early in short days 4, a mutation in Arabidopsis that causes early flowering and reduces the mRNA abundance of the floral repressor FLC

The plant shoot is derived from stem cells within the shoot apical meristem(SAM). In Arabidopsis, the primordia that form on the flanks of the SAM soon after germination give rise to leaves, while those that form later in shoot development produce flowers. This change in the identity of the lateral organs formed on the shoot is regulated by environmental conditions, such as temperature and daylength, and by the age of the plants. Early in the development of floral primordia, the mRNA of the LEAFY gene accumulates and encodes a transcription factor that activates the expression of floral homeotic genes (Weigel et al.,1992; Parcy et al.,1998). LEAFY expression is increased by environmental conditions that promote flowering(Blázquez et al., 1998;Blázquez and Weigel,2000), and this complex response is mediated by signals that act on the shoot meristem and probably directly on the developing primordium(Hempel and Feldman, 1994). Some of the signals are formed in the mature leaves and promote or repress flowering at the apex of the shoot, while others act within the apex and determine its competence to respond to these signals (reviewed byAukerman and Amasino,1998).

A systematic genetic approach to identifying genes that regulate flowering time has been taken in Arabidopsis (reviewed byAraki, 2001;Mouradov et al., 2002;Simpson and Dean, 2002). Many mutations have been identified that delay flowering, and genetic and physiological analysis has placed these mutations in at least three independent pathways that promote flowering(Koornneef et al., 1998). These are the long day pathway, the autonomous pathway and the gibberellic acid (GA)-dependent pathway. Mutations affecting the long day pathway (co,fd, fe, sha, ft, fwa, gi and lhy) delay flowering under long day conditions, whereas those affecting the autonomous pathway (fca, fpa, fve,fy and ld) delay flowering under all photoperiods. Mutations that strongly reduce GA biosynthesis delay flowering under long days, and almost abolish flowering under short days(Wilson et al., 1992). The existence of these pathways is supported by the phenotypes of double mutants(Koornneef et al., 1991;Koornneef et al., 1998). Partial redundancy between the long day, autonomous and GA pathways probably explains why no single mutation has been identified that prevents flowering. However a triple mutant, in which all three flowering pathways are impaired does not flower under long or short days, indicating that these pathways are absolutely required for flowering under these conditions(Reeves and Coupland,2001).

The cloning of several flowering-time genes, and analysis of their expression in wild-type and mutant backgrounds, have led to detailed models of the mechanisms underlying the flowering response(Lee et al., 1994;Putterill et al., 1995;Macknight et al., 1997;Schaffer et al., 1998;Michaels and Amasino, 1999a;Sheldon et al., 1999;Kardailsky et al., 1999;Kobayashi et al., 1999;Samach et al., 2000;Lee et al., 2000;Suárez-López et al.,2001; El-Assal et al.,2001; Gendall et al., 2002). These models are supported by the phenotypes of transgenic plants in which flowering-time genes are overexpressed (Kardailsky et al.,1999; Kobayashi et al.,1999; Onouchi et al.,2000; Lee et al.,2000). An endogenous circadian clock acts to control the expression patterns of genes within the long day pathway, enabling the promotion of flowering under appropriate day lengths(Schaffer et al., 1998;Fowler et al., 1999;Park et al., 1999;Suárez-López et al.,2001). The autonomous pathway appears to promote flowering by reducing the expression of the FLC gene that encodes a repressor of flowering (Michaels and Amasino,1999a; Sheldon et al.,1999; Michaels and Amasino,2001). More recently, it has become apparent that the long day,autonomous, and GA pathways converge on a common set of target genes to regulate flowering time and flower development. For example, all three pathways are involved in the regulation of expression of LEAFY(Simon et al., 1996;Blázquez et al., 1998;Nilsson et al., 1998;Blázquez and Weigel,2000), SUPPRESSOR OF OVEREXPRESSION OF CONSTANS(SOC1)/AGAMOUSLIKE 20 (AGL20)(Samach et al., 2000;Lee et al., 2000;Borner et al., 2000;Michaels and Amasino, 2001)and FT (Kardailsky et al.,1999; Kobayashi et al.,1999; Samach et al.,2000;Suárez-López et al.,2001; Ohto et al.,2001).

