LEUNIG, a putative transcriptional corepressor that regulates AGAMOUS expression during flower development

Abstract

Regulation of homeotic gene expression is critical for proper developmental patterns in both animals and plants. LEUNIG is a key regulator of the Arabidopsis floral homeotic gene AGAMOUS. Mutations in LEUNIG cause ectopic AGAMOUS mRNA expression in the outer two whorls of a flower, leading to homeotic transformations of floral organ identity as well as loss of floral organs. We isolated the LEUNIG gene by using a map-based approach and showed that LEUNIG encodes a glutamine-rich protein with seven WD repeats and is similar in motif structure to a class of functionally related transcriptional corepressors including Tup1 from yeast and Groucho from Drosophila. The nuclear localization of LEUNIG–GFP is consistent with a role of LEUNIG as a transcriptional regulator. The detection of LEUNIG mRNA in all floral whorls at the time of their inception suggests that the restricted activity of LEUNIG in the outer two floral whorls must depend on interactions with other spatially restricted factors or on posttranslational regulation. Our finding suggests that both animals and plants use similar repressor proteins to regulate critical developmental processes.

Arabidopsis flowers, like those of other dicots, are composed of four types of organs (sepals, petals, stamens, and carpels) arranged in four concentric whorls. Three classes of floral homeotic genes work in a combinatorial fashion to specify the identity of floral organs (1). These floral homeotic genes all encode DNA-binding transcriptional factors, which activate floral organ-specific developmental programs. Proper temporal and spatial expression of the C-class floral homeotic gene AGAMOUS (AG) is central to flower pattern formation. In wild-type, AG mRNA is expressed only in the inner two whorls of a flower to specify stamen and carpel development (2–4). In leunig (lug) and apetala 2 (ap2) mutants, however, AG mRNA is ectopically expressed in the outer two whorls of a flower, resulting in the homeotic transformation from sepals to carpels and petals to stamens or absent (4, 5). Thus, the activities of LUG and AP2 are required in the outer floral whorls to negatively regulate AG expression. In addition, AG RNA is expressed much earlier in lug and ap2 mutant floral meristems, suggesting that LUG and AP2 also regulate the temporal expression pattern of AG. A third gene, CURLY LEAF, is required to represses AG transcription mainly in vegetative tissues (6). Hence, transcriptional repression of AG is vital to both vegetative and reproductive development.

LUG provides a unique opportunity to study transcriptional repression in higher plants. Although LUG is required in the outer two whorls of a flower to repress AG mRNA expression, LUG is not required to specify sepal/petal identity in the outer two whorls because ag lug double mutants still develop normal sepals and petals (5). In contrast, AP2, which encodes two copies of a DNA-binding motif (7), is required to both repress AG and specify sepal/petal identity (3). Hence, unlike AP2, LUG is strictly a negative regulator of floral homeotic gene expression with no direct role in the specification of floral organ identity. In addition, LUG is distinct from CURLY LEAF, an Arabidopsis homolog of the Drosophila polycomb-group gene Enhancer of zeste, in that LUG is required for the initial repression of AG transcription at early stages of floral meristem development, whereas CURLY LEAF is required to maintain AG repression at later stages of flower development (6). Finally, lug mutants exhibit other defects that are independent of AG, including a split stigma (Fig. 1 A and B), abnormal carpel and ovule development, reduced female and male fertility, and narrower leaves and floral organs (5, 8–10). Therefore, LUG likely regulates the expression of several target genes in diverse developmental processes.

Figure 1.

Micrographs of lug flowers. (A) lug-16, a weak allele. The pair of arrows indicate the split stigma characteristic of lug mutant flowers. (B) lug-12, a strong allele. The flower has no petal and exhibits split stigma (arrows). The smaller size sepals are caused by partial homeotic transformation into carpelloid sepals. (C) The flower of a lug-16 mutant rescued by cosmid 31-G. The flower and its stigma are similar to wild type.

To further understand the molecular mechanisms on how LUG represses AG expression, we isolated the LUG gene by using a map-based approach. Unlike other genes known to regulate floral homeotic gene expression, LUG does not encode any obvious DNA-binding motifs. Instead, LUG encodes a glutamine-rich protein with seven WD repeats at the COOH terminus. LUG is similar to the yeast Tup1 and Drosophila Groucho (Gro) transcriptional corepressors both in genetic function and in motif structure. Because transcriptional repression clearly plays a major role in higher plant development, our study may contribute to the understanding of general regulatory mechanisms of higher plant development.

