In plants of Sinapis alba induced to flower by one long day, the MADS box gene, SaMADS A, is expressed initially in the central corpus (L3 cells) of the shoot apical meristem (SAM), about 1.5–2 days before initiation of the first floral meristem. We have combined a physiological approach by testing the effects of three putative floral signals on SaMADS A expression in the SAM of S. alba plants with a transgenic approach using Arabidopsis thaliana plants. A single application of a low dose of a cytokinin or a gibberellin to the apex of vegetative S. alba plants is capable of mimicking perfectly the initial effect of the long day on SaMADS A transcription. A treatment combining the two hormones causes the same activation but seems to enhance the level of SaMADS A expression. A sucrose application to the apex of vegetative plants is, on the contrary, unable to activate SaMADS A expression. None of these chemicals, alone or combined, is capable of causing the floral shift at the SAM. Since the constitutive expression of SaMADS A leads to precocious flowering in A. thaliana and antisense expression of a fragment of the A. thaliana homologue AGL20 leads to a delay in flowering time, these results are consistent with SaMADS A activation being an intermediate event in a cytokinin- and/or gibberellin-triggered signal transduction pathway that is involved in the regulation of floral transition in S. alba.
Plant development is controlled by both environmental and endogenous factors. In photoperiodic plants, exposure of leaves to a day length favourable for flowering results in the production and transmission by these leaves of one or several floral signals, collectively called ‘the floral stimulus’. Arrival of these signals at the shoot apical meristem (SAM) causes cessation of leaf formation and start of flower initiation. Molecular and genetic studies have shown that this morphogenetic switch is governed by the activation of interacting groups of genes, including for instance genes such as CO ( Simon et al., 1996 ) and FPF1 expressed early in the switching SAM ( Kania et al., 1997 ) and genes such as LFY and AP1 expressed later and responsible for floral meristem identity ( Ma, 1998; Pidkowich et al., 1999 ). So far, studies attempting to relate the arrival at the SAM of particular floral signals of leaf origin with activation of some of these genes are scarce. This is because the nature of the transmissible floral signals is still a controversial issue ( Bernier et al., 1993 ), and thus the times of arrival at the SAM of these signals are often entirely unknown.
In Arabidopsis thaliana, there is evidence that a gibberellin (GA) is one of the primary floral signals ( Koornneef et al., 1998 ; Levy and Dean, 1998). In this species, it was indeed found that flowering of mutants deficient in GAs is delayed under long days or even suppressed under short days unless GA3 is applied ( Wilson et al., 1992 ). Activation of GA-mediated signal transduction pathways, such as in plants with a mutation in SPY or over-expressing FPF1, leads to early flowering and other phenotypic changes that mimic the effects of GA application ( Jacobsen and Olszewski, 1993 ; Kania et al., 1997 ). Moreover, exogenous or endogenous GAs activate in the SAM of A. thaliana the expression of LFY, one of the major genes responsible for floral meristem identity ( Blázquez et al., 1998 ), through an identified cis element in the LFY promoter that resembles the consensus binding site for some MYB transcription factors ( Blázquez and Weigel, 2000). Up-regulation of LtGAMYB in the shoot apex of Lolium temulentum at floral transition is also believed to be caused by a GA signal originating in leaves ( Gocal et al., 1999 ).
Another floral signal in A. thaliana is sucrose, as shown by observations that the floral response to an inductive photoperiodic treatment is severely impaired in conditions where leaves are unable to provide an extra amount of sucrose to the SAM ( Corbesier et al., 1998 ). Sucrose can hasten considerably the flowering time of two late cold-requiring ecotypes and several late-flowering mutants ( Roldán et al., 1999 ), and it has been shown that sucrose acts in concert with a GA to activate LFY expression ( Blázquez et al., 1998 ).
