9.2.1.1 How to Build a Flower

Most flowers in angiosperms follow a common basic plan, which consists of a standard arrangement of four concentric whorls. The two outer whorls or perianth are composed of sterile organs, most commonly sepals and petals, and the two inner whorls comprise the male and female reproductive organs.

Two decades ago, a genetic model for specifying floral organ identity was proposed based on mutant analyses in the two eudicot model species Arabidopsis thaliana and Antirrhinum majus. In both Arabidopsis and Antirrhinum, mutants had been described that showed transformations of one class of floral organs into another (Bowman et al. 1989; Schwarz-Sommer et al. 1990). These transformations were classified as a, b, or c types and generally affected two adjacent whorls. Class a mutants showed transformations of the two outer whorls, in which sepals took carpel identity and petals had stamen identity. Class b mutants developed sepals instead of petals and carpels instead of stamens. Finally, class c mutants showed transformations of stamens into petals and carpels into sepals and, in addition, they lost floral determinacy forming new flowers from the center of the floral meristem that repeated this pattern of homeotic transformations—lost determinacy.

Haughn and Sommerville made the first attempt to explain the phenotypes of these floral homeotic mutants with some simple logical rules that involved three classes of genes (A, B, C) which in different combinations of on/off states could specify the four types of floral organs (Haughn & Somerville 1988). A few years later, when the first homeotic genes started to be molecularly identified and Schwartz-Sommer et al. had already started to rework these “abc” rules (Schwarz-Sommer et al. 1990), Coen and Meyerowitz published the seminal “War of the whorls” paper, where the ABC model of floral development was clearly delineated (Coen & Meyerowitz 1991).

The ABC model was drawn from the three classes of homeotic mutants explained earlier and proposed that three gene functions, A, B, and C, defined by these mutant classes, would be responsible for the specification of floral organ identity. A, B, and C functions should be expressed in two adjacent whorls, overlapping with each other. Thus, A-function alone would specify sepals in the outer whorl, A + B would define petals in whorl 2, B + C stamens in whorl 3, and C alone carpels in the center of the flower. Mutual repression of A and C functions also had to be included in the model to explain a and c phenotypes, as well as an additional role of C-function in preventing indeterminate floral meristem growth (Figure 9.1a). In addition to the elegance of this simple yet powerful model, an added value was that it was inferred from the phenotypes of equivalent mutants in two relatively distant species of quite different floral morphology, and therefore strongly indicated that it could represent a general mechanism for flower development in angiosperms.

Concurrent with the proposal of the model, the identification of genetic functions to which ABC roles could be assigned strongly supported the model and provided a wealth of molecular data (Bradley et al. 1993; Goto & Meyerowitz 1994; Jack et al. 1992; Jofuku et al. 1994; Mandel et al. 1992; Sommer et al. 1990; Tröbner et al. 1992). All the identified ABC genes, except the Arabidopsis class A APETALA2 gene, were closely related in sequence, and belonged to the MADS-box family of transcription factors. Moreover, MADS-box proteins were shown to interact in protein complexes, providing a molecular clue to the mechanisms that explained the combinatorial nature of the model (Davies et al. 1996b; Egea-Cortines et al. 1999). In addition, genes with equivalent functions in different species showed interspecific homology, reinforcing the idea of the general applicability of the model throughout the angiosperms.

Still, the original ABC model did not provide answers for some essential questions. For example, ectopic expression of ABC genes was shown to transform one type of floral organ into another, as predicted by the model, but was not sufficient to transform leaves into floral organs (Mizukami & Ma 1992; Davies et al. 1996a; Krizek & Meyerowitz 1996). These results were in conflict with an old but attractive theory of the German philosopher and writer Goethe, who proposed that all lateral organs shared a leaf-like developmental plan (Goethe 1790). This necessitated an extension of the basic model with an additional E-function, specific to the floral context, which was proposed and eventually assigned to a subclade of MADS-box genes, the SEPALLATA (SEP) subfamily, expressed in all floral whorls (Figure 9.1a).

Sepallata mutants showed flowers where only leaf-like organs developed in a whorled but indeterminate pattern, strongly supporting the idea of a common ground plan for leaves and floral organs (Pelaz et al. 2000; Ditta et al. 2004). E-function (SEP) proteins were subsequently shown to act as molecular “glue” in high order MADS-box complexes (Immink et al. 2009a). Thus, the ABC model was reformulated into what we know today as the ABCE or quartet model: whorls 1, 2, 3, and 4 contain, respectively, tetramers of A-A-E-E, A-B-B-E, B-B-C-E, or C-C-E-E MADS-box proteins, which bind to DNA of target genes to direct floral organ specific development (Figure 9.1a). Several excellent reviews have addressed the molecular details and the historical development of the quartet model (Causier et al. 2010; Immink et al. 2009b; Krizek & Fletcher 2005; Theissen & Melzer 2007).

The regulation of temporal and spatial expression of ABCE functions was shown to be crucial for establishing floral organ identity. The genetic pathways from floral meristem specification to the establishment of ABCE functions have been studied and it was shown how they were based on feed-forward loops and a final feedback loop that reinforced and conferred robustness to the floral program. In Arabidopsis, LEAFY (LFY), a plant-specific transcription factor with a major role in floral meristem specification, is also a key player in the upregulation of ABCE genes (Weigel et al. 1992). LFY is uniformly expressed in the young floral meristem, but achieves temporal and spatial regulation of floral organ identity genes through the action of different cofactors (Parcy et al. 1998). LFY appears to be sufficient to activate APETALA1 (AP1), a MADS-box gene with a dual role in floral meristem specification and A-function.

Once activated by LFY, AP1 upregulates SEP3. SEP3 then forms a protein complex with AP1, switching its activity from floral specification to A-function. Subsequently, SEP3 is also able to act as a LFY cofactor to upregulate the other SEP genes, and the B- and C-function genes, although SEP3 is not as important for LFY activity as the other LFY cofactors (reviewed in Liu et al. 2009). LFY directly interacts with UNUSUAL FLORAL ORGANS (UFO), an F-box protein, to activate the B-class gene APETALA3 (AP3) in the UFO domain of expression, which coincides with the B-class expression domain (Lee et al. 1997; Chae et al. 2008). Activation of the C-class gene AGAMOUS (AG) in the center of the floral meristem requires the concerted action of LFY and WUSCHEL (WUS), a homeodomain protein with a key role in stem cell maintenance of all aerial meristems. Finally, in a negative feedback loop, AG terminates floral meristem growth by repressing WUS expression in the center of the floral meristem (Lenhard et al. 2001; Lohmann et al. 2001).

