Analysis of the Transcription Factor WUSCHEL and Its Functional Homologue in Antirrhinum Reveals a Potential Mechanism for Their Roles in Meristem Maintenance

Abstract

One of the most significant features of plant development is the way in which it can be elaborated and modulated throughout the life of the plant, an ability that is conferred by meristems. The Arabidopsis thaliana WUSCHEL gene (WUS), which encodes a homeodomain transcription factor, is required to maintain the stem cells in the shoot apical meristem in an undifferentiated state. The mechanism by which WUS prevents the differentiation of stem cells is unknown. We have characterized a meristem maintenance mutant in Antirrhinum majus and shown that it arises from a defect in the WUS orthologue ROSULATA (ROA). Detailed characterization of a semidominant roa allele revealed an essential role for the conserved C-terminal domain. Expression of either ROA or WUS lacking this domain causes a failure of meristem maintenance. The conserved domain mediates an interaction between WUS and two members of a small family of corepressor-like proteins in Arabidopsis. Our results suggest that WUS functions by recruiting transcriptional corepressors to repress target genes that promote differentiation, thereby ensuring stem cell maintenance.

INTRODUCTION

Unlike that in animals, most plant organogenesis takes place postembryonically, allowing plants to modify their development in response to the environment. The shoot apical meristem (SAM), located at the tip of the shoot, gives rise to all of the aboveground organs and tissues initiated after germination. The SAM has the ability to maintain an organized structure while responding to intrinsic and extrinsic developmental signals. To accomplish this, a constant pool of slowly dividing undifferentiated stem cells is maintained within the central zone of the SAM. Cells that leave the central zone enter the flanking peripheral zone and can be recruited into lateral organ primordia, where they begin to divide more rapidly and differentiate. Cells leaving the central zone to enter the underlying rib zone are incorporated into the pith tissue. In this dynamic environment, coordination between the rates of differentiation and proliferation critically depends on signaling between the different compartments of the SAM (Clark, 2001).

The proliferation of stem cells within the SAM must be exactly matched to the rate at which daughter cells differentiate (Baurle and Laux, 2003; Carles and Fletcher, 2003; Gross-Hardt and Laux, 2003). The mechanism of meristem maintenance and coordination has been studied extensively using a series of Arabidopsis thaliana mutants that show defects in meristem homeostasis (Clark, 2001; Fletcher, 2002). Mutants in the CLAVATA genes (CLV1, CLV2, and CLV3) result in an overproduction of stem cells and a breakdown in meristem organization. By contrast, mutants of the WUSCHEL gene (WUS) result in plants that are unable to maintain a pool of undifferentiated stem cells (Laux et al., 1996). Cells expressing WUS define a new domain within the central zone, referred to as the SAM organizing center (Schoof et al., 2000). WUS activity results in signaling to the overlying stem cells, inducing CLV3, which then signals back to repress WUS. The CLV/WUS negative feedback loop ensures homeostasis of the SAM by regulating the number of stem cells present in the central zone (Brand et al., 2000; Schoof et al., 2000). The stability of WUS expression also involves signaling from leaf primordia and tissue underlying the organizing center (Stuurman et al., 2002). WUS encodes the founding member of the WUSCHEL-related homeobox (WOX) subfamily of homeobox transcription factors, several of whose members have subsequently been shown to control aspects of plant development (Matsumoto and Okada, 2001; Haecker et al., 2004; Nardmann et al., 2004; Park et al., 2005; Wu et al., 2005).

In addition to the homedomain, WUS contains three conserved short sequence motifs at the C-terminal end: an acidic domain, a WUS box (TLPLFPMH), and an EAR-like (for ERF-associated amphiphilic repression) domain (ASLELTLN) (Ohta et al., 2001). The presence of the acidic domain is consistent with a function as a transcriptional activator. Although the function of the WUS box is unknown, EAR-like domains have been found to be involved in transcriptional repression (Ohta et al., 2001; Hiratsu et al., 2004; Tiwari et al., 2004). At the end of flower development, WUS terminates stem cell maintenance, and this function involves the direct activation of the AGAMOUS (AG) gene together with LEAFY (LFY) (Lenhard et al., 2001; Lohmann et al., 2001). However, the mechanism by which WUS represses the differentiation of stem cells is unknown.

The role of WUS in maintaining the SAM appears to have been evolutionarily conserved in angiosperms, because mutants in a related Petunia hybrida gene, TERMINATOR (TER), also result in analogous meristem termination (Stuurman et al., 2002). Here, we report the characterization of an Antirrhinum majus mutant that displays an inability to maintain its apical meristems and show that it also results from defects in a WUS-like homeobox gene, ROSULATA (ROA). Analysis of the semidominant Antirrhinum roa-2 allele provides experimental evidence for the functional importance of conserved amino acid sequences in the C-terminal domain of WUS (Stuurman et al., 2002; Haecker et al., 2004). Based on the Antirrhinum roa-2 allele, we recreated a dominant wus-like phenotype in Arabidopsis by expressing a truncated WUS protein lacking the conserved C-terminal domains. We have identified two WUS-interacting proteins (WSIP1 and WSIP2) that are members of a small family of transcriptional corepressor-like proteins in Arabidopsis and that depend on the presence of the conserved C-terminal domains for full interaction with WUS. Our results suggest that WUS acts to maintain the meristem by recruiting transcriptional corepressors to repress target genes required for differentiation.

