PEP1 of Arabis alpina Is Encoded by Two Overlapping Genes That Contribute to Natural Genetic Variation in Perennial Flowering

Higher plants exhibit a variety of different life histories. Annual plants live for less than a year and after flowering produce seeds and senesce. By contrast perennials live for many years, dividing their life cycle into episodes of vegetative growth and flowering. Environmental cues control key check points in both life histories. Genes controlling responses to these cues exhibit natural genetic variation that has been studied most in short-lived annuals. We characterize natural genetic variation conferring differences in the perennial life cycle of Arabis alpina. Previously the accession Pajares was shown to flower after prolonged exposure to cold (vernalization) and only for a limited period before returning to vegetative growth. We describe five accessions of A. alpina that do not require vernalization to flower and flower continuously. Genetic complementation showed that these accessions carry mutant alleles at PERPETUAL FLOWERING 1 (PEP1), which encodes a MADS box transcription factor orthologous to FLOWERING LOCUS C in the annual Arabidopsis thaliana. Each accession carries a different mutation at PEP1, suggesting that such variation has arisen independently many times. Characterization of these alleles demonstrated that in most accessions, including Pajares, the PEP1 locus contains a tandem arrangement of a full length and a partial PEP1 copy, which give rise to two full-length transcripts that are differentially expressed. This complexity contrasts with the single gene present in A. thaliana and might contribute to the more complex expression pattern of PEP1 that is associated with the perennial life-cycle. Our work demonstrates that natural accessions of A. alpina exhibit distinct life histories conferred by differences in PEP1 activity, and that continuous flowering forms have arisen multiple times by inactivation of the floral repressor PEP1. Similar phenotypic variation is found in other herbaceous perennial species, and our results provide a paradigm for how characteristic perennial phenotypes might arise.


Introduction
Key stages in the plant life cycle are regulated in response to environmental cues.Many genes controlling these responses show allelic variation within species allowing adaptation of individuals to their environment.Short-lived annual plants exhibit natural genetic variation for seasonal flowering responses that has been studied extensively in Arabidopsis thaliana [1,2] and several crops [3].The transition from vegetative to reproductive development in annuals marks the end of their life cycle.In contrast, perennials live for many years and undergo repeated cycles of vegetative growth and flowering.Most temperate perennial species flower seasonally and the phases of flowering and vegetative growth are clearly separated.Interestingly, in perennials the environmental cues that regulate floral initiation also have an impact on the duration of the flowering episode either by regulating the growth of flower buds [4] or the return to vegetative development [5,6,7].Genetic variation also exists for these traits so that individuals of the same species flower either for a restricted period or for a more prolonged time [4,8,9].In addition, differences in activity of flowering genes can contribute to the divergence in life history between plant species, including the distinction between annual and perennial life cycles [5,10,11,12,13,14,15].Analysis of species within the Brassicaceae family provides an opportunity to extend the intensive knowledge of mechanisms of flowering-time control in the model annual species A. thaliana to closely related perennial species.
The evolution of adaptive traits that contribute to the annual or perennial life strategies can be relatively rapid and in a small number of examples has been associated with genetic alterations such as inversions, gene duplications or modifications in gene expression patterns [5,10,11,14,16,17,18,19].We showed previously that the MADS box transcription factor PEP1 regulates flowering and perennial specific traits in the Brassicaceae species A. alpina [5].So far, flowering and the perennial growth habit of A. alpina have been characterized only in the accession Pajares (Paj), which was collected in the Cardillera Canta ´brica mountain region of Spain [5].This accession exhibits an obligate requirement for prolonged exposure to cold (vernalization) in order to flower.PEP1 represses flowering before vernalization, so that pep1-1 mutants identified after mutagenesis flower rapidly without vernalization [5].In addition, pep1-1 mutant plants flower for a longer period than wild-type plants linking vernalization requirement to the duration of a flowering episode.
PEP1 is the orthologue of A. thaliana FLOWERING LOCUS C (FLC), which also encodes a floral repressor that delays flowering prior to vernalization.In A. thaliana, FLC delays flowering by repressing transcription of genes necessary for the switch to reproductive development [20,21,22].This block on flowering is overcome by vernalization, which reduces FLC expression allowing flowering to proceed [23,24].FLC mRNA levels in A. thaliana are stably repressed after vernalization ensuring that all lateral branches that are formed after floral initiation also flower and produce seeds.A. thaliana plants then die, completing the annual life cycle.Stable FLC repression after flowering is therefore important in ensuring that all branches undergo flowering, maximizing seed production within the single year of the life cycle.The stable repression of FLC is associated with epigenetic changes at the locus, which involve trimethylation on lysine 27 of histone 3 (H3K27me3) at the FLC gene.The H3K27me3 mark spreads after vernalization at the FLC locus [25,26] stably repressing its transcription.
Most natural genetic variation in A. thaliana for flowering time in response to vernalization is conferred by allelic variation at FLC or its upstream regulator FRIGIDA (FRI) [27,28,29,30,31,32,33].FRI promotes FLC transcription, so that loss of FRI activity results in low FLC mRNA levels [23,24].Early flowering A. thaliana accessions that flower without vernalization are summer annuals.Most of these show low FLC mRNA levels either because they carry lesions in FRI or mutations at the FLC locus that prevent its upregulation by FRI [27,28,29,31,34,35].Early-flowering accessions with high levels of FLC mRNA have also been reported and carry mutations at the FLC locus [28,36].By contrast, all winter-annual accessions, which are late flowering and exhibit a strong vernalization requirement, express FLC mRNA at high levels [23,24,27,29,30].Natural genetic variation among winter annual A. thaliana accessions has been described for the extent of vernalization period required to promote flowering and is also associated with allelic variation at FLC [37,38].
In A. alpina, PEP1 has a similar role to FLC conferring a response to vernalization, but in addition it contributes to perennial flowering traits not found in A. thaliana.PEP1 mRNA levels in A. alpina are not stably repressed by vernalization, even after a saturating vernalization period that initiates flowering.Rather, PEP1 mRNA levels rise again when plants experience warm temperatures after winter.Similar expression patterns of FLC orthologues have been described in another perennial Brassicaceae species, Arabidopsis halleri [19].The rise in PEP1 mRNA levels after vernalization blocks flowering of any shoots that have not already flowered, causing the plant to return to vegetative growth.The accumulation of trimethylation of lysine 27 of histone 3 (H3K27me3) increases at PEP1 locus during vernalization and is associated with repression of transcription [25,26].However, contrary to FLC in A. thaliana, the H3K27me3 mark is not maintained after cold and this correlates with reactivation of PEP1 transcription [5,25,26].
Here we identify A. alpina accessions that flower without vernalization and show that these are also naturally occurring perpetual flowering accessions.Five of these accessions were characterized at the genetic and molecular levels and in each case the natural phenotypic variation is caused by loss of function alleles at PEP1.These five accessions carry different lesions at the PEP1 gene suggesting that they have arisen independently during evolution.Analysis of this variation also demonstrated that A. alpina contains a segmental partial duplication of PEP1 that created two transcriptional starts sites and two overlapping transcripts, a more complex structure than is found at FLC in A. thaliana.We discuss this intra-and inter-species variation of PEP1 structure and function in terms of evolution of life history traits in the Brassicaceae.

