The male sexual apparatus in the order Scorpiones (Arachnida): a comparative study of functional morphology as a tool to define hypotheses of homology

Background Insemination in scorpions is carried out by means of a partly sclerotized structure, the spermatophore, which is composed of two separate halves, the hemispermatophores. In most genera these reproductive structures can be used to differentiate species. However, many taxa such as the genus Euscorpius and the family Diplocentridae lack the morphological diversity observed in the copulatory organs of many other arthropods, rendering them useless for species level taxonomy. Such structural stasis, however, suggests that hemispermatophores have evolved relatively slowly and may thus provide a stronger phylogenetic signal for recognizing supra-generic ranks than previously thought. Based on the postulate that the phenotypic stability observed in some groups is the consequence of functional constraint, the most comprehensive comparative study of the male sexual apparatus to date was conducted for a complete reassessment of the morphology, phylogenetic value and hypotheses of homology of these structures. Results Hemispermatophores, pre- and post-insemination spermatophores, as well as the inherent mechanisms of insemination, were studied across the whole order, allowing the recognition and description of a series of five basic bauplans for the capsular region. For the most part, these patterns appear to be consistent within each major taxonomic group, but several cases of incongruence between spermatophore morphology and taxonomy raises questions about the monophyly of some clades. The Bothriuridae are traditionally regarded as a basal scorpionoid family. However, except for the genus Lisposoma, bothriurid hemispermatophores and spermatophores are morphologically more similar to those of the Chactoidea than to those of scorpionoids. On the other hand, the male copulatory structures of the hormurid clade (Hormiops (Hormurus + Liocheles)) are more akin to those of Diplocentridae and Heteroscorpionidae than to those of other hormurids. Conclusions Spermatophore capsular patterns appears to be congruent with a recent phylogeny of the order Scorpiones based on phylogenomic data that placed Bothriuridae outside of Scorpionoidea and Liocheles outside of Hormuridae, in contradicton with earlier phylogenetic reconstructions based on morphology. This raises questions about the potential use of functionally constrained traits to assess the reliability of contradicting phylogenetic hypotheses and emphasizes the need for a thorough reassessment of the scorpion phylogenetic relationships. Electronic supplementary material The online version of this article (10.1186/s12983-017-0231-z) contains supplementary material, which is available to authorized users.


Morphology of the scorpion reproductive apparatus
In scorpions, sperm is not transferred directly by a sexual organ, but using a spermatophore, an external, partly sclerotized structure containing the spermatozoa, which is produced and deposited on the ground by the male [1]. This spermatophore is composed of two halves, the hemispermatophores, which are secreted by the paraxial organs [2] and joined together when they are expelled from the body during courtship.
The spermatophore is composed of four distinct parts (Fig. 1). Although their shapes and proportions can be extremely variable, each part has the same function in all taxa: (1) the pedicel is used to glue the spermatophore to the substrate, (2) the stem is composed of soft membranes and contains the spermatozoa, (3) the stalk (= flagellum and distal lamina sensu Lamoral [3]) is usually sclerotized and acts as a lever to trigger the compression of the stem in most taxa, and (4) the capsule is located between the stem and the stalk and guides semen from the spermatophore cavity to the female genital tract. This last part, the capsule, is responsible for the sperm transfer per se. It is composed of two sclerotized ridges/ carinae (distal and basal) delimiting a soft membrane, the "sperm duct" (sensu Stockwell [4]). The capsule shows the greatest structural variation in the spermatophore, ranging from simple openings to complex eversible structures.
Although hemispermatophores are usually distinct for closely related species, their morphology can also be extremely uniform in some genera ([4, 43, 44]; Monod, unpublished data) or even in some families [4], conveying little to no taxonomic information at the species level. Such structural stasis suggests that hemispermatophores have evolved relatively slowly at least in some groups and may thus potentially provide a strong phylogenetic signal at higher taxonomic levels.

Hypothesis of homology in hemispermatophores
Hemispermatophores have very rarely been considered in a more global evolutionary perspective because their study is usually purely descriptive. Lamoral [3] was the first to propose structural homologies among the hemispermatophores of Buthidae and Scorpionidae. He considered that the flagellum of buthids was an extension of the capsule and was therefore not homologous to the distal lamina of scorpionids. Stockwell [4] disagreed with this hypothesis, pointing out the very limited sampling of intermediate taxa examined by his predecessor and presented a more congruent hypothesis of homology based on comparison of a wider range of taxonomic groups. His work is thus far the only available study giving a comparative assessment of the morphology of hemispermatophores across the whole order. Several phylogenetic analyses [45][46][47][48][49][50] subsequently used Stockwell's work to define characters and character states of hemispermatophores for their respective matrices, but each of these studies focused on infraordinal taxonomic groups rather than the order as a whole.

Phylogenetic values and functional morphology of Hemispermatophores
Due to the high phenotypic dissimilarity seen in many organisms, the male sexual apparatus is widely used to diagnose species, but is also considered to have evolved too rapidly to be phylogenetically informative [51,52]. However, numerous phylogenetic studies of various arthropod lineages have successfully included male genitalic characters, demonstrating that reproductive structures may also be useful in determining supra-specific phylogenetic relationships [53].
Scorpion hemispermatophores can be divided into different 'subunits' that are arguably subject to different selective pressures and thus show different patterns of evolution and therefore various degrees of phylogenetic value [54][55][56]. While some characters follow the general trend of rapid and divergent evolution driven by strong sexual selection observed in most male genitalia, others show only limited variation among closely related taxa, making them more suitable for the reconstruction of the deeper phylogenetic relationships [56].
Mattoni et al. [56] pointed out that the traits of spermatophores, which are mechanically constrained to perform a particular task, would be less likely to evolve rapidly and randomly than, for instance, features subject to sexual selection. Functional constraint limits the morphological variability, and when a structure is involved in performing a precise mechanistic task, its shape cannot be quickly and radically modified without causing severe functional disruption [57][58][59][60].
Based on this reasoning, slowly evolving mechanically constrained characters are potentially the most informative for reconstructing deeper phylogenetic nodes. It is therefore necessary to consider not only the morphology of hemispermatophores but also their functional aspect in order to correctly assess homology. In this respect, the study of the mechanisms of insemination by comparison of pre-and post-mating spermatophores over a wide range of taxa is of paramount importance in understanding how the different features that compose the male reproductive apparatus relate to each other in distant taxa.

Post-Copulatory Spermatophores in Systematics
Despite their putative usefulness for systematics [54], pre-and post-copulatory spermatophores have been treated relatively marginally in the scorpion literature [1,2,4,, and only few studies present an analysis of functional morphology and the mechanisms involved. To our knowledge, spermatophore morphology has never been compared in detail across distantly related taxa and used in a phylogenetic study in the way that it has been for the order Amblypygi Thorell, 1883 [95,96]. The limited data available on spermatophores is probably a consequence of the difficulty in obtaining these structures, necessitating as it does the availability of living adult animals of both sexes and the observation of successful matings. This gap in our knowledge of the reproductive structures in male scorpions has probably contributed to the difficulty of establishing reliable hypotheses of homology.

Achievements of the present study
We present here the most comprehensive comparative study of the scorpion male reproductive apparatus, not limited to morphology of hemispermatophores but also including comparison of spermatophore morphology, an analysis of the mechanisms of insemination and of the function of the various structures. The largest database to date was gathered by the examination of museum material but also by searching the literature for all published illustrations of hemi-and spermatophores. Moreover, we doubled the data available for pre-and post-copulatory spermatophores and present the first comparative study of these structures over a range of distantly related taxa. This allow us (1) to elaborate a more consensual terminology applicable to the whole order, (2) to identify characters most suitable to provide strong phylogenetic signal at supra-specific phylogenetic levels, (3) to reassess homology hypotheses for hemispermatophores, with an emphasis on the capsule. Based on these new paradigms, we identify and describe five general patterns (bauplans) for the capsular region and map them onto available phylogenetic reconstructions in order to uncover inherent evolutionary trends. The roles of natural and sexual selection in the evolution of scorpion copulatory structures are then evaluated. Finally, conflicts between morphological and molecular phylogenies and the reliabilty of each hypothesis are also discussed based on congruence between these hypotheses and the reproductive morphology.

