For a nicely-formatted and better referenced PDF of the entire series, click here!
Annelids have always played a big part in our views of animal phylogeny. This post is NOT a summary of animal phylogeny explicitly, but rather about how differing views of animal phylogeny affect the annelids, and how the annelids affect those views.
For example, let’s look at their nervous system. As in the vertebrates and most invertebrate phyla (including the arthropods, molluscs and echinoderms), annelids have a central nervous system (CNS). The picture above shows expression patterns of several neurogenic proteins in (left to right) fruit fly, annelid and mouse embryos. As you can see, there is a lot of conservation, hinting that their last common ancestor – in this case, the first bilaterian – also had a CNS and shared this pattern. This is one of the goals of phylogeny: to allow us to reconstruct the last common ancestor. But as must always be stressed, phylogenies are hypotheses, not fact, and so unless they are robust and proven, the reconstruction of the last ancestor will be false, or at least very tentative.
So let’s look the gross topology of the metazoan tree of life. Except for a few exceptions (from the molecular phylogeneticists), sponges are the most basal animals. Then come the cnidarians. The bilaterians (which I’ll be referring to as animals from now) are the rest of the animals, characterised by bilateral symmetry.
First, some historical results (note that I will only highlight some of these, as several tens of papers on this are published yearly). I have purposely decided to do this in text form, so you can practice your cladogram drawing skills ;) In the very first relatively modern classification of animals, by Linnaeus, the invertebrates were split into seven classes: molluscs, crustaceans, arachnids, insects, worms, polyps and ‘Radières’ (radials; echinoderms and sipunculids). Annelids were in the worms (= Vers or Vermes), which were split into Vers extérieurs (external worms) and Vers intestins (intestinal worms). The former was split into two groups, one of which was the Annelida – he named them so after the latin annulus (little ring), in reference to their segments (linguistically, it should be Annulata, but Annelida just proved more popular).
One historically popular scheme (as in uncontroversially found in all relevant textbooks up to the 1990s; first set up by Cuvier in 1817, and popularised by Haekel in 1866 as proof of his biogenetic law, now disregarded; he also viewed the arthropods as emerging from an annelidan ancestor, a view also popular even until the 1990s, but now pretty silly; an opposite view, that annelids emerged from arthropod ancestors, has also been proposed once!) placed annelids as a sister group to the panarthropods (Panarthropoda = Arthropoda + Onychophora (velvet worms) + Tardigrada) in a clade called the Articulata, and the pogonophores as sister to the Articulata, and that the Mollusca are sister to Articulata + Pogonophora. While the Articulata are united by their segmentation, there are no convincing synapomorphies for the Annelida in this clade – this has also led to some calling for the Annelida to be dismantled, as some analyses have shown that the clitellates and polychaetes are no more related to each other than they are to the arthropods. Similarly, the polychaetes have been considered to be paraphyletic, as some groups within them seem to be more related to the pogonophores.
Some others have stressed that annelids and molluscs are sister to each other, with the arthropods being the sister of Annelida + Mollusca. Some have supported a eutrochozoan taxon consisting of Annelida + Mollusca + Nemertea and others, critically leaving out the arthropods – this would make the coelomic cavities in those groups homologous. Or how about the Neotrochozoa hypothesis, where annelids, molluscs and Platyhelminthes are placed together – or a contradictory version of it, where annelids and platyhelminths are sister groups, but exclude the molluscs?
Bütschli (1876) recognised that the annelids and arthropods are separate and, contrary to the fashion, separated them from each other – he viewed the arthropods as closer to the nematodes and explained similarities between arthropods and annelids (e.g. the nervous system’s structure) as convergent evolution due to the convergent development of segmentation.
While Ernst Haeckel argued for a separate origin of segmentation in Articulata and vertebrates, Anton Dohrn argued for an annelid ancestry of the vertebrates – for him, vertebrates were simply annelids turned upside down (in reference to the dorsal nerve cord of vertebrates). While this may sound very silly, it was first proposed by Saint-Hilaire in 1822 and is still supported, especially when looking at neurodevelopment: the genes and processes controlling the development of the neural system of annelids, arthropods and chordates are the same – right down to the fate maps! – except that chordates are inverted on their dorsoventral axis.
