The timing of flowering is important for the reproductive success of plants. Here we describe the identification and characterization of a new MADS-box gene, FLOWERING LOCUS M (FLM), which is involved in the transition from vegetative to reproductive development. FLM is similar in amino-acid sequence to FLC, another MADS-box gene involved in flowering-time control. flm mutants are early flowering in both inductive and non-inductive photoperiods, and flowering time is sensitive to FLM dosage. FLM overexpression produces late-flowering plants. Thus FLM acts as an inhibitor of flowering. FLM is expressed in areas of cell division such as root and shoot apical regions and leaf primordia.
The transition from vegetative to reproductive growth is a critical transition in the life cycle of flowering plants. In many species the timing of this transition is determined by an interaction between developmental programs and pathways that respond to environmental cues such as daylength and temperature (e.g. Bernier, 1988; Koornneef et al., 1998). These developmental programs and environmental pathways affect both the production of signals that promote or repress flowering, and the competence of shoot apical meristem to undergo the flowering transition (e.g. Bernier, 1988; Levy and Dean, 1998).
Arabidopsis thaliana is a facultative long-day plant, that is, it flowers earlier in long days (LD) than in short days (SD) (Koornneef et al., 1998). Genetic analyses have revealed several genes that, when mutated, cause delayed flowering. These mutants can be grouped according to their flowering response to photoperiod and to exposure to low temperature (the promotion of flowering by low temperature is known as vernalization.) One group of delayed-flowering mutants retains a response to photoperiod and to a vernalizing cold treatment. This group includes FCA, FLOWERING LOCUS D, FPA, FVE and LUMINIDEPENDENS (LD); these genes are thought to promote flowering through a photoperiod-independent pathway known as the autonomous pathway (Levy and Dean, 1998). Another group of delayed-flowering mutants has little or no response to photoperiod or vernalization. This group, which includes CONSTANS (CO), FD, FE, FHA, FT, FWA and GIGANTEA (GI), comprises the photoperiod pathway (Levy and Dean, 1998). A third pathway is represented by the gibberellin biosynthesis mutant ga1. These mutants are only slightly delayed under LD but do not flower in SD (Wilson et al., 1992). Thus ga1 appears to be required by a pathway that promotes flowering in SD (Levy and Dean, 1998).
In Arabidopsis, naturally occurring late-flowering accessions containing the FRIGIDA (FRI) gene and plants containing mutations in the autonomous floral-promotion pathway are relatively late flowering unless vernalized (Johanson et al., 2000; Michaels and Amasino, 2001). The late-flowering phenotype of these backgrounds is caused by up-regulation of the floral inhibitor FLOWERING LOCUS C (FLC). After vernalization, however, FLC transcript levels are down-regulated and remain low for the remainder of the plant's life (Michaels and Amasino, 1999; Sheldon et al., 1999). Thus FLC acts as a central regulator of flowering time that is up-regulated by FRI and down-regulated by vernalization and the autonomous pathway.
Phylogenetic analysis places FLC in a subfamily with two other MADS-box genes, AGL27 and ALG31 (Alvarez-Buylla et al., 2000). Here we report that AGL27 is a regulator of flowering time that acts as a repressor of the transition from vegetative to reproductive development, and we designate this gene FLOWERING LOCUS M (FLM).
Identification and description of flm mutants
FLC was first identified genetically as an inhibitor of flowering that plays a central role in the timing of the transition to flowering in Arabidopsis (Koornneef et al., 1994; Lee et al., 1994). FLC is a MADS-box gene that is up-regulated by FRI and down-regulated by the autonomous pathway or by vernalization (Michaels and Amasino, 1999; Sheldon et al., 1999). Phylogenetic analysis reveals that the two closest relatives of FLC in the Arabidopsis genome are AGL27 and AGL31 (Alvarez-Buylla et al., 2000). AGL31 is organized in tandem repeats forming a cluster of four nearly identical genes at the bottom of chromosome V (GenBank accession numbers AB019236, AB026633 and AB013395), and AGL27 is located at the bottom of chromosome I (GenBank accession number AC002291). Like FLC, AGL27 is a negative regulator of flowering in Arabidopsis (see below). Thus we have designated this gene as FLOWERING LOCUS M (FLM); the M in this designation reflects that this gene encodes a MADS-domain protein. The MADS domain of FLM is 70% identical to FLC and 74% identical to AGL31 (Figure 1a).
