ariel d. chipman
In the past decade or so, there has been a significant increase in the available data on the developmental mechanisms underlying the process of segmentation in a wide range of arthropod taxa. This large body of data makes it possible to attempt, albeit cautiously, a comparative analysis of the various aspects of the segmentation process, and to try to find which of its features and components may have been present in the arthropod common ancestor. A recent review (Peel et al. 2005) covers much of what is known about the diversity of segmentation processes in arthropods, although even at the time of this writing, less than a year later, there is already a substantial amount of newly published data not covered therein. My aim in this chapter is not to repeat the review and synthesis presented in Peel et al. (2005), but to build on it, adding the most recent data, and expand the discussion into the more speculative domain of evolutionary reconstructions. The reader is encouraged to refer to that review for more details of the currently available data and for a more complete bibliography.
When addressing a large-scale evolutionary question, such as that suggested in the title of this chapter, it is important to define the boundaries of the problem discussed. In this review, I will focus only on the mechanisms of trunk segmentation, ignoring the differentiation and segmentation of the head region, and the posterior unsegmented region. Although it is difficult to think of the trunk as a distinct compartment that stands on its own, and many of the relevant developmental processes are continuous between the trunk and the areas
Evolving Pathways: Key Themes in Evolutionary Developmental Biology, ed. Alessandro Minelli and Giuseppe Fusco. Published by Cambridge University Press. # Cambridge University Press 2008.
immediately anterior and posterior to it, not enough is known about how these areas are defined in a range of arthropods. For the purpose of the discussion, trunk segmentation will be thought of as a continuing process, without any consideration of where it begins or ends.
The phylogenetic scope of the analysis will include the Euarthro-poda only, excluding tardigrades and onychophorans. In fact, because of the almost complete lack of data on segmentation in pycnogonids, these are also excluded from the discussion, leaving us with an evolutionary reconstruction of the last common ancestor of the Cormogo-nida (arthropods excluding pycnogonids; Zrzavy et al. 1997, Dunlop and Arango 2005). For simplicity, I will continue talking about 'arthropods', even though I will, in fact, be covering a somewhat smaller group. For it to be possible to reconstruct the common ancestor of such a diverse clade, we need a sample that covers as much of its diversity as possible. There are good data about many species of insects; mostly, but not exclusively, holometabolous insects. Outside the insects, information is more limited, but segmentation has been studied extensively in the spider Cupiennius salei, as a representative che-licerate; and in two centipedes, Strigamia maritima and Lithobius forficatus, and a millipede Glomeris marginata representing the myriapods. Within the crustaceans most of the available data come from the Malacostraca, mainly Parhyale hawaiiensis, but also others.
As mentioned above, there is a wealth of data about segmentation mechanisms in diverse arthropods. One must be cautious when dealing with such a large amount of information to differentiate between characters that are specific to one system or taxon and characters that can be considered generalities and can be used for ancestral reconstruction. I will attempt to concentrate only on the latter, and try to avoid being bogged down with too many bits of specific information. It is worth pointing out that many arthropods do not generate all of the trunk segments during embryogenesis, but continue adding segments during larval stages, a phenomenon known as anamorphic development (see Fusco 2005 for a discussion). A further complication is that in a few cases, the segmentation process is not necessarily generating a single series of segments along the trunk, but a number of distinct serial structures as in the dissociation of dorsal and ventral segmentation in the pill millipede (Janssen et al. 2004). For simplicity, I will not consider these two exceptions, but will stay with the case of a single series of segmental structures formed as part of a single process. A final pitfall I hope to avoid is getting entrenched in the so-called 'Drosophila paradigm'. Although I will use what is known about segmentation in Drosophila melanogaster as a reference, it will be made clear that Drosophila is but one example, and an unusual one at that, so no general conclusions can be drawn from information on Drosophila alone.