PEP1 contributes to variation in flowering behavior among A. alpina accessions
The accession Paj carries an active allele of PEP1 and has an obligate vernalization requirement to flower (Figure 1A; [5]).EMS-induced mutations in PEP1 abolish the obligate vernalization requirement causing Paj to flower within 60 days in long days [5].To test whether allelic differences at PEP1 contribute to natural variation for perennial flowering traits, the flowering times of 24 accessions (Table S1) from diverse locations were assessed in long days.Five accessions that flowered without vernalization and within 3 months of germination were identified as candidates for carrying natural pep1 mutant alleles.Four of these accessions, Dorfertal (Dor), Totes Gebirge (Tot), West Carpathians (Wca) and Czarna Go `ra (Cza), flowered even earlier than the pep1-1 mutant, whereas the accession Muggendorf (Mug) flowered slightly later than pep1-1 (Figure 1B-1G, Figure S1).The pep1-1 mutant plants flower perpetually indicating that PEP1 also regulates the duration of flowering and return to vegetative development (Figure 1H; [5]).To test whether flowering without vernalization in the A. alpina accessions also correlates with perpetual flowering, the accessions Dor, Tot, Cza and Wca were grown under long days and their duration of flowering was compared to the pep1-1 mutant.All accessions were still flowering 28 weeks after flower initiation, demonstrating that flowering without vernalization correlates