In Sinapis alba, a long-day (LD) plant that can be induced to flower by exposure to a single long day, the timing of movement of the floral stimulus from leaves to SAM has been determined ( Bernier, 1989), and two putative floral signals, sucrose and a cytokinin, have been identified ( Bernier et al., 1993 ). For each of these signals, the timing of their increased flux from induced leaves to the SAM was determined ( Lejeune et al., 1993 ; Lejeune et al., 1994 ). Also, although the application to plants kept under non-inductive short days (SDs) of either sucrose or a cytokinin fails to cause flowering, each of these compounds causes several cellular and metabolic events in the SAM that are typically observed after floral induction by one long day ( Bernier et al., 1993 ). Whether a GA is also a floral signal in S. alba remains an unresolved question. Application of a GA under SDs or of a GA biosynthesis inhibitor under LDs yielded inconclusive results ( Bernier, 1969). However, FPF1, a gene involved in a GA-mediated signal transduction pathway, is expressed early in the S. alba SAM in response to floral induction ( Kania et al., 1997 ), suggesting that increased GA sensing is a component of the floral transition process in this species, as in A. thaliana (see above).
We have started to investigate the effects of these three putative floral signals on the transcription of genes involved in the SAM floral transition in S. alba. Genes expressed early during this transition are presumably the best candidates as targets for individual floral signals. One such gene, SaMADS A, was identified as being expressed in the SAM of S. alba well before initiation of the first floral meristem ( Menzel et al., 1996 ). SaMADS A protein presents a high degree of similarity to AGL20 (now renamed SOC1, Samach et al., 2000 ) of A. thaliana, with 100% similarity between the MADS box domains and 95.3% similarity between the K box domains. In this paper, we show that the constitutive expression of SaMADS A in A. thaliana leads to precocious flowering under SD and LD conditions and that the constitutive expression of an AGL20 fragment in the antisense orientation leads to a delay in flowering. Recent work, published after submission of this paper, has shown that AGL20, which is a target gene of CO, has a major role in promoting flowering in A. thaliana ( Onouchi et al., 2000 ; Samach et al., 2000 ). AGL20 was proposed by Samach et al. (2000) to be part of a transcriptional cascade within the SAM that results in activation of LFY and other floral meristem identity genes. In the present paper, we also report that a single cytokinin or gibberellin application, but not a sucrose application, does indeed activate the transcription of SaMADS A in the SAM of S. alba plants kept under non-inductive SD conditions, in a way that mimics perfectly the initial activation caused by an inductive LD. It is suggested that SaMADS A acts within a signal transduction pathway involved in the SAM floral transition and activated by a cytokinin and/or a gibberellin.
Transcription of SaMADS A in S. alba in response to floral induction caused by one long day
The work of Menzel et al. (1996) showed early activation of SaMADS A in the SAM of S. alba plants induced to flower by continuous 16 h LDs. Since the timing of movement of the floral stimulus and putative floral signals of S. alba were determined in an inductive system involving a single 22 h LD, it was first necessary to establish the spatial and temporal transcription of this gene in this particular inductive system.
In situ hybridizations were performed with longitudinal sections of apical buds of LD-induced plants collected from 24 to 72 h after the start of the LD, using 35S-labelled SaMADS A antisense RNA as a probe. SaMADS A was not expressed in the SAM of vegetative plants continuously grown under SD conditions ( Figure 1a). The first transcription of SaMADS A was detected in the SAM of some, but not all, plants 24 h after start of the LD, i.e. 16 h after the start of the photo-extension period of this LD; transcription was present in all plants collected at 32 h, that is about 1.5–2 days before initiation of the first floral meristem which occurs in most plants at 66–80 h after start of the LD ( Figure 1b). As described by Menzel et al. (1996) , transcription was initially localized to the central zone only. In fact, we observed that transcription started in a few cells deeply localized in the central corpus (just above the rib meristem) and then spread to the whole central zone, except the superficial two layers (L1 and L2) and also the outermost layer of L3 in some SAMs ( Figure 1b). At 56 h, about 0.5–1 day before initiation of the first floral meristem, RNA transcripts were more abundant in the central zone (although still excluded from L1 and L2 but no longer from the outermost layer of L3) and were now seen in the cells of the rib meristem just below the central zone ( Figure 1c). There was clearly no transcription in the peripheral zone. At 72 h, when the first floral meristem was emerging from one flank of the inflorescence meristem, the hybridization signal had spread, although at a reduced level, to the peripheral zone ( Figure 1d), as previously observed by Menzel et al. (1996) . Again we found that transcription was absent in the L1 and L2 cell layers of both the central and peripheral zones ( Figure 1d). There was no transcription in nascent floral meristems (left flank of the SAM in Figure 1d). In control experiments with the SaMADS A sense probe, there was no hybridization signal above background level ( Figure 1e).
