SEARCH

SEARCH BY CITATION

Keywords:

  • Arabidopsis thaliana ;
  • flowering time;
  • genetics of adaptation;
  • quantitative trait locus (QTL) mapping;
  • vernalization;
  • winter annual life history

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • To gain an understanding of the genetic basis of adaptation, we conducted quantitative trait locus (QTL) mapping for flowering time variation between two winter annual populations of Arabidopsis thaliana that are locally adapted and display distinct flowering times.
  • QTL mapping was performed with large (n = 384) F2 populations with and without vernalization, in order to reveal both the genetic basis of a vernalization requirement and that of variation in flowering time given vernalization.
  • In the nonvernalization treatment, none of the Sweden parents flowered, whereas all of the Italy parents and 42% of the F2s flowered. We identified three QTLs for flowering without vernalization, with much of the variation being attributed to a QTL co-localizing with FLOWERING LOCUS C (FLC). In the vernalization treatment, all parents and F2s flowered, and six QTLs of small to moderate effect were revealed, with underlying candidate genes that are members of the vernalization pathway. We found no evidence for a role of FRIGIDA in the regulation of flowering times.
  • These results contribute to a growing body of evidence aimed at the identification of ecologically relevant genetic variation for flowering time in Arabidopsis, and set the stage for functional studies to determine the link between flowering time loci and fitness.

Introduction

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

Throughout a species' range, populations adapt to local conditions. Understanding the genetic basis of such local adaptation has been a major goal in evolutionary biology for over a century. The last decade has witnessed profound advances in molecular genetics approaches for the identification of genomic regions associated with adaptive traits, allowing for the rigorous assessment of this long-standing question (Barton & Keightley, 2002).

Plants have developed sophisticated physiological, developmental and genetic mechanisms to optimize the time of flowering (Bastow & Dean, 2003). Because the transition from vegetative growth to flowering is crucial for plant reproductive success, plants integrate different environmental cues to achieve a flowering response that is adapted to local conditions. The observation that the flowering time of many plant populations varies with latitude or altitude suggests that this trait contributes to geographic adaptation (Lacey, 1988; Kalisz & Wardle, 1994; Olsson & Ågren, 2002). Hence, an understanding of the pathways and genetic mechanisms that contribute to the natural variation in flowering time is a central goal of many studies in plant evolutionary genomics.

The model plant Arabidopsis thaliana (hereafter referred to as Arabidopsis) serves as a workhorse for the study of plant molecular biology. This species is distributed widely throughout northern latitudes and exhibits substantial natural variation in flowering time. Two different approaches have been used to characterize the life history and flowering behavior of Arabidopsis (Nordborg & Bergelson, 1999). The ecological criterion is based on the timing of seed germination and flowering; winter annuals germinate in the fall, overwinter as rosettes and flower in late spring, whereas summer annuals germinate in the spring and flower in mid to late summer. Based on this criterion, Arabidopsis has a winter annual life history throughout most of its range, although summer annuals have also been described (Napp-Zinn, 1985; Pigliucci, 2003; Koornneef et al., 2004; Shindo et al., 2007). Climatic data from the southern edge of the native range indicate that, although soil temperatures very rarely fall below freezing, they do fall below 4°C (Ågren & Schemske, 2012), which is a sufficient temperature for a vernalization response (Nordborg & Bergelson, 1999; Shindo et al., 2005).

By contrast, the physiological criterion classifies plants as winter annuals if vernalization, that is, a period of cold temperature at the seed or rosette stage, is required for flowering, and as summer annuals if plants can flower without vernalization. These categories are also referred to as ‘early’ or ‘late’ flowering, respectively (Clarke et al., 1995; Gazzani et al., 2003), because the vegetative phase is shorter in plants that do not require vernalization. Some populations flower without vernalization in the laboratory, and would therefore be classified as summer annuals by the physiological criterion, but are winter annuals under the ecological criterion (D. W. Schemske & J. Ågren, pers. comm.). We suggest that the ecological criterion for the classification of life history should be adopted universally, because it best reflects the actual timing of growth and flowering in the field.

The genetic basis of flowering time in Arabidopsis has received considerable attention (Simpson & Dean, 2002; Sung & Amasino, 2004; Amasino, 2010). The analysis of Arabidopsis mutants has characterized multiple genetic pathways that regulate the transition from vegetative to reproductive growth. The photoperiod and vernalization pathways regulate the response to environmental signals, whereas the autonomous and gibberellin pathways respond to endogenous signals and are functionally independent from environmental cues (Simpson & Dean, 2002). These pathways form a complex network of > 100 genes that regulate flowering. However, only a few genes identified from mutant screens have also been associated with the natural variation in flowering time. Of these, FRIGIDA (FRI) and FLOWERING LOCUS C (FLC) are thought to be major regulators of flowering in natural populations (Burn et al., 1993; Lee et al., 1993; Clarke & Dean, 1994; Lempe et al., 2005). FLC, a transcription factor encoding a MADs-box protein, represses the expression of other transcription factors promoting flowering (Michaels & Amasino, 1999). Functional alleles at FRI delay flowering by activating the strong expression of FLC (Johanson et al., 2000). Vernalization represses FLC expression by epigenetic modification, reducing its sensitivity to FRI, and thus promotes flowering (Amasino, 2004).

