Internal conulariid structures unveiled using µCT

An extensive sample of well-preserved conulariids from the Pennsylvanian of the North American Midcontinent (Texas and Oklahoma, USA) have been studied using X-ray micro-Computed Tomography (µCT) and have shown structures identified as longitudinal muscle bundles and a potential gastric cavity. These unequivocal structures appear in several specimens coming from different sites. Their preservation varies from a gastric cavity with muscle bundles in some individuals to only longitudinal muscle bundles in others. The muscle bundles fuse apically or medially, normally forming V-shaped pairs, and they extend along the theca/exoskeleton, parallel to the corner, towards the aperture. Longitudinal bundles have predominant perradial positions. Although there have been some articles on conulariid soft parts, most of them refer to relic soft parts. This is the first time that these structures are shown using µCT. Discovery of conulariid soft parts contributes to knowledge of metazoan evolutionary history.


Introduction
Conulariids have been studied for more than two centuries, since the first time that a Carboniferous specimen was figured by "the father of Scottish palaeontology", Reverend David Ure, in 1793 (pl. 20: Fig. 7). Since then, when conulariids were referred to as "very rare" (Ure 1793: p. 331) until the present, much progress has been made in the study of the group's external morphology, but very little with respect to their soft parts. It has been assumed that soft parts can be inferred from preserved hard parts, but proper soft part anatomy has never been documented.
Although Barrande (1855aBarrande ( , b, 1867 already highlighted the conulariid organic deposit had to build the schott, the first authors to speculate about conulariid soft parts were Bouček and Ulrich (1929 : p. 206) when they inferred that the muscles, which closed and opened the aperture, could only be attached to internal septa. Some years later, Kiderlen (1937) was the pioneer in describing conulariid relic soft parts in detail, elucidating muscular attachments in the aperture. After observing numerous specimens from different sites around the world and some of them with several schotts, Kiderlen (1937) came to the conclusion the soft body was firmly attached to the theca and completely lined it with its "epithelium", with the exception of the aboral area. He also described how the aperture could have only been closed by muscle contraction, these muscles being attached to the septa and also responsible for the ribs' alternation, or not, at the midline. Kiderlen's concepts were fundamental to support a cnidarian affinity for the conulariids and have been echoed until nowadays. Moore and Harrington (1956), in the Treatise on Invertebrate Paleontology, also made inferences on the origin of the soft parts responsible for building the theca, writing about its ectodermal origin. This was based on observations of 1 3 individuals with encrusting organisms on their four faces, but maintaining erect positions when they were alive. If the theca had an endodermal origin, it would have been secreted covering the external parts and none of the encrusters could have been attached. A few years later, Werner (1966Werner ( , 1967Werner ( , 1971) compared the relic soft parts of conulariids to coronate scyphozoan polyps, interpretating conulariids as tentaculate animals, as Kiderlen (1937) and Moore and Harrington (1956) did as well.
The end of the twentieth century was the peak moment of the study of fossils using X-rays. Steul (1984) used X-ray radiographs to study conulariids from the Hunsrück Slate, Lower Devonian of SW Germany. She described and illustrated putative soft parts seen through X-Ray, but unfortunately was misled by a nautiloid siphuncle (see Hergarten 1994 for explanation) and the rest of the specimens could not be re-examined as their origin and whereabouts are mostly unknown. Babcock and Feldmann (1986b: Fig. 30.5) illustrated one of eight specimens from the Lower Mississippian of Kentucky and Alberta, studying soft remains using X-ray radiographs. They also showed images of specimens with remains of soft parts. They described and drew them as a single elongate tube and globular masses (Babcock and Feldmann 1986a). The elongate structure is always located near a corner, extending most of the theca length and turning at an angle of almost 90° to the long axis of the theca close to the aperture. They interpretated the longitudinal structures as remains of organ systems. These researchers were the first ones to illustrate presumed soft parts. Much later, Van Iten and Südkamp (2010) also studied Hunsrück conulariids with X-rays and observed irregular concentrations of pyrite, some oriented more or less parallel to a midline or corner or obliquely to the specimen's long axis, close to the aperture, but without a clear structure, assuming these concentrations were decayed remains of soft parts.
