Reconstructing the diet of a 505-million-year-old arthropod: Sidneyia inexpectans from the Burgess Shale fauna.

The feeding ecology of the 505-million-year-old arthropod Sidneyia inexpectans from the middle Cambrian (Series 3, Stage 5) Burgess Shale fauna (British Columbia, Canada) is revealed by three lines of evidence: the structure of its digestive system, the fossilized contents of its gut and the functional anatomy of its appendages. The digestive tract of Sidneyia is straight, tubular and relatively narrow in the trunk region. It is enlarged into a pear-shaped area in the cephalic region and stretches notably to form a large pocket in the abdomen. The mouth is ventral, posteriorly directed and leads to the midgut via a short tubular structure interpreted as the oesophagus. Anteriorly, three pairs of glands with internal, branching tubular structures open into the digestive tract. These glands have equivalents in various Cambrian arthropod taxa (e.g. naraoiids) and modern arthropods. Their primary function was most likely to digest and assimilate food. The abdominal pocket of Sidneyia concentrates undigested skeletal elements and various residues. It is interpreted here as the functional analogue of the stercoral pocket of some extant terrestrial arachnids (e.g. Araneae, Solifugae), whose primary function is to store food residuals and excretory material until defecation. Analysis of the gut contents indicates that Sidneyia fed largely on small ptychopariid trilobites, brachiopods, possibly agnostids, worms and other undetermined animals. Sidneyia was primarily a durophagous carnivore with predatory and/or scavenging habits, feeding on small invertebrates that lived at the water-sediment interface. There is no evidence for selective feeding. Its food items (e.g. living prey or dead material) were grasped and manipulated ventrally by its anterior appendages, then macerated into ingestible fragments and conveyed to the mouth via the converging action of strong molar-like gnathobases. Digestion probably took place within the anterior midgut via enzymes secreted in the glands. Residues were transported through the digestive tract into the abdominal pocket. The storage of faeces suggests infrequent feeding. The early diagenetic three-dimensional preservation of the digestive glands and abdominal pocket may be due to the capacity of Sidneyia to store Phosphorus and Calcium (e.g. spherites) in its digestive tissues during life as do, for example, modern horseshoe crabs.


Introduction
Reconstructing the diet of ca 500-million-year-old animals presents a scientific challenge with important implications regarding the level of interactivity within early marine communities and the construction of the very first modern-style trophic webs. The exceptional preservation of fossils from several Cambrian Lagerst€ atten has allowed scientists to reconstruct the feeding ecologies of early animals with astonishing accuracy, as exemplified by recent studies on carnivorous worms (Vannier, 2012), anomalocaridids , trilobites (Lerosey-Aubril et al., 2012) and other arthropods (Zhu et al., 2004;Haug et al., 2012b). Here we focus on Sidneyia inexpectans (Walcott, 1911;Bruton, 1981), an iconic arthropod of the middle Cambrian (Series 3, Stage 5; ca 505 Ma) Burgess Shale biota from British Columbia, Canada. Its feeding ecology is inferred using three lines of evidence: the structure of its digestive system, the fossilized contents of its gut, and the functional morphology of its appendages. Comparisons with modern chelicerates (e.g. spiders, horseshoe crabs) are made to reconstruct key-aspects of its biomechanics of feeding and to elucidate the phosphatic preservation of its digestive structures.

Fossil material
Most of the fossil material studied here (Table Sup. 1,2) comes from: 1) the Walcott Quarry Shale Member, characterized by fossiliferous, finely laminated, calcareous siltstones and silty graphitic mudstones; and 2) the slightly younger Raymond Quarry Shale Member, characterized by grey, greenish and brown layered blocky-slaty mudstones (see Fletcher and Collins, 1998Caron and Jackson, 2008;Collins, 2009). The Sidneyia inexpectans specimens deposited in the collections of the National Museum of Natural History, Smithsonian Institution, Washington D.C. (USNM) were mainly collected from the Phyllopod Bed within the Walcott Quarry Shale Member. The great majority of those housed at the Royal Ontario Museum (ROM) were collected from both the Raymond Quarry Shale Member and Walcott Quarry Shale Member (RQ, RT and WQ, WT numbers, respectively) in successive seasons of excavations and talus picking (RT, WT) between 1975 and 2000, by ROM parties led by D. Collins. A few specimens come from the Emerald Lake Oncolite Member, about 30 m above the floor of Raymond's Quarry (Fletcher and Collins, 1998). Additional material comes from localities that were more recently exploited by the ROM (field campaigns led by J.-B. Caron) ca 40 km southeast of the type area (Raymond and Walcott's Quarries, near Field) in the northern part of the Kootenay National Park, British Columbia: ten specimens from the Stanley Glacier area (STAN in Caron et al., 2014) in the uppermost part of the Stephen Shale ('thin' Stephen) and Burgess Shale Formations ('thick' Stephen) respectively.
A total of 136 specimens was selected for detailed observations and analyses under a stereomicroscope (see list in Table Sup. 1,2).

Recent material
Six specimens of the horseshoe crab Limulus polyphemus were purchased from the Woods Hole Oceanographic Institution (Massachusetts, United States) for anatomical comparisons (digestive system, appendages) with S. inexpectans. Horseshoe crabs were collected near the marine station, kept in the laboratory for a few days, fed with mashed Mytilus edulis and Crepidula fornicata before being euthanized with clove oil, put in 70% ethanol and shipped to Lyon. One specimen was deep-frozen in diluted ethanol (ca À57 C) and cut across its sagittal plane with a wire saw before being photographed. Other specimens were dissected in 70% ethanol in order to observe details of their internal anatomy (e.g. hepatopancreas) and to sample gut contents for microscopic observations and analyses. A specimen of the spider Leucorchestris sp. (Araneae, Sparassidae) from Namibia was dissected in 70% ethanol for observations of its stercoral pocket.

