An apparent reversal in floral symmetry in thelegume Cadia is a homeotic transformationHelene L. Citerne*†‡, R. Toby Pennington*, and Quentin C. B. Cronk§

*Royal Botanic Garden Edinburgh, 20a Inverleith Row, Edinburgh EH3 5LR, United Kingdom; †Institute of Molecular Plant Sciences, University of Edinburgh,Edinburgh EH9 3JR, United Kingdom; and §University of British Columbia Botanical Garden, Room 206, Campbell Building, 6804 Southwest Marine Drive,Vancouver, BC, Canada V6T 1Z4

Edited by Michael J. Donoghue, Yale University, New Haven, CT, and approved June 26, 2006 (received for review February 8, 2006)

Within papilionoid legumes, characterized by flowers with strongbilateral symmetry, a derived condition within angiosperms, Cadia(Cadia purpurea) has reverted to radially symmetrical flowers.Here, we investigate the genetic basis of this morphologicalreversal. Two orthologues of the floral symmetry gene CYCLOIDEA(CYC) demarcate the adaxial (dorsal) region of the flower in typicalpapilionoid legumes. In the model legume Lotus japonicus, one ofthese LegCYC genes has been shown, like CYC, to be required forthe establishment of floral bilateral symmetry. This study showsthat these genes are expressed in the adaxial region of the typicalpapilionoid flower of Lupinus, which belongs to the same papil-ionoid subclade as Cadia. In Cadia, these genes also are expressed,but the expression pattern of one of these has expanded from theadaxial to the lateral and abaxial regions of the corolla. This resultsuggests that the radial flowers of Cadia are dorsalized and,therefore, are not a true evolutionary reversal but an innovativehomeotic transformation, where, in this case, all petals haveacquired dorsal identity. This study raises a question over otherputative reversals in animals and plants, which also may be crypticinnovations.

developmental gene � floral evolution � CYCLOIDEA � Leguminosae(Fabaceae)

Most members of the plant family Leguminosae (Fabaceae)have flowers that are characterized by a single axis of

symmetry (bilateral symmetry or zygomorphy). The majority oflegume species, from the subfamily Papilionoideae and account-ing for �5% of all f lowering plant species, are characterized byhaving typical ‘‘pea’’ (papilionoid) flowers with pronouncedbilateral symmetry particularly in the corolla (illustrated here byLupinus nanus; Fig. 1a). The typical papilionoid flower has threedistinctive petal types: a single upper (adaxial or dorsal) petal(the ‘‘standard’’), two lateral petals (‘‘wings’’), and two lower(abaxial or ventral) petals (‘‘keel’’ petals). The androecium oftypical papilionoids also is bilaterally symmetrical, with stamensof different lengths along the dorsoventral axis, the adaxialstamens being shorter (Fig. 1a). However, within the Papilion-oideae, a number of unrelated genera have radially or nearlyradially symmetrical f lowers. The flowers of these atypicalspecies frequently have been considered primitive (reviewed inref. 1), but more recent phylogenetic studies suggest that they arederived from a zygomorphic ancestral state (2, 3). An exampleis Cadia Forsk., a genus of seven species of small shrubs fromArabia, Madagascar, and eastern Africa. Cadia has actinomor-phic pendent f lowers in solitary or few-f lowered axillaryracemes (ref. 4; Fig. 1b). These flowers produce abundant nectarbut no scent, suggesting these may be pollinated by birds (1).Cadia occupies a relatively isolated phylogenetic position (J. S.Boatwright and D. Edwards, personal communications) withinthe genistoid clade of papilionoid legumes (83 genera and 2,350species; ref. 5). Its closest relatives are from the tribe Podalyrieae(J. S. Boatwright and D. Edwards, personal communications),which have typical zygomorphic papilionoid flowers.

The early development of Cadia (Cadia purpurea) f lowers issimilar to that of most papilionoid species with zygomorphicflowers where sepals, petals, and stamens are initiated unidi-rectionally, starting on the abaxial side (6). In typical papilionoidlegumes, although organogenesis is asymmetric, a phase ofuniform organ growth precedes the unequal development alongthe dorsoventral axis, and differential organ development occursat an advanced stage of floral ontogeny (7). By contrast, in C.purpurea, f loral organs continue to develop equally after orga-nogenesis until maturity (6). Tucker (6) therefore interpreted thephenotype of C. purpurea as ‘‘neotenous,’’ that is, retaining thecharacteristics of early flower development (uniform growth)and not entering the asymmetric phase. In genetic terms, thisinterpretation suggests that the radial phenotype of C. purpureamay be caused by a loss of expression of genes that affect f loral

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNAS office.