9.2.1.2 The Evolution of Flower Developmental Pathways

Conceptually, the quartet model could be considered a paradigmatic example of an evolutionary framework. It was based on a modular structure, with combinatorial activities that resulted in different outcomes, and these activities were largely controlled by regulation of expression patterns. Moreover, since it was proposed from independent studies in two relatively distant species, conservation across species was strongly supported. These features of the model provided powerful mechanisms to explain evolutionary innovation, immediately suggesting predictions that could explain morphological novelties. Accordingly, from the early years when the model was initially proposed and MADS-box genes were identified as the main players involved, intensive research comprising an ever growing diversity of species has shown how the model shows both conservation and divergence and has provided a body of evidence to study the evolution of organ identity. Still, we remain limited by the comparatively small number of comprehensively studied species, especially in evolutionary nodes where innovations appeared and functional studies are challenging.

The elusive A-function. From molecular phylogenies and comparative development studies, it is clear now that the best-supported part of the model refers to the specification of stamens and carpels by B and C functions. The A-function, however, remains controversial. Arabidopsis is the only species so far where clear loss-of-function A-class mutants have been described. Arabidopsis ap1 mutants show a partial loss of floral meristem identity and also some loss of A-function, with sepals transformed into bract-like organs and petals that do not form (Bowman et al. 1993). In the other A-class mutant, ap2, sepals are transformed into carpels and petals into stamens (Bowman et al. 1989). The other few mutants in AP1-like genes characterized in different species (for example, squamosa in Antirrhinum, proliferating inflorescence meristems in legumes or macrocalyx in tomato; Huijser et al. 1992; Taylor et al. 2002; Vrebalov et al. 2002) do not show these dual defects, being mainly affected in the floral meristem specification function. Likewise, mutants in AP2 orthologs like lip1 and lip2 show only mild defects in sepal and petal development but not misregulation of C-class genes.

For these reasons, many authors have favored a variant of the ABCE model, an earlier version known as the BC model (Litt 2007; Schwarz-Sommer et al. 1990), or its more recent reformulation, the (A)BC model (Causier et al. 2010). This alternative to the quartet model proposes that the (A) function would comprise both AP1-like and SEP genes and would not have a major role in perianth organ specification. (A)-class main roles then would be to specify the floral meristem, establishing a floral context where the default identity for floral organs would be sepals, and subsequently to regulate the expression pattern of B- and C-class genes. Once established the floral program, only B and C functions would be required for the specification of floral organ identity—carpels (C), stamens (B + C), and petals (B), with sepals being defined by the absence of B and C activities (Figure 9.1b).

The debate on quartet versus (A)BC model is hampered by the scarcity of mutants in different species, which makes it difficult to distinguish the general from the particular. But even after revisiting the original A-function, postulated as required for sepal and petal identity and to restrict C-function expression to the inner whorls, the quartet model in which A-function would require AP1-like genes still deserves some credit. The evolution of a dual perianth composed of sepals and petals appears to be fixed only in core eudicots, coincidental in evolutionary time with a whole genome duplication that originated the euAP1 clade (Litt & Irish 2003; see Section “Duplication and diversification of large gene families”). Moreover, while the in vivo formation of MADS tetramers has not been definitively proven yet, indirect evidence obtained in vitro or in yeast strongly supports the quartet ensemble, being one of these AP1/SEP/AP3/PI (Honma & Goto 2001; Ferrario et al. 2003). Also, the simultaneous ectopic expression of AP1 and the B-class genes PI and AP3 has been shown to convert cauline leaves into petals, without further development of floral meristems (Pelaz et al. 2001), and, for instance, the ap1-like mutant in Medicago truncatula (mtpim) shows conversions of sepals into leaf-like organs, casting some shadows on the idea that sepals could be the “default” floral organ state (Benlloch et al. 2006).

It is thus likely that AP1-like genes are part of the floral organ identity toolkit, while not so much in charge of the spatial repression of the C-function. This role in Arabidopsis would be mainly fulfilled by AP2, in turn repressed in the center of the flower by the microRNA miR172 or other cadastral regulators. In other species, AP2-like genes have not been assigned equivalent functions. Notably, however, a similar cadastral function is performed in Petunia hybrida or A. majus by a parallel module composed of NF-YA transcription factors regulated by a different microRNA, miR169 (Cartolano et al. 2007). The parallel AP2/miR172 and NF-YA/miR169 pathway constitutes an example of variation in molecular networks, which is not translated into structural differences in the flowers. It could also be a consequence of parallel redundant networks, which differ in relative importance in different species. This is suggested by the weak ap2-like phenotype of loss-of-function mutants in LIP1 and LIP2, two AP2-related genes from Antirrhinum (Keck et al. 2003), or the common role in C-function repression performed by similar transcriptional cofactors in Arabidopsis (LEUNIG) and Antirrhinum (STYLOSA) (Conner & Liu 2000; Navarro et al. 2004)

The B function works with borders: from fading to sliding. The debate affecting the specific role of putative A-function genes in floral organ specification does not touch the widely accepted universality of B and C-function roles in specifying petals, stamens, and carpels. B-class and C-class genes have been identified from a huge number of species and corresponding mutants have been characterized for several of them, which in general exhibit the expected homeotic defects predicted by the quartet model (reviewed in Irish & Sussex 1992; Krizek & Fletcher 2005; Kater et al. 2006).

Most common differences in floral architecture affect perianth organ identity. While core eudicots appear to have fixed the whorled arrangement of sepals–petals–stamens–carpels, considerably more variation occurs in basal angiosperms and monocots. Basal angiosperms generally show a perianth where no clear distinction can be made between sepals and petals. Perianth organs are often arranged in spiral phyllotaxis and show variable merosity. In some species, all perianth organs are petaloid, but in others, outer tepals are morphologically different from inner tepals, and intergrading organs of intermediate morphology are observed. Also, in some basal angiosperms like Asimina species, clearly different outer sepals and inner petals develop.

