Content uploaded by Sophie Nadot
Author content
All content in this area was uploaded by Sophie Nadot on Jun 24, 2021
Content may be subject to copyright.
!
!
!
!
!
!"#$%&"'()*+%,'$%-+#.#/'001%(230#$"*4%#/%)"*%3--5%!"#$%&'()*%)+,-$%*&$.)/'('$0&1)
23,.45')6789)!"*%&-(1%'))'&"*4%#$%(+-6#4*4%31%70$*6#*+%8-+%)"*%'2)"-+9$%3*/*8#)%'/4%8-+%)"*%3*/*8#)%%
-8%)"*%'2)"-+9$%#/$)#)2)#-/:%8-+%/-/;&-<<*+&#'0%+*$*'+&":%'/4%*42&')#-/'0%2$*=%!"#$%#/&024*$%,#)"-2)%
0#<#)')#-/%2$*%#/%#/$)+2&)#-/%')%1-2+%#/$)#)2)#-/:%4#$)+#32)#-/%)-%$(*#&%&-00*'.2*$:%'/4%(+-6#4#/.%'%
&-(1%)-%1-2+%#/$)#)2)#-/9$%'4<#/#$)+')-+=%
!
!
!
>00%-)"*+%2$*$:%+*(+-42&)#-/%'/4%4#$)+#32)#-/:%#/&024#/.%,#)"-2)%0#<#)')#-/%&-<<*+&#'0%+*(+#/)$:%$*00#/.%
-+%0#&*/$#/.%&-(#*$%-+%'&&*$$:%-+%(-$)#/.%-/%-(*/%#/)*+/*)%$#)*$:%1-2+%(*+$-/'0%-+%#/$)#)2)#-/9$%,*3$#)*%-+%
+*(-$#)-+1:%'+*%(+-"#3#)*4=%?-+%*@&*()#-/$:%(*+<#$$#-/%<'1%3*%$-2.")%8-+%$2&"%2$*%)"+-2."%70$*6#*+A$%
(*+<#$$#-/$%$#)*%')B%
"))(BCC,,,=*0$*6#*+=&-<C0-&')*C(*+<#$$#-/2$*<')*+#'0%
%
?+-<%D')"*+#/*%E'<*+6'0:%FG0H/*%D#)*+/*:%?0-+#'/%I'33-2+%'/4%J-("#*%K'4-):%
!"*%76-02)#-/%-8%?0-+'0%J1<<*)+1=%
L/B%I*'/;D0'24*%M'4*+%'/4%N#&"*0%E*0$*/1:%*4#)-+$:%%
>46'/&*$%#/%O-)'/#&'0%P*$*'+&"%QR-02<*%STU=%
>&'4*<#&%V+*$$:%WXYX:%(=%ZS=%
LJOKB%[\Z;X;YW;]ZXZ\X;\%
^%D-(1+#.")%WXYX:%70$*6#*+%_)4=%
>&'4*<#&%V+*$$=%
!"#$%&'&()#"(*#*+,#--'",%./("'0'.",1(.*&('&2,.3%#*./(20'(#*/45((
6#3()#"("'7"#&2,3%#*8(&%03"%923%#*(#"(,#--'",%./(20'5(
Author's personal copy
The Evolution of Floral Symmetry
* †,‡
HE
´ LE
` NE CITERNE, FLORIAN JABBOUR, SOPHIE NADOT
†
AND CATHERINE DAMERVAL
*,1
*
UMR de Ge
´ne
´tique Ve
´ge
´tale, CNRS—Univ Paris-Sud—INRA—
AgroParisTech, Ferme du Moulon, 91190 Gif-sur-Yvette, France
†
Universite
´ Paris-Sud, Laboratoire Ecologie, Syste
´matique, Evolution,
CNRS UMR 8079-AgroParisTech, Orsay, F-91405, France
‡
Institute for Systematic Botany and Mycology, University of Munich,
Menzinger Strasse 67, 80638 Munich, Germany
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
II. Definitions of Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
III. Symmetry and Flower Development. . . . . . . . . . . . . . . . . . . . . . . . . 93
A. Establishment of Symmetry at Various Stages During
Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
B. Impact of Growth and Organ Elaboration on Floral symmetry . 94
C. Developmental Trajectories and Flower Symmetry . . . . . . . . . . 95
IV. Evolution of Flower Symmetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
A. Distribution of Symmetry among Extant Angiosperms . . . . . . . 97
B. Emergence of Zygomorphy during Angiosperm Evolution in
Relation to Insect Diversification . . . . . . . . . . . . . . . . . . . . . . . 98
C. Architecture of Flowers and Inflorescences—What is Their
Impact on Floral Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
V. The Significance of Symmetry in Plant–Pollinator Interactions . . . . . 108
A. Zygomorphy and Outcrossing Strategies . . . . . . . . . . . . . . . . . . 109
All authors contributed equally to this review
1
Corresponding author: E-mail: catherine.damerval@moulon.inra.fr
Advances in Botanical Research, Vol. 54 0065-2296/10 $35.00
Copyright 2010, Elsevier Ltd. All rights reserved. DOI: 10.1016/S0065-2296(10)54003-5
Author's personal copy
86 H. CITERNE ET AL.
B. Pollinator Preferences and their Perception of Symmetry . . . . . . 112
C. Floral Symmetry and Pollination Syndromes . . . . . . . . . . . . . . . 112
D. Variability of Floral Traits in Zygomorphic and Actinomorphic
Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
VI. Molecular Bases of Flower Symmetry. . . . . . . . . . . . . . . . . . . . . . . . 115
A. The Floral Symmetry Gene Regulatory Network in
Antirrhinum Majus .................................. 115
B. CYC-like Genes are Implicated in the Control of Zygomorphy
in Diverse Lineages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
C. Genetic Mechanisms Underlying Changes in Floral Symmetry . . 121
D. Evolution of CYC-like Genes: Functional Implications . . . . . . . 122
E. Beyond CYC: Conservation and Divergence of Other
Components of the Floral Symmetry Network . . . . . . . . . . . . . 124
VII. Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
ABSTRACT
Symmetry is a defining feature of floral diversity. Here we review the evolutionary and
ecological context of floral symmetry (adding new data regarding its distribution), as
well as the underlying developmental and molecular bases. Two main types of symmetry
are recognized: radial symmetry or actinomorphy and bilateral symmetry or zygomor-
phy. The fossil record suggests that zygomorphy evolved in various lineages !50 MY
(million years) after the emergence of angiosperms, coinciding with the diversification of
specialized insect pollinators. Among extant angiosperms, zygomorphy is a highly
homoplastic trait, and is associated with species radiation thereby satisfying the defini-
tion of key innovation. The evolution of symmetry may be influenced by clade-specific
floral and inflorescence characteristics, possibly indicating different underlying con-
straints. Ecological studies suggest that zygomorphy may promote cross-fertilization
through increased precision in pollen placement on the pollinator’s body. The molecular
bases of flower symmetry are beginning to be unravelled in core eudicots, and available
evidence underlines the repeated recruitment of CYC2 genes, associated with frequent
gene duplications. Future prospects are discussed, emphasizing symmetry as a model
character for understanding the evolutionary bases of homoplastic floral traits.
I. INTRODUCTION
With more than 260,000 extant species, angiosperms represent about 90% of
terrestrial plant biodiversity. The flower, which is a synapomorphy of the
group, is a fascinating structure in many respects, having a well-conserved
ground plan but tremendous diversity in the size, colour, shape and number of
its parts. As a component of human environment it participates in shaping our
feeling of beauty.
Author's personal copy
87THE EVOLUTION OF FLORAL SYMMETRY
Symmetry is one of the major features taking part in this perception.
There are two principal types of floral symmetry, radial and bilateral
(Section II), the latter having evolved several times independently in angios-
perms (Section IV). Bilateral symmetry is therefore a homoplastic trait,
which poses fascinating questions concerning the homology of underlying
developmental and genetic processes, and the evolutionary forces at work in
the different occurrences. Indeed, such recurrent innovations provide
researchers with ideal models to address the question of the relative impor-
tance of historical contingency, physical and developmental constraints and
selection, in the course of organismal evolution.
As with many other architectural traits, the type of symmetry is often an
integral part of a species’ definition, even though more or less important
deviations from the characteristic type can be observed in natural popula-
tions (Section V). The first floral symmetry mutant was described by
Linnaeus, based on an atypical sample of Linaria vulgaris harvested by
Magnus Zio
¨berg in 1742 in Roslagen (Sweden). The flower was radially
symmetric with five nectar spurs, contrasting with the normal bilaterally
symmetric Linaria flower with just a single spur. Linnaeus called it
“Peloria,” after the Greek word for monster. He proposed that it arose
through fertilization of a normal Linaria by pollen from an alien species
(Linnaeus (1744), discussed in Gustafsson, 1979). Darwin was aware of
peloric forms in a number of species, and he remarked that many Labiateae
and Scrophulariaceae species are prone to such abnormal shapes.
He supposed that pelorism was due to an arrest of development or to
reversion. He made reciprocal crosses between peloric and normal snap-
dragon, and observed that none of the offsprings exhibited peloria; he
reported that “the crossed plants, which perfectly resembled the common
snapdragon, were allowed to sow themselves, and out of a hundred and
twenty-seven seedlings, eighty-eight proved to be common snapdragons,
two were in an intermediate condition between the peloric and normal state,
and thirty-seven were perfectly peloric, having reverted to the structure of
their one grand-parent” (1868). Darwin failed to interpret this segregation
(not significantly different from a Mendelian 3:1 segregation for one domi-
nant gene) and explained the results in the context of his pangenesis
hypothesis, which has now been totally dismissed. Hugo De Vries investi-
gated extensively peloric Linaria; he observed that the typical peloria repro-
duces five times the ventral part of the normal flower, and suspected that
repetitions of other parts could also occur. Indeed, he reported a rare
regular variant with a tubular corolla lacking spurs (cited in Gustafsson,
1979). Nowadays, several peloric mutants are commercialized as horticul-
tural varieties (e.g. in Antirrhinum, Sinningia and orchids).
Author's personal copy
88 H. CITERNE ET AL.
The elucidation of the genetic bases of peloria in Anthirrhinum and Linaria
came more than two centuries after the discovery of peloria, through the
work of E. Coen and R. Carpenter’s group (Section VI). Since these ground-
breaking results, several research groups have been investigating the genetic
origin of symmetry in a growing number of plant families, relying mostly on
a candidate gene approach. In recent years, results have been obtained that
point to a key role of the TCP gene family in independent occurrences of
bilateral symmetry. These advances have prompted several excellent reviews
in the last year (Busch and Zachgo, 2009; Hileman and Cubas, 2009; Jab-
bour et al., 2009a; Preston and Hileman, 2009; Rosin and Kramer, 2009). At
the dawn of a new era in evolutionary biology opened up by high-
throughput DNA sequencing technologies and functional genomics, it may
be of interest to examine what we know about floral symmetry, not only
from a genetic but also from evolutionary, developmental and ecological
points of view. This review explores these various fields in an attempt to
summarize existing knowledge and open new prospects for future research.
II. DEFINITIONS OF SYMMETRY
Symmetry is a geometrical concept that can be applied to either living
organisms or non-living objects. In biology, rotational and reflection sym-
metries are generally sufficient to describe the range of forms (Almeida and
Galego, 2005; Manuel, 2009). Rotational symmetry is defined as the rotation
of an object by an angle of 360˚/n (n>1) that does not change the object. In
reflection (flip or mirror) symmetry, an axis can be defined such that two
points on a perpendicular line to this axis are at equal distance from it; in
other words, this axis defines two mirror images. These two types of sym-
metry have formed the basis for the discrete categories used to describe
flower symmetry.
Floral symmetry is generally defined from an “en face” view at anthesis,
taking into consideration a planar projection of the flower, which justifies
the use of the term “axis” of symmetry. Very few studies of floral symmetry
integrate the three-dimensional structure of the flower (Leppik, 1972), where
it is more appropriate to talk about “planes” of symmetry. The classification
reflecting the three dimensions is complex and unwieldy, and simple defini-
tions of flower types are generally preferred. Nevertheless, taking into
account the three-dimensional structure may be important for understand-
ing the adaptive value of particular shapes in relation to interactions with
pollinators. Although in name floral symmetry refers to the entire structure
with all its constitutive parts (sepals, petals, androecium and gynoecium), the
Author's personal copy
89THE EVOLUTION OF FLORAL SYMMETRY
descriptions apply primarily to the perianth (particularly the corolla) and
sometimes to the androecium. The symmetry of the gynoecium is often
described independently from other floral organs. It is generally defined on
the basis of ovule placentation, that is, according to internal compartmenta-
lization since the carpels are often partially or totally fused (with fused ovary
walls, styles and stigma). Moreover, the gynoecium is often affected by a
reduction in carpel number compared to the merism of the perianth. It is
therefore frequently left out when characterizing floral symmetry.
Flowers appear predominantly symmetrical and rarely asymmetrical.
