plants observed for each Arabidopsis thaliana population and survey during the field experiment. For code locations, see Fig. 1.

Download figure to PowerPoint

Quantitative trait variation measured in field and glasshouse conditions

All field traits, except the mean number of fruits per plant, differed significantly among A. thaliana populations (Table 1). AGU and GRA were the populations with the lowest and highest number of rosettes observed in field conditions, respectively (Table 2). GRA was also the population with the highest number of fruiting plants, whereas AGU and CDC had the lowest number (Table 2). SAN exhibited the earliest flowering time whereas LEO bore the latest flowering individuals (Table 2). The range of mean fecundity among populations was 2.4–4.2 fruits per plant (Table 2). All traits estimated in field conditions significantly differed among individuals within populations (Table 1). Block significantly affected some traits but the interaction between block and population or block and individual were not significant in almost all cases (Table 1), indicating the robustness of phenotypic variation at population and individual levels in field conditions.

Table 1. General linear model (GLM) testing of the effect of block, population, and individual nested within population on phenotypic traits of Arabidopsis thaliana recorded in field and glasshouse conditions
FactordfNumber of RosettesFlowering timeFruiting plantsFruits per plantdfLeaf number (NV)Flowering time (NV)Leaf number (V)Flowering time (V)
F-valueF-valueF-valueF-valueF-valueF-valueF-valueF-value
  1. Field traits: maximum number of rosettes, flowering time, total number of fruiting plants and number of fruits per plant. Glasshouse traits: leaf number and flowering time with (V) and without (NV) vernalization treatment. Degrees of freedom (df), F-values and their significance are given. Degrees of freedom of the error term: 1071, number of rosettes; 1028, flowering time; 1023, total number of fruiting plants; 1023, mean number of fruits; 924, leaf number (NV); 1071, flowering time (NV); 1240, leaf number (V); 1240, flowering time (V). Significance: ***, < 0.0001; **, < 0.01; *, < 0.05; ns, nonsignificant.

Block (B)31.85 ns2.62 ns6.53 **3.77 *28.37 ***6.61 *0.69 ns4.29 *
Population (P)77.41 ***30.37 ***7.15 ***0.20 ns722.39 ***21.27 ***16.30 ***11.47 ***
Individual (I)6317.23 ***12.37 ***4.82 ***1.41 *6363.98 ***58.07 ***33.72 ***23.91 ***
B × P211.52 ns1.39 ns1.63 *1.18 ns141.01 ns3.82 ***1.78 *1.70 *
B × I270.96 ns0.82 ns0.71 ns0.74 ns181.01 ns1.01 ns3.03 ***3.94 ***
Table 2. Mean (± SE) values of phenotypic traits quantified in eight Iberian Arabidopsis thaliana populations
Population codeNumber of RosettesFlowering timeFruiting plantsFruits per plantLeaf number (NV)Flowering time (NV)Leaf number (V)Flowering time (V)
  1. Field traits: maximum number of rosettes, flowering time (number of days), total number of fruiting plants and number of fruits per plant. Glasshouse traits: leaf number and flowering time (number of days) with (V) and without (NV) vernalization treatment. Means with different letters differ significantly from one another (< 0.05; SNK test). For code locations, see Fig. 1.

AGU13.7 (0.6) f142.6 (0.4) cd4.9 (0.3) e2.6 (0.3)74.4 (3.0) d106.4 (4.4) e23.2 (0.4) c42.3 (0.9) c
CDC16.7 (0.6) e146.2 (0.4) b4.9 (0.3) e2.4 (0.3)114.5 (1.7) b166.0 (2.7) b27.5 (0.5) b48.1 (0.6) b
GRA32.3 (0.6) a140.3 (0.4) e10.2 (0.3) a3.7 (0.4)56.5 (0.8) e108.9 (1.2) e22.4 (0.2) c42.7 (0.3) c
LEO25.0 (0.9) b154.1 (0.7) a8.3 (0.4) b3.6 (0.5)131.0 (1.0) a187.8 (1.9) a33.7 (0.8) a50.5 (1.0) a
MAR23.9 (0.9) bc131.4 (0.8) f7.8 (0.4) bc4.1 (0.8)69.6 (3.1) d117.7 (3.9) d17.9 (0.3) d39.1 (0.6) d
PRA22.2 (0.8) c143.1 (0.4) c7.4 (0.3) bcd3.1 (0.4)100.1 (2.8) c151.9 (3.1) c23.5 (0.3) c42.9 (0.5) c
QUI18.3 (0.6) de141.1 (0.6) de6.4 (0.3) d4.2 (0.8)55.2 (1.3) e78.8 (1.1) f16.3 (0.2) e30.1 (0.3) f
SAN20.0 (0.6) d127.2 (0.6) g6.8 (0.3) cd2.9 (0.2)50.0 (1.4) e72.7 (1.8) f16.8 (0.3) de33.7 (0.4) e

