Cricocosmiidae is a clade of palaeoscolecid-like worms from the Chengjiang Biota, China (Cambrian Stage 3). In contrast with palaeoscolecids sensu stricto, which exhibit tessellating microplate trunk ornamentation, cricocosmiids have larger, serially repeated sets of trunk sclerites bearing a resemblance to lobopodian trunk sclerites (e.g. Microdictyon spp.). Cricocosmiidae have therefore been proposed as stem-group Panarthropoda in some studies, but are recovered as stem-group Priapulida in most phylogenetic analyses. The affinity of cricocosmiids within Ecdysozoa is therefore of much interest, as is testing the homology of these seriated structures. We report four new specimens of the rare cricocosmiid Tabelliscolex hexagonus, yielding new details of the ventral trunk projections, sclerites and proboscis. New data confirm that T. hexagonus had paired ventral trunk projections in a consistent seriated pattern, which is also reported from new material of Cricocosmia jinningensis (Cricocosmiidae) and Mafangscolex yunnanensis (Palaeoscolecida sensu stricto). Even when the seriated sclerites and ventral projections of cricocosmiids are coded as homologous with the seriated trunk sclerites and paired appendages, respectively, of lobopodian panarthropods, our tree searches indicate they are convergent. Cricocosmiidae is nested within a monophyletic ‘Palaeoscolecida sensu lato’ clade (Palaeoscolecidomorpha nov.) in stem-group Priapulida. Our study indicates that morphological seriation has independent origins in Scalidophora and Panarthropoda.

Supplementary material: Phylogenetic character list and matrix are available at https://doi.org/10.6084/m9.figshare.c.5551565

Thematic collection: This article is part of the Advances in the Cambrian Explosion collection available at: https://www.lyellcollection.org/cc/advances-cambrian-explosion

Palaeoscolecida Conway Morris and Robinson, 1986 is a group of marine benthic ecdysozoan worms with a spiny anterior proboscis, an elongate trunk with striking ornamentation of small, annular-arranged, tessellating polymorphic plates, and bilaterally orientated terminal posterior hooks (Harvey et al. 2010). Cricocosmiidae Hou et al., 1999 are otherwise morphologically similar to palaeoscolecids, but exhibit a different, although also highly characteristic, trunk ornamentation. The cricocosmiids Cricocosmia jinningensis Hou and Sun, 1988 and Tabelliscolex hexagonus Han et al., 2003a possess larger sclerites that are serially repeated in groups along the length of the trunk, with a striking similarity in structure, position and composition to the trunk sclerites of some lobopodian panarthropods such as Onychodictyon spp. and Microdictyon spp. (Han et al. 2007a). Cricocosmiids and a few other vermiform taxa such as Maotianshania cylindrica Sun and Hou, 1987 and Tylotites petiolaris Luo et al., 1999 have been informally grouped with palaeoscolecids as ‘Palaeoscolecida sensu lato’; in this framework, ‘Palaeoscolecida sensu stricto’ is reserved exclusively for those taxa bearing the distinctive polymorphic tessellating plate ornament (see Harvey et al. 2010). Palaeoscolecida sensu stricto has a broad biogeographical and stratigraphic range from the early Cambrian to the late Silurian, whereas Cricocosmiidae, Maotianshania and Tylotites (i.e. Palaeoscolecida sensu lato) are known only from the Cambrian Stage 3 Chengjiang Biota of Yunnan, China.

Individual plates of Palaeoscolecida sensu stricto comprise the form-genera Milaculum, Hadimopanella, Kaimenella and Utahphospha (see Hinz et al. 1990). These small shelly fossils are known from the Cambrian to Silurian across a wide geographical range and are sometimes recovered in association with the scleritome (Hinz et al. 1990; Brock and Cooper 1993; Müller and Hinz-Schallreuter 1993; Zhang and Pratt 1996; Conway Morris 1997; Harvey et al. 2010; Topper et al. 2010; Duan et al. 2012; Streng et al. 2017). Individual plates of the form-genus Hadimopanella have also been identified as small carbonaceous fossils, alongside possibly conspecific pharyngeal teeth, introvert scalids and terminal posterior hooks from the early Cambrian of Scandinavia (Slater et al. 2017). Some small shelly fossil cuticle fragments even preserve in situ the posterior terminal hooks of Palaeoscolecida sensu stricto (e.g. Harvey et al. 2010, their fig. 2e–g). Cricocosmiids remain unidentified from microfossil assemblages. Articulated two-dimensional compression fossils from Konservat-Lagerstätten demonstrate the complete cuticular morphology and partial soft tissue anatomy of both groups, revealing many similarities. Palaeoscolecida sensu stricto described in detail from such articulated material include Scathascolex minor (Smith 2015), Wronascolex antiquus (García-Bellido et al. 2013) and Wronascolex yichangensis (Yang and Zhang 2016), Mafangscolex yunnanensis (Hou and Sun 1988; Luo et al. 2014; Yang et al. 2020); Guanduscolex minor (Hu et al. 2008) and Utahscolex ratcliffei (Whittaker et al. 2020). The best preserved of these Palaeoscolecida sensu stricto exhibit a three-part proboscis divisible into the traditional zonation system common to extant and fossil priapulans (Conway Morris 1977). Zone I is represented by an introvert with rings of scalids, Zone II is the collar – representing a diastema between introvert and pharynx – and Zone III is the eversible pharynx with rings of teeth. This is mirrored by the sometimes exquisitely preserved proboscis morphology of C. jinningensis and M. cylindrica (Huang 2005) and is partially known in T. hexagonus (Han et al. 2003a) and T. petiolaris (Han et al. 2003b, 2007b).

Palaeoscolecid-like worms have an interesting and varied history of phylogenetic interpretation. Palaeoscolecida was originally attributed to Annelida when first described (Whittard 1953), but later researchers favoured an affinity among the cycloneuralian phyla as a result of the annulated trunk region (Dzik and Krumbiegel 1989; Conway Morris 1993), especially after the discovery of the anterior proboscis (Hou and Bergström 1994). Alternative cycloneuralian interpretations of palaeoscolecids have been postulated, the group being allied to either nematomorphs (Hou and Bergström 1994) or priapulans (Conway Morris 1997) or as stem-group Cycloneuralia (Zhuravlev et al. 2011). Another interpretation is that palaeoscolecids represent a basal branch within Ecdysozoa, which was popularized by the influential hypothesis of Dzik and Krumbiegel (1989), wherein palaeoscolecids represent a transitional locomotor strategy linking priapulans and lobopodians. This was advocated by several researchers (Budd 2001a, b, 2003; Budd and Jensen 2003; Zrzavý 2003; Webster et al. 2006) on the basis that palaeoscolecids appear to possess a suite of characters that are expected in the ancestral ecdysozoan, such as an annulated cuticle and an armoured proboscis. Similarities between palaeoscolecids and priapulans were therefore viewed as ecdysozoan plesiomorphies by these researchers. This reconstruction of the ancestral ecdysozoan has been corroborated to a degree by ancestral state reconstructions (Howard et al. 2020a), but morphology-based phylogenetic analyses consistently retrieve a stem-group Priapulida affinity for palaeoscolecid-like worms (Wills 1998; Dong et al. 2004, 2005, 2010; Harvey et al. 2010; Wills et al. 2012; Liu et al. 2014; Ma et al. 2014a; Zhang et al. 2015; Shao et al. 2016). As such, Harvey et al. (2010) dismissed the idea that palaeoscolecid-like worms are basally diverging ecdysozoans. Nevertheless, a recent study once again supported the basal-ecdysozoan interpretation after reporting hexaradial symmetry of the rings of introvert scalids (i.e. circumoral armature) in a member of Palaeoscolecida sensu stricto (Yang et al. 2020). Hexaradial symmetry of the anterior armature is interpreted as plesiomorphic for Ecdysozoa, in contrast with the pentaradial symmetry of crown-group Priapulida (see Liu et al. 2014).

