Understanding the role of floral development in the evolution of angiosperm flowers: clarifications from a historical and physico-dynamic perspective
- 490 Downloads
Flower morphology results from the interaction of an established genetic program, the influence of external forces induced by pollination systems, and physical forces acting before, during and after initiation. Floral ontogeny, as the process of development from a meristem to a fully developed flower, can be approached either from a historical perspective, as a “recapitulation of the phylogeny” mainly explained as a process of genetic mutations through time, or from a physico-dynamic perspective, where time, spatial pressures, and growth processes are determining factors in creating the floral morphospace. The first (historical) perspective clarifies how flower morphology is the result of development over time, where evolutionary changes are only possible using building blocks that are available at a certain stage in the developmental history. Flowers are regulated by genetically determined constraints and development clarifies specific transitions between different floral morphs. These constraints are the result of inherent mutations or are induced by the interaction of flowers with pollinators. The second (physico-dynamic) perspective explains how changes in the physical environment of apical meristems create shifts in ontogeny and this is reflected in the morphospace of flowers. Changes in morphology are mainly induced by shifts in space, caused by the time of initiation (heterochrony), pressure of organs, and alterations of the size of the floral meristem, and these operate independently or in parallel with genetic factors. A number of examples demonstrate this interaction and its importance in the establishment of different floral forms. Both perspectives are complementary and should be considered in the understanding of factors regulating floral development. It is suggested that floral evolution is the result of alternating bursts of physical constraints and genetic stabilization processes following each other in succession. Future research needs to combine these different perspectives in understanding the evolution of floral systems and their diversification.
KeywordsFloral meristem Genetics Heterochrony Morphospace Organ pressure Size change
Causes for diversification of plants have become increasingly clarified by the understanding of the genetic structure of model organisms through the science of evo-devo. It is widely recognised that the building of extensive gene regulatory networks is responsible for the development and evolution of floral systems. The development of flowers, from their initiation until senescence, is regulated by a cascade of gene regulatory networks (GRN) that act in sequence and influence the build-up of floral structures (e.g. Alvarez-Buylla et al. 2010; Specht and Howarth 2015). In flowers the most important network is the ABCDE model, which regulates the identity of flowers and floral organs, and is responsible for the evolution of floral morphology (see Alvarez-Buylla et al. 2010; Bowman et al. 1989; Coen and Meyerowitz 1991; Erbar 2007; Glover et al. 2015; Kramer and Hall 2005; Litt and Kramer 2010). Major changes in the floral configurations, such as the development of symmetry or the evolution of petals can be explained by shifts in genetic expression and gene duplications (Donoghue et al. 1998; Hileman and Irish 2009; Irish 2009; Irish and Litt 2005; Litt and Kramer 2010; Specht and Bartlett 2009; Specht and Howarth 2015).
Advances in the understanding of angiosperm relationships have been extensive in the last decades, with a clearcut recognition of different plant groups, as supported by molecular phylogenies (e.g. Angiosperm Phylogeny Group 2016; Zeng et al. 2014). Since the basic principles of morphology have been formulated by Goethe, various attempts have been made to understand homology. However, homology has mostly been restricted to common ancestry (Vergara-Silva 2003; Weber 2003). Before molecular systematics, classification had relied on morphological characters, which were often poorly defined and homoplastic. With the onset of cladistics floral morphology, including floral development, has mostly been approached from the perspective of phylogenetic relationships (“historical homology”: Endress 2003; Rutishauser and Moline 2005). Characters are evaluated for their systematic impact, mainly as synapomorphies, reflecting common ancestry (e.g. Jaramillo et al. 2004; Specht et al. 2012). This oversimplification inevitably leads to a rejection of any convergences and parallelisms as homoplastic and at the same time to a loss of understanding of the fundamental significance of floral morphology.
The concept of “biological homology” (see Endress 2003; Rutishauser and Moline 2005) implies that homologous structures share a same set of developmental constraints. Indeed, concepts such as deep morphology (Scotland 2010; Stuessy et al. 2003), apomorphic tendencies (Endress 2003, 2010a; Endress and Matthews 2012), or cryptic apomorphies (Ronse De Craene 2010) emphasize the importance of the presence of characters in clades, even as they are not synapomorphic. The occurrence of parallel evolution in different taxa can be the result of a re-expression of latent genetic mechanisms that can be switched on and off (see e.g. Glover et al. 2015; Vasconcelos et al. 2017). However, the approach of homology from a genetic perspective has its limitations. The reliance on molecular processes to interpret floral evolution (e.g. Buzgo et al. 2004; Glover et al. 2015; Jaramillo and Kramer 2007; Specht and Howarth 2015) also limits the interpretation of homology and character evolution to gene interactions, without considering other factors, such as self-organisational phenomena acting on the development of structures (see Endress 2005).
Petals have been mainly approached from the perspective of B-gene expression (e.g. Glover et al. 2015; Hileman and Irish 2009; Sharma et al. 2011), where various structures are lumped under the appellation of petals. Hileman and Irish (2009) showed that petal versus sepal characteristics are regulated by downstream interactions of A and B-genes through a loophole system of activation and desactivation. Gene expression studies may clarify the mechanisms that make petals appear as they are subjectively conceived but do not clarify their homology. A good illustration of this is the identification of lodicules as “petals” (e.g. Yoshida 2012). When considering the poor definition of petals, their origin is still debatable (see Ronse De Craene 2007; Ronse De Craene and Brockington 2013). However, it is recognized that a distinction needs to be made between process homology and structural homology (e.g. Jaramillo and Kramer 2007), where the former is related to the activation of gene systems (such as petaloidy of organs), while the latter refers to more stable system, such as organ position in flowers [which relates to the basic homology criterion of Remane (1956)]. However, causes for the position and number of organs in flowers are mostly not well understood or rely on outdated models (Kitazawa and Fujimoto 2015).
Apart from genetic systems, evolution of flowers is strongly influenced by pollination systems. There are major shifts in angiosperm flower morphology linked with pollinator preferences. Increasing evidence suggests that pollinators can potentially drive diversification at different levels of the evolution of flowers (e.g. Buckhari et al. 2017; Hodges and Arnold 1995; Specht et al. 2012; van der Niet and Johnson 2012; van der Niet et al. 2014). There are various examples based on observations and experiments, showing that shifts in pollinator preference can cause a response in plant structures and lead to diversification. There is the classical example that nectar spur length of Aquilegia (Ranunculaceae) is closely correlated with bee, bird or moth pollination syndromes and has arisen sequentially in the evolution (Hodges and Arnold 1995; Whittall and Hodges 2007). In Zingiberales there is a neat correlation between floral shape and shift in pollination syndrome (Specht et al. 2012). There needs to be a balance between pollinator attraction and pollination efficiency that can induce a progressive diversification within a species and lead to flower evolution (van der Niet et al. 2014). Chinga and Pérez (2016) demonstrated that co-variation of different morphological traits in the late ontogeny of flowers of Schizanthus species (Solanaceae) is closely correlated with pollination systems.
The strong link between floral morphology and pollination systems is also very clear in transitions to wind pollination. Most wind-pollination syndromes are derived from biotic systems and are characterized by the loss of organs directly associated with pollinator attraction, such as petals and nectaries (e.g. Friedman and Barrett 2008; Leins and Erbar 2010; Linder 1998; Ronse De Craene 2010). However, several cases are known where wind-pollinated systems have undergone a reversal to insect pollination (e.g. Cyperaceae, Fagaceae, Salicaceae, Caryophyllales: Brockington et al. 2009; Friedman 2011; Ronse De Craene and Brockington 2013). However, with the loss of petals, attraction had to be recreated by the recruitment of different building blocks in the flower. This has been achieved in two ways, either by transforming remaining organs such as sepals and stamens into attractive structures, or by the aggregation of simple flowers into pseudanthial units surrounded by attractive bracts. The first possibility is clearly expressed in Caryophyllales, where petals have been re-invented many times from wind-pollinated precursors (for a review see Ronse De Craene 2013). Attraction comes either from a petaloid calyx (e.g. Phytolaccaceae, Nyctaginaceae, Cactaceae), or from transformed outer staminodes (e.g. Corbichoniaceae, Aizoaceae). The second possibility evolved in several clades (e.g. Classen-Bockhoff 1991; Endress 2010a), culminating in the genus Euphorbia (Euphorbiaceae), where pseudanthia imitate real flowers nearly to perfection (Prenner and Rudall 2007).
The development of new technologies have seen a surge in the study of three-dimensional morphometrics to clarify how plant form adapts to pollinator shape or to intrinsic forces during development (e.g. Chartier et al. 2014; Stuppy et al. 2003; van der Niet et al. 2010). These studies in the morphospace of flowers also exist in animals where their application has many practical outcomes (e.g. Arthur 2000; Mitteroecker and Huttegger 2009; Riegner 2008). Pollinator shifts will generally be correlated with micromorphological changes, such as corolla tube length, spur length, or flower colour (van der Niet and Johnson 2012), parameters that can easily change along a gradient. The influence of pollinators will be more important in flowers with specialized pollination systems (monosymmetric flowers) than in actinomorphic generalist flowers.
Significance of developmental processes in flowers
It is widely recognized that subtle developmental changes lead to diversification and evolution (e.g. Endress 1990, 1997; Glover et al. 2015; Hufford 1990, 1995, 1997, 1998, 2003; Ronse De Craene et al. 2001; Takhtajan 1972). In relation to pollination systems, these changes may be responsible to a strong diversification and are generally described as “key innovations” (Endress 2011). These changes can be triggered by different causes, such as genetic mutations, pollinator-mediated selection on flower size and shape, and mechanical forces acting directly on the position, shape and size of organs.
It is understood that different genetic pathways may be associated with highly similar phenotypes, and morphologies may differ considerably despite having similar genetic bases (Endress 2005; Jaramillo and Kramer 2007; Li and Johnston 2000; Specht et al. 2012). Comparative developmental studies may clarify the existence of a greater complexity than expected by comparing mature morphologies. For example in Cleomaceae two different floral developmental pathways lead to seemingly similar mature structures (Patchell et al. 2011).