Among symmetrical flowers, two major types are classically recognized:
radial symmetry—also called polysymmetry or actinomorphy (from the
Greek word aktis a!"#&: sunray), and bilateral symmetry—also called
monosymmetry or zygomorphy (from the Greek word zugon $%&on: a
device joining two objects together). Actinomorphy is characterized by
both rotational and reflection symmetry. In actinomorphic flowers, all
organs of a same type (i.e. sepals, petals or stamens) are identical in
shape and size, and evenly distributed around the floral receptacle
(Fig. 1B–E). Zygomorphy has only reflection symmetry along a single
axis (Fig. 1G–J). The term zygomorphy was first introduced by the German
botanist Alexander Braun (1835). Zygomorphic flowers have also been
referred to as irregular (Sprengel, 1793), which is misleading in suggesting
an absence of symmetry, or as symmetrical (Mohl, 1837; Wydler, 1844),
which does not strictly differentiate between radial and bilateral symmetry.
Although inappropriate, these terms are still being used in the literature
(see, for instance, Coen et al., 1995; Luo et al., 1996). A rarer type of
symmetry is disymmetry, where two different orthogonal symmetry
axes can be distinguished (Fig. 1N). It occurs in a few clades of magnoliids
(Winteraceae (Ronse De Craene et al., 2003)), basal eudicots
(e.g. Fumarioideae (pers. obs.) and Eupteleaceae (Ren et al., 2007)), and
core eudicots (Oleaceae (Sehr and Weber, 2009), Brassicaceae (Ronse De
Craene et al., 2002; Rudall and Bateman, 2002), Begoniaceae (Rudall and
Bateman, 2002) and Balanophoraceae (Eberwein et al., 2009)).
For most zygomorphic flowers, the single symmetry axis is vertically
oriented, passing through the inflorescence apex (adaxial or dorsal side)
and the subtending bract (abaxial or ventral side). Consistently, bilaterally
symmetrical flowers are also described as dorsoventrally asymmetrical
(Carpenter and Coen, 1990; Coen, 1991). Cases of oblique zygomorphy
(where the symmetry axis deviates from the dorsoventral position)
and transverse zygomorphy (where the symmetry axis is horizontal) occur
in some families. Oblique zygomorphy is found in Sapindaceae and
Vochysiaceae (Eichler, 1878), Musaceae (Lane, 1955; Schumann, 1900;
Author's personal copy
(A) (B) (C) (D) (E)
(F) (G) (H) (I) (J)
(K) (L)
(M)
Flag flower Lip flower
(N) (O) (P) (Q) (R)
90 H. CITERNE ET AL.
Fig. 1. Different types of floral symmetry illustrated by examples from monocots
and eudicots. Symmetry types are represented in A (actinomorphy), F (zygomorphy),
M (disymmetry) and O (asymmetry). Red dotted lines: symmetry axes. B:
Hibiscus sp. (Malvaceae, eudicot), C: Aquilegia vulgaris (Ranunculaceae, eudicot), D:
Nigella damascena (Ranunculaceae, eudicot), E: Iris pseudacorus (Iridaceae, monocot),
G: Corydalis sp. (Papaveraceae s.l., eudicot), H: Orchis militaris (Orchidaceae, monocot),
I: Lobelia tupa (Campanulaceae, eudicot), J: Alstroemeria sp. (Alstroemeriaceae,
monocot), K: flag flower: Lathyrus sp. (Fabaceae, eudicot), L: lip flower: Lamium
galeobdolon (Lamiaceae, eudicot), N: Lamprocapnos spectabilis (Papaveraceae s.l.,
eudicot), P: Vinca minor (Apocynaceae, eudicot), Q: Tibouchina urvilleana
(Melastomataceae, eudicot), R: Strelitzia reginae (Strelitziaceae, monocot). Photographs:
F. Jabbour, except 1N: C. Damerval. (See Color Insert.)
Winkler, 1930), Marantaceae (Kunze, 1985), Solanaceae (Tucker, 1999),
Moringaceae, Bretschneideraceae (now included in Akaniaceae) (Ronse
De Craene et al., 1998, 2000, 2002) and Heliconiaceae (Kirchoff et al.,
2009). Transverse zygomorphy is found in Sabiaceae (Wanntorp and
Ronse De Craene, 2007) and Papaveraceae (Corydalis and Fumaria). In
the latter, however, rotation of the flower pedicel results in vertically
oriented flowers at maturity (Fig. 1G).
Author's personal copy
91THE EVOLUTION OF FLORAL SYMMETRY
The single symmetry axis of zygomorphic flowers can originate from an
unequal distribution of organs at maturity and/or the superimposition of
secondary identities on the basic sepal, petal or stamen identity. Examples
of unequal distribution of organs of a same identity can be found in
Asteridae, where the corolla is organized following a few conserved pat-
terns, the most common being 2|3 (two petals in the dorsal position|three
in the ventral position), 4|1 and 0|5 (Donoghue et al., 1998). The concept
of secondary identity translates morphological differentiation (including
micromorphological specificities) within a given organ type. For example,
in the species Antirrhinum majus (Veronicaceae, Asteridae) where the
corolla has an upper lip formed by two fused petals and a lower lip
formed by the three other petals (2|3 type), three petal identities
(dorsal - single petal—two petals, lateral—two petals and ventral—one
petal) are recorded (Corley et al., 2005; Luo et al., 1996). Similarly, in
Fabaceae, the standard (dorsal—single petal), wings (lateral—two petals)
and keel (ventral—two petals more or less fused) can be considered as
having three different petal identities. The combination of unequal distribu-
tion and secondary identities of petals makes the corolla of many zygo-
morphic flowers appear bilabiate, leading to the definition of two main
types of flowers, namely, lip (or gullet—Faegri and van der Pijl, 1966) and
flag (Endress, 1994). The distinction comes essentially from the placement of
sexual organs in the upper (lip type) or lower (flag type) part of the flower (see
Section V). Lip flowers are essentially found in Lamiales (Fig. 1L), Campa-
nulales, Zingiberales and Orchidales. Flag flowers are encountered in Faba-
ceae (Fig. 1K), in Polygalaceae and in Papaveraceae (Proctor et al., 1996). In
rare cases such as in tribe Delphinieae (Ranunculaceae), secondary identities
develop on spirally inserted organs (Jabbour et al., 2009b).
Symmetry is not constant within natural populations, and small devia-
tions can occur around a main type (see Section V). In addition, within the
discrete categories defined above, a quantitative element can be added to
classify flowers according to the degree of differentiation or deviation from
radial or bilateral symmetry they exhibit. For instance, flowers can be
described as almost actinomorphic, slightly zygomorphic or almost
zygomorphic (Endress, 1999). There are three main developmental causes
for such deviations. First, spiral phyllotaxis implies that organs sharing the
same identity are not inserted on a same plane, resulting in flowers that are,
strictly speaking asymmetric, even though they can appear actinomorphic
or zygomorphic. This is the case in most members of Ranunculaceae, for
instance, in tribe Delphinieae with “zygomorphic” flowers, and in Adonis
and Nigella that have “actinomorphic” flowers. Second, the curvature of
organs or groups of organs can result in a heterogeneous spatial
Author's personal copy
92 H. CITERNE ET AL.
distribution of organs and can also create imperfectly symmetrical flowers
(e.g. both the androecium and gynoecium are curved in Geranium
(zygomorphic; Geraniaceae), Solanum (actinomorphic; Solanaceae) and
Gladiolus (zygomorphic; Iridaceae). Finally, the degree of zygomorphy
can depend on the position of the flower along the inflorescence. In several
groups with actinomorphic flowers, the flower can become slightly zygo-
morphic due to the bending of floral organs when compressed laterally by
neighbouring flowers (Endress, 1999).
Very few species have asymmetric flowers (Fig. 1P–R). Asymmetric
flowers with chaotic organization occur in a few basal angiosperms, where
the innermost perianth organs and the stamens are irregularly arranged
from inception (e.g. in some Zygogynum species (Winteraceae), Endress,
1999). Asymmetry can also be the result of precise developmental processes
that are reproducible among members of the same species (e.g. in Fabaceae,
Lamiales, Orchidaceae and Zingiberales). In this case, asymmetry can be
found in all floral parts (e.g. in Vochysiaceae (Tucker, 1999)) or in just a
single organ type (e.g. in Senna (Caesalpinioideae), where asymmetry affects
only the gynoecium). It can be due to a reduction in organ number, such as
in Cannaceae and Valerianaceae (e.g. flowers in Canna and Centranthus
have a single lateral stamen). Another form of asymmetry is enantiomor-
phy, an asymmetry polymorphism resulting in flowers of two types that are
mirror images. It can be due to the formation of both left- and right-
contorted (sinistrorse or dextrorse) corollas (e.g. Wachendorfia (Haemodor-
aceae), Endress, 1999, 2001a; Helme and Linder, 1992; Senna (Fabaceae),
Marazzi and Endress, 2008; Banksia (Proteaceae), Renshaw and Burgin,
2008), or the deflection of the style to the left or to the right (enantiostyly;
see Graham and Barrett, 1995) such as in Wachendorfia paniculata (Hae-
modoraceae) (Endress, 2001a; Jesson and Barrett, 2002; Jesson et al., 2003;
Ornduff and Dulberger, 1978; Tucker, 1996, 1999) and Paraboea rufescens
(Gesneriaceae) (Gao et al., 2006). In most enantiostylous species, style
deflection is associated with a single pollinating anther opposite the style.
Monomorphic enantiostyly, in which individuals exhibit both flower
morphs (e.g. Solanum rostratum (Endress, 2006)) has been described in at
least 10 monocot and eudicot families, whereas dimorphic enantiostyly,
where the two morphs occur on separate plants, has been recorded only
in seven species belonging to three monocot families (reviewed in Jesson
and Barrett, 2002, 2003). Rarely, only one morph occurs within a species
(Endress, 1999) (e.g. Strobilanthinae (Acanthaceae); Moylan et al., 2004).
Examination of the developmental process leading to enantiostyly has
shown that it resulted from unequal cell division rates at the base of the
style (Douglas, 1997; Jesson et al., 2003).
Author's personal copy
93THE EVOLUTION OF FLORAL SYMMETRY
III. SYMMETRY AND FLOWER DEVELOPMENT
Studies of flower development have benefited from the development of
scanning electronic microscopy in the mid 20th century, and various authors
have made remarkable contributions to our understanding of flower devel-
opment and symmetry (e.g. Endress, 1999; Ronse De Craene, 2003; Tucker,
2003a). The developmental processes underlying the different types of floral
symmetry at anthesis appear highly diverse and provide information regard-
ing the evolution of floral diversity.
A. ESTABLISHMENT OF SYMMETRY AT VARIOUS STAGES DURING
DEVELOPMENT
In the first stages of development, phyllotaxis and direction of organ initiation
are crucial parameters influencing meristem symmetry (Dong et al., 2005;
Tucker, 2002, 2003b). There are two types of phyllotaxis, spiral and whorled.
In some taxa there is a combination of both spiral and whorled phyllotaxis,
with some organs inserted on a spiral and others on whorls (e.g. Aquilegia
(Ranunculaceae) where all organs are inserted on whorls, except sepals
(Tucker and Hodges, 2005)). In spiral phyllotaxis, organs are initiated one at
a time, with an equal time interval (plastochron) between organs of a same
type. In whorled phyllotaxis, initiation of organs of a same type can be
synchronous or unidirectional. Usually, and provided that growth is homo-
geneous, the first organs initiated are the largest at maturity (Goebel, 1905) but
there are numerous exceptions (e.g. papilionoid corollas and androecia). An
exhaustive list of taxa spanning all major angiosperm clades in which organo-
genesis follows a unidirectional order is given by Tucker (1999).
Zygomorphy can be observed before organ initiation, and persist through-
out development, or can appear later at various stages of development.
For instance, the floral meristem of A. majus has initially the form of a loaf
(oval, thus disymmetric), then becomes pentagonal and lastly zygomorphic
(Vincent and Coen, 2004). According to the authors, zygomorphy is estab-
lished in this species at the 15th plastochron among the 58 identified, that is,
after 9% of the floral developmental sequence, with the acquisition of dorsal
and ventral identities. Another instance of early establishment of zygomorphy
during development is Lotus japonicus (Fabaceae) (Feng et al., 2006), where
the initiation of floral organs is unidirectional (Dong et al., 2005).
Zygomorphy can also be established late in development. The developmental
processes underlying late-onset zygomorphy can include heterogeneous growth,
heterochrony (a temporal shift from the ancestral condition in a developmental
process (Douglas and Tucker, 1996; Rudall and Bateman, 2004; Tucker, 1999))
Author's personal copy
94 H. CITERNE ET AL.
and/or late elaboration of structures such as glands or spurs (Tucker, 1999). Late
zygomorphy appears to be frequent in taxa embedded in groups with predomi-
nantly actinomorphic flowers (Endress, 1999), such as Ranunculaceae (Jabbour
et al., 2009b). In the tribe Delphinieae (Ranunculaceae), it originates from
heterogeneous growth of petals and sepals after ontogenesis is completed, and
the elaboration of spurs on the two petals and single sepal on the adaxial side
(Jabbour et al., 2009b). In Iberis amara, which belongs to Brassicaceae, a family
with predominantly actinomorphic flowers, dorsal and ventral petals begin to
grow in a heterogeneous way only after the onset of stamen initiation, leading to
a zygomorphic mature flower (Busch and Zachgo, 2007).
Both early and late zygomorphy occur in the Asteridae. For instance, zygo-
morphy is apparent from the onset of organ initiation in the subfamily Oroban-
chaceae, but it is preceded by an actinomorphic stage during development in the
Plantaginaceae, Bignoniaceae and Lecythidaceae (Tucker, 1999).
