Diverse Aquatic Adaptations in Nothosaurus spp. (Sauropterygia)—Inferences from Humeral Histology and Microanatomy

Mid-diaphyseal cortical bone tissue in humeri of Nothosaurus spp. consists of coarse parallel-fibered bone, finer and higher organized parallel-fibered bone, and lamellar bone. Vascular canals are mainly arranged longitudinally and radially in a dominantly radial system. Blood vessels are represented by simple vascular canals, incompletely lined primary osteons, and fully developed primary osteons. Nothosaurus spp. shows a variety of diaphyseal microanatomical patterns, ranging from thick to very thin-walled cortices. In the early Anisian (Lower Muschelkalk), small- and large-bodied Nothosaurus spp. generally exhibit bone mass increase (BMI). In the middle to late Anisian (Middle Muschelkalk) small-bodied nothosaurs retain BMI whereas larger-bodied forms tend to show a decrease in bone mass (BMD). During the latest Anisian to early Ladinian (Upper Muschelkalk), small- and few large-bodied nothosaurs retain BMI, whereas the majority of large-bodied forms exhibit BMD. The stratigraphically youngest nothosaurs document five microanatomical categories, two of which are unique among marine amniotes: One consists of a very heterogeneously distributed spongy periosteal organization, the other of very thin-walled cortices. The functional significance of the two unique microanatomical specializations seen in large-bodied nothosaurs is the reduction of bone mass, which minimizes inertia of the limbs, and thus saves energy during locomotion. Transitions between the various microanatomical categories are rather gradual. Our results suggest that small-bodied Nothosaurus marchicus and other, not further assignable small-bodied nothosaurs seem to have been bound to near-shore, shallow marine environments throughout their evolution. Some large-bodied Nothosaurus spp. followed the same trend but others became more active swimmers and possibly inhabited open marine environments. The variety of microanatomical patterns may be related to taxonomic differences, developmental plasticity, and possibly sexual dimorphism. Humeral microanatomy documents the diversification of nothosaur species into different environments to avoid intraclade competition as well as competition with other marine reptiles. Nothosaur microanatomy indicates that knowledge of processes involved in secondary aquatic adaptation and their interaction are more complex than previously believed.


Introduction
Sauropterygia was a diverse group of diapsid marine reptiles that existed from the late Early Triassic until the end of the Cretaceous [1,2]. Their Triassic radiation was thought to be restricted to the near-shore habitats of the Tethys Ocean and connected epicontinental seas. It primarily involved shallow marine forms such as Placodontia, Pachypleurosauria, Nothosauroidea, and Pistosauroidea, the latter three clades forming the Eosauropterygia [1]. Nothosauroidea include the well-known genera Nothosaurus, Lariosaurus, and Ceresiosaurus, as well as the only fragmentarily known Germanosaurus [1,3]. There is also Simosaurus, which is morphologically distinct from the other nothosauroids e.g., [1,4,5].
Alpha taxonomy of nothosaurs is primarily based on skull morphology. The nothosaur skull is dorsoventrally flattened and antero-posteriorly elongated. It has a heterodont dentition, including large fangs in the maxillary bone suggesting piscivory, although stomach contents of Ceresiosaurus also include small-bodied marine reptiles [6,7]. Marine invertebrates were also possibly preyed on by nothosaurs, depending on the absolute size of an individual [7]. Postcranial material of nothosaurs is seldom considered to be diagnostic because the postcranial skeletons of eosauropterygians are relatively uniform due to much convergence associated with secondary aquatic adaptations. The most common skeletal elements in classical Muschelkalk bonebed and condensation horizons throughout the Anisian and Ladinian outcrops of the Germanic Basin (Lower to Upper Muschelkalk deposits) [1,8,9,10,11] are isolated bones of Nothosaurus. Nothosaur humeri can be distinguish from those of other Sauropterygia due to their dorsoventrally flattened shape, their thin and sharp preaxial margin and the set-off but not constricted midshaft [12]. All nothosaur humeri have an oval to triangular cross section related to the thin and sharp preaxial margin, resembling the cross section of a hydrofoil in some taxa from the Upper Muschelkalk. Morphological details of nothosaur humeri were summarized by Rieppel, [1], Rieppel and Wild [9], Bickelmann and Sander [13], and Klein [12]. Specific taxonomic assignment of isolated Nothosaurus humeri is often impossible due to incompletely known association with diagnostic skulls and strong morphological variability, possibly involving sexual dimorphism as well [1,9,14,15]. Large Nothosaurus-type humeri also may belong to the large-bodied taxon Germanosaurus, which is based on two skulls but lacks associated humeri [1]. On the other hand, Simosaurus humeri are distinguishable from those of nothosaurids by their more slender appearance amongst other characters [1,4,5,16].