Special mention is deserved for a recent publication by Miller et al. (2022). This research is based on scanning electron microscopy (SEM) with energy-dispersive spectrometry (EDS) of Silurian Waukesha Lagerstätte conulariids. These specimens underwent a complex taphonomic history that did not leave clear soft parts preserved, but kerogenized patches and carbon films within the theca remains, assuming these kerogenized patches could have been derived from decayed organic material. Therefore, this is the first time that assumed internal structures of conulariids are seen using µCT. Here we report on the most complete, and best preserved specimens in three dimensions of conulariids from the Wewoka and Graham formations, Pennsylvanian, of Oklahoma and Texas, USA ( Fig. 1), respectively. These fossils show a distribution of longitudinal bundles inside of the theca, from the proximalto-distal regions, and what we consider remains of muscle tissue, probably muscle bands that would have been attached to corner carinae. The presence of this structure in conulariids becomes really meaningful and will be compared to recent cnidarian groups in a consecutive paper. Most of these specimens show an almost perfect radial tetramerous symmetry of their thecae.

Geological setting
The specimens described in this paper were collected from the Pennsylvanian of the North American Midcontinent (Fig. 2), in Central Oklahoma and North-Central Texas (USA). This area has been well-studied stratigraphically due to the perspective of regional exploration for hydrocarbon reserves. Pennsylvanian strata in this area are dominated by transgressive-regressive lithic sequences, or cyclothems (Heckel 1991). Maximum transgression is characterised by deposition of organic-rich black mud.
The Middle-Upper Pennsylvanian stratigraphic sequence in the Midcontinent formed north of the equator in the tropical trade wind belt of that time. Though deposition was generally slow in the North America Midcontinent during this age, it was more rapid near active detrital sediment sources, particularly in central Oklahoma. The conulariids were found in the fissile black shale (Malinky and Heckel 1998), in the offshore part of the Pennsylvanian cyclothems. This shale indicates dysoxic conditions rather than fully anoxic conditions that prevailed at least on the seafloor from time to time. This should be the reason for the low biodiversity found in these levels, probably with an oxygen content between 0.7 and 0.3 ml/l (Demaison and Moore 1980) that would be maintained during long periods. The Wewoka and Graham formations from Central Oklahoma and North Texas, respectively, (Fig. 3) have harvested hundreds of specimens that we have studied for this research. Some of these were collected near Atwood, Hughes County, probably in the Lower Wewoka Member, in what has been named beds of less resistant shales (Weaver 1954). The palaeoenvironment of these shales probably included turbidity currents which carried fine-grained material. These deposits have been able to preserve what we consider soft parts of an abundant and low diversity population of individuals which suffered from predation. Some of the individuals studied from both Pennsylvanian sites have evidence of bioturbation, which means the sea bottom was not hostile to life at least intermittently.
Hughes County is on the southeast edge of the central Oklahoma uplift, with thick deposits of lower Desmoinesian age which were deposited in a shallow sea covering this basin. Outcrops of middle and upper Desmoinesian age are spread throughout this county. Close to the Canadian River Valley, near Atwood and Holdenville Lake, there are friable sandstones and shales, mostly massive, alternating beds of about 200 m thickness (Weaver 1954) that conform to the Wewoka Formation. The shales are by far the more dominant lithology. The fossiliferous clay shales that house the conulariids can be seen in the Lower and Middle members. The Middle Wewoka Member shales stand out for the abundant and perfectly preserved fossil shells of brachiopods and molluscs (Weaver 1954;West 1970) that can be also extended to the conulariid thecae. Numerous fossil sponges, echinoderms, bryozoans, trilobites, corals, fishes and plants have been very well preserved as well. As the sandstones and shales may be discontinuous in many areas, intertonguing with the overlying and underlying shales, we are not sure if some of the conulariids collected near Atwood belong to the Lower Wewoka Member (marked with a question mark on Fig. 3). The lower part of the Graham Formation, the Finis Shale, is very well exposed at the Lost Creek Lake spillway site ( Fig. 2a close to Jacksboro Lake) with fossils in exceptional conditions, mostly found in situ, in black to dark-grey shale, with siderite nodules as well. The Finis Shale fauna is considered one of the best sampled and preserved fossil assemblages known to science. Conulariids appear in the Crurithyris-Paraconularia Community of Lobza et al. (1994), in a relatively firm mud bottom with shallow water depths changing to deeper and calm water conditions. The fact of the relatively small number of individuals per species and their size, larger than in overlying deposits, suggests a high rate of sedimentation. The energy for the transport was probably low. Although anoxic conditions within the substrate are evident from the pyritic preservation of fossils, the abundance of preserved organic matter and the few infaunal species suggest overlying water was oxygenated.
