We report the discovery of an Antirrhinum MADS-box gene, FARINELLI (FAR), and the isolation of far mutants by a reverse genetic screen. Despite striking similarities between FAR and the class C MADS-box gene PLENA (PLE), the phenotypes of their respective mutants are dramatically different. Unlike ple mutants, which show homeotic conversion of reproductive organs to perianth organs and a loss of floral determinacy, far mutants have normal flowers which are partially male-sterile. Expression studies of PLE and FAR, in wild-type and mutant backgrounds, show complex interactions between the two genes. Double mutant analysis reveals an unexpected, redundant negative control over the B-function MADS-box genes. This feature of the two Antirrhinum C-function-like genes is markedly different from the control of the inner boundary of the B-function expression domain in Arabidopsis, and we propose and discuss a model to account for these differences. The difference in phenotypes of mutants in two highly related genes illustrates the importance of the position within the regulatory network in determining gene function.
The study of Antirrhinum and Arabidopsis floral homeotic mutants produced a simple combinatorial model to explain how the regulated expression of a few genes could determine the organ composition of the flower (Coen and Meyerowitz, 1991). The model proposes that three functions are expressed in adjacent, overlapping whorls such that the A-function is expressed in whorls 1 and 2, the B-function in whorls 2 and 3, and the C-function in whorls 3 and 4. Expression of the A-function alone determines sepal development, co-expression of the A- and B-functions or the B- and C-functions specifies petals or stamens, respectively, and expression of the C-function alone results in carpel development. In B-function mutants, petals and stamens are replaced by sepals and carpels, respectively. In C-function mutants, perianth organs develop in place of reproductive organs. In both Antirrhinum and Arabidopsis, a single C-function and two B-function genes have been identified and shown to encode MADS-box transcription factors (Sommer et al., 1990; Yanofsky et al., 1990; Jack et al., 1992; Tröbner et al., 1992; Bradley et al., 1993; Goto and Meyerowitz, 1994).
The MADS-box family is defined by the presence of a 58 amino acid domain which is conserved between the founding members; MCM1 (yeast), AGAMOUS and DEFICIENS (plants) and SRF (animals) (Schwarz-Sommer et al., 1990). Although MADS-box genes have been identified and characterized in animals and yeasts (Shore and Sharrocks, 1995), they are found most extensively in plants (Theissen et al., 1996, Liljegren et al., 1998). The >200 known plant MADS-factors, from many different species, have been divided into sub-families based mainly on their predicted amino acid sequences (Doyle, 1994; Purugganan et al., 1995; Theissen et al., 1996). Often, genes from different species which fall into one of these sub-families have similar expression patterns and similar roles in development. For example, the B-function genes of Antirrhinum [DEFICIENS (DEF) and GLOBOSA (GLO)] and Arabidopsis [APETALA3 (AP3) and PISTILLATA (PI)] are expressed mainly in the second and third whorls and show similar mutant phenotypes. Similarly, the C-function genes of Antirrhinum [PLENA (PLE)] and Arabidopsis [AGAMOUS (AG)] are expressed in the third and fourth whorls and show similar mutant phenotypes. Currently identified functions of the MADS-box gene family include the control of organ identity (see above), flowering time (Sheldon et al., 1999), meristem identity (Huijser et al., 1992; Mandel et al., 1992; Kempin et al., 1995), regulation of ovule development (Angenent et al., 1995), regulation of fruit development (Gu et al., 1998) and control of root architecture (Zhang and Forde, 1998).
Here we report the isolation of a novel Antirrhinum MADS-box gene, FARINELLI (FAR), and a recessive far mutant. We show that all our current criteria identify FAR as a second C-function gene. However, far mutants display none of the typical features of C-function mutants and instead show reduced male fertility. Furthermore, the ple/far double mutant reveals that PLE and FAR act redundantly to prevent expression of the B-function genes, DEF and GLO, in the fourth whorl. We propose a model for the interactions of PLE and FAR and show that two crucial differences could account for the variation in single and double homeotic mutant phenotypes observed in Antirrhinum and Arabidopsis.
Isolation of a second Antirrhinum C-function gene
To identify further members of the MADS-box gene family, an Antirrhinum floral cDNA library was screened using the Arabidopsis C-function gene AG (Yanofsky et al., 1990) as a probe. Sixteen strongly hybridizing clones which did not correspond to PLE were identified, and DNA sequence analysis showed that they were derived from the MADS-box gene, FAR, which is closely related to, but distinct from, PLE (Bradley et al., 1993) (Figure 1A). A FAR probe was used to screen an Antirrhinum genomic library and the DNA sequence of 7 kb spanning the FAR coding region was determined. The FAR gene, like PLE, comprises nine exons, the first of which is non-coding (Figure 1B). In both cases, the MADS-box is encoded by the second exon and is preceded by 16 or 17 amino acids (Figure 1A). This N-terminal extension is a characteristic of C-function homeotic MADS-box factors, such as PLE and AG, and related proteins such as AGL1 and AGL5 (Ma et al., 1991). Within the MADS-box, each pair of factors is 95% identical and, over the entire protein sequence there is 64–67% identity (Figure 1A). Thus, based on the functions of the most closely related genes, it appeared likely that FAR is a C-function factor involved in determining the identity of the floral reproductive organs. Another feature of the C-function genes, which is also shared by FAR, is the presence of a large intron after the MADS-box-containing second exon (Figure 1B). There is evidence to suggest that this intron contains regulatory elements which contribute to the control of spatial and temporal expression of PLE and AG (Bradley et al., 1993; Sieburth and Meyerowitz, 1997). We note the presence of a 70 bp sequence (3581–3650 in the FAR genomic sequence), including a direct repeat of the sequence CCAATCA, which is 60% identical between sequences in this intron of FAR (AJ239057), PLE (D.Bradley, R.Carpenter and E.Coen, unpublished), AG (ATF13C5) and PTAG1 (AF052570), suggesting a possible role for this sequence in the negative regulation of C-function gene expression (Figure 1B).
