Crura, the calcareous support structures of the lophophore in rhynchonellide brachiopods, have historically been used to justify higher-level rhynchonellide classification and reveal major evolutionary lineages within rhynchonellides. Seventeen crural types have been described and categorized into four groups based on variation in overall structure and cross-sectional shape, but not evaluated in a quantitative or comprehensive manner. Heterochrony has been hypothesized to play a role in the evolutionary transitions among some types, but the structural, developmental, and phylogenetic context for testing these hypotheses has not yet been established. In this study, we use three-dimensional geometric morphometric techniques to quantify morphological disparity among all six crural morphs in Recent adult rhynchonellides, with the goal of delineating more objective criteria for identifying and comparing crural morphs, ultimately to test hypotheses explaining morphological transformations in ontogeny and phylogeny. We imaged the crura of seven Recent rhynchonellide species, using X-ray computed microtomography. We used landmarks and semi-landmarks to define the dimensions and curvature of the crura and the surrounding hinge area. Procrustes-standardized landmark coordinates were analyzed using a principal component analysis to test the discreteness of the individual crural morphs and named groups of morphs, and to identify features that vary most among the crural configurations.
Our results demonstrate that microCT imaging techniques provide novel ways to investigate the morphology of small features that may be otherwise impossible to quantify using more conventional imaging techniques. Although we predicted overlap among crural morphs in the 3-D shape space, the principal component analyses suggest that five of the six crural morphs differ distinctly from one another. Some but not all previously designated crural groups appear to exhibit morphological cohesion. This study establishes a quantitative morphological foundation necessary to begin an investigation of the phylogenetic significance of ontogenetic changes in crura, which will allow hypotheses of heterochrony to be tested.
Crura, the prong-like, calcareous structures that support the lophophore on either side of the mouth, are often the most conspicuous morphological features of the interior of rhynchonellide brachiopod dorsal valves (Fig. 1). The crura support and position the base of the lophophore, allowing the lophophore to filter water efficiently as it enters the mantle cavity along either side of the commissure and exits at the valve anterior (Ager 1965; Rudwick 1970; Williams et al. 1997). A broad range of crural morphological variability exists—17 types have been named even though all rhynchonellides are characterized by only one lophophore type, the helically coiled spirolophe lophophore (Rudwick 1970; Williams et al. 1997; Savage et al. 2002). The morphological diversity among crura has historically been used to organize higher-level rhynchonellide classification (Manceñido 1998, 2000; Manceñido and Owen 2001; Savage et al. 2002; Manceñido et al. 2007), but it remains unclear how different named crural configurations are related morphologically, phylogenetically, or ontogenetically (Cooper 1959; Ager 1965; Rudwick 1970; Manceñido 1998, 2000; Manceñido and Owen 2001; Savage et al. 2002; Manceñido et al. 2007; Manceñido and Motchurova-Dekova 2010). A quantitative characterization of crural morphology would facilitate reproducibility in the naming of crural types (morphs) and in the identification of specimens with respect to crural type, and allow us to test proposed evolutionary patterns of crural transformation (Manceñido and Motchurova-Dekova 2010). It would also enable quantitative comparisons among adults, throughout ontogeny, and across phylogenetic hypotheses of relationship (Cohen and Bitner 2013; Schreiber et al. 2013). Using microCT technology, we obtained 3-D images of all six named crural types expressed in Recent rhynchonellides, and statistically analyzed three-dimensional geometric morphometric measurements of crura in order to evaluate the relationship between size and shape of crura in rhynchonellides of different body (shell) size, taxonomic affiliation, and phylogenetic affinity.
Crura in Rhynchonellida
Rhynchonellida originated in the Ordovician and is the second most diverse extant brachiopod order (after Terebratulida), with over 500 fossil and extant genera (Williams et al. 2000a,b; Carlson and Leighton 2001; Savage et al. 2002; Manceñido et al. 2007). Today, forty extant species are classified into 19 genera. They are distributed globally, but are most abundant and diverse in extra-tropical regions, specifically Australia and New Zealand (Savage et al. 2002; Logan 2007; Manceñido et al. 2007; Savage 2007). Although rhynchonellide (Kuhn 1949) brachiopods are the geologically oldest and putatively the phylogenetically most basal of the extant rhynchonelliforms, they are somewhat inconspicuous in today's oceans, with many living in patchy distributions at bathyal depths. Their apparent rarity in modern faunas makes numerous species difficult to collect in abundance, and consequently they are relatively understudied by neontologists and paleontologists. Rhynchonellide extant diversity—approximately 3% of their total Phanerozoic generic diversity—is severely diminished; however, their apparently basal phylogenetic position provides, among crown-clade articulate brachiopods, critically important comparative information about the evolution of more derived rhynchonelliform brachiopods (Cooper 1959; Ager 1965; Carlson 1995, 2007; Cohen and Gawthrop 1997; Manceñido and Owen 2001; Cohen 2001a,b, 2007; Carlson and Leighton 2001; Savage et al. 2002; Cohen and Weydmann 2005; Manceñido et al. 2007).
