Dinocaridids: anomalous arthropods or arthropod-like worms?

Transcription

1 Dinocaridids: anomalous arthropods or arthropod-like worms? Hou Xianguang & Jan Bergström Jan 2006 In: Rong Jiayu, Fang Zongjie, Zhou Zhanghe, Zhan Renbin, Wang Xiangdong & Yuan Xunlai (eds): Originations, Radiations and Biodiversity Changes evidences from the Chinese fossil record. pp Herein an English version of the Chinese text. Extended abstract in English: Science Press, Beijing 2006

2 Dinocaridids anomalous arthropods or arthropod-like worms? Hou Xianguang & Bergström, Jan Hou Xianguang, Yunnan Key Laboratory for Palaeobiology, Yunnan University, Kunming ; Jan Bergström, Department of Palaeozoology, Swedish Museum of Natural History, SE Stockholm, Sweden; The history of research on Anomalocaris and related carnivorous Cambrian animals, the dinocaridids, is intriguing. Their remains have been interpreted as sponges, jellyfishes, echinoderms, annelids, flatfish-like protostomians and crustaceans. They tended to disintegrate more easily than typical arthropods, and the Chengjiang material shows that this is because they were largely soft-skinned, albeit with some sclerotised parts. It has been suggested that they swam with lateral fins, but the Chengjiang material allows us to identify biramous appendages. It is now a question whether the biramous condition is shared with arthropods because of common descent or because of convergent evolution. The Chengjiang species prove that the dinocaridids have transverse bands of lanceolate blade-like structures attached to transverse rods, and that these structures are dorsal. Dinocaridids show similarities with fossil soft-legged lobopodians, most of which were found in the Chengjiang fauna. Dinocaridids and lobopodians are important for our understanding of the origin of arthropods, and of the possible relationship between arthropods and cycloneuralian worms. Here we summarize the evidence bearing on the morphology and relationships of the dinocaridids, with particular emphasis on the Chengjiang finds. 1. Introduction The first finds of Anomalocaris and similar animals were made in the Burgess Shale, British Columbia, Canada, towards the end of the 19th Century. J. F. Whiteaves (1892) then described a supposed shrimp under the new name Anomalocaris canadensis. C. D. Walcott (1911) described other remains as seemingly quite different animals: the holothurian echinoderm Laggania cambria, later interpreted as a sponge, and the jellyfish Peytoia nathorsti. In 1979, D.E.G. Briggs realised that the shrimp must be the appendage of an animal rather than the entire animal, and Harry Whittington and Derek Briggs (1985) published proof that the different types of remains all belonged to a single general type of animal, represented by a few species in the Burgess Shale. This animal type has a head region carrying a pair of eyes and a pair of powerful, segmented and spiny arms apparently for catching prey. Behind the head, there are structures which where interpreted as segmentally arranged fins (Whittington & Briggs 1985; Collins 1996, Figs. 6, 8). No traces of leg structures were identified. Linear structures were interpreted as dorsal gills. One of us (Bergström 1986, 1987) recognised transverse striated bands associated with rods and extending from side to side over the dorsal side and also questioned the fin interpretation, suggesting that the supposed fins are part of ventral appendages. The development of the story so far was described in more detail by Desmond Collins (1996), who reverted to the fin interpretation of Whittington and Briggs. Collins also added to the story by ignoring the striated bands and shifting to the ventral side the associated segmented dorsal rods. In Collins view, the rods extend into the fins as supportive structures (Collins 1996, Fig. 8, 9). Deep impressions made by the rod ends at the steep lateral sides of the body were described as club-shaped ends of the rods. It is notable that the Burgess Shale has yielded no isolated walking legs, although there are plenty of frontal, grasping appendages. One explanation may be that, like in the Chengjiang Cucumericrus (Fig E), the legs were not sclerotised, and therefore did not easily survive

3 as fossils, if not as mere imprints in fossilised complete individuals. The Burgess Shale has also yielded the odd Opabinia regalis (Fig C), described as a crustacean by Charles Walcott in 1912, and restudied by Harry Whittington in 1975 and by Graham Budd in Superficially this animal is quite different from the Anomalocaris species, but Bergström (1986) noted important similarities, which are now generally accepted (Budd 1996). Unlike Anomalocaris, Opabinia was reconstructed with the transverse bands (interpreted as gills), which were correctly placed on the dorsal side, although restricted to the pleurites. Two genera, Kerygmachela (Fig D) and Pambdelurion, of related appearance, have been found in Greenland (for instance, Budd 1998, 1999a). These further deepen our understanding of the natural group, the Dinocaridida, of which Anomalocaris is a member. Some terms used: Aiolopoda (from Greek aiolos, shifting, changing, variable, referring to the variation in shape and function of the limbs, and pous, podos, foot ): new term for a superphylum containing taxa with segmentally arranged, paired locomotory limbs: Cambrian lobopodians, dinocaridids, onychophorans, true arthropods, and perhaps tardigrades. This term is introduced to avoid the confusion caused by the variable use of Pan-Arthropoda and Arthropoda (sensu lato). anomalocaridids: as used in this text comprise the Anomalocaris group with Anomalocaris, Amplectobelua and Cucumericrus, and the Peytoia group with Peytoia (=Laggania,? Amiella), Parapeytoia and Cassubia. Arthropoda Siebold & Stannius, 1845 (from Greek arthron, joint, and pous) (= Euarthropoda Lankester, 1904): aiolopods with sclerotised surface and segmented appendages. Mouthparts consist of modified appendages. basis, the most proximal segment in many old arthropods, see Fig D. In crustaceans, a newly formed coxal segment has displaced it to the second position. dinocaridids: anomalocaridids (see