In all basal angiosperms, homologs to ABCE genes are found and, in general, expression patterns are consistent with predictions of floral organ identity by the quartet model. AG-like gene expression is found only in reproductive organs, likely reflecting this restricted expression pattern as ancestral. However, MADS-box genes related to the B-class genes AP3 or PI are expressed in broader domains, encompassing the whole of the perianth and frequently also the carpels in many basal angiosperms (Kim et al. 2005). The presence of B-class genes in the perianth is consistent with a modified version of the (A)BC model where broad B-gene expression explains the petaloid identity of these organs in most basal angiosperms. Also, in Asimina species, B-class gene expression is restricted to the petal domain, suggesting that a reduction in the ancestral broad domain of B-class expression could have originated the Asimina dual perianth (Kim et al. 2005). To explain gradual transitions in floral organ identity found in many basal angiosperms, the “fading borders” hypothesis has been proposed. This model suggests that gradients of expression of organ identity genes, and especially of B-genes, could generate domains with different relative proportions of overlapping activities that led to organs of intermediate morphology (Figure 9.1c). Precise information of expression patterns of B-genes in different basal angiosperm species is limited, but studies in Amborella or Persea support this scenario (Soltis et al. 2007a, 2007b; Theissen & Melzer 2007).

Many nongrass monocot species, such as lilies or tulips, have whorled tepaloid perianths. In general, although with a few exceptions, these species express B-class genes related to AP3 and PI in whorls 1, 2, and 3 (Kanno et al. 2003), supporting the prediction of van Tunen et al. (1993) of a “sliding border” for the B-class genes, expanding to the whorl 1 and therefore conferring petaloid identity to these organs (Figure 9.1d).

In fact, from these and other studies in basal angiosperms and monocots, the petaloid character of the perianth together with a broad and may be undefined expression pattern of B-class genes could be inferred as ancestral. Dual perianth organ identity would have evolved independently many times in different clades of angiosperms, possibly through the refinement of B-class gene expression domains. From the perspective of the quartet model, authors like Theissen and Melzer (2007) have suggested that the establishment of positive autoregulatory loops through changes in promoters or increasingly restricted protein complex formation could have contributed to define the borders of B- and C-function activities.

B and C genes are already present in gymnosperms, so they predate the origin of angiosperm flower, suggesting an ancestral role already in the specification of male and female reproductive structures. Interestingly, B-type proteins from gymnosperms tested so far have been shown to form only homodimers (Wang et al. 2010). In contrast, in basal angiosperms and monocots, homodimerization, facultative, and obligate heterodimerization of B-class proteins have been shown to occur (Su et al. 2008). Finally, in core eudicots, B-proteins form obligate heterodimers, and only these heterodimers are able to provide B-function, both maintaining its own expression and acting on other downstream targets (reviewed in Theissen & Melzer 2007; Immink et al. 2009a). This trend toward obligate heterodimerization possibly coupled to autoregulation, could have thus provided a mechanism for restricting and sharpening the borders of B-activity domain to specify defined whorls or petals and stamens.

Variations in the quartet model and the evolution of novel morphologies. Novel floral morphologies not only necessarily included in these evolutionary trends but also caused by changes in floral organ identity can be interpreted using the ABCE model. For example, the eudicot Rumex acetosa, has a two whorled perianth composed only of sepaloid organs. Accordingly, only the stamens express B-class genes, suggesting that an inward shift of their expression has originated this new morphology (Figure 9.1e; Ainsworth et al. 2005).

Similarly, shifts in B-class gene expression appear to underlie the highly unusual floral morphology of the monocot Lacandonia schismatica. In Lacandonia, small flowers form with a tepaloid perianth of sepaloid morphology and an inverted arrangement of the reproductive organs, where numerous carpels surround a central whorl of stamens. Studies on B- and C-function genes in this species have shown how the B-class gene LsPI is expressed broadly in young floral meristems to restrict later to the domain that will give rise to stamens and carpels, overlapping with the C-class gene LsAG. In contrast, the expression of the other B-class gene, LsAP3, is restricted to the central zone of the meristems where stamens will develop. Thus, simultaneous expression of LsAP3, LsPI, and LsAG is only achieved in the central whorl of stamens, while in the surrounding carpels only LsPI and LsAG are found, and all of them are absent in the tepaloid organs (Figure 9.1f; Alvarez-Buylla et al. 2010).

An additional example of morphological innovation that can be explained by modifications in the quartet model is provided by species of the basal eudicot genus Aquilegia. Most Aquilegia species have a perianth composed of two morphologically different types of petaloid organs and an additional whorl of a novel type of organ, the staminodium, placed between stamens and carpels. Aquilegia has one PI-like gene but three AP3-like genes. Detailed studies of B-class gene function have shown how the three AP3-like genes are expressed in different spatial and temporal patterns, including petals, stamens, and staminodia, and that each organ type expresses a specific combination of B-type genes. Moreover, downregulation of AqvPI caused transformation of both stamens and staminodia into carpels, indicating that staminodia are probably derived from stamens, and suggesting that gene duplication and subsequent subfunctionalization of AP3 genes could have originated this novel organ identity program (Kramer et al. 2007).

As we have seen, variations in the ABC architecture are able to explain floral morphological diversity. However, the same ABC system has also been shown to specify very different looking organs. Grass flowers differ significantly from their eudicot relatives in the identity of the sterile perianth organs, which is composed of grass-specific organs known as glumes, lemma, palea, and lodicules. The glumes were considered as bracts, while homology of lemma, palea, and lodicules to other angiosperm floral organs has been the subject of debate. The lodicule is a small glandular-like organ, placed between the palea/lemma and the stamens. At anthesis, lodicules expand to separate the lemma and palea and thus expose the anthers to the wind. Functional studies of flowers of grasses, such as rice or maize, have shown how mutations in B-class genes lead to homeotic transformations of lodicules into palea-like organs and stamens into carpels, strongly supporting the homology of lodicules and petals (Ambrose et al. 2000; Nagasawa et al. 2003; Xiao et al. 2003; Whipple et al. 2004). The petaloid nature of lodicules represents a remarkable example of a novel type of organ derived from a preexisting identity program. The unique morphology of lodicules would then be more likely the result of a change in the set of downstream targets of B-type genes throughout the course of grass evolution (Whipple et al. 2007).

