Molecular Analysis of FRIGIDA, a Major Determinant of Natural Variation in Arabidopsis Flowering Time

See allHide authors and affiliations

Science  13 Oct 2000:
Vol. 290, Issue 5490, pp. 344-347
DOI: 10.1126/science.290.5490.344


Vernalization, the acceleration of flowering by a long period of cold temperature, ensures that many plants overwinter vegetatively and flower in spring. In Arabidopsis, allelic variation at the FRIGIDA (FRI) locus is a major determinant of natural variation in flowering time. Dominant alleles ofFRI confer late flowering, which is reversed to earliness by vernalization. We cloned FRI and analyzed the molecular basis of the allelic variation. Most of the early-flowering ecotypes analyzed carry FRI alleles containing one of two different deletions that disrupt the open reading frame. Loss-of-function mutations at FRI have thus provided the basis for the evolution of many early-flowering ecotypes.

A requirement for vernalization—the acceleration of flowering that occurs during a 3- to 8-week period of cold temperature (4°C)—has been bred in many crops to produce winter/spring varieties. Vernalization requirement is also a major factor in determining flowering time in Arabidopsis thalianaecotypes. Despite the large number of genes known to control flowering time (1), the vernalization requirement segregates as a single gene trait (2–6) mapping to theFRIGIDA (FRI) locus (7). This locus was first described by Napp-Zinn (8), who analyzed the progeny of a cross between the late-flowering ecotype Stockholm and the early-flowering ecotype Li5. The action of an active FRI allele depends on an activeFLC allele (9, 10).

To analyze the molecular basis of the allelic variation atFRI, we cloned the gene using map-based techniques (Fig. 1) (11). FRI is a single-copy gene in the Arabidopsis genome and encodes a predicted open reading frame (ORF) of 609 amino acids (GenBank accession numbers: genomic FRI, AF228499; cDNA, AF228500). The predicted protein shows no significant match to any protein or protein domain of known function in available databases. TheFRI protein is predicted to contain coiled-coil domains in two positions (between amino acids 55 to 100 and 405 to 450, respectively). FRI has been shown to increase RNA levels ofFLC, which encodes a MADS-box protein likely to act as a transcriptional repressor (12, 13). Whether the predicted coiled coils in the FRI protein are important for this function remains to be tested.

Figure 1

Cloning of FRIGIDA. (A) Genetic interval, molecular markers, recombination positions, and yeast and bacterial artificial chromosome (YAC and BAC) clones covering FRI. Markers with mi and CC prefixes are described in (22), and those with UJ prefixes are new CAPS (cleaved amplified polymorphic sequence) markers distinguishing H51; Col DNA, 40D10, and 32F5 are H51 cosmids. Breakpoints in different recombinants are indicated at ends of horizontal arrows. IGF and TAMU BAC clones carry F and T prefixes, respectively. The different YAC clones are prefixed CIC, EW, EG, and YUP. The vertical lines and rectangles represent the extent of the region covered by the respective marker (11). (B) H51 cosmid clones covering the BAC clone F6N23. Cosmids complementing the flowering time phenotype (i.e., changing it from early to late) are shaded in red. (C) Subclones of cosmid 84M13, with complementing subclones in red. JU235 is an internal deletion of 84M13. The number of late-flowering individuals per number of transformants generated is indicated at the right. (D) Extent of deletions and rearrangements in three fast neutron–induced early-flowering mutants, FN13, FN233, and FN235 [from a population of >40 early mutants in the FRI(Sf-2), FLC-Col background (9)]. Southern analysis showed that FN233 and FN235 carried intact FRI promoter regions but lacked most of the coding region of FRI. FN13 could be interpreted as having an insertion or a complex rearrangement within a 600-bp Pst I–Sac I fragment of exon 1 of FRI, internal to the Sca I fragment missing in subclone JU235. (E) Introduction of H51FRI allele into Li5. Wild-type Li5 is shown on the left. A late-flowering primary transformant followingAgrobacterium-mediated transformation of Li5 with the cosmid 84M13 is shown on the right. Both plants were photographed once they had flowered (after ∼3 months for the transformant and 3 weeks for the parent).