Despite the fact that floral genes are responsible for the morphology of the flowers, it is recognised that floral genes do not work on their own and can only function within a pre-existing floral context (Causier et al. 2010). Other aspects, such as the spatial disposition of organs, cannot be explained by the ABC model alone and require different explanations, such as biomechanical forces, stochasticity, or other epigenetic influences (e.g. Alvarez-Buylla et al. 2010; Dumais 2007; Green 1999; Kitazawa and Fujimoto 2014, 2015, 2016; Mirabet et al. 2011; Prusinkiewicz and Barbier de Reuille 2010). Studies of the morphogenesis of the shoot apex have demonstrated the importance of pressures in triggering a genetic response acting as a feedback loop in development, such as the orientation of microtubuli in the apex of Arabidopsis or the regulation of auxin production (Gordon et al. 2009; Hamant et al. 2008, 2010; Nakayama et al. 2012; Reinhardt et al. 2003; Robinson et al. 2013; Smith et al. 2006). Phytohormones such as auxin play a crucial role as mediators between genes and tensions or pressures in growing tissues, including the apical meristem. The experimental application of constraints on leaves, inflorescences or flowers (e.g. Dumais and Steele 2000; Green 1999; Green et al. 1996; Hernández and Green 1993; Iwamoto pers. Comm.) demonstrates the effect of external pressures on the development. These pressures are increasingly applied at a smaller, cellular level (e.g. Routier-Kierzkowska and Smith 2013). Computer simulations or statistical analysis have also shown that spatial constraints can autoregulate the development and growth of organs (e.g. Brady et al. 1997; Kitazawa and Fujimoto 2015; Prusinkiewicz and Barbier de Reuille 2010). Several studies on phyllotactic patterns have shown that the causes for pattern formation are purely physical and can be replicated artificially (Green 1992; Green et al. 1996; Douady and Couder 1996; Meicenheimer 1998; Smith et al. 2006). There is an increased consensus that flower development and growth cannot be solely based on genetic regulatory networks, but also depend on self-regulatory aspects such as physical parameters and the influence of external factors often acting in combination with biochemical processes (see e.g. Endress 2005; Kwiatkowska 2008; Meinhardt and Gierer 2000; Newell et al. 2008; Prusinkiewicz and Barbier de Reuille 2010).
In this review I will investigate the significance of two contrasting approaches to the understanding of floral development and evolution:
A historical approach infers that floral development is a reflection of historical (genetic) events: flowers evolve as conserved pre-regulated systems that reflect specific genetic pathways.
A physico-dynamic approach implies that floral development is affected by environmental conditions and stochastic events: flowers evolve as dynamic random systems.
The first approach follows either a neo-Darwinian philosophy, indicating that changes are gradual and the result of a progressive evolution through subtle changes led by genetic mutations, or a saltational view, that a single mutation leads to highly different morphologies (e.g. Bateman and Di Michele 2002). The second approach implies that flower evolution is linked with external factors that act in a haphazard way. In the next chapters I will discuss both perspectives, their significance and limitations.
Floral diversification and evolution are shaped by historical constraints regulating the number of building blocks present in a flower, such as the number of floral parts and their position. These constraints are important in canalizing evolution. The basic morphology of certain floral structures remains strongly conserved in the angiosperms and allows the easy identification of different groups of plants (Endress 2010a, 2011; Leins and Erbar 2010; Ronse De Craene 2010). This conserved morphology is mainly reflected in the number and position of floral organs in the flower and their sequential arrangement. A conserved modularity linked with trimerous and pentamerous flowers allows us to recognize major clades in the angiosperms, such as the monocots and core eudicots or Pentapetalae (e.g. Ronse De Craene 2010, 2016; Soltis et al. 2005; Specht and Bartlett 2009). Initiation and growth of flowers and floral organs is determined by the fact that plants are modular systems that are repeated indefinitely. Therefore the origin and development of floral organs are regulated in a series of developmental steps that arise in a given and predictable sequence.
It is often assumed that features appearing earlier in ontogeny are more conservative in the evolutionary history, and this is also reflected in the gene system (Specht et al. 2012). Tucker (1997, 1999) in her hierarchical-significance hypothesis insisted that processes that happen in early stages of development are more common to larger groups of plants (suprageneric level) than late developmental processes that are shared at the infrageneric level. This also indicates that different ontogenetic stages reflect different levels of similarity and relationship among plants. Several studies tend to support this assumption (e.g. Caris 2013; Hufford 1990, 1995, 1998; Jabbour et al. 2009; Ronse De Craene et al. 2001; Olson 2003; van Heel 1966; Vasconcelos et al. 2017; von Balthazar et al. 2006; Zhao et al. 2012). However, while this criterion is supported in some families, such as Myrtaceae, it is not absolute, as other families show a strong lability in early development, while they are much more stable at later developmental stages (e.g. initiation sequence of organs in Apiaceae: Ajani et al. 2016; Endress 2005; Erbar and Leins 1997—petal and androecium initiation in Caryophyllaceae:; Ronse De Craene 2013; Wei and Ronse De Craene submitted). Other examples are polysymmetric flowers that show an early monosymmetry (discussed by Endress 1999, 2005). Evolutionary changes that happen early in the ontogeny may be important drivers of morphological diversification (e.g. Harrison et al. 1999; Kirchoff 1997; Vasconcelos et al. 2007). Other factors, that are not necessarily genetic, may be responsible for this variability.
Within the historical perspective it is generally assumed that underlying genetic changes are mainly responsible for the regulation of developmental processes (e.g. Buzgo et al. 2004; Erbar 2007; Glover et al. 2015; Irish 2009; Specht et al. 2012). These changes are either progressive (through genetic duplications and neofunctionalisations), or sudden (saltational), where a single genetic mutation leads to strongly divergent morphologies (Bateman and DiMichele 2002; Hintz et al. 2006; Rudall and Bateman 2003; Theissen and Melzer 2007). The historical perspective also implies that phylogeny is crucial in understanding the processes of evolutionary change during development. Ernst Haeckel (1834–1919) famously wrote that “ontogeny recapitulates phylogeny”. This is the idea that developmental stages of an organism reflect evolutionary change. Although largely discredited in its original form, especially in zoology, this concept can be applied in modular plant systems. The idea implies that evolution can only progress through changes in the ontogeny over successive generations, especially in the timing and rate of developmental processes in a descendant relative to its ancestor (heterochrony—also a term first used by Haeckel) (see e.g. Box and Glover 2010; Einset 1987; Hufford 2001a; Li and Johnston 2000; Takhtajan 1972). Heterochrony implies that progressive or dramatic diversification is possible from a more generalized structure by extension (peramorphosis) or by shortening of the developmental sequence (paedomorphosis or neoteny). While heterochrony has been extensively discussed and illustrated with examples (e.g. Li and Johnston 2000), this generally refers to changes between closely related species, not to larger steps in the phylogeny. Change in flower morphology is not restricted to heterochrony (which is related to time). According to Li and Johnston (2000) size, shape, timing, and rate are the four elements that are closely linked during development (Fig. 1a). Allometry (a change in size or shape) is complementary to heterochrony and corresponds to the idea of metamorphosis first proposed by Goethe and later developed by Troll as the principle of variable proportions (Classen-Bockhoff 2001). In addition, homeosis (or the replacement of one developmental pathway by another, or of one part by another; Kirchoff 1991; Ronse De Craene 2003; Sattler 1988) is an additional force in the evolution of plants (see below). Li and Johnston (2000) also made a distinction between heterochrony and heterotopy. While heterotopy (the development of an organ or structure in an abnormal place) can be caused by homeosis (also explained as ectopic gene expression), it is also the direct consequence of heterochrony, which can cause a positional shift of an organ because of alterations of timing of initiation and rate of development. Therefore I think the term heterotopy is superfluous because it is dependent of the context of the flower and mainly caused by heterochrony. I prefer to consider three factors in establishing floral form (Fig. 1b): time, size, and pressure. Rate is defined by heterochrony and is the result of differential growth of organs. In addition, pressure is an additional element not considered by Li and Johnston (2000), responsible for the shape of floral structures (see below).
Hufford illustrated the idea of heterochrony by using several examples from Hydrangeaceae, Loasaceae, and Plantaginaceae (the genus Besseya), where subtle changes in the development of flowers, such as addition, deletion or substitution of developmental stages, lead to evolutionary diversification (Hufford 1998, 2001a, b, 2003). The three processes can occur independently, sequentially, or in unison. Takhtajan (1972) used almost similar terms (prolongation, abbreviation, deviation) to discuss ontogenetic change, but also included neoteny as a combination of deviation and terminal abbreviation.
The emergence of additional structures during development is a frequent factor in floral evolution leading to a progressive change of morphologies. The development of a hypanthium or floral tube through meristem expansion is a fundamental cause for the addition of novel features in flowers (Classen-Bockhoff 2016; Classen-Bockhoff and Meyer 2016; Ronse De Craene 2010). For example, it is through the development of a receptacular spur that Tropaeolaceae becomes distinct from Bretschneidera and Akania, while they are nested in the same clade at the base of the Brassicales and share the same oblique monosymmetry up to mid-development. The development of a median spur in Tropaeolum radically changes the morphology of the flower (Ronse De Craene and Smets 2001; Ronse De Craene et al. 2002). Another example of developmental addition is found in the genus Hypecoum (Papaveraceae), which shares a similar floral development with other Fumarioideae in the development of six stamen primordia, two transversal and four median arranged as two triplets (Ronse De Craene and Smets 1992a). While the marginal stamens of the triplets remain free and develop monothecal anthers in the majority of taxa, marginal stamens of neighbouring triplets fuse to become two large median stamens in the flowers of Hypecoum. The fusion is responsible for highly different floral morphologies between Hypecoum and other Fumarioideae (Fig. 2a, b).
Substitution is far less common than addition, but is a powerful drive in the evolution of flowers. This process has been used as evidence for saltational evolution (Theissen and Melzer 2007), leading to an abrupt change in floral morphologies. Examples of substitution are closely linked with a process of homeosis, where one type of organs is replaced by another one in the flower (e.g. Ronse De Craene 2003; Sattler 1988). Floral developmental evidence is the best indicator for the occurrence of homeosis and numerous examples exist among angiosperms. The most common substitution is the replacement of stamens by petals and vice versa. Most Papaveraceae possess two dimerous whorls of petals and the first stamens are initiated in alternation with the petals (Karrer 1991; Murbeck 1912; Ronse De Craene and Smets 1990). In the wind-pollinated genus Macleaya petals are replaced by stamens alternating with the four legitimate outer stamens (Ronse De Craene 2003; Ronse De Craene and Smets 1990). The genus Sanguinaria represents the opposite trend, where petals replace the outer stamens, leading to eight petals (Lehmann and Sattler 1993). The genus Saraca (Leguminosae) is characterized by a replacement of petals by stamens (Tucker 2000a). In Zingiberaceae part of the androecium is petaloid and represents the main attraction of the flower. Developmental evidence shows that the androecium is partly transformed in petaloid appendages, with outer stamens forming lateral lobes and inner stamens forming a labellum (Fig. 2c; Eichler 1878; Kirchoff 1991, 1997; Kress 1990). A substitution of carpels by stamens is a plausible process in unisexual flowers, when stamens occupy the exact position of carpels in staminate flowers of several Euphorbiaceae (Ronse De Craene 2010).
Deletion or organ suppression is a third process leading to evolutionary change. Deletion is rarely a sudden process, and more often the result of a progressive reduction. A distinction needs to be made between loss, when an organ fails to initiate, and suppression, when an organ is initiated but is subsequently aborted (so called phantom organs). Suppressed organs are intermediates on a spectrum of change between fully functional organs and total loss of organs. This distinction has been widely illustrated for the perianth and stamens of Legumes (e.g. Amorpha: McMahon and Hufford 2005, several genera of Detarieae:; Tucker 2000b, 2001, 2002, 2003). Monosymmetric flowers are often characterized by the loss of petals or stamens and evidence for their existence is found in early stages. In Melianthus a fifth anterior petal is initiated but fails to develop beyond the primordial stage (Ronse De Craene et al. 2001). The best examples of organ suppression and loss are found in the androecium of many Rosids and monocots. Two stamen whorls (termed diplostemony) are highly common and probably represent the plesiomorphic condition for Pentapetalae (Ronse De Craene 2010). However, one stamen whorl is frequently reduced or lost, with the existence of staminodes as an obvious transitional condition (Fig. 2d; Ronse De Craene and Smets 1995a, 1998a, 2001; Ronse De Craene et al. 1998a). This process has evolved repeatedly in parallel in the major clades of Pentapetalae and monocots.
