Sunday, March 21, 2010

Reassembling animal evolution: a four-dimensional puzzle

18.1 Introduction

Drawing from the latest literature and the con- tributions in this volume, we consider some of the recent progress made in the study of animal evolution and the hurdles that remain. Each of the disciplines considered—palaeontology, evo- devo, phylogenetics, and the incorporation of gen- omic data—have made major contributions to our understanding of how animals have diversified. Together, these pursuits are resulting in a return to whole-organism biology where the link between genotype and phenotype is considered in the con- text of changing physical and biological environ- ments. The modern approach integrates across all these sometimes disparate disciplines, with the aim of reconciling available evidence to describe the patterns and processes that have led to the existing diversity of animal life.
Arguably, there is one underlying common quest that unites the goals of individual researchers: the search for homology—recognizing it, defining it, and using it. Whether it is establishing shared common ancestry of form or function, similar challenges face those contemplating strings of nucleotides, protein structure, gene expression, biochemical pathways, organs systems, or fos- silized microstructures. As we move towards a greater understanding of evolution and the bio- logical entities undergoing selection, it is the study of homology that allows us to detect patterns and interpret processes.
Gaps in our knowledge can be daunting. At best they define the limits of our ignorance, and
at worst they prevent any meaningful or confi- dent interpretation of available information. We consider how some of the major gaps are being addressed with the renaissance of whole-organism biology, the development of improved models, and the advent of new technologies.

18.2 Phylogenies and phylogenetics

Since the first credible molecular estimate of ani- mal relationships was published by Field et al. (1988) there have been a number of significant changes in our understanding of the evolution of the animal kingdom. The largest shift has been from the widely held assumption of gradualism, whereby morphologically simpler animals such as flatworms were placed towards the base of the tree, and complex features such as coeloms and segments were thought to be homologous and to define major groups of animals higher up the tree. The tree widely accepted today has its roots firmly in Field et al.’s study, and subsequent studies adding to the sampling of small subunit (SSU) ribosomal RNA gene (rDNA) sequences; the major revolutions have, until recently, almost all come from efforts using SSU rDNA. Terms such as Ecdysozoa and Lophotrochozoa draw upon shared morphological features, but their roots stem from SSU rDNA. The new animal phylogeny, hand in hand with com- parative developmental studies of homologous gene expression, has forced a reassessment of the evolution and homology of many characteristics of animals; a recognition of the pervasive effects of the loss of characters and secondary simplification



of body plans (Copley et al., 2004; Jenner, 2004c) as
is apparent in the flatworms.
While there has been enormous progress in our
understanding of metazoan phylogeny leading to
broad agreement over the outline of the animal
tree (Halanych, 2004; Telford, 2006), there remain
a number of hotly contested questions in meta-
zoan phylogeny; with inevitability, the outstanding
questions are the hardest to answer and the dif-
ficulties encountered are likely to stem from mul-
tiple sources. The first major source of difficulty
occurs when the living phyla emerged in an explo-
sive radiation leaving little chance for the fixation
of informative substitutions; such a situation is
exemplified by the difficulty of resolving relation-
ships between the lophotrochozoan clades (Dunn
et al., 2008). The second important source of diffi-
culty arises when living exemplars are the result of
unusual patterns of genomic evolution that violate
assumptions of models used to reconstruct trees,
resulting in inaccuracies in their placement on the
tree (Philippe and Telford, 2006). This is undoubt-
edly seen in the case of the acoel flatworms, chae-
tognaths, myzostomids, gnathostomulids, and
various other ‘Problematica’.
The tendency for phylogeneticists to contra-
dict each other over the placement of problematic
groups may be rather frustrating to outsiders but
is inevitable. First, all animals that have ever been
described have also been positioned somewhere on
a phylogenetic tree. Any progress to be made inev-
itably involves changing this position and hence
introduces contradiction. Secondly, and alluded
to above, all the easily solved aspects of the tree
were answered 10 or 20 years ago, meaning any-
thing currently worth studying is by definition
problematic. A reliable phylogeny is fundamental
to comparative biology and to our understanding
of evolution, and progress continues.
The progress currently being made stems from
the combination of four approaches; much larger
data sets (phylogenomics) which avoid stochastic
error from limited samples; data from additional
representatives of problematic taxa to avoid or
reduce systematic error; alternative sources of data
(e.g. microRNAs) and, potentially, other rare gen-
omic changes which it is hoped are resistant to
homoplastic evolution (Rokas and Holland, 2000;
Boore, 2006); and finally, improved methods of tree reconstruction that more accurately model the underlying process of molecular evolution so reducing further the possibility of stochastic error (Philippe and Telford, 2006). The biggest contribu- tors to progress in terms of data are the new, cheap technologies for DNA sequencing. We are not far from the day when any given species (with a ‘nor- mal’ sized genome) will have its genome completely sequenced for less than the sum that a single gene may have cost 25 years ago. This will provide the greatest possible source of data for phylogenetic analysis and the resolution of any remaining errors will be the province of the model makers.

18.3 Palaeontology

The frustrations inherent in reconstructing the phylogeny of living animals are echoed by the problems of palaeontology. Many fossils are hard to decipher, especially for outsiders, and confusion is exacerbated by the vehement disagreements over their interpretation by the experts. As an example, the Lower Cambrian Emmonaspis cambrensis has been linked with graptolites (hemichordates), chordates and arthropods, and even with Ediacaran frond-like organisms since its description in 1886 (Conway Morris, 1993b). Beyond the well-known problems of preservation and interpretation (Budd and Jensen, 2000), the most interesting fossils— those in the stem lineages of living taxa with the potential to show the order of acquisition of clade synapomorphies—are the hardest to interpret and to relate to modern groups by their very lack of synapomorphies.
Despite the undoubted problems of palaeon- tology, fossils are unique in their ability to inform us about certain aspects of evolution (Smith, 1994). While comparisons of living taxa within an accur- ate phylogenetic framework give tremendous insight into the pattern of evolution, this approach remains limited by the fact that most of the steps of evolution leading to living clades are absent. As an example, it seems clear that the closest relatives of the arthropods are to be found amongst the cycloneuralian worms. It is not clear, however, how much a comparison of priapulids and arthropods will tell us about the stages by which segments and


jointed appendages were acquired in the arthro- pod stem; in such a case, fossils can be of enormous importance.
The importance of studying fossil lineages for our understanding of the evolution of crown groups has been discussed. Stem-lineage fos- sils make an important contribution in several ways; they break long branches leading to crown groups and show intermediate character states; they may reveal unsuspected character hom- ologies or indeed convergent evolution between extant groups; they can highlight character loss in certain groups; and, finally, they provide the sole means to calibrate evolutionary trees by giv- ing minimum divergence times of living clades. Fossils are also able to provide the ecological back- ground to specific evolutionary events, perhaps most spectacularly the great extinctions and the invasion of new habitats such as the land. All of this information is provided uniquely by fossils; it is vital that evolutionary biologists do not damn fossil evidence too readily based on the difficulties inherent in the field. Palaeontologists themselves recognize the problems they face, and efforts are being made to strengthen the objectivity of fos- sil interpretation and to understand the limits of inference; e.g. in calibrating trees (Drummond et al., 2006; Marshall, 2008), and the interpret- ation of biological evidence for historical events (Budd and Jensen, 2000; Domazet-Los et al., 2007; Peterson et al., 2007; Donoghue and Purnell, 2009). Newly discovered deposits, new tools to visualize internal and microscopic features, new methods of detecting and characterizing biomolecules, and simply returning repeatedly to problematic taxa in the light of new evidence will keep the study of fossils alive.

18.4 Developmental evolution

A phylogenetic tree can describe the relationships of species of living and fossil taxa; mapping the characteristics of those taxa onto the framework of the tree permits us to track the evolution of those characters, showing in which groups—and even at what time—key morphological novelties have evolved. While this combination of a dated phylogenetic framework and the distribution of
characters provides a historical description or pat- tern of character evolution, to understand mor- phological novelty and how such morphological change has occurred at the level of the genome and the embryo (the process of morphological evolu- tion) we need to study the genetics behind changes in ontogeny (see, for example, Moczek, 2008).
The birth of modern developmental evolutionary biology came 25 years ago with the molecular clon- ing of the homeobox motif from Drosophila home- otic genes (Carrasco et al., 1984; McGinnis et al.,
1984) alongside the amazing discovery that the same motif (and indeed the same genes) existed in vertebrates with conserved functions. Comparative molecular genetic analyses of development have since changed our view of the evolution of devel- opmental mechanisms and the origins of novel morphology, revealing surprising conservation and providing an alternative to phylogenetic prox- imity for determining homology. The promise of current evo-devo research is to expand the focus of research to new groups of organisms. While a great deal of progress continues to be made using comparisons of expression patterns (using in situ hybridization) for detecting similarity of function of homologous genes and identifying homology of characters, the export of genomics and true func- tional studies (e.g. RNA interference and transgen- esis) to animals not previously considered model organisms is extremely exciting (see, for example, Abzhanov et al., 2008, and Vera et al., 2008).
By expanding beyond the traditional model organisms, practitioners of developmental evo- lutionary biology are able to build on the discov- eries of the phylogeneticists and palaeontologists to address some of the more intriguing ques- tions in morphological evolution. Current ques- tions revealed by the new animal phylogeny and palaeontological discoveries include the origins of arthropods from the cycloneuralian worms such as priapulids and kinorhynchs, the unexpected rela- tionship of the deuterostome-like brachiopods to lophotrochozoans such as annelids and molluscs, and the possible origins of bilaterians from ani- mals resembling the acoel flatworms.
In addition to investigating specifics such as those questions mentioned above, another focus of devel- opmental evolutionary studies is the generalities


of the genetics behind morphological evolution. A current debate concerns the relative importance of changes in regulatory DNA versus coding DNA of genes (Carroll, 2008; Stern and Orgogozo, 2008; Wagner and Lynch, 2008). One thing on which both sides seem to agree, however, and perhaps this realization is more fundamental than scoring points, is that changes of small effect predomin- ate. Cis-regulatory changes are common due to the possibility of making subtle changes in independ- ent enhancers, and coding changes occur where their pleiotropic effects are minimized. There is nothing new under the sun, however (Ecclesiastes
1:9–14), and this debate harks back, of course, to R. A. Fisher’s analogy of the focusing of a micro- scope using small adjustments (Fisher, 1930).