In glasshouse conditions, population and individual nested within population significantly affected leaf number and flowering time with and without vernalization treatment (Table 1). For both vernalization treatments, LEO was the population with the highest leaf number and the latest flowering times (Table 2). GRA, QUI and SAN exhibited the lowest leaf number without vernalization, whereas QUI had the lowest leaf number with vernalization (Table 2). QUI and SAN showed the earliest flowering times without vernalization, whereas QUI was the population with the earliest flowering times with vernalization (Table 2). Block and the interaction between block and main factors generally had a significant effect on several traits estimated in glasshouse conditions (Table 1), indicating a greater sensitivity of phenotypic traits to microenvironmental variation in glasshouse conditions.

Flowering time varied significantly among field and glasshouse conditions (Table 3, Fig. 3). On average, population means did not significantly differ over the three environments (Table 3), but significant interactions were found between populations and environments, and between individuals (genotypes) nested within populations and environments (Table 3). At the population level, flowering time estimated from field and glasshouse conditions with vernalization was rather similar (Fig. S3). By contrast, the among-population pattern of variation in flowering time estimated in glasshouse conditions without vernalization showed pronounced fluctuations (Fig. S3), accounting for the significant environment × population interaction. However, population mean flowering times estimated in glasshouse conditions with and without vernalization were significantly positively correlated (= 0.93, < 0.0001), despite the reduction of among-population mean flowering time from 115 to 20 d owing to vernalization (Table 2). A similar pattern was found between flowering time estimated in field conditions, which displayed a range of among-population variation of 27 d (Table 2), and flowering time estimated in glasshouse conditions with (= 0.73, = 0.038) and without (= 0.76, = 0.030) vernalization.

Table 3. General linear model (GLM) testing of the effect of environment (one field and two glasshouse conditions), population, and individual (genotype) nested within population on flowering time of Arabidopsis thaliana
FactordfF-value
  1. Degrees of freedom of the error term: 685. Significance: ***, < 0.0001; ns, nonsignificant.

Environment (E)251.16 ***
Population (P)72.29 ns
Genotype (G)636.43 ***
P × E1476.73 ***
G × E186.48 ***
image

Figure 3. Frequency distributions of flowering times in eight Arabidopsis thaliana populations estimated in the field experiment and the two glasshouse conditions with (V) and without vernalization (NV) treatment. Data correspond to the mean values of the 10 individuals per population. Populations are ranked from early to late according to the mean flowering time observed in field conditions. Flowering time is given in number of days. For code locations, see Fig. 1.

Download figure to PowerPoint

In general, frequency distributions of within-population flowering times estimated in the field were narrower than those estimated in the glasshouse without vernalization treatment, which showed a wider range of variation in most of the populations (Fig. 3). We detected a significant individual × environment interaction, which can be interpreted as the G × E interaction. Individual flowering times from CDC, GRA, and PRA populations were not significantly correlated between any pair of environmental conditions (> 0.08 in all cases). Flowering times from glasshouse conditions with and without vernalization were only significantly positively correlated for AGU (= 0.91, < 0.0001). Flowering times from field and glasshouse conditions with vernalization were only significantly positively correlated for LEO (= 0.76, = 0.011) and MAR (= 0.85, = 0.002) and significantly negatively correlated for QUI (= −0.79, = 0.006). Finally, flowering times from field and glasshouse conditions with either vernalization treatment were significantly positively correlated for SAN (> 0.64, < 0.048).