The reconstruction of the segmentation process in the arthropod common ancestor will use both available molecular data and cellular/ morphological data. The latter data type is, in general, much poorer for diverse arthropods, and necessarily reconstruction of cellular and morphological processes relies more on deduction from basic principles and extrapolation from the little that is known. For molecular data, the approach I will take is essentially one of ancestral character reconstruction using parsimony, although the number of data points for any given molecular player or interaction is usually much too small for a formal or mathematical analysis. Of course, the ever-present spectre of convergence is a problem for parsimony-based reconstructions. Nonetheless, I will assume that the expression of homologous genes in comparable domains during similar processes is sufficient for assuming common ancestry. In many cases, homologous genes are apparently performing the same function in very different cellular environments, so gene expression patterns alone tell us little about cellular processes. Thus, molecular data and cellular/morphological data will be considered separately, to be joined only at the end when I go to the ancestral reconstruction itself.
Finally, what I will present is my own interpretation of the data. The data, despite their wealth, are still very partial, and are open to many alternative interpretations. Some readers may prefer different evolutionary scenarios, and I agree that many are possible. Nonetheless, the evolutionary scenario I will present is consistent with the data and addresses the key questions about the ancestral condition of the arthropod segmentation mechanism.
ways of making segments
Making a segmented body involves defining a main body axis and generating a repeated pattern along that axis during ontogeny. The main source of information about segmentation in arthropods is the fruit fly Drosophila melanogaster. In Drosophila, the repeated pattern is generated by a progressive subdivision of the blastoderm - a thin layer of tissue surrounding a yolky egg, which will ultimately give rise to the embryo itself. This is possible in the Drosophila embryo because almost the entire anterior-posterior extent of the blastoderm will give rise to the segmented germ band (so-called 'long germ' embryos). In most arthropods, this is not the case, and the initial embryonic rudiment is much shorter than the ultimate extent of the germ band. In embryos with this kind of short rudiment (often referred to as 'short germ'), an additional step is required before a reiterated pattern can be generated, namely axial elongation, in which the initial short rudiment is extended to give the full length of the germ band. These three processes - axis determination, axial elongation (in non-long-germ embryos) and generation of a repeated pattern - are central to the creation of segmented body plan during embryogenesis. I will start by going through these three and examining how they are manifested in diverse arthropod embryos. In what follows, I will use the term 'germ band' to refer to axially polarised tissue that has undergone a certain amount of differentiation and arrangement to distinguish it from the initial undifferentiated field of cells. The germ band includes both tissue that is overtly segmented, and tissue that is still unsegmented but may have already undergone some of the molecular processes that precede segmentation.
setting up an anterior-posterior axis
The anterior-posterior axis in the best-studied model, Drosophila melano-gaster, is set up by maternal determinants deposited in the egg during oogenesis. As with many Drosophila characters, it is difficult to extrapolate from this to other arthropods, because the existence of nurse cells, which are responsible for the loading of these maternal determinants, is not universal within arthropods. However, many of the genes and gene products that are active in axis determination in Drosophila can be found in other arthropods as well. First and foremost among these is the homeobox-containing transcription factor Caudal. Expression of caudal in the posterior of early embryos has been found in all arthropods where it has been looked for (Peel et al. 2005, Olesnicky et al. 2006), and indeed in other metazoans as well (Epstein et al. 1997, Holland 2002, de Rosa et al. 2005, Shimizu et al. 2005). This posterior expression pattern, together with functional studies in Drosophila, suggests a role in determining the posterior pole. The widespread appearance of posterior caudal in arthropods and in outgroups indicates that it probably had a similar role in the arthropod common ancestor.
A second gene that may have a conserved ancestral role is nanos. In Drosophila, it is expressed posteriorly, and represses translation of the maternally deposited transcription factor hunchback (Irish et al. 1989). Homologues of nanos can be found in most metazoans, indicating that the gene itself existed in the arthropod ancestor. Most of what is known about its role is limited to insects. The translational repression of hunchback is conserved in dipterans (Curtis et al. 1995), as well as in grasshoppers (Lall et al. 2003) and possibly in the jewel wasp Nasonia vitripennis (Pultz et al. 2005). Nothing is known about nanos in other arthropods, but in the very distantly related Cnidaria, it also has a posterior expression pattern in the developing embryo (Torras et al. 2004, Torras and González-Crespo 2005). Another homeobox-containing transcription factor that is involved in axis formation in some insects is Orthodenticle. In N. vitripennis, orthodenticle patterns both the anterior and the posterior poles of the embryo (Lynch et al. 2006), whereas in the beetle Tribolium castaneum it has a role in defining the anterior pole (Schröder 2003). However, studies in non-insect arthropods do not support such a role outside of the insects (Browne et al. 2006), so its involvement in axis determination in an arthropod ancestor cannot be confirmed.