RESULTS

ROA Regulates SAM Maintenance in Antirrhinum

Two independent roa alleles were available (Stubbe, 1974), both of which grow as short bushy plants compared with the wild type (Figure 1A). roa-1 and roa-2 homozygous seedlings both develop normally until production of the first pair of leaves (Figures 1A and 1B). Mutants show a range of severity of phenotype at this stage, producing two, one, or no leaves. The petiole base is often ventralized and appears radially symmetrical (cf. P1 in Figures 1C and 1D), resulting in varying degrees of loss of leaf blade expansion (data not shown). In wild-type plants, the next pairs of leaves are initiated by the SAM (Figure 1D). In roa mutants, the next pairs of leaves arise in the axils of the cotyledons, as a result of loss of the SAM (Figure 1C). The centers of the mutant apices are occupied by the abnormal, radially symmetrical petioles of the first pair of leaves, separated by a few rows of cells. Most of the meristem is consumed after initiation of a single pair of leaves. However, after 2 weeks, a dome-shaped structure forms (Figure 1E) composed of small, actively dividing cells containing large nuclei (Figure 1F). Longitudinal optical sections through the dome reveal its layered pattern, resembling the L1, L2, and L3 organization of the shoot meristem (Figure 1F, inset). In older mutant plants, shoot meristems and individual leaves are initiated from the apical dome in a disorganized manner (Figure 1G). In some roa-2 plants, stripe(s) of cells, running across the dome between the petioles, differentiate as epidermal cells bearing trichomes (Figure 1H).

Figure 1.

roa Phenotype.

(A) Seven-week-old plants (from left to right): roa-1, roa-2, heterozygous roa-2 (arrowhead indicates shoot branching), and wild type.

(B) Three roa-1 mutant seedlings (4 weeks) with decreasing severity of defects affecting the first pair of leaves, and a wild-type seedling for comparison (right).

(C) to (J) SAM phenotype. Where necessary, leaves were removed to gain access to the apices and petiole scars are visible.

(C) and (D) Scanning electron micrographs of a 16-d-old roa-1 apex (C) and a 16-d-old wild-type apex (D). P1 and P2 indicate the petioles of the first and second pairs of leaves, respectively. Pc, cotyledon petiole. A third set of leaves is developing from the meristem in the center of the wild type. The mutant meristem domain is now reduced to a narrow strip (arrowhead).

(E) A 30-d-old mutant.

(F) Transverse confocal laser section though a 30-d-old mutant apex. A longitudinal section shows the layered organization (inset).

(G) A later stage showing the disorganized initiation of leaves and shoots (arrowheads).

(H) A roa-2 apex with a stripe of differentiated cells connecting petioles of the first pair of leaves.

(I) Initiation of new leaves (P3) between the second pair of leaves (P2).

(J) A severe mutant apex showing arrested development.

Bars = 150 μm in (C) to (E) and (G) to (J), 100 μm in (F), and 30 μm in (F) inset.

Organogenesis between the first pair of leaves is frequent in roa-1 mutants but rare in roa-2. The leaves and meristems that form in the apical position usually arrest at an early stage and do not contribute to further development. The roa mutants continue to develop from new SAMs initiated between the pairs of leaves in the axils of the cotyledons (P3 in Figure 1I). These meristems arise at the base of petioles that show only mild defects in their dorsoventral asymmetry. Severely defective, ventralized petioles do not carry new SAMs at their base. In the most dramatic cases, the organogenic zones separating the three first pairs of leaves become merged into a large structure bearing many aborted organs (Figure 1J), and meristem activity terminates. All new shoot meristems reproduce the same defects and terminate organized activity after producing one pair of leaves. This pattern of organogenesis and termination is constantly reiterated in mutant plants, creating a complex accumulation of hundreds of leaves. Flowering has never been observed in homozygous roa plants, although morphologically normal revertant shoots are occasionally found on roa-1 plants, indicating that roa-1 results from a transposon insertion.

Thus, the principal feature of roa mutants is that the SAM is initiated normally but meristem maintenance is disrupted. After initiation, the SAM produces a pair of leaves and is almost completely consumed. The few remaining cells divide and replace the SAM by a large disorganized organogenic zone that behaves like a SAM peripheral zone. Further growth derives from SAMs that are repetitively gained and lost in the axils of older leaves. Interestingly, heterozygous roa-2 mutants also have a defect in meristem maintenance, although much weaker than that seen in homozygous mutants, so that SAM failure and the accompanying reinitiation of new SAMs is often observed at higher nodes (arrowhead in Figure 1A).

ROA Encodes a WUS-Like Homeodomain Protein

Similarities between the roa and Arabidopsis wus phenotypes raised the possibility that roa-1 could be caused by a transposon insertion in a WUS-like homeobox gene. An Antirrhinum cDNA library was screened at low stringency with a WUS cDNA probe, and a 1075-bp clone was identified. The full-length cDNA was obtained by rapid amplification of cDNA ends and was used to probe a DNA gel blot containing genomic DNA from roa-1 and wild-type plants. Cosegregating polymorphisms were identified between the wild type and roa-1, and the banding pattern of heterozygous roa-1 plants was identical to the pattern generated from roa-1 revertant shoots (data not shown). Therefore, a transposon insertion in the ROA gene was responsible for the roa-1 allele. Thermal asymmetric interlaced PCR was used to isolate the transposon insertion site. Sequence analysis confirmed the presence of a Tam7-like transposable element in the second intron (Figure 2A). DNA gel blots did not reveal a polymorphism between the wild type and roa-2. Sequencing of the ROA gene in this allele demonstrated a duplication of four nucleotides (CTAG) in the third exon, resulting in a frameshift mutation (Figures 2A and 2B).

Figure 2.

Molecular Characterization of roa.

(A) Representation of the ROA and WUS genes. Boxes represent exons, continuous lines represent introns, black boxes represent homeodomains, light gray boxes represent untranslated exon sequences, and dotted lines represent flanking genomic DNA. The dark gray box above the ROA gene (not to scale) indicates the transposon insertion point in roa-1, 529 bp into the second intron. The CTAG duplication in roa-2 is represented above its corresponding position.

