Multiple Roles of Arabidopsis VRN1 in Vernalization and Flowering Time Control

See allHide authors and affiliations

Science  12 Jul 2002:
Vol. 297, Issue 5579, pp. 243-246
DOI: 10.1126/science.1072147


Arabidopsis VRN genes mediate vernalization, the process by which a long period of cold induces a mitotically stable state that leads to accelerated flowering during later development. VRN1 encodes a protein that binds DNA in vitro in a non–sequence-specific manner and functions in stable repression of the major target of the vernalization pathway, the floral repressor FLC. Overexpression of VRN1 reveals a vernalization-independent function for VRN1, mediated predominantly through the floral pathway integratorFT, and demonstrates that VRN1 requires vernalization-specific factors to target FLC.

Many annual plants use seasonal variations in temperature and photoperiod to control the transition to flowering (1). Long periods of cold temperature (1 to 3 months of about 4°C) accelerate flowering later in development in a process known as vernalization. In Arabidopsis, the pathways that regulate vernalization requirement and response converge on FLC, a gene encoding a MADS-box repressor protein. In vernalization-responsive genotypes, reduction in the abundance of FLC mRNA, protein, and time to flower is closely correlated with the duration of cold treatment (2,3). The activity of genes in the autonomous floral promotion pathway (FCA, FY, FVE, FPA, and LD) also reduces FLC RNA levels (3, 4). Mutations in these genes conferFLC-mediated late flowering, which is reversed by vernalization to early flowering. Recent work suggests that the activity of the autonomous, photoperiod, and vernalization pathways is integrated by common downstream targets, including AGL20 andFT, and that FLC represses flowering by negatively regulating these genes (5, 6).

The molecular basis of vernalization has been investigated through characterization of a set of genes (VRN genes) defined by mutations that reduce the vernalization response (7). We recently showed that one of these genes,VRN2, encodes a nuclear-localized zinc finger with similarity to the Drosophila Polycomb group protein SU(Z)12 as well as Arabidopsis proteins FIS2 and EMF2 (8). In vrn2 mutants, FLC expression is down-regulated normally in response to cold but, instead of remaining low, FLC RNA levels increase during later development at warm temperatures. VRN2 thus functions to stably maintain FLC repression.

Another Arabidopsis mutant impaired in its response to vernalization is vrn1 (7). Under our long-day growth conditions, plants with mutant vrn1 alleles (vrn1-1 ethylmethane sulfonate–induced,vrn1-2 γ-ray induced) were not delayed in flowering but had a reduced vernalization response (Fig. 1, A and B). We conducted a detailed examination of FLC mRNA levels invrn1-1 plants to determine whether FLCrepression was altered. In both fca-1 andvrn1-1 fca-1 plants, 6 weeks of cold temperature reduced FLC mRNA levels. Invrn1-1 fca-1, unlikefca-1, FLC levels increased again after transfer to normal growth temperatures (fig. S1). Therefore,VRN1, like VRN2, is required not for the initial down-regulation of FLC by cold but for stable maintenance ofFLC repression during later development in warm temperatures.

Figure 1

Phenotype of vrn1 mutant plants and structure of VRN1 gene and encoded protein. (A) Plants (fca-1, left;vrn1-1 fca-1, right) were grown for 28 days under long-day photoperiods at 20°C, without (−) or with (+) 6 weeks of vernalization. (B) Vernalization dose-response curves. Landsberg erecta (open triangles),fca-1 (closed squares), vrn1-1 fca-1 (open squares), and vrn1-2 fca-1 (open circles) were vernalized for 0, 1, 3, or 6 weeks before transfer to growth in long days at 20°C. Flowering time was scored as total leaf number (mean ± SE) (for 20 plants) for each genotype/treatment. (C) Structure of theVRN1 gene (upper) and protein (lower). Exons are shown as black boxes and introns are shown as lines; the predicted translation start (ATG), stop (TGA), and polyadenylation [(A)n] sites are indicated; untranslated regions are boxed in white. Mutations in thevrn1 alleles are indicated by arrowheads. Ellipses indicate B3-like domains; vertical hatches; PEST-like sequences; and horizontal hatches, nuclear localization signals. See fig. S3 and GenBank accession numbers AF289051 and AF289052 for details of theVRN1 gene and transcript.

