Role of vernalization and of duplicated FLOWERING LOCUS C in the perennial Arabidopsis lyrata

Transcription

1 Research Role of vernalization and of duplicated FLOWERING LOCUS C in the perennial Arabidopsis lyrata Ulla Kemi 1, Anne Niittyvuopio 1, Tuomas Toivainen 1,2, Anu Pasanen 1,Bénédicte Quilot-Turion 1,3, Karl Holm 4, Ulf Lagercrantz 4, Outi Savolainen 1,2 and Helmi Kuittinen 1 1 Department of Biology, University of Oulu, PO Box 3000, FIN-90401, Oulu, Finland; 2 Biocenter Oulu, University of Oulu, 90014, Oulu, Finland; 3 INRA, UR1052 Génétique et Amélioration des Fruits et Légumes, F-84143, Montfavet, France; 4 Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvägen 18D, SE-752, 36 Uppsala, Sweden Author for correspondence: Ulla Kemi Tel: ulla.kemi@oulu.fi Received: 12 July 2012 Accepted: 7 September 2012 doi: /j x Key words: Arabidopsis lyrata, expression quantitative trait locus, FLOWERING LOCUS C, flowering, gene duplication, gene expression, vernalization. Summary FLOWERING LOCUS C (FLC) is one of the main genes influencing the vernalization requirement and natural flowering time variation in the annual Arabidopsis thaliana. Here we studied the effects of vernalization on flowering and its genetic basis in the perennial Arabidopsis lyrata. Two tandemly duplicated FLC genes (FLC1 and FLC2) were compared with respect to expression and DNA sequence. The effect of vernalization on flowering and on the expression of FLC1 was studied in three European populations. The genetic basis of the FLC1 expression difference between two of the populations was further studied by expression quantitative trait locus (eqtl) mapping and sequence analysis. FLC1 was shown to have a likely role in the vernalization requirement for flowering in A. lyrata. Vernalization decreased its expression and the northern study populations showed higher FLC1 expression than the southern one. eqtl mapping between two of the populations revealed one eqtl affecting FLC1 expression in the genomic region containing the FLC genes. Most FLC1 sequence differences between the study populations were found in the promoter region and in the first intron. Variation in the FLC1 sequence may cause differences in FLC1 expression between lateand early-flowering A. lyrata populations. Introduction Plant populations may have extensive ranges with highly variable environments with respect to temperature and the amount and quality of light (Munguía-Rosas et al., 2011). Organisms living at high latitudes also experience seasons with highly contrasting environmental conditions. Synchronizing developmental transitions with the correct season is important for plants to reproduce successfully. In the model plant Arabidopsis thaliana, many genetic pathways are known to mediate environmental signals and affect flowering (Ausin et al., 2005). The importance of a particular pathway may differ between different environments and this plasticity can confer an evolutionary advantage (King & Heide, 2009). Among the flowering pathways, the vernalization pathway is important in many temperate species. The vernalization requirement ensures that plants do not flower in the fall when the environmental conditions are unfavorable for reproduction. The strength of the vernalization requirement can vary within plant species. For example, Norwegian high-latitude strawberry (Fragaria vesca) populations have an obligatory vernalization requirement, unlike other strawberry populations (Heide & Sønsteby, 2007). Plants need to be of a given age or size to be able to receive vernalization or other flowering-promoting signals (Lacey, 1986; Bernier & Perilleux, 2005). Flowering is size-dependent, especially in perennials: size correlates with the amount of available resources for reproduction, vegetative growth, overwintering and forthcoming growing seasons. In contrast, annuals need to reproduce within their first year after germination, regardless of available resources (Lacey, 1986). The length of the juvenile period when plants are nonresponsive to vernalization varies between species: for example, some grasses can receive a vernalization signal even at the seed stage whereas other grasses become sensitive to vernalization when they are older (Heide, 1994 and references therein). In winter annual accessions of A. thaliana, both seeds and rosettes respond to vernalization, but seed vernalization is more effective (Nordborg & Bergelson, 1999). Many perennials in the Brassicaceae, such as Arabis alpina and Brassica oleracea, become sensitive to vernalization after several weeks of growth (Lin et al., 2005; Wang et al., 2011). The genes FLOWERING LOCUS C (FLC )andfrigida (FRI ) are important components of the vernalization pathway in A. thaliana ( Johanson et al., 2000; Sheldon et al., 2000). Several studies have shown that these two genes may account for most of 323

2 324 Research New Phytologist the observed flowering time variation among accessions of A. thaliana (Johanson et al., 2000; Gazzani et al., 2003; Shindo et al., 2005), although some studies have indicated a more limited role in governing that variation (Scarcelli et al., 2007; Brachi et al., 2010). FLC is mostly expressed in leaves, in root tips and in the shoot apex (Michaels & Amasino, 1999; Sheldon et al., 1999). FLC encodes a MADS-box transcription factor that inhibits the expression of floral pathway integrator genes a set of genes that combine information from flowering pathways and regulate the expression of flower meristem identity genes (Samach et al., 2000; Simpson & Dean, 2002). Winter annual A. thaliana accessions germinate in the fall (Nordborg & Bergelson, 1999) and flowering is repressed before winter by high FLC expression. During the cold period, FLC expression is reduced by epigenetic chromatin remodeling (Bastow et al., 2004; Sung & Amasino, 2004) and in the spring plants have the competence to flower (Wilczek et al., 2009). FLC expression is promoted by a transcription activation complex, in which FRI and SUPPRESSOR OF FRIGIDA4 (SUF4) are among the main components (Choi et al., 2011), and inhibited by the genes of the autonomous and vernalization pathways (reviewed by Henderson & Dean, 2004; Wang et al., 2012). Small regulatory RNAs transcribed from the FLC locus, COLDAIR from intron 1 and the antisense transcript COOLAIR covering the entire FLC locus, are also involved in FLC expression regulation (Swiezewski et al.,2009;heo&sung,2011).genomic regions in the promoter and in the first intron are required for FLC down-regulation and maintenance of the low expression (Sheldon et al., 2002; Sung et al., 2006). FLC orthologs are known to mediate vernalization effects also in plant species with different life-history strategies, such as in the biennial B. oleracea (Lin et al., 2005) and in the polycarpic perennials A. alpina (Wang et al., 2009) and Arabidopsis halleri (Aikawa et al., 2010). In addition to mediating vernalization effects, the FLC ortholog in A. alpina, PERPETUAL FLOWERING 1 (PEP1), was shown to have roles in limiting the duration of flowering and in governing the polycarpic flowering habit by regulating the fate of meristems (Wang et al., 2009). Studies on the effect of FLC on flowering time variation among populations within the same species are needed to reveal the role of FLC in local adaptation. FLC is triplicated in a short-lived perennial Arabidopsis arenosa and duplicated in Arabidopsis lyrata (Nah & Chen, 2010) as a consequence of sequential tandem duplications. In A. arenosa, the copies are differentially expressed, probably as a result of changes in the regulatory regions in the promoter (Wang et al., 2006; Nah & Chen, 2010). The expression and the role of the two FLC genes in A. lyrata have not been studied, although it was suggested that the two FLC genes in A. lyrata and A. arenosa have the same function as FLC in A. thaliana (Nah & Chen, 2010). Here we report on the effects of vernalization on flowering and on the expression of the downstream copy (FLC1) of the two tandemly arranged FLC genes in Norwegian, Swedish and German A. lyrata populations. We have also studied the genetic basis of between-population differences in FLC1 expression. The FLC genes (FLC1 and FLC2) are located at the top of linkage group 6 in the A. lyrata genome and the distance between the genes is c. 22kbp (Hu et al., 2011).Wefirstexaminedtheeffectsofseedlingand rosette vernalization to find out whether both are effective in this species. Secondly, we compared the expression and DNA sequences of the two FLC genes in Norwegian and German populations. We showed that FLC1 probably has the main role in the vernalization requirement in both populations, and then focused on this gene. Thirdly, we measured FLC1 expression in Norwegian, Swedish and German A. lyrata populations before and after vernalization. We hypothesized that, if FLC1 mediates vernalization effects on flowering in A. lyrata, its expression will be decreased by vernalization. Further, as the Norwegian and Swedish populations flower later and respond more to vernalization than the German population (Kuittinen et al., 2008), we expected to find higher FLC1 expression in the northern populations. Fourthly, in an expression quantitative trait locus (eqtl) mapping experiment, we searched for genetic loci affecting FLC1 expression differences between the German and the Norwegian populations. We specifically asked whether there is an eqtl in the FLC or in the FRI genomic region. Finally, we examined the FLC1 DNA sequence in the parental populations to detect causal variants for the observed eqtl. On the basis of our eqtl mapping results, we searched for sequence differences at important regulatory regions. Materials and Methods Study populations Arabidopsis lyrata (L.) O Kane & Al-Shehbaz (Brassicaceae) is a perennial, polycarpic and outcrossing plant species, which has become a model plant for ecological genetics and genome evolution (Clauss & Mitchell-Olds, 2006; Hu et al., 2011; Savolainen & Kuittinen, 2011). The species is divided into two subspecies, of which ssp. petraea inhabits Europe with a fragmented distribution. Here, three A. lyrata ssp. petraea populations were studied: the Norwegian Spiterstulen (hereafter Sp ; N, 8 24 E), the German Plech ( Pl ; N, E) and the Swedish Mjällom ( Mj ; N, E; population Storsanden in Leinonen et al., 2009). Many environmental factors differ between these locations; for example, in Plech the thermal growing season (period from five successive days with daily-average temperature >+5.0 C to five successive days with daily-average temperature <+5.0 C) is > 150 d, whereas in Spiterstulen it is < 50 d (Norwegian Meteorological Institute; Germany s National Meteorological Service; B. Quilot-Turion et al., unpublished). These populations were chosen because they represent the northern and southern parts of the distribution area of A. lyrata in Europe, because they have been shown to be locally adapted to their native environments (Leinonen et al., 2009) and because they flower at different times in common garden experiments and respond differently to rosette vernalization (Riihimäki & Savolainen, 2004; Kuittinen et al., 2008). Effects of vernalization on flowering We studied the effect of vernalization applied to plants at two different developmental stages. In the first experiment, newly

