One of the most fascinating questions in evolutionary biology is understanding how novelties such as new organs or novel body plans arise. Answering this question is all the more challenging as obvious predecessors cannot always be identified in the incomplete fossil record. Evolutionary developmental (evo-devo) studies address this issue by examining how developmental mechanisms have evolved over time. To undertake such studies, rigorous assessment of trait homology (i.e. inherited from a common ancestral trait) is necessary, and this can now be conducted in a robust phylogenetic framework thanks to the remarkable recent progress in molecular systematics (e.g. APG III, 2009; Moore et al., 2010). Close examination of the sequence of developmental events underlying homologous traits, and the genes regulating developmental changes, is critical to our understanding of the present state and evolution of biodiversity. Indeed, the emergence of novel body plans or organs very often coincides with the rise of a wealth of new species, suggesting that evo-devo concepts can revolutionize our understanding of macro-evolutionary events. The flower is doubtlessly the most beautiful example of an evolutionary innovation that is believed to have been a major contributor to the diversification of angiosperms. Flowering plants nowadays dominate terrestrial ecosystems and, as a major food source, are of outstanding importance for mankind. Floral structure is generally conserved with four main organ types (sepals, petals, stamens and carpel); however, the variation on this theme is breathtaking. The types of variation include abortion of organs, radial versus bilateral symmetry, whorled or spiral phyllotaxis, dramatic variations in the colour, arrangement, number or size of floral organs, or even evolution of extra floral organs, such as, for example, in the genus Aquilegia. In recognition of the recent advances in our understanding of the mechanisms involved in the evolution of floral phenotypes, this issue gathers together eleven articles contributing to the field of ‘flower evo-devo’, which represent diverse approaches ranging from morphological and developmental comparative studies, to phylogenetic analyses of important developmental regulators, and gene function analysis in non-model plants.
Beyond their remarkable appearance and diversity that enchant our eyes (Fig. 1), flowers function as the angiosperm reproductive system. The amazing morphological diversity results from the interplay of selection and developmental constraints; however, evaluating the relative importance of these two evolutionary forces is not straightforward. Traits that have evolved several times independently provide interesting frameworks for addressing this issue. For instance, bilateral symmetry appears to be an adaptive trait that has probably evolved jointly with specialized pollinators, increasing precision in pollen deposition on the pollinator's body and thus resulting in better cross-pollination efficiency (e.g. Citerne et al., 2010). Symmetry characteristics involved in flower–pollinator interactions are often clade-specific but with conserved general trends (e.g. bilaterally symmetrical flowers exist in Orchidaceae and Fabaceae but with different architectures: Fig. 1C, H). Bilateral floral symmetry probably emerged repeatedly as a result of selection for increased reproductive efficiency, and involved the evolutionary tinkering of inherited genetic networks (Jacob, 1977). The molecular network controlling flower symmetry is beginning to be characterized in distantly related taxa of core eudicots, revealing both shared and different factors (Costa et al., 2005; Wang et al., 2008). Similarly, selfing has evolved multiple times from outbreeding, with associated changes in several floral traits. In this Special Issue, Sicard and Lenhard (2011) examine the selfing syndrome, with a specific focus on sex allocation to male vs. female function and flower morphology. The molecular bases of these traits are still largely unknown, but this system appears promising for examining the importance of genetic constraints in evolution, especially in the early stages of the transition to selfing when selection associated with outbreeding is expected to be relieved.
Floral diversity among angiosperms. (A) Aconitum napellus. (B) Akebia quinata. (C) Lotus corniculatus. (D) Lamprocapnos spectabilis. (E) Doronicum sp. (F) Gentiana acaulis. (G) Papaver somniferum. (H) Ophrys scolopax. (I) Cleome spinosa. (J) Leptospermum scoparium. (K) Callistemon citrinus. (L) Eschscholzia californica. (M) Brassica oleracea. (N) Aquilegia alpina. (O) Heracleum spondylium. (P) Lilium martagon. Images: C. Damerval (A–H, N–P); A. Becker (I–L); K. Alix (M).
