Consistent with this observation, molecular and genetic evidence has indicated that the underlying molecular mechanisms diverge to some degree between Papilionoideae and other well-studied eudicots. According to studies in Arabidopsis and Antirrhinum, LFY and FLO are floral meristem identity homologs that are expressed throughout the floral primordium and are responsible for specifying floral fate (Coen et al., 1990; Weigel et al., 1992). They play an additional role in the recruitment of floral organs and the activation of downstream organ identity genes. However, loss of function of Uni, a pea LFY/FLO homolog, led to development of flowers lacking petals and stamens and growth of additional mutant flowers in the axils of the sepals (Hofer et al., 1997). In Arabidopsis and Antirrhinum, UFO or FIM acts synergistically with LFY/FLO (Simon et al., 1994; Ingram et al., 1995; Levin and Meyerowitz, 1995). In ufo and fim mutants, homeotic transformations of petals to sepals and stamens to carpels were found, as well as some loss of lateral and apical determinacy during floral organ development (Simon et al., 1994; Ingram et al., 1995). Similar phenotypes were observed in the loss-of-function mutants, stp and pfo, of pea and Lotus japonicus UFO/FIM homolog, respectively (Taylor et al., 2001; Zhang et al., 2003). The pfo phenotype was strikingly similar to those of ufo, fim, and stp plants, although pfo produced many more ectopic flowers. Furthermore, the pea Uni and Stp genes were also shown to function in compound leaf development (Hofer et al., 1997; Taylor et al., 2001). In both uni and stp mutant plants, the complexity of compound leaves decreased, suggesting that these orthologs may be recruited to different regulatory pathways and/or may target different biochemical factors in Papilionoideae than in the other model organisms.

The molecular mechanisms controlling floral organ specification tend to be highly conserved between species (Soltis et al., 2002). Based on genetic studies in Arabidopsis and Antirrhinum, the ABC model (Bowman et al., 1991; Coen and Meyerowitz, 1991) has been proposed, in which the combination of three functions specifies floral organ identity. Later, the function E was found to act along with the B and C functions to specify petal, stamen, and carpel identity (Pelaz et al., 2000). Many organ identity genes, which play a central role in these different functions, have been isolated, and most of them encode MADS-domain proteins (Ng and Yanofsky, 2001; Lohmann and Weigel, 2002). In Arabidopsis, LFY plays a role in activating these organ identity genes directly (Parcy et al., 1998; Busch et al., 1999; Wagner et al., 1999; Lenhard et al., 2001; Lohmann et al., 2001), and UFO acts synergistically with LFY to activate B function genes in whorls 2 and 3 (Simon et al., 1994; Ingram et al., 1995; Levin and Meyerowitz, 1995). Several homeotic mutants have been identified in pea and M. truncatula (Ferrandiz et al., 1999; Penmetsa and Cook, 2000; Taylor et al., 2002), showing that A, B, and C functions are necessary for floral patterning in legumes.

Although the ABC model is expected to be at least partially valid in legumes, characterization of ABC function orthologs is needed to examine how the ABC model can be used to explain the unique floral patterning in Papilionoideae. Functional genomic studies in the model plant, L. japonicus (Handberg and Stougaard, 1992), have taken advantage of its relatively small genome to identify several genes involved in nodulation or floral patterning (for review, see Perry et al., 2003; Somers et al., 2003; VandenBosch and Stacey, 2003). We have used similar approaches to study floral pattern formation in L. japonicus as the basis for comparing the floral regulation networks of Arabidopsis and L. japonicus. We report here the cloning of 13 homologs of LFY/FLO, UFO/FIM, and A, B, C, D, and E function genes in L. japonicus. Phylogenic studies indicated that duplication of B and C function genes frequently occurs in L. japonicus. A mutant, proliferating floral meristem (pfm), was identified from an ethyl methanesulfonate (EMS) mutant collection and associated with a nonsense mutation in LjLFY, and an allele of proliferating floral organ (pfo) was also found. Moreover, RNA in situ hybridization analysis revealed that the expression patterns of these homeotic genes during L. japonicus floral ontogeny differed from those in Arabidopsis and Antirrhinum, indicating molecular explanations for the species-specific differences in floral ontogeny.