Two types of FAR cDNA were identified in the cDNA library at different frequencies. The DNA sequence and predicted amino acid-coding potential of the rarer of the two cDNAs was identical to the more common cDNA except for an insertion of four codons (QQLF inserted between amino acids 103 and 104). Comparison of the cDNA and genomic sequences revealed that the rarer variant is derived from differential splicing which makes use of an alternative splice acceptor site within the third intron. PCR analysis of this region from the 16 independent FAR cDNA clones showed that four utilized the alternative splice site. The variation occurs within a region of the MADS-box protein known as the I-domain (Ma et al., 1991), the length and sequence composition of which have been postulated to affect the heterodimerization specificity amongst MADS-box transcription factors (Huang et al., 1996). It is not known what, if any, function is served by the inclusion of an additional four amino acids in a minor variant of FAR, but it remains possible that the presence of two forms of FAR alters its heterodimerization potential.
Two C-function-like genes have been isolated from several species including Antirrhinum (Bradley et al., 1993; this report), petunia (Angenent et al., 1993; Tsuchimoto et al., 1993), tobacco (Kempin et al., 1993; H.Sommer, unpublished), cucumber (Kater et al., 1998) and maize (Schmidt et al., 1993; Theissen et al., 1995; Mena et al., 1996). In petunia, tobacco and Antirrhinum, the two C-function-like factors form sub-families; FBP6, NTPLE and PLE form one sub-family and NAG, PMADS3 and FAR form the other (Figure 1C). AG is most closely related to the FAR family. One of the two cucumber factors, CUM1, falls into the same sub-family as AG, whereas the other, CUM10, does not fit into either sub-family. It has been suggested that the two maize genes ZAG1 and ZMM2 are members of a split C-function, each performing a separate role in floral organ development (Mena et al., 1996). However, since they are more related to each other than to members of the two dicot sub-families, it is likely that a split C-function evolved independently in maize. Discussion of the apparent implications of these observations for the evolution of class C genes and the splitting and/or switching of their role in the control of different aspects of the C-function is beyond the scope of this report. We shall focus rather on developmental implications, since Antirrhinum is the first species in which mutants for both C-function genes have been identified and where the contribution of each gene to the C-function can be studied independently and in combination.
Comparison of PLE and FAR expression patterns
Initial Northern blot and in situ hybridization experiments suggested that the expression patterns of PLE and FAR were identical, in that both are expressed in the third and fourth whorls (not shown). However, more detailed analysis, using sequential sections reveals slight differences in their expression patterns. At stage 4, when expression of PLE first becomes apparent (Bradley et al., 1993), the expression patterns of PLE and FAR are very similar (Figure 2A and C). Differences become apparent by stage 6 when PLE expression is detected in the developing stamen and the carpel primordia (Figure 2B). FAR expression is also observed in these tissues and at the lower border of PLE expression in a bowl-shaped region linking the stamen primordia (Figure 2D). At late stages of development, PLE and FAR show distinct expression patterns (Figure 2E and F). In the anther, PLE transcripts become localized to the region of the stomium (arrowed in Figure 2E) whereas FAR is expressed predominantly in the connective. In the gynoecium, PLE is expressed mainly in the developing ovules and to a lesser extent in the placenta and carpel wall (Figure 2E). In contrast, FAR expression is weaker in the developing ovules and stronger in the placenta (Figure 2F).
To investigate the possibility that PLE controls the expression of FAR, in situ hybridization experiments were carried out using two PLE mutant alleles; ple-1, a loss-of-function ple mutant, and Ple-888, a gain-of-function ple mutant in which PLE is ectopically expressed. At stage 6, PLE expression, which is confined to whorls 3 and 4 in wild-type flowers (Figure 2B), is detected strongly in all whorls of the Ple-888 mutant (Figure 2G). PLE expression can be seen on the inner surface of the organs of the first whorl (which will become carpeloid) and in the second whorl organs (which will become staminoid). In contrast, at this stage, little FAR transcription can be detected in the outer whorls of Ple-888 mutant flowers (Figure 2H). In mature Ple-888 flower buds, PLE transcripts are still detected in all floral organs (Figure 2I), and ectopic expression of FAR mimics that of PLE (Figure 2J). Thus, ectopic FAR expression is induced by ectopic expression of PLE, but there is a delay in the induction of FAR. This delay could indicate that the induction of FAR by PLE is indirect and relies on a PLE-dependent alteration of organ identity, or that PLE can only induce FAR expression in combination with other, later expressed factors. Alternatively, induction of FAR expression could require higher levels of PLE than are present at early stages of development of the outer organs in Ple-888 mutants.