Transcription of SaMADS A in S. alba caused by a cytokinin or a gibberellin under non-inductive SD conditions
Two-month-old plants, kept vegetative by growth under continuous SDs , were treated once with solutions of the three putative floral signals applied directly to the apical bud. Solution concentrations were 2% for sucrose (SUC), and 5 × 10−5m for the cytokinin benzylaminopurine (BAP) and for gibberellin A3 (GA3). The concentrations of applied solutions were chosen on the basis of previous studies ( Bernier et al., 1990 ; F. Bonhomme et al., unpublished data) . Times of application were derived from previous knowledge about the timings of movement of the floral signals in S. alba plants induced by one long day. SUC levels started to increase in the phloem sap, reaching the apical bud and the SAM itself extremely early during the LD, in fact at 10 h after its start ( Bodson and Outlaw, 1985; Lejeune et al., 1993 ). SUC was apparently a fast-moving component of the floral stimulus since the slowest components moved from the leaves to apex only 8–10 h later, as indicated by defoliation experiments ( Bernier, 1989). Accordingly, and taking into account that there is a lag period of 4–5 h between application of a chemical and its arrival at the SAM (see Experimental procedures), the times of solution application to the bud were 6 h after the start of the experimental SD for SUC and 16 h after the start of this SD for BAP and GA3. None of the applied compounds caused flowering. SaMADS A transcription was investigated at 16 (SUC only), 24, 32, 48, 56 and 72 h after the start of the experimental SD.
There were no SaMADS A transcripts detectable in the SAM of SUC-treated plants throughout the investigated period ( Figure 1f).
SaMADS A transcripts were first detectable in the SAM of all BAP-treated plants 8 h after the application of the chemical, i.e. 24 h after start of the experimental SD ( Figure 1g). Transcripts were localized in the central zone, except in L1, L2 and the outermost layer of L3, as well as in the cells of the rib meristem just below the central zone ( Figure 1g). Transcript localization was identical to that initially observed above in the SAM of LD-induced plants ( Figure 1b). Despite there being only a single BAP application, the signal was maintained in the central zone and rib meristem for at least up to 56 h after the application, i.e. 72 h after start of the experimental SD, but unlike in LD-induced plants it did not spread with time to the peripheral zone ( Figure 1h).
The effect of a GA3 application was identical to that of a BAP application: activation of SaMADS A transcription started 8 h after the GA3 treatment and was observed in exactly the same tissue ( Figure 1i). As for BAP, SaMADS A activation caused by GA3 was also long-lasting, and did not spread with time to the peripheral zone ( Figure 1j).
Cytokinin and gibberellin
As BAP and GA3 were capable of activating SaMADS A expression independently of each other, it was decided to analyse the effect of a combination of these two compounds. A mixture of 5 × 10−5m BAP and 5 × 10−5m GA3 was applied once to the apical bud of SD-grown plants 16 h after the start of the experimental SD. The combined treatment caused an activation of SaMADS A transcription with the same timing and localization as those activations produced by the individual treatments, but the hybridization signal was somewhat stronger, indicating that the two hormones possibly had additive effects ( Figure 1k, compare with Figure 1g,i).