Screening of a cDNA library identified two forms of FLM resulting from differential splicing at two sites. If differential splicing occurs randomly at these two sites, four mRNA splice variants would be formed. To determine if all splicing variants are produced, FLM cDNAs were amplified by RT–PCR, cloned and subjected to diagnostic restriction enzyme digestions that would distinguish the four forms. All four predicted transcripts were present, and their identity was confirmed by sequencing (see Experimental procedures). The four forms were designated: FLMα;FLMβ;FLMγ; and FLMδ(Figure 1a,c). All of these splice variants contain seven of nine possible exons (Figure 1b). FLMα is comprised of exons 1, 2, 4, 5, 6, 7 and 8; FLMβ, exons 1, 2, 4, 5, 6, 7 and 9; FLMγ, exons 1, 3, 4, 5, 6, 7 and 8; and FLMδ, exons 1, 3, 4, 5, 6, 7 and 9 (Figure 1c). In wild-type Wassilewskija (Ws) seedlings, all splicing variants are approximately equally represented (in 87 clones we found 22 clones for the α variant, 19 for β, 28 for γ, 18 for δ).
To determine whether FLM, like FLC, is a repressor of flowering, we identified two T-DNA-insertion mutations in FLM in the Ws accession using a reverse genetic approach (Krysan et al., 1999). Each line contained a single T-DNA locus based on DNA blot analysis of a segregating population (data not shown). The flm-1 allele has a T-DNA insertion in exon 9, and the flm-2 allele has a T-DNA insertion in intron 1 (Figure 1b). Plants homozygous for either mutant allele are early flowering (Figure 2). To confirm that the disruption of FLM affected flowering behavior, both mutant alleles were back-crossed to wild-type Ws and the flowering time and FLM genotype of plants in segregating F2 populations were analyzed. Plants from the segregating F2 population for each allele were genotyped by PCR and then analyzed for leaf number. For both alleles, under both LD and SD conditions, the F2 population contained an early, an intermediate and a wild-type class segregating in a 1 : 2 : 1 ratio. The earliest-flowering class was comprised of the homozygous mutants. The flm-1 and flm-2 homozygous mutants flowered with an average of two leaves earlier than wild type under LD, and 12 leaves earlier than wild type under SD (Table 1). Plants heterozygous for either allele exhibited the intermediate phenotype, indicating that flowering is sensitive to FLM dosage (Table 1). The intermediate phenotype was most apparent in SD (Table 1). Although the flm-1 allele has a T-DNA insertion in the last exon of the β and δ forms, both mutant alleles produced a similar early flowering phenotype and transcript was not detected in either mutants by RT–PCR, thus both are likely to represent loss-of-function mutations (data not shown).
Table 1. Leaf number of flm mutants in LD and SD conditions
Total leaf number was counted as rosette leaves plus cauline leaves. The number of plants analyzed is given in parentheses.
5.2 ± 0.5
2.1 ± 0.5
7.1 ± 0.9 (27)
5.2 ± 0.7
2.0 ± 0.6
7.2 ± 0.9 (34)
6.2 ± 0.6
2.1 ± 0.4
8.3 ± 0.8 (42)
5.9 ± 0.7
2.2 ± 0.5
8.2 ± 0.8 (38)
7.1 ± 0.9
2.5 ± 0.7
9.4 ± 0.8 (44)
10.0 ± 2.2
4.3 ± 0.8
14.3 ± 2.4 (15)
10.2 ± 2.3
4.1 ± 0.9
14.2 ± 2.9 (32)
17.1 ± 4.6
5.6 ± 1.3
21.5 ± 5.6 (24)
16.1 ± 4.6
5.4 ± 1.2
21.0 ± 5.9 (48)
22.6 ± 4.7
7.5 ± 1.2
30.6 ± 5.1 (28)
Overexpression of FLM delays flowering
The early flowering phenotype of flm mutants indicates that FLM is an inhibitor of flowering. To further investigate the role of FLM, a genomic clone of FLM was fused to the constitutive 35SCaMV promoter. In the first transformed generation (T1), 55 of 76 plants were late flowering compared to the parental line; the remaining lines showed no alteration in flowering time. Photoperiod and vernalization responsiveness was examined in the progeny of several of these transformants (T2 lines). Overexpression of FLM delayed flowering under both LD and SD conditions (Table 2). Late-flowering T2 plants flowered with an average of 25 leaves in LD and 60 leaves in SD (compared to 10 and 41 leaves in LD and SD, respectively, for wild type) and thus FLM overexpression lines retain a photoperiod response. Also, the delay in flowering resulting from FLM overexpression was not significantly affected by vernalizing cold treatments (Table 2), thus FLM does not appear to act upstream of vernalization-sensitive components of flowering.