18.5 Mind the gaps

Addressing what is missing in the study of ani- mal evolution is unavoidable and necessary, not least because it demonstrates openness, attempts to define the limits of our knowledge, and indi- cates possible directions for future research. The influence of missing empirical information can be substantial, and assessing the impact of missing fossils, missing taxa, and missing data is almost a discipline itself in systematics. What is not known can influence estimates of tree topology and stability and the biological inferences we are prepared to make (see Wiens, 2006; Geuten et al.,
2007; Fitzhugh, 2008). In phylogeny, should miss- ing features be scored as losses or simply miss- ing data, and when are multiple related missing features indicative of single losses (e.g. the dele- tion of strings of nucleotides or the loss of entire organs systems)? In palaeontology and evo-devo, when can absence of evidence be used as evidence of absence?
Incomplete information necessarily pushes us either towards caution, in the fear that any infer- ences from gappy data may be deemed premature, or towards bravery (perhaps even foolhardiness) as the constant need to take stock of available evi- dence forces phylogenetic estimates, character map- ping, taxonomic revisions, recalibrated histories, and the desire to provide a narrative that explains biodiversity through space and time. Diligent
researchers are keen to indicate the strength of their arguments by circumscribing the limits and possible influence of what is not known, at the risk of undermining any conclusions drawn from what is known. In contrast, selective sampling can pro- vide more robust arguments and may obviate the need to consider uncertainty or less compelling scenarios. Though we do not set out to sample selectively, the nature of certain data sets puts us firmly at the mercy of exemplars. Just as the early days of SSU rDNA estimates of animal phylogeny relied on single taxa as representatives of entire phyla, we have seen phylogenomic analyses suf- fering from over-representation of taxonomically biased model organisms or unbalanced data sets as more or fewer expressed sequence tags (ESTs) are recruited for analysis from unrelated research. Using all available evidence from GenBank to esti- mate animal interrelationships would be cumber- some and unwise, but that is not to say we should not consider all the available data for statements on homology, and sample them for balanced represen- tative data sets.
Balancing taxon and character sampling is diffi- cult, and has been the focus of empirical and the- oretical studies (e.g. Graybeal, 1998; Pollock et al.,
2002), but there is little doubt that with each new data set we are liable to repeat the mistakes of insuf- ficient or biased sampling. In many cases we sim- ply do not know that our sampling is insufficient or biased, or may not be able address any shortfalls until new data sets become available. Many gaps in phylogenetic data sets await attention on key taxa for known characters that need to be scored. Meanwhile, expert morphologists and taxonomists are declining in number, character coding is fre- quently controversial, archival specimens may not be available or suitable for sampling the missing data, and the animals may be difficult to sample, being rare, cryptic, geographically isolated, elu- sive, or extinct. We need to live with gaps but also to recognize the need to address them when the opportunity arises.
The age of genomics arrived with the expect- ation that knowledge of complete genetic blue- prints would provide a surfeit of phylogenetic information for robust tree reconstruction. This has yet to occur, since our efforts to uncover form,


function, and homology have been achieved for
very few components of genomes (Kuzniar et al.,
2008). For animal evolutionary biologists the
era of post-genomics is a long way off, not just
because of the lack of understanding of available
genomes, but also because of the lack of character-
ized genomes themselves. Sampling systematically
across the animal tree of life is an important strat-
egy in developing comparative genomic data sets,
but until now evolutionary biologists have rarely
dictated sampling priorities. Furthermore, even a
cursory look at the revolutions in molecular sys-
tematics show how sampling just a few key taxa
can upset the entire understanding of animal evo-
lution. For example, it was preliminary molecular
systematic surveys of flatworms that highlighted
the phylogenetic uniqueness of acoelomorph flat-
worms (Carranza et al., 1997; Littlewood et al., 1999)
and that led ultimately to their current status, their
distinctness from the Platyhelminthes and their
importance as links to our deep bilaterian past
(Baguñà et al., 2008; Hejnol and Martindale, 2008b).
Undoubtedly, denser sampling of animal genomes
will provide more surprises.
Whilst evolutionary biologists are constantly
concerned with homology either implicitly or
explicitly (see recent review by Szucsich and
Wirkner, 2007), large-scale data sets are moving
us away from an intimate understanding of all the
statements of homology that we make or rely upon.
To some, this may appear to be neglecting our
responsibility as those whose task it is to detect,
highlight, and interpret the evidence for shared
ancestry. Recently there has been a shift from por-
ing over nucleotide and amino acid alignments
with reference to secondary structures, open read-
ing frames, and function, where indels (insertion/
deletion markers) might be placed judiciously and
exclusion sets chosen carefully, to a need for auto-
mation in order to harness considerable volumes of
data (Wong et al., 2008). A plethora of data requires
the building and implementation of bioinformatic
pipelines to make many of these decisions for us,
swiftly, consistently (with given criteria), and rou-
tinely in the hope that we are minimizing noise
and maximizing signal. Whilst these routines and
algorithms might be borne of an understanding
of the underlying data, such automated efforts do
not negate the need to make evolutionary sense of the biological data, and we must be wary of open- ing new gaps in our understanding.

18.6 Learning from the past and taking advantage of the present

In an era dominated by unprecedented access to information, we have an opportunity for embracing considerable bodies of primary data, meta-data, and the thoughts and arguments of generations of researchers. Global efforts to digit- ize literature and specimens, internet tools that mine, parse, and link databases, and concerted global efforts by a generation of researchers will- ing to synthesize existing information are gener- ating new understanding, whilst complementary efforts by others to generate primary data con- tinue unabated. Indeed, the increase in rate at which gene sequence data can now be gener- ated with second-generation sequencing is phe- nomenal, and third-generation sequencing, now on the horizon, promises orders of magnitude more data (Shendure and Ji, 2008). The informa- tion revolution is vast in scale and breadth and brings with it new powers and challenges, not least for bioinformaticians (Helaers et al., 2008; Pop and Salzberg, 2008). New ways of studying genomes and inferring historical events challenge underlying philosophies and resurrect arguments against phenetics, but there is little doubt that presence/absence of genes, gene networks and biochemical pathways, relative arrangement of genes, and so on, provide an entirely new vocabu- lary with which to consider the past (Boore, 2006; Ding et al., 2008; Dulith et al., 2008).
Although we strive for pragmatic approaches to the onslaught of information, and welcome the opportunities to bring disparate fields back into the fold, caution is always at the back of our minds. For example, although we might expect to be able to access information at the click of a mouse, at what point should we select the following without a second thought: a gene sequence with no associ- ated voucher specimen, a distribution map based on inaccurate identifications or DNA barcodes, a tree topology based on data we have not seen, a cluster of genes we have not verified as being


orthologous, a supertree? Clearly, no individual can make all these decisions independently and it is as a community that we police ourselves, and the data we choose to accept as fit for purpose. Systematics continues to be about maximizing the signal and minimizing the noise, but there is a constant battle against a modern trend towards
‘one-gene-fits-all’ approaches, the undermining
of systems that ‘ain’t broke and don’t need fixing’
(e.g. the Linnean system for classification), a lack of rigour in the understanding or implementing the tools (and underlying philosophies) of the trade, and false claims as to how we will have cat- alogued or barcoded every species on the planet and resolved the position of every twig of the tree of life within the next 25 years. Rhetoric aside, there has been no better time to study animal evolution.

Conserved developmental processes and the evolution of novel traits: wounds, embryos, veins, and butterfly eyespots

17.1 Introduction

The origin and diversification of novel traits is one of the most exciting topics in evolutionary developmental biology (evo-devo). The genetic and developmental mechanisms that underlie morphological innovations are, however, poorly understood, and the fact that no unambiguous definition of novelty exists does not make things easier (see Moczek, 2008; Pigliucci, 2008). While some authors suggest that only a structure for which no homologue can be found in the ancestral species or in the same organism can be considered a morphological novelty (e.g. Müller and Wagner,
1991), others emphasize its ecological importance and define novelty as a new trait that enables new functions and opens up new adaptive zones (Mayr, 1960). The animal kingdom has numerous examples of such adaptive morphological innova- tions. Feathers of birds, spinnerets of spiders, and carapaces of turtles are unique traits that have played a crucial role in the diversification of these lineages. Analysing the genetic and developmen- tal underpinnings of novel traits, however, can be a challenge when they are not represented in model organisms and the comparative method, so successful in evo-devo, is harder to apply. The co- option of existing genes, pathways, or organs in the evolution of novel traits offers the opportunity to overcome this limitation.
17.1.1 Co-option in the evolution of novelties

Among the different mechanisms that have been proposed to explain the origin of morphological novelties, the co-option, or recruitment, of pre- existing features into performing novel functions has received a great deal of attention (e.g. True and Carroll, 2002; Sanetra et al., 2005). This phe- nomenon seems to be prevalent and includes the co-option of tissues and organs as well as of single genes and whole developmental pathways, often with modification of components therein. Avian feathers, for example, have evolved from primitive feather-like epithelial outgrowths used for thermo- regulation and/or camouflage in non-avian dino- saurs (Prum, 1999), and insect wings and spider spinnerets are both derived from the respiratory organs of the common arthropod ancestor (Damen et al., 2002). The development of horns in a num- ber of beetle species seems to rely on redeployment of the arthropod limb patterning genes Distal-less and aristaless (Moczek and Nagy, 2005). The Wnt signalling pathway, involved in various develop- mental processes in vertebrates (Logan and Nusse,
2004), has been implicated in the evolution of the turtle shell (Kuraku et al., 2005).
Studies in butterflies and moths provide some spectacular examples of pathways that are shared across insects and have been co-opted in the evo- lution of colour patterns. The remarkably diverse



lepidopteran wing patterns are built up from a mosaic of thousands of flattened pigmented scales produced by wing epidermal cells (Figure 17.1a and Plate 11). It has been proposed that these scales are homologous to insect sensory bristles, and have evolved through the recruitment of bristle-pattern- ing genes of the Achaete-Scute complex, followed by acquisition of target genes responsible for typical scale morphology (Galant et al., 1998). The variable colour patterns generated by those wing scales have themselves evolved through the co-option of a number of genetic pathways. The pigments that colour butterfly scales, for instance, derive from the redeployment of several genes from the ommochrome synthesis pathway (e.g. vermilion, cinnabar), known to function in insect eye pigmen- tation (Reed and Nagy, 2005; Reed et al., 2008). The patterns made by the spatial arrangement of col- oured scales, on the other hand, rely on genetic pathways (e.g. the Hedgehog and Wingless path- ways), involved in embryonic and wing develop- ment in butterflies and other insects (Carroll et al.,
1994; Monteiro et al., 2006). Such co-option of gen- etic pathways offers the potential to use extensive knowledge gathered from work on classical model organisms to dissect the formation of butterfly- specific colour patterns.