Cricocosmiids, although mostly recovered within stem-group Priapulida as Palaeoscolecida sensu lato in the phylogenetic studies cited here, are also of great phylogenetic interest. Despite their major similarities to Palaeoscolecida sensu stricto, cricocosmiids have been implicated in hypotheses of panarthropod origins. This is because they exhibit a dorso-ventrally differentiated trunk, equipped with serially repeated sclerites that bear considerable similarity to some lobopodian trunk sclerites in both position and composition – thus prompting hypotheses of homology (Han et al. 2007a; Steiner et al. 2012). As such, cricocosmiids are of much evolutionary significance, which has not been explored in the context of ecdysozoan-wide phylogeny. Furthermore, it is unclear which palaeoscolecid-like taxa actually belong in Cricocosmiidae beyond Cricocosmia and Tabelliscolex, and what the exact phylogenetic relationship is between Cricocosmiidae and Palaeoscolecida sensu stricto.

Cricocosmia jinningensis is one of the most common Chengjiang taxa, known from thousands of specimens (see Huang 2005; Hou et al. 2017). However, material and documentation of Tabelliscolex are far sparser, with only one complete specimen of T. hexagonus described (Han et al. 2003a) and a single incomplete specimen of T. chengjiangensis apparently distinguished by sclerite morphology (Han et al. 2007a). Tylotites petiolaris is a similarly scarce palaeoscolecid-like worm from the Chengjiang Biota (Luo et al. 1999; Han et al. 2003b, 2007b), also with serially repeated large sclerites, suggesting a cricocosmiid affinity, although the sclerites are spinose and arranged in annular rings, in contrast with the paired lateral sclerites of Cricocosmia and Tabelliscolex.

We present here new material of T. hexagonus from the Chengjiang Biota, preserving new anatomical details in high fidelity. These specimens facilitate a re-description of Tabelliscolex, and comparison with new material of Cricocosmia and Mafangscolex provides insight into the evolution of the palaeoscolecid-like worms. We conducted an extensive phylogenetic analysis of Ecdysozoa to better resolve the systematics of ‘Palaeoscolecida sensu lato’ (sensu Harvey et al. 2010). Our results facilitated a discussion of the evolution of seriated morphology in Ecdysozoa, suggesting that seriation independently evolved in Panarthropoda and Scalidophora.

Geological setting

All the specimens studied here belong to the Yunnan Key Laboratory for Palaeobiology (YKLP) collection and were collected from various locations in the Chengjiang Biota (see Hou et al. 2017), eastern Yunnan, SW China. The Chengjiang Biota is found within the Maotianshan Member of the Yu'anshan Formation at Cambrian Series 2, Stage 3 (c. 518 Ma). The hypothesized location of the Chengjiang Biota during the early Cambrian is near the shore of the South China continent, which was close to the equator (see Hou et al. 2017; Holmes et al. 2018).

Fossil material

Three specimens of Tabelliscolex were collected from Haikou County (all with counterparts; see Figs 1a–e, g, h,2, 3, 4a–c and 5g–i) and an additional specimen missing the counterpart from Anning City (Fig. 1f). One specimen (YKLP 11428) is complete with proboscis and armature, and all specimens show details of the ventral trunk projections. A total of 7131 specimens of C. jinningensis and 703 specimens of M. yunnanensis were also investigated for comparative analysis of the ventral projections (Fig. 5). Details of new palaeoscolecidomorph material reported in this study are listed in Table 1. All specimens are deposited in the Yunnan Key Laboratory for Palaeobiology, Yunnan University, Kunming, China.

All specimens are preserved in fine-grained yellowish mudstone. All are essentially flattened compressions in a lateral, dorsolateral or ventrolateral perspective, although retaining a low degree of three-dimensional relief in some structures, such as the sclerites, scalids and gut tract. The preservation of the trunk sclerites in Tabelliscolex and Cricocosmia is similar to that investigated by previous researchers (see Han et al. 2007a; Steiner et al. 2012). Dark carbon films represent gut tracts. The observations, preparation and camera lucida drawings were performed under a Nikon SMZ800N microscope. Photographs were taken using a Leica M205 microscope and a Canon EOS 6D Mark II digital camera equipped with either a 100 or 65 mm macro lens. A Keyence VHX-6000 3D imaging microscope was used to create topographic models of the ventral projections in Tabelliscolex (Fig. 6) and sclerites in various taxa were imaged under an FEI Quanta 650 scanning electron microscope (Fig. 4).

Phylogenetic methods

Phylogenetic analyses were performed to resolve the position of palaeoscolecid-like worms within Ecdysozoa (Fig. 7). The tree searches used four alternative optimality criteria (equal-weights parsimony, implied-weights parsimony, maximum likelihood and Bayesian inference) to account for the debate over the most suitable method to analyse discrete morphological characters (O'Reilly et al. 2016, 2018a, b; Puttick et al. 2017, 2019; Goloboff et al. 2018a, b).

Parsimony tree searches were conducted in TNT 1.5 (Goloboff et al. 2008; Goloboff and Catalano 2016) using the New Technology Search function with default settings. A strict consensus of the six most parsimonious trees is presented from equal character weighting (Fig. S1) and clade support was assessed by jack-knife resampling (Fig. S1; Table 2) (Farris et al. 1996). A strict consensus of the four most parsimonious trees (Fig. S2) is presented for implied character weighting (using the default concavity constant k = 3) and clade support was assessed by symmetrical resampling (Fig. S2) (Goloboff et al. 2003). Consistently recovered nodes across parsimony analyses are summarized in Figure 7b and c. Character transformations within Palaeoscolecidomorpha were optimized in a parsimony context using WinClada (Nixon 2002) (Fig. 7c).

Probabilistic tree searches used the MK model for discrete morphological character data (Lewis 2001). The maximum likelihood implementation was conducted in IQ-Tree (Nguyen et al. 2015) (Fig. 7a), with nodal support assessed by 1000 Ultrafast bootstrap (UFBoot) replicates (Minh et al. 2013; Hoang et al. 2018). The Bayesian implementation was conducted in MrBayes 3.2 (Ronquist et al. 2012). The Bayesian analysis was run until convergence of the MCMC (Markov chain Monte Carlo) chains after 6 000 000 generations. Convergence was assessed from the average standard deviation of split frequencies (<0.01), ESS (Effective Sample Size) scores (>200) and PSRF (Potential Scale Reduction Factor) values (c. 1.00). Twenty-five per cent of the samples were discarded as burn-in and a majority rule consensus was output (Fig. S3).

