Legume flowers bear fruit

After the grasses, the legume family (which includes peas, beans, and soya) is the plant group of greatest importance to mankind. The legume family is a source of products and services from tropical timber to atmospheric nitrogen fixation. As the botanist E. J. H. Corner (1) put it, “From arctic circle to tropics, desert to pergola, bacteria to plough, field to mouth, and legend to science, Leguminosae invest our lives.” The family is huge, with ≈20,000 species. Legume diversity is so great (and much of it hidden away in relatively unexplored tropical rain forest) that it has been difficult to gain an overview. Recently, a landmark study has covered the whole family (2). The result, a bird's-eye view of the legumes, is a chronicle of how the legumes have played out 60 million years of dazzling “evolutionary jazz,” riffing and improvising on the basic themes provided by the characteristic leguminous leaf, flower, and fruit. Despite the family's vastness, it is not (by angiosperm standards) particularly old. Current estimates suggest that the family underwent a diversity explosion ≈50–60 million years ago, during which most of the major legume clades arose (3, 4). One of the reasons for the early and continued diversification of legumes may be coevolution with pollinating bees (Hymenoptera, superfamily Apoidea), which are by far the most common pollinators of the legumes and appear to be extremely effective at the job. Bees recognize complex shapes readily, and so the pronounced bilateral symmetry (zygomorphy) of many legume flowers, which leads to complex flower outlines and intricate floral mechanisms, may promote exclusive flower visitation of some species by specific pollinators (5). The important study by Feng et al. (6), in this issue of PNAS, now elucidates the molecular architecture underlying the development of zygomorphy in legume flowers and so provides an important tool for understanding the diversification of the family.

Floral zygomorphy has evolved in numerous groups of angiosperms independently (7–9). Much of our knowledge of the process, until now, came from studies of Antirrhinum, or snapdragon. In snapdragon, four genes were found to be pivotal (10). Two of these genes, CYCLOIDEA (CYC) and DICHOTOMA (DICH), are closely related TCP transcription factors that are asymmetrically expressed and therefore signal the positional information needed to make the top (adaxial) petals different from the bottom (abaxial) petals (11, 12). CYC and DICH have overlapping expression domains and partially redundant function with respect to each other. A cyc/dich double mutant is required to eliminate zygomorphy entirely. These TCP genes have some direct phenotypic effect, but their main action is to activate a MYB transcription factor RADIALIS (RAD), which appears to do much of the “heavy lifting” of determining the zygomorphic phenotype. Furthermore, the RAD transcript appears to be mobile, migrating laterally in the flower to determine the phenotype of the lateral petals. Thus, RAD acts as a “lateralizing factor” (10, 13). A fourth gene, DIVARICATA (DIV), is another MYB transcription factor. DIV acts abaxially to promote abaxial identity (10, 14). An important question in the evolution of zygomorphy is whether groups that have evolved zygomorphy independently of the Lamiales (to which Antirrhinum belongs) have used the same mechanism at the molecular level. It is known that legumes possess homologues (legCYC genes) of the CYCLOIDEA genes found in Antirrhinum (15–17). Now, because of the work of Feng et al. (6), we have functional information that the establishment of zygomorphy in legume flowers does indeed involve a CYC-like TCP gene. The TCP gene pathway has therefore been independently utilized in both Antirrhinum and legumes to perform the same function in floral development.

Feng et al. (6), working with the model legume Lotus japonicus (18), examined four legCYC genes. One of these genes, LjCYC2 proves to be a major player in establishing adaxial identity. First, they looked at the LjCYC2 mutant allele squared standard 1 (squ1), identified in a screen for mutants with abnormal floral morphology in 50,000 mutagenized lines. The squ1 mutation affects only the top (adaxial) petal, affecting its shape and cell identity.

The legume group to which L. japonicus belongs (the papilionoid clade) has the most highly developed floral zygomorphy in the legumes. The adaxial petal is known as a “standard” and is usually large and erect. In most species (including Lotus japonicus), it is the main contributor to pollinator-attracting display. Furthermore, the epidermal cells of the standard are conical-papillate (i.e., have a raised center). This feature is commonly found in petal epidermal cells, and it has been shown to alter the optical properties of the petal (19). The two lateral petals (the wings) have flat epidermal cells with wavy margins, whereas the abaxial petals (keel petals) have flat epidermal cells with linear margins. In the squ1 mutant, the epidermal cells of the standard become like those of the lateral petals.