Extant rhynchonellides have a spirolophe lophophore (with the exception of Tethyrhynchia, which is trocholophous), in which the apices of the spires point dorsally and are supported posteriorly by crura (Williams et al. 1997; Savage 1996; Manceñido and Owen 2001; Savage et al. 2002). Their distinctive, roughly triangular shell morphology is often characterized by a strongly biconvex and costate shell in extinct forms, usually with a dorsal fold and ventral sulcus (though many today are rectimarginate and lack shell ornamentation). Extant adult rhynchonellides range in shell length from approximately 1 mm (e.g., Tethyrhynchia mediterranea) to 20 mm (e.g., Pemphixina pyxidata); compared to terebratulides, they are relatively small as adults.
Crura (singular: crus) are short (typically no more than 1 or 2 mm long), paired, rod- or prong-like calcareous processes (Fig. 1). Crura extend antero-ventrally from the inner socket ridge of the dorsal valve into the mantle cavity on either side of the mouth of the brachiopod, from which the lophophore arms extend, and serve as attachment sites for the body wall (Rudwick 1970; Brunton et al. 1996; Williams et al. 1997; Savage et al. 2002). Each crus supports the very proximal section of the lophophore directly adjacent to the mouth, while the remaining portion of the spirolophe is supported hydrostatically, lacking any additional mineralized support (Rudwick 1970; James et al. 1992). Because they typically are short, the crura act primarily as positioning devices for the lophophore rather than extensive support structures, and consequently, their geometry may affect the three-dimensional flow of water through the mantle cavity (Ager 1965; Rudwick 1970; Williams et al. 1997). However, the specific details of the relationship between crural morphology, lophophore geometry, and water flow patterns have yet to be studied (although see LaBarbera 1977, 1978, 1981; Emig 1992; Shiino et al. 2009; Shiino and Kuwazuru 2010). Crura vary morphologically in three primary ways: angle of projection into the mantle cavity, toward the ventral valve shell (curvature of the crus); shape of the distal tip of the crus (narrow or broad, digitate or not); and cross-sectional shape of the crus (straight or curved, and curved dorsally or ventrally). Crural morphs range from laterally to dorsoventrally compressed and can be either relatively straight or highly curved or twisted medially in a gentle helix (Fig. 2).
The crura begin to develop in juvenile rhynchonellides shortly after larval settlement (Long and Stricker 1991; James et al. 1992; Williams et al. 1997). Sheathed in outer epithelium, they consist of secondary shell material (Rudwick 1970; Williams et al. 1997) and develop from the inner socket ridge, growing by simple accretion to the distal end. Rudwick (1970) claims that crura grow through ontogeny without resorption of shell material, but this is a hypothesis that has yet to be tested. The tips of the crura may be elongated into the primary lamellae of spire-bearing brachiopods (e.g., extinct atrypides, athyridides, and spiriferides) or the descending lamellae of loops in terebratulide brachiopods (Williams et al. 1997); all are groups that have evolved from within a paraphyletic Rhynchonellida or share close common ancestry with them (Carlson 2007). Crura are thus an important component of the cardinalia of all crown-clade articulated brachiopods (Neoarticulata; Carlson 2012; Carlson and Cohen in press).
Over the past 150 years, 17 crural configurations have been named and have recently been placed into four qualitative groups (raducal, septifal, ensimergal, arcual) according to differences in overall structure and cross-sectional shape (Fig. 2) (Manceñido 1998, 2000; Savage et al. 2002; Manceñido et al. 2007; Manceñido and Motchurova-Dekova 2010). The depauperate Recent rhynchonellide fauna represents not only a small fraction of taxonomic diversity, but also a fraction of the morphological diversity of crura found in the geologic past (Manceñido and Owen 2001; Savage et al. 2002). Under the current classification, nine rhynchonellide superfamilies have the same crural type, whereas six superfamilies are characterized by multiple crural types including the four superfamilies with extant representatives (Cooper 1959; Ager 1965; Carlson and Leighton 2001; Manceñido and Owen 2001; Savage et al. 2002).