Systematic palaeontology

Phylum (stem-group) Priapulida Delage and Hérouard, 1897 

Clade Palaeoscolecidomorpha nov. (=Palaeoscolecida sensu

lato in Harvey et al. 2010)

Included taxa:

Palaeoscolecida Conway Morris and Robinson, 1986 (=Palaeoscolecida sensu stricto in Harvey et al. 2010)

Cricocosmiidae Hou et al., 1999 (emended to include Tylotites Luo et al., 1999)

Maotianshania Sun and Hou, 1987 

Markuelia Val'kov, 1983 

Diagnosis: Annulated worm-like body with an anterior armoured proboscis, an elongate trunk of consistent width and terminal posterior hooks. Anterior proboscis divided into three sections from proximal to distal: introvert with circumoral longitudinal rows of spines (presumed scalids); collar region with or without spines; and pharynx covered in rings of pharyngeal teeth. Introvert at least partially retractable into the trunk, pharynx elongate and fully eversible. Posterior terminal hooks arranged bilaterally about the sagittal plane of the trunk.

Remarks: The division of the proboscis is likely to be a plesiomorphy for Priapulida (see Conway Morris 1977). The reliable autapomorphies of Palaeoscolecidomorpha are the elongate body profile with consistent trunk width (i.e. the trunk does not vary in width throughout the length of the body and the lateral margins are virtually parallel) and the terminal posterior hooks. The number of hooks may vary across taxa, but are always arranged bilaterally, with a left hook and a right hook (or a left and right pair). Serially repeated paired ventral trunk projections are present in select Cricocosmiidae and Palaeoscolecida (Figs 5 and 8) and represent another possible autapomorphy for Palaescolecidomorpha.

Family Cricocosmiidae Hou et al., 1999 

Included genera:

Cricocosmia Hou and Sun, 1988 

Tylotites Luo and Hu, 1999 

Tabelliscolex Han et al., 2003a 

Revised diagnosis: Palaeoscolecidomorphs with serially repeated arrangements of macroscopic trunk sclerites and corresponding paired ventral projections (at least in Cricocosmia and Tabelliscolex), a single pair of terminal posterior hooks and a posterior unarmed region of the introvert. Trunk sclerites composed of many tubercles of equal size on the outer surface with corresponding pits on the inner surface (at least in Cricocosmia and Tabelliscolex).

Remarks: In the previous revision of Cricocosmiidae (Han et al. 2007a), Tylotites was not included, but our phylogenetic analysis recovered Tylotites within this clade (see Fig. 7). Therefore the family Tylotitidae Han et al., 2003b is a synonym of Cricocosmiidae. The serially repeated trunk sclerites are distinct in shape and arrangement in each genus, but their presence is autapomorphic for Cricocosmiidae. Tylotites lacks data on sclerite fine structure, so the presence of tubercles and corresponding pits is unknown. Tylotites sclerites differ considerably from Cricocosmia and Tabelliscolex in arrangement, being long conical spines along transverse rings. Cricocosmia exhibits lateral pairs and Tabelliscolex exhibits lateral pairs with an additional dorsal sclerite for each pair. Tylotites and Cricocosmia are sister taxa in our parsimony and maximum likelihood analyses (see Figs 7, S1 and S2) and share the synapomorphy of an acute distal termination to their sclerites (i.e. an overall spinose shape with a distinct apex), although in Tylotites the spines are elongated. The presence of paired ventral projections is unknown in Tylotites, but likely considering this character is present in Cricocosmia, Tabelliscolex and Mafangscolex (Palaeoscolecida), and therefore is likely to be a plesiomorphy for Cricocosmiidae. The posterior unarmed introvert region (i.e. a smooth region between the Zone I armature and the trunk annulations) is present in all cricocosmiids and also in Maotianshania (see Huang 2005). This character is also possibly present in Markuelia (see reconstruction in Dong et al. 2010), but it is inferred from coiled late-stage embryos and is therefore less reliable. Regardless, Maotianshania and Markuelia are resolved proximally to Cricocosmiidae to the exclusion of Palaeoscolecida (see Fig. 7), which appear to lack this character (e.g. Smith 2015, figs 1a, h, i and 3; García-Bellido et al. 2013, fig. 3a; and Yang et al. 2020, figs 1a, b and 2o). The phosphatic cuticle fragment taxon Houscolex Zhang and Pratt, 1996, from the early Cambrian of Shaanxi, was also referred to Cricocosmiidae by Han et al. (2007a). Although it possesses ventral projections resembling those of Cricocosmia and Mafangscolex, it lacks large sclerites, and accordingly is excluded from the Cricocosmiidae as delimited here.

Genus Tabelliscolex Han et al., 2003a 

Type species:Tabelliscolex hexagonus Han et al., 2003a 

Diagnosis: (emended from Han et al. 2003a; Huang 2005) Cricocosmiid worm with two lateral rows of ellipsoidal sclerites and a single corresponding dorsal row of ellipsoidal sclerites. Typical three-zoned proboscis exhibits at least seven circlets of elongate spines on the introvert (Zone I), stout triangular spines on the collar (Zone II) and many circlets of short, spinose pharyngeal teeth (Zone III). Paired ventral projections are preserved as teardrop-shaped from above, or spine-shaped in lateral profile, curving posteriorly, and are present beneath every third set of sclerites.

SpeciesTabelliscolex hexagonus Han et al., 2003a 

2003, Tabelliscolex hexagonus Han et al. 2003a, plate 1

2005, Tabelliscolex hexagonus Huang 2005, plate 46

2007, Tabelliscolex hexagonus Han et al. 2007a, plates 1, 4 and 6

2007, Tabelliscolex chengjiangensis Han et al. 2007a, plates 1, 4

and 6, syn. nov.

Diagnosis: As for the genus.

Holotype: Eli-0001218A, Han et al. 2003a, plate 1.1.

Paratypes: Eli-0001219A, Eli-0001220A, Eli-0001221A and Eli-0001222A, Han et al. 2003a, plates 1.2–5.

Description: Two of the four new specimens preserve the full length of the trunk: YKLP 11428 and YKLP 11431 (Figs 1a–c, g and h and 2a, b). YKLP 11428 measures 7.9 cm in length and up to 0.3 cm in width, whereas YKLP 11431 measures 3.4 cm in length and up to 0.2 cm in width, indicating either different stages of maturity or intraspecific variation. The body is slim, cylindrical and divided into an anterior proboscis, annulated trunk with net-like ellipsoidal sclerites and a pair of terminal posterior hooks on the trunk.