17.1.2 Butterfly eyespots as an example of an evolutionary novelty

The study of butterfly eyespots, characteristic pat- tern elements composed of concentric rings of dif- ferent colours (Figure 17.1b), has started to shed light on how novel patterns have arisen and diversi- fied in the Lepidoptera. Eyespots probably evolved from primitive, uniformly coloured spots through the recruitment and modification of conserved developmental genes and pathways, acquisition of signalling activity, and further diversification of colour schemes under the influence of natural selection (Brunetti et al., 2001; Monteiro, 2008). Their ecological significance in predator avoidance and sexual selection is well documented (Stevens,
2005; Costanzo and Monteiro, 2007), as is the spec- tacular diversity in eyespot morphology. Variation
in eyespot number, position, shape, size, or colour composition is found not only across species and among individuals of the same species, but often also between different wing surfaces of the same individual butterfly (Nijhout, 1991). Eyespot devel- opment is amenable to detailed characterizat- ion, ranging from the genetic pathways involved in establishing the pattern, to the molecular and cellular interactions underlying pattern specifica- tion, to the biochemical networks involved in pig- ment production (Beldade and Brakefield, 2002; McMillan et al., 2002).
Models of eyespot formation involve the pro- duction and diffusion of one or more signalling molecules from a central organizer, the focus, and the response of the surrounding epithelial cells to the signal(s) in a threshold-like fashion, eventu- ally leading to the production of rings of differ- ent pigments (Nijhout, 1980; Dilão and Sainhas,
2004). The organizer properties of the focus are supported by experiments in which transplant- ation of the focal cells into a different position on the early pupal wing induces the formation of an ectopic eyespot around the transplanted tissue (Nijhout, 1980; French and Brakefield, 1995). The molecular identity of the signal is not known, but both Wingless and Decapentaplegic have been proposed as candidate morphogens (Monteiro et al., 2006). A number of genes have also been implicated in the determination of eyespot cen- tres and colour rings, including members of the Hedgehog pathway (Keys et al.,1999), the recep- tor gene Notch (Reed and Serfas, 2004), and the transcription factor-encoding genes Distal-less, engrailed, and spalt (Carroll et al., 1994; Brakefield et al., 1996; Brunetti et al., 2001). The latter three, for example, are expressed in scale-building cells in association with the different colour rings of the eyespots of several butterfly species (Figure
17.1h). Despite the fact that these and other genes have been implicated in eyespot development, we still know little about the interactions between them, how they regulate pigment synthesis, or the extent to which they contribute to phenotypic variation in eyespot morphology (see Beldade and Saenko, 2009).


(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

(j) (k) (l)

Figure 17.1 Conserved developmental processes implicated in butterfly eyespot formation. Coloured scales covering butterfly wings
(a), and eyespot patterns formed by these scales (b), are key innovations in the lepidopteran lineage and are represented in the laboratory- tractable system, Bicyclus anynana (c, ventral view of female at rest). The formation of eyespots in B. anynana shares genetic commonalities with different conserved developmental processes such as embryonic development (d–f), wound healing (g), and wing vein patterning (j–l). Embryonic development in B. anynana has been characterized in wild-type and pleiotropic eyespot mutants (Saenko et al., 2008): wild-type embryo after completion of blastokinesis (d), the characteristic expression of the segment polarity gene engrailed at that stage (e), and a homozygous Goldeneye embryo of the same age that has failed to undergo blastokinesis (f). Expression pattern of engrailed in pupal wings in association with the gold ring of the presumptive adult eyespot (h). This expression is altered in Goldeneye eyespots (Brunetti et al., 2001; Saenko et al., 2008) in a manner that matches the change in adult eyespot colour-composition (i). Damage with a fine needle applied to the
distal part of the developing pupal wing (arrows in left panel) results in formation of ectopic eyespots around the wound site (right panel) (g). Wing venation mutants often affect eyespot patterns (all photos show the ventral surface of the hindwing): the additional vein in extra veins mutants can lead to the formation of an extra eyespot (j), while partial vein loss in Cyclops (k) and vestigial venation in veinless (l) mutants typically result in changes in eyespot size, number, and/or shape. (See also Plate 11.)


17.1.3 Bicyclus anynana as an emerging
‘eyespot evo-devo’ model

The tropical Nymphalid B. anynana (Figure 17.1c) has been established as a laboratory system for studying the reciprocal interactions between evo- lutionary and developmental processes underlying ecologically relevant phenotypic variation, with emphasis on wing patterns (Brakefield et al., 2009). This system combines knowledge of ecology, often minimal for classical genetic model species, with experimental tractability, including recently devel- oped genomic resources (Beldade et al., 2008) and transgenic techniques (Ramos and Monteiro, 2007), and allows an integrated analysis, from the devel- opmental and genetic basis of eyespot formation to the molecular underpinnings of variation in nat- ural populations.
Here we focus on butterfly eyespots as an example of a morphological novelty and review recent findings relating to the co-option of con- served developmental processes in the evolution of eyespot patterns. First, we discuss experimental evidence for the similarities between eyespot pat- terning and wound repair. We then give examples of pleiotropic mutations isolated in B. anynana laboratory populations which affect not only eyespots but also either embryonic or wing vein development, and discuss how studies of such mutations may help elucidate the genetic path- ways involved in eyespot formation and variation. A detailed analysis of such conserved genetic net- works, extensively studied in the model organism Drosophila melanogaster, in the context of eyespot formation will be invaluable for our understand- ing of the evolutionary origin and diversification of butterfly eyespots.

17.2 Wounds and eyespots

The ability to repair wounded tissue is a funda- mental property of all multicellular organisms, and a key topic of current research (see Gurtner et al., 2008). Here we review evidence suggesting that some components of this process are involved in eyespot formation.
17.2.1 Damage-induced eyespot formation in B. anynana

Local damage to pupal wing tissue has long been known to disturb colour patterns in many lepidop- terans, and has been used to study the mechanisms of pattern formation in butterflies and moths (e.g. Kühn and Von Englehardt, 1933). In eyespot-bear- ing butterfly species, for example, damage with a very fine needle applied to the presumptive eyespot in early pupae can completely eliminate eyespots in the adults (Nijhout, 1980; French and Brakefield,
1992). Also, damage to other locations on the pupal wing epidermis can result in the formation of an ectopic eyespot around the wound site (French and Brakefield, 1995). In B. anynana, for example, rings of black and/or gold, typically poorly defined and less symmetrical than those of the native eye- spots, are found around the healed wound (Figure
17.1g). Interestingly, this type of damage produces eyespots only on the distal area of the wing, and only if applied during a very narrow time win- dow (Brakefield and French, 1995), which more or less corresponds to the period when ‘colour-ring genes’, Distal-less, engrailed, and spalt, are upregu- lated in the presumptive eyespot area (Monteiro et al., 2006). The mechanisms by which the genes and pathways of the damage response machin- ery might contribute to the formation of ectopic eyespots are as yet unclear, but insights from studies in model organisms might provide some clues.