(B) ClustalW alignment of ROA, TER, and WUS. Identical residues are highlighted in black, and similar residues are highlighted in gray. Dashes represent gaps in the alignment. The 13–amino acid frameshift substitution in roa-2 is shown in italics below the wild-type sequence. The asterisk represents the point of WUS truncation in WUSΔ. The conserved acidic domain is boxed, and the homeodomain is single underlined. The two small conserved C-terminal domains, including the WUS box, are double underlined.

(C) Amplification of ROA by semiquantitative RT-PCR. Samples are duplicated, each corresponding to RNA extracted from four plant apices of homozygous roa-1 and roa-2 and heterozygous roa-2. Two wild-type and one negative control are included. The bottom panel shows a control amplification of the constitutively expressed EF1-α cDNA.

An alignment of ROA, WUS, and TER reveals that their homeodomains share 83% amino acid identity. Apart from a few short motifs, little sequence conservation is evident outside the homeodomain (Figure 2B). This is especially apparent in a comparison between ROA and TER, which show 100% identity within the homeodomain. As a result of the frameshift in roa-2, the last 61 amino acids of the mutant protein are predicted to be replaced by 13 amino acids from the +2 reading frame (Figure 2B). This mutation is not expected to abolish gene expression, and semiquantitative RT-PCR confirms that the ROA transcript is present at wild-type levels in roa-2 (Figure 2C). By contrast, no full-length ROA transcript can be detected in the roa-1 allele, suggesting that the transposon insertion abolishes transcript processing or stability. Three conserved domains lie within the final 65 amino acids of ROA (Figure 2B). One corresponds to a previously reported acidic domain (Mayer et al., 1998). The other two (TLPLFPMH and ASLELXLN) were also noted in sequence comparisons between WUS and TER (Stuurman et al., 2002), and TLPLFPMH has been named the WUS box (Haecker et al., 2004). The fact that these domains are conserved between WUS, TER, and ROA, within a region of little overall sequence conservation, suggests that they are functionally important. The loss of these conserved domains in the roa-2 allele and its consequent mutant phenotype demonstrate that they are required for ROA activity.

ROA Expression Pattern

ROA expression in the vegetative apex is only detectable by RT-PCR (Figure 3A), but later expression becomes detectable by in situ analysis in inflorescences and young floral meristems. Expression is observed in the L3 but not the L1 and L2 cell layers in a small group of cells below the central zone of the meristem (Figures 3B and 3C). This expression pattern is reminiscent of that observed for WUS in Arabidopsis (see Figure 8B). After the initiation and separation of the sepal primordia, ROA expression is no longer visible in floral meristems. RT-PCR also revealed ROA transcript in petals, stamens, and carpels. No expression was detected by in situ hybridization in petals, where the signal may be too low and/or diffuse to be detected. ROA is expressed early in developing anthers, in the region that will later separate the two locules of each theca (Figure 3D). During stamen and pollen development, ROA transcript remains expressed in the same domain, including the septum and stomium, the two tissues controlling pollen dehiscence (Figure 3E). Because roa mutants never flower, the mutant is not informative with respect to the role of ROA in anther development. However, WUS expression in Arabidopsis stamens mirrors the expression of ROA (F. Deyle and T. Laux, unpublished data). The stamens formed in wus-1 mutants produce little pollen (Mayer et al., 1998), which suggests that ROA/WUS is important for male fertility. In carpels, ROA is expressed strongly at the upper end of developing ovules in the nucellus (Figures 3F and 3G). In Arabidopsis, expression of WUS in the nucellus nonautonomously regulates the development of integuments (Gross-Hardt et al., 2002). Fasciated mutants of Antirrhinum have been described previously (Stubbe, 1966), although the affected genes are as yet uncharacterized. ROA expression expands laterally in the large inflorescence meristem of the classical fasciata mutant (Figure 3H and arrowheads in Figure 3I). Therefore, like WUS, the expression domain of ROA is enlarged in a mutant that has an abnormally large meristem. Based on mutant comparison and expression analysis, the ROA gene represents the functional homologue of WUS in Antirrhinum.

Figure 3.

ROA Expression.

(A) Semiquantitative RT-PCR for eight different tissues. The bottom panel shows a control amplification of the constitutively expressed EF1-α cDNA.

(B) to (G) In situ hybridization with a ROA probe.

(B) Longitudinal section through a wild-type inflorescence showing expression in an inflorescence meristem (arrowhead) and a young floral meristem.

(C) Expression in floral meristems at a later stage. br, bract; se, sepal.

(D) Expression between locules (asterisks) of immature anthers.

(E) Late expression between locules of maturing anthers including septum (sp) and stomium (st) and most of the connective.

(F) Longitudinal section through an immature carpel, with strong expression in the developing ovules. pt, placenta.

(G) Longitudinal section through immature ovules showing transcripts in the nucellus (nu) but not in the funiculus (fu) or integuments (in).

(H) fasciata mutant inflorescence showing strong, linear fasciation of the apical meristem.

(I) ROA expression (arrowheads) expands along the large fasciata inflorescence meristem.

Bars = 50 μm in (B) to (G) and (I) and 1 cm in (H).

Figure 8.

Comparison of WSIP1 and WUS Expression Patterns.

WSIP1 and WUS expression patterns were studied by in situ hybridization in inflorescence ([A] and [B]), stamen in the developing locules and immature pollen ([C] and [D]), and flower with immature carpel ([E] and [F]). Bars = 50 μm.

Expression of Truncated WUS in Arabidopsis Phenocopies the Semidominant roa-2 Allele

The semidominant nature of the roa-2 allele in Antirrhinum suggested that a truncated version of ROA, lacking the C-terminal domains, could act as a dominant negative allele. To test this notion, we recreated this allele in Arabidopsis using a truncated WUS fused to the glucocorticoid receptor, making its activity dependent on the presence of dexamethasone (35S:WUSΔ-GR; see Methods). Expression of inducible full-length WUS (35S:WUS-GR) has several clear morphological consequences. Compared with wild-type plants (Figures 4A to 4D), the leaves of induced 35S:WUS-GR (full-length WUS) plants are curled (Figure 4E), the plants are stunted (Figure 4F), and ectopic outgrowths are observed on the pedicels of the flowers (Figures 4G and 4H). However, the apical meristem is maintained, as it is in wild-type plants. Similar results have been reported previously using expression systems giving general overexpression (Brand et al., 2002; Xu et al., 2005). A different phenotype was reported when WUS was ectopically expressed using the CLV1 promoter, leading to the enlargement of the SAM (Schoof et al., 2000). The difference is likely to be attributable to the promoters used and suggests that meristem enlargement is dependent on either spatial or quantitative differences in WUS expression.

