FRIGIDA LIKE 2 Is a Functional Allele in Landsberg erecta and Compensates for a Nonsense Allele of FRIGIDA LIKE 1 1[W][OA]

The Landsberg erecta (L er ) accession of Arabidopsis ( Arabidopsis thaliana ) has a weak allele of the ﬂoral inhibitor FLOWERING LOCUS C ( FLC ). FLC -L er is weakly up-regulated by the active San Feliu-2 (Sf2) allele of FRIGIDA ( FRI -Sf2), resulting in a moderately late-ﬂowering phenotype. By contrast, the Columbia (Col) allele of FLC is strongly up-regulated by FRI -Sf2, resulting in a very late-ﬂowering phenotype. In Col, the FRI -related gene FRI LIKE 1 ( FRL1 ) is required for FRI -mediated up-regulation of FLC . It is shown here that in L er , the FRL1 -related gene FRI LIKE 2 ( FRL2 ), but not FRL1 , is required for FRI - mediated up-regulation of FLC. FRL1 -L er is shown to be a nonsense allele of FRL1 due to a naturally occurring premature stop codon in the middle of the conceptual protein sequence, suggesting that FRL1 -L er is nonfunctional. Compared to FRL2 -Col, FRL2 -L er has two amino acid changes in the conceptual protein sequence. Plants homozygous for FRI -Sf2, FLC -L er , FRL1 -L er , and FRL2 -Col have no detectable FLC expression, resulting in an extremely early ﬂowering phenotype. Transformation of a genomic fragment of FRL2 -L er , but not of FRL2 -Col, into a recombinant inbred line derived from these plants restores both FRI mediated up-regulation of FLC expression and a late-ﬂowering phenotype, indicating that FRL2 -L er is the functional allele of FRL2 . Taken together, these results suggest that in the two different Arabidopsis accessions Col and L er , either FRL1 or FRL2

The timing of reproductive development is an important decision during the life cycle of flowering plants. The coordination of flowering time is vital for self-incompatible plant species, because they strongly need their sexual partners to flower at the same time. Coordinate regulation of flowering time is also required for the reproductive success of self-compatible species such as Arabidopsis (Arabidopsis thaliana). For instance, it can determine whether a population of dormant seeds or plants in a vegetative state will overwinter, because Arabidopsis has evolved both naturally occurring, early flowering summer-annual ecotypes and naturally occurring, late-flowering winter-annual ecotypes (Laibach, 1937). A major difference between the two growth habits is that winterannual types overwinter as vegetative seedlings or plants, because they require vernalization (exposure to a prolonged cold period during winter) to flower in the next spring or summer. By contrast, summerannual types generally, but not exclusively, produce seeds that remain dormant during winter and germinate the next year for a summer-annual growth habit (Nordborg and Bergelson, 1999). The requirement for vernalization in natural populations of Arabidopsis is mostly controlled by the synergistic interaction of two dominant genes, FRIGIDA (FRI) and FLOWERING LOCUS C (FLC; Michaels and Amasino, 1999;Sheldon et al., 1999;Johanson et al., 2000). FRI encodes a plantspecific coiled-coil domain-containing protein required for the up-regulation of FLC, which produces a MADS domain-containing transcription factor that acts as a strong floral repressor. Vernalization is antagonistic to FRI and leads to the epigenetic downregulation of FLC expression; that is, levels of FLC transcript remain low even after removal of the cold stimulus (Michaels and Amasino, 1999;Sheldon et al., 2000;Schläppi, 2001). FRI-mediated up-regulation of FLC is reset in the next generation when progeny plants become late flowering again (Amasino, 2004).