Acquisition of spermatophores
Live specimens were purchased from the pet trade, donated by colleagues or collected in the field in compliance with the legal requirements of each country. Spermatophores were obtained from breeding specimens during field expeditions or subsequently in the laboratory. Specimens were kept in captivity at temperatures between 25 and 30°C, without seasonal changes. Each specimen was housed individually in a plastic container provided with a mixture of bark mulch and peat as substrate and pieces of cork bark for hiding. Each specimen was fed 2-4 crickets (Acheta domestica (Linnaeus, 1758), Gryllomorpha dalmatina (Ocskay, 1832), Gryllus assimilis (Fabricius, 1775) or Gryllus bimaculatus De Geer, 1773) or 1-2 cockroaches (Blaptica dubia Serville, 1838) every two weeks. Each enclosure was watered once a week or every two weeks depending on the moisture need of each taxon. For mating, a female and a male were placed in a larger plastic container with a thin sheet of cork bark on the bottom. In order to create a humidity gradient the cork bark was moistened on one side with a plant mister. When mating occurred, spermatophores were retrieved from the enclosure immediately after the pair separated and placed into 75% ethanol. In order to obtain pre-insemination spermatophores, the pairs were separated after the male had deposited the spermatophore on the substrate and the unused spermatophore was then placed in 75% ethanol.

Dissection and examination
Spermatophores and hemispermatophores were examined with a ZEISS Stemi SV8 stereomicroscope. Mature male specimens were dissected using microsurgical scissors and forceps for extracting both of their hemispermatophores. Paraxial organ tissue was removed either manually with forceps or chemically with a solution of Proteinase K (concentration: 10 mg/mL; Qiagen, Venlo, The Netherlands). Chemical extraction was performed by immersing hemispermatophores in the solution and then placing them in an oven at 45-50°C for 15 min to an hour depending on the size and sclerotization of the structure. Once the soft tissue of the paraxial organ was sufficiently digested, the hemispermatophores were retrieved from the solution and thoroughly rinsed with water.
Depending on the taxa, some parts of the hemispermatophore capsular region are composed of soft membranes that can be easily damaged during dissection whether chemical or manual. The thinner membranes can be degraded by long exposure to Proteinase K, or torn apart by strong traction of the forceps, especially after prolonged storage in alcohol which can stiffen them. In fresh specimens, these membranes remain flexible and are less likely to be damaged during dissection. Therefore, whenever possible the dissection was carried out immediately after the specimen was euthanized.
However, most of the time fresh specimens were not available. In those cases, the size and state of preservation of the specimen were evaluated in order to determine the best methods of extraction to minimize the risk of damaging the structure. Manual extraction was performed with relatively good success on large taxa, even those preserved for a long time, while chemical extraction was preferred for smaller taxa. On several occasions, several specimens of the same species were dissected in order to obtain a hemispermatophore showing the complete set of informative characters.

Evaluation of insemination mechanism
The position of a spermatophore during mating was inferred by placing the spermatophore on a preserved female of the corresponding species, following the method used by Jacob et al. [90]. For taxa in which the spermatophore remains attached to the female gonopore for some time after copulation, i.e. the genus Chiromachus, and the sub-genera Monodopisthacanthus Lourenço, 2001 andNepabellus Francke, 1974 (see corresponding paragraph for details), the position of the structure could be ascertained in situ directly after the mating occurred.
Assessment of the insemination processes also involved a precise localization of the capsular foramen, the opening through which the semen is expelled from the spermatophore. It is clearly visible under a microscope and always associated with the presence of sperm.

Photographs and illustrations
Line drawings were produced using a drawing tube mounted on the SV8 stereomicroscope. Pencil sketches were subsequently inked and scanned for further processing and editing. High-resolution images were taken with a custom-built stacking system and with a DSLR camera equiped with a VariMag II DSLR microscope adapter system (CNC Supply, cape Coral, FL, U.S.A.) and mounted on a stereomicrocope. Zerene Stacker (Zerene Systems, Richland, WA, U.S.A.) was used to assemble images taken at different focal planes into a single image with greater depth of field. Illustrations and photographs were edited (background removal and contrast adjustment) in Adobe Photoshop CS5 and plates prepared with Adobe illustrator CS5 (both from Adobe systems, San Jose, CA, U.S.A.).
Line drawings are preferred over photographs and SEM to illustrate hemispermatophores for the following reasons: (1) the small size and transparency of the structure sometimes results in photographs uninterpretable in three dimensions, whereas drawing with a drawing tube enables a precise interpretation of the complex shape of the capsule; (2) drying a hemispermatophore for SEM often causes shrivelling or deformation, even when using critical point drying. For spermatophores, photographs were deemed appropriate because the transparency is less problematic than for hemispermatophores (spermatophores are filled with semen), and they are preferred over SEM because of the paucity of material available.
Rather than illustrating all hemispermatophores examined, major patterns were identified and representative taxa were selected. As explained above, it is difficult and time-consuming to obtain spermatophores, and each specimen provides crucial information that can potentially facilitate future work. Illustrations of spermatophores from most of the genera for which the structure was previously unknown are therefore presented, even when they belong to closely related taxa. When a hemispermatophore and a spermatophore were available for a given taxon, a plate representing both structures was prepared in order to allow a direct comparison of the two structures, which we considered necessary to understand the process of capsular eversion.

Scanning electron microscopy
Spermatophores of Hormiops davidovi Fage, 1933 preserved in 75% ethanol were dehydrated in a graded alcohol series, critical point dried in a SPI-DRY critical point dryer (SPI supplies, West Chester, PA, U.S.A.), mounted on standard aluminium stubs (diameter 12.5 mm, height 6 mm; Agar Scientific, Essex, U.K.), and sputter-coated with gold in a Cressington Sputtercoater 108 Auto. The sample was examined with a Zeiss DSM940A SEM.

Terminology of positional aspect of hemispermatophores
Lamoral [3] proposed a terminology to identify the different orientational aspects of the hemispermatophore based on the anatomical position of the structure inside the body prior to dissection. Stockwell [4], on the other hand, used the position of the functional spermatophore in relation to the position of the male during copulation as the reference system. Although most researchers subsequently used Lamoral's terminology to describe hemispermatophores, Stockwell's alternative is more appropriate, as pointed out by Cauwet [97]. Depending on the taxonomic group, hemispermatophore capsules may not have the same orientation inside the male body, and its orientation is, moreover, difficult to assess with accuracy without completely removing the tergites. On the other hand, the orientation of deposited spermatophores is always the same across the order, with the capsule and sperm duct always facing away from the male towards the female gonopore. Furthermore, contrary to Lamoral's hemispermatophore terminology, Stockwell's can be applied to both hemispermatophores and spermatophores. A second terminology for spermatophores is therefore not needed and direct comparison between the two structures is straightforward. The positional terminology of Stockwell [4] adapted by Cauwet [97] ( Fig. 2) is thus used in the present contribution.

Nomenclature of hemispermatophores
The tendency among scorpion taxonomists to use different nomenclatural systems, adapted to the particular group studied but not necessarily to other taxa, has arguably hampered the establishment of a consensual terminology based on reliable interpretations of homology [97]. The terminology used in the present paper is thus adapted from several nomenclatural systems, i.e. Vachon [5], Alexander [62,63], San Martin [7], Lamoral [3], Francke [83], Stockwell [4], Peretti [85], Monod & Volschenk [98], Monod & Lourenço [41] and Monod [49]. Whenever possible, we have adopted the first term proposed. In some cases, however, older terms were considered less appropriate than more recent ones, either because they define the character concerned less accurately, or because a more recent name has gained common acceptance among researchers. Furthermore, several new names are introduced, some to designate newly recognized structures or parts, and others to designate features already described but for which all names available were considered inappropriate. For instance, numerous parts of hemispermatophore capsule have been referred to as lobes. This vague wording does not properly describe either morphology or function and was thus systematically replaced.
All the terms used in this publication and their synonymies with earlier nomenclatural systems are listed below. The synonyms are listed in chronological order with their references and a definition for each new name is provided. The list is divided into three sections corresponding to the main parts of the hemispermatophore (Fig. 1), i.e. the stalk, the capsule and the stem.