Data deposition: The sequences reported in the paper for Lupinus nanus LEGCYC1A andLEGCYC1B and Cadia purpurea LEGCYC1A and LEGCYC1B have been deposited in theGenBank database (accession nos. AY382156, AY382155, AY225826, and AY225825).

‡To whom correspondence should be addressed. E-mail: h�citerne@hotmail.com.

© 2006 by The National Academy of Sciences of the USA

Fig. 1. Mature whole and dissected flower of L. nanus and C. purpurea. (a)Detail of an L. nanus flower, showing strong bilateral symmetry typical ofpapilionoid flowers. Individual standard (D, dorsal), wing (L, lateral), and keel(V, ventral) petals are shown, as well as the monadelphous (with unitedfilaments) androecium with stamens of different lengths (lateral view on theleft, and dissected dorsal and lateral�ventral stamens on the right; the threedorsal stamens are shorter than the lateral and ventral ones). (b) C. purpureaflowers are radially symmetrical, with equal dorsal (D), lateral (L), and ventral(V) petals and stamens.

www.pnas.org�cgi�doi�10.1073�pnas.0600986103 PNAS � August 8, 2006 � vol. 103 � no. 32 � 12017–12020

EVO

LUTI

ON

Dow

nloa

ded

by g

uest

on

May

29,

202

1

organ differentiation during the latter stage of development oftypical zygomorphic papilionoid flowers.

Genes that are important for the control of f loral symmetryhave been identified in Antirrhinum majus L. (Veronicaceae,Lamiales) (8–10). Two closely related genes CYCLOIDEA(CYC) and DICHOTOMA (DICH) demarcate the dorsal regionof the flower and affect the development of floral organs alongthe dorsoventral axis. CYC and DICH expression is maintainedthroughout floral development in the adaxial region (8, 9).Double knockout mutants for CYC and DICH have a fullyradially symmetric phenotype characterized by ventralization offloral organs (i.e., resembling the ventral phenotype of wild-typeflowers) (8).

CYC�DICH homologues have been identified in legumes,referred to as LegCYC (11). There is strong functional evidencethat at least one LegCYC gene, LjCYC2 (Lotus japonicus 1 in ref.11), is involved in the control of f loral symmetry in the modelpapilionoid legume Lo. japonicus (12), a member of the robin-ioid clade. Like CYC, LjCYC2 is expressed in the adaxial regionof developing flowers and has been found to be important forestablishing dorsal identity (12). Lotus plants expressing LjCYC2constitutively have lateral and ventral petals that are similar todorsal petals in shape and epidermal cell type, whereas thosewith reduced LjCYC2 activity have lateral petals with a ventral-ized phenotype (12). We show here that two LegCYC genes,belonging to the same gene clade (LegCYC1) as LjCYC2 (11)(sequences given in Fig. 4, which is published as supportinginformation on the PNAS web site) are expressed on the adaxial

side throughout floral development in another papilionoid le-gume, Lupinus (L. nanus), from the genistoid clade. L. nanus hastypical zygomorphic pea flowers, suggesting that these genes arecandidates for the control of f loral symmetry in genistoidlegumes. The expression of these two LegCYC genes is examinedin the radially symmetrical f lowers of Cadia (C. purpurea), which,like Lupinus, is a genistoid legume (2). Comparison of LegCYCexpression pattern between an unusual legume such as C.purpurea and a related zygomorphic species can illuminate thegenetic basis of morphological reversal in relation to floralsymmetry in legumes. The reversibility of evolution is an im-portant problem, but few examples of apparent phenotypicreversals have been investigated genetically (reviewed in ref. 13).Here, we establish whether the apparent reversal of f loralsymmetry in Cadia is the result of loss of expression of candidategenes, which could be described as ‘‘true reversal,’’ or is due tomodified gene expression causing the replication of ancestralstates.

ResultsLegCYC1A and LegCYC1B RNA was detected in floral tissue ofL. nanus in a pattern similar to Antirrhinum CYC (ref. 8; Fig. 2).Both genes were detected in floral meristems before organo-genesis, on the adaxial side of the meristem (Fig. 2 d and g). Atmore advanced developmental stages, both genes were detectedin the corolla (Fig. 2 e, f, h, and i). Similar to CYC, expressionof LegCYC1B in the dorsal petal was found in the inner cell layersat the site where cell division was repressed early in organogen-esis. Although the expression domains of LegCYC1A andLegCYC1B are largely overlapping, suggesting functional redun-dancy, LegCYC1A appears to have a reduced expression domainrelative to LegCYC1B.