That PLE expression is not absolutely required for the induction of FAR can be seen by studying the transcription of FAR in ple mutants. ple-1 flowers show all the characteristics of a loss of C-function: reproductive organs are replaced by perianth organs and the flower becomes indeterminate and produces many internal whorls of organs (Figure 3F and G). Although these flowers have no stamens or carpels, reduced FAR expression can be detected in all whorls internal to the second (Figure 2K and L). ple-1 is not a null allele and occasionally produces a small amount of pollen on the third whorl petals. In a ple null mutant background, FAR expression is reduced further, but not abolished (not shown). Thus expression of FAR is not dependent on PLE although, directly or indirectly, PLE is required for full expression of FAR and can induce ectopic FAR expression at later stages.
The similarities between FAR and other C-function genes suggested a role in the determination of reproductive organ identity. However, the fact that FAR expression is observed in ple mutants lacking reproductive organs indicated that FAR cannot substitute for PLE. To discover a molecular basis for the failure of two such similar proteins to provoke the same developmental response, we analysed their protein–protein interactions as detected in a yeast two-hybrid assay. We previously had identified four MADS-box factors (SQUA, DEFH49, DEFH200 and DEFH72) that interact with PLE (Davies et al., 1996b). A similar screen was carried out with full-length FAR as bait. Ten colonies which passed all the controls were identified from 3.5×106 primary transformants (Table I). Eight of the positives corresponded to DEFH200, one to SQUA and one to DEFH49. As was the case with the screens involving PLE, we did not identify any other interacting MADS-box factors, such as those of the B-function, and neither did we identify PLE or FAR as FAR-interacting proteins. Subsequent cross-testing of interactions by introducing pairs of cloned MADS-box genes in the appropriate vectors into yeast also failed to find any evidence for interaction between FAR and FAR, PLE, DEF, GLO or DEFH72, although they confirmed the interactions with DEFH200, SQUA and DEFH49 (Table II).
Table 1. Results of the two-hybrid library screen
His+ colonies picked
Table 2. Testing for interaction between known factors
These experiments suggest that differences in the activity of PLE and FAR are unlikely to result from absolute differences in their potential to interact with other MADS-box factors, unless the apparent inability of FAR to interact with DEFH72 makes a significant difference to its ability to induce developmental changes. Further experiments will be required to determine if this differential interaction is responsible for the difference in activity. Although only a single potential difference in protein–protein interactions was detected, subtle alterations in the stability of the heterodimers formed between the common partners and FAR or PLE and/or in their DNA-binding affinity or specificity cannot be excluded.
Ectopic expression analysis in transgenic plants
As a first step to identify a function for the FAR gene in wild-type flower development, it was expressed under the control of a constitutive promoter, in tobacco plants. It has been shown previously that 35S::PLE tobacco plants show the typical features of ectopic C-function expression (Davies et al., 1996a). The plants are normal until flowering, when they produce abnormal flowers with carpeloid organs in the first whorl and stamenoid organs in the second whorl, although the homeotic conversion of the second whorl organs is less complete than that of the first whorl organs. Figure 3B shows a typical flower of a 35S::PLE tobacco plant with a dramatic conversion of the first whorl sepals towards carpeloidy (compare with wild-type flower in Figure 3A). In contrast, the first whorl sepals of 35S::FAR flowers are often completely normal, but sometimes grow to be slightly larger than is usual in the wild-type and show slight carpeloidy. The organs of the second whorl show a much more complete transition to stamens than is seen in 35S::PLE flowers (Figure 3C). Pollen collected from second whorl anthers of 35S::FAR plants has been used successfully in crosses. The filaments of the second whorl stamens are often flattened and slightly petaloid. No abnormalities were observed in the third and fourth whorls or in the vegetative tissues. These experiments suggested that the two C-function genes might play distinct and complementary roles in flower development, with PLE mainly involved in female reproductive organ development and FAR mainly involved in the development of male reproductive organs. Representative 35S::PLE plants were crossed with 35S::FAR plants and the progeny analysed. Purely additive effects were observed in the double transgenics; the plants showed the strong first whorl carpeloidy seen in 35S::PLE lines combined with strong second whorl staminoidy typical of 35S::FAR plants (not shown). Whether the phenotypic differences between the 35S::PLE and 35S::FAR transgenic plants result from the absence of their heterodimerization partners or from their inappropriate expression pattern or altered interaction potential in tobacco with the Antirrhinum proteins was not pursued.
