High Diversity, Low Disparity and Small Body Size in Plesiosaurs (Reptilia, Sauropterygia) from the Triassic–Jurassic Boundary

Invasion of the open ocean by tetrapods represents a major evolutionary transition that occurred independently in cetaceans, mosasauroids, chelonioids (sea turtles), ichthyosaurs and plesiosaurs. Plesiosaurian reptiles invaded pelagic ocean environments immediately following the Late Triassic extinctions. This diversification is recorded by three intensively-sampled European fossil faunas, spanning 20 million years (Ma). These provide an unparalleled opportunity to document changes in key macroevolutionary parameters associated with secondary adaptation to pelagic life in tetrapods. A comprehensive assessment focuses on the oldest fauna, from the Blue Lias Formation of Street, and nearby localities, in Somerset, UK (Earliest Jurassic: 200 Ma), identifying three new species representing two small-bodied rhomaleosaurids (Stratesaurus taylori gen et sp. nov.; Avalonnectes arturi gen. et sp. nov) and the most basal plesiosauroid, Eoplesiosaurus antiquior gen. et sp. nov. The initial radiation of plesiosaurs was characterised by high, but short-lived, diversity of an archaic clade, Rhomaleosauridae. Representatives of this initial radiation were replaced by derived, neoplesiosaurian plesiosaurs at small-medium body sizes during a more gradual accumulation of morphological disparity. This gradualistic modality suggests that adaptive radiations within tetrapod subclades are not always characterised by the initially high levels of disparity observed in the Paleozoic origins of major metazoan body plans, or in the origin of tetrapods. High rhomaleosaurid diversity immediately following the Triassic-Jurassic boundary supports the gradual model of Late Triassic extinctions, mostly predating the boundary itself. Increase in both maximum and minimum body length early in plesiosaurian history suggests a driven evolutionary trend. However, Maximum-likelihood models suggest only passive expansion into higher body size categories.


Introduction
The origin of Plesiosauria 200 million years ago (Ma) was a landmark event in tetrapod evolution. Spanning 135 Ma, plesiosaurians represent one of only three long-lived radiations of secondarily marine, non-mammalian tetrapods (alongside ichthyosaurs and marine turtles; duration .100 Ma). Plesiosaurians possess an unusual body plan not seen in other marine vertebrates [1]. All four limbs are enlarged and modified as propulsive flippers, the trunk is short and stiff, and proportional head size seems to vary inversely with neck length, resulting in a gradation of forms between 'plesiosauromorph' (long neck, small head) and 'pliosauromorph' (short neck, large head) extremes [2][3]. Despite its oddity, this body plan was an extraordinarily successful adaptation to life in the open ocean [1,4]. It secured plesiosaurian survival for the entire Mesozoic, in spite of the emergence of possibly competing marine tetrapods clades [5], and recurrent regression events, which decimated shallow marine tetrapod lineages [6], including basal representatives of Sauropterygia, the wider clade that includes Plesiosauria.
Despite its significance for understanding the macroevolutionary dynamics of adaptive radiations, the fossil record of the oldest plesiosaurs has been relatively understudied. This contrasts with early records of other pelagic tetrapod clades, the ichthyosaurs [7][8][9][10] and cetaceans [11][12][13][14]. Lower Jurassic (200-175 Ma) European deposits yield the earliest plesiosaurian fossils, possessing their full complement of pelagic adaptations. Abundant specimens provide snapshots of the emergence of Plesiosauria in the earliest Hettangian of Somerset and Leicestershire, UK (,200 million years ago [Mya] [15][16]), the Sinemurian of Dorset and Leicestershire UK (,197-190 Mya [17][18]) and the lower Toarcian of Yorkshire and Northamptonshire, UK, and Baden-Württemberg, Germany (,183-180 Mya [19][20]). The occurrence of multiple intensively-sampled horizons provides unique data on key biotic and ecological parameters associated with invasion of the open ocean, including body size, species richness, and morphological disparity. Furthermore, global ecosystems were in decline for much of the Late Triassic, resulting in a progressive diminution of diversity in many invertebrate clades reviewed by [21], and possibly culminating in Late Triassic extinctions, or a Triassic-Jurassic boundary extinction event (e.g. [22]). The exact cause and modality of these extinctions is uncertain [21][22]. However, the Lower Jurassic rise of plesiosaurians is one aspect of global recovery from this episode.
We present the results of a comprehensive review of the littlestudied earliest Jurassic plesiosaurian fauna from Street, Somerset, UK (and an adjacent, contemporaneous locality at Watchet, Somerset), and a new phylogenetic dataset focussed on Lower Jurassic taxa. These are used as tools to study the earliest stage of plesiosaurian evolution.
Our taxonomic revision of the plesiosaurian fauna from Street indicates the presence of six species, including two new taxa identified here for the first time (below, Systematic Palaeontology). In total 16 plesiosaurian specimens from Street are taxonomically determinate (Table 1; nine additional specimens are indeterminate). Penecontemporaneous strata at Watchet, Somerset yield an additional new taxon (below; Systematic Palaeontology). Other localities of similar age at 'Street-on-the-Fosse' (Pylle), Somerset and Barrow-on-Soar, Leicestershire have yielded additional specimens of 'Rhomaleosaurus' megacephalus [15][16], a specimen of T. hawkinsii is also known from Walton, Somerset [35].