movements. The use of the forelimbs for propulsion was recently suggested by trace fossils interpreted as swimming tracks possibly formed by a large nothosaur [22]. Klein et al. [14] hypothesized that Nothosaurus from the Lower Muschelkalk used its forelimbs for locomotion, too, based on the dorsoventrally flattened humerus with its oval to triangular cross section.
Klein [12] described a lamellar-zonal bone tissue type with a dominance of longitudinal vascular canals and occasional incipient fibrolamellar bone in nothosaurs from the middle to late Anisian (Middle Muschelkalk). She was able to distinguish two humeral morphotypes in her sample [13]. These morphotypes also differ in histology [12]. In Ceresiosaurus, Hugi [23] described lamellar and parallel-fibered bone, as well as a high amount of woven bone with longitudinally and radially arranged vascular canals. Remodelling is limited in Ceresiosaurus [23]. The medullary region of Ceresiosaurus is filled with calcified cartilage and lamellar bone deposits [23]. She further described and figured two large nothosaur humeri (note that PIMUZ AIII 0001 was incorrectly identified as a femur in Hugi [23]). Hugi [23] pointed out the thin cortex of the humeri, which she interpreted as being related to fast active swimming and preferences for a more open marine environment [23].
Krahl et al. [24] described lamellar-zonal bone tissue in adult Nothosaurus and were the first to document differences in microanatomy between small-bodied N. marchicus from the early Anisian (Lower Muschelkalk) and large-bodied Nothosaurus species from the latest Anisian to early Ladinian (Upper Muschelkalk), indicating a shift in locomotion and habitat. Small N. marchicus was adapted to diving in shallow waters by BMI, whereas the large-bodied species from the Upper Muschelkalk showed an exceptional BMD, suggesting more active swimmers that colonized the open marine environment [24]. Further, the thin-walled, triangular humeral cross sections of large-bodied nothosaurs were interpreted as adaptations for withstanding high bending loads. In agreement with the morphological studies cited above, Krahl et al. [24] hypothesized that the evolution of paraxial front limb propulsion had already taken place in Nothosaurus, well before its convergent evolution in the Plesiosauria in the latest Triassic.
The current study focuses on describing and comparing humeral histology and microanatomy of a large sample of Nothosaurus spp. spanning their entire evolutionary history from early Anisian to early Ladinian (Lower to Upper Muschelkalk deposits of the Germanic Basin) in order to document and analyze intrageneric variability. This allows assumptions about the number of possible species occurring in the different stratigraphic horizons and outcrops. Additionally, the study aims at providing insights into aquatic adaptations, detecting evolutionary changes and trends, as well as drawing paleoecological inferences for the genus Nothosaurus.