B. IMPACT OF GROWTH AND ORGAN ELABORATION ON FLORAL SYMMETRY
Reduction, suppression and differential elaboration of organs determine
structural symmetry sensu Rudall and Bateman (2003), as opposed to zygo-
morphy caused or reinforced by differential petal colouration (Fig. 1J),
displacement or unequal organ expansion during development. Organ
abortion, which can result from totally suppressed or early arrested growth,
is a major determinant of zygomorphy. As a result of heterochrony, an
organ can become progressively aborted at an earlier stage until its total
suppression (e.g. Li and Johnston, 2000; Mitchell and Diggle, 2005). One or
several organ types can be affected. A large monocot group with mostly
zygomorphic flowers by organ reduction is Poales sensu lato (Kellogg, 2000;
Rudall and Bateman, 2004). The three grass lodicules are hypothesized to be
homologous to a single perianth whorl, based on morphological, develop-
mental and genetic evidence (see, for instance, Schmidt and Ambrose, 1998).
Since the dorsal lodicule is absent from most derived grasses (e.g. Hordeum,
Pooideae), the presence of only two ventral lodicules renders the grass
flowers structurally zygomorphic (Rudall and Bateman, 2004).
The female flowers of Stephania dielsiana (Menispermaceae) have a single
sepal, two petals and a single carpel, which makes them zygomorphic due to
organ reduction, compared to the trimerous actinomorphic male flowers
(Wang et al., 2006). In Sinningia cardinalis, A. majus and Rehmannia angulata
(all belonging to different families within Lamiales), the dorsal stamen is
reduced to a staminode and the degree of reduction increases from the former
to the latter, reinforcing the zygomorphic shape of the flower (Endress, 1998).
Strong morphological differentiation at the perianth level is often associated
Author's personal copy
95THE EVOLUTION OF FLORAL SYMMETRY
with alterations in the androecium, including stamen reduction (staminodes)
or even abortion (Rudall and Bateman, 2004). In Gesneriaceae, for instance,
the strength of corolla zygomorphy was found to be associated with alteration
in stamen number (Endress, 1997). In zygomorphic Proteaceae, a bilabiate
perianth is associated with ventral (e.g. Placospermum) or dorsal
(e.g. Synaphea) staminodes (Douglas, 1997; Douglas and Tucker, 1996).
Differential organ elaboration contributes to bilateral symmetry at matur-
ity. It includes fusion, curvature (see Section II) and the formation of glands or
spurs. A well-known example of differential fusion of organs of a same identity
is found in the bilabiate corolla of A. majus, but also in the ligulate flowers of
Asteraceae (which can have two reduced and three large fused petals (2|3),
three fused petals only (Asteroideae), five fused petals (Cichorioideae) or one
reduced and four large fused petals (e.g. Barnadesia) (Ronse De Craene,
2010)). Another example is found in Proteaceae where tepals are fused post-
genitally and their partitioning is either equal, resulting in an actinomorphic
flower, or unequal, rendering the flower zygomorphic (e.g. Lomatia).
Spurs are floral appendages that appear late during development. Their
origin is highly diverse, developing on sepals (e.g. Impatiens (Balsaminaceae)),
petals (e.g. Viola (Violaceae)), receptacular hypanthia (e.g. Tropaeolum (Tro-
paeolaceae)), stamen–petal tubes (e.g. Diascia (Scrophulariaceae)) or at the base
of the ovary (e.g. Pelargonium). The formation of spurs can affect the symmetry
of a flower. When the number of spurs is equal to the merism of the flower
(e.g. Epimedium (Berberidaceae), Aquilegia (Ranunculaceae) and Halenia
(Gentianaceae)), the flower is actinomorphic. Flowers with a single spur
(e.g. Corydalis), or a pair of spurs (e.g. Diascia, Delphinium, Dicentra), are
zygomorphic or disymmetric. The presence of a single spur can also determine
the orientation of the symmetry axis. The development of a spur in species of
Tropaeolum changes the symmetry from oblique to median zygomorphy (Ronse
De Craene and Smets, 2001). It has been shown that in Asteridae the evolution
of floral symmetry is tightly correlated with that of spurs, and that zygomorphy
is a prerequisite for the evolution of single or pairedspurs (Jabbour et al., 2008).
C. DEVELOPMENTAL TRAJECTORIES AND FLOWER SYMMETRY
Following organ initiation, the major determinants of floral symmetry are organ
growth, differentiation and distribution of mature organs. The symmetry of
mature flowers can be largely independent of phyllotaxis and organ initiation,
and flowers with either whorled (with or without unidirectional initiation) or
spiral phyllotaxis can appear actinomorphic or zygomorphic.
Figure 2 proposes theoretical examples combining three developmental
processes taking part in the establishment of flower symmetry at anthesis,
Author's personal copy
Developmental processes Symmetry of adult flower
Initiation I Growth G Differential elaboration D
Homogeneous Iα
Homogeneous Gα
Heterogeneous Gβ
Absent Dα Iα | Gα | Dα
Actinomorphy without change
of symmetry during
development
Present Dβ Iα | Gα | Dβ
Late zygomorphy
Absent Dα Iα | Gβ | Dα
Zygomorphy with a change
of symmetry during
development
Present Dβ Iα | Gβ | Dβ
Zygomorphy with a change
of symmetry during
development
Heterogeneous Iβ
Homogeneous Gα
Heterogeneous Gβ*
Absent Dα Iβ | Gα | Dα
Early zygomorphy
Present Dβ Iβ | Gα | Dβ
Early zygomorphy
Absent Dα Iβ | Gβ* | Dα
Actinomorphy with a change
of symmetry during
development
Present Dβ Iβ | Gβ | Dβ
Zygomorphy with changes
of symmetry during
development
96 H. CITERNE ET AL.
Fig. 2. Theoretical developmental trajectories combining different states for three
processes (organ initiation, growth and differential elaboration) resulting in different types
of floral symmetry. Two states are considered for each process: synchronous (Ia) or
asynchronous (Ib) initiation, homogeneous (Ga) or heterogeneous (Gb) growth, and
absence (Da) or presence (Db) of differential elaboration. Combining the two states for
the three developmental processes results in eight theoretical outcomes. For example, the
Ib | Gb | Da trajectory has an actinomorphic outcome because the heterogeneous growth
compensates for the unidirectional initiation of organs (indicated by Gb*). Black circle:
floral meristem. Black/gray disk: organ primordium. Black star: elaborated organ. The
colour of disks is lighter for organs initiated later. The size of disks is proportional to the
primordium growth rate. Stars of different shapes represent differentiated organs. Red
line: the single axis of symmetry in zygomorphic stages. (See Color Insert.)
Author's personal copy
97THE EVOLUTION OF FLORAL SYMMETRY
namely, initiation, growth and differential elaboration of organs of a same
type. Two states are considered here for each process: synchronous (Ia) or
asynchronous (Ib) initiation, homogeneous (Ga) or heterogeneous (Gb)
growth and absence (Da) or presence (Db) of differential elaboration. The
combination of these three processes results in eight developmental trajec-
tories, producing either zygomorphic or actinomorphic flowers at maturity
(Fig. 2). Although this representation oversimplifies complex developmental
processes, it serves to illustrate how similar states at maturity may result
from different developmental pathways, suggesting that the underlying
molecular agents controlling floral symmetry may also be different. Actino-
morphic flowers can originate from disymmetric or zygomorphic develop-
mental stages (Endress, 1994; Ronse De Craene and Smets, 1994). Tsou and
Mori (2007) report cases where symmetry changes more than once during
flower development, such as in Cariniana micrantha (Lecythidaceae) where
flowers are successively zygomorphic (sepals initiate asynchronously, Ib),
then almost actinomorphic (when sepals are initiated, petals initiate and
grow synchronously, Gb), and finally zygomorphic (a hood is derived from
the abaxial rim of the ring meristem, Db) (Endress, 1994; Tsou and Mori,
2007). A similar situation is found in the genus Couroupita (Lecythidaceae)
in which the upper half of the developing flower is initially retarded at first,
resulting in an early zygomorphic stage. The flower becomes actinomorphic
when stamens and carpels initiate and then zygomorphic again when the
androecium proliferates and forms a tongue-like structure with sterile sta-
mens (Endress, 1999, Tsou and Mori, 2007).
Floral zygomorphy thus relies on complex and potentially numerous
developmental trajectories, and this relates to the highly homoplastic nature
of this trait in adult flowers. A detailed knowledge of symmetry changes
during development is important for (1) understanding symmetry transitions
among related species, (2) understanding the repeated establishment of
bilateral symmetry across angiosperms and (3) interpreting genetic data
underlying these morphological changes.
IV. EVOLUTION OF FLOWER SYMMETRY
A. DISTRIBUTION OF SYMMETRY AMONG EXTANT ANGIOSPERMS
Zygomorphy has always been considered a derived trait in angiosperms
compared to actinomorphy. Studies that have attempted to infer the ancestral
features of the first angiosperms (e.g. Doyle and Endress, 2000; Endress and
Doyle, 2009) conclude that the first angiosperms had actinomorphic flowers.
Author's personal copy
98 H. CITERNE ET AL.
The exact number of transitions toward zygomorphy throughout all angios-
perms is unknown. The estimated numbers given in papers that deal with the
evolution of zygomorphy vary according to the paper (e.g. more than 25 in
Cubas, 2004, at least 38 in Zhang et al., 2010). However, it generally reflects
the number of families in which zygomorphy is found, but not the actual
number of transitions from actinomorphy to zygomorphy. Indeed, such
transitions can potentially happen several times within a family. The exact
number of transitions can only be obtained through the detailed reconstruc-
tion of the evolution of the character “floral symmetry” (i.e. character
optimization) on a robust and well-resolved phylogenetic tree of angios-
perms. Variation in the number of families displaying zygomorphy may
be due to changes in the classification of angiosperms. We conducted a
detailed phylogenetic study of the evolution of zygomorphy in angiosperms
using updated phylogenies based on the latest classification (APG3: The
Angiosperm Phylogeny Group, 2009 and http://www.mobot.org/MOBOT/
research/APweb). Our results indicate that zygomorphy evolved only once in
“basal angiosperms” (a paraphyletic assemblage consisting of all angiosperm
taxa that have diverged before the divergence of monocots and eudicots), at
least 23 times independently in monocots (see Section IV.C for more detail)
and at least 46 times independently in eudicots (see Figs. 4 and 5). The number
of independent transitions from actinomorphy to zygomorphy is therefore
much higher (at least 70, almost twice the highest number given in the
literature) than all previously estimated numbers.
Many speciose taxa present strongly zygomorphic flowers (like, for
instance, Faboideae, Orchidaceae, Poaceae or the order Lamiales), which
is consistent with the hypothesis that zygomorphy could play a positive role
in speciation rates. This was rigorously tested using a phylogenetic frame-
work comparing species richness in sister clades differing in their floral
symmetry (Sargent, 2004). In 15 out of 19 sister pairs identified, the lineage
with zygomorphic flowers is significantly more diverse than its sister group
with actinomorphic flowers, which gives strong support to the hypothesis
that zygomorphy is a key innovation.
B. EMERGENCE OF ZYGOMORPHY DURING ANGIOSPERM EVOLUTION IN
RELATION TO INSECT DIVERSIFICATION
The first known angiosperm remains are pollen grains dated to the Hauter-
ivian (130–136 Ma, million years ago) (Fig. 3; Feild and Arens, 2007; Friis
et al., 2006; Frohlich, 2006). The first fossil of a whorled pentamerous flower
with both petals and sepals, considered as a eudicot representative, is
recorded in the Cennomanian (Basinger and Dilcher, 1984), while fossil
Author's personal copy
99THE EVOLUTION OF FLORAL SYMMETRY
Myr
JURASSIC
Late
Early
CRETACEOUS PALEOGENE
71
93
100
112
130
140
125
89
83
65
145
56
34
23
Valanginian
Hauterivian
Barremian
Berriasian
Albian
Aptian
Cennomanian
Turonian
Santonian
Coniacian
Campanian
Maastrichtian
Paleocene
Eocene
Oligocene
First bee fossil: Melittosphex burmensis
Fossils of monoaperturate (black) and triaperturate
(white) pollen grains
Fossils of flower; black: first remains related to
Nympheales; white: first cyclic eudicot flower; light gray:
fossils ancestral to zygomorphic flowers; dark gray:
zygomorphic flower
Fig. 3. Timescale showing the first appearance of important floral features during
angiosperm evolution, based on the fossil record. The vertical black bars indicate two
major diversification periods, which coincide with the appearance of new floral traits
(from Crepet, 2008; Crepet and Niklas, 2009; Dilcher, 2000; Friis et al., 2001, 2006,
2010; Poinar and Danforth, 2006).
flowers with spirally inserted floral parts are dated to the Barremian–Aptian
(Crepet, 2008). Transition to a whorled organization of the flower possibly
opened the way for further floral innovations, which appear especially
numerous during the Turonian geological stage, coinciding with a period
of radiation leading to angiosperm dominance in some floras of the mid-
Cretaceous (Crepet, 2008; Crepet and Niklas, 2009; Friis et al., 2010).