The presence of seven species in the lowermost Hettangian indicates high taxic diversity at a single locality immediately following the Triassic-Jurassic boundary. This is higher than the number of species known from the Late Hettangian-Sinemurian of Lyme Regis and Charmouth, Dorset and the lower Toarcian of Holzmaden, Germany and vicinity (both localities have yielded five species; Appendix S1); it is lower than the number of species known from the lower Toarcian of the UK (eight species; Appendix S1), and from the lower Toarcian of Europe (Germany and the UK combined: 13 species; Appendix S1). However, Toarcian plesiosaur faunas contain multispecific genera that inflate species counts, and the number of genera known from the lower Toarcian of Europe is in fact comparable to that from Street (seven genera).

Nomenclatural Acts
The electronic version of this document does not represent a published work according to the International Code of Zoological Nomenclature (ICZN), and hence the nomenclatural acts contained in the electronic version are not available under that Code from the electronic edition. Therefore, a separate edition of this document was produced by a method that assures numerous identical and durable copies, and those copies were simultaneously obtainable (from the publication date noted on the first page of this article) for the purpose of providing a public and permanent scientific record, in accordance with Article 8.1 of the Code. The separate print-only edition is available on request from PLoS by sending a request to PLoS ONE, 1160 Battery Street, Suite 100, San Francisco, CA 94111, USA along with a check for $10 (to cover printing and postage) payable to ''Public Library of Science''. In addition, this published work and the nomenclatural acts it contains have been registered in ZooBank ,http://zoobank.org., the proposed online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ''http://zoobank.org/''. The LSID for this publication is: urn:lsid:zoobank.org:pub:5D0B1E45-4D2D-43F7-9480-F85C9832F815. Holotype. OUMNH (Oxford University Museum of Natural History, Oxford, UK) J.10337, a skull and partial postcranial skeleton including anterior cervical and pectoral vertebrae, a partial hindlimb and ilium ( Fig. 1) from Street, Somerset, UK (likely lowermost Hettangian; see above; Plesiosaurian faunal composition at Street).

Systematic Palaeontology
Etymology. Genus name from 'Strate', the name for Street recorded in the Doomesday Book and sauros, Greek meaning lizard. Specific epithet after Michael A. Taylor, who carried out acid preparation of OUMNH J.10337 during an earlier study of Street plesiosaurians [16].
Referred specimens. AGT (Alfred Gillett Trust, Street, UK) 11, a skull, and GSM 26035 ( [39], pl. 19), a skull with anterior cervical vertebrae. GSM 26035 possesses prominent posterior processes on the posterolateral surfaces of the prezygapophyses, an autapomorphy of S. taylori (below; Diagnosis). AGT 11 is difficult to distinguish from Avalonnectes arturi (below), for which only the postorbital skull is known. However, in AGT 11, GSM 26035, and probably OUMNH J.10337, the premaxilla terminates posteriorly at approximately orbital midlength, whereas it extends further posteriorly in A. arturi. Furthermore, in AGT 11 and OUMNH J.10337 the squamosals contact each other dorsal to the parietals, unlike in A. arturi, in which this contact is more posterior. AGT 11 is otherwise indistinguishable from OUMNH J.10337 and GSM 26035 and is thus referred to S. taylori.
Short description of holotype. The skull is crushed dorsoventrally, but has been prepared using acid and is almost free of matrix (Fig. 1A, B). It shows many details that are only briefly described here. The snout tapers anteriorly and lacks a rostral constriction. Thus, the premaxilla is not transversely expanded, unlike many robust Lower Jurassic taxa, including 'Rhomaleosaurus' megacephalus [15]. The premaxilla contains five teeth of subequal size, as in many basal plesiosaurs, but unlike T. hawkinsii, which has four premaxillary teeth [35]. The maxilla contains 16 weakly heterodont alveoli that terminate ventral to the postorbital bar. A superficial flange of the maxilla extends dorsally posterior to the external naris (Fig. 1A, B). This was figured as a distinct ossification contacting the naris, and identified as a nasal, by O'Keefe ( [40], fig. 4). However, it does not contact the external naris, and we interpret the suture at the base of the 'nasal' (of O'Keefe [40]) as a crack. The parietal bears a low sagittal crest posterior to the suboval pineal foramen. The posterior part of the parietal is abruptly expanded laterally, forming the 'lateral angle' of Smith & Dyke [41] that is present in rhomaleosaurids and more basal pistosaurians (e.g. [42]). A dorsoventrally thin, sheet-like anterior extension of the squamosals overlaps the posterior surface of the parietals. A rounded squamosal bulb is present and the squamosals contact one another dorsal to the parietals. The posteroventral process of the postorbital extends far posteriorly, forming the dorsal margin of the temporal bar for most of its length ( Fig. 1A-B). This also occurs in basal sauropterygians [42][43] and rhomaleosaurids (e.g. R. thorntoni: Natural History Museum, London, UK [NHMUK] R4853). Contrastingly, in pliosaurids and plesiosauroids the length of the posteroventral process is reduced [35,44]. Eighteen anterior-middle cervical vertebrae are preserved, including most of the atlas-axis complex. The neurocentral sutures are convex, as in E. antiquior n. sp. (below), but unlike the other taxa present at Street. The cervical centra have double-headed rib facets and concave anterior and posterior surfaces that are approximately 1.2 times as broad mediolaterally as high dorsoventrally (Fig. 1C, D, E). A small 'lip' projects ventrally from the anterior surface (Fig. 1E). This has been considered to be a unique synapomorphy of derived pliosaurids, known from the Middle Jurassic onwards [45][46]. However, during the present study it was also observed in a specimen referred to 'Plesiosaurus' macrocephalus Owen, 1840 from the Lower Lias Group of Lyme Regis, Dorset (NHMUK 49202 [47]), which also has transversely broad cervical centra. Uniquely in S. taylori, prominent processes project posteriorly from the posterolateral surfaces of the anterior cervical prezygapophyses (Fig. 1C). These are confirmed as absent in most plesiosaurian taxa represented at Street: 'R.' megacephalus (LEICS G221.1851), E. arcuatus (NHMUK 2047*), T. hawkinsii [16], Avalonnectes arturi and Eoplesiosaurus antiquior (see below). The condition cannot be determined in 'Plesiosaurus' cliduchus, for which only posterior cervical vertebrae are preserved. However, S. taylori has proportionally short pectoral centra (Fig. 1G), unlike all other plesiosaurians from Street, including 'P.' cliduchus (Sedgwick Museum of Earth Sciences, University of Cambridge, Cambridge, UK [CAMSM] J.35180).
