Hinge and ecomorphology of Legumen Conrad, 1858 (Bivalvia, Veneridae), and the contraction of venerid morphospace following the end-Cretaceous extinction

Abstract The Veneridae are the most speciose modern family of bivalves, and one of the most morphologically conservative and homoplastic, making subfamily- and sometimes even genus-level classification difficult. The widespread Cretaceous genus Legumen Conrad, 1858 is currently placed in the subfamily Tapetinae of the Veneridae, although it more closely resembles the Solenoida (razor clams, Pharidae and Solenidae) in general shell form. Here we provide high-resolution images of the Legumen hinge for the first time. We confirm from hinge morphology that Legumen belongs in Veneridae, but it should be referred to incertae subfamiliae, rather than retained in the Tapetinae, particularly in light of the incomplete and unstable understanding of venerid systematics. Legumen represents a unique hinge dentition and a shell form—and associated life habit—that is absent in the modern Veneridae despite their taxonomic diversity. Veneridae are hyperdiverse in the modern fauna, but strikingly ‘under-disparate,’ having lost forms while gaining species in the long recovery from the end-Cretaceous extinction.


Introduction
The modern Veneridae are the most speciose family of the Bivalvia, with ∼750 species in 135 genera, outnumbering the next most diverse families, the Galeommatidae and Tellinidae, each with ∼500 species in 100 genera (Huber, 2015;Collins et al., 2018;Edie et al., 2018). Molecular analyses have established a backbone phylogeny for Veneridae (Mikkelsen et al., 2006;Chen et al., 2011), but species coverage within the group is still insufficient to confirm monophyly of some longestablished taxonomic groupings, such as subfamilies. A further complication (or source of interest) is the high level of homoplasy in infaunal bivalve shell form owing to the functional requirements of burrowing method and life habit. Consequently, systematics within the Veneridae has been unstable, with species frequently moved between genera and with a much-revised subfamilial classification system (e.g., Keen, 1969;Harte, 1998;Huber, 2010;Alvarez, 2019). The classification of Bivalvia in the ongoing revision of the Treatise on Invertebrate Paleontology demotes many historically accepted venerid subfamilies to tribes, some of which, including Tapetini (=Tapetinae), are thought to be paraphyletic (Carter et al., 2011), although subsequent studies continue to use subfamilies, and recent molecular work recovered a monophyletic Tapetinae Mikkelsen et al., 2006;Chen et al., 2011).
Despite their taxonomic richness, the modern Veneridae are not as morphologically diverse as other species-rich bivalve families (e.g., the Pectinidae and Tellinidae), particularly if specialized rock-boring Petricolinae (subtribe Petricolini in Carter et al., 2011) are omitted from consideration. Soft-bottom infaunal venerids are equivalve, inequilateral, and prosogyrous. Most of their variation in form derives from posterior elongation, but even in this they are conservative compared to other heterodont bivalves such as the Tellinidae or the razor clams in the superfamily Solenoidea, and Veneridae lack anteriorly elongate forms, such as occur in the fast-burrowing families Donacidae and Mesodesmatidae.
