Activation of a Floral Homeotic Gene in Arabidopsis

See allHide authors and affiliations

Science  23 Jul 1999:
Vol. 285, Issue 5427, pp. 585-587
DOI: 10.1126/science.285.5427.585


The patterned expression of floral homeotic genes inArabidopsis depends on the earlier action of meristem-identity genes such as LEAFY, which encodes a transcription factor that determines whether a meristem will generate flowers instead of leaves and shoots. The LEAFY protein, which is expressed throughout the flower, participates in the activation of homeotic genes, which are expressed in specific regions of the flower. Analysis of a LEAFY-responsive enhancer in the homeotic geneAGAMOUS indicates that direct interaction of LEAFY with this enhancer is required for its activity in plants. Thus, LEAFY is a direct upstream regulator of floral homeotic genes.

The appearance of individualArabidopsis flowers is marked by the expression of several floral meristem-identity genes, including LEAFY(LFY), APETALA1 (AP1), and CAULIFLOWER. Inactivation and overexpression of these genes cause opposite phenotypes, indicating that they act as developmental switches (1). A particularly dramatic loss-of-function phenotype is seen in lfy mutants, in which early arising flowers are replaced by leaves with associated shoots (2, 3). About 2 days after an individual flower has been initiated, three classes of homeotic genes, A, B, and C, are expressed in distinct, overlapping patterns within the developing flower. Both loss- and gain-of-function studies have shown that the patterned activities of ABC genes are responsible for the stereotypic sequence of sepals, petals, stamens, and carpels that form from the periphery to the center of the flower (4). InArabidopsis, C function is specified by the MADS box geneAGAMOUS (AG), which confers stamen and carpel identity and which is normally expressed in the center of the flower (5, 6).

Although the function of LFY as a switch for floral identity indicates that LFY is formally an upstream regulator ofAG, the effects of lfy mutations on AGexpression are complex. Not all flowers of strong lfymutants are replaced by leaves and shoots, and although AGRNA expression is delayed in the more flowerlike structures oflfy mutants, it eventually reaches levels that are similar to those in wild-type flowers (7). In addition,AG is ectopically expressed in later-developing flowers as well as in the stem of lfy mutants, indicating not only a positive but also a negative role of LFY in AGregulation. Similarly, the snapdragon LFY orthologFLORICAULA (FLO) has both positive and negative roles in the regulation of the AG ortholog PLENA(PLE) (8, 9). That FLO can activate PLE nonautonomously suggests that the effects ofFLO and LFY on PLE and AGneed not be direct (8).

Control of floral identity by LFY can be genetically separated from a role of LFY in the later process of ABC gene regulation (10). Genetic arguments also suggested that the primary effect of LFY on AG is activation (10). In plants that expressed an activated form of LFY protein, LFY:VP16, AG RNA was detected earlier, ectopically, and at elevated levels when compared with wild-type flowers, suggesting that LFY normally interacts with region-specific coregulators that restrict AG expression to a subset of LFY-expressing cells (10).

To dissect the interaction between LFY and AG, we identified enhancers responsible for transcriptional activation ofAG using transgenic plants that expressed the bacterial β-glucuronidase (GUS) gene under the control of various AGsequences. The second intron of AG, whose location largely coincides with a 3-kb Hind III restriction fragment, is required forAG-typical expression of a chimeric gene that encodes a translational fusion of AG to GUS (11). This Hind III fragment (12) contains transcriptional enhancers that are both necessary and sufficient for the wild-type AGexpression pattern, as demonstrated by placing this fragment upstream of a minimal heterologous promoter linked to GUS (reporter KB9; Figs. 1 and 2B). The KB9 reporter responded to changes in LFY activity in the same way as endogenous AG, indicating that it contains important LFY-responsive enhancer sequences. The onset of GUS expression was delayed in strong lfy mutants, whereas GUS was expressed earlier and ectopically in LFY:VP16 plants (Fig. 2, A to C) (13). The 3-kb Hind III fragment is likely to confer most, if not all, of the AG response toLFY, because a GUS fusion containing the AGpromoter, but not the large intron (11), does not respond toLFY:VP16 (14).