The ilium of S. taylori is approximately straight in lateral view and has a broadly expanded dorsal end that is rotated approximately 20 degrees relative to the proximal end (Fig. 1F). A rugose, proximodistally oriented ridge is present on the posterolateral surface of the distal expansion ( Fig. 1H-I). This is also present in 'R. megacephalus (LEICS G221.1851). A complete description of all material referred to S. taylori is beyond the scope of the present study but will form the basis of future work.
Etymology. Genus name from Avalon, an island from the legend of King Arthur, often identified with Glastonbury, near Street, and nectes, Greek meaning 'swimmer'. Species epithet after Arthur Cruickshank (1932Cruickshank ( -2011, who with M. A. Taylor initiated the restudy of British Lower Jurassic plesiosaurians in the 1990s,  and also as a reference to the legendary King Arthur of British folklore. Referred specimen. AGT uncatalogued, a partial postcranial skeleton from Street, Somerset, UK (likely lowermost Hettangian; see above; Plesiosaurian faunal composition at Street).
Short description of holotype. The postorbital portion of the skull is preserved. It has been embedded in wax so the dorsal surface is exposed ( Fig. 2A-B). The tapering posterior processes of the premaxillae extend far posteriorly compared to those of S. taylori ( Fig. 1A-B), almost to the level of the postorbital bar. However, they do not contact the parietal. The lateral surface of the maxilla is visible on the right side ( Fig. 2A-B). It underlaps the jugal and has been displaced posteriorly by crushing. The jugal is penetrated by numerous small foramina and contacts the squamosal just posterior to the postorbital bar. As in S. taylori and other basal pistosaurians, the posteroventral process of the postorbital is long, extending more than two-thirds the length of the temporal bar. The postfrontal-postorbital suture is oriented anteroventrally, allowing only a small exposure of the postorbital in the posterior rim of the orbit. The parietal bears a low sagittal crest, which is penetrated anteriorly by a suboval pineal foramen. The posterior part of the parietal is abruptly expanded laterally ( Fig. 2 A-B), forming the 'lateral angle' that is present in S. taylori and other basal pistosaurians. A low squamosal bulb is present.
Twenty-three cervical vertebrae are preserved ( Fig. 2D-E). These do not include the atlas-axis complex, so A. arturi possessed at least 25 cervical vertebrae. Because of the very small size of the anteriormost preserved centrum (13 mm long anteroposteriorly), it is unlikely that more than one or two additional cervical vertebrae were originally present. Thus, A. arturi has an estimated cervical count of 26-28, fewer than in T. hawkinsii (exactly 31 cervicals in all four sufficiently-complete referred specimens; [35], table 2), but similar to the number in the larger-bodied 'R.' megacephalus (28 cervicals, trunk length = 1820 mm; New Walk Museum and Art Gallery, Leicester, UK [LEICT] G221.1851). The cervical rib facets have two articular surfaces separated by a narrow horizontal groove. The neurocentral sutures are V-shaped in lateral aspect ( Fig. 2D-E), as in E. arcuatus (NHMUK 2047*) and 'P.' cliduchus (CAMSM J.35180). This is unlike the condition in S. taylori and Eoplesiosaurus antiquior (see below), which have rounded neurocentral sutures. It is also unlike the condition in T. hawkinsii, in which the neurocentral suture contacts the rib facet (contra Benson et al. [35]; this also occurs in the Toarcian pliosaurid Hauffiosaurus [48]). The posterior cervical neural spines of A. arturi are tall, more than 1.4 times the height of the centrum (Fig. 2D-E), as in E. arcuatus and 'R.' megacephalus. This is unlike the condition in 'P.' cliduchus, S. taylori and T. hawkinsii, in which the posterior cervical neural spines are subequal to the centrum height. The holotype of A. arturi has only 18 dorsal vertebrae (identified on the basis of rib morphology [35]), and 19 are present in the referred specimen (AGT uncatalogued). This count is low compared to T. hawkinsii (22 dorsals [35]) and 'R.' megacephalus (22-23 dorsals; LEICT G221.1851). The complete dorsal series is not known in other plesiosaurians represented at Street. The apices of the dorsal neural spines of A. arturi alternate in morphology between being transversely thin and sheet-like, and transversely thicker (Fig. 2C). This was also observed in E. arcuatus, and some rhomaleosaurids, including Meyerasaurus victor ( [49], pl. 10), and 'R.' megacephalus (LEICT G221.1851) during the present study. A qualitatively different alternating morphology was reported in leptocleidid plesiosaurians [50].
