New insights into the evolution of portunoid swimming crabs (Portunoidea, Heterotremata, Brachyura) and the brachyuran axial skeleton

Portunoidea (Heterotremata) is a morphologically disparate taxon of true crabs (Brachyura) best-known for many of its representatives being considered “swimming crabs”. The term “swimming crab”, however, sometimes refers to a distinct taxon (traditionally to Portunidae within Portunoidea), and sometimes to a certain morphotype in which the 5th pereiopod (P5) has a specific shape that facilitates swimming. We use the term “P5-swimming crab” or “P5-swimmer” herein, not only to restrict it to the morphotype, but also to distinguish the swimming in question from other kinds of swimming in Brachyura. The evolution of P5-swimming crabs has not yet been satisfactorily investigated. In particular, it is not known whether the morphotype evolved several times independently in different lineages of Portunoidea or whether it evolved only once and was lost in several lineages. Ours is the first approach combining molecular with morphological data to result in a new phylogenetic positioning of some members of Portunoidea. For the first time, data from the axial skeleton and extrinsic musculature are used. Morphological examinations reveal that the axial skeleton and extrinsic musculature in P5-swimming crabs are more diverse than previously thought, with the exception of the P5 anterior coxa muscle, which originates at the median plate in all P5-swimmers. Ancestral state reconstructions based on parsimony reveal that the stem species of Portunoidea already showed the morphotype of a P5-swimming crab, but with a long merus which probably resulted in less effective P5-swimming than in extant P5-swimming crab species with a short merus. Several other extant taxa represent a reversal of the P5-swimmer morphotype to varying degrees, with some extant species showing a complete reversal of unambiguous P5-swimming crab character states—one example being the well-known common shore crab Carcinus maenas. The absence of a connection between interosternite 7/8 and the sella turcica (the secondary loss of the “brachyuran sella turcica”) in the stem species of Heterotremata, resulting in a junction plate which forms a cavity that offers room and attachment sites for the P5 extrinsic musculature is uncovered as preadaptation to the P5-swimmer morphotype in Heterotremata. This preadaptation is missing in Podotremata and Thoracotremata, the other two traditional main taxa of Brachyura. Supplementary Information The online version contains supplementary material available at 10.1186/s12983-022-00467-8.


Introduction
Brachyuran crabs of the taxon Portunoidea Rafinesque, 1815 display a wide range of morphological disparity. The taxon is well-known for many of its representatives exhibiting a specific morphotype in which the 5th pereiopod (P5) is shaped as a swimming leg, enabling these crabs to perform a special mode of swimming (see for example [21,29,51,52]). This particular mode of swimming is referred to herein as P5-swimming to distinguish it from other modes of swimming that have been documented in crabs that show no swimming modifications of their pereiopods or that have different morphological modifications for swimming [21,66]. Accordingly, we refer to the morphotype as the P5-swimming crab (or P5-swimmer) morphotype, while referring to P5 in this morphotype as the swimming leg. However, many portunoid taxa do not exhibit this morphotype. For example, representatives of Carcinus Leach, 1814 and Chaceon R.B. Manning & Holthuis, 1989 are typical walkers, entirely lacking swimming ability (but being able to bury), while others like Thia Leach, 1815 and Portumnus Leach, 1814 show morphological features primarily considered to represent adaptations to a burying mode of life [2,7,13,14,21,22,41,51]. Some representatives of the portunid subfamily Thalamitinae Paulson, 1875 are known to live in symbiotic relationships to other marine organisms, and in some of these cases the P5 is modified for grasping [6,12,20,62,65].
The morphology and evolution of portunoid crabs has been little examined considering the high level of disparity in this taxon. The external features of the P5-swimming crab morphotype/swimming leg were characterized by Hartnoll [21] on the basis of statements by Herter [24] regarding Liocarcinus holsatus (J. C. Fabricius, 1798). They include (1) a significantly shorter P5 merus than in a walking leg, (2) a P5 propodus and dactylus that are much broader than in a walking leg, and paddle-like, (3) a rotation in the P5 thoracal-coxal articulation axis of about 90° compared to a hypothetical ancestor (i.e. a subdorsal P5 coxa position) and (4) an increased range of motion in the P5 coxal-basi-ischial and meral-carpal articulations. According to Kühl [29], the external features of the P5-swimming crab morphotype/swimming leg also include (5) a carpal-propodal articulation axis that lies on the longitudinal propodus axis (rather than oblique to it like in a walking leg; see also [22]). Schäfer [51] emphasizes the significance of (6) relatively long setae arranged in dense fringes along P5 podomere margins for P5-swimming. However, whether a short merus really should be used to characterize a P5-swimmer is a matter of debate. Steudel [66] found several portunoid genera to exhibit all the typical P5-swimming crab features mentioned above except for a short merus. One of them-Carupa Dana, 1851-was anecdotally reported to swim, but with no details given on the swimming technique. It was concluded that these genera are generally able to perform the same swimming movements as "real" P5-swimmers, but not as fast or effectively [66]. However, this hypothesis was never corroborated by behavioural observations or kinematic studies. Another feature considered to be crucial to the swimming performance of a P5-swimming crab is the ability to bend the swimming leg in an antero-dorsal direction over the carapace [21,29,51,52,66].
When it comes to internal anatomy, there is a distinct lack of information about most taxa of Portunoidea, as there is about other Brachyura. Of particular interest are features concerning the inner skeleton, which is formed by infoldings of the exoskeleton [9,19,[56][57][58]. In the P5-swimmer Liocarcinus depurator (Linnaeus, 1758) (which is morphologically very similar to L. holsatus) the endophragmal system shows significant differences to that in non-swimmers such as Cancer pagurus Linnaeus, 1758 and Carcinus maenas (Linnaeus, 1758) [22]. These differences are associated with the enlarged P5 extrinsic musculature in P5-swimmers, and with differences in muscle attachment sites. Generally speaking, it is a combination of these inner morphological traits and external features (particularly those of the swimming leg) that permit P5-swimming and that can thus be used to characterise the P5-swimming crab morphotype [22]. Schmidt et al. [52] also found an increased range of motion in the P5 thoracal-coxal articulation in P5-swimmers, which matches findings by Hazerli and Richter [22] in which a larger P5 thoracal-coxal arthrodial cavity is present in L. depurator than in C. maenas and C. pagurus. However, the latter finding has yet to be confirmed in other P5-swimmers, as do features concerning the endophragmal system and extrinsic musculature. The first aim of this study, therefore, is to examine, describe and 3D-visualize the axial skeleton (= endophragmal system + pleurum + sternum; for more information on the terminology used here, see [22]) and P5 extrinsic musculature of various portunoid taxa (together with some outgroup taxa). This study is the most comprehensive to date on these structures in Portunoidea, and the first to test whether the traits (morphemes) found in L. depurator are also present in other P5-swimmers.