Figure 4.

WUS Ectopic Expression in Arabidopsis.

Phenotypic comparison of wild-type Columbia ([A] to [D]) and wus-1 mutant ([M] to [P]) plants with lines overexpressing full-length WUS (35S:WUS-GR) ([E] to [H]) or C-terminal truncated WUS (35S:WUSΔ-GR) ([I] to [L]). Where necessary, some sepals and petals were removed to gain access to inner floral organs. Ectopic outgrowths with stigmatic papillae (arrowhead) form on 35S:WUS-GR receptacles (H). The lines in (I) to (L) have normal leaves and flowers but show meristem termination (I), inducing the early development of lateral branches, as seen in 35S:WUSΔ-GR (arrow in [J]). Note the similarity to wus-1 (N). All plants shown, except in (I) inset, were subject to dexamethasone induction (see Methods). It is possible to increase the frequency of meristem terminations observed in 35S:WUSΔ-GR plants by additionally watering, as required, with a 1 μM dexamethasone solution from the point of sowing. These conditions of increased induction do not lead to meristem termination in 35S:WUS-GR plants. Bars = 1 cm in (A), (B), (E), (F), (I), (J), (M), and (N) and 400 μm in (C), (D), (G), (H), (K), (L), (O), and (P).

In contrast with 35S:WUS-GR, induced 35S:WUSΔ-GR (truncated WUS) plants show meristem termination (Figure 4I and arrowhead in Figure 4J), similar to that observed in homozygous and heterozygous roa-2 and in wus-1 (Figures 4M and 4N). Uninduced 35S:WUSΔ-GR plants display an entirely wild-type morphology (Figure 4I, inset). This finding suggests that 35S:WUSΔ-GR acts as a dominant negative allele, inhibiting endogenous WUS function in the shoot meristem. However, 35S:WUSΔ-GR flowers appear normal (Figures 4K and 4L), in contrast with the occasional flowers produced in wus-1 mutants, which lack some organs and terminate development early (Figures 4O and 4P). Expression of a truncated version of WUS, therefore, can recapitulate the apical meristem termination aspects of the phenotypes conferred by the wus and roa mutants, but it has no influence on flower development.

Expression of WUS, like that of ROA in Antirrhinum, is lower in the seedling SAM and highest in the floral meristems. To test the possibility that 35S:WUSΔ-GR is simply unable to counteract the presence of larger amounts of endogenous wild-type WUS in floral meristems, we analyzed the effects of 35S:WUSΔ-GR in homozygous and heterozygous wus-1 backgrounds. In neither case was 35S:WUSΔ-GR capable of influencing flower development (Figure 5C). It remains possible that, even in heterozygous wus-1 mutants, WUS expression in the floral meristem is too high for the truncated WUS to have an observable effect. Alternatively, the truncated WUS protein could be less able to interfere in flower development as a result of WUS acting in different ways in the SAM and floral meristem. In either case, the conserved C-terminal domain is clearly required for WUS function in the floral meristem, because induced 35S:WUS-GR but not 35S:WUSΔ-GR can complement the loss-of-organ defect observed in wus-1 (Figures 5C to 5E).

Figure 5.

Complementation of the Phenotype Conferred by wus-1 in Lines Overexpressing WUS.

(A) Wild-type flower including four sepals, four petals, six stamens, and terminating in two fused carpels forming the gynoecium.

(B) Typical wus-1 flower missing the gynoecium and terminating in a single stamen.

(C) Flower from the wus-1 line overexpressing the C-terminal truncated version of WUS (35S:WUSΔ-GR).

(D) Complemented wus-1 flower in the line overexpressing full-length WUS (35S:WUS-GR). A full set of floral organs is visible, including a fully grown gynoecium.

(E) Abnormal floral organization in a 35S:WUS-GR plant with multiple flowers initiated from a large fasciated inflorescence meristem (double-headed arrow).

Bars = 1 mm.

The Conserved C-Terminal Domains Are Involved in Protein–Protein Interactions

Because the conserved C-terminal sequences are required to complement the wus-1 mutant and the expression of truncated WUS lacking them produces a dominant wus-like meristem termination phenotype, it was possible that the C-terminal sequences are involved in functionally essential protein–protein interactions. In this model, truncated WUS would be capable of performing part of the role of WUS, such as binding DNA via its homeodomain, but incapable of making the further protein–protein interactions necessary for WUS activity. Thus, truncated WUS would be capable of competing with endogenous, full-length WUS and acting as a dominant negative allele. To test this, we used the yeast two-hybrid assay to identify Arabidopsis proteins that can interact with full-length but not truncated WUS. Using the full-length WUS bait and a normalized yeast two-hybrid cDNA library constructed from Arabidopsis apices, 4.5 million colonies were screened. Sixty-one positive colonies came through all rounds of selection, and the prey cDNAs present in these positive colonies were identified by hybridization and sequencing. The 61 candidate interactors corresponded to just two distinct but related cDNAs, one of which was isolated 60 times.

