Why molecular biology needs palaeontology

Until very recently the fundamental problem of metazoan evolution, that of the origins and relationships of the thirty to forty phyla that are generally recognized, remained effectively intractable. This was despite more than 200 years of research and controversy, and the production of a literature that now almost defies synthesis (e.g. Hyman, 1940-1961; Willmer, 1990). This did not prevent, of course, a plethora of proposals and hypotheses. Some tackled supposedly fundamental problems: is the coelom a primitive structure (the archicoelomate hypothesis, much favoured by the German school; see Willmer, 1990) or can larval anatomy, e.g. the trochophore, provide unique clues to the ancestry and relationships of otherwise disparate phyla (e.g. Nielsen and Nørrevang, 1985)? Other proposals concerning metazoan relationships were more specific but ranged from the frankly controversial, such as Løvtrup’s (1977) analysis that proposed a relationship between arthropods, specifically arachnids, and chordates, to the apparent consensus that places brachiopods and the other lophophorates firmly within the deuterostomes. As we will see below, however, this latter idea may be less secure than popularly imagined.

For each and every proposal concerning metazoan relationships there was almost invariably a counter-suggestion. The existing literature is like the Sargasso Sea of mythology, dotted with hulks of varying decrepitude, each manned by a crazed crew or more likely ghosts. This morass persisted because whatever phylogenetic scheme was proposed had to rely on a handful of features being chosen as axiomatically suitable for identification of monophyly, e.g., segmentation, coelom, setae, blood-pigments, and the tacit admission that other characters accordingly had to be homoplasic. For various reasons, features such as segmentation and type of body cavity came to occupy effectively inviolate positions in terms of phylogenetic reliability, although in their more candid moments most zoologists will admit that there is little a priori evidence that metameric segmentation, for example, could not have evolved independently several times. The net result is that any scheme of metazoan phylogeny will inevitably involve characters that to one set of workers are crystal-clear guides to relationships, but to another group are examples of rampant convergence.

The irruption of molecular biology and specifically sequence analyses, however, has revitalized this hitherto moribund area of biology (Conway Morris, 1993a). It promises a release from the deadlock of mutually incompatible phylogenetic hypotheses by offering an independent source of evidence, most obviously to date in the reading of sequences within molecules of ribosomal RNA where there is no apparent connection between the molecule and either anatomical expression or environmental preference. These molecular methods, of course, are not foolproof. Different parts of a sequence evolve at different rates, discrepancies exist if comparisons are drawn between the 3’ and 5’ ends of the molecule (e.g. Patterson, 1989), and some taxa (e.g. dipteran insects (see e.g. Carmean et al., 1992), sipunculan worms) appear to evolve very rapidly. Further problems arise where the morphological distinctiveness of a group is echoed in its molecular sequence and so it remains phylogenetically isolated, as has been suggested for the nematodes by Wolstenholme et al. (1987; but see Brandl et al., 1992). Other disagreements arise about the most suitable molecule. For example, 5S rRNA is now generally regarded as too small to be appropriate in this context (Halanych, 1991), and differences also arise concerning methods of tree-building and their most appropriate interpretation.

Nevertheless, some degree of consistency seems to be emerging. As an increasing number of genes and their products are scrutinized the long-standing problems of metazoan phylogeny, such as the placement of the brachiopods, or the monophyly of arthropods, will be resolved. Should then wholeorganism zoologists, not to mention palaeontologists, wait patiently for the laboratories of molecular biology to issue a series of phylogenetic dicta and then hasten to provide unquestioning assent? No, and for four reasons:

• (1) Even if it is shown unequivocally, in terms of molecular biology, that two phyla are closely related, say molluscs and annelids (e.g. Ghiselin, 1988), this will tell us nothing about how the anatomical, ecological and behavioural transitions that led to these now distinct phyla were achieved. Nor will this information explicate the evolutionary processes that were involved in the origin of what we now regard as body plans. For example, if as indeed now seems likely molluscs and annelids are closely related, then how might we explain features held to be fundamental in one or both phyla such as metamery, the coelom, and chaetae? Below I will argue that fossil data are an essential component in helping to answer this question. In many cases new discoveries will provide a crucial key, but earlier material will be subject to continuing reinterpretation (e.g. Ramskold and Hou, 1991).

  In this paper I will concentrate almost entirely on the relationships between molecular biology and palaeontology so far as they concern early metazoan evolution. It would be remiss, however, if attention was not drawn to the flourishing interactions in tetrapod biology (e.g. Coates and Clack, 1990; Eemisse and Kluge, 1993). Ultimately sequence and morphological data will be complimentary, but we need to be reminded of their respective advantages and limitations. The great majority of readers will be aware of these in the context of molecular biology, but perhaps less so with respect to palaeontology. The principal advantage of fossils is that they provide taxa across the fourth dimension, many of which possess character states that otherwise have been entirely lost and so provide crucial bridges in reconstructing evolutionary trees.