Author Summary
Perennial plants live for many years and cycle between flowering and vegetative growth.These stages of the life cycle are often initiated by environmental conditions and occur seasonally.However, many herbaceous perennial species such as strawberry, rose, or Arabis alpina contain varieties that flower continuously irrespective of the seasons.Here we characterize this genetic variation in A. alpina and show that five continuously flowering accessions carry independent mutations in the PERPETUAL FLOWERING 1 (PEP1) gene.These mutations impair the activity of the PEP1 floral repressor causing the plants to flower without requirement for winter cold and to flower continuously.This result has interesting parallels with strawberry and rose, where inactivation of a different floral repressor controlling response to day length gives rise to naturally occurring perpetual flowering forms.We also show that PEP1 in A. alpina has a complex duplicated structure that gives rise to two overlapping transcripts.This arrangement differs from the simple structure of PEP1 orthologues in related annual species, such as FLC of Arabidopsis thaliana, suggesting that duplication of PEP1 might contribute to the complex transcriptional patterns associated with PEP1 function in perennials.Our work provides insight into genetic variation contributing to the perennial life history of plants.
with the perpetual flowering trait in natural A. alpina accessions (Figure 1H-1J).
In A. thaliana FLC mRNA levels often correlate with flowering time [23].Therefore we compared the PEP1 mRNA levels of the early-flowering accessions that do not require vernalization to flower with those of the accession Paj that has an obligate vernalization requirement.Most accessions exhibited PEP1 mRNA levels at least as high as those found in the accession Paj suggesting that flowering without vernalization did not obviously correlate with reduced PEP1 mRNA (Figure 2A).However, the accessions Dor and Wca had lower PEP1 mRNA levels than Paj.
Therefore reduced PEP1 mRNA levels in these accessions might explain their flowering phenotypes, but this cannot be the basis of early flowering in the accessions Mug, Cza and Tot.
These non-vernalization requiring accessions were then crossed to pep1-1 to test whether they carry a PEP1 allele that can complement the pep1-1 mutation.As control, these accessions were also crossed to Paj and the flowering times of F1 plants were measured (Figure 2B, Figure S2).All F1 plants derived from the crosses to the pep1-1 mutant flowered without vernalization, suggesting that these accessions carry non-functional PEP1 alleles that cannot complement the pep1-1 mutation (Figure 2B).By To test whether they carry mutations in the PEP1 open reading frame that might account for the non-functional PEP1 alleles we sequenced PEP1 cDNAs from each accession (Table 1, Table S2).Primers that anneal to the -59 and -39 UTRs were used to amplify 814 bp of coding sequence from apices prior to vernalization.Amplification products from each accession were then cloned in Escherichia coli and several clones were sequenced.PEP1 mRNA from Paj was shown to be alternatively spliced generating forms that retained introns or that had lost parts of or whole exons [5].Most of the sequenced clones from the accessions contained the full length ORF suggesting that this is the predominant transcript although splicing variants were also detected (Table 1, Table S2).The full-length cDNAs isolated from the Dor, Tot, Cza and Mug accessions contained polymorphisms that caused amino acid substitutions or deletions compared to the Paj allele (Table 1), suggesting that these polymorphisms might be the basis of loss of PEP1 activity and therefore the altered flowering behavior.In the accession Wca the predominant cDNA sequence did not contain any amino acid substitutions or deletions when compared to Paj and thus the reduced PEP1 activity is probably due to the low mRNA level described above.
Apart from the splicing variants the polymorphisms found within the PEP1 cDNA sequences were different for each of the non-vernalization requiring accessions (Table 1).From the Dor accession cDNAs containing polymorphisms in exon 1 were amplified.Most cDNAs contained a G to A base substitution in exon 1 that is predicted to cause an amino acid substitution in the highly conserved MADS box domain responsible for DNA binding whereas some sequences were identical to that of Paj (Figure S3).The origin of these different cDNA sequences is discussed below.The PEP1 cDNA isolated from accession Tot contained two base substitutions in exon 4, which encodes the K box, and one was predicted to cause a valine (Val) to isoleucine (Ile) amino acid substitution compared to Paj.The accession Cza contained a base substitution and a 3 bp deletion in exon 3. Compared to Paj the base substitution is predicted to cause a leucine (Leu) to arginine (Arg) change, whilst the deletion is expected to cause loss of glycine (Gly) and tyrosine (Tyr) residues and their replacement by a cysteine (Cys) amino acid.Finally, the accession Mug contained a synonymous base substitution in exon 1 and a 27 bp deletion in exon 3, which resulted in a predicted loss of 9 amino acids in the K box domain of the protein.
To test if PEP1 protein accumulation correlated with flowering behavior in the accessions an antibody was raised against the protein (Methods).PEP1 protein was then tested using Western blots of apical samples of the non-vernalization requiring accessions, Paj and pep1-1 before vernalization, and vernalized Paj plants (Figure 2C).PEP1 protein was detected in Paj before vernalization but not after 12 weeks vernalization and was absent from the pep1-1 mutant.In the accession Wca PEP1 was not detected, which correlates with low PEP1 mRNA levels compared to the accession Paj (Figure 2A, 2C).PEP1 protein in the accession Mug was smaller, consistent with the 27 bp in frame deletion in exon 3 in this accession (Figure 2C, Table 1).The accumulation of PEP1 in the accessions Dor and Tot was similar to the accession Paj before vernalization, although as described above the forms of PEP1 in Dor and Tot are predicted to contain amino acid changes compared to Paj.
Taken together these data indicate that some non-vernalization requiring accessions carry PEP1 alleles with polymorphisms in the protein coding sequence that likely impair protein function, whilst Wca shows reduced PEP1 accumulation.The genetic and molecular analyses of these accessions suggest that these natural alleles contribute to loss of vernalization requirement and perpetual flowering by reducing PEP1 protein level or activity.
Sequence variation in PEP1 cDNAs of the Dor accession reveals the complex structure of the PEP1 locus The PEP1 cDNAs isolated from the accession Dor varied in the sequence of exon 1, being either identical to Paj or carrying a nonsynonymous change (Table 1, Figure 3A).These different cDNAs could not be explained by Dor being heterozygous at PEP1, because genetic experiments demonstrated that it carries the same allele on both chromosomes.Most Dor PEP1 cDNAs contained the G to A base substitution compared to Paj at the beginning of the first exon (Table 1).The other set of cDNAs encoding an ORF identical to the accession Paj also contained a 248 bp insertion in the 59 UTR region, which was absent from Paj cDNA and absent from the predominant transcript carrying the G to A substitution (Table 1).The presence of both PEP1 cDNAs in Dor was verified using a different primer in the 59 UTR and the same reverse 39 UTR primer.Again both types of cDNA were detected, but surprisingly their ratio was altered so that most clones contained sequences similar to accession Paj and only one contained the G to A base substitution (Figure 3E).
The PEP1 locus in the accession Paj is partially duplicated and is located in a region of the A. alpina genome that shows conserved synteny with A. thaliana on chromosome 5 where FLC is located.In A. alpina Paj the distance between the orthologues of the genes that flank FLC in A. thaliana (At5g10130 and At5g10150) is increased [5], and this region includes several tandemly duplicated sequences that include parts of PEP1 (Figure 3B).The longest duplicated segment is approximately 2 kb and contains a copy of exon 1 of PEP1, as well as parts of the PEP1 promoter and intron 1.The tandem copies of exon 1, named 1a and 1b are located 6.3 kb apart (Figure 3B).To determine whether the two types of PEP1 cDNA identified in the Dor accession are encoded by these different copies of exon 1, the PEP1 locus was sequenced in the accession Dor using a BAC containing the entire PEP1 locus.The contiguous region from exon 1b to exon 7 showed a similar structure to PEP1 from Paj.Furthermore, exon 1b showed the same sequence as Paj but contained the 248 bp insertion in the 59 UTR found in some of the Dor cDNAs described above (Figure 3A, Table 1).Therefore those Dor PEP1 cDNAs encoding the same ORF as Paj and the insertion in the 59 UTR were encoded by the contiguous region from exon 1b to exon 7, but the cDNAs encoding the non-synonymous polymorphism in exon 1 were not derived from this region.A 10 kb region upstream of PEP1 exon 1b in accession Dor was then amplified, cloned and sequenced.A second copy of PEP1 exon 1, which corresponded to exon 1a in Paj, was present in this upstream region, and this contained the G to A base substitution in the first exon which was detected in the second group of PEP1 cDNAs found in the Dor accession (Figure 3C).This result indicated that exon 1a is also used to produce a full length PEP1 transcript and both exon 1a and exon 1b are spliced to the same copy of exon 2 to produce two transcripts that share exons 2-7 (Figure 3D).Which transcript is amplified preferentially from Dor accession depends on the primer in the 59 UTR used which is probably influenced by the insertion in the 59 UTR of the exon 1b (Figure 3E).