Modulation of the time to flowering in transgenic A. thaliana plants
A. thaliana is a facultative long-day plant, flowering earlier under long-day than under short-day conditions ( Koornneef et al., 1998 ; Levy and Dean, 1998; Piñeiro and Coupland, 1998). To analyse whether the SaMADS A gene from S. alba and the homologous gene from A. thaliana are involved in the control of the floral transition, we have generated transgenic A. thaliana plants that over-express the coding region of SaMADS A in the sense orientation or an AGL20 fragment in the antisense orientation under the control of the 35S CaMV promoter. We obtained six kanamycin-resistant lines with the SaMADS A coding region in the sense orientation and have selected 88 transgenic Arabidopsis lines with an AGL20 antisense construct.
Expression of the SaMADS A transgenes was monitored by Northern blotting. As shown in Figure 2, all over-expressing lines showed expression of the transgene; however, lines S1–S3 showed only weak expression compared to the RNA level of the transgene in lines S4–S6. In wild-type plants, no hybridization could be observed ( Figure 2). To analyse phenotypic changes in SaMADS A over-expressing plants, these were grown in parallel with wild-type plants in the same environment. Seedlings at the cotyledon stage appeared normal, and germination and development up to the four-leaf stage showed the same duration as in wild-type plants. From this stage onwards, the duration of the following phases in some lines was remarkably diminished, which led to precocious bolting and earlier flowering ( Figure 3). The transgenic lines S1–S3, which showed only weak expression of the SaMADS A transgene, did not flower earlier than wild-type plants under LD conditions ( Figure 4), whereas the lines S4–S6 showed a significant reduction of the vegetative phase under these conditions. Under SD conditions, the time to flowering was also remarkably reduced in lines S1–S3, but was even shorter in lines S4–S6 ( Figure 4). The transgenic line S6 showed the most dramatic reduction of the vegetative phase. The time to flowering was reduced to 52 days, compared to 86 days in wild-type plants under SD conditions. The decrease in leaf number by about 50% is mainly caused by a decrease of rosette leaves, and the number of cauline leaves is only slightly reduced.
Most of the 88 AGL20 antisense lines showed a slight delay in flowering, whereas several lines showed a markedly late flowering time compared to wild-type plants both under SD and LD conditions. From the late-flowering lines, eight were selected and further analysed. From homozygous kanamycin-resistant plants, the amount of RNA from the antisense transgenes was determined by Northern blotting. Very strong expression of the antisense transgene was seen in five lines (A10, A11, A25, A40 and A 75) and very low expression in the other selected lines (data not shown). With the exception of line A75, the lines with the strongest expression of the antisense construct were the lines that flowered latest under LD conditions, with line A40 flowering 13 days later than the wild-type plants ( Figure 4). However, under SD conditions, the flowering time was not strictly correlated with the transgene expression. Under these conditions, line A40 was also the latest line to flower, flowering 17 days later than the wild-type plants. Floral development was not affected in the transgenic plants: flowers developed normally, were fertile and the number of seeds per silique was not altered.
Evidence that SaMADS A is involved in flowering time control
SaMADS A and SaMADS B are the earliest up-regulated MADS box genes from S. alba that are expressed in the SAM upon floral induction ( Menzel et al., 1996 ). The homologous genes to SaMADS A and B from A. thaliana, AGL20 and AGL8, show a similar pattern of expression in A. thaliana (Borner et al., unpublished results ; Mandel and Yanofsky, 1995). In this paper, we have shown that the over-expression of SaMADS A in A. thaliana is sufficient to shorten the vegetative phase of transgenic plants. More direct evidence that the gene is involved in flowering time control comes from transgenic approaches with an antisense construct of the homologous gene AGL20 in A. thaliana, which leads to late-flowering phenotypes. Moreover, we have shown that the over-expression of AGL20 also causes very early flowering in transgenic A. thaliana plants, confirming similar data from Samach et al. (2000) . We have also found that transposon-induced mutations in AGL20 cause late flowering in null mutants (Borner et al., unpublished results) . Therefore, the results with S. alba and A. thaliana are consistent with the assumption that SaMADS A and AGL20 play similar and important roles in the floral transition in these two species.
