Reevaluating scorpion ecomorphs using a naïve approach

Background Ecomorphs create the opportunity to investigate ecological adaptation because they encompass organisms that evolved characteristic morphologies under similar ecological demands. For over 50 years, scorpions have been empirically assigned to ecomorphs based on the characteristic morphologies that rock, sand, vegetation, underground, and surface dwellers assume. This study aims to independently test the existence of scorpion ecomorphs by quantifying the association between their morphology and ecology across 61 species, representing 14 families of the Scorpiones order. Results Without a priori categorization of species into ecomorphs, we identified four groups based on microhabitat descriptors, which reflect how scorpion ecospace is clustered. Moreover, these microhabitat groups, i.e., ecotypes, have significantly divergent morphologies; therefore, they represent ecomorphs. These ecomorphs largely correspond with the ones previously described in the literature. Therefore, we retained the names Lithophilous, Psammophilous, and Pelophilous, and proposed the name Phytophilous for vegetation dwellers. Finally, we sought to map the morphology-ecology association in scorpions and found that the morphological regions most tightly associated with ecology are at the extremities. Moreover, the major trend in ecomorphological covariation is that longer walking legs and relatively slender pedipalps (pincers) are associated with sandy microhabitats, while the inverse morphological proportions are associated with rocky microhabitats. Conclusions Scorpion ecomorphs are validated in a naïve approach, from ecological descriptors and whole body anatomy. This places them on a more solid quantitative footing for future studies of ecological adaptation in scorpions. Our results verify most of the previously defined ecomorphotypes and could be used as a current practice to understand the adaptive significance of ecological morphology. Supplementary Information The online version contains supplementary material available at 10.1186/s12862-022-01968-0.


Introduction
Ecological adaptation is the process whereby organisms respond to environmental change by modifying their phenotype. Selection triggers adaptation to new environments and ecological niches, resulting in divergence. If the adaptations include changes in external morphology, we can observe this type of divergence in the form of recurrent morphologies -ecomorphs. Ecomorphs are groups of organisms that share a set of morphological traits associated with speci c ecological requirements. Selection for these characteristic morphologies occurs in association with enhancing the ecologically relevant performance that morphology permits, also modulated by associated behaviors [1][2][3][4][5]. Striking examples of ecomorphs are predominantly known from recent adaptive radiations, such as cichlid sh [6] and Anolis lizards [7,8]. Although with fewer examples, ecomorphs originating from older radiations are also known, such as beetles and salamanders [9,10].
Scorpions are particularly suitable for studying ecomorphology because they have a functionally compartmentalized anatomy. As such, the functions and behaviors related to intraspeci c competition, prey incapacitation, defense against predators, feeding, sensing, and sexual courtship are carried out by four anatomical regions: the pedipalps, prosoma, metasoma, and telson [11][12][13][14][15][16]. In other organisms, these functions may all be carried out using a single anatomical region, such as the head in lizards [17], while in scorpions, some functions are limited to one or more of these anatomical regions. For instance, prey capture and manipulation are restricted to the pedipalps, while prey incapacitation and defense often include the telson. Consequently, in scorpions, differences in these anatomical regions are more easily assigned to different functional demands.
Another reason makes scorpions an interesting group for studying the ecological adaptation of their morphology: they have long been assigned to ecomorphs [18][19][20][21]. These ecomorphs are considered to be associated with microhabitat use, speci cally the substrate that scorpions use. The ve major scorpion ecomorphs previously identi ed are vegetation dwellers (corticolous) [22,23], rock climbers (lithophilous) [24,25], hard soil fossorials (pelophilous) [23,26], sand fossorials (psammophilous) [27,28], and ground surface vagrants (lapidicolous) [20,29]. Among these, there are substrate specialists and substrate generalists, i.e., habitat stenotopes and habitat eurytopes, sensu [29]. Psammophiles and lithophiles are habitat stenotopes because the extensive morphological adaptations for their preferred substratesloose sand and rocks -make such species un t to occupy other substrates. Lapidicolous, corticolous and pelophilous ecomorphs are eurytopic [23], with lapidicolous being the least specialized group. Other scorpiologists labeled ecomorphs differently, yet this re ects a terminological difference rather than a conceptual one [23,30]. Some of the described morphological characteristics of these ecomorphs are considered functional adaptations. Psammophilous scorpions have longer setae and claws on their feet that may spread their mass over a larger area and reduce energy wasted on displacing loose sand during locomotion. Also, the streamlined metasoma and telson are suggested to increase sand burrowing e ciency [28] and aid in escape when animals become buried in the sand [20]. In pelophilous species, the robust, crab-like chelae are proposed to aid in burrowing, although not all species with large chelae use them for this function [29]. In addition, a lifetime spent inside a burrow is thought to reduce telson size due to its diminished use [20]. Lithophilous species are associated with dorsoventral compression and often elongation of the pedipalps, legs, and metasoma. This trend, however, seems more evident in rock crevice dwellers than in other rock microhabitats [29]. Corticolous species are also associated with some elongation of the pedipalps and legs. Other trends in corticolous scorpions seem to be more phylogenetically speci c, such as the elongation of the metasoma in buthids and a general dorsoventral compression of the body in non-buthids [29]. Nonetheless, these examples are largely empirical, featuring descriptions of both regional scorpiofauna or clade-speci c groups. No global study has tested whether scorpions inhabiting different habitats and continents can be grouped together based on their ecological preferences; and whether such groups also exhibit morphological coherence across the scorpion phylogeny. In sum, there is no overall data-based validation that scorpion ecomorphs exist.
Con rming that scorpion ecomorphs exist involves the simultaneous analysis of morphological and ecological traits in a phylogenetic comparative framework. Our main questions are: Can scorpions be grouped based on microhabitat preference? If so, are these groups morphologically distinct? Do these groups correspond to currently recognized ecomorphs? Which morphological traits co-vary more strongly with ecology? Given the recurrent association of characteristic pedipalp and "tail" (metasoma and telson) shapes for each ecomorph, we hypothesized that they are the ecologically most important anatomical parts.
We used recent transcriptome-based phylogenies of the Scorpiones order for our comparative analyses [31,32]. From there, we drew an ultrametric tree with the same topology as the higher-level classi cations.
The interfamilial relationships are congruent in both publications, except for the position of the Vaejovidae, which was revised in [32] and adopted in this study. Species that were not present in either phylogenetic analysis were treated in one of two ways: species with members of the same genus represented in the phylogenies were placed at the position of the genus; species without genus representation were placed basal inside the corresponding family clade [33,34]. In all cases, unclear relationships between species were represented with polytomies. Branch lengths were adjusted to the tree topology, calculated using Grafen's method [35].