RT-PCR results in L. nanus suggest that, in agreement withthe findings in situ, both LegCYC1A and LegCYC1B are florallyexpressed (Fig. 3) and only in the adaxial part of the developingflower (Fig. 3). RT-PCR also shows that this expression is foundin floral buds at more advanced developmental stages. Bothcopies are transcribed at this stage not only in the standard(dorsal) petal but also in the dorsal anthers.

Fig. 2. Patterns of RNA in situ hybridization of LegCYC1A and LegCYC1B inL. nanus inflorescenses (a and b, respectively), with details of floral meristemsat different development stages (d–f, LegCYC1A; g–i, LegCYC1B). Longitudi-nal sections of L. nanus inflorescences show floral meristems (fm) in the axil ofbracts (B) on the abaxial side. More mature flowers are at the base of theinflorescence. RNA from LegCYC1A and LegCYC1B is detected in the adaxialpart of floral meristems before organogenesis (d and g) and during floralorgan development (e, f, h, and i). Although both copies have a similarexpression pattern, LegCYC1B appears to have a wider expression domainthan LegCYC1A. Negative control (sense probe) is shown (c). Ad, adaxial; Ab,abaxial; St, stamen; AbS, abaxial sepal. (Scale bars: a–c, 500 �m; d–h, 200 �m;f and i, 250 �m.)

Fig. 3. RT-PCR analysis of LegCYC1A (1A) and LegCYC1B (1B) from whole-flower buds at early developmental stages (�1 mm diameter) and dissectedfloral tissue from L. nanus and C. purpurea flowers are later stages. The corollaand androecium in both species were dissected as described in Materials andMethods. Intron splicing in both genes allowed to distinguish between cDNAand genomic DNA. cDNA products amplified by actin-specific primers wereused as a positive control. gDNA, genomic DNA; �ve, negative control (notemplate); D, dorsal; L, lateral; V, ventral.

12018 � www.pnas.org�cgi�doi�10.1073�pnas.0600986103 Citerne et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

1

In C. purpurea, RT-PCR from young flower buds revealed thatLegCYC1A and LegCYC1B also are expressed in young floralbuds (Fig. 3). However, the pattern of expression of these genesin dissected tissue from the corolla and androecium at anadvanced developmental stage are different from each other andfrom their L. nanus orthologues. In the corolla of C. purpurea,LegCYC1A is expressed only in the dorsal petal, and its level ofexpression appears moderate to weak (Fig. 3). LegCYC1B,however, is expressed in all petals (Fig. 3), showing an expansionof the expression domain of this gene with respect to L. nanus,which is consistent with the radial phenotype of the corolla. Inaddition, unlike in L. nanus, neither LegCYC1A nor LegCYC1Bappear to be expressed in the androecium of C. purpurea (Fig. 3).

DiscussionThis study shows that two CYC-like genes, based on theirexpression pattern in the zygomorphic flowers of the typicalpapilionoid legume Lupinus, are likely to control f loral symme-try in the genistoid clade to which Lupinus belongs. These genes,which have duplicated independently from CYC and DICH (11),may be functionally redundant. Therefore, not only do CYC-likegenes appear to have been recruited independently for thecontrol of zygomorphy in Lamiales and legumes (12), butmaintenance of partially redundant copies also has occurredindependently in these lineages. The role of these genes in thedevelopment of zygomorphic flowers in this legume clade issupported by functional evidence from the orthologous LjCYC2in Lo. japonicus (12). This result provides a framework forexamining the genetic control of f loral symmetry in a clade thatcontains atypical species with actinomorphic flowers.