• (2) Molecular biologists (e.g. Christen et al., 1991) have been careful to emphasize that during times of rapid divergence, such as appears to typify the protostomes (e.g. Field et al., 1988), the precise order of branching may be very difficult to resolve. Nobody pretends the fossil record is perfect, not least in terms of relative timing of appearance. Nevertheless, a cautious analysis of metazoan evolution during the Vendian and Cambrian suggests that branching orders may yet be discernible. The difficulty, however, is not an over-abundance of characters but recognizing homologous structures at deep levels of metazoan branching and especially anatomical transformations from one state to another. From our perspective, the end-results of these transformations may look very different, but in the Cambrian they may have evolved by a series of rather trivial alterations.

• (3) There is more to life than molecules. If we can generate a sensible framework for metazoan phylogeny then this reopens a whole series of questions on the nature of evolutionary convergence. Whatever scheme of evolution of the phyla is ultimately accepted it inevitably must involve the independent acquisition of major features such as body cavities, respiratory pigments, skeletal hard-parts, osmoregulatory organs, and arrangements of the nervous system that include optical sensors. The forty-odd bodyplans and the myriad of lower taxa represent a massive experiment in biological occupation and the constraints that govern the evolutionary process.

• (4) Finally, old problems might receive new insights. For example, are body plans really as conservative as is usually thought? Indeed, should we abandon the essentialist concept of the body plan and the related taxon of phylum? Why do some phyla have almost invariant body plans, and why in such examples is a phylum sometimes of low diversity (e.g. sipunculans) but in others of high diversity (e.g. nematodes, see Burglin and Ruvkun, 1993, p. 619). Alternatively, why do other phyla demonstrate a much greater plasticity of form?

At first sight the field of metazoan relationships is strewn with over-turned apple carts, and beside each one stands a molecular biologist. Certainly within various groups there have been some major surprises. Avise (1994, Table 8.2), for example, lists a significant number of new insights into the inter-relationships of birds. In terms of the relationships between the phyla, there is, however, less sign of truly radical reorganizations. Thus, molecular biologists agree with many zoologists that the cnidarians are among the most primitive of the metazoans (e.g. Adoutte and Philippe, 1993; Christen et al., 1991; Field et al., 1988; Lake, 1991; Telford and Holland, 1993). Recent evidence also suggests that within the Cnidaria the anthozoans arose first (Bridge et al., 1992; see also Schuchert, 1993) a proposal that may be consistent with new evidence from Ediacaran-like fossils (see below). Shostak (1993) has reiterated the notion that the stinging cells (or cnidae) in cnidarians are symbiont acquisitions from protistan microsporidians, an idea that in principle should be easy to test by comparing molecular sequences of proposed host and cnidae.

Even more primitive, perhaps, are the sponges, although there is also some evidence that this group is polyphyletic (Lafay et al., 1992). Concerning the triploblasts, the platyhelminthes retain a rather primitive position and appear to be monophyletic (e.g. Ruitort et al., 1993; see also Telford and Holland, 1993; Adoutte and Philippe, 1993), and investigations into the inter-relationships within this phylum also continue (e.g. Riutort et al., 1992). Within the so-called higher triploblasts a clear-cut division remains between the deuterostomes (echinoderms, hemichordates, chordates) and the protostomes (including annelids, arthropods, molluscs and sipunculans; e.g. Lake, 1990). Nevertheless, despite this overall consensus there are plenty of adjustments in sight concerning existing hypotheses of inter-relationships, and most will prove controversial.

Within the protostomes it may be that the arthropods are an early branch, which is paraphyletic (e.g. Lake, 1990; see also Adoutte and Philippe, 1993). Inter-relationships within the arthropods are also in some state of flux. Recently, Ballard et al. (1992) have argued that while onychophorans are definitely arthropods, contrary to received opinion they are highly derived, whereas the myriapods are primitive. Some support for this comes from other areas of developmental biology, including Whitington et al.’s (1993) examination of neural development in arthropods. They find evidence for homologous expression in insects and crustaceans (see also Averof and Akam, 1993 for similar conclusions), but conclude that the myriapods are less close. More in accordance with established thinking is molecular evidence for the placement of the parasitic pentastomids in the crustaceans (Abele et al., 1989).

Several other phyla now appear to be close to, if not within, the protostomes. One highly significant achievement is the recognition by Telford and Holland (1993; see also Wada and Satoh, 1994; Wright, 1993) of the protostomous nature of the chaetognaths. Their almost invariant body plan without clear similarities to any other phylum has puzzled systematists for decades, who have had an ill-defined preference for placing them near to the deuterostomes (e.g. Hyman, 1959). Admittedly, this uncertainty has been contested recently by a minority view arguing for a close relationship between chaetognaths and molluscs, specifically the opisthobranch gastropods (Casanova, 1987). The new molecular view does not explicitly support a link with molluscs, but placing chaetognaths in the protostomes has some important ramifications. In particular, the characteristic radial cleavage of chaetognath embryos strongly suggests that this feature per se cannot be employed usefully to distinguish protostomes, in which cell cleavage is often spiral, from the supposedly diagnostic radial cleavage of deuterostomes (Telford and Holland, 1993).