Table 2. Leaf number of FLM overexpressors in LD and SD conditions
The levels of FLM were difficult to detect by RNA blot analysis. Therefore quantitative RT–PCR was used to analyze FLM mRNA expression. To avoid cross-hybridization with other MADS-box genes such as FLC and AGL31, a pair of primers were designed to sequences that were unique to FLM and which amplify all splice variants.
FLM expression was detected in all vegetative tissues as well as inflorescences and flowers (Figure 3a). These results are consistent with previous studies of AGL27 expression (Alvarez-Buylla et al., 2000). Transcript levels did not differ between plants growing in LD and SD, indicating that FLM expression was not affected by photoperiod (Figure 3). Transcript levels were also unaffected by plant age during vegetative development (Figure 3b).
FLC is up-regulated by FRI and down-regulated by vernalization and the autonomous pathway. Because of the sequence similarity between FLC and FLM, and their similar roles as inhibitors of flowering, we wished to determine whether FLM is regulated similarly to FLC. FLM transcript levels were unchanged by FRI, autonomous-pathway mutants or a 40-day vernalization period (Figure 4a,b). FLM mRNA levels were also not affected by the presence of the photoperiod pathway mutants co and gi (data not shown). Thus the regulation of FLM is distinct from that of FLC.
To examine more carefully the pattern of FLM expression, a construct was created that contained the β-glucuronidase (GUS) gene (Jefferson, 1987) fused in frame to an FLM genomic clone at the NheI restriction site in exon 7. Analysis of transgenic plants containing this GUS fusion revealed a pattern of fusion protein accumulation that changed during development (Figure 5). In seedlings up to 3 days after germination (DAG; germination refers to radicle emergence), GUS activity is present throughout the plant (Figure 5a). As the seedling grows, the pattern of GUS staining becomes restricted. In plants 5 DAG and older, staining is only seen in the root and shoot apex and in young leaves (Figure 5b–e). After the transition to flowering, staining in the inflorescence meristem is reduced and strong expression is restricted to floral primordia (Figure 5f). A similar expression pattern is also observed for SVP (Hartmann et al., 2000).
In this study we report the identification of a gene encoding a novel MADS-domain protein from Arabidopsis as a flowering repressor. We designated this gene FLOWERING LOCUS M (FLM) because the amino-acid sequence of the encoded protein is related to the MADS-box gene FLOWERING LOCUS C (FLC), which has previously been described as an inhibitor of flowering (Michaels and Amasino, 1999; Sheldon et al., 1999). Sequence comparison reveals that FLM, FLC and the tandem repeat of four AGL31 genes comprise a subfamily of MADS-box genes (Alvarez-Buylla et al., 2000; these authors refer to FLM as AGL27). Subfamily members often exhibit similar expression patterns and related functions (Theissen et al., 1996). The expression pattern and function of the AGL31 genes has not been determined, but it would not be surprising if one or more members of the AGL31 genes also act as flowering repressors. However, determining the role of these genes may be difficult because if two or more of these genes are redundant, classical genetic approaches may require lesions in all active genes to reveal a phenotype. Thus approaches such as overexpression and gene silencing may be required to identify the role of these AGL31 genes.