The PEP1 locus contains tandem duplications in other A. alpina accessions
The duplicated regions flanking PEP1 exons 1a and 1b in the accession Paj (grey to blue boxes in Figure 3B) are highly similar except for a 246 bp sequence after exon 1a that has been deleted from exon 1b region (yellow rectangle in Figure 3B, Figure 4A and  4B).This 246 bp deletion after exon 1b also exists in the accession Dor, indicating that it occurred before the divergence of the accessions Paj and Dor (Figure 3C).Moreover, this 246 bp sequence in Paj PEP1a shows homology with the first intron of FLC in A. thaliana, suggesting that in this respect PEP1 exon 1a is more related to FLC than is PEP1 exon 1b (Figure 4B).
To test if other A. alpina accessions also contained the tandem duplication at the PEP1 locus specific primers were designed for PEP1a and PEP1b copies annealing to the 246 bp segment specific for exon 1a or to a conserved sequence in the first intron after exon 1b (Figure 4C, Figure 4E).These primers were used in combination with a common forward primer in the MADS box region that annealed to both PEP1a and PEP1b (Figure 4A, Figure 4C, Figure 4E).The PEP1b specific primers amplified a fragment in all accessions, indicating that PEP1b is likely conserved among these accessions (Figure 4F).However, these fragments varied slightly in size, consistent with the deletion present in the accession Dor (described below) compared to Paj.By contrast the PEP1a specific primers amplified a fragment in the accessions Dor, Cza and Paj, but not in Tot, Wca or Mug (Figure 4D).This result indicates that the PEP1b region is more conserved than PEP1a among A. alpina accessions.
The genomic PEP1 loci from the accessions Cza and Tot were then sequenced to analyze their structure in detail.The PEP1 allele in the accession Cza exhibited a similar structure to the accessions Dor and Paj, containing both exon 1a and exon 1b (Figure 4G).No fragment corresponding to exon 1a was amplified from accession Tot.To provide genomic information for this region, Illumina sequencing was performed on DNA extracted from Tot.The sequence reads were assembled and searched for homology to PEP1 using BLAST.This sequence information was then used to design locus specific primers to test the structure of PEP1 in Tot and to provide fragments for Sanger sequencing.This analysis demonstrated that the Tot allele contained only exon 1b, and had suffered a deletion that includes exon 1a and the intergenic region between exon1a and 1b.Therefore loss of PEP1 function in this allele correlates with absence of the PEP1a region as well as the non-synonymous substitution in exon 4.