Another scheme had the Articulata + Kinorhyncha as a monophyletic clade, with the nematodes being their sister group. Alternatively, a clade of Pogonophora + Annelida + Mollusca + Sipuncula + Echiura was also hypothesised, with the arthropods being their sister group. To really prove that there are no holy dogmas in phylogenetics, consider that some have even proposed the polyphyly of the arthropods, by postulating that the Clitellata and Uniramia (Uniramia = Myriapoda (centipedes and millipedes) + Hexapoda (insects)) are allies, with the Hirudinea (leeches) as their sister group.
I took many trees at random, both historical and recent, and placed them lazily in the composite image above, and did dome helpful colouring: red is deuterostome, blue is Lophotrochozoa, green is Ecdysozoa, pink is Articulata, teal is annelid if none of the above apply. The two ancient ones are Haeckel 1866 (the big bushy tree) and Bütschli 1867 (the simplistic one) – note the differences between them (phylogenetic, not artistic)!
The main debate going on is, as usual, between staunch morphologists (a group which I fall into, so keep that in mind) and the molecular phylogeneticists. The latter group’s work (we will not go into their reliability and depth) completely contradicts the traditional, morphological view of bilaterian phylogeny. At a basic level, molecular phylogenies have the acoelomorphs as the basalmost bilaterians and divide the bilaterians into the deuterostomes and the protostomes. The latter is divided into Ecdysozoa and Lophotrochozoa. The Ecdysozoa includes groups like the arthropods, priapulids, loriciferans, kinorhynchs and nematodes and are characterised by the ability to moult. The Lophotrochozoa includes groups like the annelids, molluscs, sipunculids, echiurans, nemerteans, phoronids, brachiopods, bryozoans, platyhelminths and rotifers, and are characterised by having a larva that is either of the trochophore-type or which has a lophophore (a distinct feeding structure), or even those that have neither of those (welcome to phylogenetics.), such as rotifers, acanthocephalans, cycliophorans and gastrotrichs; annelids have a trochophore larva.
From a morphological point of view, annelids are most closely related to the sipunculids, forming a clade called the Pulvinifera. The fact that arthropods have been placed in the Pulvinifera is only due to the Articulata hypothesis assumed as being true. I won’t discuss how the arthropods fit in here, besides saying that they don’t. There are two autapomorphies for the Pulvinifera: a cuticle made of collagen fibers and the coelom building a hydrostatic skeleton.
Now it’s time to look at the Articulata. As I said, it was first proposed by Cuvier (1812) (first presented by him in a lecture in the 1790s) and accepted until the Ecdysozoa hypothesis reared its head in 1997. Autapomorphies for an Articulata clade – assuming it isn’t polyphyletic – would be segmentation through teloblasty (i.e. the posterior growth zone with segments added on; note that insects don’t have this!). We will look at segmentation in more detail some paragraphs later. But besides those characters, everything else is very highly modified or reduced in the panarthropods – this has always been the trouble with the Articulata hypothesis.
Here we have to bring up the Spiralia, a clade characterised by spiral cleavage and consisting of the annelids, molluscs, sipunculids, echiurans, nemerteans, myzostomids, ectoprocts, some flatworms and maybe the gnathostomulids. Comparing this with the Lophotrochozoa clade, we see that they are nearly identical. No ecdysozoans undergo spiral cleavage. Lophophorates (bryozoans, phoronids) – traditionally considered deuterostomes – don’t have spiral cleavage, neither do gastrostrichs, rotifers and parasitic flatworms, and notably, the cephalopod molluscs. It is assumed to have been lost independently in all these clades.
The Spiralia appear to be monophyletic in many analyses, with major problems being the resolution of the base of the Spiralia, i.e. the relationships between the Plathelminthes, Gnathifera and Nemertini. Involved in this debate is the possible non-monophyly of the Plathelminthes, since the Acoelomorpha do not undergo spiral cleavage. One definite monophyletic assemblage placed within the Spiralia is the Trochozoa, i.e. those animals with a trochophoran larva, including molluscs and annelids.