Do adult individuals within a single morph vary significantly in shape, or exhibit similar degrees of variability from morph to morph? After surveying rhynchonellide crural variation present in museum collections and literature sources (see complete list in Supplementary Table 1), we noted that slight qualitative shape variations in the crura, often found in only a few specimens, were used as the basis for naming new crural morphs—a fact that Cooper (1959) and Ager (1965) and others confirmed in their descriptions. A thorough comparative review of rhynchonellide crural morphs is called for because no consistent method has been used historically to identify, name, or group them, or to determine relationships among morphs or among groups of morphs. Arguably, the best way to achieve this revision is to use both qualitative and quantitative methods, as each can illuminate the other. Qualitative descriptions of individual crural morphs exist (see Rothpletz 1886; Wisniewska 1932; Cooper 1959; Ager 1962, 1965; Dagys 1968; Rudwick 1970; Baranov 1980; Savage et al. 2002; Manceñido and Motchurova-Dekova 2010) and include brief discussions of crural shape variability. However, these descriptions can vary from author to author depending on the particular specimens studied, revealing the need for quantitative analyses that can test hypotheses by using measurable data in a more objective and repeatable manner.
Our study is the first to undertake a quantitative analysis of crura in an effort to identify and classify the range of variability present in extant rhynchonellides. We have chosen to characterize Recent crural morphs by using computer generated three-dimensional surface models, which allow in-depth examination of very small crural features not easily seen with more conventional imaging and analytical methods. The 3-D surface models can be enlarged and manipulated fully in three dimensions to reveal multiple views of the crura from many perspectives (Fig. 3). We then use 3-D geometric morphometric and multivariate statistical analyses to quantify the morphological diversity within and among the six crural morphs present in Recent rhynchonellides. How distinct are these six morphs from one another and how are they related in size and shape? How does the raduliform morph, the stratigraphically oldest and most common crural (Savage et al. 2002) form, vary among adults? Because several different names have been given to the morphologically simple crura lacking quantitative analysis, we predict that crural morphs have been over-split and may occupy overlapping regions in three-dimensional shape space.
Materials and Methods
We selected extant rhynchonellide species for this initial morphometric study because crura can be imaged more precisely in three-dimensions with X-ray computed microtomography when the mantle cavity is entirely free of sediment. By using only Recent specimens, we also avoid complications from postmortem distortion of the crura, a confounding problem that will be examined in future studies. A minimum of three individuals of each of the six extant crural morphs (raduliform, falciform, arcuiform, canaliform, spinuliform, maniculiform; Fig. 3), from seven species, were selected from museum and marine laboratory collections, for a total of 23 adult rhynchonellides (see Supplementary Table 1 for a complete list). Specimens were examined from the National Museum of Natural History (Smithsonian Institution, Washington, D.C.), the California Academy of Sciences (San Francisco), Portobello Marine Laboratory (Portobello, New Zealand), University of California, Davis, and Scripps Institution of Oceanography (San Diego), and were either dried or preserved in 70% ethanol.
Images of the crura were obtained by using X-ray computed microtomography (microCT). Using X-rays, microCT scanners generate a series of digital, contiguous two-dimensional cross-sectional slices of an object by detecting differences in the attenuation of the X-rays as they pass through the object. Materials will scatter or absorb X-rays in direct relation to their density. A more dense material will appear more opaque in a microCT image than a less dense material (Elliot and Dover 1982; Flannery et al. 1987; Ketcham and Carlson 2001; Monnet et al. 2009; Shiino et al. 2009; Peck et al. 2009; Angiolini et al. 2010; Motchurova-Dekova and Harper 2010; Pakhnevich 2010; van Dam et al. 2011; Abel et al. 2012; Görög et al. 2012). The microCT scanner produces a series of sequential, adjacent two-dimensional images which, when assembled using computer software such as 3D Slicer ( http://www.slicer.org; Gering et al. 1999; Pieper et al. 2004, 2006), create a three-dimensional model of the object (Ketcham and Carlson 2001). These three-dimensional representations can then be easily manipulated digitally, by rotation in three dimensions, for ease of measurement and visualization of features. We imaged all specimens with the Scanco Medical microCT scanner located at the School of Veterinary Medicine at the University of California, Davis.The scanner is a desktop cone-beam microCT scanner with a nominal resolution of approximately 5–90 μm. Samples require no preparation and can be scanned either dried or preserved in alcohol. With this initial set of images of extant crura as a baseline, to establish proof of concept, we can then attempt to obtain images of fossil crura, from individuals preserved in sediment matrix of a range of densities. Individual image slices were assembled and surface models constructed using the software platforms Amira v. 5.2 or 3D Slicer v. 3.4. The surface models were then edited and enhanced in the program Raindrop GeoMagic Studio v. 10.0 to expose the crura and other internal features of the shell such as the sockets, hinge plates, and socket ridges (Fig. 1).