Sampled bones
Sampled humeri from the Lower Muschelkalk (early Anisian) originate from localities in Germany (Eschenbach in der Oberpfalz, Bavaria), Poland (Górny Śląsk), and The Netherlands (Winterswijk). Samples from the Middle Muschelkalk (middle to late Anisian) are from central Germany (Freyburg/River Unstrut, Oberdorla, Rüdersdorf), and those from the Upper Muschelkalk (latest Anisian to early Ladinian) originate from classical Muschelkalk bonebeds of southern Germany [8] (Table 1; S1 Text and S1 Table). All humeri were found in isolation. Additionally, usually only incomplete and damaged bones were available for histological sampling, further limiting morphological information and making specific taxonomic identification difficult. Table 1 lists all specimens sampled together with their collection numbers, locality information, and stratigraphic range. The humeri and their respective thin sections are curated in the following institutions under the collections numbers given in

Taxonomic assignment
Four humeral morphotypes were recognized for nothosaurs; three by Bickelmann and Sander [13] and a fourth by Klein [12]. While these worked well in categorizing the then relatively small sample of Lower to Middle Muschelkalk nothosaur humeri, they lost their distinction in the currently available larger sample. Only humeral morphotype II, erected by Bickelmann and Sander [13], can be linked to a specific taxon, N. marchicus [14,13]. Large humeri from the Upper Muschelkalk are often assigned to N. mirabilis or N. giganteus [24], although these assignments are not based on skeletal association because no large humerus has been found with an associated skull in the Muschelkalk deposits so far. The humerus of the complete skeleton of N. giganteus, formerly Paranothosaurus amsleri, [26], from the Alpine Triassic is difficult to compare to specimens from the Germanic Basin because it is heavily crushed and only visible in ventral view, still being embedded in the matrix. Further information on taxonomic assignment is provided in S1 Text, S1 Table. Because the difficulty of assigning humeri of Nothosaurus to named species, we first organized the histological and microanatomical description from small to large size, with midshaft width as the size proxy, and second by stratigraphy from oldest to youngest. Possible taxonomic affinities based on histological results are discussed later. In addition to newly sampled specimens, previously sampled humeri are included in the current study [14,23,24].
In addition to the Muschelkalk humeri, a medium-sized humerus of Ceresiosaurus lanzi (PIMUZ T4845) from the early Ladinian of Cassina, Meride, Canton Ticino, Switzerland, is included in this study.

Sampling
The humeri were photographed and measured before sampling (see Table 1). Where possible, bones were sectioned exactly at the narrowest point of the midshaft. However, in some specimens, sampling location is slightly proximal or distal on account of poor bone preservation ( Table 1). The thin sections were produced following standard petrographic methods [27]. The entire thin sections were scanned with an Epson V740 PRO high-resolution scanner. In addition, thin sections were studied and photographed with a Leica1 DM 750P compound polarizing microscope equipped with a digital camera (Leica1 ICC50HD). The bone histological terminology follows Francillon-Vieillot et al. [28]. All thin sections are stored together with the bone they were cut from under the bone's collection number in the respective public collection (see Table 1 and above for list of collections).
Thin sections were drawn by hand from scans, with black denoting vascular spaces and white bone tissue. In these drawings, the ratio of compact bone versus vascular spaces and cavities, i.e., bone compactness (BC in Table 1), was measured with a custom-designed pixel-counting computer program (P. Göddertz, StIPB). Additionally, the drawings were analyzed with the program Bone Profiler [29]. Three parameters provided by this software were used to characterize bone density distribution: C, P, and S (see S2 Table). C is the global bone compactness for the whole sectional area. P is the relative distance from the center of the section to the point where the most abrupt change in compactness is observed. S is the reciprocal of the slope at the inflection point, which generally reflects the width of the transition zone between cortical bone and medullary region. A principal component analysis (PCA) was performed on the parameters cited above in order to explore the distribution of the different taxa in morphospace while optimizing the variance. The analysis was performed on a selection of specimens not affected by distortion/compression, and thus showing a well preserved entire sectional area and inner organization, in order to obtain valid parameter values. Prior to the analysis, all data were Log10-transformed to meet assumptions of normality and homoscedasticity. We tested the influence of size, again using midshaft width (mw) as size proxy, on the various parameters using linear regression analyses. In order to test for differences between specimens based on their stratigraphic origin, on the one hand, and the microanatomical category in which we placed them (see below) on the other, we performed MANOVAs and MANCOVAs (taking size as the co-variable). All of these analyses were performed using the statistic software R (R Development Core Team, 2008). Histological and microanatomical features of the samples are summarized in Table 1.