From deletion analysis of the 3-kb Hind III fragment, we identified at least two redundant enhancers that could drive GUS expression in young wild-type flowers in a pattern that was similar to that of GUS under the control of the full-length 3-kb fragment (Figs. 1 and 2). For example, the nonoverlapping fragments present in KB14 (5′ enhancer) and KB31 (3′ enhancer) both conferred GUS expression in the center of young flowers, resembling the pattern of endogenous AG (Fig. 2, E and N). The response of the various reporters to changes inLFY activity revealed several additional features ofAG regulation. First, there appear to be synergistic enhancers that mediate the action of LFY-independent activators of AG, because GUS expression of the smaller reporters was generally more sensitive to loss of LFYactivity than that of the full-length reporter (Fig. 2). Second, there appear to be cryptic regulatory elements that have overlapping roles in repressing AG in different regions of the flowers. For example, GUS was ectopically expressed in the outer whorls, but much less so in the floral stem or pedicel of LFY:VP16 flowers carrying the KB18 reporter (Fig. 2L). In contrast, ectopic GUS expression was more consistent in the pedicel than in the first whorl of LFY:VP16 flowers carrying KB14 (Fig. 2F).

Figure 1

AG enhancer dissection. (A) Diagram of reporter constructs carrying AGgenomic fragments in front of a minimal promoter driving GUS (19). Reporters discussed in the text or shown in Figs. 2and 3 are in bold. MX68 carries the AG I m1 andAG II m1 mutations. MX100 carries AG I m2 andAG II m2 (Fig. 4C). The first column to the right indicates the orientation of the fragment relative to the minimal promoter. F (forward) and R (reverse) indicate whether the 3′ or 5′ end was closest to the promoter. The second column shows the distribution of primary transformants that showed no (n), weak (w), intermediate (i), or strong (s) staining [see (B)]. The third column indicates whether and how strongly a representative line stained in an AG-typical pattern during stage 3 of flower development (n/d, not determined). Although many KB11 and K13 lines showed GUS expression, the early pattern was notAG-typical, but was throughout early flowers and the shoot apex, and in the stem. The fourth column indicates relative activity in a lfy-12 background. The fifth column indicates whether or not activity was increased in LFY:VP16. The last column indicates the number of lines analyzed for each construct. A restriction map is shown below. B, BamH I; Bc, Bcl I; H2, Hinc II; H3, Hind III; N, Nla IV; Sc, Sca I; Sn, SnaB I; Sp, Spe I; Ss, Ssp I; X, Xba I. (B) Examples of weak, intermediate, and strong GUS staining in whole apices of different reporter lines (from left to right, MX68 line 38; KB31 line VIIC1; KB31 line VIIC7). A mid–stage 3 flower (20) is indicated in each apex. Faint staining in the left apex is indicated by arrowheads.

Figure 2

(A toO) GUS expression in AG reporter lines KB9, KB14, KB21, KB18, and KB31. Apices were stained for GUS activity with X-gluc (5-bromo-4-chloro-indolyl-β-d-glucuronic acid), embedded, sectioned, and viewed under darkfield illumination. Weak staining appears orange, and strong staining pink to purple. Asterisks indicate shoot apical meristems, numbers indicate stages of flower development (20). Note failure of GUS expression to expand completely into the first whorl (arrowhead) of KB14 LFY:VP16in (F), and failure of GUS expression to expand completely into pedicels (p) of KB18 LFY:VP16 in (L). In contrast, GUS expression is obvious throughout the first whorl and pedicel in KB9LFY:VP16 in (C) and KB31 LFY:VP16 in (O). All panels are at the same magnification (×58).

We focused on the 3′ enhancer because the smallest 3′ fragment that was active on its own was considerably smaller than the minimal active element that we could define in the 5′ portion of the full-length reporter (Fig. 1). In addition, GUS activity was much reduced in older flowers of plants carrying reporters containing only the 3′ enhancer, which we thought would simplify the interpretation of expression patterns in mutant backgrounds.

As is not uncommon in enhancer analyses, simple deletions progressively reduced activity of the 3′ enhancer. Nevertheless, we could trackLFY responsiveness using LFY:VP16, which apparently bypasses the requirement for other positive factors. Several reporters that were either inactive or only weakly active in a wild-type background were reactivated when introduced intoLFY:VP16 (Fig. 3, A and B). We could thus map a LFY-response element to the overlap of AG sequences in KB24 and KB28, which share a common region of 230 base pairs (bp) (Fig. 1). Because LFY can bind to DNA (10), we investigated whether LFY could bind to the 3′AG enhancer. Using immunoprecipitation of DNA-protein complexes (15) followed by electromobility shift assays with overlapping subfragments spanning the 3′ enhancer, we found that LFY bound to two sites (AG I and AG II) that are only 31 bp apart and that are located in the LFY-response element defined with the LFY:VP16 experiments (Fig. 4). The sequences of the two AG sites are similar to each other as well as to the previously defined LFY binding site in the promoter of the AP1 gene (10). LFY bound both AGsites with similar affinity, but did not bind either AG site as strongly as it did the AP1 site (15).

Figure 3

(A to I) GUS expression in AG reporter lines KB30, KB45, KB46, MX68, and MX100. Specimens were prepared as in Fig. 2.