Bilateral symmetry is thought to have first evolved during this first angios-
perm radiation, based on Turonian fossils with asymmetric flowers with
staminodal nectaries that could be considered “precursors” of zygomorphic
flowers, as suggested by their resemblance to the flowers of extant taxa
adapted to specialist pollinators (Crepet, 1996, 2008). Remains of clearly
zygomorphic flowers, as well as brush flowers (with numerous long
stamens), are recorded in Paleocene–Eocene deposits (Fig. 3; Crepet and
Niklas, 2009; Dilcher, 2000).
Author's personal copy
100 H. CITERNE ET AL.
The coevolution of plants and insects has been considered for a long time
to be the primary cause of radiation of plants and of some insect groups
(correspondence between Saporta and Darwin (1877) cited in Friedman,
2009; Grant, 1949; Grimaldi, 1999; but see Waser, 1998; Gorelick, 2001).
Reconstructing the evolution of pollination modes on the phylogeny of
extant basal angiosperms clearly indicates that insect pollination is the
ancestral state (Hu et al., 2008). Early flowering plants may have been
pollinated by a wide diversity of insects such as beetles, primitive moths,
various flies and possibly sphecid wasps ancestral to bees (Bernhardt, 2000;
Grimaldi, 1999; Hu et al., 2008). Extant bees, that comprise many extant
pollinators of zygomorphic flowers, constitute a derived natural group of
spheciform wasps (vegetarian wasps) that almost certainly originated in the
Mid to Late Cretaceous (Grimaldi, 1999; Poinar and Danforth, 2006).
Corbiculate bees (honeybees, bumblebees, orchid bees and stingless bees)
extensively diversified in the Early Tertiary (Grimaldi and Engel, 2005).
Interestingly, the emergence of floral innovations and derived pollinators
co-occurs with the angiosperm radiations of the Turonian (89–93.5 Ma) and
the Upper Paleocene Lower Eocene periods (Crepet and Niklas, 2009). In
addition, a significant correlation was observed between angiosperm species
number and insect family number during Cretaceous–Tertiary geological
stages. Even though correlations cannot be considered to necessarily reflect
causative influence of one group on the other, it may indicate reciprocal
driving mechanisms for diversification (Crepet, 1996). The fossil records
thus indicate that zygomorphy evolved in several plant lineages during the
same period as the rise of some bee families, supporting the hypotheses
of coevolution with these insects as the triggering mechanism for floral
symmetry evolution (e.g. Neal et al., 1998).
C. ARCHITECTURE OF FLOWERS AND INFLORESCENCES—WHAT IS THEIR
IMPACT ON FLORAL SYMMETRY
Perianth symmetry is only one of the numerous floral features that can
present variation. Because bilateral symmetry affects the shape of the
meristem sometimes from the earliest stages of floral development, the
issue of how changes in floral symmetry may have been constrained or
canalysed by other features of the flower or the inflorescence architecture
during the course of evolution can be raised. When flowers are grouped in
inflorescences, they become necessarily constrained by neighbouring flow-
ers during their development, which may potentially affect flower shape at
adult stage. Intrinsic features of flowers such as the number of organ
primordia could also be prone to have such an effect, by exerting sterical
Author's personal copy
101THE EVOLUTION OF FLORAL SYMMETRY
constraints on flower shape. In the following paragraphs, we examine the
relationship between floral symmetry and selected features of flowers and
inflorescences.
1. Flower Symmetry and Inflorescences
It has been suggested that the symmetry of flowers is somewhat linked to the
way they are organized in inflorescences (Coen and Nugent, 1994; Rudall
and Bateman, 2010). Inflorescence architecture among angiosperms is very
diverse, which has led to a complex and sometimes ambiguous typology
(Prenner et al., 2009 and references therein). Two basic types can be distin-
guished based on the fate of the terminal meristem. In cymose inflorescences,
the terminal meristem forms a flower, and inflorescence growth results from
the development of one or more lateral axes, which in turn reiterate this
pattern (sympodial growth). All axes terminate in a flower. In racemose
inflorescences, the terminal meristem promotes inflorescence growth by
producing lateral meristems that will produce either flowers or secondary
axes reiterating the main axis pattern (monopodial growth). Terminal
meristems do not produce flowers but eventually become exhausted. Inflor-
escences can be simple or compound, associating diversely cymose and/or
racemose modules (Prenner et al., 2009).
Inflorescence axes are observed in fossil records as early as flowers, but
their interpretation is very difficult. The particular architecture of the repro-
ductive unit of Archaefructus (Barremian–Aptian), now considered to be
related to Nympheales, has been interpreted either as a multipartite naked
flower with an elongated axis (Sun et al., 2002) or as an ebracteate racemose
inflorescence bearing simple unisexual and naked flowers (Friis et al., 2003).
Several spike-like or even compound inflorescences from the mid-
Cretaceous, densely covered with small flowers, have been found (Friis
et al., 2006). Parkin (1914) suggested that the primitive inflorescence type
is determinate, meaning in its simplest expression a solitary flower (discussed
in Rudall and Bateman, 2010). Morphological analyses of extant “basal”
taxa and fossil records in a phylogenetic framework suggest grouping of
flowers in inflorescence rather than solitary as the ancestral state, but the
ancestral state for inflorescence remains equivocal (Endress and Doyle,
2009). This result apparently comes from the authors’ interpretation of
Archaefructus and the inflorescence of Nympheales as racemose, which is a
matter of debate (Rudall and Bateman, 2010).
Classically, it is stated in the literature that radially symmetric flowers are
found in both racemose and cymose inflorescences whereas zygomorphic
flowers preferentially occur in racemose inflorescences (Coen and Nugent,
Author's personal copy
102 H. CITERNE ET AL.
1994; Dahlgren et al., 1985). Indeed, the meristems of grouped flowers are
embedded in an asymmetric morphogenetic field defined by the flower
subtending bract toward the ventral side and the terminal inflorescence
meristem toward the dorsal side (Coen and Nugent, 1994). The existence
of different cellular or physiological contexts for terminal and lateral mer-
istems can be illustrated by terminal peloria that occurs in species that
normally produce zygomorphic flowers grouped into racemose inflores-
cences. In the centroradialis mutant of A. majus, the inflorescence meristem
shifts to a floral identity, and the resulting terminal flower is radially sym-
metric, very similar to lateral ones in the cycloidea mutant (Clark and Coen,
2002). Terminal peloria in eudicots have also been reported in species
belonging to the Lamiales, Ranunculaceae (Rudall and Bateman, 2004)
and Fumarioideae (Cysticapnos vesicarius, pers. obs.). Morphogenetic gra-
dients may also account for the different symmetry of central and marginal
flowers in derived “flower-like” inflorescences, such as the radiate capitula in
Asteraceae, the corymb of I. amara or the umbels in some Apiaceae (e.g. in
Daucus carota, pers. obs.). It can be speculated that a prerequisite for the
evolution of zygomorphy is the emergence of asymmetric morphogenetic
fields in an inflorescence.
We examined the relationship between floral symmetry and inflorescence
growth pattern (monopodial versus sympodial) by conducting a detailed
comparative study of the evolution of both characters in monocots, taking
into account the most recent phylogenetic advances in this large clade.
Figure 4 presents two mirror phylogenetic trees of the monocots on which
flower symmetry (left-hand tree) and inflorescence type (right-hand tree)
have been optimized using Maximum Parsimony. It shows that zygomorphy
evolved at least 23 times independently from actinomorphy throughout
monocots, and not only in the context of a racemose (indeterminate) inflor-
escence. Zygomorphy evolved together with single flowers in various
families, for example, in Arachnites uniflora (Corsiaceae), in Thismia
americana (Thismiaceae), in Tecophilaea cyanocrocus (Tecophilaeaceae), in
Paphiopedilum appletonianum (Orchidaceae) and in Gethyllis atropurpureum
(Amaryllidaceae) and it is found in association with cymose (at least
the terminal units) inflorescences in several families of Zingiberales
(in Musaceae, Heliconiaceae and Strelitziaceae), in Commelinales (in
Haemodoraceae and Commelinaceae), in Liliales (in Alstroemeriaceae)
and in Asparagales (in Doryanthaceae and Amaryllidaceae). Flowers in
some of these taxa may be quite strongly zygomorphic, like in Zingiberales
or Gilliesia (Amaryllidaceae) for example (with organ reduction and synor-
ganization), indicating that zygomorphy is not necessarily precluded by the
sympodial growth of cymose inflorescences. In other words, in monocots,
Author's personal copy
Type of symmetry Type of inflorescence
Racemose
Actinomorphy
At least terminal units cymose
Zygomorphy
Panicle
Asymmetry
Single flowers
No perianth
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
A
103THE EVOLUTION OF FLORAL SYMMETRY
Fig. 4. (A) Mirror trees of monocots showing the evolution of perianth symmetry and
inflorescence type, optimized using Maximum Parsimony. Left tree: optimization of
perianth symmetry. Several colours on the same branch denote ambiguity in the ancestral
state. Right tree: optimization of inflorescence type. Several colours on the same branch
denote ambiguity in the ancestral state. Coloured lines refer to the orders of monocots.
Asterisks indicate taxa that produce flowers possessing more than six stamens. The topology
of the tree was established using information from the Angiosperm Phylogeny website
(http://www.mobot.org/MOBOT/research/APweb/) and detailed phylogenies obtained
from the literature when necessary. Species represented in this tree were selected according
the following criteria: (1) all monocot families are represented by at least one species, and (2)
families in which there is variation for at least one of the characters examined are represented
by two or more species. Botanical descriptions were mostly obtained from Dahlgren et al.
(1985). (B) From left to right and top to bottom: names of the terminal taxa (species)
included in the tree. (See Color Insert.)