Proboscis: YKLP 11428 shows the completely everted proboscis (Fig. 3) and facilitates a more detailed description. The proboscis is subdivided into three sections (Zones I–III) from anterior to posterior in typical fashion for palaeoscolecidomorphs (introvert, collar and pharynx). The width of the everted introvert is almost equal to the width of the trunk. The length of the fully everted pharynx reaches c. 5 mm and the pharynx features numerous circlets of tiny spinose teeth. The pharynx consists of three sections: section 1 is c. 2.2 mm in length and is slightly wider than the collar, covered with teeth pointing anteriorly; section 2 is swollen, 1.1 mm in length and also covered with teeth pointing anteriorly; and section 3 is at least 1.6 mm in length with a uniform width and covered with teeth pointing laterally (see Fig. 3b). The collar is roughly conical and ornamented by stout, triangular spines. The collar spines sit immediately below the everted pharynx and are c. 0.1 mm in length. The introvert follows the collar with a clear boundary and is ornamented by elongate curved spines (presumably scalids). The spines are arranged circumorally, forming longitudinal rows. From anterior to posterior, each longitudinal row of spines becomes shorter in length. The spines of the first three circlets are c. 0.7 mm in length and are elongate and curved with sharp tips (Fig. 3); the spines in the following circlets become shorter and their lengths are about one-third that of the scalids in the first three circles, with at least seven circlets in total. The fidelity of preservation is not sufficient to determine the exact number of spines per row or their mode of radial symmetry.

The detailed morphology of the proboscis in T. hexagonus remains incomplete, but is improved by the present study. Our observations show that YKLP 11428 appears to exhibit three distinct fields of pharyngeal teeth, as in C. jinningensis (see Huang 2005). However, C. jinningensis appears to have a markedly different tooth morphology and arrangement in those distinct fields, but this cannot be confirmed or rejected in T. hexagonus due to the quality of preservation. The pharynx of the putative cricocosmiid T. petiolaris (see Han et al. 2007b) is too poorly preserved/figured to be compared for this character. Other palaeoscolecidomorphs with detailed available pharyngeal morphology data do not share this configuration. The pharynx of M. cylindrica certainly differs, with simple uniform spinose teeth covering most of it, and some more elongate basal teeth (Huang 2005). Mafangscolex yunnanensis exhibits two fields of teeth with differing arrangements, but seemingly uniform morphology (Yang et al. 2020), and S. minor exhibits just one field with cusped morphology (Smith 2015).

Trunk: The annulated trunk is cylindrical and elongate with no discernible subdivision, with about four annulations per millimetre. YKLP 11428 shows that the annulations in the anterior-most portion of the trunk are denser and lacking in sclerites (Fig. 2a). T. hexagonus exhibits a single bilateral pair of hooks at the posterior termination of the trunk, each hook being curved and bent laterally with a broad base.

Sclerites:T. hexagonus bears 110–120 sets of serially repeated sclerites along the majority of the length of the trunk. Each set is composed of three ellipsoidal flat sclerites arranged transversely (one dorsal, two lateral). The dorsal sclerite is equal in size to its corresponding lateral sclerites (Figs 2c, d and 4a). The length to width ratio of the sclerites decreases gradually from anterior to posterior (Fig. S4). Each sclerite is ornamented with outer tubercles (c. 120) and corresponding inner pits. The pit/tubercle diameter ranges from 20 to 30 μm. There is a clear and smooth gap between each set of sclerites, with boundaries running either side of the gap transversely across the trunk to the body margin, which is best seen in some parts of YKLP 11428 and 11429 (Figs 1a, b, d and 8c; also seen in Han et al. 2003a, plate I, fig. 1d, and Huang 2005, fig. 46a–d). If each set of sclerites is assumed to be situated in the middle of every annulation, as shown in the reconstructions of Huang (2005, fig. 45) or Han et al. (2007a, fig. 6b), then the annulation boundaries should run across the middle of the gap between each set of sclerites, but this is not observed. Therefore each of these gaps is interpreted as a non-sclerite-bearing annulation and sclerite sets occur on every two annulations (Figs 1a, b, d and 8c). It appears that the sclerite-bearing annulations are slightly narrower than those without sclerites, indicating heteronomous annulation, which is known from varied ecdysozoan taxa.

Ventral projections: Pairs of projections are serially repeated along the length of the ventral trunk (see Figs 2 and 5g–i), seemingly occurring beneath every three sets of sclerites (Fig. 8c). The pair is usually preserved as teardrop-shaped, with one convex projection and one concave projection (Fig. 6). The depth–height ratio of the ventral projections ranges from 25 to 90 μm and the length of the projection can reach up to 0.5 mm. Laterally compressed projections reveal that these ventral projections are spine-shaped in profile, curving posteriorly, with a strong base (Fig. 5i).

Remarks: Han et al. (2007a) described a second species, Tabelliscolex chengjiangensis, differing from T. hexagonus by having concentric lamina in some pits and possibly lacking dorsal sclerites. No complete nor articulated fragmentary specimen of T. chengjiangensis has yet been reported. This additional species is described from suspected moulted sclerites with a concentric lamina pit interpreted as absent in the type species T. hexagonus. In general, a concave structure is more likely to gather clay, pyrite particles and other minerals in late diagenesis, and therefore the concentric lamina may not be primary biogenic structures. As such, the moulted sclerites described as T. chengjiangensis may be synonymous with T. hexagonus; more specimens are needed to confirm the validity of T. chengjiangensis.

New observations in Cricocosmia and Mafangscolex

In addition to T. hexagonus, we report similar paired ventral structures in M. yunnanensis (Palaeoscolecida sensu stricto) and C. jinningensis (Cricocosmiidae). The ventral projections of Mafangscolex are spike-like (Fig. 5b), 0.20–0.25 mm in length, with an expanded base (Fig. 5c), occurring roughly every two annulations (Figs 5a, b and 8a). The ventral projections of Mafangscolex occur on the median zone between two transverse plate bands. The ventral projections of Mafangscolex are perpendicular to the body wall (Fig. 5b). The shape of the ventral projections of Cricocosmia appears similar to those of Mafangscolex in dorsal view, but there are marked differences. Cricocosmia ventral projections correspond to paired lateral sclerites per annulation and, in lateral view, it can be observed that the tip of the ventral projection of Cricocosmia points posteriorly (Fig. 5d–f). The ventral projections of Tabelliscolex are clearly larger than those of Cricocosmia and Mafangscolex, are preserved in high relief (Fig. 6) and overall are more similar in shape to those of Cricocosmia than Mafangscolex.

Phylogenetic characters

The phylogenetic character matrix (included in NEXUS format as a supplementary data file) was used in all tree searches and consisted of 95 fossil and extant ecdysozoan taxa scored for 179 morphological characters. This character sample derives primarily from two previous matrices: the cycloneuralian-focused matrix of Harvey et al. (2010) and the panarthropod-focused matrix used and updated in Smith and Ortega-Hernández (2014), Smith and Caron (2015), Yang et al. (2015) and Zhang et al. (2016). The full list of character descriptions is included as an appendix in the Supplementary Material. The following newly defined/modified characters are of specific relevance to the present study (i.e. pertaining to the seriated structures common to palaeoscolecid-like worms and lobopodians) and are listed in the order in which they appear in the character list.

