The significance of developmental robustness for species diversity

Abstract

Background The origin of new species and of new forms is one of the fundamental characteristics of evolution. However, the mechanisms that govern the diversity and disparity of lineages remain poorly understood. Particularly unclear are the reasons why some taxa are vastly more species-rich than others and the manner in which species diversity and morphological disparity are interrelated.

Scope and Conclusions Evolutionary innovations and ecological opportunities are usually cited as among the major factors promoting the evolution of species diversity. In many cases it is likely that these factors are positively reinforcing, with evolutionary innovations creating ecological opportunities that in turn foster the origin of new innovations. However, we propose that a third factor, developmental robustness, is very often essential for this reinforcement to be effective. Evolutionary innovations need to be stably and robustly integrated into the developmental genetic programme of an organism to be a suitable substrate for selection to ‘explore’ ecological opportunities and morphological ‘design’ space (morphospace). In particular, we propose that developmental robustness of the bauplan is often a prerequisite for the exploration of morphospace and to enable the evolution of further novelties built upon this bauplan. Thus, while robustness may reduce the morphological disparity at one level, it may be the basis for increased morphological disparity and for evolutionary innovations at another level, thus fostering species diversity.

The disparity of plant and animal body plans

Evolution has brought about an enormous disparity of forms (Minelli, 2016). The variation in animal body plans is vast, from sponges to cnidarians, annelid worms, molluscs, nematode worms, arthropods (including insects) and vertebrates; land plants are similarly varied in their construction, ranging from diverse bryophytes (liverworts, mosses and hornworts) to lycophytes, ferns and their allies, gymnosperms and angiosperms (flowering plants). However, huge variations exist in the numbers of species that belong to different taxonomic groups. For example, approximately 70 % of all described extant animal species are insects (Zhang, 2013). Similarly, approx. 90 % of all land plant species are angiosperms (Roskov et al., 2015). Moreover, and surprisingly, these differences in diversity (i.e. the number of species in a taxon) are not necessarily reflected in the disparity (i.e. the variety of form) encompassed by clades, as diversity and disparity levels appear to be decoupled (Hughes et al., 2013; Oyston et al., 2016).

The asymmetry of diversity patterns in the tree of life was recognized almost 100 years ago by the British botanist J. C. Willis and the statistician G. U. Yule who introduced the term ‘hollow curve’ to describe the distribution of species numbers across different taxa (Willis and Yule, 1922). They also demonstrated that this pattern is observed across a diversity of taxonomic ranks and phylogenetic groups (Willis and Yule, 1922). Since then, this phenomenon has attracted considerable attention, and several hypotheses have been formulated to explain it (e.g. Hutchinson, 1959; Minelli et al., 1991; Farrell, 1998; Mayhew, 2007; Bokma et al., 2014). In general, these hypotheses fall into two categories: either asymmetries are the result of stochastic events or, alternatively, biological explanations have been sought to account for the differential success of various taxa (Scotland and Sanderson, 2004; Bokma et al., 2014). Overall, the mechanisms promoting and limiting diversity and disparity remain poorly understood (Hughes et al., 2013; Oyston et al., 2015, 2016). However, understanding why different taxa are vastly more diverse than others is one of the most fundamental questions in evolutionary biology and biodiversity research because this is intimately linked to our understanding of how evolution proceeds and how new species and forms originate. Here, we argue for the importance of developmental robustness as a major biological driver of diversity patterns. We suggest that evolutionary innovations (novelties), ecological opportunities and developmental robustness constitute a self-reinforcing mechanism that may have been critical for the success of many taxa.

Random processes may generate substantial differences in species distribution across taxa

Although the differential success of taxa provokes biological explanations, marked asymmetries in species diversity might also be generated by random processes. This is well illustrated by the broken stick model (MacArthur, 1957; Dial and Marzluff, 1989; Ricklefs, 2003). In this model, a stick is randomly broken at several points. Each of the resulting pieces represents a taxon, and the length of the piece is proportional to the species richness of that taxon. The process results in few relatively large pieces and many smaller ones, thereby generating a pattern that resembles the inequalities in species richness observed in nature (Dial and Marzluff, 1989; Wilkinson, 2011). Beyond the broken stick model, the principle of maximum entropy, power laws and a broken plate model, among others, have also all been shown to reproduce patterns of species diversity that are often very similar to those observed in nature (Minelli et al., 1991; Wilkinson, 2011; Bokma et al., 2014). Some of these models can also be applied to other seemingly unrelated phenomena such as the frequency of different words in a book or the differential size of cities (Baek et al., 2011; Bokma et al., 2014). Moreover, many of the models do not take biological or evolutionary considerations into account yet describe species distributions sometimes very accurately. This has led to the proposition that differential species richness is often created just by random processes (Wilkinson, 2011; Bokma et al., 2014). However, even when these stochastic processes are factored in, many groups are significantly more species-rich than expected, indicating that additional explanatory factors are needed (Dial and Marzluff, 1989; Ricklefs, 2003; Sims and McConway, 2003; Harmon, 2012). Random processes may therefore offer a suitable ‘null hypothesis’ for subsequent analyses of species richness.