Additional flowering pathways promote flowering specifically in response to vernalization. The vernalization response shares common targets with the autonomous pathway, and acts to repress FLC mRNA abundance(Michaels and Amasino, 1999a;Sheldon et al., 1999). However, vernalization is not mediated by any of the three pathways described above, because the vernalization response is not abolished in mutants representative of each pathway (Koornneef et al., 1991; Chandler et al.,2000; Michaels and Amasino,1999b), nor in a co-2 fca-1 gal-3 triple mutant in which all three pathways are impaired (Reeves and Coupland, 2001). The FRI gene confers a vernalization response on naturally occurring varieties of Arabidopsis(Johanson et al., 2000), and is involved in the promotion of FLC mRNA levels(Michaels and Amasino, 1999a;Sheldon et al., 1999;Johanson et al., 2000). Stable repression of FLC by vernalization requires VERNALIZATION 2, which was proposed to act within a protein complex similar to Polycomb group complexes of Drosophila (Gendall et al., 2001). Both the promotion of flowering by the autonomous pathway and the delay in flowering by strong FRI alleles are absolutely dependent on active FLC, although the vernalization response also has an FLC independent component(Michaels and Amasino,2001).

A diverse group of mutations causes early flowering in Arabidopsis(Alvarez et al., 1992;Zagotta et al., 1996;Hicks et al., 1996;Goodrich et al., 1997;Telfer and Poethig, 1998;Somers et al., 1998;Soppe et al., 1999;Scott et al., 1999;Hartmann et al., 2000;Michaels and Amasino, 2001;Gomez-Mena et al., 2001). For example, the phyB mutation disrupts the gene encoding the red/far-red light receptor PHYTOCHROME B (PHYB) and causes early flowering under both long and short days (Reed et al.,1993; Koornneef et al.,1995). The elf3 mutation causes early flowering under short days so that mutants flower at the same time irrespective of daylength(Hicks et al., 1996). Under continuous light, elf3 also disrupts the rhythmic expression of the circadian clock-regulated gene CAB2, and the rhythmic movement of leaves. ELF3 is proposed to act by gating light signalling to the circadian clock (McWatters et al.,2000). The early-flowering mutant, hasty, forms adult leaves earlier in vegetative development than wild type(Telfer and Poethig, 1998). The apical meristem of the hasty mutant can also respond more rapidly to expression of the floral meristem-identity gene LEAFY, suggesting that HASTY reduces the competence of the shoot to respond to flowering signals. EMBRYONIC FLOWER mutations cause extreme early flowering, probably by inactivating transcriptional repression complexes that in wild-type plants repress the expression of floral identity genes such asAPETALA1 (Chen et al.,1997; Kinoshita et al.,2001).

Combining mutations causing early or late flowering can provide information on how the functions of the affected genes are inter-related(Yang et al., 1995;Koornneef et al., 1995;Koornneef et al., 1998a; Weller et al.,1997; Soppe et al.,1999; Michaels and Amasino,2001). For example, in Arabidopsis, the efs gene was suggested to act within the autonomous pathway on the basis of double mutant analysis (Soppe et al.,1999). However, in general, little information is available on how mutations causing early or late flowering interact in Arabidopsis and the molecular basis for such interactions is even less clear.

Here we describe a novel mutation, early in short days 4(esd4), that causes an extreme early-flowering phenotype inArabidopsis. The analysis demonstrates genetic and molecular relationships between ESD4 and genes in the autonomous flowering pathway.

The early in short days mutants were isolated in theArabidopsis thaliana ecotype Landsberg erecta(Ler). Ler seeds were subjected to 90 kRad gamma irradiation by Dr Mary Anderson, University of Nottingham. Around 40,000 M₂plants derived from approximately 2,200 M₁ parents were screened under short-day conditions and three early-flowering mutants were recovered. Together with two mutants isolated independently by Maarten Koornneef(University of Wagningen, Netherlands), these were named early in short days 1 to early in short days 5 (esd1 toesd5). The two mutants provided by M. Koornneef were esd2and esd3.

Mutant seed stocks were all in Ler and were provided by the following individuals: fca-1, fve-1, co-2, fwa-1, ft-1 M. Koornneef(University of Wageningen), ag-3 J. Goodrich (University of Edinburgh), gai-1, ga1-3, (Nottingham Stock Centre).

Flowering time was measured under controlled conditions as described previously (Reeves and Coupland,2001). Short days consisted of a photoperiod of 10 hours light,whereas long days consisted of 10 hours light with a 6 hour daylength extension supplied by incandescent bulbs. At least 15 plants were used to examine the flowering time of each genotype. For analysis of FT, SOC1and CO mRNA expression patterns, plants were grown in cabinets under true long days of 16 hours light with no daylength extension.

The map position of ESD4 was defined using the CAPS markers 326(5′-GTGACGTACTCGGTGAAG, 5′-CTCTACTACACACCACAC; StyI) and SC5 (5′-TCGACGACTCTCAAGAACCC, 5′-CACAAGCTATACGATGCTCACC;AccI).

The sample was mounted on an aluminium stub which was then plunged into liquid nitrogen slush and transferred onto the cryostage of a CT1000 Hexland cryo-transfer system at -85°C (Oxford Instruments, Oxford) fitted to a CamScan Mark IV scanning electron microscope (Gresham-CamScan, Cambridge). The sample was transferred to the pre-chamber, placed on a stage at -195°C and sputter coated with gold for 6 minutes at 2 mA. It was returned to the main cryo-stage of the microscope and viewed at 16 kV.