Methods

Mutant Strains and Mapping Population.

lug-1 and lug-3 have been described (5). lug-12 and lug-16 are ethyl methanesulfonate (EMS)-induced mutation in the Landsberg erecta (L-er) background. lug-14 is a Ds transposon-induced allele in L-er. Two mapping populations were screened to identify recombinant chromosomes in F₂ or F₃ by using the genetic markers cer2 and ap2 that flank LUG at 5 and 16 map units, respectively. A cer2 allele in Wassilewskija ecotype was crossed into lug-1 in L-er ecotype. F₂ cer2 families and F₂ lug-1 families were screened for segregating cer2 lug-1 double mutants in F₃. These double mutants were used for subsequent fine mapping. A second mapping population was similarly constructed by crossing an ap2 mutant in Columbia ecotype with lug-8 in L-er ecotype.

Positional Cloning of LUG.

Fifteen recombinants between cer2 and lug and 52 recombinants between ap2 and lug were further screened with cleaved amplified polymorphic sequence (CAPS) markers shown in Fig. 2A. The CAPS markers F10N7H and F4D11S flank lug with one recombinant on each side. Information for these markers is at http://www.arabidopsis.org/maps/CAPS_Chr4.html. The region from F10N7H to F4D11S is covered by the yeast artificial chromosome (YAC) clone CIC1F9 and five overlapping bacterial artificial chromosome (BAC) clones (Fig. 2A). A cosmid library was made from the YAC CIC1F9 and the binary cosmid transformation vector pCLD04541. The procedure is essentially the same as described (11). Cosmid clones were isolated by screening the cosmid library with each of the five BAC clones as probes. The cosmid contig, which spans 284 kb, was constructed by using restriction enzyme analyses. A total of 23 cosmids were transformed into the weak and fertile lug-16 mutants via Agrobacterium strain GV3101 and vacuum infiltration (12). Complementation rescue by the 31–1 and 31-G cosmid clones was observed by the wild-type primary transgenic plants and subsequently was verified in the next generation by 100% (n = 44) cosegregation of the rescued plants with corresponding cosmid clones.

Figure 2.

Molecular cloning of LUG. (A) A physical map of LUG on chromosome 4. Open bars represent yeast artificial chromosome (YAC) and bacterial artificial chromosome (BAC) clones. The relative position of several PCR-based markers is placed above the chromosome. Cosmids 31–1 and 31-G both complemented the lug-16 mutant. (B) The protein sequence of LUG from L-er. The position of five lug mutations is indicated above the amino acid sequence. The two glutamine-rich regions are in bold. The seven WD repeats are underlined and are identified by using http://pfam.wustl.edu/hmmsearch.shtml. The specific base pair changes in each lug allele are described below based on the sequences of cosmid L23H3 (AL050398). lug-3 causes a C to T change at 25309; lug-12 causes a C to T change at 25198; lug-16 causes a G to A change at 24693; lug-1 mutation causes a G to A change at 26267; lug-14 mutation inserts an A after 29080. (C) Structural similarity between LUG, LUH (AC 003974), the S. cerevisiae Tup1 (19), and the Drosophila Gro (20). Numbers above and in parentheses correspond to amino acids. The percentage between the amino acid numbers indicates the level of identity between LUG and LUH. The LUFS domain in LUG and LUH is indicated by a dotted region. A second highly conserved region between LUG and LUH immediately precedes the WD repeats and is marked by wavy lines. The level of similarity is identified by using http://www.bioinformatix.com/sas/. (D) Sequence alignment in the LUFS domain among LUG, LUH (AC003974), Flo8, and a human single-strand DNA-binding protein (AF077048). Numbers correspond to amino acids. Flo8 has extra 21 aa indicated by (X)21. Amino acids conserved in all four genes are in bold and are indicated by *. Amino acids conserved in three of the four genes are in bold and are indicated by :.

Isolation and Analyses of LUG mRNA.

Three LUG cDNA clones were isolated by screening a total of ≈500,000 plaques from two cDNA libraries: Weigel L-er flower library and Kieber and Ecker Columbia cDNA library (3–6 kb). A 5′ rapid amplification of cDNA ends kit version 2.0 (GIBCO/BRL) was used to identify the 5′ sequence of LUG mRNA. Three nested primers from the 5′ gene-specific region of LUG cDNA were used. They are 5′-TCTCATTGGTCCTAGC-3′, 5′-CCAGAAGACAGACCACCACTC-3′, and 5′-CAGGTGCGTCAATAGCAACTG-3′. The PCR products were cloned into pGEM-T vector (Promega) and sequenced.