Drosophila melanogaster cannot provide any clues about the evolutionary history of axial elongation mechanisms in arthropods, since as a longgerm insect it already has the entire anterior-posterior extent of the germ band present at very early developmental stages. Looking at expression patterns and functional studies in other arthropods can give some hints as to the ancestral players in this process.
I have already mentioned caudal as a key player in the determination of the posterior pole during axial specification in many arthropods. In addition to this role, it also has a central role in axial elongation. Experimental knock-down of caudal expression results in a complete disruption of the segmentation process and a truncation of the growth of the embryo posterior to the gnathal segments (Copf et al. 2003, Shinmyo et al. 2005).
A similar phenotype to the caudal knock-down is seen when knocking down even-skipped in Oncopeltus fasciatus (Liu and Kaufman 2005a). In several insects even-skipped is expressed in a broad posterior domain (reviewed in Liu and Kaufman 2005b). In the brine shrimp Artemia franciscana it is also expressed broadly in the posterior (Copf et al. 2003), and in the centipede Strigamia maritima an even-skipped homologue, eve2, is one of a group of genes expressed very early in the segmentation process (A. Chipman and M. Akam, unpublished data). These results suggest an early role and possible involvement in axial elongation for even-skipped, a surprising suggestion, given that in Drosophila, even-skipped is generally thought of as a gene that is involved in the segmentation cascade at a relatively late stage. However, looking outside of arthropods, it has been suggested that even-skipped and caudal are jointly involved in axial elongation even in the annelid Platynereis dumer-ilii (de Rosa et al. 2005).
At the cellular level, there are two possible mechanisms of axial elongation. Elongation can be done either through the activity of a growth zone, in which cell proliferation contributes new tissue throughout the elongation process as seen in many malacostracan crustaceans (Scholtz et al. 1994, Wolff and Scholtz 2002), or through rearrangement and recruitment of existing tissue to the elongating germ band as seen in the centipede Strigamia maritima (Chipman et al. 2004a). These two possibilities represent two extremes of a continuum, and in most cases, axial elongation is probably achieved through an intermediate process incorporating contributions by both mechanisms. It should be pointed out that the details of axial elongation in most arthropods are very poorly known, and have not been studied extensively (Liu and Kaufman 2005b). Keller (2006) discusses, in a much wider phylogenetic scope, the many different types of cellular mechanisms that are involved in elongation processes in development. Although he does not give much information about arthropods, he provides possible clues to the types of processes we could look for in axial elongation. What seems clear is that in probably all cases, the elongation zone is subterminal, that is, there is a terminal zone, which remains constant and does not participate in the elongation process.
generating a repeated pattern
Once again, the mechanism for generating a repeated pattern in Droso-phila melanogaster cannot give us information about other arthropods, since in higher Diptera - unusually - all segments are formed almost simultaneously through a stepwise subdivision of the germ band. Generating a repeated pattern in an elongating embryo, as is more common in arthropods, can be accomplished in two ways, dependent in part on the mode of axial elongation. One possibility is that one end of the rudiment is made of a population of cells that is constantly proliferating. As new cells are formed, each is given an identity - based on the timing of its birth - that corresponds to a specific role or position within the forming segment, and thus a reiterated pattern is created simultaneously with the generation of new tissue. The alternative possibility is that tissue addition is independent of generating a repeated pattern, and the pattern is generated through a periodic input acting on unsegmented tissue that has already been recruited to the germ band. The first alternative is only possible in embryos where the process of elongation is tied to the generation of new cells, as seen in the extreme case of malacostracan crustaceans. Since this type of elongation is probably a malacostracan apomorphy (Scholtz et al. 1994, Scholtz 1998), the ancestral arthropod pattern is more likely to be one of patterning an undifferentiated population of cells through a periodic signal.