Methods, techniques and terminology
Microscopic observations (Leica MZ125 stereomicroscope equipped with Plan 1.0Â and Planapo 1.6Â lenses, digital camera and Leica LAS 3.7.0 imaging system with multifocus option) and digital light photography (D3X-Nikon camera with Nikon Micro-Nikkor 60 mm lens) with cross-polarized light techniques to increase contrast of anatomical features were used routinely. Scanning electron microscopy (FEI Quanta FEG 250) and Energy dispersive X-ray spectroscopy (EDX) analysis were used to study the detailed morphology and chemical composition of the fossil and Recent materials. Images were acquired with Secondary Electron (SE) and Back-Scattered (BSE) Electron detectors at 15 KV and under high vacuum, with no sputter coating. Some low-relief specimens were coated with ammonium chloride (NH 4 Cl) sublimate or immersed in water before being observed under the stereomicroscope and photographed.
Head, thorax, abdomen and their adjectives are used here as descriptive terms and do not imply homologies with their equivalents in extant arthropods. The head of S. inexpectans corresponds to the shield-bearing anterior tagma (cephalon), the thorax to the limb-bearing trunk region and the abdomen to the limbless trunk region.

External morphology and affinities of Sidneyia inexpectans
The most accurate description of S. inexpectans Walcott, 1911 remains that of Bruton (1981) who reconstructed the overall external anatomy of this iconic arthropod of the Burgess Shale and discussed its possible lifestyle. Only Stein (2013) challenged some aspects of Bruton's interpretation, especially the head and appendage structure.
Sidneyia is a low-vaulted, oval-shaped arthropod with its dorsal exoskeleton clearly divided into three regions: a cephalon, a thorax and an abdomen (Fig. 1). The anteriormost exoskeletal element forms a short hemielliptical cephalic shield (CS) that bears lateral notches from which multi-annulated antennulae and small stalked eyes protrude. It is followed by nine imbricated dome-shaped thoracic tergites each composed of a broad medial area and pointed pleurae directed posteriorly. The body of Sidneyia terminates with two cylindrical abdominal sclerites and a tail fan. The latter consists of a central, triangular, moderately swollen telson flanked with paired (possibly two-folded) thin flap-like structures arising from the rear margin of the second abdominal somite (Fig. 1C and D). Rare specimens preserved in lateral view ( Fig. 2) confirm that CS has a rounded frontal margin and extends ventrally but do not confirm the presence of a ventral cephalic doublure sensu Bruton (1981). We agree with Stein (2013) that Sidneyia had most probably a hypostome attached to its CS via a hypostomal suture that probably functioned during molting as in modern horseshoe crabs.
Sidneyia has nine pairs of post-antennal appendages (Fig. 1E,F) that are characterized by particularly large basipods and robust attachments to the body (Fig. 3). The basipods exhibit large gnathobases directed towards the animal's midline in natural position (Fig. 3E). These gnathobases are made of strongly sclerotized elements which are teeth-like in the anterior appendages and seem to be more spine-like in the posterior ones. The endopod has seven podomeres, the four proximal ones bearing a bunch of thin spines, the 3 distal ones being armed with stout claw-like elements (Fig. 3H). The endopods of at least the 3 anteriormost post-antennal appendages seem to end up as a three-folded pincer-like structure suggesting a possible prehensile function (Fig. 3JeL). Post-antennal appendages 4e9 are characterized by well-developed exopods fringed with numerous blade-like lamellae interpreted as gills ( Fig. 3I; see Bruton, 1981 andStein, 2013 for more details). The presence of such gills in appendages 1e3 (Stein, 2013) is not confirmed here. Juveniles have an overall external morphology comparable to that of adults (Fig. 1F,G).
Although the phylogenetic position of S. inexpectans within the Arthropoda has been debated for more than a century (see Stein, 2013 for a recent review), no real consensus has emerged. This taxon has often been associated with Emeraldella to form the Xenopoda (sometimes with a third arthropod from the Burgess Shale, Molaria), but most phylogenetic analyses have failed to resolve this group as monophyletic (e.g. Wills et al., 1995;Cotton and Braddy, 2004;Ortega-Hern andez et al., 2013;. More recently, Ortega-Hern andez et al. (2013) have formally characterized the clade Vicissicaudata, which regroups  Sidneyia and the other 'xenopods', and two well-supported groups of Palaeozoic arthropods, the Aglaspidida and the Cheloniellida. However, this new clade only relies on a single shared character and none of the phylogenetic analyses that have identified it agreed on the relationships between its different components. Moreover, while the vicissicaudates were resolved as the sister taxon of a clade regrouping crustaceans and marrellomorphs by Ortega-Hern andez et al. (2013), other studies have resolved them as the sister taxon of chelicerates (e.g.  or of a clade regrouping chelicerates and megacheira (Edgecombe et al., 2011). Finally, recent studies suggest that Sidneyia might in fact be a stem-chelicerate rather than a vicissicaudate (Legg, 2014a, b).

General features
The gut tract of S. inexpectans is straight, tubular and relatively narrow through the thoracic region, but stretches both anteriorly and notably posteriorly to form a large elliptical pocket. Rare specimens preserved in lateral compression show that the digestive tract runs parallel and close to the axial roof of the exoskeleton and slopes down in the abdominal region ( Fig. 2A,B). The anus opens ventrally at the telson level and seems to be covered by a circular anal plate attached along the posterior margin of the second abdominal sclerite (Fig. 1C). The digestive glands and the abdominal pocket are by far the best preserved features of the digestive system of Sidneyia due to the early diagenetic mineralisation of their soft tissues and lumina in phosphate.