Trunk with serially repeated paired ventral/ventrolateral structures (absent or present)

This character suggests the homology of the paired ventrolateral trunk lobopods of panarthropods with the paired ventral projections in Mafangscolex, Cricocosmia and Tabelliscolex. This is justified by the consistent metameric pattern of the ventral projection pairs relative to the trunk annulations (Fig. 5) and, in the case of Cricocosmia and Tabelliscolex, relative to lateral/dorsolateral sclerites shared by lobopodians (see reconstructions in Fig. 8). As such, the projections are consistent with lobopods in topological position, occur in a serially repeated sequence and share a metameric relationship to lateral/dorsolateral sclerites with many lobopodian taxa. Tylotites is coded as unknown(?) because it has serially repeated ventral spines (Han et al. 2007b), but it is unknown whether these are distinct paired structures that can be discriminated from the transverse rings of spinose sclerites that are unique to this taxon.

An alternative to this character formulation would draw a homology between the ventral projections of palaeoscolecidomorphs and the claws of panarthropods. We regard this as less plausible than a homology with the lobopod/appendage as a whole for the following reasons. First, in terms of character ontology, the whole–part relationship between an appendage and its claw is appendage (whole)–claw (part), so the claw evolving before the appendage violates the character ontology, whereas a proposal that appendages existed in palaeoscolecidomorph ancestry, but were lost apart from the claws, is ad hoc. Second, evidence for claws – a recalcitrant morphological element (Murdock et al. 2016) – is missing from many lobopodian taxa, challenging an assumption that claws are a panarthropod plesiomorphy. Some taxa are missing claws, including Diania (see Ma et al. 2014b) and Xenusion (see Dzik and Krumbiegel 1989), which are potential stem-group Panarthropoda according to some of our phylogenetic analyses. Third, the ventral projections of palaeoscolecidomorphs are not sclerotized, but rather are preserved in the same colour and style as the body wall, whereas sclerotized claws are darker and more opaque than the lobopod. In addition, claws are usually paired (except Aysheaia, Hallucigenia hongmeia and Luolishaniidae), whereas the ventral projections are single: one left and one right.

Form of serially repeated paired ventral/ventrolateral trunk structures (simple projections or limbs)

This character discriminates the paired spine-like projections of Mafangscolex, Cricocosmia and Tabelliscolex from lobopods. Although the paired ventral projections of the worms mirror those of lobopods in a topological context, lobopods are markedly different in their more flexible, conical form, with a broad circular attachment to the body wall (Whittington 1978; Hou et al. 2004; Liu et al. 2008; Howard et al. 2020b).

Serially repeated epidermal specializations (absent or present)

In addition to many lobopodian taxa, this is coded present for Cricocosmia (as in Yang et al. 2015) and Tabelliscolex, which have serially repeated pairs/triplets of lateral/dorsolateral sclerites (Han et al. 2007a), and Tylotites, which has serially repeated transverse rings of spinose sclerites (Han et al. 2007b). Mafangscolex and other palaeoscolecid-like worms are coded as absent.

Position of serially repeated epidermal specializations (longitudinal rows, incomplete transverse rings or complete transverse rings)

In most lobopodians, Cricocosmia and Tabelliscolex, the serially repeated epidermal specializations (e.g. nodes, spines and sclerites) are arranged in longitudinal rows in a dorsal, lateral or dorsolateral perspective. Tylotites (see Han et al. 2007b) and the luolishaniid lobopodian Acinocricus (see Caron and Aria 2020), however, exhibit their respective sclerites in rings. In the case of Tylotites, the sclerites form complete rings around the body, whereas in Acinocricus the rings do not conjoin ventrally. Two taxa therefore exhibit serially repeated spinose sclerites in rings: Tylotites (complete ring) and Acinocricus (incomplete ring). This character also differentiates the two by the dorso-ventral differentiation exhibited by Tylotites, wherein the more ventral spines of the transverse ring are considerably shorter than more dorsal spines (see Han et al. 2007b, fig. 1.1).

Correspondence of serially repeated dorsolateral epidermal specializations to ventral paired structures (1:1, 2:1, 3:1 or 4:1)

In most lobopodians and Cricocosmia, there is a 1:1 relationship between trunk lobopod pairs and the metameric epidermal specializations – that is, for every serially repeated group of epidermal specializations (e.g. nodes or spines) there is one pair of ventral structures (lobopods or ventral projections; see Figs 5d and 8b for Cricocosmia). In Tabelliscolex, the ventral projection pairs occur beneath every third serial sclerite triplet (see Figs 1, 2 and 8c), in Luolishania beneath every second (Ma et al. 2009) and in Thanohita beneath every fourth (Siveter et al. 2018). Coded 1:1 for Collinsovermis (Caron and Aria 2020), inapplicable (–) for Mafangscolex and uncertain (?) for Tylotites and Acinocricus.

Phylogenetic trees

The majority of our phylogenetic analyses (all parsimony tree searches and maximum likelihood) recovered an exclusive monophyletic grouping of the palaeoscolecid-like worms, located within stem-group Priapulida that we name Palaeoscolecidomorpha (see Fig. 7). This clade is essentially equivalent to ‘Palaeoscolecida sensu lato’ as defined by Harvey et al. (2010), which was ambiguous in terms of monophyly in that study. Only Bayesian inference failed to recover Palaeoscolecidomorpha, but it also did not resolve Scalidophora beyond the extant priapulan taxa and Palaeoscolecida sensu stricto. Full topologies with 95 taxa and support values for all tree searches are presented in the Supplementary Material.

When Palaeoscolecidomorpha was recovered, it always contained a monophyletic clade comprising Wronascolex, Mafangscolex and Scathascolex (representatives of Palaeoscolecida = ‘Palaeoscolecida sensu stricto’ in Harvey et al. 2010) and another containing Cricocosmia, Tylotites and Tabelliscolex (Cricocosmiidae). The position of Maotianshania and Markuelia within Palaeoscolecidomorpha was slightly variable across optimality criteria, but they were always closer to Cricocosmiidae than to Palaeoscolecida. Maximum likelihood resolved Maotianshania and Markuelia as sister taxa and that clade, in turn, was a sister taxon to Cricocosmiidae. However, in the parsimony strict consensuses, Markuelia and Maotianshania form a polytomy with Cricocosmiidae.

No tree search yielded a relationship between Palaeoscolecidomorpha and Panarthropoda. Stem-group Panarthropoda comprises only Aysheaia under equal-weights parsimony and maximum likelihood, whereas Aysheaia is resolved in the stem-group of Tardigrada (as sister to Onychodictyon) under implied-weights parsimony. The stem-group of Panarthropoda comprises Microdictyon, Xenusion, Diania and Paucipodia under implied-weights parsimony, whereas other methods recovered these lobopodians in stem-group Onychophora. Bayesian inference did not resolve the stem-group of Panarthropoda.