3 New Phytologist Research 325 germinated seeds from the three study populations were exposed to vernalization (seedling vernalization). In the second experiment, plants experienced vernalization at 4 wk of age (rosette vernalization). Both experiments had controls that did not receive vernalization. In the seedling vernalization experiment, 40 seeds from 10 to 20 families per population (generated by crosses in the laboratory) were germinated for 12 d on Petri dishes (+18 C; light:dark cycle (LD) 8 : 16 h), planted in pots with fertilized peat and gravel and vernalized for 9 wk (+4 C; LD 8 : 16 h). As a result of a timer failure, the light was continuous in the cold room for a period of unknown length within the first 4.5 wk of vernalization (and within the first 5.5 wk of rosette vernalization). As vernalization is effective both in short and in long days in A. lyrata (Kuittinen et al., 2008; P. H. Leinonen et al., unpublished) and as the photoperiod was correct for several weeks at the end of the 9-wk vernalization, this should not have had much influence on the results. Moreover, the plants in both vernalization experiments experienced the same photoperiod. After vernalization the pots were placed in the growth room (+22 C; LD 20 : 4 h; Power Star HQI bulbs, OSRAM, Augsburg, Germany) together with germinated and planted seeds (40 per population) that had not received vernalization and served as controls. The vernalized and control plants were randomized within eight blocks. The rosette vernalization experiment was similar, except that the plants were grown in a growth chamber (+22 C; LD 20 : 4 h) for 29 d before vernalization. Control plants were planted 9 wk later and were not exposed to vernalization. The results for days to emergence of flower bud are briefly described in Kuittinen et al. (2008). Here we analyze additional traits from the same experiment and compare the results with those from the seedling vernalization experiment. Additional randomized plants for the expression study were included in this experiment. In both vernalization experiments, days to flowering (first open flower), excluding time in vernalization, was scored, as well as the number of leaves and the rosette size at the start of flowering. To find out whether the populations respond differently to vernalization treatments, the interaction between population and vernalization was studied with likelihood ratio tests between linear models (lm function in R) within each vernalization experiment. If the interaction term was not significant, the main effect of population and vernalization was also tested. Log 10 transformation was used for traits whose residuals were not normally distributed. The block did not have a significant effect on any traits (Kruskal Wallis tests; results not shown), and thus the data were analyzed as a single randomized block. The likelihood ratio tests were also used to study the effect of rosette and seedling vernalization on the studied traits within each population. v 2 tests were used to study the effect of vernalization on flowering probability within the populations. The analyses were performed in the statistical environment R v (R Development Core Team, 2010). Expression studies FLC1 expression in populations The effect of vernalization and population on FLC1 expression was studied in two experiments by quantitative PCR (qpcr). In the first experiment (experiment 1; rosette vernalization experiment described in the paragraph Effects of vernalization on flowering ), the apical meristem and the youngest leaves of 8 Pl, 10 Sp and 10 Mj plants before vernalization and of 12 Pl, 11 Sp and 9 Mj plants 4 d after vernalization were collected for the expression measurement. The samples were disrupted by TissueLyser (Qiagen) and total RNA was isolated with the RNeasy Mini Kit (Qiagen) from the lysate. An optional on-column DNase digestion (RNase-Free DNase Set; Qiagen) was included in the extraction procedure, according to the manufacturer s instructions. cdna was synthesized in reverse transcription reactions containing random primers (Invitrogen), DTT (Invitrogen), dntp mix and SuperScript III RT enzyme (Invitrogen). A dilution of 1/100 was used for qpcr reactions, which were run with the 7000 Sequence Detection System ver 3.0 (Applied Biosystems, Carlsbad, CA, USA). The qpcr product was detected with Platinum SYBR Green qpcr SuperMix-UDG (Invitrogen), and on average three technical replicates were run. FLC1 expression was normalized with that of the reference gene b-tubulin6 (TUB6) (primers are shown in Supporting Information Table S1). Details of the sample storage, reverse transcription, qpcr and primer design are given in Notes S1. In the second experiment (experiment 2), Sp and Pl plants experienced two rounds of vernalization (+4 C; LD 8 : 16 h, for 8 wk). The pre-growing conditions were identical to those for studying the effects of rosette vernalization, except that the day length was 14 h. Between vernalization treatments, plants were allowed to flower in the growth room (+22 C; LD 20 : 4 h) for 3.5 months. Young leaves were collected for FLC1 expression quantification before and after the first vernalization (sampling times 1 (19 Pl and 19 Sp samples) and 2 (17 Pl and 20 Sp samples)), and then before and after the second vernalization (sampling times 3 (8 Pl and 14 Sp samples) and 4 (10 Pl and 13 Sp samples)). Samples were disrupted in the way described for experiment 1 and RNA was extracted with the RNeasy 96 Kit (Qiagen) or the RNeasy Plant Mini Kit (Qiagen); cdna synthesis was performed in experiment 1 and qpcr reactions were run with the 7000 Sequence Detection System ver 3.0 (Applied Biosystems) or with LightCycler 480 (Roche) (Notes S1). These plants represented parental populations in the common garden experiment that was also used for eqtl mapping of FLC1 expression (see next paragraph). The effects of rosette vernalization and population on FLC1 expression in both experiments were analyzed in R (v ; R Development Core Team, 2008) with analysis of covariance. The response variable was the FLC1 cycle threshold (C t )valueand explanatory variables were the TUB6 C t value, sampling time and/ or population. The covariance approach generally results in similar estimates but lower standard errors and confidence intervals than the traditional DDC t method (Yuan et al., 2006). Instead of subtracting the estimates of the reference gene from each target gene estimate before statistical analysis, reference gene estimates were included as a covariate in the model. The regression coefficients of the TUB6 C t value (0.51 and 0.53) were taken into account when the FLC1 expression was normalized for Figs 5 and 6, respectively.