Screening for transposon insertions
Sequence analysis and yeast two-hybrid experiments had shown similarity between PLE and FAR, yet in situ and transgenic studies suggested that the two genes may play different roles in flower development. Furthermore, large-scale Antirrhinum mutagenesis programmes consistently had failed to produce a C-function-like mutant which was not allelic with ple, providing further evidence that the developmental role of FAR was not analogous to that of PLE. In order to resolve these questions, a reverse genetic PCR-based approach was taken to identify transposon insertions in the FAR gene. Pools of DNA (E.Keck, R.Carpenter and E.Coen, unpublished results) were screened by PCR, leading to the identification of two individual plants, far-1 (Stock No. JIT2) and far-2 (Stock No. JIT3), with insertions of the transposon Tam3 (Figure 4A). The positions and orientation of the Tam3 insertions were mapped by PCR using gene-specific primers and primers specific for the right and left borders. In both cases, the orientation of the Tam3 element was the same, with the left border of the element towards the 5′ end of the FAR gene, an orientation which previously has resulted in loss of gene function, even when insertion has occurred within introns (Bradley et al., 1993). Both insertions were within the large second intron, with the insertion in far-1 being further towards the 5′ end than that in far-2. The insertion in far-2 is associated with a deletion which has removed part of the coding region (Figure 4A). Southern blot analysis of the two plants confirmed the insertion of foreign DNA into the FAR gene (not shown). far-2 homozygous plants were crossed with wild-type, the resulting F1 plants were selfed and co-segregation of the far mutant phenotype with the far-2 mutant allele was observed in the segregating F2 population grown at low temperature. Loss of expression of the FAR gene was shown for the homozygous far-2 plants by Northern blot analysis (Figure 4B). In far-2 mutant plants, the FAR transcript is undetectable, although the actin control demonstrates the integrity of the RNA.
Phenotype of far mutants
The two far mutant plants showed identical developmental abnormalities and, therefore, only far-2 was used for further morphological and genetic studies. In contrast to our expectations, homeotic changes were not observed and the plants developed normally. Close inspection of the mutant flowers revealed a variable degree of male sterility. The anthers of homozygous far mutant plants, particularly when grown under warmer conditions, appeared white (at early stages) or brown (at later stages) and were slightly shrunken (Figure 3E). At lower temperatures, some pollen was formed in the mutant anthers but always less than was seen in wild-type control plants. Pollen development appears normal up to the early stages of microspore development when callose begins to surround the microspores. After this, a variable range of defects appear, ranging from apparently normal development to complete degradation of microspores and tapetal cells (not shown). In extreme cases, the locules are either completely empty or contain some empty, microspore-like structures surrounded by a filamentous material, which might be the remains of degenerated tapetal cells. Occasionally, normal, fertile anthers were formed on far mutant plants. When this happened, it was usually the upper pair which were normal, whilst the lower pair often remained empty, suggesting that the developmental process controlled by FAR is also influenced by the floral symmetry genes (Luo et al., 1996).
The male-sterile phenotype of far plants is similar to a pre-existing Antirrhinum mutant, albidiflora (albi). In order to test whether albi is also caused by a defect in FAR, the FAR gene was PCR amplified from homozygous and heterozygous albi plants and wild-type sisters, all of which were derived from selfing the progeny of a cross between albi (in a Si50 background) and the wild-type 165E. A FAR polymorphism was detected by digesting the resulting PCR fragment with the enzyme AluI. The polymorphism did not co-segregate with the albi mutation, showing that, although similar in phenotype, albi and far are unlinked (data not shown). Thus, far represents a novel male-sterile mutant.
Expression of PLE in far mutant plants
Since far mutants appear morphologically normal, with the exception of empty anthers, it was likely that PLE expression would also be normal in far mutants. However, in situ hybridization experiments revealed that in the absence of FAR, the expression pattern of PLE expands. PLE expression, which is confined to the region of the stomium in the mature wild-type anthers (Figures 5A and 2E), can now be detected throughout the anther, in a pattern which looks like the superimposition of the expression patterns of FAR and PLE (compare Figures 2E and F, and 5B). Similarly, in the carpel, expression of PLE, which had been detectable mainly around the developing ovules in the wild-type (Figures 5A and 2E), is now much stronger and spread throughout the carpel wall and placenta (Figure 5B). This increase and expansion of the PLE expression pattern in far mutants suggests that FAR negatively regulates PLE. It is possible that in far mutants the expansion of the PLE expression domain, into tissues which would normally express FAR in wild-type flowers, compensates for the loss of FAR activity.
Genetic analysis of far mutants
ple-1/far double mutants were constructed and analysed. Superficial examination of the double mutants suggested that ple-1 is epistatic to far since, like ple (Bradley et al., 1993), ple-1/far flowers are indeterminate and show homeotic conversion of reproductive organs to perianth organs. However, more detailed investigation of the organs produced in the third and fourth whorls of ple-1 and ple-1/far flowers revealed a surprising interaction between the two genes. The third whorl of ple-1 mutant flowers consists of laterally non-fused petaloid organs which resemble stamens in that their lower part is filament-like and narrow and the upper part is slightly broadened and flag-like (Figure 3F). In contrast, the upper part of the third whorl organs of the ple-1/far double mutant are broader and the organs fully or partially fuse to the second whorl at their lower part, similarly to a ple null mutant (Figure 3J). This more pronounced petaloid morphology of ple-1/far double mutants compared with ple-1 mutants suggests that although FAR is unable to compensate for the loss of PLE, it still plays some positive role in stamen organ identity.