PEP1a and PEP1b transcripts are unstably and differentially expressed after vernalization in the accession Dor
The sequences upstream of PEP1 exon 1a and exon 1b are highly diverged (Figure 3B), suggesting that the two PEP1 genes might be expressed from different promoters.To test whether the PEP1 transcripts encoding exon 1a and exon 1b use two different transcription start sites 59 RACE was performed on the accession Dor in which the two copies of exon 1 can be differentiated.This analysis showed that exon 1a and exon 1b have distinct transcription start sites located 65-186 bp upstream of the ATG for exon 1a and 61-190 bp upstream of the ATG of exon 1b (Figure 5A, 5B).
To determine whether these different transcripts are differentially expressed 59 RACE was performed in the accession Dor in apices and leaves before and after vernalization (Figure 5C).PEP1a polymorphism in the accession Dor was mainly found in transcripts from vegetative apices before vernalization representing 26% of the total PEP1 mRNA pool.In leaves before and after vernalization Dor PEP1a was detected only in 3% of the transcripts representing only a small proportion of the total PEP1 transcripts whereas was virtually undetectable in flowers five weeks after vernalization.By contrast Dor PEP1b was strongly expressed in apices, flowers and leaves in all stages.
To compare PEP1a and PEP1b transcript levels more thoroughly after vernalization, their expression patterns were followed through a time course in 3 week old Dor seedlings vernalized for 12 weeks and then returned to normal temperatures for an additional 1, 3 or 5 weeks.Both transcripts were repressed during vernalization and increased in expression after vernalization (Figure 5D, 5E, 5F).Dor PEP1b mRNA stayed at high levels for several weeks after vernalization, whereas Dor PEP1a mRNA levels fell again after 1 week in warm conditions.
These experiments indicate that the PEP1a and PEP1b genes are independently transcribed from different transcription start sites.Both transcripts show very similar patterns of transcriptional regulation before, during and immediately after vernalization, but PEP1b expression persists more strongly in flowers and leaves later after vernalization than PEP1a transcript.
the G to A substitution in the exon 1.A few clones that did not contain the G to A base substitution also contained a 248 bp insertion in the 59 UTR.When primer PEP1_5UTRF2 was used most clones did not contain G to A base substitution.doi:10.1371/journal.pgen.1003130.g0036) and water control (7).(E) Alignment using part of the

Discussion
A. alpina is an alpine perennial and the only previously characterized accession, Paj, flowers for a limited period in the spring after prolonged exposure to winter chilling temperatures.Flowering and perennial specific traits were studied in detail in this Spanish accession and shown to be under the control of PEP1, which is the orthologue of the A. thaliana gene FLC [5].Here, we identified several A. alpina accessions that carry inactive PEP1 alleles and hence do not require vernalization to flower.These accessions also flower perpetually similar to pep1-1 mutant plants [5].Therefore, among A. alpina accessions allelic variation at PEP1 contributes to phenotypic differences in the duration of flowering as well as to differences in vernalization response.This is in contrast to loss of FLC function in A. thaliana accessions which is not associated with perpetual flowering due to its annual, monocarpic life cycle.The inactive alleles of PEP1 are all different indicating that this phenotypic divergence has occurred many times in A. alpina.We found that the PEP1 locus is partially duplicated including two copies of the first exon as well as parts of the proximal promoter and intron 1.These sequences are arranged as a tandem duplication creating two transcriptional start sites that generate two overlapping transcripts that are differentially expressed.This increased complexity at the PEP1 locus might contribute to the contrasting transcriptional regulation of PEP1 and FLC observed in perennial A. alpina and annual A. thaliana respectively.