The two isolated prey cDNAs (At1g15750 and At3g15880; renamed WSIP1 and WSIP2 here) represent different members of a small gene family in Arabidopsis and other plant species (Figure 7C). Both WUS-interacting proteins are predicted to contain N-terminal LisH (for Lis1-homologous) and CTLH (for C-terminal to LisH) domains, a Pro-rich region, and two domains containing multiple tandem WD repeat motifs (Figures 6C and 7A). The LisH domain was originally identified as a conserved motif in a series of proteins associated with human disease (Emes and Ponting, 2001) and subsequently has been shown to promote dimerization, tetramerization, and interaction with other proteins (Cerna and Wilson, 2005). The CTLH domain is a conserved motif that is sometimes found associated with LisH domains (Emes and Ponting, 2001). The LisH domain corresponds to the first half of a domain originally called LUFS that is found at the N-terminal end of the Arabidopsis transcriptional corepressor LEUNIG (LUG), which also contains a C-terminal WD repeat motif (Conner and Liu, 2000). Tandem WD repeats are found in a variety of proteins with a wide range of unrelated functions, where they combine to provide an interface for specific protein–protein interactions (Smith et al., 1999). It is striking that several well-characterized transcriptional corepressors contain an N-terminal LisH domain and a C-terminal WD repeat domain, including yeast Sif2p (Cockell et al., 1998; Cerna and Wilson, 2005), human TBL1 and TBLR1 (Guenther et al., 2000; Li et al., 2000; Zhang et al., 2002), and Arabidopsis LUG (Conner and Liu, 2000). The corepressors Groucho (Drosophila) and TUP1 (Saccharomyces cerevisiae) also contain C-terminal WD repeats and N-terminal tetramerization domains, although their N-terminal domains are unrelated both to each other and to the LisH domain (Chen et al., 1998; Jabet et al., 2000). Transcriptional corepressors rely on specific interactions with transcription factors to identify their target genes (Courey and Jia, 2001). The organizational similarity of the WUS-interacting proteins to these diverse transcriptional corepressors strongly suggests that they are recruited by WUS to repress the expression of target genes that need to remain inactive in the stem cell niche.

Figure 7.

WUS-Interacting Proteins.

(A) Predicted exon (box) and intron (line) structure of WSIP1 and WSIP2. Black shading denotes 5′ and 3′ untranslated regions.

(B) WSIP1 protein structure as determined using SMART4.0 (http://smart.embl-heidelberg.de/) (Letunic et al., 2004): LisH domain (box), CTLH domain (double underlined), Pro-rich region (dotted underlined), and putative WD domains (underlined).

(C) Phylogram of the WSIP family. At, Arabidopsis thaliana; Os, Oryza sativa; Mt, Medicago truncatula CTV.2-like (accession number AC148304); Sb, Solanum bulbocastanum (Q7XA22); Pt, Poncirus trifoliata CTV.2 (AAN62336). The Arabidopsis WUS-interacting proteins are represented in boldface.

Figure 6.

WUS-Interacting Proteins.

(A) Full-length WUS and two modified versions, used as bait in a yeast two-hybrid assay, are shown schematically with conserved domains shaded. WUSΔ lacks the three conserved domains. ΔWUS truncates the homeodomain. All baits were tested against three prey proteins coding for C-terminal truncated versions of WSIP1 (WSIP1Δ1), WSIP2 (WSIP2Δ2), and full-length WSIP1. WSIP1Δ1 was also tested for interaction with ROA and two ROA truncations. ROAΔ lacks the C-terminal conserved domain, and ΔROA truncates the homeodomain.

(B) In vitro confirmation of the interaction between WUS and the WSIP1 protein by glutathione S-transferase (GST) pulldown. Lanes WUS* and WUSΔ* contain in vitro synthesized, radiolabeled WUS and WUSΔ proteins, either alone as controls or after incubation and pulldown with truncated WSIP1-GST fusion proteins (^GWSIP1Δ1) or GST alone as a negative control.

(C) Domains involved in the WSIP1–WUS interaction. The WSIP1 protein is represented schematically, showing the LisH domain (black), the CTLH domain (light gray), the Pro-rich region (dark gray), and two regions (shaded and labeled WD) containing seven and six putative WD motif repeats. Full-length WSIP1-GST (^GWSIP1) fusions and five truncated versions (^GWSIP1Δ1 to ^GWSIP1Δ5) were tested for interaction with radiolabeled WUS (WUS*). The WUS* lane contains an in vitro synthesized, radiolabeled WUS protein control.

In both cases, the cDNAs isolated in the two-hybrid screen were found to be truncated; the isolated cDNAs are referred to here as WSIP1Δ1 and WSIP2Δ1 to distinguish them from the WSIP1 full-length cDNA, which was subsequently obtained from the ABRC stock center. Both isolated prey clones and full-length WSIP1 were tested in the yeast two-hybrid system for interaction with a variety of full-length and truncated WUS proteins. Interaction was observed in all cases, but only when the expressed WUS protein contained the conserved C-terminal domains (Figure 6A). The specificity of the interaction between WSIP1 and WUS was confirmed by GST pulldown in vitro (Figure 6B), but in this assay a small degree of interaction was also observed between WSIP1Δ1 and a C-terminally truncated WUS. Therefore, WUS is capable of interacting with at least two related Arabidopsis proteins, and yeast two-hybrid and pulldown evidence suggests that the conserved C-terminal domains of WUS are important for this interaction. A similar result was obtained when WSIP1 was tested in the yeast two-hybrid screen for interaction with Antirrhinum ROA; once again, a strong interaction was obtained, but the interaction was dependent on the presence of the C-terminal conserved domains of ROA (Figure 6A). These data support our dominant negative model to explain the phenotype observed on expression of truncated WUS (in transgenic Arabidopsis experiments) and ROA (in the Antirrhinum roa-2 allele).