The main goal of the present study is to deduce the transformations involved in morphological character evolution in Portunoidea, and to test whether the P5-swimmer morphotype evolved several times independently or whether it appeared only once and was lost in some lineages. To answer this question, a robust phylogenetic hypothesis is necessary, which to date has not been achieved. Previous phylogenetic analyses of Portunoidea remain controversial, at least in part [12,26,54,63,64]. In the present study, a new phylogenetic hypothesis is formulated by combining available gene sequences with new morphological data pertaining to the inner anatomy and newly conceptualized characters of external structures (for the most part based on Karasawa et al. [26], though statements by Evans [12] and Spiridonov [64] are also considered in our character conceptualization). To understand the evolution of morphological traits, the ancestral states of characters are reconstructed on the basis of parsimony [49,55,67].

Micro-computed tomography (μCT), 3D reconstruction
At least one female and one male specimen per species (if available) were used for X-ray imaging using a XRadia Versa 410 X-ray microscope (ZEISS, Oberkochen, Germany) and the program Scout and Scan v.11 (Table 1). In species in which only one sex was represented in the voucher material, at least two specimens of the same sex (if available) were used. All specimens were adults except for one juvenile specimen of Necora puber and immature specimens of Callinectes sapidus and Portunus inaequalis. If extrinsic pereiopod musculature was barely visible in μCT scan images, the respective specimen was bathed in alcoholic Lugol's iodine solution for several days to improve the visibility of the musculature during μCT scans. Digital image stacks obtained via μCT were processed using the 3D reconstruction software Amira 6.4/6.5/6.6/6.7 (by FEI). Scans for 3D reconstruction of the axial skeletons and P5 extrinsic musculature were chosen on the basis of the quality of the scan and the condition of the extrinsic musculature. μCT scan image stacks were checked to see whether differences in the extrinsic musculature of pereiopods 2-4 (P2-P4) could be found between species. Some of these muscles were reconstructed in 3D models as well. The remaining scans were used to check whether there were differences between specimens of the same sex or between specimens of different sexes (if available) in each species (and optionally, the axial skeleton and extrinsic musculature were roughly reconstructed). Some μCT scans, 3D models and photographs of the specimens were used to produce images and drawings using CorelDraw 2020 and common graphics programs.

Terminology
The morphological terminology used here is basically the same as in Hazerli and Richter [22], with some adjustments pertaining to the details of the term endophragma (pl. endophragmata). In Hazerli and Richter [22], we adopt the definition of endophragma used by Guinot et al. [19] as infoldings of the axial skeleton "composed of two layers that are confluent with their cuticle at its outer margin" (and "technically inseparable at moult"). However, we found that in many species, large parts of axial skeleton infoldings generally defined as endophragmata consist of two separate, clearly distinguishable layers of cuticular infoldings. This applies in particular to large areas of the junction plate and median plate. In fact, the wall of the junction plate cavity always appears to be formed by only one layer of cuticular infolding. Plus, in many of the species in which interosternite 4/5, 5/6, and/or 6/7 is connected to the median plate, the two layers of median infoldings connected to interosternites are predominantly not confluent at their outer margin. These infoldings are not true endophragmata according to the definition used by Hazerli and Richter [22]. However, here we use a less strict definition of the term endophragma, simply labelling the median plate and interosternites as median and lateral sternal infoldings, and interopleurites as pleural infoldings, which together with the junction plate and sella turcica are referred to as endophragmata herein.

Molecular data set, phylogenetic analysis and ancestral state reconstruction
We created a combined dataset containing 2141 characters comprised of 59 newly conceptualised morphological characters and 2082 nucleotide sites. All characters were equally weighted. Gaps were treated as missing data. In all analyses, Sternodromia monodi was set as the most basal outgroup. Sequences were obtained from GenBank and derived from 63 vouchered specimens listed in Table 1. The final molecular data set comprises 33 operational taxonomic units (OTUs). When sequence data were not available for a certain species analysed for morphology, OTUs were concatenated of the species studied morphologically and sequence data of closely related species (see Table 1), an approach which has been used previously (e.g. [15] [8] and Evans [12]. The dataset was analysed using maximum parsimony (MP) and Bayesian inference (BI). For BI analyses, the best fit model of evolution was determined using the implemented model test in MEGA X. BI analyses were performed using MrBayes 3.2.7a [50] on CIPRES Science Gateway [36]. The GTR + G + I Model was applied. Each Bayesian analysis included four runs with four differentially heated chains and the analysis was run for 30 × 10^6 generations, sampled every 3000 generations. The first 10% were discarded as burnin. All parameters were checked with Tracer version 1.7 [44]. The MP analyses were performed in TNT v1.5. [16] using New Technology. In "New Technology Search" the following parameters were used: initial addseq = 30, find minimum tree length 100 times and random seed = 2000. Default settings were used for sectorial searches, tree drifting and tree fusing. Unsupported nodes were collapsed. Nodal support was assessed using 1000 replicates of standard bootstrap support ("New Technology Search" with sectorial search, ratchet, drift and tree fusing [default], initial addseqs = 30, find minimum length tree 2 times). The resulting trees were visualized using FigTree v.1.4.2 [43] and Corel Draw 2020. Ancestral state reconstructions were traced using the software Mesquite (build 927) by Maddison and Maddison [33]. They were based on the phylogenetic tree deduced from MP analysis by choosing the parsimony model 'unordered' as the most conservative option for optimizing character states.

Character conceptualization
We here follow Hennig ([23]; see also [17,18,48,74]) in assigning character states belonging to a transformation series (which in turn represents the character). We further treat the assignment of the same state to different taxa as "character state identity hypothesis", while the assignment of two (or more) different states to the same character represents one (or more) "transformational hypothesis" (or "hypotheses"). Both types of hypotheses can be seen as homology hypotheses of a kind (e.g., [23,47,68,72]). Characters are phrased as suggested by Sereno [59]. Character dependencies are expressed by "inapplicable (-) if " followed by the state(s) of the character(s) which make(s) the character in question inapplicable for a certain taxon. This work focuses on characters concerned with the axial skeleton, the extrinsic musculature of pereiopods 2-5 (P2-P5) and the external morphology of the 5th pereiopod (the extensively modified swimming leg in P5-swimmers), which are discussed in detail herein. With respect to remaining external features, characters are only briefly discussed and conceptualized mainly on the basis of statements by Karasawa et al. [26] and Spiridonov [64]. A complete character state data matrix is suppilied in the "Appendix" in Additional file 1.