The Pennsylvanian area of North-Central Texas has been described in reference to the sedimentary formations of the Midcontinent as two great inliers of Carboniferous sediments that protrude through Cretaceous strata on the east and dip beneath Permian rocks on the west and north of North-Central Texas (Plummer 1919). In the vicinity of Jacksboro, Jack County, there are rocks which are part of the Canyon and Cisco groups of Pennsylvanian age. The Cisco Group was deposited during increased orogenic activity in the mid-continent. Its deposition was controlled by the complex interaction of delta progradation, basin subsidence, shift stability and terrigenous clastic input from the surrounding mountains to the east in conjunction with eustatic sea-level fluctuations derived from glaciations in Gondwana (McLeod et al. 2003). It involves the north-west one-third of the county. The basal Cisco Group outcrops near Lost Creek Reservoir Spillway, at the north end of Lake Jacksboro (northeast of Jacksboro city), contain the Finis shales ( Fig. 3; lower section of the Graham Formation, Upper Pennsylvanian), very prolific in well-preserved fossils. It has a thickness of about almost 10 m. The Finis Shale has been interpreted to be the deep-water shale component of a typical Midcontinent cyclothem (Boardman et al. 1984). A cyclothem sequence usually consists of, from base to the top, a nearshore shale, a transgressive limestone, a deepwater shale, a regressive limestone and another nearshore shale. At the base, phosphatic, black shales were deposited in anoxic waters below a thermocline, with a high rate of organic decomposition and bacterial interaction. At this level there are mostly pelagic and nektonic remains, and occasional presence of conulariids, meaning that dysoxic, instead  with correspondence with the regional stages and conulariid occurrences of anoxic, conditions prevailed on the seafloor from time to time (Malinky and Heckel 1998). Overlying the phosphatic black shales are medium to dark gray shales, normally deposited under shallower oxygenated waters with high biodiversity (Boardman et al. 1984). Low current activity and sedimentation rates allowed good preservation in situ. Finally, towards the top of the Finis Shale, there are medium to light gray shales that are highly fossiliferous, with pyritisation of some specimens. The paleofauna found at this level is typical of shallower water deposition (Kocurko 1993), restricted to shallow-shelf muddy siliciclastic sediments. This makes the lower Virgilian Finis Shale one of the most richly fossiliferous exposures of marine shelf sediments in the United States (McLeod et al. 2003) and they have yielded more than 100 conulariids of the same species. The best preserved specimens, 74 individuals, have been studied. The other upper half of the outcrop is also fossiliferous, but with thin discontinuous sandstones lenses.

Materials and methods
This study is based mainly on material collected from the Wewoka and Graham formations, Pennsylvanian, of Oklahoma and Texas, USA. A total of 298 isolated specimens of 'Conularia crustula' (White 1880) from the Wewoka Formation and Paraconularia sp. from the Graham Formation, all preserved in three dimensions, have been studied. 'Conularia crustula' is common in the Carboniferous of Texas and Oklahoma. Most of the specimens are small, a few centimetres in length as adults. Although C. crustula has been traditionally assigned to the genus Paraconularia by many palaeontologists, White (1880: p. 170, 1881: p. XXVIII) described this species with transverse ribs that cross the sulcate corners, bending slightly adaperturally ("striae …. across the angle-furrows, in crossing which they bend slightly backward"). Consequently, if the transverse ribs are not interrupted at the sulcate corners as White's (1880: pl. 42: Fig. 4a, 1881 illustrations showed, this species would not belong to Paraconularia Sinclair, 1940. We therefore recommend redescribing this species after examining the type specimens and will name these specimens under open nomenclature as 'Conularia crustula'. The Finis Shale specimens are smaller, with lower apical angle, and whose transverse ribs are interrupted in the corners. We did not find any species already described which fit with our specimens from the Finis Shale. The North American specimens were distributed between the Wewoka Formation (168) and the Graham Formation (130). All these North American specimens have been collected and donated to the Natural History Museum, London, by one of the authors (WR). Two additional individuals of Paraconularia 'quadrisulcata' (see Sendino 2021: Fig. 9 for explanation on open nomenclature here) from the Silurian of Dudley, Worcestershire (UK), and the Carboniferous of East Renfrewshire (Scotland) have been studied. The reason for studying these individuals is because they are the best preserved specimens in the Conulariida collection at the Natural History Museum, London (NHMUK), preserving their original three-dimensional shape and aperture, making them candidates to study with Micro-CT scanning. All this material used in this study is deposited in the NHMUK.