Until recently, FRI and FLC were considered the major determinants of flowering time in natural populations of Arabidopsis. This is because most early flowering accessions were shown to have either defects in FRI (Johanson et al., 2000;Le Corre et al., 2002;Gazzani et al., 2003), weak alleles of FLC (Koornneef et al., 1994;Lee et al., 1994;Sanda and Amasino, 1996;Schläppi, 2001;Gazzani et al., 2003;Michaels et al., 2003) or nonfunctional FLC transcripts (Shindo et al., 2005;Werner et al., 2005). However, recent studies have identified late-flowering Arabidopsis accessions that do not fit this pattern. Those accessions have either high levels of FLC expression in the absence of a functional FRI allele or are late flowering without a functional FLC allele (Schläppi, 2001;Werner et al., 2005). This suggests that there is naturally occurring variation in flowering time genes other than FRI and FLC. Through mutagenesis experiments with summerand winter-annual ecotypes, several classes of flowering time genes were identified that might be candidates for natural variation in FRI-or FLC-independent late flowering. Those are the six genes of the autonomous floral promotion pathway, LUMINIDEPENDENS, FCA, FLOWERING LOCUS D, FPA, FY, and FVE, that repress up-regulation of FLC expression in the absence of FRI (Boss et al., 2004); or the three FLC paralogs, FLOWERING LOCUS M/MADS AFFECTING FLOWERING 1 (MAF1), MAF2, and MAF3, that together with FLC have an additive effect on floral repression (Y. Pan and M.R. Schläppi, unpublished data) or repress flowering when overexpressed (Ratcliffe et al., 2001(Ratcliffe et al., , 2003Scortecci et al., 2001). Another class of more pleiotropic suppressors of FRI activity and FLC up-regulation includes chromatin regulators such as encoded by ACTIN-RELATED PROTEIN 6, PHOTOPERIOD-INDEPENDENT EARLY FLOWERING 1, or the VERNALIZATION INDEPENDENCE genes (Noh and Amasino, 2003;Oh et al., 2004;Choi et al., 2005;Deal et al., 2005).
In addition to finding accessions that are late flowering in the absence of active alleles of FRI or FLC, recent mutagenesis experiments led to the identification of FRI LIKE 1 (FRL1), a FRI-related gene that is required for the winter-annual growth habit of Arabidopsis (Michaels et al., 2004). A single frl1 mutant suppresses FRI-mediated late flowering and up-regulation of FLC expression in the Columbia (Col) ecotype of Arabidopsis. FRL1 is part of a gene family of six FRI-related genes, including FRI LIKE 2 (FRL2), which has some functional redundancy with FRL1 in the Col ecotype (Michaels et al., 2004). In this study, it is shown that there is naturally occurring variation at FRL1 and FRL2 between the Col and Landsberg erecta (Ler) accessions of Arabidopsis. In Col, FRL1 is functional, whereas in Ler, FRL2 is shown to be functional but not vice versa. This suggests that natural variation at these two loci could potentially modify or even suppress FLC upregulation and late flowering in Arabidopsis accessions that have functional alleles at both FRI and FLC.

Genetic Identification of a Modifier Gene Required for FRI-Mediated Late Flowering in the Ler Background
In previous work, the effect of FRI on flowering time and its interaction with FLC was investigated in different genetic backgrounds of Arabidopsis (Schläppi, 2001). In a series of genetic experiments, control test crosses were performed between the very late-flowering Col-FRI-San Feliu-2 (Sf2) line (rosette leaf no. [RLN] range 51-72) and the moderately late-flowering Ler-FRI-Sf2 line (RLN range 12-28), both of which contained the active FRI-Sf2 allele in the Col or Ler background, respectively (Schläppi, 2001). While F 1 plants (RLN range 46-68) were almost as late flowering as the late Col-FRI-Sf2 parent, about 1/16th (17/ 255) very early flowering F 2 plants (RLN range 3-5) were recovered from this cross. These early flowering Col/Ler-FRI-Sf2 F 2 plants were considered as transgressions, because they flowered significantly earlier than the earliest Ler-FRI-Sf2 parent. Test crosses and mapping analyses showed that all early flowering F 2 plants were homozygous for the active and dominant FRI-Sf2 gene, the weak and recessive FLC-Ler gene, and an unlinked recessive Col-specific suppressor gene of FRI-mediated late flowering (Schläppi, 2001). The flowering time phenotype suggested that the naturally occurring dominant Ler variant of this suppressor gene was required for FRI-mediated late flowering in the Ler background. Therefore, the gene was named ACTIVATOR OF FRI-MEDIATED LATE FLOWERING IN LER (AFL).