Author's personal copy
Costus speciosus (Costaceae)
Hedychium coronarium (Zingiberaceae)
Canna glauca (Cannaceae)
Maranthochloa cuspidata (Maranthaceae)
Heliconia magnifica (Heliconiaceae)
Orchidenta maxillarioides (Lowiaceae)
Ravenala madagascariensis (Strelitziaceae)
Phenakospermum guinanense (Strelitziaceae)
Strelitzia reginae (Strelitziaceae)
Musa acuminata (Musaceae)
Haemodorum corymbosum (Haemodoraceae)
Anigozanthos flavidus (Haemodoraceae)
Wachendorfia paniculata (Haemodoraceae)
Heteranthera callifolia (Pontederiaceae)
Pontederia lanceolata (Pontederiaceae)
Philydrum lanuginosum (Phylidraceae)
Hanguana malayana (Hanguanaceae)
Tradescantia sillamontana (Commelinaceae)
Commelina forskalaei (Commelinaceae)
Dasypogon bromeliifolius (Dasypogonaceae)
Poa trivialis (Poaceae)
Oryza sativa (Poaceae)
Ochlandra stridula (Poaceae)
Arundinaria gigantea (Poaceae)
Imperata cylindrica (Poaceae)
Joinvillea plicata (Joinvilleaceae)
Ecdeicolea monostachya (Ecdeiocoleaceae)
Flagellaria guineensis (Flagellariaceae)
Dapsilanthus disjunctus (Restionaceae)
Centrolepis fascicularis (Centrolepidaceae)
Anarthria prolifera (Anarthriaceae)
Lipocarpha occidentalis
Evandra aristata (Cyperaceae)
Scirpus californicus (Cyperaceae)
Carex praeclara (Cyperaceae)
Distichia sp. (Juncaceae)
Juncus castaneus (Juncaceae)
Thurnia sphaerocephala (Thurniaceae)
Mayaca fluviatilis (Mayacaceae)
Eriocaulon taishanense (Eriocaulaceae)
Eriocaulon decangulare (Eriocaulaceae)
Abolboda linearifolia (Eriocaulaceae)
Orectanthe sceptrum (Xyridaceae)
Xyris lacerata (Xyridaceae)
Rapatea paludosa (Rapateaceae)
Pitcairnia xanthocalyx (Bromeliaceae)
Dyckia remotifolia (Bromeliaceae)
Billbergia nutans (Bromeliaceae)
Typha latifolia (Typhaceae)
Sparganium erectum (Sparganiaceae)
Retispatha dumetosa (Arecaceae)
Nypa fruticans (Arecaceae)
Caryota mitis (Arecaceae)
Phoenix dactylifera (Arecaceae)
Phytelephas macrocarpa (Arecaceae)
Phytelephas aequatorialis (Arecaceae)
Synechantus warscewiczianus (Arecaceae)
Cocos nucifera (Arecaceae)
Howea balmoreana (Arecaceae)
Dypsis lutescens (Arecaceae)
Dypsis lantzeana (Arecaceae)
Dypsis mirabilis (Arecaceae)
Neuwiedia inae (Orchidaceae)
Apostasia odorata (Orchidaceae)
Paphiopedilum appletonianum (Orchidaceae)
Vanilla planifolia (Orchidaceae)
Ophrys insectifera (Orchidaceae)
Eulophia andamanensis (Orchidaceae)
Aspidistra dodecandra (Asparagaceae)
Asparagus officinalis (Asparagaceae)
Yucca baccata (Asparagaceae)
Hosta japonica (Asparagaceae)
Aphyllanthes monspeliensis (Asparagaceae)
Sowerbaea juncea (Asparagaceae)
Lomandra insularis (Asparagaceae)
Trichlora lactea (Amaryllidaceae)
Miersia chilensis (Amaryllidaceae)
Leucocoryne purpurea (Amaryllidaceae)
Gillesia graminea (Amaryllidaceae)
Solaria miersiodes (Amaryllidaceae)
Allium vineale (Amaryllidaceae)
Gethyllis atropurpureum (Amaryllidaceae)
Gethyllis ciliaris (Amaryllidaceae)
Sprekelia formosissima (Amaryllidaceae)
Habranthus robustus (Amaryllidaceae)
Lycoris aurea (Amaryllidaceae)
Galanthus nivalis (Amaryllidaceae)
Asphodelus aestivus (Xanthorrhoeaceae)
Haworthia integra (Xanthorrhoeaceae)
Hemerocallis fulva (Xanthorrhoeaceae)
Simethis planifolia (Xanthorrhoeaceae)
Phormium cookianum (Xanthorrhoeaceae)
Arnocrinum gracillimum (Xanthorrhoeaceae)
Xanthorrhoea preissii (Xanthorrhoeaceae)
Xeronema callistemon (Xeronemataceae)
Moraea aristata (Iridaceae)
Isophysis tasmanica (Iridaceae)
Iris germanica (Iridaceae)
Geosiris aphylla (Iridaceae)
Aristea biflora (Iridaceae)
Gladiolus segetum (Iridaceae)
Crocosmia masoniorum (Iridaceae)
Crocosmia paniculata (Iridaceae)
Freesia laxa (Iridaceae)
Crocus angustifolius (Iridaceae)
Romulea citrina (Iridaceae)
Doryanthes palmeri (Doryanthaceae)
Doryanthes ensifolia (Doryanthaceae)
Tecophilaea cyanocrocus (Tecophilaeaceae)
Zephyra elegans (Tecophilaeaceae)
Conanthera bifolia (Tecophilaeaceae)
Cyanella lutea (Tecophilaeaceae)
Cyanella hyacinthoides (Tecophilaeaceae)
Cyanastrum johnstonii (Tecophilaeaceae)
Ixiolirion montanum (Ixioliriaceae)
Borya spetentrionalis (Boryaceae)
Astelia pumila (Asteliaceae)
Lanaria plumosa (Lanariaceae)
Pauridia longituba (Hypoxidaceae)
Curculigo latifolia (Hypoxidaceae)
Curculigo racemosa (Hypoxidaceae)
Hypoxis decumbens (Hypoxidaceae)
Blandfordia grandiflora (Blandfordiaceae)
Calochortus nuttallii (Calochortaceae)
Corsia unguiculata (Corsiaceae)
Arachnites uniflora (Corsiaceae)
Smilax aspera (Smilacaceae)
Heterosmilax japonica (Smilacaceae)
Heterosmilax longiflora (Smilacaceae)
Heterosmilax seisuiensis (Smilacaceae)
Gagea lutea (Liliaceae)
Ripogonum scandens (Ripogonaceae)
Philesia magellanica (Philesiaceae)
Paris quadrifolia (Melanthiaceae)
Chamaelirium luteum (Melanthiaceae)
Chionographis chinensis (Melanthiaceae)
Veratrum album (Melanthiaceae)
Campynema lineare (Campynemataceae)
Colchicum automnale (Colchicaceae)
Petermannia cirrosa (Petermanniaceae)
Luzuriagaria radicans (Luzuriagaceae)
Bomarea pardina (Alstroemeriaceae)
Alstroemeria aurantiaca (Alstroemeriaceae)
Asplundia multistaminata (Cyclanthaceae)
Pandanus candelabrum (Pandanaceae)
Croomia pauciflora (Stemonaceae)
Pentastemona egregia (Stemonaceae)
Triuris hyalina (Triuridaceae)
Barbacenia purpurea (Velloziaceae)
Vellozia prolifera (Velloziaceae)
Dioscorea communis (Dioscoreaceae)
Dioscorea melanophyma (Dioscoreaceae)
Dioscorea convolvulacea (Dioscoreaceae)
Trichopus zeylanicus (Dioscoreaceae)
Stenomeris cumingiana (Dioscoreaceae)
Burmannia madagascariensis (Burmanniaceae)
Thismia americana (Thismiaceae)
Afrothismia pachyantha (Thismiaceae)
Oxygyne triandra (Thismiaceae)
Narthecium ossifragum (Nartheciaceae)
Japonolirion osense (Petrosaviaceae)
Petrosavia stellaris (Petrosaviaceae)
Potamogeton pectinatus (Potamogetonaceae)
Althenia filiformis (Potamogetonaceae)
Zannichellia palustris (Potamogetonaceae)
Zostera marina (Zosteraceae)
Posidonia oceanica (Posidoniaceae)
Cymodocea nodosa (Cymodoceaceae)
Ruppia spiralis (Ruppiaceae)
Maundia triglochinoides (Juncaginaceae)
Triglochin maritimum (Juncaginaceae)
Lilaea scilloides (Juncaginaceae)
Aponogeton hexatepalus (Apotonogetonaceae)
Aponogeton proliferus (Apotonogetonaceae)
Aponogeton madagascariensus (Apotonogetonaceae)
Aponogeton distachyos (Apotonogetonaceae)
Scheuchzeria palustris (Scheuchzeriaceae)
Sagittaria platyphylla (Alismataceae)
Wiesneria triandra (Alismataceae)
Hydrocleys nymphoides (Limnocharitaceae)
Limnocharis flava (Limnocharitaceae)
Butomopsis latifolia (Limnocharitaceae)
Halophila ovalis (Hydrocharitaceae)
Thalassia testudinum (Hydrocharitaceae)
Vallisneria americana (Hydrocharitaceae)
Hydrilla verticillata (Hydrocharitaceae)
Najas marina (Hydrocharitaceae)
Egeria densa (Hydrocharitaceae)
Elodea nuttallii (Hydrocharitaceae)
Stratiotes aloides (Hydrocharitaceae)
Hydrocharis morsus-ranae (Hydrocharitaceae)
Limnobium spongia (Hydrocharitaceae)
Butomus umbellatus (Butomaceae)
Pleea tenuifolia (Tofieldiaceae)
Tofieldia pusilla (Tofieldiaceae)
Lemna minor (Araceae)
Cryptocoryne crispatulata (Araceae)
Pistia stratiotes (Araceae)
Pothos chinensis (Araceae)
Anthurium ramoncaracasii (Araceae)
Acorus calamus (Acoraceae)
Zingiberales
Commelinales
Dasypogonaceae
Poaceae
Arecaceae
Hosta japonica (Asparagaceae)
Liliales
Pandanales
Dioscoreaceae
Petrosaviales
Alismatales
Acorales
B
104
Fig. 4. (Continued)
H. CITERNE ET AL.
Author's personal copy
105THE EVOLUTION OF FLORAL SYMMETRY
bilateral symmetry does not occur exclusively in flowers produced by lateral
meristems. In eudicots, zygomorphic flowers are generally assembled in
racemose inflorescences. A rare exception is the cymose inflorescence of
the zygomorphic Capnoides sempervirens (Fumarioideae), even though
zygomorphy is more fluctuant in the terminal flower than in lateral flowers
(pers. obs.).
2. Floral Constraints on the Evolution of Symmetry
A previous study exploring the relationships between floral symmetry,
merism, number of stamens and presence of spurs in Ranunculales, the
earliest-diverging order in the eudicots, showed that zygomorphy evolved
three times independently and in very different architectural contexts in
this group (Damerval and Nadot, 2007). Another study conducted in the
large Asterid clade, where numerous transitions toward zygomorphy
have occurred (sometimes followed by reversals to actinomorphy) has
shown that zygomorphy is almost never associated with polyandry (i.e. a
number of stamens higher than twice the merism) in this derived eudicot
clade (Jabbour et al., 2008). This study highlights the fact that floral
symmetry may not evolve completely independently from other floral
features. In particular, it suggests that an increase in stamen number
could impede the dorsoventralization of the flower. A similar situation
was found in monocots (Fig. 4) in which only one co-occurrence of
polyandry (defined here as more than six stamens, six being twice the
most widespread type of merism in monocots) and zygomorphy is
observed, within the genus Aponogeton from the basal order Alismatales.
In Rosids, however, several co-occurrences of zygomorphy and polyan-
dry are recorded (Fig. 5). Among the 11 (at least) transitions toward
zygomorphy and the more than 25 transitions toward polyandry (defined
as over twice the merism) recorded in the phylogenetic tree of rosid
families, co-occurrences of both character states are observed five
times. They are found in Emblingiaceae (which produce dimerous flow-
ers with eight or nine stamens), in Begoniaceae, which have dimerous
disymmetric rather than truly zygomorphic flowers, in Resedaceae,
Cleomaceae (in which however, most zygomorphic genera have flowers
with few stamens) and in Chrysobalanaceae. Truly zygomorphic flowers
with numerous stamens are found only in Resedaceae and Chrysobala-
naceae, suggesting that the establishment of zygomorphy might be con-
strained in a polyandrous context, like in the Asterids. Furthermore, like
in the Asterids the presence of spurs (here a single spur) is conditioned
to zygomorphy (Fig. 5). The main difference lies in the fact that in
Author's personal copy
Type of perianth symmetry Number of stamens
Polysymmetry Twice merism or less
Monosymmetry More than twice merism
No perianth (polyandry)
Variable
*
*
*
*
*
*
*
*
*
A
106 H. CITERNE ET AL.
Fig. 5. (A) Mirror trees of Rosids showing the evolution of perianth symmetry and the
state of the androecium (number of stamens) relatively to the merism, optimized using
Maximum Parsimony. Left tree: optimization of perianth symmetry. Several colours on the
same branch denote ambiguity in the ancestral state. Right tree: optimization of the state of the
androecium (number of stamens). Several colours on the same branch denote ambiguity in the
ancestral state. Coloured lines refer to the orders of Rosids. Asterisks indicate taxa that produce
spurred flowers. The topology of the tree was established using information from the
Angiosperm Phylogeny website (http://www.mobot.org/MOBOT/research/APweb/). All
families are included and represent the terminal taxa of the tree. When zygomorphy is present
in addition to actinomorphy within a family, it concerns closely related taxa, therefore the
number of transitions at the family level is a good proxy for the actual number of transitions.
Botanical descriptions were obtained from the AP website, from eFloras (http://www.efloras.
org/), and from Delta (http://delta-intkey.com/angio/www/index.htm). (B) From left to right
and top to bottom: names of the terminal taxa (families) included in the tree. (See Color Insert.)
Author's personal copy
Celastraceae
Lepidobotryaceae
Huaceae
Oxalidaceae
Connaraceae
Brunelliaceae
Cephalotaxaceae
Elaeocarpaceae
Cunoniaceae
Linaceae
Irvingiaceae
Ixonanthaceae
Humiriaceae
Pandaceae
Ochnaceae
Hypericaceae
Podostemaceae
Calophyllaceae
Bonnetiaceae
Clusiaceae
Centroplacaceae
Malpighiaceae
Elatinaceae
Peraceae
Rafflesiaceae
Euphorbiaceae
Picrodendraceae
Phyllanthaceae
Balanopaceae
Chrysobalanaceae
Euphronaceae
Dichapetalaceae
Trigoniaceae
Caryocaraceae
Achariaceae
Goupiaceae
Lacistemataceae
Salicaceae
Violaceae
Passifloraceae
Putranjivaceae
Lophopyxidaceae
Ctenolophonceae
Erythroxylaceae
Rhizophoraceae
Fabaceae-Faboideae
Fabaceae-Mimosoideae
Fabaceae-Caesalpinioideae
Suraniaceae
Polygalaceae
Quillajaceae
Rosaceae
Rhamnaceae
Eleagnaceae
Dirachnaceae
Barbeyaceae
Ulmaceae
Cannabaceae
Moraceae
Urticaceae
Corynocarpaceae
Coriariaceae
Cucurbitaceae
Tetramelaceae
Begoniaceae
Datiscaceae
Anisophylleaceae
Nothofagaceae
Fagaceae
Myricaceae
Rhoipteleaceae
Juglandaceae
Ticodendraceae
Betulaceae
Casuarinaceae
Geraniaceae
Melianthaceae
Francoaceae
Ledocarpaceae
Vivianaceae
Combretaceae
Onagraceae
Lythraceae
Penaeaceae
Alzateaceae
Crypteroniaceae
Melastomataceae
Vochysiaceae
Myrtaceae
Stachyceraceae
Crossosomataceae
Guatemalaceae
Staphyleaceae
Geissolomataceae
Ixerbaceae
Strasburgeriaceae
Aphloiaceae
Picramniaceae
Nitrariaceae
Kirkiaceae
Burseraceae
Anacardiaceae
Simaroubaceae
Meliaceae
Rutaceae
Sapindaceae
Biebersteiniaceae
Gerrardinaceae
Tapisciaceae
Dipentodontaceae
Neuradaceae
Thymeleaceae
Sphaerosepalaceae
Bixaceae
Dipterocarpaceae
Sarcolaenaceae
Cistaceae
Cytinaceae
Muntingiaceae
Malvaceae
Akaniaceae
Tropaeolaceae
Moringaceae
Caricaceae
Setchellanthaceae
Limnanthaceae
Koeberliniaceae
Bataceae
Salvadoraceae
Emblingiaceae
Pentadiplandraceae
Gyrostemonaceae
Resedaceae
Tovariaceae
Capparaceae
Brassicaceae
Cleomaceae
Krameriaceae
Zygophyllaceae
Vitaceae
Peridiscaceae
Cercidiphyllaceae
Daphniphyllaceae
Hamamelidaceae
Altingiaceae
Paeoniaceae
Crassulaceae
Aphanopetalaceae
Tetracarpaceae
Penthoraceae
Haloragaceae
Iteaceae
Grossulariaceae
Saxifragaceae
Dilleniaceae
Gunneraceae
Myrothamnaceae
Celastrales
Oxalidales
Malpighiales
Fabales
Rosales
Cucurbitales
Fagales
Melianthales
Myrtales
Crossosomatales
Picramniales
Sapindales
Huerteales
Malvales
Brassicales
Zygophyllales
Vitales
Saxifragales
Dillenialese
Gunnerales B
107THE EVOLUTION OF FLORAL SYMMETRY
Fig. 5. (Continued)
Author's personal copy
108 H. CITERNE ET AL.
Rosids, unlike in Asterids, zygomorphy has evolved in a limited number
of families and does not characterize large clades (with the exception of
Faboideae (Fabaceae)). Polyandry has evolved frequently throughout the
group and represents a synapomorphy of the speciose subfamily Mimo-
soideae (Fabaceae) as well as of Rosaceae and of the genus Begonia
(Begoniaceae). One striking feature is that many families display varia-
tion in the number of stamens among genera. Rosids are mostly char-
acterized by corollas with free petals whereas Asterids have corollas with
fused petals. Could it be that the former allows more flexibility in floral
organ number than the latter?