Transcription of SaMADS A in S. alba in response to floral induction caused by one long day
Our analysis of SaMADS A gene transcription during the floral transition caused in S. alba by one long day has generally confirmed the observations of Menzel et al. (1996) . Thus, activation of SaMADS A occurs well before initiation of the first floral meristem, and this activation is initially restricted to the central zone of the SAM. It spreads to the rib meristem and peripheral zone only at later times. A distinctive observation of our study is that the activation of SaMADS A occurs only in L3 cells, not in the L1 and L2 cells of the SAM. This was not reported by Menzel et al. (1996) , perhaps because they induced their plants to flower by continuous LDs , i.e. an inductive treatment much stronger than our single LD. Indeed, in plants of S. alba induced by successive LDs, we have observed that activation of this gene eventually occurs in L1 and L2 cells (data not shown). However, our results show that the floral transition can proceed when SaMADS A is activated only in L3 cells.
The initial activation of SaMADS A in the central L3 cells or corpus and its later activation in other zones of the SAM are interesting. The central corpus is known to consist of relatively undifferentiated cells, including a small number of self-maintaining pluripotent stem (initial) cells ( Evans and Barton, 1997). These cells replenish continuously both the peripheral zone, which produces the leaf or flower primordia, and the rib meristem, which gives rise to the pith of the shoot axis. It is tempting to speculate that SaMADS A could initially switch on at least part of the floral morphogenetic programme in the undifferentiated cells of the central corpus and then this new programme could extend progressively to the surrounding zones through cell divisions.
Transcription of SaMADS A in S. alba caused by a cytokinin or a gibberellin under non-inductive SD conditions
SaMADS A activation under non-inductive SD conditions may be achieved by applying to the apex a single low dose of a cytokinin, a GA or a combination of these two hormones, but not by a sucrose application. Activation by the cytokinin or the GA is restricted to the central L3 cells (corpus) and rib meristem cells, just as in the initial steps of activation in LD-induced plants. Activation is first detectable in all cases 8 h after hormone application, and since the application was made 16 h after the start of the experimental SD, activation of SaMADS A is first detectable at 24 h after the start of the experimental SD. Activation of SaMADS A expression in LD-induced plants also started 24 h after the start of the LD. Thus, relative to the start of the light period, the timings of activation caused by the LD or the hormones are identical. Thus, both spatially and temporally, each hormone is capable of mimicking almost perfectly the initial effect of the LD on SaMADS A transcription. Interestingly, when the two hormones are applied together, the localization and timing of SaMADS A activation remain unchanged but activation is apparently greater than after application of the individual hormones. These observations are consistent with SaMADS A activation being an intermediate event in a signal transduction pathway activated by a cytokinin and/or a GA and involved in the floral transition of the S. alba SAM.
Activation of SaMADS A is consistently detectable 8 h after hormone applications. As it takes compounds applied to the apical bud 4–5 h to reach the SAM itself (see Experimental procedures), we infer that SaMADS A transcription is activated within 3–4 h following arrival of the hormones at the SAM. Similar time ranges have been reported for activation of other genes by cytokinins or GAs ( Davies, 1995; Schmülling et al., 1997 ).
Despite there being only a single application of the hormones, the activation of SaMADS A is long-lasting, persisting for at least 56 h after application. The late spreading of this activation to the peripheral zone of the SAM, observed after one LD, does not occur after the hormone applications under SD conditions, and this might be one of the reasons why these hormones do not cause the floral transition under SD conditions . Presumably, the late spreading process under LD conditions is dependent on the activation (or deactivation) of other genes.