above), Opabinia, Kerygmachela, Pambdelurion, formally united in the Class Dinocaridida Collins, 1996 Dinocarida). (name corrected from endopod: the medial branch of a bifid ventral appendage in arthropods. It extends from the segment called basis. See Fig D, en. exopod: the lateral branch of a bifid ventral appendage in arthropods. It extends from the segment called basis. Se Fig D, ex. gill: as used in the literature, any lineated or striated, straight but also curved, structure, dorsally or ventrally, on the body or on the appendages. lanceolate blade: the individual element in the in the transverse dorsal bands that cover the back and/or the pleurites in each body segment. lobopod: the unsegmented, not sclerotised locomotory limb of many aiolopods. lobopodian: an aiolopod with lobopod limbs. An old synonym is malacopod. peytoiid mouth: the ring of sclerites surrounding the mouth in dinocaridids, named after the genus Peytoia. pleurite: a more or less thin fold extending ventrally to laterally from the body side in many arthropods as well as in dinocaridids; the upper side is an extension of a tergite (dorsal sclerite). propod: the proximal and lateral part of the ventral appendage in dinocaridids. Where known, it has 5 endite projections on the medial side. Se Fig B, prp. telopod: the walking leg in dinocaridids. It extends from beyond the 5th endite in the propod. See Fig B, tel. veins: thin lines extending over part of the swim flaps in some anomalocaridids. xenusians: informal collective name for lobopodian aiolopods other than the onychophorans and tardigrades, from the formal class name Xenusia Dzik & Krumbiegel, 1989 (Fig ). 2. Contributions of the Chengjiang fauna Hou et al. (1995) described four new dinocaridids from the Chengjiang biota. Of these, one was referred to the old genus Anomalocaris, whereas three species

4 represented body constructions new to science. The latter three were placed in three new genera. Like other dinocaridids, some species from China reach large size, up to an estimated maximum of m. An estimation of a greater length (e.g., Chen & Zhou 1997, p. 81, figs ; Maas et al. 2004, p. 160) is based on a misidentification of both taxon and structure (Hou et al. 2006). The species from China confirm important observations based on the Burgess Shale species. Thus, powerful grasping appendages are noted in Amplectobelua symbrachiata, Anomalocaris saron and Parapeytoia yunnanensis (Figs C, D and F). There is no evidence for or against such appendages in the fragmentarily preserved fourth species, Cucumericrus decoratus which, however, is close to A. saron and A. symbrachiata. The large lateral flaps typical of Anomalocaris are seen in all species. In the species of Anomalocaris, Amplectobelua and Cucumericrus, the flaps have a characteristic pattern of regularly spaced lines or veins (Figs E, 2.8.1B). The peytoiid mouth known from the Burgess Shale species is known from Anomalocaris saron and Parapeytoia yunnanensis. There are distinct differences between the species and genera in the morphology of the grasping appendages. Thus Anomalocaris and Amplectobelua have double rows of medially directed spines, with one spine pair on each of the many segments (Figs C-D). [The spines are simple in Amplectobelua, but branched in Anomalocaris. Amplectobelua has a single, strongly prolonged spine on one of the proximal segments.] Parapeytoia differs in having only a single row of flat, finger-like spines (Fig F). Also, the appendage has only five segments, the last four of which have single, medially directed, long and flat fingers (Hou et al. 1995, p. 175, Figs. 12A-C). Contrary to the account of Chen et al. (2004), all fingers are serrated on the edge facing a neighbouring finger (Fig F; Hou et al. 1995, p. 169 and fig. 12A-C). Parapeytoia yunnanensis is reminiscent of the Burgess Shale Peytoia (or Laggania) nathorsti in having a single row of long flat fingers, but the latter has nine fingers rather than four, and they are strongly spinous (on one side only) rather than serrated (Fig G; Briggs 1979, Appendage F ; see Collins 1996 for identification). To many colleagues, the greatest surprise in the description of the Chengjiang dinocaridids would have been the find of ventral legs, and the realisation that the lateral fin is a lateral projection of the limb. The most derived leg is so far known from Parapeytoia (Fig F; Hou et al. 1995, figs 9, 10, 11A, D, 12D, 13, 19). In this genus there is a propod with five enditic projections. The propod extends laterally without any articulation into the very large lateral flap, the fin or lateral lobe of authors. Given that each endite corresponds to one segment, which is indicated by the similarity in lengths of endites and leg segments in Parapeytoia, this is a notable difference from the condition in arthropods. In the latter, there is only a single segment (the basis) proximal to the two limb branches (endopod and exopod) except in crustaceans, where a coxal segment has been secondarily inserted proximal to the basis. The dinocaridid flap was a fairly stiff, sclerotised structure. A pattern of parallel veins had an unknown function. A fully segmented leg (telopod) extends distally from the propod in Parapeytoia. It has at least eight segments excluding the distal element. In an informal way, the leg segmentation thus makes Parapeytoia an arthropod in the sense that this term means joint-leg. Next surprise is that Cucumericrus has a leg that shows diffuse indication of segments only in the propod and towards the distal end (Fig E; Hou et al. 1995, fig. 16). This animal therefore is somewhat intermediate between a lobopodian and an arthropod as far as the leg is concerned. It may be worth noting, however, that the propod has the same number of endites as we have noted in Parapeytoia (Hou et al. 1995, Figs. 13, 16C). The Chengjiang material has made it possible to observe the nature of the lanceolate blades (variously called gill-like structures, pleated structures, gills, comb-like gills, lanceolate dorsal scales ) in the repeated dorsal squirts of Anomalocaris (Hou et al. 1995, figs 4-6). Each lanceolate blade is flattened, but the blades are so tightly spaced that each of them must have rested with an edge against the body, rather than with a flat surface. Thus, as preserved, they tend to form imbricated series of blades (Hou et al. 1995, Fig. 6, sc ). There is an indication of such blades also in both Parapeytoia (Fig A; Hou et al. 1995, figs 9-10) and Amplectobelua (Fig B; Chen et al. 1994, fig. 3A). There can hardly be any doubt that such squirts were attached along the distinct transverse rows of pits seen in Burgess