The molecular analysis of natural accessions has revealed considerable allelic variation for FRI and FLC, and many mutations are known that cause inactivation of either FRI or FLC. Such mutants are often significantly associated with early flowering (Johanson et al., 2000; Le Corre et al., 2002; Gazzani et al., 2003; Lempe et al., 2005). Over 70% of early-flowering accessions contain loss-of-function alleles at the FRI locus (Shindo et al., 2005). In a geographic survey of the relationship between flowering time and FRI, active FRI alleles were associated with a latitudinal cline in flowering, but no such cline was detected in accessions with nonfunctional FRI (Stinchcombe et al., 2004). In the presence of functional FRI alleles, FLC allelic variation contributes to a latitudinal cline in flowering, and FLC expression is most highly correlated with flowering time variation (Caicedo et al., 2004; Lempe et al., 2005; Shindo et al., 2005). These results suggest that FRI and FLC are targets of natural selection (Stinchcombe et al., 2004; Izawa, 2007). Although FRI and FLC are important determinants of natural variation in flowering time for some Arabidopsis populations, extensive molecular and genetic analysis of Arabidopsis accessions has indicated that other loci also contribute to flowering time (Lempe et al., 2005; Werner et al., 2005a,b; Li et al., 2006).

Quantitative trait locus (QTL) mapping has proven to be a powerful method for the identification of the genetic basis of quantitative trait variation (Tanksley, 1993; Barton & Keightley, 2002; Mitchell-Olds & Schmitt, 2006), and numerous QTL studies have been conducted examining flowering time variation in Arabidopsis. The majority of these studies include the mapping of populations derived from one or both of the laboratory strains Columbia (Col) and Landsberg erecta (Ler), both of which could be classified as summer annuals based on a physiological criterion alone (Lister & Dean, 1993). Knowing where in the life cycle variation in flowering time is manifest is a critical step in assessing the mechanisms responsible for population differences in flowering time. For example, in plants requiring vernalization, the developmental and physiological pathways that contribute to the variation in flowering time among populations may differ from those of populations that do not require vernalization. It is thus important to understand the life histories of populations grown under conditions of their native environments.

Recently, Ågren & Schemske (2012) have presented the results of a multi-year reciprocal transplant study between Arabidopsis populations originating from Sweden and Italy – the geographic limits of the native distribution. This experiment is the first of its kind involving Arabidopsis (Lowry, 2012), and has demonstrated strong adaptive differentiation of these populations to their source environment. In this study, freezing tolerance and flowering time were identified as putative adaptive traits conferring local adaptation. At the Italian field site, the Italy population flowered 33 and 50 d before Sweden in the 2 yr of study; at the Swedish field site, the Italy population flowered 3 d before Sweden in both years (Ågren & Schemske, 2012). Here, we report the results of extensive genetic mapping experiments designed to examine the genetic basis of variation in flowering time differences between these populations. Given the winter annual life history of the study populations, we employed a two-tiered approach to dissect the genetic basis of flowering time. We first searched for the QTLs required for plants to flower in the absence of vernalization, and then identified the QTLs that contribute to variation in flowering following vernalization. In this way, we are able to decouple flowering time variation from a vernalization requirement.

Our study addressed the following questions. What are the number, location and magnitude of effect for QTLs contributing to flowering time? What are the genetic regions responsible for differences in vernalization response and do these contribute to flowering time variation after vernalization? Do QTLs co-localize with FRI and FLC as seen in other studies? Do candidate genes underlying QTLs differ between parental populations in coding region sequence?

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 system

This experiment is part of an ongoing effort to understand the mechanisms of adaptation in natural populations of Arabidopsis. In this study, we used one population collected from central Italy, Castelnuevo di Porto (42°07′N, 12°29′E), and one from north-central Sweden, Rödåsen (62°48′N, 18°12′E), as mapping parents. These individuals are derived from the same maternal lines as utilized by Ågren & Schemske (2012) – additional information on these populations is presented there. Both the Italy and Sweden populations are winter annuals; plants germinate from late August to early September in Sweden and from late October to mid-November in Italy, and overwinter as rosettes in both regions. The flowering phenology differs between the two areas, with the Italy population flowering and fruiting much earlier (March–April) than the Sweden populations (May–June) (Ågren & Schemske, 2012). A plant from Italy was utilized as the maternal parent and was crossed to Sweden to produce an F1 generation. A large F2 population was then generated by autonomous self-pollination of an F1.

Growing conditions

Arabidopsis seeds were cold stratified for 4 d at 6°C and were then sown on nutrient agar and allowed to germinate in a growth chamber under long-day conditions (16 h). Twenty-day-old seedlings were transplanted into 2-inch pots and were allowed to grow for an additional 4 wk (total age, c. 7 wk), at which time they were randomly assigned to nonvernalization or vernalization treatments. The nonvernalization treatment proceeded at 22°C under a long-day condition with 16 h of light (total photosynthetically active radiation (PAR) = c. 125 μmol m−2 s−1]. Plants in the vernalization treatment experienced 5 wk of vernalization at 6°C with a short-day condition of 10 h of light (total PAR = c. 50 μmol m−2 s−1). After this 5-wk period, the conditions were switched back to long days, as described above in the nonvernalization treatment. For each treatment, there were 50 plants from each parental population and 384 F2 hybrids.