Digestive glands and anterior gut
Three pairs of conspicuous features often occur in the cephalic region and the anterior thorax up to the third segment. They are preserved in three dimensions in contrast to most anatomical and exoskeletal features that are strongly compressed. They are mineralised in calcium phosphate with frequent cauliflower-like microcrystals of ca 0.5 mm in diameter (see Fig. 4G and Energy Dispersive X-ray Spectroscopy (EDX) analysis in Fig. 4HeJ). The features are interpreted as paired digestive glands owing to their symmetrical arrangement on both sides of the sagittal plane of the animal (Fig. 4AeF). They seem to attach along the wall of the anterior part of the gut although the exact outlines of this wall are not clearly defined. These assumed digestive glands have morphofunctional equivalents in various Cambrian arthropods such as Squamacula (Paterson et al., 2012) and naraoiids (Vannier and Chen, 2002;Zhang et al., 2007). Their relatively irregular outline suggests that they were soft organs possibly resembling the digestive glands of extant crustaceans (e.g. Icely and Nott, 1992;Felgenhauer, 1992) and chelicerates (Fage, 1949;Fahrenbach, 1999;Botton and Shuster, 2003). The anteriormost pair has the largest size, the broadest base, and forms wing-shaped pointed extensions into the head region. The second pair is shorter and splits into two lobes distally and the third one has a more elongated spindle-shape. They are located between the cephalic shield and the third trunk segment. Each gland is made up of closely packed, curved and relatively long tubules of ca 150e300 mm in diameter. The tubules ramify distally and end up blindly along the margin of the gland (  disposition of these glands is suggested by their insertion sites on the gut at the boundaries between cephalon/tergite 1, tergite 1/ tergite 2, and tergite 2/tergite 3 (Fig. 4AeH). A similar distribution pattern of digestives glands has been described in various Cambrian arthropods (Lerosey-Aubril et al., 2012;Ortega-Hern andez and Brena, 2012;Vannier et al., 2014;Lerosey-Aubril, 2015). By their lobate shape, tubular network and tubule average diameter, the digestive glands of S. inexpectans recall the hepatopancreatic tissue of modern crustaceans ( Fig. Sup. 1G,H) that secretes digestive enzymes and assimilates food. The anterior part of the gut between the glands, although poorly delimited, is pear-shaped (Fig. 4C,F) and is interpreted as the anteriormost region of the midgut. Its swollen shape suggests a possible function in food storage and processing.

Oesophagus and mouth
Two specimens of Sidneyia show a short cylindrical feature within the sagittal plane of their body that runs across the boundary between the cephalic shield (CS) and at least the first trunk tergite (Fig. Sup. 2). This tube-like feature is less than 10 mm long and is lined with a ca 1 mm thick wall rich in carbon (EDX analysis and SEM backscattered images; Fig. Sup. 2B,C) and slightly expands posteriorly in a funnel-like manner. This feature is interpreted as the possible oesophagus of Sidneyia. In most modern arthropods, the oesophagus is lined with cuticle and is associated with ring muscles whose peristaltic contractions convey food into the gut ( Fig. Sup. 2F,G). The thick wall of the possible oesophagus of Sidneyia may represent the remains of a comparable musculature. In the laterally compressed specimen (Fig. 2), the assumed oesophagus lies ventral to the anterior part of the gut but how it connects to it remains unclear, probably because of the overlap of various anatomical features present in the head region. However, the oesophagus seems to loop upwards to join the midgut. The flared end of the assumed oesophagus ( Fig. Sup. 2B) is likely to represent the mouth of Sidneyia that clearly opened ventrally and posteriorly in a sagittal position, and probably close to the posterior margin of the first trunk segment. The anterior part of the digestive system of trilobites displayed a similar loop and posteriorly oriented mouth opening (Chatterton et al., 1994;Lerosey-Aubril et al., 2011).

Abdominal pocket
The gut of Sidneyia stretches markedly and relatively abruptly at the level of the posterior margin of the last trunk tergite to form a spacious, elongated abdominal pocket (AP) that fills a large portion of the abdomen being encased by the two cylindrical abdominal sclerites (Figs. 4A and 6; Fig. Sup. 3). The AP tapers gently at the telson level down to the anus opening. In dorsal view, the AP has a consistent elliptical or inverted pear-shaped outline. In numerous specimens, the AP is the only part of the digestive system to be preserved. By contrast with the digestive glands and the middle part of the gut, the AP has sharp, well-defined boundaries (Fig. 6AeC). The AP is typically filled with three-dimensionally preserved material (called gut contents here) that consists of more or less abundant exoskeletal elements of invertebrates (e.g. trilobite trunk tergites, cephala and pygidia) embedded in a mineralized matrix. No other region of the gut concentrates undigested elements in such a way, indicating that the AP had the probable function of storing food residuals. Moreover, very few specimens of Sidneyia have an empty AP. These two characteristics suggest that food did not run through the gut in a continuous flow and that the undigested fraction accumulated in the AP for some time before defecation. This feature is unique among Cambrian arthropods described so far and no exact equivalent of the AP of Sidneyia occurs in extant arthropods. However, interesting comparisons can be made with modern Arachnida such as the Amblypygi (Weygoldt, 2000), Anactinotrichida (Coons and Alberti, 1999), Araneae (Felgenhauer, 1999;Foelix, 2010), Pseudoscorpiones (Vachon, 1968), Ricinulei (Talarico et al., 2011) and Solifugae (Klann and Alberti, 2010). In these groups, the posterior portion of the midgut or the hindgut (in the Solifugae only) forms a pouch, the socalled stercoral pocket, that serves as a reservoir for fecal material ( Fig. Sup. 4). In the Araneae and Solifugae (at least), this pocket is surrounded by variously orientated muscle fibres. The stercoral pocket is assumed to perform various functions such as the storage of digested food and excretory material or even the reabsorption of water (Ueda, 1974;Butt and Taylor, 1986;Tillinghast, 1986;Coons and Alberti, 1999;Felgenhauer, 1999;Farley, 1999;Foelix, 2010;Klann and Alberti, 2010;Talarico et al., 2011). When empty, the AP of Sidneyia shows lineated transverse or criss-crossed structures ( Fig. 6DeI) that clearly belong to the wall of the AP and cannot be mistaken for gut contents. These structures represent either transverse folds in the wall of the AP and/or muscle fibres. One Sidneyia specimen displays a particularly well-preserved crissed crossed network of fibre-like structures ( Fig. 6DeI) that seems to be lining the wall of the AP. They are comparable in size and overall arrangement to the muscle fibres that surround the stercoral pocket of modern chelicerates ( Fig. Sup. 4F,I). The AP was probably flexible, being able to stretch in order to accommodate the successive input of food residuals, as indicated by its more or less inflated shape (Fig. 6;Figs. Sup. 3,4). In rare specimens (see also Caron et al., 2014), the AP appears to have been extruded from the abdomen (Fig. Sup. 5) as the possible result of a stress that may have preceded the animal's death. In conclusion, the AP is interpreted here as a possible functional analogue of the stercoral pocket of some modern chelicerates, the role of its assumed musculature being to expel food residuals and excretory material outside.