5 Shale Anomalocaris, and there associated with linear impressions of the lanceolate blades (Whittington & Briggs 1985, fig. 34; Bergström 1986, fig. 3B; Hou et al. 1995, fig. 17A; Collins 1996, fig Still another great surprise was the realisation that the body (at least in Cucumericrus) is largely covered with soft, wrinkled skin rather than by stiff sclerites (Hou et al. 1995, fig. 16DE; Bergström & Hou 2004, fig. 6), as would have been expected in an arthropod. The wrinkling comes in different patterns, one with close to ten wrinkles in a millimetre, another with about two wrinkles in a millimetre, and a third with a wider meshwork. To this comes the wrinkling in the leg, referred to above. This lack of a complete cover of body sclerites may explain the poor preservation of dinocaridids in general. We get a picture of dinocaridids as animals with a range in sclerotisation which can be listed roughly as follows: 1, strongly sclerotised: grasping appendages and adjoining sternit(es), mouth ring (cone), segmental sternites in Parapeytoia; 2, well sclerotised: lateral limb flap, telopod in Cucumericrus, pleurites (pleural lobes) in Opabinia; 3, weakly or not sclerotised: walking leg generally, general body surface except in Opabinia. 3. Close relatives? Anomalocaris is known from different parts of the world, but the other Chengjiang and Burgess Shale genera are so far unknown from places other than their type localities. Instead, there are some possibly related genera from Greenland, namely Kerygmachela (Figs A, 2.8.1D) and Pambdelurion. There is also the genus Opabinia from the Burgess Shale. Opabinia (Fig C) is unlike the Chengjiang dinocaridids in several respects, but there are also notable similarities including the tripartite tail fin, lanceolate blades, grasping appendages and, possibly, a peytoiid mouth. These similarities suggested to Bergström (1986) that Opabinia is related to Anomalocaris, a view that now appears to be generally accepted (see Budd 1996, p. 6). Deviating features include the unique opabiniid proboscis. It is clear, however, that the distal end consists of a pair of Anomalocaris-like grasping appendages (for instance, Budd 1996, fig. 7). It appears that a median corm has evolved to enhance the range and flexibility of the grasping appendages. Large flaps are seen on the sides in both genera, but those in Anomalocaris are extensions on the ventral appendages, those in Opabinia pleurites from the body side (Bergström 1986; Budd 1996, p. 2-3 and fig 3). Both genera have dorsal lanceolate blades. These are situated on the pleurites in Opabinia (Fig C). In anomalocaridids, in which there are no pleurites, the lanceolate blades form bands crossing the back (Fig A-B), where the rod-like attachment structure is similar to that described by Budd (1996, p. 3, fig. 4) for Opabinia. In his reconstructions of Peytoia (or Laggania), Collins (1996, fig 8, 9 left side) placed the rods on the ventral side and had them prolonged with the segmented legs, which in his drawing are club-shaped. Budd (1996, p. 3, fig. 4) thought that the lanceolate blades are gill blades and that this would preclude the flaps from being pleurites. They may, or may not, have functioned as gills, we think not, because of their degree of sclerotization. As a last point, Opabinia is usually thought of as lacking a peytoiid mouth similar to that in anomalocaridids. This may not be so some illustrations give the impression that there is such a ring (Whittington 1975, figs 4, 7, 55, 59; cf. Hou et al 1995, p. 178). Thus the differences are hardly great enough to preclude a relationship between Opabinia and anomalocaridids. It is more to the point to note that there are indeed unique similarities uniting them (Table 2.8.1). Pambdelurion is fairly poorly known. Budd mentioned and illustrated lateral lobes, and thought that the lobopodous limbs (Fig B) were probably unconnected to lateral flaps (Budd 1998, p. 126). However, his illustrations (Budd 1998, Figs. 11.6B-C, ventral views) show that each lateral lobe overlaps its neighbour in front, just as it does in Anomalocaris (e.g., Whittington & Briggs 1985). In arthropods, any pleurite overlaps its posterior neighbour. This would speak against a pleurite nature of the lateral lobes and in favour of them being ventral appendage flaps. Budd (1998, Table 11.1) listed both lateral lobes (character 11) and biramous limbs (his character 12) as being present in anomalocaridids, whereas Kerygmachela, Opabinia and Pambdelurion would have only lateral lobes. This is confusing. If, instead, we distinguish between pleurites and limb paddles, distinct, ventro-

6 laterally extended pleurites are seen only in Opabinia, whereas all the others have long, strictly lateral flaps, perhaps consistently being limb paddles. Therefore, Opabinia may be alone in having both limb paddles and pleurites (cf. Table 2.8.1). Pambdelurion definitely has somewhat lobopod-like legs, which differ from those of typical lobopodians in having distinct, narrow sclerite rings, making them pseudo-segmented (Budd 1998, Figs. 11.5B, 11.6A, C, 11.8). This condition is not known from any lobopodian. A distinct similarity with advanced anomalocaridids is the presence (if correct) of a Peytoia-like mouth cone. The double series of spines on the grasping appendages is a character shared with Anomalocaris and Amplectobelua (Fig B). Kerygmachela (Fig D) has a narrow and parallel-sided, annulated body similar to that of Cambrian lobopodians (xenusians, Fig A-G). Similarities with the anomalocaridid group are said to include large grasping appendages, a conical mouth apparatus, and large lateral lobes with striation, the gills of Budd. The grasping appendages are different in all details, for instance in being lateral rather than ventral, and in having a set of very long terminal structures (Fig E). The latter are thought by Budd to be possibly defensive spines, but they have variable degrees of bending and look more like tentacles provided with sensory papillae (Budd 1999a, Figs. 2a, 6a, b, d, 28a). The striated pattern (Fig D) differs from that of anomalocaridids in not being parallel to the body axis and in lacking association with segmented, transverse rods (Budd 1999a, Fig. 13), and in being probably fused ( intimate ) to the lateral lobe (Budd 1999a, Table 2 and Fig 30). The mouth apparatus, although being circular, is not impressively similar to that of the anomalocaridids. 