DISCUSSION

Here we discuss the relevance of the new data on Mellopegma and other stenothecids to the following topics: functional morphology; escalation with predators; phylogeny; and shell characteristics including microstructure, the presence of pores, and the periostracum.

Shell Damage and Cambrian Escalation between Molluscs and Their Predators

Numerous specimens of the most abundant species of Mellopegma, M. georginense and M. schizocheras sp. nov., show imprints of shell scars, typically expressed as indentations on the internal moulds (Figure 6.4, 6.10-11, 6.14, Figure 9.5, 9.10-12). Several specimens also show missing regions of shell, often with a smooth margin (Figure 8.18-21). The smooth borders around most signs of damage, both scars and missing pieces, suggest that many breaks were healed during the lifetime of the animals. While it is difficult to rule out a mechanical cause of this damage, several lines of evidence support predation as a cause of most of the wounds: (1) the attacks occur at a higher frequency in the sub-apical region (Figure 17), the area with the largest aperture width and hence easiest access to the animal's flesh; (2) in many cases the wound tapers upward (Figure 6.4, Figure 8.18, Figure 9.5, 9.10-11), a pattern that would be produced by the tip of an appendage, versus mechanical damage which would produce more variation in form; (3) specimens interpreted as coprolites that contain bradoriid shells occur in these beds (Figure 7.12), revealing the presence of a predator on small animals; and (4) the many other signs of predatory activity in Cambrian fossils (see below). It must be kept in mind, though, that such a potential shell-breaking predator of Mellopegma must have been small, as Mellopegma was on the order of a millimetre long.

Mellopegma was a tiny burrowing animal and so may have escaped the senses of larger predators, but Swedmark (1968) pointed out that modern interstitial organisms even in subtidal environments need mechanical protection from damage caused by shifting sediment and smaller predators. Although the shell of Mellopegma is revealed in thin sections to have been quite thin (c. 10-20 µm total shell thickness; Figure 7.1-4), the organic-rich shell with a laminar inner layer may have helped it resist some attacks (see below). Moreover, the highly organic prismatic shell layer in M. georginense probably provided additional flexibility to the shell that helped prevent shell fracture during attack.

Clear evidence of predation is well-known from the Cambrian, including bite marks, healed scars, drill holes, and predatory appendages (e.g., Alpert and Moore 1975; Miller and Sundberg 1984; Jensen 1990; Conway Morris and Bengtson 1994; Skinner 2005; Vannier and Chen 2005) and in fact traces of predation have been described in fossils as old as the Late Precambrian (Bengtson and Zhao 1992; Hua et al. 2003). It is clear that large predators such as Anomalocaris roamed Middle Cambrian seas (Whittington and Briggs 1985), although smaller predators are also known. For example, the priapulid Ottoia ranged from 2-16 cm (Briggs et al. 1994) and has been found with hyolith and other shells in its gut; in these cases though the hyoliths are about four times the length of a typical specimen of Mellopegma. The strongest evidence for predation being a powerful selective factor at small sizes during the Cambrian comes from traces such as boreholes in small fossils (e.g. Bengtson and Zhao 1992; Conway Morris 1998). However, fossils of predators in the same rocks that preserve Mellopegma have not been found.

In modern molluscs predation is the predominant cause of shell damage, with molluscs in turbulent waters without predators suffering much lower rates of shell injury than molluscs in calm waters with predators (Vermeij 1987). Shell damage from predatory attack has been well documented in Paleozoic brachiopods (Alexander 1986) and molluscs (Schindel et al. 1982; Peel 1984; Lindström and Peel 1997, 2003). Skovsted et al. (2007) demonstrated shell repair in an early Cambrian mollusc deposited in a low energy environment, concluding that predation must have caused the injuries. In addition there is a high frequency of angular fragments of fossils in the Gowers Formation assemblage, and such fragments are typically produced by durophagous predation rather than physical factors (Oji et al. 2003).

Specimens of Mellopegma georginense show a relatively high frequency of healed damage (Figure 5.7, Figure 6.4, 6.10-11, 6.14, Figure 9.1, 9.5, 9.10-12), possible bite marks (Figure 8.18-21), and fragmentation. Although the apparent bite marks and fragmentation may have been caused by taphonomic factors, the healed injuries clearly occurred during the life of the animal. Although shell crushers in the early Paleozoic were relatively inefficient (Vermeij 1987), predators with the capability to damage the thin shells of early molluscs like Mellopegma likely existed in the early Cambrian (Skovsted et al. 2007).

The shell of Mellopegma was especially thin, but its innermost shell layer consisted of calcitic semi-nacre, which is very similar in form to aragonitic nacre (Weedon and Taylor 1995), a very strong microstructure (Currey 1977; Jackson et al. 1988, 1990). Nacre is thought to be energetically costly and time consuming to produce, largely because of is high organic component (Currey 1977). It is unclear to what extent the calcitic semi-nacre in Mellopegma had similar costs and benefits to aragonitic nacre, but Mellopegma had a flexible outer prismatic shell layer that would have helped it withstand shell-crushing predators. Mellopegma also possessed many of the other defensive traits against shell-crushing predators that were described by Vermeij (1987), including several characters that would have assisted in escape via vertical burrowing (Table 2).

