New insights into the sea spider fauna (Arthropoda, Pycnogonida) of La Voulte‐sur‐Rhône, France (Jurassic, Callovian)

Three species of sea spider (Arthropoda, Pycnogonida) have been described from the Middle Jurassic (Callovian) Konservat‐Lagerstätte of La Voulte‐sur‐Rhône: Palaeopycnogonides gracilis, Colossopantopodus boissinensis and Palaeoendeis elmii. These fossils were initially attributed to three extant families or superfamilies, justifying their use as calibration points in a recent tree‐dating analysis. However, their taxonomic affinities are still debated. Our knowledge of the morphology of these Jurassic sea spiders is limited by prior investigation with only light microscopy and radiographs, such that further morphological details such as cephalic appendages (chelifores, palps, ovigers) remain poorly documented. Here, we reinvestigate the La Voulte‐sur‐Rhône fossils using x‐ray microtomography and reflectance transformation imaging. A new specimen, tentatively attributed to P. gracilis, was found using x‐ray microtomography, while another fossil initially interpreted as Palaeoendeis elmii may also be related to this species. We attribute all fossils to Pantopoda, crown‐group Pycnogonida. Palaeopycnogonides gracilis had ovigers in at least one sex and had chelifores and palps that were either reduced or absent. Together, the cephalic appendage set and the structure of the ovigers are unique among Pantopoda, and this species is reassigned to the new family Palaeopycnogonididae. Colossopantopodus boissinensis lacked chelifores but had palps and ovigers, the latter with the typical structure shown by extant Colossendeinae, to which we attribute the fossil. The absence of chelifores and palps in Palaeoendeis elmii and the structure of its ovigers indicate affinities with Endeidae. The impact of these new taxonomic assignments on the way Jurassic sea spiders can be used as fossil calibrations is discussed.

S E A spiders (Arthropoda, Pycnogonida Latreille, 1810) are an enigmatic class of living arthropods. Their body plan includes a proboscis made of three antimeres and with tri-radial cross-section (Wagner et al. 2017); internal organs such as the digestive tract and gonads extend into the legs (e.g. Miyazaki & Makioka 1991Miyazaki & Bili nski 2006;Frankowski et al. 2022) with genital pores opening on the coxae 2 of legs (Arnaud & Bamber 1987); median eyes are carried on an ocular tubercle while lateral eyes are absent (Lehmann et al. 2017); and earliest life stages typically take the form of a protonymphon larva (Brenneis et al. 2017). Sea spiders are morpho-anatomically unique among arthropods and in the past this has led to their interpretation as the sister group of all other living arthropods (see a review in Dunlop & Arango 2004), the so-called Cormogonida hypothesis (Zrzav y et al. 1998). But phylogenetic studies now confidently support a position of pycnogonids within Chelicerata, as sister to the Euchelicerata (Regier et al. 2010;Rehm et al. 2011;Ballesteros & Sharma 2019;Lozano-Fernandez et al. 2020;Ballesteros et al. 2021), showing that the body plan of sea spiders has undergone radical transformations from the last common ancestor of the Arthropoda crown-group. This evolutionary transformation cannot be understood solely through the study of living pycnogonids given that they largely share the same body plan. Fossils are integral to elucidating the evolutionary origin of the sea spider's morpho-anatomy, as well as in calibrating its evolution against geological time. This has been attempted on a few occasions (Arango & Wheeler 2007;Ballesteros et al. 2021), but the known fossil record of sea spiders is currently too poorly characterized to obtain material insights into their evolution.
Six fossil sites have provided pycnogonid fossils, for a total of 13 species described (for a review, see Sabroux et al. 2019). In the Palaeozoic, the Hunsr€ uck Slate accounts for five species (Early Devonian, c. 400 Ma;Bergstr€ om et al. 1980;Poschmann & Dunlop 2006;K€ uhl et al. 2013), and Herefordshire provided one more (early Silurian, c. 425 Ma; Siveter et al. 2004). Older sites of William Lake (Late Ordovician, c. 450 Ma;Rudkin et al. 2013) and Orsten (late Cambrian, c. 500 Ma; Waloszek & Dunlop 2002) also potentially yielded sea spiders; but the status of these fossils remains doubtful because their preservation or developmental stage do not allow the presence of diagnostic characters to be con- During the Palaeozoic, sea spiders showed an extraordinary diversity of forms and morphological traits, including paddling legs adapted to swimming, clawbearing palps, a long and segmented abdomen, a posterior flagellum, a post-anal telson, and enigmatic annular structures at the base of the legs (Bergstr€ om et al. 1980;Siveter et al. 2004;Poschmann & Dunlop 2006). These characters are unevenly distributed among the fossils and whether they are plesiomorphic or apomorphic is not always clear. These characters are unmatched in extant species, which led to support for the order Pantopoda Gerst€ acker, 1863 encompassing crown-group Pycnogonida, but from which most, if not all, Palaeozoic fossils are excluded (Hedgpeth 1954(Hedgpeth , 1978Bergstr€ om et al. 1980;Sabroux et al. 2019). After the Devonian-Jurassic hiatus, such a patchwork of morphologies was no longer present, at least as seen in the fossil record. Jurassic fossils are all morphologically close to the taxa alive today and all fossils have been attributed to extant families or superfamilies of Pantopoda (Charbonnier et al. 2007a;Sabroux et al. Charbonnier et al., 2007a, to Endeidae Norman, 1908and ?Eurycyde golem Sabroux et al., 2019 to Ascorhynchidae Hoek, 1881. As such, La Voulte-sur-Rhône material was critical for the first tree-dating analysis ever performed on sea spiders (Ballesteros et al. 2021), in providing three of the four in-group fossil calibrations for the molecular clock. However, taxonomic affinities as proposed by Charbonnier et al. (2007a) need confirmation: the 11 pantopod families that are most generally accepted in the literature (Bamber et al. 2023) are characterized by different combinations of presence/absence of the cephalic appendages (chelifores, palps, ovigers) and the characters borne on those appendages. In most of the Jurassic fossils, the presence or absence of these appendages remains ambiguous, mostly because only the surface of the fossils could be explored extensively. Here, we reinvestigate the fossils of La Voulte-sur-Rhône described by Charbonnier et al. (2007a). The sea spider fauna of La Voulte-sur-Rhône is one of the richest with a relatively high number of specimens (14 specimens available to this study), often with exquisite preservation. In contrast to Solnhofen fossils, they are not simple imprints but have a threedimensional (3D) pyritized structure, such that hidden features may be enclosed in the sediment. So far, these fossils were investigated only using radiographs and light microscopy. To confirm, amend and complete their original description, we investigated the La Voulte-sur-Rhône sea spider material using reflectance transformation imaging (RTI) to interpret the exposed features of the fossils, and x-ray computed microtomography to investigate further details that have remained hidden in the sediment.