Organs may be absent at maturity without indication that they ever existed, but based on phylogenetic relationships of plants one may assume a process of loss if one species possesses a structure and another does not, even if there is no developmental evidence for this (cf deep homology). In some cases there is an empty space on the floral apex, where the missing organ is expected to arise (e.g. Tropaeolum: Fig. 5d; Duparquetia: Prenner and Klitgaard 2008; Rhynchocalyx:Schönenberger and Conti 2003).
Combination of different developmental processes
Floral evolution is possible by the integration of subtle or more abrupt changes in the development of the flower that become genetically fixed over sequential generations of a species. The fate of organs may become altered and even reversed by a combination of external pressures (mainly from pollinators) and intrinsic changes linked to alterations of genes. Developmental processes, such as addition, substitution and deletion can act in unison or sequentially within different taxa.
In the Zingiberales, different processes act in parallel and lead to a high diversification of the clade. Substitution of part of the androecium into petaloid staminodes leads to a high variability in floral forms. This is accompanied by a deletion of one (Musaceae, Zingiberaceae) or more (Cannaceae, Marantaceae) staminodes (Kirchoff 1991; Rudall and Bateman 2004; Specht et al. 2012).
The Primuloid clade is a perfect example illustrating the evolution of staminodial structures through time. While the ancestral androecium was probably diplostemonous, the antesepalous stamen whorl is either absent or staminodial. Suppression and loss of the staminodes is reflected in different families of the clade (e.g. Primulaceae, Theophrastaceae, Myrsinaceae: Fig. 3d). The suppression of stamens is generally accompanied by a stepwise reduction, starting with a sterilisation of the anther tissue, and resulting in a stublike staminode. Passing a threshold, the staminode might appear as a small stub but will ultimately disappear completely. The genus Samolus represents a good illustration of this trend, with some species having staminodes (Fig. 2e), which are absent in other (Wanntorp and Anderberg 2011). However, in some Theophrastaceae, the deletion process appears to be reversed, as the staminodes are substituted for petaloid staminodes, strongly resembling the petals (Caris and Smets 2004). In Soldanella (Primulaceae) the staminodes are even incorporated into the petal tube as distinct lobes (Caris 2013; Saunders 1937).
In Lamiales, a fifth posterior stamen is often reduced or lost (Donoghue et al. 1998; Endress 1998; Ronse De Craene 2010) and this reduction is linked to shifts in the expression of genes linked to monosymmetry (e.g. Citerne et al. 2010). However, several cases are known where the reductive trend is reversed with the reassignment of the staminode for novel roles in floral biology, such as Penstemon (Walker-Larsen and Harder 2001), Jacaranda (Guimarães et al. 2008) or Scrophularia (López et al. 2016).
These examples illustrate that the genetic potential for different developmental steps may be switched on or off, or even diverted in evolution. A historic perspective of floral evolution makes it possible to understand how different organs are interrelated over time.
The development of flowers implies that floral organs arise in a specific sequence on the floral receptacle. “Initiation precedes appearance” (Endress 1999: S9). Organ initiation appears to be prepatterned, i.e. the site of initiation of organs is predetermined before the actual appearance of primordia. Different prepatterning models have been presented in the past, as a recognition that the position of organs is influenced centripetally by previously initiated organs, such as the reaction model of Wardlaw (1957; in Rudall 2010), or Hofmeister’s rule (Kirchoff 2000), or even the controversial ideas of “hélices foliaires” by Plantefol (1949). Prepatterning models have been reinterpreted in a genetic context, understanding that certain genes (not ABC genes) are responsible for signaling the spatial arrangement of organs in the flower (e.g. Chandler 2014; Running and Hake 2001).
Besides an acropetal prepatterning, it is also understood that the floral apex has a basipetal prepatterning influence (Choob and Penin 2004; Rudall 2010). This is very clear in the influence of carpels on stamen position that can be detected in the floral development (Ronse De Craene 2013). However, it is more likely that size of the floral apical meristem is the major factor influencing a basipetal change (see below). Floral development appears to be regulated by the early establishment of a carpel zone antagonizing the acropetal influence of the outer organs (Rudall 2010). Choob and Penin (2004) demonstrated that the ABC model can only predict the flower organ type, but not the position of organs, and they suggested that a new model be developed based on genetic regulation of space. Meristem size increase, organ initiation, organ spacing, and floral meristem determinacy appear to be regulated by the activity of several genes acting independently (Running and Hake 2001).
The current models of prepatterning only consider a genetic approach. In the prepatterning process of flowers it is debatable whether gene expression is precursory to the development of the flowers or the result of mechanical pressure acting at the onset of floral meristem specification. Although higher auxin concentrations have been associated with floral meristem initiation (e.g. Reinhardt and Kuhlemeier 2002), the position of the FM could not be clarified (Chandler 2014). In the case of Arabidopsis and Brassicaceae in general, a subtending bract is absent although the floral meristem position appears to be prepatterned in the axil of the non-existing bract, with the abaxial sepal replacing the bract in early development (Endress 1992; Erbar and Leins 1997). The bilateral shape of the flower appears to be correlated with the influence of transversally inserted carpels, also influencing the size and position of the transversal stamens. A smaller size of these stamens or their possible loss in some Brassicaceae (e.g. Lepidium: Bowman and Smyth 1998; Endress 1992) appears to be linked with mechanical pressure, caused by carpels and lateral sepals, and not with a basipetal patterning as suggested by Choob and Penin (2004). Although the gene interactions may roughly affect meristem size and organ spacing, the arrangement and number of floral organs is ultimately linked with a combination of mechanical forces and biochemical signals during development (Chandler 2014). Kitazawa and Fujimoto (2016) demonstrated that the transition between a helical initiation pattern and a whorled pattern can be artificially simulated by changing the strength of mutual inhibition created by developing primordia. The establishment of the floral morphospace is strongly linked to the prepatterning process, as forces operating at different sites of the meristem are intimately interrelated.
Although genetic systems regulate much of the identity of flowers and their organs, other fundamental developmental processes, such as the geometry of the floral meristem, the order of initiation and position of organs, are driven by physical factors. Experimental physical interventions on floral meristems or computer simulations have demonstrated that space availability is the most important factor guiding development (see above). Classen-Bockhoff (see Classen-Bockhoff 2016; Classen-Bockhoff and Bull-Hereñu 2013) demonstrated a fundamental difference between vegetative and floral systems, as floral systems are characterized by two processes absent in the vegetative parts of the plants: loss of apical growth and a resulting meristem expansion. This spatial shift is responsible for a change in the geometry of the floral meristem and results in a wide range of evolutionary possibilities that are not genetically programmed. Floral evolution results from the subsequent genetic fixation of specific shifts in developmental parameters regulated by spatial constraints. Changes in floral morphology are induced by three criteria: timing of initiation, external and internal pressure, and alterations of size in the floral meristem (expansion and regression) (Fig. 1b).
The importance of time (heterochrony)
As discussed above, shifts in the timing of initiation of floral organs are a fundamental cause of flower diversification (see Li and Johnston 2000). In the shoot apical meristem (and also floral meristem) the initiation of organs creates an auxin shift that prevents the initiation of other primordia within a radius around the first-formed organ (see Hamant et al. 2010; Reinhardt and Kuhlemeier 2002; Traas 2013). Time is a crucial factor, as subtle changes in the sequence of initiation of successive organs will change their spatial relationships and alter their interaction with neighbouring organs (in size and shape). A delay or acceleration of organ initiation may further lead to organ reduction and loss, because of increased spatial constraints. This becomes important in Pentapetalae with well differentiated whorls of organs that behave more or less independently of each other, under different genetic influences.
I will explore the process of time for three important developmental processes in flowers, viz. phyllotaxis, the occurrence of common primordia, and obdiplostemony.
Phyllotaxis refers to the insertion of phyllomes (leaves or floral organs) along an axis. The phenomenon of phyllotaxis has been extensively studied in the past with various interpretations and models to explain the phenomenon, including physico-mechanical causes (for a review see Rutishauser and Sattler 1985). Phyllotaxis is a good illustration of the model of Prusinkiewicz and Barbier de Reuille (2010), where shape, space, and time interact in the creation of floral patterns. Spiral phyllotaxis is characterized by organs arising with a continuous plastochron and equal divergence angles between succeeding primordia (e.g. Endress 1987, 2011; Endress and Doyle 2007). The consequence of a spiral phyllotaxis is the arrangement of organs in tight parastichies following Fibonacci numbers. Fibonacci series are linked to the size of the individual organs in relation to the size of the apical meristem. A small floral meristem with larger organs will result in a lower Fibonacci number (1/2, 1/3, 2/5; Figs. 4b–e, 5a), while a larger floral meristem with smaller organs will usually be associated with a higher Fibonacci number (3/8, 5/13, 8/21). Within flowers these series may change between organ categories, as has been illustrated for Ranunculaceae (Endress 1987; Schöffel 1932). Kitazawa and Fujimoto (2016) demonstrated that intraspecific variation in organ number is statistically coherent among species of Ranunculaceae and Asteraceae, showing a clear phenotypic stability.
Whorled phyllotaxis is characterized by organs having equal divergence angles and a very short plastochron within a whorl, followed by a longer plastochron in the transition to the next whorl (Endress 2011). A spiral and whorled phyllotaxis are easily interchangeable and recent discussions have favoured an origin of spirals from whorls in the phylogeny of angiosperms (Doyle and Endress 2011; Sauquet et al. 2017). However, the most plausible developmental evidence points in the opposite direction. The increase in size of the perianth parts in Magnoliales has been interpreted as causal for a shift of a spiral phyllotaxis to a whorled phyllotaxis (see Erbar and Leins 1983, 1994). Larger perianth primordia reduce the space available on the floral meristem and lead to a reduction of the number of organs that can initiate with a same plastochron. The reason for a shift of spiral to whorled phyllotaxis is linked to a rhythmic interruption of a continuous plastochron, leading to a sequential alternation of a longer and shorter plastochron. Because of the interruption of the sequence, organs within a time-series have the opportunity to grow in size and influence the position of the organs in the next series. As a result the organs of the next series will alternate with the previous series and whorls are borne. A trimerous merism is the most common whorled pattern in basal angiosperms and monocots and appears to be structurally extremely stable (Kubitzki 1987). The transition between larger perianth parts and the much smaller stamens leads to an arrangement of outer stamens in pairs (or in the spaces between the larger perianth organs; Fig. 3a, b). Numerous examples of basal angiosperms, Alismatales and Ranunculales illustrate the arrangement of stamens in alternating whorls of paired and unpaired stamens (Endress and Armstrong 2011; Leins and Erbar 1985; Leins and Stadler 1973; Ronse De Craene and Smets 1993b; Ronse De Craene et al. 2003; Sattler and Singh 1978; Staedler and Endress 2009; Tucker and Hodges 2005). Whorled phyllotaxis leads to the creation of stable systems with increased synorganisation and lies at the base of the diversification of angiosperms (Endress 2011, 2016).