Most of the healed marks on Mellopegma occur along the ventral margin (Figure 6.10, Figure 8.18-21, Figure 9.1, 9.10-11), and usually the damage was centered on the apical half and near the aperture (Figure 17). These observations suggest that the ventral margin, and likely the sub-apical side, of Mellopegma was exposed above the sediment surface, lending support to Runnegar's (1996) and Peel's (1991b) interpretation of the life position of the similar Eurekapegma. The main difference between Mellopegma and Eurekapegma is the presence in the latter of an internal plate called the zygion (MacKinnon, 1985). This plate extended from the apex to the ventral margin in the area where most of the healed damage in Mellopegma occurred. It, therefore, seems likely that at least one function of the zygion of Eurekapegma, an inferred descendent of Mellopegma (MacKinnon 1985; Peel 1991b; Runnegar 1996), was to support the shell in resisting crushing forces of predators. The zygion, a relatively thick internal plate that extended from one side of the shell to the other, would have resisted perpendicular lateral crushing forces at the easiest region of the shell to clamp. Moreover, a long region of the ventral margin on the apical side of Eurekapegma is sealed by a convergence of the two sides of the shell (MacKinnon 1985, fig. 3l,m,s,v), providing more evidence that this was a vulnerable area in its immediate ancestors.

Preliminary analyses of damage frequency provide some additional evidence in support of escalation between micromolluscs and predators over different time scales in the Cambrian (Figure 18). Proportions of damage in stenothecids were compared between the early Cambrian Parara Limestone and the middle Cambrian Gowers Formation, and also between two beds within the Gowers Formation. The damage rate of specimens from the Parara Limestone was 0.137 (n=51). The damage rate of specimens from the lower bed in the Gowers Formation was 0.347 (n=101) and from the upper bed 0.494 (n=89). The z-value for the difference between the early Cambrian versus the lower of the two middle Cambrian beds is 2.531 (99.4% confidence level); the z-value for the difference between lower and upper of the middle Cambrian beds is 1.917 (97.2% confidence level). Thus the results of this preliminary study of damage proportions in stenothecids through the Cambrian are consistent with escalation in predation intensity, although additional beds should be studied to confirm this pattern.

At the small sizes and miniscule shell thicknesses that characterize stenothecids, it is difficult to estimate the full range of predation pressure. Perhaps predation was just as high in the early Cambrian as in the middle Cambrian, but shell strength was lower then and shells may have often been totally demolished (and thus not preserved). In any case the frequency of damage caused by small shell-crushing predators that were not able to obliterate the thin, small shell of stenothecids increased through this time period, and other morphological trends in this family indicate increasing defense. For example, the Mellopegma lineage shows a trend toward increasing narrowness of the aperture (Figure 19), a defensive characteristic (Vermeij 1987) with high expense as it severely limits the space for organs inside the shell. This narrowing reached its zenith in the youngest form, Eurekapegma cooperi. Apertural narrowing would allow for faster burrowing to escape predators, and it would have made it more difficult for predatory appendages to reach into the shell. Thus the combined evidence indicates that Mellopegma was involved in a Cambrian arms race between small predatory arthropods or worms and tiny shelled molluscs.

Shell Characteristics

The phosphatic moulds reveal new details about the shell morphology of Mellopegma, including the form and organization of the outer prismatic and inner laminar (calcitic semi-nacre or similar) shell layers, the nature of the shell pore system, the form of the periostracum, and the shape and size of the protoconch.

Shell microstructure. Preservation of shell microstructure in Mellopegma has been known for many years. Runnegar (1985) described polygonal imprints near the aperture margin and angular imprints elsewhere in Mellopegma georginense, and suggested this species had nacreous inner and prismatic outer shell layers. New data presented here and in Vendrasco et al. (2010, table 2) confirm the presence of an inner laminar layer and an outer prismatic layer. However, the results of our comparisons of many aspects of the fossil imprints – including interfacial angles – with the variation in modern shell microstructures largely support the hypothesis that the inner shell layer was calcitic semi-nacre instead of (aragonitic) nacre (Vendrasco et al. 2010). In addition, studies here suggest that (1) Mellopegma had an inner shell layer composed of calcitic semi-nacre (the calcitic version of the shell microstructure defined by Carter et al. 1990, p. 611, as "laminae consisting of polygonal tablets which show more abundant screw dislocations and less lateral continuity of the laminae than in typical nacreous structure") that was highly organized and consisted of many stacks of laminae; and (2) Mellopegma had a prismatic outer shell layer that varied widely in form and region of exposure on the inner shell surface (where the inner shell layer was thin or missing) between the three species from the Gowers Formation – M. georginense, M. simesi, and M. schizocheras sp. nov. The type of calcitic semi-nacre expressed in Craniiformean brachiopod shells is high magnesian (Cusack et al. 2008), which probably also characterized the inner shell layer of Mellopegma, in accordance with the Tommotian transition to calcite seas.

Internal moulds of Mellopegma georginense, Mellopegma schizocheras sp. nov., Mellopegma simesi comb. nov., Mellopegma uslonicum, and Mellopegma? illustrated herein show imprints of shell microstructure. Specimens of Mellopegma indecorum that we examined have less distinct imprints of shell microstructure, but the limited evidence suggests a microstructure consistent with that found in the other species of Mellopegma.