17.2.2 Genetic mechanisms of wound repair

Comparative studies on the genetic and cellular mechanisms of wound repair and regeneration in representatives of various animal phyla have suggested their evolutionary conservation (see Woolley and Martin, 2000). For example, some steps in the wound healing process are regulated by the same transcription factor, Grainyhead, in flies and mice (Mace et al., 2005; Ting et al., 2005). Also, wound healing seems to recapitulate some aspects of embryonic morphogenesis, such as dorsal closure in flies and eyelid fusion in mice, raising the possibility of co-option of genetic path-


ways active during embryogenesis into wound repair processes during adult life (see Martin and Parkhurst, 2004).
The mechanism by which damage results in the formation of ectopic eyespots is an intriguing question. It is known that candidate genes for eye- spot signalling also perform functions related to wound healing. Studies in fly larvae, for example, have shown that wounds act as sources of short- and long-range signalling molecules and activate downstream pathways in a gradient-like man- ner, as has been proposed for eyespot develop- ment. In Drosophila, the Jun N-terminal kinase (JNK) pathway is activated in a gradient centred around the wound (Galko and Krasnow, 2004) and up-regulates the transcription factor AP-1 which, in turn, leads to induction of decapentaple- gic, a transforming growth factor (TGF)-E family member and one of the candidate eyespot-induc- ing signals (Monteiro et al., 2006). Some evidence also exists for the involvement of the Wnt proteins in wound repair in mammals (e.g. Okuse et al.,
2005), whereas their insect homologue, Wingless, is a candidate morphogen in eyespot formation (Monteiro et al., 2006).
To investigate whether the same genetic mechan- isms that underlie wound healing are implicated in damage-induced eyespot formation in butter- flies, Monteiro et al. (2006) used immunostaining to detect the expression of known ‘eyespot genes’ around damage sites in developing pupal wings. They showed that these genes were upregulated in the epidermal cells surrounding the wound (Monteiro et al., 2006) and proposed that the for- mation of both native and ectopic eyespots relies on at least some identical molecules. Their data, however, do not distinguish between the possibil- ity that the rings of expression of the tested eyespot genes around wound sites are induced by known candidate eyespot morphogens (Monteiro et al.,
2006), or whether they result from some other, as yet unidentified, signals produced by wounded cells. Future work will focus on the identification of genes and pathways that are upregulated upon wing damage and lead to the formation of ectopic eyespots.
17.3 Embryos and eyespots

Among the spontaneous mutants maintained in laboratory populations of B. anynana (Brakefield et al., 2009), some have been characterized that are embryonic lethal in the homozygous state and have a dramatic effect on eyespot morphology in het- erozygotes (Saenko et al., 2008). Comparative ana- lysis of disturbed embryonic development in these mutants with similar phenotypes described in model insects might help identify genes involved in eyespot formation. The mechanisms of early embry- onic development are well studied in the dipteran D. melanogaster (Peel et al., 2005), and are becoming increasingly well understood in representatives of other insect orders, such as the coleopteran Tribolium castaneum and the hemipteran Oncopeltus fasciatus (reviewed in Liu and Kaufman, 2005), the hymenop- teran Nasonia vitripennis (e.g. Pultz et al., 2005; Lynch et al., 2006a), and the lepidopterans Bombyx mori (e.g. Nagy, 1995) and Manduca sexta (e.g. Kraft and Jackle,
1994). To the extent that the genetic mechanisms of embryogenesis are conserved across insects (see Peel et al., 2005; Damen, 2007; see also Chapter 16), a comparison of disturbed embryonic develop- ment in B. anynana eyespot mutants with mutants in insect model species can help identify signalling pathways and/or specific genes involved in eyespot formation and variation.
Embryonic development in B. anynana (described in Saenko et al., 2008) is characterized by a long- germ mode of segmentation, as in most other lepi- dopterans and dipterans (Kraft and Jackle, 1994), but differs from that of D. melanogaster in some important aspects. For example, at about half way through embryonic development, lepidopterans go through a characteristic movement within the egg that results in the reversal from a ventral to a dorsal flexion, called blastokinesis (Broadie et al., 1991; Figure 17.1d). Other aspects of embry- onic development, such as segment patterning by segment polarity and Hox genes (Figure 17.1e) and limb patterning by Distal-less, are similar in B. anynana and other insects (Saenko et al., 2008). This suggests that the direct comparison of dis- rupted embryonic development in pleiotropic B. anynana eyespot mutants with described mutants


in model insects could be a useful approach for identifying new pathways and candidate genes for eyespot development (Saenko et al., 2008).

17.3.1 The homozygous lethal mutation

Goldeneye is one such pleiotropic mutation that is embryonic lethal in homozygotes (Saenko et al.,
2008) and affects eyespot colour composition in het- erozygous adults (Brunetti et al., 2001). In Goldeneye butterflies, the scales that typically form the black inner ring of B. anynana eyespots are replaced with gold-coloured scales characteristic of the outer ring (Figure 17.1i), and the expression of genes associ- ated with the different colour rings changes accord- ingly (Brunetti et al., 2001). Analysis of embryonic development in Goldeneye homozygotes has shown that these embryos do not undergo blastokinesis, subsequently become shorter and thicker than wild-type embryos of the same age, and end up dying shortly after the time at which blastokinesis would normally have occurred (Saenko et al., 2008; Figure 17.1f).
The study of disturbed embryogenesis in Goldeneye has suggested that genes involved in blastokinesis may play a role in eyespot formation (Saenko et al., 2008). Unfortunately, the specific gen- etic regulation of embryonic movements in insects, including blastokinesis in the Lepidoptera, is poorly understood, despite the fact that several mutations have been identified that disturb these processes (e.g. Ueno et al., 1995; Schock and Perrimon, 2002; Van der Zee et al., 2005; Panfilio et al., 2006). Even though the Goldeneye embryonic phenotype did not show resemblance to described mutants in other species, and despite the fact that we do not know whether embryonic movements are regulated by similar mechanisms across insect lineages, analysis of candidate genes affecting this process seems like a valuable starting point for further genetic dissec- tion of variation in eyespot morphology.

17.3.2 Conservation versus divergence in insect embryonic development

The strategy of comparing disturbed embryonic phenotypes between B. anynana eyespot mutants
and mutants in genetic model systems will be useful only to the extent that the genetic mechan- isms of embryonic development are conserved across insects. Recent studies extending the ana- lysis of insect embryonic development outside D. melanogaster (see Peel et al., 2005, Damen, 2007; Peel, 2008) have shown that while some aspects of embryonic development are indeed highly conserved (e.g. the functions of segment polarity and Hox genes), others appear to be unexpectedly diverged (e.g. the functions of gap and pair-rule genes; see Chapter 16).
The analysis of embryonic lethality in three other pleiotropic B. anynana mutants, in which development appears to be disturbed during the segmented germband stage, is more prom- ising for identifying mutated genes (P. Beldade and S. V. Saenko, unpublished data). Unlike blas- tokinesis, this stage of embryogenesis is highly conserved among arthropods (e.g. Farzana and Brown, 2008), and the genes and developmental pathways that regulate it have been studied in greater detail in model organisms (see Galis et al.,
2002). Still, whatever the embryonic stage affected in any pleiotropic eyespot mutant, because direct comparison of disturbed eyespot phenotypes with
‘eyespot mutants’ in model species is impossible (model insects do not have eyespots!), comparative analysis of embryonic phenotypes in such mutants remains a valuable first approach.

17.4 Veins and eyespots

Models for butterfly wing pattern formation pro- pose an important role for wing veins and the wing margin (see Nijhout, 1991). An association between wing venation and either patterns of col- ourful stripes and bands or eyespot formation has been supported by the phenotypic characteriza- tion of spontaneous venation mutants in Papilio and Heliconius species (Koch and Nijhout, 2002; Reed and Gilbert, 2004) and in B. anynana (Saenko et al., 2008), respectively. Models of eyespot forma- tion have suggested that wing veins act as sources of diffusible molecules involved in the determin- ation of eyespot-organizing centres, the foci (see Nijhout et al., 2003; Evans and Marcus, 2006), but this role, as well as the nature or even the existence


of the proposed diffusible signals, has not yet been
shown experimentally.

17.4.1 Parallels between fruitfly and butterfly vein development

The mechanisms of vein patterning in Drosophila are fairly well understood (reviewed in Blair, 2007), as is the role of veins in the distribution of melanin precursors in newly eclosed fruitflies (True et al.,
1999). As is often the case for work in non-model systems, this knowledge is an invaluable starting point for our understanding of the mechanisms behind vein establishment and its role in pattern formation in butterfly wings.
Unsurprisingly, positional specification in butterfly wing discs seems to be achieved in a man- ner very similar to that described for Drosophila. Division of the developing wing discs into antero- posterior and dorsoventral compartments is marked by the expression of the genes engrailed and apterous, respectively, and proximal–distal patterning is regulated by Distal-less and wing- less (Carroll et al., 1994). The signalling pathways that are involved in the positioning and differenti- ation of longitudinal and cross veins in Drosophila (reviewed in Marcus, 2001, and Crozatier et al.,
2004) might also be conserved between the lin- eages of Diptera and Lepidoptera (De Celis and Diaz-Benjumea, 2003). Functional analysis of homologues of known Drosophila vein patterning genes during butterfly wing development will be instrumental in our understanding of vein estab- lishment and role in butterfly wings.

17.4.2 Wing venation and eyespot patterns in B. anynana mutants

Study of B. anynana mutants with disturbed ven- ation has started to explore the functional relation- ship between wing veins and eyespot formation, and has suggested that eyespot patterning depends on normal formation of veins and tracheae (Saenko et al., 2008). Mutations that lead to the addition of veins, such as extra veins (Figure 17.1j) can also lead to the appearance of extra eyespots, presum- ably when the ectopic vein bisects an existing
eyespot signalling centre or because the additional vein itself acts as an inducer of eyespot formation. Conversely, partial or complete loss of wing veins often leads to formation of smaller, fewer and/or misshaped eyespots, as in the Cyclops (Figure 17.1k) and veinless (Figure 17.1l) mutants of B. anynana (Saenko et al., 2008).
Surgical manipulations on developing wings of B. anynana venation mutants have provided the first insights into the mechanisms by which loss of veins can interfere with eyespot formation. The absence of eyespots on the dorsal surface of vein- less individuals, investigated by transplanting the signalling focus of a wild-type pupa into a vein- less wing, has shown that vestigial venation leads to impaired determination of the eyespot focus or lack of focal signal (Saenko et al., 2008). The molecu- lar mechanisms of such relationship are yet to be explored.

17.5 Concluding remarks

Here we have reviewed knowledge of the genetic and developmental mechanisms of eyespot forma- tion, and discussed new approaches to the study of these lineage-specific structures, based on the commonalities with conserved developmental pro- cesses such as wound healing, embryonic develop- ment, or vein patterning.
How much the study of laboratory populations can tell us about variation in natural populations remains a crucial issue in evo-devo. In particular, the extent to which the mutants of large effect identified in the laboratory are relevant for nat- ural variation within and across species is still debated (see Haag and True, 2001). Whilst it seems unlikely that pleiotropic mutations with negative effects on other traits (e.g. embryonic lethality in Goldeneye, or fragile wings in mutants with ves- tigial venation) will contribute to variation in natural populations, it is possible that the same loci harbour other, less deleterious, alleles rele- vant to variation in eyespot patterns. Future work will explore the extent to which loci identified in laboratory eyespot mutants contribute to vari- ation segregating in natural populations and to variation across species.