Characters concerned with the axial skeleton and extrinsic musculature of pereiopods 2-5
Shape and proportion of overall axial skeleton An overview of axial skeleton morphology in most of the taxa is given in Figs. 1, 2, 3, 4, 5 and 6. High resolution 3D models of these species' axial skeletons including some of the extrinsic musculature of the pereiopods are supplied in the "Appendix" in Additional files 2, 3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27. Low resolution 3D models of some of the species' axial skeletons (Medorippe lanata, "Polybius" henslowii, Thia scutellata, Ovalipes ocellatus, Libystes nitidus) without extrinsic musculature are supplied in the "Appendix" in Additional files 28,29,30,31,32. No apparent differences in the shape of the axial skeleton or the extrinsic musculature of the pereiopods are found between sexes, except in Medorippe lanata (see below). No distinct differences were found between a juvenile and adult specimen of Necora puber either. The axial skeleton in all taxa consists of a ventral sternum, a left and right latero-dorsal pleurum, and infoldings deriving from them. Interosternites and interopleurites are, respectively, pairwise sternal and pleural infoldings between thoracomeres. The interosternites and interopleurites of each lateral body side are fused to form the two junction plates of the endophragmal system, the two lateral halves of which are connected by the sella turcica, which also seperates the cephalothorax from the pleon. Most species have a median plate, which is a medially uprising endophragma of the sternum (for more information on the terminology of morphemes, see [22]). Examples of the states concerned with axial skeleton proportions (character 1) are shown in Fig. 7A-D. The degree of dorsoventral sternal curving, which is conceptualized as two discrete states herein (character 2), can be seen in Fig. 7E-L. Characters concerned with sternum proportions (character 3) and shape (character 4) are shown in Fig. 8. Characters 1-3 are conceptualized with respect to thoracomeres 5-8 only, since the shape of thoracomere 4 depends on the size of the cheliped, which is not included here. Ashtoret lunaris is the only taxon in which the pleurites of thoracomeres 5, 6 and 8 expand noticeably in a medial direction, with interopleurites (= infoldings of the pleurum) 5/6 and 6/7 almost reaching the median plane of the axial skeleton. Although this is restricted to a single taxon, we conceptualize this as a character here (character state 5(1); Fig. 9). A minor medial expansion of pleurite 8 is also present in P5-swimming crabs (Fig. 9D, F, H).  However, we do not conceptualize this as a character state here since variation in this morphological feature is very high between species, with a more or less distinct medial expansion of pleurite 8 also being present in several non-P5-swimmers (Fig. 9A, G).

Medial margins of interosternites 4/5 to 6/7
The medial margin of an interosternite is characterised by an upper, medially directed interosternal expansion we herein term "interosternal process" (Fig. 14; for more information on the terminology of morphemes, see [22]). As it is difficult to conceptualize the morpheme variability seen in the different taxa into distinct character states, we only distinguish here between three discrete states which are not prone to subjective perception (character 18, Fig. 14). In all cases, the distinctiveness of processes decreases progressively from interosternite 4/5 to interosternite 6/7. For this reason, this character is conceptualized by the process shape of interosternite 4/5 only.

Shape of junction plate and sella turcica
In all ingroup taxa (in which interosternite 7/8 is not connected to the sella turcica; character state 19(1); Figs. 15G, H, 16A-D), each junction plate forms a cavity whose extension in an anterior direction is variable (Figs. 15H, I, 17). This is absent in all non-heterotrematan outgroup taxa (in which interosternite 7/8 is connected to the sella turcica; character 19(0); Fig. 15A-F). The shape of the "junction plate cavity" (for more information on the terminology of morphemes see [22]) is very variable between taxa, in most cases forming an anteriorly closed calyx with an asymmetrical posterior margin of which the ventral part is more anteriorly situated than the dorsal part (Fig. 15H, I, 17A-C, E). In some taxa, the ventral part of the calyx is completely missing, with the junction plate cavity rather resembling a convex roof (Fig. 17D). We here conceptualize the extension of the junction plate cavity in an anterior direction as a character, with states corresponding to the interosternites which are reached by the anterior end of this cavity (character 22; Fig. 17). In contrast to Guinot et al. [19], who suggested that in Podotremata a junction plate is lacking, we found in Sternodromia monodi that interosternites and interopleurites are indeed fused, forming a structure we here consider as a junction plate. However, the fusion is not as complete as in the other taxa, so that the junction plate appears to have large gaps (Fig. 1A). The sella turcica in S. monodi differs from that in all other taxa not only by not being directly connected with the two junction plates (character 21; Figs. 15A, D, G, 16A-D, 18I-L), but also by being fused with interosternites 4/5 to 7/8 (Fig. 1A). As borders between the elements of this fusion cannot be distinguished unambiguously, they are considered as diffuse transitions in 3D reconstructions, with the fusion reconstructed as being part of the sella turcica. The shape of the sella turcica is variable between taxa, making character conceptualization difficult (Fig. 18A-H). We here conceptualize its shape in only one unambiguous character which expresses whether the sella turcica covers the dorsal median plate margin (character 23; Fig. 18A-H). In outgroup taxa Eriocheir sinensis, Varuna litterata, and Medorippe lanata, the whole medial margin of interopleurite 7/8 is confluent with the sella turcica, while in all other taxa, the medial margin of interopleurite 7/8 is not or is only partly connected to the sella turcica (character 24; Fig. 18I-L). Extrinsic musculature of pereiopods 1-4 In contrast to the extrinsic musculature of P5 (see below), differences in the origin positions of the extrinsic musculature of P1-P4 in the taxa examined are limited in most cases to single terminals. We found no apparent differences between species in the origins of the extrinsic musculature of P1. In all taxa, the P2 anterior coxa muscle originates at interosternite 4/5, interosternite 5/6 and interopleurite 4/5 (character state 25(0); Fig. 19A, B), while the P3 anterior coxa muscle originates at interosternite 5/6, interosternite 6/7 and interopleurite 5/6 ( Fig. 19C). Only in Ovalipes ocellatus were parts of the P2 anterior coxa muscle additionally found to originate at interopleurite 3/4 (character state 25(1); Fig. 19D). In most of the species, the P2 and P3 anterior coxa muscles have a branch which originates at interopleurite 4/5 and 5/6, respectively, and runs along (but