To determine the epistatic interaction between the naturally occurring Col and Ler alleles of AFL, three randomly selected early flowering Col/Ler-FRI-Sf2 plants containing AFL-Col (lines #1, #3, and #5; RLN range 3-4) were backcrossed with AFL-Ler-containing Ler-FRI-Sf2 (RLN range 18-27). As shown in Figure 1A, the flowering time phenotypes of F 1 plants (RLN range 5-18) from two of the three crosses were intermediate between that of the two parental lines, whereas F 1 plants from the third cross (RLN range 16-28) flowered as late as the Ler-FRI-Sf2 parent. As shown in Figure  To identify the chromosomal position of AFL, bulked segregant analysis was done with the 17 very early flowering F 2 individuals and simple sequence length polymorphism (SSLP) molecular markers (Supplemental Fig. S1). This analysis suggested that AFL was linked to the nga248 marker on chromosome 1. Individual analysis of the 17 very early flowering plants using six molecular markers on chromosome 1 indicated that AFL was located about 2 cM south of UNUSUAL FLORAL ORGANS (UFO; Supplemental Table S1). While mapping of AFL was in progress, work on FRI-related genes was published, describing that FRL1 on chromosome 5 was required for FRImediated up-regulation of FLC in the Col-FRI-Sf2 background (Michaels et al., 2004). Interestingly, the AFL locus mapped very closely to At1g31814, another FRI-related gene on chromosome 1. At1g31814 single mutants had no effect on flowering time; however, frl1 At1g31814 double mutants flowered slightly earlier than frl1 single mutants (Michaels et al., 2004). Therefore, At1g31814 was interpreted to be functionally redundant with FRL1 and was named FRL2. Thus, the map location of FRL2 and the absence of an frl2 single mutant phenotype in Col raised the possibility that FRL2-Col was identical to the suppressor allele AFL-Col, and conversely, that AFL-Ler was a functional allele of FRL2.
To test whether AFL-Ler was a functional allele of FRL2, a molecular-genetic approach was taken. Toward this end, a partial FRL2-Ler sequence was retrieved from the Monsanto Ler single-pass shotgun sequencing database (Jander et al., 2002) and com-pared to FRL2-Col. Four nonsynonymous polymorphisms were found in the coding regions between FRL2-Ler and FRL2-Col, one of which resulted in an AluI site in the Ler allele but not in the Col allele. As shown in Figure 2A, this AluI polymorphism was a useful cleaved amplified polymorphic sequence (CAPS) marker to distinguish between the Col and Ler alleles of FRL2. A prediction from the hypothesis that AFL is identical to FRL2 was that the very early flowering Col/Ler-FRI-Sf2 plants had the Col-specific suppressor allele of FRL2. Indeed, CAPS analysis of the 17 pooled very early flowering F 2 plants from the cross of Col-FRI-Sf2 with Ler-FRI-Sf2 indicated that there was an apparent bias for FRL2-Col, the Col allele of FRL2 (   2B) was heterozygous for both alleles. Individual CAPS analysis of the 17 very early flowering plants indicated that 16 out of 17 plants were homozygous for FRL2-Col and one plant was heterozygous. The heterozygous plant was subsequently shown to segregate early and late-flowering progeny, suggesting that it was misidentified during the initial selection (data not shown). Taken together, these results were an indication that the early flowering phenotype was associated with the Col allele of FRL2. To rule out that FRL2-Col was differentially expressed compared to FRL2-Ler, semiquantitative reverse transcription (RT)-PCR was performed on RNA isolated from Col, Ler, and FRL2-Col-containing, very early flowering Col/Ler-FRI-Sf2 F 2 plants. As shown in Figure 2C, FRL2 transcript levels were comparable in the Ler and Col genetic backgrounds, indicating that FRL2-Col allele was not a hypomorph.