4 326 Research New Phytologist eqtl mapping of FLC1 expression To identify the genomic regions explaining the difference in FLC1 expression between the Sp and Pl populations, we performed eqtl mapping in a family of F2 plants. A random set of 123 F2 plants from a common garden experiment was used (see details of the crossing design in Notes S2 and Fig. S1). The plants were grown in a similar way than in experiment 2 (see the paragraph FLC1 expression in populations ). One young leaf from each plant was sampled at the age of 4 wk, before the vernalization treatment. After collection, sample storage and disruption, cdna synthesis and qpcr were performed in a similar way as in experiment 2. Here, the regression coefficient for TUB6 that was used to normalize FLC1 expression was 0.9. The linkage map was constructed with 40 markers (Notes S2; Table S2; Fig. S2). The markers were microsatellites, or CAPS (cleaved amplified polymorphic sequences) and dcaps (derived cleaved amplified polymorphic sequences) including some in flowering-time genes, such as FRI. No markers were positioned in FLC genes themselves, but there were microsatellite markers close to them. FLC1 expression was normally distributed and the eqtl analysis was performed by genome-wide interval mapping with MAPQTL 5 (van Ooijen, 2004). A 5% LOD significance level for the whole genome was defined with permutations. FLC1 and FLC2 expression comparison A subset of 72 F2 plants from the eqtl mapping experiment was used for measuring both FLC1 and FLC2 expression. The cdna samples synthesized for eqtl mapping were used and qpcr reactions were performed as described for experiment 2. The expression levels were measured in individuals from each of the four parental allele combinations, as deduced from the markers flanking the FLC region. To find out whether FLC1 is differentially expressed in Pl and Sp homozygotes among the F2 plants, we tested the expression differences between these genotype classes as for population samples. The response variable was the FLC1 C t value and the explanatory variables were the TUB6 C t value and the genotype class (Pl 1 Pl 2 or Sp 1 Sp 2 ). Before analysis, the data were log e - transformed to make the distribution of the residuals closer to normal. When the expression levels of both genes were normalized for Fig. 4, a regression coefficient of 0.83 was used. Sequence analysis The promoter and gene sequences of both FLC genes of two randomly chosen Sp and Pl individuals were sequenced (FLC1: from 2894 to (sequence set 1); FLC2: from 832 to (sequence set 2); site information based on the published A. lyrata sequence (Hu et al., 2011) and indicated from the start of the translation start codon; for some individuals the obtained sequence was slightly shorter). As a consequence of technical issues, we did not obtain the full gene sequence of FLC2 for both Pl individuals, and the Pl sequence was obtained by combining the sequences of the two individuals (individual Pl56: from 809 to and from to +6071; individual Pl59: from to +3808). The full FLC1 gene region of one of each F2 homozygote for parental alleles was also sequenced (genotype classes Pl 1 Pl 2 and Sp 1 Sp 2 ; sequence set 3). Note that these parental alleles represent two random alleles from each natural population. Finally, we studied the FLC1 sequence polymorphism by sequencing a shorter region of 3529 bp, including 1134 bp from the end of the promoter, the first exon and 2210 bp from the beginning of intron 1 (from 1134 to bp) in 13 randomly chosen Pl and seven Sp individuals (sequence set 4). To amplify the target sequences, we used Phusion DNA polymerase products and the Piko Thermal Cycler (Finnzymes, Vantaa, Finland). The full promoter and gene region of FLC1 was amplified in three overlapping fragments and that of FLC2 in four overlapping fragments and the shorter FLC1 sequence in one fragment using primer pairs designed based on the A. lyrata sequence (Table S3). The sequencing reactions were carried out with an ABI Prism 3730 (Applied Biosystems) using the Big Dye Terminator kit v.3.1 (Applied Biosystems). Sequence data were verified by visual inspection and aligned with CODONCODEALIGNER v (CodonCode Corporation; Li-Cor, Inc, Lincoln, Nebraska, USA) and edited with MEGA 4 (Tamura et al., 2007). The sequences are deposited to NCBI/ GeneBank under accession numbers JX JX We calculated the divergence between A. lyrata FLC1 and FLC2 at nonsynonymous (K A ) and synonymous (K S ) sites and their ratio (K A /K S ; Nei & Gojobori, 1986; sequence sets 1 and 2). In addition, divergence between the A. thaliana FLC locus and the A. lyrata FLC1 and FLC2 genes was studied separately, pooling data from the two populations. The level of polymorphism in FLC1 within Sp and Pl populations was quantified separately using the whole FLC1 gene sequences (sequence sets 1 and 3) and using the shorter sequences (sequence set 4), as the number of segregating sites (S) and as the average number of pairwise differences at silent sites (p; Tajima, 1983). The divergence between Pl and Sp populations was quantified, using the same sequence sets, as the average number of nucleotide substitutions per site between populations (Dxy; Nei, 1987), as F ST (a measure of population differentiation based on mean pairwise differences; Hudson et al., 1992), and as the number of fixed differences between populations. Summary statistics for sequence data were calculated using DNASP v. 5.0 (Librado & Rozas, 2009). Results Vernalization at rosette stage is more effective in A. lyrata than at seedling stage To study the effectiveness of vernalization given at different development stages in three A. lyrata populations, flowering of plants exposed to vernalization at the seedling and rosette stages was compared. Seedling vernalization reduced days to flowering in the Pl and Sp populations and leaf number at flowering in the Mj population. However, rosette vernalization had a much bigger influence (Fig. 1a,b). When plants were vernalized at the rosette stage, they started flowering at a smaller rosette size than the control plants. This effect was not seen when vernalization treatment

5 New Phytologist Research 327 (a) (b) was given to the seedlings (Fig. 1c). The seedling and rosette vernalizations had contrasting effects on the flowering probability: in the Mj population, rosette vernalization increased and seedling vernalization decreased the probability. In the Pl and Sp populations, vernalization at neither stage affected the flowering probability (Fig. 1d). The populations responded differently to rosette vernalization in all traits (significant vernalization 9 population interaction; Table 1). Vernalization had a larger effect on the flowering time in the late-flowering Mj and Sp populations than in the earlyflowering Pl population, so that after vernalization the time to flowering was very similar among populations (Figs 1a, 2a). In contrast, all populations showed similar, small responses to the vernalization that was given to the seedlings (Table 1; Figs 1a, 2b). In addition to reducing the time to flowering, vernalization also reduced the variance of flowering time, but only when given to the rosettes (Fig. 2a,b). (c) (d) Fig. 1 Effects of rosette (left) and seedling (right) vernalization on (a) days to flowering, (b) leaf number at the start of flowering, (c) rosette size at the start of flowering, and (d) flowering probability, in German Plech (Pl), Norwegian Spiterstulen (Sp) and Swedish Mjällom (Mj) Arabidopsis lyrata populations. Mean values are shown with 95% confidence intervals. Control plants are shown as white bars and vernalized plants as gray bars. Significant differences between vernalization and control treatments within populations are indicated with asterisks (flowering probability: v 2 test; other traits: likelihood ratio tests between linear models). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, P > 0.05; na, not analyzed because of a low number of observations. FLC1, but not FLC2, is functional in both study populations To find out whether the two FLC genes in A. lyrata have similar DNA structures, their DNA sequences were compared (sequence sets 1 and 2). The comparison revealed that the genes show high similarity from c. 350 bp before the first exon to the end of the gene. Upstream of this, the genes were so different that they could not be aligned. The exons of the genes were very similar: there were nine synonymous and seven nonsynonymous differences between FLC1 and FLC2 (Fig. 3b; Table S4). Most of the nonsynonymous differences were detected in exon 4, whereas the synonymous changes were distributed more equally between the exons (Fig. 3b; Table S4). An insertion of 1651 bp, introducing stop codons, was detected in exon 4 of FLC2 in Sp individuals (Fig. 3a,b; Table S4). Intron 1 of FLC1 was bp shorter and intron 6 was bp longer than those of FLC2 (Fig. 3a), depending on the population. The divergences for nonsynonymous (K A ) and synonymous sites (K S ) between FLC1 and FLC2 were and 0.073, respectively. A comparison of FLC1 and FLC2 with A. thaliana FLC showed that the divergence in these genes were quite similar based both on nonsynonymous and synonymous sites (K A for FLC1 = and for FLC2 = 0.028; K S for FLC1 = and for FLC2 = 0.115) and on the K A /K S ratio (for FLC1 = and for FLC2 = 0.243). FLC1 and FLC2 expression was measured in a subset of F2 plants to find out whether the Sp and Pl alleles of both loci are expressed and whether the two loci show similar expression patterns. We found that both FLC genes were expressed within F2 plants, but they showed different expression patterns among the F2 genotype classes (Fig. 4a,b). Expression of FLC1 was highest in Sp homozygotes (mean value for normalized FLC1 expression: 6.67) and lowest in Pl homozygotes ( 9.24; analysis of covariance, df = 1, F = , P < 0.001; Fig. 4a). Interestingly, the likely nonfunctional Sp alleles of FLC2 were also expressed, but the Sp homozygotes showed a very low expression. In the other genotype classes, FLC2 was expressed at a higher and fairly similar level (Fig. 4b).