With a higher number of small organs (e.g. by a secondary stamen increase or by the development of smaller leaf primordia on vegetative shoots) the phyllotactic pattern is easily disturbed (e.g. Papaveraceae, Annonaceae, Ranunculaceae, Fabaceae). The result is the formation of complex whorls or irregular spirals with non-standard Fibonacci numbers (e.g. Endress 1990; Rutishauser 2016; Staedler and Endress 2009; Zhao et al. 2012). Rutishauser (2016) demonstrated that in Acacia a regular phyllotaxis pattern of vegetative shoots becomes easily disrupted by a reduction in size of leaf primordia, leading to whorled and irregular initiation patterns. A disruption of the phyllotaxis is particularly clear in Annonaceae, where an increase of stamens and carpels (with much smaller organ sizes relative to the floral meristem) leads to pronounced disruptions in the arrangement of stamens and carpels (e.g. Leins and Erbar 1996; Rutishauser 2016; Xu and; Ronse De Craene 2010; Fig. 3a, b). However, in all Annonaceae outermost stamens are always alternating with the inner petals, and this regularity is maintained in all species. Irregularities of further stamen series arise by the strong compression of the floral apex linked with small primordia.
Vegetative systems are relatively stable with a continous plastochron (reflected in a spiral leaf arrangement), or with a regular alternation between shorter and longer plastochrons, leading to the frequency of a decussate leaf arrangement. In flowers the inflation of the floral meristem leads to a shift in the phyllotaxis by a change of the plastochron. This transition is usually mediated by bracts, bracteoles or a calyx. The decussate arrangement of the leaves is rarely continued in the tetramerous or dimerous calyx of terminal flowers. More generally there is a transition of a 1/2 phyllotaxis to a 2/5 phyllotaxis. The transition from decussate to spiral has been shown to be linked to different factors, such as asynchrony between two decussate primordia (Douady and Couder 1996), or an expansion of the apex linked to this asynchrony (Meicenheimer 1998). Simulation studies (Smith et al. 2006) have demonstrated that different phyllotactic patterns can be created by changing the inhibition functions which determine the sequential initiation of primordia during development.
In conclusion, changes in phyllotaxis are subtle but with a great impact on floral evolution. The causes for these changes are physico-dynamic by the interaction between time and space with no immediate genetic causes. There is a clear inverse relationship between the size of the floral meristem and the size of organ primordia. With greater size of the meristem and smaller organs, organ initiation switches to an irregular pattern; with smaller size of the meristem and larger organs initiation is clearly following spiral or whorled phyllotactic patterns. This follows the same principle as for changes of merism that are affected by the relationship between the size of primordia versus the floral meristem (Ronse De Craene 2016). A shift from a spiral to a whorled initiation by the interaction of two contrasting parameters of initiation, viz. inhibition and mutual repulsion of organs, is also the obvious cause for the stable pentamery of Pentapetalae, as supported by mathematical modeling (Kitazawa and Fujimoto 2015).
Common stamen-petal primordia
The importance of time in floral evolution is clearly demonstrated in the development of common primordia, which are linked to a whorled phyllotaxis. Common primordia can be seen as an ontogenetic division of a single primordium into two daughter primordia. This phenomenon has been studied and discussed repeatedly in the past as dédoublement (reviewed by Ronse De Craene and Smets 1993a) and was mainly seen as the potential of organs to increase in number. This may be plausible when the two daughter primordia are equal, but this is generally not the case. Common primordia can divide tangentially (within whorls) or radially (affecting superposed whorls or organs). A tangential subdivision is observed in the androecium and is often responsible for a stamen increase (as paired stamens) (see Ronse De Craene and Smets 1995b, 1996). A radial subdivision of a common primordium may occur between flower and bract, two superposed stamens, or stamen with sepal or petal. The latter is the most commonly observed in angiosperms and is known as common stamen-petal primordia.
The most plausible interpretation for common stamen-petal primordia is linked with time. Petals are often delayed in their development, only reaching a larger size prior to anthesis (Ronse De Craene 2008; Figs. 3d, f, 4g, 5d, e, h). The delay in development usually happens after initiation of the petal primordia. The petal whorl is crammed between two organ series that are not delayed in their development, considerably delaying its further development. However, the delay may also become such that the petal primordium is virtually overtaken by the next organ series. As a result the petal whorl becomes absorbed in the superposed stamen whorl and both develop into a common primordium. Sattler (1962) observed the variation of common primordia in the Primuloid clade of Ericales. After the division of the common primordium, petals generally grow into full organs (Fig. 3c, d). However, the delay of petal initiation may also be read in a sequence leading to petal reduction and loss. Pushing the delay further leads to greater competition with the surrounding organs and a shift to the ultimate disappearance of petals. Even in the primuloid clade there are cases of apetaly, such as Glaux maritima with no sign of any petal primordia (e.g. Caris 2013). In Caryophyllaceae, petals may arise independently, but they are often delayed and linked with antepetalous stamens in common primordia, or missing altogether (see Ronse De Craene 2013; Ronse De Craene et al. 1998b; Wei and Ronse De Craene submitted).
In Alismatales petals are often closely positioned with stamen pairs and have been described as complex petal-stamen (CA) primordia (Sattler and Singh 1978). Whether these CA primordia represent discrete entities or are the consequence of spatial (phyllotactic) arrangement of organs in the flower has been considerably discussed (Charlton 1999, 2004; Ronse De Craene and Smets 1995a; Sattler and Singh 1978). However, a recent investigation into the development of petaloid Alismatales (Iwamoto et al. 2018) demonstrates that the occurrence of so-called stamen-petal primordia is linked to the delay in initiation and growth of the petal primordia becoming spatially linked with the paired outer stamens.
Ronse De Craene et al. (1993) described the phenomenon of common stamen-petal primordia as a case of negative dédoublement, implying that a dividing primordium does not lead to the creation of an additional organ, but to its absorption and loss. This clearly illustrates how timing of organ initiation affects the fate of flower diversification.
Obdiplostemony and loss of whorls
Obdiplostemony is the androecial configuration consisting of two stamen whorls, where the arrangement of whorls is inversed, as the outer stamen whorl is inserted opposite the petals (Ronse De Craene and Bull-Hereñu 2016). Obdiplostemony represents a disruption in the diplostemonous flower with a regular alternation of whorls (Figs. 3f, 5a, b). The causes for obdiplostemony are diverse and are related to a differential development of the antesepalous and antepetalous stamens (including petals). Changes in the growth rate between the antepetalous and antesepalous sectors lead to an obdiplostemonous arrangement, and this is strongly linked with variations in the timing of initiation of organ whorls. Either the alternisepalous sector is delayed leading to the ultimate loss of the antepetalous stamens, or the antesepalous stamen is reduced, ultimately leading to obhaplostemony. Reduction of the antepetalous stamens may be linked with common stamen-petal primordia and a reduction of petals (Fig. 3e, f). Ronse De Craene and Bull-Hereñu (2016) discussed the different developmental pathways that cause obdiplostemony in the flower of Pentapetalae and demonstrated that fundamental shifts in the configuration of the androecium are intimately linked with the timing of primordium initiation. The loss of either the antepetalous or the antesepalous stamen whorl represents fundamental changes in the morphology of the flowers that became consequently genetically stabilized.
Shifts in the timing of initiation of stamen whorls may lead to an inversed initiation sequence: the antepetalous stamens emerge before the antesepalous whorl, and the antesepalous stamens arises centrifugally externally to the antepetalous whorl. This pattern is responsible for an obdiplostemonous stamen arrangement in mature flowers. Several examples have been presented and discussed by Rudall (2010, 2011) and represent cases of heterochrony. Causes for the centrifugal initiation of whorls are linked with availability of space on the floral apex and can indicate a pattern of evolution in certain clades. This pattern of initiation is observed in certain Commelinaceae (Hardy and Stevenson 2000). The outer antesepalous stamens emerge low on the slopes of the floral apex and undergo more pressure from the sepals. In the related Poales there is a strong tendency for the antesepalous stamen whorl to become reduced or completely lost (Remizowa et al. 2010). Similarly in Fumarioid Papaveraceae the second stamen whorl arises centrifugally to the four first initiated stamens (Ronse De Craene and Smets 1992a; Fig. 2a). This pattern of development is related to the dimerous merism, but is also linked to stronger pressure in the median radius of the flower (see below).
These three examples demonstrate that changes in the timing of organ initiation greatly affect the progression of floral evolution. It is indubitable that these changes operate in conjunction with pollinator pressure and genetic shifts, although very little is known about what causes heterochrony to happen. Spatial constraints or external pressures (see below) are probable triggers in the changing relationships of organs in the floral meristem. The consequences of altered timing of primordium initiation can be evolutionarily translated in reductions and losses of organs, often with a high phylogenetic significance. It is no coincidence that families with a delayed petal initiation (with common stamen-petal primordia) often contain species with reduced or absent petals (see Ronse De Craene 2010; Ronse De Craene and Smets 1995b).
The importance of pressure
Developing floral meristems undergo various pressures caused by organs external to the flower or as part of flowers. Organs often show pressure marks, reflecting their development in a confined space and an economic use of this space in the bud (Endress 2008; Fig. 4c, e). Space limitations may mould the shape of organs in buds, such as the development of a five-angled stigma in bicarpellate Apocynaceae (Endress 2016). It is possible that the extrorse orientation of anthers in the third stamen whorl of Lauraceae is induced by space restrictions in the small bud. However, these are late-developmental pressures that affect already formed organs without influencing their position an initiation. Pressures within buds will alter the shape of organs, but will also influence their position and induce further changes, such as losses or duplications of organs (see below).
Pressures on the developing flowers are related to the prepatterning process (see above), which is exercised centripetally (from the external organs) or centrifugally (from the floral apical meristem). The genetic basis for these antagonistic influences has been presented by Choob and Penin (2004), but non-genetic factors are as important in transforming the structure of flowers sometimes preceding a genetic stabilization (see also Schwander and Leimar 2011). Pressure of organs external to the flower (bracts and bracteoles) or outer organs in the flower (sepals and/or petals) cause a strong physical constraint that can influence or disrupt the shape, position, and size of the next-formed organs, and also induce the loss, increase or chaotic initiation of inner organs. This pressure operates centripetally.