Legumen Conrad, 1858, a Cretaceous venerid genus, comprises ∼22 valid species, most of which are small, extremely posteriorly elongate, and very compressed, and thus a morphological outlier to the family. The genus has been placed in the Tapetinae since Stoliczka (1871; as "Tapesinae"), following the designation of Venus (Tapes) fragilis d'Orbigny, 1845 as the type of his new genus Baroda, subsequently synonymized with Legumen by Stephenson (1923). That synonymy was retained by Keen (1969, p. N682), who diagnosed Tapetinae as "Ovate to elongate, shell surface somewhat polished, inner margins smooth on at least posterior third; hinge plate narrow, with cardinals 3a entire, 3b normally entire, others frequently bifid; lateral teeth wanting." The first appearance of Legumen is given as 'lower Cretaceous' in Keen (1969), which may have been a reference to "Tapes" parallela Coquand, 1865 (Aptian) or to an unnamed Albian Angolan Legumen sp. (Rennie, 1929; see Dartevelle and Freneix, 1957). Since the Treatise on Invertebrate Paleontology, a potential Albian species has been named as L. iraniense Collignon, 1981, but this is based on a single poor specimen with no hinge preserved. Because of the extremely thin, delicate nature of the shell of most Legumen species, this is not an uncommon situation: few specimens are available in museum collections and even fewer with interior characters (i.e., hinge plate, teeth, muscle scars, pallial line, and sinus) preserved. Illustrations of internal characters in the literature are sparse for all known species. Our aim here is to provide, for the first time, high-resolution images of the hinge of this genus, and to place this unusual venerid in a broader context in terms of shell and hinge morphology. We find that the form of the dentition in Legumen, while it mostly conforms to Keen's broad diagnosis of Tapetinae, is very distinctive.
We use 3D micro-CT scans to digitally 'excavate' the hinge of two specimens of L. ellipticum Conrad, 1858, one embedded in matrix and the other articulated. This technique, along with a well-preserved single valve of L. ellipticum and two of L. carolinense (Conrad, 1875), has allowed us to include Legumen in a 3D morphometric dataset. Using landmark/semilandmark morphometrics on the interior surface of the shell and on hingeplate configurations, we quantify and compare the form and hinge position in Legumen to an exemplar species from each extant genus in Veneridae (excluding the rock-boring Petricolinae) and to the related and superficially Legumen-like heterodont families Pharidae and Solenidae, colloquially known as 'razorclams.' This analysis of form, plus close observation of hinge anatomy, suggests that Legumen fits poorly within the Subfamily Tapetinae (=Tribe Tapetini). This analysis also suggests that Legumen was a fast, vertically burrowing, deep-infaunal siphonate, silty-sand dweller similar to the smaller species of Pharidae. This represents a distinct life-habit within the Veneridae that was vacated at the Cretaceous-Paleogene boundary when Legumen went extinct, and has not been re-occupied by other, surviving lineages within the family.

Materials and methods
Specimens.-The dataset consists of micro-CT scans of 151 Recent and Cretaceous bivalve specimens (138 non-petricoline venerids, 10 pharids, and three solenids), representing all extant sediment-burrowing genera in the three families, plus a small number of morphologically representative Cretaceous specimens from the southeastern USA. We exclude the Petricolinae because they are a monophyletic and functionally restricted group that occupies a hard substratum, requiring special modifications of the shell. We treat bivalve subgenera as operational genera (hereafter "genera"), following general paleobiological practice (e.g., Sepkoski, 2002). Genera are represented by one valve of a single adult specimen, most often of the type species (where available), and usually a left valve. Right valves were operationally mirrored about the plane of the commissure for inclusion in shape analyses when a left valve was unavailable for study.  1). Other species of Legumen and other Cretaceous venerids (N = 114) that were unavailable to us for scanning are included in our dataset using published shell heights and lengths, which we use to compare them to all other taxa using the commonly used metrics of aspect ratio and size, as geometric mean of length and height, a metric closely correlated to centroid size (Kosnik et al., 2006).
Shape analyses.-Shapes of the interior surface of the shell are characterized using an automated semi-landmarking procedure adapted from the 'eigensurface' procedure of Polly and MacLeod (2008), following the method detailed in Collins et al. (2019). The interior surface of the shell best represents the general shell form used to infer life modes within the Bivalvia, without adding noise related to sculptural variation (Collins et al., 2019). A grid of semilandmarks is placed over the digitized 3D surface of the shell, superimposed across specimens using a Generalized Procrustes Analysis, and then summarized into a morphospace using Principal Components Analysis (PCA).