Systematics of palaeoscolecidomorph worms

According to our phylogenetic analyses, Palaeoscolecidomorpha is most likely a clade within the stem-group of Priapulida, forming the sister-group to the clade comprising crown-group Priapulida and the paraphyletic assemblage of ‘archaeopriapulids’ (e.g. Ottoia, see Conway Morris 1977; Eximipriapulus, see Ma et al. 2014a). The hexaradially ornamented phosphatic microfossil Eopriapulites (Liu et al. 2014) was usually the most stemward branch of total-group Priapulida, lending support to the hypothesis that Ecdysozoa is ancestrally hexaradial in circumoral armature (Liu et al. 2014). The armature of M. yunnanensis has been determined to be hexaradial (Yang et al. 2020) and therefore it is inferred that the transition to pentaradial circumoral armature occurred in the more crownward paraphyletic ‘archaeopriapulid’ assemblage. The crown-group Priapulida plus ‘archaeopriapulids’ clade is constrained to a minimum divergence point of the Ediacaran–Cambrian boundary by the ichnospecies Treptichnus pedum (Vannier et al. 2010; Kesidis et al. 2019), which indicates that Palaeoscolecidomorpha had also diverged by this point. As such, the fossil record shows Palaeoscolecidomorpha probably existed from at least the terminal Ediacaran until at least the Silurian.

Palaeoscolecidomorpha is supported by two unambiguous autapomorphies: bilaterally arranged terminal posterior hooks and an elongate trunk of consistent width (see systematic palaeontology). Bilateral terminal posterior hooks are present in all included taxa, although with variations in the number of hooks/spines. Duan et al. (2012) rejected a palaeoscolecid affinity for Markuelia on the basis of the number of posterior spines (three pairs) and the absence of polymorphic tessellating plate ornament. Our analysis partly supports the hypothesis of Duan et al. (2012) in that Markuelia is not a member of Palaeoscolecida. Markuelia is recovered here in the more inclusive Palaeoscolecidomorpha, in which all taxa share the presence of posterior bilateral spines/hooks, also including taxa with seemingly derived numbers of hooks (e.g. the triplet set of Wronascolex and the double pair of Scathascolex). The inclusion of Markuelia within Palaeoscolecidomorpha suggests that palaeoscolecidomorphs were direct developers (see Dong 2007; Dong et al. 2004, 2005, 2010), but are bracketed by loricate taxa in our phylogeny (Loricifera, Sicyophorus and crown-group Priapulida), suggesting that direct development is a derived trait in Markuelia/Palaeoscolecidomorpha. The elongate trunk of consistent width is also present in all included palaeoscolecidomorph taxa and presents a reliable character for identifying poorly preserved palaeoscolecidomorph fossils that preserve little or no cuticular ornamentation (e.g. Conway Morris and Peel 2010).

One additional potential palaeoscolecidomorph autapomorphy is the seriated paired ventral projections (see Fig. 5). Ventral projections are clearly present in the cricocosmiids Cricocosmia and Tabelliscolex, but also apparently in the palaeoscolecids Mafangscolex and Houscolex (see Zhang and Pratt 1996, fig. 2), which implies that this character is plesiomorphic to both groups (hence an autapomorphy of Palaeoscolecidomorpha). However, some caution is required, as this character is clearly absent in all known stages of development in Markuelia (Dong et al. 2010). Furthermore, it is unknown whether this character occurs in the remaining two-dimensional compression taxa because it is rarely preserved, even in the taxa in which it definitely occurs other than Tabelliscolex, but the presence of ventral structures is also not precluded in these without further study. For example, only 21 of 703 specimens of Mafangscolex preserved the ventral projections in the present study (see Table 1) and most palaeoscolecid studies include substantially fewer specimens than the present study.

Palaeoscolecida is supported as a group within Palaeoscolecidomorpha by a single, but highly distinctive, autapomorphy: the annular-arranged tessellating polymorphic plate ornamentation. Within Palaeoscolecida, Scathascolex and Wronascolex each show a derived number of posterior hooks. Mafangscolex exhibits a single pair of hooks similar to cricocosmiids and Maotianshania, whereas Scathascolex has a double pair of hooks (Smith 2015, fig. 1j) and W. antiquus has a triplet (García-Bellido et al. 2013, fig. 3f). A single ring of introvert spines/scalids is autapomorphic to S. minor. The introverts of all other palaeoscolecidomorph taxa (and the vast majority of scalidophorans) consist of multiple transverse rings. Smith (2015) speculated that the single ring in Scathascolex may be indicative of a nematoid affinity because rings of circumoral structures on nematoid proboscises are much fewer than those of scalidophorans, but our analyses do not support this. Scathascolex exhibits several other characters allying it to Scalidophora and Palaeoscolecida (see Table 3). Cricocosmiidae is supported by at least one distinctive autapomorphy, the variable seriated large trunk sclerites (see systematic palaeontology) and appears to share the unarmed posterior proboscis region with at least Maotianshania.

Seriated sclerites in Cambrian Ecdysozoa

Dzik and Krumbiegel (1989) predicted that palaeoscolecid-like worms represented a transitionary locomotor grade between infaunal priapulans and errant epibenthic lobopodians (Dzik and Krumbiegel 1989, fig. 6). This was presented as evidence of the shared ancestry of lobopodians (‘xenusiids’) and priapulans, which was confirmed in the following decade by the advent of molecular phylogenetics and the Ecdysozoa hypothesis (Aguinaldo et al. 1997; Giribet and Edgecombe 2017). In the model of Dzik and Krumbiegel (1989), a hypothetical worm is depicted to represent the ‘missing link’ between palaeoscolecids and the relatively simple lobopodian Xenusion auerswaldae, which was resolved in stem-group Panarthropoda by implied-weights parsimony in our study (Fig. S2). This hypothetical worm had paired, seriated, dorsolateral and ventrolateral structures. This depiction therefore resembles our current interpretation of cricocosmiids in cross-section and lateral profile (see Fig. 8e). Cricocosmiids were poorly known at the time of publication of Dzik and Krumbiegel (1989), but subsequent studies recognized the similarity of the seriated sclerites of Tabelliscolex, Cricocosmia and several lobopodians (Han et al. 2007a; Steiner et al. 2012). This naturally provoked hypotheses in these studies of panarthropod ancestry among cricocosmiids, seemingly corroborated by Dzik and Krumbiegel's predictions. This has not, however, been supported by previous morphology-based phylogenetics (e.g. Harvey et al. 2010) and is not supported by our phylogenetic analyses (see Fig. 7), even with the ventral projections newly accounted for and a significantly better representative taxon and character sample for Ecdysozoa as a whole. Furthermore, there is evidence that cricocosmiids were infaunal burrow dwellers (Huang et al. 2014), whereas Dzik and Krumbiegel (1989) expected such a worm to be epibenthic.

Like Tabelliscolex, Cricocosmia bears two lateral rows of sclerites (Figs 5d and 8b) with two distinct surfaces (Fig. 4a–f). However, Cricocosmia lacks the additional dorsal row of sclerites and, rather than ellipsoidal plates as in Tabelliscolex, Cricocosmia sclerites are posteriorly directed, smoothly curved, hollow spines with an apex and a broad base. Regardless, the outer surface in both taxa bears hexagonally arranged tubercles, with the inner surface covered by corresponding pits, which are interpreted together as holes ornamenting the sclerite. This gives cricocosmiid sclerites a distinctive ‘net-like’, porous appearance (see Fig. 4) that is at least superficially similar to sclerites of the lobopodians Onychodictyon spp. (Topper et al. 2013), Microdictyon spp. (Topper et al. 2011; Pan et al. 2018) and seemingly H. hongmeia (Steiner et al. 2012), but not H. sparsa or H. fortis (see Caron et al. 2013). In addition, many other lobopodian taxa exhibit seriated nodes, spines and plates, but have not been investigated by scanning electron microscopy. There are, however, several aspects where cricocosmiid sclerites differ from lobopodian sclerites in structure. In Microdictyon the ‘net-like’ appearance of the sclerite is similarly produced by hexagonally arranged pore holes, but they vary in size towards the periphery (whereas cricocosmiid pores are consistent all over) and each pore is ornamented by a further hexagonal ring of nipple-like tubercles (Fig. 4g, h; see Topper et al. 2011; Pan et al. 2018). This can also be observed in Onychodictyon, although the secondary hexagons of ornamentation are not nipple-shaped, but scarped platforms (Topper et al. 2013).