6 328 Research New Phytologist Table 1 Effect of vernalization, Arabidopsis lyrata population and their interaction on traits of interest in rosette and seedling vernalization experiments Vernalization 9 population Vernalization Population df F P df F P df F P Rosette vernalization Days to flowering < Leaf number at < flowering start 1 Rosette size at flowering start 1 Seedling vernalization df F P df F P df F P Days to flowering < < Leaf number at < flowering start 1 Rosette size at < flowering start 1 Data were analyzed with likelihood ratio tests between linear models. 1 Log 10 transformation. (a) (b) Fig. 2 Distributions of timing of flowering in Arabidopsis lyrata populations from Plech (Pl), Spiterstulen (Sp) and Mjällom (Mj) in (a) rosette and (b) seedling vernalization experiments. Upper row, controls; lower row, vernalized. The medians are shown by the arrowheads. Note the different scales on the y-axes. FLC1 expression differed between populations and was decreased by rosette vernalization FLC1 expression was first studied in Pl, Sp and Mj populations to find out whether the expression is reduced by vernalization in A. lyrata and whether it differs between the populations. The overall effect of rosette vernalization was to decrease FLC1 expression (Table 2; Fig. 5). Before vernalization, plants from the northern Sp population had a tendency to show the highest FLC1 expression (mean value for normalized FLC1 expression: 12.15) and plants from the southern Pl population the lowest ( 13.06). After taking into account the effect of vernalization, the between-population difference in FLC1 expression was close to significant (Table 2). When FLC1 expression was analyzed separately within nonvernalized or vernalized plants, significant differences between populations were not detected in these small samples (Table 2). FLC1 expression was further studied in Pl and Sp populations in an additional experiment, where plants experienced two rounds of vernalization. In this experiment, vernalization treatments also decreased FLC1 expression significantly (Table 2; Fig. 6). Moreover, the overall difference between the populations was significant (Table 2). When the expression was compared between the populations within each sampling time, the Sp

7 New Phytologist Research 329 (a) Fig. 3 (a) Structure of FLOWERING LOCUS C1and 2 (FLC1 and FLC2) genes in Arabidopsis lyrata. Open boxes indicate the promoter and the 3 untranslated regions, closed boxes exons and solid line introns. (b) Exons (E1 E7) of FLC1 and FLC2 in A. lyrata. Nonsynonymous substitutions are shown by open circles and synonymous substitutions by closed circles. The FLC gene in which the substitution has occurred was determined by comparing the DNA sequence with that of Arabidopsis thaliana. The location of the 1651-bp insertion in the Spiterstulen (Sp) population is indicated in both panels, but note that the insertion is not to scale in (b). Sequence sets 1 and 2 (see the Materials and Methods section for details) were used in this figure. (b) (a) (b) Fig. 4 (a) FLOWERING LOCUS C 1 (FLC1) and (b) FLOWERING LOCUS C 2 (FLC2) expression in Spiterstulen (Sp) and Plech (Pl) homozygotes and two heterozygotes for parental alleles of the Sp 9 Pl F2 population of Arabidopsis lyrata for eqtl mapping. Expression levels were measured in the samples collected before the vernalization by qpcr and normalized with expression of b-tubulin6 (TUB6), using the regression coefficient 0.83 (normalized FLC1/FLC2 expression = C t TUB6 C t FLC1/FLC2 ). The median is indicated by the thick line, lower and upper quartiles by boxes and outliers by dots. The significance of the FLC1 expression difference between Pl and Sp homozygotes was determined with analysis of covariance. population showed higher expression before both vernalization treatments, and after the first one (Table 2). Between the vernalization treatments, when plants were grown in the photoperiod of 20 h for 3.5 months, FLC1 expression was increased (Table 2; Fig. 6). FLC1 expression within the F2 plants was explained by one eqtl The genomic regions explaining the FLC1 expression variation in the Pl 9 Sp F2 cross were located by eqtl mapping. One significant eqtl with LOD score was found (Fig. 7). This eqtl was located in the FLC gene region in linkage group 6 and explained 55.4% of the expression variation within the F2 plants. Interestingly, possible FRI variation did not explain variation in FLC1 expression in this cross, as we did not observe any indication of an eqtl in linkage group 8, where FRI is located (Fig. 7). At the eqtl peak, the FLC1 expression of different genotype classes showed a pattern similar to that of the parental populations: the expression was highest in the Sp homozygotes (mean value for normalized FLC1 expression: 2.63) and lowest in Pl homozygotes ( 3.70) and the expression of heterozygote classes was intermediate ( 3.32 and 3.03). The main between-population differences in FLC1 sequence are located in the promoter region and in the first intron We compared the FLC1 sequence in Pl and Sp populations (sequence sets 1 and 3) to detect putative causal variants for the observed eqtl. The FLC1 sequences of the populations were very similar. The most notable difference between the Pl and Sp individuals was a 350-bp deletion in the promoter in Sp. Interestingly, the deletion included the area where the binding site for SUF4 is located in A. thaliana. In addition to this deletion, there