New data presented here reveal more details about the outer prismatic shell layer and its variation among Mellopegma species. For example, M. georginense and M. schizocheras sp. nov. differ significantly in the form of the polygonal organic framework of prismatic microstructure and the distribution of imprints of this texture over the surface of internal moulds. In M. georginense the polygons are small (a few µm diameter), thick-walled, and occur over most of the surface of the internal mould, except at or near the apex (Figure 5.13, 5.16, 5.22, Figure 6.12-13, 6.15). In M. schizocheras the polygons are large (about 20 µm diameter), thin-walled, and occur only at the sub-apical and to a lesser extent the supra-apical regions of the aperture margin (Figure 9.3-4, 9.7). This disparity in polygon size, form, and distribution between M. georginense and M. schizocheras sp. nov. represents a major difference between these two similar taxa.

The polygonal network in the case of M. schizocheras sp. nov. is clearly an imprint of prismatic shell microstructure, as its pattern is very similar to that in epoxy moulds made of prismatic microstructure in modern molluscs (Vendrasco et al. 2010, pl. 10, figs. 1-4), and it occurs near the aperture of the internal mould, where the adpressed shell typically thins out and expresses the outer shell microstructure on the inner shell surface. The polygons in M. schizocheras sp. nov. might be the infill of prisms whose organic walls decayed away (sensu Kouchinsky 1999), or may represent an active replacement of the organic conchioloin walls (sensu Vendrasco et al. 2010).

In specimens of M. georginense the polygonal texture looks different from the polygonal texture in M. schizocheras and many other Cambrian molluscs, and instead looks similar to pores on the external surface of some molluscs such as chitons. This observation indicates the possibility that the polygonal texture in M. georginense might instead be external ornament, and that these specimens are shell replacements or casts instead of internal moulds. However, evidence that this texture is an imprint of inner shell microstructure and not external ornament includes the observations that: (1) imprints of calcitic semi-nacre tablets – a clear indicator of shell interior – overlie the polygonal network (Figure 6.12-13, 6.15); (2) the polygonal network is visible on every specimen in an assemblage heavily dominated by internal moulds; and (3) tubercles similar to those on internal moulds of other taxa, which represent the cast of internal shell tunnels (Kouchinsky 2000a; Parkhaev 2006; see below), occur within the polygonal network (Figure 5.16).

The polygonal network in M. georginense is perplexing in that it occurs over most of the surface of the internal mould, except at or near the apex, instead of just at the aperture margin. The polygonal network has the overall appearance of an organic matrix, which is common in all shells of molluscs and helps control crystal deposition in shell formation. The polygonal network is interpreted here as an imprint of an outer prismatic shell layer, in which case the inner laminar calcitic semi-nacre shell layer must have been very thin so that on internal moulds its imprints would not fully cover the prismatic layer. If this interpretation is correct, the prismatic shell layer consisted of small crystals embedded in a thick, flexible conchiolin matrix. Alternatively, this organic matrix perhaps was not part of a prismatic shell layer but instead seeded the incipient crystals of calcitic semi-nacre. Imprints of calcitic semi-nacre occur over the polygonal network (Figure 6.12-13, 6.15), and so clearly this organic layer was either part of that inner laminar shell layer or it occurred right at the boundary between the outer prismatic and inner laminar shell layers.

Distinct traces of prismatic shell microstructure have not been observed in specimens of Mellopegma simesi comb. nov. or in the early Cambrian Stenotheca drepanoida or Mellopegma indecorum. M. simesi is known from many specimens with well-preserved laminar shell microstructure, so the lack of polygonal texture in this species is perplexing. Perhaps the shell of M. simesi did not thin out as much distally as in other species of Mellopegma, and so the outer prismatic shell layer would only be seen in the inner shell surface if it were significantly abraded. Although the few known specimens of Mellopegma uslonicum lack well-preserved polygons, tubercles are prominent (Figure 12.6) and in some cases (Figure 13.6) the tubercles are linked up in a way reminiscent of the faintly preserved polygons of M. schizocheras sp. nov. (Figure 9.3-4, 9.6-7). Such faint merging of tubercles is seen to a lesser extent in M. simesi (Figure 11.1), but is less convincing a reflection of prismatic microstructure than is the comparable texture in M. uslonicum.

The new fossils of Mellopegma also reveal additional information about the inner laminar shell microstructure in members of this genus. All species of Mellopegma have a laminar inner shell layer. In M. georginense and M. simesi, the laminar microstructure is calcitic semi-nacre (Vendrasco et al. 2010). Based on the overall similarity in tablet imprints with these species, the other species of Mellopegma probably had an inner shell layer of calcitic semi-nacre as well.

The imprints of laminar microstructure in Mellopegma show that the crystal tablets in the inner shell layer were highly organized, with consistent orientation of the crystals within the layer (Figure 9.8-9, Figure 11.5-9). This structure suggests that Cambrian molluscs had a precise ability to control the microstructure of their shell, and that crystal nucleation was strongly guided, most likely by an organic matrix. This high degree of organization of the inner laminar layer also characterizes modern bivalve sheet nacre (compared with the less organized gastropod columnar nacre) (Taylor et al. 1969), providing support for a close relationship between Mellopegma and bivalves.