The LisH Domain of WSIP1 Mediates Interaction with WUS

The truncated versions of WSIP1 and WSIP2, which were identified in the yeast two-hybrid screen, both lacked the WD domains, suggesting that these are dispensable for WUS interaction. To test this idea, a series of WSIP1 truncations were tested in a GST pulldown assay for interaction with full-length WUS. Full-length WSIP1 and WSIP1Δ1 retained the ability to interact with WUS, but all deletions in which the LisH domain was removed failed to interact (Figure 6C). These results demonstrate that the interaction between WSIP1 and WUS requires the C-terminal conserved domain of WUS and the N-terminal LisH domain of WSIP1.

Published microarray analyses revealed that both WSIP1 and WSIP2 are expressed throughout Arabidopsis development (Schmid et al., 2005) in all tissues tested. In situ hybridization was used to analyze the temporal and spatial coexpression of WSIP1 and WUS in greater detail, because coexpression would be expected of genes encoding interacting proteins (Figures 8A to 8F). In situ hybridization verified initial RT-PCR findings (data not shown) that WSIP1 is strongly expressed in inflorescences and flowers. Expression of WSIP1 is most strongly observed in inflorescences and floral meristems (Figure 8A). Expression is also observed in developing anthers and ovules (Figures 8C and 8E). In comparison, WUS expression is detected at lower levels and in more restricted tissues (Figures 8B, 8D, and 8F). However, expression of WUS and WSIP1 overlaps in SAMs, floral meristems, and developing carpels. Coexpression of WUS and the WUS-interacting proteins is consistent with an interaction in vivo.

DISCUSSION

Meristem Regulation in Different Species

Although phenotypically similar, the roa mutant is more severe than wus-1. The differences fall into four areas: loss of dorsoventral identity, site of organ production, frequency of internode elongation, and impairment of flowering. These differences can all be explained as a consequence of a more complete loss of SAM activity in the Antirrhinum mutant, possibly attributable to the presence of redundant factors specifying stem cell fate in Arabidopsis. Signals from the SAM are necessary to establish or maintain the dorsoventral asymmetry of leaves (Waites et al., 1998). Because the SAM is lost in roa mutants, the lack of these signals results in a failure to fully establish the dorsoventral axis of the first leaf pair. By contrast, wus-1 mutants produce a normal first pair of leaves, leaving a flat apical surface with some residual meristematic activity (Laux et al., 1996). These differences in the establishment of a dorsoventral axis subsequently affect the site of further organ production. In roa, the next meristems form in the axils of the cotyledons, whereas in wus-1, they arise between the leaves. Meristem production is not observed in the axils of the more severely ventralized petioles of roa leaves. Further evidence for the more severe disruption of the SAM in roa mutants is provided by the different frequency of internode elongation. Mature wus-1 plants have a bushy appearance, reflecting frequent internode elongation. In comparison, roa mutants remain compact, as internode elongation is very rare. Although roa never flowers, abnormal flowers are produced occasionally on wus-1 plants (Laux et al., 1996). The ter mutant of Petunia also produces only occasional flowers with a reduced organ number (Stuurman et al., 2002). The absence of flower production in roa could be caused by a shorter functional duration of the SAM in Antirrhinum, resulting in a reduced accumulation of factors such as FLORICAULA (FLO) (Coen et al., 1990). Flowering is also more severely inhibited in Antirrhinum meristem identity mutants, such as squamosa (Huijser et al., 1992) and flo, than in the corresponding Arabidopsis mutants apetala1 (Mandel et al., 1992) and lfy (Weigel et al., 1992). Therefore, the inability of roa mutants to flower could be attributable to the stricter control of flowering in Antirrhinum combined with a more complete loss of SAM activity. Together, the evidence indicates that WUS-like genes play an evolutionarily conserved central role in SAM maintenance in angiosperms but that subtle differences will be apparent in different species. It is likely that differences in phenotype severity and growth habit made it easier to identify the weak and variable semidominant effect of the roa-2 allele in Antirrhinum.

Sequence Conservation and Dominant Negative Models

Sequence comparisons between the orthologues WUS, TER, and ROA identified three conserved domains apart from the DNA binding domain: an acidic domain and two additional short motifs. A similar acidic domain is also found in the WOX genes WOX1, WOX4, and WOX5 (Haecker et al., 2004). The first of the two short motifs, the WUS box, is present in all of the WOX genes except WOX13. The second short motif, ASLELXLN, is found only in WUS and the WUS orthologues TER and ROA. We have shown that the expression of truncated WUS or ROA, lacking all three domains, results in a dominant negative phenotype. Furthermore, protein–protein interaction analyses have shown that removal of all three domains reduces or eliminates the interaction between WUS and the two related WUS-interacting proteins. Further experiments will be required to dissect the exact role of each component of the conserved C-terminal domain sequences in this interaction. It is interesting, however, that the ASLELXLN domain bears a striking similarity to motifs that act as potent transcriptional repressors in plants. The EAR motif (L/FDLNL/FxP) is required for the repressive function of class II ethylene-responsive element binding factors (ERFs) and TFIIA-type zinc finger repressors (Ohta et al., 2001). A related minimal repression domain (DLELRL) was identified in the zinc finger factor SUPERMAN and was further characterized as XLxLXL (Hiratsu et al., 2004). The repression motif found in Aux/IAA proteins was also characterized as LxLxL (Tiwari et al., 2004). Active repressors are thought to work by interacting with other proteins, such as transcription factors, coactivators, and corepressors. Therefore, the conservation of this motif in WUS lends support to the idea that WUS can act as a transcriptional repressor and provides a potential site for interaction with WSIP corepressors (see below). Because both the WOX proteins and the WUS-interacting proteins described here belong to small protein families in Arabidopsis, it is possible that other interactions between different members of these families might also be found. There is a precedent for family sequence-specific interactions between transcriptional corepressors and their associated transcription factors. For example, Groucho interacts with Hairy family basic helix-loop-helix factors, via a conserved C-terminal WRPW motif in the Hairy family proteins, to exert its repressive effect (Paroush et al., 1994).