To directly show that FRL2-Ler was required for FRImediated late flowering, a genomic clone of FRL2-Ler was transformed into the very early flowering, recombinant inbred line (RIL) Col/Ler-FRI-Sf2#1, which had been selfed and propagated from single seeds for five generations (Supplemental Table S2). To more easily identify transgenic plants that were later flowering than Col/Ler-FRI-Sf2#1, but earlier flowering than Ler-FRI-Sf2, the flowering time of 26 T 1 plants was analyzed under less inductive 12-h photoperiods. As shown in Figure 3A, individual T 1 plants had a range of flowering time phenotypes, from flowering as early as the Col/Ler-FRI-Sf2#1 parent to as late as the Ler-FRI-Sf2 control. Only six out of 26 T 1 plants flowered as early as Col/Ler-FRI-Sf2#1, whereas the rest flowered later (Fig. 3A). As a control, a genomic clone of FRL2-Col was transformed into the same very early flowering RIL, and 14 T 1 plants were recovered. All 14 plants flowered as early as the RIL control, producing only four to five rosette leaves at the time of bolting (data not shown). Taken together, these experiments demonstrated that introduction of FRL2-Ler into the very early flowering line Col/Ler-FRI-Sf2#1 restored late flowering in a majority of T 1 plants.
The late-flowering time phenotype of individual T 1 plants transformed with a genomic clone of FRL2-Ler was generally maintained in T 2 progeny plants and cosegregated with the transgene. For instance, lateflowering T 1 line 8 had a single kanamycin resistance (Km R ) gene locus and thus a single FRL2-Ler transgene locus. Kanamycin-selected T 2 progeny plants from the selfed T 1 parent were always late flowering, whereas unselected T 2 progeny plants segregated 3:1 late flowering:very early flowering (28 late:11 early; x 2 5 0.21; P . 0.5). The T 3 progeny of 10 unselected T 2 plants was subsequently analyzed for Km R segregation, indicating that all chosen early flowering plants were kanamycin sensitive (three plants), and that the seven chosen late-flowering plants were either homozygous (two plants) or heterozygous (five plants) for Km R . This suggested that the late-flowering phenotype cosegregated with the single FRL2-Ler transgene locus.
Moreover, as shown in Figure 3B, FRL2-Ler transgenecontaining T 2 progeny from T 1 line 16 that flowered late under 12-h photoperiods was also late flowering under long-day photoperiods. Conversely, T 2 progenies from T 1 plants that were early flowering remained early flowering. The T 2 flowering time distribution under long-day photoperiods of several transgenic plants is summarized in Table I. It is interesting to note that early flowering T 1 lines generally had a range of kanamycin-resistance cosuppression phenotypes in the T 2 generation, whereas selected T 2 progenies from lateflowering T 1 lines were fully kanamycin resistant.

Allelic Variation at FRL2 and FRL1
When a partial genomic FRL2-Ler sequence obtained from the Monsanto database (Jander et al., 2002) was compared with the standard FRL2-Col sequence, four nonsynonymous polymorphisms were found in the 473-amino acid sequence of FRL2. However, when the genomic FRL2-Ler fragment used to transform plants was sequenced for this study, only the following two polymorphisms were confirmed: the functional Ler allele has an Ala at position 132 (Ala-132;  AluI CAPS marker, Fig. 2) and a Leu at position 401 (Leu-401), whereas the nonfunctional Col allele has a Pro (Pro-132) and a Gln (Gln-401), respectively. It remains to be determined whether both Ala-132 and Leu-401 or only one of those substitutions is critical for the function of FRL2-Ler. An alignment of the two FRL2 proteins and their polymorphisms together with FRI is shown in Figure 4.