8 330 Research New Phytologist Table 2 Effect of sampling time and Arabidopsis lyrata population on FLOWERING LOCUS C 1 expression in two vernalization experiments Experiment 1 Experiment 2 df F P df F P Sampling time < < Population < Population before first vernalization Population after first vernalization Population before < second vernalization Population after second vernalization Sampling time < (before vs after first vernalization) Sampling time < (after first vs before second vernalization) Sampling time (before vs after second vernalization) < Data were analyzed with analysis of covariance. 1 Experiment 1: Plech, Spiterstulen and Mjällom; experiment 2: Plech and Spiterstulen. Fig. 6 Expression levels of FLOWERING LOCUS C 1 (FLC1) in Plech (Pl; gray) and Spiterstulen (Sp; white) populations of Arabidopsis lyrata before and after two successive vernalization treatments. Both vernalization treatments lasted for 8 wk and between the treatments, the plants were allowed to flower for 3.5 months in a photoperiod of 20 h at 22 C. The expression was measured by qpcr and normalized with expression of b-tubulin6 (TUB6) and the regression coefficient 0.53 was used (normalized FLC1 expression = C t TUB6 C t FLC1 ). The median is indicated by the thick line, lower and upper quartiles by boxes and outliers by dots. to bp (sites compared with translation start codon in the Pl and Sp sequences). Further sequencing of 13 Pl and seven Sp individuals (sequence set 4) indicated that the observed indels and SNPs are fixed in the study populations. Overall, the nucleotide polymorphism at FLC1 was lower in Sp than in Pl: the silent p was on average 5.5 times lower in Sp for the whole sequenced region (Table 3; Fig. 8b) and 15 times lower for the shorter sequenced region (Table 3). Divergence was also highest in the same region where the level of polymorphism differed most markedly (Fig. 8a). F ST between the populations was very high (0.62 for the whole gene and 0.85 for the shorter region; Table 3). Discussion Fig. 5 Expression levels of FLOWERING LOCUS C 1 (FLC1) in Plech (Pl; light gray), Spiterstulen (Sp; white) and Mjällom (Mj; dark gray) populations of Arabidopsis lyrata before and after vernalization of 9 wk. The expression was measured by qpcr and normalized with expression of b-tubulin6 (TUB6) and the regression coefficient 0.51 was used (normalized FLC1 expression = C t TUB6 C t FLC1 ). The median is indicated by the thick line, lower and upper quartiles by boxes and the outlier by a dot. were also several smaller differences between the populations (five indels and 27 SNPs), mostly located in intron 1 (Fig. 8a). The most differentiated region of intron 1 was in the area from +425 Flowering is accelerated and synchronized by rosette vernalization in A. lyrata Vernalization is known to have effects on flowering in several taxa. For instance, vernalization has been shown to decrease heading/flowering time in Phleum pratense (Fiil et al., 2011), in A. thaliana (Shindo et al., 2006) and in A. lyrata (Kuittinen et al., 2008) as well as flowering probability in Bromus tectorum (Meyer et al., 2004). Here, we showed that, like in many perennials, adult A. lyrata plants responded more to vernalization than juvenile ones. Similar results were found when flowering responses were studied in three A. lyrata populations after seed vernalization

9 New Phytologist Research 331 Fig. 7 One eqtl for FLOWERING LOCUS C 1 (FLC1) expression before the vernalization was found in linkage group 6 in the gene region containing the FLC genes (FLC1 and FLC2; location shown) in the F2 cross between the German Plech and Norwegian Spiterstulen populations of Arabidopsis lyrata. The LOD score is shown by the solid line and 5% LOD significance level (3.6) by the dashed line. No eqtl was found in linkage group 8 (or any other linkage groups), where FRIGIDA (FRI), an important regulator of FLC, was located. Genetic distances of the markers are shown in cm on the x-axis. Note that there was a marker placed in FRI, but not in the FLC genes. (a) and after overwintering as rosettes (Riihimäki & Savolainen, 2004), as natural winter increased the flowering probability more than seed vernalization. In contrast, vernalization given at seed stage is more effective than at rosette stage in A. thaliana (Nordborg & Bergelson, 1999) and seed and rosette vernalization were shown to have similar effects on flowering in the annual grass B. tectorum (Meyer et al., 2004). Annual plants can invest most of their resources into flowering and reproduction, but for perennial plants it is important to flower after gaining enough resources for flowering as well as for forthcoming growing seasons. Vernalization also synchronizes flowering of the individuals of the same population, as, in this study, variation in flowering time was greatly reduced after rosette vernalization. (b) Fig. 8 (a) Divergence, measured as Dxy, between the Spitertulen and Plech populations of Arabidopsis lyrata in FLOWERING LOCUS C 1 (FLC1). (b) Sliding window for silent nucleotide diversity (p) across FLC1 in Arabidopsis lyrata populations from Plech (solid line) and Spiterstulen (dashed line). Analyses are based on sequence sets 1 and 3 (see the Materials and Methods section for details). The structure of the FLC1 gene is shown below both figures (generated using Exon-Intron Graphic maker by Nikhil Bhatla; Open boxes indicate the promoter and 3 untranslated regions, closed boxes exons and the solid line introns. The 350-bp deletion in the promoter in the Spiterstulen population is shown by a thick horizontal black bar. The FLC1 gene may have the main role in vernalization requirement in A. lyrata It was recently suggested that all three FLC genes are functional in A. arenosa and that the FLC1 and FLC2 genes in A. arenosa and A. lyrata share the same function as FLC in A. thaliana (Nah & Chen, 2010). Here, we showed that there is a long insertion giving rise to premature stop codons in FLC2 in A. lyrata Sp plants, which indicates that FLC2 is not functional in these individuals. The insertion was not observed in the Pl sequence. However, it should be noted that we had sequence data for this area of FLC2 only from two Sp and one Pl individual and thus the insertion may be polymorphic in both populations. Interestingly, our unpublished whole-genome sequence data for several A. lyrata populations suggest that the insertion is present only in the Sp population, although not fixed (T. M. Mattila et al., unpublished). Apart from the insertion in FLC2, no other signs suggesting that this gene is nonfunctional were observed (K A /K S between the genes: 0.219). This suggests that the insertion is quite recent and additional deleterious changes have not accumulated yet. Separate comparison of FLC1 and FLC2 with A. thaliana FLC did not strongly support the hypothesis that one of the genes would be more divergent. The two FLC genes had a different expression pattern within the F2 plants: the Sp alleles showed higher FLC1 expression than the Pl alleles, but for FLC2 the opposite result was obtained. If