However, it should also be remembered that even a well-fitting model does not simultaneously provide a mechanistic explanation for its own fit (Adamic, 2011). For example, the fact that the principle of maximum entropy reflects some biodiversity patterns very well does not explain why this is the case (Adamic, 2011). Either way, biological explanations must be sought.

Species diversity can be studied at different levels

The size of a taxonomic group is determined by the speciation rate, extinction rate, carrying capacity and age of the clade (together, these factors are sometimes referred to as proximate explanations of species diversity; Mayhew, 2007). Studying diversification rates and clade age will therefore explain how certain clades became more species-rich than others. This begs the question of why speciation rates and clade ages differ among taxa. Two factors that we focus on here are ecological opportunities and evolutionary innovations (both sometimes referred to as ultimate explanations of species diversity; Mayhew, 2007). Ecological opportunities may, for example, increase speciation rates, thus leading to an increase in species richness.

Cladogenesis and clade age determine species diversity

Studying patterns of cladogenesis (i.e. speciation and extinction rates) and the age of clades can yield important insights as to how some taxa became more species-rich than others. If the speciation rate exceeds the extinction rate, species numbers in any given taxon will increase over evolutionary time. Conversely, if the extinction rate exceeds the speciation rate, species numbers will decline (Mayhew, 2007). The carrying capacity (i.e. a limit to the species richness of a given clade) is probably also an important factor affecting diversity and may lead to a decoupling of clade age and clade diversity (Mayhew, 2007; Rabosky et al., 2012). For some of the most speciose taxonomic groups, hypotheses have been raised as to which of these factors explains their species richness. For example, beetles [comprising approx. 390 000 species (Zhang, 2013) and representing by far the largest order of animals] may have attained such spectacular diversity because of their evolutionary age (many modern beetle lineages originated some 200 Mya) and their impressive rates of lineage survival (Hunt et al., 2007; but for an alternative view see Rabosky et al., 2012). In contrast, the speciation rate of beetles appears comparable to that of other taxa (Hunt et al., 2007).

Studying patterns of cladogenesis is complicated by the fact that the current biodiversity is only a snapshot of the evolutionary history of a taxonomic group, and at best permits only inferences regarding the processes shaping this diversity. An intriguing example illustrating this problem is provided by a recent study of cycads (Nagalingum et al., 2011). These gymnosperms are considered ‘living fossils’. They originated approximately 270 Mya, remained morphologically largely unchanged since then and constitute a relatively small group of approx. 300 extant species (Nagalingum et al., 2011). However, molecular clock estimates indicate that the living species diversity of cycads originated in a ‘burst’ of speciation that began only about 12 Mya (Nagalingum et al., 2011). This illustrates that detailed studies of clade ages, speciation and extinction rates can often yield surprising insights into the processes that shape biodiversity. Molecular clock estimates and/or a solid fossil record are important means to infer the cladogenetic patterns that determine the evolutionary success of taxa. A detailed understanding of cladogenesis patterns and clade age is therefore essential.

Ecological opportunities and species diversity

A change in environmental conditions can often expose new ecological niches or resources that can be an enormous trigger for speciation, leading to adaptive radiations (Losos, 2010). There are several definitions for the term ‘adaptive radiation’, but here we take it to be what happens when ‘natural selection drives divergence of an ancestral species into descendants that are better able to exploit ecological opportunity’ (Glor, 2010). Adaptive radiations can cause bursts in speciation rates (Glor, 2010). Plants and animals colonizing hitherto uninhabited areas provide prominent examples (Glor, 2010). However, bursts in speciation rates may not always be necessary for adaptive radiations to occur (Glor, 2010). As described above, the high species diversity of beetles might be related to the age of the clade, not to exceptionally high speciation rates. In this context, the coevolution of beetles and angiosperms may be considered as a self-perpetuating and reinforcing ecological opportunity (Rabosky, 2009; Losos, 2010; Nyman, 2010).