Gut contents of Sidneyia inexpectans
Recognisable, whole or fragmented exoskeletal elements (e.g. trilobite sclerites) and undetermined debris of most probably biological origin (e.g. black patches suggesting organic-rich material) were observed in the gut of 94 specimens of Sidneyia (Table 1). The vast majority of these elements are concentrated in the AP but some of them occasionally occur in the anterior and middle parts of the gut. SEM observations and EDX analysis of gut contents indicate that the material contained in the AP is particularly rich in Calcium and Phosphorus and therefore likely represents calcium phosphate ( Fig. Sup. 6). Natural sections through this material show that it is often capped with a thin layer of phyllosilicates that are similar (elemental composition, texture) to those of the host rock ( Fig. Sup.  6A,B). Elemental mapping of the exposed surface of gut contents confirms that calcium phosphate does not occur outside the boundary of the AP (Fig. 7, Fig. Sup. 6G). It is likely to have precipitated during early diagenesis within the AP and before the post-mortem collapse of the gut structures. Carbon is scattered all over the AP with patches of carbon films in places. Some exoskeletal elements have a phosphatic composition perhaps inherited from their original chemistry ( Fig. Sup. 6E,F), typically the phosphatic shell of inarticulated brachiopods. Even in those cases, however, it cannot be entirely ruled out that diagenetic phosphatisation occurred, considering that they are embedded within a matrix of calcium phosphate. More surprisingly, EDX analyses of trilobite sclerites show no clear Phosphorus or Calcium signals (Fig. 7A,C,D) suggesting that their 3D-preservation may be due to aluminosilicate coatings or replacement (see Butterfield et al., 2007). Pyrite, identified by the co-occurrence of Sulphur and Iron (Fig. 7G,I), is present in isolated spots and rarely prevalent. The origin of calcium phosphate in the gut contents and digestive glands of Sidneyia is discussed later in this paper (see section 6.4).

Food remains within the anterior gut
Remains of brachiopod shells concentrated between the first digestive glands were found in one specimen of Sidneyia (Fig. 8EeJ). They consist of broken elements and an almost complete valve (length and width approx. 2.5 and 1.5 mm) bearing symmetrical radiating ribs. These brachiopod elements may belong to Nisusia burgessensis Walcott, 1924(see Briggs et al., 1994. Their occurrence suggests that the enlarged interspace between the digestive glands received freshly ingested food. More questionable is the presence in one specimen of Sidneyia of a cluster of 5 agnostids (Ptychagnostus praecurrens Westergaard, 1936) in the central part of the cephalic shield ( Fig. 8AeD; see also Bruton, 1981, Fig. 99). These agnostids have a small size (cephalon width less than 3.5 mm) and are neither disarticulated nor broken apart, except for one specimen. This rounded cluster is located under the central part of the cephalic shield (Fig. 8B), thus matching the assumed position of the anteriormost part of the midgut housing freshly ingested food. However, the lack of visible landmarks such as the remains of digestive glands or gut boundaries makes it difficult to ascertain whether these agnostids are effectively ingested items preserved within the anterior midgut or result from chance superposition with Sidneyia. Agnostids such as Ptychagnostus are extremely abundant in the Burgess Shale (Caron and Jackson, 2008) and are often superimposed on the external surface of larger animals or within cavities (Chatterton et al., 2003). Based on other Cambrian occurrences, Fatka et al. (2009) and Fatka and Szabad (2011) also suggested that agnostids, because of their small size, could move into the cavities of carcasses for sheltering or scavenging. A group of agnostids may have effectively penetrated into the decaying carcass of Sidneyia, its cephalic shield offering a potential shelter and source of food. These two hypotheses (ingestion versus sheltering/scavenging behaviour) need to be tested by additional fossil evidence.