Information on the construction of double mutants can be obtained from the authors.

To examine the flowering times of esd4 gal-3 and gal-3plants seeds were germinated without applying exogenous GA as described previously (Reeves and Coupland,2001).

RNA extraction, northern blotting and hybridisation was as described by Suárez-López et al.(Suárez-López et al.,2001). The FLC probe was provided by A. Gendall and C. Dean (JIC, Norwich), and consisted of a 403 bp PCR fragment corresponding to nucleotides 298-700 of the FLC cDNA sequence. The FT probe was described previously (Samach et al.,2000). The UBQ10 probe was as described by Wang et al.(Wang et al., 1997). TheSOC1 probe was a 459 bp PCR fragment, amplified from theSOC1 cDNA with the primers 5′-AATATGCAAGATACCATAGATCG-3′and 5′-TCTTGAAGAACAAGGTAACCCAAT-3′. The strength of hybridisation signals was assessed using a Phosphorimager (Molecular Dynamics). The ratio of signal intensity compared to the UBQ10 control was calculated for each sample using ImageQuant software (Molecular Dynamics).

RT-PCR analysis of CO mRNA abundance was as described by Suárez-López et al.(Suárez-López et al.,2001).

Early-flowering mutants were identified under short-day conditions after mutagenesis with gamma rays (see Materials and Methods). To exclude the possibility of recovering previously analysed hy or elf3mutants (Reed et al., 1993;Hicks et al., 1996), only early-flowering individuals that did not show an elongated hypocotyl were selected. The early in short days 4 (esd4) mutant was the most extreme. None of the other mutations recovered were alleles ofesd4, because F₁ plants derived from crossing each of them to esd4 flowered at the same time as wild type.

The esd4 mutant was back-crossed to Ler three times before further phenotypic analysis, and crossed to Columbia to enable its position to be determined relative to RFLP markers. For the mapping, DNA was extracted from 200 F₂ plants that were homozygous for esd4and analysed with several CAPS markers. Linkage was detected to markers on the lower arm of chromosome 4. For example, esd4 was located approximately 1.8 cM distal to marker 326, and 0.5 cM proximal to SC5. Relative to phenotypic markers, esd4 is therefore located around 2.3 cM distal to cop9 and 0.5 cM proximal to fca. The map position of esd4 excluded the possibility that it was an allele of any previously described mutation causing early flowering.

Plants homozygous for esd4 were grown under long and short days,and their flowering time compared with that of Ler(Table 1;Fig. 1). The early-flowering phenotype was most dramatic under short days where wild-type plants flowered after forming around 49 rosette and cauline leaves compared to only 10 foresd4. The mutants also flowered slightly earlier than wild-type plants under long days, forming only 5 leaves compared to 9 for Ler. The esd4 mutation therefore causes early flowering under long and short days, and the flowering time of the mutant is influenced by daylength,although less strongly than that of wild-type plants(Table 1).

Arabidopsis plants exhibit heteroblasty, forming juvenile rosette leaves that have trichomes only on their adaxial (upper) side; adult rosette leaves that develop trichomes on their adaxial and abaxial (lower) sides, and cauline leaves that form on the stem above the rosette and have trichomes on both surfaces (Chien and Sussex,1996; Telfer et al.,1997). The early-flowering hasty mutant forms fewer juvenile rosette leaves than wild type but approximately the same numbers of adult rosette and cauline leaves (Telfer and Poethig, 1998). As shown inFig. 2, esd4 showed a reduction in all types of leaf. However, the most dramatic effect was on adult rosette leaves, which were absent from esd4 mutants grown under long days and dramatically reduced in number under short days.

The main inflorescence of esd4 mutants showed several abnormalities. Internodes between the cauline leaves and solitary flowers are shorter, and the leaves are smaller than in wild-type plants(Fig. 1). Also at the nodes at which the last leaf or the first solitary flower form there are often alterations to phyllotaxy compared to wild-type plants(Fig. 1). For example, in approximately 13% of esd4 mutants, two solitary flowers develop at the first node after the last cauline leaf, while in another 10% of mutants the last cauline leaf to develop and the first solitary flower form at the same node on opposite sides of the stem. More complex abnormalities also occur at this point of transition from cauline leaves to solitary flowers(Fig. 1).