The models that best predict the appearance of a periodic pattern in an undifferentiated field of cells are models including a cellular oscillator or clock, in which cells oscillate between a series of cell states with a fixed periodicity. The cell states can be thought of as specific expression levels of a set of genes or their products. The first formulation of such a model was the clock and wavefront model (Cooke and Zeeman 1976), in which all the cells of the pre-segmented tissue oscillate with a linked phase (i.e. they are all simultaneously at the same phase of the cycle) at a relatively high frequency. A slow moving wave of cell states, or expression of a different set of genes and gene products, passes over the oscillating tissue in an anterior-posterior direction, with the front of the wave including a rapid change of cell state (i.e. a steep gradient of expression). Each cell in the pre-segmented tissue gets fixed in a specific state, depending on when it meets the wavefront. More recent models are based on a cell-autonomous oscillator, in which cells emerge from a growth zone or progress zone and continue oscillating between states. In this model the cells are not locked in the same phase, but each cell communicates with cells anterior and posterior to it, creating a travelling wave of oscillating states that moves in an anterior direction. These oscillations slow and eventually stop in a specific state based on the time since they emerged from the growth zone (Jaeger and Goodwin 2001).
The existence of a somitogenesis clock in vertebrates has been demonstrated by a number of workers on different vertebrate systems (Palmeirim et al. 1997, Pourquie 2003, Dubrulle and Pourquie 2004, Giu-dicelli and Lewis 2004), and vertebrate segmentation seems to conform to the aforementioned theoretical models (Aulehla and Herrmann 2004). The main components of the vertebrate clock are genes in the Notch signalling pathway that are expressed in a cycling fashion through a negative feedback loop (Collier et al. 1996, Rida et al. 2004).
The discovery that Notch and its ligand Delta are involved in segmentation in the spider Cupiennius salei (Stollewerk et al. 2003) set off a flurry of interest in the similarities between vertebrate somitogenesis and arthropod segmentation. Involvement of Notch pathway genes in segmentation has since been found in centipedes (Chipman et al. 2004b; A. Chipman and M. Akam, unpublished data) and in the cockroach Peri-planeta americana (J.-P. Couso, unpublished data). Although several attempts have been made to find similar genes in other insects, there have been no additional conclusive data. In the case of the centipede Stri-gamia maritima the Notch-ligand gene Delta and the Notch target odd-skipped are expressed in what looks like an oscillating pattern of travelling waves (Chipman et al. 2004b; A. Chipman and M. Akam, unpublished data), as would be expected if the segmentation clock models were true for centipede development. There is no current evidence to indicate the existence of a gradient or wavefront that would interact with an oscillator in arthropod segmentation, nor a clear indication as to its identity. Jaeger and Goodwin's (2001) cellular oscillator model does not require a specific gradient, and in the case of S. maritima it may be sufficient for the oscillating signal to become fixed when it reaches the germ band after travelling through the undifferentiated posterior zone. However, if we were to speculate about such a gradient, a possible candidate would be caudal, which is present in a graded pattern at the correct place and time. Circumstantial support for this hypothesis is provided by the fact that caudal is expressed in periodic stripes at exactly the point where the oscillating pattern of S. maritima is fixed (Chipman et al. 2004b), possibly through some feedback from the oscillator.
translating a repeated pattern into segments
The segmented germ band is a highly conserved stage in arthropod development, and has even been called the arthropod 'phylotypic stage' (Raff 1996, Galis et al. 2002). This conservation of a morphological stage is also represented at the molecular level (Peel et al. 2005), and unlike earlier stages in the process, information from Drosophila segmentation is applicable to other arthropods as well. The involvement of a series of segment polarity genes, namely engrailed, hedgehog, wingless and others in generation of segmental boundaries (von Dassow et al. 2000, Larsen et al. 2003), is conserved in all arthropods where it has been studied (Peel et al. 2005). Indeed, engrailed is so ubiquitously conserved that it is the standard marker for segmental boundaries in almost all studies of arthropod segmentation. Directly upstream of the segment polarity genes are a group of genes that in Drosophila are referred to as pair-rule genes. This group includes even-skipped, odd-skipped, hairy, runt and several others. In Drosophila, they are initially expressed in a two-segment periodicity, which is then split to give the single segment periodicity of the segmented germ band. Homologues of pair-rule genes have also been found wherever they have been looked for within the arthropods (Peel et al. 2005, Choe et al. 2006). However, their expression in a two-segment periodicity is not universal. In most insects, at least some of these genes are involved in generating a two-segment repeat. In non-insect arthropods, with the exception of the geophilomorph centipede S. marítima, they are expressed segmentally (Peel et al. 2005). The highly conserved involvement of pair-rule gene homologues in the final stages of arthropod segmentation suggests they may represent the primary output of the oscillator. Furthermore, in many arthropods there is a functional division within the pair-rule genes between primary and secondary genes (Coulter and Wieschaus 1988, Damen et al. 2005, Choe and Brown 2007), in which the primary pair-rule genes are upstream of the secondary ones, and hence possibly the immediate output of the oscillator. The exact members of each of these subgroups vary from species to species. It may be that such a functional division existed in the common ancestor of arthropods, but that individual genes have moved between primary and secondary roles with relative ease throughout evolution (Choe and Brown 2007).