A more dramatic effect of FAR on organ identity is observed in the fourth whorl. In ple-1 mutants, the degree and kind of homeotic fourth whorl transformation is variable (Figure 3F and G), but usually two to three sepaloid/petaloid/carpeloid organs are formed. Within this whorl, another mutant flower develops to a variable extent (Figure 3G), initiated by slightly sepaloid petals composed of a mixture of petal and sepal tissue. This sepaloid feature is observed more easily in the double mutant between ple-1 and a weak allele of deficiens, def-nic (Schwarz-Sommer et al., 1992). In the ple-1/def-nic double mutant, the fourth whorl consists of two sepaloid carpels and, internal to them, borne on a pedicel, a new flower is initiated by a whorl or spiral of five or more sepals (Figure 3h and I), followed by petals with morphological features characteristic of def-nic second whorl organs (Figure 3I). In contrast, the fourth whorl of ple-1/far mutant flowers consists of four to five petals forming a tube and corolla similar to the second whorl petals, followed by further spirally arranged petals (Figure 3K). Thus, in the double mutant, both the identity and the number of organs produced in the fourth whorl are altered.
Since petal organ identity in Antirrhinum is determined by the expression of the B-function genes DEF and GLO, it seemed likely that the conversion of the fourth whorl to petals in ple/far double mutants resulted from ectopic expression of DEF and GLO. To test this, in situ experiments were carried out on ple and ple/far double mutant flowers using DEF and GLO as probes. At stage 4, expression of both DEF and GLO in ple-1 buds is similar to that normally observed in wild-type buds (Figure 5C and E). At the same stage in the ple-1/far double mutant, expression of both genes is markedly stronger, although the tissue specificity is maintained; GLO is confined mainly to the developing second and third whorls (Figure 5F) and DEF is clearly detectable in the inner three whorls (Figure 5D). In mature ple-1 flower buds, the expression pattern is maintained, DEF being expressed mainly in the second and third whorls, reduced in the fourth whorl (arrowed in Figure 5G) and increased again in internal whorls. Expression of GLO in mature ple buds is similar to that of DEF, but as in younger stages, GLO transcripts are less detectable in the fourth whorl than are those of DEF (Figure 5H). In the ple-1/far double mutant, the reduced expression of DEF and GLO in the fourth whorl is no longer apparent and both transcripts are strongly expressed in the organs internal to the first whorl (Figure 5I and J, fourth whorl arrowed in I). Thus, PLE and FAR act independently to prevent the expression of the B-function in the fourth whorl.
The developmental role of FAR
In Antirrhinum, two MADS-box genes, PLE and FAR, share typical features of C-function genes, including gene organization, sequence similarity and protein–protein interaction potential. In ple mutant flowers (Bradley et al., 1993), the reproductive organs are converted homeotically to perianth organs, and further perianth organs develop inside the fourth whorl (Figure 3F and G). PLE is therefore required to specify the identity of the stamens and carpels and to confer floral determinacy. In the newly isolated far mutant, the only apparent phenotype is a failure to complete normal pollen development. Analysis of ple/far double mutants reveals an unexpected interaction between the two genes and demonstrates that they are partially redundant. In ple single mutants, the fourth whorl develops as two sepaloid/carpeloid/petaloid organs (Figure 3G). In contrast, the fourth whorl organs of ple/far double mutants develop as four or five well-formed petals (Figure 3J and K). Furthermore, the B-function genes, DEF and GLO, which normally are mostly excluded from the fourth whorl, become ectopically expressed there in the ple/far double mutant (Figure 5I and J). Since the B-function genes are largely confined to the second and third whorls in both ple and far single mutants, it seems that PLE and FAR are redundant with respect to their ability to exclude the B-function genes from the fourth whorl. In addition, combination of a weak ple allele with far enhanced third whorl petaloidy, suggesting a role for FAR in the control of wild-type stamen identity.
The relative contribution of each gene to the development of the reproductive organs is more difficult to assess. FAR is incapable of substituting for the organ identity function of PLE, since no reproductive organs are formed in ple mutants even though FAR is expressed. FAR is involved in pollen development, but the small amount of pollen which occasionally is produced in far mutant anthers shows that it is not absolutely required. In contrast, PLE can direct the formation of male and female reproductive organs in far mutants, although the anthers are frequently empty. It could be that PLE, perhaps indirectly, leads to the production of the limited amount of pollen found in far mutants. Conversely, it could be the expansion of PLE expression within the developing anthers of far mutants which is responsible for the failure to undergo normal pollen development. The phenotypic effect of the loss of FAR function could be limited because PLE expression expands in far mutants. In contrast, in ple mutants, FAR expression decreases and cannot compensate for the loss of PLE. This dependence of FAR on the expression of PLE is circumvented in transgenic plants where FAR is controlled by the 35S promoter. Indeed, ectopic expression of Antirrhinum FAR in transgenic tobacco plants is sufficient to cause the development of stamens in the second whorl of tobacco flowers. Preliminary results with transgenic Antirrhinum plants confirm this, although the extent of homeotic transformation of the first and second whorls is less pronounced (I.Heidmann, Z.Schwarz-Sommer and B.Davies, unpublished). In summary, FAR appears capable of functions similar to those of PLE, but this is not apparent in the single mutants because of the regulatory interactions between the two genes.