Duplication and diversification of large gene families. Gene duplication is a common phenomenon in plants which, when followed by neo- and/or subfunctionalization, can lead to large gene families and gene networks, enabling functional diversification and the subsequent evolution of morphological novelties. Given the central role of MADS-box genes in floral development and evolution, considerable efforts have been directed to identify a vast number of members of this family across the land plants and to reconstruct their phylogeny. These studies have revealed the highly complicated structure of the family where relationships of orthology and paralogy give rise to multiple cases of neo- and subfunctionalization, gene co-option, and even convergent evolution (Jaramillo & Kramer 2007). Still, general patterns on MADS phylogenies correlate diversification of the MADS family with major steps in floral evolution. These studies have shown how key MADS-box genes related to floral organ specification could have undergone simultaneous duplication, probably through polyploidization, just before or coincident with the origin of the angiosperms (Irish & Litt 2005; Zahn et al. 2005). Moreover, the patterns of duplication of ABCE genes (AP1, AP3/PI, AG, and SEP subfamilies) are related and basically coincident with the origin of major clades within angiosperms, therefore suggesting a role for these duplications in angiosperm diversification.

For instance, in gymnosperms only one copy of a B-class homolog is present, whereas at least two copies (related to AP3 and PI) are found in angiosperms (Kim et al. 2004). Then, two clades of AP3-like genes are found only in core eudicots (euAP3 and TM6) indicating a second event of duplication before the origin of this clade (Kramer et al. 1998). Likewise, angiosperms possess two lineages of C-like genes, one comprising AG-like genes related to C-function and another involved in ovule identity specification originated from a duplication predating angiosperm origin (Kramer et al. 2004). AG-like genes have also undergone more duplication events before and within the monocot and the dicot clades. These duplications have resulted in paralogs where changes mainly in expression patterns but also in protein function have given rise to functional diversification (Hasebe et al. 1998; Davies et al. 1999; Causier et al. 2005; Airoldi et al. 2010). Likewise, for E-function genes we found a similar scenario. Two lineages of SEP genes (SEP1/2/4 and SEP3) are thought to be present in the common ancestor of angiosperms while no clear homologs of SEP genes have been found in gymnosperms, suggesting that SEP genes were lost in the ancestor of extant gymnosperms (Zahn et al. 2005; Theissen & Melzer 2007). The essential role of E-class genes in the specification of all floral organs, as key components of the floral quartets, suggests that the origin and diversification of the SEP-clade might have played a key role in the origin of the angiosperm flower (Wang et al. 2010).

Notably, the clade where the putative A-function AP1-like genes belong (euAP1) appears at the base of the eudicots as the result of a duplication in the more ancient AP1/FUL lineage (Litt & Irish 2003). The origin of the euAP1 clade coincides with that of the euAP3 lineage (see Section “The elusive A-function”) and with an important morphological innovation—the fixation of the whorled structure of the eudicot flower, with a dual perianth of clearly distinct sepals and petals. Although as previously discussed, the role of AP1-like genes in the specification of perianth organs is still debateable, the temporal concurrence of euAP1/euAP3 clades, sepals, and petals suggests a role for these eudicot gene clades in this morphological innovation.

9.2.1.3 The Origin of the Flower

Although floral organ identity is well studied and the quartet model constitutes a powerful working framework, the evolutionary origin of the flower remains an open question. The fossil record is still far from comprehensive, and we have not yet identified fossils unequivocally representing stem-group angiosperms. Therefore, we lack a clear picture of how the ancestral parents of extant flowers looked like, what was the temporal sequence of the morphological novelties that appeared in the first flowers, and what were the molecular changes responsible for these innovations. In spite of these uncertainties, Goethe's theory of leaves and floral organs sharing a common ground plan is nowadays accepted and strongly supported by the sepallata mutant phenotypes (Goethe 1790; Pelaz et al. 2000; Ditta et al. 2004). Thus, flowers are viewed as “compressed” branches and floral reproductive organs (stamens and carpels) derive from leaves bearing reproductive structures (sporophylls) along the axis of these branches.

Three major evolutionary steps can be proposed to transform these ancestral reproductive branches into “modern” flowers. Since most reproductive branches in gymnosperms are unisexual, the first step could be the origin of bisexuality. In addition, reproductive branches in gymnosperms are usually indeterminate, so determinacy would have to be acquired. Finally, sterile perianth organs should have originated. Although there is no definitive data on the order or even the nature of these steps, this sequence appears to be a plausible scenario.

Several recent publications have discussed current theories on the origin of bisexuality based on paleobotanic data and the rapidly increasing amount of molecular data (Frohlich & Chase 2007; Specht & Bartlett 2009; Melzer et al. 2010). Briefly, three different hypotheses are currently debated, which basically differ with respect to the nature (male or female) of the ancestral branch of sporophylls.

The Mostly Male Theory (MMT; Frohlich & Parker 2000) proposes that flowers are derived from male strobili of a gymnosperm-like ancestor. In the male strobili of this ancestor, the male sporophylls near the apex would have first become bisexual by developing ectopic ovules, and subsequently, these bisexual sporophylls would have lost male sporangia and become functionally female. Additional evolutionary steps toward the flower would be the loss of the residual female strobili, the transformation of sporophylls in the male/female strobili by closure of the apical female sporophylls to form carpels and the basal male sporophylls into stamens (Figure 9.2).