does not attach at) the medial side of the junction plate (Fig. 19A, C). This branch was not found in some taxa (Sternodromia  , but since it is very thin and the musculature in the voucher material representing these taxa is in parts poorly preserved, this might be artificial. We thus do not implement these findings in the character statements. In both T. cheiragonus and C. cassivelaunus the P2 posterior coxa muscle originates at interopleurite 4/5 and 5/6 (character state 26(1); Fig. 20A), while Libystes nitidus A. Milne-Edwards, 1867 is the only representative in which the muscle originates at interosternite 6/7 in addition to interopleurite 5/6 (character state 26 (2) ; Fig. 20C). In all other species, it originates at interopleurite 5/6 alone (character state 26(0); Fig. 20B). The P2 dorsal basi-ischium muscle originates at interosternite 4/5 and 5/6 in all terminals, but in Thia scutellata and Ashtoret lunaris, it additionally originates at the median plate, which is conceptualized here as a neomorphic character (character 27; Fig. 20D, E) that is scored as inapplicable in taxa in which the median plate is absent or does not extend to pereiomere 2/thoracomere 5 (character states 7(0), 8(1) or 8 (2)). In all taxa in which the median plate reaches thoracomere 5, the P3 and P4 ventral basi-ischium muscle originates at the median plate and interosternites 4/5 and 5/6, respectively. However, T. scutellata differs from the other taxa in that the P2 ventral basiischium muscle does not originate at interosternite 4/5 and the P3 muscle does not originate at interosternite 5/6  Fig. 23B, D). In all the other taxa the muscle originates at interosternite 5/6 and 6/7, but not at the median plate (character state 30(1); Fig. 23C), and additionally at the junction plate, near interopleurite 5/6, sometimes with some fibre bundles attached to it. However, as the attachment sites at interopleurite 5/6 are ambiguous in many specimens of different species, they are not included in the character statements. The origin positions of the P4 ventral basi-ischium muscle (character 31) in species in which the median plate reaches interosternites 5/6 or 6/7 (character states 8(1) and 8(2)) largely depend on whether the whole median plate is occupied by the P5 extrinsic musculature, which is the case in all the typical P5-swimming crabs (assigned to the morphotype on the basis of the criteria mentioned above, including a short merus; character state 31(2); Fig. 24F) except "Polybius" henslowii. In the latter and all the other species in which a median plate is present (character state 7(1)), part of the P4 ventral basi-ischium muscle also originates at the median plate (character state 31(1) ; Fig. 24E).  Extrinsic musculature origins at the junction plate are not considered in characters 25-33 because, as the junction plate is a product of the fusion of interosternites and interopleurites, with no visible sutures, muscular attachment sites here are ambiguous.  1); at interosternite 6/7 + 7/8 + interopleurite 6/7 (2). 30. Pereiopod 3, dorsal basi-ischium muscle, origin (Fig. 23): at interosternite 6/7 + interopleurite 5/6 (0); at interosternites 5/6 + 6/7 (1); at median plate + interosternite 6/7 (2); at median plate + interosternites 5/6 + 6/7 (3). 31. Pereiopod 4, ventral basi-ischium muscle, origin (Fig. 24C, E, F, G): at median plate (if present) + interosternite 7/8 (0); at median plate + interosternites 6/7 + 7/8 (1); at interosternites 6/7 + 7/8 (2). 32. Pereiopod 4, dorsal basi-ischium muscle, anterior branch, origin (Fig. 24A, B, D, F, H): at sternum (0); at interosternite 6/7 (1); at median plate + interosternite 6/7 (2); 33. Pereiopod 4, dorsal basi-ischium muscle, posterior branch, origin (Fig. 24A, B Extrinsic musculature of pereiopod 5 The extrinsic musculature of P5 is especially interesting as it is responsible for moving the swimming leg in P5-swimmers. Remarkably, in all P5-swimming crabs a branch of the anterior coxa muscle originates at the median plate. The muscle fibres in this branch are long because of the distance between the insertion point at the coxa and the origin at the median plate or its anterior process (which is present in all typical P5-swimmers assigned to the morphotype on the basis of the criteria mentioned above;    Fig. 28D). In outgroup taxa lacking a median plate, the ventral basi-ischium muscle only originates at the sternum (character state 38(0); Fig. 28A). With regard to the dorsal basi-ischium muscle, in all taxa but Sternodromia monodi and M. lanata, two distinct branches can be distinguished (character 39; Fig. 29), with the dorsal branch always originating at the sella turcica and in some species additionally at the sternum near the border between the sella turcica and the sternum (as the border is indistinct this is not considered a separate character state). In C. granulata and A. lunaris only, the dorsal branch also originates at the median plate (character state 40 (1) In typical P5-swimming crabs (assigned to the morphotype on the basis of the criteria mentioned above), the extrinsic musculature of P5 is considerably more voluminous than that of P2-P4. Voluminous musculature is also present in A. lunaris and Coelocarcinus foliatus Edmondson, 1930, though here it is similarly expanded anteriorly and dorsally, while in the P5-swimmers it is mainly expanded anteriorly. The expansion of musculature is associated with specific features of the axial skeleton, such as the anteriorly expanding process of the median plate (characters 9 and 10) in P5-swimmers. The dimension of the extrinsic musculature of P5 is thus not considered an independent character here.
As is the case with the extrinsic musculature associated with P2-P4, origins at the junction plate are not considered in characters 34-41.

Characters concerned with the external morphology of pereiopod 5
There is much variability between species in the external features of pereiopod 5. The occurrence and modes of arrangement of long setae on P5 can be seen in Figs. 1, 2, 3, 4, 5, 6 and 30A, B. In most species, long setae are arranged in dense fringes along podomere margins, as is also the case in a typical swimming leg (character state 43(0); Fig. 30B). We here express the shortened merus of a typical swimming leg through a comparison to P5 propodus length (character state 44(1) ; Fig. 30B). The latter has been shown to be relatively independent of merus length    . IntPl interopleurite (with number pair indicating thoracomeres between which it is situated), IntSt interosternite (with number pairs indicating thoracomeres between which it is situated), MP median plate, VentrBas P2 2nd pereiopod ventral basi-ischium muscle, VentrBas P3 3rd pereiopod ventral basi-ischium muscle [52]. The insertion of the propodus in the disto-dorsal carpus margin is associated with an alteration in the position of the carpal-propodal articulation axis relative to the longitudinal propodus axis (character state 45(1); Fig. 30B; see also [22,29]). This state is present in all typical P5 swimmers, but also (with varying distinctness) in some other taxa.