Food remains within the middle part of the gut
The middle part of the gut is relatively narrow and cylindrical (Fig. 2). Unlike the AP and the digestive glands, its external boundaries are usually poorly defined, discontinuous, if not absent. However, one specimen shows closely packed clusters of disarticulated sclerites of ptychopariid trilobites, possibly Ehmaniella burgessensis Rasetti, 1951, distributed along the midline of the body between the 5th and 8th trunk tergites (Fig. 9A,C). Their location, their disarticulated and/or fragmentary state, and the fact that they form clusters strongly suggest that they represent ingested elements that were transiting through the midgut and do not result from chance superposition. Similarly, an agnostid cephalon (presumably belonging to Ptychagnostus praecurrens) and a cluster of trilobite elements were found in the middle part of the gut of two other specimens (Fig. Sup. 7DeG). Some doubts remain however when elements overreach the assumed outer boundaries of the gut. This is typically the case of some elongated, curved, exoskeletal elements (e.g. Fig. Sup. 7AeC). Rather than storing food, the main function of this part of the digestive tract was most probably to ensure the transit of undigested residuals towards the AP, which would explain that gut contents rarely occur there. Insufficient concentration of Phosphorus and Calcium in the middle part of the gut due to its transit function may explain its poor preservation compared with that of the digestive glands and the AP. Trilobite elements are by far the most frequent exoskeletal remains within the AP of S. inexpectans. They were found in at least 13% of the Sidneyia specimens where gut contents are present in the AP ( Table 1). The trilobite elements are typically whole or broken cephala and pygidia, and more or less disarticulated thoracic tergites. Their overall morphology (e.g. cephalon) and ornament (e.g. pustules) best compare to Ehmaniella burgessensis, suggesting that it was the most frequent trilobite ingested by Sidneyia. E. burgessensis is common in the Burgess Shale biota and is very often associated with Sidneyia on numerous rock slabs. The majority of ingested trilobites belong to juvenile stages (probably meraspis).
The size of exoskeletal elements found in the AP is a good indicator of the size range of trilobites that Sidneyia had the ability to ingest. One complete subadult specimen of Sidneyia (ca 90 mm long) displays a cephalon in its AP that corresponds to a juvenile stage of E. burgessensis of approximately 5.5 mm long and 3.5 mm wide (Fig. 9AeE). Similarly, an almost complete undetermined juvenile trilobite (not Ehmaniella; possibly a corynexochid such as Olenoides or Kootenia, based on its long pleural spines) of about 3 mm long and 2.3 mm wide occurs within the AP of another specimen of Sidneyia (Fig. Sup. 7L,M). The assumed oesophagus of Sidneyia had a diameter of about 1 mm with a possible funnelshaped mouth opening reaching about 2.5 mm wide or more ( Fig. Sup. 2B). This size range is consistent with the ingestion of such small trilobites provided that the oesophagus could slightly stretch open to allow food to pass into the anterior midgut. It is the case in the majority of modern crustaceans in which the oesophagus is a tube equipped with ring muscles that can expand and contract (e.g. in myodocope ostracods; Fig. Sup. 2F,G). The trilobite elements often show obvious signs of disarticulation and breakage (Fig. 9EeG) that probably resulted from the grinding action of gnathobases (Fig. 3) prior to ingestion. No appendages of trilobites were found in the AP of S. inexpectans. Their non-biomineralized cuticle and internal soft tissues (e.g. muscles) were probably resorbed through enzymatic digestion in the anterior part of the gut. The absence of large exoskeletal fragments in the AP indicates that Sidneyia neither preyed upon nor scavenged on adult or subadult trilobites. This selectivity towards feeding on immature individuals might have been related to the fact that the appendages of Sidneyia, especially their gnathobases, were not powerful and/or resistant enough to break the heavily mineralized cuticle of adult trilobites. We must also consider that adult trilobites could probably escape more easily and were possibly less abundant than juveniles.

Brachiopods
Brachiopods are less frequent (ca 6%; Table 1) than trilobites in the AP of Sidneyia and are represented by various species, none of them being dominant. Shell fragments of Micromitra burgessensis (Resser, 1938), with their readily recognizable reticulated ornament are present in one specimen (Fig. 10AeC). Costellate, articulate brachiopods also occur in a few specimens and most probably belong to Diraphora bellicostata, Walcott, 1924 or Nisusia burgessensis (Walcott, 1889) (Fig. 10FeH). Another form represented by three valves bearing fine concentric growth lines was found in the AP of one specimen of Sidneyia (Figs. 6B and 10D,E) and might belong to Lingulella waptaensis Walcott, 1924.

Bradoriid arthropods?
More enigmatic are tiny skeletal elements of less than a millimetre length that occur in at least two specimens of Sidneyia (Fig. 10M, Fig. Sup. 8). Although partly covered by sediment, their outline seems to be circular with no dividing line possibly representing a hinge of a dorsal fold such as in the shell/carapace of typical 'bivalved' invertebrates (e.g. brachiopods, arthropods). They are preserved in calcium phosphate and are characterized by a very fine, concentrically arranged reticulated ornament and a ridge running parallel to the margin. Their overall morphology, ornamented pattern and chemical composition recall those of bradoriid arthropods, especially Flumenoglacies michaeli Andersson, 2014 from the middle Cambrian Stephen Formation of British Columbia, Canada (Andersson, 2014). However, this bradoriid species has not been reported so far in the Burgess Shale Formation. Resemblances with the ornament of Zepaera sp. (Wrona, 2009) and Kunmingella sp. (Zhang, 2007) are also worth being noted. Perhaps these enigmatic tiny remains belong to a new bradoriid species (J.-B. Caron personal communication). The lack of dorsal split (e.g. fold) and anteroeposterior polarity that normally characterize the carapace of subadult bradoriids might indicate that these tiny shells are juvenile stages. Univalved, subcircular carapaces with an almost circular outline have been reported in the juvenile stages of bradoriids such as Kunmingella douvillei (Mansuy, 1912) from the early Cambrian of China (Zhang, 2007; pl. 1). The AP  of one specimen of Sidneyia contains a pustulated shelly fragment also of possible bradoriid origin (Fig. 10IeL). Finally, it is important to note that the bradoriid Liangshanella burgessensis Siveter and Williams, 1997 which is by far the most abundant animal in the Greater Phyllopod Bed of the Walcott Quarry Shale Member (Caron and Jackson, 2008) does not occur in the AP of Sidneyia.

Other animal groups
Approximately five-to-eight millimetre long, slender, slightly curved elements occur in the AP of a few specimens of S. inexpectans or are adjacent to it (Fig. Sup. 9). They are preserved in aluminosilicates, iron oxides, calcium phosphate ( Fig. Sup. 9C,D) or partly as carbon films (Fig. Sup. 9F,H). Whether some of these elongated elements belong to gut contents or are superimposed to them by chance is uneasy to determine (e.g. Fig. Sup. 9B,C). Their shape recalls that of various sclerites of Burgess Shale animals such as the helens of hyolithids (e.g. Haplophrentis carinatus (Matthew, 1899) or the long spiny sclerites of the halwaxiid Orthrozanclus reburrus (see Conway Morris and Caron, 2007)). However, considering the lack of conical hyolithid shells within the AP (see Vannier, 2012) and the fact that the curved elements are never numerous, it seems unlikely that Sidneyia actually fed on either of these two animals. Some of the slightly curved elements superficially resemble the tubes of Tubulella flagellum Matthew, 1899 (see Van Iten et al., 2012), a possible sessile but poorly documented cnidarian from the Burgess Shale (Fig. Sup. 9FeH). The arthropod Marrella splendens Walcott, 1912 (Garcia-Bellido andCollins, 2006) would appear as an interesting alternative if we consider that the curved elements found in the AP of Sidneyia might be the undigested head spines of this extremely abundant arthropod of the Burgess Shale. However, the presumed high mobility of Marrella on and above the seafloor probably strongly reduced its chances to be captured alive by epibenthic consumers such as Sidneyia. Regardless of their origin, the curved elements are relatively rare and simply may have been ingested by chance along with consumable food. Rare annulated fragments were also found within the AP of two specimens of Sidneyia (Fig. Sup. 9IeK). Based on their tubular shape and wrinkled appearance, they are tentatively interpreted as cuticular remains of worms, possibly of juvenile stages of Ottoia prolifica, a common priapulid worm of the Burgess Shale fauna (Conway Morris, 1977;Vannier, 2012).