We used 3-D geometric morphometric techniques to quantify the disparity among the six crural morphs found in extant rhynchonellides. Landmarks, along with semi-landmarks, defined the dimensions of the crura, cardinalia, and the curvature of the crura (Fig. 4). A landmark is a discrete, geometrically homologous anatomical point that can be accurately identified on all individuals, whereas a semi-landmark is a constructed point on a geometric feature, often a curve or surface, defined by its relative position on that feature (Bookstein 1991; Zelditch et al. 2004). We defined nine homologous landmarks (Types 1 and 2 [Bookstein 1991]). Three-dimensional Cartesian coordinates were collected for all landmarks and semi-landmarks (Mitteröcker and Gunz 2002; Zelditch et al. 2004), using the morphometric program Landmark v. 3.6 (Wiley et al. 2007).
Crural curvature and the shape of the distal tip are important characteristics for defining crural morphs; therefore, we used semi-landmarks to delineate the curved areas of the crus (e.g., distal tip morphology; Bookstein 1997; Gunz 2001, 2005; Gunz et al. 2005; Mitteroecker and Gunz 2009) (see Table 1 for a complete description of all landmarks and semi-landmarks). Semi-landmarks allow information about the curvature of a feature to be incorporated into a geometric morphometric analysis (Zelditch et al. 2004). Each curve consists of three equally spaced semi-landmarks anchored by two landmarks. Bilateral symmetry allowed landmarks to be digitized on one crus per specimen, useful in cases in which one crus was damaged or broken off entirely.
Landmark and semi-landmark descriptions. Landmark type denoted. Each curve has a total of five points. Landmarks are illustrated in Figure 4.
Following data collection, we used the morphometric program Morphologika v. 2.5 (O'Higgins and Jones 1998), to perform a generalized Procrustes analysis (Gower 1975; Rohlf and Slice 1990), which removed any variation between sets of landmarks due to differences in scale, rotation, or translation. A generalized Procrustes analysis performs a Procrustes superimposition, which minimizes the Procrustes distance among all landmark configurations in the data set by using centroid size (Gower 1975; Rohlf and Slice 1990; Zelditch et al. 2004). The Procrustes-fitted coordinates served as input variables for multivariate statistical analyses. We first examined shape distinct from size, and later added size back into the analysis by comparing shape with centroid size of landmark and semi-landmark data.
We used multivariate statistical analyses to explore the nature of morphological variation among crural morphs in order to locate the areas of the crura that vary most among Recent morphs and to test statistically the morphological distinctiveness and examine within-morph variability of the six Recent crural morphs. A principal component analysis (PCA) was used to locate and explore areas of the crura that exhibit the most variability and to study the variation of landmark positions between the Recent crural morphs, allowing shape parameters that vary among crural morphs to be identified. The PCA of the measured variables was completed in the program PAST v. 1.94b (Hammer et al. 2001) with the variance-covariance matrix of the unstandardized data (i.e., the variance of the data is not standardized), allowing the areas of maximum shape variation to be identified. We also performed cluster analyses, both single linkage and neighbor-joining, based on the Euclidean distances between specimens, as measured using scores derived from the first three principal components of the PCA, in order to test whether individuals in the same crural type cluster together and whether different types cluster together.
We evaluated morphological variability within and among six crural morphs in adults of seven species (representing four superfamilies) of rhynchonellides, variability among the raduliform crura of adults of two species, and, to a more limited degree, variability within and among the arcual and raducal groups (Manceñido and Motchurova-Dekova 2010). Adult morphological variation of crural morphs was assessed using a PCA of all crura from adult rhynchonellides using a combination of landmark and semi-landmark data. Differences in crural shape have been deemed to be more important than absolute changes in size in naming crural morphs (Savage et al. 2002; Manceñido et al. 2007). Shape and orientation also appear to influence the way in which the crura contact and support the lophophore (Cooper 1959; Manceñido and Owen 2001; Savage et al. 2002; Manceñido et al. 2007). To assess within-morph variability, we performed a PCA on the Procrustes coordinates derived from specimens having raduliform crura (adult Notosaria nigricans and Hemithiris psittacea). Shape differences found among the crural groups designated by Manceñido and Motchurova-Dekova (2010) also were investigated using a PCA. Qualitative differences among the crural morphs, and the biological implications of those differences, were also evaluated.