Almost as surprising as the reassignment of the chaetognaths is the recent molecular evidence concerning the nemerteans. This phylum, long imagined as a separate offshoot of the platyhelminthes that paralleled coelomate development, is now also placed firmly in the protostomes (Turbeville et al., 1992), a point that accords with Turbeville’s (1986) earlier discussion of the similarity between the nemertean coelom and that of the polychaete Magelona. The final example of probable recruitment of a phylum to the protostomes concerns the brachiopods, and by implication the related ectoprocts (bryozoans) and phoronids. Prior to the molecular evidence for a protostomous position (Field et al., 1988), the zoological consensus had hovered between a place close to the base of the deuterostomes, or possibly intermediate between this superphylum and the protostomes (see Willmer, 1990). However, the fate of the blastopore, radial cleavage, and the trimerous coelom with the lophophore arising from the second segment, persuaded most investigators that a deuterostomous placement was appropriate, with specific similarities being drawn with the rhabdopleurid hemichordates. If the molecular evidence (Field et al., 1988; Lake, 1990) is correct, then despite the similarity, for example, of feeding mechanisms in the lophophore of hemichordates and ectoprocts (Halanych, 1993) this and other features are convergent.

The next few years should, therefore, see some interesting developments. Some progress may be made in unravelling the orders of branching. For example, amongst arthropods are onychophorans really primitive (cf. Ballard et al., 1992), but amongst lophophorates could phoronids transpire to be derived (e.g. Emig, 1982), rather than basal to this radiation, as is generally thought? For too many phyla we still lack crucial information. Various lines of evidence indicate priapulid worms to be protostomes, and comparisons between haemerythrin pigments argue for a close relationship to brachiopods (Runnegar and Curry, 1992). Will new molecular data support this relationship? Concerning the possibly polyphyletic aschelminthes, little is yet known, apart from the nematodes, which appear to be quite primitive but perhaps fairly close to the arthropods (Brandl et al., 1992). One clue comes from some preliminary evidence that the acanthocephalans may also occupy a relatively primitive position with respect to the main group of protostomes (Telford and Holland, 1993).

What then is the role of palaeontology in unravelling the interrelationships of metazoan phyla? Employment of fossils in this specific area involves a paradox. On the one hand the geologically abrupt appearance of a wide variety of fossils near to the base of the Cambrian is consistent with a major radiation (e.g. Lipps and Signor, 1992). This is most notable in the geologically abrupt appearance of skeletal hard-parts (Fig. 1), but is also evident in a parallel diversification of trace fossils, many of which would have been made by animals with a very low preservation potential (Frey and Seilacher, 1980). On the other hand, it is received wisdom that a phylum maintains its integrity as far back as it can be traced, which in many cases is to the Cambrian, and that the fossil record provides no evidence for recognizing intermediates between phyla (e.g. Bergstrom, 1989, 1990). In the Cambrian, although they are by no means restricted to sediments of this age, there is a plethora of so-called “bizarre”, “enigmatic”, “problematic” or simply “weird” animals. These fossils have attracted quite widespread attention, and have even earned the cognomen of extinct phyla. Such additions to the Cambrian bestiary have not only reinforced the perception of the magnitude of this evolutionary explosion, but have been used to imply the necessity to search for new evolutionary mechanisms to explain this seeming disparity of anatomies (e.g. Gould, 1989). The recent demonstration that celebrated members of this bestiary, such as Hallucigenia and Wiwaxia, are not as peculiar as once thought (Butterfield, 1990; Ramskold and Hou, 1991) has provided satisfaction to those who view zoology as largely an exercise in the correct filling of taxonomic pigeonholes. More significant is the placement of these taxa in schemes of phylogeny (Conway Morris, 1993a). For example, it is argued below that the assertion that Wiwaxia may be regarded as a “true polychaete” (Butterfield, 1990) is a less useful statement than treating it as part of a protostome stem group.

A further complication to the above discussion is a widespread perception that fossils are ultimately irrelevant to this type of phylogenetic discussion, in as much as “instances of fossils overturning theories of relationship based on Recent organisms are very rare, and may be non-existent” (Patterson, 1981, p. 218). This critique has received wide attention and as stated can hardly be faulted: fossils are relatively poor in characters when compared with the richness of data obtainable from living organisms. Nevertheless, the utility of fossils in deciding between phylogenies has been underestimated. In some cases fossil data make a decisive difference, a point made forcibly by a number of palaeontologists (e.g. Doyle and Donoghue, 1987; Gauthier et al., 1988; Novacek, 1992; Eemisse and Kluge, 1993). Note, however, that this new-found enthusiasm for fossil data has more to do with the evolutionary significance of character states of morphology and their placement in a cladistic framework, than their antiquity or the elusive search for ancestors. In this latter context a particularly useful concept is ghost lineages, which correspond to the, as yet, unobserved predictions in a phylogenetic branching (Norrell, 1992). These entities can be tested against the fossil record.