Genetic diversity, genetic structure and genotype-phenotype correlations

Both sets of 14 microsatellites and 198 SNPs showed consistently that PRA, CDC and SAN bore the highest number of multilocus genotypes, whereas GRA had the lowest number of multilocus genotypes (Table 4). A very similar hierarchy was found with regard to the mean number of alleles per locus and gene diversity (Table 4). The 57 polymorphic sites segregating across flowering genes indicated that GRA was also the less diverse population, whereas AGU, CDC, MAR and SAN were highly variable for these genes (Table 4).

Table 4. Genetic diversity of Arabidopsis thaliana populations
PopulationMarker type N G PL (%) n a H S
  1. SNP, single-nucleotide polymorphism; SSR, simple sequence repeat. Markers: 14 nuclear microsatellites, 198 nuclear SNPs and 57 polymorphic sites across four flowering genes (FRI, FLC, CRY2, and PHYC). Genetic parameters: NG, number of unique multilocus genotypes; PL, percentage of polymorphic loci; na, mean (± SE) number of alleles per locus; HS, mean gene diversity (± SE). = 10 individuals genotyped per population. For code locations, see Fig. 1.

AGU14 SSRs978.63.00 ± 1.470.443 ± 0.288
198 SNPs540.91.41 ± 0.490.163 ± 0.218
Flowering genes931.61.32 ± 0.470.128 ± 0.201
CDC14 SSRs1092.94.43 ± 2.170.655 ± 0.310
198 SNPs1064.61.65 ± 0.480.237 ± 0.208
Flowering genes742.11.42 ± 0.500.121 ± 0.165
GRA14 SSRs550.01.71 ± 0.830.233 ± 0.255
198 SNPs48.61.09 ± 0.280.030 ± 0.103
Flowering genes10.01.00 ± 0.000.000 ± 0.000
LEO14 SSRs778.63.07 ± 1.490.532 ± 0.308
198 SNPs561.61.62 ± 0.490.257 ± 0.228
Flowering genes443.91.44 ± 0.500.194 ± 0.230
MAR14 SSRs101004.50 ± 1.650.639 ± 0.250
198 SNPs732.81.33 ± 0.470.102 ± 0.163
Flowering genes833.31.33 ± 0.480.179 ± 0.256
PRA14 SSRs91004.36 ± 1.600.707 ± 0.186
198 SNPs968.21.68 ± 0.470.289 ± 0.230
Flowering genes629.81.30 ± 0.460.104 ± 0.181
QUI14 SSRs892.93.07 ± 1.210.531 ± 0.234
198 SNPs535.91.36 ± 0.480.157 ± 0.225
Flowering genes21.81.02 ± 0.130.004 ± 0.026
SAN14 SSRs91004.14 ± 1.560.690 ± 0.179
198 SNPs967.21.67 ± 0.470.274 ± 0.217
Flowering genes728.11.28 ± 0.450.119 ± 0.198

Genetic structure analyses based on a total of 54 multilocus genotypes identified with the 198 SNPs detected five genetic clusters (Fig. S4). Overall, LEO, PRA and AGU genotypes grouped together, those from MAR and GRA formed a homogeneous group and QUI, CDC and SAN genotypes represented the other three genetic clusters, respectively.

Significant correlations between genotypic and phenotypic distances were found only for flowering time in field conditions (= 0.16, = 0.003), leaf number with (= 0.10, = 0.025) and without (= 0.17, = 0.003) vernalization and flowering time in glasshouse conditions with vernalization (= 0.14, = 0.012).

Heritability estimates and QSTFST comparisons

The traits that showed the lowest mean heritability values were mean number of fruits per plant (mean ± SE = 0.03 ± 0.01) and total number of fruiting plants (0.22 ± 0.05; Table 5). By contrast, the traits with the highest mean heritability were leaf number (0.71 ± 0.10) and flowering time (0.67 ± 0.11), both without vernalization (Table 5). The populations with the lowest and highest mean heritability values were GRA (0.17 ± 0.07) and MAR (0.64 ± 0.10), respectively (Table 5).