Feng et al. (6) further confirm the importance of LjCYC2 in floral zygomorphy by creating transgenic plants. Their transgenic plants expressed either LjCYC2 constitutively or expressed an antisense (silencing) construct. Constitutive expression produced specific effects on the adaxial and lateral petals, such as the production of conical-papillate cells in the lateral petals, implying that enhanced expression of the adaxial identity gene can adaxialize the lateral petals. Conversely, antisense constructs caused the lateral petals to adopt both the shape and the epidermal cell type of the abaxial petals, implying that petal development defaults to an abaxial-type pathway in the absence of the adaxial identity gene (Fig. 1). Interestingly, the antisense constructs also affect the inflorescence morphology, suggesting a wider role for this legume CYC-like gene in comparison with its counterpart in Antirrhinum. The occurrence of fusion between the paired flowers in the antisense lines may be significant, because it implies that CYC may be involved in setting the boundaries between floral primordia in the adaxial region.

Fig. 1.

Exaggerated diagram showing the three petal forms of the wild-type papilionoid legume flower (B) and the floral transformations (arrows 1 and 2) (A and C) reported in Feng et al. (6). The standard petal (developmental domain of LjCYC2) is in blue, the wing petals (domain of action of the uncharacterized gene KEW1) are in yellow, and the keel petals are in orange. Suppression of LjCYC2 and KEW1 activity (arrow 1) leads to an abaxialized flower (A). Constitutive expression of LjCYC2 (arrow 2) leads to an adaxialized flower (C).

A key question is whether there is a legume counterpart of RAD involved as a lateralizing factor. Feng et al. (6) identify a candidate for such a lateralizing factor with the mutant keeled wing in lotus 1 (kew1). This mutant has abnormal lateral petals, which are abaxialized into keel-like petals. The characterization of the underlying gene is not complete, but the pattern of interaction reported here seems very different from RAD. The squ1/kew1 double mutant shows a much more severely abaxialized phenotype than either alone, implying some redundancy between LjCYC2 and the kew1 gene. In this respect, they behave more like CYC and DICH than CYC/DICH and RAD. This apparent difference in how lateral identity is achieved in Lotus, as opposed to Antirrhinum, implies that, although the mechanism is similar (at least in respect to the common use of CYC genes), it will probably be shown to embody major differences. Further characterization of the legume zygomorphy pathway will thus be very exciting for the light it will shed on the extent of the evolutionary convergence.

The independent recruitment of CYC-like genes for zygomorphy in unrelated plant groups is curious, to say the least. However, a study of the CYC-like gene TCP1 in Arabidopsis indicates there may be an ancestral and more widespread function of TCP genes. Specifically, they may be involved in signaling adaxial identity during the morphogenesis of lateral branches (20). There thus seems to be a TCP-gene-determined “prepattern,” to which the floral development pathway can come to respond during the evolution of zygomorphy. Therefore, the key step in the convergent evolution of floral zygomorphy in legumes and snapdragons may not lie in the independent recruitment of TCP genes to a novel function. Rather, it may lie in other genes coming to respond to TCP control, independently in different lineages.

The work of Feng et al. (6) opens up exciting possibilities to understand the colossal variation in legume flower morphology in the 20,000 or so species. There are strongly zygomorphic flowers in the caesalpinioid legumes (for instance Cercis, the redbud tree) as well as the papilionoid (pea-flowered) clade. However, caesalpinioid legumes are generally less zygomorphic than legumes in the papilionoid clade, and it will be interesting to see whether these patterns are the result of altered legCYC gene function. A particularly promising opportunity arises in those legumes that have strongly radialized flowers but that occur within legume groups that are otherwise highly zygomorphic (see ref. 21; Fig. 2). Another question involves the evolutionary radiation within groups of closely related species. Such radiation often involves small changes in floral morphology among species, such as the size and shape of the wings and standards, possibly as an adaptation to different pollinating insects. How do such small changes in legume flowers occur? It is possible that minor alterations in CYC-gene regulation or function could be involved (17), as has been shown in Antirrhinum relatives (22). On the other hand, it is equally possible that only downstream genes are involved. Feng et al. (6) have provided important tools with which to start answering these questions.

Fig. 2.

Diversity of legume flowers in the papilionoid clade. The photograph shows lupin (Lupinus nanus) with typical papilionoid flowers (only the standard and wings are visible, because the wing petals enclose the keel petals) (B), and Cadia, with a radial flower (A). Despite the atypical floral morphology, Cadia belongs to the papilionoid legume clade. [Reproduced with permission from Citerne et al. (15) (Copyright 2003, American Society of Plant Biologists).]