what about gap genes?
Readers who are familiar with Drosophila segmentation will have noticed the conspicuous absence of gap genes in my discussion up to this point. In the conceptual sequence of events leading to a segmented body plan, such as I have outlined in this chapter, the gap gene phase, which forms the crucial early pattern in Drosophila segmentation, is not necessary. However, gap gene homologues have been found in insects other than Drosophila and they are claimed to have a role in the generation of the segmented body plan (Peel et al. 2005). The changing role of gap genes in different insects has been discussed in Peel et al. (2005), and I will not repeat that discussion here. Suffice it to say that the evidence suggests that the gap genes in sequentially segmenting insects are not involved in segment generation per se, but rather in generating segmental identity. The only data on gap gene homologues outside of insects come from the centipede S. maritima, and for hunchback only, from the brine shrimp Artemia franciscana. In the centipede, two of the gap genes, Krüppel and hunchback, are suggested to have a role in neural precursor identity, long after the segments have formed (Chipman and Stollewerk 2006) - a role that is conserved in Drosophila as well (Isshiki et al. 2001). Both of these genes, as well as knirps and giant, have no relevant expression during segmentation stages (A. Chipman and M. Akam, unpublished results). In the brine shrimp, hunchback is expressed in already segmented mesoderm (as is also seen transiently in the centipede) but has no gap-like pattern, or obvious role in the segmentation process (Kontarakis et al. 2006). With this rather sketchy evidence, it is difficult to draw firm conclusions about the ancestral role of gap genes. I would suggest, however, that whatever this role was, gap genes were not significant players in the sequential segmentation process itself, although gap genes could have had a role in patterning the anterior part of the embryo, including the head segments.
reconstructing the ancestor
Having pointed out the conserved aspects in each of the stages of the segmentation process in arthropods, I now move on to reconstructing a series of hypothetical segmentation events that may be similar to the process in the common ancestor of all arthropods (Figure 18.1).
Virtually nothing is known about what embryos of early arthropods looked like, and the fossil record has been silent on this question to date. Many aspects of the earliest phases of the segmentation process are dependent on the size, shape and yolk content of the egg and on the extent of the embryonic rudiment relative to the egg. Leaving these considerations aside, for lack of information, I will stay with the most simplistic generalities based on what we do know.
Initially, the embryo would include a uniform, unpatterned field of cells. One end of the field would have to be slightly different, either through an inherent asymmetry in the egg, or because of the point of sperm entry, or by random gravitational orientation. This end would form one of the poles of the anterior-posterior axis. The posterior pole would be defined by the expression of caudal and possibly nanos, as is probably the case in all extant arthropods. The target of one or both of these genes might be hunchback, which may have been initially distributed uniformly.