The fossils also reveal that each lamina was quite thin, as suggested by the imprints of tablets on multiple vertical levels (Figure 11.7) and the observations of replaced sheet-like laminae overlying the internal mould surface in some specimens (Figure 5.17, Figure 6.8, 6.14, Figure 9.8-9, Figure 11.7). There also appears to have been a tall stack of many laminae of calcitic semi-nacre that made up the inner shell layer of M. schizocheras sp. nov. and M. simesi (Vendrasco et al. 2010, pl. 1, fig.10, pl. 2, figs. 9-10, 12).

The occurrence of a laminar microstructure in Mellopegma uslonicum (Figure 12.5-9) similar to the calcitic semi-nacre in M. georginense, M. simesi, and M. schizocheras adds to the evidence that it is closely related to other Mellopegma species and extends back the history of this unusual type of laminar shell microstructure to the early Cambrian. Faint traces of a similar laminar shell microstructure made up of angular tablets can be seen in the other early Cambrian stenothecids as well (Mellopegma indecorum (Figure 14.5-6) and Stenotheca drepanoida (Figure 15.14)), although the preservation in these cases is too poor to allow a detailed comparison with other species of Mellopegma. Acanthotheca junior (Figure 16.3, 5-7) also shows distinct laminar microstructure that appears to be calcitic semi-nacre, strengthening the link between this taxon and Mellopegma. Calcitic semi-nacre thus characterizes Mellopegma and perhaps the Stenothecidae overall.

Calcitic semi-nacre is also known in platyceratoid gastropods from the Paleozoic (Carter and Hall 1990) but is otherwise rare in the Mollusca. Calcitic semi-nacre is more common in brachiopods (Williams and Wright 1970) and bryozoans (Weedon and Taylor 1995), and its occurrence in molluscs reveals a fundamental similarity in biomineralization between these lophotrochozoan taxa. The new data here show that calcitic semi-nacre dates back at least to the early Cambrian in molluscs, providing more evidence that it may have been a primitive shell microstructure in both molluscs and calcitic brachiopods. Similarity in shell formation between brachiopod and early molluscan shells was noted by Carter (1979), who suggested both groups had shell microstructure where the component crystals are not uniformly oriented in three dimensions (vertically and horizontally). This ancestral mode of biomineralization became more extensively modified in later molluscs than in brachiopods (Carter 1980). Molluscs and brachiopods are in the same major clade of Lophotrochozoa (Dunn et al. 2008, fig. 1, clade C), but the many soft-bodied taxa in this clade (i.e., molluscs and brachiopods each have soft-bodied taxa as their close relatives) make it unreasonable to assume that a shell was primitive in this clade. However, it is possible that the most recent common ancestor of brachiopods and molluscs had similar organic precursors in shell formation, leading to similar shell microstructures in their early history. In spite of the differences between modern mollusc and brachiopod shells (e.g., some brachiopod shells are phosphatic whereas no mollusc shells are; calcareous brachiopods are calcitic whereas molluscs are more often aragonitic; and molluscs have a greater diversity of shell microstructures), there are many similarities, including: (1) extensive shell pore system seen in chitons and many early Cambrian molluscs – see "Shell pores" below; (2) organic-rich shell; (3) mantle; (4) periostracum; (5) a complex shell with different types of shell microstructure in different layers; and (6) similar types of shell microstructure (all types of brachiopod shell microstructure are also seen in molluscs – Carter and Clark 1985).

Many different lineages of animal appear to have independently evolved a shell over the geologically short "Cambrian explosion" (Bengtson and Conway Morris 1992), suggesting that the evolutionary precursors to shell formation occurred in these taxa. Evidence for underlying homology in the shells of metazoans has been provided by Jacobs et al. (2000), who demonstrated that engrailed expression is involved in skeletal formation in a wide range of bilaterians. There was probably a high degree of homology in the precursors to shell formation in molluscs and brachiopods, explaining the great number of similarities in the shell and its formation between these taxa. The common ancestor of these taxa likely had a similar organic coat and genetic framework for constructing a shell. The occurrence of calcitic semi-nacre in molluscs from the early Cambrian described here reveals a stronger homology in shell formation among molluscs and brachiopods than previously realized.

Shell pores. Mellopegma is characterized by a pore system that extended through much or all of the thickness of the shell, connecting with the conchiolin of the prism sheaths in the outer prismatic shell layer. Kouchinsky (2000a) had previously noted the occurrence of "tubercles" or small protrusions on the surface of internal moulds of many early molluscs. He interpreted these structures as in-filling of pores on the inner surface of the shell. Feng and Sun (2006) and Parkhaev (2006) described numerous additional observations of pores in early molluscs, revealing that many early Cambrian molluscs had pores that infiltrated the shell, in some cases extending through the entire shell thickness (Parkhaev 2006, fig. 3). The evidence for pores consists of casts of the entire vertical canals extending from the surface of internal moulds, and, more commonly, tubercles, interpreted as partially broken casts.

The surface of internal moulds of Mellopegma uslonicum contains large-diameter (~ 3-4 µm), conical tubercles with a blunt end, suggesting that this species had relatively large shell pores that either ended abruptly or, more likely, were incompletely preserved (Figure 12.4; Parkhaev 2006). Runnegar and Jell (1976; fig. 8b8) noted tubercles on the internal molds of Mellopegma georginense and concluded this species had depressions on the internal surface of the shell. These depressions are interpreted here as a shell pore system, or a remnant of one that was more extensive in the early representatives of Mellopegma.