We made a selection of the best specimens, because they preserve the aperture or keep the most external layer of the theca, or because they have an unusual structure, such a conulariid theca inside of another conulariid theca (Fig. 4), which was already reported in living coronates (Morandini and Jarms 2010). A total of 60 individuals from the Pennsylvanian of the North American Midcontinent and two from Silurian and Carboniferous of England and Scotland, respectively, were chosen, which were firstly radiographed in order to make another selection with those candidates to show soft parts (Fig. 5). The specimens studied here were unprepared except for mechanical removal of minor amounts of rock matrix. 33 individuals were scanned with Carl Zeiss Xradia Versa 520 Micro-CT scanner, which combines classic X-ray geometric magnification with different optical lenses. They were analysed at a range of source conditions; 60 to 160 kV (adjusted depending on specimen size). Data were reconstructed using the ZEISS software from 1601 to 3201 projections taken over 360°. The reconstructed images were then segmented and rendered using Avizo software (ver. 2019.1 Amira-Avizo, Thermo Fisher Scientific, USA), highlighting the internal structures in different colours.
For scanning electron microscopy (SEM) and energydispersive X-ray spectrometry (EDS), selected samples were embedded in epoxy resin, cut, polished and coated with 10 nm carbon. A ZEISS EVO LS15 SEM and Oxford Instruments AZtec EDS system with an XMax 80 mm 2 silicon drift detector was operated at an accelerating voltage of 20 kV and a probe current of 1.5 nA resulting in an output count rate of ~ 50,000 counts per seconds. A large area elemental map was acquired for one sample. 120 fields were acquired using automated stage control. Each field covered an area of 563 × 422 μm and was analysed for ~ 25 min. Secondary electron (SE) and back-scattered electron (BSE) images were acquired at a resolution of 1024 × 768 pixels, corresponding to a pixel size of 0.55 μm. EDS spectra were stored as hyperspectral imaging datasets at a resolution of 512 × 384 pixels, corresponding to a pixel size of 1.1 μm. AZtec 5.1 software was used to stitch the individual fields into one hyperspectral imaging dataset with an EDS resolution of 5511 × 2984 pixels and SEM resolution of 11,023 × 5968 pixels. The distribution of elements is displayed as net intensity maps where the X-ray background has been subtracted and peaks with overlapping X-ray lines have been deconvolved.
Most of the images shown in this article (except SEM and EDS images), from µCT and macrophotographs under incident natural illumination, have been edited in Adobe Photoshop CS6. The macrophotographs were obtained using a Canon 5Drs camera mounting 100 mm and Sigma 50-mm lens. All the data can be seen as Supplementary material.

Description
Thirty-two individuals, from both North America Midcontinent sites, have internal structures susceptible to be conulariid soft parts, after discarding specimens that were heavily bioturbated (Fig. 6a), and others with sand grains (Fig. 6b) and further bioturbated with remains of soft parts (Fig. 6c). As these longitudinal structures are common in the scanned individuals and appear in both the Wewoka and Graham formations, they are not considered to be taphonomic artifacts. They are also consistent with previous studies considering conulariids as scyphozoan polyps.
Ten of these individuals, all of them scanned, have a longitudinal structure in "V" inside the theca cavity, occasionally with one of the branches more accused than the other, running parallel to the internal theca sides or can be seen with a more central position in the theca (Fig. 7c, d). These individuals have mostly damaged thecae, probably due to sublethal and lethal injuries. The "V" structure can start at the apical part or at the middle of the theca length and ends at the aperture hinge level. The width of each branch of the bundle is about 1.5 mm. These bundles can be interconnected with possible muscular fibers that extend between bundles (Fig. 7d). Retraction of this muscular system had to facilitate closure of the aperture.
Four additional specimens have a similar structure to that shown by Babcock and Feldmann (1986b: figs. 30.2-30.3, 30.5-30.6), with a main longitudinal structure that at the middle of the theca length widens and turns adaperturally. It runs in a tubular shape from the apical part to the middle of the theca where it widens to an oval-shape and narrows again at its end, before the hinge level (Fig. 7a). At the adapical area, it is 1 mm wide and at the middle of the theca the width is at least three times more. All of these specimens show injuries (NHMUK PI CL 602; CL 613; CL 803; and CL 855) (Fig. 7a, b and Suppl. material) and are originally mostly from the Finis Shale except the last one cited, NHMUK PI CL 855 from the Wewoka.