Ancestral state reconstructions
The five bauplans identified during the course of this study were optimized onto the currently available phylogenetic trees of the order Scorpiones proposed by Stockwell [4], Coddington et al. [104], Soleglad & Fet [46],/ Soleglad et al. [105], and Sharma et al. [106] under the parsimony criterion with Mesquite version 2.75 [107]. The presence/absence of invagination on the basal edge of capsule was also optimized onto the phylogenies of Stockwell [4], Prendini [45], and Sharma et al. [106] using the same method. Character matrices were produced for each of the cladograms (Additional files 1, 2, 3, 4, 5). The different bauplans seems to reflect a gradual evolutionary transition from a simple plain sperm duct towards more complex evertible capsules by successive foldings. The invagination of the capsular basal edge represents the ultimate step of this process eventally resulting in the establishment of the more complex pattern. For these reasons, both characters were assigned linear transformation series and treated as additive (ordered) [108]. The consistency index and retention index of each character were calculated on each of the cladograms in order to evaluate their levels of homoplasy on the different trees.

Sampling
Hemispermatophores of 122 species (5% of known species) belonging to 71 genera (35% of described genera) and 17 families (85% of known families) were studied (Fig. 3ac), incorporating material from various museum collections [see Additional file 6 for a complete list of specimens]. The relevant literature was thoroughly checked to inventory all published illustrations of hemispermatophores and spermatophores. Illustrations from published taxonomic descriptions and morphological studies were used to complement specimen examination, allowing us to confirm the constancy of defined patterns within taxonomic groups for which only a limited number of taxa had previously been examined. Although we tried to be as exhaustive as possible, some illustrations may have been missed. Photographs or drawings of hemispermatophores and/or spermatophores of 578 species (24% of known species), belonging to 119 genera (58%) and 19  showing the proportion of families for which data were obtained from examination of specimens or from the literature during this study, and the proportion of taxa for which data was missing. b, e Pie charts showing the proportion of genera for which data were obtained from examination of specimens or from the literature during this study, and the proportion of taxa for which data was missing. c, f Pie charts showing the proportion of species for which data were obtained from examination of specimens or from the literature during this study, and the proportion of taxa for which data was missing Akravidae Levy, 2007 which is only known from hollowedout exoskeletons and probably extinct [109,110]) were found in the literature [see Additional file 7 for the complete list and Additional file 8 for references]. The combined dataset (data from examined material and from the literature) represent 619 species (26% of known species), 132 genera (65%) and 19 families (100%) (Fig. 3a-c). The hemispermatophore bauplan for each of these taxa was coded [see Additional file 9 for the complete matrix]. In the case of hemispermatophores for which no specimen was available, the vast majority of published illustrations are good enough to assess similarity with the material examined, and, in most cases, to unambiguously assign hemispermatophores to one of the recognized patterns, even if the capsular region was not accurately depicted.
For some taxa no material or only incomplete hemispermatophores were examined, and in a few case the bauplan remains difficult to determine with certainty based on published illustrations alone, i.e. the genus Lisposoma Lawrence, 1928, and the families Pseudochactidae and Superstitioniidae. Hemispermatphores of these taxa were nonetheless assigned to a recognised pattern, and the reasons for each decision are given (see relevant paragraphs for detail), pending the study of specimens to confirm the proposed hypotheses.
In addition to the hemispermatophores, 119 spermatophores from 36 species (1% of known species) belonging to 14 genera (7%) and 7 families (35%) (Fig. 3d-f ) [See Additional file 10 for the complete material list] were obtained and studied in the course of this study. This structure was previously unknown for 30 of these species, for 9 of these genera, for two of the families and for one of the sub-families [see Additional file 10]. Our dataset does not generally overlap with material from earlier publications and thus represents major progress in our knowledge of scorpion spermatophores, presenting information about more taxa than in all previous publications combined (spermatophores of 34 species belonging to 13 genera and 9 families). The combined data represent 66 species (3% of known species) belonging to 33 genera (16%) and 11 families (55%) (Fig. 3d-f). Moreover, 22 of the spermatophores obtained are from Hormuridae and Scorpionidae, two families for which very little data was previously available in the literature. These spermatophores are among the most complex and they provide essential data for a thorough comparison with the better-known spermatophores of the families Bothriuridae, Buthidae and Euscorpiidae.

Architecture of the capsule
The primary function of spermatophores is to ensure insemination. The mechanisms by which this is achieved involve two sets of characters: (1) characters pertaining to the overall architecture, which are responsible for the ejection of semen from the spermatophore, and (2) characters of the capsule, which ensure the transfer of semen into the female genital tract. In all scorpion taxa, semen expulsion is carried out by the same mechanism, i.e. an increase of pressure in the internal cavity of the spermatophore induced by a bending of the whole structure [2,63,64,84,90]. As a result of this conserved mechanism, the general groundplan of the spermatophore with four distinct parts remains the same across the order despite a considerable morphological diversity (Fig. 1).
The capsule, that ensures the insemination per se, shows an even greater structural variation, but its architecture follows a linear evolutionary pathway (see discussion) probably due to underlying mechanical constraints. As for the overall groundplan, while proportions and sizes of the various capsular features may differ significantly, the variation of the structural pattern is actually quite limited, mainly consisting of a gradual complexification of the invaginations and foldings of the sperm duct. Five basic bauplans (Fig. 4), accounting for the structural changes of the capsule observed in the examined material, were identified. Each of these bauplans roughly corresponds to an additional fold of the sperm duct. These are described below from the simplest to the most complicated architecture. The term 'bauplan' is used here to designate an overall architectural pattern or organization invariant among a wide range of taxa [60].
In most illustrations of the capsule presented, the different parts are color-coded as follows: (1) the distal carina is in green, (2) the basal carina in blue, (3) the sperm duct is given in yellow in the first bauplan, (4) the tectum is in yellow and the subex in orange in the following bauplans, and (5) the terminal membrane is in pink. Corresponding parts in hemispermatophores and spermatophores, and homologous parts in hemispermatophores/spermatophores of different taxa are thus easily identifiable.

I. 'No-fold' bauplan
This is the simplest pattern observed, with a non-folded sperm duct (Fig. 4a). This is present in the families Buthidae, Chaerilidae (Figs. 5-6), and putatively also in the Pseudochactidae.
The monogeneric family Chaerilidae possess hemispermatophores with the simplest capsule, consisting of two carinae surrounding an unfolded sperm duct ([4]; Fig. 5a-c yellow). Although there is no visible foramen, the semen can nevertheless be expelled because the membranes of the two hemispermatophore are not fixed together in the spermatophore between the capsular ridges (arrow in Fig. 5d). The same basic "no-fold" pattern is observed in the family Buthidae ( [3,4]; Fig. 6).
However, in contrast to chaerilid hemispermatophores, a foramen (cf in Fig. 6c and g) is clearly visible in buthid hemispermatophores, because the duct membrane is shifted away from the longitudinal axis as a result of wider capsular carinae. In some other buthid taxa, like observed here in Hottentotta jayakari (Pocock, 1895), the duct membrane even extends above the basal carina (sdm in Fig. 6g-h yellow). The insemination process is similar in the Chaerilidae (this publication) and Buthidae [64, 65, 68-72, 75, 78, 80-84, 88], without alteration of the structure of the capsule. Furthermore, unlike other spermatophores, buthid and chaerilid spermatophores do not have a fixed truncal flexure; the stem is simply bent backward during mating, medially in chaerilids (dotted arrow in Fig. 5d) and more basally in the elongated buthid spermatophores.
According to illustrations in Prendini et al. [42], the hemispermatophore of Pseudochactidae appears to be morphologically intermediate between those of buthids and chaerilids. The stem is short as in chaerilids, and the stalk is flagelliform but much thicker than that observed in buthids. The capsule is not shown in detail in Prendini et al. [42], therefore the bauplan could not be determined and potential similarities with buthid and chaerilid capsules could not be assessed accurately. An apophysis at the base the stalk was interpreted by Prendini et al. [42] to be homologous with the basal hook of the basal carinae of the buthid hemispermatophores (bh in Fig. 6), but this cannot be unambiguously confirmed at the moment. Therefore, solely based on gross morphology, pseudochactid hemispermatophores are putatively considered here to possess the same bauplan as buthids and chaerilids, although this needs to be verified.