We cloned the VRN1 gene by a map-based approach (figs. S2 to S4). VRN1 encodes a protein of 341 residues comprising two putative B3 DNA binding domains, originally described in the maize transcription factor VIVIPAROUS1 (VP1) (9); two predicted PEST regions involved in proteasome-dependent protein degradation (10); and a putative nuclear localization signal (Fig. 1C; figs. S4, S5). Analysis of a GFP-VRN1 chimeric protein in onion epidermal cells showed that VRN1 was nuclear localized (fig. S6).

VRN1 mRNA was detected at moderate levels in different tissues and developmental stages (Fig. 2, A and B) and was undetectable in vrn1-1 fca-1 and vrn1-2 fca-1(Fig. 2C); as with VRN2 (8), VRN1 RNA levels were unaffected by vernalization (Fig. 2D). Rat polyclonal antibodies were raised against the NH2-terminal 192 amino acids of VRN1 produced in Escherichia coli. VRN1 protein was detected at equivalent levels in wild-type andfca-1 plants, was not detected invrn1-1 fca-1 andvrn1-2 fca-1 seedlings (Fig. 2E), and was unaffected by 6 weeks of cold (Fig. 2F). The role of two predicted PEST regions in the VRN1 sequence thus remains unclear. The 53 residues encompassing the two PEST regions have an overall acidic charge (pI = 3.8 versus about 9.2 for the whole protein), so this region may play a role in transcriptional activation, as suggested for a similar region in ARF1, another B3-containing protein (11,12).

Figure 2

Abundance of VRN1 RNA and protein during development and in response to vernalization: northern analysis of VRN1 RNA from fca-1tissues (A), fca-1 seedlings of different ages (B), fca-1 andvrn1 mutants (C), and fca-1seedlings without (−) or with (+) 6 weeks of vernalization (D). Equal amounts of total RNA were loaded per lane as judged by ethidium bromide staining and reprobing the blot with 18S rDNA. (E) Immunoblot analysis of VRN1 protein from 14-day-old seedlings of different genotypes. Twenty micrograms of total protein was loaded on a SDS–12% polyacrylamide gel and stained with Coomassie blue (left) or blotted and probed with a rat VRN1 polyclonal antiserum diluted 1:2000 (right). Arrowhead indicates VRN1 protein. No truncated form of VRN1 was detected with the extracts prepared from vrn1-1 fca-1 orvrn1-2 fca-1. (F) Immunoblot analysis of VRN1 protein from fca-1 seedlings harvested 0, 3, 6, 14, or 20 days after germination (DAG) without (−) or with (+) vernalization.

We fused VRN1 to the strong cauliflower mosaic virus35S promoter (13, 14) and transformed it into vrn1-2 fca-1plants. The most striking phenotype caused by the35S::VRN1 transgene was early flowering without vernalization (Table 1,Fig. 3A). When35S::VRN1 plants were vernalized before growth, flowering was further accelerated (Table 1). Under noninductive short-day conditions, the35S::VRN1 plants flowered later than under long photoperiods, and vernalization accelerated flowering (Table 1). This retention of photoperiod response in35S::VRN1 plants suggests thatVRN1 is not functioning through the photoperiod pathway. The35S::VRN1 plants showed additional morphologic phenotypes compared with control plants, including reduced flower abscission, a less acute angle between the siliques and the inflorescence stem, shortened pedicles (Fig. 3B), more variable and flattened silique shape (Fig. 3C), and a looser flower structure with somewhat enlarged petals (Fig. 3D). These characteristics suggest that VRN1 functions more broadly than vernalization. The role of VRN1 in vernalization-independent flowering control is also revealed when loss-of-function mutants are grown in different conditions (4). In 16-hour high-light photoperiods [with a red/far-red (R/FR) ratio of 2.3], vrn1 mutants do not flower late (Fig. 1B); however, in extended short-day (ESD) conditions (15), in which the long photoperiod is composed of 10 hours of high light (R/FR ratio of 2.3) and 6 hours of low-intensity incandescent light (R/FR ratio of 0.66), vrn1 is late flowering (Table 1).