Our phylogenetic analyses suggest that these differences in structure are explained by the fact that these sclerites are a result of evolutionary convergence (analogy), as opposed to shared inheritance (homology). Cricocosmiidae is deeply nested in Scalidophora in our trees and lobopodian taxa exhibiting seriated sclerites are a diverse polyphyletic assemblage of stem-group panarthropods, stem-group onychophorans and stem-group tardigrades. The most parsimonious explanation appears to be that seriated dorsolateral trunk structures are plesiomorphic to the panarthropods, but have also evolved independently within Scalidophora. Our phylogenetic interpretation does not necessarily preclude some of the ideas postulated by Dzik and Krumbiegel (1989), however, as the lower reaches of the panarthropod stem-group remain poorly resolved. The elongate, annulated trunks of lobopodians undoubtedly reveal that panarthropods are derived from worm-like ancestors, but no fossil has yet been constrained to stem-group Panarthropoda that lacks paired appendages, with a recent study rejecting Facivermis yunnanicus as a cycloneuralian-panarthropod transitionary example (Howard et al. 2020b). Our study does conclude, however, that cricocosmiids, like Facivermis, are also not representative of the cycloneuralian-panarthropod transition (i.e. stem-group Panarthropoda). Cricocosmiids and other palaeoscolecidomorphs exhibit an array of scalidophoran synapomorphies, in addition to the ecdysozoan plesiomorphies and homoplastic characters they share with lobopodian panarthropods. These scalidophoran and/or priapulan synapomorphies, including the retractable anterior introvert, multiple transverse rings of spinose circumoral armature on that proboscis (although their identity as hollow scalids is technically unconfirmed), the eversible pharynx and, at least in some Palaeoscolecida, the array of characteristic scalidophoran external sensory structures (e.g. flosculi, tumuli and tubuli; see Harvey et al. 2010).

Ventral projections in palaeoscolecidomorphs

Ventral spines and nodes were originally described in Tabelliscolex (Han et al. 2003a) and subsequently reported to occur in a sequence beneath every third set of sclerites (Huang 2005), although fossil material supporting this hypothesis was not figured. Han et al. (2007a) also reported ventral spines in Cricocosmia, but did not figure them, and compared them with the ventral spines of the extant meiofunal priapulan Tubilichus, interpreted as an indication of sexual dimorphism. We have demonstrated here that the hypothesis of Huang (2005) is likely to be the most correct, but the spines of Tabelliscolex are paired, curved and posteriorly directed (see Figs 5g–i, 6 and 8c). In addition, we report that similar, smaller paired ventral projections are also present in Cricocosmia and Mafangscolex. We interpret the ventral projections as a probable autapomorphy for Palaeoscolecidomorpha (see earlier discussion and Table 3). Incompletely preserved members of Palaeoscolecida from other localities also appear to demonstrate corroborating evidence of this character (Zhang and Pratt 1996; Hu et al. 2012). Hu et al. (2012) described large protuberances being irregularly present in the posterior ventral of Yunnanoscolex magnus and Wudingscolex sapushanensis from the Guanshan Lagerstätte.

The Wuliuan Burgess shale ‘archaeopriapulid’ Louisella pedunculata also has two longitudinal rows of long filamentous ventral spines/papillae (Conway Morris 1977). The shape and arrangement of these structures in Louisella have clear differences, however, running only along a posterior portion of the trunk and with no corresponding relation to the trunk annulations. As is the case for seriated sclerites, the phylogenetic distance between cricocosmiids (stem-group Priapulida) and lobopodians (stem-group Panarthropoda) means that even when a primary homology between ventral projections and lobopods is coded, these seriated ventrolateral structures optimize on the phylogeny as homoplastic. The ventral projections of Cambrian scalidophoran worms are convergent with panarthropod appendages, even to the degree that both show a correlated seriality with dorsolateral sclerites.

Palaeobiology of Tabelliscolex

Locomotion

Ventral projections of palaeoscolecidomorphs have previously been discussed with reference to sexual dimorphism, to increase friction in burrowing or for mucus secretion from glands (Han et al. 2007a). Given that ventral projections are observed in all specimens of Tabelliscolex, but are seen in only a minute fraction of the samples for Cricocosmia and Mafangscolex, relative abundance data are inconsistent with sexual dimorphism. We regard it as most plausible that the ventral projections of T. hexagonus were used to increase friction because these structures are larger and harder than those of other palaeoscolecidomorphs. The ventral projections of T. hexagonus might therefore have had the ability to support the body and be used in locomotion.

All priapulans move using the hydrostatic skeleton to push the body via peristalsis, but the movement of different species reveals some minor differences. This behaviour may result in distinctive probing burrows, which are comparable between present day priapulans and late Ediacaran–Cambrian trace fossils called treptichnids (Vannier et al. 2010; Kesidis et al. 2019), such as T. pedum, the occurrence of which defines the boundary between the Ediacaran and Cambrian (Buatois 2018). The complete movement cycle of Priapulus caudatus has four stages: (1) trunk peristaltic contraction, then propagating rapidly from posterior to anterior; (2) invagination; (3) powerful eversion; and (4) proboscis inflation (Vannier et al. 2010). The complete movement cycle of the Cambrian priapulan Eximipriapulus globocaudatus was inferred to involve seven stages (the introvert, trunk and terminal region have different shapes in different stages) (Ma et al. 2014a). The movement of the cricocosmiid T. hexagonus is hypothesized to have involved a double anchor fixing strategy like other palaeoscolecidomorphs, making use of both the proboscis scalids and the posterior hooks. T. hexagonus has a long trunk and a clear difference between the dorsum and abdomen. This species is therefore regarded as engaging in a demersal and epifaunal lifestyle. When it burrowed, the parts of the trunk close to the proboscis or the posterior end underwent peristaltic contraction. The ventral projections pointing posteriorly increased friction to prevent sliding and accelerate movement, as well as controlling the direction of movement.