GEOLOGICAL SETTING
The site of La Voulte-sur-Rhône (Ard eche, France;Fischer 2003;Charbonnier et al. 2007bCharbonnier et al. , 2010Charbonnier et al. , 2014Charbonnier 2009 and references therein) is situated on the right bank of the Rhône River and belongs to the eastern sedimentary cover of the Massif Central (see Charbonnier et al. 2014, fig. 1). La Voulte-sur-Rhône includes two Jurassic localities: the Ravin du Ch enier (dominated by siliceous sponges and stalked crinoids), and the Ravin des Mines. This latter corresponds to the La Voulte-sur-Rhône Konservat-Lagerst€ atte. Its stratigraphy (see Charbonnier et al. 2010, fig. 2) consists of a lower 15 m layer of thick iron deposits that were economically exploited in the 19th century, topping a basal marl including a thin layer (5-6 m) with many sideritic nodules including soft-bodied, uncompacted animals; and soft-bodied, flattened fossils. Outcrops belong to the Gracilis ammonite Biozone. During the early Callovian (c. 160 Ma), this was probably a marine, bathyal environment (Charbonnier et al. 2007b(Charbonnier et al. , 2010 located along the western Tethyan margin. It was situated on an active, major subvertical fault (the La Voulte fault), so that the sea bottom was steep.

Studied material
This study focuses on the sea spider fossils from La Voulte-sur-Rhône Konservat-Lagerst€ atte (Charbonnier et al. 2007a). The material studied consists of 14 specimens (plus one hidden specimen) distributed over six slabs (Fig. S1). Although the La Voulte-sur-Rhône strata that yielded sea spiders are known (Fischer 2003;Charbonnier 2009;Charbonnier et al. 2010), the information about the precise stratigraphic origin of each separate fossil is, to our knowledge, most often lost (S. Charbonnier, pers. comm.). Specimens are preserved as 3D pyritized elements exposed on marly slabs. All fossils are curated in the palaeontological collections of the Mus eum national d'Histoire naturelle, Paris, France (MNHN.F).

Microtomography
X-ray computed microtomography was performed on five fossils. The specimens selected for scanning were limited to those in which the fossils were similar in size to the host slab, the size of which dictates the resolution of the scan. All five fossils were scanned with the AST-RX platform of the Mus eum national d'Histoire naturelle using a v|tome|x 240 L scanner (Baker Hughes Digital Solutions). Parameters were adapted for each fossil to provide the best possible results (Table S1): the current was set to 200-355 lA, voltage to 200-225 kV and exposure time to 1-2 s. A tungsten reflection target and 0.50 mm tin and 0.25 mm copper filters were used. The voxel sizes obtained ranged from 45.2 lm to 80.7 lm depending primarily on the size of the slab hosting the specimen.
Contrast was enhanced on conversion to 8-bit stacks using Drishti Import v3.1 (Limaye 2012). Segmentation was performed using the SPIERS software suite (Sutton et al. 2012). Because there was low attenuation contrast between the specimens and the surrounding matrix, it was necessary to manually remove artefactual structures from the masks. Some specimens were preserved in contact with high-density nodules or fossils of other sea spiders or brittle stars Ophipinna elegans (Heller 1858). In these cases it was not possible to determine the exact boundary between the included objects and the studied pycnogonid. Here, the shape of the sea spider part in contact with the alien object was defined using interpolation of the boundary from adjacent slices. Finally, in a few occurrences, appendages lost in the specimen were preserved as counter-imprints. Here, the shape of the hollow left in the matrix was used to reconstruct their approximate shape (as, for example, in Garwood & Dunlop 2011). The resulting models were rendered as isosurfaces in SPIERS, with the parts reconstructed from imprints rendered semi-transparent. Very weak smoothing and large island removal were applied.