For Northern analyses, poly(A)⁺ RNA was isolated from respective tissues by using the fasttrack 2.0 RNA isolation system (Invitrogen). Both root and shoot mRNA was isolated from the same 10-day-old seedlings. mRNA from the stems, cauline leaves, and inflorescences was isolated from the same 21-day-old plants that had just bolted. mRNA from the lug-1, lug-12, and lug-16 was isolated from inflorescences of 21-day-old plants. A total of ≈1 μg poly(A)⁺ RNA was loaded per lane and fractionated in a 1.2% agarose gel. The Northern filters were probed with the 3′ 520-bp LUG cDNA sequence. The actin probe is a 1.6-kb actin cDNA from Brassica oleracea (13). The northern images were generated by using a Cyclone Storage Phosphor System (Packard) and quantitated by using optiquant image analysis software (Packard).

The in situ sense and antisense probes were derived from a 220-bp EcoRV/SspI fragment from the 5′ untranslated region of LUG, which was cloned into the pSK vector (Stratagene). The probes were labeled with digoxigenin-UTP as described (14), and the fixation and hybridization procedure was based on www.wisc.edu/genetics/CATG/barton/protocols.html.

Green Fluorescent Protein (GFP)–LUG Chimeric Gene Construction and Transient Expression.

A full-length LUG cDNA was PCR-amplified with primers: 5′-GAGATCTATGTCTCAGACCAACTG-3′ and 5′-TAAGATCTTCGTCAGATCATACAACAAC-3′ by using the Platinum High Fidelity Taq polymerase (GIBCO/BRL). The amplified fragment was cloned into the pGEM-T vector (Promega) and sequenced. Full-length LUG cDNA in the pGEM-T vector was digested with BglII and cloned into the pAVA393 vector (15). The resulting plasmid construct, verified by sequence analyses, carries an in-frame fusion of LUG to the carboxyl terminus of GFP. Transient expression assays in onion epidermal cells were carried out by using a Biolistic Particle Delivery System (Bio-Rad) (16), and the tissues were visualized and photographed as described (14). Images were obtained with a ×20 objective and a Nikon 2000 35-mm camera.

Results

Positional Cloning of LUG.

We isolated the LUG gene by positional cloning. The position of LUG on chromosome 4 was established by using recombinant chromosomes and PCR-based markers (Fig. 2A and Methods). A 284-kb cosmid contig of 23 cosmid clones was transformed into lug-16 mutants to test for rescue of the mutant phenotype. lug-16, a weak lug mutant with the highest fertility among all lug alleles, allows the direct transformation of homozygous lug-16 plants. Two overlapping cosmid clones, 31–1 and 31-G, complemented the lug-16 mutant as shown by the wild-type flowers formed by both primary and secondary transgenic plants (Figs. 1 A and C and 2A; Methods). The overlapping region of these two cosmids contains one gene that spans the genomic sequences from nucleotides 23704 to 29200 of cosmid L23H3, whose sequence was released by the European Union Arabidopsis Sequence Consortium (AL050398).

Three cDNA clones residing in the overlapping region of the two cosmids were isolated by screening cDNA libraries (Methods). 5′ Rapid amplification of cDNA ends was used to identify the 5′ sequences of the mRNA (Methods). Sequence analyses of five lug mutant alleles all revealed single base changes in this gene. lug-3 and lug-12 mutants both exhibit strong phenotypes (ref. 5; Fig. 1B), and both possess a C to T change that results in a glutamine (Q) to a stop codon near the NH₂ terminus (Fig. 2B). Northern analyses of lug-12 mRNA revealed that the level of lug-12 transcript is only at about 38% of the wild-type level (Fig. 3A, lane 7), indicating that the premature termination of translation may have caused a reduced stability of the lug-12 transcript. lug-16 causes a G to A change at the splicing acceptor site of the third exon. This single base substitution apparently alters splicing as shown by a reduced size of lug-16 transcript (Fig. 3B, lane 8). However, the level of the abnormally spliced lug-16 transcript is increased to about 200% of wild-type level (Fig. 3B). One possible explanation may be that LUG is involved in repressing its own transcription and that an increased LUG mRNA level is seen only in those lug mutants that do not suffer from premature translational termination. The intermediate-strength lug-1 mutation causes a G to A change at the splicing acceptor site of the eighth exon. Sequence analyses of reverse transcription–PCR products indicated that the lug-1 transcript uses an alternative splice acceptor site only one base 3′ to the original site (data not shown). This results in a frame shift followed by 44 new amino acids and then a stop. Northern analyses revealed that the level of the lug-1 transcript is reduced to 45% of wild-type level (Fig. 3A, lane 6), probably because of the premature translational termination. Finally, the strong lug-14 mutation inserts an A and changes the last 39 amino acids of the protein (Fig. 2B). These data indicated that we had isolated the LUG gene.