Figure 1. Ontogeny of Gifu flower. a, Gifu flower; bar = 1 mm. b, Dissection of Gifu flower; bar = 1 mm. c to k, SEM showing the floral ontogeny of wild-type L. japonicus. Sp, Sepal; Pt, petal; Sti, inner stamen; Sto, outer stamen; Ca, carpel. Bar (more ...)

At Stage 0, while the meristem of the I2 is degenerating, floral primordia are initiated from the peripheral region of I2 and trichomes are formed in the marginal area between the floral primordia (Fig. 1c). At Stage 1, bracts have developed on one side of each floral primordium. This is defined as the abaxial position of a floral primordium, just opposite of the adaxial position where the floral primordia are separated from each other (Fig. 1d). The bract adjacent to a developed floral primordium will degenerate quickly, and only the one adjacent to a degenerated floral primordium will develop fully. At Stage 2, a sepal primordium is formed in the abaxial position of the floral meristem, following the rule of unidirectional initiation order from abaxial to adaxial side (Fig. 1e). At Stage 3, one ellipsoid anlage (or abaxial common primordium, which will give rise to the abaxial petal and stamen primordia at later stages) can be recognized in the axil of the abaxial sepal. At the same time, the other four sepal primordia appear at the lateral and adaxial regions of the floral meristem (Fig. 1f). At Stage 4, the carpel primordium appears at the center of the floral meristem. Two ellipsoid anlagen (or lateral common primordia) are initiated in the axils of the lateral sepals, and two abaxial petal (keel) primordia and one abaxial outer stamen primordium develops from the abaxial anlage. However, the shape of the lateral anlagen is less obvious than that of the abaxial ones (Fig. 1g). At Stage 5, two lateral petal (wing) primordia and two outer stamen primordia develop from the two lateral anlagen (Fig. 1h). At Stage 6, one adaxial petal (standard) primordium and two adaxial outer stamen primordia are initiated asynchronously (Fig. 1i). At Stage 7, the five inner stamen primordia are formed; the adaxial-most one has developed asynchronously and will be the only one not forming the fused stamen tube with the other nine stamens (Fig. 1j). After Stage 7, all organ primordia have developed and started elongating (Fig. 1k).

Hence, L. japonicus shows some features common to the Papilionoideae, including noncentripetal and unidirectional organ initiation order, a heterogeneous whorl with petal and stamen identity, and an inner whorl of stamens. However, unlike pea and M. truncatula, the standard (adaxial petal) and the two adjacent outer stamens were initiated separately in L. japonicus, and thus the common primordium in the adaxial position was not observed.

In wild-type plants, the compound leaf located above the sixth node was a complete leaf consisting of five leaflets, two basal leaflets, two lateral leaflets, and one terminal leaflet. In the corresponding positions of pfm mutants, about 85% of these complete compound leaves lacked 2 basal leaflets, 14% lacked 1 basal leaflet, and only 1% were normal. In pfo-2 mutants, the phenotype was different, with about 70% of the complete compound leaves lacking at least 1 leaflet. However, the missing leaflets in pfo-2 mutants were not contained in the basal position of the compound leaves, and petioles of the compound leaves were frequently missing (data not shown).

During the reproductive phase, instead of producing flowers, the I2 in pfm plants produced sepal-like proliferating structures (Fig. 2a). These mutant flowers lacked petals and stamens and withered as the plant aged. Dissection of the pfm I2 under stereomicroscope revealed a few ball-like structures with pedicels, which are defined here as the primary flowers. The primary flowers each had 5 sepals, which were narrower than those from wild type (Figs. 2b and 1b, respectively). Removal of sepals revealed another round of floral-like structures, which were made up of sepals and a successive ring of floral-like structures (Fig. 2c). The sepals and nested floral-like structures were repeated continuously depending on the growth condition of the plants (Fig. 2d). Basically, the same phenotype was observed in pfo-2 mutant plants (Fig. 2, e–h).