Homeotic mutants in model organisms have been crucial for the discovery of genes controlling key steps in developmental processes. The evolutionary potential of homeotic mutants is still strongly debated. Goldschmidt (1940) was the first to propose the concept of ‘hopeful monsters’, which has been strongly challenged by the tenants of the Synthetic Theory of evolution. It seems, however, that discrete phenotypical differences between species cannot always be explained by the accumulation of small changes and that, in at least some cases, evolution may have proceeded in a saltationist rather than a gradualist way (Rudall and Bateman, 2003; Theißen, 2009). Gould underlined that mutations in key developmental genes could result in drastic phenotypical changes in adult forms without conflicting with the Darwinian theory, and that the evolutionary potential of such phenotypes could not be simply dismissed (Gould, 1980). In plants, few mutants of floral phenotype are known (e.g. the ‘peloria’ flower in Linaria; Gustafsson, 1979): in this issue, Wang et al. (2011, who also review other cases) describe a double-flowered mutant of periwinkle, and underline the need for fitness studies in natural habitats to evaluate the evolutionary potential of such anomalous phenotypes. Ronse de Craene et al. (2011) address the question of the evolutionary role of carpeloidy, by comparing the hermaphrodite flowers of Carica papaya and the superman1 mutant of Arabidopsis thaliana, which both display stamen-to-carpel conversions. They are able to show that the developmental mechanisms differ in the two species and suggest that the way the conversion takes place in arabidopsis may be reminiscent of early events in the evolution of bisexuality in plants. This comparative developmental study highlights the importance of in-depth knowledge of how organs are formed in order to formulate relevant evolutionary hypotheses. In connection with this topic, a review by Peter Endress (2011) is devoted to the development and evolution of angiosperm ovules. These organs can be traced back to the origin of seed plants, being shared by angiosperms and gymnosperms. Angiosperm ovules have evolved specific traits, some of which may be related to their inclusion in the carpels. The review points towards the diversity and morphological trends of ovules in major angiosperm clades, providing evo-devo researchers with a new framework to investigate the molecular bases of ovule evolution.
More than 20 years ago A. thaliana was established as a convenient plant system to study the molecular mechanisms of diverse physiological and developmental processes. The molecular genetics of flower development has been a major research focus from very early on, and yielded a wealth of information about the regulation of floral organ identity, and floral meristem identity, maintenance and termination. Studies on A. thaliana in combination with Antirrhinum majus, for which large collections of mutants have been available for a long time, have allowed the identification of a limited number of transcription factor families such as the MADS, TCP, NAM/CUC and YABBY families that have played crucial roles in floral morphogenesis in core eudicots. Orthologs and/or homologs of the A. thaliana and A. majus genes have been identified in other angiosperm species as being related to specific traits, and have been characterized in terms of sequence conservation and expression patterns. This so-called candidate-gene approach has turned out to be very insightful, suggesting that many genes have functions that are conserved among higher eudicots, and sometimes between them and other lineages of angiosperms, in accordance with the monophyly of angiosperms (Soltis et al., 2008; APG III, 2009). An emblematic gene family for flower development is the MIKC-type MADS-box transcription factor family: indeed, most genes involved in the control of floral organ identity in arabidopsis and antirrhinum are MADS-box genes (Coen and Meyerowitz, 1991; Causier et al., 2010). Studies of MADS-box gene expression in key species for floral diversity suggest that their role in floral organ identity specification is widely conserved among angiosperms (Kim et al., 2005). Additionally, MADS-box genes have been shown to play roles in a variety of developmental processes in plants, such as floral meristem specification, regulation of flowering time, embryo and fruit development, and are involved in nodulation during symbiosis with nitrogen-fixing bacteria (Zucchero et al., 2001; Seo et al., 2009; Vrebalov et al., 2009; Zheng et al., 2009, Kaufmann et al., 2010). Considering the reproductive phase, SOC1 is a major flowering pathway integrator in arabidopsis (Moon et al., 2003); in this Special Issue, Ruokolainen et al. (2011) characterize three homologs of SOC1 in Gerbera hybrida, and describe the contribution of a paralog of SOC1 to the floral and inflorescence phenotype in this species. Indeed, the candidate-gene approach is currently the only way to tackle unusual variation in phenotypic traits. For example, Clianthus maximus is a Fabaceae species with an unusual order of floral organ differentiation that produces inflorescences all year round, all of which abort except during a limited time in the autumn. Song et al. (2011) address the molecular bases of these unusual traits by characterizing MADS-box genes of the ABC model of floral organ identity and the homolog of Leafy.