Both expression of truncated ROA in the semidominant roa-2 allele of Antirrhinum and overexpression of truncated WUS in Arabidopsis leads to a dominant meristem termination phenotype, reminiscent of that seen in homozygous roa and wus-1 mutants. We show here that the conserved region missing from the truncations is involved in specific protein–protein interactions with proteins that show similarity to transcriptional corepressors in other systems. This raises the possibility that WUS functions, at least in part, by recruiting corepressors via its conserved C-terminal domains and bringing them to the sites of transcriptional regulation of WUS target genes. Interestingly, the truncated proteins only affect SAM maintenance and appear not to interfere with normal flower development. We cannot exclude the notion that both truncated WUS and truncated ROA are subject to a floral meristem–specific negative regulation that does not affect the full-length versions. However, it is more likely that the inability of the truncated versions of ROA and WUS to affect normal flower development simply results from the higher levels of expression of wild-type, endogenous ROA or WUS present in floral meristems. We favor an alternative model in which WUS acts differently in the indeterminate SAM and the determinate floral meristem. For example, WUS is an activator of the homeotic gene AG in floral meristems but not in SAMs (Lenhard et al., 2001; Lohmann et al., 2001) (see below).

A Model for WUS Function

The interaction between WUS and the transcriptional corepressor-like WSIP proteins suggests a model in which WUS recruits WSIP to prevent the expression of target genes promoting differentiation, thus maintaining the undifferentiated state of the SAM. However, WUS also acts as an activator of gene expression. Such an alternative mode of function could be achieved either by disruption of the WUS–WSIP interaction or by the WSIP proteins acting like the human LisH/CTLH/WD corepressor TBLR1 and mediating an exchange between repressor and activator complexes (Hong and Privalsky, 2000). In the case of the activation of AG (Lenhard et al., 2001; Lohmann et al., 2001), the requirement for adjacent WUS and LFY binding sites in the regulatory intron of AG (Hong et al., 2003) makes it conceivable that the role of WUS could switch from repression of AG in the SAM to activation of AG in the floral meristem by a LFY-mediated disruption of the WUS–WSIP interaction. Although we do not know at present how the WUS–WSIP interaction could act to repress target genes, transcriptional corepressors such as Groucho can exert their repressive activity via histone deacetylation (Courey and Jia, 2001). If the WSIP proteins use a similar mechanism, there should be a link between histone acetylation status and SAM maintenance. Antisense reduction of histone deacetylase activity results in the formation of aerial rosettes, reminiscent of those seen in wus-1, although other pleiotropic effects were also observed (Tian and Chen, 2001). Termination and differentiation of the SAM also occurs in plants grown in the presence of the histone deacetylase inhibitor trichostatin A (Figure 9). These and other models for WUS action can be tested once we have identified the positively and negatively regulated target genes.

Figure 9.

Disruption of Meristem Maintenance by the Histone Deacetylase Inhibitor Trichostatin A.

(A) Arabidopsis (Columbia) plant grown in vitro for 45 d.

(B) Scanning electron micrograph of the SAM of the plant in (A).

(C) Plant grown under the same conditions but on medium supplemented with 10 μM trichostatin A. This plant produced only two mature leaves, as opposed to the 22 leaves produced in the plant shown in (A).

(D) Scanning electron micrograph shows the demise of the SAM in (C) and the production of small abnormal leaves at the base of the cotyledon petiole.

Bars = 250 μm in (A) and (C) and 30 μm in (B) and (D).

METHODS

Plant Material

Two Antirrhinum majus mutants, rosulata and rosulans, initially described as nonflowering, compact plants lacking stem elongation (Stubbe, 1974) (Figure 1), were obtained from the Gatersleben seed collection. roslulata was identified as a recessive mutant, and rosulans was described as semidominant. Our genetic and molecular analysis revealed that rosulata and rosulans are allelic, and we have renamed them roa-1 and roa-2, respectively. The semidominance of the roa-2 (rosulans) allele displays low penetrance and is influenced by the culture temperature. A third mutant, originally named rosularis (Stubbe, 1974), was also characterized and found to be allelic and phenotypically identical to roa-2 and to share an identical lesion in the ROA gene (data not shown). The fasciata line was kindly provided by Richard Waites (University of York). Arabidopsis thaliana transformation was performed in Columbia plants by the floral dip method (Clough and Bent, 1998). The previously described wus-1 mutant (Mayer et al., 1998) was used in the complementation experiment. Arabidopsis seeds were cold-treated at 4°C for 2 d before culture in a growth chamber. For the trichostatin A experiment, plants were grown in Petri dishes on Murashige and Skoog medium, 3% sucrose, and 1% agar, pH 5.6, with or without 10 μM trichostatin A (Sigma-Aldrich; T8552). All plants were grown at 24°C with a 16-h photoperiod.

ROA Isolation and Analysis

The ROA cDNA was isolated from a λNM1149 inflorescence cDNA library using a full-length WUS probe. The 5′ end was obtained by RNA ligase–mediated rapid amplification of cDNA ends (Ambion FirstChoice). The ROA cDNA was used to hybridize a DNA gel blot with wild-type, roa-1, and roa-2 homozygous and heterozygous samples as well as roa-1 revertant shoot DNA, digested with EcoRI, HindIII, or BclI. Polymorphisms were identified in roa-1 samples with all three enzymes used. Heterozygosity of the polymorphisms in the roa-1 revertant samples revealed the presence of a transposon in the ROA second intron. The ROA genomic sequence was obtained by PCR on genomic DNA using cDNA flanking primers. Thermal asymmetric interlaced PCR (Liu and Whittier, 1995) was used to isolate the roa-1 transposon insertion site.