Figure 3.

LUG mRNA expression. (A) LUG mRNA is expressed in all tissues including roots (lane 1), shoots (lane 2), stems (lane 3), cauline leaves (lane 4), and inflorescences (lane 5). A decreased level of LUG mRNA is detected in lug-1 (lane 6) and lug-12 (lane 7) mutant inflorescences. The mRNA level is corrected with the actin loading control, and the relative level is derived by comparing all signals to the stem (lane 3), which equals to 1.0. (B) LUG mRNA expression in inflorescences of wild-type (lane 1), ap1–1 (lane 2), ap2–1 (lane 3), ap2–2 (lane 4), pi-1 (lane 5), ag-2 (lane 6), lfy-6 (lane 7), and lug-16 (lane 8). With the exception of ag-2, LUG mRNA level in most mutants appears unchanged. In lug-16, LUG transcript is reduced in size and increased in abundance (arrowheads).

LUG Protein Is Q-Rich and Contains Seven WD Repeats.

Based on the LUG cDNA and 5′ rapid amplification of cDNA ends, the LUG gene encodes a protein of 931 aa (Fig. 2B). Near the NH₂ terminus of LUG, two Q-rich regions are identified (Fig. 2B). The first Q-rich region (residues 89–184) contains 66 Qs, including three uninterrupted stretches of 10, 10, and 20. The second Q-rich region (residues 449–470) contains 12 Qs. The COOH terminus of the protein contains seven WD repeats (Fig. 2B). Finally, the region between the Q-rich and the WD repeats is rich in serine, glycine, leucine, and proline.

blast search revealed an Arabidopsis chromosome 2 gene with the highest sequence similarity to LUG. This homolog of LUG (named LUH; AC 003974) exhibits 44% overall sequence identity to LUG (Fig. 2C). The presence of several Arabidopsis LUH expressed sequence tag clones indicates that LUH is an expressed gene. The NH₂-terminal 88 aa of LUG show 80% identity to LUH. This high level of conservation suggests that the NH₂-terminal 88 aa may define an important functional domain. Database search with the NH₂-terminal 88 aa revealed 35% and 31% sequence identity to the yeast Flo8 gene (17, 18) and a human single-strand DNA-binding protein (AF077048), respectively (Fig. 2D). Thus, we named this conserved domain LUFS (LUG/LUH, Flo8, single-strand DNA-binding protein), which is also present in genes of unknown function in organisms such as Schizosaccharomyces pombe, Drosophila, and Caenorhabditis elegans. The functional nature of this domain is unknown.

The second motif of LUG that shows sequence similarities to other proteins in databases is the WD repeat. Among WD repeat-containing proteins, the yeast transcriptional corepressor Tup1 (19) exhibits additional levels of similarity to LUG. Tup1 protein shows 21% overall sequence identity to LUG. Like LUG, Tup1 protein has Q-rich regions near the NH₂ terminus and seven WD repeats at the COOH terminus (Fig. 2C). However, Tup1 has a unique NH₂-terminal domain of 72 aa, which bears no sequence similarity to the LUFS domain of LUG. Tup1 belongs a class of functionally related transcriptional corepressors including Drosophila Gro, which also possesses a NH₂-terminal Q-rich region and COOH-terminal WD repeats (Fig. 2C; refs. 20 and 21).

LEUNIG Is Expressed in Both Vegetative and Reproductive Tissues.

Because lug mutants exhibit defects in both reproductive and vegetative tissues, we examined LUG mRNA expression in different plant tissues and organs by Northern blots. We found that LUG is expressed in all tissues tested including roots, shoots, stems, cauline leaves, and inflorescences (Fig. 3A). The expression level is highest in inflorescences that contain flowers at all stages. The broad mRNA expression profile of LUG is consistent with a role of LUG in regulating multiple developmental processes. To identify transcriptional regulators of LUG, we examined LUG mRNA expression in several floral homeotic mutants including apetala1 (ap1), ap2, pistillata (pi), and ag (2–3, 22) and in floral meristem identity mutant leafy (lfy) (23). Using mRNAs isolated from floral tissues, we found that, with the exception of ag-2, LUG mRNA level appeared unchanged in all of the mutants tested (Fig. 3B). In the ag-2 mutants (Fig. 3B, lane 6), however, LUG mRNA level is slightly reduced. Because the ag-2 mutant flowers do not form any stamens and carpels, which express high levels of LUG mRNA (see below), it is not surprising that LUG mRNA appeared reduced in the ag-2 mutants.