Adult Morphological Variation
The principal component that accounts for the greatest amount of variation in this analysis, PC 1, is associated with landmarks and semi-landmarks that describe the ventral position of the medial edge of the crus (Fig. 5). The first three PC axes account for 64.83% of the total variance in the data: PC 1 accounts for 27.73%, PC 2 accounts for 25.37%, and PC 3 accounts for 11.73% (complete PC scores are available from the authors for all analyses). The morphological variation illustrated along PC 1 is associated with the width and the medial twist of the distal end of the crus relative to the proximal end (Fig. 5A). Falciform crura represent one morphological extreme with broad, medially convex crura. Arcuiform crura represent the opposite extreme with narrower crura twisted medially. All other crural morphs are concentrated around the origin, indicating that the width and twisting of the distal end of the crura dominate variation along PC 1. The morphological variation illustrated along PC 2 is associated with crus length and width and ventral curvature (Fig. 5A). Maniculiform crura represent one end-member with narrow, straight, and elongated crura. They are also the smallest crura in absolute size (Fig. 3). Canaliform represent the opposite end-member morphology with short, wide crura, and are among the largest crura that occur in the largest individuals. Crural morphs are more or less equally distributed along PC 2, indicating slight variations in crural width and length and ventral curvature from one end-member to the other. Morphological variation along PC 3 is associated with crural curvature and medial twisting (Fig. 5B); Hemithiris distal tips are horizontal; Frieleia are nearly vertical, and only slightly medially tilted. Variation along PC 3 ranges from relatively straight and laterally compressed spinuliform crura to dorsoventrally compressed, medially twisted, and ventrally curved in raduliform crura. Crural morphs are more or less equally distributed along PC 3, indicating slight variations in crural curvature from one extreme to the other. Semi-landmarks along the medial edge of the crus have a significant effect on the outcome of the analysis by capturing the variability of the medial edge of the crus and subsequently outweighing the variability associated with crural length. Without semi-landmarks, the variation of the medial edge among Recent crural morphs is not captured fully. This suggests that the shape and curvature of the medial edge, in all three dimensions, is particularly important for distinguishing Recent crural morphs. Delineating Recent crural morphs depends on the degree of medial twisting from proximal to distal ends of the crura, a transformation that is expressed ontogenetically.
Statistical analyses of landmark and semi-landmark coordinates for all adult individuals indicate that those with the same crural morph generally occupy a volume of morphospace that is smaller than the volume of morphospace that separates groups of different crural morphologies. The canaliform crural morph is an exception (Fig. 5), in that it consistently groups with the raduliform crura, supporting the grouping of both these crural morphs into the raducal group. The Euclidean distance between Notosaria and Hemithiris (calculated from the first ten PC scores), both considered to have raduliform crura, is as large as the distance between most crural morphs (Fig. 5). Canaliform crura overlap only with the raduliform crura of Notosaria, not those of Hemithiris (Fig. 5).
Major axes of shape variation are potentially related to size; therefore, we performed a multivariate regression analysis to test the degree of association between crural centroid size and the first three principal components of the landmark and semi-landmark analysis. The analysis shows that there is no general dependence between size and shape (adjusted R2 = 0.22; p = 0.05), but crural size and PC 1 are significantly correlated. The linear dependence of PC 1 on size indicates that it describes allometric size-related variation (among adults) among the crural morphs (r2 = 0.20; p = 0.03). The dependence, however, is not a uniform one among morphs, but instead is a function of the exceptional differences of the small-sized Neorhynchia profunda crura and the larger Basiliola lucida crura from an otherwise isometric similarity among the remaining taxa. Size is not correlated significantly with PC 2 (r2 = 0.12; p = 0.11) or PC 3 (r2 = 0.001; p = 0.87). Comparing a simple linear measure of crural length with overall shell length (Fig. 6A), it is clear that smaller individuals, in general, have shorter crura than larger individuals, as might be expected. And yet, the relationship between centroid size of the crural region and overall shell length among all species is not necessarily as clear; adults of species in some genera (e.g., Pemphixina) have a much different allometric relationship between crural region and shell length than closely related adults of the same shell length in other genera (Fig. 6B).
Previous authors (e.g., Rothpletz 1886; Muir-Wood 1934; Wisniewska 1932; Cooper 1959; Ager 1965; Savage et al. 2002) have noted the variable morphology of the raduliform morph, including variation in size, distal end morphology, and angle of curvature. We performed a second PCA of the landmark and semi-landmark coordinates of the raduliform crura in adult Notosaria nigricans and Hemithiris psittacea specimens only, to investigate within-morph variability among species (Fig. 7). PC 1 accounts for 64.93% of the total variance in the data. The crura of Notosaria nigricans are thicker and more robust than those in Hemithiris psittacea, even though they have the same curvature and distal tip morphology. This PCA, along with the Procrustes distance information, supports the results of the all-adult crural morph PCA (Fig. 5), which illustrates that the two raduliform species are as different from one another as are any two different morphs, as discussed earlier. It is unclear whether other morphs might exhibit comparable variability; additional species per morph are being investigated currently to test this possibility as are additional adults in other species with raduliform crura.
