Biomechanical analyses of pterygotid sea scorpion chelicerae uncover predatory specialisation within eurypterids

Introduction

Feeding toolkits of proposed fossil predators are typically explored through functional morphology, often with comparison to modern analogues. In the last two decades, there has been a focus on modelling animals using two-dimensional (2D) and three-dimensional (3D) biomechanical analyses, including finite element analyses (FEA) (Ross, 2005; Rayfield, 2007; van Heteren et al., 2021; Rowe & Snively, 2022). This latter method has been an effective approach for modelling fossil vertebrates and, as such, applications of FEA in palaeontology have largely focused on select vertebrate groups (Rayfield et al., 2001; Wroe, McHenry & Thomason, 2005; Wroe et al., 2007, 2018; Kupczik et al., 2009; Strait et al., 2009; Attard et al., 2016). By comparison, fossil arthropods have not been examined as thoroughly with 3D FEA (Bicknell et al., 2018a, 2021; Esteve et al., 2021). Recent focus on modelling extinct arthropods has increased knowledge of the biomechanical capability of these fossil forms.

Fossil predatory arthropods are epitomised by the large, aquatic eurypterid family Pterygotidae, known from Silurian and Lower Devonian deposits (Braddy, Poschmann & Tetlie, 2008; McCoy et al., 2015; Lamsdell & Selden, 2017). Some forms have bodies reaching ~2.5 m in length and represent the largest known marine arthropods, living or extinct (Braddy, Poschmann & Tetlie, 2008; Lamsdell & Braddy, 2010; Vermeij, 2016). A key feature of pterygotids is their large, anteriorly-directed chelicerae that often show enlarged denticles (Ciurca & Tetlie, 2007), likely used to capture and subdue prey (Kjellesvig-Waering, 1964; Waterston, 1964; Miller, 2007b; Braddy, Poschmann & Tetlie, 2008; Kennedy, Miller & Gibling, 2012; Bicknell, Smith & Poschmann, 2020). Given this, pterygotids are presumed to have been apex predators within their respective ecosystems (Selden, 1984; Plotnick & Baumiller, 1988; Braddy, Poschmann & Tetlie, 2008). Despite the striking appearance of pterygotid chelicerae, to date, biomechanical investigations of these structures have been limited to 2D lever arm studies (Selden, 1984; Laub, Tollerton & Berkof, 2010). Examination of these chelicerae using 3D biomechanical analysis could therefore present a more complete understanding of their functional morphology. Importantly, modern scorpion pedipalp chelae can inform on the mechanical performance of pterygotid chelicerae, as the former are perhaps the closest functional analogue for the fossil forms. Furthermore, as scorpions are a diverse chelicerate group that are phylogenetically closer to eurypterids than other arthropod groups with chelate appendages, such as decapod crustaceans (Shultz, 2007; Legg, Sutton & Edgecombe, 2013; Haug, 2020), and likely have comparable cuticular properties to eurypterids, scorpions are one of the more informative groups to function as modern analogues.

Pterygotid chelicerae are occasionally preserved with sufficient anatomical fidelity to allow for detailed reconstructions. In particular, Acutiramus bohemicus, Erettopterus bilobus, Jaekelopterus rhenaniae, and Pterygotus anglicus are four exceptionally-preserved and well-documented species, making them ideal for biomechanical analysis (Waterston, 1964; Chlupáč, 1994; Poschmann & Tetlie, 2006; Miller, 2007b; Braddy, Poschmann & Tetlie, 2008; Dunlop, Penney & Jekel, 2020). Furthermore, modern scorpions provide a diverse and well-studied group for comparative purposes (Van der Meijden, Kleinteich & Coelho, 2012; Lourenço, 2021). Building on previous FEAs of scorpions (Van der Meijden, Kleinteich & Coelho, 2012) and analyses of extinct arthropods (Bicknell et al., 2021), here we present 3D finite element models (FEMs) of the chelicerae of Ac. bohemicus, Er. bilobus, Ja. rhenaniae, and Pt. anglicus and compare them to models of chelae of 16 scorpion taxa.