From sequence to function
Various criteria have been used to assign MADS-box genes to sub-groupings within the family. These include overall sequence similarity, the presence of specific amino acid residues at characteristic positions within the MADS-domain, the length of the I-region, the presence or absence of an N-terminal extension and a high degree of amino acid sequence identity amongst the five or six C-terminal amino acids. It has been supposed that some of these indicators could be used to suggest functions for newly discovered MADS-box genes (Theissen et al., 1996). In the case of FAR, these indicators, together with the expression pattern, suggested a typical C-function gene. Ectopic expression experiments supported this, since homeotic conversions of petals to stamens were observed, although the more typical sepal to carpel conversions were not apparent. Yeast two-hybrid screens for proteins capable of interacting with FAR resulted in the isolation of almost exactly the same subset of cDNAs as was identified when the known C-function factor PLE was used in similar experiments. From our understanding of the C-function, it would therefore be expected that a loss-of-function far mutant should result in homeotic conversion of the perianth to reproductive organs and a loss of floral determinacy. In fact, the far single mutants showed no homeotic changes and no effects on determinacy. The defect in male fertility, which is the only change apparent in far mutants, could not have been predicted from either the sequence classification of FAR or from any of the subsequent biochemical experiments. It also could not have been predicted that mutation of the FAR gene would lead, in a ple mutant background, to alterations in organ identity and number in whorls which are unaffected in far single mutants. The complications which obscured the developmental role of FAR in Antirrhinum, redundancy and regulatory interaction, have been observed for other MADS-box genes (Bowman et al., 1993; Liljegren et al., 1998) and demonstrate that the developmental role of a gene is dependent on both its primary sequence and its position within the regulatory hierarchy. The implication of this result is that it will be impossible to determine the developmental role of other members of the MADS-box gene family unless the appropriate mutants and mutant combinations are first isolated. Caution will be needed when interpreting the effects of ectopic expression of MADS-box genes, especially when inferring their functions from heterologous systems. As a consequence, reverse genetic strategies such as the one used in this report will be required to understand the contribution of these genes to plant development.
Control of floral determinacy in different species: beyond the ABC model?
Although the Antirrhinum and Arabidopsis C-function mutants ple and ag both consist of indeterminate whorls of perianth organs, they differ with respect to the identity and organization of certain whorls. In ple-1 flowers, the fourth whorl consists of two to three organs which surround a new flower, borne on a pedicel and initiated by sepaloid organs (Figure 6A). A new flower also develops in the Arabidopsis ag mutant, but in this case the four sepaloid organs of the first whorl of the internal flower form inside the third whorl (Figure 6A). Thus, in Arabidopsis, the new internal flower of the C-function mutant is initiated after three whorls, not after four whorls as in Antirrhinum. In this interpretation of C mutant phenotypes in either species, there is no homeotic transformation of carpels to sepals, as proposed in the ABC model (see Introduction), but rather initiation of a new flower, albeit at a different site, as shown above and discussed below. The different phenotypes of the two C-function mutants become identical when they are combined with other mutants. In ple/far double mutants, there is no initiation of a new internal flower (Figure 6A); all internal whorls consist of petals. This is similar to the phenotype of the double mutant between ag and superman/flo10, which also fails to initiate a new internal flower and consists of sepals surrounding multiple whorls of petals (Schultz et al., 1991; Bowman et al., 1992). SUPERMAN/FLO10 (SUP) acts at the boundary of whorls three and four to prevent expression of the B-function in the fourth whorl, probably by controlling the proliferation of B-containing cells (Sakai et al., 1995). We show here that in Antirrhinum the B-function is negatively regulated by PLE and FAR.
An analogous phenotypic difference is apparent when the B-function mutants of Antirrhinum (def and glo) and Arabidopsis (ap3 and pi) are compared. In the Antirrhinum B-function mutants, no fourth whorl organs develop due to the antagonistic effect on determinacy between the B- and C-functions (Tröbner et al., 1992); the B-function prevents C-function-dependent termination of development (Figure 6A). In the Arabidopsis B-function mutants, a fourth whorl develops, indicating that, unlike in Antirrhinum, the mechanism to prevent premature termination by the C-function must be at least partially independent of the B-function (Figure 6A). It has been reported that double mutants between sup/flo10 and pi or ap3 produce only three whorls (Schultz et al., 1991), making them more similar to the Antirrhinum single B-function mutants def and glo (Figure 6A). The fact that supflo10 single mutants produce a proliferation of stamens within the third whorl shows that SUP and the B-function act together to prevent C-function-dependent termination. Taken together, these observations suggest that initiation of fourth whorl primordia and subsequent termination of flower development in the centre of the flower are controlled differently in the two species and indicate that the difference could centre on the role of SUP.