During mapping of the Col-specific suppressor gene, it appeared that besides FLC (near nga249), another Ler-specific region linked to FLC on chromosome 5, near marker nga139, cosegregated with the very early flowering phenotype (Supplemental Table  S1). Moreover, the previously reported complete absence of FLC expression in the RIL Col/Ler-FRI-Sf2#1 (Schlä ppi, 2001) was similar to the phenotype of Col-FRI-Sf2 plants containing single frl1 mutations (Michaels et al., 2004). This raised the possibility that the Ler accession had a nonfunctional allele of FRL1.
To determine whether FRL1-Ler had a lesion, a partial FRL1-Ler sequence from the Monsanto database (Jander et al., 2002) was compared to FRL1-Col. Interestingly, this comparison suggested that FRL1-Ler contained a premature stop codon in the middle of the protein sequence; that is, the codon GAG  in FRL1-Col was changed to TAG [stop-279] in FRL1-Ler. Cloning and sequencing of FRL1-Ler confirmed this significant polymorphism, together with two other codon changes reported by the Monsanto sequence (a Pro instead of a Thr at position 141 and a deletion of a Lys at position 387). Whether the other two amino acid changes affect FRL1-Ler function in the absence of the stop codon remains to be determined. An alignment of the two FRL1 proteins together with FRI is shown in Figure 5.
To genotype the stop codon in FRL1-Ler, a derived CAPS (dCAPS) marker was generated that allows cleavage of FRL1-Ler, but not of FRL1-Col, by the restriction enzyme SpeI (acTAGt recognition site). As shown in Figure 6A, SpeI indeed cleaved the dCAPS site in FRL1-Ler, but not in FRL1-Col, indicating that the dCAPS marker was functional. A dCAPS marker analysis was then done to test whether early flowering Col/Ler-FRI-Sf2 plants had the nonfunctional FRL1-Ler allele, as predicted from their early flowering phenotypes. As shown in Figure 6B, all early flowering Col/Ler-FRI-Sf2 plants assayed indeed had the FRL1-Ler nonsense allele. Taken together, these data suggest that in FRI-containing Ler, the functional FRL2-Ler allele is the major FRI-related gene required for FRImediated late flowering.
To determine whether the nonsense allele of FRL1-Ler was a result of mutagenesis in Ler, dCAPS analysis on LER, the unmutagenized parent of Ler (Rédei, 1962), was performed. The result indicated that FRL1-LER also has a premature stop codon and is thus a nonsense allele of FRL1 (Fig. 6C). Moreover, a CAPS analysis of the LER allele of FRL2 showed that it, too, contained the AluI site observed in FRL2-Ler (Fig. 6C). This suggests that the premature stop codon in FRL1 and the AluI polymorphism in FRL2 are naturally occurring in Ler and not the result of mutagenesis.

FRL2-Ler Promotes FRI-Mediated Activation of FLC
To determine whether introducing the genomic copy of FRL2-Ler into the early flowering line Col/ Ler-FRI-Sf2#1 restored FRI-mediated up-regulation of FLC, as suggested by the late-flowering phenotypes of most transformants, a semiquantitative RT-PCR analysis using RNA from late-flowering T 1 plants and untransformed control plants was done. As shown in Figure 7, the level of FLC transcript was very low in line Col/Ler-FRI-Sf2#1, comparable to Ler controls lacking active FRI-Sf2. This was consistent with previous results from RNA gel-blot analyses (Schläppi, 2001). By contrast, compared to Col/Ler-FRI-Sf2#1, three individual T 1 plants that were late flowering after transformation with FRL2-Ler ( Fig. 3; Table I) had higher levels of FLC transcript. This demonstrated that the late-flowering phenotype, after introducing a genomic copy of FRL2-Ler into Col/Ler-FRI-Sf2#1, indeed correlated with increased levels of FLC transcript. Taken together, the results of these experiments were in agreement with the hypothesis that FRL2-Ler is a functional FRI-related gene and required for FRImediated up-regulation of FLC transcripts in Ler.

