Portal de Periódicos da CAPES

Segmentation in animals

2008; Elsevier BV; Volume: 18; Issue: 21 Linguagem: Inglês

10.1016/j.cub.2008.08.029

1879-0445

Seth S. Blair,

Marine Ecology and Invasive Species

Segmentation is the serial repetition of similar organs, tissues, cell types or body cavities along the anterior-posterior (A-P) axis of bilaterally symmetric animals (bilaterians). You, like other vertebrates, are segmented — consider the skeleton, musculature and nervous system of your trunk. Segmentation has provided a fascinating puzzle for developmental biologists and mathematical modelers interested in complex patterns of differentiation. And to those interested in the evolution of different animal body plans, segmentation has provided another important character to discuss. How often has segmentation been invented, how often lost, how easily can it be altered (and why), and what does this mean for the history and mechanisms of animal evolution? We have a deep understanding of the molecular bases of segmentation for just a few model species: the insect Drosophila melanogaster (in the phylum Arthropoda, along with crustaceans, chelicerates and myriapods), and various vertebrates (mostly zebrafish, amphibians, chicks and mice, all members of the phylum Chordata). It is striking just how different Drosophila and vertebrate segmentation appear. In Drosophila, the entire length of the embryo is simultaneously subdivided into segments, in large part at an early stage in which nuclei are not separated by cell membranes (the syncytial blastoderm). The syncytium is critical because the early steps of segmentation rely upon the movement of transcription factors through the cytoplasm (Figure 1A). Gradients of anterior and posterior transcription factors roughly subdivide the A-P axis into 'gap' domains. Gap transcription factors overlap to define smaller pair-rule domains. Pair-rule transcription factors, such as Hairy, are expressed in alternating 'parasegments', which have boundaries in the middle of each prospective segment. By the time cell membranes are formed, parasegments are already defined, and signaling between cells plays a limited, local role. Segment polarity proteins, such as Engrailed, are expressed in a portion of each parasegment, stabilizing the parasegmental boundaries and regulating short-range signals that further subdivide each parasegment into different domains, including the ectodermal invaginations that mark segment boundaries. Vertebrates, in contrast, have no early syncytial stage in their development. They add somites, the bilaterally paired mesodermal blocks of the developing trunk, from a posterior pool of cells called the pre-somitic mesoderm. Somitogenesis is sequential, not simultaneous: somites separate one after the other from the anterior end of the pre-somitic mesoderm as the pre-somitic mesoderm elongates. Somitogenesis relies heavily on signaling between cells, apparently using a clock and wavefront mechanism (Figure 1B). The clock is provided by cyclic changes in gene expression in the pre-somitic mesoderm that mirror the timing of somite formation. There are likely to be several linked clocks driven by cyclic expression of regulators of the Notch family of membrane-bound signals and receptors, of Hes-family transcription factors and of Wnt and FGF signals. The wavefront is provided by gradients of signaling molecules along the pre-somitic mesoderm. When a cell in the elongating pre-somitic mesoderm moves into a zone (the wavefront) where posterior signals (FGFs and Wnts) are low and an anterior signal (retinoic acid) is high the cell responds to the cycling clock. Cells respond to different phases of the clock by taking on anterior or posterior somite fates and separating from the pre-somitic mesoderm. Vertebrates have homologs of some of the Drosophila segmentation genes, but most do not play a role in their segmentation. One exception is that the Drosophila pair-rule transcription factor Hairy is in the same protein family as the Hes transcription factors of the vertebrate clock. Conversely, there is no evidence of clock-like cyclic changes in gene expression in Drosophila or of wavefront gradients of extracellular signals along the A-P axis; indeed, what would these be doing in a syncytium? Looking at these two groups, we would feel quite comfortable with the traditional view that segmentation was invented separately in the ancestors of chordates and arthropods as the result of convergent evolution. However, vertebrates and Drosophila provide only a small sample of segmented bilateria (red taxa in Figure 2) and the mechanisms they use may be only distantly related to the mechanisms evolved by their segmented ancestors. Therefore researchers have been examining segmentation in other arthropods and chordates, and in the third major segmented phylum, the segmented worms of the