Methods

Extant models

Information on the muscle that adducts the moveable finger in a scorpion chela follows Snodgrass (1952), Gilai & Parnas (1970), and Van der Meijden, Kleinteich & Coelho (2012) (Fig. 1A). One primary muscle mass fills the majority of the proximal manus region and is responsible for adducting the moveable finger. There is also a secondary closing muscle in the patella (Gilai & Parnas, 1970). We have not considered this muscle here because: (a) scans of patellar sections needed to estimate muscle size were not made; and (b) there is no fossil evidence for this muscle (such as scars or fibres) in eurypterids. Data for calculating force of the primary closing muscle were collected from micro-computed tomography (micro-CT) scans. The scans were made with a Skyscan 1076 micro-CT scanner using a source voltage of 31 kV and a source current of 187 µA at 35 µm resolution, as detailed in Van der Meijden, Kleinteich & Coelho (2012). Sixteen specimens, each representing a different scorpion genus and spanning six families, were scanned and chelae segmented with Amira/Avizo 5 (Thermo Fischer Scientific, Waltham, MA, USA). The external cuticle and internal manus content were segmented either manually or using the “Thresholding” tool in Amira. Internal manus content, excluding regions where adducting muscles were not present, was segmented as a globular structure to estimate muscle volume, surface area, and cross-sectional area. Values of volume and area were calculated with the “Surface Area Volume” module in Avizo, while the cross-sectional area was calculated by bisecting the muscle volume transversely (i.e., perpendicular to the proximal-distal axis of the manus and approximately perpendicular to muscle fibre directions) in Meshlab v. 2020.12 (Cignoni et al., 2008). Analysed specimens were preserved in ethanol and housed in the Centro de Investigação em Biodiversidade e Recursos Genéticos, Vila do Conde, Portugal (CIBIO) collection and assigned the prefix of ‘Sc’.

Maximum pinch force data were only known from the micro-CT scanned specimens of Androctonus bicolor and Pandinoides cavimanus (Simone & Van der Meijden, 2018). As we did not measure in-vivo pinch forces of the micro-CT scanned individuals of other species, pinch forces were estimated for those specimens using data available from other individuals of the same species. As such, chela length, height, and width (Stahnke, 1970) were used to predict pinch force for our FEMs based on known pinch forces of other individuals of the same species (Van der Meijden, Herrel & Summers, 2010; Van der Meijden, Kleinteich & Coelho, 2012; Table S1). The data were log₁₀ transformed to linearise variables with different dimensionality and used to produced species-specific linear models, and associated regression coefficients (Table S2).

The mechanical advantage of the moveable finger was calculated from the micro-CT scans following Simone & Van der Meijden (2018). However, as in-vivo pinch forces were measured approximately two thirds of the finger length from the joint and not the distal-most point of moveable fingers, the mechanical advantage was corrected by shortening the length of the out-lever by one third. Muscle force at insertion was then calculated for scanned specimens by dividing pinch force by the specimens’ mechanical advantage (Table S3). Muscle stresses used here were derived from Van der Meijden et al. (2012), where the stress of Galeodes sp. (203 kPa) and Rhagodes sp. (905 kPa) were employed as upper and lower bounds, respectively (Table S4). The manus and moveable finger reconstructions were exported from Amira as .STL files for analysis. The manus and moveable finger sections for assessed specimens were then imported into 3-matic version 12 (Materialise, Leuven, Belgium) and solid-meshed as distinct solid homogeneous structures consisting of tet-4 elements and the gape angle was set to biologically realistic values between 10–30°. These models were then exported as Nastran files for import into Strand7 (Strand7 Pty Ltd, NSW, Sydney, Australia) FEA software. Material properties used are a Young’s modulus of 7 GPa and a Poisson’s ratio of 0.3—values used for scorpion cuticle following Van der Meijden, Kleinteich & Coelho (2012). Muscle origins were tessellated as beam elements onto the Nastran models (Bicknell et al., 2018b). This was done following successful applications to other arthropods (Bicknell et al., 2018b, 2021) and allowed for a large muscle origin area to be modelled. Muscle forces (Table S3) were assigned to trusses directed toward the insertion site—the most proximal section of the moveable finger. These insertions were treated as static points at the beam terminus. Three denticles along both the manus and moveable finger were constrained in all directions at their most apical node (Fig. 1A); selected denticles were located at the proximal, mid-length, and distal regions of the manus and finger. A hinge between the manus and moveable finger was constructed using two sections: one link on either side of the proximoventral section of the moveable finger. This emulates a simplified action of manus and moveable finger closure. A colour-coded von Mises (VM) stress map was generated after solving models. Loaded Strand7 models are presented as Data S1–S16. These data are found at 10.17605/OSF.IO/GV8J5 Additionally, the analysed .STL files were used to generate 3D PDFs using Tetra4D (Adobe Systems, Mountain View, CA, USA). Scorpion 3D PDFs are available from 10.17605/OSF.IO/GV8J5.