DISCUSSION
The focus of this study was to characterize AFL, a gene required for FRI-mediated late flowering and FLC up-regulation in the Ler genetic background of Arabidopsis. The activity of AFL was discovered, because 1/16th of F 2 plants from the cross of Col-FRI-Sf2 with Table I. Flowering time analysis in long-day photoperiods The flowering time of individual T 2 plants derived from the very early flowering RIL Col/Ler-FRI-Sf2#1, transformed with a genomic copy of FRL2-Ler, was determined by counting the RLN at the time of bolting. Plants were selected for Km R (linked to the FRL2 transgene) and grown in 16 h of light and 8 h of dark conditions. CLN, Number of cauline leaves.   (Michaels et al., 2004). Consistent with this idea, all true-breeding, very early flowering transgressions tested were homozygous for the Col allele of FRL2. Therefore, a Ler-specific genomic fragment of FRL2 was genetically transformed into Col/Ler-FRI-Sf2#1, an RIL derived from one of the very early flowering transgressions, which resulted in restoration of a lateflowering phenotype and FLC up-regulation in a majority of T 1 plants (Figs. 3 and 7). This suggested that AFL is identical to the Ler allele of FRL2 and was thus renamed FRL2-Ler. By contrast, transformation of a Col-specific genomic fragment of FRL2 did not restore late flowering in the Col/Ler-FRI-Sf2#1 RIL, indicating that the Ler allele of FRL2 is functional and that the Col . CLUSTAL alignment of FRL2 and FRI. Polymorphisms in the amino acid sequence of the FRL2 proteins are indicated with the # sign. The underlined sequences above and below the alignment indicate the predicted coiled-coil domains of FRL2 and FRI, respectively. The COILS software at http://www.ch.embnet.org/software/COILS_form.html was used to predict the coiled-coil domains in FRL2 and FRI (Lupas et al., 1991). variant FRL2-Col is nonfunctional. This interpretation explains, at least in part, why a single frl2 mutant does not suppress the very late-flowering phenotype of Col-FRI-Sf2 (Michaels et al., 2004). Taken together, these results suggest that FRL2 is an active FRI-related gene in the Ler genetic background and required for FRImediated late flowering.
The reason why FRL2-Ler, but not FRL2-Col, is an active allele of FRL2 is not known at the moment. It does not appear that FRL2-Col is expressed at lower levels than the FRL2-Ler allele (Fig. 2), and it is thus more likely that either one or both of the amino acid substitutions in FRL2-Col has a negative effect on FRL2 function. It is, however, interesting to speculate that the Ala-132 to Pro-132 change has a more dramatic effect on FRL2 function than the other substitution. This is because Ala/Pro-132 is located between the two putative protein interacting coiled-coil domains of FRL2 (Lupas et al., 1991;Fig. 4). As shown in Figure 8, the closest homologs of FRL2 in Arabidopsis and other plants all have a very conserved Pro two amino acids prior to Ala/Pro-132, but none of them has an additional Pro in a Pro-X-Pro sequence as FRL2-Col does. It is, therefore, conceivable that an additional Pro at this position might change overall protein conformation of FRL2-Col and affect its ability to interact with protein partners. Because the closest FRL2 homologs have similarity to ABI3-interacting protein 2 (AIP2) in other plants (Fig. 8), it is intriguing to speculate that the proposed Pro-induced conformational change of FRL2-Col compromises its ability to interact with an ABI3-type protein in Arabidopsis. These questions can be addressed in future experiments using chimeras between the two protein sequences and might define a region or regions critical for the function of FRI-related proteins and potentially for FRI itself. Contrary to the absence of a phenotype for single frl2 mutants in Col-FRI-Sf2, mutations in the FRL2-related gene FRL1 have a strong effect and significantly suppress late flowering in Col-FRI-Sf2, indicating that FRL1 is an active FRI-related gene in the Col genetic background (Michaels et al., 2004). By contrast, the premature stop codon in the middle of FRL1-Ler indicates that it is a nonsense allele in the Ler genetic background and suggests that it is nonfunctional. It is interesting to note, however, that FRL1-Ler has a similar Pro substitution as FRL2-Col in the region between the two putative coiled-coil domains (Fig. 5), leading to a Pro-X-Pro sequence unique for FRL2-Col and FRL1-Ler. It is possible that this polymorphism leads to a nonfunctional protein even in the absence of the premature stop codon further downstream. This possibility, or whether the deletion of a Lys in the C terminus has an additional effect on protein function, can be addressed in future studies using protein chimeras.
