Hemichordates are central to our understanding of deuterostome evolution. The two classes (tubicolous Pterobranchia and vermiform Enteropneusta) are monophyletic [1–3], but are morphologically disparate (however, see [4, 5] for an alternate viewpoint of Pterobranchia as sister to the family Harrimaniidae within a paraphyletic Enteropneusta). Accordingly they give only generalized clues as to both the anatomy and mode of life of the last common ancestor as well as its connections to the sister phylum Echinodermata (collectively Ambulacraria). The resistant tubaria of pterobranchs (notably the Paleozoic graptolites [6]) provide an adequate fossil record, but in contrast that of the enteropneusts is almost non-existent [7–9]. One exception is a tubicolous taxon (Spartobranchus tenuis) from the middle Cambrian Burgess Shale [10]. This enteropneust is closely comparable to extant harrimaniids, although its organic tube finds no modern counterpart [11]. The coeval Oesia disjuncta Walcott [12] has been compared to groups as diverse as annelids [12], appendicularian tunicates [13] and chaetognaths [14, 15], thus remaining in phylogenetic limbo. The proposed chaetognath affinity was refuted by Conway Morris [16] and a hemichordate affinity briefly suggested instead, albeit without detailed re-observation of original specimens or consideration of new material. On the basis of hundreds of specimens from the newly discovered Marble Canyon fossil locality (Kootenay National Park, British Columbia) [17], we not only identify Oesia as a primitive enteropneust but also demonstrate that it constructed the perforated tube-like fossils previously assigned to Margaretia dorus and interpreted as thalli of a green alga similar to Caulerpa [18].

Oesia possesses the canonical enteropneust body plan of proboscis, collar and elongate trunk (Figs. 1, 2a, b) but is unusual in that posterior to the pharynx there is a bilobed structure, rather than a vermiform intestine. Body length averages 53 mm (n = 187, size range 2.4–120 mm), but the width seldom exceeds 10 mm. The proboscis is relatively elongate (ratio of length to width is 1.35 ± 0.58) and variable in shape (Figs. 2a, d, e, g, h; Additional files 1A, D–F, 2A, F–I, 3C, 4). A conspicuous ovoid area at the medial base of the proboscis appears darker or more reflective than the surrounding area (Fig. 2a–c, f; Additional files 2F–I, 3C). This is interpreted as the heart-kidney-stomochord complex [11]. More irregular structures across the proboscis probably represent decayed musculature (Fig. 2c; Additional file 2F–I). The collar is rectangular, but with rounded edges (Fig. 2a–c, f, g; Additional files 1A, D–F, 2F–I, 3C, 5D, E, G). In proportion it is shorter than the proboscis (average ratio is 0.39 ± 1.12) but has an equivalent width (average proboscis to collar width is 1.08 ± 0.23 mm). At the posterior margin of the collar (Fig. 2b, c, f; Additional file 2F–I), a dark or reflective band probably represents the circum-collar ridge, while a thin, longitudinal structure between the proboscis base and collar (Fig. 2d, e; Additional file 5A, B, D–I) is interpreted as the nuchal skeleton. The pharyngeal region houses a series (about 3 bars/mm) of approximately U-shaped gill bars (Fig. 2g–j; Additional file 5A, C) but is remarkable in that it occupies approximately 80 % of the trunk length (Fig. 2a, g; Additional files 2A, C, 5D, E, G). The posterior end of the trunk is bulbous (Fig. 2b, g; Additional files 1D–E, 2H, 4, 5D, E, G) and sometimes terminates in a bilobed structure (Fig. 2a, f; Additional files 1B, C, 2A–D, F, H) that is usually wider than long (average width-to-length ratio is 1.48 ± 0.63).

The preservation of Oesia (n = 45 in Marble Canyon, n = 6 in Raymond Quarry) inside Margaretia (now a junior synonym of Oesia disjuncta) suggests an original association (Fig. 3a; Additional file 6). Only single worms are found within tubes, suggesting a solitary mode of life, although due to breakage during transport, it is conceivable that tubes may have been inhabited by more than one worm (Fig. 3b–d). Typically the tube is at least twice the width of the worm, suggesting the worm could move freely within its dwelling (Fig. 3e–j; Additional files 3B, 7, 8). Three-dimensional preservation of both sides of the tube (Fig. 4d, e) shows that the internal cavity of the tube was spacious, and that the tube was at least semi-rigid. The total length and extremities of the tubes are poorly known. This is because of either prior breakage or concealment (Fig. 4a), but at least one end (presumably the top of the tube) appears rounded and closed (Additional file 9B, C).