With regard to P5 dactylus shape, we here prefer a character concept with fewer states than in Karasawa et al. [26] as the variability of this character can be very high not only between species (Fig. 31) but also within species, implying that it is prone to subjectivity. For example, we found in Carcinus maenas that there were specimens that could be assigned to the state "ensiform" (Fig. 31D) based on Karasawa et al. [26], while others exhibited a "lanceolate" dactylus shape (Fig. 31E). Consequently, the character we conceptualize here has only two states reflecting a clearly apparent, intersubjective difference: whether or not the P5 dactylus is broader than the P5 merus (a broad "paddleshaped" dactylus being character state 47(1); Figs. 30B, 31C, F, G). A second character related to dactylus shape concerns its proximo-ventral margin. In species in which the margin is concave, the disto-ventral propodus area is not immersed in the arthrodial cavity when adducted (character state 48(1); Fig. 30B, E, F; see also [22]). This is the case in all the broad, paddle-shaped dactyli, but also in some of the narrow ones (see below).
The P5 propodus and dactylus are broader than the respective P4 podomeres not only in taxa with a broad "paddle-shaped" dactylus, but also in C. maenas (albeit  Fig. 32C; see also for example [21,22]). However, we do not conceptualize the P5 propodus and dactylus being broader than the respective P4 podomeres as an additional character herein. This is because in P5-swimming crabs and several other taxa with a paddle-like dactylus, the extreme broadening of the propodus is formed by a large lobe-like projection of the disto-ventral propodus margin (character state 46 (1); Figs. 30B, 32B, D) which is absent in C. maenas. In C. maenas, the broader propodus correlates with the relative volume of the musculature inserting at the dactylus, while in taxa with the propodus projection, this is not the case; i.e., the dactylus musculature is small in relation to propodus width (Fig. 32B, C,  D). Furthermore, the P5 proximo-ventral dactylus margin in C. maenas is that of a normal walking leg, which means that the broader dactylus is not associated with a concave dactylus margin (Fig. 30C, D, 31D, E). For this reason, we consider the broader distal P5 podomeres in C. maenas not to be homologous to those in P5-swimmers (i.e. neither an identical character state nor a different state of the same character-in other words, not part at all of the transformation series suggested by Hartnoll [21]. Schäfer [51] also noted the P5 carpus in C. maenas to be broader than that of the P4, which again is the consequence of the relative volume of the musculature inserting at the propodus (Fig. 32C). We found slightly broader distal P5 podomeres (even less pronounced than in C. maenas) associated with relatively more voluminous intrinsic musculature also to be present in Eriocheir sinensis and Cancer irroratus (Fig. 32A). An examination of external morphological features in another species of Eriochier (Eriocheir japonica (De Haan 1835)) even showed distal P5 podomeres there to be broader (compared to P4 podomeres) to a smiliar degree to those in C. maenas (Fig. 36 in "Appendix").
An extrinsic P5 anterior coxa muscle originating at interopleurite 6/7 (morpheme property 26 in [22]) is not considered in our conceptualization either since borders between the junction plate, interopleurite 6/7 and the sella turcica are not as apparent in many taxa as in those examined by Hazerli and Richter [22].

Other external characters
Several morphological features used in taxonomic approaches are not conceptualized as characters here because of their high variability between species. A discussion of this variability is beyond the scope of this paper, but suffice to say it makes the distinction of discrete states covering more than one species difficult (for a detailed discussion of many of the features in question, see [64]). They are concerned with the shape of the carapace front, the number of spines and fissures on the ventral orbit margin, the presence or absence of a spine or lobe at the inner angle of the dorsal orbit margin, the shape and number of anterolateral carapace spines, the presence or absence and configuration of carapace tubercles and ridges, the shape of the basal antenna article, the presence and shape of a "portunid lobe" on maxillipede 1 (touched upon by [66]) and the overall shape of the male pleon. Thus, only a few previously used external morphological characters are considered herein. Spiridonov et al. [63], who considered morphological features of the chelae an important character complex, described them in detail in many portunoid species but deduced no character concepts. We found a high degree of morphological chela variety in Carcinus maenas alone, with specimens exhibiting varying degrees of heterochely and heterodonty (Fig. 37A, B in "Appendix"). Homochelic specimens (homodontic or somewhat heterodontic) were observed too (Fig. 37C in "Appendix"). In C. maenas, this can perhaps be explained by morphological plasticity (see, for example, [3,11]). As data concerning intraspecific variety are lacking for most of the other species examined herein, we did not implement chela morphology into the character matrix either.
Karasawa et al. [26] drew distinctions concerning not only the distinctiveness of the sutures between male pleomeres, but also concerning pleomere movability. As the movability was difficult to determine in conserved specimens, we only conceptualize suture distinctiveness here (character 53). We determined character states that were different from Karasawa et al. [26] regarding the mode of connection of the basal antenna article to the suborbital region in Caphyra loevis, Raymanninus schmitti and Macropipus rugosus (character 54). The presence or absence of a subterminal spine on the first gonopod is adapted from Karasawa et al. [26], if data for the species were available (character 58).
Although we do not claim to present an exhaustive list of external morphological characters, we nevertheless conceptualize a few other conspicuous morphological differences as characters herein. They include (1) a putative respiratory canal formed by the prolongated endopodite of maxillipede 1 (character state 55; Fig. 38A-C in "Appendix"), which occurs in Ashtoret lunaris, Calappa granulata and Medorippe lanata (and constitutes a morphological feature that once was used to assign these species to Oxysytomata, a taxon that is widely refuted nowadays; [4,10,13]), and (2) the degree to which maxillipede 3 obscures the mandibles when adducted to cover the buccal cavern. This latter character constitutes an

Results and discussion of phylogenetic analysis
Both phylogenetic analyses of the combined data set (BI and MP) resulted in nearly the same topologies (Fig. 33).
In both analyses, the Portunoidea are recovered as monophyletic and composed of three distinct clades we herein name according to the classification system established by Evans [12]: Portunidae is the sister group to a taxon comprising the monophyletic Carcinidae and Geryonidae. The position of Portunidae deviates from the phylogenetic hypothesis put forward by Evans [12], in which Geryonidae is the sister group to Carcinidae and Portunidae. In our analysis, Parathranites orientalis is part of the Portunidae, being a sister taxon to the remaining portunid representatives which is s consistent with morphological data showing that all representatives of Portunidae including Parathranites orientalis have a median plate extending up to interosternite 5/6 (character state 8(1)), including a transverse sternal ridge from interosternite 6/7 to the anterior end of the median plate (character state 12(1); Fig. 12). In contrast, Evans [12], assigned this species to Carcinidae, though with the caveat that the support was not strong. As a personal communication Nathanial Evans offered some doubt about the information content of the 16S rRNA, particularly that Tsang et al. [70] excluded some hypervariable regions because of their focus on Brachyura in general, which nevertheless might have been informative for the relationships of Portunoidea. In conclusion, the placement of P. orientalis remains problematic and should be revised taking new genetic data into account.