Reflectance transformation imaging
In order to uncover further details of the exposed features of La Voulte-sur-Rhône fossils already described by Charbonnier et al. (2007a), we used reflectance transformation imaging (RTI; also known as polynomial texture mapping). RTI is an imaging technique primarily developed by archaeologists and art historians (Padfield et al. 2005;Earl et al. 2010) that is now increasingly used in palaeontology (e.g. B ethoux et al. 2016;Decombeix et al. 2021). An item of interest (here, a fossil specimen exposed in a rock matrix) is photographed repeatedly, each time changing the position of the light source that is ideally centred on the item. The light source turns around the item at a fixed distance. In contrast to photogrammetry, the position of the camera relative to the item does not change. After computation, the resulting model is not a 3D image but a 2D image that can be visualized under all light orientations. This methodology is particularly relevant for fossils that are largely flat and can be observed in a single plane, but that show 3D features on their surface. RTI was performed for 14 fossil specimens, using an RTI dome (a hemispheric rig placed over the fossil with an automated lighting series of LEDs evenly spaced around its convex surface). Here, we used a 28-cm-diameter light dome equipped with 54 LEDs distributed over three rings. Image capture and lighting changes were automated by a dedicated control box. Light orientation was recorded on each image by a black reflective hemisphere (all RTIdedicated equipment by FlyDomes). Photos were taken with a Canon EOS 700D digital camera, with Canon 50 mm and Canon EF 100 mm macro lenses (Table S2).
Images were processed with the software RTIbuilder (Cultural Heritage Imaging) using the Highlight Based (HSH Filter) operation sequence. The resulting models were then visualized using the RTIviewer software (Cultural Heritage Imaging). RTIviewer enables the user to vary the direction of illumination at will. 'Specular enhancement' enables the contrast between illuminated and shadowed surfaces of the item to be enhanced by estimating the normal for each pixel, and using this to render surfaces using synthetic specular highlights. Finally, 'normals visualization' enables visualization of all normals at once with false colours depending on the position of the light source that most strongly illuminates a given part of the item.
Podomere nomenclature in post-cheliceral appendages and noticeable features Sea spiders typically have a pair of chelifores, a pair of palps, a pair of ovigers and four pairs of legs (cephalic appendages can be independently lost at the adult stage and up to two additional pediferous segments can be found in some species; Arnaud & Bamber 1987). Although chelifores are very different in structure compared with other appendages, post-cheliceral appendages (palps, ovigers and walking legs) of sea spiders have an overall similar architecture, schematizable as a Z-shape. Although the present work does not seek to address the serial homology for the podomeres of the different appendages, these similarities justify the adoption of a homogenized vocabulary for these three types of appendages (Fig. 1A), which we do here for the sake of simplification. The few cases in which this ground-plan cannot be readily applied (perhaps linked to fusion of podomeres, as suggested for the palps of Nymphon Walking legs are composed proximally of three podomeres, coxae 1, 2, 3, that are generally short. The articulation between coxae 1 and 2 is the only one able to perform a promotor-remotor movement (Manton 1978). The two joints around coxa 3 are extremely flexible (Schram & Hedgpeth 1978), so that these create a ventral geniculation of the leg hereafter referred to as the 'coxal geniculation'. In palps and ovigers a similar geniculation is generally observed distal to three proximal podomeres of variable size (Fig. 1B), hence we name these three podomeres coxae 1, 2 and 3 as well.
Distally to coxa 3 articulate three long walking leg podomeres, the femur, tibia 1 and tibia 2. Although these terms are the most common in Pycnogonida taxonomic literature by far, we propose instead to adopt here the more convenient nomenclature used by Schram & Hedgpeth (1978), which calls these podomeres, respectively, femur, patella and tibia. The joints between these three podomeres all show a geniculation (respectively named the 'femoro-patellar geniculation' and 'patello-tibial geniculation'). These are always oriented dorsally in the walking legs of living pantopods. In most ovigers and palps the femur is elongated and creates a femoro-patellar geniculation with the following patella, the size of which is more variable. The patello-tibial geniculation, when present, is often less marked (Fig. 1B). Because the number of podomeres distal to the tibia varies, we simply refer to these as 'tarsal podomeres'. In the walking legs there are two tarsal podomeres: the tarsus and the propodus, between which the joint has no muscle (Schram & Hedgpeth 1978). Instead of these, ovigers typically have four tarsal podomeres (forming a hook-like structure called the strigilis) although this number may be reduced (e.g. Endeidae). Palps typically have a group of 0-3 tarsal podomeres (Fig. 1B). In walking legs, a terminal claw constitutes the most distal podomere. This so-called 'main claw' is often accompanied by two lateral auxiliary claws, forming together a typical triple claw that is thought to be plesiomorphic among Chelicerata (Dunlop 2000). Ovigers may or may not carry one terminal claw (Fig. 1B), while extant species never have a palpal terminal claw (but see Haliestes dasos Siveter et al. 2004).

Microtomography
The computed tomography was successful (Figs 2, 3) and showed parts of the fossils hidden in the matrix. We detected ventral features in specimens MNHN.F.A49277 (holotype of Palaeoendeis elmii) and MNHN.F.A52384 (paratype of Palaeopycnogonides gracilis) that were not seen using x-rays by Charbonnier et al. (2007a). These structures correspond in position and shape to ovigers. Articulations were poorly preserved.
We also detected one additional, hidden specimen (MNHN.F.A88079; Fig. 2C) enclosed in the same slab as MNHN.F.A52384, just underneath the fossil's surface. This specimen presents the typical outline of modern sea spiders (slender body, eight slender legs, absence of segmented post-ambulatory tagma).
There is often a marked density contrast between the outer part of fossils and their internal counterparts. In most cases, the outer, higher contrast parts of the fossils form only a simple outline. In the case of fossils MNHN.F.A52381 and MNHN.F.A52384, however, this outer part is much thicker (Fig. S2).

