The Arabidopsis floral organ identity genes APETALA3 (AP3) and PISTILLATA (PI) encode related DNA-binding proteins of the MADS family. Considerable evidence supports the hypothesis that a heterodimer of AP3 and PI is an essential component of B class activity. All ap3 and pi alleles characterized to date exhibit equivalent phenotypic defects in both whorls 2 and 3. In strong ap3 and pi mutants, petals and stamens are missing and sepals and carpels develop in their place. Weak ap3 and pi mutants exhibit partial conversions of petals to sepals and stamens to carpels. In this report, we describe the isolation and characterization of pi-5, an unusual B class mutant that exhibits defects in whorl 2 where sepals develop in place of petals, but third whorl stamens are most often normal. pi-5 flowers resemble those from 35S::SEP3 antisense plants. pi-5 contains missense mutation in the K domain (PIE125K). PIE125K exhibits defects in heterodimerization with its partner protein AP3. Via a reverse yeast two-hybrid screen, AP3K139E was isolated as a compensatory mutant of PIE125K. The compensatory interaction between PIE125K and AP3K139E is observed both in yeast two-hybrid assays and in planta. On its own, AP3K139E exhibits defects in specifying both petal and stamen identity. In addition, PIE125K is defective in interaction with SEPALLATA proteins in both two- and three-hybrid assays suggesting that PIE125K is defective in forming higher order complexes of MADS proteins. The decreased concentration of PI/AP3/SEP complexes offers an explanation for the petal defects observed in both pi-5 and 35S::SEP3 antisense plants.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Most ABC genes, and the SEP genes encode proteins belonging to the MADS family of transcription factors (Riechmann and Meyerowitz, 1997; Theißen 2001). The A class protein AP1, the B class proteins AP3 and PI, the C class protein AG, and the three SEP proteins SEP1, SEP2, and SEP3, all contain the conserved 56 amino acid MADS domain. The functionally characterized plant MADS proteins have a stereotypic organization of functional domains (Figure 1). The MADS domain is located at the amino-terminal end of the protein and encodes DNA binding and dimerization functions. A second conserved domain, the K domain, is located in the central region. Secondary structural analyses predict that the K domain encodes three amphipathic α-helices, K1, K2, and K3. Yeast two-hybrid assays and domain swapping experiments suggest that the K domain is important for dimerization and functional specificity (Davies et al., 1996; Fan et al., 1997; Moon et al., 1999; Riechmann and Meyerowitz, 1997; Yang et al., 2003). Between the MADS and K domains is a variable region referred to as the I (intervening) domain. The carboxy-terminal (or C) domain has been postulated to mediate the formation of higher order MADS multimers (Egea-Cortines et al., 1999; Honma and Goto, 2001). The C domain of a subset of MADS proteins has also been demonstrated to encode a transcriptional activation domain (Cho et al., 1999; Honma and Goto, 2001; Moon et al., 1999).
In this study, we describe the isolation and characterization of pi-5, an unusual B class mutant. To date, 12 ap3 and four pi alleles have been described and these alleles can be ordered in an allelic series. Strong ap3 and pi alleles (e.g. ap3-3 (Jack et al., 1992) and pi-1 (Bowman et al., 1989)) result in complete conversions of petals in the second whorl to sepals, and stamens in the third whorl to carpels. In weak alleles (e.g. ap3-1 (Bowman et al., 1989; Sablowski and Meyerowitz, 1998), ap3-11 (Yi and Jack, 1998)), partial organ identity transformations are present in both whorls 2 and 3. pi-5 is unlike all other ap3 and pi alleles described to date in that it exhibits strong phenotypic defects in specifying petal development but not stamen development. pi-5 is not the only floral organ identity mutant with differential phenotypic effects in different floral organs. The C class mutant, ag-4, also exhibits differential effects (Sieburth et al., 1995). In ag-4 mutants, the third whorl develops as stamens or stamen-like organs but the fourth whorl develops as sepals and the flowers exhibit a loss of determinacy. The ag-4 mutation affects the splicing of AG mRNA and leads to the deletion of 12–14 amino acids in the K domain. In molecular terms, it is not clear how AG-4 functions to direct stamen development, but not carpel development and floral determinacy.
A second example of a floral mutant that exhibits an organ identity defect in a single whorl comes from studies in petunia. Like pi-5, the green petals (gp) mutant exhibits a transformation from petals to sepals in whorl 2 but the whorl 3 stamens are normal (Angenent et al., 1995; Halfter et al., 1994). By sequence comparison, GP is a putative ortholog of AP3. RNA for GP is localized to both whorls 2 and 3, but gene function is required only in whorl 2. Two petunia genes, FBP1 and pMADS2, are PI homologs (van der Krol et al., 1993). It is postulated that an AP3-like gene in addition to GP functions in whorl 3 but not whorl 2 (Tsuchimoto et al. 2000). Thus, in petunia there appears to have been a diversification of the function of B class genes after a gene duplication event. Such a gene duplication has not occurred in Arabidopsis.