The above results strongly support the idea that a cytokinin and a GA are floral signals in S. alba. As far as cytokinins are concerned, previous analyses of (a) the changes in their endogenous levels and fluxes in plants induced by a long day, and (b) the effects of their application to the bud of non-induced plants, have provided strong evidence that this hormone family participates in the control of the floral transition in this species (summarized in Bernier et al., 1993 ). Dewitte et al. (1999) have reported that the cytokinin levels in the SAM of tobacco decrease markedly at the floral transition. This result does not disprove the above conclusion concerning S. alba as it is well known, from both in vivo and in vitro studies, that if flowering is to occur, there is a permissive range of cytokinin concentrations in many species, with too low or too high concentrations being inhibitory depending on the plant species and age and prevailing growing conditions (reviewed in Bernier et al., 1981 ; Bernier, 1988). Thus, the direction of changes in cytokinin levels at floral transition may be opposite in different experimental systems.
In the case of the GAs, evidence for their participation in the control of the floral transition in S. alba is based on the observation of Kania et al. (1997) that FPF1, a gene involved in a GA-dependent signalling pathway, is activated in the SAM at an early step of the floral transition, well in advance of initiation of the first floral meristem. Thus, GA sensing appears to be implicated in the SAM transition in S. alba, just as in A. thaliana (see Introduction). In A. thaliana, an increased GA level is a critical factor in the control of the floral transition ( Blázquez and Weigel, 2000; Blázquez et al., 1998 ; Levy and Dean, 1998). Whether a similar situation prevails in S. alba is unknown as the levels and fluxes of GAs at floral transition in this species have not yet been investigated.
Interestingly, SaMADS A transcription is activated by two different kinds of hormones, and the effects of these hormones are apparently additive. This is a situation frequently encountered in the field of plant hormones which are known to substitute for each other in many cases and also often act in concert ( Crowell and Amasino, 1994; Davies, 1995; Izhaki et al., 1996 ). As far as the expression of a gene such as SaMADS A is concerned, a possibility is that the transduction pathways triggered by the two hormones partially overlap. This may happen when a pathway crosses another at some point ( Genoud and Métraux, 1999) or when one hormone is capable of altering the level of the other hormone ( Davies, 1995). Alternatively, the transduction pathways may be independent but finally affect the same target gene.
S. alba plant materials
Plants of S. alba were kept vegetative by growth under continuous 8 h SDs (light period: 08:30–16:30 h) as described by Lejeune et al. (1988) . The irradiance at the top of the plants was 150 µmol m−2 sec−1 and was provided by very-high-output Sylvania fluorescent white tubes (Zaventem, Belgium). When the plants were 2 months old, they were exposed to the following treatments:
1maintained under SD conditions;
2induced to flower by exposure to one 22 h LD (08:30–06:30 h) and then returned to SD conditions;
3maintained under SD conditions and treated once with an aqueous solution of 2% sucrose (SUC) applied directly to the apical bud at 14.30 h, i.e. 6 h after the start of the experimental SD;
4maintained under SD conditions and treated once with an aqueous solution of 5 × 10−5m benzylaminopurine (BAP) (Sigma B-5898) applied directly to the apical bud 16 h after the start of the experimental SD, i.e. at 00:30 h;
5maintained under SD conditions and treated once with an aqueous solution of 5 × 10−5m gibberellin A3 (GA3) (Sigma G-7645) applied directly to the apical bud 16 h after the start of the experimental SD, i.e. at 00:30 h; or
6maintained under SD conditions and receiving a combination of the two previous treatments.
Time was computed from the start of the experimental SD or LD. All experiments were repeated 2–4 times. Apical buds were collected at various time intervals after the start of the experimental SD or LD. Aqueous solutions of the chemicals were applied to a small cotton plug inserted between the young leaves of the apical bud. An amount of about 0.3 ml was applied to each bud and 0.1% Tween-20 was added to all solutions. Tween-20 did not affect floral induction ( Gonthier, 1987). Experiments using 3H-thymidine or red eosine applied in the same way allowed estimation of the lapse of time required for an applied chemical to reach the SAM itself (4–5 h) ( Gonthier, 1987).