What sources of palaeontological data are most relevant for understanding metazoan inter-relationships? Recent marine communities are predominantly composed of soft-bodied animals or at least ones with such delicate skeletons as to have a minimal preservation potential. If the same situation applied to ancient communities, and there is evidence that it did (Conway Morris, 1986), then it is these stratigraphic horizons where soft-tissues are fossilized that in principle will be the most valuable because they will preserve the most representative cross-section of former life. As discussed below, the Vendian assemblages of Ediacaran fossils (Fig. 1) are highly controversial in that some workers (e.g. Seilacher, 1989, 1992; see also Bergstrom, 1989, 1990) interpret them as separate “experiments” in multicellularity, independent of the Metazoa, which they term vendobionts. While this conceivably applies to some soft-bodied Ediacaran fossils, it is my contention that many are metazoans and so could provide crucial information on the early stages of diversification. Either way, interpretation of Ediacaran fossils remains highly controversial. Recently a compromise solution has emerged whereby the vendobionts are now interpreted as cnidarian-like metazoans, but lacking the stinging cells (cnidae) that were subsequently acquired by symbiosis with microsporidians (Buss and Seilacher, 1994; see also Shostak, 1993). The only point of general agreement on Ediacaran assemblages is that the co-existing trace fossils (Glaessner, 1969) are definitely the product of metazoan activity, albeit as unspecified “worms”.

Burgess Shale-type faunas are more straightforward. Shelly animals with a high fossilization potential form a small proportion of the total assemblage: in the Burgess Shale itself probably less than 5 per cent of individuals in the standing crop had robust hard-parts (Conway Morris, 1986). Significantly, soft-part preservation in the Lower and Middle Cambrian is quite widespread (e.g. Butterfield, 1994; Chen et al., 1991; Conway Morris et al., 1987). Burgess Shale-type faunas are dominated by arthropods (which here are taken to include the lobopods, anomalocarids and opabinids, but trilobites are relatively unimportant), but also contain a significant proportion of sponges, priapulid worms (including the palaeoscolecidans), and sometimes polychaete annelids (Conway Morris, 1989). Particularly striking is the general similarity between the Burgess Shale and Chengjiang faunas, although they are separated by perhaps as much as 10 Myr (the faunas are midMiddle Cambrian and mid-Lower Cambrian respectively) and occupy separate tectonic plates (Laurentia and South China) whose geographical separation in the Cambrian was thousands of kilometers (Conway Morris, 1989; see also Shu and Chen 1994).

It is worth stressing that simply because skeletal remains are relatively unrepresentative of the original nature of Cambrian life, they are by no means a redundant source of information. A serious problem in their study, however, is the tendency for many of the skeletons to disaggregate into dozens, if not hundreds, of component parts. In the case of the trilobites, this is seldom a problem because the group has such a rich fossil record, which often includes articulated specimens, and in any event a large part of trilobite palaeontology concentrates on the head-shield. In many other cases, however, we still have only the vaguest notion of the original skeletal configurations. In the Cambroclavida, a class of uncertain phyletic position, the sclerites are usually found isolated, but specimens are known that demonstrate unequivocally articulated rows of sclerites, juxtaposed in staggered arrays and sometimes back-to-back to produce arm-like structures (Conway Morris and Chen, 1991; Yue, 1991) vaguely reminiscent of the echinoderms (Conway Morris, 1993a). But the overall appearance of the cambroclave animal remains totally unknown. In the Tommotiida (an order of uncertain phyletic position) it is again reasonable to infer a skeletal arrangement of numerous, juxtaposed sclerites. Evans and Rowell (1990) have proposed that the tommotiid was broadly similar to an armoured slug, but some sclerites are strikingly similar to certain brachiopods. Perhaps this is simply convergent, as is, more probably, the general similarity of some tommotiid plates such as those of Dailyatia to the plates of barnacles (Bischoff, 1976; see also Bengtson, 1977). Nevertheless, the suspicion remains that tommotiids might have been sessile rather than a vagrant crawler. In yet other Cambrian groups, such as the Mobergellidae (a family of uncertain systematic position) (Bengtson, 1968; Missarzhevsky, 1989), it is not even clear if the animal carried a single phosphatic plate, two such or perhaps even hundreds of such elements.

These outstanding problems need to be set against some recent successes. Halkieriids, long known only from isolated sclerites, are now seen to have been provided with a dorsal coating of such sclerites mantling a slug-like animal, but against all expectation the scleritome also houses a prominent shell at both anterior and posterior ends (Conway Morris and Peel, 1990; see Fig. 2D). This articulated material was collected from the Lower Cambrian Sirius Passet fauna, a Burgess Shale-like assemblage exposed in Peary Land, North Greenland (Conway Morris et al., 1987). A broadly similar disposition of associated sclerites is now claimed for the hitherto enigmatic Volborthella, which is otherwise known from conical sclerites built largely by accretion of sediment grains (Signor and Ryan, 1993).

The interpretation of some of these sclerite-bearing animals is hindered because of different body parts showing variable preservation potential as fossils. Thus, disarticulated calcareous sclerites of halkieriids retain a relatively high preservation potential, especially if subject to secondary diagenetic phosphatization. In contrast, the related wiwaxiids (Conway Morris, 1985), the sclerites of which are unmineralized, have a much lower chance of becoming fossilized. The conspicuous exception in this category of variable preservation potential are the brachiopods, which appear to have been invariably equipped with either a calcareous or phosphatic shell. The brachiopod radiation in the Cambrian exemplifies many of the aspects of the general metazoan divergence, including a variety of geologically short-lived “enigmatic” taxa. Properly understood Cambrian brachiopods, with their high preservation potential, could provide one of the leading exemplars of the Cambrian explosion.