Molecular analysis presented here suggests that the PI protein encoded in pi-5 mutants (PIE125K) is defective in dimerization with its partner protein AP3. PIE125K is also defective in interaction with the SEP proteins in both yeast two- and three-hybrid assays. Given the similarity in phenotype between pi-5 and 35S::SEP3 antisense plants, we propose that a decreased concentration of AP3/PI/SEP multimers is the cause of the defects in petal development in pi-5. A compensatory mutant, AP3K139E, isolated in a reverse yeast two-hybrid screen has in vivo defects in specifying petal and stamen organ identity suggesting that the region of K2 from amino acids 125–139 is important for the functional specificity of AP3 and PI.
pi-5 exhibits a novel B class mutant phenotype
pi-5 was isolated as an enhancer of the weak ap3 allele ap3-11 (Yi and Jack, 1998). In a wild-type AP3 background, pi-5 exhibits a novel B class mutant phenotype. At 23°C, third whorl stamens of pi-5 flowers are almost completely wild type (Figure 2b) but second whorl organs resemble wild-type sepals (compare Figure 3a,d). Although the third whorl of pi-5 flowers is primarily staminoid, these organs sometimes develop with slight carpelloid features, such as the appearance of stigmatic papillae at the tips of the anthers causing them to be pointed rather than rounded at the tips (compare Figure 3c,f). Similar to other B class mutants, the severity of the pi-5 phenotype varies acropetally with flowers developing later on the inflorescence exhibiting a more severe mutant phenotype in the third whorl. Although some third whorl organs develop with carpelloid characteristics, the vast majority of third whorl organs are primarily staminoid (Table 1). In summary, pi-5 functions differentially in whorls 2 and 3 resulting in loss of petal organ identity in whorl 2 while stamen organ identity in whorl 3 is largely unaffected.
Table 1. Third whorl organs of pi-5 flowers at 23°C
1–10 (n = 92)
11–20 (n = 98)
21–30 (n = 93)
>30 (n = 96)
The phenotype of pi-5 suggests that B class activity is differentially active in whorls 2 and 3. One trivial explanation for the lack of B class activity in whorl 2 is that PI is not expressed in whorl 2. To test whether PI RNA is expressed in whorl 2 of pi-5 flowers, we performed in situ hybridization. As in wild-type flowers, PI RNA is expressed in precursor cells of the petals, stamens, and carpels of pi-5 flowers during the early stages of flower development demonstrating that the defects in pi-5 are not due to the failure of the gene to be transcribed in the second whorl (Figure 4a,b). However, during later stages of flower development, PI RNA is not detectable in the second whorl of pi-5 flowers, but is detectable at high levels in third whorl stamens (Figure 4c). The failure to detect PI RNA in the second whorl at later floral stages is likely due to disruption of the PI autoregulatory circuit. In the third whorl, by contrast, PI autoregulation appears normal. Based on this, we conclude that PI RNA is transcribed normally in whorl 2 in pi-5 flowers at early floral stages thus the failure of PI to function is not due to a failure of PI to be expressed in whorl 2.
To distinguish whether the defects in pi-5 are specific to whorl 2, or alternatively, are specific to the petal developmental program, we crossed pi-5 to the C class mutant ag-3. In ag-3 mutants, petals develop in whorl 3 in place of stamens and whorl 4 develops as a new flower that consists of a repetition of sepals and petals. ag-3 pi-5 mutant flowers consist entirely of sepals (Figure 2d) and are identical to the double mutant between the strong pi-1 allele and ag-3 (Bowman et al., 1991). The phenotype of ag-3 pi-5 demonstrates that pi-5 is unable to direct petal development in whorl 3. Based on this, we conclude that pi-5 is primarily defective in directing the petal developmental pathway.
PI-5 contains a missense mutation in K domain and is defective in heterodimerization with AP3
To gain insight into potential molecular defects in pi-5, we sequenced the genomic copy of the PI gene from pi-5 as well as an RT–PCR product amplified from pi-5 inflorescence RNA. Sequencing revealed the presence of a single nucleotide change that results in a missense mutation (E125K) in the K domain at a position that is conserved as glutamic acid in most plant MADS proteins (Theißen et al., 2000). In PIE125K, an acidic amino acid is substituted by a basic one. In the meristem identity protein CAULIFLOWER (CAL), a glutamic acid at the corresponding position is mutated to lysine in the strong cal-4 allele suggesting that Glu-125 is important for the function of other MADS proteins as well (Kempin et al., 1995).