It is striking that certain taxa appear to have radiated successfully on several occasions in several different ecosystems (Losos, 2010). Insects, for example, are important constituents of most ecosystems on land (Fisher, 1998). Similarly, flowering plants are the dominant plant group in most terrestrial ecosystems (Berendse and Scheffer, 2009). It thus appears evident that different taxa ‘respond’ differently to ecological opportunities. Angiosperms co-evolved with insects to a considerable extent, but extant gymnosperms (the closest living relatives of angiosperms) did not (Grimaldi, 1999; Frohlich and Chase, 2007; Crepet and Niklas, 2009). This demonstrates that ecological opportunities alone are often not sufficient to drive species diversification. Rather, ecological opportunities can often only be ‘explored’ if the developmental genetic programme of the organism is capable of evolving the required morphological diversity and evolutionary adaptations. Evolutionary innovations (adaptations or pre-adaptations) are therefore important requirements for seizing an ecological opportunity (Glor, 2010; Losos, 2010).

Evolvability, evolutionary innovations and species diversity

Evolvability describes the ability of an organism to ‘acquire novel functions through genetic change, functions that help the organism survive and reproduce’ (Wagner, 2005). This definition ties evolvability to the propensity of the organism to generate key innovations. In turn, a key innovation ‘opens up a new character space (or breaks constraints) that potentially allows the occupation of more niches’ (Galis, 2001). According to this definition, key innovations are of critical importance for exploring ecological opportunities. Prominent examples include the wings of insects or the flowers of angiosperms. The wings of insects helped to explore new ecological niches that were hitherto unapproachable (Mayhew, 2007). The petals of flowering plants helped to attract pollinators, establishing plant–pollinator interactions that presumably fostered the diversification of both groups (Grimaldi, 1999; Fenster et al., 2004; Bronstein et al., 2006). The latter example also illustrates that the interplay between evolutionary innovations and ecological opportunities is not a unidirectional process. Evolutionary innovations themselves may create new ecological opportunities that can subsequently be further explored by new innovations (Laland et al., 2011, 2014).

Nevertheless, in many cases the relevance of evolutionary innovations for species diversity remains controversial (Galis, 2001; Glor, 2010). Even the wings of insects may not have led instantly to an increase in species richness as the early-diverging winged insects (dragonflies and damselflies) are not particularly more species-rich than more basal wingless groups (Mayhew, 2002, 2003, 2007; Nicholson et al., 2014). Similarly, early-diverging angiosperms are not particularly species-rich, although the key innovations of flowering plants – the petal, bisexuality and carpels – were established at the base of (or very early during) angiosperm evolution (Theißen and Becker, 2004; Endress and Doyle, 2009). One interpretation for this observation is that evolutionary innovations provide the intrinsic potential for diversification, but that this potential can only be explored if an extrinsic ecological opportunity is present or if additional innovations are established (Galis, 2001; Mayhew, 2007).

However, in some cases, the evolutionary innovation as well as the ecological opportunity seemed to be available, yet species diversity did not increase. That is most obvious in sister taxa that share the same evolutionary innovations, are exposed to similar environments and yet possess very different species numbers. The Austrobaileyales, for example, constitute a group of early-diverging angiosperms that consists of some 90 species (Soltis and Soltis, 2004; Roskov et al., 2015). The sister group of the Austrobaileyales are the magnoliids + eudicots + monocots that together comprise more than 300 000 species (Roskov et al., 2015). Both groups possess flowers, both had the chance to diversify in conjunction with pollinators, yet the species numbers differ by three orders of magnitude.

Developmental robustness

Developmental robustness might be defined as the ‘persistence of an organismal trait under perturbations’ (Felix and Wagner, 2008). Robustness entails the persistence of traits against environmental, stochastic and genetic perturbations (Felix and Wagner, 2008; Felix and Barkoulas, 2015; Mestek Boukhibar and Barkoulas, 2016). It is evident that the robust determination of certain traits is of critical importance for the organism (Kitano, 2004; Lachowiec et al., 2016). Returning to insects, the wings represent a well-adjusted functional unit, and almost all developmental deviations in wing number would almost certainly have adverse effects on the ability to fly (Grodnitsky and Morozov, 1993). A robust genetic mechanism that programs the development of two and not three or more pairs of wings appears therefore as essential as the evolutionary innovation ‘wing’ itself [but note that in a number of clades such as Dipterans (flies, mosquitos) and Coleopterans (beetles) one functional pair of wings evolved, which is robustly expressed within those clades, however]. From a developmental perspective, robustness is probably always important if two or more structures need to function in a highly integrated manner with each other. In analogy to the insect wing, many flowers might be optimized for a specific number of floral organs positioned in a specific orientation to each other to attract pollinators, and deviations from that pattern may usually lead to reduced pollination success.