Constraints in the evolution of morphological traits may stem from three
different sources that are not necessarily independent: physical, selective and
genetic. Our results suggest that inflorescence and floral architecture do not
influence the evolution of floral symmetry in the same way in all clades,
which invalidates a general role of physical constraints on the evolution of
zygomorphy per se. We focused on the possible evolutionary antagonism
between polyandry and zygomorphy in flowers. From a physical point of
view, it is possible to conceive that the spatial constraints exerted by numer-
ous stamen primordia on the floral meristem are strong at the beginning of
development, but can become relaxed as development proceeds, allowing for
late-onset zygomorphy. From an adaptive point of view, polyandry and
zygomorphy may be viewed as redundant for pollination efficiency. We
argue that polyandry emerging in a zygomorphic context (or the reverse)
may not be positively selected. Indeed, there are few examples of taxa
associating both traits. The unequal distribution of this association between
plant groups (near absent in Asterids but present in Rosids and Ranuncula-
ceae) could suggest variation in the genetic networks underlying both traits.
For instance, in Asterids, the antagonism of polyandry and zygomorphy
could be linked to the role of symmetry genes in inhibiting stamen develop-
ment (see Section VI). It would be of major interest to decipher the genetic
mechanisms involved in taxa where zygomorphy and polyandry co-occur,
such as in Resedaceae (Rosids) or in the Delphinieae (Ranunculales).
V. THE SIGNIFICANCE OF SYMMETRY IN
PLANT–POLLINATOR INTERACTIONS
In this section, we explore the ecological aspects of symmetry and its possible
adaptive value. For ease of comparison, we consider only the flower as the
study object, not lower- (bilabiate structures within flowers such as the
meranthia defined by Westerkamp and Classen-Bockhoff (2007)) or higher-
Author's personal copy
109THE EVOLUTION OF FLORAL SYMMETRY
order (flower-like inflorescences such as capitula of Asteraceae) structures.
We examine to what extent bilateral symmetry can be considered as one
among many mating strategies increasing gene flow and thus potentially
genetic diversity within species. For insects (as for other animal flower
visitors), flowers are potential energetic food sources. In this context, sym-
metry can be perceived as an indicator of the quality and/or quantity of
reward/food, even though floral mimicry may alter this potential relation-
ship (pollination deceit). The capacity of pollinators to perceive symmetry
and discriminate between different types lays the foundation for pollinator-
mediated selection of flower shape, which gives an insight into the potential
role of symmetry in plant population dynamics and species diversification.
A. ZYGOMORPHY AND OUTCROSSING STRATEGIES
Zygomorphy results in a polarized visual signal emitted by the flower, to
which participates the orientation of the symmetry axis. This axis is generally
vertically oriented, thus matching the symmetry plane of flying visitors in
approach. Some studies showed that flower orientation plays a role per se in
orienting the approach and landing behaviour of pollinators (Fenster et al.,
2009; Ushimaru and Hyodo, 2005; Ushimaru et al., 2009), and vertical
orientation has been suggested as being the first evolutionary step toward
the evolution of zygomorphy (Fenster et al., 2009). Morphological differ-
entiation further restricts pollinator access and movement within flowers,
often resulting in improved precision in pollen placement and subsequent
increase in cross-fertilization. Zygomorphy thus appeared as one of numer-
ous contrivances for decreasing selfing and its detrimental effects on
offsprings. In addition, precise pollen placement could form the basis for
reproductive isolation, and thus may promote species diversification.
1. Attributes of Zygomorphic Flowers Promoting Cross-Pollination
The visual signal emitted by zygomorphic flowers is generally borne by the
corolla, with its brilliant colours and polarized morphology. Consistently,
among 38 insect-pollinated Mediterranean species, zygomorphic ones allo-
cate significantly more biomass to the corolla than actinomorphic ones
(Herrera, 2009).
In order to ensure reproductive success, a balance must be achieved
between the amount of pollen deposited on visitors and especially pollina-
tors, and the amount actually transferred to the stigma of another flower.
This is all the more crucial when pollinators are pollen feeders. This is
achieved by adaptations aiming to limit pollen wastage (e.g. poricidal
Author's personal copy
110 H. CITERNE ET AL.
Fig. 6. Interaction between a solitary bee and a flower of Agapanthus africanus
(Amaryllidaceae). Due to the ventral position and curvature of the stamens, pollen
deposition is sternotribic. The style is longer than the filaments, so that the pollinator
comes into contact with the stigma before reaching the anthers. This arrangement favors
cross-pollination. Photograph: S. Nadot.
anthers in buzz-pollinated flowers, which is encountered in some bee-
pollinated species) and to increase precision in pollen placement on the
pollinator’s body (Fig. 6). In this context, zygomorphy of the perianth is
very often supplemented by various devices. For example, rewards—nectar
or oil, less often resins—may be more or less concealed or not easily
accessible, in nectar spur or flower throat. In Antirrhinum and Linaria,
for example, the lower lip is inflated and pressed against the upper lip
(“personate” flower), creating a physical obstacle in front of the nectaries.
Such flowers select for strong bees able to insert their head between the two
lips and open the corolla. Nectar guides are especially elaborate in zygo-
morphic flowers, participating in the internal symmetry, are often yellow—
possibly mimicking anther colour—and attractive to bees (Endress, 1994).
Bilabiate flowers of the lip and flag types (see Section II) are characterized by
contrasted placement of sexual organs. In both types, the lower part of the
flower serves as a landing platform for non-hovering pollinators. Stamens
and stigma are protected by the upper lip in lip-type flowers, and pollen
deposition on pollinators is usually nototribic (on the back). In flag flowers,
Author's personal copy
111THE EVOLUTION OF FLORAL SYMMETRY
stamens and style are enclosed in the lower part of the bilabiate flower,
and pollen deposition is sternotribic (on the ventral part of pollinators).
In addition, special mechanical devices can ensure pollen application,
powered by pollinators as they land on the flower (e.g. trigger system in
Medicago sativa) or as they move in during visitation (e.g. the motile stamens
of Salvia) to reach the reward.
2. Comparison of Zygomorphy with other Mating Strategies Promoting
Outcrossing
Mating strategies promoting cross-pollination include herkogamy (spatial
separation of sexual organs, including various types of stylar polymorph-
isms), dichogamy (temporal separation of male and female maturity,
i.e. protandry or protogyny), self-incompatibility systems, unisexual flowers,
borne on the same (monoecy) or different (dioecy) individuals, and various
combinations of both uni- and bisexual flowers (e.g. gynomonoecy,
gynodioecy). These systems coexist with zygomorphy to a variable extent.
Darwin (1877, cited in Barrett, 2010) considered heterostyly somewhat
functionally redundant with zygomorphy as morphological adaptations
promoting cross-pollination, which is consistent with the rare occurrence
of both characters simultaneously. Barrett et al. (2000) found distyly in a
rare species of the zygomorphic genus Salvia, possibly as a response to a new
environment where protandry was not sufficient to limit intrafloral mating.
In some zygomorphic species, differential spatial arrangements of reproduc-
tive parts have been observed, such as flexistyly (a reciprocal combination of
herko- and dichogamy) in Alpinia species (Zingiberaceae), inversostyly
(reciprocal vertical positioning of sexual organs) in Hemimeris species (Scro-
phulariaceae) or enantiostyly (Section II, and reviewed in Barrett, 2010). In
most enantiostylous species, style deflection is associated with a single polli-
nating anther in opposite direction to the style. This particular configuration
results in pollen deposited on the pollinator’s flank by one type of flower
coming into contact with the stigma of its mirror-image flower (Jesson and
Barrett, 2003). In addition, in some enantiostylous species, anther dimorph-
ism evolved, with the non-pollinating anthers specialized in pollinator feed-
ing. An association between zygomorphy and enantiostyly has been
observed in monocots (Jesson and Barrett, 2003).
Among the two forms of dichogamy, protogyny is common in wind-, bee-
and fly-pollinated flowers, while protandry is predominant in flowers polli-
nated by bees and butterflies (e.g. Endress, 2010). Consistently, an associa-
tion between protandry and zygomorphy has been observed in Asteridae
(Kalisz et al., 2006).
Author's personal copy
112 H. CITERNE ET AL.
B. POLLINATOR PREFERENCES AND THEIR PERCEPTION OF SYMMETRY
Insects constitute the most speciose group of extant plant pollinators, even
though pollination by specific groups of birds, bats and non-flying mammals
also occurs (Cronk and Ojeda, 2008; Endress, 1994; Fleming et al., 2009).
Symmetry as a visual cue may be recognized because of innate preferences or
learning abilities.
Insect were already diverse by the Permian (Whitfield and Kjer, 2008),
which means that their vision began to evolve well before the emergence of
angiosperms, and innate preferences or visual bias may have been recruited
to improve plant–insect relationship up to flower pollination. For instance,
Biesmeijer et al. (2005) established a parallel between floral guides (high
frequency of a dark centre, with radial stripes or dots), insectivorous pitchers
(dark centres, stripes and peripheral dots) and the appearance of the
entrance of the nest of stingless bees. They proposed that plants exploit the
perceptual bias of insects to attract them to specific displays such as flowers.
In tests with artificial flowers, many insect species belonging to Lepidop-
tera, Coleoptera, Hymenoptera and Diptera have been found to prefer the
largest and most symmetric flowers (Mo
¨ller, 2000; Mo
¨ller and Sorci, 1998;
Wignall et al., 2006). Preference for larger flowers is most probably related
to the low resolution of the composite insect eye (Chittka and Raine, 2006).
Bees as a whole constitute the most important group of pollinators with
about 20,000 species (Grimaldi and Engel, 2005), including insects with
different social behaviour (solitary or social), size and various adaptations
for nectar and pollen collection (e.g. Krenn et al., 2005; Thorp, 2000). Bees
have high learning abilities. They are able to discriminate bilateral and radial
symmetry from asymmetry. At a short distance, internal flower symmetry
marked, for instance, by nectar guides may reinforce symmetry perception
(Lehrer, 1999). Among bilaterally symmetrical patterns, bees prefer the
patterns with vertically oriented symmetry plane, and among radially sym-
metric patterns, the ones with radiating bars rather than concentric circles
(Giurfa et al., 1999). Preference for bilaterally symmetric shapes was demon-
strated to be innate in flower-naive bumblebees (Rodriguez et al., 2004). In
many other studies, it is not always clear whether discrimination is based on
innate preference or experience from natural conditions where particular
shapes may be linked to the availability of different rewards (Lehrer, 1999).
C. FLORAL SYMMETRY AND POLLINATION SYNDROMES
The concept of pollination syndrome has been widely debated since its
definition in the 19th century by Federico Delpino (Fenster et al., 2004;
Author's personal copy
113THE EVOLUTION OF FLORAL SYMMETRY
Ollerton et al., 2009; Tripp and Manos, 2008 and references therein). It
translates the observation that similar suites of flower traits can be found in
evolutionarily unrelated taxa as a result of convergent selection by the same
pollinating agent (Faegri and van der Pijl, 1966; Fenster et al., 2004; Proctor
et al., 1996). Functional groups of pollinators have been defined to account
for the observation that many species have flowers visited by large arrays of
pollinator species, and conversely some pollinators visit a large array of
species with different flower shapes. Analysing the Carlinville (Illinois)
flora, Fenster et al. (2004) found that 61% of 86 zygomorphic species were
pollinated by one functional group, significantly more than the 52% observed
among 192 actinomorphic species. The traditional bee pollination syndrome
includes a well-marked tridimensional form—more or less tubular flowers
and most commonly zygomorphic—yellow, blue or purple colour, and nectar
and pollen rewards (Faegri and van der Pijl, 1966; Proctor et al., 1996). This is
not to say that all zygomorphic flowers are bee-pollinated. Indeed, it is
believed the shape associated with bee pollination may have paved the way
for further diversification, for example, bird pollination consistently evolved
from bee pollination, and some bird-pollinated species have strongly
zygomorphic flowers (e.g. Lotus maculatus—Cronk and Ojeda, 2008).