Many of the new specimens of M. georginense, M. schizocheras, and M. simesi have tubercles on the surface of the internal mold, sometimes merged together (Figure 5.15-16, Figure 9.6), and in many cases clearly in line with growth lines (Figure 5.15). These tubercles are typically best preserved at the apex (Figure 5.8) and along the dorsal ridge (Figure 5.9, Figure 10.17, Figure 11.1). At the anterior and posterior margins of M. schizocheras sp. nov., such tubercles can be seen at polygon boundaries (Figure 9.3-4, 9.6) and so pores may have occurred in between prisms of the outer shell layer. The pores in Mellopegma, best seen in M. uslonicum, are smooth-walled (Figure 11.4), which suggests they may have had an inner lining of – or were entirely filled with – tissue.

Some specimens of M. georginense and M. schizocheras appear to show that the organic sheaths around the prisms of the outer prismatic shell layer were in contact with the vertical pores (Figure 5.16, Figure 9.3-4, 9.6-7). This configuration makes it clear that the pores extended through most or all of the thickness of the shell, as they open at the inner surface of the shell (bumps over much of surface in internal moulds) and extend above the base of the outer prismatic shell layer (bumps at nodes of polygons near aperture where outer prismatic shell layer is preserved). However, the function of this pore system in Mellopegma and the significance of a connection to the conchiolin sheaths are at present unclear.

Shell pores occur in many modern molluscs. They are extensively developed in chitons, including the earliest known representatives (Pojeta et al. 2010), and also occur in gastropods (Reindl and Haszprunar 1996), the Palaeozoic monoplacophoran Tryblidium (Erben et al. 1968), and in many bivalve superfamilies where they penetrate the entire shell (Taylor et al. 1969). The pores in Mellopegma are most similar to those in bivalves; as with bivalve tubules, those in Mellopegma are inferred to have possessed a smooth surface and a cylindrical shape, and to have extended through much or all of the shell.

The current list of early Cambrian molluscs with evidence of shell pores is provided in Appendix 2. It appears that shell pores were either primitive in the Mollusca or that pores independently and rapidly evolved in different early Cambrian lineages (note in particular the occurrence of pores in the possible gastropods Barskovia and Philoxenella as well as diverse groups of helcionellids). Evidence in support of the hypothesis that shell pores in molluscs are primitive include: (1) the occurrence of shell pores among representatives of many groups of modern molluscs (Reindl and Haszprunar 1996); (2) their prevalence among early Cambrian molluscs (Parkhaev 2006); (3) their prominence in lophotrochozoans closely related to molluscs, including brachiopods (Reindl and Haszprunar 1996) and sipunculans (Ruppert and Rice 1995); and (4) their prevalence in other taxa thought to be molluscs or closely related to them, such as hyoliths (Kouchinsky 2000b) and coeloscleritophorans (Bengtson 1992) like halkieriids (Vinther 2009). Although some of these pore systems in various modern molluscs appear structurally different and so perhaps are not homologous (Reindl and Haszprunar 1996), their widespread occurrence in the earliest known molluscs and other evidence listed above suggests pores may be primitive in this phylum, and were subsequently lost in major branches of molluscs. Pores are probably even more widespread among Cambrian molluscs than the data currently suggest because: (1) pores might not have been common or large in some shells and so were not commonly fossilized; (2) some pore openings may not have been filled by the phosphate that coated the shell (evidence for this can be seen in the variable preservation of pores in single specimens); (3) some Cambrian species are not known from well-preserved specimens such as fine-grained internal moulds; and (4) in some cases pore fillings might not have been noted in descriptions and cannot be identified in pictures from the literature.

Periostracum. One specimen of Mellopegma georginense shows prominent radiating ridges over the surface of an apparent cast (Figure 5.18-19). These ridges originate at the apex and curve downwards to the aperture. Similar structures can be seen in other middle Cambrian molluscs, including Pseudomyona queenslandica (Gubanov et al. 2004, fig. 9g-h), Yochelcionella ostentata (Gubanov et al. 2004, fig. 6s), and Anabaroconus sibiricus (Gubanov et al. 2004, fig. 5n). Gubanov et al. (2004) did not provide an interpretation of these structures, but we interpret them to be a replacement of a portion of the periostracum layer. The periostracum is the outermost, entirely organic, layer of the molluscan shell that in modern forms consists mostly of quinone-tanned proteins. We interpret the fossil structures to be replaced periostracum because: (1) in each case the structure covers all other textures on the fossils, consistent with it being the external periostracum layer; (2) it is rarely preserved (seen in only one specimen of Mellopegma out of more than 100 examined via SEM), consistent with being an organic structure that degraded quickly after the animal's death, and inconsistent with being external ornament; (3) the radial ridges extend the full height of the shell (from apex to aperture), consistent with an organic covering but inconsistent with shell microstructure; and (4) on specimens where it is preserved there is variation in prominence of ridges from one region of the shell to the next, and often it is only faintly preserved in spite of consisting of thick ridges. The last point is consistent with the hypothesis that this structure is a replacement of an organic layer and inconsistent with the hypothesis that it represents external shell ornament; a thick-ridged external ornament should be more conspicuous and evenly preserved in the fossils.