The frl1 and frl2 single mutant phenotypes thus suggest that active FRL1 is the main requirement for FRI-mediated late flowering in the Col genetic background. However, FRL2-Col may have at least some partially overlapping function with FRL1, because it was reported that the frl1frl2 double mutant was slightly earlier flowering than the frl1 single mutant (Michaels et al., 2004). If FRL2-Col is indeed partially functional in Col, then its activity may not be strong enough for FRI-mediated up-regulation of the weak FLC-Ler allele. Therefore, if the strong FLC-Col allele requires mainly FRL1 activity to affect its FRI-medi-ated up-regulation, is it then possible that FLC-Ler is so weak that it requires the cooperation of both FRL1 and FRL2 for its FRI-mediated FLC up-regulation? From this study, the simple answer is no, because FRL1-Ler has a premature stop codon at position 279 (Figs. 5 and 6) and is, therefore, an apparent null allele of FRL1. This suggests that FRL2, but not FRL1, is necessary for FRI-mediated up-regulation of FLC transcripts and late flowering in Ler, and, conversely, that FRL1, but not FRL2, is active in Col. It is important to note that both the Ler-type nonsense allele of FRL1 and the AluI polymorphism of functional FRL2-Ler were also found in the LER accession (Fig. 6), the unmutagenized parent of Ler (Rédei, 1962). This indicates that the described variations at FRL1 and FRL2 in Ler are naturally occurring and not the result of mutagenesis.   That FRL1 is nonfunctional in Ler is most likely the reason why about 1/16th of the F 2 progeny from the cross of Col-FRI-Sf2 with Ler-FRI-Sf2 were very early flowering, because nonfunctional FRL1-Ler is closely linked to weak FLC-Ler. Thus, early flowering transgressions homozygous for FRL2-Col are probably not only produced in the presence of a weak FLC-Ler, as assumed before (Schläppi, 2001), but rather because FRL1 is linked to FLC on chromosome 5. The most likely scenario, therefore, is that FRL1-Ler and FRL2-Col are the two recessive genes with complementary gene action necessary for an early flowering transgression phenotype. However, FLC-Ler always cosegregated with FRL1-Ler in the limited number of transgressions tested here and was thus responsible for the very early flowering phenotype of those plants. If this interpretation is correct, then slightly later-flowering transgressions with recombination events between FLC-Col and FRL1-Ler should be identified when larger populations of early flowering transgressions are analyzed in future studies.