The feeding mode of Sidneyia inexpectans
Numerous Cambrian arthropods with assumed predatorial habits (e.g. anomalocaridids, Isoxys, Kodymirus, Yohoia; Vannier et al., 2009Vannier et al., , 2014Haug et al., 2012a,b;Daley et al., 2013;Lamsdell et al., 2013;Legg and Vannier, 2013;Daley and Edgecombe, 2014;Cong et al., 2014) had well-developed grasping frontal appendages. S. inexpectans lacked such prehensile appendages pointing forwards. Its flexible, flagelliform antennae (Fig. 8A,B) had an obvious sensory function and were not involved in manipulating food. Moving food to the ventral gnathobases (Fig. 3E) seems to have been the initial step of the feeding process. Although the primary function of ventral endopods was locomotion, we assume that the most anterior ones played an important role in manipulating food and moving it to the gnathobases. Their clawlike termination, articulated design and possible pincer-like structure (Fig. 3) strongly suggest capacities for pushing, seizing and directing small prey or food particles to the mouth area. One specimen of Sidneyia preserved in ventral view clearly shows that the basal parts of the four anteriormost appendages converged towards the sagittal axis of the animal where the posteriorly directed mouth was supposed to open (Fig. 11A,B). The converging motion of the gnathobases had the effect of disarticulating and crushing food such as the exoskeleton of small trilobites and the Table 1 Count data of the gut contents of Sidneyia inexpectans Walcott, 1911 from the middle Cambrian (Series 3, Stage 5) Burgess Shale, British Columbia, Canada. The three columns correspond to the number of Sidneyia specimens with specific gut contents (e.g. trilobite elements) in the anterior part (between the digestive glands), the middle part of the gut and the abdominal pocket. The total number of Sidneyia specimens with gut contents is 94. USNM, collections of the National Museum of Natural History, Smithsonian Institution, Washington D.C.; ROM, collections of the Royal Ontario Museum, Toronto, Canada. RQ/RT, Raymond Quarry/Talus; WQ/WT, Walcott Quarry/Talus; others, specimens from other localities (see Table Sup. 2).

Anterior part
Middle part Abdominal pocket  shells of brachiopods prior to its ingestion through the oesophagus. The physical breakdown of food exerted by the gnathobases was sufficient for millimetre-sized items to be sucked into the gut via the assumed peristaltic contractions of the oesophagus (Fig. Sup. 2). The grinding power of gnathobases was however limited since no thick sclerotized or mineralised fragments are found within the gut contents of Sidneyia. A comparable mechanism, involving opposing and converging pairs of gnathobases, has been observed in modern horseshoe crabs (Fig. 11C,E), that easily crush the shells of small bivalves and soft-bodied animals (Botton and Shuster, 2003). In L.
polyphemus, food is pushed into the funnel-like mouth opening by the spiny tips of gnathobases and pumped into the oesophagus (Fig. 12). In Sidneyia, ingested food passed from the oesophagus into the pear-shaped anterior part of the midgut that probably functioned as a temporary storage pocket. The enzymatic chemical breakdown of food most probably took place within this structure, enzymes being secreted within the tubules of the digestive glands attached to it and drained into the anterior midgut through ducts.
The digestive glands of Sidneyia performed the same function as the hepatopancreas of modern horseshoe crabs and crustaceans (EeJ) ROM 56945,1, brachiopod fragments between the first pair of digestive glands, interpreted as being located within the anterior part of the gut; general view, details of the whole cluster and of individual fragments. Scale bars: 1 cm in A, B and E; 5 mm in C and D; 1 mm in F-J. Abbreviations: ag, agnostid cluster; an, antenna; as, abdominal sclerite; br, brachiopod fragments; cs, cephalic shield; dg1edg2, 1st and 2nd pair of digestive glands; e, eye; pa, postantennal appendage; tt4, 4th trunk tergite.
( Fig. Sup. 1G,H). The relatively small diameter of their tubules (less than 200 mm) was not consistent with food particles entering the ramified network of the glands. Where assimilation took place is less clear. In modern horseshoe crabs, lipids, glycogen and proteins resulting from digestion are stored in connective tissues surrounding the tubules of the hepatopancreas (Fahrenbach, 1999). Similarly, the digestive glands of Sidneyia formed relatively large organs that most probably housed tubules and storage tissues with various cellular types for both digestion and assimilation. After a short transit through the middle part of the gut, digested food residuals (mainly shell fragments and sclerotized exoskeletal elements) and excretory material accumulated within the abdominal pocket. Although this pocket may be seen as a functional equivalent to the stercoral pocket of modern arachnids, important differences should be noted. In some arachnids (Araneae, Ricinulei), the pocket is a diverticle of the gut (extensible blind sack; Fig. Sup. 4; Ludwig et al., 1994;Foelix, 2010;Talarico et al., 2011), whereas it runs in straight continuity with the digestive tract in Sidneyia. The abdominal pocket of Sidneyia was probably lined by muscle fibres (Fig. 6) arranged in different directions allowing residues to be excreted through circular and/or longitudinal muscular contractions. Although morphofunctional comparisons with horseshoe crabs are valuable (Figs. 11 and 12), the digestive system of Sidneyia seems to differ markedly from that of L. polyphemus ( Fig. 12A; Fahrenbach, 1999). Sidneyia displays a swelling in its anterior midgut but does not seem to have possessed a gizzard for mashing food material as in modern horseshoe crabs (Fig. 12A). The hepatopancreas, which occupies much of the prosoma in Limulus (Packard, 1880;Fahrenbach, 1999;Shuster and Sekiguchi, 2003), is represented in Sidneyia by much smaller glands in proportion to the body size of the animal (Fig. 4). The storage of faeces within the abdominal pocket suggests infrequent feeding.