Bracts and bracteoles
Bracts are extremely important in shaping and orienting the floral meristem (Chandler 2014; Kwiatkowska 2008). Endress (1999) refers to the interaction of an acropetal gradient caused by the main shoot, and a basipetal gradient caused by the bract, acting in opposite directions on the floral primordium. These opposing forces are responsible for unidirectional development of organs and the symmetry of the flower. Even in the Brassicaceae where a bract is absent at maturity, a residual bract primordium is present and influences the onset of the flower initiation (Kwiatkowska 2008). The adaxial activation of cycloidea genes and the dorso-ventral polarity of flowers is also induced by the presence of bracts (Chandler 2014), although this is generally not considered in genetic studies. CYC-like genes are present in actinomorphic flowers as well, as these genes were acquired very early in angiosperm phylogeny and maintained in certain clades while lost in others including Arabidopsis (Citerne et al. 2010; Zhang et al. 2013). Several studies have demonstrated the correlation between expansions or regressions of the CYC-like gene expression and changes in the symmetry of flowers (e.g. Gesneriaceae: Wang et al. 2010; Malpighiaceae:; Zhang et al. 2013; Caprifoliaceae:; Howarth et al. 2011). However, these studies have not investigated the developmental relationship of flower primordia with axis and enclosing bract that may influence these shifts. The presence of peloric terminal flowers, which are actinomorphic with a variable merism (e.g. in Digitalis) demonstrates this clearly, and relates to genetic and mechanical factors shaping the flower (Rudall and Bateman 2003). It is possible that the lack of subtending bracts allows for a different merism in the terminal flower compared to lateral flowers of cymose inflorescences (e.g. Adoxa, Ruta: Ronse De Craene 2016).
A strong pressure of bracts in early stages of flower development can influence the shape of flower meristems as to become elliptical or zygomorphic, eventually leading to disymmetric flowers (Endress 1999). Strongly monosymmetric flowers generally show a unidirectional initiation sequence of floral organs, with the first organs arising in areas with least pressure, either adaxially or abaxially. Flowers with an elliptical shape develop in a variety of groups with variable causes, such as Eupteleaceae (Ren et al. 2007), Buxaceae (von Balthazar et al. 2006), Winteraceae (Doust 2001, 2002; Doust and Drinnan 2004), Bataceae (Fig. 4a; Ronse De Craene 2005); Rubiaceae (Rutishauser et al. 1998); Begoniaceae, or Papaveraceae (Ronse De Craene and Smets 1990; Fig. 1a, b). Pressures are strongest on lateral flowers in compound inflorescences, leading to elliptical shapes and a dimorphism between lateral and terminal flowers (e.g. Drimys: Doust 2001; Adoxa:; Erbar 1994; Trochodendron: Endress 1986). Developmental studies in Cyperaceae have demonstrated that bracts have a strong pressure on the median side of the flowers, leading to a delay of median organs, or loss of median tepals and stamens (Richards et al. 2006; Vrijdaghs et al. 2005). In Salicornia (Chenopodiaceae) lateral flowers of dichasia are displaced almost parallelly with the main flower due to the pressure of the subtending bract (Beer et al. 2010). The pressure of outer bracts may also influence the sequence of initiation of floral organs. This has been illustrated for Astragalus (Leguminosae) by Naghiloo et al. (2012), who found four different patterns of sepal initiation linked to the pressure of outer bracts, as sepals arise preferentially where the pressure is lowest. The strong influence of the development of bracteoles on the flower shape has been described by Tucker (2000b, 2001, 2002), who recognized the existence of two kinds of bracteoles in tribe Detarieae of Caesalpinoid legumes. The “circular type” consists of small bracteoles and five sepals are usually present and arise helically. The “omega-type” is characterized by very large bracteoles and the first sepal arises in abaxial position. The species with omega-type buds are generally characterized by loss of perianth organs. Space constraints caused by the bracteoles are obviously responsible for the arrangement and loss of organs. Similarly in Duparquetia a fifth sepal fails to initiate (Prenner and Klitgaard 2008). In some cases, such as Colophospermum a distinction between bracteoles and lateral sepals becomes debatable (Krüger et al. 1999). Bracteoles may also become integrated in the flower, leading to a petaloid transformation of lateral sepals and the occasional loss of the lateral petals in Polygalaceae (Bello et al. 2010; Prenner 2004b). In Petiveria (Rivinaceae) bracteoles are much smaller compared to other members of the family. As a result the four sepals become arranged diagonally and linked with an absence of a well developed petiole, the symmetry of the flower becomes strongly distorted (Ronse De Craene and Smets 1991a). Decussately arranged bracts subtending terminal flowers, or bracteoles of lateral flowers have a mediating role in the transition to the spiral initiation of the calyx (Endress 1994). Their position is also essential in guiding the position of the first two sepals, which become positioned in a median position. The direction of initiation of the bracteoles (clock- or counterclockwise) is often a random process (except in Asterids) with effect on the direction of initiation of the calyx (Endress 1999).
A special case of external pressures affecting the shape of flowers is an involucre, or the presence of two median outer organs tightly enclosing flowers. An involucre is found in several angiosperms, where it has been variously described as sepals or bracts, reflecting different interpretations or origins (e.g. Winteraceae: Doust and Drinnan 2004; Papaveraceae, Begoniaceae:; Ronse De Craene and Smets 1990; Bataceae:; Ronse De Craene 2005, Fig. 4a; Theligonum in Rubiaceae:; Rutishauser et al. 1998; Montiaceae:; Dos Santos and Ronse De Craene 2016). Flowers with an involucre generally have flowers with an elliptical shape compressed between the two median lobes, and organs preferentially develop in transversal position where pressure is less strong. The compression is often increased by a combination of involucre and subtending bract (e.g. in Tasmannia: Doust and Drinnan 2004). An involucre is a synapomorphy for the portulacaceous clade of Caryophyllales, where it is described as bract-derived (Dos Santos and Ronse De Craene 2016; Ronse De Craene 2013). In most Portulacinae with an involucre, initiation of lateral sepal lobes precede the median sepals that are delayed in their development (Ronse De Craene 2013; Vanvinckenroye and Smets 1996). In Lewisia, apart from a delay in initiation of median sepals, the involucre is responsible for an increase in median sepal and stamen numbers by extending the shape of the floral meristem (Dos Santos and Ronse De Craene 2016). A similar chaotic increase of stamens is observed in Theligonum (Rubiaceae) linked to the elliptical shape of the floral primordium of staminate flowers (Rutishauser et al. 1998).
In several families the reduction or loss of bracteoles leads to changes in the developmental sequence of flowers, including the precocious initiation of the transversal sepals (e.g. Ribes, Grossulariaceae: Eichler 1878; Streptocarpus, Gesneriaceae: Haston and Ronse De Craene 2007; Pyrola, Bruckenthalia, Ericaceae:; Caris 2013). In core Brassicales, where an absence of bracteoles appears to be plesiomorphic, lateral sepals often initiate before the median sepals (Ronse De Craene and Haston 2006). In Leguminosae there are several reported cases of bracteoles being initiated and progressively becoming aborted during development (Prenner 2004a). Except for most Mimosoideae the odd petal (i.e. banner) occupies an adaxial position, contrary to the majority of Pentapetalae. This unusual inversed position may be caused by the strong pressure of bracteoles in flowers, pressing the first sepal in abaxial position and leaving more space for the development of the petals on the adaxial side of the flower. Where bracteoles have become reduced or lost this floral arrangement is maintained and is genetically stabilized. Buckhari et al. (2017) observed that a median adaxial petal has been derived at least 30 times from zygomorphic flowers with a median abaxial petal without analyzing the developmental causes for this.
Outer floral organs
In flowers the calyx has the strongest influence on the inner organs, as it is usually the organ whorl that encloses and protects the young bud. Putting cases with involucral sepals aside (see above), a spirally initiated calyx will be mostly imbricately arranged and cause pressure on the next whorls (Endress 1994; Ronse De Craene 2010). Heterochrony is an important factor in changing the degree of pressure of the perianth organs. With a continuous helical sequence the outer sepals will be much larger than the innermost sepals, and this is often reflected in the quincuncial arrangement of sepals in flowers. In Cistaceae heterochrony changes the size of sepals in mature flowers, influencing the arrangement of petals (Saunders 1936; Ronse De Craene 2010). In some families of Pentapetalae heterochronic shifts in the plastochron of perianth initiation leads to spiral sepals and petals influencing the onset and size of inner organs (e.g. Pentaphyllaceae: Fig. 4b, c; Zhang and Schönenberger 2014; Theaceae:; Tsou 1998; Clusiaceae: Fig. 4d, e:; Gustafsson 2000; Hochwallner and Weber 2006). In these cases bracts and perianth are generally indistinct and exercise equal pressure on the rest of the flower. Even in relatively large buds, the proportion taken up by the perianth exceeds the rest of the flower (Fig. 4b–e.). Strongly developed, massive sepals, which take on an early valvate aestivation, can mould the flower in the shape of a starfish, as there is less pressure on the meristem in the area between the sepal lobes. As a result petal sectors tend to develop faster and are shifted more externally of sepal sectors, leading to a condition resembling obdiplostemony. Stamens tend to proliferate along the arms of the starfish, while the inner stamens are compressed by the pressure of sepal lobes, become staminodial, or disappear. Carpels, when isomerous, are positioned in the petal sector where more space is available. Such development is found in Elaeocarpus (Elaeocarpaceae) and Byttnerioideae of Malvaceae (Ronse De Craene and Bull-Hereñu 2016). A shift to a valvate aestivation is caused by heterochrony and alters the influence of the calyx on the inner whorls. In Mimosoid legumes and Myrtaceae imbricate sepals become valvate by equalization of the sepals lobes (De Barros et al. 2017; Ramirez-Domenech and Tucker 1990; Vasconcelos et al. 2017). A shift to a valvate aestivation can have the opposite effect in reducing the size and pressure of the calyx. Reduced pressure of the sepals caused by their valvate arrangement can also influence the shape of the petals, which become elaborate as they are less obstructed in their development (Endress and Matthews 2006). In addition, a valvate calyx favours an increase in floral merism (Ronse De Craene 2016).
External and internal pressures can be subtle, influencing the sequence of initiation or disrupting the position of neighbouring organs. However, increased pressure has the effect of throwing the regularity of flowers in disarray. Increased pressure is an important cause of heterochrony of organs immediately affected by the force, leading to a disrupted development. Tucker (1988) implied that organ loss in one whorl leads to a disruption in the next successive whorls. This has also been described as an irregular or chaotic phyllotaxis by Endress (1994), emphasizing the consequence of reduction of organs (mostly a perianth) on the structure and regularity of the androecium. The pressure of bracts can induce an increase and chaotic initiation of perianth parts and stamens, as is visible in the variability of the flower of Lewisia in Montiaceae (Dos Santos and Ronse De Craene 2016). There is a close correlation in the pressure of bracts and a development of bristle-like appendages in lieu of the calyx. Compared to calyx lobes, bristles develop in high numbers after inner organs have been formed and their number and position is variable (e.g. Cyperaceae: Vrijdaghs et al. 2005; Asteraceae:; Leins and Erbar 1987; Caprifoliaceae:; Naghiloo and Classen-Bockhoff 2017).