Figure 2. Ontogeny of mutant flowers. a, An inflorescence of pfm; bar = 1 mm. b to d, Dissection of a pfm mutant inflorescence; bar = 1 mm. e, An inflorescence of pfo-2; bar = 1 mm. f to h, Dissection of pfo-2 mutant inflorescence; bar = (more ...)

The phenotypic differences between the wild-type (Gifu) and pfm flowers were further analyzed by scanning electron microscopy. During the reproductive stage in both plants, the SAM became the I1 meristem, which initiated the I2 primordia at its periphery (Fig. 2i), with the floral meristems initiating progressively from the I2 meristems (Fig. 2j). There were no significant differences observed between Gifu floral and pfm primary floral development in the early stages (Figs. 1, d–f, and 2, k–m). When the floral meristems of Gifu and pfm had produced 5 sepal primordia at whorl 1 with a carpel (or a carpel-like) primordium at the center, there were obvious alterations in the mutant plants: trichomes initiated at the boundary of anlagen; the shape of the primary floral meristem was much rounder; the size of two lateral anlagen were obviously larger than the abaxial one; and 3 whorls of organ primordia were observed clearly (Fig. 2n).

Although these alterations indicated the primary floral meristems were developmentally abnormal in the mutants, the arrangement of the primordia in the middle whorl was basically the same as the wild type: three ellipsoid anlagen, one in the abaxial and two at the lateral position, with one or two primordia in the adaxial region (Fig. 2n). These primordia later contracted into four or five ball-like structures (Fig. 2o) and then were transformed into the secondary floral-like meristems (Fig. 2p). From then on, the pattern of floral-like structures was repeated and primordium initiation in the floral-like meristems was not initiated unidirectionally. In the mutants, therefore, numerous iterations of floral-like meristems proliferated in a highly organized manner. The repeated pattern of the typical floral-like meristems and their structures with three whorls (sepals, successive meristems, and carpel; Fig. 2, q and r) normally could be readily recognized for four to five rounds under scanning electron microscopy (SEM). Thus, our analysis of the mutant floral structures could be focused on the primary floral meristems. In a typical floral-like structure, the carpel-like structures produced at the center of the floral-like meristems had the characters of a carpel that cease developing at various stages but always failed to become pistils (data not shown). The SEM analysis confirmed that the phenotypes of pfm and pfo-2 were similar (data not shown).

The phenotype of pfm and pfo-2 was very similar to those of stp and uni in pea. It has been shown that stp/pfo and uni are the legume homologs of UFO and LFY, respectively, in Arabidopsis. Hence, we PCR amplified and sequenced the L. japonicus genomic sequences of UFO and LFY homologs from both the wild type and pfm mutants. Two genes, LjUFO/Pfo and LjLFY, from the wild type were obtained, respectively (see next section of this paper), and were used as the functional homologs, to conduct a comparison with the mutants (see “Materials and Methods”). The pfm and Gifu genomic sequences showed no differences in the LjUFO/Pfo gene, but those for the LjLFY gene contained a nonsense mutation in pfm at position 142 (cga to tga) that was predicted to truncate the LjLFY protein to 47 amino acids (Fig. 3a), indicating that the protein was nonfunctional in the pfm mutant plants, whereas the LjLFY sequence was unimpaired in the normal M2 plants whose progeny did not segregate a mutant phenotype.

Figure 3. Point mutations in the LjLFY and Pfo sequences of the pfm and pfo-2 mutants, respectively. a, Schematic representation of the LjLFY ORF. Nucleotide and predicted amino acid sequences are shown for the pfm mutant and the corresponding region in Gifu. b, (more ...)