SYSTEMATIC PALAEONTOLOGY
The synonymy style follows the recommendations of Matthews (1973).  Redescription. Trunk robust, completely segmented. Lateral processes separated by less than their own diameter, touching at base. Two dorso-median tubercles on cephalon, one a large dome at anterior margin of cephalon, the other a small dome carried equidistantly to anterior and posterior margins of cephalon. Position of ocular tubercle ambiguous. Proboscis cylindrical, about as long as first two trunk segments, possibly mobile, with slightly thickened proximal part and distal opening. Abdomen not visible dorsally. Anus opening ventrally. Chelifore absent or inconspicuous. Palp absent or inconspicuous. Oviger developed, with femoro-patellar geniculation visible ventro-posteriorly between 3rd and 4th lateral processes. Strigilis, number of podomeres, spination and sexual dimorphism unknown. Four pairs of legs, moderately robust. Coxae 1, 2, 3 indistinct. Patella c. 1.3-fold as long as femur or tibia. Tarsus indistinct, probably short. Propodus + tarsus c. 0.4-fold as long as patella. Spination and claws of propodus unknown.
Remarks. All of the fossils (or imprints) included here in the list have a typical stout trunk outline. Identification of these specimens as the same species as the holotype is further supported for two of them (MNHN.F.A52383 and A88075) by a similar femur/trunk length ratio. MNHN.F.A52381, A52383 and A52384 all have a characteristic central dorso-median tubercle, although this is attenuated in MNHN.F.A52384 probably because of the different extraction method that was used (micro-chisel and sandpaper) and/or heavy pyritization. The proboscis of P. gracilis is found at different angles: it is for example almost horizontal in MNHN.F.A52381, while specimen MNHN.F.A52383 has it with a c. 45°angle to the body's antero-posterior axis, and in specimen MNHN.F.A88073 it is oriented at a c. 90°angle. Either this indicates that the proboscis was directed downward in living specimens (and was flattened in some specimens during fossilization), or that it was mobile (as found in modern Ascorhynchidae Hoek, Ammotheidae Dohrn, 1881).
In specimen MNHN.F.A88075, a conspicuous pit is positioned medio-posteriorly on the ventral surface of the fourth trunk segment (Fig. 4). A similar pit (although possibly artefactual) is observed at the same position in ventrally preserved MNHN.F.A52382 and A88073 (Figs 5, 6). Given the absence of a dorsally visible abdomen, we suggest that it was ventralized in P. gracilis, so that the pit is the probable remains of the anus. The abdomen of P. gracilis is never clearly visible, although it is worth noting that there is a round swollen outline around the probable anus of MNHN.F.A88073 (Fig. 5).
The inferred position of leg articulations is peculiarly distinct in the paratype MNHN.F.A52383 by bends between podomeres ( Fig. 7): the most conspicuous are the three long podomeres that unequivocally correspond (from proximalmost to distalmost) to femur, patella and tibia. Proximally, the femur is joined to the lateral processes by coxae 1, 2 and 3, although it is not possible to confidently identify them separately. The distalmost bend may correspond to the joint between the tarsus and propodus. On the proximal part of preserved legs of MNHN.F.A88075, a depression can be observed (Fig. 4): it is tempting to interpret these depressions as the gonopores, which are found (most often) on the ventral surface of coxae 2. Gonopores are wider in the female specimens of extant species F I G . 4 . RTI of Palaeopycnogonides gracilis, MNHN.F.A88075, body region (preserved ventrally, with imprint of dorsal region anteriorly). A, default view. B, specular enhancement (circle on bottom right indicates light orientation, see Table S2 for details). C, 'normals visualization'. D, interpretative drawing; plain black lines correspond to the outline of the fossil; dashed black lines correspond to the specimen imprints; dotted black lines correspond to the putative ovigers; grey lines correspond to the main breaks in the fossil. Abbreviations: an, anus; gp, gonopore. Scale bars represent 5 mm. (Arnaud & Bamber 1987), and fossils with these depressions could be female. However, the shape of these depressions is very variable and it cannot be excluded that they are preservation artefacts.
The cephalon of P. gracilis has two characteristic ornamentations ( Fig. 8): an anterior cephalic feature (initially interpreted as chelifores by Charbonnier et al. 2007a) and a median cephalic feature (the ocular tubercle in Charbonnier et al. 2007a). Three hypotheses can be envisaged to interpret these cephalic features: (1) Charbonnier et al.'s interpretation is correct; (2) the anterior feature is a bulge or a tubercle positioned near the proboscis insertion, and the median cephalic feature is the ocular tubercle as suggested by Charbonnier et al.; and (3) the anterior cephalic feature is the actual ocular tubercle, and the median cephalic feature is an ornamental dorso-median tubercle. The first hypothesis is weakened by the anterior cephalic feature not being clearly paired: in holotype MNHN.F.A52381 the anterior cephalic feature instead appears as a rounded structure with a central depression; in the paratype MNHN.F.A52383 it clearly appears as a central unpaired bulge. It is not visible in MNHN.F.A52384, probably due to the extraction methods (micro-chisels and sandpaper). In addition, it does not seem to extend beyond the anterior cephalic margin as in extant Ammotheidae and Ascorhynchidae. The anterior cephalic feature most probably corresponds to an unpaired tubercle as in the second and third hypotheses. Both the anterior and median cephalic features are good candidates to be the ocular tubercle given that each is in a position that is present in extant taxa; for example, as in Ammotheidae Dohrn, , Ascorhynchidae Hoek, 1881, Pycnogonidae Wilson, 1878and Callipallenidae Hilton, 1942, Endeidae Norman, 1908 and Nymphonidae Wilson, 1878, respectively. However, no extant species is known to possess an unpaired tubercle on the anterior margin of the cephalon, while a median tubercle on the first trunk segment is found in Dromedopycnon acanthus Child, 1982 and several species of Pycnogonum Br€ unnich, 1764 (e.g. Pycnogonum pustulatum Stock, 1994). Hypothesis 3 is therefore the likeliest, although heavy pyritization of the fossils may hide further morphological details that could contradict our interpretation.
No unambiguous trace of a palp is observed. Only the paratype MNHN.F.A52382 has a slender structure (Fig. 6) near the right side of the proboscis that could be interpreted as a palp, but this same specimen also misses the right leg of the first trunk segment, and it is likelier that this structure corresponds to remains of the missing leg. Palps were therefore probably absent (as in e.g. Pycnogonidae, Endeidae) or reduced (e.g. Pallenopsidae Fry, 1978;Callipallenidae Hilton, 1942;Queubus Barnard, 1946).
RTI and x-ray tomography both showed ovigers on the ventral surface of P. gracilis: they are visible in the 3D model of paratype MNHN.F.A52384 (Fig. 2B). They are better resolved in specimen MNHN.F.A88073 (Fig. 5), in which their structure is clearly V-shaped (i.e. with a marked femoro-patellar geniculation posteriorly). Further preservation of the ovigers' articulations, or the ovigerous spines and claws are missing. Due to the absence of conspicuous sexual dimorphism of the legs (see above) and the small number of specimens with well-preserved ventral surfaces (three) it is not possible to determine whether P. gracilis shows sexual dimorphism in the ovigers, either as a complete absence in female specimens (as in extant Endeidae Norman, 1908, Pycnogonidae Wilson, 1878 and Phoxichilidiidae Sars, 1891) or as a different shape (as in Ammotheidae, Callipallenidae or Nymphonidae). In specimen MNHN.F.A88075 (Fig. 4) the presence of ovigers is ambiguous. A slight thickening of the ventral surface between the third lateral process could correspond to femoro-patellar geniculation of the ovigers; in the 3D-model of the holotype MNHN.F.A52381 ( Fig. 2A) the ventral surface is poorly preserved and strongly flattened, so that the apparent absence of ovigers cannot be regarded as significant. Other fossils are preserved either on the dorsal side or as an imprint, so that it is not possible to tell whether these had ovigers.  Remarks. This specimen was resolved inside the rock on the same slab as MNHN.F.A52384 and could be studied only using x-ray microtomography. It is associated with several nodular structures (small dorsal ones, larger ventral and anterior ones). The dorso-ventral orientation can be determined based upon the coxal geniculation. The fossil is strongly compressed dorsoventrally.
This fossil shares numerous similarities with P. gracilis and is likely to belong to the same species. The specimen is about the same size as other specimens of P. gracilis. The body is stout; the abdomen is not visible dorsally and is likely to be ventral, as suggested by a marked pit on the ventral surface of the fourth trunk segment that may be the anus; the trunk is segmented at least between the first, second and third trunk segments. The shape of the cephalon is the same as in P. gracilis, except in that it does not show the two characteristic dorso-median tubercles. However, the dorsal part of the cephalon is clearly damaged, perhaps because of compression, and two marked pits are positioned where the tubercles are expected in P. gracilis.
Two ventro-median features are observable (Fig. 2C): an anterior one, thick, almost square, crossing the entire length from F I G . 5 . RTI of Palaeopycnogonides gracilis, MNHN.F.A88073, body region (preserved ventrally): A, default view. B, specular enhancement (circle on bottom right indicates light orientation, see Table S2 for details). C, 'normals visualization'. D, interpretative drawing; plain black lines correspond to the outline of the fossil; dashed black lines correspond to the specimen imprints; grey lines correspond to the main breaks in the fossil. Abbreviations: an, anus; fpg, femoro-patellar geniculation of ovigers; gp, gonopore; ov, oviger. Scale bars represent 5 mm. the first trunk segment; and a median one, shallow and ovoid, spreading between the second and the third trunk segments. We suggest that the anterior one corresponds to remains of the proboscis. It is articulated to the distal margin of the head and bent ventrally rather than inserted ventrally on the trunk, as in species in which the proboscis originates ventrally (Pallenopsidae and Phoxichildiidae) and never originates so close to the first trunk segment's posterior margin. We regard this unusual proboscis position as possibly linked to the proboscis mobility in P. gracilis (see above) and subsequent dorso-ventral compression. As for the median feature, it is either the remains of ovigers or an artefact hiding the ventral surface of the specimen. We cannot therefore determine whether the specimen had ovigers or not. Redescription. Trunk slender, completely segmented, unornamented. All lateral processes well separated along their length by less than their own diameter. Basal process of palps and ovigers widely separated. Proboscis about as long as trunk, slender and cylindrical, without visible swelling. Abdomen inconspicuous.
Chelifores absent. Palps present, poorly preserved, with basal process emerging latero-ventrally near anterior margin of head. Ovigers long, inserted just ventral to first lateral processes. Strigilis present. Three proximalmost podomeres very short. Number of podomeres, spination and sexual dimorphism unknown. Four pairs of legs, slender, longer than trunk and proboscis together. Podomeres indistinct due to preservation.
Remarks. The coxal geniculation of legs is clearly identifiable and shows that the single known specimen is preserved in ventral view (see Charbonnier et al. 2007a, fig. 1c). This is also consistent with the right oviger being preserved over the second right leg as well as the observability of the ovigers' proximal structures (Fig. 9). RTI also shows that the trunk was completely segmented. Cephalic knobs already spotted by Sabroux et al. (2019) are interpreted as palps or palp basal process sensu Cano-S anchez & L opez-Gonz alez (2017) (Fig. 9). The right 'palpal knob' is connected to a short lateral structure that disappears after reaching the first right leg; it is also recovered using x-ray tomography (Fig. 3B), suggesting that this was a part of the living animal. We interpret this feature as remains of the right palp, indicating that it was present but poorly preserved, rather than reduced or absent as in modern Pallenopsidae.
The fossil has ovigers. The right oviger shows a conspicuous distal, hook-like feature (Fig. 9), interpreted as the strigilis. Articulation of podomeres is not preserved well enough to be interpreted, except for the proximalmost parts using RTI ( Fig. 9): we are able to identify at least three distinct structures, which correspond either to the three proximalmost podomeres of the ovigers (coxae 1, 2 and 3) or the two first proximalmost podomeres of the ovigers (coxae 1 and 2) and the ovigers' basal process (sensu Cano-S anchez & L opez-Gonz alez 2017). The geniculation at the oviger's mid-length is interpreted as the femoropatellar geniculation. Podomeres distal to this geniculation are together subequal to the set of anterior podomeres. Redescription. Trunk slender, segmentation absent or unpreserved. No dorsal ornamentation. Ocular tubercle rounded, dome-like, positioned near anterior margin of first lateral processes. Lateral processes well separated along their length by about their diameter (distally) or less (proximally). Abdomen visible dorsally, horizontal, reduced, not reaching beyond fourth lateral processes. Proboscis long, straight, cylindrical, widening distally, about as long as first two trunk segments. Chelifore absent. Palp absent. Ovigers present ventrally, podomeres indistinct. Podomeres proximal to femoro-patellar geniculation long, reaching beyond third trunk segment. Podomeres distal to femoro-patellar geniculation altogether shorter than proximal ones. Strigilis absent. Four pairs of legs. Coxae 1, 2, 3 not distinct, altogether c. 0.5-fold as long as trunk. Femur c. 0.9-1.1fold as long as trunk. Articulation between tibia, tarsus and propodus indistinct, total length c. 1.2-1.6-fold that of trunk, longer than femur. Tibia conspicuously compressed when distinct (compared with extended legs), deduced to be c. 1.5-fold as long as trunk by comparison with extended legs.
Remarks. Our interpretation of Palaeoendeis elmii relies only on the holotype MNHN.F.A49277; it is likely that specimen MNHN.F.A52386, originally designated as a paratype by Charbonnier et al. (2007a), belongs to another species (see below). As already noted by Charbonnier et al., the abdomen is visible dorsally but chelifores and palps are apparently absent (Fig. 10).
The legs and body are slender. However, we see no trace of segmentation on the trunk (Figs 3A, 10; cf. Charbonnier et al. 2007a), although it is possible this detail was lost while extracting the specimen with chisels and sandpaper.
Articulation of legs can be determined by the bending of the second, third and fourth right legs. Femur and patella are clearly individualized. It is possible to identify coxae 1, 2 and 3, although it is not possible to distinguish the articulation between the three podomeres. It is also difficult to determine the boundary between tibia and tarsus, as well as to differentiate the tarsus from the propodus, if ever preserved. Interestingly, the patella is found to be much shorter than femur or tibia in the second, third and fourth right legs. However, we suggest that this resulted from taphonomic deformation, given that the sum of the length of the podomeres as shown in these legs is shorter than the length of the fully extended legs.
Tomography showed ovigers on the ventral side of the specimens (Figs 3A, 11B). Podomeres are not discernible, but it is possible to confidently identify the femoro-patellar geniculation. The two ovigers cross in their middle. Distal to the femoro-patellar geniculation, the podomere set is shorter than proximally. There is no evidence of a strigilis composed of four podomeres, instead there is evidence of a single distal podomere (possibly corresponding to tibia or the first tarsal podomere; Fig. 1).
PANTOPODA indet.  Remarks. The coxal geniculations of the legs can be identified on the surface of the fossil, while femoro-patellar geniculations are always directed in the opposite direction. The patello-tibial geniculation is also conspicuously directed downward. Given that it is unlikely that taphonomic processes could have resulted in twisting all sea spider legs the same way, we regard the specimen as preserved ventrally, despite the abdomen being visible at the surface. Abdomens positioned ventrally are rare but are known, for example, in the extant species of the subfamily Hedgpethiinae Pushkin, 1990 (Colossendeidae). The short length and the absence of visible segmentation of the abdomen is typical of Pantopoda. A ventral abdomen is incompatible with P. elmii (see above). The distance between lateral processes is also larger in MNHN.F.A52386 than in P. elmii, in which they  Table S2 for details). C, 'normals visualization'. D, F, interpretative drawing; plain black lines correspond to the outline of the fossil; dashed black lines correspond to the imprint of the leg; dotted lines correspond to putative position of articulations; grey lines correspond to the main breaks in the fossil. Abbreviations: acf, anterior cephalic feature; cx1-3, coxa 1-3; f, femur; mcf, median cephalic feature; pa, patella, pr, propodus; ta, tarsus; ti, tibia. Scale bars represent 5 mm.
are almost touching. The femora/trunk ratio is also different (0.8 in P. elmii holotype, 0.6 in MNHN.F.A52386). The structure of the cephalon is puzzling. In the absence of the proboscis (or its imprint) it is not possible to know whether the cephalon is complete. Ambiguity prevents us from interpreting the absence of cephalic characters, such as ocular tubercle, palps and chelifores, which can be carried in extant pantopods on the distal part of an elongated head (e.g. Phoxichilidiidae and Pallenopsidae, or the genus Dromedopycnon). The ocular tubercle is also absent in several extant bathyal species (e.g. Ascorhynchus ovicoxa Stock, 1975).
There is no evidence of ovigers in MNHN.F.A52386, nor of any oviger insertion. This cannot be attributed to the possible incompleteness of the cephalon because the ovigers' insertion always occurs near the first lateral processes in Pantopoda. Therefore, ovigers were probably absent in the living specimen. This is shared with the female specimens of three extant families, Endeidae, Phoxichildiidae and Pycnogonidae. This specimen is therefore likely to be female, although it is worth noting that in some species of Pycnogonum the ovigers are missing in both male and female specimens, and that in Austrodecus Hodgson, 1907 the ovigers can be absent either in both sexes or in the male sex only.