Centrifugal influence of the ovary
The central floral meristem terminates the development of the flower. As presented before, there is a strong prepatterning effect of the gynoecium in flowers, as carpels tend to influence the initiation of the androecium. The influence of the ovary is centrifugal, but its effect is mainly spatially induced by the size of the floral apical meristem. Changes in the space occupied by the gynoecium can influence the number and position of the stamens by exercising pressure in the immediate proximity of the carpellary lobes. Isomerous flowers are relatively rare in angiosperms (see Ronse De Craene 2016), and carpels generally follow Hofmeister’s rule in an alternation with the previous stamen whorl. The occurrence of obdiplostemony has been linked to the position of isomerous carpels in antepetalous position (Endress 2010b), although the main cause is heterochrony (Ronse De Craene and Bull-Hereñu 2016; see above). In the case of a trimerous gynoecium, the forces exercised by the carpels will determine the stamen arrangement, initiation, and reduction and become a constant factor in evolution. This pattern is especially clear in Caryophyllales, where we find a strong correlation between stamen and carpel numbers, with stamens fluctuating from ten to one (Ronse De Craene 2013; Ronse De Craene et al. 1998b). In Frankenia (Frankeniaceae) the three carpels have a guiding influence on the number, size and position of the six stamens, which become arranged in two trimerous whorls of different length, although they arise as a combination of a complete antesepalous whorl and a single antepetalous stamen (Fig. 4f, g; Ronse De Craene 2010). In the pentamerous Caryophyllaceae flowers a reduction to three carpels is always correlated with an inversed antesepalous stamen initiation, and the last initated stamens are those situated in the radius of two of the carpels. Limited space often results in loss of these two stamens (Ronse De Craene 2013). The remaining antesepalous stamens always alternate with the carpels, where there is sufficient space for their development. This arrangement is constant and is also found in several other Caryophyllales (Ronse De Craene 2013; Wei and Ronse De Craene submitted). In Montiaceae the ovary also has a strong influence on the arrangement and position of stamens (Dos Santos and Ronse De Craene 2016; in prep.), as upper stamens alternate with the carpels. A combination of the external pressure of the involucre and internal pressure of carpels causes a shift in the position of the androecium and loss of stamens. When carpel numbers are further reduced to two units, the reduction of stamens often follows a similar pattern.
Pressure is only one aspect of factors influencing flower development, but it has a great influence on the timing of initiation of organs, and the space availability in the flower.
The importance of size
Meristem expansion and regression are linked to spatial changes on the floral meristem. Such alterations of space represent fundamental factors favouring floral evolution by providing new opportunities for floral variation not found in vegetative systems (Classen-Bockhoff 2016). Although most flowers have a limited growth potential that appears to be genetically fixed, floral systems have the potential for change. In closed synorganized systems, size changes (allometry) may affect all organs in equal proportions, leading to smaller or larger flowers (e.g. Brugmansia: Endress 1994), or there is a co-variation in organ differentiation affecting different organs in different proportions (e.g. Chinga and Pérez 2016; Endress 2016). However, in other (less synorganized) systems flower organ number and size can strongly fluctuate (Fig. 3b). Experimental studies in Papaver (Papaveraceae) have demonstrated that floral size and floral organ number are directly linked to available nutrients (Murbeck 1912). Expansion of a terminal meristem is also responsible for an increase of floral parts in Myosurus (Ranunculaceae). However, these floral systems relate more strongly to vegetative systems with extended growth of a shoot apical meristem. The apical expansion is limited in flowers of Pentapetalae, which lack the central zone with stem cells of the shoot apical meristem (Kwiatkowska 2008); instead floral meristems can expand by internal cell division (Classen-Bockhoff 2016). Linked with the initiation of floral organs, the expansion of the meristem provides multiple opportunities for diversification by creating space for additional organs. Four different possibilities can be recognized, implying that floral meristems have the potential to expand or regress in size throughout the evolution of flowers:
A vertical expansion creates space for the initiation of novel structures (hypanthia and floral tubes, stamen increases, inferior ovaries, coronas and appendages);
A horizontal expansion creates space for an increase in the number of organs within whorls (merism increase, stamen and carpel increase);
A horizontal regression leads to limited horizontal space on the floral meristem resulting in a lower merism, reduction of carpel number, or fusion and loss of organs;
A vertical regression leads to limited vertical space on the floral meristem resulting in fusion of opposing organs (common stamen-petal primordia) or a loss of whorls.
In the literature there is a common misconception of fusion and fused organs and this is ultimately linked to the concepts of “congenital” and “postgenital” fusion (e.g. Endress 1994, 2006; Ronse De Craene 2010; Sattler 1978). Hypanthial growth or expansion of the receptacle is often confused with fusion (as in sympetaly connecting stamens and petals in a common tube). I believe that any “congenital” fusion is linked to hypanthial growth (often described as receptacular or zonal growth: Ronse De Craene 2010). This fusion is especially clear in carpels that may arise as free primordia and remain free or may fuse postgenitally (e.g. Rutaceae, Rosaceae), free carpel primordia that become connected by a basal rim, leaving the original carpels as stylar or stigmatic lobes (e.g. Rosaceae), or fully connected carpellary tissue without clearly distinctive primordia (e.g. Primulaceae). Carpel fusion will be ultimately linked with the relationship between the central meristem and the carpel primordia being initiated. It is a spatial matter linked with size.
A change in merism is strongly connected to the expansion of the floral meristem (Ronse De Craene 2016). Merism increase is possible when the size of the organ primordia decreases relative to the floral meristem. An increase in merism tends to be correlated with a reduction in size and valvate aestivation of the outer organ whorl. There is a treshold in the size of primordia beyond which they become rearranged in two alternating whorls. A combination of increase of the floral meristem linked with organ size reduction leads to high jumps in the merism of the flower (e.g. Paris, Sempervivum, Fig. 5a, b: Ronse De Craene 2016). An increase in merism is generally unstable, with easy transitions between tetramery, pentamery, and hexamery. However, in several cases a high stability is achieved partly by a rearrangement of the organs in two alternating trimerous whorls (e.g. Fagaceae, Polygonaceae), or by a stabilization of the size of organ primordia relative to the floral meristem (e.g. Dryas octopetala in Rosaceae, Lythraceae). Naghiloo and Classen-Bockhoff (2017) also demonstrated the close correlation between meristem size and organ size in Dipsacoideae of Caprifoliaceae.
The horizontal expansion of the floral meristem is also responsible for carpel multiplications (Endress 2014; Ronse De Craene 2016). Carpel multiplications are possible when the fusion of individual carpels is not pronounced and the apical meristem has a considerable size. There is a direct correlation between the size of carpels, the width of the floral meristem and the number of carpels formed. This correlation has been quantitatively recorded in Medusagyne (Ronse De Craene 2017) and Eucryphia (Bull-Hereñu et al. 2018). The limited space on the floral apex is responsible for the rarity of additional carpel whorls (e.g. Punica, Rutaceae, Malvaceae-Ureneae) (Ronse De Craene and Smets 1998b; van Heel 1978). An increase in the number of carpels is generally accompanied by an increase in the number of stamens (Fig. 5e, g; Ronse De Craene 2016). This increase may also involve a vertical expansion of the floral meristem. In these cases stamens develop on a ring meristem in association with an increase of carpels in a single whorl (e.g. Fig. 4e, g; Papaver: Ronse De Craene and Smets 1990; Nelumbo:; Hayes et al. 2000; Medusagyne:; Ronse De Craene 2017; Eucryphia:; Bull-Hereñu et al. 2018; Victoria, Nymphaea:; Schneider et al. 2003; Actinidia:; van Heel 1987; Couroupita:; Endress 2014).
Within legumes the number of carpels is strongly regulated by the size of the floral receptacle. While a single carpel is the plesiomorphic condition, several exceptions are known with carpel increases especially in Mimosoideae. Carpel increase is linked with polyandry and a larger floral receptacle (Paulino et al. 2014).
A distinction between vertical and horizontal expansion appears artificial, as meristem expansion occurs in three dimensions. Contrary to horizontal expansion affecting mainly the merism of a whorl, or by extension all whorls of a flower, vertical expansion is mainly caused by hypanthial growth. The development of a hypanthium is a key-innovation in angiosperm evolution, allowing the flower to develop infinite possibilities of shape and form (Ronse De Craene 2010). Hypanthial growth is variously described in the literature as zonal growth (Sattler 1978), or intercalary meristematic growth (Basso-Alves et al. 2017; Leins and Erbar 2010). Hypanthial growth triggers diversity in floral form by creating opportunities for addition of structures, fragmentation of primordia, and shifts away from organ determinacy. In several families, hypanthial growth leads to increased diversification and variable stamen numbers, such as floral morphs of Rosaceae (Fig. 5c–e, Kania 1973), Hydrangeaceae (Ge et al. 2007; Hufford 1998, 2001b), and Myrtaceae (Ronse De Craene and Smets 1991b; Vasconcelos et al. 2015, 2017), or the building of elaborate appendages such as coronas in various Passiflora species (Classen-Bockhoff and Meyer 2016). Shifts in ovary position in Melastomataceae are attributed to extended growth affecting the tissue surrounding the ovary (Basso-Alves et al. 2017).
In androecia with a secondary increase of stamens, the direction of development is strongly linked with the shape of the floral bud, as organs will arise preferentially where there is sufficient space for their initiation (Ronse De Craene and Smets 1992b). In convex floral buds vertical expansion of the meristem creates more space between the perianth and first-initiated stamens; as a result additional stamens develop centrifugally (Figs. 4e, 5f, g). The initiation of a concave flower with a cup-shaped hypanthium creates more space between first-initiated stamens and the gynoecium and additional stamens will preferentially develop in centripetal direction (Fig. 5d, e). The correlation between the changing direction of stamen development and hypanthial growth has been illustrated for Myrtales (e.g. Ronse De Craene and Smets 1991b; Vasconcelos et al. 2017). With lack of space between perianth and gynoecium, stamen increase will be mainly lateral, or variously centrifugal and centripetal (Fig. 4c; Marcgravia in Marcgraviaceae, Omphalocarpum in Sapotaceae: Caris 2013; Decumaria in Hydrangeaceae:; Hufford 1998, 2001b). In Myrtaceae and Melastomatacae the curvature of stamens in bud is strongly influenced by the expansion of a hypanthium and the number of stamens initiated (Basso-Alves et al. 2017; Vasconcelos et al. 2015). Curved filaments are characteristic for Myrtales and anthers can become imbedded in spongious tissue surrounding the ovary (Ronse De Craene 2010). Curved stamens have been identified as a common phenomenon in several Rosids (Myrtales, Rosales) and are intimately linked with the extent of hypanthial growth. In Napoleonaea floral buds (Lecythidaceae), fertile anthers are curved over the nectar disk and fit between style and hypanthium (Ronse De Craene 2010, 2011).
Vertical expansion allows for important morphological changes in flowers but also vegetatively. In some Acacia a vertical expansion of the vegetative apex between a first set of leaves that are initiated spirally allows for the initiation of smaller phyllodes in between (Rutishauser 2016). Classen-Bockhoff and Meyer (2016) demonstrated that the expansion of the corona in Passiflora is intimately linked to the available space: the formation of additional space by the vertical expansion of a hypanthium triggers the origin of new elements. This is considered to be an autonomous process. The development of the staminodial corona of Napoleonaea is also the result of a limited expansion of the receptacle (Ronse De Craene 2011). In Certain palms the floral receptacle presents much flexibility in size and shape, allowing for the increase in the number of stamens and perianth parts. Correlated with smaller stamens, this may lead to a very high polyandry as in Palandra (Uhl and Moore 1977, 1980).