The periostracum aids in the initial formation of the shell, in isolating the mantle cavity from the surrounding seawater, and in protection (Taylor et al. 1969). Its importance plus its widespread occurrence in conchiferans and chitons suggest that this layer originated early in the evolution of shelled molluscs. The findings in Gubanov et al. (2004) and herein provide direct evidence for this hypothesis, revealing that Cambrian molluscs had a thick periostracum. These observations suggest that the mechanism of biomineralization that characterizes modern molluscs was already in place during the Cambrian, a mechanism wherein the chemistry in the mantle cavity – isolated from the external environment by the periostracum – is controlled to induce crystal initiation and growth.

Protoconch. The protoconch is obvious in most well-preserved specimens of Mellopegma. It is a simple cone whose internal mould is smooth with scattered tubercles (e.g., Figure 5.8, Figure 6.7). The protoconch in Mellopegma is typically ~150-200 µm in length, in line with the range in modern monoplacophorans (Marshall 2006). The large size of protoconch 1 (the initial part of the protoconch, formed inside the egg capsule) suggests lecithotrophic development. Bivalves with lecithotrophic development tend to have a small or non-existent Prodissoconch 2 (Jablonski 1985), which might explain why this boundary is not readily identifiable in Mellopegma. Similarly, the early Cambrian bivalve Pojetaia runnegari had only a large (~150 µm), single prodissoconch (Runnegar 2007).

Nützel et al. (2006) suggested that Cambrian molluscs had lecithotrophic larvae and that planktotrophic larvae evolved in the Ordovician. Freeman and Lundelius (2007) instead argued that planktotrophy was primitive in molluscs. Both teams used as evidence measurements of Cambrian fossils, mostly phosphatic internal moulds. Runnegar (2007) criticized the validity of these measurements but noted that the few clear cases of protoconchs in Cambrian molluscs suggest they were lecithotrophic. Mellopegma provides another example of a Cambrian mollusc with a clear, large protoconch suggestive of non-planktotrophic larvae.

Martí Mus et al. (2008) provided fossil evidence that some helcionellids were juvenile shells of much larger animals with broader, limpet-like shells. Extrapolating this idea to a large fraction of helcionellids is problematic, however. In the case of Mellopegma, the distinct larval shell, consistent size range of specimens, lack of larger calcitic specimens in the rocks, and unbroken ventral margin in many specimens all indicate that it is the adult shell that is preserved.

Functional Morphology of Mellopegma

The laterally compressed shell suggests that Mellopegma was at least semi-infaunal, or – given its small size – interstitial. The life position of Eurekapegma MacKinnon, 1985, which had a very similar shell to Mellopegma, has been debated. MacKinnon (1985, figure 6a) postulated that the internal plate (zygion) of Eurekapegma provided an area for attachment of muscles from the foot and that the sub-apical region of the shell was buried in sediment. Peel (1991b) reversed this orientation, postulating that in Eurekapegma the supra-apical surface was buried, and that the zygion delimited the posterior mantle cavity with gills to the sub-apical region exposed above the sediment. Runnegar (1996) agreed with Peel that the supra-apical surface was buried, but disagreed with Peel's (1991b) assertion that it was the posterior of the animal that was exposed. Runnegar (1996) also noted that at these small sizes drawing water into and through the body would have been comparable to honey sucked through a straw. Either Mellopegma did not actively draw in water, or it spent significant metabolic energy to do so.

Many specimens of Mellopegma georginense and Mellopegma schizocheras sp. nov. exhibit caved-in portions or missing regions of shell below the apex (Figure 17). We interpret this to reflect the action of predators, indicating that this part of the animal was probably exposed above the sediment surface and thus exposed to predators (Figure 1).

Mellopegma as Ancestor of Rostroconchs

Rostroconchs, bivalves, and Cambrian stenothecids such as Mellopegma share a ventrally curved lateral margin (Waller 1998) and significant lateral compression. In addition, Runnegar and Jell (1976) described rostroconch-like features in two internal molds referred to "Mellopegma?", that are not as laterally compressed as M. georginense or M. schizocheras sp. nov., but are otherwise similar in form. Both specimens have a shallow depression on the supra-apical end near the margin (Runnegar and Jell 1976, figure 8c7, 8c9). Runnegar and Jell (1976) interpreted this as a muscle insertion similar to what is seen in ribeirioid rostroconchs, a hypothesis that Waller (1998) said needs more testing. This feature is also seen in specimens of Mellopegma simesi (Figure 10.9-10; Figure 11.3 with arrow) and Stenotheca drepanoida (Figure 15.1, 15.8-9, 15.16-17 with arrows). Specimens of Mellopegma also show an incurved portion of the internal mold beneath the apex (best developed in Mellopegma simesi; Figure 11.11, 11.14) which Runnegar and Jell (1976) interpreted as a small pegma, the internal shell projection characteristic of rostroconchs.

Such observations led to speculation and debate about the role of Mellopegma in the early evolution of rostroconchs. Similarities between Mellopegma and early rostroconchs led Runnegar and Jell (1976) to postulate that Mellopegma is transitional between narrow early Cambrian forms such as Anabarella and rostroconchs, an idea echoed in Runnegar (1978), who referred to Mellopegma as a "pararostroconch." In Wagner's (1997) cladistic analysis (Figure 3), Mellopegma/Eurekapegma is transitional between other stenothecids and a clade that includes bivalves and rostroconchs. In our cladistic analysis herein (Figure 4; Table 3, Table 4; Appendix 1), Mellopegma and Acanthotheca form a clade that is the sister group to a clade that comprises rostroconchs and Pseudomyona/Tuarangia.