Tubes with irregular undulations and lacking the spiral pattern were previously interpreted as prostrate subterranean rhizomes (Fig. 4.2-3 [18]). While the reassignment from alga to organically produced tube invalidates this identification, it remains plausible that subterranean, lateral extensions of the tube could serve as an anchor. In any individual the width of the tube is usually consistent along the length, but otherwise it varies considerably (4–20 mm). Occasionally a tube shows one (Fig. 3i; Additional files 7A, C, 8A–C, 9D–F) or, more rarely, two bifurcations (Fig. 4c). Each bifurcates at approximately the same angle and has the same width as the primary tube. The tube wall is perforated by spirally arranged pores (about 10 openings per revolution; Fig. 4a, b). In a single tube pore size varies. Some may be almost closed, but others have diameters equivalent to about a third of the tube width (Fig. 4a, b, d, e; Additional file 9A). Pore shape varies from circular to oblong ellipse and rhombic. That these might simply be taphonomic variations is less likely given that the specimens are preserved parallel to the bedding plane (Fig. 4a, b, g–e; Additional file 9A). The margins of the pores tend to be raised, imparting a semi-corrugated texture to the external surface of the tube (Fig. 4a, b; Additional file 9A). The tube is composed of narrow fibres (about 7 μm) that are braided and/or overlain in bundles (Fig. 4f, g).

Margaretia dorus is unlike any known species of Paleozoic algae. In particular, the combination of a fibrous composition and elaborate pore architecture are inconsistent with an algal grade of organization, as are its biotic associations and size in relation to well-established Cambrian macroalgae [19]. This in turn argues against Oesia being an example of inquilinism. While the dozens of co-occurrences of O. disjuncta and its tube strongly suggest an original association, the preservation of large numbers of isolated Oesia specimens on single bedding surfaces (Additional files 3, 4) at Marble Canyon also needs an explanation. One possibility is that the association was facultative and Oesia could alternate between a tubicolous and non-tubicolous existence. Alternatively the worm may have been forced to vacate the tube as an en masse evacuation prior to final burial. This may be related to both the high-energy burial events [17] and the resultant dysoxic conditions that such events create [20], although this hypothesis is weakened by the lack of obvious exit structures (i.e. there is no evidence the worms could enter or leave the tubes at either end).

In this context, fragmentation of the tubes and dispersal during transport is perhaps a more plausible explanation as to how the worms became isolated. This appears to be reasonable given the observation that although tubes with a length of up to 544 mm are known (Fig. 4c), tubes of comparable width can be not only significantly shorter (e.g. Figs. 3b, 4c), but sometimes are even smaller than the worms themselves. A related observation is that along the tube margins showing evidence for breakage, the bundles of fibres may exhibit a pattern of ’unbraiding.’ This suggests that originally the tubes were vulnerable to damage (Fig. 3b).

The second factor is that in at least some cases the tube evidently serves to conceal the worm. For a worm to be readily visible, the tube either needs to be prepared mechanically, split more or less along the axis or be sufficiently degraded so as to allow a view of the interior. Accordingly, tubes showing such evidence of degradation also contain worms in an evident state of decay (Fig. 3b–h). In such cases worms are poorly preserved and are effectively reduced to a narrow band of reflective carbon (Fig. 3k–m). Worms in such late stages of decay also show a tendency to bend at sharp angles into semi-discrete sections (Figs. 2g, 3e, f, l, m). This appearance may represent adjacent sets of gill bars maintaining their articulation through attachment to the collagenous basal lamina, but at points where this basal lamina has degraded, the more acutely angled bending occurs [11].

Establishing Oesia as an enteropneust that inhabited the tube previously identified as the alga Margaretia has significant implications for the Cambrian paleogeography and paleoecology of this group. Until now, Oesia was one of the rarest of Burgess Shale taxa and was restricted to the Walcott Quarry [21]. At the coeval Marble Canyon locality, however, it is amongst the five most abundant taxa [17] and occupied a key trophic position. In marked contrast, Margaretia is recorded from various Burgess Shale sites in Laurentia (including the Stephen Formation of British Columbia and the Spence and Wheeler Shales of Utah [18] — Additional file 11: Table S1), eastern Yunnan, China [22] and further afield in Siberia (originally referred to as Aldanophyton [18]) [23]. This expanded distribution suggests that enteropneusts were a significant component of many Cambrian communities (Additional file 6, Additional file 11: Table S1).