To confirm that the point mutation in the K domain is the direct cause of phenotypic defects in pi-5, a cDNA containing the PIE125K mutation was ecotopically expressed in Arabidopsis under the control of the constitutive 35S promoter from cauliflower mosaic virus. In 35S::PIwt, first whorl organs develop as petaloid sepals rather than sepals (Figures 2f and 3g) (Krizek and Meyerowitz, 1996). Unlike 35S::PIwt, whorl 1 floral organs in 35S::PIE125K are primarily sepaloid (Figures 2e and 3h). Furthermore, when 35S::PIE125Kis crossed to a pi mutant, the second whorl floral organs are sepaloid (Figures 2h and 3k) indicating that PIE125K fails to rescue the petal defects of pi mutants. This is in sharp contrast to 35S::PIwt, which fully rescues second whorl petal identity in a pi background (Figure 2g). The first whorl phenotype of 35S::PIE125K, as well as the first and second whorl phenotypes of 35S::PIE125Kpi, provides evidence that PIE125K is defective in specifying the petal developmental pathway.
To gain insight into the molecular defects in PIE125K, we tested protein–protein interaction utilizing yeast two-hybrid assays. Full-length versions of the Antirrhinum majus homologs of AP3 and PI (DEF and GLO) interact strongly and specifically in yeast (Davies et al., 1996). Unlike DEF and GLO, full-length AP3 and PI fail to interact strongly in yeast (Yang et al., 2003). However, a strong interaction is detected in yeast with MADS domain-deleted versions of AP3 and PI (AP3(IKC) and PI(IKC)). AP3 and PI homodimer combinations do not interact above background levels demonstrating that AP3 and PI homodimers do not form in yeast. Similarly, neither AP3 nor PI interacts strongly with the floral organ identity MADS proteins AP1 or AG (Yang et al., 2003).
To test whether PIE125K is defective in protein–protein interaction with AP3, we tested the interaction between AP3(IKC) and PIE125K(IKC). PIE125K(IKC) interacts with AP3(IKC) at 25–40% of the activity of wild type demonstrating that PIE125K is defective in heterodimerization with AP3 (Figure 5a).
We also tested the ability of PIE125K to bind to a MADS-binding site together with wild-type AP3 in an electrophoretic mobility shift assay (EMSA) (Figure 6a). Wild-type AP3 together with wild-type PI bind with high affinity to a MADS-binding sequence (Hill et al., 1998; Tilly et al., 1998), but neither AP3 nor PI alone exhibits detectable DNA binding supporting the hypothesis that AP3 and PI form a highly specific heterodimer. PIE125K/AP3 combinations are able to bind to a MADS consensus sequence, but more weakly than wild type, at 20–50% of wild-type levels (compare lanes 4 and 5 in Figure 6a).
Low temperature or overexpression of AP3 or PIE125K partially suppresses the pi-5 phenotype
Several lines of in planta evidence support the hypothesis that the heterodimerization defects of PIE125K are the cause of the petal to sepal phenotype in whorl 2. First, both the second and third whorl phenotypes in pi-5 are temperature sensitive. Compared to pi-5 plants grown at 23°C, plants grown at 16°C exhibit whitish petaloid sectors in the second whorl organs. These second whorl organs are distinct from wild-type petals since they are shorter than wild-type petals and contain green sectors and elongated cell types characteristic of sepals (Figures 2c and 3e). A second line of evidence that is consistent with the phenotypic defects in pi-5 being due to the defects in dimerization of PIE125Kwith AP3 is the fact that overexpression of either AP3 or PIE125K partially suppresses the pi-5 petal defects. In the ectopic expression line 35S::AP3, flowers exhibit a conversion of carpels to stamens in whorl 4 (Figure 2m) (Jack et al., 1994). In 35S::AP3 pi-5, the second whorl defects of pi-5 are partially suppressed (Figures 2o and 3j). SEM analysis reveals that second whorl organs of 35S::AP3 pi-5 consist of cell types intermediate between the elongated cells characteristic of sepals and the small, rounded, hexagonally packed cells characteristic of petals (compare Figure 3j with Figure 3d).
Similarly, overexpression of PIE125K leads to partial PI function in whorl 2; this is assayed by analyzing 35S::PIE125K in a pi background (Figures 2h and 3k). The second whorl organs of 35S::PIE125Kpi-4 flowers exhibit features of both sepals and petals (Figure 3k). Second whorl organ identity is more dramatically rescued when both AP3 and PIE125K are overexpressed simultaneously (Figure 2l). Although the first whorl organs of both 35S::AP3 and 35S::PIE125K develop as sepals, the first whorl of doubly transgenic 35S:: PIE125K 35S::AP3 (Figure 2j) or 35S:: PIE125K 35S::AP3 pi-4 (Figures 2l and 3l) develops as organs that resemble petals.
It is surprising that the fourth whorl of 35S::AP3 pi-5 develops as carpels rather than stamens (Figure 2o) since it is likely that both AP3 and PIE125K are expressed in whorl 4. A similar difference is observed between the development of the third and fourth whorls in 35S:: PIE125K 35S::AP3 pi-4 (Figures 2l and 3l); the third whorl develops as stamens while the fourth whorl is primarily carpelloid.