In pfo-2 (line F0795), we identified a nonsense mutation at position 241 (cga to tga) causing a truncation of the deduced LjUFO/PFO protein within the putative F-box (Fig. 3b). This point mutation gave rise to a cleaved amplified polymorphic sequence (CAPS) marker that was used to distinguish the Pfo alleles in both the pfo-2 mutant and the wild type (see “Materials and Methods”). The homozygous point mutation cosegregated with the mutant flower phenotype in the M3 descendents of F0795. However, there was no observable difference in the LjLFY genomic sequence between pfo-2 and Gifu. Taken together, these data indicate that we had identified another allele (pfo-2) of the Pfo gene. We refer to the original pfo mutant (Zhang et al., 2003), therefore, as pfo-1 from here on.

LjLFY consists of three exons that are conserved with LFY, FLO, and Uni in terms of exon size and number. Southern blotting (data not shown) indicated that L. japonicus contained only one copy of the LFY/FLO homolog. LjUFO encodes an F-box protein, and the amino acid sequence is consistent with that of Pfo (Zhang et al., 2003). Southern blotting (data not shown) of LjUFO and Pfo showed the same pattern as each other, suggesting that LjUFO is Pfo, the UFO/FIM ortholog in L. japonicus. From here on, therefore, we refer to this gene as Pfo.

Most of the A, B, C, D, and E function genes belong to the plant-specific MIKC-type MADS box gene superfamily, which encodes transcriptional regulators with MADS domain, intervening domain, keratin-like domain, and C-terminal domain. We obtained sequences for 11 MIKC-type MADS box genes/fragments from L. japonicus. Based on phylogenic analysis and comparison of gene structures, we grouped them into different homologs of A, B, C, D, and E function genes, respectively (Fig. 4a; Supplemental Table I). All of the identified A, B, C, D, and E homologs possessed typical MADS domains (Fig. 4b) due to our primer-designing strategy, but varied to different extents in their C-terminal sequences (Fig. 4c). Two of them, LjAP1a and LjAP1b, were identified as homologs of AP1/SQUA (a floral meristem identity and A function gene) and CAL. Like PEAM4, the AP1/SQUA homolog in pea, LjAP1a and LjAP1b encode proteins that lack the characteristic C-terminal CaaX motif and are thus unable to be prenylated (Fig. 4c). LjPIa and LjPIb are homologs of PI/GLO (a B function gene) that share 90% similarity through the 170 amino acids at their N termini, including the M, I, and K domains. The C domain of LjPIb is about 30 amino acids shorter than that of LjPIa and other PI/GLO homologs (Fig. 4c). Therefore, LjPIb, but not LjPIa, lacks the canonical PI motif that was shown to be necessary for PI as a B function protein in Arabidopsis (Lamb and Irish, 2003). Analysis of soybean (Glycine max) expressed sequence tags (ESTs) in silico showed a similar phenomenon: GmPI2 (deduced from soybean ESTs together with GmPI1) has the PI motif in its C terminus, whereas GmPI1 lacks the PI motif. The PI/GLO gene duplication and C-terminal truncation appears to have occurred widely in legumes. The phylogenic tree shown in Figure 4a indicates that the LjAP3-like protein is located at the root of the AP3/DEF (another B function protein) subfamily, while LjAP3, ALFBMP (the AP3 homolog in Medicago sativa), and GmAP3 (deduced from soybean ESTs) form a legume clade in the AP3/DEF subfamily. Moreover, LjAP3 contains the PI-derived motif in its C terminus, but the LjAP3-like protein does not. LjAGa and LjAGb appear to have resulted from a recent duplication in L. japonicus; we found partial genomic sequences for LjAGa and LjAGb in GenBank (accession nos. AP004549 and AP004519) that contained 5 of their 3′ exons. The exon positions and sizes were conserved among LjAGa, LjAGb, and AG/PLE. A Q-rich motif and extra 20 amino acids were found at the C terminus of LjAGa but not in the AG/PLE protein. Such motifs are considered to belong to transcriptional activators (Tiwari et al., 2003), implying that LjAGa may act as an activator. The sequence of LjSEP3 was deduced from the L. japonicus genomic sequence (accession no. AP004516). Comparisons between the cDNA and genomic sequences revealed that LjSEP3 had 8 exons conserved in number and size with SEP3 (an E function gene) from Arabidopsis.