Developmental robustness and species diversity might be correlated

In plants, the importance of developmental robustness is especially evident for floral structures involved in interactions with specific pollinators (Møller, 1995). Flower colour and floral form are key features recognized by pollinators (Van der Niet et al., 2014) and hence their robust specification is vital for successful propagation. This is probably best exemplified in orchids, where developmental precision is of critical importance for pollinator visitation (Armbruster, 2014). The flowers of the vast majority of orchids possess six perianth organs, termed tepals (often also misleadingly called sepals and petals) organized in two whorls. Whereas the three outer-whorl organs are usually relatively similar to each other, the three organs of the inner perianth whorl develop into distinct shapes, with two lateral tepals and, most characteristically, a median inner tepal called the lip, giving the orchid flower its zygomorphic appearance (Rudall and Bateman, 2002; Mondragon-Palomino and Theissen, 2008). This floral bauplan appears remarkably robust: flowers from a single species are very similar to each other, reiterating the same number of perianth organs and the same floral structure (although intraspecific variations in, for example, the size of flowers are well documented) (Bateman and Rudall, 2006; Morales et al., 2010; Ackerman et al., 2011).

This is in stark contrast to the degree of developmental robustness of the floral bauplan and perianth organ number in some other angiosperms, particularly in early-diverging species where considerable variation can be observed, even for flowers from the same plant (Endress, 2001; Warner et al., 2008, 2009). In Nuphar lutea, for example, perianth organs vary in number, and floral organ identity is sometimes blurred and appears to depend on environmental conditions (Warner et al., 2008, 2009). Organ identity is poorly defined and perianth organ numbers vary strongly also in other species of Nymphaeales (Schneider et al., 2003; Warner et al., 2008, 2009).

Intuitively, one might assume that taxa with a low degree of robustness are very species-rich, simply because they occupy a wide spectrum of the phenotypic space. However, at least in terms of the floral bauplan, the opposite seems to be the case: the aforementioned orchids with their highly standardized floral bauplan comprise more than 25 000 species whereas Nymphaeales are a relatively small order with only approx. 100 species (Roskov et al., 2015).

A positive relationship between developmental robustness and species diversity might also be observed in animals. For example, the most speciose taxa of Pancrustaceae (hexapods and crustaceans) are the hexapods (insects and allies) with more than 1 000 000 species (Zhang, 2013). Hexapods have a highly uniform body plan with virtually no variation in the number of leg-bearing segments (Fusco and Minelli, 2013). In contrast, Branchiopoda and Remipedia, two crustacean groups which are considered to be closely related to hexapods (von Reumont et al., 2012; Oakley et al., 2013; Edgecombe and Legg, 2014) show substantial variations in the number of appendage-bearing segments (Longhurst, 1955; Koenemann et al., 2006; Fusco and Minelli, 2013). Intriguingly, those groups are very species-poor when compared with hexapods, with Branchiopoda comprising only some 1200 and Remipedia hardly 20 species (Zhang, 2011). Clearly, Branchiopoda and Remipedia differ in several morphological characters from hexapods (most evidently adaptations to aquatic vs. terrestrial life) that may well be responsible for differences in species numbers; but we consider it conceivable that differences in the robustness of segment number determination contributed to the differential success of the taxa. Along these lines of reasoning, it is interesting to note that the Malacostraca, which represent another huge group of Pancrustaceae with approx. 40 000 species (Zhang, 2011), appear to have a quite fixed number of thoracic segments bearing appendages (Fusco and Minelli, 2013).