Identity of specimen MNHN.F.A52386 and preservation state in Palaeopycnogonides gracilis
Palaeopycnogonides gracilis has a ventrally directed abdomen, which is rare among sea spiders. But surprisingly, it is shared with the unidentified specimen MNHN.F. A52386, inviting the question of whether they could be the same species. The formerly misidentified fossil is quite similar in length to P. gracilis material. MNHN.F.A52386 is incomplete because it lacks both the proboscis (unpreserved) and the ovigers (most probably absent in the living specimen, given that no traces of insertions are observed). Therefore, the distinction between MNHN.F. A52386 and P. gracilis relies on two characters: (1) the cephalic ornamentation; and (2) the body being much stouter in P. gracilis than in MNHN.F.A52386. However, neither of these arguments are robust as a consequence of the preservation state of the specimen: 1. Due to the absence of a proboscis, it is not possible to tell whether the cephalon is complete distally, or if it is broken. If complete, the general aspect of the cephalon, in size and proportions, is similar to P. gracilis.
However, the anterior and median cephalic features (interpreted here as two median tubercles, see above) are inconspicuous in MNHN.F.A52386. But the dorsal surface of MNHN.F.A52386's cephalon is not flat, and a few bulges could correspond to poorly preserved tubercles. No such bulges are observed elsewhere on the trunk of the specimen. These bulges are not positioned mediodorsally, although the cephalon was probably deformed, as indicated by the left crack between the cephalon and the first lateral process, and by the twisting of the anterior section of the cephalon. Although none of these observations provides support to link MNHN.F.A52386 with P. gracilis, it is not possible to regard cephalic features as arguments against this hypothesis. 2. The fact that the abdomen is hardly visible and that ovigers are preserved only as an outline in MNHN.F.A52384, may indicate that P. gracilis fossils appear thicker than they would have been in life due to heavy pyritization. This is possibly supported by x-ray tomography data that suggest zones of lesser density inside the fossils (Fig. S2). The shape of these zones is very similar to that of MNHN.F.A52386. However, this inner outline could alternatively be either a taphonomic artefact, the internal surface of the cuticle or the remains of the digestive organs. We are unable to identify any diagnostic features to conclusively support either hypothesis. There is no clear reason why the specimens identified as P. gracilis should be preserved differently from other fossils. The sea spider fossils in La Voulte have been sampled in the same location (Ravin des Mines) but in four different layers: layers g-9, g-7, g-3 (all grouped within a 146-156 cm layer), and layer a (c. 190-485 cm deeper;Fischer 2003;Charbonnier 2009). It is therefore possible that fossils in different layers went through different taphonomic processes, leading to different states of preservation. Unfortunately, the information about the level at which each fossil was sampled has been lost (S. Charbonnier pers. comm.). It seems therefore impossible to reach a conclusion about the identity of MNHN.F.A52386 with the sole available specimen.