Morphological measurements
We aimed for an unbiased sampling of scorpion anatomy by measuring all anatomical regions of the scorpion body. Speci cally, we did not restrict our sampling to known eco-functional traits. We used digital calipers (Absolute IP67, Mitutoyo Inc., Kawasaki, Japan) to measure 70 lengths from six anatomical regions, namely in the prosoma, mesosoma, metasoma, telson, walking legs, and pedipalps, to the nearest 0.01 mm, following Stahnke [36]. Our measurements did not include structures such as the chelicerae, carinae, tarsi, or the setal hairs. The measured species exhibit considerable size variation, and our sampling represents total body lengths from 22.1 mm in Microbuthus sp. to 158.05 mm in Ha. granulatus. To reduce the number of variables for statistical analysis, we summed trochanter and femur lengths into a "proximal leg" part and the patella, tibia, and metatarsus lengths into a "distal leg" part. This division corresponds to biomechanically functional units of the leg: on one side, the distal muscles, responsible for pretarsal ("foot") movement occupy all distal segments until and including the patella; on the other side, most of the leg motion occurs around the femoropatellar joint [37]. The lengths of the ve metasoma segments were added together, while their heights and widths were averaged. Pedipalp measurements represent averages between left and right pedipalps. For details about the measurements (description, abbreviation, and illustration), see Supplementary Table 1 Table 2). In general, scorpions are not characterized by strong sexual dimorphism, so we did not differentiate specimens by sex. However, in those species with more substantial sexual dimorphism, males and females may have subtly different ecological roles [38].
Unlike in, e.g., herpetology, no consensually accepted single linear measurement corresponds well with overall body size in scorpions [39]. Therefore, an isometric body size (IsoSize) was calculated by projecting the 36 linear measurements on an isometric vector. We then calculated each linear measurement's regression residuals using IsoSize to obtain size-corrected traits, following [40]. Last, we calculated the degree of phylogenetic signal present in the morphological variables given the phylogeny using the function physignal of the of R package GEOMORPH. physignal provides a mathematical generalization of the Kappa statistic [41] appropriate for highly multivariate data [42].