The MMT was proposed mainly on the basis of evolutionary studies on LEAFY gene function. LFY plays a central role in the specification of floral meristems in angiosperms and, subsequently, in the upregulation of floral organ identity genes. Although in angiosperm species only one LFY gene is found, in gymnosperms two paralogs are present, one LFY ortholog, and NEEDLY (NLY), for which no direct ortholog has been found in angiosperms. Initial expression studies in certain gymnosperms showed that each paralog was expressed predominantly in male (LFY) or female cones (NLY). Thus, the absence of NLY in angiosperms pointed to the female strobilus having been lost from the common ancestor to flowering plants and therefore to the male origin of the ancestral flower (Mouradov et al. 1998). Several independent observations supported this theory. For example, the Arabidopsis lfy mutant is able to produce carpels but no stamens, suggesting that LFY is only essential for male specification. In addition, it has been described that ectopic ovules form in different single mutants of several angiosperm species and even in the leaves of a gymnosperm like Ginkgo, suggesting that ectopic ovules are easily produced. However, recent conflicting data showing coexpression of LFY and NLY in both female and male cones from other gymnosperms (Carlsbecker et al. 2004; Vazquez-Lobo et al. 2007) have weakened this theory.

Two other sister hypotheses on flower origin, the out-of-male (OOM)/out-of-female (OOF) theories, both proposed by Theissen et al. (2002), are based on the observation that some extant conifers possess bisexual cones. These theories incorporate molecular data and the ABCE model of floral organ identity in an attempt to explain how bisexuality of the flower was originated by temporary or spatial shifts in patterns of expression of genes determining the male/female identity of reproductive organs, downstream of LFY/NLY (Figure 9.2). The most important difference of OOM/OOF and MMT is that the former do not involve extensive loss of female developmental programs and therefore, postulate that carpels have female origin instead of the ectopic formation of ovules on male organs proposed by MMT. The OOM hypothesis proposes that in an ancestral male strobilus, the expression of male-identity B-class MADS-box genes could have been downregulated in the distal part to leave female organs in these apical positions (Theissen et al. 2002) by an unknown signal of hormonal or other nature. Conversely, the OOF theory postulates that B-class genes could have been upregulated in the basal portion of a female cone, likewise producing a bisexual axis. Some recent studies on pollen versus seed cones in extant conifers show most spontaneous bisexual cones appear to be modified seed cones rather than modified pollen cones, thus favoring the out-of-female hypothesis (Rudall et al. 2011).

A further theory has been proposed by Baum and Hileman (2006; Baum and Hileman Theory (BHT)), which can be considered a variant of the OOM hypothesis, but with much more molecular detail. BHT incorporates increasing knowledge of molecular mechanisms of flower identity specification in extant species to formulate a highly speculative model that, in addition to the origin of bisexuality, goes further in an attempt to explain the acquisition of floral determinacy and even perianth origin. The BHT takes into account the central role of LFY in the upregulation of B- and C-function genes and the quartet model of floral organ identity. According to this hypothesis, LFY protein levels would increase with time in male reproductive branches, as has been observed in extant angiosperm inflorescences (Blázquez et al. 1997, 1998; Schmid et al. 2003). The BHT incorporates two more assumptions: that this tendency of temporal LFY accumulation would have been reinforced during evolution, and also that C-function proteins should be able to accumulate at much higher quantitative levels than B-function proteins.

Increasing levels of LFY together with this higher capacity of C-function accumulation would create a distal maxima of C-function. This apical excess of C-function would then sequester all available SEP proteins, preventing the formation of B-SEP complexes and thus resulting in the development of megasporophylls at the apex of the strobilus (Figure 9.2). The BHT also suggests that a change in WUSCHEL cis-regulatory sequences during evolution could have occasioned WUS repression by the apical excess of C-function, resulting in floral determinacy. This hypothesis also speculates further on the origin of the sterile perianth, involving a newly acquired essential role of WUS as a LFY cofactor to spatially define the activation of C-function genes. The final requirement of WUS for C-function activation, together with the central spatial pattern of WUS in the meristem, would be its exclusion from the outer/basal region of the ancestral flower. This would result in the loss of stamen identity in these regions and the development of sterile organs of petaloid appearance and therefore implying the andropetaloid origin of these organs. An attractive aspect of the BHT is that it makes a huge number of predictions that potentially could be experimentally tested. However, as already discussed, functional studies in representative species of most interesting clades, including gymnosperms and basal angiosperms, are currently out of reach, and we will have to wait before more conclusive evidence is gathered.

9.2.2.1 What is a Leaf?

Leaves are one of the most conspicuous features of land plants and arise as lateral determinate organs from an indeterminate apical meristem. The leaves of most vascular plants are all lateral organs that are vascularized and have determinate growth (Gifford & Foster 1989), with a broad lamina with adaxial–abaxial asymmetry, which is well suited for photosynthesis and respiration. Although leaves can be easily recognized by their position and structure, the leaves that are found throughout the land plants are not considered homologous (Bower 1935).

9.2.2.2 Leaf Evolution and Homology Across the Land Plants

The evolution and homology of leaves has been a longstanding question among botanists. Leaves in the bryophytes are found only in the gametophyte generation and are not considered homologous to the leaves found in any of the vascular plants (Domin 1931). Vascular plant leaves are found only in the sporophyte generation and are classified as either microphylls or megaphylls, the former found in lycophytes and the latter in the euphyllophytes (Figure 9.3). Microphylls are defined by their generally small size, single unbranched vascular strand, and lack of a leaf gap. Megaphylls are generally large, have complex venation, and are associated with a leaf gap. However, there are exceptions to these definitions, some lycophyte microphylls have branched vasculature and some fossil lycophyte microphylls were large (Gifford & Foster 1989). In spermatophytes, exceptions to the megaphyll definition also exist, such as the needles found in pines. Microphylls and megaphylls are not considered homologous, and megaphylls are proposed to have arisen anywhere from two to seven times independently in the euphyllophytes. There are several hypotheses regarding the evolution of megaphylls, the most popular being the telome theory, which proposes that megaphylls evolved through a transformation of branches (Wilson 1953; Kenrick & Crane 1997).

Paleobotanical studies have been important in constructing hypotheses of leaf evolution in vascular plants. Fossils of vascular plants from the Silurian and Devonian have highly branched structures that lack laminate leaves and Paleozoic fossils indicate that laminate leaves evolved independently in the euphyllophyte clades (Gifford & Foster 1989). Although there is a wide range of vascular plant leaf morphology, these morphologies exist in all euphyllophyte clades indicating a high level of convergence in leaf development (Boyce & Knoll 2002). Analyses of fossil leaf morphology and vasculature show that spermatophytes and monilophytes underwent similar leaf developmental trajectories. Boyce and Knoll (2002) suggest that this indicates modifications of a common developmental network and that this shared developmental network constrained the early evolutionary direction of leaf development and the number of leaf morphologies that could be attained. Comparative development of angiosperm leaves provides some support for these hypotheses, however, molecular genetic support across the euphyllophytes is still lacking.