How then might palaeontological data contribute to understanding early metazoan evolution and complement the insights gained from molecular biology? Below, I discuss three examples where some progress has been made.

Ediacaran fossils (Fig. 1) lack obvious skeletal parts, yet they may reach substantial sizes: some of the frond-like organisms approach a metre in length. Seilacher (1989, 1992) was puzzled by the ubiquity of soft-part preservation in generally shallowwater, turbulent and presumably well-aerated environments which represent the very antithesis of the anoxic muds, so typical of exceptional tissue-preservation in younger sediments. Seilacher proposed a novel solution and interpreted the Vendobionta as a separate branch of eukaryote multicellularity. He argued that a unique composition of a tough exterior and an internal anatomy with a mattress-like construction, that lacked the metazoan features such as digestive, muscular and nervous tissue, could explain their high fossilization potential. This hypothesis was in dramatic contrast to the previous consensus that identified within the Ediacaran assemblages a preponderance of cnidarians, with anthozoans, cubozoans, hydrozoans and scyphozoans all identified with varying degrees of confidence (Glaessner, 1984; Jenkins, 1992). To this roster were added, again with fluctuating degrees of certainty, representatives of the so-called articulates, i.e. annelids and/or arthropods, as well as a possible echinoderm (Gehling, 1987).

The Vendobionta hypothesis has won both attention and adherents, but is it correct? A major problem in the interpretation of the Ediacaran fossils is their preservation in typically fairly coarse-grained sediments, making the resolution of some fine anatomical features controversial or even impossible. Even proponents of these fossils as metazoans will admit comparisons with known phyla are seldom straightforward. Nevertheless, circumstantial evidence for such features as muscular contraction and a circulatory system (Runnegar, 1982) cannot be ignored. In the Burgess Shale, moreover, a frond-like fossil (Thaumaptilon walcotti, Fig. 2C) closely approaches a number of Ediacaran taxa, most notably Charniodiscus (Conway Morris, 1993b). These latter frond-like fossils resemble anthozoan sea-pens (pennatulaceans), although Seilacher (1989) chose to emphasize the difference between some living examples, where the branches arising from the central rachis are separated, and those from the Ediacaran in which the branches are fused to a common blade (as is the case for Thaumaptilon). When the overall disparity of living sea-pens is considered, however, this difference seems relatively unimportant. Most significant is the recognition in Thaumaptilon of possible zooids and internal canals (Conway Morris, 1993b). Similar canals have been identified in Ediacaran fronds and the apparent absence of zooids can be reasonably attributed to entombment in sediments of a substantially coarser grain than the Burgess Shale. It cannot be disproved that Thaumaptilon is convergent with the Ediacaran fronds, but the onus of proof has shifted to supporters of the Vendobionta hypothesis. Moreover, if the Thaumaptilon-Chamiodiscus connection is accepted, then given that the Ediacaran fronds are preserved in the same way as the other co-occurring fossils, this suggests that although the preservational circumstances of Ediacaran assemblages still require an explanation, it is unlikely to be the consequence of a unique body-organization.

Molecular evidence also appears to be consistent with an early radiation of the cnidarians, and although the division between diploblasts and triploblasts remains deep (Christen et al., 1991; Wainwright et al., 1993) there is now little support for a diphyletic origin of cnidarians and other metazoans as originally proposed (Field et al., 1988). Evidence for metazoan monophyly comes from a variety of sources (e.g. Degnan et al., 1993; Erwin, 1993; Lake, 1990; Morris, 1993). Further molecular data indicate a primitive status for the anthozoans (Bridge et al., 1992) and this would be consistent with the abundance of frond-like fossils in Ediacaran sediments (and the later Thaumaptilon). The evidence for hydrozoans, in the form of chondrophorines, is moderately compelling, and Kimberella may be an early cubozoan (Jenkins, 1992). Despite the abundance of discoidal fossils, comparisons with the scyphozoans seem distinctly more tenuous. More implausible, however, is the recent resurrection by Valentine (1992) of an older idea that Dickinsonia is a polyploid cnidarian.

Molecular evidence also places cnidarians fairly close to the ctenophores (e.g. Christen et al., 1991), and it is possible that some of the bag-like Ediacaran fossils deserve consideration as early ctenophores. Nevertheless, the chances of identifiable comb-rows of cilia surviving seems remote, and again it is only the exceptional quality of preservation of these structures in Fasciculus from the Burgess Shale that demonstrates ctenophores extend back to at least the Cambrian (Conway Morris, 1993a; Conway Morris and Collins, unpublished observations). The precise relationships of ctenophores with early metazoan phyla nevertheless remain enigmatic and in some ways they approach more closely the triploblasts (e.g. Ehlers, 1993; Willmer, 1990).