II. 'One-fold' bauplan
The sperm duct membrane folds itself longitudinally, forming two perpendicular surfaces (Fig. 4b) that are designated here according to their positions: the basal surface is named subex (Fig. 4b orange) and the more distal plane, tectum (Fig. 4b yellow). The sperm duct of the hemispermatophore capsule in the family Iuridae (Fig. 7) show significant modifications of the capsule proportions compared to the pattern observed in buthids/chaerilids/pseudochactids. The capsule is extended anteriorly, protruding conspicuously. This membranous process is mostly composed of the subex (Fig. 7 orange) and is terminated by a thin membrane (tm in Fig. 7 pink) where the foramen is usually located. A broad apophysis is located at the base of the basal carina (bh in Fig. 7a) and may be homologous with the basal hook of Buthidae. Hemispermatophore morphology is extremely conservative in the Iuridae, with only slight differences between species and even genera [29,[111][112][113][114][115][116].
The capsular region of hemispermatophores in the Typhlochactidae (Fig. 7e-i; [34,117]) is strikingly similar to that of iurids: the capsule is markedly elongated anteriorly into a membranous process. Moreover, the transverse ridge is curved towards and very close to the lateral hook (tr in Fig. 7i). For these reasons typhlochactid hemispermatophores are assigned to the 'one-fold' bauplan. Hemispermatophores of the family Superstitioniidae are also putatively assigned to the 'one-fold' bauplan on the basis of illustrations from Francke [34]. As first mentioned by Francke & Soleglad [29], superstitionid hemispermatophores possess an extended, weakly sclerotized sperm duct similar to that of iurids.
A deep longitudinal folding clearly divides the sperm duct into two surfaces, subex and tectum (  (Fig. 8a-d orange). The terminal membrane observed in Iurus is also present in these taxa but remains small (tm in Fig. 8 pink). The hemispermatophores of Caraboctonus and Hadruroides were first illustrated in Francke & Soleglad [29] and Stockwell [4]. Ochoa & Prendini [118] later presented a revision of the Peruvian Hadruroides species, with illustrations of spermatophores for six of these taxa showing that the structure is extremely conservative within the genus.
The 'one-fold' bauplan is the only one for which the insemination mechanisms are still unclear. Although the spermatophores of Iuridae and Caraboctoninae Kraepelin, 1905 remain unknown, the post-insemination spermatophore of Superstitionia donensis Stahnke, 1940 was illustrated in Francke [83]. It shows the absence of protruding features from the capsular region, suggesting  (2) there is probably no intromission of this membrane into the female genital tract because it is too soft. The terminal membrane may nonetheless limit semen loss during mating by covering the sides of the gonopore. Obtaining spermatophores is necessary to test these hypotheses.

III. 'Two-folds' bauplan
Here, the tectum is folded longitudinally twice, in opposite directions (Fig. 4c), which results in the complete invagination of the capsule into the interior of the spermatophore. In hemispermatophores (i.e. non-everted state of spermatophore halves), the subex and tectum sit parallel to each other in accordion-pleats (Figs. 9b-c, significantly more pronounced, leading to a larger intromittent structure when everted (Fig. 13d-k) Fig. 12a-d) is radically modified in comparison with capsules observed in other Chactoidea. The sperm duct appears to be elongated and twisted inward, with the subex (Fig. 12a-d orange) forming a rather deep invagination (is in Fig. 12d) Fig. 12e-g) present a morphology intermediate between the chactid genera mentioned in the previous paragraph and other Chactoidea, i.e. the tectum is extended and twisted inward as in the brotheins, but the rest of the capsule remains similar to the basic chactoid morphology. Furthermore, hemispermatophores of Chactopsis, Chactopsoides and Megachactops share some morphological similarities with scorpiopids. Unlike Brotheas, Broteochactas, Hadrurochactas, Neochactas and Teuthraustes, they possess an extended terminal membrane covered with spicules, and their subex is not completely folded over itself [30].
A terminal membrane is present in all taxa examined. It is covered with minute spicules in Chactopsis [30], and Scorpiopidae ( [4,121]; tm in Fig. 12g pink). When the eversion of the capsule is triggered, the terminal membrane forms a sort of externally inflated membranous pouch [4,90], referred to here as the physema (ph in Fig. 13d-f pink and g-k), which helps to widen the female genital atrium and to maintain a tight fit between spermatophore and female genital tract, preventing sperm backflow [90]. A clump of spinules on the distal part of the basal carina, the crown-like structures (sensu Jacob et al. [90]), was first reported in Euscorpius Thorell, 1876 ( [4,90]; cls in Fig. 13) and Megacormus (Karsch, 1881) [4,119,120]. This feature is also present in Chactas Gervais, 1844 (cls in Fig. 11e) and Scorpiopidae (cls in Fig. 12f ), but in these taxa, it is more reduced and basally extended by a membranous crest that runs along the basal carina (bcr in Figs. 11d-e and 12e-f).
Hemispermatophore capsules of Bothriuridae (Figs 14c and d, 15,16) presumably also belong to the 'two-folds' type. Bothriuridae is probably the scorpion family for which hemispermatophores, spermatophores and insemination mechanisms are best known as a result of the extensive work of several generations of Latin American researchers [4, 8-22, 55, 56, 66, 74, 76, 77, 79, 85-87, 89, 92, 122-151]. In this family the capsule folding pattern is quite similar to that of the Chactoidea. The sperm duct is bent on two planes, i.e. subex and tectum (Figs. 15 and 16 orange and yellow respectively), and possess a rather large terminal membrane (Figs. 15 and 16 pink). In everted spermatophores this membrane forms an inflated vesicle, the physema, from which the semen is expelled ( [55,66,74,76,77,79,[85][86][87]89]; Figs. 13c-d, 15e and j pink). A pair of long sclerotized processes, referred to as claspers (cl in Figs. 14c-d, 15b-c, g-h, 16b-c), is present on the basal part of the capsule. We consider these to be modifications of the basal carinae. Like the chactid crown-like structures, they are extended anteriorly by wide membranous crests that merge with the antero-basal edge of the sperm duct (bcr in Figs. 15c and h, 16b-c), forming a capsular concavity sensu Peretti [85] (cc in Figs. 15c and h, 16b-c) basally.
We requested a loan for the paralectotype of the African bothriurid Lisposoma elegans Lawrence, 1928 (SAMC B6077) dissected by Lamoral [152] and for which the right hemispermatophore was relatively well illustrated in Lamoral [3] and Prendini [152]. The left hemispermatophore appears to be the only remaining part of this material; the staff could not locate the right half. Its examination revealed that the capsule has been completely torn off, preventing analysis of the capsular structure. Although the published illustrations [3; 152] are not sufficient to determine the bauplan unambiguously, they nevertheless allow the identification of several important differences between the hemispermatophore of Lisposoma and those of non-African bothriurids and suggest that the former is morphologically closer to hemispermatophores of non-bothriurid scorpionoids.
Stockwell [4] and Prendini [45,152,153] recorded the presence of a "semi-lunar shelf" or "internal cresentic shelf" on the internal wall of the sperm duct invagination in Lisposoma (arrow 1 in Fig. 16d-f yellow,), a character they considered synapomorphic for the Bothriuridae. This structure was referred to by previous authors as 'Lóbulo interno' (= internal lobe) and corresponds to the tectum. Its occurrence however, is not restricted to the capsular pattern of Bothriuridae; it is present in all hemispermatophores with a two-folds, three-folds and four-folds capsule, but does not protrude anteriorly in all taxa, making it less conspicuous.
The distal lamina of the Lisposoma hemispermatophore does not bear the typical latero-distal bothriurid crest [3,4,152,153] and the capsule also seems to be different from the bothriurid pattern. There is no basal capsular concavity and basal crest (arrow 2 in Fig. 16d). The capsular foramen at the base of the tectum is also absent (arrow 3 in Fig. 16e) and the subex is slightly more invaginated than in other bothriurids (arrow 4 in Fig. 16e-f orange). These characters suggest that the sperm duct fold in Lisposoma is homologuous to the hemisolenos of non-bothriurid scorpionoids rather than to the bothriurid claspers.