A. arturi possesses four sacral vertebrae, with short, robust ribs that converge laterally (Fig. 2D-E). Sixteen proximal-middle caudal vertebrae are preserved. They have flat ventral surfaces, widely-spaced chevron facets, and the dorsal portion of the caudal rib facet is formed by the neural arch. The scapular blade expands dorsally (Fig. 2D-E). It lacks a prominent posterodorsal kink in the outline in lateral view, unlike in E. arcuatus (NHMUK R1317) and T. hawkinsii (GSM 26035). It lacks the ventral projection at the distal end of the scapular blade of 'P.' cliduchus ( [51], fig. 14). The ilium is approximately straight in lateral view, and has an anteroposteriorly expanded distal end similar to S. taylori and many basal plesiosaurians. However, unlike in S. taylori, a deep fossa bounded anteriorly by a prominent proximodistally oriented ridge is present on the medial surface ( Fig. 2D-E). The femur is approximately straight and slightly more expanded posterodistally than anterodistally (the left femur has a symmetrical distal expansion). The tibia is slightly longer than the fibula.  (Fig. 3) from Watchet, Somerset UK (likely lowermost Hettangian; see above; Plesiosaurian faunal composition at Street).
Etymology. Genus name from Eo, Greek meaning 'dawn', and Plesiosaurus, referring to a proportionally long neck, which is also present in Plesiosaurus. The species epithet is Latin meaning 'more ancient', in reference to the old geologic age of the specimen and Duria antiquior ('a more ancient Dorset'; 1830), a watercolour depiction of Lower Jurassic fauna, including plesiosaurs and ichthyosaurs, by Henry De la Beche.
Diagnosis. Proportionally long-necked plesiosaurian (ratio of neck:trunk length.1.2) with at least 38 cervical vertebrae and one autapomorphy: small, conical lateral projections on the bases of the anterior cervical prezygapophyses.
Short description of holotype. TTNCM 8348 is an almost complete postcranial skeleton, although many features are obscured by matrix (Fig. 3). Thirty-six cervical vertebrae are preserved. These do not include the atlas-axis complex, so at least 38 were present originally, more than in A. arturi (estimated 26-28 cervicals), 'R.' megacaphalus (28 cervicals; LEICT G221.1851) or T. hawkinsii (31 cervicals [35]). The cervical rib facets are divided by a narrow horizontal groove. A small, conical eminence projects from the base of the prezygapophyses in anterior cervical vertebrae of E. antiquior (Fig. 3B). This differs from the posterior projection of the prezygapophyses of S. taylori (Fig. 1C) and is an autapomorphy of E. antiquior. Unlike in A. arturi, E. arcuatus, 'P.' cliduchus (CAMSM J.35180), S. taylori and T. hawkinsii, only the anteriormost 12 preserved cervical neural spines curve posterodorsally. More posterior neural spines are inclined straight posterodorsally (Fig. 3). The dorsal vertebrae are enclosed in matrix. Thirty-three possible caudal vertebrae are preserved, although a few distal elements may be missing, and it is possible that the anterior few possible caudals are sacral vertebrae; this is difficult to determine due to incomplete preparation. The ventral surfaces of the middle caudal centra are exposed. They bear prominent chevron facets posteriorly.
The pectoral girdle is preserved, although the clavicle-interclavicle complex is incompletely prepared and difficult to observe, and the coracoids are covered by the humeri (Fig. 3). The right scapula is partly visible. It lacks the ventral projection of the distal blade that is present in 'P.' cliduchus ( [51], fig. 14). The humerus curves posterodistally. A prominent longitudinal ridge is present on its anterior surface (Fig. 3), as observed in E. arcuatus and 'R.' megacephalus by Smith & Dyke ( [41], character 92). A possible ilium is covered by the right ischium. The left pubis is approximately as long anteroposteriorly as it is wide mediolaterally. Fore-and hindflippers are well-preserved and partly articulated (Fig. 3). A small postaxial ossicle is present between the right ulna and ulnare.
Microcleididae   Diagnosis. Plesiosauroids possessing: widely separated posterior cervical rib facets (character 123.1 in the Phylogenetic analysis; below); posteriormost dorsal rib facets split between centrum and neural arch ('sacralised', but bearing a dorsal rib) (146.1); medial surface of the iliac blade anteroposteriorly concave (178.0; also present in some rhomaleosaurids); and a prominent flange extends anteriorly from the proximal half of the radius (197.1; also present in Hauffiosaurus and some rhomaleosaurids). This diagnosis focuses on unambiguous postcranial synapomorphies because cranial material of basal microcleidids is unknown.