4. Limbs and locomotion Anomalocaris has been illustrated as swimming by means of undulating lateral fins, roughly as a flatfish or a squid (Whittington & Briggs 1985; Collins 1996). On the evidence from Chengjiang material, we can instead interpret the ventral appendages as the locomotory organs. Swimming, as well as walking, could therefore be accomplished in a way more typical to arthropods, with limbs swinging back and forth. The dinocaridid legs appear to range from irregularly wrinkled lobopods via poorly segmented intermediates to fully sclerotised and segmented legs. Pseudo-segmented lobopods constitute another type of intermediate structure. All these types are adapted for standing and walking on the substrate. Budd s concept of lateral lobes needs an interpretation. He illustrated and mentioned imbrication of the lobes, which in the case of Kerygmachela means that each lobe overlaps its posterior neighbour (Budd 1993, p. 709, Fig. 1b, f; 1999a, Fig. 30). Therefore, lateral lobe in this case would appear to be a pleurite rather than a ventral structure. If so, it is possible that its function was protective. However, Budd (1999a, p. 266) suggested that it might have been used in swimming. Perhaps the lobe is not a pleurite but a propod swim flap, just as in Anomalocaris. In Pambdelurion, each lateral lobe overlaps its anterior neighbour, similarly as in Anomalocaris. The general configuration is quite similar to that in Anomalocaris, with the pediform branch extending along the posterior edge of the lateral lobe (Budd 1999a, Fig. 11.6C). This indicates a similar morphology of the appendage, with a large propod giving rise laterally to the large lobe and ventrally to the pediform branch. Pambdelurion has a wide body, like that in anomalocaridids but unlike the parallel-sided body of Kerygmachela. It is likely that Pambdelurion was an active swimmer. The interpretation of Opabinia is problematic. It is clear that the lanceolate blades are attached anteriorly to the dorsal surface of pleurites (Bergström 1986; Budd 1996). We are convinced that the row of blades extend also onto the body flanks (Fig C). We feel satisfied with Budd s conclusion that Opabinia has lobopod legs. However, the presence of tail fins indicates a capacity to swim, which would be difficult with lobopods as the only locomotory organ. Despite the lack of hard evidence, we therefore suggest that Opabinia was provided with ventral appendages similar to those of anomalocaridids, with a flat propod for swimming and a distal telopod for walking (Fig C). It is possible that the propod flaps could be more efficient in the partly closed space under the pleurites and that, therefore, they could be smaller than in other dinocaridids.

7 We are therefore inclined to believe that most, perhaps all, dinocaridids had limbs with a swim flap. We can imagine swimming much as in branchiopod crustaceans, which have also flattened appendages. The presence of a triplet of tail fins indicates that the appendages allowed only poor manoeuvrability, however. It should be noticed that, although such tail fins are known only from Opabinia, Anomalocaris and, less certainly, Parapeytoia, their angled posture might be a likely cause for non-preservation in many cases. 5. Evolution and relationships 5.1 There are four main interpretations of the phyletic position of the dinocaridids in relation to arthropods. According to the first one, the dinocaridids are ancestral to true arthropods. In a second interpretation, dinocaridids and uniramian arthropods form a sister-group of biramous arthropods. In a third interpretation, dinocaridids are stem-group chelicerate arthropods. A fourth interpretation have dinocaridids as a closed group outside of the arthropods. This leads to a consideration of the relationships with other animal groups, in particular lobopodian taxa. i. First interpretation. Budd (1998a, fig ) published a cladistic analysis, with lobopodians as an outgroup, showing the following branching sequence (cladistic formula): [Recent lobopodian Peripatus (Cambrian lobopodian Aysheaia (Kerygmachela (Opabinia (Pambdelurion (Parapeytoia + arthropods))))]. A few of the key characters are difficult to understand. Thus, no. 3, lateral branched frontal appendages, is said to be shared by the groups beyond Peripatus, but to our knowledge neither of these have branched frontal appendages. Parapeytoia is said (character 9) to have lost its Peytoia-like mouthparts (although they were illustrated by Hou et al. 1995, Figs 10m and 18A) and to have a complete cuticular sclerotization with tergite articulation (we know only of ventral sternites, not of any dorsal tergites). Dewel & Dewel (1998) published a similar analysis, with lobopodians and dinocaridids positioned as stem-group arthropods. ii. Second interpretation. Budd (1993) originally suggested a phyletic tree with two terminal sister clades, one consisting of arthropods exclusive of the Myriapoda (and Hexapoda), the other of an Anomalocaris-like clade (Kerygmachela, Opabinia, perhaps Anomalocaris). This means that conventional arthropods would have evolved twice from lobopodians, as the uniramians and the biramous arthropods. Next interpretation has a similar double origination, but the resulting arthropod groups are different. iii. Third interpretation. - Maas et al. (2004) sketched a phylogenetic sequence leading from anomalocaridids to a node with lineages leading directly