10 332 Research New Phytologist Table 3 Summary statistics for polymorphism and divergence in FLOWERING LOCUS C 1 in Arabidopsis lyrata Polymorphism within populations Divergence between populations 6 Population 1 Sequence set 2 L 3 N Silent S 4 Silent p 5 Dxy F st differences N of fixed Pl 1 and Sp 1 and Pl Sp Pl, Plech (Germany); Sp, Spiterstulen (Norway). 2 Set 1: randomly chosen Pl and Sp individuals, whole FLC1 gene; set 3: F2 plants (from Pl 9 Sp cross) homozygous for parental alleles, whole FLC1 gene; set 4: randomly chosen Pl and Sp plants, shorter region of FLC1; see the Materials and Methods section for details. 3 Length (bp). 4 Number of segregating sites. 5 Pairwise nucleotide heterozygosity. 6 See the Materials and Methods section for details. flowering time and vernalization requirement were regulated by an FLC gene in A. lyrata, it would be expected to have higher expression in the northern Sp population as a result of its stronger vernalization requirement compared with the more southern Pl population. Based on this reasoning, our results indicate that FLC1 has the main role in governing the vernalization requirement for flowering in A. lyrata. However, we are currently studying the expression of both FLC genes in a larger set of A. lyrata populations in different conditions to examine their functional divergence in more detail. In A. thaliana, FLC has been shown to affect flowering via the ambient temperature pathway, and to be involved in the control of the circadian clock (reviewed by D Aloia et al., 2008) and in many other developmental pathways (Scarcelli et al., 2007; Deng et al., 2011). It is thus possible that both FLC genes are functional at least in some populations of A. lyrata, but that they are involved in different developmental pathways. The two genes may still share the same function, but they may be expressed in different tissues or their expression may be reduced compared with the progenitor gene and their combined expression may be important. This may maintain both gene duplicates, as has been demonstrated in yeasts and in mammals (Qian et al., 2010). However, gene duplicates maintaining identical functions are quite rare. Silencing of one copy, partitioning of the tasks between the copies, or the development of a new function for the other copy is more likely (e.g. Lynch et al., 2001). As expression patterns differed between the FLC genes here, it seems possible that the genes are on their way to differentiate or that one of the copies is becoming a pseudogene. If the FLC duplicates have achieved different roles, it is possible that the functionally different gene regions may be in exon 4, where the genes are most clearly differentiated. Earlier it was estimated that the FLC duplication occurred c. 2.5 Mya (Nah & Chen, 2010). Based on the current mutation rate estimate of (Ossowski et al., 2010), A. lyrata and A. thaliana diverged c. 10 Mya (Hu et al., 2011) and thus the estimated time for the FLC duplication is c. 5 Mya, which is a relatively short time in the evolution of a gene. Flowering time differences between A. lyrata populations may be affected by differences in FLC1 expression Variation in FLC expression has been shown to affect natural flowering time variation in A. thaliana (Gazzani et al., 2003; Lempe et al., 2005) and also in other species in Brassicaceae, such as Brassica napus (Tadege et al., 2001). The FLC1 alleles from the early-flowering southern Pl population had lower FLC1 expression in nonvernalized F2 plants than alleles derived from the lateflowering Sp population. The same trend was seen when FLC1 expression was compared between Pl and Sp populations in two separate experiments. Our results indicate that FLC1 may be involved in the between-population differences seen in flowering time and in vernalization requirement in A. lyrata, and thus could also be involved in local adaptation. The flowering time differences between populations are likely to be affected also by genes of the other flowering pathways, as was indicated by a gene expression profile study in Capsella bursa-pastoris (Huang et al., 2012). In our experiment, no difference in flowering time was observed after the adult plants were vernalized, indicating that the vernalization pathway genes play an important role in population differentiation in flowering time in these conditions in A. lyrata. In addition to FLC, there are also other vernalization-responsive genes in A. thaliana (Michaels & Amasino, 2001) that may be involved in the differentiation of the vernalization responses in natural populations. These include VERNALIZATION INSENSITIVE 3 (Sung & Amasino, 2004), VERNALIZATION 1 (Levy et al., 2002) and VERNALIZATION 2 (Gendall et al., 2001). Flowering time variation in A. thaliana is mainly explained by several large-effect loci (e.g. Salomé et al., 2011), for instance the FLC, FRI and MADS AFFECTING FLOWERING (MAF) genes. In A. thaliana, FRI and autonomous pathway genes balance each other to achieve the appropriate level of FLC expression (Choi et al., 2011). Here, we found no indication that any upstream genes would be involved in population differentiation in FLC expression, as the only eqtl was in the FLC region itself. However, in the QTL mapping experiment the results depend on the genotypes of the parents of the F2 cross. For