A specimen from the Finis Shale (NHMUK PI CL 601: Fig. 8) preserves four longitudinal bundles that divert from the apical part (Fig. 8). This specimen is displayed on its four sides showing how these four bundles are distributed and how they start to fuse and divert towards the apertural area, keeping a parallel-like position to the corners and not to the midline. They are closer to the external side of the theca adaperturally and more distant adapically due to the theca's thickness. As the theca is thicker adapically, the bundles are more distant from the external layer of the theca or periderm, opposite to the adapertural part with thinner theca. On the other hand, the bundles are thinner adapically (approx. 1 mm wide) and thicker adaperturally (approx. 1.5 mm). It seems they extend adaperturally to form a layer. This individual shows a sublethal scar that has been repaired with transverse ribs, partly preserved in its apical part, and other lethal injuries close to the aperture (Fig. 8 with arrows showing the injuries). Both of these scars are of puncture type, with a circle-like outline. It is possible to see under CT scan that soft tissues are behind the injury marks. A couple of individuals have foraminifera and sponge spicules inside their theca remains (e.g. NHMUK PI CL 803, Suppl. material), also with bivalve shells, and a brachiopod shell. They could be introduced during a storm. The non-compaction of the specimens suggests that the internal part of the animal was filled in at the time of burial, and these remains were introduced inside the central cavity before closing their aperture, e.g. with turbiditic currents. The brachiopod shell inside of a conulariid (NHMUK PI CL 815, Suppl. material) probably suffered from predation as has a clear circular drill hole placed at the level of the brachiopod soft parts, on the external part of the theca. This probably represents a unique and successful predation attempt.
There is an individual without bioturbation marks in the interior of the theca that has cylindrical pellets running underneath and parallel to the theca to form bands (NHMUK PI CL 801, Suppl. material). The longitudinal bundles of the conulariid are hardly preserved, with friable tissue left. These pellets have different sizes and do not keep any arrangement. They may be faecal pellets that are not related to the organism with which they are associated. They could be produced by scavengers that fed on soft parts (Bruthansová and Kraft 2003) after the death of the conulariid, or by animals which used the empty theca as a hiding place. Observations on living coronates show some annelid worms, crustaceans and pycnogonids can feed on the soft tissues, oftentimes using the periderm tube (exoskeleton) of the polyps as a hiding place (ACM, personal observations). As the soft part remains are not well defined, we believe that the theca was used as shelter. The animal used the theca mainly to refuge from predators and, secondly, was fed with the conulariid soft parts. This would be consistent with the thesis that carcasses in shallow marine environments provide shelter (Vinn et al. 2018) and would explain the preservation of the longitudinal bundle remains. The conservation of these pellets may have been produced through phosphatization and lithification that allowed them to be preserved three-dimensionally.
The British specimens did not show any clear internal structure. One of the specimens has crinoid ossicles ( Fig. 9) close to the aperture. These ossicles are common throughout the Much Wenlock Limestone Formation (Homerian) of Dudley, in the shallowest water deposits represented by crinoidal grainstone beds. These may have been caught by the living conulariid during a period of high energy or storm that, with a rapid burial of the specimen, allowed its preservation in three dimensions, including the aperture.
Elemental mapping of the specimen NHMUK PI CL 855, from the Middle Wewoka Member, shows a predominant composition of calcium phosphate, also with SiO 2 , K-feldspar and Na-feldspar grains up to 40 μm in size. The soft part remains are mainly represented by Fe 2 O 3 , and some minor contribution of Zn. There are also framboidal iron oxide granules which also represent soft remains (Fig. 10).

Discussion
The four longitudinal bundles diverting from the apical part that have been observed in one of the specimens (NHMUK PI CL 601: Fig. 8, Finis Shale) compare well with the muscle bundles described in fossil and extant cnidarians. These bundles are arranged in a quadrilateral pattern, diverting dichotomously along the oral-aboral axis. They start at the apical area with a perradial position and extending adaperturally in a slightly different position, probably due to contraction after the organism's death. The fact they can be seen as a layer adaperturally relate them to the ectodermal longitudinal and endodermal circular muscle layers of the cnidarian column. The specimen is preserved in three dimensions, like the rest of the individuals studied, due to entrance of sediment inside the central cavity, probably during a storm episode or because of turbidity currents. It may be then when the muscles closed the aperture. The preservation of these bundles may have taken place via authigenic replacement of muscular tissues enabled by rapid burial, probably at the same place where the individuals lived. In case, they were exposed after burial, the soft parts would have hardly been preserved and the thecae would have bioturbation structures (Fig. 6a). This could happen when the individuals were transported to more oxygenated waters. In this last case, the thecae would keep their three dimensions because cnidarians decay from outside-in, retaining their three-dimensionality long after all internal characters have fully decayed (Hancy and Antcliffe 2020).
