First African thylacocephalans from the Famennian of Morocco and their role in Late Devonian food webs

Thylacocephalans are enigmatic arthropods with an erratic Palaeozoic and Mesozoic fossil record. In many of the few localities where they occur, they are quite abundant. This also holds true for the Famennian Thylacocephalan Layer in the Maider (eastern Anti-Atlas of Morocco), a small epicontinental basin hosting some strata with taphonomic properties of a conservation deposit yielding exceptionally preserved gnathostomes and non-vertebrates. In a thin argillaceous interval in the earliest middle Famennian, thylacocephalans occur in such great numbers that they became eponyms of this unit. Therein, we discovered a new taxon of thylacocephalans, Concavicaris submarinus sp. nov., which represent the oldest records of thylacocephalans from Africa. In the CT-imagery, the holotype of Concavicaris submarinus sp. nov. revealed anatomical details including its eyes, appendages and other soft parts. Sedimentary facies and faunal composition of the Thylacocephalan Layer suggest that these animals populated the water column above the low-oxygen sea floor. Thus, thylacocephalans likely represented an important component of the diet of chondrichthyans and placoderms, which are quite common as well. The abundance of thylacocephalans in other conservation deposits like the Cleveland Shale (USA) and the Gogo Formation (Australia) underline their pivotal role in Late Devonian pelagic food webs.

Thylacocephalans constitute a group of small to medium-sized marine arthropods of unclear affinities. Their first undisputed appearance in the fossil record dates back to the Silurian 1 , although some fossils of Early Cambrian age 2 may also belong to this group. Thylacocephalans lived until the Late Cretaceous period; the youngest fossils were found in the Cenomanian of Lebanon [3][4][5] .
Their key anatomical features include a body almost entirely enclosed within a bivalved carapace (in the wide sense) that usually carries a pointed rostrum in the front. Below the rostrum, a pair of multi-facetted compound eyes was situated. These eyes reached giant proportions in Mesozoic forms 1,6 . Three pairs of subchelate raptorial appendages are located at the front of the body, but it is not entirely clear from which segments they arise 1 . Eight pairs of lamellate gills and a series of at least eight pairs of small paddle-like limbs were present at the posterior part of the body 1 . Although various anatomical features are known from exceptionally well-preserved specimens, many questions related to their palaeobiology and systematic position are yet to be resolved 6 . For example, their systematic position remains disputed due to unknown anatomical features such as their cephalic appendages. A recent finding from the Silurian of the US suggested the presence of antennae in a very-well preserved specimen, which would increase the likelihood that they really belong to the crustaceans 1 , but this character may not be very reliable 1 . Additionally, Haug et al. points out that the presence of several enditic structures on the legs supports a position in the Eucrustacea 1 . Thylacocephalans from the Devonian are known to occur in a few places in Europe 5,7-9 , Australia 10 and North America (e.g.) 11 . The Moroccan occurrence described here represents, as far as we know, the first record of thylacocephalans from Africa. This is also the second report of African thylacocephalans., the other being from the Early Triassic of Madagascar 12 .
Late Devonian strata of the eastern Anti-Atlas are known to be highly fossiliferous, yielding both vertebrates and invertebrates [13][14][15][16] . The Middle Famennian outcrops contain a layer attributed to the Maeneceras genozone 13,[17][18][19] , erroneously called 'Phyllocarid Layer' in an earlier paper (Ref. 15 , renamed in 20 ). The specimens Description. The holotype, PIMUZ 37349, is 56 mm long, 26 mm high and 15 mm wide (Fig. 1a). The biggest specimen available to us, PIMUZ 37348, lacks the posterior end, but its total length can be estimated to have reached nearly 100 mm at a height of 46 mm (Fig. 1c). PIMUZ 37354 lacks the rostrum and is the smallest specimen, which is 27 mm long and 10 mm high (Fig. 1f). AA.MEM.DS.4 and PIMUZ 37354 lack their rostrum. All specimens reach the maximum carapace height anterior to the centre.
There is no posterior spine in any of the specimens. When the rear end is preserved, it appears always truncated and gently rounded. The anteroventral process of the optic notch is small in size and visible in most specimens (Fig. 1).
The most noticeable diagnostic feature of Concavicaris submarinus lies on the ventral margin ( Figs. 1 and 2). This edge has a section anterior to the centre, which is bent outwards, i.e. laterally (e.g. Fig. 1a,d,e). This is best visible in the least deformed specimens (AA.MEM.DS.3 and 4, PIMUZ 37348, 37349, 37353, 37354). Along the ventral margin, a depression runs longitudinally from the anterior starting point of the bent margin along a third of the carapace length (Fig. 1a,b).
The large compound eyes of Concavicaris submarinus are well preserved in the holotype and reach a seventh of the body length (see on Fig. 2). Their structure is clearly visible under the microscope and shows a clear and regular pattern of convex, slightly elongated, hexagonal facets ( Fig. 3a-c).