In addition, esd4 mutants formed fewer flowers than wild-type plants (Table 1). The main inflorescence of wild-type plants contained approximately 37 solitary flowers under long days, and around 59 under short days. However, the main inflorescence of esd4 mutants formed many fewer solitary flowers:approximately 19 and 34 under long and short days, respectively. This reduction in flower number was in part associated with the conversion of the shoot apical meristem into a carpelloid, pistil-like structure(Fig. 1) that does not occur in wild-type plants (Fig. 1). Under long days, the main inflorescence of approximately 80% of esd4mutants terminated with this structure. Under short days, this phenotype was less severe, and was only observed in around 25% of plants.

After fertilisation, the gynoecium of Arabidopsis plants elongates to form a silique that contains the developing seeds(Bowman, 1993). The siliques ofesd4 mutants are shorter than those of wild type, and are broader at the tip (Fig. 1). The siliques of esd4 mutants contain 2 valves, as do those of wild-type plants(Bowman, 1993).

Plants heterozygous for esd4 were generated by back-crossing to Ler. The shoot of the heterozygotes showed a wild-type phenotype, and the heterozygous plants flowered at the same time as wild-type under long and short days (Table 1). Theesd4 mutation is therefore recessive with respect to all aspects of the mutant phenotype.

To determine whether the esd4 phenotype was due to a mutation at a single locus, 78 F₂ plants derived from a back-cross ofesd4 to wild-type plants were examined. Nineteen plants flowered at a similar time to esd4 and showed all of the shoot phenotypes previously described for the esd4 mutant, while 59 plants flowered at a similar time to the Ler wild-type, and showed a wild-type shoot phenotype. The ratio of esd4-like plants to wild-type like plants was approximately 3:1, suggesting that the esd4 phenotype is caused by a mutation at a single genetic locus.

Ectopic expression of the floral-organ identity gene AGAMOUS was previously shown to cause early flowering(Mizukami and Ma, 1992;Goodrich et al., 1997). To test whether AG was required for the early flowering ofesd4, double mutants carrying both esd4 and ag-3were made. These plants flowered at the same time as esd4 under long and short days. AG is therefore not required for the early flowering of esd4 plants. However, ectopic expression of other genes that encode MADS box containing proteins can also cause early flowering. We therefore constructed the esd4 ap3, esd4 pi and esd4 ap1double mutants and they all flowered at the same time as esd4,indicating that AP3, PI and AP1 are also not required for early flowering ofesd4.

To test the relationship between ESD4 and genes that promote flowering, double mutants were made containing esd4 and mutations causing late flowering. The co, ft and fwa mutations were proposed to affect the long-day pathway(Koornneef et al., 1998). The double mutants esd4 co-2, esd4 ft-1 and esd4 fwa-1 were constructed to test the effects of the mutations on the esd4phenotype.

Under long days, all three double mutants flowered later than esd4(Table 2;Fig. 3). The flowering times of the esd4 co-2, esd4 fwa-1 and esd4 ft-1 double mutants were similar to that of wild type. Under short days, the co-2, ft-1 andfwa-1 mutants do not show late flowering compared to wild type. However, these mutations delayed flowering of the esd4 mutant under these conditions (Table 2,Fig. 4). The ft-1 andfwa-1 mutations caused a more severe delay in the flowering time ofesd4 than co-2, and under short days the esd4 ft-1and esd4 fwa-1 plants flowered almost as late as wild-type and the late flowering parent. The latest flowering genotype was esd4 fwa-1that formed around 23 rosette leaves under short days compared to approximately 9 for esd4 (Table 2).

The pleiotropic effects of esd4 were also reduced in severity in the esd4 co-2, esd4 ft-1 and esd4 fwa-1 plants. More flowers were formed on the primary inflorescences of the double mutants, and the proportion of plants in which the primary inflorescence was determinate was reduced. For example, in long days esd4 co-2, esd4 ft-1 and esd4 fwa-1 produced over 35, 72 and 61 flowers, respectively, compared to the 20 flowers produced by esd4. Only around 37% of esd4 co-2mutants showed a determinate main inflorescence, whereas no esd4 ft-1or esd4 fwa-1 plants showed this phenotype. The frequency with which abnormalities in floral phyllotaxy were observed at the node representing the transition from cauline leaves to flowers(Fig. 1) was also reduced. These were visible in fewer than 8% of esd4 co-2, and 3% of esd4 ft-1 mutants. No esd4 fwa-1 mutants showed these abnormalities. The double mutants retained the altered silique shape of esd4, and remained slightly dwarfed with respect to wild type, with the exception ofesd4 ft-1, which was taller than wild type.

The fca and fve mutations affect the autonomous flowering-time pathway (Koornneef et al.,1998). The double mutants esd4 fve-1 and esd4 fca-1 were made and their flowering times scored under long and short days (Table 2; Figs4 and5).