In summary, the development of petaloid features in whorl 2 of 35S::AP3 pi-5, 35S::PIE125Kpi-4, 35S::AP3 35S::PIE125Kpi-4, and pi-5 grown at 16°C demonstrates that PIE125K can function to specify petal identity in whorls 1 and 2. However, lower temperature or overexpression of AP3 or PIE125K results in only partial rescue of the petal developmental pathway.
AP3K139E, a compensatory AP3 mutant of PIE125K
Since PIE125K exhibits an impaired ability to heterodimerize with AP3, we screened for an AP3 mutant that could compensate for the PIE125K dimerization defect. To isolate compensatory AP3 mutants, we used a variation of the yeast two-hybrid assay. In yeast two-hybrid assays, wild-type AP3 and wild-type PI interact strongly resulting in a blue colony color phenotype on an X-gal plate when lacZ is used as a reporter. PIE125K interacts less well with AP3 in the yeast two-hybrid assay resulting in light blue colonies on an X-gal plate. We randomly mutagenized the I, K, and C domains of AP3 via error-prone PCR mutagenesis (Fromant et al., 1995; Vidal, 1997) and screened for mutants, which when placed together with PIE125K, exhibited a dark blue colony color. We screened several thousand colonies and identified a small number of dark blue colonies. Upon retesting, one AP3 mutant exhibited a greater interaction with PIE125K than PIE125K did with wild-type AP3 (Figure 5a). Sequencing revealed a mutation in the K domain of AP3 resulting in a change of a positively charged lysine to negatively charged glutamic acid (K139E). Specifically, AP3K139E interacts with PIE125K at approximately 85% of the levels observed for wild-type AP3 and PI. By comparison, PIE125K interacts with AP3wt at only 25–40% of AP3wt/PIwt levels and AP3K139E interacts with PIwt at approximately 70% of AP3wt/PIwt levels. These results demonstrate that AP3K139E interacts with PIE125K in a compensatory manner in yeast.
The compensatory interaction between AP3K139E and PIE125K is also observed in planta. In 35S::AP3K139E, multichambered gynoecia develop in the fourth whorl demonstrating that AP3K139E possesses partial activity (Figure 2p). As described in the previous section, in 35S::AP3 pi-5 flowers, wild-type carpels develop most often in whorl 4 indicating that the combination of AP3wt and PIE125K fails to result in B class activity in whorl 4 (Figure 2o). By contrast, in 35S::AP3K139Epi-5 flowers (Figure 2r), fourth whorl organs are more staminoid and staminoid carpels develop more frequently than in 35S::AP3K139E or 35S::AP3 pi-5. We interpret this enhanced conversion of carpels to stamens in 35S::AP3K139Epi-5 as evidence of an in planta compensatory interaction between AP3K139E and PIE125K.
Unlike PIE125K, AP3K139E exhibits phenotypic defects in both petal and stamen development in a wild-type PI background. As described above, 35S::AP3K139E exhibits a much weaker whorl 4 phenotype than 35S::AP3 (compare Figure 2m,p). The defect of AP3K139E in specifying stamen organ identity in whorl 3 is revealed in 35S::AP3K139Eap3-3. 35S::AP3wt fully rescues stamen development in ap3-3, but 35S::AP3K139E only partially rescues stamen development in ap3-3 (compare Figure 2n,q). AP3K139E is also defective in specifying petal identity; this is best exemplified by the whorl 1 phenotype of 35S::AP3K139E 35S::PI (Figure 2s). In 35S::AP3 35S::PI, petals develop in both whorls 1 and 2 (Figure 2i). By contrast, in 35S::AP3K139E 35S::PI, the first whorl floral organs develop as petaloid sepals. Similarly, in 35S::AP3K139E 35S::PI ap3-3 (Figure 2t), short greenish floral organs develop in both whorls 1 and 2 indicating that AP3K139E is defective in specifying petal development.
The importance of Lys-139 in AP3/PI heterodimerization is further confirmed by another AP3 mutant, AP3K139A,K141A, constructed via site-specific mutagenesis. In a yeast two-hybrid assay, AP3K139A,K141A interacts with PIwt at approximately 70% of wild-type levels (Figure 5a). AP3K139A,K141A exhibits similar defects to AP3K139Ein planta; specifically, the first whorl of 35S::AP3K139A,K141A 35S::PI is sepaloid and is similar to the phenotype of first whorl organs in 35S:AP3K139E 35S::PI (data not shown). Surprisingly, AP3K139A,K141A can compensate for the dimerization defects of PIE125K in yeast; AP3K139A,K141A/PIE125K interacts at 75% of wild-type levels, much higher than AP3wt/PIE125K, which interacts at only 30% of wild-type levels (Figure 5a).