Affinities with extant taxa
The name Pantopoda Gerst€ acker, 1863 was originally synonymous with Pycnogonida Latreille, 1810; its use to designate an order within Pycnogonida was proposed for the first time by Hedgpeth (1954) and is now widely accepted  1954, 1978Bergstr€ om et al. 1980): both species have characteristic ring-like structures near the legs' insertion, and a long, segmented abdomen. Strongly sclerotized ring-like structures are not seen in the leg series of any other arthropods: Bergstr€ om et al. suggested that these had different origins in P. problematicus and P. maucheri, and that they belonged to the legs in the former and to the trunk in the latter. As such, this character requires reinvestigation, and it is of little relevance to justify the order Pantopoda given that we cannot tell whether the absence of this feature in extant species is apomorphic or plesiomorphic. Instead, the segmented abdomen in P. problematicus and P. maucheri (five-segmented in the former and three-segmented in the latter) is arguably plesiomorphic; so that its reduction is the sole unambiguous synapomorphy of Pantopoda. The La Voulte-sur-Rhône fossils have none of the features typically found in Devonian fossils (paddling legs, annular structures at the base of legs). More importantly, their abdomen is either extremely reduced or inconspicuous (Palaeopycnogonides gracilis, Colossopantopodus boissinensis), or unambiguously unsegmented (Palaeoendeis elmii, and MNHN.F.A52386). Therefore, their placement within Pantopoda is supported. In addition, Charbonnier et al. suggested that these fossils had affinities with extant families or superfamilies: P. gracilis with either Ascorhynchoidea (sensu Bamber 2007) or Pycnogonidae, C. boissinensis to Colossendeidae, and P. elmii to Endeidae. However, these affinities were so far only tentatively proposed. Drawing on the new morphological data provided by x-ray microtomography and RTI, we were able to reinforce or correct the initial proposals of Charbonnier et al. (2007a).