Under long days, esd4 fca-1 and esd4 fve-1 double mutants flowered earlier than wild type, and were earlier flowering than the esd4 co-2, esd4 ft-1 and esd4 fwa-1 plants described above. Some of the double mutant plants were indistinguishable from esd4, although on average esd4 fve-1 and esd4 fca-1 formed 1 or 2 rosette leaves more than esd4 (Table 2). esd4 fve-1 double mutants flowered slightly earlier than esd4 fca-1 double mutants, probably because fve-1causes a weaker phenotype than fca-1(Table 2)(Koornneef et al., 1991).

Under short days, esd4 fca-1 and esd4 fve-1 mutants again flowered earlier than wild-type Ler and the late-flowering parent(Table 2). The double mutant plants flowered later than the esd4 mutant, with esd4 fca-1plants and esd4 fve-1 plants forming a total of 19 and 15 leaves,compared to 12 in the esd4 mutant. Furthermore, although thefca-1 and fve-1 mutants flowered much later than any of the long-day pathway mutants described in the previous section(Table 2; see above), the relative severity of their phenotypes was in general reversed in the presence of the esd4 mutation: the esd4 fca-1 and esd4 fve-1double mutants flowered much earlier under short days than esd4 fwa-1and esd4 ft-1.

Under both long and short days, the double mutants still showed the pleiotropic effects caused by the esd4 mutation; the siliques retained their club-like appearance, the shoot terminated in a carpelloid structure, and the phyllotaxy of flowers on the shoot was disrupted. The frequency with which an abnormality occurs at the node representing the transition from cauline leaves to flowers was reduced from 45% inesd4 to 19% in esd4 fca-1 and 15% in esd4 fve-1. The number of flowers formed on the main inflorescence was also affected:under long days, 29 flowers were formed by esd4 fca-1 plants and 21 by esd4 fve-1 plants compared to 20 for esd4. However, thefca-1 and fve-1 mutations reduced the severity of the pleiotropic effects of esd4 to a much lesser extent than the long-day pathway mutations did (see previous section).

Mutations that affect GA synthesis or signal transduction delay flowering weakly under long days, and severely under short days(Wilson et al., 1992). The severe mutation ga1-3 disrupts an early step in GA biosynthesis, and prevents flowering under short days (Sun and Kamiya, 1994; Wilson et al., 1992). The gai mutation affects GA signalling(Peng et al., 1997). To determine whether GA synthesis and response pathways are required for the early flowering caused by esd4, plants carrying esd4 ga1-3and esd4 gai were constructed.

Under both long and short days the esd4 gai double mutants flowered earlier than wild type, particularly under short days, and at approximately the same time as esd4 mutants(Table 2). The gaimutation therefore has almost no effect on the early-flowering phenotype caused by esd4. The esd4 ga1-3 plants showed a flowering time intermediate between the esd4 and ga1-3 parents,indicating that GA synthesis is required for the extreme early-flowering phenotype caused by esd4. However, esd4 still has a dramatic effect on the flowering time of ga1-3 mutants, particularly under short days where ga1-3 flowered after forming approximately 80 leaves, while the esd4 gal-3 double mutants formed approximately 15(Table 2;Fig. 4).

A co-2 fca-1 ga1-3 triple mutant, in which all three flowering time pathways are impaired, does not flower under long days(Reeves and Coupland, 2001). To determine whether esd4 promotes flowering in this triple mutant background, esd4 co-2 fca-1 ga1-3 quadruple mutants were constructed and their flowering time examined under long days. As previously shown, theco-2 fca-1 ga1-3 control plants did not flower under long-day conditions (Table 2)(Reeves and Coupland, 2001). However, the esd4 co-2 fca-1 ga1-3 quadruple mutant flowered after the production of 28 leaves.

The esd4 mutation most effectively suppressed the late-flowering phenotype of mutations that impair the autonomous pathway(Fig. 3;Table 2). Therefore,esd4 might cause early flowering by increasing the activity of the autonomous pathway downstream of FCA and FVE, or by bypassing the requirement for the autonomous pathway. FLC, which encodes a repressor of flowering, is a downstream target of both the autonomous and the vernalization-dependent floral promotion pathways(Michaels and Amasino, 1999a;Michaels and Amasino, 2001;Sheldon et al., 1999;Sheldon et al., 2000).FLC mRNA abundance is increased in mutants impaired in the autonomous pathway and this increase is responsible for their late-flowering phenotype. Therefore, esd4 may suppress the effect of autonomous pathway mutations by reducing FLC mRNA levels.

The abundance of the FLC mRNA was compared in wild-type, esd4,fca-1, esd4 fca-1, fve-1 and esd4 fve-1 seedlings that were 7 days old and had been grown in long days(Fig. 5A). FLC mRNA abundance was higher in both fca-1 and fve-1 mutants than in wild-type plants, as previously shown(Sheldon et al., 1999;Michaels and Amasino, 1999a). However, in esd4 fca-1 and esd4 fve-1 double mutants, the level of FLC mRNA was reduced compared to the late floweringfca-1 and fve-1 parents. Nevertheless, FLC mRNA abundance was still higher than in wild-type plants, which flower later thanesd4 fca-1 and esd4 fve-1 mutants. FLC mRNA levels were also compared between wild-type plants and esd4 mutants. Although FLC is expressed at a low level in Ler wild-type plants, this was further reduced in esd4 mutant plants. Similar reductions in the level of FLC mRNA were also observed under short days (data not shown).