Phenotyping

The dates of bolting and of first flowering were recorded for each plant. The date of bolting was defined as the date on which the inflorescence extended beyond the rosette. The date of first flowering was the date on which the first flower opened, defined as the stage at which the petals reflexed to expose the stigma. In the nonvernalization treatment, plants were coded as ‘1’ if they flowered and ‘0’ if they did not flower. For the vernalization treatment, plants were coded as above, and the date on which plants flowered was recorded. For data analysis, plants were scored as the number of days relative to the first plant that bolted or flowered.

Primer design and marker selection

Simple sequence repeats (SSRs) were utilized to genotype the mapping population. Ninety-nine SSR primer pairs were obtained from The Arabidopsis Information Resource (TAIR) website and another 202 SSR primer pairs were designed after searching for microsatellites in the Arabidopsis genome of the Columbia accession. The total, 301 SSR loci, were screened for polymorphisms between the Italy and Sweden mapping parents; 103 polymorphic markers were identified; 71 of these were used for segregation analysis in 96 F2 plants. Segregation distortion was examined by chi-squared tests (P < 0.05) to identify markers that deviated from an expected 1 : 2 : 1 ratio (Lu et al., 2002). In the end, 62 nondistorted markers were selected to construct the linkage map, providing even coverage of the Arabidopsis genome. DNA was isolated using FastDNA kits (Q-BIOgene, Carlsbad, CA, USA). SSRs were amplified by polymerase chain reaction (PCR) and were visualized on 3% agarose gels stained with ethidium bromide.

Genetic map and QTL analysis

The genetic map was constructed on the basis of 62 SSR markers and 768 F2 plants from the combined vernalization and nonvernalization treatments. The linkage analysis was performed on the basis of 706 F2 individuals using JoinMap 3.0 (Van Ooijen & Voorips, 2001). The linkage groups were separated using the highest logarithm of odds (LOD), 10.0. Recombination values were converted to genetic distances using the Kosambi mapping function. Five linkage groups were obtained, corresponding to the five Arabidopsis chromosomes. For QTL analysis, the mapping populations included 384 F2 plants in both the vernalization and nonvernalization treatments. In the vernalization treatment, QTL analysis was conducted with the raw flowering time data and quantile-normalized flowering times. Both the raw and transformed data yielded essentially the same results, with the same number of QTLs being identified and the effect sizes and peak LOD positions changing only slightly. Here, we present the QTL results with the raw flowering time data. QTL analysis was performed with the R/qtl package (Broman et al., 2003). We used a Haley–Knott regression (Haley & Knott, 1992) in the ‘stepwiseqtl’ function to detect QTLs, with the normal and binary models being utilized in the vernalization and nonvernalization treatments, respectively. Genome-wide significance (P = 0.01) thresholds for QTL detection were determined with 10 000 permutations. For all QTLs, we determined 2-LOD confidence intervals, as well as estimates of the additive and dominance allelic effects. Epistasis between QTLs was tested using two-way analysis of variance (ANOVA) (Westerbergh & Doebley, 2002; Morgan & Mackay, 2006). The markers closest to the LOD peaks of QTLs were tested for interaction.

Candidate genes

A comprehensive list of potential candidate genes was compiled by searching TAIR 10 (Brachi et al., 2010). Candidates were selected that contained ‘flowering’ or ‘vernalization’ in their name and/or description. This list included a total of 165 unique genes. To reveal candidate genes underlying a QTL, the sequence of SSR markers flanking a QTL was located within the Columbia genome. The locations of the flowering/vernalization genes were located in the Columbia genome using the Chromosome Map Tool, available on TAIR. Genes that lie within markers flanking a QTL were considered as potential candidates. Candidate genes largely fell into three categories: (1) genes that are involved in basic biological processes with mutants/transgenics showing a suite of abnormal phenotypes, including flowering; in addition, published work is lacking or there is no direct evidence of a role in regulating flowering time or development; (2) genes that are involved in development during the transition from vegetative growth to flowering; typically, these genes have been well characterized, but have no known role in controlling the initiation of flowering; (3) genes that are members of the photoperiod or vernalization pathway in controlling flowering time, most of which have been well studied and identified as candidates in published QTL studies, or associated with natural variation in flowering time. These categories were classified as weak, medium or strong candidates, respectively (Supporting Information Table S1).

Gene sequencing

The coding regions of several strong candidate genes underlying QTLs were sequenced in an effort to identify putative causal polymorphisms (Salome et al., 2011). Sequencing primers were designed to amplify the gene regions based on the Columbia genome. Gene regions were PCR amplified using a high-fidelity polymerase (PfuUltra II Fusion HS DNA Polymerase, Stratagene; La Jolla, California, USA). Successful amplification was determined by running PCR products on a 1.5% agarose gel. DNA bands were then cut from the gel and purified (Wizard SV Gel and PCR Clean-Up, Promega). PCR products were sequenced at the Michigan State University Genomics Facility. To ensure that the correct gene region was amplified, the sequence was compared with the Columbia genome via a BLAST search. Sequence data were analyzed with the Staden package Pregap4 and Gap4 (http://staden.sourceforge.net/). Polymorphisms were compared between the mapping parents in BioEdit (http://www.mbio.ncsu.edu/bioedit/page2.html). Potential polymorphisms between mapping parents were confirmed by re-amplification and sequencing of the specific region. A search of natural accessions available on TAIR was conducted to identify common polymorphisms with the mapping parents.