to the anomalocaridid Parapeytoia, to separate non-lamellipedian arthropod lineages (ending in Fortiforceps, Tanglangia, Yohoia, Jianfengia, Alalcomenaeus and Leanchoilia), and to chelicerates. This diagram is based entirely on the morphology of the grasping appendage. A basic idea is that predation was a dominant mode of life in early arthropods. They refer to Butterfield (2002) for gut contents thought to indicate a predatory mode of feeding in Chengjiang arthropods. However, Butterfield made a wild guess on Chengjiang material entirely from a study of Burgess Shale material. He did not know that the gut content of Chengjiang arthropods typically consists of sedimentary quartz grains and the mica mineral(s) muscovite/illite in the same proportions as in the surrounding sediment (mineralogical and radiographic analyses, Ulf Hålenius, Swedish Museum of Natural History, Stockholm). The authors accept that convergent development of predatory limbs is not uncommon among arthropods and that therefore it may be that the anterior appendages modified into raptorial organs more than once early in arthropod evolution (Maas et al. 2004, p. 159, right column). It follows a logical somersault: because of the similarity of the grasping appendage from anomalocaridids over various Cambrian arthropods to chelicerates, this kind of first appendage must have formed only once (Maas et al. 2004, pp Their conclusion puts us in the awkward situation that the peytoiid mouth devised for carnivory must have been lost seven times in their diagram (Maas et al. 2004, Fig. 2), while the carnivory, for which the peytoiid mouth may have been

8 specially designed, was retained. Similarly, endites must have been lost repeatedly, while a typical arthropod dorsal exoskeleton was invented seven times. The authors do not discuss this problem, nor do they explain where other arthropods came from. Correcting two misunderstandings removes much of the similarity the authors claim to exist along the series of appendages dealt with: the authors (Chen et al. 2004; Maas et al. 2004) have not noticed that spines/fingers are paired in Anomalocaris and Amplectobelua (Fig B; Hou et al. 1995), but single in chelicerates. Nor have they recognised that the fingers in Peytoia were not designed for grasping between the fingers, but for making a basket to confine its food (Fig G; Briggs 1979, text-figs , Pl. 80, figs. 1-4, 8; Collins 1996, Fig. 7-2). We note also that a grasping appendage with repeated fingers is known from the second appendage in the Cambrian Occacaris (Hou et al 2004, pp ), demonstrating that evolution could bring forward the same morphology more than once. It is confusing that the authors have not included the stem-lineage crustaceans inclusive of agnostids (see Bergström & Hou 2005) into their discussion. This would have been very relevant. It has, in fact, been illustrated and pointed out in a number of contributions that these animals, or at least their larvae, had 1st antennae with strong spines useful in food capture (e.g., Müller & Walossek 1986, 1987; Walossek & Müller 1990; Walossek & Szaniawski 1991). The spines are thinner than the fingers in the arthropods that were treated by Maas et al. (2004), but this is not a relevant argument against a possible relationship with chelate appendages. The further development of a chelate morphology from a stem-group crustacean 1st antenna would be a fairly small step. A serious criticism against the approach in Chen et al. (2004) and Maas et al. (2004) is the lack of accuracy. [Their references to the literature are often so distorted, and their reinterpretations are based on insufficient photographs rather than actual specimens, that text and conclusions are misleading. If we select an easily demonstrable example out of many, they claim that that our illustrations of Parapeytoia yunnanensis show approx. 7 segments (Chen et al. 2004, p. 14), although according to our photograph, drawing and text there are 8, excluding the terminal element (Hou et al. 1995, figs. 11D, 12D, and p. 169). - This section probably rewritten in the Chinese version, with new references] 5.2 Relationships with other taxa According to other interpretations, the dinocaridids form a discrete, monophyletic group. Differences between opinions here concern merely a question of the phyletic distance between dinocaridids and typical arthropods. Apart from arthropods, it is lobopodians that are most often mentioned as being possibly related to the dinocaridids. Budd suggested a stem-group of lobopodians giving rise either to biramous arthropods and dinocaridids as sister-groups (Budd 1993), or to a dinocaridid plexus positioned as stemgroup arthropods (e.g., 1996, 1999a, Fig. 37). Modern onychophoran lobopodians are shown as an early branch in a phyletic tree based on Cambrian lobopodians (Budd 1996, Fig. 9). Chen et al. (1994) produced a cladogram with lobopodians as outgroup and dinocaridid taxa nested within the (eu-)arthropods. Budd (1999a, Table 1), rejected 13 of the 17 characters used in their cladogram to elucidate the possible relationship between dinocaridids and arthropods. Of the remaining 4 characters, comb-like gills is rejected here because there is no way the setae of trilobite-type ventral limb exopods can be homologised with the lanceolate blades or gills on the dorsal side of the dinocaridid body or pleurites. Their furcae is a misnomer there is only one furca ( fork ). A similar furca is found, for instance, in some trilobites, crustaceans and insects as well, and the significance is questionable. The only remaining character ( central body of 11 segments ) has no bearing on the relationship between dinocaridids and arthropods. Wills et al. (1998b, Fig. 2.1; see also 1998a) used 97 morphological characters to construct phyletic trees involving dinocaridids and arthropods. A 50% majority rule consensus tree is produced with equal weighting. Annelids and molluscs are used as outgroups. A first branch (that is, the smaller component of the first two sister-groups) includes the Cambrian Anomalocaris and Opabinia, a second the lobopodians (Recent onychophorans, Cambrian Aysheaia and Kerygmachela), a third the tardigrades. The sister-group of the tardigrades is the (true) Arthropoda. The authors play around with the characters, but the basal