The characterization of candidate-genes in a growing number of species affords the possibility to reconstruct the evolutionary history of gene families. In many cases, such studies point to the frequency of independent gene duplications and losses. This must be considered in the context of whole-genome duplication (WGD), a major process of the evolutionary history of angiosperms (Van de Peer et al., 2009), which has opened the way for the emergence of novel functions and/or patterns of expression by neo- or sub-functionalization of duplicated genes. Several papers in this Special Issue underline the importance of these phenomena. Vialette-Guiraud et al. (2011) focus on the evolutionary origin of the NAC gene family (NAM, ATAF1/2 and CUC3) in seed plants. Members of this large transcription factor family are essential for the formation of organ boundaries in plants. The members of the NAM subfamily are regulated by micro-RNAs (miRNAs), while CUC3 subfamily members are not. Intriguingly, Vialette-Guiraud et al. demonstrate that separate NAC and CUC3 lineages already exist in the three earliest diverging angiosperm lineages (Amborellales, Nympheales, Austrobaileyales: ANA-grade) but a common ortholog of both groups, possessing a miRNA regulation site, was present in gymnosperms. The most recent common ancestor of angiosperms and gymnosperms thus most likely contained a NAC gene expressed under the control of small, non-coding RNAs. Members of the TCP transcription factor gene family have crucial roles in the regulation of cell division, controlling organ size and bilateral symmetry in several plant species (Martin-Trillo and Cubas, 2009). This family has been studied for a long time in eudicots, where it has been shown to undergo repeated duplication events (e.g. Citerne et al., 2003; Howarth and Donoghue, 2006; Damerval et al., 2007; Zhang et al., 2010). Here, Howarth et al. (2011) compare expression patterns of orthologs of specific TCP genes in radially and bilaterally symmetrical members of the Dipsacales, and find meaningful associations between gene duplication, changes in expression patterns, and flower symmetry and shape. Mondragón-Palomino and Trontin (2011) establish a phylogenetic framework for the TCP family in monocots, and suggest that expansion of these genes in grasses originates from two rounds of WGD. Genetic redundancy as a by-product of small-scale or genome-wide duplication is a mechanism that should ensure robustness of the gene networks underlying developmental processes, while neo- or sub-functionalization would contribute to ‘evolvability’ of these networks. Evolutionary possibilities offered by multiple duplicates is then reviewed by Geuten et al. (2011) in the case of B-class genes, whose role in petal and stamen identity appears conserved in a large group of angiosperm species. Through mathematical modelling, they tackle the role of diversification in activation mechanisms and protein–protein interactions on downstream gene regulation.