Annelida. One conclusion of such studies is that the mechanisms underlying segmentation may vary considerably, even in species that have evolved from a common segmented ancestor. For example, two prominent features of Drosophila segmentation, simultaneous formation of segments and segmental gene expression in syncytial nuclei, are certainly not the rule in arthropods or even insects. While many arthropods also have an early syncytial stage, some or all of their segments (and segmental gene expression) appear after cells are formed. These segments arise in an anterior-to-posterior sequence from a pool of posterior cells. Some malacostracan crustaceans have an odd variation on sequential segmentation using posterior stem cells termed teloblasts (Figure 1C). Each posterior teloblast produces a bandlet of blast cells, and each blast cell in that bandlet divides in a stereotyped fashion to contribute to exactly a segment's worth of tissue. In a sense, the production of a blast cell is the earliest sign of segmentation in these animals. Teloblastic segmentation has not been reported in other arthropods and so is likely an evolutionary novelty. Teloblasts may have evolved first as stem cells for producing a posterior pool of unspecified cells in sequentially segmenting species. Only later did cell-lineage based mechanisms link each blast cell to a single segment, perhaps as a way of reinforcing other mechanisms of segmentation. In many sequentially segmenting arthropods the small number of cells in the posterior pool, together with the results from some experimental studies, make it unlikely that these cells already have segmental identities inherited from the syncytium. Homologs of the transcription factors that operate in the syncytial Drosophila embryo might provide some rough A-P positioning in these other arthropods, but not the fine detail needed to break a pool of cells into segments. Some crustaceans never have any syncytial stage at all. Thus, homologs of the gap transcription factors, which work in the syncytium in Drosophila, play little role in segment formation in some other arthropods. The expression of pair-rule and segment polarity homologs is better conserved throughout the arthropods, but in many cases this expression is likely to be patterned by signaling between cells. Researchers have therefore started to look in arthropods for something resembling the cell signaling of the vertebrate pre-somitic mesoderm, and especially for the characteristically cyclic expression of either Hes-Hairy transcription factors or regulators of the Notch pathway. The sampling of taxa and genes is still quite sparse, and some caution is warranted since the vertebrate clock can in some cases still cycle without cyclic Notch or Hes. That said, there is no evidence for a Hes- or Notch-based clock or even patterned Notch signaling during insect segmentation. However, a spider (a chelicerate) and a centipede (a myriapod) have been found to exhibit striped expression of Notch regulators during the sequential formation of segments, and the expression patterns appear to be cycling rather than static. Loss of Notch signaling blocks spider segmentation. Couple that with recent suggestions that some arthropods have a wavefront-like posterior-to-anterior gradient of Wnt signaling, and the possibility exists that at least some arthropods use a clock and wavefront mechanism. And yet, using the same criteria, it is not clear that the clock and wavefront mechanism is shared by all of the chordates, the phylum of segmented animals that includes the vertebrates. Like vertebrates, the cephalochordate Amphioxus forms mesodermal somites in an anterior to posterior sequence (Figure 1D). But these do not arise from a large pool of mesenchymal pre-somitic mesoderm cells. Instead, anterior, early-arising segments are formed by periodic 'enterocoelic' outpocketing from the wall of the archenteron, the embryonic invagination that will also form the gut. Posterior, later-arising 'schizocoelic' segments pinch off periodically from a small group of tail bud cells at the posterior of the archenteron wall. While some of the striped gene expression patterns observed in forming vertebrate somites are also seen in Amphioxus, so far there are no signs of the cyclic Hes or Notch regulator expression. Oddly, Amphioxus expresses an Engrailed segment polarity homolog in the posterior of every enterocoelic somite, something lacking from vertebrates. What of our third large group of segmented animals, the worms of the annelids? The best-studied examples are hirudineans (leeches) and oligochaeates (for example, Tubifex), related taxa collectively termed the clitellates. Detailed cell-lineage tracing has shown that the segmented ectodermal and mesodermal tissues of clitellates are derived from blast cell bandlets produced by ten large posterior teloblasts (Figure 1C). There