17.6 Acknowledgements

We thank Max Telford and Tim Littlewood for organizing the Novartis Foundation Symposium on Animal Evolution and inviting us to contribute
to this book. We also thank Paul Brakefield, Vernon French, and Antónia Monteiro for many inspiring discussions on butterfly wing patterns.

The evolution of developmental gene networks: lessons from comparative studies on holometabolous insects

Recent comparative studies have revealed significant differences in the developmental gene networks operating in three holometabolous insects: the beetle Tribolium castaneum, the para- sitic wasp Nasonia vitripennis, and the fruit fly Drosophila melanogaster. In this chapter I discuss these differences in relation to divergent and con- vergent changes in cellular embryology. I speculate on how segmentation gene networks could have evolved to operate in divergent embryological con- texts, and highlight the role that co-option might have played in this process. I argue that insects represent an important example of how diversifi- cation in life-history strategies between lineages can lead to divergence in the genetic and cellular mechanisms controlling the development of hom- ologous adult structures.

16.1 Introduction

Arthropods are defined by a segmented body plan consisting of a series of anteroposteriorly arrayed segmental units with associated jointed append- ages. The insects are traditionally viewed as one of the four major monophyletic arthropod groups, the other three being crustaceans, myriapods, and chelicerates. However, recent molecular phyloge- nies suggest that crustaceans are paraphyletic with respect to the insects; i.e. insects could reasonably be regarded as a monophyletic clade of terrestrial crustaceans (Carapelli et al., 2007). While many insect species have retained the ancestral condi- tion of undergoing metamorphosis from larva to adult through a series of intermediate nymphal
stages (the hemimetabolous insects; Figure 16.1), the holometabolous insects undergo complete metamorphosis from larva to adult via a pupal stage (Brusca and Brusca, 2003). This is consid- ered a derived life-history trait that arose only once during insect evolution (Brusca and Brusca,
2003) (see Figure 16.1). The vast majority of holom- etabolous insects belong to four speciose orders: the Diptera (two-winged flies), the Lepidoptera (butterflies and moths), the Coleoptera (beetles), and the Hymenoptera (wasps, bees, ants, etc.). We currently have a better understanding of the devel- opmental genetic network underlying segmen- tation in a member of the Diptera—particularly the fruitfly D. melanogaster—than for any other insect, or indeed arthropod (Lawrence, 1992) (see Figure 16.2). However, a representative of the Coleoptera, the beetle T. castaneum, and a repre- sentative of the Hymenoptera, the parasitic wasp N. vitripennis, are rapidly being established as powerful insect model systems (Choe et al., 2006; Brent et al., 2007). Recent studies have revealed significant differences in the segmentation gene networks operating in these insects when com- pared with each other and with D. melanogaster (Schröder, 2003; Bucher and Klingler, 2004; Cerny et al., 2005; Choe et al., 2006; Lynch et al., 2006a,b; Olesnicky et al., 2006; Brent et al., 2007; Choe and Brown, 2007). In this chapter, I review and discuss these differences in relation to the modes of cellu- lar embryogenesis exhibited by these insects. Both D. melanogaster and N. vitripennis have evolved a rapid mode of development that required major changes in embryogenesis at the cellular level



(Bull, 1982; Lawrence, 1992; Davis and Patel, 2002). In contrast, T. castaneum has retained a more ances- tral mode of cellular embryogenesis (Handel et al.,
2000; Davis and Patel, 2002). I speculate on how insect segmentation gene networks have evolved to operate in these divergent embryological con- texts. A recent molecular phylogeny suggests that the rapid mode of cellular embryogenesis exhib- ited by D. melanogaster and N. vitripennis evolved convergently (Savard et al., 2006). I go on to ask whether convergent changes in gene networks might have underpinned these apparent parallel transitions in cellular embryology. First I review the modes of cellular embryogenesis found within the insects, and discuss the role that life history has played in their evolution.
16.2 The influence of life-history strategy on insect embryogenesis

16.2.1 An evolutionary biologist’s view on development

The principal aim of a developmental biologist is to work towards establishing a more complete picture of how the genetic information contained within an organism’s genome is deployed over developmen- tal time to transform a single cell into a functional multicellular organism. In contrast, the principal aim of many evolutionary developmental biolo- gists is to identify—through comparative analysis of developmental data within a phylogenetic frame- work—the changes in developmental mechanisms


Schistocerca sp. Gryllus bimaculatus

Oncopeltus fasciatus

Hemi No

Hemi ?

Seq ?

Seq ?

Nasonia vitripennis Yes Long Yes
Bracon hebetor Yes Long ?
Aphidius ervi No Seq ?
Apis mellifera Yes Long* ?



Tribolium castaneum
Callosobruchus maculates

Bombyx mori
Manduca sexta

Drosophila melanogaster
Anopheles gambiae







Seq* Long*

? Yes

Figure 16.1 A phylogeny of the insect species discussed in this chapter with embryological features mapped on. The relative relationships of the four holometabolous insect orders follows the study by Savard et al. (2006). Some insects do not fit comfortably into the categories
‘sequential’ or ‘long-germ’ segmentation; for caveats in relation to the categorization of specific species (*) see Davis and Patel (2002). Character states have been left clear where there are uncertainties, i.e. when there is a lack of gene expression data and/or dye injection experiments to ascertain the existence of an extended syncytial blastoderm stage.


Anterior Posterior




Krüppel Knirps Gaint





Parasegment boundaries

Figure 16.2 The Drosophila melanogaster segmentation gene cascade. Adapted from Peel et al. (2005).

(1) Maternal genes. Maternal transcripts of the segmentation genes caudal and hunchback are uniformly distributed, whereas maternal bicoid mRNA is tethered to the anterior pole of the egg. Localized at the posterior pole is a complex of maternal proteins and RNAs that includes transcripts of the gene nanos. On fertilization, maternal mRNAs are de-repressed, translated and Bicoid and Nanos proteins form gradients at either end of the egg. Bicoid activates zygotic hunchback expression and represses caudal translation in the anterior, whereas Nanos represses the translation of maternal hunchback in the posterior. As a result Hunchback protein is restricted to the anterior of the egg and protein gra- dients of Bicoid (decreasing posteriorly) and Caudal (decreasing anteriorly) form. In parallel, a maternally encoded terminal patterning system operates during embryogenesis; the product of torso-like—which is expressed within specialized follicle cells situated at both egg poles during oogenesis—catalyses the localized cleavage of a protein encoded by trunk within the perivitelline fluid. The trunk cleavage product acts as a ligand on the receptor tyrosine kinase encoded by torso, triggering a signalling cascade that regulates the zygotic expression of downstream segmentation genes, such as tailless, at either pole of the egg.
(2) Gap genes. The net result of maternal signalling is the activation along the anteroposterior axis of the egg of a series of zygotic gap genes (i.e. giant, Krüppel, tailless), thus named because their mutation leads to gaps in the region of the embryo in which they are normally expressed. The protein products of the gap genes themselves diffuse within the syncytial blastoderm, regulate each other, and thus further refine their expression. Gap genes also play an important role at this stage in regulating the expression of the Hox genes, whose proteins products confer identity to segments.
(3) Pair-rule genes. In the next tier of the Drosophila segmentation cascade are three genes—even-skipped, runt, and hairy—whose expression is driven by the maternal and gap gene transcription factor products. All three genes possess complex regulatory sequences that interpret the aperiodic expression of maternal and gap gene products and drive expression in a periodic pattern of seven stripes. These genes are collect- ively referred to as pair-rule genes since their mutation often leads to abnormalities in alternate segments. The three ‘primary’ pair-rule gene products in turn regulate expression of ‘secondary’ pair-rule genes, such as fushi-tarazu, paired, sloppy-paired, and odd-skipped. Black curve, even-skipped; grey curve, fushi-tarazu.


that underpin divergence in body architecture between lineages. Rather than thinking in terms of developmental time—with the rather arbitrary starting point of zygote or germ cell—evolutionary developmental biologists consider evolutionary timescales, and as such development is not viewed as a linear process but rather as continuous devel- opmental cycles undergoing constant modification in response to selection and drift.
Natural selection can act independently on dis- tinct stages of an organism’s developmental cycle. This is obvious when considering insects. There has clearly been divergence in segment form between insect species, particularly with respect to append- age morphology; compare, for example, the suck- ing mouthparts of the phytophagous milkweed bug Oncopeltus fasciatus (Hughes and Kaufman,
2000) with the mandibles of some carnivorous beetles (Konuma and Chiba, 2007). However, it is clear that, on the whole, the basic insect segmental unit has been conserved. In contrast, oocytes and early eggs exhibit significant morphological differ- ences, a consequence of the numerous and diverse life-history strategies that have evolved within the insects.

16.2.2 All eggs are different, but some eggs are more different than others

Evolutionary shifts in insect life-history strategies often correlate with changes in cellular modes of embryogenesis. This was dramatically illustrated in a study by Grbic and Strand (1998) on two para- sitic wasps belonging to the hymenopteran fam- ily Braconidae. Bracon hebetor is an ectoparasite that lays yolky eggs on the integument of moth larvae. In the lineage leading to Aphidius ervi, however, there has been a transition to an endo- parasitic life history; A. ervi lays a single yolkless egg into the haemocoel of an aphid host. Grbic and
Strand (1998) studied the cellular embryology of these insects and found significant differences. In the eggs of B. hebetor, the cellularization of early cleavage nuclei is delayed until after they form a blastoderm, and all segments develop more or less simultaneously. In contrast, in A. ervi eggs, com- plete cytokinesis (the formation of cell membranes) occurs from the fourth round of nuclear divisions onwards, the early embryo ruptures from the chor- ion within the host haemocoel, and segments form one by one in an anterior to posterior progression. One can only speculate on why the transition to an endoparasitic life history required such dra- matic changes in cellular embryology, but is seems likely they are associated with the transition from receiving nutrients in the form of maternal yolk to the use of nutrients available from the haemo- lymph of the unfortunate host.
Similarly dramatic cellular transitions in embryogenesis have occurred within non-para- sitic insect lineages (for an in-depth review see Davis and Patel, 2002). Although the precise eco- logical reasons remain speculative, it seems likely that in many cases these transitions occurred in response to selection for increases in the speed of embryogenesis. Here I discuss two specific cellular adaptations and how they might have facilitated the faster development of an insect segmented body plan.