The diet of Sidneyia inexpectans
Sidneyia fed upon a variety of organisms such as trilobites, brachiopods, possibly agnostids and priapulid worms and various undetermined invertebrates represented by their exoskeletal elements (e.g. possible bradoriids, Marrella). Juvenile stages of trilobites seem to have been the most frequent prey of Sidneyia as indicated by the relatively large number of their exoskeletal fragments stored in the AP. Some elements might have been ingested incidentally (e.g. some elongated/curved elements; see section 5.3.4). Sidneyia was not a sediment eater, no significant proportion of sediment being ever found in its gut. The overall shape and appendage design of Sidneyia is typical of an epibenthic crawler that fed on small sessile or relatively slow moving animals living at the water-sediment interface. Bradoriids overwhelmingly represented by Liangshanella, agnostids and small trilobites such as Ehmaniella were among the most abundant arthropods of the Burgess Shale biota (e.g. Caron and Jackson, 2008 and J.-B. Caron, personal communication) as indicated by their numerical abundance on rock slabs. Whereas Sidneyia did feed on small trilobites and possibly agnostids, Liangshanella was absent from its diet. This might stem from the possibility that Liangshanella was a vagile, swimming arthropod that Sidneyia could not capture.
Hyolithids that were a major food source for the carnivorous worm Ottoia prolifica (Vannier, 2012) are absent from the gut contents of Sidneyia. Whether hyolithids were avoided by Sidneyia relative to other food items is impossible to ascertain. Contrasting with Sidneyia, the feeding mechanism of Ottoia involved no appendages and was based on the pharyngeal eversion/inversion that drew prey into the gut, a mechanism that may have favoured the ingestion of slender, conical organisms such as hyolithids. Perhaps, the functional morphology of the mouth and oesophagus of Sidneyia did not permit such food items to be ingested easily. Although the ventral appendages of Sidneyia were obviously able to manipulate a wide range of live and dead food items (e.g. via their gnathobases), only a small proportion of them could probably find its way to the mouth to be eventually ingested via the oesophagus. Sidneyia was possibly a predator but may have had scavenging habits as well. We see no biomechanic constraints that would have prevented this arthropod from processing carcasses and extracting food from them through the tearing and grinding action of its gnathobases. The food items that could be identified with certainty in the gut contents of Sidneyia may account for a small fraction of the actual diet of Sidneyia, which possibly fed upon a much more diverse food source including a great variety of soft-bodied animals and larval stages. We must keep in mind that soft tissues (e.g. plucked from carcasses) and thin cuticular elements may have been completely resorbed through digestion, leaving no identifiable traces in gut contents. No fossil evidence suggests that Sidneyia was a selective feeder. Its indiscriminate feeding mode, however, must have been strongly constrained by biomechanical factors, such as the diameter of the oesophagus and the grinding power of ventral appendages.
Sidneyia provides by far the most detailed and valuable source of information concerning the diet and the feeding mode of a Cambrian arthropod. Indeed, if we except the pioneer descriptions of Sidneyia by Bruton (1981), ingested food items have been described in three Cambrian arthropods only (one specimen in each case). Eodiscoid sclerites (Pagetia; Zhu et al., 2004) were recognized in the gut a "Fuxianhuia-like" arthropod from the middle Cambrian Kaili biota (ca 508 Ma, China;Zhao, 2011). This specimen has a series of well-developed paired digestive glands preserved in phosphate. But unlike Sidneyia, they are uniformly distributed along the major part of its gut and the gut tract is filled with trilobite sclerites throughout its length. The holotype of the arthropod Utahcaris from the middle Cambrian Spence Shale of Utah preserves a cluster of fragmented sclerites of trilobites under its cephalon, which was interpreted as stomach contents by Conway Morris and Robison (1988). Information concerning the diet of early Cambrian arthropods is extremely fragmentary. However, a new arthropod from the Emu Bay Shale Lagerst€ atte (Cambrian Series 2, Stage 4), probably allied to Sanctacaris and Utahcaris (Jago et al., 2014), has identifiable fragments of the trilobite Estaingia bilobata (Pocock, 1964) in its gut. Whether this arthropod was a predator or a scavenger could not be determined.
were an important source of food for much larger arthropods in Cambrian ecosystems.

Ecological impact of Sidneyia inexpectans
Sidneyia was a relatively large (maximum length ca 160 mm) epibenthic arthropod with predatory and possibly scavenging habits. Although it is difficult to have a precise idea of its population density, Sidneyia is a frequent arthropod throughout the Greater Phyllopod Bed (Walcott Quarry) of the Burgess Shale (Caron and Jackson, 2008) and is very common in the Raymond Quarry (J.-B. Caron, personal communication). The impact of Sidneyia predation on the populations of small epibenthic arthropods, such as trilobites (e.g. Ehmaniella), should be considered, because it especially affected juveniles. Although it is impossible to make a quantitative assessment of predation pressure in the Burgess Shale biota, recent models support this hypothesis. Thin-shelled bivalves (e.g. Mytilus, Mya and Spisula) are one of the major sources of food for modern horseshoe crabs, such as L. polyphemus (Botton, 1984;Botton and Shuster, 2003). Quantitative studies demonstrate the predation impact of horseshoe crabs on these bivalve populations (Botton and Shuster, 2003). The fact that they preferentially feed on individuals within a specific size range seems to have an important effect on the composition, growth and survival of these populations. That Sidneyia may have exerted a comparable pressure on certain trilobite populations should not be excluded. Sidneyia may have been an occasional scavenger on various carcasses of Abbreviations: an, antenna; ap, abdominal pocket; as1eas2, 1st and 2nd abdominal sclerite; ch, chelicera; cs, cephalic shield; e, eye; gn, gnathobasis; gz, gizzard; lp, leg podomeres in section; mo, mouth opening; oe, oesophagus; op, opisthosoma; pa1, 1st pair of postantennal appendages; tc, terminal claw; tt9, 9th trunk tergite; 1e5, 1ste5th pair of leg. small or larger invertebrates. Its role as a possible epibenthic recycler should be considered.