9.2.2.3 Genes and Pathways Necessary for Simple Leaf Development in Angiosperms

Molecular genetic studies in angiosperms have elucidated a large part of the regulatory network necessary for the development of a simple angiosperm leaf. The development of lateral organs involves the specification of organ position on the flanks of the meristem, establishment of boundaries between the determinate organ and indeterminate meristem, and specification of the cell types of the lateral organ. This would appear to be a logical sequence of events but there is in fact a lot of feedback between the different processes composing the leaf regulatory network. In many angiosperms, the KNOX-ARP (Class I KNOTTED1-like HOMEOBOX (KNOX) and ASYMMETRIC LEAVES1/ROUGH SHEATH2/PHANTASTICA (ARP MYB domain proteins) regulatory module is important for setting the domains between an indeterminate shoot apex and a determinate lateral organ (Figure 9.3; Vollbrecht et al. 1991; Lincoln et al. 1994; Long et al. 1996; Waites et al. 1998; Timmermans et al. 1999; Tsiantis et al. 1999; Byrne et al. 2000; Reiser et al. 2000; Semiarti et al. 2001). And in many angiosperms, the interaction between Class III homeodomain leucine zipper (HD-ZIPs), KANADI (KAN), and YABBY (YAB) proteins are important for vascular development, meristem identity, and adaxial–abaxial asymmetry in lateral organs (Baima et al. 1995; McConnell & Barton 1998; Siegfried et al. 1999; Eshed et al. 2001; McConnell et al. 2001; Emery et al. 2003; Eshed et al. 2004; Prigge et al. 2005; Byrne 2006). Simple leaf development studies have focused on the KNOX/ARP and HD-ZIP/KAN/YAB modules. We will review the developmental network involving these two genetic modules mainly from work done in the angiosperm Arabidopsis thaliana unless otherwise specified.

Lateral determinate organs are produced on the flanks of the indeterminate meristem, and determinate primordia initiate at sites of auxin maxima (Reinhardt et al. 2000; Benkova et al. 2003; Reinhardt et al. 2003). Auxin influx and efflux carriers (e.g., PIN-FORMED1 (PIN1)) direct auxin flow in the shoot (Reinhardt et al. 2000; Reinhardt et al. 2003; Bainbridge et al. 2008), and plants unable to synthesize auxin or those treated with auxin transport inhibitors have defects in lateral organ formation (Okada et al. 1991). In addition, pin1 mutants, that are unable to create auxin maxima, have defects in lateral organ formation (Okada et al. 1991; Galweiler et al. 1998).

PIN1 and KNOX homeodomain transcription factors are expressed in a complementary pattern in the shoot (Heisler et al. 2005). Generally, KNOX proteins are expressed in the vegetative and floral meristems and downregulated in leaf primordia and floral organs (Vollbrecht et al. 1991; Kerstetter et al. 1994; Long et al. 1996; Reiser et al. 2000). The Class I KNOX mutant shootmeristemless (stm) germinates with no apical meristem and although adventitious branches may form, the meristems of these branches soon terminates as well (Barton & Poethig 1993; Long et al. 1996). Class I KNOX proteins are required for the maintenance of indeterminacy and prevention of cell differentiation in the shoot apical meristem. ARP and KNOX are also expressed in complementary patterns in the shoot. The ARP protein, ASYMMETRIC LEAVES1 (AS1) is expressed in leaf primordia and developing leaves (Figure 9.3; Byrne et al. 2000, 2002). A loss of ARP function results in KNOX expression in the leaves and the resulting plant resembles a KNOX overexpression phenotype. The ARP protein AS1 forms a heterodimer with the LOB domain protein, AS2, which binds to the cis-regulatory region of Class I KNOX genes and represses KNOX transcription (Guo et al. 2008). In addition, Class I KNOX negatively regulate AS1 (Byrne et al. 2000).

The complementary expression domains and antagonism of ARP and KNOX proteins have been proposed as an important developmental module to specify determinate leaves from the flanks of the indeterminate SAM. In addition, auxin and the ARP protein, AS1, have been shown to act together to repress KNOX expression in the leaves (Hay et al. 2006). The auxin defective and as1 double mutants produce leaves with more indeterminate features than either single mutant and KNOX genes are ectopically expressed in these indeterminate regions of the double mutant.

The boundary between the indeterminate apical meristem and determinate lateral organs is finely tuned by three NAC-domain transcription factors, CUP-SHAPED COTYLEDON1–3 (CUC1, CUC2, CUC3) (Aida et al. 1997; Takada et al. 2001; Vroemen et al. 2003). CUCs are expressed in the embryonic shoot apical meristem and in the boundary region between the apical meristem and lateral determinate organs (Aida et al. 1999; Ishida et al. 2000; Takada et al. 2001; Vroemen et al. 2003; Hibara et al. 2006). Plants with mutations in any two of the three Arabidopsis CUC genes have fused lateral organs while plants overexpressing CUC1 have meristems forming on cotyledons (Aida et al. 1997; Takada et al. 2001; Vroemen et al. 2003; Hibara et al. 2006). The 35S:CUC1 plants have ectopic STM expression in the cotyledons while a loss of CUC function results in the loss of STM expression and therefore a loss of meristem formation during embryogenesis (Aida et al. 1999; Takada et al. 2001). In addition, a loss of STM results in an expansion of CUC gene expression and a plant with no apical meristem and fused cotyledons (Aida et al. 1999). Therefore, CUC genes promote KNOX expression during embryogenesis and KNOX expression represses CUC expression in the meristem. CUCs have redundant roles in SAM formation during embryogenesis and lateral organ boundary formation and maintenance throughout the plant lifecycle (Aida et al. 1997, 1999; Ishida et al. 2000; Takada et al. 2001; Vroemen et al. 2003; Hibara et al. 2006). Given the relationship between KNOX and auxin and KNOX and CUC, it is not surprising that there is an interaction between CUCs and auxin. Plants mutant for the auxin efflux protein, pin1, have expanded CUC2 expression suggesting that auxin represses CUC expression to the boundary of the initiating primordia (Vernoux et al. 2000).