Oesia also throws important new light on the early evolution of the hemichordates. Construction of a large, complex and presumably metabolically costly tube is consistent with a sessile lifestyle. Given that pores appear to be present on all sides, we suggest that the tube (and branches) stood vertically, with the basal region embedded in the substrate and the top presumably closed (Fig. 5a). The porosity of the tube would have prevented dysoxia and also allowed access to the water column for filter feeding. Given that the tubes could exceed 50 cm (Fig. 4c), this suggests a tiering level at least equivalent to (if not above) the tallest sponges known from the Burgess Shale.

The strikingly extended pharynx and numerous gill bars that were employed in suspension feeding are functionally convergent with the hyperpharyngotremy seen in the tunicates [24], cephalochordates [25] and some Paleozoic jawless fish [26] (Fig. 5a). More generally, however, the pharyngeal arrangement seen in Oesia suggests that within the hemichordates as a whole suspension feeding was primitive. Whilst a few members of the basal harrimaniids facultatively filter interstitial pore water [27, 28], in extant enteropneusts the primary mode is deposit feeding, consistent with their mostly infaunal existence. Such a migration from an epifaunal existence may have been in response to increased predation pressure and as a consequence entailed significant anatomical changes. Notably in Oesia the post-pharyngeal trunk appears to lack the esophageal organ, which in extant forms serves to remove excess water from the food cord (and presumably performed the same function in Spartobranchus tenuis where it is also present; a summary of the main differences between S. tenuis and O. disjuncta can be found in Additional file 10). So too in the more derived taxa the hepatic caeca increase the absorptive area, presumably reflecting the increased demands of deposit feeding.

Oesia shares with the co-eval tubicolous S. tenuis [10] a bulbous posterior structure which may have acted as an anchor. In Oesia, however, the claw-like arrangement points to a more active role in attachment and release, perhaps as a consequence of its inhabiting a commodious tube. This interpretation draws potential comparisons to the juvenile post-anal tail of harrimaniid enteropneusts. This tail serves in ciliary locomotion and as an attachment device and may also be the homologue of the pterobranch stalk [27]. In this context, the specialized posterior structures seen in S. tenuis and O. disjuncta may actually be ancestral features. If so, these were ultimately lost in the crown group Enteropneusta, but in the Pterobranchia they helped to pave the way towards coloniality.

While too few morphological characters are available to permit a meaningful cladistic analysis, the unique combination of characters found in O. disjuncta encourages us to present a preliminary re-interpretation of early hemichordate evolution. First, a tubicolous, epifaunal and solitary habit are evidently primitive. The fibrous filaments of the Oesia tube have some resemblance to the fusellar fibres seen in graptolites such as the Cambrian Mastigograptus [29], as well as the comparable periderm of rhabdopleurid [30] and cephalodiscid pterobranchs [31]. An important inference is that Oesia (and Spartobranchus) possessed secretory glandular cells, presumably homologous with those located on the cephalic shield of the pterobranchs. The apparent absence of fibres in the tubes of Spartobranchus suggests that their loss may have preceded the loss of the tube itself. Concurrent with a shift to a burrowing and deposit feeding existence, the crown group enteropneusts abandoned the construction of such tubes. In contrast, the tubes of pterobranchs (and correspondingly the posterior stalk) were elaborated in parallel with their miniaturization and sessile coloniality (Fig. 5b). Crucially, the unique mix of pterobranch and acorn worm characteristics seen in Oesia suggests that an extensive pharynx and undifferentiated trunk are basal to the hemichordates, whereas Spartobranchus is more derived and is basal to the acorn worms [11]. Future discoveries of new Cambrian hemichordates will help elucidate the hypothesized transformation of the posterior structures into the pterobranch stolon and critically unveil the order of both trait acquisition and loss during the early diversification of this phylum.

Finally, the evidence that primitive enteropneusts were suspension feeders is congruent with the hypothesis that suspension feeding represents the primitive mode of life in deuterostomes [32] as a whole. In particular, it is notable that this lifestyle is seen in early stem-group echinoderms [33] and stem-group ambulacrarians [34], and is inferred in the ur-ambulacrarians [35] as well as the more problematic vetulicolians [36] and yunnanozoans [37].

Sediment overlaying sections of some specimens was removed using a micro-engraving tool with a carbide bit. Specimens were observed using a stereomicroscope and photographed using different illuminations, using direct or cross-polarized light on dry or wet specimens. Backscatter scanning electron images were obtained to visualize fine anatomical features. Measurements of morphology were made using the program ImageJ. A list of specimens used in the analysis can be found in Additional file 11: Table S1 [38–47].