Figure 4

Mapping of LFY binding sites in AG. (A) In vitro binding of LFY to a 160-bp fragment (12) from the AG 3′ enhancer (21). At low protein concentration, one DNA-protein complex (α) is seen, whereas at higher protein concentration, a second, more slowly migrating complex (β) appears. We interpret the α complex as representing one site and the β complex as representing two sites occupied by LFY. (B) Binding to double-stranded, 30-bp oligonucleotides spanning the AG I and AG II sites (wt, wild type). (C) Comparison of LFY binding sites in AG with the previously reported site in AP1(10). The central bases form an inverted repeat in all three sites (“consensus”).

We tested the in vivo relevance of the LFY binding sites defined in vitro by introducing two small deletions into the 3′ AGenhancer. When the AG I site was deleted in reporter KB45 (16), the GUS expression pattern was still in anAG-typical manner, but the expression levels were weak and only detectable in whole mounts, not in sectioned tissue (Fig. 3C). KB45 expression increased in response to LFY:VP16 (Fig. 3D), although the response was attenuated compared with other reporters without an internal deletion (compare, for example, to Fig. 3, A and B). This result suggests that both LFY binding sites are required for in vivo activity of the 3′ AG enhancer, and that the LFY binding sites are only partially redundant. One possible explanation is cooperative action of adjacent LFY binding sites in vivo, even though there is no evidence for cooperative in vitro binding of LFY toAG I and AG II sites (Fig. 4A).

The activity of the KB46 reporter, in which both theAG I and AG II sites were deleted (16), was even further reduced than that of theAG I deletion, KB45 (Fig. 3E). And in contrast to KB45, KB46 expression could not be restored by crossing it to LFY:VP16(Fig. 3F). To corroborate that the loss of enhancer activity was indeed due to loss of LFY binding sites, we tested the effects of a series of point mutations in the LFY binding sites. A 2-bp mutation eliminates binding of LFY to a site in the AP1 promoter (10). The equivalent 2-bp mutations (m1) eliminated in vitro binding of LFY to either AG site (Fig. 4, B and C). Single–base pair mutations (m2) in the center of the consensus LFY binding sites (Fig. 4C), although only 2 bp away from the m1 mutations, did not affect LFY binding (Fig. 4, B and C). The mutant AGI and AG II sites were introduced into the 3′ AGenhancer, yielding reporters MX68 (AG I m1 and AGII m1) and MX100 (AG I m2 and AG II m2). The m1 mutations, which abolished in vitro binding of LFY, largely eliminated in vivo activity of the 3′ enhancer (Figs. 1B and 3G). In contrast, the m2 mutations did not interfere with in vitro binding of LFY and did not affect in vivo activity of the 3′ AG enhancer (Fig. 3I). To confirm that the LFY binding sites identified in vitro were indeed responsible for LFY action in vivo, we crossed several MX68 lines to LFY:VP16. In contrast to the control reporter KB31, or even reporters such as KB24 and KB28 that had the wild-type LFY binding sites but were otherwise too small to be active in a wild-type background, MX68 expression did not increase in LFY:VP16(Fig. 3H).

In contrast with many other transcription factor genes,LFY is not part of a gene family. LFY homologs have been cloned from several species of dicotyledonous plants (17), and each species seems to contain only one functional LFY homolog per haploid genome. Thus, we have little concern that related family members with similar DNA binding specificities as LFY bind in vivo to the LFY binding sites identified in vitro. Furthermore, there was a close correlation between the ability of LFY to bind to mutated sites in vitro and the in vivo activity of the corresponding reporters. Finally, we have been able to validate our results by probing DNA-protein interactions in vivo with the activated form of LFY, LFY:VP16. Together, these results support the hypothesis that binding of LFY is critical for transcriptional activation mediated by the 3′ AG enhancer. Although this finding does not necessarily imply that LFY interacts directly with the basal transcription machinery to stimulateAG RNA expression, it makes LFY formally an AGactivator.

Having recognized LFY as a direct activator of AG opens the way to identifying the coregulators that determine why AG is normally activated only in a subset of LFY-expressing cells. Previous studies have emphasized negative control of AG byAP1, APETALA2, LEUNIG, and CURLY LEAF, as well as the presumed Arabidopsisorthologs of the snapdragon genes FISTULATA andSTYLOSA (6, 9, 18). None of our reporters was ectopically expressed in wild-type flowers in a pattern that resembledAG expression in plants mutant for negative regulators, suggesting that activation and repression of AG are intimately linked. To discover how region-specific expression ofAG is achieved, the sequences through which AGregulators other than LFY act need to be delineated.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: weigel{at}


View Abstract

Navigate This Article