Phylogenetic analysis
We constructed a new phylogenetic data matrix based on a thorough review of all previously published characters and direct observation of most Lower Jurassic, European plesiosaurians. This resulted in analysis of 207 characters (109 cranial, 98 postcranial; including 46 new characters; Appendix S2) and 32 taxa (5 outgroups+24 Lower Jurassic ingroup taxa+3 Middle Jurassic taxa; Appendix S1), 27 of which were examined directly. Two Lower Jurassic British plesiosaurians known from relatively complete remains were not included in the analysis: Sthenarosaurus dawkinsi from the Toarcian of Yorkshire is incompletely described [53] and was not examined; the holotype of 'Plesiosaurus' macrocephalus is a young juvenile and was not included, but a referred specimen, NHMUK 49202 [47], of currently uncertain taxonomic affiliation (e.g. [56]) was included. 'Plesiosaurus' cliduchus is currently difficult to access and was not examined closely or included in the phylogenetic analysis. Plesiospterys wildi O'Keefe 2004 [57] was scored separately from Seeleyosaurus guilelmiimperatoris (Dames, 1895) [58] to test the hypothesis that they are conspecific [20].
The matrix (Appendix S3) was analysed in PAUP* 4.0b10 for Macintosh [59] using 500 random addition replicates with TBR branch swapping, saving an unlimited number of trees at each step. Bremer support was calculated using the same search strategy, implemented using the 'Decay Index PAUP File' function of MacClade [60]. Yunguisaurus is likely the most basal pistosaurian included in our analysis [61] and was formally designated as the outgroup.
The analysis resulted in 42 most parsimonious cladograms, each 604 steps long with an ensemble consistency index (CI) of 0.4343, retention index (RI) of 0.5983 and rescaled consistency index (RC) of 0.2744. The strict consensus is well-resolved. Although basal pistosaurians form a polytomy, this results from the uncertain phylogenetic position of the skull of Pistosaurus. When Pistosaurus is pruned from the set of most parsimonious cladograms, Bobosaurus and Augustasaurus are resolved as successive outgroups to Plesiosauria (Fig. 4). A surprising, novel topology is recovered, in which Pliosauridae and Plesiosauroidea form a clade (Neoplesiosauria sensu Ketchum & Benson [56]) excluding Rhomaleosauridae. Many previous classifications, and all phylogenetic analyses have united Pliosauridae with rhomaleosaurids in a monophyletic 'Pliosauroidea', sometimes also including Cretaceous clades such as Polycotylidae (e.g. [40,44,52,56,62]). Neoplesiosauria receives only moderate branch support here (decay index = 2). However, it is supported by 12 synapomorphies, of which seven are unambiguously optimised: the presence of a short posteroventral process of the postorbital (character 35.2), a mediolaterally narrow parietal vault, lacking the 'lateral angle' (38.0), posteromedian ridge of the supraoccipital absent (58.1), parasphenoid terminates within the anterior one-third of the posterior interpterygoid vacuity forming diamond-shaped ventral platform (64.0), surangular transversely narrow and 'blade-like', lacking a prominent medial crest and dorsomedial fossa (99.2), distal anteroposterior width of scapular blade subequal to width at midlength (161.0; not expanded), humeral shaft has pronounced dorsodistal curvature (185.0). Strong character support for Neoplesiosauria contrasts with a previously-hypothesised 'Pliosauroidea', comprising Pliosauridae and Rhomaleosauridae, most proposed synapomorphies of which are actually plesiomorphies [35]. NHMUK 49202 was found basal to Rhomaleosauridae on the phylogenetic 'stem' leading to Neoplesiosauria as in Ketchum & Benson (2010 [56]). It possesses several plesiomorphies not found in more derived plesiosaurians, including: the premaxilla does not divide the anterior processes of the frontals (14.0); the occipital condyle lacks a constricting groove around its base, even ventrally (49.2); and the paraoccipital process is inclined dorsolaterally relative to the ventral surface of the exoccipital-opisthotic, resulting in a narrow cranioquadrate passage (54.1).
Specimens from Street and other earliest Jurassic localities include basal plesiosauroids (Eoplesiosaurus) and pliosaurids (Thalassiodracon), but most represent rhomaleosaurids (Stratesaurus, Avalonnectes, Eurycleidus, 'Rhomaleosaurus' megacephalus). Branch support for nodes within Rhomaleosauridae is low, but Pliosauridae, Microcleididae, and nodes within both families, are generally wellsupported (a monophyletic Microcleidus comprising M. homalospon-dylus, M. brachypterygius, and M. tournemirensis received exceptionally strong support [decay index = 6]). This is part of a temporal pattern in the distribution of branch support. Older nodes are generally less well-supported, probably reflecting poor knowledge of Triassic pistosaurians: 12 nodes within Plesiosauria occur prior to the Triassic-Jurassic boundary (i.e. probably within the Triassic) and have an (mean) average decay index of 1.33; six younger nodes occur within or prior to the Sinemurian and have an average decay index of 1.67; and seven younger nodes occur prior to or within the Toarcian and have an average decay index of 2.86. Most nodes within Rhomaleosauridae split in the Triassic, whereas most pliosaurid and plesiosauroid nodes split in the Jurassic.