D. VARIABILITY OF FLORAL TRAITS IN ZYGOMORPHIC AND
ACTINOMORPHIC SPECIES
Because of their specific interaction with a limited number of different
pollinators, it has been proposed that species with zygomorphic flowers
should experience stronger pollinator-mediated stabilizing selection for
flower shape and size than species with actinomorphic flowers (Berg, 1959;
Gong and Huang, 2009; Wolfe and Krstolic, 1999). Consistently, various
studies demonstrate lower variability in flower size in zygomorphic species
than in actinomorphic ones (Herrera et al., 2008; Ushimaru and Hyodo,
2005; van Kleunen et al., 2008; Wolfe and Krstolic, 1999).
While the type of symmetry is generally consubstantial with species defini-
tion, within-species variability around a main type exists, and has been
reported to be partly genetically controlled (Mo
¨ller and Shykoff, 1999).
Departure from perfect symmetry is generally measured as the difference
between the longest and the shortest petal in actinomorphic flowers, and
between the “right” and the “left” petal in zygomorphic ones (e.g. Mo
¨ller
and Eriksson, 1994). More integrative approaches have been attempted,
relying on geometric modelling of shape (Frey et al., 2007; Go
´mez et al.,
2006), which capture more spatial information.
Author's personal copy
114 H. CITERNE ET AL.
Small randomly directed deviations from perfect symmetry in natural
populations are defined as fluctuating asymmetry (Endress, 1999; Mo
¨ller,
2000; Rudall et al., 2002). Factors limiting such asymmetry are synorganiza-
tion and bilateral symmetry. Highly synorganized flowers such as those
encountered in orchids (zygomorphic—Rudall and Bateman, 2002) or
Apocynaceae (actinomorphic) exhibit low fluctuating asymmetry (Endress,
1999). Low fluctuating asymmetry is also observed in zygomorphic species
compared to actinomorphic ones (Mo
¨ller, 2000), while leaf asymmetry
does not differ between the two categories of plants, suggesting that repro-
ductive traits are subject to differential selective pressures in the two groups,
in contrast to vegetative traits. However, zygomorphic flowers tend to be
larger than actinomorphic ones, and larger flowers generally exhibit less
fluctuating asymmetry than smaller ones; thus, it is difficult to separate the
actual effect of size and symmetry on the level of fluctuating asymmetry
(Mo
¨ller, 2000).
The capacity to better control random variation may be an indication
of “genotype quality,” and the most symmetrical flowers of some species
have been shown to be the richest in nectar (Mo
¨ller, 1995, 2000). In some
species, a low degree of asymmetry was associated with a better seed set
(Mo
¨ller, 2000 for review), but in other ones this association does not hold
(Botto-Mahan et al., 2004; Frey et al., 2005; Weeks and Frey, 2007). Flower
visitation and reproductive success can be affected by a large number of
uncontrolled causes, from environmental factors to biological ones, which
may explain the lack of consistency of these results.
In addition to the variability of individual traits, the level of floral inte-
gration measured by the correlations between the size of floral parts, is also
expected to be higher in zygomorphic than in actinomorphic species because
of the fit with pollinator morphology. Harder and Johnson (2009) found
such a trend in their compilation of 56 studies on 43 animal-pollinated
species.
An integrative view of corolla shape and symmetry has been obtained by
means of geometric morphometrics in the Brassicaceae species Erysimum
mediohispanicum (Go
´mez et al., 2006, 2008b). Shape variations are mainly
found in the width of the petals and their relative distribution, generating
symmetry ranging from actinomorphy to disymmetry and zygomorphy.
The first study, conducted over 2 years in a single population, demonstrates
that the main beetle pollinator preferentially visits disymmetric and zygo-
morphic corollas. In addition, the zygomorphic shape exhibits a higher
fitness, measured by seedling survival (Go
´mez et al., 2006). In a more
extensive study involving three different populations visited by a larger
diversity of pollinator assemblages, it was found that different pollinators
Author's personal copy
115THE EVOLUTION OF FLORAL SYMMETRY
preferred different flower shapes, and that between-population variability in
shape can be accounted for by preference of the major local pollinator.
Pollen and nectar production also varied significantly with corolla shape
(Go
´mez et al., 2008a). The most rewarding flowers matched the artificial
flower shape preferentially visited by bees, suggesting that bees use the visual
cue as an indicator of reward amount. Significant phenotypic selection on
flower shape was observed in all populations of this species (Go
´mez et al.,
2006, 2008b), thus giving an insight in the mechanisms of flower shape
evolution mediated by reward and driven by pollinator preference.
To summarize, zygomorphy results in tighter flower–pollinator interac-
tion than actinomorphy, and probably contributes to increased outcrossing
rates. Several lines of arguments thus support the hypothesis that zygomor-
phy is an adaptive trait that may have brought about species divergence and
species radiation in the past. However, extant populations generally exhibit
low diversity in floral symmetry, making it difficult to compare the selective
values of different types of symmetry.
VI. MOLECULAR BASES OF FLOWER SYMMETRY
A. THE FLORAL SYMMETRY GENE REGULATORY NETWORK IN
ANTIRRHINUM MAJUS
The molecular signals controlling floral symmetry were first described, and are
best understood, in A. majus (Veronicaceae, Lamiales). Wild-type A. majus
flowers have strongly differentiated organs along the dorsoventral axis parti-
cularly in the second and third whorls (petals and stamens). The two dorsal,
two lateral and single ventral petals differ in size, shape, epidermal cell type
and internal symmetry; in particular, the dorsal petal lobes are large and
asymmetric whereas the ventral petal lobe is smaller and bilaterally symme-
trical. The dorsal stamen is arrested to form a staminode, whereas the lateral
and ventral stamen pairs differ in filament length and pilosity. Unequal devel-
opment along the dorsoventral axis is apparent at the start of organogenesis,
with dorsal organs delayed in their initiation (Luo et al., 1996).
Two closely related genes CYCLOIDEA (CYC) and DICHOTOMA
(DICH) have been identified as master control genes for bilateral symmetry
by forward genetic screens (Luo et al., 1996, 1999). Cyc:dich double mutants
have completely radially symmetric flowers with all organs resembling the
ventral phenotype. Single cyc mutants have ventralized lateral organs and
dorsal organs with lateralized features, while dich mutants display altera-
tions of the internal symmetry of the dorsal petals. CYC and DICH are two
Author's personal copy
116 H. CITERNE ET AL.
closely related DNA-binding transcription factors belonging to the TCP
gene family (Cubas et al., 1999b; Luo et al., 1996, 1999). Both genes are
expressed in the dorsal region of the floral meristem throughout its devel-
opment (Luo et al., 1996, 1999). CYC and DICH expression is detectable
prior to organogenesis at the junction of the inflorescence and floral mer-
istem. After all organs are initiated, their expression is limited to the two
dorsal petals and staminode, with DICH having a more restricted expression
in the dorsal half of the dorsal petals (Luo et al., 1996, 1999). CYC, like other
members of the TCP gene family, is believed to affect development by
regulating patterns of cell growth and proliferation (reviewed in Cubas
et al., 2001; Martı
´n-Trillo and Cubas, 2009). In the dorsal staminode, CYC
expression correlates with the downregulation of cell cycle genes such as
HISTONE H4 and CYCLIN D3B (Gaudin et al., 2000). Although the initial
effect of CYC expression on the floral meristem is growth retardation, at
later stages of development its effect as a growth suppressor or promoter is
dependent on organ identity rather than positional cues (Clark and Coen,
2002; Coen and Meyerowitz, 1991; Luo et al., 1996).
CYC and DICH promote dorsal identity in A. majus flowers. By contrast,
ventral identity is controlled by DIVARICATA (DIV), a gene encoding
an MYB transcription factor with two imperfect repeats (R2R3) of the
DNA-binding MYB domain (Almeida et al., 1997; Galego and Almeida,
2002). In loss-of-function div mutants, the ventral region of the corolla
acquires lateral identity (Almeida et al., 1997). DIV is transcribed in all
floral organs early in development and is inhibited post-transcriptionally in
the dorsal and lateral regions through the expression of CYC and DICH
(Galego and Almeida, 2002). At later stages of development when ventral
petals become differentiated from lateral petals, DIV is strongly induced in
the inner layer of epidermal cells of the ventral and adjacent parts of the
lateral corolla lobes (Galego and Almeida, 2002). DIV promotes the expres-
sion of a MIXTA-like MYB gene AmMYBML1 required for the develop-
ment of ventral-specific petal epidermal cell types, in conjunction with
B-class MADS box genes (Perez-Rodriguez et al., 2005).
A gene regulatory network has been proposed for the control of floral
symmetry in A. majus (Costa et al., 2005). CYC is activated upon floral
induction; the molecular trigger is unknown but appears to be independent
of floral meristem identity genes, as CYC is also expressed in the adaxial
region of young axillary shoots adjacent to the inflorescence (Clark and
Coen, 2002). Asymmetric expression in axillary meristems suggests that
CYC responds to a positional cue or gradient within these meristems
(Clark and Coen, 2002). The persistent expression of CYC during floral
development is thought to be maintained by B- and C-function MADS
Author's personal copy
Floral
induction
CYC
DICH
RAD
RAD – independent
pathway
CYCLIN D3B
B-and C-function
MADS box genes
DIV
DIV
DIV AmMYBML1
B -function
MADS box genes
DIV
117THE EVOLUTION OF FLORAL SYMMETRY
proteins such as DEFICIENS and PLENA (Clark and Coen, 2002) as well
as self-positive feedback (Costa et al., 2005). One direct target of CYC and
DICH is RADIALIS (RAD), a single-repeat MYB transcription factor that
has TCP-binding sites in its promoter region and intron (Corley et al., 2005;
Costa et al., 2005). RAD is required to mediate most of the effects of CYC
and DICH; however, residual asymmetry is found in rad mutants suggesting
some effects of CYC are independent of RAD (Corley et al., 2005; Costa
et al., 2005). RAD is closely related to DIV, but has lost the C-terminal MYB
II domain (Corley et al., 2005). Although direct antagonism of RAD and
DIV remains to be demonstrated, this could operate by direct competition
for molecular targets (Corley et al., 2005). RAD is believed to act non-
autonomously on lateral organ development by inhibiting DIV (Corley
et al., 2005). This may occur by cell-to-cell movement of RAD proteins, or
alternatively by the activation of a downstream signalling molecule that
affects lateral development (Corley et al., 2005). The gene interactions
described above are summarized in Fig. 7.
Fig. 7. Major gene interactions regulating floral symmetry in Antirrhinum majus.
Gene transcription and proposed interactions are shown in the different regions (dorsal
(blue), lateral (green) and ventral (orange)) of the floral meristem. Arrows indicate
upregulation, lines terminated by a perpendicular line indicate repression, and dashed
lines for the repression of DIV by RAD in lateral regions represent putative RAD protein
movement or indirect interaction. (See Color Insert.)
Author's personal copy
118 H. CITERNE ET AL.
B. CYC-LIKE GENES ARE IMPLICATED IN THE CONTROL OF ZYGOMORPHY
IN DIVERSE LINEAGES
The extent to which these genes are implicated, and their interactions con-
served, in the elaboration of bilaterally symmetrical flowers has been exam-
ined in diverse groups of angiosperms (Fig. 8). Most studies have focused on
ASTERID
Asterales
Apiales
Dipsacales
Aquifoliales
Lamiales
Solanales
Gontianales
Garryales
Fabales
Rosales
Malpighiales
Myrtales
Brassicales
Malvales
Santalales
Caryophyllales
Saxifragales
Gunnerales
ROSID
d-CYC/DICH e-VmCYC1/VmCYC2
f,g,h-LegCYC1/LegCYC2 (LST1) f,g-LegCYC3 (KEW1)
k-IaTCP1
a,b,c–CYC2 (1-2)
i,j-CYC2B (1-2) i-CYC2A, j-CYC2A/CYC2B-3
Fig. 8. Summary of expression patterns of CYC-like genes (CYC2 clade) during late
developmental stages in the corolla of representative zygomorphic core eudicot species
(phylogeny derived from the Angiosperm Phylogeny website). Asterales: a. Gerbera hybrida
(Broholm et al., 2008), b. Senecio squalidus (Kim et al., 2008), c. Helianthus annuus (Chapman
et al., 2008); Lamiales: d. Antirrhinum majus (Luo et al., 1996, 1999), e. Veronica montana
(Preston et al., 2009); Fabales: f. Lotus japonicus (Feng et al., 2006), g. Pisum sativum (Wang
et al., 2008), h. Lupinus nanus (Citerne et al., 2006); Malpighiales: i. Byrsonima crassifolia,
j. Janusia guaranitica (Zhang et al., 2010); Brassicales: k. Iberis amara (Busch and Zachgo,
2007). Although the predominant expression domain is dorsal (and lateral), ventral expression
is found in Asterales. Expression is also detected on the abaxial side in I. amara but is weaker
(in yellow) than on the dorsal side (orange). The effect on petal growth and development
(acting as growth promoter or suppressor) varies across species. (See Color Insert.)