Crural “Cognate” Groups
We used the results of the PCA of landmarks and semi-landmarks on adults to test the morphological integrity of the four crural groups proposed by Manceñido et al. (2007) (Figs. 2, 5). PC 1 and PC2 (Fig. 5A) separate representatives of the four groups from one another; PC 1 and PC 3 separate the septifal and arcual groups from the others, but the ensimergal and some members of the raducal group overlap one another completely. Representatives of the arcual group (spinuliform and arcuiform crura) occupy two distinct areas of morphospace (Fig. 5). The raducal group (canaliform and raduliform crura) shows a similar pattern, with the greatest separation between the two raduliform species, as noted above. This suggests that these two groups are not necessarily morphologically cohesive and the variation between raduliform species is as great as, or greater than, that between two different morphs. However, the crural groups put forth by Manceñido et al. (2007) and Manceñido and Motchurova-Dekova (2010) appear to be grouped mainly according to hypothesized evolutionary transformations, not necessarily morphological cohesion, so it is perhaps not unexpected that the crural morphs placed in one group do not cluster in statistical space.
Cluster Analysis of Adults
Single linkage and neighbor-joining cluster analyses of the Euclidean distances between adults in principal component space consistently generated four main clusters (Fig. 8). Individuals with the same crural morph cluster together, as expected from the distributions in Figure 5, with one exception. Pemphixina (canaliform) clusters with Notosaria (raduliform), whereas Hemithiris (raduliform) clusters with Basiliola (falciform); these two clusters themselves cluster together more closely than do either of the other two clusters. Frieleia (spinuliform) and Cryptopora (maniculiform) form the third main cluster, and Neorhynchia (arcuiform) forms a cluster that is most dissimilar to all the others. The current landmark configuration was unable to capture the serrated distal end of the maniculiform crura, a feature that distinguishes them from all other crural types.
Summary of Results
Crura vary in their morphology among adults within a single species, genus, or superfamily, among adults in different species assigned to the same crural morph, and among adults assigned to different crural morphs. Adult individuals in the same species, having the same crural morph, typically cluster together in the shape space defined here. Groups of morphs recognized previously (Manceñido and Motchurova-Dekova 2010) are often but not always distinct from one another in this morphospace. Data on more species representing the only six extant crural morphs are needed to test this conclusion more rigorously.
Crura are a fundamentally important feature of all rhynchonellate brachiopods (sensu Williams et al. 1996; Williams and Carlson 2007) because they function to support the lophophore within the mantle cavity. Crura first appear, phylogenetically, in derived syntrophiidine pentamerides, the camerelloids and pentameridines, which share common ancestry with the Rhynchonellida (Carlson et al. 2002). Rhynchonellida is a large, ancient paraphyletic group from which the various spiralia-bearing and loop-bearing groups have evolved (Carlson 2007). Because crura form the structural base of both spiralia and loops, characterizing their morphological variability in ontogeny (and phylogeny) informs our understanding of the evolutionary history of Neoarticulata (Carlson and Cohen in press), the crown clade of articulated brachiopods. This study was designed as a preliminary test of the morphological integrity of named crura types (morphs) and the grouping of crural types into “cognate groups.” Our study provides a quantitative morphological foundation for more comprehensive tests of hypotheses of heterochrony (currently ongoing) that have been suggested to play a role in these evolutionary transformations (Manceñido and Motchurova-Dekova 2010).
The small size and delicate structure of crura have hindered detailed study of their morphology for many years. MicroCT imaging techniques provide novel ways to investigate the morphology of such very small features. Three-dimensional computer models have been generated from CT-scanned images of extinct spire-bearing brachiopods, from which physical models were made to investigate water flow through the mantle cavity (Shiino et al. 2009; Shiino and Kuwazuru 2010); however, our study is the first to quantify morphological variability among crura by using these techniques. The traditional method of serial sectioning (e.g., Ager 1965; Motchurova-Dekova et al. 2002; Savage et al. 2002; Manceñido and Motchurova-Dekova 2010) destroys shell material and thus informative morphological detail between each section, which makes it difficult to interpret the complex 3-D geometries of very small crura. Scanning electron microscopy (SEM), a common imaging technique, yields highly resolved, but static, 2-D views of crura. Furthermore, in order to capture an unrestricted SEM image of the cardinalia and crura, the valves must first be disarticulated, which can damage brachiopods like rhynchonellides with cyrtomatodont (interlocking) hinge structures (Jaanusson 1971; Carlson 1989). Three-dimensional surface models created from successive, closely spaced CT scans allow the digital capture and dynamic manipulation of the entire hinge area of the brachiopod in three dimensions without the need for disarticulation, so that more detailed quantitative and qualitative analyses can be undertaken.