In C. purpurea, like in Lupinus, both LegCYC1A andLegCYC1B are transcribed in early developing flowers. Takingontogeny into account, this result is not surprising because earlydevelopment of C. purpurea f lowers is similar to that of mostpapilionoid species with zygomorphic flowers, with unidirec-tional organ initiation starting on the abaxial side (6). It remainsto be demonstrated whether LegCYC genes have the sameexpression pattern in early floral development in C. purpureacompared with typical papilionioids. During the latter stages offloral development, whereas floral organs undergo unequalgrowth along the dorsoventral axis in typical papilionoid legumessuch as Lupinus, they develop equally in C. purpurea (6). Thisequal development could suggest that genes that control dorso-ventral asymmetry are not expressed after organogenesis in C.purpurea. The molecular data presented here, however, suggestthat rather than failing to maintain CYC expression during thelater stages of f lower development, one CYC-like gene,LegCYC1B, is expressed in all five petals of C. purpurea. Anothercopy, LegCYC1A, is expressed adaxially but may be down-regulated. The expansion of LgCYC1B expression appears to bewhorl-specific. In the androecium, neither LegCYC1B orLegCYC1A mRNA was detected in C. purpurea, although ex-pression was found in the dorsal region in L. nanus. The apparentloss of CYC expression in the stamens is intriguing but difficultto interpret. Mature stamens in C. purpurea are identical mor-phologically, and the Cadia androecium is actinomorphic, but itis hard to establish whether the stamens have a ventralizedphenotype. Without more information on the role of CYC instamen development of other legume species, it is difficult toplace our observations in context.

The expression of LegCYC1B in the corolla is reminiscent ofthe backpetals mutation in Antirrhinum (9). This mutant hasectopic expression of CYC in the lateral and ventral petals. Atransposon insertion in an AT-rich site �4.2 Kb upstream of startcodon is believed to affect a cis-acting region that normallysuppresses CYC transcription during the latter stages of devel-opment in wild-type Antirrhinum f lowers (9). It may be that a

change in cis regulation also has led to the expansion of theexpression domain of LegCYC1B in C. purpurea.

The occurrence of a putative ancestral state such as radialsymmetry within a clade that has a derived character (zygomor-phy) is frequently referred to as an evolutionary reversal. Ourresults indicate that from a genetic point of view, the radialsymmetry of Cadia is an evolutionary innovation correlated, inpart, by the expansion of the expression domain of a CYC-likegene in the corolla. The functional significance of this expressionpattern is suggested by transgenic experiments in Lo. japonicus(12). Overexpression mutants showing an expansion of expres-sion of LjCYC2 displayed a dorsalized phenotype (12), support-ing the hypothesis that changes in LegCYC1B expression in Cadiahave an effect on flower development. Because CYC is a markerfor dorsal identity, these changes can be considered homeotic,with the lateral and ventral petals of Cadia assuming a dorsalphenotype. This interpretation also is supported by morphology.In Cadia, the five petals are large and bilaterally symmetric,features that are typical of the papilionoid standard petal. Bycontrast, wing and keel petals in typical papilionoid flowers areasymmetric and small relative to the standard. Partial dorsal-ization also has been noted in Mohavea confertiflora, wherestamen number is reduced from four to two compared with itsclose relative A. majus and is correlated with expansion of CYCand DICH expression from the adaxial to the lateral region (14).Although loss of function is often perceived as the simplestscenario to explain possible reversals to radial symmetry, it hasbeen suggested that, on close morphological inspection, manyshifts zygomorphy to actinomorphy involve other genetic ordevelopmental changes (15). Our findings support the hypoth-esis that such homeotic-like transformations may play an im-portant role in establishing morphological diversity.

Materials and MethodsRNA in Situ Hybridization. RNA in situ hybridization was carriedout as described by K. Barton (available upon request; TheCarnegie Institute of Washington, Stanford, CA) on 7-�mlongitudinal sections of L. nanus inflorescences fixed in 4%paraformaldehyde. Digoxygenin-labeled probes specific toLegCYC1A and LegCYC1B were prepared from linearized PCRfragments, amplified from the 5� region of the ORF by usingprimers LEGCYC1-F2 (5�-CTT TCY TTA ACC CTG AAAATG CTT C-3�) at the start of the ORF and LEGCYC1A-R1(5�-CTA CYA CTA CCC CTT CTG G-3�)�LEGCYC1B-R1(5�-CAA GCS GGT TCC TTY TGT G-3�), respectively (prim-ers and PCR conditions described in ref. 16), and cloned intopCR4 plasmid (Invitrogen, Carlsbad, CA).

Extensive and exhaustive attempts to provide in situs of Cadiaf lowers were unsuccessful. Fixation and sectioning were veryproblematic, probably because of the adaptation of the flowersto their xeric habitat (tough sepals covered in filamentoustrichomes).