As a means to decipher the molecular origin of evolutionary novelties, the candidate-gene approach suffers from several limitations. The first one is that experimental designs in non-model plants often lack the ability to specifically down-regulate genes in order to test functional hypotheses, because transformation protocols are available for a few species only, and are mostly time-consuming and laborious (often involving tissue culture steps). Recently, however, this limitation has partly been relaxed by using VIGS (virus-induced gene silencing), which employs the plant's innate defence system against viruses not only to silence the viral genome but also to target genes of interest. As most plants are susceptible to viruses and many virus genomes can be converted into vectors suitable for VIGS, this valuable technique has been successfully applied to elucidate the role of floral homeotic genes in Nicotiana tabacum and basal eudicots such as Papaver somniferum, Eschscholzia californica, Thalictrum sp, and Aquilegia sp (Drea et al., 2007; Gould and Kramer, 2007; Wege et al., 2007; Becker and Lange, 2010; Di Stilio et al., 2010; Yellina et al., 2010). In this Special Issue, Hands et al. (2011) report the functional analysis of two differentially spliced Papaver somniferum AGAMOUS orthologs, and demonstrate that beyond an expected redundant function in specifying carpel and stamen identity, one of the splicing products has unique roles in septum, ovule and stigma development. This finding is especially important as it is one of the rare cases where a functional significance can be attributed to alternative splicing.
A second limitation to the candidate-gene approach is that the ‘classical’ model species A. thaliana and A. majus are members of the highly derived clade of core eudicots, and their genetic repertoire has most likely changed tremendously since the lineage split from the remaining angiosperms. In addition, the genome of A. thaliana is rather reduced in size and might not be a suitable representative for most angiosperms. Thus, the regulation of flower development in plants distant from arabidopsis might involve genes that are not merely orthologs to A. thaliana genes, and/or the regulatory networks may be different. Flower evo-devo thus needs to move towards a less ‘arabidocentric’ view, even more so when taking into consideration recent studies that show that significant functional differences exist between orthologs of transcriptional regulators. One example for such a phenomenon is the arabidopsis YABBY gene CRABS CLAW (CRC), which is required for carpel and nectary development. Its rice ortholog DROOPING LEAF (DL) is required for leaf midrib formation and carpel organ identity specification and floral meristem termination. The basal eudicot Eschscholzia californica's ortholog EcCRC is also required for floral meristem termination and carpel development but it is not involved in carpel organ identity and leaf development (Alvarez and Smyth, 1999; Yamaguchi et al., 2004; Orashakova et al., 2009).
Alternative approaches to study regulation of floral development in non-model plant species could be the establishment of collections of mutants. This would allow screening for novel mutants and subsequent identification of the mutant locus and the characterization of novel genes. Another option is the generation of organ-specific transcriptome databases for phylogenetically relevant species, followed by large-scale VIGS analysis to transiently knock-down genes and screen for developmentally impaired phenotypes. These candidate-gene-free approaches would allow assumption-free analysis of development in a large phylogenetic context, and should lead to the discovery of novel components of regulatory networks.
The large number of sequences deposited in publicly available databases, derived from sequencing projects and individual gene-identification approaches, has already yielded a wealth of sequence information even for genes that are usually under-represented in transcriptomes, such as transcription factors regulating development that have generally low transcript abundance (Boavida et al., 2011). Such information is available for species that are interesting for their particular phylogenetic position, for example the Floral Genome Project (FGP) has provided floral transcriptome sequences for members of the important ANA-grade that diverged very early in angiosperm evolution. With the advent of next-generation-sequencing techniques at a more affordable price, transcriptome – but also genome – sequencing projects will become feasible for an even wider range of species in the near future. These genomic data will enable reconstructions of more detailed gene phylogenies, allowing analyses of gene birth and death rates, molecular clocks and evolutionary forces acting on developmental regulators. The resulting information will be highly valuable, for example in assessing the contribution of homeotic mutants to morphological innovation and speciation, or in choosing genes for comparative functional analyses, which then will help in understanding ancestral gene functions. Additionally, it will be possible to elaborate new tools to investigate in depth important evolutionary questions, such as the relative importance of selection and developmental constraints, and the evolution of modularity and its role in robustness of genetic networks. The possibility to model gene networks and their morphological outcomes, and to simulate perturbation in network functioning as a result of genetic mutation or environmental cues, should be illuminating in this respect.
ACKNOWLEDGEMENTS
We thank Dr Hélène Citerne for fruitful discussions on flower evo-devo, and for critical reading of an earlier draft of this manuscript.