Expression Analysis

ROA expression was assessed by semiquantitative RT-PCR on first-strand cDNA (Omniscript reverse transcriptase; Qiagen) using primers F12 (5′-AACAAGTTCTCTAACATGAATAGTG-3′), which spans the first intron, and R8 (5′-ATTGAGGGTGAGTTCCAGTGAAG-3′) in the final exon. Material for in situ hybridization was prepared and hybridized with digoxigenin-labeled RNA probes as described (Zachgo et al., 2000). ROA and WUS templates for transcription were either full-length or N-terminally truncated, removing the homeodomain coding region. Both probes gave the same expression pattern. The WSIP1 RNA probe comprised only the C-terminal 280 bp of coding sequence, a region that shows maximum sequence divergence across the family.

Construction of 35S:WUS-GR and 35S:WUSΔ-GR

PCR-generated full-length and C-terminal truncated WUS primers (5′-ACTCTAGACTGACTATGGAGCCGCCACAGCATCAG-3′ and 5′-TTAGGATCCGCGTTCAGACGTAGCTCAAGAG-3′ or 5′-ATTAGGATCCGCACCTTCTAGACCAAACAGAGG-3′) (Figure 2) were inserted into pBi-ΔGR (Simon et al., 1996) to create in-frame C-terminal GR fusions. The use of the pBi-ΔGR system enables nuclear localization to be regulated by the addition of dexamethasone (Simon et al., 1996).

Dexamethasone Treatment

Dexamethasone (Sigma-Aldrich; D-1756) was prepared as a 10 mM stock solution in 100% ethanol and stored at −20°C. Transformed lines were sprayed five times per week with a solution of 30 μM dexamethasone and 0.005% Silwet L-77 in pure water. Spraying commenced 1 week after germination and was continued throughout the experiment.

Microscopy

Samples for scanning electron microscopy were fixed, dehydrated, critical point-dried, dissected, and sputter-coated with gold as described (Bowman et al., 1989) and observed with a Camscan series III scanning electron microscope. Confocal microscopy samples were stained with propidium iodide and analyzed according to Running et al. (1995) using a Leica TCS SP microscope.

Yeast Two-Hybrid Analysis

A range of WUS and ROA bait plasmids, including full-length and two truncated versions, were produced (Figure 6) by PCR with the addition of flanking EcoRI and BamHI restriction sites: WUS (primers WUSF1 and WUSR1), WUSΔ (primers WUSF1 and WUSROL), ΔWUS (primers WUSF3 and WUSR1), and ΔROA (primers ROAF2 and ROAR1). The vectors pBDWUS to pBDΔWUSΔ were produced by ligating the EcoRI-BamHI DNA fragments into pGBT9 (Clontech) in-frame with the GAL4 DNA binding domain. All vectors were transformed into the Y187 and AH109 yeast strains and tested for autoactivation. pBDWUS was found to be a weak autoactivator, but this could be dampened by adding 5 mM 3-aminotriazole (3AT) to the culture medium. A full-length WSIP1 prey vector was constructed by recombining a PCR-generated fragment into pDEST-GAD (pGADT7-Rec [Clontech] Gateway-enabled vector). The vector was transformed into the AH109 yeast strain.

Library Screening

Yeast strain Y187 containing the bait vector pBDWUS was mated overnight with an amplified normalized Arabidopsis apical meristem library (pGADT7-Rec) in the yeast mating strain AH109. The library was kindly provided by Simona Masiero and Hans Sommer (Max Planck Institute, Köln, Germany). Mating and screening were done according to the Clontech MATCHMAKER protocol A (K1615-1). Yeast diploid selection was performed on minimal medium (SD/−Ade/−His/−Leu/−Trp/+X-α-Gal) plus 5 mM 3AT, and ∼1500 putative positives were selected. Subsequent stringent testing on 50 mM 3AT, hybridization, sequencing, and retransformation into the bait vector were used to reduce this to 61 distinct prey cDNAs, of which 60 corresponded to WSIP1 and 1 corresponded to WSIP2.

Small-Scale Two-Hybrid Analysis

Bait (Y187) and prey (AH109) yeast strains were mated by growing 5 μL of bait strain culture on solid medium for 1 d, then overlaying 5 μL of prey strain culture and incubating at 30°C for 2 d. Mated diploids were selected on SD/−Leu/−Trp for 4 d. Interaction was tested by growing 5 μL of the resuspended diploid strain (100 cells/μl) on minimal medium (SD/−Ade/−His/−Leu/−Trp/+X-α-Gal) plus 20 mM 3AT at 30°C for 4 d.

Phylogenetic Analysis

A distance tree was generated using PAUP 4.0b (Swofford, 2002) from a ClustalX full-length protein alignment using the neighbor-joining algorithm (see Supplemental Table 1 online for alignment). LUG was used as an outgroup. Branch bootstrap values are indicated based on 1000 replications.

In Vitro GST Pulldown

Radiolabeled WUS and WUSΔ proteins were produced with the TNT coupled reticulocyte lysate systems (Promega) using 0.5 μg of pASL/WUS and pASL/WUSΔ expression vectors, respectively. Both vectors were constructed by digesting pBDWUS and pBDWUSΔ with EcoRI and BamHI and ligating the fragments generated into pASL514 (Causier et al., 2003). A range of GST fusion WSIP1 constructs were used, including full-length and five truncated versions (Δ1 to Δ5) (Figure 6). The six gene fragments were produced by PCR with the addition of flanking restriction sites or Gateway sequences. All fragments were cloned into pGEX-4T-1 (Amersham Pharmacia Biotech) or recombined into pDEST-GEX (a Gateway-enabled pGEX derivative) in-frame with the GST sequence. All vectors were transformed into Escherichia coli BL21 cells (Stratagene). Fusion proteins were produced and GST pulldowns performed in combination with the radiolabeled WUS and WUSΔ proteins as described previously (Causier et al., 2003).

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers AY162209 (ROA), AF481951 (TER), At2g17950 (WUS), At1g15750 (WSIP1), and At3g15880 (WSIP2).

Supplemental Data

The following material is available in the online version of this article.

• Supplemental Table 1. ClustalX Full Protein Alignment.