The carapaces of eight specimens preserve the cuticle in several places. Its good preservation in various areas of the carapace (e.g., PIMUZ 37349) reveals the polygonal imprints of cuticle cells, bordered by thin grooves Scientific RepoRtS | (2020) 10:5129 | https://doi.org/10.1038/s41598-020-61770-0 www.nature.com/scientificreports www.nature.com/scientificreports/ (Fig. 3d). The polygons are irregular in the number of sides, overall shape and size. Some cells also show a more or less central depression (see black arrows on Fig. 3d).
In PIMUZ 37349 eight pairs of paddle-like limbs are preserved at the posteroventral end (only visible in the CT-images) -one leg per side being incomplete (Fig. 2). The legs on its left side are taphonomically deformed,  making them appear much thicker (6.1 mm compared to 2.4 mm on the other side) than long (6.5 instead of 14 mm on the other side). Because of the deformation, it is conceivable that the legs have been pushed outwards, which could explain why many of them seem to be detached from the body. The legs expand from posterior of the curved ventral margin until shortly posterior of the carapace where they reach out of the carapace. The trunk segments are not visible in the scan (Fig. 4), either because they did not preserve or because of an insufficient density contrast.
One of the most remarkable features seen in the holotype is the presence of thin structures on both anterior sides of the specimen (Fig. 4a,b). They appear to surround the area where the stomach is assumed to be, which could imply that these thin structures could be gastric muscles.
The CT scan revealed another structure near the putative gastric muscles. Three pairs of lamellar structures are visible (Fig. 4c,d). They are thin, divided in leaf-like sections and resemble gills.
Additionally, a pair of elongated organs are located dorsally to the presumed gastric muscles and anteriorly to the supposed gills (Fig. 4e,f). There is a clear separation between the two parts of this tissue, thus suggesting that it is a paired structure.
Remarks. Thylacocephalans are undoubtedly arthropods, but their affinities within this clade are still widely debated 1-3 . This lack of systematic resolution roots in the very incomplete knowledge of thylacocephalan anatomy. Therefore, exceptionally preserved materials such as those presented here can provide valuable anatomical details helping to better resolve trees comprising thylacocephalans. In the following, we discuss the structures that became visible in the CT-scan.
The cuticle displays a polygonal cuticular pattern separated by thin grooves, which are characteristic of thylacocephalans 9 . Secrétan 24 and Broda et al. 9,25 also suggested that the presence of a central depression in these polygons, the lumen (Fig. 3d), may represent pores where sensory setae inserted.
The eyes of C. submarinus have features strongly resembling the eyes of the Jurassic thylacocephalan Dollocaris ingens from La Voulte 26 , although the eyes of Dollocaris are significantly larger. Similarly, the Moroccan material has eyes composed of a dense and relatively regular pattern of hexagonal facets (Fig. 3b,c). However C. submarinus shows facets a little bit more elongate and a more hexagonal shape than Dollocaris in a similar surface area of the eye (Fig. 3e,f). The facets are also convex while those of Dollocaris appear to be more concave, probably due to post-mortem collapse 26 . www.nature.com/scientificreports www.nature.com/scientificreports/ The thin and flattened, yet slightly inward curved structures (Fig. 4a,b) resemble muscles. The location on these structures (anterior and lateral) implies that there could have been an organ or other internal structure between. The position of the putative muscles surrounds the space where we would expect the stomach 3,26 . Thus, we interpret these structures as gastric muscles. Further remains of musculature are found on the anterior right side of the holotype. They lie above the area where the stomach likely used to be. We think that this could be remains of an anterior gastric muscle. Among the presumed gastric muscles (Fig. 4a,b), some can be tentatively interpreted as external mandible adductor muscles, similar to what has been described from decapod crustaceans 27 , but the possibility that these could be lateral gastric muscles cannot be ruled out until better preserved specimens reveal more anatomical details. No mandibles or other mouth parts have been reported from thylacocephalans so far, implying that the gastric muscle interpretation is more probable than being part of the mouth parts.
The leaf-like lamellar structures show three branches on each side (Fig. 4c,d). This could imply paired organs that are close to one another. The position of these structures is compatible with their interpretation as gills. However, gills normally should consist of eight pairs of lamellae, which is characteristic of thylacocephalans. Also, they are not as long as expected. This suggests that some lamellae are probably simply not preserved, just like most other internal organs and structures, or they were not visible in the CT-imagery due to an insufficient contrast in density.
The shape of the paired structure segmented in green (Fig. 4e,f) resembles that of gonads in some decapod crustaceans, which also concurs with the phylogenetic position of thylacocephalans. Kienbaum et al. 28 discussed the gonads of decapods and explained that the ovaries are a pair of elongated organs located dorsally in the cephalothorax, a description fitting well with the structures found in the holotype. Although it is expected that gonads appear further posteriorly, a few species seem to have them more anteriorly than thought. Nagaraju 29 detailed the same physiology and position in the body as Kienbaum et al. 28 , and located the ovaries on top of the stomach and hepatopancreas in some decapod crustaceans. The description of the hepatopancreas in crustaceans 30 could fit the location of such structure, but this can be omitted here because thylacocephalans have a large hepatopancreas located ventrally in the thorax 26 . Although the gonad hypothesis stands a bit more than other interpretations like the hepatopancreas, further findings will contribute to revealing the correct interpretation of this structure.