Morphologic, Taxonomic, and Phylogenetic Variation Among Adult Crura
Morphologically, the crura of adult extant rhynchonellides vary mainly in five parameters: height, width, and length of each crus; degree of curvature of the entire crus, particularly along the dorso-medial edge; and the angle of divergence between the two crura (Fig. 3). Even small variations in these parameters may significantly affect the position and orientation of the spirolophe, and thus influence the 3-D geometry of water movement through the mantle cavity (see Ager 1965; Rudwick 1970; LaBarbera 1977; James et al. 1992; Williams et al. 1997). The particular functional significance of minor variations in position and orientation has not yet been investigated, and is not the focus of this study, but would yield interesting insights into patterns of water flow between the valves, and the effect of those differences on rhynchonellide feeding behavior among adults of different overall body size.
We studied multiple individuals per species, representing seven different species; individuals of the same crural morph (and species) clustered together, with six of the seven species clusters occupying a distinctly different region in the shape space constructed (Fig. 5). Taxonomically, this confirms the morphological integrity of six of the seven species with respect to crural morphology, as well as the morphological integrity of five of the six named extant crural morphs. This result suggests that our original prediction—that crural morphs had been oversplit—is not borne out among the six extant crural morphs. More individuals from additional extant (and extinct) species must be analyzed to test these preliminary conclusions, but most extant crural morphs appear to be quantitatively distinct from one another, and their relative position in morphospace is now established. There are exceptions to this pattern: two raduliform species analyzed are as different from one another as any two other crural morphs, and canaliform individuals largely overlap one of the two raduliform species clusters.
With respect to higher taxonomic affiliation, three of the four superfamilies form distinct morphological clusters separate from the others (Fig. 5). Three species in the superfamily Hemithiridoidea cluster relatively closely together in the morphospace, but the two species in Norelloidea do not, which indicates that crural morphology varies among extant representatives per superfamily. Several extinct superfamilies have been characterized by the same crural morph (raduliform), whereas others, particularly the four superfamilies with extant representatives, are characterized by multiple morphs, rarely including raduliform. These four superfamilies might have experienced a diversification in crural morphs from a raduliform ancestral state, which may have contributed to their evolutionary success. It is also possible that we are simply better able to image and study the diversity of these crural types because some are extant.
Phylogenetically, the raduliform crural morph (Figs. 3, 7) is the most basal (Schreiber et al. 2013) among all Rhynchonellida, extant and extinct; it is also the morph that appears to be the most variable morphologically among constituent species (given our limited sampling regime so far). It is the morph that first appears stratigraphically as well (Manceñido and Owen 2001; Savage et al. 2002). Very little is known about the nature of morphological variability (both within and among species) of the stratigraphically early raduliform crura—shape of the distal ends, angle of curvature, cross-sectional shape—owing to poor preservation and the difficulties of imaging crura in fossils; it has been questioned whether these early crura should even be considered raduliform (Ager 1965; Savage 1996; Savage et al. 2002). However, the presence of raduliform-like crura in many well-preserved pentameride brachiopods supports the basal phylogenetic position of raduliform crura among all the rhynchonellides (Carlson 1993; Carlson et al. 2002).
Among crown-clade (extant) Rhynchonellida only, the basal members of three of the four subclades recognized in morphological phylogenetic analyses possess spinuliform crura (Fig. 3F); the fourth, raduliform (Schreiber et al. 2013) (Fig. 3C). Molecular analysis of 12 species of extant rhynchonellides discovered three subclades (Cohen and Bitner 2013); basal members of each of these three subclades have either spinuliform or arcuiform (Fig. 3E) crura. Phylogenetic analyses using either type of data support similar ancestral character-state reconstructions of crural types among the extant taxa: spinuliform appears to be the ancestral crural morph. Raduliform crura are clearly the stratigraphically oldest and most common morph, suggesting that the spinuliform type evolved as a shared derived feature of the crown-clade Rhynchonellida. The nature of the evolutionary transition from raduliform to spinuliform crura has not yet been investigated morphologically or phylogenetically in detail, but is currently under investigation.
It is intriguing that the cluster analysis of adult crural morphology (Fig. 8) produces a branching pattern that is quite different from the current classification (Savage et al. 2002) and from both recent phylogenetic analyses (Cohen and Bitner 2013; Schreiber et al. 2013). The cluster analysis includes only features of crural morphology, however, whereas the classification and the phylogenetic analyses include essentially all morphological features or a large number of molecular characters simultaneously, so differences between them should perhaps be expected. Furthermore, the cluster analysis is purely distance-based, and takes no account of polarity determined from outgroups and the sequential acquisition of apomorphies that are suggested by a phylogenetic analysis.