Figure 4. Sequence analysis of ABC function gene homologs in L. japonicus. a, A phylogenetic tree generated using neighbor-joining method from the predicted amino acid sequences of genes in the MIKC subfamily of MADS from Arabidopsis, Antirrhinum, pea, M. sativa (more ...)

In summary, based on the analysis of 13 homologs of LFY/FLO, UFO/FIM, and the ABCDE function genes, it appears that gene duplication must have taken place in the PI/GLO and AG/PLE clades in the L. japonicus genome, and some duplication products, such as LjPIb and LjAGa, have lost or gained protein domains that are functional in Arabidopsis.

Figure 5. Expression patterns of floral patterning genes in Gifu. Probe: a1 to 4, LjLFY; b1 to 4, Pfo; c1 to 4, LjAP1a; d1 to 4, LjAP1b; e1 to 4, LjPIa; f1 to 4, LjPIb; g1 to 4, LjAGa; h1 to 4, LjAGb; i1 to 4, LjSEP3; and j1 to 3, LjAP3. SAM was labeled as a cross. (more ...)

Like Stp, the UFO/FIM homolog in pea plants (Taylor et al., 2001), Pfo was also expressed at the margin between the compound leaf primordium and the SAM (Fig. 5b, subsection 1), and later at the boundary of the leaflets (data not shown). Pfo was expressed in the floral meristem with a similar pattern to that of FIM, UFO, and Stp (Simon et al., 1994; Ingram et al., 1995; Taylor et al., 2001). At floral Stage 0, Pfo was expressed at the center of the floral meristem; by Stage 1, the expression domain had extended throughout the floral meristem (Fig. 5b, subsection 2). By Stage 2, the transcript had ceased at the center and edge of the floral meristem and retained where the petal and stamen later formed (Fig. 5b, subsection 3). During Stage 3, Pfo expression was confined to the base of floral organs (Fig. 5b, subsection 4). Interestingly, Pfo was expressed similarly in both wild-type and pfm mutant plants. In the primary floral-like meristems, Pfo expression appeared first at the center of the meristem, then changed to be expressed in the peripheral region (Fig. 6h) and eventually formed a ring. This pattern was repeated in the secondary floral-like meristem (Fig. 6i). Almost the same expression pattern was seen for Pfo in pfo-2 mutant plants (Fig. 6, q and r). Moreover, RNA in situ hybridization showed consistent results between pfo-1 and pfo-2 plants.

Figure 6. Expression patterns of floral patterning genes in pfm and pfo-2. a to i, pfm and j to r, pfo-2. a to c and j to m, LjAP1a probe; d to f, n, and o, LjAP1b probe; g and p, LjAGa probe; and h, i, q, and r, Pfo probe. Bar = 50 μm.