Table 5. Broad sense heritability (h2) and quantitative genetic differentiation (QST) values for study traits of eight Iberian Arabidopsis thaliana populations
PopulationNumber of rosettesFlowering timeFruiting plantsFruits per plantLeaf number (NV)Flowering time (NV)Leaf number (V)Flowering time (V)
  1. Field traits: maximum number of rosettes, flowering time, total number of fruiting plants and mean number of fruits per plant. Glasshouse traits: leaf number and flowering time with (V) and without (NV) vernalization treatment. The 95% confidence intervals for QST and h2 values are given in parentheses. For code locations, see Fig. 1.

AGU0.56 (0.47–0.63)0.14 (0.07–0.19)0.36 (0.27–0.42)0.04 (0.002–0.07)0.92 (0.89–0.94)0.91 (0.87–0.93)0.48 (0.39–0.55)0.71 (0.64–0.76)
CDC0.44 (0.34–0.51)0.50 (0.41–0.57)0.29 (0.21–0.36)0.04 (0.006–0.07)0.61 (0.52–0.67)0.70 (0.62–0.75)0.59 (0.41–0.68)0.51 (0.46–0.55)
GRA0.31 (0.23–0.38)0.0 (0.0–0.0)0.0 (0.0–0.0)0.0 (0.0–0.0)0.56 (0.47–0.62)0.07 (0.03–0.11)0.20 (0.13–0.25)0.09 (0.04–0.12)
LEO0.65 (0.63–0.66)0.61 (0.52–0.67)0.19 (0.12–0.24)0.0 (0.0–0.0)0.15 (0.09–0.20)0.44 (0.35–0.51)0.74 (0.67–0.79)0.68 (0.59–0.73)
MAR0.69 (0.67–0.70)0.57 (0.48–0.63)0.40 (0.31–0.47)0.09 (0.05–0.13)0.97 (0.96–0.98)0.92 (0.89–0.94)0.84 (0.79–0.87)0.48 (0.39–0.55)
PRA0.50 (0.41–0.57)0.28 (0.20–0.34)0.28 (0.20–0.35)0.02 (0.0–0.05)0.74 (0.67–0.78)0.65 (0.56–0.70)0.27 (0.19–0.33)0.08 (0.04–0.12)
QUI0.23 (0.15–0.30)0.54 (0.45–0.61)0.18 (0.12–0.24)0.0 (0.0–0.0)0.81 (0.74–0.84)0.69 (0.61–0.74)0.70 (0.62–0.75)0.36 (0.27–0.42)
SAN0.47 (0.37–0.54)0.25 (0.18–0.31)0.08 (0.04–0.12)0.05 (0.02–0.08)0.91 (0.88–0.93)0.94 (0.91–0.95)0.83 (0.77–0.86)0.45 (0.36–0.52)
Q ST 0.40 (0.27–0.48)0.77 (0.67–0.82)0.46 (0.33–0.54)0.06 (0.00–0.18)0.66 (0.54–0.72)0.65 (0.53–0.72)0.61 (0.48–0.68)0.52 (0.39–0.60)

Genetic differentiation among populations given by FST estimates was 0.35 (95% CI = 0.30–0.40) and 0.51 (95% CI = 0.49–0.54) for the sets of 14 microsatellites and 198 SNPs, respectively. The QST estimates ranged from a low of 0.06 (95% CI = 0.0−0.18) for the mean number of fruits per plant to a high of 0.77 (95% CI = 0.67−0.82) for flowering time estimated in field conditions (Table 5). Comparisons between QST values of phenotypic traits and FST values obtained with the two sets of molecular markers yielded different results based on the overlap between their 95% CI estimates (Fig. 4). The QST estimates for the mean number of fruits per plant was always significantly lower than any FST value, whereas the QST estimates for flowering time estimated in field conditions were significantly higher than all FST values (Fig. 4). The 95% CI of the rest of the QST estimates for the other traits fell within those of the FST estimates (Fig. 4). These results show that A. thaliana populations are genetically more differentiated for flowering time in field conditions than for presumably neutral markers.

image

Figure 4. Quantitative genetic differentiation (QST) and neutral genetic differentiation (FST) values (± 95%CI) of Arabidopsis thaliana populations estimated in field and glasshouse conditions. Vernalized (V) and nonvernalized (NV) traits from glasshouse experiments are indicated. The FST values are given for 14 nuclear microsatellites and 198 single nucleotide polymorphisms (SNPs).