The timing of cellularization
In most insect species early nuclear divisions
are superficial; the formation of cell membranes
around early cleavage nuclei is delayed until they
have migrated to the egg surface and formed the
blastoderm. For example, dye injection experi-
ments have demonstrated this to be the case in
the locust Schistocerca gregaria (Ho et al., 1997)
(Figure 16.1). However, this delay is particularly
pronounced in some holometabolous insect lin-

(4) Segment-polarity genes. The pair-rule gene products activate the final tier in the Drosophila segmentation gene cascade, the segment polar- ity genes. These are the genes encoding proteins that actually initiate the formation of segment boundaries, and, as the name suggests, confer polarity to segments. Segment polarity genes are expressed in a series of 14 stripes, with odd and even stripes regulated by a different combin- ation of the pair-rule proteins. The boundary between the expression of two of these genes, engrailed and wingless, becomes the parasegmental boundary, whereas segment boundaries form later, posterior to engrailed expression. Black bar, engrailed; grey bar, wingless.


eages, creating an extended syncytial blastoderm stage: examples of such insects include N. vitripen- nis (Bull, 1982), T. castaneum (Handel et al., 2000), and D. melanogaster (Lawrence, 1992) (Figure 16.1). Within a syncytium, gradients of patterning mol- ecules can form quickly across a field of nuclei, without the need for complex intercellular signal- ling pathways.

The allocation of cells to segments
The temporal dynamics by which cells are allo-
cated to segments varies across insect species. In
insects exhibiting primitive modes of develop-
ment, anterior segments are patterned through the
subdivision of blastoderm nuclei/cells, while pos-
terior segments are patterned sequentially after
the blastoderm stage, within a posteriorly located
cellular zone of extension. Examples of such
insects include the hemimetabolous insects Gryllus
bimaculatus, Schistocerca sp., and O. fasciatus and the
holometabolous insect T. castaneum (see Davis and
Patel, 2002) (Figure 16.1). I shall refer to these as
‘sequentially segmenting’ insects. In many insect
lineages there has been an increase in the number
of anterior segments patterned through subdiv-
ision in the blastoderm (Davis and Patel, 2002); this
has occurred, for example, in some coleopteran lin-
eages (Patel et al., 1994). In many holometabolous
insects this trend has reached its logical extreme,
and all segments form through early subdivision
of embryonic blastoderm nuclei. These insects are
said to exhibit ‘long-germ’ embryogenesis, since
the embryonic germ rudiment typically occupies
almost the entire length of the egg. Examples of
such insects include D. melanogaster (Lawrence,
1992) and N. vitripennis (Bull, 1982) (see below and
Figures 16.1 and 16.2).

16.3 Molecular transitions underlying the evolution of long-germ embryogenesis

During long-germ embryogenesis in D. melanogaster, a cascade of transcription factors acts within a syn- cytium to divide the embryo into progressively smaller domains such that segments develop more or less simultaneously. The D. melanogaster segmentation gene cascade is briefly outlined in
Figure 16.2, but for a more thorough understand-
ing the reader is referred to Lawrence (1992).
In order to identify the changes in gene net-
works that underpinned the evolution of long-germ
embryogenesis, a good understanding of the seg-
mentation mechanisms operating in insects that
have retained sequential segmentation is required.
One such insect is the beetle T. castaneum (Handel
et al., 2005). Recent studies on this holometabolous
insect have revealed significant differences in the
genetic circuitry underlying segmentation when
compared with D. melanogaster (Schröder, 2003;
Bucher and Klingler, 2004; Bucher et al., 2005; Cerny
et al., 2005; Choe et al., 2006; Choe and Brown, 2007;
Schinko et al., 2008). Together, comparative studies
on T. castaneum, N. vitripennis, and D. melanogaster
suggest that the evolution of long-germ embryo-
genesis required distinct molecular transitions to
occur in concert at either pole of the egg (Choe
et al., 2006; Brent et al., 2007).

16.3.1 Molecular transitions at the anterior egg pole: the evolution of maternally encoded anterior patterning gradients

In T. castaneum, head and thoracic segments are patterned through the subdivision of blastoderm nuclei located near the posterior egg pole; the anterior blastoderm forms extra-embryonic tis- sue (Handel et al., 2000) (Figure 16.3a,d). In insects exhibiting long-germ embryogenesis, however, head and thoracic segments are patterned further towards the anterior egg pole (Figure 16.3b,c,e,f). It has been proposed that this spatial shift in anterior patterning required the evolution of an instruct- ive anterior patterning gradient to complement the action of existing posterior determinants (Lynch et al., 2006a). The localization of maternal mRNAs to the anterior pole of the oocyte is observed in T. castaneum, N. vitripennis, and D. melanogaster (Lawrence, 1992; Bucher et al., 2005; Olesnicky and Desplan, 2007) (Figure 16.1). In N. vitripennis and D. melanogaster, mRNA and/or translated protein from anterior and posterior sources of maternal mRNAs form largely non-overlapping, and oppos- ing, instructive patterning gradients (Lawrence,
1992; Lynch et al., 2006a; Olesnicky et al., 2006; Brent


Anterior-Posterior patterning Terminal patterning

Tribolium castaneum
Hunchback orthodenticle-1
Unknown factor(s)



Torso- like


Torso ?
Tribolium castaneum


Torso- like




Anterior head
Thorax Growth zone
Anterior head
Thorax Growth zone


Nasonia vitripennis



Nasonia vitripennis


Hunchback Krüppel
Other anterior factors

Tailless Tailless

Anterior head
Thorax Abdomen
Anterior Gnathal head
Thorax Abdomen



Other anterior factors
Drosophila melanogaster





Torso- like


Drosophila melanogaster

Tailless Tailless

Torso- like



Anterior Gnathal Thorax Abdomen Anterior Gnathal Thorax Abdomen
head head

Figure 16.3 A schematic representation of the variation in maternal patterning observed between the holometabolous insects Tribolium castaneum, Nasonia vitripennis, and Drosophila melanogaster (as described in the text and summarized in Table 16.1). Particular focus is placed on variation in the maternal regulation of the zygotically expressed central gap gene Krüppel (a)–(c) and terminal gap gene tailless (d)–(f). Note that in T. castaneum Krüppel is expressed like, but does not function as, a canonical gap gene (see text and Cerny et al., 2005), and tailless is not expressed in the anterior (see text and Schröder et al., 2000, and Schoppmeier and Schröder, 2005). Maternally expressed genes are shown within shaded rectangles, above or beside representations of zygotically expressed genes: gradients of shading within
the rectangles depict expression gradients (not to scale). The mRNA of maternal genes written in bold is known to be anteriorly and/or posteriorly tethered during oogenesis. Genetic interactions depicted for T. castaneum and N. vitripennis are determined from RNA interference experiments and so cannot be assumed to reflect direct regulation (hence the dotted lines). Recent models suggest that the establishment
of the Krüppel gap domain in D. melanogaster can largely be explained by both positive (at low concentrations) and negative (at high concentrations) regulation by maternal hunchback (Papatsenko and Levine, 2008).


et al., 2007) (see Table 16.1 and Figure 16.3b,c,e,f). In T. castaneum, on the contrary, neither of the anteri- orly localized mRNAs identified to date play a significant role in anterior–posterior patterning (Bucher et al., 2005), and the maternal mRNAs of two genes known to be important anterior deter- minants in T. castaneum—hunchback and orthoden- ticle-1—are initially distributed uniformly in the egg (Wolff et al., 1995; Schröder, 2003; Schinko et al.,
2008) (see Table 16.1 and Figure 16.3a,d). It is pos- sible that an anteriorly localized maternal mRNA, whose protein product forms an instructive pat- terning gradient exists in T. castaneum but has been overlooked. However, it is tempting to speculate that there is an association between the retention of sequential segmentation in T. castaneum, and the lack of an instructive anterior patterning gra- dient. It will be interesting to determine whether an instructive anterior patterning gradient has evolved in those beetle lineages in which there has been an increase in the number of segments patterned in the blastoderm prior to gastrulation (Patel et al., 1994).