C. submarinus differs from all concavicarids with its near mid-ventral fold. Species like C. milesi has a deep "U"-shaped optic notch 10 , which is not the case in C. submarinus. Also, the cuticle of C. milesi is terraced, unlike that of C. submarinus. C. sinuata has a deep carapace and a thick rostrum that extends to the anteroventral process 10 . This is not visible in C. submarinus. C. submarinus has a shorter carapace than C. elytroides 10,31 and both have different types of cuticle; the cuticle of C. elytroides is striated while that of C. submarinus is much more similar to C. bradleyi. C. submarinus is a close relative of C. bradleyi 31 . They both have a convex carapace with a broad arch forming the dorsal margin. Both species have a similar outline with a rounded posterior end, a minute anteroventral process and a moderately-sized rostrum. Three differences distinguish these two species. First, the ventral margin of C. submarinus curves outwards while in C. bradleyi, the subcentral ventral margin is inflected inwards and upwards 31 . Secondly, C. submarinus shows a depression on the carapace reaching from near the curve to almost the rear end. Thirdly, the interior angle of the eye is sharper in C. submarinus than in C. bradleyi, here it is more rounded.  15,20 , we also assume that the exceptional preservation is linked with low oxygen levels of the bottom waters in the Maïder Basin. In the case of thylacocephalans, they demonstrated the presence of phosphate in the carapace, eyes and appendages, while the coarse crystals filling the void in the carapaces of most specimens are composed of yellow calcite. All thylacocephalans are embedded in flat red nodules; the red colour likely derives from haematite, which, in turn, probably goes back to weathered pyrite 20 .
Because their carapaces are thin or flexible, thylacocephalans often get deformed during diagenesis. The exceptionally well-preserved but slightly deformed material of Concavicaris submarinus corroborates that compaction of the sediments and the embedded fossils played a role (Fig. 1). Originally, the holotype of C. submarinus (PIMUZ 37349) was presumably broader when it was alive; parts of the legs are no longer attached to the body, the raptorial appendages are missing entirely, the eyes slightly shifted out of their position and one actually shifted towards the anteroventral process.
palaeoecology. Frey et al. 15 reported the remarkable abundance of gnathostomes (mainly chondrichthyans) in the Thylacocephalan Layer, especially when compared to other regions and time periods where other non-vertebrates dominated. These thylacocephalans probably were important food sources for at least some of the Late Devonian chondrichthyans or large fishes as evidenced from, e.g., stomach contents in sharks 32,33 and fish corpolites 34 . This is supported by previous studies that showed the co-occurrence of gnathostomes and crustaceans in the same layers in, e.g., Australia 32 and the United States 33 and by the association of thylacocephalans with remains of the shark Cladoselache from the Famennian Cleveland Shale 33 .
The invertebrate fauna from the Thylacocephalan Layer includes genus Guerichia 35 (elliptica and venusta) and Buchiola (Buchiola) 36 , which are both bivalves known to have inhabited water bodies with a well-oxygenated upper water layer and low-oxygen conditions near the sea floor or they settled on the sediment in phases of better oxygenated bottom waters. Frey et al. 15 highlighted the fluctuation of oxygen availability and its correlation with the global sea-level variation. This coincides with their hypothesis that oxygen levels decreased during this period due to reduced water exchange linked with a global regression 15 . Low oxygen conditions may have caused the sudden loss of numerous benthic and demersal species that require a higher level of oxygen to strive, while opportunists like Guerichia fared well. Thylacocephalans form the major part of the fauna during that time, thus indicating that these animals likely lived higher in the water column with normal oxygen content above the oxygen poor water.
Although the carapace of these thylacocephalans is slender and thus had a low drag, it enclosed most of the body including much of the appendages; this and the size of these paddle-like limbs do not suggest that they were capable of rapid swimming movements 26 . Their apposition eyes suggest an adaptation to normal light conditions in the euphotic zone 26 rather than to dark environments as previously suggested 6 . Their round eyes facing in most directions including downwards corroborate a life in the water column. This is supported by the associated fauna and palaeoecological interpretations of other thylacocephalans 2,3 . Therefore, Concavicaris submarinus was most likely spending most of its life in the water column well above the low oxygen layers (Fig. 5). Since we did not find the raptorial appendages (possibly ripped off after death), we cannot infer much about their feeding behaviour. In accordance with findings from other localities and ages, it is the most parsimonious to assume that these animals did have large raptorial appendages and that they likely were ambush predators. Stomach contents are Figure 5. Reconstruction of Concavicaris submarinus, preying on a conodont animal, with a chondrichthyan in the background. Raptorial appendages are reconstructed with a size and shape intermediate between older (Silurian) and younger (Mesozoic) forms. Although direct evidence is missing that these thylacocephalans fed on conodonts, these are one of the few groups of which fossil remains are known from these strata. Predatory behavior of thylacocephalans was shown for younger relatives of C. submarinus.