Results

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

Phenotypic data

The days to bolting and flowering were shown to be highly correlated (R2 = 0.94), with plants flowering an average of 6.2 d after bolting. Because of this, only the number of days to flowering was used for analysis. In the nonvernalization treatment, all (42/42) of the parental plants from Italy (Fig. 1) and 43% (165/384) of the F2 plants (Fig. 2) flowered, but none of the Sweden parents flowered (0/42). Thus, the Italy population is able to flower without vernalization. For the vernalization treatment, all of the parents from Italy (44/44) and Sweden (40/40), as well as all of the F2s, flowered. As all plants flowered in this treatment, the 5-wk cold treatment was sufficient for saturating vernalization. Although all plants flowered in this treatment, there was considerable flowering time differentiation between populations, with the mean flowering date for Italy being 28 d earlier than Sweden (Fig. 3). Transgressive segregation was not detected in this mapping population, which is in contrast with other studies (El-Lithy et al., 2006; O'Neill et al., 2008; Brachi et al., 2010; Salome et al., 2011; Seymour et al., 2012).

image

Figure 1. Proportion of Arabidopsis thaliana populations that flowered in the nonvernalization treatment. The number above each bar indicates the percentage of each group that flowered.

Download figure to PowerPoint

image

Figure 2. Days to flowering of the Arabidopsis thaliana F2s in the nonvernalization treatment. (Note: this only includes the plants that flowered.)

Download figure to PowerPoint

image

Figure 3. Days to flowering of the Arabdiopsis thaliana F2s in the vernalization treatment. The gray bars represent the flowering duration of the respective parents, with the black triangle representing the mean flowering date.

Download figure to PowerPoint

QTL results

Nine QTLs were identified in this study, with three and six QTLs being identified in the nonvernalization and vernalization treatments, respectively (Table 1). Of these, eight were unique, as one of the QTLs in the vernalization (VF4) and nonvernalization (NVF2) treatments mapped to the same chromosomal location, with overlapping LOD support intervals and identical QTL peaks. Although this shared QTL co-localizes between treatments, the effect size differs, with the percentage phenotypic variance explained by the QTL (PVE) of 32.8% and 10.5% in the nonvernalization and vernalization treatments, respectively. Two other pairs of QTLs between treatments (NFF1/VF1 and NVF3/VF6) have slightly overlapping LOD support intervals, but distinct LOD peaks, and therefore are considered as separate QTLs.

Table 1. Results of quantitative trait locus (QTL) analysis of flowering time between Arabidopsis thaliana populations
TreatmentQTLChPositionSupport intervalLOD scorePVEAdditive effect (SE)Dominance effectNearest markerCandidate gene
  1. Ch, chromosome on which the QTL is located. Position, peak logarithm of odds (LOD) position on the chromosome in cM. Support interval, 2 LOD. PVE, percentage phenotypic variance explained by the QTL. Nearest marker, simple sequence repeat (SSR) marker closest to the QTL peak. Candidate genes, known flowering genes (see text for definitions) underlying the 2-LOD support interval of a QTL.

NonvernalizationNVF1170.0060.00–82.009.05.9−1.50 (0.25)−0.16 (0.34)F5114 FT
NVF258.005.96–10.0041.132.8−3.80 (0.49)−0.42 (0.50)NGA158 FLC
NVF3560.0048.01–68.09.16.0−1.30 (0.24)0.78 (0.32)C1W9 LATE
VernalizationVF1180.0469.71–83.594.83.1−1.50 (0.33)0.36 (0.51)MSAT1.5 MAF1
VF2318.8011.6–31.446.74.41.62 (0.36)−1.84 (0.50)MSAT3.2VIL1, VRN2, DNF
VF3363.1956.74–63.199.26.1−2.0 (0.33)−1.08 (0.47)NGA6 
VF457.901.99–12.0515.310.5−3.42 (0.40)0.27 (0.50)NGA158 FLC
VF5537.8826.09–44.494.52.9−1.88 (0.45)−0.95 (0.52)T5E15 
VF6568.5462.71–74.04.52.9−1.78 (0.40)−0.39 (0.48)MCK7-1VIN3, VIP4, ELF5, MAF2–5

QTL effect sizes ranged from 2.9% to 32.8% (PVE), with seven of the nine QTLs having effect sizes under 10% PVE. The positions of the QTLs in each treatment are provided in Fig. 4. No QTLs were detected on chromosomes 2 and 4. With the exception of one QTL (VF2), the Italy alleles result in early flowering and the Sweden alleles in later flowering. Significant epistasis between QTLs was identified in both treatments (Table S2). In the nonvernalization treatment, all three of the QTLs were involved in an epistatic interaction. In this treatment, the large effect QTL (NVF2), which co-localizes with FLC, interacts with both of the other QTLs. In the vernalization treatment, there is a marginally significant interaction of the QTL that co-localizes with FLC (VF4) and VF1, as well as one other interacting pair of QTLs (Table S2).

image

Figure 4. Linkage map with the flowering time quantitative trait loci (QTL) and candidate genes. The ruler provides the genetic distance in cM. The green bars represent the five Arabidopsis thaliana chromosomes. The labeled black bars to the right of the green bar indicate the name and position for each of the microsatellite markers. The labeled black bars to the left of the green bars indicate the position of specific candidate flowering time genes. The blue and red bars represent QTLs identified in the vernalization and nonvernalization treatments, respectively. The length of these bars represents the 2-LOD support interval, with the black triangle indicating the logarithm of odds (LOD) peak.