Intra-specific variation for seasonal flowering in herbaceous perennials
In A. alpina the correlation between perpetual flowering and the lack of vernalization requirement for floral induction was demonstrated by characterizing pep1-1 which is an EMS induced mutant derived from mutagenesis of the Paj accession.Here we show that natural A. alpina accessions which do not require vernalization to initiate flowering also flower perpetually.This suggests that in nature A. alpina exhibits two life histories, a vernalization-requiring form that flowers for a limited period and a non-vernalization-requiring form that flowers continuously.Similar variation in the duration of flowering season also occurs in other herbaceous perennials such as strawberry and rose [8,9,39].Cultivated strawberries and their wild relatives exhibit two distinct flowering habits and are classified either as seasonal-flowering (june bearing) producing one harvest per year or as perpetualflowering (ever bearing) that produce two harvests per year [8,39].Interestingly, perpetual-flowering types in these Rosaceae species show different environmental requirements for flower induction compared to seasonal-flowering ones.Strawberries that follow the seasonal-flowering habit require short days and low temperatures to induce flowering, whereas perpetual-flowering strawberries have lost photoperiodic and temperature control of flowering or become long-day plants [7,40,41].These types differ from A. alpina where accessions that flower perpetually because they lost vernalization requirement to flower.Interestingly, continuous flowering is also associated with subalpine forests compared to lowland forests [42] and the wild strawberry (Fragaria vesca) which follows the perpetual flowering habit is also called alpine strawberry because it is thought to have originated in the alps [43].
The duration of flowering season in F. vesca is controlled by a single gene, the SEASONAL FLOWERING LOCUS (SFL) [44].Recently, sequence variation in the rose and F. vesca homologues of the A. thaliana TERMINAL FLOWER 1 (TFL1) gene was shown to correlate with differences in the duration of flowering observed in seasonal and perpetual flowering types [6,8,45].F. vesca accessions that flower perpetually carry a deletion in F. vesca TFL1 (FvTFL1) that causes a premature stop codon whereas perpetual flowering in rose correlates with the presence of a retrotransposon in the second intron of the rose TFL1 homologue (KSN) [6,45].TFL1 encodes a protein related to phosphatidyl ethanolamine binding proteins and is a floral repressor that is specifically expressed at the shoot apical meristem [46,47,48,49].Transgenic experiments in F. vesca demonstrated that FvTFL1 does not only regulate the duration of the flowering episode but also the induction of flowering in response to photoperiod, with short days in the autumn downregulating FvTFL1 mRNA levels to initiate flowering [6].In A. alpina reduction of TFL1 function does not affect the duration of flowering season but rather causes plants to flower in response to vernalization earlier in their life cycle [10].Thus although it does not define the duration of a flowering episode as it does in Rosaceae species, TFL1 retains an important function in the perennial cycle of A. alpina.By contrast our work described here demonstrates that natural phenotypic variation between seasonal and perpetual forms of A. alpina is caused by loss of function alleles of PEP1.Therefore, genetic variation at genes encoding distinct classes of floral repressors appears to be responsible for differences in the duration of the flowering season among accessions of perennial species from different plant families.Recruitment of different flowering repressors to this function presumably explains how seasonal flowering phenotypes similar to those observed in A. alpina Pajares can occur in perennial species that do not contain FLC homologues such as PEP1.
Allelism tests indicated that all of the analyzed A. alpina accessions that flower perpetually and without vernalization carry inactive PEP1 alleles.The perpetual flowering accessions carry independent mutations at PEP1 that either cause amino acid substitutions, alter the structure of the locus or prevent protein accumulation and demonstrate that PEP1 activity was lost independently multiple times.In A. thaliana most natural phenotypic variation for flowering in response to vernalization can be explained by allelic variation at FRI [28,29,31,34,36].Most summer annual accessions that flower without vernalization carry lesions at FRI and therefore flower early because they do not accumulate FLC mRNA to high levels [27,29,31].Similarly, in Arabidopsis lyrata two alleles at the FRI orthologue were identified and were suggested to be associated with flowering time differences, although they did not appear to explain large differences in flowering time between populations [50].FRIindependent variation in vernalization requirement also occurs among A. thaliana summer annual accessions and largely correlates with allelic variation at FLC.However, loss of function FLC alleles occur to a lesser extent than those in FRI [28,35,37,38].Some early-flowering summer annual accessions carry active FRI alleles but still show reduced FLC mRNA levels, which in some cases is associated with insertions in the first intron of FLC [34,35].Other accessions that express FLC mRNA at high levels carry mutations within the FLC protein coding sequence impairing its activity [28,31,36].Our results provide no evidence for natural variation at A. alpina FRI playing a role in perennial flowering or vernalization requirement.Instead all early flowering A. alpina accessions tested carried mutations at the FLC orthologue PEP1, suggesting that natural variation at PEP1 is the major source of variation in flowering response to vernalization and seasonal perennial flowering.Even in the accession Wca where PEP1 accumulation was low, consistent with reduced FRI activity, genetic complementation analysis demonstrated that the early flowering of this accession was due to a mutation at PEP1, similar to A. thaliana accessions that carry weak FLC alleles [34,35].
In agreement with our data, QTLs affecting flowering time in Brassica crop species co-segregate with regions containing FLC homologues [51,52,53,54,55,56], although there is also evidence for variation at FRI [57].Studies on selected paralogues in Brassica.rapa, Brassica napus and Brassica oleraceae support the idea that some FLC genes in these species might function in a similar way to FLC in A. thaliana and introduction of some of them, such as BrFLC1-3 and BnFLC1-5 into A. thaliana delayed flowering [58,59].Moreover, mRNA levels of some FLC paralogues correlate with flowering and vernalization requirement [56,59].
Natural loss of function alleles of PEP1 have arisen independently in several of the A. alpina populations analyzed.This might indicate that flowering without vernalization and/or perpetual flowering provide a selective advantage in some environments, perhaps because they allow production of more seeds over a longer growing period.Alternatively, PEP1 loss of function alleles might provide no selective advantage but occur as neutral variation when selection on PEP1 is relaxed at certain altitudes or in particular habitats.Understanding the evolutionary forces determining the prevalence of active and inactive PEP1 alleles will require a future more detailed ecological analysis.