The reduction in FLC mRNA levels in esd4 may contribute to the early-flowering phenotype of the mutant. The repression of flowering by FLC is probably caused in part by reduced expression of the flowering-time genes SOC1 and FT (P.Suárez-López and G.Coupland, unpublished results) (Michaels and Amasino, 2001; Ohto et al., 2001). Therefore, whether the early flowering and reduced expression of FLC in genotypes containing esd4 was also associated with increased expression of SOC1 and FT was tested. Over a 24-hour long day cycle, FT and SOC1 mRNA abundance was compared between wild-type and esd4 plants, and betweenfca-1 and esd4 fca-1 plants.

In wild-type plants, FT showed the expected pattern of expression,with the main peak in mRNA abundance occurring between 12 and 16 hours after dawn (Fig. 5B)(Suárez-López et al.,2001). In esd4 single mutants, FT mRNA abundance was slightly higher than in wild-type plants(Fig. 5B). The fca-1mutation caused a reduction in the level of FT mRNA, so that a lower level peak was detectable at the same time as in wild-type plants. This reduction in FT expression is probably caused by increasedFLC expression in fca-1 mutants(Michaels and Amasino, 2001),and was partially suppressed by the esd4 mutation so that FTmRNA levels were increased in the esd4 fca-1 double mutant compared to fca-1, and similar to those of wild-type plants(Fig. 5B).

Similar observations were made with SOC1. In wild-type plantsSOC1 mRNA peaked around 8 hours after dawn, although the peak was of lower amplitude than that of FT(Fig. 5C)(Samach et al., 2000). Inesd4 single mutants, SOC1 expression was increased, and thefca-1 mutation caused a severe reduction in the level ofSOC1 mRNA, although a low level peak was still detectable at the same time as in wild-type plants (Fig. 5C). As for FT, the esd4 mutation partially suppressed the effect of fca-1 on SOC1 expression(Fig. 5C), so that the level ofSOC1 mRNA in the double mutant was similar to that of wild-type plants.

The increases in FT and SOC1 expression in esd4 fca-1 plants compared to fca-1 mutants are consistent with the proposal that esd4 causes earlier flowering by reducing FLCexpression and thereby increasing the expression of genes that are repressed by FLC.

To test at the molecular level whether esd4 influenced the activity of the long-day pathway, CO mRNA abundance was compared in wild-type and esd4 plants. CO acts downstream of many of the other long-day pathway genes and is not repressed by FLC, although its expression is increased in several early-flowering mutants or transgenic plants that affect the long-day pathway(Onouchi et al., 2000;Suárez-López et al.,2001). CO shows a diurnal pattern of expression, with the main peak of mRNA abundance occurring around 16 hours after dawn in long-day grown plants (Suárez-López et al., 2001). In wild-type plants, CO mRNA levels showed the expected pattern (Fig. 5D)(Suárez-López et al.,2001). No significant difference in either the diurnal expression pattern or the amplitude of expression of CO was observed in theesd4 mutant.

The location of esd4 on chromosome 4 indicates that it is not an allele of a previously described mutation causing early flowering. Theesd4 mutation is recessive and segregates as a single locus,suggesting that the role of the ESD4 gene is to delay flowering. However, the pleiotropic effects of the mutation suggest that ESD4also has a broader role in plant development.

The esd4 mutation has pleiotropic effects on the architecture of the shoot and on silique shape (Fig. 1). These pleiotropic effects of esd4 may be related to the early flowering of the mutant, perhaps due to ectopic expression of flowering time or floral meristem identity genes. Some of the previously described early flowering mutants show pleiotropic effects(Goodrich et al., 1997;Soppe et al., 1999;Gomez-Mena et al., 2001)whereas others do not (Scott et al.,1999; Michaels and Amasino,2001). Moreover, transgenes causing ectopic expression of flowering time genes can also result in developmental defects(Kardailsky et al., 1999;Kobayashi et al., 1999;Onouchi et al., 2000). Some of the mutations causing late flowering largely suppressed the effects ofesd4 on flowering time, shoot determinacy and floral phyllotaxy, but none of these mutations abolished the effect of esd4 on plant height or silique shape. It is therefore unlikely that all the pleiotropic effects ofesd4 can be explained by altered expression of flowering time genes,and therefore ESD4 probably plays a broader role in the regulation of plant development.