Histological description
General histology. The primary cortex of all sampled nothosaur humeri, independent of size or stratigraphic horizon, is basically composed of parallel-fibered bone tissue. The parallelfibered bone occurs in different degrees of organization and is locally intermixed with woven bone tissue or grades into lamellar bone. Tissue organization varies within a single section as well as between sections but tendentially increases from the inner towards the outer cortex. However, the degree of organization and vascularization can be irregular within a single cross section. The cortex is divided by cyclical growth marks (zones, annuli, and lines of arrested growth), whose analysis will be the topic of another study. Evidence that asymptotic size was reached, or that the individual was most likely close to it, is given by the presence of a distinct layer of lamellar bone visible all around the outer cortex. However, in many samples, a clear reduction in growth rate is only visible locally, whereas other regions of the outer cortex can show continuing growth.
The bone tissue contains small to medium-sized osteocytes. Locally, they may become more numerous and thicker (e.g., in regions with woven or coarse parallel-fibered bone tissue). Longitudinal and radial simple vascular canals dominate the tissue and are arranged in a radial system. In comparison to other Sauropterygia, vascular density is low to moderate [12,16,[23][24][25][30][31][32]. Many samples show locally or more broadly a funnel-shaped arrangement of crystallites around simple vascular canals [12,24]. The greater the organization of the bone tissue, the more numerous are the funnel-shaped canals. Mature and immature (yet incompletely lined) primary osteons occur in nearly all samples, but their number remains low in Lower and Middle Muschelkalk nothosaurs. Upper Muschelkalk nothosaurs seem to have a higher number of primary osteons. Figs 1 and 2 depict details of the medullary region and Figs 3-5 depict details of the periosteal bone tissue of Nothosaurus spp.
Lower Muschelkalk samples. SMNS 54317 consists of woven and coarse parallel-fibered bone, as well as more highly organized parallel-fibered bone tissue. Some large radial and longitudinal vascular canals have been transformed into primary osteons. Bone tissue of SMNS 80154 is much more highly organized. The Winterswijk humeri differ among themselves in the degree of tissue and vascular canal organization, and no clear groupings can be discerned. In Wijk13-259, vascular density and the number of osteocytes are low and tissue organization is high. Although larger than Wijk13-259, Wijk11-87 has in its postaxial inner cortex remains of a tissue that contains large simple longitudinal vascular canals together with woven and coarse parallel-fibered bone ( Fig 3A). This tissue differs from the rest of the highly organized cortex and is here interpreted as representing an early ontogenetic stage (i.e., possible perinatal tissue). The medullary region of Wijk11-87 contains postaxially and in its middle part calcified cartilage and is surrounded by a sharp line. In Wijk05-9 and TWE 480000320, the entire tissue is made of woven and coarse parallel-fibered bone with large, longitudinal, and mainly simple vascular canals indicating an early ontogenetic stage as well. The more proximal sample of  (Table 1) (Fig 3B).
Middle Muschelkalk samples. The Middle Muschelkalk samples IGWH 25 ( Fig 5E) and IGWH 28 are similar in bone tissue organization to the Winterswijk humeri. Except for MB.R. 414 (Fig 3F), the tissue of the remaining Middle Muschelkalk samples is more organized ( Fig  3D) than in the Winterswijk specimens (more highly organized parallel-fibered bone, higher amount of lamellar bone), and contains a higher number of funnel-shaped canals. Vascular canals are less radially arranged, more irregular in shape, and are generally larger than in samples from the Lower Muschelkalk. In MB.R. 539 and MB.R. 477, vascularization consists mainly of radial canals. The inner half of the cortex of MB.R. 539 shows parallel-fibered tissue, whereas the outer half largely consists of lamellar bone.
In MB.R. 414, approximately two-thirds of the cortex is made up of coarse parallel-fibered bone, whereas the outer cortex consists nearly exclusively of lamellar bone that still contains some simple radial vascular canals (Fig 3F). In MB.R. 941, vascular canals are very irregular in size, shape, and arrangement; the specimen contains a high number of primary osteons, which are mainly longitudinally oriented.   (Table 1).