11 New Phytologist Research 333 example, it was shown earlier that a polymorphism in FRI affects flowering time variation within A. lyrata populations (Kuittinen et al., 2008). However, this polymorphism did not segregate in the F2 cross we studied. We have also measured FRI expression in a subset of plants used for the population comparison in FLC1 expression (results not shown), but the preliminary results did not show any differences among these three particular populations. Our results indicate that FLC1 may affect the perennial polycarpic flowering habit in A. lyrata, as its expression was decreased by vernalization and increased between the two successive vernalization treatments. Similar rhythms in the expression of FLC orthologs have also been found in A. halleri and in A. alpina (Wang et al., 2009; Aikawa et al., 2010). Signs of directional selection in the regulatory regions of FLC1 in the northern Spiterstulen population We found one large-effect cis-eqtl explaining the FLC1 expression variation within the F2 population between the parental lines in the genomic area containing the FLC genes. Detection of cis-eqtl indicates that the sequence polymorphism in the regulatory region of the gene causes the observed expression variation, whereas a detection of trans-eqtl would suggest that the variation is caused by changes in trans-acting elements, that is, in transcription factors (Gibson & Weir, 2005). Thus, our eqtl mapping result together with the sequence analysis suggests that the differential FLC1 expression may be caused by changes in the regulatory region of FLC1. Nucleotide variation was strongly reduced in the Sp population and differentiation between the Sp and Pl populations was increased, consistent with recent directional selection (Smith & Haigh, 1974). The observed F ST (0.62) is much higher than the F ST between the Pl population and a Swedish Stubbsand population (0.2) based on almost 80 different loci (Ross-Ibarra et al., 2008). The 5.5-fold difference in the level of neutral sequence variation in Sp compared with Pl greatly exceeded the twofold reduction detected at neutral microsatellite loci (Muller et al., 2008). Moreover, in the promoter region, exon 1 and part of intron 1, the populations diverged more compared with other regions of the FLC1 gene. Further, these regions were also highly diverged compared with the corresponding regions in A. thaliana (i.e. they were impossible to align), which can be interpreted as a high mutation rate at these regions. Thus, low variation in the promoter region and in the first intron in Sp is in contrast to divergence. Hence, it is possible that these expression-regulating regions of FLC1 have been under directional selection in recent history. Previous studies in A. thaliana have identified regions important for FLC expression in nonvernalized plants, particularly in the promoter. Moreover, regions affecting both the FLC repression during cold periods and the maintenance of the repression at normal temperatures have been located to both the promoter and the first intron (Sheldon et al., 2002). For instance, a protein complex including FRI binds to the FLC promoter and leads to an active chromatin state and high expression (Choi et al., 2011). Furthermore, Choi et al. (2011) identified a 15-bp binding site sequence for SUF4, one of the FRI complex components, at 363 to c. 331 from the transcription start site of FLC. A similar motif (differing by two substitutions) was identified in the promoter sequence of FLC1 from the Pl population. This putative motif is located in the region where the Sp FLC1 sequence has a 350-bp deletion and thus lacks the motif. However, as the motif in A. thaliana confers high expression, the expectation would be that the Sp deletion would lead to lower FLC1 expression, which is not supported by our data. Clearly, functional studies of the promoter and intron sequence variation observed in FLC1 from Sp and Pl are needed to clarify its effect on FLC1 regulation. Population genomic analysis of these regions can also provide clues as to the importance of the different loci. Acknowledgements We thank Jarkko Vehkaoja for his help in the glasshouse, Johanna Leppälä for her contribution to the marker development and map construction, the undergraduate students in the Savolainen laboratory and lab technicians both in the Savolainen and in the Lagercrantz laboratories for their help with the lab work, Tanja Pyhäjärvi for her contribution to the sequence analysis and Esa Aalto, Tiina Mattila and Antti Virtanen for their advice on CodonCodeAligner. Päivi Leinonen and other members of the Plant Genetics Research Group and the anonymous referees are acknowledged for their valuable comments on the manuscript. The research was supported by the Finnish Population Genetics Graduate School, Biocenter Oulu, ERA-net Plant Genomics, Academy of Finland and the Swedish Research Council. Oulun Luonnonystävät is acknowledged for a travel grant for U.K. References Aikawa S, Kobayashi MJ, Satake A, Shimizu KK, Kudoh H Robust control of the seasonal expression of the Arabidopsis FLC gene in a fluctuating environment. Proceedings of the National Academy of Sciences, USA 107: Ausin I, Alonso-Blanco C, Martínez-Zapater J Environmental regulation of flowering. International Journal of Developmental Biology 49: Bastow R, Mylne J, Lister C, Lippman Z, Martienssen R, Dean C Vernalization requires epigenetic silencing of FLC by histone methylation. Nature 427: Bernier G, Perilleux C A physiological overview of the genetics of flowering time control. Plant Biotechnology Journal 3: Brachi B, Faure N, Horton M, Flahauw E, Vazquez A, Nordborg M, Bergelson J, Cuguen J, Roux F Linkage and association mapping of Arabidopsis thaliana flowering time in nature. Plos Genetics 6: e Choi K, Kim J, Hwang H, Kim S, Park C, Kim SY, Lee I The FRIGIDA complex activates transcription of FLC, a strong flowering repressor in Arabidopsis, by recruiting chromatin modification factors. Plant Cell 23: Clauss MJ, Mitchell-Olds T Population genetic structure of Arabidopsis lyrata in Europe. Molecular Ecology 15: D Aloia M, Tocquin P, Périlleux C Vernalization-induced repression of FLOWERING LOCUS C stimulates flowering in Sinapis alba and enhances plant responsiveness to photoperiod. New Phytologist 178: Deng W, Ying H, Helliwell CA, Taylor JM, Peacock WJ, Dennis ES FLOWERING LOCUS C (FLC) regulates development pathways throughout the life cycle of Arabidopsis. Proceedings of the National Academy of Sciences, USA 108: