Reconstructing the ancestral flower of extant angiosperms: the ‘war of the whorls’ is heating up

Abstract

The origin of the angiosperm flower is a long-standing problem of botany and evolutionary biology. One widely accepted milestone towards solving it is the reconstruction of the ancestral flower of extant angiosperms, here termed ‘AFEA’. A recent approach employing novel methods gave results that were not anticipated. Most notably the reconstructed phyllotaxis of AFEA soon was criticized and sparked a heated debate in the literature. To better explain, clarify, and perhaps cool the debate, we first summarize the results of previous attempts to reconstruct AFEA and contrast them with the more recent, controversial prediction of its structure. We then outline the major arguments made by contrasting parties in the recent debate. Finally, we discuss two key topics, the molecular mechanism of phyllotaxis and the role of gene regulatory networks during flower development and evolution, that may help to clarify the issue in the intermediate future.

Introduction

‘The war of the whorls’ is the title of a seminal paper by Coen and Meyerowitz (1991) that presented the ABC model of floral organ identity specification to a broader audience and attracted numerous scientists (including one of us) into the field of floral developmental and evolutionary genetics. The title is of course a pun but one that developed over the years into some kind of running gag and inspired phrases such as ‘the war of the words’ and ‘brave new whorls’ (Theißen, 2001; Malcomber and Kellogg, 2005). We decided here to play around with this title further because of a recent controversy in the field that revolves not least around the question of whether some organs in a flower, which has only been reconstructed on a computer, are arranged in whorls or in a spiral. So the ‘war of the whorls’ is heating up, but why do whorls mean the world to some scientists? In this article we try to explain the recent controversy and provide some discussion points that we hope will help to dispel it. For some specialized terms we use in the text, the reader may refer to the glossary (Box 1). The controversy started with a thought-provoking (and to some perhaps even provocative) paper by Sauquet et al. (2017), in which the authors reported the results of their project termed ‘eFlower’. The authors conducted a large data set of floral traits from 792 angiosperm (flowering plants, Box 1) species and combined these data with mainly pre-existing phylogenies based on molecular and fossil data. Applying complex statistical models, they inferred the ancestral states of different floral traits eventually to determine the morphology of the flower of the most recent common ancestor (MRCA) of all extant angiosperms, henceforth termed here the ‘ancestral flower of extant angiosperms’ (AFEA). Angiosperms are of utmost ecological and agronomic importance. This may explain why many biologists appear to be obsessed with their origin and that of their most characteristic structure, the flower. Indeed, the origin and early evolution of angiosperms is a long-standing and immensely challenging problem of evolutionary biology. Some reasons, among others, are that flowering plants are morphologically quite different from their molecularly determined sister group, extant gymnosperms (acrogymnosperms), and because the fossil record of angiosperms and gymnosperms is not sufficiently informative (reviewed by Bateman et al., 2006; Frohlich and Chase, 2007). The MRCA of extant seed plants, namely acrogymnosperms together with angiosperms, existed ~310–350 million years ago (Ma), but, according to recent gene- and genome-based estimates, crown group angiosperms originated much later, ~140–240 Ma (Fig. 1) (Magallón et al., 2015; Murat et al., 2017; Sauquet and Magallón, 2018). For comparison, according to recent molecular estimates, the MRCA of acrogymnosperms existed ~270 Ma (Fig. 1) (Magallón et al., 2015), though much older dates, 354 (308–427) Ma, have also been determined (Lu et al., 2014). Since the oldest reliable fossil flowers are also no older than ~200 million years (Myr), at most (Sauquet and Magallón, 2018; Fu et al., 2018; Taylor and Li, 2018), the question arises as to which morphological transformations occurred in which order in the ‘dark tunnel’ of angiosperm evolution—in the stem group that eventually led to AFEA after separation from the lineage that led to extant gymnosperms (Sauquet and Magallón, 2018). During that time, reproductive structures of some gymnosperms, probably a kind of ‘seed ferns’, must have been transformed, very probably in several steps, into a genuine flower. This puts AFEA in a spotlight of biological interest. However, considering AFEA as ‘earth’s first flower’ or ‘the world’s first flower’ (e.g. Monahan, 2017; De-Paula et al., 2018; Ledford, 2018) is a misconception; flowers may well have originated in the stem group of extant angiosperms long before the MRCA of crown group angiosperms existed (Sauquet et al., 2017). Alternatively, it even cannot be completely ruled out that the complex structures and syndrome of features that we recognize as flowers today originated independently more than once in different lineages of angiosperms after they had branched off from the MRCA of extant angiosperms. In the first case, AFEA could have been quite different from the Earth’s real first flower; in the second case, which appears extremely unlikely at present, the reproductive structure of the MRCA of extant angiosperms would not have been a flower yet but some other sort of ancestral reproductive structure. According to Sauquet et al. (2017), AFEA was most probably hermaphroditic and had an undifferentiated perianth of >10 tepals, an androecium comprising >10 stamens, and a gynoecium of >5 carpels (Fig. 1). Both the perianth and the androecium probably had whorled phyllotaxis with three organs per whorl, implying at least four whorls in each of the three organ categories. Moreover, according to the recent reconstruction, the perianth of AFEA was actinomorphic, the stamens had introrse anthers, the carpels were superior and probably arranged in a spiral, and all floral organs were free from each other. Note that organ identity of the undifferentiated tepals (whether they are petaloid or sepaloid) was not inferred in the analysis of Sauquet et al. (2017).

Most features of AFEA, such as bisexuality and actinomorphy, had already been suspected previously by others based on more simple approaches and thus did not spark a controversy (reviewed by Baum and Hileman, 2006; Scutt, 2018; Theißen and Rümpler, 2018). However, AFEA’s phyllotaxis, especially the specific combination of whorled and spiral organ arrangements, soon became heavily debated (De-Paula et al., 2018; Ledford, 2018; Sauquet et al., 2018; Sokoloff et al., 2018). The questions appears as to why some claims by Sauquet et al. (2017) spark such an intense debate; why do leading figures in the field of flower evolution feel compelled to enter into the discussion of a computer model? To answer these questions, we first need to know about previous attempts and the novel approach used by Sauquet et al. to reconstruct AFEA. We then discuss some aspects that have been neglected in the ongoing debate so far, but that eventually may help to clarify the issue.

Early attempts to reconstruct AFEA

Attempts to reconstruct the ancestral flower of extant angiosperms are not new. The floral diversity of extant angiosperms had inspired botanists throughout the 20th century to postulate the structure of AFEA (Engler, 1904; Cronquist, 1988; Takhtajan, 1991). However, these early attempts were ‘merely’ based on ‘expert opinion’ (Sokoloff et al., 2018), and were hampered by the fact that the branching order of extant angiosperms was not known. The situation changed in 1999 when several research groups more or less independently identified the species of the ‘ANITA grade’ (now termed ‘ANA grade’) as the earliest diverging extant angiosperms (reviewed by Kuzoff and Gasser, 2000; Specht and Bartlett 2009; Sauquet and Magallón, 2018). This finding soon boosted fresh attempts to reconstruct AFEA by focusing on floral diversity of ANA grade species and employing simple plausibility arguments or the principle of maximum parsimony. For example, Kuzoff and Gasser (2000) postulated that AFEA probably had bisexual flowers with an undifferentiated perianth arranged in more than two cycles or series, and that differentiated sepals and petals evolved after the origin of the clade of extant flowering plants. Similarly, Endress (2001), by trying to identify potential plesiomorphies among the flowers of extant early diverging angiosperms, suggested that AFEA and the flowers of other early angiosperms were small, pollinated by small insects (dipters, thrips, and moths), and had moderate or low number of floral organs, in probably a spiral, but possibly a whorled arrangement, with a tendency to form organ series in low Fibonacci numbers (3, 5, 8). Moreover, these flowers had tepals (in spiral flowers) with gradual transitions between bract-like, sepal-like, and petal-like forms. They most probably were bisexual but easily evolved unisexuality due to the low level of synorganization between organs (Endress, 2001).

Unfortunately, despite further intensive attempts (e.g. Doyle and Endress, 2000, 2011; Endress and Doyle, 2007, 2009, 2015), reconstruction of ancestral states remained equivocal for many of the most interesting key characters in these parsimony-based approaches. For example, it could not be clarified whether AFEA was unisexual or bisexual, and whether its phyllotaxis was whorled or spiral. The main reason for this ambiguity is the high structural diversity of extant angiosperms, especially among ANA grade lineages. For instance, whereas the earliest branching lineage, Amborella, produces functionally unisexual flowers with spirally arranged floral organs, some representatives of the next lineage, Nymphaeales (e.g. Cabomba), develop bisexual flowers with a uniformly whorled arrangement of organs (Endress, 2001). Another problem in the reconstruction of AFEA is the inapplicability of most floral characters to potential gymnosperm sister groups, both extant and extinct, because of questionable homologies (Bateman et al., 2006; Frohlich and Chase, 2007; Sokoloff et al., 2018). There is evidence, for example, that the extremely diverse group of acrogymnosperms, comprising Ginkgo, cycads, gnetophytes, and conifers, is the sister group of angiosperms, thus hampering homology assignments (Fig. 1) (Frohlich and Chase, 2007).

Due to the difficulties experienced by traditional botany to clarify the origin of the angiosperm flower, evo-devo approaches have also been brought to bear on the problem. For example, based on the analysis of floral organ identity genes and their orthologues in gymnosperms, some authors have postulated that the origin of the flower bauplan occurred in several steps. These studies suggested that, starting from unisexual axes in a gymnosperm, hermaphrodite structures may have evolved first, again implying that the perianth originated later (Theißen et al., 2002; Theißen and Becker, 2004; Baum and Hileman 2006).

The debate

The reconstruction of AFEA by Sauquet et al. (2017) was first criticized by Sokoloff et al. (2018), who questioned the hypothesized switch in phyllotaxis from a whorled perianth and androecium to a spiral gynoecium. Flowers in which organs in whorled and spiral patterns co-occur are termed hemicyclic, and many such cases have been documented, albeit together comprising <1% of angiosperm diversity (Sokoloff et al., 2018). For flowers such as AFEA with outer whorled and inner spiral phyllotaxis, the term cyclospiral has been proposed (Sokoloff et al., 2018). The criticism by Sokoloff et al. (2018) did not refer to the change in phyllotaxis per se but rather to the region where it supposedly occurs. Among the 792 species examined by Sauquet et al. (2017), only four species (Androstachys johnsonii, Fostera bidwillii, Limeum africanum, and Peumus boldus) were considered as illustrating a transition in floral phyllotaxis between androecium and gynoecium. After Sokoloff et al. (2018) reviewed the data set of Sauquet et al. (2017), they stated that none of the mentioned species constitutes a reliable example of a hemicyclic flower. Furthermore, they maintained that hemicyclic flowers are generally very rare among extant as well as fossil angiosperms, and that transitions in floral phyllotaxis are exclusively observed between perianth and androecium but never between androecium and gynoecium. The absence of a certain combination of reconstructed character states among extant taxa might not be unreasonable, and it is certainly not a killer argument against the quality of a reconstruction. Evolution does often proceed in a mosaic fashion through forms that are succeeded by deviant organisms and may soon die out. For example, there is certainly no living animal or plant on earth anymore that shares certain character combinations of the MRCA of extant arthropods, mammals, land plants, vascular plants, and so on. However, as the phyllotaxis of different floral parts is developmentally tightly linked (as discussed more in detail below), the floral organ arrangement of the androecium and gynoecium should not be considered individually. Therefore, Sokoloff et al. (2018) questioned the reconstructed AFEA phyllotaxis and hypothesized that developmental constraints may exist that reliably prevent a change in phyllotaxis between androecium and gynoecium.

Although the criticism by Sokoloff et al. (2018) focused on a very specific feature of flower morphology, it launched a more general debate about the use of extensive data sets and mathematical algorithms to reconstruct and understand morphological evolution (De-Paula et al., 2018; Ledford, 2018; Sauquet et al., 2018). In their reply, Sauquet et al. (2018) argue that albeit the results of their statistical predictions might be reductionist, they should not be rejected summarily simply because they appear unlikely. New ‘unbiased’ approaches may indeed prove beneficial in challenging longstanding dogmas and inevitably biased expert knowledge.

The second major point of criticism regarding the approach of reconstructing AFEA was the homology assessment that was used to delimit the characters of different floral organ types (De-Paula et al., 2018). Sauquet et al. (2017) predominantly used parameters such as shape, number, and function of the anthetic floral organs to assign homology to different floral parts. However, De-Paula et al. (2018) argued that especially for the reconstruction of the perianth and stamens, complementary data on relative organ position, morphological progression during development, and vascularization need to be considered. Furthermore, genetic studies such as the expression profiles of floral organ identity genes should also be used to discriminate between different floral organ types. As an alternative to the approach of Sauquet et al. (2017), De-Paula et al. (2018) therefore suggested an integrative model for homology assessment based on seven specified biological parameters: gene expression, vascularization, relative time of organ initiation, organ initiation pattern, phyllotaxis in early stages, phyllotaxis in maturity, and relative organ position. The suggested approach would certainly allow for a more profound homology assessment. However, currently the number of species for which the different biological parameters are sufficiently researched is rather low. It would thus require a large number of preliminary studies in order to build up a data set that is of sufficient size to include representative species from diverse orders and families. Once such a database is available, it would indeed prove very helpful as not only could it be used for profound homology assessment but also the biological parameters themselves may be utilized to reconstruct their ancestral states. The reconstruction of ancestral gene expression profiles, in particular, would probably allow to better reconcile morphological changes of the flower during early angiosperm evolution (as discussed more in detail below).

Future perspectives

The structure of AFEA has obviously not been clarified to everyone’s satisfaction up to now, so it is likely that the debate about its reconstruction will continue for a while. We will therefore discuss possible future approaches that may help to better understand the origin of the angiosperm flower, including the structure of AFEA.

The molecular mechanism generating phyllotaxis

Due to its regularity and mathematical properties, the fascination of phyllotaxis has extended way beyond botany for centuries (Kuhlemeier, 2017). Different arrangements are well known, such as distichous, decussate, spiral, and whorled. In many extant angiosperms, phyllotaxis changes during development, for example at the transition from the vegetative region to the flower. In ANA grade angiosperms, a change from decussate to spiral phyllotaxis is common at the vegetative to floral transition (e.g. in Austrobaileya and Trimenia); changes in other taxa are from distichous to spiral (e.g. in Eupomatia) and from distichous to whorled phyllotaxis (e.g. in Annonaceae) (reviewed by Endress and Doyle, 2007). Changes in phyllotaxis within the flower at the transition from one category of floral organ to another have also been well documented. The most common change within the flower is found in many core eudicots where the sepals of the calyx are in a spiral arrangement whereas the other organs are whorled; this change is accompanied by considerable shortening of plastochrons and narrowing of organ shape (Endress and Doyle, 2007). In contrast, in many Magnoliaceae, the perianth is whorled but both the androecium and gynoecium are spiral (Endress and Doyle, 2007). Such changes in phyllotaxis are, at least in part, consequences of changes in the sizes of the initiated organs—floral organs are usually narrower than vegetative organs, and stamens are narrower than perianth organs (sepals, petals, or tepals).

The hypothesis offered by Sokoloff et al. (2018)—that a fundamental developmental constraint prevents the transition in phyllotaxis between androecium and gynoecium—is tempting, even though the mechanistic basis of such a constraint remains unknown. The authors speculate that the class C floral homeotic genes, which control stamen and carpel identity, ‘also influence mechanisms that regulate floral phyllotaxis’ (Sokoloff et al., 2018). However, in a class C mutant of Arabidopsis thaliana, for example, floral organs still appear in whorled phyllotaxis, except when, under conditions suboptimal for flowering, the floral meristem is transformed into an inflorescence meristem (Mizukami and Ma, 1997). Moreover, almost nothing is known about how any of the floral organ identity (‘ABCDE’) genes interact with the regulatory mechanisms of phyllotaxis (Kuhlemeier et al., 2017). This fundamental ignorance exists despite the fact that much is known about both the developmental genetics of floral organ identity (reviewed by Stewart et al., 2016; Theißen et al., 2016) and the molecular mechanisms generating phyllotaxis (for a primer, see Kuhlemeier, 2017).

The currently available experimental evidence, as well as computational modelling, suggest that the regular spacing of organs as observed in phyllotaxis is largely controlled by a reiterative mechanism that depends on the phytohormone auxin. The mechanism is based on the active transport of the auxin by cellular influx and efflux carriers, such as AUX1 and PIN1. Models of phyllotaxis invoke the accumulation of auxin at organ initials and removal of auxin through developing vascular strands (Kuhlemeier, 2017). The combined experimental results suggest models in which an autoregulatory loop between auxin and subcellular PIN localization generates auxin maxima. How auxin and PIN interact is not exactly known. Several different hypotheses yield the same phyllotactic pattern during computational modelling and thus cannot distinguish between particular underlying molecular mechanisms (Kuhlemeier, 2017). What would be needed to better understand phyllotactic patterning would be accurate quantitative data about auxin concentrations and polar auxin transport at high spatial (cellular) resolution. However, experimental methods to obtain such data are currently not available, implying that we do not exactly know yet how phyllotactic patterning is driven (Kuhlemeier, 2017). Once such methods have been developed and applied, a better understanding of the development of flower phyllotaxis at deep mechanistic, molecular levels may eventually provide strong evidence as to whether a cyclospiral flower like the AFEA reconstructed by Sauquet et al. (2017) could have existed, or whether there are indeed developmental and mechanistic constraints that appear almost impossible to overcome. If the latter is the case, the structure of the recently reconstructed AFEA might indeed be considered too bizarre to be true. It should be noted here that the phyllotaxis of AFEA was probably not even entirely fixed as at least the perianth organs of some early diverging angiosperm species such as Nuphar advena show a tendency for irregular arrangement.

Reconstruction of gene regulatory networks

Another approach that may help to clarify the morphological transitions during the origin of the angiosperm flower and the structure of AFEA is comparison of the gene regulatory networks (GRNs) that control reproductive organ formation in extant seed plants. All aspects of plant development, including flower development, are largely controlled by GRNs that are mainly composed of transcription factors (TFs), miRNAs, and the genes that encode TFs and miRNAs (Vialette-Guiraud et al., 2016; Wils and Kaufmann, 2017). In recent years, numerous techniques enabled determination of protein–DNA and protein–protein interactions at large scale. Chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) or hybridization to tiling arrays (ChIP-chip), the yeast two-hybrid system (Y2H), as well as gene perturbation and genome-wide expression analyses (transcriptomics) have all been used to gain detailed insights into the GRNs controlling flower development (for a recent review focused on A. thaliana, see Wils and Kaufmann, 2017). These GRNs comprise genes involved in the transition to flowering, and in the specification of inflorescence, floral meristem, and floral organ identity.

Since the identity of floral organs is specified by conserved floral homeotic genes (‘ABC genes’, later extended also to ‘D and E genes’), initial evo-devo studies on the origin of the flower were focused on a better understanding of the evolution of these genes, almost all of which encode MADS-domain TFs (e.g. Theißen and Saedler, 1995; Theißen et al., 2000). Putative orthologues of class B and C floral homeotic genes (DEFICIENS/GLOBOSA-like genes, also known as APETALA3/PISTILLATA-like genes and AGAMOUS-like genes, respectively) could meanwhile be identified in all major groups of extant gymnosperms, whereas such genes appear to be absent from non-seed plants (Gramzow et al., 2014). The expression patterns of putative gymnosperm B and C genes are characteristically similar to those of class B and class C genes in angiosperms; class C genes are generally expressed in both male and female reproductive organs, whereas class B genes are exclusively expressed in male reproductive organs (Winter et al., 1999; Theißen et al., 2002). These data suggest that a kind of ‘BC system’ of reproductive organ specification had been established in the stem group of extant seed plants >300 Ma, after the lineage that led to extant ferns had already branched off, but before the separation of the lineages that led to extant gymnosperms and angiosperms (Fig. 1) (Winter et al., 1999; Theißen et al., 2002; Baum and Hileman, 2006; Theißen and Rümpler, 2018). This system may have worked by generally distinguishing reproductive organs from vegetative organs via the expression of C genes, and distinguishing male from female reproductive organs by differential expression of B genes (B genes ‘off’, female organs will develop; B genes ‘on’, male organs will develop).

Recently, evidence was provided, that, as long suspected, the link between class B genes (and probably also class C genes) and one of their upstream activators, the meristem identity gene LEAFY (LFY), was probably already established in the stem group of extant seed plants (Moyroud et al., 2017). However, understanding flower origin may require more than just the recognition of this mini-network comprising the link between some meristem and organ identity genes. It would be helpful to reconstruct the GRNs controlling the development of extant gymnosperm reproductive structures in the same detail as those controlling flower development in the angiosperm A. thaliana. In an ideal case, this would be done for at least one representative of all the five presumably monophyletic groups of gymnosperms that still exist today (gnetophytes, cycads, Ginkgo, Pinaceae, and the remaining conifers) (Lu et al., 2014), though even starting with just one model would already be extremely helpful. Admittedly, there would be high hurdles to overcome; gymnosperms have very large genomes and are not yet easily amenable to genetic transformation. However, some major techniques, such as ChIP-seq, ChIP-chip, Y2H, and transcriptomics, are also applicable to gymnosperms. These methods thus might be used to reconstruct the GRNs that control the development of reproductive structures in a ‘model’ gymnosperm species such as Norway spruce (Picea abies). Involving more angiosperm and gymnosperm species may enable the reconstruction of the respective GRNs at the base of extant gymnosperms and angiosperms roughly about 270 Ma and 200 Ma, respectively. One then could try to reconstruct the GRNs of the ancestral reproductive structure of the MRCA of extant seed plants that may have existed ~330 Ma. As gymnosperms are paraphyletic, the MRCA of extant seed plants was certainly also some kind of gymnosperm. Subsequently one could identify changes that occurred during the origin of the flower, by analysing the differences between the different ancestral GRNs. The addition or loss of components (‘nodes’ in network terminology, in this case genes or proteins), changes in the structures (nucleotide or amino acid sequence) of components, and the loss or gain of interactions (‘edges’ of the networks) all could have contributed to the critical changes that occurred when the GRN of an ancestral gymnosperm reproductive structure turned into an ancestral flower GRN. As indicated above, it is already quite obvious that major gymnosperm and angiosperm nodes, such as meristem and organ identity genes, have been conserved, whereas major structural changes occurred during flower origin, such as the transition from unisexual to hermaphroditic structures. We are tempted to speculate, therefore, that the rewiring of pre-existing components by changes in, for example, cis-regulatory elements of network components rather than major changes in the composition of the components were most crucial for the origin of the angiosperm flower.

New fossils

Last, but not least, one should never forget that the serendipitous finding of new, informative fossils could help a lot. Whereas previous genuine fossils of angiosperm flowers were not older than 130 Myr and thus considerably younger than AFEA as reconstructed by Sauquet et al. (2017), the recently described and carefully reconstructed Nanjinganthus dendrostyla, which lived ~174–204 Ma (probably closer to 174 Ma, i.e. latest Early Jurassic), appears to be almost as old as AFEA (Fig. 1; Fu et al., 2018; Taylor and Li, 2018). Most interestingly, according to Taylor and Li (2018), the flower of N. dendrostyla shares 23 out of 29 (i.e. 79%) floral characteristics with the ancestral flower reconstructed by Endress and Doyle (2009), but shares only 6 out of 13 (46%) floral characteristics with the reconstruction by Sauquet et al. (2017). Does that imply that the traditional ‘low-tech’ reconstruction by Endress and Doyle (2009) a decade ago has yielded better results than the highly sophisticated ‘big data’ approach applied by Sauquet et al. (2017)? Not so fast, especially because the exact phylogenetic position of N. dendrostyla with respect to the species that carried AFEA remains unknown. Together with some other reservations outlined above, however, the new fossil may tell us that we might still be quite far away from a general consensus about how AFEA really looked.

Concluding remarks

AFEA represents an important stage in the evolution of the flower of angiosperms. Nevertheless, the intensity of the debate about a mere computer reconstruction of a flower that has been extinct for >100 Myr may surprise many. It appears likely that a myriad of reasons contributes to this. Two points might be of more general interest. It is a recurrent theme in the history of science that when new methods and concepts achieve results that contradict long-held views, these novel findings are considered with considerable scepticism, for good reasons. Numerous examples are provided by the employment of molecular markers to clarify phylogenetic relationships. For example, when based on DNA markers, Amborella and the other ANA grade angiosperms were identified as the earliest branching flowering plants (reviewed by Kuzoff and Gasser, 2000; Specht and Bartlett 2009), this claim raised quite some debate (e.g. Soltis and Soltis, 2004). Likewise, when it was shown that gnetophytes are not the sister group of angiosperms (thus making the morphology-based anthophyte hypothesis less likely), but are more closely related to other extant gymnosperms, and that even all extant gymnosperms are very probably monophyletic (e.g. Winter et al., 1999; Frohlich and Chase, 2007), many traditional botanists found it hard to swallow this result and its many implications. Today, these findings are already almost textbook knowledge. One cannot reverse the logic of the argument, of course, since many apparently bizarre ‘findings’ in fact eventually turned out to be wrong. Given the serious and reasonable concerns by Sokoloff et al. (2018) and De-Paula et al. (2018), it seems that the jury is still out with respect to both the methods used and the results obtained by Sauquet et al. (2017).

However beyond the questionable structure of AFEA, there might be an even deeper reason for the debate. In a way, the whole debate documents how frustratingly little we still know about even major aspects of flower development and evolution, such as how exactly simple phyllotactic patterning works and evolves. Even more innovative approaches might be required to clarify such issues, and the required methods may at best only now be appearing on the horizon.

Acknowledgements

We thank Frank Wellmer for his kind invitation to write a Flowering Newsletter Review. We are indebted to Richard Bateman and two anonymous reviewers for numerous helpful comments on a previous version of the manuscript. We are also grateful to H.G. Wells (1898), Aldous Huxley (1932), Enrico Coen, and Elliot Meyerowitz for inspiration, and to the OVG Münster for its courage to see the extant forest for the fossil trees.

References