IV. 'Three-folds' bauplan
An additional folding of the sperm duct is observed in this structural type, translating morphologically into a pronounced invagination of the subex (Figs 4d, 17b-c, 18b-c and e-f, 19b-c orange). Three different groups are recognized within this bauplan depending on the presence/absence of accessory appendages, i.e. mating plug and terminal membrane. The genus Uroctonus Thorell, 1876 has very peculiar hemispermatophores with characters specific to Chactoidea and Vaejovoidea. In addition to the deeply invaginated subex (Fig. 17b-c orange), the capsule possesses a rather large terminal membrane (Fig. 17b-c pink) that forms a physema in everted spermatophores ( [4]; Figs. 14b, 17d pink), as in the euscorpiins (Fig. 13d-k). The basal carina only bears a small process, which is deemed homologous to the crown-like structure of the euscorpiids (cls in Fig. 17b-d).

V. 'Four-folds' bauplan
This structural pattern (Fig. 4e-f ), characterised by a sperm duct presenting four major folds is only observed in the hemispermatophores of non-bothriurid Scorpionoidea Latreille, 1802 (Figs. 14e-h, 20,21,22,23,24,25,26,27,28,29,30,31,32,33). In addition to the three folds in the previous bauplan, a medial invagination of the basal part of the capsule opposite the deep subex (arrow 1 in Fig. 4e-f ) leads to the narrowing of the basal edge into a pipe-like structure (arrow 2 in Fig. 4e-f).  [41] which is renamed here hemisolenos. In the spermatophore, the hemisolenos of the hemispermatophores are fused together to form a pipe-like structure referred to as the holosolenos (hos in Figs. 14e-h, 21f-i, 23f-i, 25, 26f-i, 27f-j, 28d-g, 29f-i, 30e-h, 31d-g, 32d, and f-g). Like the chactoid physema and the vaejovid mating plug, the holosolenos is directly responsible for sperm transfer and insemination: it is an evertible intromittent appendage through which semen is expelled into the female genital tract (hos in Fig. 14e-h). Furthermore, in the hormurid Iomachus politus Pocock, 1896 and in the sub-genus Opisthacanthus Peters, 1861, the holosolenos bears accessory hooks and apophyses (ah and aa in Fig. 27) that prevent it from slipping out of the female genital tract once inserted. The holosolenos detaches from the spermatophore at the end of a succesfull mating (Fig. 27i) and remains in the female genital tract.
The spermatophore capsule probably also acts as a detachable mating plug in the families Hemiscorpiidae and Urodacidae, as suggested by the morphology of their respective hemispermatophores. In hemiscorpiids, the capsule pattern is very similar to that of Iomachus politus and of the subgenus Opisthacanthus, with hemisolenos bearing an accessory hook and apophysis [41], but a bifid laminar hook is also present as in Hadogenes Kraepelin, 1874 [41]. In urodacid hemispermatophores ( the propensity of the hemisolenos to detach if the dissection is not performed carefully, and the presence of an accessory hook in some species (ah in Fig. 24b) suggest that the holosolenos probably also acts as a mating plug in this family, as already mentioned by Stockwell [4]. In many scorpionoid taxa the anterior extremity of the basal carina is also modified into an apophysis (cl in Figs. 20, 21, 22a-d, 24,26,27,28,30,31,33) that can have a hook-like shape (cl in Figs. 24,26,27,28,30,31,33). This feature, previously referred to as the distal lobe [41], is considered here as homologous to the bothriurid claspers (see discussion) and is thus called the same.
In addition to the development of accessory processes (clasper and hemisolenos), the overall shape of the basal carina is also modified in non-bothriurid scorpionoids. In all these taxa, except for the Diplocentridae, the basal edge of the capsule is invaginated between clasper and hemisolenos. This fold (= anterior edge of basal lobe sensu Monod [49], arrow 1 in Fig. 4e) can be very shallow as in Heteroscorpion Birula, 1903 (white arrow in Fig. 20c), Hormiops Fage, 1933 ( [94,182]; white arrow in Fig. 21c, and arrow 2 in Fig. 25a-d), Hormurus Thorell, 1876 [183] and Liocheles Sundevall, 1833 [181], but in other scorpionoids the invagination is much deeper (white arrow in Figs. 24c, f, 26c, 27c, 28c, 29c, 30d, 31c, 32c, 33c and arrow 2 in 25f-g). The depth of this additional invagination has major repercussions on how the capsule is everted. If the invagination is absent, as in Diplocentridae (Figs. 22g-h, 23c), or weak as in Hormiops (white arrow in Fig. 21c and arrow 2 in Fig. 25a-b), Hormurus and Liocheles (white arrow in Fig. 22b), the claspers and holosolenos form a solid block, and thus capsule eversion triggers their combined forward rotation (Fig. 25c-d). In post insemination spermatophores (Figs. 21f-i, 23f-i) the holosolenos is thus reoriented anteriorly and sits approximately parallel to the stalk. Given the shallowness of their basal capsular invagination (white arrow in Fig. 20c), Heteroscorpion spermatophores presumably follow the same unfolding process. The deeper carinal invagination observed in other scorpionoid taxa (white arrow in Figs. 24c, f, 26c, 27c,  28c, 29c, 30d, 31c, 32c, 33c) translates into a more balanced architecture with deep folds on both sides of the holosolenos (Fig. 25g). In simple terms, the structure is comparable to a double origami sink fold. The carinal invagination creates a weak spot on the basal carina between clasper and hemisolenos. This point of least resistance acts as an additional hinge when the sperm duct membrane is unfolded during capsular eversion. As a result, the holosolenos does not rotate together with the claspers but is pulled out with only a slight rotation (Fig. 25h-i). The additional fold also translates into an increase of the outward extension of the sperm duct. Post-insemination spermatophores then show a holosolenos sitting more perpendicularly relative to the axis of the stalk (hos in Figs. 26f-h, 27f-j, 28d-g, 29f-i, 30e-h, 31d-g, 32d, f-g), and a sperm duct that is more elongated. This last point is especially obvious in Chiromachus ochropus (Koch, 1837) ( Fig. 31d-g), Opisthacanthus (Nepabellus) cf. asper (Peters, 1861) ( Fig. 30e-h) and Monodopisthacanthus spp. (Fig. 32d-g).
In Chiromachus ochropus, Monodopisthacanthus and Nepabellus the spermatophore remains attached to the female genital tract after insemination, sometimes for several hours. The distal lamina in all these taxa bears a pronounced antero-distal crest ('lateral crest' sensu Lamoral [3]) (adc in Figs. 29,30,31,32). This crest appears to be adhesive and will stick to the female coxapophyses II (Fig. 32e), holding the spermatophore in place and the sperm duct inside the female genital tract, probably to ensure that insemination is carried out properly and that the female does not reject semen. Although the courtship of Palaeocheloctonus Lourenço, 1996 and Indian Iomachus Pocock, 1893 was not observed, a similar antero-distal crest is present on hemispermatophores of these taxa (adc in Fig. 33a) and, given their close relationship with C. ochropus and Monodopisthacanthus [49,50], probably fulfils the same function. In Opisthacanthus (Nepabellus) validus Thorell, 1876 the male detaches the spermatophore from the female with its anterior legs after insemination (Monod, four unpublished observations).
Hemispermatophores of Indian Iomachus possess a capsule more complex than in other scorpionoids (Fig. 33b-c). In addition to the invagination between clasper and hemisolenos, the anterior part of the carina that bears the clasper also extends inward around the hemisolenos, thus creating an additional inner fold between the hemisolenos and the subex (icf in Fig. 33c). Although spermatophores remain unknown for these Indian taxa, the inner carinal fold (icf) is probably also unfurled during capsular eversion, as indicated by the presence of small spinules (sp in Fig. 33c) on its inner wall. Like other features present on the sperm duct, these spinules probably help to create a seal between the spermatophore and the female genital tract, and thus must be in contact with the walls of the latter during insemination. Because they are situated on the side of the fold concomitant with the hemisolenos, the usual unfolding observed in scorpionoids is not sufficient to expose the spicules and an unfolding of the inner carinal fold is thus necessary. It is also important to point out that the hoplurus-group of the genus Urodacus [23] possesses hemispermaphores with a strongly modified morphology (Fig. 24d-f) in comparison to the rest of the non-bothriurid scorpionoids. In hemispermatophores of this species group the tectum is flattened against the distal carina and forms a very slender process. The hemisolenos is compressed in a similar fashion, and the claspers are much bigger than in other genera, sometimes having serrated edges. Despite these pronounced differences in the proportions of the capsular elements, the bauplan is the same.