Upper Muschelkalk samples. The medullary region contains calcified cartilage surrounded by a sharp line in some of the smaller humeri from the Upper Muschelkalk (GPIT/ RE/1339d, GPIT/RE/1339f, GPIT/RE/1590c, GPIT/RE/1339c, MHI 1906, and SMNS 2557). A sharp line is missing in GPIT/RE/1590a and GPIT/RE/1339b, but their medullary region also contains large amounts of calcified cartilage ( Fig 1F). GPIT/RE/1339a only locally shows remains of calcified cartilage between the intertrabecular spaces. No other nothosaur sample from the Upper Muschelkalk shows calcified cartilage or sharp lines, except for MB.R. 272. In MB.R. 272 remains of calcified cartilage are encompassed within secondary trabeculae. Bone tissue in GPIT/RE/1339c and GPIT/RE/1590c consists largely of coarse parallel-fibered bone, which is interrupted by thin layers of more highly organized tissue. In GPIT/RE/1590c, the primary cortex forms only a thin layer around a large core of calcified cartilage.
Except for GPIT/RE/1339a and GPIT/RE/1590d, the histology of the GPIT/RE/ sample is similar to that of the humeri from Winterswijk. GPIT/RE/1590d and MHI 633 are dominated by a radial vascular organization (Figs 4B and 5C). Radial vascular canals also occur in SMNS 81884, SMNS 53012, and SMNS 2557, but in these samples it is restricted to the preaxial half. MHI 1906 differs from the others because it postaxially has a high number of large open, round to oval vascular canals that are mainly arranged radially. SMNS 50221 has a low primary vascular density, and MB.R. 282 and SMNS 84772 have only a moderate one. The spongy appearance of these bones is due to secondary resorption of the periosteal cortex ( (Fig 5A). MB.R. 269 shows a high amount of lamellar bone throughout its entire cortex but has locally areas with poorly organized parallel-fibered bone tissue and a high primary vascular density (although this was already locally secondarily altered) (Fig 5E)

Microanatomical description
Lower Muschelkalk samples. SMNS 54317 has a spongy medullary region that is filled by calcified cartilage and endosteal bone (Fig 1A). The medullary cavity in SMNS 80154 is very small and additionally filled by endosteal deposits. The humeri from the locality of Winterswijk all share a free medullary cavity at midshaft surrounded by a compact cortex (Figs 1C, 6A-6E; Table 1). However, despite a similar general internal structure, the relative size of these cavities and of the medullary area more generally is variable. MB.R. 782 has a small free cavity that is surrounded by a thick layer of endosteal bone, which also coats the inner parts of the nutrient canal. MB.R. 780 is peculiar in having a medullary region completely filled by endosteal bone, resulting in an extremely compact section (Figs 1D and 6F). MB.R. 817.1 shows a spongy and rather large medullary region (in accordance with its more distal sampling location by comparison with other specimens in this study) with large and randomly shaped cavities and secondary trabeculae. Most of the Lower Muschelkalk samples show relatively high bone compactness, with values above 80%; only Wijk05-9 (from a probably juvenile specimen) and the distally sectioned MB.R. 817.1 display compactness values below 80% (Table 1)  Ceresiosaurus lanzi. The humerus of Ceresiosaurus lanzi (PIMUZ T4845) has very compact cortical tissue, and the medullary region is nearly completely filled by endosteal deposits (Fig 7C).
Results of principal component analysis (PCA). The PCA shows that the two main axes explain most of the variation (83.6%; Fig 8). While the Upper Muschelkalk specimens cover the whole variation observable in the Nothosaurus spp. sample (Fig 8A), the Lower and Middle Muschelkalk samples occupy more restricted distribution areas. Interestingly, they are rather distinct from each other despite some limited overlap. C (compactness) and P (point of most abrupt change in compactness) are the main variables discriminating the taxa along the first PCA axis (Fig 8A). Conversely, the width of the transition zone (S) discriminates specimens along the second axis. Although these trends are naturally associated with general bone microanatomical features (Fig 8B), no clear grouping is observed. This reflects the continuous variation in Nothosaurus microanatomical features and the gradual transitions between the major microanatomical types (Fig 8B). It is worth noting that S has no real impact on the distinction between the various microanatomical types in our Nothosaurus sample.