Hinge configurations are captured using a minimal set of user-defined landmarks designed to capture the mechanical shape of the hingeplate without relying on specifics of tooth homology, which is violated by the comparison of species from two distantly related superfamilies (or even simply inclusion of both left and right valves in the dataset). The first three landmarks describe the beak and cardinal teeth, delineating the area of the hinge that acts to reduce shear during active burrowing. In order to capture variation in the position of this cardinal area relative to the rest of the hinge area, four more landmarks are included: the junctions of the adductor muscle scars with the hingeplate and the anterior-posterior-most points on the shell commissure parallel to the vector joining the adductor muscle landmarks (i.e., the vector commonly used to define the shell length). These anterior and posterior commissure points were computationally determined by finding the largest magnitude vector connecting two points along the shell commissure that is parallel to the vector defined by the two adductor muscle scar landmarks (Fig. 2).

Results
The Appendix contains descriptions of the morphology and dentition of Legumen ellipticum and L. carolinense, and additional images of their dentition (Appendix Fig. 1). Maastrichtian and the Recent (although the Veneridae first appear in the Jurassic and the Tapetinae in the Albian, these occurrences are omitted from the plot for clarity). In the Maastrichtian, these other species are encompassed wholly within the same area of the size/aspect ratio space as the rest of the Veneridae, and in the Recent, they are scattered, but still Each specimen is illustrated as the whole valve with scalebar (10 mm) and an oblique close-up of the hingeplate (below, with a 2.5 mm scalebar) to facilitate comparison of dentition. Note that in Legumen ellipticum, unlike in the other modern and fossil Tapetinae illustrated here, teeth 1, 2b, and 3b are not bifid, and tooth 4b is extremely elongate. In Legumen carolinense, tooth 3b only is bifid, and the bifurcation is unlike that of true Tapetinae in orientation and character (see text). Confirmed occurrences of Legumen begin in the Cenomanian, where specimens are already small and elongate relative to most of the modern venerids, and their morphospace expands to include a less elongate, more 'veneriform' extreme (Fig. 3). By the Campanian, the full range of aspect ratios known from this genus is in place, from the "veneriform" Legumen ooides (Gabb, 1864) ("L. o". in plots) to the "phariform" Legumen ellipticum. By the Maastrichtian, diversification and/or improved sampling shows a wide range of Legumen forms between these extremes, clearly separable from the rest of the Veneridae.
Shell form morphospace.- Figure 4 illustrates the shell form morphospace (the shapes of the interior surface of the shell), Principal Component (PC) 1 describes 49% of the total variance in shell form, mostly anteroposterior elongation of the shell with a small component of inflation; equilateral species such as Tivela tripla (Linnaeus, 1771) plot at high positive values, and posteriorly elongate forms such as the Solenoida plot at extreme negative values. PC2 describes 16% of the total variance, dominantly shell height and inflation, with tall and inflated forms plotting at extreme negative values and low, compressed forms at high positive values. It is immediately apparent that the Veneridae and the Solenoida are well separated in this morphospace-except for Legumen ellipticum (Conrad, 1858), which plots close to Siliqua and Neosiliqua (Pharidae). Its congener, L. carolinense (Conrad, 1875), is much more like a "typical" venerid in outline, although notably compressed compared to other Tapetinae of similar outline. The callocardiine venerid Macrocallista nimbosa (Lightfoot, 1786) also plots in this space. Macrocallista is elongate, but considerably larger and more inflated than Legumen.
Hinge morphospace.- Figure 5 shows the morphospace of the cardinal area relative to the hingeplate and overall shell elongation ("hinge geometry"). PC1 describes 39% of the variance in shape, mostly posterior elongation (unsurprising, given that the vast majority of heterodont bivalve shape variation is related to this trait), but also relative size of cardinal area: shells with large cardinal areas have high PC1 scores and are more equilateral overall. Shells with small cardinal areas have low PC1 scores and tend to be posteriorly elongate. PC2, which explains 15% of the variance, includes a component of cardinal area size, but dominantly describes the rotation of the cardinal area relative to the hingeplate-the cardinal area being delineated by a triangle where the "apex" (relative to the shell) is the proximal end of whichever cardinal tooth is closest to the beak (in practice, almost always the anterior cardinal 2a), and the "base" of the triangle is formed by the distal end of the anterior cardinal and the distal end of the posterior cardinal. Low PC2 scores indicate a small cardinal area with a base subparallel to the line of the hingeplate, and high scores on PC2 indicate a larger cardinal area with a base at ∼45°to the line of the hingeplate. In this hinge geometry space, a pattern similar to the shell morphospace ( Fig. 4) emerges, where L. ellipticum groups with the Pharidae and L. carolinense with the Veneridae, but in this space various other venerids (including non-tapetine M. nimbosa), also approach the Solenoida.