Palaeopycnogonides gracilis
In their original description, Charbonnier et al. (2007a) proposed to group P. gracilis with Ammotheidae sensu Stock (1994) [= Ascorhynchoidea sensu Bamber 2007], a group that is now regarded as probably paraphyletic (Arabi et al. 2010;Ballesteros et al. 2021;Sabroux et al. 2023). However, this hypothesis relied on the interpretation of anterior cephalic features as chelifores, an interpretation that our new data do not support. Alternatively, Charbonnier et al. (2007a) suggested that this species could also be related to Pycnogonidae, and Bamber (2007) proposed Endeidae affinities, interpretations that are more consistent with the lack of chelifores and palps. Endeidae affinities would be strengthened if one accepts the median cephalic feature as an ocular tubercle (hypothesis 2), which is typically positioned between lateral processes in extant Endeidae (as well as Callipallenidae, Nymphonidae and some Colossendeidae, but these are unlikely to be closely related given that all have either chelifores or palps). Endeidae often show two swellings (often called a 'collar'; see for example Stock 1994) on the cephalic anterior margin, but this structure is paired and always extends laterally; as in Endeis charybdaea (Dohrn, 1881), for example; small spines can also stand at this position (e.g. as in Endeis spinosa (Montagu, 1808)) but these are also always paired. As explained above, we do not know of other cephalic tubercle-like structures that could lay in this position in extant species. This makes this hypothesis less likely than the alternative: that the anterior cephalic feature is the ocular tubercle (hypothesis 3), which fits better with the Pycnogonidae hypothesis. But both Endeidae and Pycnogonidae have different ovigers from P. gracilis: in Endeidae, the ovigers have very long proximal podomeres, shorter distal podomeres and no strigilis, making them project posteriorly; in Pycnogonidae, ovigers are very small and remain in the anterior region of the body. In P. gracilis instead, the ovigers are V-shaped as found for example in Ammotheidae, Nymphonidae and Colossendeidae. Both Pycnogonidae and Endeidae lack ovigers in the female sex, while it is impossible to say whether there was such a sexual dimorphism in P. gracilis. A third potential candidate as a relative to P. gracilis is the enigmatic Queubus: the shape of the bulbous proximal part of the proboscis as observed in MNHN.F.A52384 may F I G . 9 . RTI of Colossopantopodus boissinensis, MNHN.F.A52385, body region (A, C-F) and ventral surface of the head of Colossendeis macerrima Wilson, 1881 (B). A-B, default view. C-D, specular enhancement (circles on bottom right indicate light orientation, see Table S2 for details). E, 'normals visualization'. F, interpretative drawing; plain black lines correspond to the outline of the fossil. Abbreviations: ob, oviger base; pb, palp base. Scale bars represent: 5 mm (A, C-F); 1 mm (B).
correspond to the basal flexible peduncle of Queubus; chelifores are missing; palps of Queubus are either absent or present but are strongly reduced in the male sex only; and ovigers are present in both sexes and have a typical V shape (Bamber & Steffani 2007). The position of Queubus in the phylogeny of sea spiders is not yet known. Palaeopycnogonides is therefore, at best, on the stem of either Pycnogonidae or Endeidae, may be related to Queubus, or the species may belong to a completely extinct Pantopoda lineage. In any case, the combination of cephalic appendage characters of P. gracilis differs from any other described family, such that P. gracilis should be regarded as belonging to a new family, which we name here Palaeopycnogonididae.

Colossopantopodus boissinensis
The gigantic size of Colossopantopodus boissinensis (11.5 cm leg span) recalls the gigantism in the extant family Colossendeidae. Representatives of this family routinely have leg spans beyond 10 cm, some species even reach 40 cm (Arnaud & Bamber 1987). All La Voulte fossils are 'gigantic' to some extent (Palaeopycnogonides gracilis and Palaeoendeis elmii holotypes have a leg span of c. 7 cm) when compared with extant fauna, therefore size alone cannot be conclusive for the taxonomic assignation of C. boissinensis. The cephalic appendage set of C. boissinensis (i.e. chelifores absent, palps present and ovigers carrying a strigilis) is typical of Colossendeidae Jarzynsky, 1870, Rhynchothoracidae Thompson, 1909and the genus Pantopipetta Stock, 1963(Austrodecidae Stock, 1954. But austrodecids are characterized by a pipette-like proboscis and rhynchothoracids by a pyriform one; C. boissinensis shows instead a long, cylindrical proboscis, which is more consistent with Colossendeidae. Colossendeids are divided into two subfamilies, Colossendeinae Hoek, (including Colossendeis Jarzynsky, 1870Decolopoda Eights, 1835;Dodecolopoda Calman & Gordon, 1933;Pentacolossendeis Hedgpeth, 1943) and Hedgpethiinae Pushkin, 1990(including Hedgpethia Turpaeva, 1973and Rhopalorhynchus Wood-Mason, 1873. Colossopantopodus does not share the proximally narrowed proboscis typical of modern Hedgpethiinae (Bamber 2007), which excludes the fossil from its crown-group. The bases of the ovigers touch each other and the palps' bases are positioned ventrally to the head, which are typical of modern Colossendeinae. These characters suggest that Colossopantopodus is related to Colossendeinae, as already suggested by Sabroux et al. (2019). The trunk segments of Colossopantopodus being individualized rather than fused as in most Colossendeinae, it is possible that Colossopantopodus belongs to the Colossendeinae stemgroup rather than its crown-group. This character is shared with Pentacolossendeis, a genus that is traditionally classified in Colossendeinae (e.g. Bamber 2007) although its relationships have never been formally addressed in phylogenetic studies. Other characters of C. boissinensis, such as the absence of the mid-length swelling of the proboscis, and the wide distance that separates the palpal and ovigeral bases, are not found in other Colossendeidae. The preservation of C. nanus does not enable observation of these latter characters in the Solnhofen species, which prevents us from considering these characters as diagnostic for Colossopantopodus.