Most of the individuals that preserve longitudinal bundles (45% of the scanned individuals) at the apical part can be seen diverting from this area and extending through the theca. This occurs indistinctly in both North American sites. Ten individuals (30%) show a unique longitudinal structure as a "V" inside the theca cavity (Fig. 7c, e). This structure has both sides with a tubular form of similar size. Four individuals (12%) display a much thicker longitudinal tubular bundle than the rest of the bundles that extends and turns adaperturally (e.g. NHMUK PI CL 602: Fig. 7a) in the same way as Babcock and Feldmann (1986b: Fig. 30.5) illustrated. This structure starts close to the corner adapically and extends in oval-shape adaperturally and centripetally. There is another individual (NHMUK PI CL 855: Fig. 7b) which has only preserved a longitudinal bundle that extends from the apical area to the apertural hinge level. The asymmetric position of this structure prevents us from suggesting it could be the potential gastric cavity. But, could this asymmetric position be triggered by the death of the organism? McMahon et al. (2017) studied how a sea anemone decays and documented the morphological changes the animal suffered. They detailed how the column contracted near the time of death and changed its shape dramatically afterwards. Moreover, the muscular bundles were the most resistant to degradation. This matches with conulariid soft parts, where muscle bundles have been preserved. The dysoxic conditions in which these specimens have been buried would have favoured their preservation. Although McMahon et al.'s (2017) work could be extrapolated to putative soft-bodied organisms, conulariids are biomineralized; their thin thecae (composed of carbonate-rich apatite) show an alternation of organic poor and organic-rich microlamellae that could not help with the preservation of soft tissues in the fossil record, with few exceptions such as some of these specimens from the Pennsylvanian of the North American Midcontinent. These longitudinal bundles are interpreted as part of the muscular system and the oval-shaped structure as the gastric cavity. The latter's preservation in a few individuals may have been produced because of altered remains of partially digested food matter and/or sediment left in its interior when the animals died. The soft tissues are mainly preserved by iron minerals which replicate the original shape. Some longitudinal bundles can be seen partially pyritised (NHMUK PI CL 803, Suppl. material). SEM analyses on iron mineral grains with preserved soft tissue show a difference in crystal size and morphology inside the theca. On the other hand, McMahon et al's. (2017) work could explain why tentacles are hardly preserved in the fossil record and have been only described to date in putative conulariids. It seems tentacles are no longer discernible 6 days after the death of cnidarian polyps.
Although conulariid muscle bundles have been assumed to have interradial positions, these specimens, from both Graham and Wewoka sites and of two different taxa, have predominant perradial positions agreeing with Babcock and Feldmann (1986b). The individual that preserves four longitudinal muscles keeps three out of the four bundles at the perirradii (NHMUK PI CL 601, Fig. 8, Suppl. material). There are also specimens with a couple of narrower and shorter interradial longitudinal bundles (NHMUK PI CL 602, Suppl. material). Both kind of bundles may extend dichotomously. We cannot forget that the morphology of these muscle bundles has to be affected by the contraction at the time of the burial. Those longitudinal structures from the Graham Formation seem to be better preserved, probably because of the lower oxygen content as the grey shales are darker than the Wewoka ones. More dysoxic conditions would also explain why there are fewer predation marks on the Finis Shale thecae. Furthermore, the thecae are also better preserved. The only specimen with four "V-shaped" longitudinal bundles has been found in Finis Shale deposits. The Wewoka specimens inhabited more shallow waters and wave action would certainly be sufficient to break their thecae after the organism's death. Moreover, Wewoka conulariids have suffered from transport (Olszewski and West 1997).
The fact that those specimens that keep the aperture/operculum, and this is closed, means the individuals were able to contract their muscular system to close their aperture. This ability to perform muscle contraction is one of the most important and distinctive features of eumetazoans, holding important information and informative phylogenetic position Fig. 10 A Energy-dispersive spectrometry net intensity elemental map mixed with the back-scattered electron image (11,203 × 5968 pixels). Calcium phosphate is represented by phosphorous (green) and the soft tissue by iron (red) oxide. B Detail of the outlined rectangle in A showing framboidal iron oxide in the soft part remains. Silicon (blue) dioxide, sodium (turquoise) feldspar (Nafs) and potassium (magenta) feldpar (Kfs) grains are present within calcium phosphate for understanding muscle evolution (Leclère and Röttinger 2017). Any movement, not only contraction, but also extension, was regulated by the nervous system. The study of their muscular system may give new insights into the organisation of a muscle system in a non-bilaterian organism and help with molecular clock studies of cnidarians. The longitudinal bundles responsible for closing the aperture seem to reach the aperture hinge (Fig. 7c-e). Most of the specimens that keep the aperture do not have soft tissue or longitudinal bundle remains at the apertural area. There is one specimen (NHMUK PI CL 849, Suppl. material) with a fragment of longitudinal bundle detached from other area such as a corner. The reason there is no soft tissue at the apertural area may be due to contraction at the time of the burial (Frickhinger 1994) and the preservation state, probably with sediment in this area.