A possible explanation for the observed differences in mutant phenotypes described above could be that Antirrhinum lacks a SUP function and instead uses PLE and FAR to regulate the B-function negatively. However, the role of SUP in Arabidopsis cannot be taken wholly by PLE and FAR in Antirrhinum because the B-function is expressed in the third whorl, although both PLE and FAR are active there, suggesting that PLE and FAR co-operate with a spatially restricted function at the boundary of whorls three and four and do not act independently to prevent B-function expression there. A mutant with a similar phenotype to sup/flo10 has been found recently in Antirrhinum and named octandra (oct) (R.Carpenter, personal communication), suggesting that OCT performs the function of a SUP-like gene in Antirrhinum. In order to explain the observed differences between Antirrhinum and Arabidopsis mutant phenotypes, we need to postulate that OCT differs from SUP in two key ways. First, OCT requires PLE or FAR in order to exclude the B-function from the fourth whorl, whereas SUP can act independently of AG. This is supported by the fact that the B-function does not expand into the fourth whorl of ag mutants (Jack et al., 1992) but does in ple/far double mutants (this report). Secondly, OCT expression or function is more dependent on the B-function in Antirrhinum than is SUP in Arabidopsis, although other factors must also be involved to control the expression of OCT at the boundary between the third and the fourth whorl. Figure 6B illustrates how these differences in the regulation of and by OCT/SUP could account for all the observed wild-type and mutant phenotypes of Antirrhinum and Arabidopsis. For simplicity and clarity, Figure 6B and the following paragraphs do not consider that factors other than SUP/OCT and the organ identity controlling functions are involved in the control of organ initiation and determinacy and that the regulatory relationships are likely to be indirect and more complex.
In the fourth whorl of Antirrhinum and Arabidopsis, only the C-function, controlled by PLE/FAR or AG, is expressed (Figure 6B). This leads to the formation of carpels and termination of floral development. In the third whorl of Arabidopsis, the co-expression of the B-function genes, AP3 and PI, with AG leads to the production of stamens, whilst AP3/PI and SUP prevent AG from terminating flower development in this whorl. Similarly, in Antirrhinum, DEF/GLO and OCT prevent PLE or PLE/FAR from terminating floral development in the third whorl. If OCT expression in Antirrhinum is dependent on DEF/GLO, OCT will not be expressed in def and glo mutants and development will terminate in the third whorl because there is no DEF/GLO or OCT activity to counteract the terminatory influence of PLE or PLE/FAR. In contrast, Arabidopsis ap3 and pi mutants produce four whorls because SUP, being partly independent of AP3/PI, will still act to prevent AG-dependent termination in the third whorl. In ap3/sup and pi/sup double mutants, termination occurs with the production of third whorl carpels because, as in Antirrhinum def and glo mutants, both the B-function and SUP/OCT are absent.
In Arabidopsis ag mutants, AP3 and PI remain excluded from the fourth whorl due to the activity of SUP. Thus, in the fourth whorl, there is no C- or B-function activity and a new flower is produced. In Antirrhinum ple mutants, DEF and GLO, which we suggest to be under the control of PLE/OCT and FAR/OCT, expand slightly into the fourth whorl as a result of a loss of PLE and a concomitant reduction in FAR. Some residual control over DEF and GLO is maintained in ple mutants due to the reduced activity of FAR/OCT. The slight expansion of DEF and GLO into the fourth whorl is sufficient to prevent the initiation of a new mutant flower until after initiation of the fourth whorl. Loss of FAR does not have similar consequences because in the absence of FAR, PLE expression increases to compensate. In contrast, the double mutants ag/sup and ple/far show the same phenotype because in neither case is the B-function prevented from being expressed in the fourth whorl. The model thus predicts that OCT expression or function is absent in the def and glo mutants and that these Antirrhinum B-function mutants would produce four whorls if an OCT transgene were to be expressed in the third whorl. This hypothesis can be tested when OCT is isolated from Antirrhinum. It also predicts that the concommitant loss of ple, far and oct will not confer changes to floral morphology other than those observed in the ple/far double mutant.
Materials and methods
cDNA libraries, in situ hybridization and Northern blotting
In situ and Northern blot expression analysis and cDNA library construction and screening were carried out as previously described (Sommer et al., 1990; Huijser et al., 1992). Photographs were taken in dark field to detect the silver grains and superimposed on calcofluor-stained epifluorescence to visualize the underlying tissue. Sequencing was carried out using the fmol cycle sequencing system (Promega) with sequence analysis by MacVector (Oxford Molecular Group).