Palaeoendeis elmii
The absence of palps and chelifores is compatible with two extant families, Endeidae and Pycnogonidae. However, the shape of the ovigers of P. elmii differs from the tiny ovigers of Pycnogonidae and is instead compatible with Endeidae (although it is not possible to make a podomere-to-podomere identification owing to the preservation) (Fig. 11). Affinities with Endeidae were already suggested by Charbonnier et al. (2007a) andBamber (2007). Interestingly, the oviger's patella is not curved, as observed in all extant species of Endeidae, which may justify including this species in a different genus than Endeis. It is, however, not to be excluded that the absence of curving is either a derived character (given that the oviger's patellae are often variably curved in most extant families), or a result of taphonomic deformation. The morphology preserved in the fossil is consistent with the diagnostic characters of Endeis as stated by Child (1992) (which are the same as Endeidae as a whole): trunk slender, lateral processes short and well separated, proboscis long, inflated distally, chelifores and palps absent in adults.

Fossil calibrations for the Pycnogonida tree of life
In their phylogenetic reconstruction of the Pycnogonida tree of life, Ballesteros et al. (2021) were the first to address F I G . 1 0 . RTI of Palaeoendeis elmii, holotype MNHN.F.A49277, body region (preserved dorsally). A, default view. B, specular enhancement (circle on bottom right indicates light orientation, see Table S2 for details). C, 'normals visualization'. D, interpretative drawing; plain black lines correspond to the outline of the fossil. Abbreviations: ab, abdomen; ov, oviger. Scale bars represent 5 mm. F I G . 1 1 . Comparison of the ovigers of Palaeoendeis elmii with some extant pantopods. A, ovigers as found in the two sexes in eight examples of modern sea spiders, using the same colour code as in Figure 1. B, close-up of the ventral view of the three-dimensional reconstruction of P. elmii, holotype MNHN.F.A49277. C, interpretative drawing of the ovigers of P. elmii; plain black lines correspond to the outline of the fossil; dotted lines correspond to the putative position of articulations. Abbreviation: fpg, femoro-patellar geniculation. Scale bars represent 2 mm. the timing of pycnogonid evolution. Their molecular clock calibration relied inter alia on four Pycnogonida fossils: Haliestes dasos (Siveter et al. 2004) and the three species of La Voulte-sur-Rhône. Calibrated nodes based on the Jurassic material followed the systematic interpretations of Charbonnier et al. (2007a) (Fig. 12).
F I G . 1 2 . Fossil occurrences and calibration intervals for nodes in the Pantopoda phylogenetic tree based on La Voulte-sur-Rhône fossils. Coloured circles indicate sea spider fossils (for a review see Sabroux et al. 2019). When possible, fossils are set on the branch to which they are the most closely related. Palaeozoic fossils are either regarded as stem Pantopoda or their status as pantopod is dubious, and are not set on any branch. Bars on the tree nodes indicate, for the node they are overlapping, the fossil-calibrated time intervals between hard minima and soft maxima, as should be used to define priors in molecular clock analyses. Justifications for these calibration intervals are provided in the main text. Phylogenetic relationships are those supported by Sabroux et al. (2023)

CONCLUSION
Because they are the best preserved of the known pantopod fossils, a better understanding of the La Voultesur-Rhône sea spider morphology is a keystone in the understanding of the early evolution of Pycnogonida. In aiming to resolve the timing of sea spider evolution through tree-dating analyses, Ballesteros et al. (2021) calibrated their molecular clock based, inter alia, on these three fossils, relying on the taxonomic interpretations of Charbonnier et al. (2007a). Here, we have supported their interpretation of Colossopantopodus boissinensis as a member of the family Colossendeidae and suggest further affinities with the subfamily Colossendeinae. We also support Charbonnier et al.'s hypothesis of Palaeoendeis elmii being an Endeidae. In neither case can a robust argument be made that they are crown-group members of these clades, as opposed to being in the respective stem-groups. In the phylogenetic framework of Sabroux et al. (2023), these fossils are most appropriately used to calibrate the nodes Colossendeidae and Endeidae + Phoxichilidiidae. Finally, we suggest that rather than attributing Palaeopycnogonides gracilis to Ammotheidae, this fossil should be assigned to its own family: Palaeopycnogonididae. Affinities of Palaeopycnogonididae to extant pantopods are not yet clear, although we suppose a closest relationship to either the families Endeidae or Pycnogonidae or to the genus Queubus, rather than to Ammotheidae. The role of the sea spider fossils of La Voulte-sur-Rhône in calibrating nodes of Pycnogonida phylogeny is refined by these new interpretations.
Data for this study, including raw RTI pictures, RTI files, computed tomography and segmentation data, are available in the University of Bristol Research Data Repository: https://doi.org/10.5523/bris.24dj1xxj4fvi022hm6mpq5gyx1.

SUPPORTING INFORMATION
Additional Supporting Information can be found online (https:// doi.org/10.1002/spp2.1515): Figure S1. Full view of the six La Voulte-sur-Rhône fossil slabs included in this study. Figure S2. CT slices of the trunk of Palaeopycnogonides gracilis, specimens MNHN.F.A52381 and MNHN-F-A52384 showing internal low density outline inside the pyritized fossils. Table S1. X-ray tomography parameters used with the v|tome|x 240 L scanner for the La Voulte-sur-Rhône sea spider fossils.