Structure of the PEP1 locus
PEP1 in A. alpina has an unexpectedly complex structure including a partial tandem duplication giving rise to two overlapping mRNAs.The duplicated tandem segments include the first exon and adjacent sequences.Both copies of exon 1 (exon 1a and exon 1b) are spliced to the same copy of exon 2 and in the accession Paj are predicted to encode identical proteins.Alternative splicing of the single FLC copy in A. thaliana has also been described [60].Most A. alpina accessions that carry inactive PEP1 alleles have mutations in exons 2-7 that contribute to both proteins, and therefore the mutations inactivate both forms of the gene.Related overlapping plant gene configurations have been reported to form tissue specific isoforms of proteins based on the use of alternative promoters and/or alternative splicing of exons [61,62,63,64,65].Studies in tropomyosins, for example, suggested that intragenic duplication events generated new exons that are combined in different ways utilizing alternative promoters, translational initiation sites or polyadenylation sites to produce tissue specific protein isoforms [62,63,66,67,68].Nevertheless, in the accession Paj the two overlapping PEP1 genes are predicted to encode identical proteins and therefore this structure has not evolved to encode different protein isoforms.However, the sequences upstream of each copy of exon 1 share little homology suggesting that the two overlapping copies are expressed from different promoters, and consistent with this interpretation they have independent transcription start sites.Furthermore, the expression patterns of the two PEP1 transcripts are different suggesting that this complex structure might confer PEP1 activity in a broader range of tissues or developmental stages.In A. thaliana FLC functions to regulate flowering in leaves and apices [20] and other MADS box transcription factors have been shown to contribute to different transcriptional complexes at various stages in development [69].A broader expression pattern of PEP1 might allow it to contribute to a wider range of regulatory pathways.
Tandem duplications of FLC paralogues are found in A. thaliana and have been described for FLC orthologues in other members of the Arabidopsis genus [70].A cluster of four genes encoding MADSbox transcription factors related to FLC occurs in A. thaliana.These genes, MADS AFFECTING FLOWERING 2 (MAF2) to MAF5, delay flowering time but to a markedly lesser extent than FLC [71].The tandem duplication of these four genes spans around 24 kb and allelic variation occurs among A. thaliana accessions, including a fusion of MAF2 and MAF3 to generate a chimeric gene [72].The partial gene duplication of PEP1 found in A. alpina might have arisen from such a rearrangement within an ancestral structure that includes a tandem array of full-length genes.Interestingly in A. lyrata and Arabidopsis arenosa tandem arrays of full-length FLC orthologues have been described in detail [70] and in B. oleracea a tandem duplication of BoFLC1 was reported [73], suggesting that the single copy of FLC present in A. thaliana might be derived during evolution from an ancestral FLC locus that contained two or three tandem copies.A. lyrata harbours two full-length FLC genes whereas A. arenosa contains two copies and a partial copy.These were proposed to have arisen within the Arabidopsis genus and therefore would represent an independent duplication event to the one we describe in A. alpina.Nevertheless, as A. arenosa, A. lyrata and A. alpina are all perennials, and PEP1/FLC shows a more complex expression pattern in perennials than annuals, tandem duplication of FLC may be one mechanism that contributes to the complex transcriptional patterns associated with the perennial cycle.Gene copy number expansion by tandem gene duplication and functional divergence mediated by the accumulation of mutations in cis regulatory regions has been proposed to contribute to evolution of complex traits in both plants and mammals [74,75,76].In the Brassicacae acquisition of metal hyperaccumulation in A. halleri compared to A. thaliana was attributed to the amplification of HEAVY METAL ATPASE 4 (HMA4) [77].In this case, gene amplification increases the expression level of the metal pump encoded by HMA4 due both to the higher copy number of the gene and to cis-acting changes in promoter sequences.Conceivably the increased copy number of PEP1 may also be associated with an increased expression level and this might be partially responsible for the stronger repression of flowering found in obligate vernalization requiring A. alpina accessions such as Pajares.

Plant materials and growth conditions
A. alpina accessions screened are listed in Table S1.The accession Pajares (Paj) was originally collected in the Cordillera Canta ´brica mountain in Spain, selfed for six generations by single seed descent and characterized in [5].The pep1-1 mutant was previously characterized in [5].The accessions Dor, Tot and Wca were selfed by single seed descent for six generations.The accession Cza was selfed for two generations and accession Mug was selfed once.F1s for allelism tests with pep1-1 and Paj were generated using pep1-1 and Paj as mother plants.
Flowering time was measured when first flower opened and demonstrated as number of days to flower (from the time seeds were put on soil) or number of leaves at flowering.For all flowering time experiments plants were grown in long days cabinets (16 hours light at 20uC and 8 hours dark at 18uC).Duration of flowering season was measured in long day (16 hours light) controlled environment glasshouse.

Gene expression analysis
For gene expression studies RNA was extracted from apices and leaves using the RNeasy Plant Mini kit (Qiagen) and purified the DNA-free Kit (Applied Biosystems).For expression studies, first strand cDNA was synthesized from 1 mg RNA using different primers depending on the aim of the experiment (Table 2).59 RACE was performed using the Invitrogen kit on 2.5 mg RNA according to manufacturer's instructions.PCR products after 59 RACE were cloned in a pGEM-T easy vector and several clones were sequenced using M13 primers.
shown in the ''Clones'' column.PEP1 in the accession Paj is also differentially spliced but splicing forms are rare compared to the canonical Pajares ORF [5].The full-length PEP1 cDNA sequence of the vernalization-requiring accession Paj is used as a reference (row highlighted in grey).Nucleotide polymorphisms compared to Paj PEP1 cDNA sequence obtained for each accession are presented.Nucleotide (nucl.)position and aminoacid (a.a.) changes compared to Paj are mentioned in rows above the grey row.* indicates sequences containing a stop codon.(PDF)

Figure 1 .
Figure 1.Flowering behavior of A. alpina accessions.(A) Accession Paj grown vegetatively for four years in long day glasshouse.Paj has an obligate requirement for vernalization to flower.(B)-(I) Non-vernalization requiring A. alpina accessions at flowering under long days.Accession Dor (B), Tot (C), Wca (D), Cza (E) and Mug (F).(G) Flowering times of non-vernalization requiring A. alpina accessions under long days (16 hours light) compared to pep1-1 mutant and the accession Paj.Flowering time is measured as days to flower (DTF).pep1-1 mutant (H) and the accession Dor (I) flower perpetually after 6 months in long days.(J) Duration of flowering in non-vernalization requiring A. alpina accessions.doi:10.1371/journal.pgen.1003130.g001