The observation that neither Col nor Ler have fully active alleles of both FRL1 and FRL2 may also explain, at least in part, why the FLC-Col allele appeared dominant in F 1 plants from the cross of Col-FRI-Sf2 with Ler-FRI-Sf2 but semidominant in other crosses (Lee et al., 1994;Schläppi, 2001). The reason for this dominance may be that F 1 plants from the Col 3 Ler cross have active alleles of both FRL1 and FRL2, which together might effect stronger up-regulation of either FLC-Col, FLC-Ler, or possibly both, thus compensating for the weak FLC-Ler copy in the F 1 hybrid. This question can be addressed in future experiments designed to determine whether FRL1-Col alone or a combination of FRL1-Col and FRL2-Ler enhances FRImediated up-regulation of weak FLC-Ler expression. Conversely, it can also be determined whether a combination of both active alleles produces very lateflowering Col-FRI-Sf2 plants that need a longer vernalization period to induce early flowering or whether both active alleles partially up-regulate FLC even in the absence of active FRI. It is also interesting to note that some F 1 plants from backcrosses between very early flowering transgressions and the late-flowering Ler-FRI-Sf2 tester had flowering times between the two parents, whereas other F 1 plants were as late as Ler-FRI-Sf2 (Figs. 1 and 2). One explanation for this observation is that FRL2-Ler may be semidominant or that the nonfunctional FRL2-Col protein somehow interferes with full FRL2-Ler activity in some crosses. This does not explain, however, why in other crosses FRL2-Ler can be fully dominant (Fig. 1). An alternative explanation is that accession-specific variants of other FRI-related genes such as At1g14900, At2g22440, At5g27230, and At5g48385 (Michaels et al., 2004) or other flowering time genes interact with FRL2 and thus regulate its activity in a dosage-dependent manner.
In summary, this study presents an example that naturally occurring variation of flowering time genes in Arabidopsis can be uncovered in very well-studied laboratory strains such as Col and Ler, which were previously used for quantitative trait loci mapping of flowering time loci Koornneef et al., 1998;Alonso-Blanco and Koornneef, 2000). It is thus possible that FRL1 and FRL2 correspond to some previously identified quantitative trait loci such as FLG on chromosome 5 or AD.121C on chromosome 1, which were identified in crosses of Ler to the Cape Verde Island accession of Arabidopsis . It is important to point out, however, that the large effect on flowering time of these naturally occurring suppressors of FRI-mediated late flowering in Col and Ler was uncovered in this study only because the active FRI-Sf2 allele had been introgressed into these laboratory stains (Lee et al., 1994). From this Col 3 Ler analysis it appears that there is some selection pressure to maintain an active copy of at least one FRI-related gene, even in the absence of an active FRI allele. It is thus likely that some of the previously observed, FRI-independent, flowering time variations (Gazzani et al., 2003;Werner et al., 2005) could be attributed to natural variation in FRL1 and FRL2. This may be especially true in the case of F 2 plants from the Ler 3 Col cross where earliness was linked to the recessive ms1 allele of Ler, which maps near FRL1 (Koornneef et al., 1994). It is now possible to test in future studies whether the FRL1 and FRL2 polymorphisms identified here can be correlated with flowering time differences and the adaptation to ecological niches of the large number of available Arabidopsis accessions collected from the wild.

Plant Growth Conditions
Per sterile petri dish (90-mm plate), about 100 surface-sterilized seeds were grown on 0.8% agar-solidified medium (Difco) containing half-strength Murashige and Skoog (Gibco BRL) salts without Suc. Petri dishes were placed at 4°C for up to 2 d to break seed dormancy, then grown under cool fluorescent light with a 16-h-light/8-h-dark long-day photoperiod or a 12-h-light/12-hdark short-day photoperiod with approximately 100 mmol m 22 s 21 photon flux and about 22°C day/night temperature. After 10 to 14 d, plantlets were transferred from petri dishes to soil (2:1:1 mix of peatmoss:vermiculite:perlite) into 2-inch pots (four plants/pot; 32 pots/flat) and grown under cool fluorescent light with a 16-h-light/8-h-dark long-day photoperiod, 20°C 6 1°C day/night temperature, and about 60% to 70% relative humidity. Flats were watered three times per week with 0.1 g/L 15-16-17 Peters fertilizer (Grace Sierra).