Download figure to PowerPoint

To further compare the patterns of population differentiation between quantitative traits and molecular markers, we analysed isolation-by-distance using FST or QST estimates. The significant correlations detected between FST and geographic distances (= 0.61 and = 0.66, = 0.012 and = 0.001, for microsatellites and SNPs, respectively) indicated the presence of a geographic pattern derived from the demographic history of these populations. By contrast, these populations did not show a significant pattern of isolation-by distance when comparing QST values for any trait (> 0.09 in all cases), except for the number of vegetative rosettes estimated in the field experiment, which showed a positive significant relationship (= 0.52, = 0.03). Therefore, other factors different from demography, such as divergent local adaptation, probably contribute to the larger population differentiation observed for flowering time variation in field conditions.

Association analyses between flowering traits and flowering genes

Sequence analysis of FRI, FLC, CRY2 and PHYC flowering genes detected a total of 164 polymorphisms, but only 57 of them were nonsingletons and did not show within-gene linkage disequilibrium. Two common FRI truncations (FRI-294 and FRI-428; Méndez-Vigo et al., 2011) showed significant association with all flowering traits estimated in glasshouse conditions (Table 6). One common FLC polymorphism (FLC-765), showing strong linkage disequilibrium with FRI truncations, also affected all flowering traits estimated in glasshouse conditions except for flowering time without vernalization (Table 6). The rare allele of these polymorphisms accelerated flowering and accounted for a large amount of phenotypic variation in flowering traits (Table 6).

Table 6. Association analyses between flowering traits and flowering genes in Arabidopsis thaliana
GenePolymorphismGene locationTraitsMAF (%)P-valueR2 (%)
  1. Traits: flowering time for nonvernalized (FT) and vernalized (VFT) plants, and number of leaves at flowering for nonvernalized (LN) and vernalized (VLN) plants. The gene, its polymorphism, its effect, and the corresponding flowering traits are indicated. The minor allele frequency (MAF) of polymorphisms, the range of P-values and R2 of association analyses are also given.

FRI TruncationsFRI-294, FRI-428FT, LN, VFT, VLN16.23.0 × 10−12 – 1.9 × 10−540.1–20.9
FLCSNP-765First intron substitutionLN, VFT, VLN17.51.1 × 10−8 – 1.1 × 10−534.3–22.0

None of the phenotypic traits from field experiments showed significant associations with gene polymorphisms. However, A. thaliana genotypes bearing the FRI truncations that significantly affected flowering traits in glasshouse conditions exhibited lower recruitment in the field experiment (Fig. 5). Furthermore, the relationships between flowering time estimated in field and glasshouse conditions were also affected by the presence of such FRI truncations. In the case of individuals with functionally active FRI alleles (= 67), the correlation between flowering time in the field and in the glasshouse with or without vernalization was positive and significant in both cases (> 0.68, < 0.0001; Fig. 6). By contrast, individuals bearing FRI truncations (= 13; all 10 individuals from QUI plus three individuals from AGU – AGU-0, AGU-2 and AGU-18) exhibited a significant negative correlation between flowering time estimates in the field and in the glasshouse without vernalization treatment (= −0.62, = 0.025; Fig. 6b), whereas this correlation was not significant when using data from the vernalization treatment (= 0.12; Fig. 6a).

image

Figure 5. Difference in recruitment between sets of Arabidopsis thaliana individuals with different FRI alleles. Data indicate the mean (± SE) number of vegetative rosettes per pot for the set of individuals with (= 13; orange bars) and without (= 67; blue bars) FRI truncations observed during the field experiment. Significance of one-way ANOVAs between groups of individuals for each survey: **, < 0.001; *, < 0.05.