After leaf initiation, the angiosperm leaf soon acquires polarity, which is important for proper cell type specification in this lateral organ. The Class III HD-ZIP, KAN (GARP-type transcription factors), and YABBY proteins are important for meristem identity and adaxial–abaxial asymmetry in lateral organs (Figure 9.3; Siegfried et al. 1999; Eshed et al. 2001; Kerstetter et al. 2001; McConnell et al. 2001; Emery et al. 2003; Prigge et al. 2005). HD-ZIPs are expressed in the SAM, vasculature, and adaxial region of lateral organs and are necessary for the initiation of the SAM and axillary meristems, specification of adaxial identity in lateral determinate organs such as leaves and floral organs, and determination of vascular tissue patterning (Baima et al. 1995; McConnell & Barton 1998; McConnell et al. 2001; Emery et al. 2003; Prigge et al. 2005; Byrne 2006). KANs are expressed in the abaxial region of lateral organs and are necessary to specify abaxial identity and vascular patterning (Eshed et al. 2001; Kerstetter et al. 2001; Emery et al. 2003; Eshed et al. 2004). YABBYs are expressed in the abaxial region of lateral organs and are necessary for abaxial identity in lateral organs (Sawa et al. 1999; Siegfried et al. 1999; Sarojam et al. 2010).

A loss of HD-ZIP function results in peg-shaped cotyledons with abaxial epidermal features while a loss of KAN or YABBY proteins results in radial lateral organs that are adaxialized (Siegfried et al. 1999; Eshed et al. 2001; Emery et al. 2003; Eshed et al. 2004; Sarojam et al. 2010). KANs are ectopically expressed in the HD-ZIP loss of function mutants and HD-ZIPs are ectopically expressed in the KAN loss of function mutants (Eshed et al. 2001). HD-ZIP proteins and KAN proteins are mutually antagonistic (Eshed et al. 2001; Kerstetter et al. 2001; Reinhart et al. 2002; Rhoades et al. 2002; Emery et al. 2003). Mutations that result in abaxialized lateral organs also have meristems that cease to function early in development (Siegfried et al. 1999; Eshed et al. 2001; Emery et al. 2003). Micro RNAs 165/166 are important for repressing HD-ZIP activity and restricting its action to the adaxial side of the leaf (McConnell et al. 2001; Rhoades et al. 2002; Emery et al. 2003). In model angiosperm species, the juxtaposition of abaxial and adaxial identity in leaves has been proposed to be necessary for laminar outgrowth (Waites & Hudson 1995; McConnell et al. 2001; Rhoades et al. 2002; Emery et al. 2003; Eshed et al. 2004).

Molecular genetic studies have been important to further elucidate the developmental network necessary to specify the leaves. Genes involved in auxin signaling along with KANs promote abaxial identity (Pekker et al. 2005). Proteins that are important for leaf determinacy also play a role in the axial patterning of the leaves. The ARP protein AS1 along with the LOB domain protein AS2 promotes adaxial identity by positively affecting HD-ZIP expression (Lin et al. 2003, 2005). In addition, AS1 and AS2 repress the activity of YABBY genes in the adaxial region (Iwakawa et al. 2007).

9.2.2.4 Genes and Pathways Necessary for Compound Leaf Development in Angiosperms

Comparative molecular genetic studies have shown that the leaf regulatory network that specifies the leaf primordia from the shoot apical meristem is similar to the regulatory network that species leaflets from the rachis of a compound or dissected leaf (Figure 9.4). Differences in KNOX and ARP expression patterns and regulation have provided a molecular genetic explanation for the morphological differences between simple versus compound leaves in some angiosperm species (Kessler & Sinha 2004; Hay and Tsiantis 2006; Kellogg 2006; Kimura et al. 2008). In simple-leaved angiosperms, KNOX proteins are expressed in the SAM, downregulated in incipient leaf primordia, and not expressed in mature leaves (Vollbrecht et al. 1991; Lincoln et al. 1994; Nishimura et al. 1999).

However, in angiosperms with compound leaves (e.g., tomato), KNOX proteins are expressed in the incipient leaf primordia but not in mature leaves (Janssen et al. 1998). In tomato, overexpression of KNOX produces young leaves that are coiled like a fern fiddlehead (circinate vernation) and mature leaves that are highly dissected (Hareven et al. 1996; Chen et al. 1997; Janssen et al. 1998). This coiled morphology in tomato KNOX overexpression plants is hypothesized to be due to faster cell growth on the abaxial compared to the adaxial side of the leaf. In addition, the natural variation in the level of leaf dissection found in various tomato cultivars is due to differences in KNOX expression and its interactions (Kimura et al. 2008).

Developmental genetic studies in Cardamine hirsuta have nicely illustrated how changes in the leaf developmental network result in changes in leaf morphology (Hay & Tsiantis 2006). C. hirsuta is closely related to Arabidopsis but has compound leaves. Similar to what is found in other compound leaved angiosperm species, C. hirsuta KNOX expression is reactivated in the leaflet primordia (Bharathan et al. 2002; Hay & Tsiantis 2006). In addition, an increase in KNOX expression in C. hirsuta results in more complex leaves while a reduction in KNOX expression gives simple leaves (Hay & Tsiantis 2006). Hay and Tsiantis (2006) also found that the difference in C. hirsuta KNOX expression in leaves compared to Arabidopsis KNOX expression are due to changes in the cis-regulatory region of C. hirsuta KNOX.

Studies on the regulation of these leaf developmental networks are important for understanding how changes in this regulatory network can result in different leaf morphologies. Analyses of angiosperm KNOX sequences have shown that a conserved noncoding sequence in the 5' regulatory region termed the K-box is found in angiosperms with simple and complex leaves (Uchida et al. 2007). Deletion of the K-box in simple leaved species results in an as1 (ARP) mutant phenotype. These results show that the K-box is necessary for the persistent repression of KNOX in leaves.