Putative arthropods, such as Spriggina, and more uncertainly taxa such as Onega, Vendomia, Praecambridium and an asyet-undescribed “soft-bodied trilobite” (Jenkins, 1992, fig. 15) have been identified in Ediacaran assemblages, although proponents of the Vendobionta hypothesis (Bergstrom, 1989; Seilacher, 1989) have presented radically different hypotheses. Valentine (1988) has stressed how animals such as Spriggina are consistent with the paraphyletic and early branching of the arthropods as seen in some molecular trees (e.g. Lake, 1990). Unfortunately, despite its almost iconographie status as an Ediacaran representative, Spriggina remains remarkably poorly known. Another animal, Vendia, has also been persistently promoted as a primitive arthropod (e.g. Jenkins, 1992, p.168). Self-evident displacement in the only known specimen of left and right segments casts doubt on Vendia being an arthropod (Bergstrom, 1989; Fedonkin, 1985). Nevertheless, Jenkins (1992, p.168) notes that such displacement “could have occurred on burial”, and Runnegar (1995) has now scotched the reiterated proposal that Spriggina (Bergstrom, 1989) and Dickinsonia (Bergstrom, 1990) showed a comparable left-right asymmetry. The bilateral symmetry of their flexible bodies was evidently distorted during burial.

Attention, moreover, should be given to a number of other Ediacaran fossils. Bomakellia kelleri, for example, is only known from the Ediacaran localities of the White Sea, northeast Russia. Identified as an arthropod by Fedonkin (1985), surprisingly this fossil appears to have escaped almost any mention (Bergstrom, 1990) in recent discussions of arthropod phylogeny. Bomakellia appears to possess an anterior shield and segmented trunk that bears lateral structures and a more axial series of tubercles. Fedonkin (1985) interpreted the lateral extensions as appendages, but they may be better considered as pleurae. The only known specimen of Bomakellia is incomplete, but it appears to have been bilaterally symmetrical, as does the possibly related taxon Mialsemia (Fedonkin, 1985).

The Cambrian record of arthropods has been revitalized by the analysis of the Burgess Shale and subsequent supplements addressing the Chengjiang (e.g. Bergstrom, 1993; Hou and Bergstrom, 1991; Hou et al., 1991) and Sirius Passet faunas (Conway Morris et al., 1987) (see Fig. 1). At a relatively early stage of the investigation into these early arthropods it emerged that representatives of all four major clades (chelicerates, crustaceans, trilobites, uniramians) were present, but were greatly outnumbered by a seemingly bewildering array of forms whose precise relationship to any of the above groups was enigmatic. This uncertainty was graphically expressed by Whittington (1979) in a “phylogenetic lawn” that depicted a myriad of lineages arising from an unspecified Precambrian ancestor. More recently in a series of papers Briggs and Fortey (1989; Briggs, 1990; see also Wills et al., 1994) have formulated a cladistic analysis of arthropods that is being regularly updated and from which is emerging a reasonably consistent pattern (Fig. 3).

Our understanding of early arthropod evolution, however, is far from complete as is apparent from two notable developments. First, the diversity of lobopod animals is now known to be considerable, especially on the basis of the Chengjiang finds (Hou et al., 1991). Recent recruits to this regiment include the hitherto enigmatic Hallucigenia from the Burgess Shale (Ramskold and Hou, 1991). The consensus opinion would link, in some way, this lobopod radiation to the surviving terrestrial onchyophorans, although this may be in conflict with the derived position inferred from molecular biology (Ballard et al., 1990). A related and important development is the description of Kerygmachela kierkegaardi from the Sirius Passet fauna (Budd, 1993). In this arthropod (Fig. 2A,B) the lobopods are also associated with dorsal flaps, comparable to gills. At the anterior is a spectacular grasping apparatus. A full description of Kerygmachela is not yet available, but the biramous arrangement of lobopod and gill suggests a route by which the biramy of the chelicerates, crustaceans and trilobites may have arisen, by transformation of the flexible lobopods into cuticular, jointed appendages. Budd (1993) also emphasizes that Kerygmachela may help to constrain the systematic position of the hitherto enigmatic Burgess Shale animals Anomalocaris and Opabinia. The former is known to be equipped at its anterior with jointed appendages, and it remains possible that the prominent flaps arising from the trunk region now conceal legs. Such are depicted in a popular review of the Chengjiang fauna (Chen et al., 1991), although Chen et al. (1994) were unable to recognize legs or lobopods in newly discovered anomalocarids from this fauna. Opabinia is also known to bear an array of broad flaps, and the possibility of lobopods concealed beneath them needs re-investigation (G. Budd, personal communication).