In situ hybridization was used to examine further the LUG mRNA expression pattern in young 14-day-old seedlings and during flower development (Fig. 4). A low level of LUG mRNA was detected both in shoot apical meristems and inflorescence meristems (Fig. 4 A and D). LUG mRNA level is low in the first few young leaf primordia in seedlings but increases rapidly in older leaves (Fig. 4A). In cauline leaves of bolted plants, LUG mRNA is more abundant at the adaxial side of the leaves (Fig. 4C). In stems and carpel valves, LUG mRNA is prominently expressed in the vascular tissues (Fig. 4 C, G, and I). Because one of the most important function of LUG is to repress AG RNA expression in the outer two whorls of a flower, we tested whether LUG RNA is expressed only in the outer two floral whorls. We found that LUG mRNA is detected in all four floral whorls at the time of their inception (Fig. 4 C–F). LUG is highly expressed in the entire stage 1–2 floral primordia, and in young floral organ primordia (sepals, petals, stamens, and carpels) (Fig. 4 C–F). Thus, the ability of LUG to repress AG expression in the outer two whorls must depend on either posttranslational regulation or interaction with other spatially restricted regulators. During later stages of flower development, LUG expression subsides from sepals (at stage 7) and is detected in the developing stamen and carpel primordia (Fig. 4F). Interestingly, the expression of LUG in petals persists until at least stage 12 (Fig. 4F). In addition, LUG mRNA is found prominently in the placenta/ovules (Fig. 4 G–I) and in locules of anthers (Fig. 4H). This high level of LUG mRNA expression during both female and male gamete development suggests that LUG plays an important role in ovule and pollen development and is consistent with the observed abnormality in lug female and male fertility (5, 8–10).

Figure 4.

In situ expression pattern of LUG mRNA. (A) LUG expression in 14-day-old seedlings. A low level of LUG mRNA is present in the shoot apical meristem and in the first few young leaves. LUG mRNA level increases dramatically in more developed leaves (arrow). (B) LUG sense probe to 14-day-old seedlings. (C) LUG mRNA is detected in the secondary inflorescence meristem and in the stage 1 and 3 floral meristems (numbers indicate the stage of respective floral meristems; ref. 33). LUG mRNA is detected in vascular tissues (arrowhead). LUG mRNA level is also more abundant in the adaxial side of the cauline leaf (arrow). (D) LUG mRNA level is low in the inflorescence meristem but increases in young flowers (stages 3 and 5, respectively). (E) LUG is strongly expressed in the sepal primordia and the central dome of the stage 4 and 5 flowers. (F) At stage 7, LUG mRNA is not detected in sepals, but is present in the carpel (ca) and stamen (st) primordia. LUG mRNA is persistently detected in petals (pe) as shown in this stage 11 flower. (G) In stage 10 carpels, LUG mRNA is strongly expressed in the placenta/ovule primordia (arrows) and weakly expressed in the carpel valves. (H) A cross section of a stage 9 flower. A high level of LUG mRNA is detected in placenta (p) and locules (lo) of the anther. (I) LUG mRNA is detected in developing ovules at stage 12.

GFP–LUG Fusion Protein Is Localized in the Nucleus.

If LUG functions as a transcriptional regulator, LUG protein should be located in the nucleus. However, LUG does not appear to encode any obvious nuclear localization signal (Fig. 2B). To determine the subcellular location of LUG, we made a fusion protein between LUG and the GFP. A full-length LUG cDNA was inserted COOH-terminal to the GFP, and the fusion construct was transiently expressed in onion epidermal cells. The GFP–LUG fusion protein localizes to the nucleus of onion epidermal cells (Fig. 5 A and C), whereas GFP alone localizes to both cytoplasm and nucleus (Fig. 5 B and D). The nuclear localization of GFP-LUG supports LUG as a transcriptional regulator.

Figure 5.

Nuclear localization of GFP-LUG. (A and B) Dark-field images. (C and D) Corresponding phase-contrast microscopic images. (A and C) An onion epidermal cell transiently expressing the GFP-LUG chimeric protein. (B and D) An onion epidermal cell transiently expressing the GFP protein (vector pAVA 393; ref. 15). Arrows mark the location of nucleus.