Several individuals displaying healed injuries may contribute to understand the distribution of the living soft parts as these tissues were affected by predation and covered the damaged theca area, displaying incredible plasticity in their regeneration. The thecae are the canvases where the predatory attacks are recorded. There are other individuals with very damaged thecae that we cannot differentiate if they are due to predation or postmortem breakage, as the lethal attacks could destroy the prey, or could be mistaken predation, in which predators break the thecae of dead organisms believing they were alive. On the other hand, it is possible to see that there are individuals with puncture injuries (e.g. NHMUK PI CL 604: Fig. 4) which could be non-lethal. Their soft parts in the theca interior cavity protrude lining with the theca. In case of lethal injuries (e.g. NHMUK PI CL 639: Fig. 11), the puncture injuries do not have soft tissues underlying in the theca cavity. The ability to heal injured areas coincides with extant coronates (Chapman and Werner 1972).
Most of the North American individuals have very damaged thecae, due to mechanical damage, produced mainly by severe predatory attacks (mostly Wewoka Formation specimens), and by biostratinomic factors (mainly Graham Formation specimens) such as bioerosion and dissolution typical of muddy offshore environments. Very different is the case of the British individuals which have very subtle repaired thecae, with scalloped scars. These latter could have been produced as the previous ones by unsuccessful predation, but also by friction between individuals of the same or different taxa. The biotic origin of the marks found in conulariids is distinguished by their geometric shape as happens with puncture injuries and their non-random distribution being preferentially located at the corners. Some thecae show multiple scars, after failed attacked and reconstruction of the thecae, and also lethal puncture injuries. Drilling millimeter holes related to bioturbation have been seen (NHMUK PI CL 851, Wewoka Fm, Suppl. material) close to the aperture. Although the hole seems not penetrate deep inside the theca, the theca interior cavity is very bioturbated. This drilling hole runs inside at an acute angle. Scalloped and cleft scars can be discerned thanks to the theca ornamentation when this is affected. The fact that Wewoka individuals suffered more from predation than the Finis Shale ones is reflected in their preservation. Once the thecae have perforations, they are more vulnerable to theca breaks at the weak areas in case of storms; and give less resistance to taphonomic forces before or after burial. Abiotic processes, such as pressure solution due to compaction of the sediment and abrasion, are facilitated.
It is possible that predation was selective and affected young individuals more than adult ones as happens with corals (Wood 2003) and also with more predisposition for some taxa than other. If predation more strongly affected youngsters, this would explain why most of the specimens studied, which have predation marks, have very small sizes, of several centimeters long ( Fig. 1) with the exception of a few individuals that do not have clear predation marks. Probably, the percentage of individuals which reached the adulthood was very low due to predation, triggering significant changes to its population regarding mean size, as a result of selective removal of small prey by predators. In the case of 'Conularia crustula', this species did not survive the Pennsylvanian. This could be due to slow rate of growth, considered as disadvantageous to sessile organisms (Wood 2003) and their ability to regenerate soft tissue and theca, both depending on species' abilities of regeneration, concerning lesion size and environmental factors. One of the Wewoka individuals (NHMUK PI CL 635, Suppl. material) suffered from predation, with sublethal injuries from which the animal recovered and healed its theca, but probably not completely its soft tissue as seen in the µCT scan (Fig. 12). Although the specimen is fragmented, there are thicker longitudinal bundle remains in other parts of the theca interior cavity. Alternately the individual could have been attacked again, being broken by the weakest area and withdrawn part of their soft tissues, or maybe the theca was broken and the soft tissues were not preserved.
The fact that the individuals lived in relatively shallow waters would expose them to more predation than deep waters. Bioturbation with scavenging should have occurred when the specimens were exposed on the bottom and would be responsible, in part, for the lack of soft tissue preserved. On the other hand, there are numerous individuals in both Pennsylvanian sites that keep their apertures complete and closed. By analogy with some recent Cnidaria, the conulariids could have closed their aperture as a way of defense when they felt threatened by predators or any other risk (Werner 1970;Holst and Jarms 2006). It had to be vital for the conulariid survival to preserve their soft parts. Apart from predation and bioturbation, additional causes of the theca incompleteness could be storm breakage, crushing and dissolution while buried (Klompmaker et al. 2019).