The rapid-cycling ecotypes Columbia (Col) and Landsbergerecta (Ler) carry recessive FRIalleles. To analyze the basis of the recessivity, we compared ∼3.6 kb of genomic sequence from the dominant H51 [a derivative of the late-flowering ecotype Stockholm (8)] FRI allele with the same region from Col and Ler. Ten polymorphisms were found between H51 and Col. Two result in amino acid differences (Gly146 → Glu and Met148 → Ile, respectively, the former resulting in loss of a Bsm FI restriction site); another is a 16–base pair (bp) deletion at the end of exon 1, which changes the reading frame and terminates the ORF immediately at the beginning of exon 2 (Fig. 2).

Figure 2

Molecular structure of different FRIalleles. The FRI gene contains two introns, 393 and 89 bp long, positioned 954 and 1520 bp downstream of the Met residue of the likely translation start codon. The genomic regions of H51, Col, and Ler alleles are represented schematically, with the changes in nucleotide and amino acid sequence shown and the positions of the deletions indicated.

Three differences were detected between H51 and Ler FRIalleles: two single-base changes that do not alter the ORF, and a 376-bp deletion combined with a 31-bp insertion that disrupts the beginning of the ORF, removing the putative translation start codon (Fig. 2). The additional 31 bp appear to be a partial duplication of the subsequent 52 bp, and it carries an ATG codon. If translation occurred, this would very likely act as the start codon and would yield a short out-of-frame 41–amino acid peptide. No FRItranscript is detected in Ler seedlings when transcript can be detected in both H51 and Col seedlings.

To survey a more extensive set of FRI alleles, we designed polymerase chain reaction–based tests to score the 16-bp and 376-bp deletions and the Gly → Glu amino acid polymorphism, the latter detected by the presence or absence of the Bsm FI restriction site. We analyzed 38 additional ecotypes that had been classified as early- or late-flowering in the absence of vernalization (10, 14,15). The results of this analysis (Table 1) enable a classification of the ecotypes into five groups. Group 1, represented by Col, contains early-flowering ecotypes having FRI alleles with the 16-bp deletion and lacking the Bsm FI site. Group 2, represented by Ler, contains early-flowering ecotypes having FRIalleles with the 376-bp deletion, the 31-bp insertion, and the Bsm FI site. Group 3, represented by H51, and group 4 are all late-flowering ecotypes and have FRI alleles with or without the Bsm FI site, respectively, but none have either of the two deletions. Group 5, however, contains six early-flowering ecotypes that (on the basis of the polymorphism tests) carry functional FRI alleles.

Table 1

Classification of Arabidopsis ecotypes based on FRI alleles and flowering time. Flowering times (FT) of different ecotypes are from (10, 14, 15); ecotypes were classified as flowering early (E) or late (L), the latter being scored if the flowering time without vernalization was >75 days or the plants had more than 10 leaves. Where discrepancies occurred in the published data, flowering time was taken from our own unpublished observations. Stock center accession numbers of the ecotypes are included where available. The promoter difference was scored (+, present; –, deletion) using primers 5′-AGTACTCACAAGTCACAAC-3′ and 5′-GAAGATCATCGAATTGGC-3′. The presence of the Gly residue was scored by a CAPS marker (primers 5′-CCATAGACGAATTAGCTGC-3′ and 5′-AGACTCCAGTATAAGAAG-3′) and the presence of a Bsm F1 site. The 16-bp difference was scored (+, present; –, deletion) using primers UJ26 (5′-AGATTTGCTGGATTTGATAAGG-3′) and UJ34 (5′-ATATTTGATGTGCTCTCC-3′). Longitude and latitude coordinates are given for collection points of different ecotypes, except for Columbia whose geographical origin is unclear.

View this table:

These results suggest that the FRI allele from the late-flowering ecotypes is the ancestral form of the gene, with early flowering evolving independently at least twice from late-flowering ecotypes through deletion events leading to loss of FRIfunction. Group 1 early-flowering ecotypes would seem to be derived from a group 4 ecotype and group 2 from a group 3 ecotype. The late flowering of groups 3 and 4 supports the interpretation that the Gly → Glu and Met → Ile difference does not contribute to the earliness of group 1 ecotypes.

FRI was sequenced from the ecotypes Sf-2 (group 3) and EDI-0 (group 4). The only difference between the EDI-0 and Col alleles was the 16-bp deletion, which suggested that the deletion occurred relatively recently. Four polymorphisms distinguished Sf-2 and Ler, one of which was the 376-bp deletion and 31-bp insertion. One other led to a change in the protein (a conservative amino acid substitution, Leu79 → Ile). The greater number of differences between Sf-2 and Ler FRI alleles as compared to EDI-0 and Col suggests that the deletion that gave rise to the group 2 ecotypes predated the deletion that gave rise to the group 1 ecotypes. Ler also carries recessive FLC alleles, so it is possible that changes at FLC were the primary cause of earliness in Ler and the FRI deletion was secondary. It will be important to determine the dominance of theFLC alleles in the other group 2 ecotypes to further address this question.