Author's personal copy
119THE EVOLUTION OF FLORAL SYMMETRY
homologues of CYC/DICH. The Lamiales have evolved zygomorphic flow-
ers from an ancestor with actinomorphic flowers (Coen and Nugent, 1994;
Donoghue et al., 1998; Endress, 2001b), and it is therefore unsurprising that
CYC-like genes are implicated in the control of bilateral symmetry in other
members of this clade. In particular, the persistent expression on the dorsal
side of the developing flower of CYC homologues has been described in
other zygomorphic species of Veronicaceae (Cubas et al., 1999a; Hileman
et al., 2003; Preston et al., 2009) and Gesneriaceae (Du and Wang, 2008; Gao
et al., 2008; Song et al., 2009; Zhou et al., 2008). Notably, variations in the
pattern of stamen development and the degree of petal differentiation along
the dorsoventral axis have frequently been associated with modifications of
CYC-like gene expression. For example, in Mohavea confertiflora (Veroni-
caceae), the abortion of both dorsal and lateral stamens coincides with an
expansion of the expression domain of CYC and DICH homologues from
the dorsal region to the lateral stamen primordia (Hileman et al., 2003).
Similarly, in Chirita heterotricha (Gesneriaceae), an expanded expression
domain (i.e. in both dorsal and lateral regions of the flower) of one CYC
homologue coincides with the abortion of dorsal and lateral stamens (Gao
et al., 2008). In the Lamiales, however, stamen abortion per se is not
necessarily associated with CYC expression, particularly on the ventral
side (Preston et al., 2009; but see Song et al., 2009).
CYC-like genes have been recruited for the control of floral symmetry in
families that have evolved zygomorphy independently of the Lamiales.
Within Rosids, these have been implicated in the control of dorsal (and
sometimes lateral) petal identity in Fabaceae, Brassicaceae and Malpighia-
ceae (Busch and Zachgo, 2007; Feng et al., 2006; Wang et al., 2008; Zhang
et al., 2010). In Papilionoideae (Fabaceae), two closely related CYC-like
genes are expressed in the dorsal region of developing flowers (Citerne
et al., 2006; Feng et al., 2006; Wang et al., 2008); one of these, LOBED
STANDARD 1 (LST1), is an important determinant of dorsal petal identity,
promoting cellular proliferation and epidermal cell differentiation (Feng
et al., 2006; Wang et al., 2008). The other copy appears to have less effect
on phenotype, but may act redundantly to control dorsal petal development
(Wang et al., 2008). A third CYC homologue expressed in the dorsal and
lateral regions of the developing flower, KEELED WINGS 1 (KEW1), is also
a regulator of dorsoventral asymmetry, and determines lateral petal identity
(Feng et al., 2006; Wang et al., 2008). The petals of lst1:kew1 double mutants
have ventral identity in both L. japonicus and Pisum sativum (Feng et al.,
2006; Wang et al., 2008).
Similar expression is found in duplicate CYC-like genes in zygomorphic
Malpighiaceae (Zhang et al., 2010). As in Fabaceae, paralogues exhibit
Author's personal copy
120 H. CITERNE ET AL.
either dorsal or dorsolateral expression late in floral development, a pattern
that is not found in their closest relatives with actinomorphic flowers. Gene
duplication, and consequently functional divergence, has occurred indepen-
dently in Fabaceae and Malpighiaceae (Citerne et al., 2003; Zhang et al.,
2010), and the extent of functional redundancy and specificity remains to be
demonstrated in Malpighiaceae.
Within Brassicaceae, I. amara has been a case study for the control of
late-onset zygomorphy (Busch and Zachgo, 2007). I. amara flowers are
tetramerous with two reduced dorsal petals and two enlarged ventral petals.
A shift occurs during flower development: petals are initiated simultaneously
and grow equally until relatively late in development when, at the onset of
stamen differentiation, unequal adaxial–abaxial petal growth becomes
apparent. A shift is also observed in the expression of the homologue of
CYC in I. amara IaTCP1, which is expressed equally early in development
but becomes strongly expressed in the two dorsal petals relative to the
ventral petals at later developmental stages (Busch and Zachgo, 2007). The
effect of IaTCP1 decreases petal growth and is the opposite of what is
observed in Antirrhinum and Fabaceae (i.e. promoter of petal growth during
late developmental stages), indicative of functional divergence. Constitutive
expression of IaTCP1 in Arabidopsis produces a similar phenotype to when
the endogenous gene TCP1 is constitutively expressed, that is, repressed cell
division reducing vegetative and petal growth, suggesting that DNA targets
and interacting proteins are conserved in Brassicaceae (Busch and Zachgo,
2007). By contrast, the effect on petal growth of heterologous expression of
Antirrhinum CYC in Arabidopsis is enlargement by cell expansion suggesting
that targets and interacting proteins are not conserved between Antirrhinum
and Brassicaceae (Busch and Zachgo, 2007; Costa et al., 2005).
In Asteraceae (Asterid clade like Lamiales), CYC-like genes also regulate
dorsoventral asymmetry but in a novel manner, as a ventralizing factor
(Broholm et al., 2008; Kim et al., 2008). In radiate inflorescences, both
actinomorphic (disc) and zygomorphic (ray) flowers are present: the outer-
most flowers develop enlarged fused petal lobes on the ventral side (the
ligule), and have aborted stamens. Expression of a subset of CYC-like
genes was found predominantly in ray flowers (Broholm et al., 2008; Chap-
man et al., 2008; Kim et al., 2008), in particular on the ventral side promoting
ligule development (Broholm et al., 2008). In Gerbera hybrida, the effects of
constitutive expression of GhCYC2 differ not only with organ type (increas-
ing growth of petals and reducing growth of stamens) but also according to
flower type and position along the capitulum radius (Broholm et al., 2008).
There is less evidence for the involvement of CYC-like genes in the control
of zygomorphy outside the core eudicots. In rice, RETARDED PALEA1
Author's personal copy
121THE EVOLUTION OF FLORAL SYMMETRY
(REP1) promotes the differentiation of the palea and the lemma (which
together function as a calyx surrounding the stamens and carpel) by regulat-
ing cellular expansion and differentiation (Yuan et al., 2009). In Fumarioi-
deae (Papaveraceae), bilaterally symmetric flowers are characterized by the
development of a nectar spur in one of the two outer petals. The asymmetric
expression of one CYC-like gene in the spurred petal of C. sempervirens
could indicate a role in floral zygomorphy but remains to be demonstrated
functionally (Damerval et al., 2007).
C. GENETIC MECHANISMS UNDERLYING CHANGES IN FLORAL SYMMETRY
Modification of key development regulators appears to underlie morpholo-
gical evolution (e.g. Doebley and Lukens, 1998; Wilson et al., 1977; Rosin
and Kramer, 2009). Changes in the timing, duration and localization of
CYC-like gene expression have repeatedly been implicated in changes in
floral symmetry. Case studies have provided examples of different muta-
tional mechanisms. In L. vulgaris, naturally occurring radially symmetrical
mutants have lost CYC expression through extensive methylation of pro-
moter and ORF (Cubas et al., 1999a). Surveys of epigenetic alteration of
gene expression in plants suggest this mode of regulation may play a role in
morphological evolution (Kalisz and Purrugannan, 2004; Rapp and Wendel,
2005); however, no other example has been described so far in the context of
floral symmetry.
In Senecio, interspecific hybridization has been shown to have played a
part in the evolution of a floral symmetry polymorphism (Kim et al.,
2008). In Senecio vulgaris, a species with typically non-radiate inflores-
cences bearing only disc florets, a radiate form has evolved by introgres-
sion of an allele at the RAY locus from Senecio squalidus with radiate
inflorescences. The RAY locus consists of two CYC2 paralogues, and as in
Gerbera, one of these genes appears to promote ventral identity in ray
florets. These genes are specifically expressed in the outer florets, and are
differentially expressed in the two forms. It is believed that changes in cis-
regulatory regions, rather than the ORF, may underlie the differences
between the two morphs.
There are numerous cases of species derived from zygomorphic lineages
that have evolved actinomorphic flowers secondarily. Diverse types of
changes in CYC-like gene expression have been described. In Plantago
lanceolata (Veronicaceae), a wind-pollinated genus with radial tetramerous
flowers, expression of the CYC-homologue PlCYC is detected in flowers
only at later stages of development in all four stamens (in the anther con-
nective and stamen filament) and transiently in the ovaries (Reardon et al.,
Author's personal copy
122 H. CITERNE ET AL.
2009). The actinomorphy in P. lanceolata is therefore correlated with a lack
of both early expression and asymmetric expression in petals. The function
of PlCYC is unknown, but has been proposed to delay stamen development
and therefore promote dichogamy. Unlike other members of Veronicaceae
(Preston et al., 2009), P. lanceolata has only one CYC-like gene, which could
suggest a functionally significant gene-loss event (Reardon et al., 2009).
In Bournea leiophylla (Gesneriaceae), the transition from a zygomorphic
pattern in the early stages of floral development to actinomorphy at anthesis
correlates with the downregulation of the dorsal expression of a CYC-like
and a RAD-like gene (Zhou et al., 2008). By contrast, in Cadia purpurea
(Fabaceae), the derived radial symmetry of the corolla coincides with an
expansion of the expression domain of one CYC-like gene to all petals
(Citerne et al., 2006). It remains to be determined whether these heterochro-
nic and heterotopic changes in gene expression are caused by modifications
in their cis-regulatory regions or in the function or nature of their trans-
acting regulators.
D. EVOLUTION OF CYC-LIKE GENES: FUNCTIONAL IMPLICATIONS
It is believed that morphological evolution proceeds by tinkering of existing
genetic pathways (Jacob, 1977). What is the context of CYC-like gene
evolution that makes them a common player in the repeated evolution of
floral zygomorphy in many lineages? Members of the TCP gene family are
transcription factors that bind to DNA through their characteristic basic
helix–loop–helix domain (bHLH) (Martı
´n-Trillo and Cubas, 2009). CYC
together with its homologue in maize TEOSINTE BRANCHED 1 (TB1)
belong to a clade of class II TCP genes (the ECE clade), whose members are
generally characterized by a second short conserved hydrophilic domain (R
domain) and a conserved motif of amino acids termed “ECE”. Character-
ized genes in this clade appear to have a predominant role in growth repres-
sion (Martı
´n-Trillo and Cubas, 2009). TB1 is a suppressor of axillary
meristem growth (Doebley et al., 1997), but also affects floral development
by suppressing stamen growth in female flowers (Hubbard et al., 2002).
Two major duplication events have occurred in the ECE clade, prior to the
divergence of the core eudicots (Howarth and Donoghue, 2006). All genes
implicated so far in dorsoventral asymmetry of flowers belong to the same
CYC2 clade (Howarth and Donoghue, 2006), whereas genes from the CYC1
and CYC3 clade in Arabidopsis appear to have a role like TB1 in the
development of axillary buds (Aguilar-Martı
´nez et al., 2007; Finlayson,
2007). This could reflect sub/neofunctionalization of major ECE-CYC
lineages in the core eudicots, where the effects on floral development such
Author's personal copy
123THE EVOLUTION OF FLORAL SYMMETRY
as stamen suppression of the CYC/TB1 ancestor were retained and subse-
quently modified in the CYC2 clade.
The dorsal expression of many CYC2 genes is believed to be shared by the
common ancestor of Rosids and Asterids (Cubas et al., 2001). In Arabidopsis
thaliana (Brassicaceae), which has radially symmetrical flowers, the homo-
logue of CYC TCP1, is transiently expressed in the dorsal region of the floral
meristem prior to organogenesis (Cubas et al., 2001). Modification of this
incipient asymmetry through its persistent expression during organ primor-
dia development could account for the repeated evolution of zygomorphy
(Cubas et al., 2001). However, evidence of ventral and radial expression of
CYC2 genes in different lineages suggests lability in the response to local
signals along the dorsoventral axis in the floral meristem. For example, early
expression of IaTCP1 in I. amara is very weak and ubiquitous (Busch and
Zachgo, 2007), and differs from that of Arabidopsis TCP1, which is transi-
ently expressed on the dorsal side of the floral meristem. Without expression
data from other Brassicaceae, it is not clear whether the early asymmetric
pattern is ancestral or derived. Similarly in Malpighiales, the actinomorphic
relatives of the zygomorphic members of family Malpighiaceae (which have
“typical” CYC dorsal expression) differ in their expression of CYC-like
genes; in the closest relative these are expressed uniformly in late-stage
flowers, whereas in the next closest relative no CYC expression is detected
at this stage (Zhang et al., 2010). The role of CYC2 genes in petal develop-
ment also appears to be labile, probably reflecting differences in their inter-
action with other proteins. In different lineages, these genes can either
promote or repress growth through cell proliferation and/or expansion,
and are often associated with cellular differentiation.
Independent duplication of CYC2 genes appears to be a common phe-
nomenon in core eudicots, for example, in Veronicaceae (Preston et al.,
2009), Gesneriaceae (Citerne et al., 2000; Smith et al., 2004, 2006),
Asteraceae (Broholm et al., 2008; Chapman et al., 2008), Dipsacales
(Howarth and Donoghue, 2005), Fabaceae (Citerne et al., 2003; Fukuda
et al., 2003) and Malpighiales (Zhang et al., 2010). Correlation between
floral form and copy number has been postulated in Dipsacales but the
significance of duplications specific to zygomorphic lineages remains to