Ecological data and microhabitat clustering
Since the conceptualization of the ve scorpion ecomorphs, not all species have been assigned to one. For example, less than 50% of the taxa selected here are unambiguously assigned to an ecomorph in the literature. In cases where an ecomorph assignment can be found in the literature, the assignment is often made based on the morphological habitus of the specimens, risking circular reasoning. To overcome these limitations, we chose to forego the assignments to classical ecomorphs entirely in our analysis.
Instead, we retrieved descriptors of microhabitat use from the literature. We selected descriptors referring to substrates, their arrangement, and the activities scorpions perform in them. Those most frequently encountered in the literature were used to record presence (= 1) vs. unreferenced presence (= 0) for the following 12 parameters: "compact soil", "loose sand", "rock surface", "leaf-litter", "under rocks", "under vegetation", "dug burrow", "passive shelter", "climb bushes", "climb trees", "rock crevices" and "hanging upside down". Each species was assigned one or more microhabitat uses (see Supplementary Table 2).
We grouped species with similar microhabitat uses together into clusters. To achieve this, we used the matrix of 12 microhabitat traits in the following three steps. First, we calculated Jaccard distances between pairs of species based on all microhabitat descriptors together, using the function vegdist of R package vegan [43]. Secondly, using the obtained distance matrix, we selected the number of habitat clusters based on the Bayesian Information Criterion (BIC), using the function Mclust of R package mclust [44] (Supplementary Figure 2). We selected the number of clusters corresponding to the rst k preceding a plateau in BIC values. The number of clusters, four (see results), was used to compute a kmeans clustering. The distance matrix was used as input for a Principal coordinate analysis (PCoA), resulting in 26 axes. Lastly, we reviewed the ecological composition of each microhabitat cluster by calculating two matrix correlations (Spearman's as well as Pearson's ρ): 1) between the microhabitat traits and the PCoA scores to obtain PCoA-to-microhabitat correspondence; 2) between the resulting matrix from 1) and the k-means cluster centers to obtain microhabitat-to-cluster correspondence. Cluster terminology reproduces the different associations of each cluster to the microhabitat traits.
To visualize the scorpion ecospace, we performed multiple correspondence analyses (MCA) using function MCA of R package FactoMineR [45]. MCA uses the PCoA scores of microhabitat traits to plot barycenter points of categories (n-individual mean scores) and barycenter points of individuals (ncategory mean scores) simultaneously; for visual clarity, however, only the latter were plotted.

Ecology-morphology associations
We searched for strongly eco-covarying traits by exploring which morphological traits co-vary most with ecology across all species. To this end, we applied a phylogenetic Partial Least Squares (PLS, using the phylogeny shown in Figure 2) to the multivariate sets of ecological and morphological variables using the function phylo.integration [46-48] of R package GEOMORPH [49]. Here, multivariate ecology consisted of 26 microhabitat traits (PCoA axes), and multivariate morphology consisted of 36 size-corrected morphological traits. Permutations with 10,000 cycles were used to test for signi cance of the multivariate correlation between vectors of morphology and ecology. The resulting matrix of morphological traits with maximized ecological covariance is hereafter referred to as the matrix of ecoprojected morphology.

Ecomorphological distinctiveness between microhabitat clusters
To corroborate the existence of ecomorphs in scorpions, we examined whether microhabitat clusters exhibit distinctive phenotypes as represented by ecologically correlated morphology. For this purpose, we performed a MANOVA with the matrix of eco-projected morphology as the dependent variable and microhabitat cluster as a predictor while accounting for phylogenetic autocorrelation using generalized least squares (GLS). To test for signi cance, we used randomization of residuals over 10,000 permutations, as implemented in the function lm.rrpp of the RRPP R package [50,51]. Then, to identify which microhabitat clusters differed signi cantly, we employed distance-based testing of pairwise differences between microhabitat cluster means, as implemented in the function pairwise of RRPP [52,53]. For illustration, we plotted group means rotated to their principal components and with 95% con dence ellipses around them, using the plotting tools of the RRPP R package.