In a similar manner to those of LjLFY, transcripts of LjAP1a were first detected in a group of cells in the I2s (Fig. 5c, subsection 1) where the first floral anlage formed. Thereafter, transcripts were confined to the next floral anlagen. During Stages 1 to 3, the transcripts gradually decreased in the center of the floral meristem (Fig. 5c, subsection 2). At Stage 4, LjAP1a was expressed in the sepals and petals, but was undetectable in the outer and inner stamens (Fig. 5c, subsection 3). This pattern was maintained until the flower matured (Fig. 5c, subsection 4). LjAp1b showed a similar expression pattern to that of LjAP1a (Fig. 5d, subsection 1–4), although its signal was always weaker in sepals and petals. However, the expression patterns of LjAP1a and LjAP1b were altered in pfm mutant plants. The expression of LjAP1a and LjAP1b in pfm I2s was similar to that in Gifu (Fig. 6d). When abaxial sepal primordium was initiated in the primary floral meristem, LjAP1a and LjAP1b were expressed throughout the floral meristem (Fig. 6, a and e). Later on, LjAP1a and LjAP1b were ectopically expressed in the region next to the sepal primordia, where another round of floral-like meristems would form, whereas their expression decreased in the sepal and carpel primordia (Fig. 6, b and f). The expression of LjAP1a and LjAP1b was maintained in the primary floral meristem until the formation of successive rings of floral-like meristems (Fig. 6c). The expression patterns of LjAP1a and LjAPb in pfo-2 mutant plants were similar to those in pfm. LjAP1a and LjAP1b were expressed in the pfo-2 I2, similar to the expression pattern in wild type (Fig. 6, j and n). Later, LjAP1a and LjAP1b were expressed throughout the primary floral meristem (Fig. 6, k and n), and they were ectopically expressed next to the sepal primordia (Fig. 6, l and o) where the secondary floral-like meristems initiated, whereas their expression decreased in the sepal and carpel primordia.

Interestingly, LjPIa was expressed in a unidirectional pattern over time. At Stage 2, LjPIa transcripts first appeared at the abaxial side of the floral meristem, covering an area 2 to 3 cells wide (Fig. 5e, subsection 1) where abaxial anlage would form. By Stage 3, the expression pattern had shifted into the region where lateral anlagen would form (Fig. 5e, subsections 2 and 3). LjPIa was consistently expressed in petals and stamens throughout flower development (Fig. 5e, subsection 4). LjPIb displayed almost the same expression pattern as LjPIa (Fig. 5f, subsections 1–4). LjAP3 was first expressed at Stage 4 between the sepals and the carpel (Fig. 5j, subsection 1), where the petals and stamens would form. Thereafter, the LjAP3 transcripts were found strictly within the petals and stamens as the flower matured (Fig. 5j, subsections 2 and 3). Expression of the 3 B function orthologs, LjPIa, LjPIb, and LjAP3, were not detected in either pfm or pfo-2 mutant plants (data not shown). This is consistent with the absence of petals and stamens in pfm and pfo-2 mutants, indicating that their expression was linked to the identity of petals and stamens in L. japonicus.

LjAGa and LjAGb were expressed in a similar pattern during floral ontogeny. Unlike the C function gene AG or PLE in Arabidopsis and Antirrhinum, these genes were first expressed at Stage 2, in the center of the floral meristem in an area of 5 to 6 cells (Fig. 5g, subsection 1, and 5h, subsection 1). At Stage 3, the expression domains of both genes extended toward a position where the inner and outer stamen primordia would form and decreased at the center where the carpel primordium had initiated (Fig. 5g, subsections 2 and 3, and 5h, subsections 2 and 3). In the heterogeneous whorl, LjAGa, LjAGb, and LjAP1a, LjAP1b displayed complementary expression patterns: LjAGa and LjAGb were expressed only where the outer stamen primordia would form, whereas LjAP1a and LjAP1b were expressed only where the petal primordia would form (Fig. 7). LjAGa and LjAGb were continually expressed in the outer and inner stamens until the flower matured (Fig. 5g, subsection 4, and 5h, subsection 4). In the pfm and pfo-2, LjAGa was transiently expressed at the center of the primary floral meristem at the Stage 2 (Fig. 6, g and p) at the time when the carpel-like primordia were initiated, but then its expression subsequently disappeared. Our observations that LjAP1a/LjAP1b and LjAGa were expressed in a complementary manner in all the Gifu, pfm, and pfo-2 plants, indicates the antagonistic relationship between the A and C function genes and further supports that the identity conferred at the center of the floral-like meristems in the mutants was carpel like.

Figure 7. Summary of the expression patterns of A, B, C, and E function genes during floral ontogenesis. The false colorings superimposed on SEM of the SAM indicate the expression patterns of different function genes. Bar = 50 μm.