Results

The linear model used to estimate muscle force for the scorpions has a statistically significant adjusted R² value of 0.657 (p-value: 3.8e⁻⁰⁴; Fig. 2). Furthermore, estimated muscle volume has a significant correlation to the pinch force estimation (F-value: 29; p-value: 1.1e⁻⁰⁴). The pterygotid muscle forces predicted from muscle cross-sectional area and muscle stress were considerably higher than those based on the linear model (Table S5).

The VM stress distributions for the examined scorpion chelae are categorised into three main groups. The first group comprises Androctonus amoreuxi, Caraboctonus keyserlingi, Chactas sp., Hadogenes paucidens, Opistophthalmus boehmi, and Scorpio maurus, each of which exhibit relatively high VM stress across the entire model, with the highest VM stress proximal to the articulation between the manus and moveable finger, and along chelae (Figs. 3A–3F). These forms are typically stouter and have more pronounced proximal manus regions. Notably, Ca. keyserlingi, Ha. paucidens, and Sc. maurus have higher stress along the moveable finger. The second group consists of Androctonus australis, An. bicolor, Grophus flavopiceus, Hadrurus arizonensis, Hottentotta gentili, Leiurus quinquestriatus, and Parabuthus transvaalicus. These models have overall lower VM stress in the manus and variable degrees of VM stress along elongate chelae fingers (Figs. 3G–3M). Androctonus australis, An. bicolor, Ha. arizonensis, Le. quinquestriatus and Pa. traansvalicus have high VM stress proximal to the articulation between the manus and moveable finger, contrasting with Gr. flavopiceus and Ho. gentili that have lower VM stress over the entire model. The third group contains Bothriurus sp. and Pandinoides cavimanus. These models show low VM stress in the proximal manus section, high VM stress at the articulation between the manus and moveable finger, and lower VM stress along stout chelae (Figs. 3N and 3P). Finally, Opisthacanthus maculatus is unique, with low VM stress in the proximal manus region and higher VM stress along the length of the chela fingers (Fig. 3O). Overall, these groupings are comparable to those presented by Van der Meijden, Kleinteich & Coelho (2012), where chelae models were standardized for scale using force to surface area ratio.

Distributions of VM stress in the pterygotid models are broadly comparable to those of scorpion FEMs (Fig. 4). Although these appendages are not homologous—representing the first and second limb-bearing segments in eurypterids and scorpions, respectively—the functional analogy of the two can be tested with FEA. The observed VM stress similarity strongly supports the idea that pterygotid chelicerae functioned like scorpion chelae, indicating an informative analogue for these fossil forms. Considering the mean VM stress distributions of the scorpion and pterygotid models, the former have higher mean stress values (Fig. 5). There is limited clustering of family groups in bivariate space, reflecting the range of input force values for the modelled scorpions (Table S9). Taxa within the Buthidae in particular have a large spread of stress within similar volume values. Within the pterygotids, the larger models (Jaekelopterus rhenaniae and Acutiramus bohemicus) show the lowest mean stress.

Of the pterygotid models, Erettopterus bilobus and Pterygotus anglicus have comparable VM stress along the fixed and free rami and proximal to the region of articulation between the fixed and free rami (Figs. 4A and 4B). Pterygotus anglicus shows high VM stress on the large denticles and Er. bilobus has lower VM stress along the distal section of the free ramus. Jaekelopterus rhenaniae has the lowest overall VM stress and its only high VM stress region is proximal to the point of articulation between the fixed and free rami and along the proximal section of the free ramus (Fig. 4C). Acutiramus bohemicus has high VM stress in the region proximal to the point of articulation between the fixed and free rami and along the proximal section of the free ramus (Fig. 4D). Further, the highest strain areas are located along the oblique, serrated denticles.

In all FEMs, the VM stress is predictably concentrated in areas with high loads, such as muscle insertions and around the fixed vertices on the denticles. In certain scorpion models, higher VM stress areas are visible on the proximal manus region—the location of simulated muscle origins. However, scorpion chelae muscle origins are distributed across the majority of the proximal manus region. As such, VM stress concentration on the proximal manus region reflect where the models were constrained.

The large difference in estimated muscle force and stress distributions for the pterygotids between the linear model and the cross-sectional area method is striking. As both methods are commonly used, and both have weaknesses and assumptions, we opted to present the results of both approaches to see if similar values could be predicted. The linear model predicted high muscle force values for the pterygotids. However, as no extant chelicerates have comparably-sized chelae or chelicerae, muscle forces were estimated by extrapolation rather than interpolation. An unavoidable result of extrapolation is increased error. The cross-sectional area method of muscle force estimation has more explicit assumptions, each with a level of uncertainty. The range of lowest to highest muscle stress estimates illustrate this uncertainty (Table S5). For these reasons, we did not use the muscle force estimates derived from cross-sectional area in the FEMs, and did not focus on absolute magnitudes of stress or force in our analysis and discussion.