16.3.2 Molecular transitions at the posterior egg pole: changes in the regulation of pair-rule gene homologues

The primary pair-rule genes (even-skipped, hairy, and runt) are the first genes within the D. mela- nogaster segmentation cascade to be expressed in a periodic pattern of stripes (Jaynes and Fujioka,
2004) (see Figure 16.2). The primary pair-rule genes activate a suite of secondary pair-rule genes, that includes paired, sloppy-paired-1 and -2, and odd-skipped (Jaynes and Fujioka, 2004). They are collectively referred to as pair-rule genes since their mutation often leads to abnormalities in alternate segments. Recent work on T. casta- neum has revealed divergent regulatory interac- tions between the homologues of D. melanogaster pair-rule genes. Choe et al. (2006) showed that it is the homologues of D. melanogaster even-skipped, D. melanogaster runt and D. melanogaster odd- skipped that comprise the primary tier of pair-rule genes in T. castaneum. The authors disrupted the expression of each primary pair-rule gene in turn
using parental RNA interference (RNAi) and then examined the expression of the remaining genes in knockdown embryos. Surprisingly, rather than canonical pair-rule phenotypes, the individual knockdown of each of these genes result in aseg- mental phenotypes in which all but a few anterior segments are deleted. Although direct regulatory interactions were not proven, the results suggested to the authors that T. castaneum even-skipped acti- vates T. castaneum runt, which in turn activates T. castaneum odd-skipped, which completes a regula- tory cycle by repressing T. castaneum even-skipped. On the basis of these data, it was proposed that the genes comprise a regulatory gene circuit, each cycle of which sequentially patterns pairs of seg- ments via the downstream regulation of a second- ary tier of pair-rule gene homologues composed of T. castaneum paired and T. castaneum sloppy- paired (note that T. castaneum hairy does not appear to play a role in trunk segmentation; Choe and Brown, 2007).
This model and the data it is based on raise an
interesting question. How could a regulatory cir-
cuit of transcription factors—that by definition
must operate intracellularly—pattern segments
within a cellularized zone of extension? One pos-
sibility I suggest is that the proposed transcription
factor circuit—or perhaps just some components
of it—constitute an intracellular molecular oscilla-
tor with an analogous role to the molecular oscil-
lators that control sequential segmentation of the
pre-somitic mesoderm during vertebrate develop-
ment (see reviews by Pourquié, 2004, and Gridley,
2006). The vertebrate segmentation clock relies on
the Notch intercellular signalling pathway, both as
a component of the molecular oscillator in some
cases (Pourquié, 2004; Gridley, 2006) and to coord-
inate oscillations amongst neighbouring cells
(Masamizu et al., 2006). Work on myriapods and
chelicerates has shown that some pair-rule gene
homologues and members of the Notch intercellu-
lar signalling pathway are expressed in a dynamic
fashion during sequential segmentation in a man-
ner reminiscent of that seen during vertebrate
sequential segmentation (Stollewerk et al., 2003;
Chipman et al., 2004; Schoppmeier and Damen,
2005), leading to the exciting hypothesis that a
segmentation clock, analogous if not homologous


Table 16.1 A summary of some Tribolium castaneum and Nasonia vitripennis segmentation gene homologues shown by recent studies to exhibit differences in expression and function when compared to Drosophila melanogaster (see also Figures 16.2 and 16.3).

Gene T. castaneum N. vitripennis

bicoid1 bicoid not present in the genomes of insects outside the cychlorrhaphan flies
Evolved from a zerknüllt (zen)–like precursor
Orthodenticle-1 appears to play an analogous role to D. melanogaster Bicoid in these species

orthodenticle-12– 4 mRNA is maternally inherited—unlike in D. melanogaster where expression is purely zygotic—but not localized to the egg poles as in N. vitripennis
Maternal plus zygotic expression becomes anteriorly restricted
Functions with Hunchback to pattern the anterior of the embryo

giant4– 6 No maternal expression as is the case in D. melanogaster
Expressed in two zygotic gap-like domains as in
D. melanogaster, except that the posterior domain is positioned much more to the anterior
Anterior domain controls segment identity via the regulation of Hox genes, but unlike in D. melanogaster, it is not required for segment formation
Required for the formation of all thoracic and abdominal segments, not just the segments in which it is expressed

cauda13, 4,7–10 Maternal mRNA initially uniformly distributed as in
D. melanogaster
Posterior protein gradient forms through translational repression by an unknown factor/factors (i.e. not Bicoid)
Expressed in the cellularized growth zone
Unlike in D. melanogaster, required for the formation of all but the most anterior few segments

Unlike in D. melanogaster, maternal mRNA is localized to the anterior and posterior poles (via distinct mechanisms at either pole)
On fertilization, mRNA is released and forms an opposing anterior and posterior mRNA and, through translation, protein gradients
Anterior gradient functions with maternal Hunchback to activate anterior gap genes: empty spiracles, giant, and (zygotic) hunchback
Posterior gradient functions with Caudal to pattern posterior segments
Largely conserved zygotic head gap gene role

Unlike in D. melanogaster, maternal mRNA is localized to the anterior during oogenesis
On fertilization, mRNA is released and forms an anterior mRNA and, through translation, a protein gradient
Represses central gap gene Krüppel in anterior, preventing repression of anterior gap gene hunchback by Krüppel, and thus plays a permissive role in anterior development
A similar zygotic gap gene role to D. melanogaster

Unlike in D. melanogaster, maternal mRNA localized to posterior during oogenesis
On fertilization, mRNA is released and forms a posterior mRNA and, through translation, a protein gradient
Functions with Orthodenticle-1 to pattern posterior via activation of posterior gap genes
Influence extends further to the anterior than in
D. melanogaster and includes activation of the central gap gene Krüppel

tailless and
the terminal patterning system3,11–13

tailless expressed by small group of cells at the posterior pole of the blastoderm.
In contrast to D. melanogaster, there is no expression at the anterior pole of the blastoderm
Terminal patterning system (torso and torso-like) required for sequential segmentation and formation of anterior extra-
embryonic tissue

tailless expression is activated in anterior and posterior by Orthodenticle-1 (i.e. there is no evidence for a terminal patterning system in N. vitripennis)
Unlike D. melanogaster, the anterior domain is not required for segmentation
Posterior domain has more extensive influence on
posterior patterning than in D. melanogaster

References: 1, Stauber et al. (2002); 2, Schröder (2003); 3, Lynch et al. (2006a); 4, Olesnicky and Desplan (2007); 5, Bucher and Klingler (2004);
6, Brent et al. (2007); 7, Schulz et al. (1998); 8, Wolff et al. (1998); 9, Copf et al. (2004); 10, Olesnicky et al. (2006); 11, Schröder et al. (2000);
12, Schoppmeier and Schröder (2005); 13, Lynch et al. (2006b).


to that operating in vertebrates, controls sequential
segmentation in these arthropods (Peel and Akam,
2003; Stollewerk et al., 2003). However, there is no
evidence, as yet, for the involvement of the Notch
signalling pathway during insect sequential seg-
mentation. Wingless signalling also plays a central
role in the vertebrate segmentation clock (Pourquié,
2004; Gridley, 2006). Perhaps wingless signalling
forms the basis of a possible segmentation clock
in insects (Miyawaki et al., 2004) or alternatively
other signalling pathways might be involved.
In D. melanogaster the periodic expression of
primary pair-rule genes is, somewhat curiously,
activated by an aperiodic series of anteroposteri-
orly restricted domains of gap gene expression
(Figure 16.2). Gap genes play an additional role
in D. melanogaster development; they regulate the
anteroposteriorly restricted domains of Hox gene
expression that confer identity to segments (Irish
et al., 1989). In T. castaneum, most D. melanogaster
gap gene homologues are expressed in restricted
anteroposterior domains, consistent with a gap
gene function, and in a roughly conserved antero-
posterior order, albeit shifted towards the anterior
(Schröder, 2003; Bucher and Klingler, 2004; Cerny
et al., 2005). However, the knockdown by RNAi of
at least two of the D. melanogaster gap gene homo-
logues—Krüppel and giant—does not result in
canonical gap gene phenotypes: i.e. the loss of the
segments in and around their domains of expres-
sion (Bucher and Klingler, 2004; Cerny et al., 2005).
Instead, these segments take on abnormal iden-
tities as a result of the misexpression of Hox genes
(Bucher and Klingler, 2004; Cerny et al., 2005). This
might imply that one of the ancestral roles of gap
gene homologues in insects was to position Hox
gene domains correctly, and that they were only
later recruited to pattern pair-rule genes (Peel and
Akam, 2003). Under this model, gap gene recruit-
ment is correlated with the transition to activating
pair-rule stripes simultaneously in a syncytium,
where control by intercellular signalling becomes
redundant and where a spatial rather than tem-
poral regulatory input is required (Peel and Akam,
2003). Presumably transcription factors expressed
at the right time and place in the posterior blasto-
derm were co-opted to regulate progressively
more posterior pair-rule stripes, thus explaining
the complex nature of the regulatory sequence of D. melanogaster primary pair-rule genes. And per- haps the co-option of D. melanogaster gap gene homologues was favoured, since they had already evolved a spatially and temporally corresponding role in Hox gene regulation.

16.4 Molecular transitions underlying the convergent evolution of long-germ embryogenesis

The four major holometabolous insect orders all contain species exhibiting long-germ embryo- genesis, for example the dipteran D. melanogaster, the lepidopteran Manduca sexta, the coleopteran Callosobruchus maculates, and the hymenopterans Apis mellifera, N. vitripennis, and Bracon hebetor (Grbic and Strand, 1998; see also Davis and Patel,
2002). However, the Lepidoptera, Coleoptera, and Hymenoptera also contain species that have retained, or re-evolved in the case of parasitic hymenopterans (Grbic and Strand, 1998), differing degrees of sequential segmentation; for example, the lepidopteran Bombyx mori, the coleopteran T. castaneum, and the hymenopteran A. ervi (Grbic and Strand, 1998). This has led to the idea that long-germ development evolved multiple times independently during the holometabolous insect radiation (see Davis and Patel, 2002). This scenario now seems more likely due to a recent re-evalua- tion of holometabolous insect phylogeny. The gen- eral consensus surrounding the relationship of the major holometabolous insect orders used to be that the Diptera and Lepidoptera are sister groups, and that the Coleoptera form the basal branch in the tree (Whiting, 2002). However, a recent molecu- lar phylogenetic study supports a reversal in the position of the Hymenoptera and Coleoptera, such that the Hymenoptera now form the basal branch (Savard et al., 2006). Thus, the Diptera and Hymenoptera are now separated by an order that contains many species with clear sequential seg- mentation; the Coleoptera.
Work by the Desplan and Pultz laboratories has begun to reveal the molecular basis to segmen- tation in N. vitripennis, which has an embryonic fate map almost identical to that of D. melanogaster (Bull, 1982; Brent et al., 2007). If N. vitripennis and