FLM has two sites at which differential splicing occurs. The differential splicing at each site is independent, and thus there are four forms of FLM mRNA. Comparison of the predicted amino-acid sequence from the four different forms of FLM with FLC and AGL31 reveals that the β and δ forms are more similar to FLC, whereas the α and γ forms are more similar to AGL31. It is not yet known whether one or several of the FLM splice variants encode a protein that is active in flowering-time control. The splice variants are present in approximately equal proportions in RNA prepared from seedlings. However, it is possible that regulation of splicing may occur in specific cell types. Future work will address the activities of the individual splice variants and whether regulation of splicing may occur.
The mutant and overexpression phenotypes indicate that FLM acts as an inhibitor of flowering. Two independent flm mutants are early flowering in both LD and SD conditions. Also, overexpression of FLM from the constitutive CaMV 35S promoter delays flowering.
Two mutations that affect flowering time, fe and efs, are located near FLM on the bottom of chromosome I (Koornneef et al., 1998). Because these genes have not yet been molecularly identified, it is possible that they correspond to lesions in FLM. The fe mutant, however, delays flowering whereas flm mutants are early flowering. Although the efs mutant flowers early like flm, it has pleiotropic effects not seen in flm mutants. Furthermore, efs is located 3.4 cm from ADH (Soppe et al., 1999), but ADH is located very close to FLM (≈20 kb). Thus it seems unlikely that either fe or efs is an allele of FLM. In addition to FE and EFS, several quantitative trait loci have also been mapped to this region (Koornneef et al., 1998), but whether FLM corresponds to any of these loci has not been established.
Despite the close sequence similarity between FLC and FLM, the function of FLM may be most similar to that of SVP. For example, flm and svp mutants have quite similar phenotypes; both mutants flower early in non-inductive SD photoperiods as well as inductive LD photoperiods, and this early flowering phenotype is observed in rapidly flowering summer-annual accessions such as Ws and Col. In contrast, the early flowering phenotype of flc mutations in inductive LD photoperiods is observed only in genetic backgrounds that are otherwise late-flowering, such as FRI-containing winter annuals or in autonomous-pathway mutants (Michaels and Amasino, 1999; Michaels and Amasino, 2001). Furthermore, in summer-annual accessions such as Col, flc mutations have only a slight affect on flowering in SD (Michaels and Amasino, 2001). Another distinction between FLM and FLC is the regulation of mRNA levels. FLC expression is positively regulated by FRI and negatively regulated by autonomous-pathway genes such as LD, FCA, FPA and FVE. FLM, in contrast, does not exhibit any regulation by FRI or the autonomous pathway and, as in SVP, expression is also not affected by photoperiod. Given the similarities between FLM and SVP, it is likely that they act in the same flowering pathway, and it is possible that they act as a heterodimer to repress flowering in Arabidopsis.
Plant growth conditions
Imbibed seeds were incubated on agar-solidified medium containing 0.65 g l−1 Peters Excel 15-5-15 fertilizer (Grace Sierra, Milpitas, CA) for 2 days at 4°C for stratification before being transferred to soil. Growth conditions were 22°C and 60% relative humidity under approximately 100 µE m−2 sec−1 cool-white fluorescent light. Long-day conditions consisted of 16 h light and 8 h dark; short days were 8 h light and 16 h dark. Plants used to obtain roots for RNA extraction were cultured in flasks with liquid Murashige–Skoog medium with 1% sucrose under continuous light with gentle shaking. For vernalization treatment, seeds were incubated on the agar medium mentioned above for 24 h at 21°C and then shifted to 4°C for 40 days under SD conditions.
The flm-1 allele was detected using the primers (5′-GGATAG AAGCGCTGTTCAAGCCGGA-3′) and (5′-TGTCTCCGAAGGAGGT ACAACACTG-3′) which gives a fragment of 1550 bp for FLM in wild-type and heterozygous plants, and no amplification in homozygous mutant plants. To verify the presence of the T-DNA in flm-1 mutants, a PCR reaction was done using the primers (5′-GGATAGAAGCGCTGTTCAAGCCGGA-3′) and (5′-CATTTTATA ATAACGCTGCGGACATCTAC-3), which gives a fragment in the homozygous mutant and heterozygous plants. The flm-2 allele was detected using the primers (5′-TCCTTTTCTGGGTCTCACT CGA-3′) and (5′-TGTGTGGCGAGTATCAATGT GG-3′) which gives a fragment of 700 bp for FLM in wild-type and heterozygous plants, and no amplification in homozygous plants. To verify the presence of the T-DNA in flm-2, a PCR reaction was done using the primers (5′-TGTGTGGCGAGTATCAATGTGG-3′) and (5′-AGC ACGGGAACTGGGATGAC-3′), which gives a fragment in both homozygous mutant and heterozygous plants.