Figure 3

Phenotypes of transgenic plants overexpressingVRN1. (A) Plants (21 days old) grown in ESD without vernalization. Left to right: wild-type, vrn1-2 fca-1 empty vector, two35S::VRN1 vrn1-2 fca-1lines, 13A and 12B (all Landsberg erecta background). (B) Compared with wild-type (left),35S::VRN1 (right) inflorescences showed a less acute angle between the siliques and the pedicles and a defect in floral organ abscission. (C) Silique morphology of35S::VRN1 lines (right) was variable compared with wild type (left). (D) Flowers of overexpressing lines (lower) were larger than those from the parental line (vrn1-2 fca-1) (upper) and had a looser structure. (E) Northern blot analysis ofFLC and AGL20 expression. Seedlings were grown without vernalization and harvested after 7 days. Total RNA samples were hybridized with gene-specific probes as described in (5, 8). β-Tubulin (TUB) mRNA served as a loading control. (F) Effect of35S::VRN1 transgene on FTexpression was investigated by reverse transcriptase PCR (RT-PCR), as described in (20, 21), on RNA prepared from 7-day-old nonvernalized seedlings. Duplicate samples were reverse transcribed [without (−) or with (+) enzyme] and hybridized withTUB- or FT-specific probes. (G) Northern blot analysis of FLC and TUB expression in line 35S::VRN1 6C with and without vernalization. (H) RT-PCR analysis of FT andTUB expression in loss-of-functionvrn1-2 seedlings grown for 3 weeks in ESD and harvested 2 and 19 hours after dawn. (I) Northern blot analysis of FLC and TUB expression in loss-of-function vrn1-2 seedlings grown in ESD. (J) Model for multiple roles of VRN1 in controlling the floral transition. In the vernalization response, VRN1 in association with cold-induced factors maintains repression ofFLC. VRN1 also acts in a pathway that does not affect FLC RNA levels but positively regulatesAGL20 and FT. This is drawn as an independent pathway to the photoperiod pathway, as35S::VRN1 plants flower later in short days and their flowering is further accelerated by vernalization.

Table 1

Effects of mutations and transgenes on flowering time of plants under different conditions [ESD, extended short-day, and SD, short-day photoperiods; V, 5 weeks of vernalization (15)]. Flowering time is recorded as mean total leaf number at flowering ± SE. At least 10 plants were counted unless indicated.

View this table:

To investigate the molecular basis of the acceleration of flowering by35S::VRN1 and the vernalization-independent function of VRN1, we examined the expression of FLC, AGL20, and FT. The abundance of AGL20 and FT RNA was increased in35S::VRN1 plants compared with wild-type and empty vector control plants (Fig. 3, E and F). One possible cause of elevated expression of FT andAGL20 in 35S::VRN1 plants is an enhancement of VRN1 repression of FLC. However, FLC RNA levels were unchanged in35S::VRN1 plants compared with controls (Fig. 3, E and G). Therefore, the accelerated flowering of35S::VRN1 lines under long- and short-day photoperiods without vernalization (Table 1) is through a pathway that does not regulate FLC RNA levels but does activate AGL20 and FT, targets of multiple floral pathways (Fig. 3J). Vernalization causes additional earliness in35S::VRN1 lines due to down-regulation of FLC RNA by vernalization (Fig. 3G). To investigate the consequences of ectopic FT and AGL20 expression, we introduced ft-7 and soc1 (a loss-of-function AGL20 allele) mutations into the35S::VRN1 line 6C introgressed into Landsberg erecta (Ler). In greenhouse conditions,soc1 ameliorated the phenotype of the35S::VRN1 lines only slightly, whereasft-7 35S::VRN1 lines flowered late (with between 19 and 23 leaves), similar toft-7. Therefore, in these conditions, the early flowering of 35S::VRN1 is mainly due to ectopic expression of FT. Consistent with this,vrn1-2 plants in ESD conditions showed reducedFT but wild-type FLC expression (Fig. 3, H and I).