A FRI allele from a group 5 ecotype was also sequenced to establish whether a different mutational event accounted for the earliness of this group. The FRI allele from the ecotype Shakhdara, from Tadjikistan, did not contain any deletions or rearrangements relative to the H51 allele. The Shakhdara allele showed six nucleotide differences in comparison to H51, with only one polymorphism resulting in an amino acid substitution (Phe55→ Ile). Northern analysis showed that Shakhdara FRI is expressed at the same level as FRI in the late-flowering H51 parent. In addition, FRI genotype analysis of late- and early-flowering F2 individuals from a Shakhdara × Columbia cross showed that the Shakhdara FRI allele confers late flowering (16). These data suggest that the Shakhdara FRI allele is fully functional and cannot account for the early flowering of Shakhdara. Crosses of Shakhdara to FRI flc and fri flc genotypes [an flc mutant from the FRI/FLC mutagenesis (13) with or without the active FRI allele] resulted in F1 plants that were early flowering (∼12 leaves), whereas crosses to fri FLC (wild-type Col) resulted in late-flowering F1 plants (>35 leaves). Thus, Shakhdara, like Ler, appears to carry recessive FLC alleles, which could account for its early-flowering phenotype. Another ecotype in group 5, Cvi, has dominant FLC alleles (10,17), so FLC variation cannot account for the earliness of all the group 5 ecotypes. Whether the other group 5 ecotypes carry FRI alleles with an as yet unidentified lesion, or whether genes other than FRI and FLCaccount for their earliness, remains undetermined. It is interesting that mutations at FRI are the basis of earliness in the majority of early-flowering ecotypes, as frimutations accounted for relatively few (3 from >40) of the early-flowering mutants in the FRI/FLCmutagenesis. This may indicate that loss of function of other genes incurs a fitness penalty in the natural environment, or that the strong phenotypic change caused by fri mutations conferred the greatest selective advantage.

The independent appearance of early-flowering ecotypes in the evolution of Arabidopsis suggests that there has been strong selection in some environments for ecotypes that do not require vernalization. The longitude and latitude coordinates for the collection points of the ecotypes are shown in Table 1. The ecotypes show a north-south distribution that is significantly different from random (0.1 >P > 0.05, Wilcoxon two-sample test). The majority of the late-flowering ecotypes are from northern latitudes, whereas most of the early-flowering ecotypes were collected from central and eastern Europe. This might suggest that a strong vernalization requirement conferred by active FRI alleles is a selective advantage in northern regions but a disadvantage in localities where winters are milder. However, there was no correlation between flowering time and mean temperatures (December–February or September–October, 1961 through 1990) or altitude for the ecotypes shown in Table 1. Variation in flowering times of ecotypes collected from the same locality indicates the complexity of the selective forces in different microenvironments (15, 18). Flowering early without vernalization may be an advantage where severe winter conditions prevent germination [e.g., the Shakhdara ecotype (18)], in climates that would support more than one generation per year, or where there is a selective advantage in escaping agricultural harvesting, succession, or summer drought (19). The selective forces driving the molecular evolution of flowering time genes can now be examined in more detail.

There was no clear geographical association of the ecotypes within the groups defined by the FRI alleles, or between the early-flowering group 1 ecotypes and the late-flowering group 4 ecotypes from which they may have arisen. Human-induced dispersal has been a major factor in the recent spread of Arabidopsisecotypes and has resulted in little association of geographical and genetic distance (20, 21). Human disturbance regularly exposes Arabidopsis ecotypes to novel environments, thus maintaining a strong selective pressure for adaptive mutations. This may account for loss-of-function mutations providing the basis for the evolutionary changes in Arabidopsis flowering time.

  • * Present address: Department of Plant Biochemistry, Lund University, Post Office Box 117, SE-221 00 Lund, Sweden.

  • To whom correspondence should be addressed. E-mail: caroline.dean{at}

  • Present address: Department of Plant Sciences, University of Cambridge, Cambridge CB2 3EA, UK.


View Abstract

Navigate This Article