Ancestral state reconstructions of the Bauplans
Four phylogenies of the order Scorpiones are currently available. Three are based on morphological data [4, 46,  and one on phylogenomic data [106]. In order to assess the evolution sequence of the male reproductive apparatus, character optimization of the bauplans was conducted on these four cladograms. In all of these hypotheses, the 'four-folds' bauplan is evolved only once; the 'no-fold' bauplan is recovered as plesiomorphic for the order Scorpiones, and the 'one-fold' bauplan as plesiomorphic for the parvorder Iurida Solegad & Fet, 2003 (Fig. 34).
In phylogenies by Stockwell [4] (Fig. 34a) and Coddington et al. [104] (Fig. 34b), the 'one-fold' and 'three-folds' bauplans each evolve only once, respectively from a 'no-fold' and 'one-fold' ancestor. On the other hand, the 'two-folds' bauplan is evolved four times in Stockwell's [4] (three times from a 'one-fold' ancestor and one reversal from a 'four-folds' ancestor) and three times in Coddington et al.'s [104] (twice from a 'one-fold' ancestor and one reversal from a 'four-folds' ancestor). Because of the lower number of independent emergences, the bauplan obtains better consistency and retention indexes on Coddington (Fig. 34c), the 'two-folds' bauplan is evolved three times (once from the 'one-fold' ancestor, one reversal from a 'three-folds' ancestor and one reversal from a 'four-folds' ancestor), while the 'three-folds' bauplan is evolved twice (once from a 'one-fold' ancestor and one reversal from a 'two-folds' ancestor). Contrary to Stockwell's [4] [104].
In the cladogram by Sharma et al. [106] (Fig. 34d), the evolution from the 'no-fold' to the 'four-folds' bauplan is incremental, each pattern evolving from the precedent in a linear fashion. The numbers of parallel evolution and reversal events is lower than in the preceding trees: the 'three-folds' bauplan is evolved twice independently from 'two-folds' ancestors, and there is one reversal from 'two-folds' to 'one-fold' pattern. The consistency and retention indexes of the bauplan (CI = 0.66666667 / RI = 0.92307692) are thus higher than in preceding trees.

Ancestral state reconstructions of the invagination of the basal edge of the capsule
In the three phylogenies (Fig. 35), a weak or absent invagination of the basal capsular edge is retrieved as plesiomorphic.
In the phylogeny by Sharma et al. [106] (Fig. 35c), the deep invagination evolves only once with a subsequent reversal to the plesiomorphic state in Diplocentridae.

Reliability of data gathered from published illustrations
A large part of the present dataset is based on the interpretation of illustrations from published taxonomic descriptions and one can question the reliability of these data, especially given the difficulty of accurately depicting the tridimentional shape of the capsule region in photographs or drawings.
The strong conservatism of in-toto shapes generally observed within taxonomic groups enables easy classification of hemispermatophores according to their gross morphology. There is no reason to believe that hemispermatophores from closely related taxa with a similar overall shape have different capsules. Therefore, based on comparison with known general morphologies, published illustrations of hemispermatophores generally allow accurate determination of the bauplan even if the capsule is not properly depicted.
For only a few taxa, i.e. the genus Lisposoma, the families Pseudochactidae and Superstitioniidae, were the illustrations insufficient to assign a bauplan with certainty. These few ambiguous cases, however, do not weaken the overall results and conceptual framework of the present study; the five structural patterns proposed here were unambiguously identified in a sufficiently large range of taxa to confirm their validity. It nevertheless remains possible that the few hemispermatophores mentioned above represent further bauplans to be added to the ones characterised here.

Intromittent structures used for insemination
The comparative study of pre-and post-insemination spermatophores allowed accurate determination of the structure performing the insemination per se for each taxon studied. It appears that distantly related taxonomic groups use different intromittent features of the capsule to guide the semen into the female gonopore. Based on their position on the capsule, it was then possible to reassess homology between these different traits.
Across the order, insemination is carried out by three different intromittent structures on the spermatophore: the chactoid physema, the vaejovid mating plug and the non-bothriurid scorpionoid holosolenos. Only the nonbothriurid scorpionoid holosolenos and the vaejovid mating plug are here considered to be homologous because they are sclerotized parts of the basal capsular carina. Moreover, the fact that the holosolenos acts as a detachable mating plug in several non-bothriurid scorpionoid taxa (I. politus, the sub-genus Opisthacanthus Peters, 1861 and probably also in the genera Hemiscorpius and Urodacus), gives more support to this hypothesis. On the other hand, the physema of chactoids is neither considered homologous to the holosolenos and nor to the mating plug, because it is not derived from the basal carina but is formed by an enlargement of the terminal membrane of the sperm duct.

Claspers and holosolenos
A conspicuous eversible sperm duct that protrudes from the capsule and guides semen into the female genital tract is traditionally considered synapomorphic for the spermatophores of Scorpionoidea [4,45]. However, analysis of the insemination mechanism suggests that the protruding capsular feature observed in Bothriuridae is not the same structure as in the other taxa of the superfamily. Bothriurid claspers were previously considered to be homologous to the holosolenos of non-bothriurid Scorpionoidea [4]. This, however, is here refuted by the comparison of used spermatophores and the insemination processes in the two groups. Claspers and holosolenos appear to have different functions. The holosolenos forms a channel for semen expulsion in scorpionoids (hos in Fig. 14e-h), whereas the paired bothriudid claspers do not contribute directly to sperm transfer but rather act as oversized hooks that help widen the female genital atrium and provide secure anchoring for the spermatophore, thus preventing sperm backflow ( [55,[85][86][87]; cl in Fig. 14c-d). In bothriurids insemination is carried out by the physema, as is the case in Chactoidea ( [55,[85][86][87]; ph in Fig. 14a-b).
Jacob et al. [90] had previously suggested that the bothriurid claspers are homologous to the crown-like structures observed in some chactoid spermatophores, because they are in a similar position beneath the capsular foramen and have a similar function, i.e. widening of the female genital atrium and helping to anchor the spermatophore. The presence of membranous crests extending basally from the crown-like structures in chactoids (bcr in Figs 11d-e, 12e-f ) or from the claspers in bothriurid (bcr in Figs. 15c and h, 16b-c)   corroborate this interpretation. In non-bothriurid scorpionoids a pair of capsular apophyses (= distal lobe sensu Monod & Lourenço [41], see next paragraph) assumes the same position (cl in Fig. 14e-h) and presumably also the same function as the bothriurid claspers during insemination (cl in Fig. 14c-d). For this reason these scorpionoid structures are deemed homologous to the claspers, and are also referred to as claspers in the present contribution.