Once the anterior-posterior axis had been set up, a germ band would form along this axis. At the time of the beginning of axial elongation, it is likely that a head or head rudiment would already be in existence, but as mentioned at the beginning of the chapter, I will not discuss how this could have been accomplished. Tissue would be recruited to the posterior of the embryo by cell rearrangements with cells moving from a pool of undifferentiated tissue, which would be
Figure 18.1 A schematic representation of the three main stages in the segmentation process of the hypothetical arthropod common ancestor. Cells are represented as ovals with a dot. Undifferentiated cells are drawn as larger than germ-band cells, but this is only for illustrative purposes and is not meant to imply that such a difference existed. Shades of grey represent gene expression levels. Captions referring to cellular events or domains are in bold while captions referring to molecular events are in italics. A, A gradient of nanos and/or caudal with a higher concentration at the posterior defines the posterior pole in an undifferentiated field of sparsely packed cells. B, The germ band has formed in the anterior (left) side of the embryo, and cells from the undifferentiated field are recruited for its elongation. Cell divisions are scattered throughout the undifferen-tiated area. Expression of caudal and/or even-skipped in the posterior is involved in the elongation of the germ band. C, Oscillating expression of Notch pathway genes and their downstream targets moves through the undifferentiated field, with the progress of the wave slowing as it moves towards the germ band, and eventually becoming fixed in the germ band.
replenished by unordered cell proliferation. The movement of cells and their addition to the growing germ band would be orchestrated by the activity of Caudal and Even-skipped. Exactly how these two would interact and what their targets would be remains unclear. It is possible that one or both of these would maintain posterior cells in an undifferentiated state, and as cells moved out of the control of these genes they would gain germ-band properties.
As the germ band is extending a series of genes would start expression in an oscillating pattern that overlays the process of germband extension. The primary oscillation would probably be through a Notch-Delta mediated negative feedback loop, with several other genes and gene-products oscillating with these two at different phases. These additional genes could either be direct components of the clock or immediate targets of some component of the clock. It is likely that the focus of the oscillating pattern would be in the very posterior of the embryo, which does not participate in the elongation process. The cycling expression patterns would move as a wave through the undiffer-entiated tissue, slowing as they entered the germ band and becoming fixed at different phases in different cells. The different phases of the cycle would be manifested by different combinations of pair-rule gene homologues - some activated directly by the main components of the Notch pathway and others secondarily by the earlier group of pair-rule homologues. The question of whether each cycle of the Notch-Delta oscillator and the downstream pair-rule homologues would represent one or two segments in the segmented germ band cannot be answered conclusively, since examples of both are found in different arthropod classes. I tend towards a single-segment periodicity as the output of the initial cycler, since a two-segment periodicity seems to be a derived feature, appearing convergently in specialised groups (the holo-metabolous insects and geophilomorph centipedes).
The combinations of pair-rule homologues would be read by segment polarity genes, such as engrailed and wingless, the products of which would activate the cellular components involved in morphological differentiation of the segments and the establishment of segmental boundaries.
Although the process has been described as including a series of discrete and separable steps, in reality all these steps are linked and difficult to tease apart. This is true both in the hypothetical ancestor and in real extant arthropods. Axis specification is closely tied to axial elongation (as is evident by the involvement of caudal in both processes). Axial elongation is closely linked to the generation of a repeated pattern.
Translating a repeated pattern into segments is a gradual process and not a series of leaps from stage to stage. Nonetheless, these are conceptually different process, and could probably vary independently throughout evolution.
It may be true that my description of the hypothetical arthropod ancestor reflects my own personal biases. The process presented above bears striking similarities to the segmentation process in the geophilo-morph centipede Strigamia maritima. My favourite animal has much to recommend it, and was chosen as a research organism for unrelated reasons (Arthur and Chipman 2005). However, it turns out to have many features that make it useful for speculations about ancestry. The simplicity of its overall morphology - a large number of homonomous trunk segments, which are all generated during embryogenesis -makes it possible to draw generalities from the specifics of S. maritima development. Still, the general reconstruction I have presented is based on a wide comparison, and not just from my own work on centipede development. No feature of S. maritima development can be assumed (or has been assumed) to be ancestral, without corroborating comparisons with distantly related arthropods.
The ancestral reconstruction I have presented above might seem just an entertaining exercise in scenario building. However, it is more than that. First, by setting out the available data and building upon them, it clarifies exactly where there are gaps in our knowledge and points out interesting and potentially useful avenues of research. Second, it provides a testable hypothesis about what features in the arthropod segmentation process can be deemed as generalities. As the evo-devo community expands and we learn more about the development of different arthropod taxa, this scenario will be corrected and refined to give a more reliable picture of how the distant ancestors of the arthropods made segments.
Was this article helpful?