Gubanov et al. (1999) described an alternative evolutionary sequence for the origin of rostroconchs, from Oelandiella through Anabarella to Watsonella in the early Cambrian of Siberia. Gubanov et al. (1999) argued that Mellopegma could not be considered ancestral to Watsonella because at the time Mellopegma was only known from the middle Cambrian whereas Watsonella is from the Tommotian. Mellopegma is now known from the early Cambrian (see above), but the assumption that Watsonella is the oldest rostroconch is now questioned.

More recent evidence suggests that Watsonella may be a quasi-bivalved helcionellid mollusc, not a rostroconch. The closer relationship of Watsonella to bivalves than rostroconchs is supported by Dzik's (1994, figure 12g) observation of a divided larval shell in Watsonella, similar to bivalves but dissimilar to rostroconchs. The idea that Watsonella is not a rostroconch but that it may be ancestral to bivalves has received recent support (Runnegar 1996; Wagner 1997; Carter 2001), although other specimens of Watsonella appear to have an undivided or incompletely divided larval shell (AVK, personal observation). More work is needed to better elucidate the range of form in this important genus.

The middle Cambrian Acanthotheca junior has also been interpreted as a rostroconch (Runnegar 1996), but this species lacks the breakdown of coiling and anterior-posterior elongation of shell that characterize ribeirioid rostroconchs, and the greatest width of Acanthotheca occurs in the supra-apical region of the shell, not the sub-apical region as in rostroconchs. If neither Watsonella nor Acanthotheca are rostroconchs, then the oldest known members of this class are from the early late Cambrian.

Runnegar (1996) proposed a morphological transition between laterally compressed stenothecids like Stenotheca and Mellopegma through Acanthotheca junior to younger riberioid rostroconchs. The occurrence of calcitic semi-nacre in Acanthotheca junior (Runnegar, 1996) from the Gowers Formation (Vendrasco et al. 2010, pl. 3, figs. 1, 6-9, 24) is an additional similarity to Mellopegma, adding to others such as a curved ventral margin, a pegma, and a slight coil. These similarities suggest a close relationship between A. junior andRunnegar (1996) Mellopegma. In particular, a transitional sequence can be envisioned from a form like Mellopegma schizocheras sp. nov. through a form like Mellopegma simesi to Acanthotheca junior. M. simesi is intermediate with respect to elongation, lateral compression, flaring of sub-apical margin, and development of the pegma. In addition, these three taxa share a similar pattern of shell microstructure (Vendrasco et al. 2010).

Although Mellopegma and Acanthotheca lack the uncoiled nature of the shell that characterizes rostroconchs, the stenothecid Eurekapegma, the youngest member of the lineage, has a range in form from slightly coiled as is typical for Mellopegma to uncoiled, typical for ribeirioid rostroconchs (MacKinnon 1985, figure 3k, o, q, u, w). Thus Eurekapegma has a form that may represent the ancestral state to rostroconchs.

The shell microstructure of ribeirioid rostroconchs is unknown if Acanthotheca junior is considered outside the group. Such information would help in testing the hypothesis that stenothecids are ancestral to rostroconchs. Shell microstructure is known from the Carboniferous conocardioidean Apotocardium, shown to have a fine prismatic outer layer, crossed lamellar middle layer, and porcellaneous/matted inner layer (Rogalla et al. 2003). This configuration differs significantly from that of Mellopegma and its kin, although there are some parallels with what occurs in Anabarella and Watsonella (Kouchinsky 1999; Rogalla et al. 2003). Peel (2004) noted the striking difference between the protoconchs of ribeirioids and conocardioids, indicating that these two rostroconch groups may have had independent origins from different types of helcionellids. In Peel's (2004) model a group of exogastric helcionellids gave rise to ribeiroids whereas endogastric forms gave rise to conocardiods. The possible polyphyly of the Rostroconchia may help explain the difference in shell microstructure between conocardioids and the stenothecids that may have been ancestral to ribeirioids.

MacKinnon (1985) argued that the middle Cambrian Enigmaconus may have been ancestral to rostroconchs. Enigmaconus has distinct pegma-like structures similar to what occurs in rostroconchs, structures that MacKinnon suggested originated in rostroconchs prior to lateral compression. However, there appears to be convergent evolution in the development of pegma-like structures (Runnegar 1996), as well as other rostroconch characters such as elongation/scaphopodization (Peel 2006) and a breakdown in coiling (in Eotebenna, Pseudomyona, Eurekapegma, and rostroconchs). Conflicting evidence typifies all the possible ancestors of rostroconchs, including:

Watsonella: the split shell, pegma-like structure underneath the apex, and overall shape—including curved ventral margin—are consistent with rostroconchs, but the significant stratigraphic gap between it and undoubted rostroconchs is evidence against this link. The apparent split larval shell likewise is evidence against the link between Watsonella and rostroconchs, although Dzik's (1994) interpretation of the juvenile shell in Watsonella has been questioned by Rogalla et al. (2003), and other specimens of Watsonella appear to have a univalved juvenile shell (AVK, personal observations).