Like LjPIa and LjPIb, expression of LjSEP3 was first observed at Stage 2 in an area 4 to five cells wide near the abaxial side of the floral meristem (Fig. 5I, subsection 1) where the abaxial anlage would be formed later. At Stage 4, its expression disappeared from the carpel primordium but was maintained in the petals and stamens until the flower matured (Fig. 5I, subsections 2–4). The transcripts of this SEP homolog were not detected in either pfm or pfo-2 (data not shown), suggesting that LjLFY and Pfo were also necessary for the transcription of this SEP homolog.

LjLFY is functionally divergent from its homolog in Arabidopsis with respect to the determination of the floral meristem identity and activation of the floral organ identity genes. Shoot-like structures replaced flowers in the early stages of floral development of Arabidopsis lfy plants (Weigel et al., 1992), whereas in pfm plants, the I2 meristem could produce floral meristems with sepal and carpel primordia without a functional LjLFY gene. There are at least 2 explanations for this: (1) LjLFY is unnecessary for the formation of the floral meristem, or (2) some other components are involved, such as AP1 homologs, but there is redundancy between them. In pfm, LjAP1a and LjAP1b are ectopically expressed in the floral meristem, and one of the TFL1/CEN (inflorescence identity genes in Arabidopsis and Antirrhinum) homologs, LjCEN1 was expressed normally (Guo et al., 2004; data not shown), indicating that regulation among LFY, TFL1, and AP1 homologs (Ratcliffe et al.,1999; Hempel et al., 2000) in L. japonicus differs from that in Arabidopsis. Furthermore, we found that 3 B and 1 E function genes were not expressed in pfm at detectable levels, whereas LjAP1a/LjAP1b were ectopically expressed and LjAGa/LjAGb expression was down-regulated in a spatially specific manner. These data suggest that LjLFY may be required for the activation of B function genes in a similar manner to LFY in Arabidopsis (Parcy et al., 1998), but, unlike LFY, it is not necessary for the initiation of A and C function genes. LjLFY may be required, however, for maintenance of their correct expression patterns and levels.

In L. japonicus, pfm and pfo mutants give rise to repetitive floral meristems between sepals and carpel. However, the determinacy in L. japonicus may still require the C function since (1) LjAGa and LjAGb maintain their expression in the outer and inner stamen primordia during carpel formation, (2) in petalous, the pea C function loss mutant, floral meristem determination is lost, and (3) LjAGa was found to be down-regulated in pfo and pfm where an indeterminate ectopic floral meristem was produced. It has been observed that ectopic flowers could be produced to different extents in fim mutant of Antirrhinum, stp of pea, and pfo of L. japonicus, respectively, and these differences could be truly species specific (Zhang et al., 2003). The very stable indeterminacy in both pfm and pfo, showing a highly regulated repetitive pattern of floral meristems, is in contrast to the phenotype of lfy and ufo mutants in Arabidopsis, where only some loss of lateral and apical determinacy during floral organ development was observed in ufo (Simon et al., 1994; Ingram et al., 1995). This difference is probably due to the functional differences between both Uni/LjLFY and Stp/Pfo in the two species.

It is well recognized that the ABC function genes establish specific expression domains for floral patterning. We summarize the expression patterns of L. japonicus homologs of ABC function genes in Figure 7. In L. japonicus, sepal, carpel, and common primordia were initiated at Stages 3 and 4. LjAGa and LjAGb were expressed no later than the B function genes at Stage 2, potentially allowing the carpel primordium to be initiated earlier than the petal and stamen primordia. The expression of LjAGa and LjAGb weaken in the carpel primordium after Stage 2, further indicating that the determination of carpel identity could occur earlier and more quickly in L. japonicus than in Arabidopsis and Antirrhinum. Moreover, LjAGa and LjAGb are expressed in both whorls 2 and 3 beginning at Stage 3, leading to the development of 2 whorls of stamens. In whorl 2, LjAP1a/LjAPb and LjAGa/LjAGb are expressed in a complementary manner to each other, which is consistent with the heterogeneous nature of whorl 2. Finally, the unidirectional activation of LjPIa, LjPIb, and LjSEP3 from the abaxial to adaxial sides is consistent with the observed initiation of petal and stamen primordia in the same direction.