Capsular and laminar hooks
Many hemispermatophores possess strong hooks on either or both of the capsular carinae and/or on the basal half of the stalk. These prongs have many different shapes and proportions across the order, but they can easily be divided into two different groups depending on their position on the capsule, i.e. the basal hook (located below the capsular foramen) and the laminar hook (located above the capsular foramen), which are here considered as non-homologous.
Although not homologous, basal and laminar hooks have the same function, i.e. to pry open the female genital operculum and atrium [29,55,63,90]. Alexander [64] suggested that bending of the buthid spermatophore is triggered by the hooking of the basal prongs to the inner edges of the female pectines. However, the size of the spermatophore and the position of the basal hooks contradict this hypothesis; if the prongs were hooked to the pectines, the capsular foramen of the spermatophore would be situated between the pectines, far away from the female gonopore, just above the basal plate, where semen would be expelled, failing to inseminate the female. The basal hooks of buthids are more likely to be jammed against the bipartite operculum, ensuring that the spermatophore foramen is in contact with the female gonopore.
"Safe sperm transfer", "revealing obstacles", "female choice by mechanical fit" and the coevolution of female genitalia and spermatophore The respective importance of natural and sexual selection in the evolution and diversification of genitalia are still debated in most animals [185]. It is usually accepted  ). Hemispermatophore, contra-lateral aspect (a, f). Spermatophore, lateral aspect (e, j). Evertion of the sperm duct, detail of the mechanism (b-d, g-i). Abbreviation: hos (holosolenos). The numbered circles with arrow show the invagination of the subex (1) and the fold of the basal carina between clasper and hemi/holosolenos (2). The circles represent the axes of rotation and the dots are fixed points along which membranes are unfolding that the morphological evolution of scorpion copulatory male structures is the result of complex synergetic interactions between a variety of selective pressures [54]. Among those hypotheses, the "safe sperm transfer" [54,90,186,187] is particularly important to understand the evolution of the capsular general architecture. This model suggests that natural selection guides spermatophore evolution, favouring morphologies that would provide a tighter copulatory lock with the female genitalia in order to improve sperm transfer. Data collected during this study suggest that evolution of the spermatophore capsule tends towards structures that arguably provide a better anchor to the female gonopore in agreement with this hypothesis. However, it seems that sexual selection was also of paramount importance in the emergence of more efficient capsules. The female may actually exert passive selection by her copulatory behaviour and genital morphology, thus channeling the evolution of the male copulatory structures. While the hemispermatophore capsule of more basal taxa such as chaerilids and iurids [46,104,106] forms a simple non-intromittent opening (Figs 5 and 6), successive foldings and invaginations of the sperm duct membrane has led to the establishment of increasingly long, eversible intromittent structures in the more derived groups such as scorpionoids [46,104,106] (Figs. 20,21,22,23,24,25,26,27,28,29,30,31,32,33). A similar evolutionary sequence from a non-intromittent to an increasingly complex intromittent reproductive apparatus is observed in opilionids, where primitive groups employ non-intromittent spermatophores whereas more derived groups possess intromittent penises [188]. Natural and sexual selection are both regarded here as major driving forces in the evolution of scorpion hemispermatophores. The "safe sperm transfer" hypothesis [90,186,187] accounts in part for the gradual complexification of the sperm duct folds; natural selection might be expected to drive the evolutionary tendency toward longer eversible ducts likely to improve sperm transfer success by minimizing loss. However, sexual selection in the form of "revealing obstacles" [189] and "female choice by mechanical fit" [190] probably also played a prominent role in the evolution of the morphological diversity of the capsule.
"Revealing obstacles" are female adaptations, usually behavioural, that increase "resistance" toward mating males but are independent of male phenotypes. They constrain all males to perform a difficult task in order to be able to mate successfully regardless of female preferences, because the female behaviour is always the same towards all males. The fitness of each male is thus indirectly assessed by overcoming the female resistance and demonstrating its physical and behavioural qualities, or the lack thereof.
In order to mate successfully, the male scorpion, after having deposited the spermatophore on the substrate, needs to pull the female over it to trigger its flexion and thus the sperm transfer. The female does not usually move over the spermatophore on her own accord, even when receptive to the male, but rather sits still and lets herself be guided by the male. In the laboratory several mating attempts failed because of the small size of the males, which were unable to coerce their larger mate over the spermatophore (Monod & Cauwet, unpublished data). Moreover, females are also known to sometimes disrupt courtship by moving away from the axis of the deposited spermatophore either voluntarily or not [87]. Therefore selective pressures are likely to favour spermatophores with features that enable a steady fit onto the operculum and that prevent the female from interrupting insemination [54,55].
In derived spermatophores eversion of the capsule precedes the expulsion of the spermatozoa [85], suggesting that interactions between spermatophore and female genitalia are primordial to ensure successful insemination. The morphological evolution of the male copulatory structures is probably tightly correlated to that of the female. In the basal taxa successful insemination only seems to be enabled by the various capsular hooks, whereas eversible sperm ducts, which became longer over evolutionary time, appear to be necessary in more derived taxa. It is postulated here that more complex spermatophore structures emerged at least in part in response to evolution of the female genital morphology. Eberhard [190] proposed that females could influence the evolution of male reproductive structures by passive mechanical discrimination. According to this model, the scorpion females may passively favour those males with spermatophores which provide the strongest copulatory grip to their genitalia and are less likely to be dislodged during copulation. A tight mechanical connection between spermatophore and female genitalia is necessary to ensure sperm transfer and the maintenance of such a copulatory fit requires coevolution of the two structures [191]. The constant adaptation of the spermatophore capsular design to the morphological changes of the female genitalia (for instance the transition from a bipartite towards a fused operculum) is thus needed to maintain a functional interface between the two structures. Evolution acts here to preserve rather than improve safe sperm transfer. Each bauplan is arguably only suited to a certain type of female genitalia, like a lock-key system. The more derived spermatophore architectures, with long eversible holosolenos, are thus not necessarily adapted to every female genitalic pattern. This translates into the persistence of intermediate 'suboptimal' forms of capsules over time, as is the case in the evolution of other complex structures [192]. The most elongated ducts, i. e. the holosolenos, are only known from taxa where female opercula are partly or completely fused (non-bothriurid Scorpionoidea), whereas spermatophores with simpler or shorter ducts are prevalent in taxa where females possess a bipartite operculum. Furthermore, in taxa where the operculum is divided but joined by a membrane, as in the Chactinae, Bothriuridae, Vaejovidae, Scorpiopidae, Uroctonus and Uroctonites, the intromittent part of the spermatophore appears to be longer than in taxa with completely disjoint opercular sclerites. This suggests that interactions between the groove formed by a divided operculum and the capsular hooks in the basal groups are sufficient to prevent the spermatophore from moving sideways during insemination, whereas a mono-sclerite operculum does not permit an efficient lateral blockage of the structure, hence the need for the development of additional features (holosolenos, mating plugs and adhesive laminar crests) to secure the anchoring of the spermatophore to the female gonopore.