Download figure to PowerPoint

Candidate gene analysis

In this study, seven of the nine QTL intervals harbor at least one strong candidate gene (Table 1). Of these, all but two, DAY NEUTRAL FLOWERING (DNF) and LATE FLOWERING (LATE), have been discussed as candidates in previous QTL studies. The QTL VF2 encompasses DNF, as well as two other common candidates, Vernalization Insensitive 3-like 1 (VIL1) and REDUCED VERNALIZATION RESPONSE 1 (VRN1). LATE is the only strong candidate underlying NVF3, which has recently been characterized as controlling the expression of certain flowering genes in leaf vasculature and floral meristem identity genes at the shoot apex (Weingartner et al., 2011). Two QTLs in the vernalization treatment (VF3 and VF5) do not possess a strong candidate gene. A possible, yet unlikely, candidate for VF5 is SERRATED LEAVES AND EARLY FLOWERING (SEF); SEF mutants show a suite of abnormal flowering phenotypes (March-Diaz et al., 2007). Two developmental genes underlie VF3, LATE MERISTEM IDENTITY2 (LMI2) and GLABROUS INFLORESCENCE STEMS (GIS), but they have not been studied with regard to the control of flowering time.

Recently, Salome et al. (2011) have identified flowering time QTLs across a range of Arabidopsis accessions and found that these genomic regions include the candidate genes FRI, FLC, and MADS AFFECTING FLOWERING1 (MAF1) and MAF2–5. They sequenced and compared the coding regions of these candidates across accessions (Salome et al., 2011). For comparison with their results, we sequenced the coding regions of the same genes (FRI, FLC, MAF1 and MAF2), and three other strong candidates underlying QTLs in our mapping population, VERNALIZATION INSENSITIVE 3 (VIN3), VERNALIZATION INSENSITIVE 3-LIKE 1 (VIL1), and FLOWERING LOCUS T (FT). Polymorphisms between the mapping parents are presented in Table 2. All of the genes, except for FLC, have at least one nonsynonymous substitution between the Italy and Sweden mapping parents. Many of the polymorphisms found between the mapping parents were also identified by Salome et al. (2011). Only one indel was identified; a 3-bp indel in exon 5 of VIN3 (Table 2).

Table 2. Polymorphisms of candidate gene coding regions between the Italy and Sweden Arabidopsis thaliana mapping parents
GenePositionExonAAItalySwedenColItaly AASweden AACol AAOther accessions
  1. Position, position of the polymorphism in the Col genome. AA, amino acid number of the polymorphism. Italy, Sweden and Col, sequence of these respective accessions. Italy, Sweden and Col AA, amino acid abbreviations of these respective accessions. Other accessions, other accessions that share the same polymorphism as the Italy (It) or Sweden (Sw) mapping parent as indicated.

FLC 3 173 8058156GAGVVVSw: Tamm-2
FRI 269 4621146CGAAGESw: Bur-0, Tsu-1, H51
269 4691148GGAMMISw/It: Bur-0, Tsu-1, H51
269 962–269 9631312-  16-bp del   Sw/It: H51
FT 24 331 561118GAGVIV 
24 333 5484129GGCGGGSw/It: Bay-0, Bor-4, Est-1, Fei-0, Lov-5, NFA-8, RRS-7, Sha, Tamm-2, Ts-1, Tsu-1, C24, RRS-10
MAF1 28 958 9686141AGGKEE 
28 959 8219189AGGPPP 
MAF2 25 982 475121GTTVFF 
VIL1 8 877 6172159TGGVGG 
8 876 9283363TGGIRRIt: Tsu-1
8 876 6003472GAARRRIt: Tsu-1
8 876 5983473AGGQRRIt: Tsu-1
8 876 4643518CCTPPSSw/It: Br-0, Fei-0, Ler-1, Sha, Tamm-2, Ts-1, Bur-0
VIN3 23 248 316379CAALMM 
23 248 305382GGAEEESw/It: Bur-0
23 247 6444274ACAMLMSw: Lov-5
23 247 5724298AAGTTASw/It: Ct-1, Edi-0, Kn-0, Oy-0, Po-0, Sf-2, Wil-2, Ws-0
23 246 987–23 246 8935502–503 3-bp del QGRQG 

Discussion

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

Vernalization requirement

Here, we report a QTL mapping study of flowering time differentiation utilizing a large mapping population generated from two natural accessions of winter annuals. In the present study, QTL mapping was conducted with and without a vernalization treatment, allowing us to decouple a vernalization requirement from flowering time variation per se. In the nonvernalization treatment, all of the parental plants from Italy and none from Sweden flowered. Plants from Italy are therefore able to flower in the absence of vernalization; yet, in nature, they behave as true winter annuals, overwintering as rosettes and flowering in the spring (Ågren & Schemske, 2012). Shindo et al. (2005) examined the vernalization response in accessions collected over a wide latitudinal range. When plants did not receive vernalization, all of the accessions collected above 62°N did not flower, all of the accessions collected below 45°N did flower, and intermediate latitudes displayed a range of responses. The Italy and Sweden populations in this study were collected at 42°07′N and 62°48′N, respectively.