Discussion
Evaluation of the position of Legumen within the Tapetinae.-These new images of the dentition of species assigned to Legumen suggest two genus-level groups of species have been included under this name. A revision of the group is accordingly now in progress-however, we accept monophyly faute de mieux in this study because the species included are undoubtedly each other's closest relatives, regardless of how many genera are involved (Fig. 1). Hinge tooth differences strongly suggest that all species currently assigned to Legumen should be excluded from Tapetinae (Fig.1). The narrow ligamental area and extreme elongation and perfect straightness of the posterior cardinal (teeth 4b [LV], 3b [RV]) even in the more "veneriform" L. carolinense contrasts with the hinge morphology of modern Tapetinae, in which a broad, curved ligamental area has a slightly raised edge and the posterior cardinal is confined to be short, and directed closer to subvertical. The striking, widely separated cardinals are described as bifid in both Legumen ellipticum and L. carolinense by Stephenson (1923Stephenson ( , 1941, but despite examining and scanning some of the specimens he illustrated ( Fig.1.1-1.6), we cannot confirm this in L. ellipticum, and only cardinal 3b (usually entire in Tapetinae according to Keen, 1969) is bifid in L. carolinense. Cardinal teeth 2a and 2b in Legumen are very dissimilar in orientation and form to the bifid cardinals of other Tapetinae (see Fig.1.4-1.6, 1.10-1.12 Legumen versus Fig. 1.16-1.18, 1.22-1.24 other Tapetinae) where a single tooth is grooved or divided only part of the way down to the hingeplate (Fig.1). A selection of  Fig. 1.20, 1.23) are also illustrated in Keen (1969), all of which fit the diagnosis of Tapetinae better. The grouping of Cretaceous tapetines Cyclorismina, Cyclorisma, and Amakusatapes with other Cretaceous venerids in the size/aspect ratio plot (Fig. 3) suggests that the group was well established by mid-Cretaceous time and highlights the unusual character of Legumen. Developmentally based tooth homologies and more densely sampled molecular analyses are needed to truly understand the systematics of modern Tapetinae and their precursors. We suggest that Legumen is certainly a venerid, but placement in the Tapetinae is difficult to defend morphologically. Given the currently uncertain view of subfamilies in the Veneridae and the ongoing molecular work attempting to build a more stable systematic understanding of the family (e.g., Lemer et al., 2019 and references therein), we refer Legumen to Veneridae incertae subfamiliae, rather than erect a monogeneric subfamily for it. We contend that Legumen represents a venerid lineage that was lost in the K/Pg extinction. The similarities of form between Legumen and the Solenoida are due to convergence, promoted by similar modes of life, as discussed in the next section.