9 branchings are stable. This sounds promising, but there are severe flaws caused by superficial similarities or dissimilarities, for instance in the placing of trilobite-like stem-crustacean agnostids with trilobites. In trees with equal weighting, this kind of mistake will necessarily concern exactly those positions where we have problems beforehand. We must remember this also regarding the dinocaridids. Bergström & Hou (2002, Fig. 2) have dinocaridids possibly related to tardigrades, and these two groups possibly related to kinorhynchs, all of them possibly being non-coelomate cycloneuralians. Other fossil and Recent lobopodians and arthropods are placed in a coelomate clade. Bergström & Hou (2004, Fig. 10) suggested a possibly common origin (poor resolution of branches) of dinocaridids, xenusians, onychophorans, and a clade with tardigrades and arthropods as sister-groups. [This paragraph is somehow modified in its Chinese version.] 5.3 Conclusions based on the morphological evidence. The contrasting results indicate a poor understanding of the nature of the dinocaridids and of their relationships. A primary lack of understanding concerns the relationships between Recent groups. Arthropods used to be regarded as coelomates related to annelids. There is hardly any reason to consider annelid segmentation as homologous to arthropod segmentation. To make judgements even more difficult, we are now uncertain whether arthropods are related to coelomates or to cycloneuralians, or to both. The same can be said about onychophorans. This affects deductions on the origin of the groups. A study of evolution within the dinocaridid cladus could possibly reveal something about their origin and relationships, even though the lack of an obvious outgroup is a dilemma. Some guesses on evolutionary directions are relevant, however (Fig ). For example, it is logical to assume that soft lobopods came before legs with a segmented exoskeleton. We can observe that Kerygmachela and probably Opabinia (Budd 1996) had soft lobopods, whereas the legs in Pambdelurion had regular, thin rings indicating sclerotization and pseudosegmentation (Fig A), and the legs in Parapeytoia had a segmented exoskeleton (Fig F; Hou et al. 1995, Figs. 9-13). Cucumericrus is intermediate, with a wrinkled leg surface (Fig E; Hou et al. 1995, Fig. 16). Where legs are not known, the anterior grasping appendages can give a hint at the evolutionary direction. They are soft-skinned in Kerygmachela (Fig E), Pambdelurion, and Opabinia (Fig C), but have a segmented exoskeleton in Anomalocaris (Fig D), Amplectobelua (Fig C), Peytoia (Fig G), and Parapeytoia (Fig F). Thus, we have a plesiomorphic condition in Kerygmachela, Pambdelurion and Opabinia, a somewhat intermediary condition in Cucumericrus, and an apomorphic condition in Anomalocaris, Amplectobelua, Peytoia and Parapeytoia. This conclusion is consistent with the distribution of body annulation. At least in Recent onychophoran lobopodians, each annulus indicates the position of a haemal channel (Robison 1985; Budd 1999a). At least if we believe in an origin in xenusian lobopodians (Fig ), Kerygmachela is apparently plesiomorphic in retaining body annulation. This genus is also similar to xenusians in having a worm-like, parallel-sided body. How do we know that all these taxa are related? There are few, if any, characters that are shared by all members (Table 2.8.1). A pair of grasping appendages is one of the best candidates, but the morphology admittedly varies widely. Another candidate for a shared structure is the circular mouth cone. Series of dorsal lanceolate blades attached to transverse bands are reported from all genera except for Pambdelurion. An additional argument for the interpretation of all these taxa as a natural phyletic group is the cumulative change in morphology along the series of genera (Fig ). This gives a general shape to the phyletic tree, but it should be noted that our knowledge has serious gaps. It is clear that the evidence from Chengjiang and other faunas indicate that a segmented and pseudosegmented walking leg evolved within the group from a soft lobopod. This is, indeed, not to say that the dinocaridids evolved from xenusian lobopodians, or from onychophorans. We must remember that parallel evolution is the rule, not the exception this is proven by virtually every cladogram (compare Bergström & Hou 2004 for compelling examples of parallelism). Still, a leading principle in cladistic practice is that similarities are regarded as synapomorphies as long as not disproved an approach that is often misleading. Given the plasticity of life, it is by no means impossible that soft podia evolved

10 more than once. In fact, we know that they did at least twice, since they exist also in the shape of echinoderm podia. Budd (1999a) identified certain lobopodian characters in Kerygmachela in addition to the lobopod. One such character is the external signs of a dorsal pericardial sinus (Budd 1999a, Figs. 7-9, 31, 33). Another is the annulation, regarded as a sign of circular haemal channels typical of onychophorans (Fig D; Budd 1999a). We agree on these similarities. The possible affiliation with the dinocaridid group is more difficult to evaluate. For instance, if there is a relationship, why did the haemal channels disappear before there was a firm exoskeleton? We are left with only few similarities (- the number of segments is hardly an argument): lobopods, pleurites, and a vaguely similar mouth region. Could they all be parallel developments? Budd (1999a) listed a number of good characters that he used in his analysis of the relationships of the dinocaridids. Two more questionable characters play an important role in keeping together the lobopodian-type Kerygmachela, anomalocaridid-like taxa, and arthropods. These are the lateral lobe and comb-like gills. In our view, these two unfortunately are based on misunderstandings. As argued above, the lateral lobe is a pleurite in Opabinia, but part of a ventral appendage in anomalocaridids. In