Although most of the specimens from the Wewoka sites have sublethal injury marks, most of them have what seems to be remains of soft parts, longitudinal bundles in case of non-lethal injuries. As they have been preserved, this may be due to a low proportion of burrowers in the dysoxic waters in which these conulariids where deposited and more importantly due to the exceptional conditions of the deeper environment in which these organisms would lay (Conway Morris 1986). The reason why the soft tissues were preserved should have also been because they were alive at the time of their rapid deposition, in fine claystones, and rapid burial (Hagadorn et al. 2002). Some of the individuals suffered storm events or turbidity currents. This is indicated by grain ingestion, possibly sand grains, by the thecae (Fig. 6b). It has been reported that recent coronate polyps locked by grains in more than 80% could survive months (Holst and Jarms 2006). In our case, the specimen is almost 100% filled by grains. Its survival should have been much more reduced. For a good preservation, it is also necessary to have had the right conditions for authigenic replacement. Some of the soft tissues have been replaced by pyrite. This is seen in blocky and subtle concentrations at the longitudinal bundles, indicating decaying soft tissues. Only 23% of the scanned individuals present bioturbation structures and very subtle longitudinal structures. All the studied specimens are preserved in three dimensions, although they are incomplete. Their preservation was facilitated by infilling the theca with mud that later would become a concretion that preserved from crushing during the biostratinomic processes. The preservation of such soft bodies requires low hydrodynamic energy and rapid burial. These are among the best preserved conulariids known due to their soft tissues being preserved (see reconstruction of the conulariid soft parts seen under CT scan: Fig. 13). Their abundance in the Hughes (Oklahoma) and Jackson (Texas) counties implies they may have been very common in the Pennsylvanian.
Possible predators susceptible of predation under low water oxygen levels are some gastropods, cephalopods, arthropods (probably phyllocarids), fishes and sharks. All of these groups have been found associated with conulariids in both North American sites and could be responsible for the attacks suffered by the conulariids. Cladodontid sharks have already been reported as conulariid crushers in the Pennsylvanian by Mapes et al. (1989).
The act that there are some individuals with only a V-shaped longitudinal bundle could be explained by predation and preservation. If the individuals which were predated withdrew some of their soft parts, the remains left would retract as much as they could. When they were able to build a scar, their soft parts would reinforce the theca (NHMUK PI CL 604, Suppl. Material). There are a couple of individuals (NHMUK PI CL 603, CL 812, Suppl. material and Fig. 7c, d for CL 812) in which it is possible to see the V-shaped longitudinal bundles with a third bundle partially preserved. All the bundles merge apically. Other specimens keep only intermittent sections of the longitudinal bundles (NHMUK PI CL 611, Suppl. material), probably by bioturbation and scavenging.
Those individuals that we consider to have remains of the gastric cavity (NHMUK PI CL 602, CL 613, CL 803 and probably CL 855: Fig. 7a, b, Suppl. material) would have the mouth associated with that cavity and tentacles would be responsible for the conulariid feeding. These tentacles would be extended during their feeding and could have been formed of different tissues from the mouth. In extant coronates, the mouth plate and tentacles are made of the same ectodermal tissue layer, but with different types of cells depending on the area. On the other hand, the cavity occupies the whole internal space, with the endoderm layer lining the theca. Any kind of muscle bundles are inside the cavity as outgrowths of the endodermal tissue layer (see Fig. 13 for reconstruction).

Conclusions
The Pennsylvanian conulariids of the North American Midcontinent (Texas and Oklahoma, USA) have been studied by µCT scanning and revealed for first time muscular bundles, conulariid soft parts, in three dimensions and a potential gastric cavity. These individuals have been preserved thanks to the exceptional conditions of the Wewoka and Graham deposits. Most of the Wewoka specimens have predation scars that have affected the theca and soft parts, and others have lethal injuries. The soft parts have been preserved in different ways, some showing the potential gastric cavity with muscle bundles and others keeping only the muscle bundles. The presumed gastric cavity is oval-shaped and can be seen fused to a muscle bundle. The muscle bundles fuse apically or medially, normally forming V-shaped pairs, and they extend along the theca, parallel to the corner, towards the aperture. Longitudinal bundles have predominant perradial positions. The understanding of these structures will help to elucidate the relationship of conulariids in cnidarian phylogeny.