PIE125K is defective in interacting with SEPALLATA proteins
Higher order complexes of MADS proteins (e.g. trimers, tetramers) have been postulated to play a functional role in specifying floral organ identity (Honma and Goto, 2001; Pelaz et al., 2001a; Theißen, 2001). The three SEP genes, SEP1, SEP2 and SEP3, function redundantly and are necessary for petal and stamen development. One of the SEP proteins, SEP3, has been shown to interact with AP3/PI in a yeast three-hybrid assay (Honma and Goto, 2001). When AP3, PI, SEP3, and AP1 are ectopically expressed together in Arabidopsis plants, cauline leaves are converted to petal-like organs (Honma and Goto, 2001; Pelaz et al., 2001a). Taken together, these results suggest that a MADS multimer consisting of AP3, PI, SEP3, and AP1 is sufficient to direct petal identity. To test whether PIE125K is defective in forming higher order MADS complexes, we tested the ability of PIE125K to interact with SEP proteins. Before testing PIE125K in yeast three-hybrid assays, we tested the interaction of the SEP proteins with AP3 and PI in yeast two-hybrid assays as a control. Previous reports, utilizing MADS domain-containing versions of AP3 and PI, demonstrated that the SEP proteins did not interact strongly with AP3 or PI in yeast two-hybrid assays. However, using MADS domain-deleted versions of PI, we detected an interaction between PI/SEP1, PI/SEP2, and PI/SEP3 in yeast two-hybrid assays without the mediating effects of AP3 (Figure 5b, data not shown). Since we could detect an interaction between the SEP proteins and PI, we tested whether PIE125K was defective in interaction with the SEP proteins in a yeast two-hybrid assay. PIE125K interacts with all three SEP proteins at around 25% of wild-type levels (Figure 5b, data not shown).
PIE125K is also defective in interaction with AP3 and SEP1 (or SEP3) in yeast three-hybrid assays. The combination of AP3(IKC), PI(IKC), and SEP1(MIKC) yields higher activity than the sum of the SEP1 + PI and SEP1 + AP3 interactions supporting the hypothesis that these three proteins interact in a higher order complex (Figure 5b, data not shown). When PIE125K(IKC) is substituted for PI(IKC) in the three-hybrid assay, the level of interaction decreases suggesting that PIE125K is defective in forming higher order complexes with AP3 and SEP1 (or SEP3).
We also performed the yeast three-hybrid experiment utilizing full-length AP3-DB and PI. To detect interaction of full-length proteins we made use of the his3 reporter. If GAL4-AD and GAL4-DB are brought together to form an active transcriptional complex, the HIS3 gene is transcribed and yeast cells are able to grow in the absence of histidine. In this experiment, we fused SEP1(MIKC) to GAL4-AD and AP3(MIKC) to GAL4-DB. In a yeast two-hybrid assay, SEP1(MIKC)-AD and AP3(MIKC)-DB do not interact as measured by the failure to allow growth in the absence of histidine in the presence of 5–25 mm 3-aminotriazole (3AT) (Figure 5c). However, when full-length PI(MIKC), fused to neither the GAL-AD nor GAL4-DB, is placed together with SEP1(MIKC)-AD and AP3(MIKC)-DB, yeast are able to grow in the absence of histidine and in the presence of 3AT (Figure 5c). By contrast, PIE125K(MIKC) was not able to support growth under these conditions in this version of the three-hybrid assay demonstrating that PIE125K is defective in forming ternary complexes with full-length AP3 and SEP3. Similar results were obtained when SEP3(MIKC) was substituted for SEP1(MIKC).
In this report, we describe the isolation and characterization of pi-5, a B class mutant that exhibits differential function in directing petal and stamen development. pi-5 mutant flowers exhibit strong phenotypic defects in the second whorl where sepals develop in place of petals, but the third whorl stamens develop almost normally. Not only is PIE125K unable to direct petal development in the second-whorl, but also in the third-whorl, as evidenced by the third-whorl phenotype of pi-5 ag-3 double mutants. The phenotype of both pi-5 ag-3 and pi-5 suggests that PIE125K is compromised in its ability to direct petal development regardless of where the petals form in the flower.
pi-5 encodes a PI protein with a missense mutation in the K domain. Protein–protein interaction assays demonstrate that PIE125K is defective in dimerization with its B class partner protein AP3. The fact that the pi-5 phenotype can be partially suppressed by both low growth temperature and overexpression of either AP3 or PIE125K is consistent with the hypothesis that PIE125K is defective in heterodimerization with AP3.