Phylogenetic diversity estimates
'Phylogenetic diversity estimates' were calculated for the lower Hettangian, Sinemurian and Toarcian (Fig. 5A), counting both taxon occurrences and 'ghost lineages' implied by the phylogeny [63]. All Hettangian-Sinemurian operational taxonomic units (OTUs) are monotypic genera (or single specimens) whereas Toarcian OTUs include multiple species within Hauffiosaurus, Microcleidus and Rhomaleosaurus. Thus, genus-and species-level phylogenetic diversity estimates are identical until the Toarcian. 'P.' cliduchus (lower Hettangian) and Sthenarosaurus dawkinsi (Toarcian) were not included in our phylogeny but were added to the phylogenetic diversity estimate. This resulted in phylogenetic diversity estimates of 13 in the lower Hettangian; 12 in the Sinemurian and 10 (genera) or 16 (species) in the lower Toarcian. We prefer the genus-level estimate because German and British representatives of Hauffiosaurus (other than H. longirostris) and Microcleidus are not contemporaneous and may represent chronospecies that did not exist contemporaneously [48]. The lower Toarcian diversity estimate may be underestimated due to the absence of well-sampled horizons in the immediately following stages (few early Middle Jurassic plesiosaurians are known). However, unlike older taxa, lower Toarcian plesiosaurians are known from multiple, intensely sampled horizons in both Germany and the UK, likely compensating for this effect. These results confirm that plesiosaurian diversity immediately following the Triassic-Jurassic boundary was comparable to that later in the Lower Jurassic, including up to 20 million years later in the early Toarcian.

Character disparity
Faunal disparity was estimated using our cladistic data matrix, following the protocol described by Wills et al. (1994 [64]) and employed by several recent studies of tetrapod disparity (e.g. [65][66][67]). Pairwise dissimilarity between all Lower Jurassic taxa was calculated as the proportion of comparable cells (those cells not scored as '?' for either taxon) that differed between the taxa. Taxa known only from postcranial material such as Eoplesiosaurus, Eretmosaurus and Westphaliasaurus could not be compared with Hauffiosaurus longirostris, which has 100% postcranial missing data (a postcranial skeleton probably belonging to the holotype [68] was not scored). Thus, H. longirostris was deleted from the matrix before further analyses to eliminate problems of missing entries in the resulting symmetric dissimilarity matrix. This matrix formed the basis of a principal co-ordinates analysis (PCo) in Ginkgo (Universitat de Barcelona, http://biodiver.bio.ub.es/ginkgo/) using a negative eigenvalue correction (Cailliez method). Two disparity metrics were calculated for three narrow Lower Jurassic stratigraphic intervals (earliest Hettangian; Sinemurian; lower Toarcian): the mean pairwise dissimilarity (as used by e.g. Wagner 1997 [69]), and the sum of variance of scores on all 24 principal coordinate axes. The sum of variances was chosen instead of product-or range-based metrics, because it is robust to variation in sample size [66,70]. It was computed using the freeware program RARE (M. Wills, pers. comm.), which also allows 95% confidence intervals to be computed using rarefaction. Both character-based disparity measures show monotonic increase through the Lower Jurassic (Fig. 5B).
The scree plot exhibits a break in slope between the third and fourth PCo axes, following which, each axis describes 5.73% or less of the Gower-transformed dissimilarity. The first three principal coordinate axes encompass 38.1% of the Gowertransformed dissimilarity and are plotted in Figure 6

Body size evolution
Trunk length was used as a proxy for overall body size (Fig. 5C), defined as the distance between the anterior surface of the anteriormost vertebra with an elongate, 'dorsalised' rib and the Figure 5. Early evolution of Plesiosauria. Plots of A, phylogenetic diversity [62]; B, disparity (main pairwise dissimilarity and sum of variances of PCO axes with 95% confidence intervals [64]); C, body size, based on the proxy trunk length (in metres); D, body proportions, based on the proxy neck:trunk length ratio. In C-D, triangles represent non-neoplesiosaurian plesiosaurians (mainly rhomaleosaurids), squares represent pliosaurids and circles represent plesiosauroids; unfilled shapes represent earliest Hettangian taxa, grey-filled shapes indicate late Hettangian-Pliensbachian taxa, and black shapes indicated Toarcian taxa. doi:10.1371/journal.pone.0031838.g005 posterior surface of the posteriormost sacral vertebra. The appearance of ribs with a 'dorsal' morphology (long, curving, rod-like ribs) was used because these ribs define the trunk directly, forming its lateral wall. Other criteria for identifying the necktrunk boundary rely only on indirect proxies such as rib facet morphology. Only measurements from the largest individuals of each taxon were counted, and taxa only known from juveniles were not included in our analysis (Plesiopterys, Rhomaleosaurus zetlandicus; indicated by rounded, incompletely ossified margins of slowly-ossifying bones such as the limb girdles [Plesiopterys], ulna and tarsals (Plesiopterys and R. zetlandicus; compare with figures in [71]), and dorsal neural spines in R. zetlandicus; SMNS 16812; The Yorkshire Museum, York, UK [YORYM] G503). Minimum body size in the Triassic-Jurassic boundary fauna, as indicated by the proxy trunk length (680 mm, Thalassiodracon) is comparable to that among Triassic pistosaurians (680 mm, Yunguisaurus [61]) and low compared to later intervals (Plesiosaurus [Sinemurian], 1100 mm; Seeleyosaurus [Toarcian], 900 mm). Maximum body size also increases through time. The largest individual of 'Rhomaleosaurus' megacephalus from the earliest Hettangian fauna has a trunk 1820 mm long compared to 1500 mm for the Triassic pistosaurian Bobosaurus [72] and 3400 mm for the Toarcian Rhomaleosaurus thorntoni (or 2460 mm for R. cramptoni). Sinemurian representatives of the large-bodied rhomleosaurid lineage are not currently known. Increase in the maximum and minimum values through time is suggestive of a driven trend of size increase [73].