Ontogeny and moulting

Information on moulting in fossil scalidophorans is often lacking, with fossils preserving an animal and its exuvium together being rare, because moult events were of short duration and the exuviae are thin (Daley and Drage 2016). Little is known about the ontogeny and moulting of palaeoscolecidomorphs specifically, aside from some isolated cuticular fragments apparently showing a double layer interpreted as having been buried shortly before ecdysis (Müller and Hinz-Schallreuter 1993). However, the moulting process has been described in phosphatic scalidophoran cuticle fragments from the Fortunian-aged Kuanchuanpu Formation of Shaanxi Province, China (Wang et al. 2019). These scalidophorans are thought to occupy a position in stem-group Scalidophora and to moult in a similar manner to extant priapulans, wherein the body is extricated smoothly from the old tubular cuticle or the exuvium is turned inside out like the finger of a glove (Wang et al. 2019). An early Cambrian stem-group loriciferan from Greenland appears to show a similar mode of ecdysis, with the body preserved emerging from an apparent tubular exuvium (Peel et al. 2013). Therefore it may be inferred that the process of ecdysis has been conserved across Scalidophora. However, cricocosmiid trunk sclerites, although resolved as convergent here, are highly similar to those of lobopodians, which may indicate additional complexity in the moulting process in cricocosmiids. Isolated lobopodian sclerites with two conjoined elements, wherein a larger sclerite underlies a smaller one, have been reported in Microdictyon spp. (Zhang and Aldridge 2007) and Onychodictyon spp. (Topper et al. 2013) from the Cambrian of China and Greenland, respectively. These conjoined specimens are interpreted as successive moults, indicating that the ecdysis of trunk sclerites may occur through a gradual replacement process (Topper et al. 2013).

T. hexagonus appears to be represented by at least one juvenile specimen (YKLP 11431, Fig. 1g). It is apparently smaller than the other specimens, but similar in morphology. Figure 1d and 1e show specimens that are different from the others and also the suspected outlines (long black arrow) of sclerites beside some sclerites may be indicative of moulting.

Mode of feeding

In most specimens of T. hexagonus, the intestine is completely flat and preserved in dark colour, whereas in one specimen the alimentary canal is preserved three-dimensionally (Figs 1d, e and 2c, d). Three-dimensional gut structures found in Burgess Shale-type arthropods have been suggested to be a taphonomic artefact due to either sediment infilling caused by panic behaviour during live burial (Edgecombe and Ramsköld 1999) or combined early diagenetic mineralization and severe weathering processes (Butterfield 2002; Vannier and Chen 2002). However, more recent studies rejected these scenarios in Chengjiang fossil preservation because geochemical analyses showed that these three-dimensional gut contents consist of genuine sediment with no evidence of mineralization (Hou et al. 2004; Bergström et al. 2007; Ma et al. 2014a). In the case of Chengjiang vermiform taxa, the three-dimensional structures often only occur partially in the gut and hence they are unlikely to be the result of sediment infill during live burial, but instead are indicative of at least occasional sediment digestion (Hou et al. 2004; Bergström et al. 2007; Ma et al. 2014a).

In contrast with the complex digestive systems seen in Cambrian arthropods (e.g. the presence of mid-gut glands/diverticulae), the palaeoscolecids and other Cambrian scalidophorans have a relatively simple alimentary tract (Vannier et al. 2014; Ortega-Hernández et al. 2018). However, the detailed alimentary tract morphologies and gut contents can shed light on their feeding ecology. The alimentary tract of most Cambrian scalidophoran worms is composed of a terminal mouth, a muscular pharynx lined with teeth and an elongated gut, whereas some species even have a division of oesophagus, mid-gut and hind gut (Ma et al. 2014a). Diverse skeletal animal fragments and cololites have been reported in the gut tracts of the stem-group priapulans Ottoia prolifica from the Burgess Shale (Vannier 2012), Singuuriqia simoni from Sirius Passet (Peel 2017) and Selkirkia sinica from Xiaoshiba (Lan et al. 2015), each instance interpreted as indicative of a generalist feeding regime including predation and deposit feeding, as in modern day macropriapulans. The data at hand do not allow us to infer whether T. hexagonus was a deposit feeder or a carnivore, or both, as no individual gut element could be confidently identified.

Tabelliscolex hexagonus is a palaeoscolecidomorph (stem-group Priapulida) belonging to the subgroup Cricocosmiidae, along with at least C. jinningensis and T. petiolaris. Tabelliscolex is distinguished from other cricocosmiids by the configuration and morphology of its sclerites and the correspondence of its well-developed ventral projections with the serially repeated sets of trunk sclerites compared with other taxa. Palaeoscolecidomorphs, and especially cricocosmiids, demonstrate a remarkable parallel evolution of ectodermal seriation that is strikingly similar to that of lobopodian panarthropods. Paired ventral projections that occur in a seriated pattern relative to the lateral/dorsolateral trunk sclerites are a topological mirror of the seriated trunk nodes/sclerites of lobopodians, which are positioned in a dorsolateral position above the ventrolateral lobopodous appendages in taxa ranging from stem-group panarthropods to stem-group onychophorans and stem-group tardigrades.

Seriation is more pervasive in Ecdysozoa than just Panarthropoda and Palaeoscolecidomorpha. Kinorhynchs have seriated cuticular structures (unpaired tergal plates and paired sternal plates) that are, in some cases, notably in the well-studied genus Echinoderes, correlated internally with seriated components of the trunk musculature and the nervous system. These include paired sets of dorsal, ventral, dorsoventral and diagonal muscles (Herranz et al. 2014) and seriated transverse neurites and ganglia (Herranz et al. 2019). As such, seriation is manifest in both the ectoderm and mesoderm; kinorhynchs exhibit covariation of seriated organ systems to a degree that they are not uncommonly described as segmented (but see Scholtz 2020 for a cogent distinction between seriation and segmentation). Seriation has multiple origins in Ecdysozoa and its different instances involve different character systems. Kinorhynchs and arthropods shared striated tergites, sternites, ganglia and trunk muscles, whereas palaeoscolecidomorphs and lobopodians shared seriated dorsolateral sclerites and ventrolateral projections.

We thank Javier Ortega-Hernández and another anonymous reviewer for constructive reviews that improved this paper.

: XS: conceptualization (equal), data curation (supporting), formal analysis (equal), investigation (equal), methodology (equal), writing – original draft (equal), writing – review & editing (equal); RJH: conceptualization (equal), data curation (supporting), formal analysis (equal), investigation (equal), methodology (equal), writing – original draft (equal), writing – review & editing (equal); GDE: conceptualization (supporting), formal analysis (supporting), investigation (supporting), methodology (supporting), supervision (supporting), writing – original draft (supporting), writing – review & editing (supporting); XH: data curation (supporting), funding acquisition (supporting), resources (supporting); XM: conceptualization (supporting), data curation (lead), funding acquisition (lead), investigation (supporting), methodology (supporting), project administration (lead), resources (lead), supervision (lead), writing – original draft (supporting), writing – review & editing (supporting).

Yunnan Provincial Research Grants (Grant Nos. 2015HA021, 2015HC029 and 2019DG050 for X-GH and X-YM) supported the Yunnan Key Laboratory for Palaeobiology group, including fossil collecting, supporting students and research expenditure. NERC Independent Research Fellowship (Grant No. NE/L011751/1) provided salary and research expenditure for X-YM. NERC GW4 + Doctoral Training Partnership provided stipend and research expenditure for RJH. The funding bodies had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Scientific editing by Zongjun Yin