If the Lophotrochozoa is a valid monophylum, then the presence of spiral cleavage is of utmost importance in defining relationships between them – all we know now is that spiral cleavage is ancestral to the Trochozoa, not to all the Lophotrochozoa, and this is highly-problematic for the entire Lophotrochozoa hypothesis: such foundational developmental patterns are not so easily lost. Additionally, we have to think of the Polyzoa, another molecular clade that unites the ectoprocts, bryozoans and cycliophorans. Since Polyzoa includes the bryozoans, it must be determined if it’s the sister group of the Trochozoa, thus forming the Lophotrochozoa. If they are sister to the Platyzoa (another molecular clade, comprising of various Platyhelminths and, variably, the myzostomids, gastrotrichs, gnathiferans and acoels), then Spiralia is exactly the same as Lophotrochozoa, and Lophotrochozoa can effectively be discarded (older names take preference).
By contrast, the traditional (some would say old) bilaterian phylogeny has platyhelminths as the basal taxon, with a more advanced coelomate group (e.g. nematodes) branching off later, with a monophyletic coelomate group comprising arthropods, annelids, lophophorates and deuterostomes as the final clade – hence why you can’t reconstruct the urbilaterian by looking at what characters are shared by chordates and insects.
Just by looking at annelids and arthropods (that would be my disciplinary bias shining through here), we notice some glaring problems – notice I don’t call them errors, since the morphologists could be wrong, or we could all be wrong and have not yet stumbled on the proper answer! In the traditional view, the annelids are allied with the panarthropods as the Articulata, sharing the characteristic of segmentation. The implication of the molecular hypothesis is that the last common ancestor of the arthropods and the annelids is the last common protostomian, a group with over 10 phyla. What this means is that segmentation must have been maintained in the annelids and arthropods, but lost in all the others – hardly as parsimonious as saying that segmentation is ancestral to just those taxa that have it. We can also look at the trochophore larva: either it was gained in the Trochozoa, or it was independently lost by the arthropods – a loss that would be difficult to explain.
But this view is rather limited. Traditionally, discussions of segmentation have been as above: concentrated only on arthropods and annelids, in most cases bringing up the chordates as well. The problem is that ‘segmentation’ is not really defined, and that we don’t really know how the transition from non-segmented to segmented looks like. Molluscs are said to show primitive segmentation when compared to annelids; onychophorans are said to have lost features of segmentation when compared to the euarthropods. Let’s put some myths to rest. Annelids were not originally a single segment that just serially repeated itself in a chain. Instead, we think of the evolution of segmentation in annelids as more of a parcellation process: their organs simply organised themselves into serially-repeating structures, according to what selective pressures were present at the time. In the face of this, we have to bluntly say that the use of ‘segmentation’ is not relevant anymore: it’s typological and vague. Do we refer to segmented organisms (no, because no animal is divided so strictly into these modules!) or to segmented organs (organs may be serially repeated, but it’s rare for the organism to be that way too!)? It has only achieved such prominence as a shorthand – it’s practical to refer to arthropods, annelids and chordates as segmented organisms; but as I said, this ignores the wealth of other taxa where segmentation may be present but only in a partial way (cf. molluscs, onychophorans). In phylogenetics, we have always asked whether an organism is segmented, and is this segmentation the same as in the other organism. This is wrong. The question really is “what parts are segmented, and is this segmentation in these parts homologous to that in other organisms?”
So let’s go through with this. In annelids and arthropods, the body wall, the coelom, the nervous system, the metanephridia and the musculature is segmented; we can add an appendage pair to the list as well. But there are many variations on this theme. Echiurans are now largely considered to be annelids, but have no segmentation in the body wall. Arthropods don’t have a segmented coelom, except during somite formation in early embryogenesis. Onychophorans have no segmented musculature, except that associated with the limbs (which is expected). Tardigrades have no coelom and their body is only partially segmented. In the annelids themselves – the apparently prototypical segmented animals – leeches don’t have a large coelom. The segmentation of the septa in larger polychaetes is nowhere to be seen, having been replaced by musculature. Going further out, consider the molluscs: poly- and monoplacophorans have segmented musculature and segmented sclerites on their body, but no segmented coelom. Other groups with such partial segmentation include nemerteans, flatworms and kinorhynchs. The basic point is, presence of segmentation should refer to individual systems, not whole organisms!
However, it is also true that body wall segmentation occurs – this is what we colloquially refer to as body segments when looking at an earthworm or an insect. And here, it’s worth noting that developmental biology supports a common origin of sergmentation in arthropods and annelids, as the same master regulatory signalling pathway that controls segmentation in insects, Hedgehog, also does the same in annelids.