The only species with this combination of character states not assigned to Portunidae was Macropipus rugosus (Carcinidae in both our analyses and that of [12]; see below). We further recovered monophyletic Thalamitinae (represented by Caphyra loevis, Caphyra rotundifrons and Lissocarcinus orbicularis) and monophyletic Carupinae Paulson, 1875 sensu lato (represented by Carupa tenuipes, Catoptrus nitidus and Libystes nitidus) following the nomenclature of Evans [12]. Only the positions of Callinectes sapidus and Portunus inaequalis differ between our MP and BI analyses (Fig. 33).
In accordance with Evans [12], the Geryonidae of our taxon sampling include Chaceon mediterraneus RB Manning & Holthuis, 1989, Ovalipes ocellatus and Raymanninus schmitti. The relationships between them differ depending on the analysis (Fig. 33). Within Carcinidae, the status of Carcinus maenas and Portumnus latipes as sister taxa is consistent with previous molecular-based phylogenies [12,54,63]. The positioning of the carcinid Pirimela denticulata (Montagu, 1808) as more closely related to Coelocarcinus foliatus is unexpected, as P. denticulata has repeatedly been identified as the sister taxon to the Carcinus-Portumnus clade [12,54,63]. Actually, the phylogenetic placement of C. foliatus has repeatedly caused problems due to the unusual external morphology of this species ("unusual portunid crab"- [37]; "certainly not a portunoid"- [26]), but Evans [12] placed it as a basal carcinid taxon on the basis of molecular data. Interestingly, the identification of C. foliatus as the sister taxon to carcinid Xaiva biguttata was based herein solely on morphological data, since no genetic data were available for this rare European species. However, C. foliatus lacks the P5 anterior coxa muscle originating at interopleurite 7/8 (character states 34(1), 34(3); Fig. 25) which is present in all other taxa of the Carcinus-Portumnus-Pirimela-Coelocarcinus-Xaiva clade, and finally, it is the only species of this clade not native to Europe [37]. Liocarcinus navigator, Liocarcinus depurator and "Polybius" henslowii form a monophyletic clade, but with L. depurator being more closely related to "Polybius" henslowii than to L. navigator, which is consistent with earlier findings recognizing the genus Liocarcinus to be polyphyletic [12,40,54]. In accordance with Evans [12], Macropipus rugosus and Thia scutellata form a monophyletic clade within Carcinide (Fig. 33). Interestingly for carcinid species, both taxa have an unusual axial skeleton configuration: in M. rugosus, the transverse sternal ridge runs from interosternite 6/7 to the anterior end of the median plate (character state 12(1); Fig. 12, otherwise only present in Portunidae), and in T. scutellata the median plate extends further than interosternite 4/5 (character state 8(0)), with all interosternites connected to the median plate (character states 13(0), 14(0), 15(0), 16(0); Fig. 11E), a configuration otherwise only present in the heterotrematan outgroup. The latter was represented by a clade comprising all non-portunoid Heterotremata but not Medorippe lanata (Fig. 33), which is positioned as sister to a clade comprising Heterotremata and Thoracotremata (not as a heterotrematan taxon as suggested by [25]). However, this and the positon of Ashtoret lunaris as the sister taxon to all remaining Heterotremata ( Fig. 33; instead of being sister to Calappa Weber 1795; [31]) should be taken with caution due to the limited outgroup taxon sampling.

Discussion of the evolution of the P5-swimming crab morphotype
Characterising the P5-swimming crab morphotype Using our analysis, the P5-swimming crab morphotype characterised by Herter [24], Kühl [29], Schäfer [51] and Hartnoll [21] can now be evaluated by checking it for the character states that all P5-swimmer species in our taxon sample share. Character states occurring in all P5-swimming crabs are summarised in Fig. 34 and Table 2 (the character matrix for the entire taxon sampling is included in the Additional file 1). It is interesting to note that P5-swimmers only occurred in Portunoidea. With respect to how merus length (character 44) should be interpreted, Steudel [66] noted that species displaying the P5-swimming leg with a long merus were generally able to perform the same swimming movements as P5-swimmers with a short merus. This ties in with AntCox P5 5th pereiopod anterior coxa muscle, IntPl 6/7 interopleurite between thoracomeres 6 and 7, IntSt interosternite (with number pair indicating thoracomeres between which it is situated), MP median plate findings by Schmidt et al. [52], who discovered that theoretical ranges of motion in P5 articulations in Carupa tenuipes with a long merus were similar to those in P5-swimming crabs with a short merus. In Liocarcinus pusillus (Leach, 1816), a species morphologically similar to Liocarcinus depurator but with a longer merus, swimming movements were principally the same as in L. depurator (although somewhat less effective; see highspeed recordings of Additional file 34: Video S1, Additional file 35: Video S2, Additional file 36: Video S3 in Supplementary Information). Liocarcinus navigator, in which the merus is even longer, also swam in a similar way to typical P5-swimmers, although not as fast or with as much agility (personal observation). In all these species, the P5 anterior coxa muscle originates at the median plate, as is also the case in all typical P5-swimmers with a short merus (character state 34 (2)), corroborating the hypothesis by Hazerli and Richter [22] that this is crucial for P5-swimming. Consequently, we do not consider a short merus mandatory in the identification of a swimming leg, although we are well aware that as a result of the variability of merus length there may be high morphological variation in species assigned to the morphotype, presumably resulting in a high level of variation in the effectiveness of P5-swimming ability. In summary, P5-swimming crabs can unambiguously be identified by the character states shown in Fig. 35, with merus length (character 44) representing a state that "merely" influences the effectivity of P5-swimming. The position of the P5 coxa (subdorsal or ventral, see for example [21]) doubtlessly also influences P5-swimming effectivity, but it was not considered here because of the difficulty of conceptualising discrete character states due to high interspecific variability. We further consider Xaiva biguttata an (albeit unusual) P5-swimmer since it displays the swimming leg with a long merus (but with no propodus lobe) and its P5 anterior coxa muscle originates at the median plate, though unlike in all other P5-swimmers it additionally originates at interopleurite 7/8 (as in, but not homologously to Thia; character state 34(3); Figs. 25C, 35). It should also be mentioned that with merus length, dactylus width (character 47) also varies between species (Figs. 4, 5, 6). In Raymanninus schmitti and Caphyra loevis we even found species that exhibited all the unambiguous P5-swimmer states mentioned above except for a broad paddle-like dactylus. In R. schmitti, however, the origin of the P5 anterior coxa muscle at the median plate needs to be confirmed in other specimens because the musculature was poorly preserved in the only specimen available (Additional file 31).