Although all of the parents from Italy flowered, it should be noted that there was variation in the time required to initiate flowering, with plants flowering over a period of 36 d (Supporting Information Fig. S1). Under the constant temperature and day length conditions of the nonvernalization treatment, plants initiate flowering through an autonomous mechanism which probably does not provide as consistent a flowering response as environmental cues, such as photoperiod and temperature. In the vernalization treatment, all of the Italy and Sweden plants flowered, as did the F2s. Following vernalization, the parental plants from Sweden flowered, on average, 28 d later than those from Italy (Fig. 3). The earlier flowering of Italy plants after a vernalization treatment suggests that the Italy plants are able to initiate the transition from vegetative to flowering stages after a much shorter period of cold exposure than are the Sweden plants. This is consistent with data from the Italian field site, where the Italy populations flower substantially earlier than those of Sweden (Ågren & Schemske, 2012).

Genetic architecture of flowering time

A major goal of evolutionary biology has been to characterize the number and effect size of genes responsible for adaptive variation. Numerous QTL studies of adaptive traits have provided support for Orr's (1998) model, predicting an exponential distribution of effect sizes with few of large effect (Barton & Keightley, 2002). For nearly 20 yr, Arabidopsis researchers have been dissecting the genetic basis of flowering time. We outline the traditional linkage (QTL) mapping publications in Table S3; this includes 26 publications, with 98 different QTL experiments. These studies differ in the mapping parents utilized, size and type of mapping populations, vernalization treatment, photoperiod and growing conditions. The average number of QTLs identified is 4.03; we detected a typical number of QTLs, with studies ranging between one (O'Neill et al., 2008; Balasubramanian et al., 2009) and 10 (Weinig et al., 2002). Our results are in agreement with most studies in which at least one large effect QTL is identified, but in contrast with studies that have detected a greater occurrence of large effect QTLs (Loudet et al., 2002; Salome et al., 2011; Strange et al., 2011). The ability to detect QTLs of small effect is limited by the size of the mapping population (Beavis, 1998). Many previous QTL studies for flowering time have involved mapping populations with < 150 individuals, and therefore limited ability to detect small effect QTLs. Here, we are able to identify QTLs with effect sizes as low as 2.9% PVE. The distribution of QTL effect sizes in the present study provides additional support for Orr's (1998) model. In addition, we detected a moderate amount of epistasis between QTLs (Table S2). The QTLs that co-localize with FLC (NVF2 and VF4) are involved in the majority of the significant QTL × QTL interactions. This may be expected because of the central role of FLC in genetic pathways of floral initiation (Simpson & Dean, 2002).

Effect of environment on QTLs

Understanding the role of genotype–environment interaction has been a long-standing goal in evolutionary biology, as this can influence local adaptation (Kawecki & Ebert, 2004). Several studies in Arabidopsis have characterized significant QTL × environment interactions (Ungerer et al., 2002; Weinig et al., 2002; Li et al., 2006). The ability to detect QTLs co-localizing with FRI can depend on the environmental conditions (Table S3). For example, in the extensively utilized mapping population, Col × Ler, the detection of FRI varies with vernalization and photoperiod (Stratton, 1998; Weinig et al., 2002, 2003; Juenger et al., 2005). Here, we report distinct QTL profiles between the vernalization and nonvernalization treatments, with only one QTL shared between them. The effect size of this shared QTL differs substantially between the treatments (Table 1). In the nonvernalization treatment, the QTL is of large effect, with a PVE of 32.8%, as opposed to 10.5% in the vernalization treatment. FLC stands out as a candidate gene for the shared QTL (NVF2 and VF4). FLC is a strong inhibitor of flowering and is routinely identified in studies examining flowering time variation (Table S3). Another candidate in the nonvernalization treatment is FT; this QTL was not detected after vernalization.

In an analogous experiment, Strange et al. (2011) examined the effect of different vernalization treatments on mapping populations of four accessions from Sweden crossed with Col. Overall, they also found that the effect size of QTLs decreased after vernalization. More specifically, the effect size of the QTL co-localizing with FLC and FT went from having a strong effect with no vernalization, to being insignificant after 14 wk of vernalization (Strange et al., 2011). Our results suggest that differences between Italy and Sweden at the FLC locus allow for populations from Italy to flower in the absence of vernalization. Vernalization acts to repress FLC by epigenetic modification, allowing for the transition from vegetative to flowering states (Amasino, 2004). FLC allelic variation has been shown to be associated with a latitudinal cline in flowering time variation. In southern populations, alleles that weaken FLC allow for flowering without vernalization. It is to be expected that FLC would have a smaller effect size in the vernalization treatment because vernalization would provide the strong repression of FLC in both the parental populations. The nonvernalization treatment yields QTLs with candidates involved in the photoperiod regulation of flowering (FLC and FT), whereas the vernalization treatment involves candidates of the vernalization pathway (VRN1, VIL1, VIN3). In the present work, only vernalization was varied between treatments; future studies should include photoperiods similar to those in the habitats of both populations.