Phosphatic preservation of the digestive system of Sidneyia inexpectans
Phosphatic mineralisation is responsible for the threedimensional preservation of the major part of the digestive system of Sidneyia, especially its digestive glands and AP (Figs. 4 and 7,respectively). This questions the origin of phosphorus and calcium that led to the mineralisation and preservation of the digestive soft tissues. In numerous extant arthropods, the epithelial cells of the midgut or the midgut glands contain calcium phosphate mineral concretions (spherites). These spherites play an important role in the storage of P and Ca and are usually released into the gut lumen to be eventually recycled in vital processes such as moulting (hardening of the new cuticle). This is clearly exemplified by horseshoe crabs in which the absorptive cells of the hepatopancreas release important quantities of calcium phosphate spherites into the lumen of tubules (Fahrenbach, 1999;Botton and Shuster, 2003) that pass into the alimentary canal and add to faeces (Lockhead, 1950). An adult specimen of L. polyphemus was dissected in order to analyse its gut contents (Fig. 12). They appeared in the form of a shapeless mass of undigested residues wrapped within a thin translucent membrane (Fig. 12D,E), the posteriormost part being filled with spherites ( Fig. 12F,G). These spherites of about 5e10 mm in diameter have a phosphatic composition (EDX ana-lysis). If we assume that Sidneyia had comparable physiological capacities to produce, store and recycle phosphate, then the 3D-preservation of its digestive anterior midgut; an, anus; do, opening of hepatopancreatic duct; du, hepatopancreatic duct opening into the gut lumen; gc, gut content; gn, gnathobasis; gu, gut; gw, gut wall; gz, gizzard; he, hepatopancreas; le, leg; mo, mouth opening; oe, oesophagus; op, opisthosoma; pr, prosoma; pv, pyloric valve. glands could be easily explained by early mineralisation processes driven from an internal in situ source of P and Ca, possibly in the form of spherites. However, the clustered apatite microcrystals (<0.5 mm; Fig. 4G) found in the digestive glands of Sidneyia are much smaller than the phosphatic spherites of Limulus. They might represent a recrystallized form of original spherites that were resorbed immediately after death. The concentration of spherites in the posterior gut of Limulus (Fig. 12) also offers a convincing explanation for the presence of phosphate in the abdominal pocket of Sidneyia. Again, the presence of P and Ca in situ among undigested residues might explain the post-mortem mineralization of both the surrounding muscles of the AP and the gut contents themselves. A comparable process has been proposed to explain the 3D-preservation of the digestive structures of Leanchoilia from the Burgess Shale (Butterfield, 2002) and trilobites from the Weeks Formation (Lerosey-Aubril et al., 2012). The midgut glands of Leanchoilia contain numerous microspheres (diameter 2e3 mm) that may correspond to the intracellular spherites found in the midgut glands of modern arthropods (Butterfield, 2002). The replacement of soft tissues with calcium phosphate, whatever its origin, requires special physicochemical conditions. Decay experiments with shrimps show that the mineralisation of soft tissues in phosphate starts within two weeks after death and requires an elevated concentration of phosphate in sediment pore waters and a drop in pH, and is favoured by a system closed to diffusion Kear, 1993, 1994;Hof and Briggs, 1997). The abdominal pocket and the blind tubular network of the digestive glands of Sidneyia both show all the characteristics of relatively closed systems in which phosphatic precipitation may have been optimal.

Conclusion
Our study reveals major functional aspects of the digestive system of a Cambrian arthropod with unprecedented accuracy allowing precise comparisons with extant arthropods. S. inexpectans displays a remarkable set of digestive glands and a spacious abdominal pocket for storing undigested elements. This pocket has no functional equivalent in extant arthropods except some arachnids. The analysis of its gut contents tells us about the diet of Sidneyia that included juveniles of trilobites, brachiopods and other epibenthic invertebrates.
Taking Sidneyia as an example, we show that significant information on the feeding ecology and digestive systems of early arthropods can be obtained through the detailed observation of fossils from Konservat-Lagerst€ atten, which invites further research on the Cambrian and Ordovician Burgess Shale-type exceptional biotas. Our approach, which considers fossils more as living entities than as dead organisms preserved in rocks, has great potential to unveil many aspects of the early organization and functioning of animal life that are still unknown or have to be exploited. This type of approach has been used recently to characterize key-anatomical features of early animals, such as the nervous, sensory and digestive systems of arthropods (e.g. Ma et al., 2012;Strausfeld et al., in this issue;Lerosey-Aubril et al., 2012, Vannier et al., 2014, to analyse their gut contents (e.g. worms; Vannier, 2012), and to reconstruct their functional morphology via computer modelling (e.g. Haug et al., 2012b). Recent studies have attempted to reconstruct Cambrian food webs via the computerized analyses of data obtained from the Chengjiang and Burgess Shale biotas (Dunne et al., 2008). They provide the opportunity to compare the trophic complexity at different stages in the evolution of life on Earth. However, the validity of these computerized models is dependent on the quality of data obtained from the literature and often suffers from erroneous or incomplete assumptions and crude extrapolations based on recent biological models. Future palaeoecological models need to be directed by new and more reliable information sources such as those presented here.