The morphological features characterising the P5-swimming morphotype which Hazerli and Richter [22] identified included muscular features, features of the axial skeleton and also external features of L. depurator. However, the greater taxon sampling in the current study revealed that many axial skeleton features in particular mentioned by Hazerli and Richter [22] differ between P5-swimmer taxa. These include the shape and configuration of the interosternite 7/8 process (character 20; morpheme 25 in [22]), the presence and extension of the median plate process (characters 9, 10; morpheme property 18 in [22]), and the distance from the medial edge  Character state combinations of each extant species and several ancestral stem species are indicated by different coloured dots, with dot position representing the character and colour (sometimes with symbol) representing the states. Note that symbols above taxa names indicate whether an extant species is assigned to the P5-swimming crab morphotype or considered to be specialised in burying and/or swimming that is different from P5-swimming and that symbols above the nodes of the cladogram indicate whether any reconstructed stem species can be assigned to the P5-swimming crab morphotype of interosternite 7/8 to the median plane (character 17; morpheme propertiy 22 in [22]). However, it should be mentioned that all P5-swimmer taxa with a short merus, at least, have the same character states with regard to the latter two characters. Apart from that, it is conspicuous that all unambiguous P5-swimming crab character states shown in Fig. 35 solely refer to features of the P5, including the origination of the P5 anterior coxa muscle at the median plate.
Ancestral state reconstructions and evolutionary transformations of P5-swimmer character states Figure 35 shows ancestral state reconstructions and transformations of character states that unambiguously characterise P5-swimmers (raw Mesquite data that show transformations of all morphological character states are supplied in the "Appendix" in Additional file 33). The most important finding is that the P5-swimming morphotype (with a long merus) already evolved in the stem species of Portunoidea (Fig. 35). Starting from this condition, evolutionary transformations proceeded into two general directions. One direction represents evolution into a more efficient P5-swimmer with a short merus. Based on our data, this happened several times independently within Portunoidea (at least twice each in Portunidae and Carcinidae, once in Geryonidae). The second direction is characterised by the loss of P5-swimming crab character states to varying degrees. A complete reversal of the swimming leg into a walking leg occurred independently in Chaceon mediterraneus, Carcinus maenas, and Pirimela denticulata (provided that the position of P. denticulata as sister taxon to the clade Coelocarcinus-Xaiva is correct; see "Results and discussion of phylogenetic analysis" section), from ancestors that had already "lost" one or several P5-swimming leg states (Fig. 35). Catoptrus nitidus evolved a P5 walking leg but retained the origin of the anterior coxa at the median plate and interosternite 7/8 (typical of P5-swimmers; character state 34 (2)) from an ancestor that had already "lost" its short merus and broad dactylus (Fig. 35). Generally, within Portunidae, P5-swimmer character states were reversed to varying degrees, but noticeably, in all taxa except for Libystes nitidus (which has a significantly smaller median plate), this anterior coxa muscle configuration is retained. It is interesting to note that a broad paddle-like dactylus (character state 47(1); Fig. 30B) is always associated (= coherent; see for example [48]) with a carpal propodus insertion (character state 45(1) ; Fig. 30B) and a concave proximo-ventral dactylus margin (character state 48(1); Fig. 30B, F), but not vice versa (Fig. 35). Furthermore, if a paddle-like dactylus (character state 47(1); Fig. 30B) evolved back to a slender dactylus (state 47(0); Fig. 30A) and the carpal propodus insertion was retained (state 45(1) ; Fig. 30B), a concave proximo-ventral dactylus margin (state 48(1) ; Fig. 30B, F) is retained, too (which also represents some sort of coherence). This is the case in the common ancestor of both Caphyra species, the common ancestor of C. nitidus and L. nitidus, and in Raymanninus schmitti (Fig. 35). Caphyra rotundifrons, in which the proximo-ventral margins of the P5 propodus and dactylus form a subchela, is an exception.
The site of origin of the P5 anterior coxa muscle is certainly one of the key characters in the evolution of P5-swimming. However, it is difficult to interpret how this site of origin evolved. Our data show that the conditions typical of a P5-swimmer (origin at median plate and interosternite 7/8; state 34(2); Fig. 25D, E, F) form the plesiomorphic state in Portunoidea (Fig. 35), but it is unclear from which state this evolved (Fig. 40 in "Appendix"). Either it evolved from an origin at interosternite 7/8 only (state 34(0); Fig. 25A) or from an origin at both interosternite 7/8 and interopleurite 7/8 (state 34(1); Figs. 25B, 26). In this case, assuming that an alteration in muscle origin took place during the course of evolution by a gradual shift of one (or several) muscle fibre(s) in its proximal attachment site(s), the connection between interosternite 7/8 and the median plate has to be considered in evolutionary scenarios in order to be able to interpret muscle origin transformations. It is well thinkable that during the course of evolution, one or several of the fibres that originally originated at interosternite 7/8 shifted with their origin to the median plate. However, this step in evolution is only imaginable as long as interosternite 7/8 was connected to the median plate (character state 16(0)). Consequently, even if, based on our ancestral state reconstruction, the state of this character 16 is ambiguous for the ground pattern of Heterotremata ( Fig. 41 in "Appendix"), we prefer a scenario in which interosternite 7/8 was connected to the median plate in the heterotrematan stem species (i.e. character state 16(0)). A reversal of the P5 anterior coxa muscle originating at both the median plate and interosternite 7/8 (state 34(2); Fig. 25D, E, F) back to it originating at interosternite 7/8 (state 34(0); Fig. 25A) or at interosternite 7/8 and interopleurite 7/8 (state 34(1); Figs. 25B, 26) in taxa in which a connection between interosternite 7/8 and the median plate was absent is imaginable if a degeneration of the muscle fibres attached to the median plate is assumed. This must have happened in the stem species of the Carcinus-Portumnus-Pirimela-Coelocarcinus-Xaiva clade (Fig. 35). If this was the case, Xaiva biguttata regained a P5 anterior coxa muscle origin at the median plate, but its phylogenetic positon should be treated with caution as it is only based on morphological data.