And this brings up another issue: the two paragraphs up there came from my own training as a morphologist. A molecular cell biologist will have a completely different view of segmentation, as will anyone else with a different scientific background. One model of segmentation refers to the state of the cells; another simply refers to the “developmental mechanism that subdivides a tissue into repeating structures” (Rauskolb, 2001); yet another says that it’s the “formation of a periodic pattern of paralogous blocks of cells” (Kroiher et al., 2000). Let’s make one thing clear: we need precise definitions. Mere repetition does not make a segment. Nematodes and flatworms have repetition in their nervous systems; monoplacophorans and kinorhynchs in their musculature; polyplacophorans in their shell; nemertines in their gonads; monoplacophorans in their nephridia. We cannot say these are all examples of segmentation without completely ruining the usefulness of the term. This is why it’s safest not to stray from the traditional definition, as we know it from arthropods and annelids (second sentence, two paragraphs ago). But then we run into logical problems: by applying that definition, we are a priori assuming homologous segmentation in the Annelida and Arthropoda and just end up running in circles; and when we detach ourselves from that logical fallacy, we get into a Catch-22-like situation where individual variations in the segmentation pattern (cf. 2 paragraphs ago) can be either over- or underinterpreted (cf. also the picture below).
We could also go into the earliest fossil record of animals and ask whether segmentation was present in the last common ancestor of the Bilateria, but then I will ramble even more than now. Suffice it to say that many fossils from the Early Cambrian are unsegmented and are interpreted variably as stem annelids, molluscs or even lophotrochozoans. The idea of a segmented urbilaterian is cute, but doesn’t really hold up to scrutiny, especially not under a phylogenetic framework.
Other notable differences between the traditional and ‘new’ animal phylogeny include the paraphyly of the protostomes under morphological systematics, as well as the placing of the lophophorates as deuterostomes and the non-monophyly of coelomates (again, implying multiple losses or rampant convergence). The arthropods, in the new system, are most closely related to various nemathelminth groups, which are grouped as Cycloneuralia.
Just to show how creative phylogeneticists can get, consider Nielsen in 2003 “proposing a solution to the Articulata-Ecdysozoa controversy” (Zool. Scr. 32, 475-482): All ecdysozoans are articulates, with a single ancestral loss of segmentation in the ancestor of the non-arthropodan Ecdysozoa.
The systems give us drastically different views of how the original bilaterian looked like. Some examples include: in the molecular system, any similarities we find between insects and humans will be due to common ancestry – more than half the protostome clades lost segmentation, as the last common ancestor between arthropods and annelids was also the first protostomian and thus must have also been segmented. A segmented urbilaterian is hard to imagine though, as segmentation is very different, both from a molecular and embryological level, in the articulates (I say that out of convenience!) and in chordates. Flatworms, long thought to be representative of a primitive bilaterian, are instead derived trochozoans who have lost their anus, coelom and circulatory system, as well as their gut (!). The animal that looks the most like the original bilaterian is the annelid.
There are even differences that affect the annelid tree directly. This is the case with the echiurans and the pogonophores. The latter have long been considered to be a separate phylum, but it has always been obvious that they are very closely related to the annelids (due to their larvae being similar), and during the resurgence of morphological analyses in the 1960s and 1970s, many similarities were discovered, so their downgrading to the family Siboglinidae is not a big deal. But saying the echiurans are derived from an annelid is quite a stretch, simply because echiurans have no form of segmentation – the only thing close to it is that their nerve cord develops serially in the embryo. That an annelid would lose its segmentation, which is one of the defining characters of annelids, is hard to swallow; that said, studies on neurogenesis do also somewhat support the relationship, and it is common to see taxa that undergoing a simplification of their body structure when changing to a sessile lifestyle. If this hypothesis turns out to be accurate, then that would shake up our view of segmentation, specifically its plasticity and variability, which in turn will help in clearing up the true nature of segmentation (in annelids at least). The sipunculids have also sometimes been included within the annelids. In these analyses, the echiurans, pogonophores (and myxostomids) are nested within the polychaetes – and this agrees with the modern view of polychaetes being paraphyletic. Other results from molecular analyses include the clitellates as polychaetes (also making the polychaetes paraphyletic). Remarkably, there is at least one published tree in which a brachiopod was found to be an annelid – clearly the result of a thorough analysis.