We tested the hypothesis of a driven trend by fitting maximum likelihood models of trunk length evolution representing stasis, Brownian motion (BM), and BM+trend, to our phylogenetic tree with temporal branch lengths estimated using stratigraphic age. This analysis was conducted in R version 2.10.1 [74], using both untransformed, and ln-transformed data. Taxa for which trunk was not known, and taxa occurring after the Lower Jurassic, were not analysed. Branch lengths were calibrated following the protocol of Brusatte et al. [65] using code available from http:// www.graemetlloyd.com/pubdata/functions_2.r (accessed 23 December 2011). Evolutionary models were fitted using the 'fitContinuous' function of the Geiger package [75] as described by Hunt & Carrano [76], specifying a trait standard deviation of 55 mm, based on four individuals of Thalassiodracon hawkinsii. The results indicate overwhelming support for the BM+trend model when untransformed trunk length is analysed (Table 2). However, ln-transformed trunk length fits all tested models subequally well (Table 2). Thus, the appearance of driven trend in the untransformed data may arise from overweighting of extreme large sizes in taxa like Rhomaleosaurus thorntoni, and the true signal cannot be distinguished from white noise or Brownian motion.

Body proportions
Neck length excluding the skull was measured for the same individuals measured for trunk length (above). The ratio of neck length:trunk length was used as a proxy for overall body proportions (Fig. 5D). Other studies have used larger sets of measurements to quantify body plans [2][3]. The approach employed here was selected to maximize taxonomic coverage as it can be applied to some relatively incomplete specimens (e.g. those missing limbs and skulls).
Body proportions, as indicated by the proxy neck length/trunk length, occupy a narrow range in the earliest Hettangian fauna
Earliest Jurassic plesiosaurians possess their full suite of adaptations for specialised marine locomotion: reduction of intralimb flexibility, hyperphalangy, interlocking phalanges, and the appearance of supernumary ossiciations in the limbs resulting in the formation of a flipper; increased thoracic rigidity by modification of the limb girdles to form large ventral plates, and enlargement of the gastral basket; and shortening of the tail [1]. More basal, Triassic sauropterygians, including basal pistosaurians, possessed only a subset of these adaptations, demonstrated different swimming dynamics, and were restricted to shallower water facies [1,61,93].
Our data allow an assessment of the evolutionary response of plesiosaurs to full pelagic adaptation in the Lower Jurassic. They show that the advent of plesiosaurian locomotion was followed by a gradual increase in body size spanning approximately 17 million years between the Triassic-Jurassic boundary and the lower Toarcian. Thus, the largest lower Toarcian plesiosaurian, Rhomaleosaurus thorntoni, has a trunk almost twice as long as 'Rhomaleosaurus' megacephalus from the lower Jurassic fauna (Fig. 5C). If linear dimensions scale isometrically in rhomaleosaurids, this implies an eight-fold increase in body mass (this is inexact, but provides an estimate of the order of magnitude of the change). This trend of increasing maximum body size continued in later plesiosaurian evolution; the largest pliosaurids, known from fragmentary Late Jurassic remains, were substantially larger than R. thorntoni. For example, 'Stretosaurus' ( = Pliosaurus) macromerus has a humeral length of 840 mm and femoral length of 960 mm [90], compared to 720 mm and 680 mm in R. thorntoni (NHMUK R4853). Minimum body size also increased through the Lower Jurassic, though less prominently (Fig. 5C). Seeleyosaurus, the smallest Toarcian plesiosaurian, has a trunk length 1.32 times that of Thalassiodracon, the smallest plesiosaurian from the lowermost Jurassic. It is noteworthy that small-bodied individuals and taxa (Thalassiodracon, Avalonnectes, Stratesaurus) are abundant in the earliest Jurassic fauna, but rarer in the Toarcian (Fig. 5C). However, analyses of ln-transformed trunk length data suggest that the pattern of body size increase in early plesiosaurian evolution cannot be distinguished from a Brownian motion, or 'passive expansion' model (Table 2). This is similar to the pattern observed in Mesozoic birds [94], and during early dinosaur evolution [95], suggesting that driven trends of body size change ('Cope's Rule') do not always occur during Mesozoic ecological radiations.
The range of plesiosaurian body plans also increased through the Lower Jurassic (Fig. 5D). Although more basal pistosaurians exhibit high variation in cervical vertebral counts (e.g. 20 in Bobosaurus [72], 44 anterior to the pectoral girdle in Yunguisaurus [61]), this is not reflected in their overall body proportions, as even Yunguisaurus has an 'intermediate' neck length that is subequal to that of the trunk [61]. The earliest plesiosaurians include taxa with intermediate and short neck lengths, as well as Eoplesiosaurus, with a neck at least 1.2 times the trunk in length (Figs 3, 5D). The range of body proportions increased progressively through the Lower Jurassic, and Toarcian plesiosaurians include long-necked taxa such as Microcleidus tournemirensis, in which the neck is 1.9 times the trunk in length [96]. These body proportions are similar to those of some Late Jurassic plesiosaurians, such as the cryptoclidid Muraenosaurus [40,97], but Cretaceous elasmosaurids exhibited substantially longer necks (e.g. ratio = 2.75 in 'Alzadasaurus pembertoni' ( = Styxosaurus snowii) [98]).