Developmental robustness as an important mechanism to foster the origin of morphological complexity and species diversity

When a trait is stably expressed, genetically well-defined and constrained within a narrow phenotypic space, it provides a well-suited substrate for natural selection (Peterson et al., 2009). This is in contrast to developmentally less robust systems, where stochastic factors and environmental fluctuations influence the phenotype to a larger extent, giving rise to a broad phenotypic space that is genetically not well constrained. Consider, for example, a scenario in which perianth organ number varies between different flowers of the same plant because it is not well constrained by the developmental genetic programme (Fig. 1A). The origin of a more complex, synorganized floral structure in which the different perianth organs possess different morphologies to form one characteristic ‘superorgan’ adaptively honed to attract specific pollinators is difficult to imagine in such a situation. This is because the evolution of such a synorganized floral structure probably requires a genetic programme that coordinates and integrates the development of all perianth organs. However, only if the number of perianth organs is robustly determined in the first place can such a dedicated gene regulatory network (integrating the development of all organs but also enabling the formation of distinct organ morphologies in a coordinated manner) evolve (Fig. 1B, C) [as presumably happened during orchid evolution (Mondragon-Palomino and Theissen, 2008)]. Indeed, a correlation between floral synorganization and developmental robustness of the floral organ number is observed for at least some taxa: whereas flowers from, for example, Papaver, that possess a low degree of synorganization, produce fewer organs under starved conditions, flowers with a higher degree of synorganization (e.g. Mimulus) are smaller but may show no variations in floral organ number when grown under starved conditions (Endress, 2006).

Fig. 1.

Hypothetical scenario of how developmental robustness may contribute to the origin of species diversity. In (A), the developmental system determining perianth organ number is not very robust and a diversity of organ numbers is observed. When developmental robustness increases during evolution, a certain organ number becomes fixed (B). Superimposed on that fixed organ number, a new ‘dimension’ in the phenotypic space can be explored, i.e. perianth organs with distinct morphologies can evolve (C). Interactions with specific pollinators may stabilize the different morphologies, thus further increasing robustness and eventually contributing to the origin of new species.

Similarly, if the number of trunk segments varies in an arthropod, the evolution of functional wings might be complicated because the developmental genetic programme does not allow a precise differentiation between segments that are destined to develop wings from those that are not.

Robustness at the level of the floral bauplan or at the level of the insect body plan may therefore facilitate the evolution of morphological disparity ‘superimposed’ on this bauplan (Fig. 1C). Orchid flowers do indeed possess a great degree of morphological disparity: the size, form and colour of the perianth organs differ considerably between different species (Rudall and Bateman, 2002; Bateman and Rudall, 2006; Endress, 2016). It is mainly the general floral bauplan that appears to be rather robust between, as well as within, species. The same applies to insects: the insect body plan appears to be developmentally and evolutionarily very robust, yet superimposed on that groundplan insects explore a vast range of forms (from beetles to mantids, flies, true bugs and butterflies). We suggest that the conservation and developmental robustness of the floral bauplan or the insect body plan actually underpin their impressive diversity and morphological disparity (Fig. 1). In the case of orchids, the robustness of the bauplan (e.g. low and fixed number of perianth organs) may have enabled the evolution of a perianth that functions as one higher-level module (one ‘superorgan’) within which the individual whorls and even individual elements within a whorl (lower-level modules) develop in a coordinated manner. At the same time, subfunctionalization of genetic programmes ensures that different organs can, to some extent, evolve independently from each other. Together, this may have given rise to the astounding diversity and disparity of orchid flowers.

The robustness of the floral bauplan may be further increased by interaction with specific pollinators. Pollinators may recognize only specific floral forms and may not tolerate significant deviations from the underlying patterns (Møller, 1995, 2000). Conversely, an increased robustness of the floral bauplan may also foster the evolution of developmentally robust structures within the pollinators, enabling them to efficiently exploit the resources on offer (e.g. nectar) (Møller, 2000). Co-evolution between plants and pollinators may thus lead to a reciprocal increase in developmental robustness of both partners.

Developmental robustness vs. developmental constraints

The necessity of developmental robustness for the evolution of the zygomorphic orchid perianth appears to be relatively straightforward to explain. However, in many cases where substantial robustness is observed, the developmental and evolutionary relevance is less clear. Robustness in perianth organ numbers in Brassicaceae might be a good case in point. Virtually every Arabidopsis thaliana flower possesses four petals and four sepals. However, the floral structure is much less complex than that of an orchid flower. Although the sepals and petals are distinct from each other, there is usually no further specialization within the petal or sepal whorl and the perianth is often considered to be radially symmetric (Endress, 1992; Busch and Zachgo, 2009). In line with this absence of specialization, A. thaliana is (even though being predominantly self-pollinating) likely to be pollinated by a diversity of insects (Hoffmann et al., 2003), and it is not immediately clear how a specific number of perianth organs is advantageous for pollination success. Interestingly, however, the stability in perianth organ number is not only characteristic for A. thaliana but for most of the species in the family Brassicaceae, the majority of which possess four sepals and four petals (Endress, 1992). Indeed, most eudicotyledonous and monocotyledonous families possess such a stability in perianth organ number, with eudicots usually having a tetramerous or pentamerous perinath and monocots having a trimerous perianth (Soltis et al., 2003; Endress, 2011). The Brassicaceae are an intermediately sized flowering plant family, comprising approx. 3800 species (Roskov et al., 2015). It is unclear whether robustness of the floral bauplan has been evolutionarily selected for in this case or whether it arose as a by-product of another, developmentally more important process. However, although floral merism is remarkably uniform in this family, inter- and intraspecific heritable variations in corolla shape are well documented, and some of these differences may have been the result of pollinator-mediated evolutionary divergence (Gomez et al., 2009, 2015). Also in this case, we hypothesize that differential corolla colours and shapes (e.g. the form and size of the petals and their position relative to each other) can only be established because the underlying genetic system robustly determines the tetramerous nature of the perianth.