The third example concerns the halkieriids (Fig. 2D), best known from the articulated specimens of the Sirius Passet (Fig. 1) fauna (Conway Morris and Peel, 1990). Research into these specimens is continuing in collaboration with Professor J.S. Peel (Uppsala), and our principal conclusions will only be touched upon. The slug-like appearance together with a dorsal coating of calcareous sclerites and two shells that grew by marginal accretion recalls in outline the appearance of primitive molluscs such as the chitons and aplacophorans. Halkieriids may indeed throw significant light on the derivation of early molluscs from a turbellarian ancestor (Bengtson, 1992; Conway Morris and Peel, 1990) and so support a recurrent proposal of invertebrate zoologists such as Vagvolgyi (1967; see also Stasek, 1972). Halkieriids appear to show homologies of sclerite arrangement with the somewhat younger wiwaxiids, best known from the Burgess Shale (Conway Morris, 1985). These animals were also compared to primitive molluscs (Conway Morris, 1985), although Butterfield (1990) made the significant discovery that the ultrastructure of the wiwaxiid sclerites is closely comparable to that of the chaetae of polychaetes, including those from the Burgess Shale (see Conway Morris, 1979). Butterfield’s (1990) claim, however, that wiwaxiids are true polychaetes is much more questionable, not least because of the absence of both parapodia, and the inter-ramal space (the latter is occupied by sclerites), as well as a feeding apparatus that is more like a molluscan radula than any comparable jaw in the polychaetes. Interpreting halkieriids as part of the stem group that led to polychaetes appears to be a distinctly more informative exercise. Finally, Conway Morris and Peel (1990) noted that the shells of the Sirius Passet halkieriid, especially that of the posterior (Fig. 2D), are remarkably brachiopodlike. This was regarded by us as a superficial convergence, but further research suggests that the once unpopular idea of a close link between annelids and molluscs (Field et al., 1988; Ghiselin, 1988) also needs to be supplemented by the longoverlooked proposal (e.g. Morse, 1873) of a near-relationship between annelids and brachiopods. Such a proposal is consistent with evidence from molecular biology, but is highly controversial amongst whole-organism zoologists (e.g. Willmer, 1990) who persist in allying brachiopods with other deuterostomes.

Palaeontological data suggest that significant new insights into metazoan evolution are already available. First, in at least two areas of protostome evolution (the annelid-brachiopod-mollusc connection and the early divergence of arthropods), it seems reasonable to invoke evolutionary transitions from turbellarians, albeit by two rather different routes. If correct, then this suggests that the metameric segmentation of annelids and arthropods arose separately, although they may still share homologous coding instructions that control the so-called pseudometamery of turbellarians (see also Holland, 1990; Newman, 1993). Second, fossil evidence of what may be termed Ediacaran survivors, principally Thaumaptilon from the Burgess Shale (Fig. 2C), appear to undermine the Vendobionta hypothesis and reaffirm an early origin and radiation of cnidarians. Problems also remain. Despite a rich record of Cambrian deuterostomes, including primitive echinoderms, rhabdopleurid hemichordates (e.g. Bengtson and Urbanek, 1985; Durman and Sennikov, 1993) and the cephalochordatelike Pikaia, their origins are still poorly understood (although see Jefferies, 1986, for a review of his controversial discussion of the calcichordates). Other outstanding problems include the origin and early divergence of the aschelminthes and ctenophores.

Here are five topics of mutual and reciprocal interest to palaeontologists and molecular biologists:

• (1) There is an urgent need to expand the roster of examined metazoans. For example, in terms of molecular sequencing, next to nothing is known of groups such as priapulid worms, polychaete annelids, rotifers, articulate brachiopods, and chiton molluscs. If the comments given above concerning halkieriids, for example, win assent then we predict further molecular evidence for a close affinity not only between annelids and molluscs (Ghiselin, 1988), but also with brachiopods. Nearly all the available information on annelids refers to the highly derived leeches (e.g. Wedeen et al., 1991; Wedeen and Weisblat, 1991; Wysocka-Diller et al., 1989). Amongst the molluscs almost nothing is published on chitons, aplacophorans or monoplacophorans. The so-called “living fossil” status of Lingula, together with its availability, explains the interest in this supposedly primitive brachiopod, but forms such as Crania certainly deserve investigation.

  The expansion of the data-base across all known phyla, however, must be accompanied by more extensive investigations among smaller clades. Examination of molecular trees often shows, unsurprisingly, that closely related taxa are separated by very short branch-lengths. But there seem to be some puzzling exceptions: Rosenberg et al. (1992) found remarkable variability in 28S rRNA (in the D6 domain) of seven species of truncatellid gastropod.

• (2) The wide employment of rRNA is now being supplemented by other gene sequences or products, although their applicability to unravelling deep relationships within the metazoans, of course, will vary (see Kumazawa and Nishida, 1993). Nevertheless, at present there are relatively few examples (e.g. Kojima et al., 1993; Miller et al., 1993; Suzuki et al., 1993; Raff et al., 1984) where the congruence of molecular trees can be tested against unrelated sequences.

• (3) There is a rich literature on biochemistry and physiology, much of which has escaped being placed in an evolutionary context. In part this is because many metabolic pathways are indeed fundamental to cellular processes. However, a comparison of more specific bioproducts and their utilization may reveal either unexpected examples of convergence or instances of supposed independent biochemical innovation actually reflecting shared ancestry. One example might be represented by the respiratory pigment haemocyanin, which occurs only in arthropods and molluscs. Accordingly, this molecule has been regarded as an important phylogenetic indicator. Evidence from both molecular biology and palaeontology, however, does not support a particularly close relationship between arthropods and molluscs. More critical analysis of haemocyanin now reveals that its higher order structure is fundamentally different in these two phyla (Mangum, 1990).