Models for the genetic control of flowering time in Arabidopsispropose that three independent genetic pathways promote flowering under long photoperiods (see Introduction) (reviewed byAraki, 2001;Mouradov et al., 2002;Simpson and Dean, 2002). Double mutants carrying esd4 and mutations previously shown to affect each of these three pathways were constructed to determine whetheresd4 promotes early flowering by acting through one or more of these pathways. No true epistatic relationships were identified betweenesd4 and mutations causing late flowering, although epistatic relationships have previously been reported between Arabidopsismutations causing early and late flowering(Yang et al., 1995;Soppe et al., 1999). However,mutations that affect the autonomous pathway, such as fca andfve, had only weak effects on the esd4 phenotype under long days. One interpretation of these results is that ESD4 acts within the autonomous pathway downstream of FCA and FVE(Fig. 6A). In this model FCA and FVE promote flowering by repressing the activity of ESD4, as was proposed previously for efs (Soppe et al.,1999).

Construction of double mutants demonstrated that under both long and short days mutations in the long-day pathway (co, ft and fwa)caused a severe delay in flowering of esd4 mutants in comparison to the effect of autonomous pathway mutations. A similar effect was observed for GA pathway mutations under long days. ESD4 could therefore act in the long-day pathway before the CO, FT and FWA genes to reduce their expression, or within the GA pathway to reduce GA signalling or synthesis. This is consistent with previous observations that overexpression of the long-day pathway gene CO from the 35S promoter causes early flowering and largely suppresses the late-flowering caused by fca(Onouchi et al., 2000), and that mutations in suppressors of GA signalling cause early flowering(Jacobsen and Olszewski,1993). However, no increase in CO mRNA abundance was detected in an esd4 mutant, and these mutant plants are not resistant to the GA biosynthesis inhibitor paclobutrazol (data not shown) nor do they show elongated internodes, which are effects characteristic of mutations such as spindly that cause GA-signal transduction independently of GA(Jacobsen and Olszewski,1993).

An alternative explanation for the partial suppression of mutations in the long-day and GA pathways by esd4 is that ESD4 acts predominantly in pathways related to the autonomous pathway(Fig. 6), and that increased activity of these pathways in esd4 mutants can partially suppress the effect of mutations in the long-day or GA pathways. The esd4 mutation would then be proposed to activate the autonomous pathway downstream ofFCA and FVE, because of the early flowering of esd4 fca and fve esd4 double mutants, or would somehow bypass the effect of these mutations on the autonomous pathway. The autonomous pathway appears to facilitate the action of the long day and GA pathways, but on its own to have only a weak floral promotion activity(Nilsson et al., 1998;Reeves and Coupland, 2001). Thus, if esd4 acts predominantly through the autonomous pathway(Fig. 6A), then mutations within other floral promotion pathways would be expected to reduce the severity of the esd4 phenotype. This reduction in severity ofesd4 by mutations in the long-day and GA pathways did occur, but not for ga1 under short days, suggesting that under these conditions there may be a closer relationship between esd4 and the GA pathway. Nevertheless, we propose that the effect of ESD4 on flowering time is most closely associated with the autonomous pathway.

The esd4 mutant flowers later under short than long days, and is therefore still responsive to daylength. In wild-type plants the daylength response is conferred by the long-day pathway. Therefore, esd4mutants that are mainly affected in the autonomous pathway would be expected to retain a response to daylength. Such an argument can also explain whyesd4 suppresses mutations that impair the autonomous pathway much less effectively under short-day conditions, because the activity of the long-day pathway would be reduced under these conditions and this would further delay flowering of esd4 fca-1 plants. The esd4 co-2double mutant is only slightly later flowering under short days compared to long days, consistent with the long-day pathway not having a strongly promotive effect on flowering under short days even in an esd4mutant. The co mutation had a similarly weak effect on the flowering time of the efs mutant under short days(Soppe et al., 1999).

Mutations within the autonomous pathway cause an increase in the expression of the floral repressor FLC (Michaels and Amasino, 1999;Sheldon et al., 1999). To test the genetic model of ESD4 function, the effect of esd4 on the regulation of FLC mRNA was examined. FLC mRNA abundance was reduced in esd4 compared to wild-type plants, and in esd4 fca-1 and esd4 fve-1 double mutants compared to the late flowering mutants. This suggests that ESD4 may act to delay flowering by increasing the abundance of FLC mRNA. However, FLC mRNA levels are still higher in esd4 fca-1 double mutants than in wild-type plants, although the esd4 fca-1 plants flowered earlier than wild type. This indicates that ESD4 is unlikely to act solely through FLC. However, we cannot rule out the possibility thatesd4 may reduce FLC mRNA abundance to a very low level in a small subset of cells that are critical for the regulation of flowering or that ESD4 may have an additional post-transcriptional effect on FLC protein. Nevertheless, the observation that flc null mutants do not flower as early as esd4 mutants under short days(Michaels and Amasino, 2001),supports our view that the regulation of FLC is not the only role forESD4 in the regulation of flowering time.