So far, we have found no interaction of VRN1 and VRN2 by using yeast two-hybrid assays, so whether they function as part of the same multisubunit complex remains an open question. With electrophoretic mobility shift assays and surface plasmon resonance (SPR), we have pursued whether VRN1 could act in recruiting a protein complex to theFLC locus by investigating whether VRN1 specifically binds to FLC. Soluble E. coli expressed histidine-tagged and untagged whole VRN1 protein bound to 27FLC fragments in a salt- and concentration-dependent manner (Fig. 4, fig. S7). The FLCfragments used represented 95% of the FLC genomic region [beginning 1856 base pairs (bp) upstream of the ATG codon and ending 695 bp downstream of the translation stop codon, containing all noncoding sequence and some of the exon sequences] (table S1, fig. S7). We observed the strong non–sequence-specific interaction for VRN1 with a wide range of unrelated DNA sequences from Arabidopsis VRN2, AGL20, and TUB and E.coli GyrB as well as pBR322 and the DNA polynucleotide poly(dIdC). Such in vitro non–sequence-specific DNA binding has been observed for Polycomb/trithorax group proteins (16) and high mobility group box (HMG-box) proteins, including theArabidopsis FILAMENTOUS FLOWER protein, which is involved in meristem and organ identity (17). However, in vivo many proteins showing in vitro DNA binding activity are targeted to specific genomic regions (18). An important issue is to identify the range of VRN1 targets in vivo and to address whether VRN1 regulates FLC, FT, and AGL20directly. FLC and FT appear to be regulated by changes in chromatin structure (8, 19), so the non–sequence-specific DNA binding of VRN1 is consistent with it being a component of many multisubunit complexes involved in chromatin regulation. Another important issue to address is the nature of the vernalization-specific factors required to specify FLC as a target of VRN1. Cold-induced posttranslational modifications of VRN1 and changes in cellular location may change VRN1 activity. Alternatively, vernalization-induced accessory proteins may recruit VRN1 into a complex, possibly including VRN2, that can targetFLC.

Figure 4

VRN1 protein binds to DNA in a non–sequence-specific manner in vitro. (A) BIACORE SPR sensorgrams ( showing the binding of protein (equal 5-μg amounts) to a double-stranded DNA (140 bp) fragment from pBR322 immobilized on a SA chip. Asterisks mark beginning and end of DNA injections. VRN1 protein and the first B3 domain alone (VRN1 B3A) bind strongly compared with horse heart cytochrome c (CYT C) (similar overall pI to VRN1) and bovine serum albumin (BSA). (B) Typical electrophoretic mobility shift from VRN1 binding to an FLC fragment. VRN1 protein (0 to 5 μg), poly(dIdC) (2 μg), bovine serum albumin (6 μg), and a 360-bp radiolabeled double-stranded FLC probe (10,000 cpm) were incubated at room temperature for 30 min in 100 mM KCl, 10 mM Hepes·OH (pH 7.9), 1 mM EDTA, 10 mM MgCl2, 10% glycerol, 1 mM dithiothreitol and then separated on an 8% tris-borate EDTA–polyacrylamide gel. (C) DNA binding by VRN1 is salt dependent. Spiking 10-μmol injections of FLC DNA fragment 8 with increasing amounts of NaCl reduced the ability of the DNA to bind VRN1. Because SPR is affected by the addition of salt, the amount of DNA bound to the chip after the injection (RU after injection) was plotted.

  • * These authors contributed equally to this work.

  • Present address: Laboratoire de Recherche Moléculaire sur les Antibiotiques, 15 rue de l'Ecole de Médecine, 75006 Paris, France.

  • Present address: Department of Botany, La Trobe University, Bundoora, Victoria, 3083, Australia.

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


View Abstract

Navigate This Article