The formation of AP3/PI heterodimers follows the principle of equilibrium driven by the concentration of individual monomers. In wild-type plants, the second and third whorls are not especially sensitive to the levels of AP3 and PI proteins since whorls 2 and 3 develop normally in both 35S::PI and pi-1/+ heterozygotes. By contrast, pi-5 is sensitive to both gene dosage and temperature. Increasing the concentration of either AP3 or PIE125K shifts the AP3/PIE125K equilibrium resulting in the formation of a larger number of AP3/PIE125K heterodimers and thus increased B class activity. Likewise, lowering the temperature stabilizes the AP3/PIE125K heterodimer resulting in increased B class activity.
The phenotype of pi-5 suggests that PIE125K is defective in directing petal development, but not stamen development. One explanation for the pi-5 phenotype is that the second whorl is more sensitive to levels of B class activity than the third whorl. This model postulates that second whorl petal development requires a higher level of B class activity than third whorl stamen development. For a number of reasons, we do not think there is a differential requirement for B class activity in whorls 2 and 3. First, if this were the case, we would expect many intermediate and weak B class mutant alleles to exhibit a more severe phenotype in whorl 2 compared to whorl 3; of the 16 ap3 and pi mutant alleles characterized, pi-5 is the only allele that exhibits a dramatically stronger phenotypic defects in the second whorl. Second, we might expect to observe different levels of B class gene expression in whorls 2 and 3, for example higher levels of RNA and/or protein expression in whorl 2. Based on comparison of in situ signals and signals from immunohistochemical staining of sections using AP3- and PI-specific antisera, AP3 and PI RNA and protein appear to be expressed equivalently in whorls 2 and 3 in both wild type and pi-5 at early floral stages (YZY and TJ, unpublished data).
The idea that whorls 2 and 3 are sensitive to different levels of B class activity has been raised previously to attempt to explain the failure of 35S::AP3 and 35S::PI to fully rescue ap3 and pi mutants. Both 35S::PI and 35S::PI 35S::AP3 fail to completely rescue the third whorl defects of pi-1 mutants (Krizek and Meyerowitz, 1996). One explanation offered to explain the failure of rescue is that the third whorl is more sensitive to overall levels of PI than the second whorl – exactly opposite from the sensitivity that would be required to explain the pi-5 phenotype. A similar argument has been made to explain the failure of 35S::AP3 to rescue the second whorl of ap3-3 mutants (Samach et al., 1997) but in this case the postulated sensitivity is opposite to that proposed for 35S::PI pi-1, i.e. a higher level required in whorl 2 than whorl 3. The third whorl-specific phenotypic defects of the def-101/def-gli mutant from Antirrhinum also does not support a model where the second whorl is inherently more sensitive to overall levels of B class activity (Zachgo et al., 1995). The fact that there is little consistency for the apparent differential requirement for B class activity makes such a model unlikely.
Although the dimerization defects offer an explanation as to why PIE125K fails to function to direct petal development in whorl 2, it does not explain why PIE125K functions almost normally in whorl 3 to direct stamen development. A comparison of PIE125K with other dimerization-defective PI mutants (Yang et al., 2003) raises the possibility that PIE125K may have a second defect. We have characterized a large number of dimerization-defective PI mutants, which when compared to PIE125K exhibit weaker interactions with AP3 in yeast two-hybrid assays. However, these dimerization-defective PI proteins are partially functional in directing petal development in planta, while PIE125K exhibits very little function in directing petal development in planta. For example, in yeast two-hybrid assays, the dimerization-defective PIL121A,V124A mutant interacts with AP3 at 10–20% of wild-type levels. When PIL121A,V124A is ectopically expressed in Arabidopsis, whorl 1 floral organs are petaloid suggesting partial in planta petal function. Although PIE125K interacts with AP3 in yeast more strongly (25–40% of wild-type levels), 35S:: PIE125K whorl 1 floral organs are sepals suggesting that PIE125K does not possess the ability to direct petal development in planta. The limited in planta function of PIE125Ksuggests that PIE125K might have a second defect in addition to the defect in dimerization with AP3.
One possibility is that PIE125K is defective in interacting with ternary/accessory factors. The AP3/PI heterodimer is an essential component of the active B function complex. We think it is likely that the AP3/PI heterodimer associates with accessory factors to form a complex that leads to the activation of the petal and stamen development pathways. PIE125K affects the formation of the active B function complexes at two levels. First, PIE125K leads to a reduction in the concentration of the AP3/PI heterodimer due to its defects in dimerization with AP3. Second, we postulate that PIE125K leads to defects in interaction with accessory factors such as SEP1, SEP2, and SEP3. The three SEP genes function redundantly to specify the identity of petals, stamens, and carpels in the flower. Evidence suggests that the SEP proteins interact directly with the ABC MADS proteins such as AP3 and PI. One model, the ‘quartet’ model (Theißen 2001) proposes that tetramers of MADS proteins direct floral organ identity; specifically, petals are specified by a combination of AP3 + PI + SEP + AP1 while stamens are specified by a combination of AP3 + PI + SEP + AG. Our two- and three-hybrid results suggest that PIE125K is defective in forming complexes with AP3 and SEP3 (or SEP1).