Although the reduction in FLC mRNA levels is not the only cause of the esd4 phenotype, it is likely to contribute to the early flowering of esd4 mutants and the suppression of the late flowering phenotype of autonomous pathway mutations. Therefore we examined the effect ofesd4 on FT and SOC1, two flowering time genes whose expression levels are repressed by high levels of FLC (P. Suárez-López and G. Coupland, unpublished results)(Michaels and Amasino, 2001;Ohto et al., 2001). Inesd4 plants, an increase in FT and SOC1 mRNA levels relative to wild-type was observed, suggesting that these genes may contribute to the early flowering of esd4. esd4 partially suppressed the reduction in FT and SOC1 mRNA caused by the fca-1mutation, consistent with the partial reduction in FLC mRNA levels observed in the esd4 fca-1 genotype. In esd4 fca-1 plants the levels of FT and SOC1 were similar to those of wild-type plants. This suggests that it is unlikely that the increase in FT andSOC1 expression alone explains the effect of esd4 on flowering time, as esd4 fca-1 plants flower earlier than wild-type despite having similar levels of FT and SOC1. Furthermore,it is unlikely that the effects on FT and SOC1 are mediated solely through FLC as esd4 fca-1 have higher levels ofFLC mRNA compared to wild type. Thus, the genetic and molecular data indicate that ESD4 has a role in increasing FLC, and consequently decreasing FT and SOC1 expression, but also that it has additional functions in the control of flowering.

Other genes have previously been shown to increase the level ofFLC. For example, the FRIGIDA gene, which is responsible for the vernalization requirement of many naturally occurring winter varieties ofArabidopsis promotes FLC expression(Michaels and Amasino, 1999a;Sheldon et al., 1999;Johanson et al., 2000).ESD4 may therefore act together with FRI to promoteFLC. However, esd4 was isolated in the Landsbergerecta ecotype, which lacks an active FRI allele(Johanson et al., 2000). Therefore, it is unlikely that the early flowering of esd4 is due to effects on FRI. The vernalization response also acts throughFLC and mutations that impair vernalization cause increasedFLC mRNA levels (Chandler et al.,1996; Sheldon et al.,1999). Moreover, vernalization does not act solely throughFLC (Michaels and Amasino,2001). ESD4 may therefore act to antagonize the activity of genes that are required to promote flowering in response to vernalization,such as the VRN genes (Chandler et al., 1996; Gendall et al.,2001). In this case loss of esd4 would result in upregulation of their activity, in effect resulting in esd4 behaving as a constitutively vernalized mutant as has been proposed for the early flowering hos1 mutant (Lee et al., 2001). However, vernalization appears more effective in repressing FLC than loss of ESD4 function(Michaels and Amasino, 1999a;Sheldon et al., 1999). The identification and cloning of additional genes that control the vernalization response should enable the interaction of ESD4 with the vernalization pathway to be examined in more detail.

We propose that ESD4 is closely associated with the regulation ofFLC within the autonomous pathway. Mutations in ESD4 promote flowering at least partly by decreasing FLC mRNA levels and thereby increasing expression of downstream flowering-time genes such as FTand SOC1. Increases in FT and SOC1 expression would be expected to lead to earlier expression of floral meristem identity genes,such as LEAFY and APETALA1(Ruiz-Garcia et al., 1997;Kardailsky et al., 1999;Kobayashi et al., 1999). However, the late flowering phenotype of autonomous pathway mutations such asfca and fve is completely dependent on functionalFLC: fca and fve mutants do not flower any later than wild type in a flc null mutant background(Michaels and Amasino, 2001),whereas the early flowering of esd4 cannot be explained solely by its effects on FLC. We therefore propose that ESD4 has an additional role in a pathway that by-passes the requirement for the autonomous pathway(Fig. 6B). In this model,ESD4 has two roles. The first is to ensure high FLCexpression that leads to repression of flowering through repressingFT and SOC1, and probably other genes (represented by X inFig. 6B). The second is to regulate flowering-time genes independently of FLC (represented by Y inFig. 6B). This model is consistent with the analysis of gene expression in esd4 mutants as well as with the single and double mutant phenotypes under long and short days.

We thank Kim Findlay for her guidance in performing the scanning electron microscopy. We also thank P. Suárez-López for her guidance in performing the RT-PCR reactions and for sharing unpublished data on the effect of flowering-time mutants on the expression pattern of FT. P. R. was supported by a BBSRC studentship, G. M. by an EC fellowship and the work was supported by a BBSRC research grant within the CAD initiative and an HFSPO grant to G. C.