Download figure to PowerPoint

image

Figure 6. Relationships between flowering times (number of days) of Arabidopsis thaliana individuals estimated in the field experiment and in the glasshouse. Flowering time in the glasshouse was estimated with (a) and without (b) vernalization treatment. Relationships are given for individuals with (= 13; orange dots) and without (= 67; blue dots) FRI truncations. Trend lines are shown when Pearson's correlations were significant.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

In this study, we evaluated the extent of among- and within-population variation in life-history traits of A. thaliana in a natural field setting. In general, the results indicated that natural A. thaliana populations contain significant genetic variation for life-history traits. The results also revealed that individuals from the GRA population exhibited the highest performance for relevant demographic traits, such as the maximum number of vegetative rosettes and the total number of fruiting plants (Table 2). As GRA represents the local population, this result suggests adaptation of GRA individuals to the local environment. Nevertheless, given the strong spatial component of local adaptation to abiotic environmental conditions, further research is needed to address this important issue in A. thaliana. For example, a macro-ecological approach may assess the effects of environmental similarities between native and experimental environments on fitness components on a large-scale spatially explicit collection of A. thaliana accessions.

We have shown that flowering time estimated in field and glasshouse conditions were all significantly positively correlated in this set of Iberian A. thaliana populations. At the population level, flowering time in field and glasshouse conditions with vernalization exhibited a very similar pattern while the among-population pattern of variation in flowering time in glasshouse conditions without vernalization was more variable (Fig. S3). This result indicates that the environmental conditions at El Castillejo Botanical Garden in southwest Spain also promoted the vernalization pathway for flowering in A. thaliana. Despite that fact that winters are mild at El Castillejo Botanical Garden (Fig. S2), the mean monthly minimum temperatures of 5°C recorded in January may have been enough to activate the response to vernalization of A. thaliana study individuals. A recent study has shown that vegetative rosettes from populations from much colder environments, such as high-altitude montane locations, hardly survive the winter and that such populations are chiefly composed by spring-germinated plants that overwinter as seeds in the soil seed bank (Picó, 2012). Hence, we hypothesize that natural environments with mild and moderately cold winters activate the vernalization pathway to promote flowering during winter. When winters are too severe, A. thaliana mostly behaves as a spring annual to complete its life cycle in such environments.

The evolutionary relevance of flowering time over the remaining life-history traits estimated in the field experiment was highlighted in this study through QSTFST comparisons, an approach that has previously been conducted in A. thaliana, but only for flowering time estimates under glasshouse conditions (Le Corre, 2005; Stenøien et al., 2005; Porcher et al., 2006). Given the inherent caveats of the QSTFST approach (Leinonen et al., 2008), these comparisons must be treated with caution. However, different results from our study support the conclusion that flowering time in A. thaliana is likely to be a trait under local divergent selection in our set of Iberian populations. First, flowering time estimated in field conditions was the only trait exhibiting significantly higher QST values than the FST value estimated with SNPs (Fig. 4). In addition, QST for flowering traits estimated without vernalization in glasshouse conditions were also significantly higher than the FST value derived from microsatellites, in agreement with previous studies (Le Corre, 2005). It must be emphasized that the comparison of different molecular markers to estimate FST values revealed the strong dependence of this approach on marker type. Our results support the suggestion that SNPs are more suitable than microsatellites for QSTFST comparisons because high mutation rates of microsatellites tend to upwardly bias the difference between QST and FST (Edelaar et al., 2011). Second, the comparison of different flowering time values estimated in different conditions indicated that flowering time or closely related traits, such as leaf number, exhibit high QST values in comparison with the other life-history traits (Fig. 4). Third, the distinct isolation-by-distance pattern observed for FST and QST values obtained for flowering time in the field support the idea that other evolutionary factors different from demography contribute to the strong population differentiation for this trait. Overall, our results reinforce the widely accepted view that flowering time plays an important role in shaping A. thaliana's life history and illustrate how inherent demographic processes (i.e. isolation-by-distance) and divergent selection mainly on flowering time may jointly account for phenotypic and genotypic population differentiation in A. thaliana.

Our experiments also revealed that there exists an important amount of within-population genetic variation for flowering time in natural A. thaliana populations whose expression varies among environments. Studies addressing the genetic and molecular basis of flowering time variation in A. thaliana have successfully detected latitudinal (Caicedo et al., 2004; Stinchcombe et al., 2004; Lempe et al., 2005; Shindo et al., 2005; Samis et al., 2008, 2012; Méndez-Vigo et al., 2011), and recently altitudinal (Méndez-Vigo et al., 2011), geographical patterns of variation. It is noteworthy that we still largely ignore the underlying mechanisms that generate and maintain within-population variation in flowering time, a key component to comprehensive understanding of evolutionary change in this important trait.