Although KNOX gene expression is reactivated in the leaflet primordia of many compound leaved angiosperms, the compound leaved Pisum sativum (pea) is an exception (Hofer et al. 1997; Bharathan et al. 2002). The pea leaf mutant unifoliata has a simple leaf and the mutation mapped to an ortholog of the meristem identity gene LEAFY (LFY) (Hofer et al. 1997). This likely represents a co-option of a protein necessary for flower development to leaf development in pea.

Not only are Class I KNOX proteins reactivated in the leaflet primordia of most compound leaved angiosperms, but research so far indicates that the entire leaf developmental network is redeployed for leaflet formation in compound leaved species (Figure 9.4). For example, auxin maxima are detected at the sites of leaflet primordia initiation of several compound-leaved angiosperm species (Barkoulas et al. 2008; DeMason & Polowick 2009; Koenig et al. 2009). In addition, in C. hirsuta the accumulation of auxin coincides with a reduction in Class I KNOX expression in developing leaflet primordia similar to what is found in developing leaf primordia on the flanks of the SAM (Barkoulas et al. 2008). The NAC transcription factors also play a role in defining the boundary between the leaflets and rachis of the compound leaf similar to their role in defining the boundary between the leaf primordia and SAM (Blein et al. 2008, 2009). Surprisingly, pea LFY interacts in the developmental network with auxin and NAC transcription factors to form compound leaves similar to the role played by KNOX with auxin and NAC transcription factors in forming compound leaves in other angiosperms (Blein et al. 2008, 2010). The developmental genetic research on compound leaves illustrates how the simple leaf developmental network has been redeployed to specify leaflets in compound leaved species.

9.2.2.5 Leaf Genes and Pathways Across the Land Plants

The identification of the leaf regulatory networks throughout the land plants will be important for understanding when these networks evolved and to elucidate the evolution of land plant leaves. KNOX homologs have been isolated throughout the land plants (Bharathan et al. 2002; Harrison et al. 2005). Functional analyses of KNOX genes in Physcomitrella patens indicate that KNOX proteins are essential for proper development of the diploid sporophytes but not for the development of the leafy gametophyte (Sakakibara et al. 2008). These results support the idea that the leaves of bryophytes are not homologous to the leaves of vascular plants (Domin 1931). HD-ZIP and KAN transcription factors have been identified in bryophytes (P. patens), lycophytes (Selaginella spp.), and seed plants, and are likely to be found in ferns (Sakakibara et al. 2001; Floyd et al. 2006; Prigge & Clark 2006; Floyd & Bowman 2007). YABBY transcription factors have only been identified thus far in seed plants and it has been suggested that the evolution of the YABBY protein family was important for the evolution of the seed plant leaf (Floyd & Bowman 2007; Sarojam et al. 2010). KNOX expression patterns have been assessed in three fern species with compound leaves (Ceratopteris richardii, Osmunda regalis, and Anogramma chaeophylla) and found to be expressed in the apical meristem, leaf primordia, and leaf margins (Figure 9.3; Bharathan et al. 2002; Harrison et al. 2005; Sano et al. 2005). Unlike angiosperms, it appears that KNOX genes are not downregulated in incipient leaf primordia in ferns. ARP expression has only been studied in O. regalis and found in the apical meristem and leaf primordia (Harrison et al. 2005).

9.2.2.6 Leaf Genes and Leaf Homology

Studies of KNOX, ARP, and HD-ZIP expression patterns have been used to provide evidence for various hypotheses on microphyll and megaphyll origins and megaphyll homology (Harrison et al. 2005; Floyd & Bowman 2006). The leaves of vascular plants are all lateral organs that are vascularized, have determinate growth, and have adaxial–abaxial polarity (Gifford & Foster 1989). The KNOX-ARP and HD-ZIP-KAN interactions are important for the specification of these defining leaf characteristics. KNOX proteins are expressed in the apical region in the lycophyte Selaginella kraussiana but not in its leaf primordia (Figure 9.3; Harrison et al. 2005). ARP homologs were also isolated from S. kraussiana and are expressed in the apical region, leaf primordia, and developing leaves.

The expression of KNOX and ARP genes in the lycophyte S. kraussiana has been interpreted as support for a common genetic mechanism specifying leaf determinacy in microphylls and megaphylls (Harrison et al. 2005). Although, it has been argued that the KNOX-ARP interaction is important for the determinate growth of leaves, others have suggested that KNOX genes are not important for the determinacy of a leaf but whether a leaf is simple or compound (Harrison et al. 2005; Floyd & Bowman 2006). A thorough investigation of KNOX and ARP expression patterns in ferns and gymnosperms with simple and compound leaves will provide further data to compare to lycophyte and seed plant expression patterns.

The HD-ZIPs and KANADIs are important for leaf development, vascular patterning, and adaxial–abaxial polarity in leaves of seed plants (Floyd & Bowman 2007). HD-ZIP homologs have been isolated from across the streptophytes, and phylogenetic analyses indicate that one HD-ZIP protein was present in the common ancestor of all embryophytes (Floyd & Bowman 2004; Floyd et al. 2006; Prigge & Clark 2006). Two Class III HD-ZIP proteins each have been isolated from the lycophytes Selaginella moellendorffii, S. kraussiana, and C. richardii, and three from Psilotum nudum (Floyd & Bowman 2004, 2006, 2007; Floyd et al. 2006; Prigge & Clark 2006). In situ hybridization in S. kraussiana showed that one HD-ZIP gene is strongly expressed in the apical cells and the provascular tissue, whereas the other HD-ZIP gene is weakly expressed in the apical cells, adaxial region of the developing microphyll where the ligule will form, and the provascular tissue (Figure 9.3; Floyd & Bowman 2006). This suggests that HD-ZIP proteins have a role in apical meristem function and vascular development. A comparison of HD-ZIP expression patterns in seed plants and the lycophyte S. kraussiana has been interpreted as support for the independent origin of megaphylls and microphylls (Floyd & Bowman 2006). HD-ZIP and KAN expression patterns have not been assessed in any ferns but are integral to our understanding of megaphyll evolution and development. These comparative studies are crucial to address questions of homology and may illustrate how regulatory networks underlying development constrain the evolution of form.