Bone Profiler is a software that was designed to analyze microanatomy in bone sections that have a vaguely circular outline and two concentric zones. The inner structure of the nothosaur sample is very variable and many of the large samples show a rather random distribution of large resorption cavities with no distinguishable medullary area (e.g., MB.R. 414, MB.R. 282, SMNS 50221, SMNS 84772, and GPIT/RE/1339a), resulting in the values attributed to the S and P parameters having no structural meaning. The specimens showing a thin-walled cross section are nevertheless clearly distinct from the ones having a thick and compact cortex (Fig 8B).
Tests on the microanatomical parameters. Linear regressions show the influence of size (medullary width) on C (r = -0.59, p <0.001) but not on S (r = 0.07, p = 0.65) and P (r = 0.16,

Choice of sampling location
In bone histological and microanatomical studies, it is crucial to compare homologous sections because of the variation in microanatomy throughout the bone caused by developmental origin of the tissue and morphogenesis. Thus, thin sections of long bones were usually taken where the midshaft is the narrowest, because this represents the plane intercepting the growth center (center of ossification) or neutral zone [12,25,[32][33][34][35]. All our sections are thus comparable among nothosaurs and with those of other taxa, notably among other Sauropterygia [12,16,[23][24][25][30][31][32]36]. In nothosaurs, the nutrient canal as the indicator for the position of the growth center [37,38] is not reliable. The nutrient foramen, i.e., the opening of the canal to the outer bone surface, is located at the transition from the proximal head to the midshaft, below the deltopectoral crest. All samples of Nothosaurus were taken distal to the nutrient foramen, and only some thin sections show a canal in the cortex that might be identifiable as part of the nutrient canal and which is intersected in different planes (e.g., SMNS 54317, MB.R.782, GPIT/RE/1339a, MHI 754, GPIT/RE/1339b, GPIT/RE/1339d). The growth center, however, is at midshaft in nothosaurs as indicated by no or limited remodeling of periosteal bone at the narrowest midshaft, whereas samples that were taken more proximally (closer to the nutrient canal) or further distally display distinctly more endosteal remodelling. The absence of calcified cartilage and the absence of the sharp line representing the boundary between the endosteal and the periosteal domains further indicate the narrowest midshaft as the area that has the most complete growth record in Nothosaurus. This is also supported by a large, serially sectioned humerus of Nothosaurus from the Upper Muschelkalk from its proximal head to its distal end [24]. The thin sections from the region of the nutrient canal [24] show less primary cortex preserved than the midshaft thin section [24]. Although microanatomy varies along the shaft, this series of thin sections furthermore documented that the structure remains essentially constant (tubular with a very large but not well delimited medullary region). This is contrary to what is observed in the stratigraphically older Nothosaurus specimens, where a free cavity at midshaft proximally and distally turns into spongy bone.
Independent of which region of the midshaft is compared; variation in sampling location along midshaft alone cannot explain the high microanatomical variability observed in the current sample of Nothosaurus, specifically in the Upper Muschelkalk forms that show a wide range of relative cortical thicknesses.

Calcified cartilage
In Nothosaurus spp., the presence of calcified cartilage depends on sampling location along the shaft as well as on ontogeny, and on intra-and interspecific differences in endochondral ossification. In adult Lower and Middle Muschelkalk nothosaurs, calcified cartilage persists in some individuals at the narrowest part of the midshaft, but not in others (Table 1). When present, remains of calcified cartilage are embedded in the endosteal bone. None of the midshaft samples keep an entire core of calcified cartilage, as opposed to the more distal samples (Table 1). Small humerus SMNS 54317 and some of the smaller humeri (mw <25 mm) from the Upper Muschelkalk show a core of calcified cartilage at midshaft, indicating an early ontogenetic stage. Among the larger humeri (mw >25 mm) from the Upper Muschelkalk, only one (MB.R. 272) shows locally remains of calcified cartilage.