is a strict relationship between the formation of an individual blast cell and the eventual formation of a morphologically recognizable segment: depending on the bandlet either one or two blast cells forms exactly a segment's worth of tissue. These segmental fates appear largely invulnerable to cell ablation and transplantation, suggesting that the segmental patterns of cell lineage depend on inherited determinants rather than signaling between cells. But the teloblastic segmentation of clitellates may be an evolutionary novelty, as there are as yet no clear examples of teloblastic segmentation in the polychaete annelids from which the clitellates are thought to have evolved. While mesodermal teloblasts have been described in polychaetes, the existence of ectodermal teloblasts is still open to question. There is no evidence that a single blast cell constitutes a segmental unit. Instead, in some polychaetes segments appear to be formed from pools of cells. If so, might this again use something like the clock and wavefront of the vertebrate pre-somitic mesoderm? There is as yet no sign of the cyclic Hes and Notch gene expression of the vertebrate segmentation clock. Striped expression of Engrailed segment polarity homologs was observed in the forming segments of one polychaete and in late stages of leech segmentation, but not in several other annelids. One other form of variability bears mentioning: the different segmentation mechanisms used in different parts of the same embryo. Differences between anterior and posterior regions are quite common. In some insects, anterior segments appear in a syncytium and posterior segments appear from a pool of cells. In some malacostracans, anterior segments are formed by recruitment from a pool of cells and posterior segments are formed by teloblasts. In some polychaete annelids, the anterior larval segments appear nearly simultaneously, well before the sequential formation of posterior segments. Anterior and posterior somites in Amphioxus have different sources and can express different genes. The head mesoderm of vertebrates has segment-like features but is not produced from the pre-somitic mesoderm and shows no evidence of the clock and wavefront mechanism. Rhombomeres, the repeating subunits of the vertebrate hindbrain, appear nowhere else in the body. Even stranger are the cases in which a single embryo produces more than one set of segments, such as the independent dorsal and ventral segmentation of some arthropods. The mechanisms that result in a similar-appearing segment can thus show a surprising degree of flexibility, both within a single organism and in a group of organisms that have evolved from a common segmented ancestor. Given that flexibility, can we say anything at this point about the evolution of segmentation within the bilateria as a whole? Some older phylogenies suggested that arthropods evolved from an annelid-like ancestor — the occurrence of segmentation in both phyla was actually a major reason for linking these phyla. Almost all recent molecularly-based phylogenies, however, have concluded that annelids and arthropods are not closely related. Some go further, strongly supporting the view that annelids and arthropods are members of two very distantly related groups of bilateria (Figure 2). The Lophotrochazoa includes phyla with spiral cleavage and trochophore-like larvae, including annelids, molluscs (such as snails, shellfish and squid) and platyhelminthes (flatworms), along with phyla with a lophophore feeding structure. The Ecdysozoa all shed their cuticles during growth (ecdysis means molting) and include arthropods, onychophorans (velvet worms) and nematodes (roundworms). These two groups are linked in a larger group called the Protostomia. The third large group of bilaterian phyla is the Deuterostomia, which includes chordates, hemichordates (acorn worms) and echinoderms (such as starfish and sea urchins, which are radially symmetric as adults but bilaterally symmetric as larvae). The conservative view is that segmentation arose at least three times, in the ancestors of the chordates, the annelids and the panarthropoda (the common ancestor of the arthropods, its sister phylum the onychophora and perhaps the tardigrades or water bears). However, some have concluded from the molecular similarities that segmentation existed in the urbilaterian ancestor of the protostomes and deuterostomes. Engrailed stripes are found in panarthropoda, cephalochordates and some annelids; Notch cycling is found in vertebrates and likely some arthropods; Hairy-Hes stripes are found in arthropods, chordates and portions of a polychaete segment; Lb-Lbx stripes are found in one annelid and in late Drosophila segments. A segmented urbilaterian would require some rethinking, at least for those of us reared on the phylogenies popular in most English language textbooks. These