In situ hybridization
Shoot apical buds of S. alba plants were fixed in 2% formaldehyde in 100 m m phosphate buffer, pH 7.2 (16 h at 4°C). Fixed tissues were dehydrated and embedded in paraffin according to the procedure described by Bonhomme et al. (1997) . Longitudinal sections (8 µm) were mounted onto poly- l-lysine-coated slides. 35S-UTP-labelled antisense and sense riboprobes were synthesized from truncated cDNA without the MADS box and poly(A) tail, and inserted in the SmaI site of pBS SKII+ (Stratagene), using T3 or T7 RNA polymerase according to the manufacturer's instructions (Promega, Madison, WI, USA). The labelled RNA probes were hydrolysed to give fragments with an average length of 150 nucleotides. Pre-treatments, hybridizations and washes were performed as previously described ( Bonhomme et al., 1997 ). Slides were exposed for 8 weeks. Four to eight SAMs (from two to four independent experiments) were examined for each experimental time point.
RNA was isolated from two- to four-leaf stage seedlings of A. thaliana ground in liquid nitrogen and stored at −80°C according to Kania et al. (1997) . For Northern blot analyses, 20 µg RNA per sample were separated on an agarose gel containing formaldehyde. After blotting to a GeneScreen membrane (NEN, Boston, MA, USA), the RNA was fixed to the membrane by baking at 80°C under vacuum for 1 h.
Hybridization was performed overnight at 65–68°C in 1.5 × SSPE, 7% SDS, 10% PEG 8000, 100 µg ml−1 salmon sperm DNA and 250 µg ml−1 heparin, using an SaMADS A cDNA fragment without the MADS box as a probe ( Menzel et al., 1996 ). Filters were washed with 0.1 × SSC/0.1% SDS and exposed to Kodak X-Omat AR or BioMax MR-1 films (Eastman Kodak, Rochester, NY , USA) for 5 days.
Transgenic approaches in A. thaliana
For over-expression of the coding region of SaMADS A in A. thaliana under the control of the 35S CaMV promoter, the EcoRI fragment of the SaMADS A cDNA ( Menzel et al., 1996 ) was inserted into the pRT104 vector ( Töpfer et al., 1987 ) and transformed into E. coli DH5α cells. After partial digestion with HindIII, a fragment containing the expression cassette was isolated and inserted into the BIN19 vector ( Bevan, 1984). The recombinant vector pBIN19–SaMADS A was transformed into the Agrobacterium tumefaciens strain C58C1 by a standard transformation procedure ( Höfgen and Willmitzer, 1988). The obtained recombinant bacteria were used to transform A. thaliana ecotype Columbia by in planta transformation ( Bechtold et al., 1993 ). After in planta transformation, plants were grown in a growth chamber under LD conditions. Plant seeds were harvested 5 weeks after infiltration, sterilized, and plated on germination medium containing 500 mg l−1 timenten to suppress growth of Agrobacterium and 50 mg l−1 kanamycin for selection of transformants. Kanamycin-resistant transformants were transferred to soil, seeds were harvested, and homozygous lines were selected with kanamycin.
For the antisense approach, an EcoRI fragment of an AGL20 cDNA which does not include the MADS box (Borner et al., unpublished results) , was cloned in the pRT104 vector in antisense orientation and further treated as described for the SaMADS A over-expression.
For analysis of the transgenic lines, seeds were sown on soil, singled out in lawn soil, and grown in climate chambers at 20°C under fluorescent tubes emitting a photon flux density of 160 µmol m−2 sec−1. Illumination time was 8 h for SD conditions and 16 h for LD conditions. The time to flowering was measured in days until the opening of the first flower and by total leaf number .
We thank H. Sommer, R.F. Lyndon, F. Cremer and J. Chandler for critical reading of the manuscript, and A. Pieltain and N. Detry for technical assistance. This work was supported by a grant from the Fonds Spéciaux (1994–1997) of the University of Liège and Pôles d'Attraction Interuniversitaires Belges (Service du Premier Ministre, Services Fédéraux des Affaires Scientifiques, Techniques et Culturelles, P4/15). F.B. was a research assistant of the ‘Fonds National de la Recherche Scientifique’ of Belgium during part of this work.