Diversity dynamics in Lower Jurassic plesiosaurian evolution
The first appearance of plesiosaurians in the earliest Jurassic is characterised by a high diversity of rhomaleosaurids (five taxa). Despite the absence of derived plesiosaurians in older deposits, rhomaleosaurids from the earliest Jurassic are widely spread across morphospace (Fig. 6), and represent nodes deeply nested within Rhomaleosauridae. Thus, most rhomaleosaurid divergence likely occurred in the Late Triassic. In contrast, although the basal neoplesiosaurian split between Pliosauridae and Plesiosauroidea probably occurred in the Triassic, most nodes within these clades occur in the Jurassic, and both clades make their earliest Jurassic first appearance at low diversity (one taxon each). Plesiosauroidea became increasingly diverse and disparate through the Early Jurassic, especially at small-medium body sizes (Fig. 5C). This diversification may have been at the expense of rhomaleosaurids and represents a clear example of a candidate clade replacement event [99]; although the earliest rhomaleosaurids occupy the entire range of contemporaneous body sizes and are widely spread across morphospace, later (Toarcian) rhomaleosaurids are exclusively large-bodied (trunk length.1.5 metres) and are restricted to a local area of morphospace (Fig. 6). Although all rhomlaeosaurids have proportionally short necks (Fig. 5D), the youngest rhomaleosaurids have the shortest necks, and thus the most extreme 'pliosauromorph' body plans. This may be associated with a restriction of rhomaleosaurids to a macropredaceous niche prior to their final demise after the Middle Jurassic (last occurrences: Maresaurus, Bajocian [100]; Borealonectes, Callovian [101]).

Disparity dynamics in Lower Jurassic plesiosaurian evolution
Despite their high diversity (Fig. 5A), the earliest Jurassic plesiosaurians are characterised by low disparity (Fig. 5B). In fact, mean pairwise dissimilarity (the number of character differences among comparable cells) is approximately half that 17 million years later in the lower Toarcian. Plesiosaurian disparity based on character data and the range of body proportions represented (see above) both increase monotonically through the Lower Jurassic. This mode of disparity accumulation, coupled with more or less constant species diversity, differs from that shown in the origins of major groups such as the Cambrian 'explosion' of metazoan disparity (e.g. [102]) and the rapid evolution of early tetrapod disparity in the early Carboniferous [103]. Although further analyses of plesiosaurian disparity over a longer time interval are required to confirm this pattern, the results obtained here suggest that invasion of morphospace by plesiosaurs following invasion of pelagic habitats was gradual, not explosive, and comparable to that in other tetrapod subclades such as dinosaurian and crurotarsan archosauromorphs [65].

Late Triassic marine tetrapod extinctions
The observation of high taxic diversity in earliest Jurassic plesiosaurians, especially among rhomaleosaurids, suggests that an extinction event precisely at, or shortly before (,100-150 kyrs [22]) the Triassic-Jurassic boundary had little effect on Jurassic plesiosaurians. It is impossible to date these rhomaleosaurid divergences accurately, but generally low pairwise dissimilarity exhibited by earliest Jurassic plesiosaurians suggests relatively recent divergence, and may represent the influence of earlier Late Triassic extinctions, which are concentrated at the end of the Norian [21]. Our understanding of the effect of Late Triassic extinctions on plesiosaurian evolution will remain incomplete until the discovery of more complete Late Triassic plesiosaurian fossils.
Thorne et al. [67] attributed a monumental decimation of ichthyosaur disparity between Late Triassic and Lower Jurassic time bins to a Triassic-Jurassic boundary extinction event. However, the 'Late Triassic' time bin of their analysis includes six taxa from the Carnian, 23.4 million years before the end of the Triassic (235-228 Ma [104]). Only four taxa represent the younger Norian stage (28 Ma;, and zero represent the latest Triassic Rhaetian (2.4 Ma; 204-201.6 Ma). Clearly, this coarse level of temporal and taxonomic resolution is too poor to attribute a Late Triassic decline in ichthyosaur disparity to a sudden extinction event at the Triassic-Jurassic boundary. Indeed, a progressive, not precipitous, decline is suggested by the staggered last appearances of other marine reptile clades including thalattosaurs (Carnian, early Late Triassic) and basal sauropterygian clades such as nothosauroids (Ladinian, Middle Triassic) and pachypleurosaurs (Carnian) [105]. This decline occurred concur-rently with lowering sea levels and a reduction in shallow marine area [4,6,106]. A gradualistic, or stepwise pattern of extinction is also exhibited by well-sampled marine invertebrate clades, including bivalves, conodonts, and ammonoids, which show intervals of concentrated extinction prior to the end of the Triassic, most notably at the end of the Norian (e.g. [21,[107][108][109][110][111]) and throughout the Rhaetian [112].

Supporting Information
Appendix S1 Measurements, age, provenance data and lists of specimens examined for taxa included in the phylogenetic analysis. [XLS] Appendix S2 Character list for the phylogenetic analysis. [DOC] Appendix S3 Phylogenetic data matrix. [PDF]