Figure 2 .
Figure 2. Non-vernalization requiring accessions do not rescue the early flowering phenotype of the pep1-1 mutant.(A) PEP1 mRNA levels in leaves of non-vernalization requiring accessions compared to Paj. (B) Flowering time of F1 hybrids resulted from crosses of non-vernalization requiring accession with pep1-1 mutant and Paj in long days without vernalization.The pep1-1 mutant and Paj were used as controls.Flowering time is measured as days to flower (DTF).(C) PEP1 accumulation in different accessions compared to the accession Paj before vernalization.pep1-1 and Paj after 16 weeks in vernalization were used as negative controls.A cross reacting protein acts as a loading control.doi:10.1371/journal.pgen.1003130.g002

Figure 3 .
Figure 3. Analysis of sequence variation in PEP1 cDNA and at the genomic locus of Dor accession demonstrates a complex structure for the PEP1 gene.(A) PEP1 cDNAs in Dor is a mixture of transcripts that contain a G to A substitution in exon 1 compared to Paj or have a similar sequence to accession Paj but have an insertion of 248 bp in the 59 UTR.(B) Sequence of the PEP1 genomic locus in the accession Paj shows that the locus is highly duplicated.Exons are indicated with black boxes, UTRs with white boxes and solid lines the inter-and intra-genic regions.Upstream and downstream genes of PEP1 are 35 kb apart.Colored boxes indicate relative positions of the duplicated regions.Overlapping boxes indicate overlapping homologous sequences.Numbers besides duplicated boxes show the length of the duplicated fragment and percentage of homology.Duplicated exon 1 copies are indicated as 1a and 1b.Dotted box shows the PEP1 locus region sequenced in the accession Dor.(C) Sequence of the PEP1 genomic locus in the Dor accession reveals that G to A base substitution is in exon 1a.Grey arrows indicate insertions, black arrows indicate deletions and vertical dotted lines indicate SNPs relative to Paj PEP1 locus.The 248 bp insertion upstream in the 59 UTR is upstream of exon 1b.Colored boxes indicate relative positions of duplicated regions.(D) Structure of the PEP1 locus and predicted splicing events (E) PEP1 transcripts in the accession Dor detected with two different primers in the 59 UTR using the same reverse primer in the 39 UTR.Black and grey arrows indicate the position of two different primers in the 59 UTR relative to the 248 bp insertion.When primer PEP1_5UTRF1 (black) was used most clones contained

Figure 4 .
Figure 4. PEP1 locus is tandemly duplicated in several A. alpina accessions.(A) Structure of the PEP1 locus and the position of exon 1a and exon 1b specific primers.(B) Sequence comparisons of FLC and tandem duplicated copies 1a and 1b in accession Paj (grey to blue boxes in Figure 3A).Vista plot using Calc.window 25, Min cons width 25 and Cons identity 70%.(C) Alignment using part of the 416 bp sequence (yellow box) specific for exon 1a from different FLC homologues.Intr1aR primer was designed in a consensus sequence.(D) PCR test using PEP1a specific primers (Ex1F and Intr1aR) in different accessions.Template used is Dor (1), Tot (2), Wca (3), Cza (4), Mug (5) Paj (6) and water control(7).(E) Alignment using part of the

Figure 5 .
Figure 5. PEP1a and PEP1b genes in the accession Dor are independently transcribed and have different transcriptional start sites.(A,B) Number of clones containing G to A polymorphism on exon 1 (A) or not (B) after 59 RACE using apices from Dor plants growing for 3 weeks in long days.Schematic representation of exon1 and 59 UTR regions (top), exon1a and exon 1b (black boxes), sequence present in 59 UTR upstream of both exon 1a and 1b (white box), sequence specific to 59 UTR upstream of exon 1a (red box), sequence specific to 59 UTR upstream exon 1b on the 248 bp insertion (grey box).Horizontal lines represent individual clones.Numbers on the top represent bp upstream of ATGs.(B,C) Percentage of clones with the A or G polymorphism after 59 RACE in apices and leaves before (3 week long days) and after vernalization (5 weeks in long days after 12 weeks vernalization).(D)-(F) PEP1 mRNA levels on 3 week old Dor plants, vernalized for 12 weeks and subsequently grown for several week in long days.(D) PEP1 (a+b) expression, primers used similar as in [5] to detect both transcripts (E) PEP1a expression, primers used to detect only PEP1a transcripts.(F) PEP1b expression, primers used to detect only PEP1b transcripts.doi:10.1371/journal.pgen.1003130.g005

Table 1 .
Comparison of PEP1 coding sequences of five early-flowering A. alpina accessions and the obligate-vernalization requiring Paj accession.
Multiple cDNAs were analyzed for the accessions Dor, Tot, Wca, Cza and Mug and the numbers of cDNA sequences recovered is shown in the ''Clones'' column.The fulllength PEP1 cDNA sequence of the vernalization-requiring accession Paj is used as a reference (row highlighted in grey).Nucleotide polymorphisms compared to Paj PEP1 cDNA sequence obtained for each accession are presented.Nucleotide (nucl.)positionandaminoacid (a.a.) changes compared to Paj mentioned in rows above the grey row.For most accessions, different PEP1 spicing forms were also recovered and are presented in TableS2.doi:10.1371/journal.pgen.1003130.t001