In accordance with previous studies, we found that the molecular factors controlling floral organ development are well conserved between L. japonicus, Antirrhinum, and Arabidopsis. We have proposed, however, that four processes may contribute to the unique floral patterning in L. japonicus: alteration of the function of key genes, gene duplication, loss or gain of functional protein domains, and changes in gene expression patterns. These processes form the molecular basis for the specific order of initiation, the number of and the position of the petals and stamens during the floral ontogeny among Papilionoideae.

The pfm mutant was isolated from an EMS-mutagenized M2 population from The Sainsbury Laboratory and John Innes Center (Perry et al., 2003). This line, designated SL1203, contained 3 siblings in the M2 generation. One member (SL1203-1) displayed wild-type phenotype containing homozygous wild-type LjLFY allele, which did not segregate any mutants with floral or leaf phenotypes at the M3 generation (135 progenies). Another member (SL1203-2) displayed dwarf phenotype and was infertile. In contrast, SL1203-3 showed stable floral defects and a reduced compound leaflet number. SL1203-3 was propagated and maintained by cutting, and the resultant cuttings displayed phenotypes consistent with that of the original plant. Samples used for SEM, RNA in situ hybridization, and DNA analysis were harvested from cuttings.

The pfo-2 mutant was isolated from a separate EMS-mutagenized M2 population (n = 4,000) generated in Shanghai. The mutant line was designated F0795; there were 11 siblings in the M2 line, designated F0795-1 to F0795-11. F0795-1 and F0795-2 showed phenotypes similar to that of SL1203-3 (above). Samples used for RNA in situ hybridization and DNA analysis were harvested from F0795-1, F0795-2, and mutant plants segregating in the M3 generations from the heterozygous plant (F0795-11) for pfo-2.

The pfo (renamed pfo-1 in this paper) mutant plants were kindly provided by Dr. Pierre R. Fobert (National Research Council Canada).

The cloning of 13 homologs of LFY/FLO, UFO/FIM, and the ABCDE function genes was conducted by the reverse transcription-PCR and cDNA or genomic library screening (for details, see Supplemental Table I). The PCR products were cloned into the pGEM-T easy vector (Promega) and sequenced. Sequences were analyzed using the Vector NTI v.6.0.0.0 and homologous alignments were performed using Bioedit v.5.0.9. Amino acid alignments, including M, I, K, and C domains were used to obtain the phylogenetic with the neighbor-joining ClustalX program (version 1.83, February 2003).

Primers SL0805 (462 nucleotides upstream of the putative start codon) and SL0806 (198 nucleotides downstream of the putative stop codon) were designed according to the genomic sequence and used to PCR amplify the LjUFO/Pfo genomic fragment from mutant plants SL1203-1, SL1203-3, F0795-1, and F0795-2. The resulting PCR products were sequenced, and the identified point mutation gave rise to a CAPS marker and was subsequently used for the genetic linkage assay. The LjLFY genomic fragment was amplified from Gifu genomic DNA using degenerate primers SL0799 and SL0800. The genomic DNA fragment was additionally isolated 3 independent times from SL1203-3 cuttings as well as the pfo-2 mutant plants. Sequencing of these PCR products showed point mutation consistently. Simultaneously, LjLFY from the normally flowerings plant SL1203-1 showed unimpaired at the same site.

Sequence data from this article have been deposited with the EMBL/GenBank data libraries under accession numbers AY770393 to AY770405.

We thank Dr. Julie Hofer for critical reading and Dr. Pierre Fobert for providing the pfo mutant. We also thank two anonymous reviewers for providing much advice.