Kerygmachela, the lobe has the extended, winglike shape typical of anomalocaridids, and it has no trace of the dorsal body ornament (Fig D). These are two arguments against interpreting the lobes as pleurites. On the other hand, the comb-like gills of Kerygmachela appear to be a striated pattern in the surface of the pleurites, with an orientation differing from that of anomalocaridid lanceolate blades, and devoid of an anomalocaridid-type rod. In anomalocaridids, the gills are lanceolate blades, attached anteriorly to a segmented, transverse rod, on the dorsal side of either the pleurites (Opabinia) or body axis (for instance, Anomalocaris, Peytoia, Parapeytoia). In arthropods, the term comb-like gills is used for the flattened setae of the ventral limbs. It may be added that, even if the lanceolate blades were homologous with the arthropod setae, the latter are found only in fairly advanced lamellipedian arthropods (Bergström & Hou 2004, Fig. 4, node 16), not in more plesiomorphic arthropods. As demonstrated in Fig , the reinterpretation of Budd s comb-like gills and lateral lobe leads to a radically different view of the relationships between dinocaridids and arthropods. The scattered distribution within the two taxa of corresponding structural details is a strong argument against a derivation of arthropods from dinocaridids. At the most, the dinocaridids and arthropods could be sistergroups, but it is problematic to find an outgroup that could help us analyse the situation. Bergström & Hou (2004, Fig. 10; see Fig herein for a simplified version) used the Cycloneuralia (represented by Rotifera, Gastrotricha, Nematomorpha, Priapulida, Kinorhyncha) as an outgroup for a possible taxon with ventral locomotory limbs (dinocaridids, xenusians, onychophorans, tardigrades, arthropods). The ingroup tree was poorly resolved, with dinocaridids, various xenusian lobopodians, onychophorans and a tardigrade-arthropod taxon as separate basal clades (Fig ). This indicates that the dinocaridids and arthropods are not sistergroups. 5.4 Results of molecular studies, and interpretation. Recent molecular studies have not confirmed the old idea that arthropods and other aiolopods are closest to annelids. Instead, the message is that aiolopods are closest to a large clade of worms usually having a cuticle and moulting their skins, the Cycloneuralia (for overview, see Nielsen 2001). Wheeler (1998) summarised recent work and presented cladograms based on morphological and nonsequence genetic data, 18S and 28S rdna, and ubiquitin in part using an insertion-deletion cost. There is a surprising stability in the trees, whether based on one or the other source, or being the result of combinations (Wheeler 1998, Figs ). Some variability is present in the position of the myriapods. The relationships can be written as [Mollusca + Annelida (Onychophora (Tardigrada (Chelicerata (Crustacea (insects and myriapods))))] [This sentence unfortunately omitted in the Chinese text.]. In some tree alternatives, myriapods turn up as a sister-group of the Chelicerata. This tree is consistent with our classical biological understanding. However, neither cycloneuralians nor deuterostomians were included.

11 This was done by Eernisse (1998) and Peterson & Eernisse (2001). Their molecular cladograms, based on 18S rrna, can be summarised roughly as follows: [most flatworms (Rotifera + Acanthocephala + flatworms (Gastrotricha + Mesozoa + Acoela (Eutrochozoa (Deuterostomia (Scalidophora (Nematoida (Arthropoda)))))]. The Eutrochozoa are taken to include the Nemertea, Annelida, Sipunculida, Bryozoa, Echiura, Entoprocta, Brachiopoda, Phoronida, Pogonophora and Mollusca. The Deuterostomia include the Enteropneusta, Echinodermata and Chordata. The Scalidophora include the Priapulida, Kinorhyncha and Loricifera. The Nematoida include the Nematoda, Nematomorpha and Chaetognatha. The Tardigrada show up either with the Arthropoda or with the Nematoida, and the Onychophora within the Arthropoda. We have considered the biological evidence to see the evolutionary consequences of the relationships, if they are those indicated by the molecular studies. There are three great differences between the molecular tree and the conventional evolutionary tree. First, the close relationship between cycloneuralian worms and aiolopods, that has given rise to the concept of the Ecdysozoa, the moulting animals (- we should recall that aiolopods are regarded as coelomates, cycloneuralians as pseudocoelomates). Second, the derivation of these groups from coelomate ancestors. Third, the position of the deuterostomians clearly within the coelomates rather than as a basal, isolated group. If this kind of molecular tree is close to the true condition, it is clear that the ecdysozoans, including the coelomate aiolopods and noncoelomate cycloneuralians, were derived from typical coelomates. Such a coelomate would be worm- or slug-like. It would have a circulatory system distributing blood throughout the body, with main arteries and veins in the sagittal plane and a dorsal heart pumping the blood forwards. It would have a primary pelagic larva provided with external cilia for propulsion and feeding. This would mean that the ecdysozoan branch started with reductions. Thus, both aiolopods and cycloneuralians lost their cilia and the ciliated larva. In addition, cycloneuralians lost all the other coelomate characters, perhaps as a result of miniaturisation. Instead, a cuticle was developed, and moulting, perhaps independently in the two groups. It may be noted that a comparable reorganisation probably occurred at the root of the deuterostomians. Basal members are ciliaryfeeders and indicate an origination through pedomorphy, in which the adult retained the larval feeding method. A second unconventional feature in the cladograms produced by Wheeler (1998) and Peterson & Eernisse (2001, Figs. 2, 7B) is that the deuterostomians are closer to the ecdysozoans (Aiolopoda + Cycloneuralia) than are, for instance, annelids and molluscs. This is in accord with a new understanding of the deuterostomians. It has been shown that a misunderstanding causes the odd deuterostomian characters. Vertebrates have stood