As described by Pelaz et al. (2001b), 35S::SEP3 antisense plants exhibit strong defects in petal development, but not stamen development. The petal defects in 35S::SEP3 antisense plants are more severe than those observed in sep3 loss-of-function mutants. Pelaz et al. (2001b) postulate that expression of the other SEP genes (i.e. SEP1 and SEP2) might be downregulated in 35S::SEP3 antisense plants. If so, presumably there would be a lower concentration of SEP1, SEP2, and SEP3 proteins available in 35S::SEP3 antisense plants to form higher order MADS complexes.
What is less clear is why strong phenotypic defects in pi-5 and 35S::SEP3 antisense plants are observed in second whorl petals but not third whorl stamens. Since the SEP proteins are expressed in both whorls 2 and 3 and are required for both petal and stamen development, the SEP genes alone do not explain why PIE125K functions to direct stamen development but not to direct petal development. A molecular explanation of the pi-5 phenotype requires the existence of a factor that is differentially expressed or differentially active that allows PIE125K to direct the stamen developmental pathway or prevents PIE125K from specifying the petal developmental pathway. One possibility that explains both the pi-5 and 35S::SEP3 antisense phenotypes is that higher levels of SEP activity are required for development of petals than stamens. Thus, the defects of PIE125K in interacting with the SEP proteins might result in a more drastic effect on the development of petals than stamens. The third whorl phenotype of pi-5 ag-3 flowers demonstrates that there is no third whorl-specific factor that is sufficient to allow PIE125K to function. However, there is evidence that PIE125K activity is optimal in the third whorl. In 35S::AP3 pi-5, stamens develop in whorl 3 and carpels in whorl 4 suggesting that the combination of PIE125K and AP3wt is unable to direct stamen development in whorl 4. Similarly, the third whorl of 35S::PIE125K 35S::AP3 pi-4 develops as stamens while fourth whorl positions develop as multichambered carpels, again demonstrating that the combination of PIE125K and AP3wt in whorl 3 functions to direct stamen development in whorl 3, but not whorl 4. These results suggest that there may be a whorl-specific factor that enhances PIE125K function in whorl 3.
Mutant isolation and double mutant construction
In a screen for enhancers of the petal and stamen defects in the very weak ap3 allele, ap3-11 (Yi and Jack, 1998), we isolated an allele of PI called pi-5. In an ap3-11 background, pi-5 exhibits a phenotype similar to intermediate B class mutants; pi-5 ap3-11 flowers exhibit a conversion of petals to sepals or sepal-like organs and stamens to carpels or staminoid carpels (Table S1). Backcrossing to wild type revealed that pi-5, in an AP3+ background, exhibited a petal to sepal phenotype in whorl 2 while stamens developed in whorl 3.
Table S1. Phenotype of flowers.
To construct the ag-3 pi-5 double mutant, pollen from AP3::AG ag-3 (Jack et al., 1997) was crossed to homozygous pi-5. To isolate ag-3 pi-5, seeds were collected from wild-type F1 plants, i.e. plants that did not exhibit the petal to stamen conversion characteristic of AP3::AG. In the F2, ag-3 pi-5 double mutants segregated as plants with flowers composed of all sepals.
To construct 35S::AP3 pi-5, pollen from 35S::AP3 was crossed to pi-5 homozygotes. Seeds were collected from kanamycin-resistant 35S::AP3 F1 plants. 35S::AP3 pi-5 plants were identified among the F2 kanamycin-resistant plants as those that lacked second whorl petals.
For pi mutant rescue, we used the pi-4 allele in addition to the previously characterized pi-1 and pi-3 alleles (Bowman et al., 1989; Bowman et al., 1991; Goto and Meyerowitz, 1994). pi-4 is a strong pi allele generated in the Ler background. pi-4 contains a mutation in the 5′ splice site for intron 1. pi-4, unlike pi-1 and pi-3, is fully rescued by 35S::PI 35S::AP3.
To construct 35S::PIE125Kpi-1, 35S::PIE125Kpi-3, and 35S::PIE125Kpi-4, pollen from kanamycin-resistant 35S::PIE125Kplants was crossed to pi homozygotes. Seeds from F1 kanamycin-resistant plants were collected. F2 seeds were selected on kanamycin plates and 35S::PIE125Kpi/pi were identified as F2 kanamycin-resistant plants that lacked second whorl petals. A similar strategy was used to construct 35S::AP3K139Eap3-3.
To construct 35S::PIE125K35S::AP3 pi-1, 35S::PIE125K35S::AP3 pi-3, and 35S::PIE125K35S::AP3 pi-4, pollen from 35S::PIE125Kpi/pi was crossed to 35S::AP3 pi/pi. In the F1, 35S::PIE125K 35S::AP3 pi plants were identified because they exhibited a vegetative phenotype (slightly curled leaves) and lacked wild-type petals.