• (4) Another fundamental problem is the way and extent by which developmental mechanisms have evolved. The central paradox at present is that flies and mice, for example, differ phenotypically in self-evident ways, but they share homologues in their developmental machinery, including homeotic genes (Holland, 1990). Some of these genes are remarkably widespread, including antennapedia-like sequences in cnidarians (Miles and Miller, 1992; Miller and Miles, 1993; Murtha et al., 1991; Schierwater, 1991; Schummer et al., 1992; Shenk et al., 1993a,b). Particular interest also lies in the recent detection of homeoboxes in the platyhelminthes (Bartels et al., 1993; Garcia-Fernandez, 1993; Oliver et al., 1992; Webster and Mansour, 1992), given that their primitive position relative to all other triploblasts may now be close to general acceptance. To date detailed proposals of how developmental mechanisms might have evolved are rather limited (e.g. Averof and Akam, 1993; Holland, 1990, 1992; Kappen and Ruddle, 1993; Raff, 1992), although gene duplication may have been an important, albeit fortuitous, factor in allowing the co-opting of genes for new instructions (see Holland, 1990, 1992). Jacobs (1990) has proposed an ingenious model to link arrangement of developmental instructions, by what he labels as “selector genes”, and perceived contraints of morphological expression as indicated by rates of ordinal origination. He argues that arthropods and annelids with clear serial construction underwent initial exuberance of morphological experimentation in the Cambrian, but were subsequently constrained by a regulatory system that had to remain simple owing to dispersal of the “selector” genes. Apart from general problems of testability, one difficulty with Jacobs’ hypothesis is the bias introduced by the large number of supposedly enigmatic taxa (orders) of Burgess Shale arthropods, whose high-level taxonomic status is probably exaggerated (see Briggs, 1990). What is emerging from these preliminary discussions of the evolution of developmental mechanisms, however, is that, not withstanding the antiquity of such structures as the antennapedia-class (Fig. 1), the complexity of higher metazoans in part is founded on gene duplications and subsequent co-option.

  To date little is known about the genome of putative ancestors, notably in either the protistan ciliates or the fungi (e.g. Baldauf and Palmer, 1993; Wainwright et al., 1993) in terms of possible homologues of developmental genes. Such information will help to constrain the nature of the ur-metazoan (see Shenk and Steele, 1993). Equally intriguing is whether a single function for primitive homeobox genes will be identified. Two items come to mind. Some evidence exists for certain genes having a primary neurogenic role, especially in the central nervous system (Patel et al., 1989; Wedeen and Weisblat, 1991). In this context perhaps we should recall that Stanley (1992) has proposed that the initiation of the metazoan radiations can be traced to the invention of the neuron. Perhaps equally fundamental, or more so, is the role of some homeoboxes in determining axis and orientation, crudely head and tail (e.g. Bartels et al., 1993). Directionality and nervous control may be the hallmarks of metazoans, and the latter might be the key to the initial diversification which was then fuelled by feedbacks, including more complex ecologies marked by the spread of predation (Vermeij, 1990) and grazing (Butterfield, 1994).

• (5) The subsequent history of metazoan evolution has often been depicted as the establishment of a remarkable stability of body designs, perhaps best exemplified in the insects. This has been linked to a vague notion that somehow the genome becomes “congealed”, thereby precluding the morphological experimentation that is said to characterize the Cambrian faunas. This stability may be more apparent than real. As a whole, the genome is recognized to be highly dynamic, and this appears to apply with equal force to rates of evolution of development (Wray, 1992). It is also a simplification to identify early stages of the embryology as conservative and later ones as more flexible. Wray (1992; see also Wray and Raff, 1991; Raff, 1992) presents a cogent rebuttal of this simplification, and stresses both the degree of developmental variation in some closely related taxa as well as evidence for rapid and geologically recent changes in early development. Most important, however, is Wray’s (1992, p.131) emphasis that there is no evidence that the “developmental differences that distinguish phyla and classes” are materially different from those that divide lower taxa, and neither is there any reason to accept the popular notion that “developmental programmes have become too constrained by interaction since the early radiation of metazoans to allow the origin of new body plans”. One could argue that the barnacle is a new body plan, but because its relationship to other arthropods, specifically the crustaceans, is clear, such a manoeuvre serves no useful purpose. But I would argue that Wray’s (1992) prescient observations apply with equal force to understanding the Cambrian diversifications, although the reader may wish to consult Erwin (1994) in support of opposite views.

It may be naive to imagine that the fossil record will reveal transitions between all the phyla, but this matters little if key examples such as the role of the halkieriids in protostome diversification continue to provide useful insights. What applies to the origin of groups as disparate as annelids, brachiopods and molluscs should be equally applicable in principle across the Metazoa. Such data, combined with an understanding of the evolution of developmental mechanisms and co-option of pre-existing genes, suggests that one of the central problems of biology is close to solution. Is it now time to consider what questions will arise from this advance?

Jeff Levinton and another anonymous referee provided exceptionally helpful reviews. I thank Sandra Last for typing several versions of this paper, Hilary Alberti for assistance with drafting, and Dudley Simons for help with photography. Graham Budd kindly made available Fig 2A,B, while Derek Briggs made available the cladogram which appears in simplified fashion in Fig. 3. Earth Sciences Publication 3817.