Horizontal and vertical regression
Meristem reduction causes a reduction in carpel and stamen number as well as a reduction of the merism of the flower. Most pressures occur centripetally from the outer organs to the inner organs. However, pressures of the ovary tend to be as important, influencing the position and inducing the loss of the innermost stamens (Figs. 4f, g, 5h). The pressure of the gynoecium is strongest in the radius of the carpels, and this is the reason why stamens will preferentially alternate with the carpels (see above). A reduced meristem size can also influence the whole flower and lead to a shift in merism. Trimerous flowers are rare in the Pentapetalae (Kubitzki 1987; Ronse De Craene 2016; Ronse De Craene and Smets 1994). A restriction of space can be extended centrifugally from the ovary to the rest of the flower. In Tripetaleia (Ericaceae), the androecium and petals become trimerous as to become spatially accommodated with the ovary (Nishino 1988). Space reduction is responsible for the difference between staminate and pistillate flowers in Apuleia (Leguminosae): the trimerous pistillate flowers have two stamens, while staminate flowers have three stamens, with the abaxial stamen occupying the space of the ovary (Zimmerman et al. 2013). The strict arrangement of eight stamens in Sapindaceae and Tropaeolaceae (Fig. 4g) follows a similar pattern, linked with a trimerous gynoecium. The loss of two stamens in comparison with the common diplostemonous androecium is linked with the confinement of space at the time of stamen initiation (Cao et al. 2018). Alternatively stamens may become fused in groups alternating with the lower carpel numbers. The rearrangement and loss of stamens acts independently of a secondary stamen increase, as shown for some species of Hypericum. Most species of Hypericum have five carpels and five alternating stamens fascicles. In species with three carpels, the number of stamen bundles is reduced to three by the pairwise fusion of four fascicles (Classen-Bockhoff 2016; Leins 1964; Ronse De Craene and Smets 1994).
Shifts between pentamery and tetramery are clearly linked with size correlations between floral organs and floral meristems (Ronse De Craene 2016). This correlation has been quantitatively demonstrated in Rubiaceae (Naghiloo and Classen-Bockhoff 2016) and Dipsacoideae of Caprifoliaceae (Naghiloo and Classen-Bockhoff 2017).
The origin of tetramerous flowers from monosymmetric flowers is clearly correlated with space restrictions and not to changes in gene expression. Loss of the posterior stamens and a connation of the adaxial petals has been recognized as the preferred route for a shift to tetramery in Lamiales (e.g. Donoghue et al. 1998; Endress 1999; Ronse De Craene and Smets 1994). In several Caesalpinoids tetramery can be obtained by the connation of the adaxial sepals (Tucker 1998, 2000c).
Especially in basal angiosperms, a vertical regression of the size of the convex floral meristem can lead to a reduction of stamens and carpels. In certain families (e.g. Annonaceae, Alismataceae, Nymphaeaceae), the number of stamen whorls may fluctuate strongly and lower numbers may be caused by a shortening of the meristem size, while higher numbers may be the result of an opposite trend (Ronse De Craene and Smets 1993b).
Stamen-petal primordia are generally caused by heterochrony and a delay in the development of the petals (see before: time), but this regression is also expressed in space and can be caused by a reduction of the size of the floral meristem.
Time, size and pressure in a historical context: the shaping of the floral morphospace
Space in flowers results from the interchange of different processes acting during the development of flowers with the regularity of a clockwork system (Fig. 1b). Alterations of any of these parameters, such as heterochrony, pressure changes, or size shifts, will be sufficient to cause shifts in the Bauplan and these alterations appear to be regulated in an autonomous way.
In the case of Montiaceae, the combination of the timing of initiation, size of the floral meristem, and external pressures influence the floral diversification to a great extent. The diversification of Montiaceae is dependent on space constraints caused by an external involucre, a reduced floral meristem, and the potential of meristematic expansion (Dos Santos et al. 2012; Dos Santos and Ronse De Craene 2016, in prep.). This represents an escape from the historical constraints of the flower that is guided by forces in the floral development. The development of two median bracts creates an elliptical shape to the floral meristem and favours a more rapid development of organs in transversal position. Due to the pressure of the bracts, median petaloids are delayed, leading to a disconnection of the stamens from the perianth in Lewisia or their absorption in stamen tissue in Claytonia (Dos Santos et al. 2012; Dos Santos and Ronse De Craene 2016). The expanded receptacle also allows the transversal increase of stamens and petals independent of the constraints of an outer perianth; pressure of the carpels leads to a preferential development of stamens away from the radius of carpellary lobes. As the developmental parameters may change at different stages of development, different outcomes of mature morphologies may be expected. In Lewisia, pressures act preferentially in early stages of development, and a certain level of regularity is restored in mature flowers by a rapid expansion of the median petaloids, due to reduced pressure of the involucre.
Hypertelis (Kewa) salsoloides (Kewaceae) illustrates the significance of physical changes on established historical developmental patterns. The floral development of Hypertelis is highly unusual for Pentapetalae in general and Caryophyllales in particular. The androecium consists of three alternating stamen whorls, where the outer stamens occupy the position expected for petals (Fig. 6c; see Brockington et al. 2013). Initiation of the stamens proceeds centrifugally, starting with the inner alternisepalous stamens, followed by the antesepalous stamens, and finally the outer alternisepalous stamens, which arise from common primordia derived from the upper alternisepalous stamens (Fig. 6a–d). This unusual androecium could be explained in three ways: (A) the initial androecium consists of three whorls and the outer whorl becomes reduced and is absorbed in the upper stamen (as in common stamen-petal primordia; see above; Fig. 7a), (B) petals have become replaced by stamens through a process of homeosis (Fig. 7b), and (C) an increase of the available space on the floral meristem has allowed the antesepalous stamens to subdivide and fill the available space (Fig. 7c). Spatial constraints from the sepals and inner carpels in antesepalous position prevent the antesepalous stamens to increase in number.
There is little support for the first interpretation, as the existence of three independent stamen whorls is not found in other Caryophyllales or Pentapetalae. Homeosis is plausible, as the outer stamen occupies the space of petals in flowers, and arises in a comparable way as petals of e.g. Caryophyllaceae. However, the fact that Hypertelis is well nested within the raphide clade of Caryophyllales precludes the transformation of petals into stamens, as “normal” petals are absent from the clade (see Ronse De Craene 2013). The third interpretation appears to be the most plausible and indicates that vertical expansion of the floral meristem is directly responsible for a stamen increase. A similar increase of stamens is found in some Rivinaceae belonging to the same clade (Trichostigma, Ledenbergia: Ronse De Craene and Smets 1991a).
Flower evolution is the result of an interchange between historical processes and physico-dynamic events during development. It is remarkable that a stable floral Bauplan was established in both monocots and Pentapetalae that follows a strict modular plan and is highly similar in being pentacyclic. The only difference is that in the monocots flowers evolved as trimerous units, while they were pentamerous in Pentapetalae. This unity in shape implies two conclusions:
Perianth whorls (sepals and petals) are basically homologous (following the principle of process homology), with a later distinction responsible for their divergence (cf. Ronse De Craene 2007, 2008; Ronse De Craene and Brockington 2013).
The existence of two stamen whorls appears as a fundamental homoplasy between both groups, supporting the distinction made by Ronse De Craene and Smets (1987, 1993b, 1995a, b, 1998a) between the two-whorled androecium in Pentapetalae and monocots.
The close association between stamen and perianth parts appears to be a reflection of the modular nature of plant systems, consisting of a phyllome and axillary bud or short shoot. In this respect a carpel should be seen as the phyllome of the axial ovular system. This interpretation clearly links with older discussions of the axillary versus phyllomatic nature of floral parts (see also Rutishauser and Sattler 1985). These parameters are firmly established and regulated by the tight control of gene regulatory systems, allowing for a predictive establishment of floral forms.
The occurrence of a physico-mechanical variability may form the base of evolutionary changes in various angiosperm families. The development of the floral meristem is heavily influenced by physical factors, which are dependent on the timing of initiation of organs and the availability of space. However, these forces have become canalized through genetic fixation leading to the establishment of specific flower types that are highly conserved. However, random changes in the floral meristem, linked to physico-dynamic forces may break the strong historical constraint and lead to new floral morphs that become consequently stabilized. Several examples were presented in this review, where the reduced size of organs in relation to the apical meristem leads to abrupt changes of the floral tissue (see also Rutishauser 2016). In several cases the physico-dynamic process precedes the genetic establishment of a specific pattern (cf. Dumais 2007; Schwander and Leimar 2011).
Certain floral patterns may have been historically established, without any visible mechanical forces acting on the development of the flower. Brassicaceae is a good example of this phenomenon. Flowers are disymmetric and tetramerous with an almost generalized floral Bauplan consisting of four sepals, four petals, six stamens, and two carpels (Eichler 1878; Endress 1992). The antesepalous stamens are absent in the median plane and the antepetalous stamens are displaced relative to the petals. However, the immediate causes for the generalized disymmetry and the loss of two antesepalous stamens are lacking, except for the strong pressure exercised by the median sepals on the bud during development (see Erbar and Leins 1997; Smyth et al. 1990). Flowers with an elliptical shape are linked with a strong pressure of bracts, but in Brassicaceae bracts are absent, and we need to assume that the floral form was established before the divergence of the family and has been consequently conserved. The cause and origin of the disymmetric flower is found in other Brassicales that are more basal to Brassicaceae (e.g. Pentadiplandra: Ronse De Craene 2002), and this was more or less “memorized” and genetically established in derived families, such as Capparaceae, Cleomaceae, and Brassicaceae.
Are a historical and physico-dynamic perspective excluding each other? In fact, there is a strong complementary between both perspectives as a reflection of the complementarity in biology (sensu Meyer-Abich 1954, see Rutishauser and Sattler 1985). Both approaches are related to time, size and pressure as determining factors. Different factors may act as intermediates in the establishment of the floral morphospace. The role of auxin as a mediator in the apical meristem between genes and the positioning of microtubuli in the developing organs has been presented earlier. Floral evolution is the result of subtle developmental changes leading to significant changes in the morphology of flowers. The historical context explains the diversity at a given stage in time, but influences of physical forces operate before, during, and after the ontogeny of the flowers. These forces become canalized because a close interaction of organs become genetically fixed. Genetic changes may precede or follow the input of physical processes, which become unchanged by the establishment of a conserved genetic development regulating the space in flowers.
In the future it will be increasingly important to observe correlations of physical change in flowers, as these are important indicators of the direction of evolution. This highlights the importance of a morphological approach to understanding flowers and their evolution.
I am grateful to Dr. Akitoshi Iwamoto for inviting me to participate in the symposium entitled “Floral development, re-evaluation of its importance” organized at the 80th Annual Meeting of the Botanical Society of Japan in Okinawa. I thank the Japanese Society for Promotion of Science for funding my travel to Japan. I also thank the editors of JPR for suggesting this special volume for Flo-Re-S. I am grateful to various colleagues including Dr. Akitoshi Iwamoto, Dr. Kester Bull-Hereñu, Prof. R. Classen-Bockhoff, and Dr. Wei Lai for helpful conversations on morphological topics. The Royal Botanic Garden Edinburgh (RBGE) is supported by the Scottish Government’s Rural and Environmental Science and Analytical Services Division.