Site-directed mutagenesis (Yukenberg et al., 1991) was utilized to construct the pi-5 (PIE125K) mutant in the context of the PI cDNA. The PI cDNA containing the pi-5 mutation (PIE125K) was cloned into plant transformation vector pCGN18 (Jack et al., 1994), which contains the 35S promoter and 3′ NOS. A cDNA containing PIE125K was cloned into pSPUTK and used to produce PIE125K protein for EMSA. AP3K139A,K141A and SEP1K143E were generated by overlap extension PCR mutagenesis (Horton and Pease, 1991).
AP3 and PI constructs and the yeast reporter strains (pJ69-4A) are identical to those used in Yang et al. (2003).
pACT2-SEP1(MIKC), pACT2-SEP2(MIKC) and pACT2-SEP3(MIKC) (Honma and Goto 2001) utilize leu2 as the nutrition marker. AP3(IKC) was cloned into pGBD, a GAL4 DNA-binding domain vector that contains trp1 as an auxotrophic selectable marker (James et al., 1996). PI(IKC) was cloned into pGBDU, a GAL4 DNA-binding domain vector with ura3 as an auxotrophic selectable marker. pACT2-SEP1(MIKC), pGBD-AP3(IKC), pGBDU-PI(IKC) were transformed into pJ69-4A MATα cells and yeast cells were plated on Sc-Leu-Ura-Trp triple drop-out plates to select for the triple transformants. Four independent colonies from each combination were subjected to β-galactosidase liquid assays (Reynolds et al., 1998).
To obtain PI(MIKC) expressed under the control of the truncated ADH promoter, but fused to neither GAL4-AD nor GAL4-DB sequences, we engineered the pGBDU-PI(MIKC) plasmid to eliminate the GAL4 DNA-binding domain via an approach that utilized overlap extension PCR and gap repair creating the plasmid pU-PI(MIKC). pACT2-SEP1(MIKC) carries a leu2 marker, pGBD-AP3(MIKC) carries a trp1 marker, and pU-PI(MIKC) carries a ura3 marker. The reporter yeast strain (pJ69-4A) utilizes both HIS3 and ADE2 as reporter genes. Yeast cells were selected on Sc-Leu-Trp-Ura triple drop-out plates (for the pACT2-SEP1/pGBD-AP3/pU-PI combination) or Sc-Leu-Trp double drop-out plates (for pACT2-SEP1/pGBD-AP3 combination). Four colonies from each plate were inoculated in liquid drop-out media and shaken overnight at 30°C. Each test plate (Sc-Leu-Trp-His, Sc-Leu-Trp-His + 5 mm 3AT, and Sc-Leu-Trp-His + 10 mm 3AT) was divided into three sections, and each section contained streaks of 2 μl of each of the four liquid cultures. The test plates were incubated at room temperature for 1 week. While PIwt can support the yeast growth on Sc-Leu-Trp-His plus 3AT plates, PIE125K can not. Similar results were obtained for the pACT2-SEP3/pGBD-AP3/pU-PI combination.
Reverse yeast two-hybrid screen and isolation of AP3K139E
To screen for compensatory mutants of PIE125K, we screened a randomly mutagenized AP3(IKC) library (Yang et al., 2003). On X-gal plates, compensatory mutants were selected as dark blue colonies on a background of light blue colonies (PIE125K/AP3wt). PIE125K consistently exhibited stronger interactions with AP3K139E than with AP3wt.
For EMSA, a 49-bp double-stranded probe containing the CArG3 from AP3 promoter was utilized (Tilly et al., 1998). pSPUTK plasmids containing wild-type AP3, wild-type PI, and PIE125Kwere used in vitro transcription/translation using TNT® coupled reticulocyte lysate (Promega, Madison, WI) to produce protein for EMSA assay (Riechmann et al., 1996a,b). Multiple independent EMSA assays were performed testing the binding of PIwt/AP3wt and PIE125K/AP3 and in all cases PIE125K/AP3 bound at 20–50% of wild-type levels, as estimated by comparing the intensities of shifted bands in multiple experiments.
Plants were grown in a 3 : 1 : 1 mixture of promix:perlite:vermiculite under constant light at 16 or 23°C.
A MADS-deleted PI antisense 700-base DIG-labeled RNA probe was produced by transcription using SP6 RNA polymerase and StyI linearized plasmid, pD1007, using a kit from Roche-BM following the manufacturers instructions. In situ hybridization was carried out according to standard protocols (Long and Barton, 1998).
We thank Koji Goto for the gift of the SEP1, SEP2, and SEP3 yeast two-hybrid plasmids. We also thank Chuck Daghlian for help with the SEM and Kankshita Swaminathan for help with in situ procedures and comments on the manuscript. This work is supported by grants from the NSF (IBN-9405884 and MCB-0090742).