posit that the urbilaterian was a simple animal resembling one of the existing taxa. This is still a popular hypothesis, and to many acoel flatworms seem the best simple candidate. These were once considered part of the platyhelminthes, but recent work suggests they are a separate phylum, possibly a basal sister-group to all other bilaterians. Acoel flatworms lack not only segments but also the coelomic or pseudocoelomic body cavities typical of most other bilateria. In contrast, an ursegmentarian would have to have been a fairly complex bilaterian, likely with a fully segmented coelomic body cavity. And most bilaterian phyla, including the acoels, would have to have lost segmentation. Is there evidence of this? There are other examples of A-P repetition that could possibly represent vestiges of ancestral segmentation, although there is as yet no evidence that these arise by similar mechanisms. The formation of a string of proglottids, the reproductive body units in tapeworms (platyhelminthes), is by most considered an evolutionary novelty. But Kinoryncha, a phylum apparently related to nematodes, have repeating subdivisions of the epidermis, nervous system and mesoderm, and are thus segmented by anyone's definition. And there are many taxa with more superficial repetitions. Some molluscs, such as chitons, have repeats in shell structure, musculature and other structures. Bristle rows, epidermal annulation or cuticular joints repeat along the A-P axis in several taxa. Even the nematode Caenorhabditis elegans has A-P repeats in some of its neuronal precursors. And there are the tripartite coelomic subdivisions of hemichordates, chaetognaths (arrow worms) and echinoderm larvae. Recent molecularly-based phylogenies also suggest that segmentation can be lost. Echiuran worms now appear to be descendents of polychaete annelids, and the same is likely to be true of the Sipuncula. They lack mesodermal segmentation, although they retain some segmentation in their nervous systems. The unsegmented urochordates (tunicates) appear to be more closely related to vertebrates than the segmented cephalochordates, strongly suggesting that urochordates evolved from a segmented chordate. The flexibility observed within phyla also in a sense lowers the bar, as any mechanism ancestral to all the bilateria might have been lost or radically altered in some lineages. If the common ancestor had a Notch-based cyclic clock, the clock was retained in vertebrates and a few arthropods but perhaps lost in annelids and cephalochordates. If the ancestor also used segmentation transcription factors like Hairy-Hes and Engrailed, this was retained in arthropods and cephalochordates, but use of Engrailed was lost from vertebrates and use of both likely lost from most (but not all) annelids. But flexibility cuts both ways, making it more likely that similar-appearing mechanisms arose through convergent evolution. Perhaps Notch signaling and particular transcription factors are especially well-suited to the task of segmentation and so have been independently co-opted into this process multiple times. Notch signaling is certainly widely used to subdivide developing tissues, and Hairy-Hes transcription factors are common Notch targets. The repeated recruitment of a transcription factor like Engrailed into segmentation may seem a bit more arbitrary. But Engrailed has been posited to have ancestral roles regulating cell specification (in the nervous system), cellular affinities and Hedgehog and Wnt signaling, all useful things during segmentation. Or one could go further and posit that our urbilaterian already used these molecules to specify tissues in a non-repetitive, non-segmented fashion — for example, for specification along the A-P axis, something undoubtedly present in the urbilatian. These molecules were then co-opted several times for different developmental functions, including, in a few cases, the repeated subdivisions of segmentation. And finally there are those many mechanisms we do not understand. Our analyses to date have been based on similarities to vertebrates and flies, and so finding similarities is something of a self-fulfilling prophecy. More detailed, mechanistic information from more taxa could certainly help in this debate, especially information based on unbiased screens instead of candidate genes. If mechanisms vary greatly, however, it will still come down to an argument about plausibility, and one scientist's homolog is often another's convergence. How do we analyze a character with so much mechanistic flexibility? Can some level of mechanistic similarity ever rule out convergence, or is that wishful thinking? The identity of ancestral organisms has been the subject of intense debate since the 1800s and it is interesting to think about what kind of data it would take to settle that debate. On the other hand, what fun would that be?