Candidate genes underlying QTLs

Many of the QTLs identified here have major candidate genes that map to the same chromosomal location (Table 1). It is possible that these candidate genes are the causal genetic elements underlying QTLs; however, functional experiments are required to confirm their role in flowering time variation. As indicated above, FLC is a prime candidate for the shared QTL in both the vernalization and nonvernalization treatment. We did not identify any nonsynonymous polymorphisms in the FLC coding region, which is in agreement with other studies examining FLC variation (Gazzani et al., 2003; Caicedo et al., 2004; Strange et al., 2011).

In addition to FLC, the nonvernalization treatment yields FT as a possible candidate gene. FT has been identified as a key component of the photoperiod pathway of floral induction across plant species (Kardailsky et al., 1999; Kobayashi et al., 1999; Böhlenius et al., 2006; Tamaki et al., 2007) and is highly conserved (Kobayashi et al., 1999). We have identified one nonsynonymous polymorphism in the FT coding region; however, it is unlikely to have an effect as both amino acids are neutral and nonpolar (Table 2). Recent studies have failed to detect nonsynonymous polymorphisms in the coding region of FT, but rather have found considerable cis-regulatory variation (Schwartz et al., 2009; Adrian et al., 2010; Strange et al., 2011). In our study, the QTL (NVF1) that encompasses FT is of relatively small effect size and was only identified in the nonvernalization treatment. FLC acts to repress FT expression until it is inhibited by vernalization or an autonomous mechanism. It is logical that these candidates were detected in the nonvernalization treatment as allelic variation between the mapping parents would provide differential sensitivity of FT to FLC in the absence of vernalization. With vernalization, the effect of FT allelic variation could be negligible once FLC is sufficiently inhibited.

In the vernalization treatment, there are several candidate genes identified, including all members of the FLC/MAF clade, which Salome et al. (2011) have highlighted as being common across QTL mapping studies. MAF1, also known as FLM, is a candidate underlying VF1. MAF1 is a homologue of FLC and also acts to inhibit flowering in a similar fashion (Scortecci et al., 2001). We have identified two synonymous substitutions and one nonsynonymous substitution in the coding region (Table 2). Given that there is functional redundancy between FLC and MAF1, it would be expected that weakened alleles at both genes would be necessary to allow for earlier flowering.

The QTL on the end of chromosome V, VF6, encompasses seven potential candidate genes (Table S1). MAF2–4 act to repress flowering after short cold spells (Ratcliffe et al., 2001), and polymorphisms at these loci have been shown to be associated with diverse flowering times in Arabidopsis accessions (Caicedo et al., 2009). Other strong candidates at the locus include VIN3, VERNALIZATION INDEPENDENCE 4 (VIP4) and EARLY FLOWERING 5 (ELF5) (Table S1). We have sequenced the coding regions of MAF2 and VIN3. In MAF2, there is one nonsynonymous substitution between Italy and Sweden; however, there is no change in polarity or charge. However, in the VIN3 coding region, there are two nonsynonymous substitutions, as well as a 3-bp deletion (Table 2).

Three genes, VRN1, VIL1 and DNF, underlie the LOD support interval of VF2. VRN2 and VIL1 act to promote flowering and are upstream members of the vernalization pathway (Gendall et al., 2001). DNF acts in the repression of CO during the early part of the day (Morris et al., 2010). For successful vernalization, it is important to note that two distinct processes must occur: the initial repression of FLC and the maintenance of FLC inhibition following vernalization. VIL1 (together with VIN3) is necessary for the chromatin modification of FLC/MAF1 repression.

The genetics of flowering time in Arabidopsis have been examined extensively, and the molecular genetics pathways have been well characterized by a variety of means, including QTL mapping. However, almost all of these QTL studies have involved the laboratory strains Col, Ler or both as mapping parents, and have been performed under controlled laboratory conditions (Table S3). In the present study, we have identified two QTLs that do not have a clear candidate gene (VF3 and VF5). Two floral developmental genes underlie these QTLs, but neither would be considered a strong candidate, or have been described in previous QTL studies. A possible candidate for VF5 is SEF, which has not been thoroughly studied to date. Although it is rare to reveal novel loci involved in flowering time, it is possible, as demonstrated in the present work through the mapping of natural populations, and in a recent study that investigated natural accessions crossed with Col in the field (Brachi et al., 2010).

The elucidation of the role of cis-regulatory and protein coding variation is a central goal in studying the genetic basis of adaptation. Hoekstra & Coyne (2007) highlighted that considerable emphasis has been placed on cis-regulatory evolution; however, sufficient data are lacking to make generalizations. Here, we examined only the protein coding regions of several known candidate genes underlying QTLs. Rigorous functional experiments are required to identify the causal genes and mutations controlling flowering time adaptation. Nonetheless, in several cases, a lack of protein coding variation suggests the role of cis-regulatory changes. In FT and FLC, there are no structural polymorphisms between the mapping parents, which is in agreement with studies characterizing the regulatory variation of these genes (Gazzani et al., 2003; Michaels et al., 2003; Schwartz et al., 2009; Adrian et al., 2010; Strange et al., 2011). By contrast, VIN3 could potentially represent protein coding changes responsible for flowering time variation, as has been well described in FRI (Johanson et al., 2000; Gazzani et al., 2003).