Having said all that, there are cases in deep animal phylogeny where molecular and morphological phylogenies agree, such as the sister group relationship of the Kinorhyncha and Priapulida, and of the Nematoda with the Nematomorpha, as well as the brachiopods with the phoronids. And since some may have read between the lines and noticed a huge undercurrent of skepticism of molecular phylogenetics and *gasp* dogmatism in my morphology-zealotism, I will just state that molecular phylogeneticists are just as arrogant in their positions – except that we morphologists can actually conclusively prove our hypotheses without having to rely on faulty methodologies. </rimshot>
But the case remains that the Articulata-Ecdysozoa debate is nowhere near as settled as some would have you think (many simply disregard the Articulata as only of historical value). I will summarise the issues here.
Except for the gastrotrichs, all Nemathelminth (= Cycloneuralia) taxa moult their cuticle, as do the arthropods. This is the main autapomorphy shared in the Ecdysozoa clade (Cycloneuralia + Arthropoda), which always emerges from molecular phylogenetic analyses – which almost always suffer from limited taxon sampling and unsuitable gene sequences (e.g. 18s rDNA), although recent phylogenomic (complete genome) analyses take care of the latter problem and still recover Ecdysozoa. Detailed ultrastructural analyses have also demonstrated very strong similarities in the ultrastructure of the chitinous cuticles from these taxa, and the hormonal control of moulting (ecdysone) is also very similar – but nothing worthy enough of being true homologies.
The autapomorphies shared by the Articulata clade (Arthropoda + Annelida) are: segmentation; teloblastic formation of segments; longitudinal muscle strands; segments with parapodia; ladder-like nervous system. These are so complex that the possibility of them having evolved convergently and so similarly (they are near-identical) is highly unlikely. This point may be moot, however, as there is evidence that segmentation itself is not a single character, but a mix of many, meaning its gradual loss or gain is more ‘simple’ than previously thought. For example, a ladder-like nervous system is automatically part of segmentation – but it must be said that neurogenesis (the development of the nervous system) in annelids is eerily similar to that in arthropods. Molecular developmental studies are increasingly showing that teloblasty is actually convergent, not homologous. So the jury is still out.
The most significant implication is that segmentation of arthropods and annelids is either convergent or that it is an ancient feature from the protostoman or even bilaterian stem – although the latter is highly unlikely (in my opinion: too many losses, therefore unparsimonious; also, chordate segmentation is way too different from annelidan and arthropodan segmentation in both its final appearance and its ontogeny to be derived from the same ancestor).
What is definitely a hinderance to progress, though, is the uncritical, almost automatic, and way too excited adoption of the Ecdysozoa (from textbooks to science museums, such as the American Natural History Museum!), and the feeble unquestioned rejection of the Articulata by most of the scientific community. This is brought about by a fallacious mindset wherein gene sequences are seen as they key to the phylogenetics – that genes are the “written text of evolution” and by comparing the texts, we can decipher the history of life – making morphological analyses not needed (if not completely useless, as some have suggested). For a lack of a more diplomatic way of refuting this, I will simply state that a hemorrhoid is more intellectual than those suggestors.
Arendt, D., Denes, A. S., Jékely, G. & Tessmar-Raible, K. 2008. The evolution of nervous system centralization. Philosophical Transactions of the Royal Society B 363, 1523-1528.
Scholtz, G. 2002. The Articulata hypothesis – or what is a segment? Organisms Diversity & Evolution 2, 197-215.
Bütschli, O. 1876. Untersuchungen über freilebende Nematoden und die Gattung Chaetonotus. Zeitschrift für wissenschaftliche Zoologie 26, 363-413.
Cuvier, G. 1812. Sur un nouveau rapprochement à établir entre les classes qui composant le Règne Animal. Annales du Musée d’Histoire Naturelle 19, 73-84.
Kroiher, M., Siefker, B. & Berking, S. 2000. Induction of segmentation in polyps of Aurelia aurita (Scyphozoa, Cnidaria) into medusae and formation of mirror image medusa anlagen. International Journal of Developmental Biology 44, 485-490.
Rauskolb, C. 2001. The establishment of segmentation in the Drosophila leg. Development 128, 4511-4521.