Microhabitat clusters as a proxy for ecomorphs
Based on the BIC, the ecological data supports the existence of four clusters (Supplementary Figure 2). In terms of ecological composition, the microhabitat clusters show different associations with microhabitat traits (Table 1, Supplementary Figure 3): Cluster 1 is primarily dominated by species occurring on loose sand and digging burrows; Cluster 2 mainly contains species found under vegetation, in leaf litter and using passive shelters; Cluster 3 is mostly comprised of species found in compact soil and digging burrows and Cluster 4 has a higher contribution of scorpions exploring both the crevices and the surface of rocks ( Figure 1, Table 1). It is worth noting that scorpions that rest hanging upside down are closest to cluster 4, while those climbing trees and bushes only positively correlate with cluster 2. Therefore, we named these clusters "Psammophilous", "Phytophilous", "Pelophilous" and "Lithophilous", respectively, following the most common ecomorph terminology (Figure1, Table 1, Supplementary Table 2).
Dimensions 1 and 2 of the MCA together capture about 40% of the microhabitat variation. Across this ecospace, the microhabitat clustering produced ecomorph groups with reduced overlap (Figure 1).
The phylogenetic tree of scorpions reveals that most scorpion families contain representatives of at least two ecomorphs (Figure 2). The Vaejovidae and Diplocentridae families are the exceptions, only composed of Psammophilous and Pelophilous species, respectively. This is likely an artifact of the limited sampling of these families, especially of the highly speciose Vaejovidae. At the superfamily level, the Iuroidea has representatives of all ecomorphs, while the Buthoidea lacks the Lithophilous ecomorph in our sampling. As a reminder, the absence of the Lithophilous ecomorph does not imply that Buthoidea scorpions are not associated with rocks. Although more prevalent in Lithophilous species, rock surfaces are also part of the Pelophilous niche (Table 1). Table 1 Microhabitat cluster a liation with microhabitat traits. Legend: Values correspond to Pearson's ρ correlation, on which the terminology adopted for the ecomorphs was based.
The phylogenetic history of scorpions has a signi cant in uence in their morphology. As such, the metasoma average width and height are the traits with the highest phylogenetic signal among the sampled taxa (Supplementary Figure 5).

Ecology-morphology associations
Examination of the multivariate association between morphology and ecology yielded signi cantly correlated PLS vectors (r = 0.698, z = 3.228, p < 0.001) (Figure 3). Vectors of ecology and morphology revealed that scorpions living under and on rock surfaces are associated with pedipalps with (dorsoventrally) higher femurs, wider patellas, and both wider and higher chelae. They are also associated with leg pairs 1 and 2 having shorter proximal parts and leg pairs 2 and 3 having shorter distal parts. Conversely, scorpions living in loose sand and digging burrows have inverse anatomical proportions ( Figure 3). As a reminder, we attributed the term eco-projected morphology to the morphological dataset resulting from this PLS; in it, the morphological traits with the highest ecological correlation, i.e., strongly eco-covarying traits, are located in the walking legs and the pedipalps of scorpions (Figure 3).
Ecomorphological distinctiveness between ecomorphs MANOVA comparisons show signi cant differences in eco-projected morphology across the four microhabitat clusters (  Table 3). Mean morphologies and respective con dence ellipses, estimated from the MANOVA, con rm the reduced overlap between the different ecomorphs ( Figure 4). The illustrations of four scorpions approximate the mean morphologies of each ecomorph.

Discussion
In this paper, we independently corroborate the existence of scorpion ecomorphs, and show that they are largely congruent with, but not identical to those previously de ned. Moreover, we found that some parts of the scorpion body vary more strongly with ecology than others, yet not those body parts we expected.

Microhabitat clusters are ecomorphs
To start with a clean slate and avoid circular reasoning, we summarized microhabitat use in clusters rather than assigning species to previously de ned ecomorphs. The number of clusters recovered in this study, four, differs from the ve ecomorphs previous literature would have assigned our sampling to [29,54]. We validated that our microhabitat clusters represent ecomorphs by showing that morphological variation is signi cantly different between them (Table 2). This validation accounted for two factors that mask anatomical similarity: body size and phylogeny.
Our results corroborate most ecomorphs already de ned in bibliography, and thus we generally use the same names for the sake of continuity. However, we propose changing the designation "Corticolous" to "Phytophilous" as we feel that this name is more apt given the microhabitat descriptors associated with that ecomorph: "under vegetation", "leaf-litter" and "passive shelter". Contrary to the ve ecomorphs from literature, we found our ecospace was best segmented into four clusters, where adding a fth cluster did not improve the model. Our classi cation of scorpion ecomorphs, therefore, lacks the ecologically "catch-all" Lapidicolous ecomorph. The Phytophilous ecomorph, positively associated with six different microhabitats, contains the most generalist species, followed by the Pelophilous ecomorph. This trend corroborates Prendini's [29] hypothesis that vegetation climbers and hard soil fossorials are microhabitat generalists. The Psammophilous and Lithophilous ecomorphs remain as habitat specialists, with strong correlations with a maximum of two microhabitats ( Figure 1, Table 1).
We acknowledge that these results, particularly on the number of clusters recovered and the species they consist of, are dependent on the choice and number of sampled taxa. We sampled 61 of the approximately 2.500 species of scorpions described: less than 3%. Although we strove for a broad sampling, families such as the species-rich radiation of the Vaejovidae and the Australo-Papuan radiations of Hormurids, Scorpiopids, and Urodacids, among others, were underrepresented. Moreover, the clustering results may be further in uenced by not including representatives of the troglophilous ecomorph, and future studies should attempt to include a su cient sample of this minor, yet more specialized, ecomorph.
Walking legs and pedipalps are rich in eco-covarying traits The strongly eco-covarying traits, highlighted in PLS axes (Figure 3), span two quite different anatomical regions: scorpions with relatively slenderer pedipalps and relatively longer legs in one extreme and scorpions with more robust pedipalps and shorter legs in the other. These shapes are the theoretical limits of morphological variation associated with sand fossorial or rock-dwelling microhabitats, respectively (but not rock crevice microhabitats). In fact, such rock eurytopes seem to be ecologically closer to the semilithophilous ecomorph sensu [29]. These trends, shown by the PLS, are relative. While in absolute terms, e.g., a loose sand burrowing scorpionid (a family of scorpions with typically robust pedipalps) may still have more robust pedipalps than, e.g., a rock-dwelling buthid, it will have slenderer pedipalps relative to other scorpionids (taxonomical relationships are depicted in Figure 2).
The extremities are of major importance for scorpions to interact with the environment: the walking legs operate on the substrate in locomotion, and the pedipalps are used to navigate, obtain prey, defend and mate [55-57]. Different microhabitats may pose different demands on the functions of these two anatomical modules. Nonetheless, pedipalps and walking legs also share many developmental pathways [58], potentially leaving them less free to vary with ecology. Developmental linkage is more evident for structures along the body axis, as they share many developmental pathways [58,59]. The ecomorphological paradigm predicts that the various ecomorphological associations uncovered here are likely functional, but this cannot be concluded from correlation alone. Nevertheless, these results can be an informed starting point for experimental validation of the functional value of those traits.
Based on the numerous mentions of the ecological role and morphological distinctiveness that the pedipalps and the "tail" (metasoma and telson) get in descriptions of scorpion ecomorphs, we hypothesized that in our dataset they would co-vary the most with ecology. To our surprise, variation in the metasoma and telson is not as strongly correlated with microhabitat use as the walking legs. As mentioned before, it looks like the appendages that directly interact with the substrate (walking legs) or to navigate through the environment (pedipalps), are more strongly associated with ecology. A potential reason for "tail" traits not emerging as part of a general ecomorphological trend in scorpions involves phylogenetic-speci c associations. In fact, the phylogenetic signal of the metasoma average height and width are the highest among all the morphological measurements; and the width of the telson vesicle within the ten highest values (Supplementary Table 5). Accordingly, when the phylogenetic history of scorpions is not taken into account, the pedipalps and both the metasoma and the telson are the anatomical regions with the strongest association with ecology (Supplementary Table 4). Our data shows that morphological variation in the metasoma and telson has a stronger phylogenetic signal than an ecological signal. [29] noted that the elongation of the metasoma in corticolous (i.e., phytophilous) scorpions was only veri ed in the Buthidae family. This observation highlights that morphological variation in the "tail" of scorpions may have a strong phylogenetic signal. Conversely, this may indicate that the ecomorphological importance of walking legs proportions may have been partly obscured by phylogenetic signal.
Because we found several morphological traits associated with ecology, the use of such characters for phylogenetic inference is not advised. For example, the individuals placed at the extremes of Figure 3 are Hadrurus arizonensis and Caraboctonus keyserlingi, two species from the same family, the Caraboctonidae. These species are, therefore, quite divergent in both ecology and morphology. Phylogenetically neutral morphological traits would place these two species closer together than with species from other taxonomic families.

Conclusion
This study is the rst broad quantitative approach to scorpion ecomorphs. We were able to identify associations between ecology and morphology, which transcend taxonomical relatedness. The ecomorphological paradigm predicts that the various associations we uncovered are likely functional, but correlation alone is not conclusive evidence. Nevertheless, these results can be an informed starting point for experimental validation of the functional value of those traits.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.     Ecomorph means ( lled dots) rotated to their principal components, and 95% con dence ellipses around them. Legend: Illustrations approximate the mean morphologies of each ecomorph, and colors correspond to ecomorph a liation.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.