Total RNA was extracted using the TRI reagent (Sigma, St Louis, MO) as indicated by the manufacturer.
The total RNA was first treated with DNAseI as recommended by the manufacturer (Promega, Madison, WI). cDNA synthesis was performed using 5 µg of total RNA with superscript reverse transcriptase (Life Technologies, Grand Island, NY). The FLM gene was amplified using the primers 5′-GTGAGCTAGGAA GGCAGAACTGA-3′ and 5′-CCGAAGGAGGTACAACACTGATCC-3′; for FLC analysis the primers 5′-CCGAACTCATGTTGAAGCTTG TTG-3′ and 5′-AAACGCTCGCCCTTATCAGCGG-3′ were used. PCR conditions were as follows: 3 min at 95°C, then 25 cycles of 95°C 30 sec, 65°C 20 sec, 72°C 30 sec, and then 72°C for 2 min. Polyubiquitin was used as a control (primers 5′-GATCTTTGC CGGAAAACAATTGGAGGATGGT-3′ and 5′-CGACTTGTCATTAG AAAGAAAGAGATAACAGG-3′) using the same PCR conditions but with 20 cycles. ExTaq polymerase (Pan Vera/Takara, Madison, WI) was used in these reactions. The amplified fragments were separated on a 1.2% agarose gel, blotted onto a nylon membrane (Biotrace HP, Gelman, Ann Arbor, MI), and hybridized with radiolabeled FLM, FLC and UBQ probes. Signals were visualized using the Molecular Dynamics phosphorimager (Sunnyvale, CA). The RT–PCR analysis was repeated at least three times with independent RNA preparations.
GUS staining and sectioning
Histochemical analyses for GUS activity were conducted according to Jefferson (1987), and sectioning was done according to Patterson (1998) and Fernandez et al. (2000). The GUS expression in the sections was analyzed using dark field in a Zeiss Microscope with a 20× magnification lens.
Analysis of FLM splice variants
FLM transcripts were amplified using the same conditions as described above using the primers 5′-GGCCATGGGAAGAAG AAAAATCGAGATC-3′ and 5′-CCGAAGGAGGTACAACACTGA TCC-3′. The RT–PCR product was cloned into pGEMT-Easy (Promega, Madison, WI). Plasmids were digested with EcoRI to release the insert from the vector and with BglII which cleaves two of the four splice variants. Thus the four forms can be distinguished as follows: α will be cleaved by BglII producing fragments of 243 bp and 577 bp; β is not cleaved by BglII and gives a fragment of 710 bp; γ will be cleaved by BglII producing fragments of 290 bp and 580 bp; δ is not cleaved by BglII and gives a fragment of 757 bp. These digestions were analyzed in 2% metaphor agarose gels (ISCBioexpress, Kayville, UT). These plasmids were also sequenced by Perkin Elmer ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction (PE Applied Biosystems, Foster City, CA) using the T7 primer.
We are grateful to Martin Yanofsky and Sherry Kempin for providing cDNA sequences and clones that were found for FLM/AGL27, to Jim Busse for sectioning the FLM::GUS lines, and to Ed Himelblau for assisting in figure preparation. We thank the National Science Foundation-sponsored Arabidopsis Knockout Facility at the University of Wisconsin for assistance in identifying insertion mutations in FLM. This work was supported by the College of Agricultural and Life Sciences of the University of Wisconsin, and by grants to R.M.A. from the US Department of Agriculture NRI Competitive Grants Program and the National Science Foundation. K.C.S is supported by a Brazilian fellowship from FAPESP (98/15276-0).