Ecomorphology of Legumen and the contraction of shell disparity within the Veneridae.-The intermediate position of Legumen between the Solenoida and the rest of the Veneridae in all three morphospaces  indicates that, whereas the genus contains members that are similar in outline to modern infaunal venerids, although distinctively compressed (e.g., L. carolinense, L. ooides), it also includes a number of species that represent a form-group that is entirely absent in the modern venerid fauna. We will devote most of this discussion to the ecomorphology of the scanned representative of this group, Legumen ellipticum Conrad, 1858. Bivalve shell morphology is tightly linked to life habit (Stanley, 1970(Stanley, , 1975Seilacher and Gishlick, 2014); anecdotal examples of convergence in form can be seen in even a glance at any large collection of shells. Shell anatomy can be broken down into a number of functional structures related to life habit, such as overall form and sculpture, which interact with the substratum, and the ligament, adductor muscles, and hinge teeth, which operate together to open and close the shell. The position of the hinge relative to the muscle scars and the overall longest axis of the shell controls both the orientation of the shell during the burrowing period, and the amount of "rocking" the animal is capable of during a burrowing sequence.  Figure 2, not on the full specimen shown (for the morphospace of the full internal shell-shape, see Fig. 4). The lower panel shows the simulated shapes for points across the space. The central configuration (shown at [0,0] and enclosed in a box) is the mean shape. Note that in this morphospace the Veneridae are much closer to the Solenoida, and that Legumen bridges the space between groups. Point shapes and colors are as given in the legend to Figure 3. Macrocallista nimbosa is indicated ("M. nm."), though not illustrated (see Fig. 4 for an image) as a point of comparison.

Journal of Paleontology 94(3):489-497
Legumen ellipticum is strikingly close to the smaller Pharidae (e.g., Siliqua) in shell form and hinge geometry, and given the strong relationship between shell form and life habit, we infer similar life modes for the two genera. The reduced cardinal area in L. ellipticum, which would have much diminished resistance to shear during burrowing compared to, for example, Dosinia, which are fast burrowers, but have large cardinal areas, suggests that, like the pharids, its burrowing method did not have a significant rocking component, so that it did not need to expend resources on structures for counteracting shear.
Pharids and the related solenids have an anterior pedal gape, which allows them to burrow almost vertically in semipermanent tubes, and to make fast escapes from their burrows if disturbed (Stanley, 1970). Lack of a pedal gape indicates that Legumen were likely not tube dwellers, but their thin, smooth shell, small size, and long posterior all suggest adaptation for unconsolidated, muddy sediment (Stanley, 1970), and the small size and truncation of the posterior suggests a more vertical final life position than that occupied by Macrocallista nimbosa, which is otherwise the most comparable venerid. Limited literature exists on the life habits and ecology of the smaller Pharidae, but L. ellipticum plots most consistently with Siliqua, Neosiliqua, and Sinonovacula, all of which live buried subvertically in shallow-marine or estuarine, muddy or silty sand substrata, and L. ellipticum evidently lived in similar sediments (e.g., Sohl and Koch, 1983, 1984, 1987Ebersole, 2016).
Modern Veneridae occupy a wide variety of habitats today across most available shelf sediments. Many species of venerid co-occur, with co-existence perhaps mediated by predation pressure, by micro-partitioning their habitats, or simply because competition is low between filter-feeders (e.g., the striking diversity of Chioninae in western North America; Roopnarine 2001; Coan and Valentich-Scott, 2012). Despite this species richness, in the 66 Myr since the K-Pg extinction, no living venerids have recapitulated the shell form of Legumen to re-occupy this area of morphospace and exploit this life-habit, currently occupied by the smaller Pharidae. Cretaceous venerids were less diverse taxonomically, with 21 known genera in the upper Maastrichtian , but encompassed a considerable range of shell forms, including the one described and illustrated here that has been lost to the family. Diversitydisparity relationships are an important aspect of biodiversity (Jablonski, 2017), and the current "under-disparity" of this hyperdiverse bivalve family is a mystery that warrants further study.

Conclusions
The Cretaceous venerid Legumen exhibits a hinge morphology that despite superficial similarity is not closely comparable to the modern subfamily Tapetinae, and represents a shell-form not seen in the modern venerid fauna. Based on morphological comparison to the superfamily Solenoida, we hypothesize that Legumen was a fast burrower, adapted for living subvertically in soft, muddy sediment. No Cenozoic or extant species of venerid exhibits this combination of morphological features. Legumen appears to represent both a lineage and a unique mode of life for venerids that was lost in the K/Pg extinction.