Discussion

Our molecular isolation of LUG reported here indicates that unlike other regulators of floral homeotic genes such as AP2 and ANT (7, 24, 25), LUG does not possess any obvious DNA-binding motifs. Instead, LUG encodes seven WD repeats, a LUFS motif, and two Q-rich regions. The WD repeats, which mediate protein–protein interactions, are found in proteins with a wide variety of biochemical functions (26). Our finding of the single base insertion that eliminates the last WD repeat in the strong lug-14 mutant suggests that the last WD repeat is indispensable for the function of LUG. Although the function of LUFS motif is unknown, the S. cerevisiae Flo8, which also possesses the LUFS domain near its NH₂ terminus, is a transcriptional activator of Flo1 (17). However, there is no evidence that Flo8 binds to Flo1 DNA directly, and the LUFS domain is unlikely a DNA-binding motif. Hence, it is highly unlikely that LUG directly interacts with AG cis-regulatory elements to repress AG transcription, and LUG must interact with or recruit other proteins to exert its negative effect on transcription.

LUG Is Similar in Motif Structure to the Yeast Corepressor Tup1.

The overall motif structure of LUG is similar to a class of functionally related transcriptional corepressors, including the yeast Tup1, Drosophila Gro, and mammalian TLE (transducin-like enhancer of split) (Fig. 2C; refs. 19–21 and 27). The mechanism of how this class of transcriptional corepressors regulates target gene expression is relatively well understood (reviewed in refs. 21 and 27). Tup1 is brought to target promoters by sequence-specific DNA-binding proteins, and Tup1 regulates a wide array of independent sets of genes such as a-cell-specific genes, glucose-repressed genes, flocculation genes (such as Flo1), and DNA damage-induced genes (27, 28). The NH₂-terminal 72 aa of Tup1 forms repression complexes with Ssn6, a tetratricopeptide repeat protein (29). Interactions between the Tup1/Ssn6 complex and the specific DNA-binding factors (via the Tup1 WD repeats) bring about transcriptional repression to specific target genes. For example, the homeodomain protein α2 and the MADS box protein Mcm1 bind cooperatively to the promoters of a-cell-specific genes and recruit the Tup1/Ssn6 complex to repress a-cell-specific genes (30). Likewise, the Drosophila Gro, which possesses a Q-rich region at the NH₂ terminus and WD repeats at the COOH terminus (Fig. 2C), is a maternally contributed protein that interacts with a variety of regulatory proteins such as Hairy, Dorsal, and Engrailed to repress different target genes expression during segmentation, dorsal-ventral patterning, and neurogenesis (21).

A Proposed Model on How LUG Regulates AG.

With the molecular isolation of LUG, we can begin to understand the mechanism of how LUG regulates the spatial and temporal expression pattern of AG and in what context LUG exerts its functional specificity in different tissues. The similarity in structure, transcriptional repression, and the regulation of multiple processes suggests that LUG may function via a similar mechanism as Tup1 and Gro. In outer whorls of Arabidopsis flowers, LUG may exert its transcriptional repression by interacting with transcription factors that bind to the cis-regulatory elements of AG. AP2 and the more recently identified Sterile Apetala (SAP) (31), for example, could be such candidate transcription factors that mediate the repression by LUG. In particular, ap2 and lug mutations exhibit synergistic interactions with respect to defects in AG repression (5), and both AP2 and LUG were shown to repress AG through the same enhancer sequences that span most of the second intron of AG (32). However, in vitro coprecipitation assays failed to detect physical interactions between LUG and AP2 (unpublished work). Either AP2 does not directly mediate the effect of LUG or additional proteins may be required for LUG to physically interact with AP2. Ubiquitously distributed LUG and AP2 mRNA in all four floral whorls (Fig. 4; ref. 7) points to the need of other whorl-specific factors to confer their spatially restricted activity. In addition, because of the diverse functions of LUG revealed by its phenotype and by its broad mRNA expression profile, LUG may interact with several different DNA-binding factors in different tissues or at different developmental stages to confer its regulatory specificity. The molecular isolation and analyses of LUG provided us with a unique opportunity to further investigate mechanisms underlying spatially and temporally restricted gene expression and to identify interacting genes.

Abbreviations

AG

AGAMOUS

lug

leunig

ap2

apetala 2

EMS

ethyl methanesulfonate

L-er

Landsberg erecta

GFP

green fluorescent protein

Gro

Groucho