Crural “Cognate” Groups
Manceñido and Motchurova-Dekova (2010) organized 15 of the 17 named crural types into four groups (Fig. 2): raducal, arcual, septifal, and ensimergal. There are two components to these groups: the assignment of types to a particular group, based generally and qualitatively on crural morphology; and hypotheses of morphological, developmental, and/or phylogenetic transformations between types. With respect to the first component, our main focus in this study, we predicted, on the basis of Manceñido and Motchurova-Dekova's (2010) work, that crural morphs in the same group would cluster together morphologically, and that crural groups would be separate from one another in the shape space constructed. Our results reveal that some morphs cluster together by group, but others do not. Crural morphs in three of the four named crural groups do occupy distinctly different regions of shape space (Fig. 5), but we investigated more than one morph in each of only two groups, necessarily so because our study focused on extant species, which represent only six of the 17 types. Raduliform and canaliform morphs cluster together as predicted (Fig. 8), but falciform morphs cluster with them as well, which is not consistent with our predictions. Spinuliform and arcuiform morphs do not cluster together, as we predicted that they would. Morphometric analyses of additional species representing each morph are clearly required to test the generality of these preliminary findings, but it appears that most (not all) crural groups are quantitatively distinct, supporting the morphological distinctions among these “cognate” crural groups.
As described by Manceñido and Motchurova-Dekova (2010), the configuration of crural groups provides a rich source of evolutionary hypotheses to test, many of which involve heterochrony, or the evolutionary consequences of changes in developmental rate or timing, leading to changes in size and shape from ancestor to descendant. Three distinct types of information are required in order to test hypotheses of heterochrony: qualitative and quantitative data on size and shape; data on the nature of and sequence of developmental transformations over ontogeny; and phylogenetic hypotheses that enable comparisons between putative ancestors and descendants (minimally, identification of sister-group pairs). As noted previously, this study is focused primarily on establishing a foundation based on the first of these three types of data.
Crura are a fundamentally important feature of all crown-clade articulated brachiopods because they function to support the lophophore within the mantle cavity. Crura form the structural base of both spiralia and loops, and studying their morphological variation can give us valuable insights into the evolutionary history of crown-clade articulated brachiopods. Our study provides a quantitative morphological foundation for more comprehensive tests of possible mechanisms (e.g., heterochrony) generating the evolutionary changes we see.
MicroCT imaging techniques provide novel ways to investigate the morphology of very small “hidden” features, such as the crura. Three-dimensional surface models created from CT scans allow the digital capture and dynamic manipulation of the entire hinge area of the brachiopod in three dimensions, thus enabling more detailed quantitative and qualitative analyses.
Morphologically, the crura of adult extant rhynchonellides vary primarily in five parameters: height, width, and length of each crus; degree of curvature of the entire crus, particularly along the dorso-medial edge; and the angle of divergence between the two crura. This study confirms the morphological integrity of six of the seven species with respect to crural morphology, as well as the morphological integrity of five of the six named extant crural morphs; extant crural morphs at least do not appear to have been oversplit. However, the two raduliform species analyzed are as different from one another as any two other crural morphs, and canaliform individuals largely overlap one of the two raduliform species clusters. Furthermore, three of the four superfamilies form distinct morphological clusters separate from the others. Stratigraphically and phylogenetically, the raduliform crural morph is the most basal among all Rhynchonellida, extant and extinct; it is also the morph that appears today to be the most variable morphologically among constituent species.
Crural morphs in three of the four named crural cognate groups occupy distinctly different regions of morphometric shape space, supporting the qualitative morphological distinctions among them, but sampling of additional species in the morphs and groups must be increased to test these preliminary results.
We thank R. Motani for helpful discussions and advice throughout this project, which grew out of one portion of the first author's dissertation research. Anonymous reviewers provided helpful suggestions that improved the quality of the manuscript. We also thank Scripps Institution of Oceanography, California Academy of Sciences, and D. E. Lee (University of Otago, New Zealand) for access to specimens. We thank J. Thompson and D. Levin (National Museum of Natural History) for providing access to specimens, and other assistance during research visits. Finally, we thank Tanya Garcia-Nolan (J.D. Wheat Veterinary Orthopedic Research Laboratory at the University of California, Davis School of Veterinary Medicine) for all microCT scanning of specimens. We gratefully acknowledge support for this project provided by National Science Foundation grant EAR 1147537. Support for this research was also provided by Durrell funds from the Department of Geology, University of California, Davis, and a Geological Society of America Graduate Student Research Grant to the first author.