Discussion

The results of the scorpion FEAs show similar groupings as in Van der Meijden, Kleinteich & Coelho (2012). The differences in magnitudes of VM stresses (Fig. 3) indicate that some species (e.g., Pandinoides cavimanus, Ophistacanthus maculatus, Hottentotta gentili) could exhibit a higher ‘safety factor’ (i.e., a measure of how much stronger a morphology needs to be compared to the input forces; Hicks & Wang, 2021) than others (e.g., Caraboctonus keyserlingi, Hadogenes paucidens, Parabuthus transvaalicus), although the pattern does not seem to correspond to defensive behaviour (Van der Meijden et al., 2013) or relative pinch force (Simone & Van der Meijden, 2018). Considering the modelled scorpions in the context of possible ecomorphs, there is limited overlap between groups identified here and recently proposed ecomorphologies (Coelho et al., 2022). This suggests that the FEMs produced here present more of an insight into the effectiveness of these feeding tools, rather than inform on possible ecological groupings and associated microhabitats. Combining these biomechanical analyses with detailed examination of scorpion life modes will undoubtedly uncover new links between chelae mechanics and the diets of these ecomorphs. However, this is beyond the scope of this work, as we have focused on using scorpions to inform aspects of the pterygotid models. On this point, it is worth considering other possible modern analogues for understanding and modelling pterygotids. Of particular note are the camel spiders (Solifugae), which have large, dentate chelicerae that can grasp and disarticulate prey for consumption (Van der Meijden et al., 2012). These chelicerae are anteriorly directed, but do not extend notably beyond the prosoma (Harms & Duperre, 2018). While they may be informative for understanding how chelicerae can break prey, they are less useful than scorpion pedipalps as functional analogues for pterygotid chelicerae.

The morphology of pterygotid chelicerae suggests that these structures were used in capturing prey for subsequent mastication by the coxal gnathobases (Waterston, 1964; Miller, 2007a). The solifuge comparisons made above suggest that the dentition of pterygotid chelicerae may have allowed for some initial tearing of prey. This inferred feeding mode aligns with the co-occurrence of pterygotids and proposed prey species, including other eurypterids, fishes, and a likely array of soft-bodied animals that are not preserved (Rolfe, 1973; Chlupáč et al., 1980; Kennedy, Miller & Gibling, 2012). Furthermore, links between cheliceral and lateral compound eye morphologies have been drawn to propose an apex predatory life mode for some pterygotids (Anderson et al., 2014; McCoy et al., 2015).

Results of the FEA support previous inferences that pterygotid chelicerae, while having a similar overall morphology, had differing functional capabilities (McCoy et al., 2015). The comparable VM stress distributions between the chelicerae of Erettopterus bilobus and Pterygotus anglicus suggest that these taxa would have experienced similar stresses while feeding and therefore probably targeted similar prey. Based on the known faunal assemblages that co-occur with these pterygotid species (Dunlop, Braddy & Tetlie, 2002; Miller, 2007b; Lebedev et al., 2009; McCoy et al., 2015; Fyffe, Johnson & van Staal, 2016; Blieck, 2017; Tables S10 and S11), coupled with the large body lengths of Er. bilobus (0.7 m) and Pt. anglicus (1.6 m), proposed prey includes other eurypterids and fishes; in the case of the Devonian Pt. anglicus, the latter could include armoured forms, such as ostracoderms and placoderms. This is further supported by the moderate to high visual acuity in these pterygotid genera (McCoy et al., 2015), allowing them to identify and pursue such mobile, possibly smaller prey. Indirect evidence for Er. bilobus consuming fish includes coprolites rich in agnathan fragments from the Lesmahagow Inlier that have been previously attributed to eurypterid predation (Rolfe, 1973; Selden, 1984). It therefore seems that Er. bilobus and Pt. anglicus were apex predators within their respective ecosystems (Selden, 1984; Dunlop, Braddy & Tetlie, 2002; Kennedy, Miller & Gibling, 2012). This contradicts a previous suggestion that Er. bilobus was a more generalised predator than Pt. anglicus (McCoy et al., 2015).