- Arthur W (2000) The origin of animal body plans: a study in evolutionary developmental biology. Cambridge University Press, CambridgeGoogle Scholar
- Bateman RM, DiMichele WA (2002) Generating and filtering major phenotypic novelties: neoGoldschmidtian saltation revisited. In: Cronk QCB, Bateman RM, Hawkins JA (eds) Developmental genetics and plant evolution. Systematics Association Special Volume Series 65. Taylor and Francis, London, pp 109–159Google Scholar
- Brady MS, Stein W, Crane PR (1997) Exploring patterns in floral morphology: mathematical modeling of floral development using an inhibitory morphogen and meristematic growth parameters. Am J Bot Suppl 84:6–37Google Scholar
- Brockington SF, Roolse A, Ramdial J, Moore MJ, Crawley S, Dhingra A, Hilu K, Soltis DE, Soltis PS (2009) Phylogeny of the Caryophyllales sensu lato: revisiting hypotheses on pollination biology and perianth differentiation in the core Caryophyllales. Int J Plant Sci 170:627–643CrossRefGoogle Scholar
- Cao L, Liu J, Lin Q, Ronse De Craene LP (2018) The floral organogenesis of Koelreuteria bipinnata and its variety integrifoliola (Sapindaceae): evidence of floral constraints on the evolution of monosymmetry. Plant Syst Evol (in press)Google Scholar
- Caris P (2013) Bloemontogenetische patronen in the Ericales sensu lato. Dissertation, Katholieke Universiteit Leuven, BelgiumGoogle Scholar
- Citerne H, Jabbour F, Nadot S, Damerval C (2010) The evolution of floral symmetry. In: Kader JC, Delseny M (eds) Advances in botanical research. Elsevier, London, pp 85–137Google Scholar
- Classen-Bockhoff R (1991) Anthodien, Pseudanthien und Infloreszenzblumen. Beitr Biol Pflanz 66:221–240Google Scholar
- Dos Santos P, Brockington S, Glover B, Ronse De Craene LP (2012) Micromorphological evidence for androecium origin of Claytonia (Montiaceae) petaloids. Modern Phytomorph 1:23–25Google Scholar
- Eichler AW (1875–78). Blütendiagramme 1 and 2. Wilhelm Engelmann, LeipzigGoogle Scholar
- Einset JW (1987) Botany: the state of art. How development’s clock guides Evolution. Arnoldia 47:20–25Google Scholar
- Endress PK (1987) Floral phyllotaxis and floral evolution. Bot Jahrb Syst 108:417–438Google Scholar
- Endress PK (1994) Diversity and evolutionary biology of tropical flowers. Cambridge University Press, CambridgeGoogle Scholar
- Endress PK (1998) Antirrhinum and Asteridae—evolutionary changes of floral symmetry. Soc Exp Biol Symp Ser 51:133–140Google Scholar
- Endress PK (2003) What should a “complete” morphological phylogenetic anaylsis entail? In: Stuessy TF, Mayer V, Hörandl E (eds) Deep morphology: towards a renaissance of morphology in plant systematics. ARG Gantner, Liechtenstein, pp 131–164Google Scholar
- Endress PK (2005) The role of morphology in angiosperm evolutionary studies. Nova Acta Leopoldina NF 92:342:221–238Google Scholar
- Erbar C (1994) Contributions to the affinities of Adoxa from the viewpoint of floral development. Bot Jahrb Syst 116:259–282Google Scholar
- Erbar C, Leins P (1983) Zur sequenz von Blütenorganen bei einigen Magnoliiden. Bot Jahrb Syst 103:433–449Google Scholar
- Erbar C, Leins P (1994) Flowers in Magnoliidae and the origin of flowers in other subclasses of the angiosperms. I. The relationships between flowers of Magnoliidae and Alismatidae. Plant Syst Evol suppl 8:193–208Google Scholar
- Farrar J, Ronse De Craene LP (2013) To be or not to be a staminode: the floral development of Sauvagesia (ochnaceae) reveals different origins of presumed staminodes. In: Berntsen T, Alsvik K (eds) Flowers, morphology, evolutionary diversification and implications for the environment. Nova Science, New York, pp 89–103Google Scholar
- Hufford LD (2001a) Ontogenetic sequences: homology, evolution, and the patterning of clade diversity. In: Zelditch ML (ed) Beyond heterochrony: the evolution of development. Wiley-Liss, OxfordGoogle Scholar
- Irish VF (2009) Evolution of petal identity. J Exp Biol 60:2517–2527Google Scholar
- Kania W (1973) Entwicklungsgeschichtliche Untersuchungen an Rosaceenblüten. Bot Jahrb Syst 93:175–246Google Scholar
- Karrer AB (1991). Blütenentwicklung und systematische Stellung der Papaveraceae und Capparaceae. Dissertation, University of ZürichGoogle Scholar
- Kirchoff BK (2000) Hofmeister’s rule and primordium shape: influences on organ position in Hedychium coronarium (Zingiberaceae). In: Wilson KL, Morrison DA (eds) Monocots: systematics and evolution. CSIRO, Melbourne, pp 75–83Google Scholar
- Leins P (1964) Die frühe Blütenentwicklung von Hypericum hookerianum Wight et Arn. und H. aegypticum L. Ber Deutsch Bot Ges 77:112–123Google Scholar
- Leins P, Erbar C (1985) Ein Beitrag zur Blütenentwicklung der Aristolochiaceen, einer Vermittlergruppe zu den Monokotylen. Bot Jahrb Syst 107:343–368Google Scholar
- Leins P, Erbar C (1987) Studien zur Blütenentwicklung an Compositen. Bot Jahrb Syst 108:381–401Google Scholar
- Leins P, Erbar C (1996) Early floral developmental studies in Annonaceae. Biosystematics and Ecology Series 10. In: Morawetz W, Winkler H (eds) Reproductive morphology in Annonaceae. Österr. Akademie der Wissenschaften, Wien, pp 1–27Google Scholar
- Leins P, Erbar C (2010) Flower and fruit. Schweizerbart’sche Verlagsbuchhandlung, StuttgartGoogle Scholar
- Linder HP (1998) Morphology and the evolution of wind pollination. In: Owens SJ, Rudall PJ (eds) Reproductive biology in systematics, conservation and economic botany. Royal Botanic Gardens, Kew, pp 123–135Google Scholar
- Murbeck S (1912) Untersuchungen über den Blütenbau der Papaveraceen. Kungl Sv Vet Akad Handl 50:1–168Google Scholar
- Nishino E (1988) Early floral organogenesis in Tripetaleia (Ericaceae). In: Leins P, Tucker SC, Endress PK (eds) Aspects of floral development. J. Cramer, Berlin, pp 181–190Google Scholar
- Plantefol L (1949) l’Ontogénie de la fleur, fondements d’une théorie florale nouvelle. Masson et Cie, ParisGoogle Scholar
- Reinhardt D, Kuhlemeier C (2002) Phyllotaxis in higher plants. In: McManus MT, Veit BE (eds) Meristematic tissues in plant growth and development. Academic, Sheffield, pp 172–212Google Scholar
- Remane A (1956) Die Grundlagen des natűrlichen Systems, der vergleichenden Anatomie und der Phylogenetik, ed. 2. Akademische Verlagsgesellschaft, LeipzigGoogle Scholar
- Ronse De Craene LP, Smets E (1991a) The floral ontogeny of some members of the Phytolaccaceae (subfamily Rivinoideae) with a discussion of the evolution of the androecium in the Rivinoideae. Biol Jb Dodonaea 59:77–99Google Scholar
- Ronse De Craene LP, Smets E (1995a) The androecium of monocotyledons. In: Rudall PJ, Cribb P, Cutler DF, Hymphries CJ (eds) Monocotyledons: systematics and evolution. Royal Botanic Gardens, Kew, pp 243–254Google Scholar
- Ronse De Craene LP, Smets E (1996) The morphological variation and systematic value of stamen pairs in the Magnoliatae. Feddes Rep 107:1–17Google Scholar
- Ronse De Craene LP, Smets EF (1998b) Meristic changes in gynoecium morphology, exemplified by floral ontogeny and anatomy. In: Owens SJ, Rudall PJ (eds) Reproductive biology in systematics, conservation and economic botany. Royal Botanic Gardens, Kew, pp 85–112Google Scholar
- Rutishauser R (2016) Acacia (wattle) and Cananga (ylang-ylang): from spiral to whorled and irregular (chaotic) phyllotactic patterns—a pictorial report. Acta Soc Bot Pol 85: https://doi.org/10.5586/asbp.3531
- Rutishauser R, Sattler R (1985) complementarity and heuristic value of contrasting models in structural botany I. General considerations. Bot Jahrb Syst 107:415–455Google Scholar
- Sattler R (1962) Zur frühen Infloreszenz und Blütenentwicklung der Primulales sensu lato mit besonderer Berücksichtigung der Stamen-Petalum-Entwicklung. Bot Jahrb Syst 81:385–396Google Scholar
- Sattler R (1978) “Fusion” and “Continuity” in floral morphology. Notes Roy Bot Gard Edinb 36:397–405Google Scholar
- Saunders ER (1937) Floral morphology, a new outlook, with special reference to the interpretation of the gynoecium II. Heffer and Sons, CambridgeGoogle Scholar
- Soltis DE, Soltis PS, Endress PK, Chase MW (2005) Phylogeny and evolution of angiosperms. Sinauer, SunderlandGoogle Scholar
- Stuessy TF, Mayer V, Hörandl E (eds) (2003) Deep morphology: towards a renaissance of morphology in plant systematics. ARG Gantner, LiechtensteinGoogle Scholar
- Takhtajan A (1972) Patterns of ontogenetic alterations in the evolution of higher plants. Phytomorphology 22:164–171Google Scholar
- Tucker SC (1988) Loss versus suppression of floral organs. In: Leins P, Tucker SC, Endress PK (eds) Aspects of floral development. J. Cramer, Berlin, pp 69–82Google Scholar
- van Heel WA (1966) Morphology of the androecium in Malvales. Blumea 13:177–394Google Scholar
- van Heel WA (1978) Morphology of the pistil in Malvaceae-Ureneae. Blumea 24:123–137Google Scholar
- van Heel WA (1987) Androecium development in Actinidia chinensis and A. melanandra (Actinidiaceae). Bot Jahrb Syst 109:17–23Google Scholar
- von Balthazar M, Schönenberger J, Alverson WS, Janka H, Bayer C, Baum DA (2006) Structure and evolution of the androecium in the Malvatheca clade (Malvaceae s.l.) and implications for Malvaceae and Malvales. Plant Syst Evol 260:171–197Google Scholar
- Wang Y-Z, Liangh R-H, Wang B-H, Li J-M, Qiu Z-J, Li Z-Y, Weber A (2010) Origin and phylogenetic relationships of the Old World Gesneriaceae with actinomorphic flowers inferred from ITS and trnL-trnF sequences. Taxon 59:1044–1152Google Scholar
- Weber A (2003) What is morphology and why is it time for its renaissance in plant systematics? In: Stuessy TF, Mayer V, Hörandl E (eds) Deep morphology: towards a renaissance of morphology in plant systematics. ARG Gantner, Liechtenstein, pp 3–32Google Scholar