model for deuterostomian orientation dorsal and ventral are no obvious features in echinoderms and tunicates. Nübler-Jung & Arendt (1994) noted that deuterostomians (their generalisation) are upside-down in comparison with protostomians. Already earlier, Malakhov (1977) had noted that chordates are upside-down, but not hemichordates. After a study of live acranians, Bergström (1997) and Bergström et al. (1998) concluded that it is only vertebrates that live upside-down, although acranian chordates and urochordate larvae are also shown upside-down in the textbooks. When the vertebrate embryo is compared upside-down with other embryos, the mouth and other structures do not develop where they are expected hence the idea that there are deuterostomians, animals which lost their primary mouth and developed a new one. Deuterostomians are in many respects typical coelomates. In fact, the enteropneust larva is more similar to annelid trochophora larvae than are most protostomian larvae (Bergström et al. 1998). Provided that the molecular trees referred to above are reasonably correct, we can thus state that the original radiation of some major animal groups must have been associated with fundamental body reorganisations, including reductions and simplifications, that gave starting-points for major animal clades. This model is different from the old one, which saw a steady growth of complexity as the norm. As noted above, onychophorans and arthropods have retained a number of basic coelomate characters, which have been lost in

12 cycloneuralian worms. The presence of a pericardial sinus in the Cambrian Kerygmachela (Budd 1999a, Figs. 4, 7-9) demonstrated the existence of a coelomate circulatory system in this genus. It is therefore crystal clear that aiolopods did not evolve from cycloneuralian worms. It does not rule out a common ancestry. It also does not rule out the possibility that cycloneuralian worms evolved from an aiolopodan origin, but such an alternative appears to produce unnecessary complications. The Cambrian lobopodians, the xenusians (Fig ), probably represent an evolutionary lawn from which evolved dinocaridids, onychophorans, arthropods and perhaps tardigrades. 6. Conclusions and summary [Basal groups with lobopod legs are well represented in the Chengjiang fauna. This is a circumstance that makes this fauna uniquely important for the interpretation of the earliest radiation of legged animals. Omitted?] The dinocaridids are represented in the Chengjiang fauna by four formal genera, more than half of all known genera. One group, with Anomalocaris, Amplectobelua (Fig B) and Cucumericrus, has grasping appendages with two spine rows and a neat vein pattern on the lateral lobes. Parapeytoia (Fig A, 2.8.2F) belongs to a group with a single row of spines and without lateral lobe patterning. The Chengjiang material of dinocaridids has revealed the presence of ventral appendages consisting of a proximal propod and a distal walking limb that is either soft or skeletonized. The so-called lateral lobe, previously interpreted as a fin used in locomotion, are reinterpreted as a swim flap formed by lateral expansion of the propod. The sloping lateral lobe in the Burgess Shale genus Opabinia is a pleurite. Narrow, transverse, segmented bands have been interpreted as ventral supports for the lateral lobes or fins (Collins 1996). A transverse striated pattern has been interpreted as showing gills homologous to the flat setae on trilobite exopods. We could show that the striae correspond to lanceolate blades attached to the segmented bands, and that these structures are dorsal rather than ventral. In anomalocaridids, which lack pleurites, the bands extend over the body midline, whereas in Opabinia they may be discontinuous in the dorsal mid-line. Strongly sclerotised parts include the grasping appendages, the peytoiid mouth circle, and the ventral side between the appendages. The pleurites were sclerotised enough to be stiff, as were the lanceolate dorsal blades. Opabinia has a somewhat different solution, with a soft shaft to the grasping appendages, and sclerotised tergites covering the back. In-group evolution includes a shift from probably soft-skinned lobopods to fully segmented and sclerotised legs. The origination of the group appears demarcated by the shift from walking to swimming, with the development of flat paddles on the ventral appendages. Arthropod-like characters originated in an order different from that in true arthropods, and it is therefore impossible to derive one group from the other (Fig , numbers). The appendages indicate an origin from lobopodians. There are at least superficial similarities with Kerygmachela; however, the grasping appendages appear to be ventral rather than lateral. A majority of the known genera of Cambrian lobopodians are from the Chengjiang fauna. It is most likely that they represent a wide evolutionary bush that gave rise to other types of animals with locomotory limbs (Fig ). It is difficult to get a firm hold on the evolutionary pattern, both since characters are somewhat diffuse and since they are distributed in a mosaic way. Some are known from a single species. There is a possible group with Aysheaia and Onychodictyon (Fig D and G), defined on the presence of an antenna-like appendage set on the side of the head rather than on the underside. This is potentially interesting, since such an appendage is also present in modern onychophorans and arthropods. Anyway, the essential message is that legged animals probably are all derived from a lineage that had radiated wildly already at the time of the Chengjiang biota but there we see it early imprints. Morphological and molecular studies have tended to give strongly different and variable answers to the question where in the evolutionary tree legged animals took their origin. Whereas the dominant morphological wisdom has indicated a relationship with annelids or molluscs, many recent molecular trees have placed arthropods and soft-legged animals close to cycloneuralian worms such as nematodes and priapulids, but also close to molluscs and deuterostomians. A problem is that cycloneuralians on one hand, molluscs and