The role of FRI in controlling flowering time

Although many genes control flowering time, few have been shown to be associated with natural flowering time variation. The regulatory gene FRI has received the most attention as a major determinant of flowering time in natural populations of Arabidopsis (Clarke & Dean, 1994; Johanson et al., 2000; Le Corre et al., 2002; Simpson & Dean, 2002). Functional FRI alleles result in the accumulation of FLC mRNA, which inhibits flowering. Vernalization acts to reduce the sensitivity of FLC to FRI, and thereby promotes flowering. Several lines of evidence support the role of FRI in controlling natural flowering time variation, including the identification of several naturally occurring, independently derived null FRI alleles (Le Corre et al., 2002), a large number of nonsynonymous polymorphisms in the first exon of FRI (Le Corre et al., 2002), the identification of QTLs co-localizing with FRI in several QTL studies with different parental populations varying in flowering time (Salome et al., 2011), and evidence for FRI alleles influencing fitness depending on the FLC allelic background and environmental conditions (Korves et al., 2007). Despite this substantial body of research demonstrating the contribution of FRI in flowering time variation, we found no role of FRI allelic variation in controlling flowering time differentiation in the Italy and Sweden populations.

It has been well established that coding region variation in exon 1 of FRI results in nonfunctional alleles and early flowering (Johanson et al., 2000; Le Corre et al., 2002; Gazzani et al., 2003). We have identified one polymorphism in FRI between the Italy and Sweden mapping parents which results in a glycine to alanine substitution at amino acid 146 (Table 2). Both of these amino acids are neutral and nonpolar, and are unlikely to affect protein function. Furthermore, no QTLs were found to co-localize with FRI.

This may be surprising, given that a latitudinal cline in flowering time has been predicted to be associated with FRI allele functionality (Karlsson et al., 1993; Nordborg & Bergelson, 1999; Pigliucci & Marlow, 2001). That is, populations in southern latitudes are expected to harbor null FRI alleles providing flowering in the absence of vernalization, and northern populations will contain functional FRI alleles. However, several studies have been unable to detect such a cline in flowering time (Karlsson et al., 1993; Nordborg & Bergelson, 1999; Pigliucci & Marlow, 2001; Stinchcombe et al., 2004). In the most extensive study, Stinchcombe et al. (2004) identified functional FRI alleles to be associated with a cline in flowering, but no association was detected for nonfunctional FRI alleles, as predicted. Furthermore, the populations from southern latitudes with functional FRI alleles actually flowered earlier than those from northern latitudes. The variation in flowering time is explained by an epistatic interaction between the functional FRI alleles and FLC (Caicedo et al., 2004). Our results are in agreement with those of Stinchcombe et al. (2004) and Caicedo et al. (2004). In the present study, FRI is not identified as a candidate gene, but FLC is in both the vernalization and nonvernalization treatments. This makes sense when taking life history into consideration. Given that both of the parental populations are winter annuals, FRI functional alleles would be favored to delay flowering until the appropriate time in the spring (Simpson & Dean, 2002). Weak alleles of FLC in the Italy mapping parent reduce the sensitivity to FRI and promote flowering after mild periods of cold compared with Sweden (Amasino, 2004).

Conclusion

In summary, we have presented a QTL mapping study of flowering time variation between two winter annual accessions of Arabidopsis collected from the northern and southern geographic limits of their native range. Our study benefits from a large mapping population size and is thus capable of accurately estimating effect sizes. Although we identified both large and small effect QTLs, it is almost certain that we have missed many QTLs of very small effect (≈ 1% PVE). We found the genetic architecture of adaptation to depend heavily on the environmental conditions, namely the vernalization treatment. Previous QTL studies have emphasized the role of FRI and FLC in determining flowering time differentiation. Although substantial variation is explained by a QTL co-localizing with FLC, we found no role of FRI. Furthermore, it should be noted that the genetic basis of flowering time in this system is complex and includes several QTLs with numerous other candidate genes. Future research should be conducted to thoroughly examine the causal genetic elements underlying QTLs, as well as field studies to determine the relationship of flowering time and fitness. An integration of these pursuits will provide a comprehensive understanding of the genetic basis of flowering time in Arabidopsis.

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 would like to thank M. Cameron for assistance with figure development, K. Califf for assistance with sequence alignment, C. Oakley for help with R/qtl, and R. Amasino and D. Weigel, as well as two anonymous reviewers, for helpful comments on the manuscript.

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
nph12109-sup-0001-FigS1.docxWord document78KFig. S1 Days to flowering of the Arabdiopsis thaliana parental plants from Italy in the vernalization treatment.
nph12109-sup-0002-TableS1.xlsxapplication/msexcel14KTable S1 Arabidopsis thaliana candidate flowering time genes underlying the quantitative trait loci (QTLs) listed in Table 1
nph12109-sup-0003-TableS2.docxWord document15KTable S2 Epistatic interactions between Arabidopsis thaliana flowering time quantitative trait loci (QTLs)
nph12109-sup-0004-TableS3.xlsxapplication/msexcel21KTable S3 Summary of Arabidopsis thaliana flowering time quantitative trait locus (QTL) studies