RT-PCR. Tissue for RT-PCR was collected from young floral buds(�1 mm diameter). Tissue also was collected from the corollaand androecium of flowers at more advanced and comparabledevelopmental stages in L. nanus and C. purpurea, during thelatter stages of organ enlargement where these had undergonea high degree of dorsoventral differentiation in L. nanus andwhere the flowers, still in bud, could be easily dissected. Floralorientation in C. purpurea was determined with respect to thegynoecium, which is the only part of the flower showing dorso-ventral asymmetry. Total RNA was extracted by using QiagenRNeasy RNA extraction kit (Valencia, CA) from the followingtissues: the dorsal, lateral, and ventral petals in L. nanus, andeach individual petal from a single flower in C. purpurea, thethree adaxial stamens separated by a groove from the sevenlateral and ventral stamens (combined) in L. nanus, and the

Citerne et al. PNAS � August 8, 2006 � vol. 103 � no. 32 � 12019

EVO

LUTI

ON

Dow

nloa

ded

by g

uest

on

May

29,

202

1

dorsal (three uppermost) lateral (two on either side) and ventral(three lowermost) stamens in C. purpurea. cDNA was synthe-sized by using Qiagen Omniscript RT kit. RT-PCR was carriedout by using the locus-specific primers LEGCYC1A-F1 (5�- CCAGAA GGG GTA GTR GTA G-3�)�LEGCYC1B-F1 (5�-CACARA AGG AAC CWG CTT G-3�) and primer LEGCYC1-R1(5�-CAC TCY TCC CAR GAY TTT CC-3�), which is down-stream of the short intron at the 3� end of the LEGCYC ORFs(primers described in ref. 16). Locus specificity was confirmed bysequencing of RT-PCR products. RT-PCRs were repeated withcDNA from separate RNA extractions of tissue from individual

f lowers. Actin was used as a positive control (using primersdescribed in ref. 17). To ensure that the quantity of PCRproducts was comparable between samples, actin products wereassayed between 20 and 35 cycles. PCR products are shown in thelinear phase after 30 cycles.

We thank E. Coen and D. Luo for making this project possible; M.Hollingsworth, A. Clark, C. Baxter, and K. Coenen for technicalassistance; and the editor and two anonymous reviewers for constructivecomments on the manuscript. This work was supported by the CarnegieTrust for the Universities of Scotland.

1. Pennington, R. T., Klitgaard, B. B., Ireland, H. & Lavin, M. (2000) in Advancesin Legume Systematics, eds. Heredeen, P. S. & Bruneau, A. (Royal BotanicGardens, Kew, U.K.), Part 9, pp. 233–248.

2. Pennington, R. T., Lavin, M., Ireland, H., Klitgaard, B., Preston, J. & Hu, J.-M.2001. Syst. Bot. 26, 537–556.

3. McMahon, M. & Hufford, L. (2004) Am. J. Bot. 91, 1219–1230.4. Van der Maesen, L. J. G. (1970) Acta Bot. Neerl. 19, 227–248.5. Lewis, G., Schrire, B., Mackinder, B. & Lock, M., eds. (2005) Legumes of the

World (Royal Botanic Gardens, Kew, U.K.).6. Tucker, S. C. (2002) Am. J. Bot. 89, 748–757.7. Tucker, S. C. (2003) Plant Physiol. 131, 911–926.8. Luo, D., Carpenter, R., Vincent, C., Copsey, L. & Coen, E. (1996) Nature 383,

794–799.

9. Luo, D., Carpenter, R., Copsey, L., Vincent, C., Clark, J. & Coen, E. (1999) Cell99, 367–376.

10. Galego, L. & Almeida, J. (2002) Gene Dev. 16, 880–891.11. Citerne, H. L., Luo, D., Pennington, R. T., Coen, E. & Cronk, Q. C. B. (2003)

Plant Physiol. 131, 1042–1053.12. Feng, X., Zhao, Z., Tian, Z., Xu, S., Luo, Y., Cai, Z., Wang, Y., Yang, J., Wang,

Z., Weng, L., et al. (2006) Proc. Natl. Acad. Sci. USA 103, 4970–4975.13. Porter, M. L. & Crandall, K. A. (2003) Trends Ecol. Evol. 18, 541–547.14. Hileman, L. C., Kramer, E. M. & Baum, D. A. (2003) Proc. Natl. Acad. Sci. USA

100, 12814–12819.15. Donoghue, M. J., Ree, R. H. & Baum, D. A. (1998) Trends Plant Sci. 3, 311–317.16. Citerne, H. L. (2006) Edinb. J. Bot. 63, 119–126.17. Golz, J. F, Keck, E. J. & Hudson, A. (2002) Curr. Biol. 12, 515–522.

12020 � www.pnas.org�cgi�doi�10.1073�pnas.0600986103 Citerne et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

1