The predominant FAR cDNA was amplified using oligonucleotide primers to introduce cloning sites. Amplification was carried out using the oligonucleotide primers 5′-ATAGGAATTCGCGTCTCTAAGCGATC-3′ and 5′-GGAAGGATCCATTTTCTCCAAGCGCC-3′ followed by digestion with EcoRI and BamHI and insertion into pGBT9 to form the bait. The bait included all the FAR coding sequences excluding the initiation codon. Two-hybrid screens were carried out as previously described (Davies et al., 1996b). Interactions were investigated both by screening an Antirrhinum floral cDNA yeast expression library and by directly testing for interaction between the FAR bait and previously isolated MADS-box preys.
PCR pools screening
The reverse genetic screen used here was established at the John Innes Centre, Norwich, UK (E.Keck, R.Carpenter and E.Coen, unpublished). Integrations of the transposable element Tam3 into the FAR gene were detected by hybridization of the FAR probe to blots of PCRs containing a gene-specific primer, a Tam3-specific primer and several pools of genomic DNA from defined sets of plants. Two DNA pools from the John Innes Institute, Norwich, UK, were screened for integrations into individual plants. The two pools were from the years 1992 and 1993 and were derived from 6350 and 6300 plants, respectively. The pools were screened as 31 superpools of 450 plants, each of which was created by combining three subpools of genomic DNA prepared from the leaves of 150 plants. Positives were first confirmed on the subpools of 150 plants and subsequently on 10 new subpools generated from batches of 15 separately collected leaves from those plants which contributed to the positive subpool. Finally, seed collected from the 15 plants in each positive pool was sown and the seedlings were tested to confirm integrations into FAR.
From the genomic sequence of FAR, four oligonucleotide primers were chosen to match approximately the length and base composition of the two primers specific for the left and right borders of the Antirrhinum transposable element Tam3. The primers used were: FAR1F, 5′-CGTTTCTATTTTTGACTCTATGG-3′; FAR4F, 5′-CCATCCTCTTTGACAGTTCC-3′; FAR4R, 5′-GAGGATGGACAAATGAGAAC-3′; FAR7R, 5′-GATTTAGTTTGTGTGCGATTC-3′; Tam3R, 5′-CTCGGCACGTTTCACATCTTTA-3′; and Tam3L, 5′-CACGGCCCAATTCACATCTTTA-3′.
PCR was carried out in 50 μl reactions, and 5–10 μl of each superpool reaction was dot blotted onto a nylon membrane using an eight channel micropipette and the filter was baked and hybridized with a FAR probe. Subsequent subpool screening was by Southern blotting of the PCR products.
Genetic stocks and transgenic plants
Antirrhinum majus plants were grown in a greenhouse at 18–25°C with additional light during winter. A climate chamber at 15°C was used for growth at lower temperatures. The following mutant and wild-type lines were used in this study: ple-1 and def-nic (Stubbe, 1966) from the Gatersleben seed collection, Ple-888 (Macho) (Schwarz-Sommer et al., 1990; Lönnig and Saedler, 1994), albidiflora Gatersleben, Sippe50 (Stubbe, 1966) and Gatersleben, 165E (niv-98::Tam3) Rosemary Carpenter, Norwich. ple/far double mutants were isolated by selecting plants in the F2 of a cross using ple-1/ple-1 flowers, which occasionally produce some pollen, to pollinate far-2/far-2 plants. The double mutants were identified by selecting for a ple phenotype and testing for homozygosity of far by PCR. Ple-1/def-nic double mutants were obtained in the selfed progeny of plants heterozygous for both mutant alleles, produced by crossing a def-nic plant with pollen from ple-1. Mutants and double mutants were identified by their phenotype and segregation ratio (ple-1:def-nic:ple-1/def-nic = 28:22:5). In addition, plants with elongated internodes and pedicels segregated in this population together with the decrease of the apical dominance, most likely due to the presence of a trait carried by the def-nic line. Segregation of this trait conferred considerable variation to the mutant phenotypes, revealed by the presence or absence of an internal pedicel when ple-1 was mutant, and by decreased or enhanced male features of def-nic flowers.
Tobacco transformation was as previously described (Davies et al., 1996a). The FAR cDNA was amplified by PCR, cloned as a transcriptional fusion behind the cauliflower mosaic virus (CaMV) 35S promoter into the binary vector pPCV701, and the insert sequenced. Following agrobacterial-mediated transformation of tobacco, 26 kanamycin-resistant plants were isolated, 18 of which showed clear morphological changes. Selection for 35S::PLE and 35S::FAR transformants and for the double transgenic was on hygromycin, kanamycin or both antibiotics, respectively. All abnormalities segregated as dominant traits.
The authors are indebted to Enrico Coen and Rosemary Carpenter for materials and generous help in screening their Antirrhinum PCR pools. Thanks also to Des Bradley and Sandra Doyle for help during the screening process, Rosemary Carpenter for information on the oct mutant prior to publication, Iris Heidmann and Sabine Clemens for their help in identifying the far mutant by PCR in segregating populations, and Barry Causier, Enrico Coen, Marcos Egea-Cortines, Phil Gilmartin, Peter Huijser and Mark Wilkinson for critical reading of the manuscript. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 243) to Z.S.-S., EU-Biotech (BIO4-CT97-2217) to B.D. and H.S., and The British Council (project No. 900) to B.D. and Z.S.-S.