However, the question remains why robustness is even retained in species that are predominantly selfing and hence do not require well-defined corolla morphologies. Importantly, developmental robustness also entails robustness against mutations (Felix and Wagner, 2008). The selfing syndrome evolved relatively recently in many Brassicaceae lineages (Vekemans et al., 2014). The relatively recent origin of self-compatibility, coupled with the robustness of petal number determination against mutational changes, may have preserved the four-fold perianth in selfing Brassicaceae.

Admittedly, the scenario we propose is quite speculative and the four-fold merosity of the Barassicaceae perianth is also compatible with the hypothesis that developmental constraints restrict floral organ numbers. That would be similar to the almost invariant number of seven cervical vertebrae in mammals (including humans), which has been proposed to arise from severe pleiotropic effects if the number of vertebrae deviates from seven (Galis, 1999; Varela-Lasheras et al., 2011). Interestingly, the determination of vertebrae number appears not to be especially robust in humans. Homeotic transformations of the seventh cervical vertebra into a rib-bearing vertebra are relatively frequent, but the affected individuals usually die early in their development (Galis et al., 2006).

In A. thaliana, mutants showing deviations in petal number have often also other developmental defects (Brewer et al., 2004; Maier et al., 2009), which supports the notion that developmental constraints restrict petal number variation. However, in the related Brassicaceae species Cardamine hirsuta, natural variation in petal numbers is observed without obvious adverse pleiotropic effects (Hay et al., 2014; Monniaux et al., 2016; Pieper et al., 2016). Clearly, more studies are needed to determine the evolutionary significance of these observations and determine the genetic mechanisms of floral organ number control. In general, it will be important to better understand the relationship between developmental constraints and robustness.

Developmental robustness, evolutionary innovations and ecological opportunities

We propose that developmental robustness facilitated synorganization and subfunctionalization and therefore was an important prerequisite for the evolution of complexity (i.e. the evolution of new types of organs or structures) (Fig. 1). These processes may in turn have fostered the evolution of species diversity. In this light, the pentamerous flowers of core eudicots, the trimerous flowers of monocots or the three pairs of legs of insects might be ‘frozen chance events’ that may not have been of direct adaptive value, but nonetheless triggered the diversification of taxa. It is conceivable that the ancestors of core eudicots and monocots were much more developmentally labile, with variable numbers of perianth organs (Endress, 2001, 2011; Soltis et al., 2003). Increasing developmental robustness might have given rise to species possessing pentamerous flowers in one lineage and trimerous flowers in another lineage. Subsequent increases in floral complexity were facilitated by the robust specification of the floral bauplan and hence further increased species diversity.

The necessity to robustly implement a developmental programme may also explain why evolutionary innovations do not always immediately lead to an increase in species diversity. For example, although early-diverging angiosperms may possess all the key innovations associated with flowering plants, they still show some lability in expressing these characters (Endress, 2001, 2011; Warner et al., 2008, 2009). The robust implementation of the key innovation is only established in later arising lineages, and may be as important as the key innovation itself for seizing ecological opportunities (e.g. the interaction with specific pollinators) and hence for an increase in species diversity.

We thus propose that robustness is an important prerequisite for diversifying selection to take place (Fig. 2). Diversifying selection, in turn, can facilitate the evolution of reproductive isolation and hence the origin of new species. Beyond ecological opportunities and evolutionary innovations we thus consider developmental robustness to be an important factor that increases species diversity.

Fig. 2.

Proposed scenario as to how the interplay between evolutionary innovations, developmental robustness and ecological opportunities contributes to the origin of species diversity.