Functional constraints and phylogenetic value
Traits directly involved in biological mechanisms are less likely to evolve quickly and randomly because functional constraints limit their morphological variability [57][58][59][60]. When a structure is integrated into a biological mechanism, its shape cannot be quickly and radically modified without causing severe functional disruption. Therefore the significant morphological variability of spermatophores is usually limited to proportions and shapes, whereas each bauplan described in the present contribution is conserved across a wide range of taxa. This suggests that hemispermatophores, and probably also other functionally constrained structures, can be more informative at higher taxonomic levels than external not functionally constrained morphology.
On the other hand, functionally constrained characters are also expected to show higher level of homoplasy than characters with an infinite morphological space [193][194][195]. However, as evolutionary distance increases, convergent or parallel evolution of similar complex structures and reversals towards lost structures tends to become highly unlikely. As a result, homoplasy between very distantly related taxa should in theory remain rare [196]. Ancestral states reconstructions of the bauplan presented here (Fig. 34) seem to confirm these postulates. Despite several occurences of convergence, parallelism, and reversal in each of the phylogenies, there is never a reversal towards the 'no-fold' bauplan. This suggests that initiation of the sperm duct folding prevents reversal towards the most plesiomorphic state. Moreover, the more complex pattern, the 'four-folds' bauplan, is evolved only once in all phylogenetic reconstructions, confirming the improbability of parallel or convergent evolution of complex features. Therefore, morphological dissimilarities of spermatophore bauplans within the major taxonomic groups are not expected or should at most remain rare.
However, two cases of spermatophore morphological incongruence within higher-level clades raise questions: (1) Except for the genus Lisposoma, the Bothriuridae, traditionally regarded as a basal scorpionoid family (Fig. 34a-c), possess spermatophores with the 'two-folds' bauplans rather than the 'four-folds' bauplan observed in the rest of the superfamily; (2) The clade (Hormiops Fage, 1933 (Hormurus Thorell, 1876 + Liocheles Sundevall, 1833)) (HHL clade), usually retrieved as derived in Hormuridae (Fig. 35a-b), possess hemispermatophores and spermatophores   Fig. 35 Ancestral state reconstruction of the absence/presence of a capsular basal invagination on phylogenies by Stockwell [4] (a), Prendini [45] (b), and Sharma et al. [106] (c). Taxa, traditionally recognized as hormurids, are indicated in brown, and taxa possessing a weak or absent invagination of the basal capsular edge in green. Consistency and retention index values are indicated for each cladogram more akin to those of Diplocentridae and Heteroscorpionidae than to those of other hormurids. Both these cases are discussed below in more detail. Interestingly, the phylogenetic position of Bothriuridae and of the HHL clade inferred in a recent analysis based on transcriptomic data [106] (Figs. 34d, 35c), is more congruent with the grouping of taxa based on the morphological similarity of their hemispermatophores than with earlier phylogenetic reconstructions based on morphology [4,45,46,104] (Fig. 34a-c and 35a-b). Although the cladogram from Sharma et al. [106] should be taken with some skepticism given the very limited number of taxa included in the study and the propensity for severe systematic bias in phylogenomic studies [197][198][199], it nonetheless emphasizes potential flaws in earliers morphological phylogenies.

Hemispermatophore morphological incongruence in Bothriuridae
In all morphological phylogenies, Bothriuridae are retrieved as basal Scorpionoidea (Fig. 34a-c) [4,45,46,104], whereas they are placed very basally in Iurida outside of Scorpionoidea in the phylogeny presented by Sharma et al. [106] (Fig. 34d). On the other hand, the capsule morphology of most of bothriurid genera shows a 'two-folds' pattern like that in the Chactoidea, whereas spermatophores of the remaining scorpionoids show the 'four-folds' bauplan. Among the bothriurids, only the African Lisposoma can be considered as a true scorpionoid if hemispermatophore morphology alone is considered.
This genus, considered as the most basal bothriurid, possesses hemispermatophores with a morphological pattern different from that of other bothriurids and more similar to that non-bothriurid scorpionoids. Moreover, whereas the female genital sclerites are always disjunct in bothriurids, they are fused in Lisposoma [3,4,45,152,153], a character only observed in nonbothriurid scorpionoids. The morphology of the female genital operculum therefore indicates that the insemination mechanism is different to that of bothriuridae and is probably more akin to that of the other scorpionoids (see paragraph above for details). Among scorpionoids, diplocentrid hemispermatophores with relatively simple hemisolenos ( Fig. 22e-h, 23) appear to be morphologically closest to that of Lisposoma.
Three hypotheses can be considered here: (1) Lisposoma is not a bothriurid: it remains in Scorpionoidea while Bothriuridae are placed more basally in Iurida; (2) Lisposoma is a bothriurid and Bothriuridae are scorpionoids: Lisposoma hemispermatophores would then represent the plesiomorphic state for the family with subsequent reversal towards the chactoid morphology in the other more derived genera; (3) Lisposoma is a bothriurid but Bothriuridae are not scorpionoids, placing more basally in Iurida as in Sharma et al. [106] (Fig. 34d). When considering hemispermatophore capsular pattern, the first hypothesis appears to be the most parsimonious with no convergent or reversal event involved. The second hypothesis imply a reversal from the 'four-folds' towards the 'two-folds' bauplan, while an additional independent evolution of the 'four-folds' bauplan from a 'two-folds' ancestor is needed in the third. Unfortunately, Lisposoma has not been included in any molecular phylogeny. Thus the position of this basal bothriurid genus needs first to be empirically tested in a molecular phylogenetic framework before unambiguous conclusions can be drawn regarding the phylogenetic position of the Bothriuridae.

Hemispermatophore morphological incongruence in Hormuridae
The HHL clade is placed as derived in Hormuridae according to several morphological phylogenies [4,45,46,49,50]. However, the hemispermatophore capsular pattern suggests that the genera Hormiops, Hormurus and Liocheles are actually not hormurids. The folding of the sperm duct in the HHL clade (Figs 21, 22a-d, 25a-e) is very similar to that of diplocentrids (Figs 22e-g, 23) and heteroscorpionids (Fig. 20), more so than to hemispermatophores of other hormurids (Figs. 27,28,29,30,31,32,33). In the phylogeny presented by Sharma et al. [106] Liocheles is not retrieved as the most derived hormurid, but is instead placed as a basal scorpionoid, outside of the remaining hormurids, in congruence with the conclusions drawn from morphology of the male copulatory structures. However, very few scorpionoid taxa were sampled in this study, which omitted important ones such as the family Heteroscorpionidae, and a more extensive phylogenetic analysis is needed to confirm the basal position of the HHL clade in Scorpionoidea as retrieved by Sharma et al. [106].
Interestingly, the phylogenetic placement of the HHL clade outside of Hormuridae is also supported by the morphology of the book lungs. As an internal organ, the book lungs can be considered as a functionally constrained complex structure. They are thus probably prone to more phenotypic stability than external nonfunctional morphology, because they form a complex system within which the response of individual characters to environmental selection is limited by functional interaction [57][58][59][60]. The book lungs present a pattern congruent with that derived from the hemispermatophore morphology. Whereas the book lung lamellae and the posterior spiracle edges of the HHL taxa are completely different from those of other hormurids, they share some similarity with those of diplocentrids and heteroscorpionids/urodacids. Hormiops, Hormurus and Liocheles possess lamellae with arcuate distal edges and posterior spiracle edges covered with hillock-like, flattened or chisel-like structures [94,200], whereas all others hormurids, as well as hemiscorpiids and scorpionids, have lamellae with distal edges covered with bristles or spines, and the posterior edge of their spiracles is covered with hexagonal tiles [200]. On the other hand, diplocentrids have lamellae with arcuate distal edges and heteroscorpionids/urodacids have flattened chisel-like structures on the posterior edges of spiracle, like the HHL taxa [94; 200].

Conclusion
The comparative study of hemispermatophores and spermatophores provides new insight into the respective and synergetic roles of natural and sexual selection in the evolution of scorpion copulatory structures. Moreover, morphological similarities between spermatophore capsular patterns of groups previously thought to be distant from each other raise questions about the phylogenetic value of functionally constrained traits and their potential use to assess the reliability of contradicting phylogenetic hypotheses. In the present study, the lack of a robust phylogenetic framework for the order Scorpiones has prevented the elaboration of a sound interpretation of the evolutionary sequence of the male reproductive apparatus. This emphasizes the urgent need for a thorough reassessment of the scorpion phylogenetic relationships based on molecular data that would provide the fundamental layout for an accurate understanding of the evolution of reproductive morphology and other complex character systems, and potentially contribute to uncover and define inherent macroevolutionary trends.