Enigmaconus: pegma-like structures are present and in the same general areas as in rostroconchs, but this form is much wider than rostroconchs and lacks the curved ventral margin. Moreover, the unusual scaly shell microstructure in Enigmaconus (Kouchinsky 2000a) is quite different from the known microstructure of Apotocardium.

Pseudomyona/Tuarangia: pseudobivalved, laterally compressed, uncoiled shell like rostroconchs, but without pegma. Although these taxa differ from typical ribeirioids in being taller and having a more centrally located larval shell, they share with ribeirioids other aspects of form: for example, loss of coiling/straight dorsal margin and lack of aperture constriction—characters not seen in stenothecids (Figure 4; Table 3, Table 4; Appendix 1). The inner shell microstructure of foliated calcite has the same mineralogy as stenothecids, but this is different from the known microstructure of Apotocardium. Wagner (1997) argued that Pseudomyona and Tuarangia are more closely related to rostroconchs than is Watsonella, noting that Pseudomyona, Tuarangia, and undisputed rostroconchs share denticles and an extensive elongation of the anterior part of the shell.

Acanthotheca gen. nov./Mellopegma/Eurekapegma: these taxa share the same curved ventral margin and lateral compression as rostroconchs. However, they lack a pegma and have a shell microstructure different from Apotocardium. Moreover, they show coiling unlike rostroconchs, although some specimens of Eurekapegma lost coiling and are very similar in overall appearance to ribeirioid rostroconchs.

The origin of rostroconchs remains a mystery. Shell microstructure data from some of the earliest undisputed rostroconchs would provide evidence for or against the hypotheses described above, but so far this information is lacking.

Origin of Bivalves

The morphology and stratigraphy of the earliest bivalves, stenothecids, undisputed rostroconchs, and Watsonella are consistent with the hypotheses that Watsonella or a close relative gave rise to bivalves and a different helcionellid gave rise to rostroconchs. In addition to significant lateral compression and a curved ventral margin, stenothecids and Watsonella share with bivalves a number of similarities, including an extended sub-apical dorsal ridge (anterior in bivalves), and an inner shell microstructure that is laminar, sheet-like, and composed of highly organized sub-units (see "Shell microstructure" above).

Pojetaia and Fordilla, taxa that appear to be the earliest bivalves (Pojeta 2000), may have originated from a genus like Watsonella, a taxon that is either a stenothecid (Parkhaev in Gravestock et al. 2001) or descended from one (Anabarella; Kouchinsky 1999). Dzik's (1994) observation of a split shell in Watsonella has shifted the predominant view of this genus from being a rostroconch to an ancestor of bivalves.

Additional support for a link between Watsonella and bivalves comes from Carter (2001), who noted similarities between the shell microstructure of Watsonella and Anabarella (described by Kouchinsky 1999) and that of the earliest bivalves Fordilla and Pojetaia. The putative earliest bivalves had an equivocal microstructure interpreted as prismatic (Runnegar 1985) or large tablet nacre (Carter 2001). Carter (2001) noted a similar microstructure in Anabarella and Watsonella that he likewise classified as large tablet nacre, strengthening the argument that these taxa are ancestral to bivalves. This shell microstructure is different than the calcitic semi-nacre of Mellopegma.

The Cambrian taxa Psuedomyona and Tuarangia have been considered questionable bivalves (Pojeta 2000, Carter 2001, Elicki and Gürsu 2009) in spite of superficial similarities in form such as a bivalved or pseudo-bivalved shell. They have a foliated calcite inner shell microstructure (Runnegar 1985), which is more similar to the calcitic semi-nacre of Mellopegma than to the laminar inner shell microstructure of Watsonella, Fordilla, or Pojetaia. Hinz-Schallreuter (1995, 2000) considered Tuarangia a bivalve and Pseudomyona a rostroconch, but the striking similarities between Tuarangia and Pseudomyona indicate they are closely related (Runnegar and Pojeta 1992). The nature of the relationships of Pseudomyona and Tuarangia to bivalves and rostroconchs is unresolved.

Runnegar (1996) suggested Watsonella may have been a link between stenothecid monoplacophorans and the Bivalvia. He also argued that the long dorsal margin of laterally compressed univalves such as M. georginense was the precursor to the bivalve ligament. Likewise, Wagner (1997) suggested Watsonella (as Heraultipegma) is not a rostroconch but instead is on the lineage leading to bivalves.

The stratigraphy of these fossils is consistent with the hypotheses that Watsonella or a close relative gave rise to bivalves and a different helcionellid, possibly Mellopegma, gave rise to rostroconchs. Watsonella occurs in the N. sunnaginicus Biozone of the Tommotian Stage and in beds likely deposited earlier in Siberia, China, and Avalonia (lower Stage 2). In fact, Watsonella was suggested as an index fossil for the base of Stage 2. Mellopegma indecorum co-occurs with Watsonella in samples M303/2 (basal Petrotsvet Formation; Rozanov et al. 1969) and 183e (Rassokha River; Egorova and Savitzky 1969). The earliest bivalves Fordilla and Pojetaia are known from the lower Stage 3 (Elicki and Gürsu 2009). In contrast, the oldest rostroconchs are from the early late Cambrian.