The FEM of Acutiramus bohemicus produced a unique VM stress distribution associated with its different cheliceral morphology. All Acutiramus species have elongate (some hypertrophied), oblique, anteriorly-directed, serrated denticles (Chlupáč, 1994; Laub, Tollerton & Berkof, 2010; McCoy et al., 2015). High VM stress along the free ramus and within the constrained denticles suggest that Acutiramus was not well adapted to capturing armoured or thick-shelled prey and may have experienced failure when doing so (Laub, Tollerton & Berkof, 2010). Indeed, Acutiramus does not co-occur with a diverse fish fauna, but rather an array of other eurypterid taxa (Laub, Tollerton & Berkof, 2010; Anderson et al., 2014; Table S12). The angled, serrated denticles would be more consistent with piercing eurypterid cuticle (Laub, Tollerton & Berkof, 2010; Anderson et al., 2014; Fyffe, Johnson & van Staal, 2016) and the proximal denticles on the free ramus may have impaled prey (Laub, Tollerton & Berkof, 2010). After impaling, the prey would have been sliced by the serrated denticles as the chelicerae closed. The serrations documented by Laub, Tollerton & Berkof (2010) are proximally orientated, suggesting that after impaling, a victim would have had to tear itself from the serrations, likely causing more damage. This conforms to the idea that Acutiramus was likely an ambush predator or scavenger that fed on soft-bodied and lightly cuticularized taxa, based on its limited vision and cheliceral morphology (Anderson et al., 2014; McCoy et al., 2015).

The Jaekelopterus rhenaniae model has overall low VM stress compared to the other pterygotid models. These data, combined with a proposed 2.5 m body length (Braddy, Poschmann & Tetlie, 2008), high visual acuity (McCoy et al., 2015; Poschmann, Schoenemann & McCoy, 2016), good swimming abilities (Plotnick & Baumiller, 1988; Tetlie, 2007), and a diverse co-occurring fauna containing several eurypterid and armoured fish species (Fyffe, Johnson & van Staal, 2016; Poschmann, Schoenemann & McCoy, 2016; Table S13), suggest that Ja. rhenaniae was capable of capturing large, highly mobile, armoured prey. The robust chelicerae would have enabled the initial grabbing and manipulation of food items, while the heavily sclerotised gnathobases on large coxae would have been employed in prey mastication (including crushing of thick cuticle or biomineralised structures), comparable to extant xiphosurids (Waterston, 1964; Botton, 1984; Poschmann, Bergmann & Kühl, 2017; Bicknell et al., 2018a, 2018b). The large reinforced chelicerae, giant body size, and phylogenetically-derived position of Ja. rhenaniae (Fig. 6; Braddy, Poschmann & Tetlie, 2008) appear to represent a peak in pterygotid cheliceral evolution that coincides with the rapid diversification of placoderms (Randle & Sansom, 2019). As such, predatory Devonian eurypterids, epitomised by Ja. rhenaniae, likely developed a toolkit to target armoured fish in the form of robust chelicerae for grabbing prey and stout gnathobases for masticating said prey (Poschmann, Bergmann & Kühl, 2017), as opposed to eurypterids driving the radiation of placoderms (contra Romer, 1933). The extinction of pterygotids in the Late Devonian is consistent with them ultimately being outcompeted by vertebrates (particularly jawed fishes), cephalopods, and various other groups during the Devonian Nekton Revolution (Lamsdell & Braddy, 2010; Klug et al., 2010, 2018), as well as falling victim to environmental changes (Lamsdell & Selden, 2017).

Conclusions

Finite element models of chelicerae from four pterygotid eurypterids (Acutiramus bohemicus, Erettopterus bilobus, Jaekelopterus rhenaniae, and Pterygotus anglicus) are presented and compared with FEMs of modern scorpion chelae. Overall similarity in VM stress distributions indicates that pterygotid chelicerae are functionally analogous to scorpion chelae, as suggested by the morphological similarity of these structures. Considering the pterygotid VM stress distributions in the context of their palaeoecology (including visual capabilities) and overall cheliceral morphology, we have demonstrated, with FEA, the morpho-functional ability of these ancient marine predators. We conclude that Er. bilobus and Pt. anglicus were apex predators within their respective ecosystems and likely targeted other eurypterids and fishes. Stress distributions of the Ac. bohemicus model, together with its unique denticle morphology, suggest that the chelicera of this species was adapted to piercing and slicing the cuticle of other eurypterids. Finally, Ja. rhenaniae, the largest known pterygotid, experienced low VM stress across the chelicera, suggesting that it was well adapted to capturing large, highly mobile, armoured prey. These results demonstrate how 3D FEA can be applied across a range of morphologies and used to explore the mechanical performance of extinct predatory arthropods.

Supplemental Information