D. melanogaster did evolve long-germ embryogen- esis independently, as the latest molecular phyl- ogenies suggest, comparisons between these two species are a first step to determining the extent to which underlying gene network changes were also convergent.
As in D. melanogaster (Figure 16.2), the segmen-
tation cascade operating in N. vitripennis can be
divided into four distinct tiers of maternal, gap,
pair-rule, and segment-polarity genes. The major
genetic differences identified so far between
D. melanogaster and N. vitripennis are found right
at the top of the segmentation cascade and relate to
changes in the maternal contribution to patterning.
These differences are summarized in Table 16.1
and Figure 16.3 and are discussed below.
In D. melanogaster, anterior development is
largely under the control of the Bicoid morphogen
gradient (Lawrence, 1992) (Figures 16.2 and 16.3c).
However, bicoid is known to be an invention of the
higher Diptera (Stauber et al., 2002). In N. vitripennis
anterior patterning is accomplished by two distinct
anterior gradients of patterning molecules (Lynch
et al., 2006a; Brent et al., 2007) (Figure 16.3b). In con-
trast to D. melanogaster, where their expression is
purely zygotic, maternal mRNAs of both orthode-
nticle-1 and giant are tethered to the anterior pole
during oogenesis in N. vitripennis (for details see
Lynch et al., 2006a, Brent et al., 2007, and Olesnicky
and Desplan, 2007). Following fertilization these
mRNAs are translated to form anterior gradients
of Orthodenticle-1 and Giant protein. The mater-
nal Orthodenticle-1 gradient functions to activate
anterior segmentation genes, such as the gap genes
empty spiracles, (zygotic) giant, and hunchback (Lynch
et al., 2006a), while the maternal Giant gradient
functions to set the anterior expression boundary
of the central gap gene Krüppel (Brent et al., 2007)
(Figure 16.3b). The repressive role of maternal Giant
is permissive for anterior development, since, in its
absence, Krüppel expression spreads anteriorly to
repress the anterior gap gene hunchback (Brent et al.,
2007) (Figure 16.3b).
The influence of Orthodenticle-1 and Giant on
embryonic patterning does not extend as far to the
posterior as does Bicoid in D. melanogaster (Lynch
et al., 2006a; Brent et al., 2007). Instead, the influ-
ence of the posterior determinant caudal extends
further towards the anterior; in D. melanogaster hunchback regulates (probably both positively and negatively; Papatsenko and Levine, 2008) (Figure
16.3c) the central gap domain of Krüppel, but in N. vitripennis it is caudal that activates this gap domain (Olesnicky et al., 2006) (Figure 16.3b). Indeed, poster- ior patterning in N. vitripennis exhibits significant differences when compared with D. melanogaster. A posterior gradient of Caudal is established in N. vitripennis—in the absence of translational repression by Bicoid (Figure 16.2 and 16.3c)—via the tethering of caudal mRNA to the posterior pole during oogenesis (Olesnicky et al., 2006; Olesnicky and Desplan, 2007) (Figure 16.3b). The mRNA of orthodenticle-1 is also tethered to the posterior pole during oogenesis, such that, together, gradients of Caudal and Orthodenticle-1 protein control poster- ior patterning (Lynch et al., 2006a; Olesnicky et al.,
2006) (Figure 16.3b).
The large degree of variation between D. mela-
nogaster and N. vitripennis in maternal patterning
suggests that the changes in the gene network
underlying the putative independent evolution of
long-germ embryogenesis within the hymenop-
teran and dipteran lineages might have been very
different. However, many of the observed dif-
ferences could be attributed to the evolution of a
maternal Bicoid gradient within the higher Diptera
(Stauber et al., 2002). Indeed, bicoid and orthodenticle
are both homeobox-containing genes, and Bicoid is
predicted to have usurped the role of Orthodenticle
as an anterior determinant (as exemplified by
T. castaneum; Schröder, 2003; Schinko et al., 2008)
and N. vitripennis (Lynch et al., 2006a) orthodenticle-1
(Table 16.1, Figures 16.3a,b) through gaining affin-
ity for Orthodenticle DNA-binding sites via the
convergent acquisition of a lysine residue at pos-
ition 50 in its homeodomain (Treisman et al., 1989).
Further data from additional long-germ dipterans
(e.g. Anopheles gambiae) and hymenopterans (e.g.
A. mellifera) will be required to map divergent (and
convergent) gene network changes to the dipteran
and/or hymenopteran lineages. For example, it
is interesting that regulation by zygotic giant is
involved in the maintenance and refinement of the
Krüppel gap domain in D. melanogaster (Papatsenko
and Levine, 2008) (Figure 16.3c). Was giant recruited
to play an earlier role in the establishment of the


Krüppel gap domain within the lineage leading to N. vitripennis, or was a role for giant at the mater- nal level a character that has been lost during the dipteran radiation within the lineage leading to D. melanogaster?
There does, however, already appear to be a strong case for differences in the evolution of the terminal patterning system during the putative independent evolution of long-germ development within the hymenopteran and dipteran lineages. In D. melanogaster a terminal patterning system acts to determine the most anterior and posterior regions of the embryo (Lawrence, 1992) (Figure 16.2). The product of the gene torso-like is maternally restricted to the egg poles where it cleaves the protein encoded by the gene trunk (Casali and Casanova, 2001). The Trunk C-terminal then acts as a ligand, binding to the receptor tyrosine kin- ase encoded by torso and triggering a signalling cascade that regulates the zygotic expression of downstream segmentation genes, such as tailless, at either pole of the egg (Casali and Casanova,
2001) (see Table 16.1 and Figure 16.3f). In T. casta- neum, both torso and torso-like are required at the posterior pole for the activation and/or mainten- ance of sequential segmentation (Schoppmeier and Schröder, 2005) (Figure 16.3d). However, in N. vitripennis the expression of tailless has been shown to be dependent on the anterior and pos- terior gradients of maternal Orthodenticle-1 and thus perhaps not a terminal patterning system (Lynch et al., 2006b) (Figure 16.3e). The absence of a D. melanogaster-like terminal patterning system in hymenopterans is further supported by the failure to find homologues of torso or trunk in the A. mellifera genome (Dearden et al., 2006).
The apparent lack of a Drosophila-like ter- minal patterning system in long-germ hymen- opteran insects is intriguing, particularly since one appears to be required for sequential seg- mentation in T. castaneum (Schoppmeier and Schröder, 2005; Lynch et al., 2006b) (Figure 16.3e). Studies on more holometabolous insects, as well as non-holometabolous insects, will be required to determine if a terminal patterning system is a character that has been lost within the hymen- opteran lineage or gained in the lineage leading to the other major holometabolous insect orders.
One possibility is that in some of the insect line- ages that underwent the transition to long-germ embryogenesis (i.e. in dipteran lineages) a ter- minal patterning system that ancestrally played an important role in initiating or maintaining sequential segmentation was co-opted to activate posterior segmentation genes (Schoppmeier and Schröder, 2005), whereas in other lineages under- going parallel transitions (i.e. in hymenopteran lineages) this did not occur.

16.5 General conclusions

16.5.1 The role of co-option in the evolution of segmentation gene networks

A theme emerging from comparative studies on insects is the role that co-option plays in evolution, at different levels of complexity. At the regulatory sequence level, it seems possible, if not likely, that some gap gene homologues—expressed at the right time and in the right place due to an ancestral role in Hox gene regulation—have been co-opted into regulating pair-rule gene homologues, perhaps via the simple acquisition of binding sites (Bucher and Klingler, 2004; Cerny et al., 2005; Choe et al., 2006). At the protein level, Bicoid—or perhaps more accurately Zerknüllt—was co-opted to an anterior patterning role within the higher Diptera via a sim- ple coding mutation that allowed it to recognize the regulatory targets of an existing anterior determin- ant; Orthodenticle (Stauber et al., 2002; Lynch et al.,
2006a). At the intracellular level, existing cytoskel- etal machinery may have been co-opted during the evolution of instructive anterior patterning gradients (Bucher et al., 2005). All these cases are consistent with a long series of simple modifica- tions to developmental gene circuits having, over evolutionary time, underpinned diversification at the cellular level and above. The results of a recent whole-genome study are consistent with the widespread occurrence of gene co-option during insect evolution. Dearden et al. (2006) looked at the presence/absence of homologues of D. melanogaster developmental genes in the A. mellifera genome. They found that, of the developmental genes involved in processes that are known to be diver- gent between these two species (i.e. sex determin-


ation, dosage compensation, meiosis, and germ-cell development), a significant number of those con- served in A. mellifera (c = 19.03; P < 0.001, n = 78) have multiple (i.e. pleiotropic) functions in D. mela- nogaster. This suggests either that the homologues of genes with multiple functions in D. melanogaster have been lost less frequently in the honeybee lin- eage and/or that genes with ancestral conserved functions have been frequently co-opted into new roles in the fruitfly lineage (Dearden et al., 2006).

16.5.2 The evolution of developmental gene networks in relation to adult morphology

The comparative studies reviewed in this chap- ter clearly demonstrate that genetic networks controlling the development of conserved adult structures (i.e. homologous insect segmental units) can diverge significantly over time due to lineage-specific transitions in cellular embryology
associated with changes in life-history strategy. They also suggest that distinct changes in gene networks might underlie convergent transitions in modes of cellular embryogenesis. This implies that over the course of hundreds of millennia, developmental gene networks and the adult morphology they pattern can become ‘decoupled’. Assigning homology, or otherwise, to adult mor- phological features based on comparative devel- opmental genetic data alone is therefore risky. Accurately reconstructing the evolution of animal body plans will require a holistic approach, which includes adequate and intelligent sampling of spe- cies (i.e. perhaps less focus on species exhibiting highly derived modes of embryogenesis), a more thorough understanding of the embryological con- texts in which gene networks operate, and a better appreciation of how evolutionary changes in life- history strategy (i.e. changes in species ecology) can influence the evolution of development.