Until now, few hypotheses have been put forward on the evolution of P5-swimming crabs. Hartnoll [21] and Steudel [66] were the first to formulate explicit hypotheses, but not on the basis of phylogenetic relationships. Hartnoll [21] suggested a transformation series to swimming crabs largely on the basis of one single morphological feature, the dactylus width of pereiopods 2-5. As far as we understand ( [21] was rather imprecise in his statements), Carcinus was considered a "basal" genus, with its P5 podomeres which are just slightly broader than those of P4 being interpreted as a plesiomorphic character state. However, in accordance with previous phylogenies, our combined analysis identified Carcinus maenas as a highly derived taxon within Portunoidea (Figs. 33, 35; [12, 26, 54]). We consider the P5 podomere broadening in C. maenas not to be homologous to that in P5-swimmers and other taxa with a paddle-like dactylus (see also "Characters concerned with the external morphology of pereiopod 5" section). "Polybius" henslowii, in which Hartnoll [21] considered pereiopods 2-4 to be modified for swimming, was considered a more advanced P5-swimmer than those in which only the P5 was modified. It was assumed that "Polybius" henslowii evolved from an ancestor in which only the P5 was modified, which is well possible considering our data (Fig. 35).
Steudel [66] distinguished between several swimming crab morphotypes (or "Konstruktionstypen" as they were termed) denoted after distinct genera, namely the Liocarcinus-type, Polybius-type, Ovalipes-type and Portumnus-type, but she recognised that only the former three types were capable of the typical P5-swimming movements. Steudel [66] also suggested that the Ovalipestype evolved from the Portumnus-type, independently of the Liocarcinus-and Polybius-types, an assumption based mainly on the shape of the sternum (see character state 4(1) herein; Fig. 8B, E). In contrast, our data show that the similar sternum shapes in Ovalipes ocellatus and Portumnus latipes evolved independently of each other (Fig. 42 in "Appendix"). Steudel [66] further suggested that the Liocarcinus-and Polybius-types both independently evolved from a non-swimming morphotype adapted to effective underwater running (our translation, Steudel used the term "Unterwasserrenner") with a straight, broad sternum (character states 3(1) and 4(0) herein) and well-developed walking legs. Subsequent evolution into the Liocarcinus-and/or Polybius-type was simply interpreted as a further "optimization" of underwater running. However, since our analysis recovered the underwater runners Carcinus maenas and Chaceon mediterraneus as derived species within Portunoidea that lost P5-swimmer character states secondarily (Figs. 33,  35), Steudel's assumption is herewith refuted. Spiridonov et al. [63] also briefly discussed the evolution of the P5-swimmer morphotype, taking into consideration the statements in Steudel [66] and a phylogeny deduced from molecular data but no morphological characters. In contrast to our findings, Spiridonov et al. [63] concluded that the P5-swimmer morphotype most probably evolved three times independently in Carcinidae, Geryonidae and Portunidae, respectively.
Spiridonov [64] speculated that a broad paddle-like dactylus adapted to burying represented a preadaptation to P5-swimming. Several of the species we examined are generally considered to be taxa specialised in a burying mode of life, namely Ashtoret lunaris, Corystes cassivelaunus, Calappa granulata, Portumnus latipes and Thia scutellata [13,14,46,51,71], and perhaps also Coelocarcinus foliatus [37]. However, of these species, only A. lunaris, P. latipes and C. foliatus have a paddlelike P5 dactylus (character state 47(1)). A. lunaris probably evolved this state independently of Portunoidea, and P. latipes and C. foliatus are derived portunoids whose specialisation in burying represents an apomorphy which evolved after P5-swimming (Fig. 35). In the light of these data, Spiridonov [64]'s suggestion is rejected here. As already mentioned, Varuna litterata, A. lunaris, and also probably the portunoid P. latipes (perhaps also C. foliatus) are known to be effective swimmers, exhibiting morphological features which facilitate swimming similar to those in P5-swimmers (like the paddle-like dactyli of pereiopods), but also some which are quite different from those of P5-swimming crabs (like the P5 anterior coxa muscle not originating at the median plate). P5-typical swimming movements above the carapace can thus not be performed by these taxa, as was shown by Schmidt et al. ([52]; in C. foliatus, however, this still has to be tested).
With regard to axial skeleton features, an interesting aspect of P5-swimming crab evolution is the "brachyuran sella turcica" (sensu [19]), a sella turcica connected to interosternite 7/8. However, in the species examined herein, this connection was only present in the outgroup taxa Sternodromia monodi, Medorippe lanata, Eriocheir sinensis and V. litterata (character state 19(0); Fig. 14A, B, D, E). In all other taxa, interosternite 7/8 is without a direct connection to the sella turcica (character state 19(1); Figs. 14G, H, 15A-D), instead possessing an interosternal process which sometimes touches interosternite 6/7. This suggests that in the ground pattern of Heterotremata, the direct connection between interosternite 7/8 and the sella turcica (the "brachyuran sella turcica") was secondarily lost. Interestingly, in the taxa examined, the absence of a connection between interosternite 7/8 and the sella turcica was always associated with the presence of a junction plate cavity, which offers space and attachment sites for the large extrinsic P5 musculature of P5-swimmers. We thus interpret this as a preadaptation for the evolution into a P5-swimming crab, and as an explanation of why the P5-swimming crab morphotype evolved in Heterotremata only.

Conclusions
We demonstrate that a detailed morphological examination on the basis of accurate character conceptualisation can be used in combination with genetic data (as total evidence analysis) not only to formulate a robust phylogenetic hypothesis, but also to reconstruct ancestral morphologies and evolutionary transformations. A careful selection of terminal taxa that display the morphological disparity within this group is a prerequisite for convincing statements. Our data suggest that the stem species of Portunoidea already showed the typical P5-swimming crab morphotype, but with a merus that was not as short as in the highly effective P5-swimmer morphotype that is represented by several extant species. The axial skeleton and extrinsic musculature configuration in different P5-swimming crab species can be highly diverse, although all have in common that the extrinsic anterior coxa muscle originates at the median plate. A species which shows this character state in combination with certain external states of the swimming leg (P5) can unambiguously be identified as a P5-swimming crab. The lack of a connection between interosternite 7/8 and the sella turcica is identified as an autapomorphy of Heterotremata and a possible preadaptation to the P5-swimmer morphotype. Earlier hypotheses on the evolution of the morphotype, namely that a paddle-like dactylus in Portunoidea evolved as preadaptation for burying, and that the broader P5 podomeres relative to the P4 podomeres in Carcinus maenas (which is well-known for its lack of P5-swimmer features) were homologuous to the broader P5 podomeres in P5-swimming crabs are rejected on the basis of morphological and phylogenetic data.   37 Variability in the degree of heterochely in Carcinus maenas (dorsal view). A Specimen showing moderate heterochely and heterodonty with right crusher chela being somewhat larger than the left and with molariform denticles, B specimen showing pronounced heterochely and heterodonty with left crusher chela being distinctly larger than the right and with molariform denticles, C specimen showing homochely and homodonty with no molariform denticles