One way to evaluate the causes of within-population variation in flowering time observed in our experiments is illustrated by the analysis of the effect of polymorphisms in important flowering genes. In this study, we found two loss-of-function FRI alleles and a previously described potential FLC polymorphism (Méndez-Vigo et al., 2011) that were significantly associated with flowering time variation in this set of Iberian A. thaliana populations. Interestingly, these FRI and FLC polymorphisms, and the two major haplogroups described for PHYC and CRY2 (Olsen et al., 2004; Balasubramanian et al., 2006) segregated within one to three populations, indicating their contribution to within-population variation. In addition, we have been able to observe several important demographic and phenotypic effects of some of these polymorphisms. For example, individuals bearing FRI truncations exhibited lower recruitment rates throughout the field experiment (Fig. 5), which suggests FRI pleiotropy on major life-history traits in A. thaliana. This finding does not represent an exception because FLC has also been found to play a role in temperature-mediated germination (Chiang et al., 2009) and seed dormancy may also influence flowering time in A. thaliana (Rubio de Casas et al., 2012). Furthermore, FRI truncations altered the relationship between flowering time values estimated in different environments (Fig. 6), as individuals carrying FRI truncations flowered earlier in the glasshouse but later in field conditions. This result is in agreement with the early flowering behavior observed under natural over-wintered conditions in accessions carrying FRI functional alleles in combination with FLC alleles of haplogroup A (Caicedo et al., 2004; Samis et al., 2012), the only haplogroup found in the Iberian Peninsula (Méndez-Vigo et al., 2011). Therefore, FRI may also contribute to the observed G × E interactions, reinforcing the importance of plasticity in flowering time to the understanding of gene function in A. thaliana. Nevertheless, given the low number of individuals carrying FRI truncations included in these analyses and the close genetic relationships among them, further studies are necessary to ensure that the observed FRI effects are not caused by other genes in linkage disequilibrium with FRI.

Overall, this study stresses the need to adopt multidisciplinary and integrative approaches to study comprehensively the evolutionary genetics of flowering time variation in A. thaliana. The increase in available genomic data for world-wide natural A. thaliana accessions (Weigel, 2012) will facilitate the simultaneous study of natural variation in multiple gene polymorphisms. However, further efforts are needed to generate phenotypic data and reaction norms from different environments, including more populations and especially more individuals per population, which represent an important limitation for the advance of evolutionary genetics of ecologically important life-history traits.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We are grateful to C. Cárdenas and the staff of the El Castillejo Botanical Garden for field assistance, and M. Ramiro, J. Pozas and E. Manzano for laboratory assistance. We also thank the administration of the Sierra de Grazalema Natural Park for permission to work in the El Castillejo Botanical Garden. The CEGEN Genotyping Service (Barcelona, Spain) carried out the marker genotyping. Adrian C. Brennan and two anonymous reviewers improved the manuscript. This work was supported by Ministerio de Ciencia e Innovación of Spain (grants CGL2006-09792/BOS and CGL2009-07847/BOS to F.X.P.; grant BIO2010-15022 to C.A-B.) and the Ministry of Higher Education and State of Scientific Research of Egypt (ParOwn Grant 1207 Cycle) to N.H.G. Data is deposited in the Dryad repository: http://dx.doi.org/10.5061/dryad.3gk26.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

FilenameFormatSizeDescription
nph12082-sup-FigS1-S4.pdfapplication/PDF3755K

Fig. S1 Photographs of the experimental setting at El Castillejo Botanical Garden, Spain, including the general view of the four blocks, dimensions of each block, vegetative rosettes and fruiting plants of Arabidopsis thaliana.

Fig. S2 Daily minimum and maximum temperatures and total precipitation at El Castillejo Botanical Garden, Spain, during the field experiment.

Fig. S3 Population mean flowering times of Arabidopsis thaliana estimated in the field and glasshouse with and without vernalization illustrating the significant population × environment interaction found in this study.

Fig. S4 Genetic structure of nonredundant multilocus Arabidopsis thaliana genotypes based on 198 polymorphic single-nucleotide polymorphisms (SNPs), and estimated with the structure program.