Calcified cartilage at midshaft resulting from incomplete endochondral ossification is a typical feature of osteosclerosis [39], as seen in adults of the stratigraphically somewhat younger pachypleurosaurs from the Alpine Triassic [31]. However, few of the adult sampled Nothosaurus specimens show this histological feature, which does not seem to have been fixed during nothosaur evolution.

General histological observations and resorption activity
Samples of Nothosaurus spp. may show woven bone in combination with some primary osteons or incompletely lined primary osteons. This tissue can then be interpreted as incipient fibrolamellar bone or true fibrolamellar bone, respectively [12,23,24]. However, fibrolamellar bone tissue of Nothosaurus spp. is always restricted to layers or regions and never forms the entire cortex, as opposed to some other Triassic Sauropterygia [24,25,32,40,41]. Coarse parallel-fibered bone tissue is more common than woven bone. This bone tissue was already described for several other aquatic tetrapods (the temnospondyl Plagiosaurus, [42]; mosasaurs, [43]; ichthyosaurs, [44]; placodonts, [25,32]; Simosaurus, [16]). Organization of this tissue together with a high density of osteocytes indicates fast deposition and thus suggests increased growth rates when compared to the deposition of typical parallel-fibered bone tissue [16].
Nothosaurus spp. share with other sauropterygian taxa such as placodonts [25,32], pachypleurosaurs [12,30,31,36], pistosaurs [24], and plesiosaurs [24,41] dominantly radially organized vascularization as expressed by a high amount of radial vascular canals in addition to longitudinal ones. The exception is the nothosaur Simosaurus, which does not show radial vascular organization [16]. Independently of stratigraphic age or size, many Nothosaurus samples show preaxially a high number of large radial vascular canals, indicating faster growth in this region than elsewhere. Radial vascularity similar to that of sauroptergyians was also described for early synapsids [45,46], indicating that this tissue is not restricted to marine reptiles.
Cortical resorption patterns are highly variable in Nothosaurus, ranging from nearly no resorption to strong resorption, the latter resulting in BMD. The resorption activity in some large-bodied Nothosaurus spp., resulting in very thin-walled humeral cross sections, is unique in its intensity and absence of secondary deposits among other marine reptiles or any other amniote except birds, as already pointed out by Krahl et al. [24].
In a few samples (SMNS 84772, SMNS 50221, MB.R. 282, MB.R. 414), much of the primary cortex is resorbed in a peculiar pattern, resulting in large irregular cavities and in a spongy periosteal tissue, whereas the medullary cavity-if distinguishable at all-retains a moderate size. Resorption in these specimens did not follow the standard pattern starting from the inner cortex and continuing into the outer cortex, but occurred randomly throughout the cortex.
Because of the lack of any modern analogue for these two different but unusual resorption activities, their meaning remains unresolved. Biomechanical explanations and those invoking reproductive or physiological stresses (e.g., salinity fluctuations) must remain hypothetical.

Microanatomical patterns in Nothosaurus spp.: Increase vs. decrease in bone mass
Bone mass increase (BMI) [47] is a mechanism by which secondarily aquatic tetrapods living in shallow waters increase their specific bone density to counteract the positive buoyancy caused by the lungs [39]. BMI is achieved through osteosclerosis, which can be the result of different mechanisms, such as incomplete endochondral ossification (i.e., persistence of calcified cartilage), the filling of inner cavities by excessive secondary bone deposits, and/or the inhibition of cortical resorption [39,47]. Osteosclerosis, often in combination with morphologically observable pachyostosis, has been described for many marine reptiles and some marine mammals that are interpreted as slow swimmers moving about by long but shallow dives [39,[48][49][50]. In Nothosaurus, pachyostosis has been documented only in ribs [1], which have not been histologically studied yet. Ceresiosaurus humeri show pachyostosis and osteosclerosis [23] (this study). Bone mass decrease (BMD) is caused by an imbalance between strong resorption and limited secondary deposits. BMD occurs in pelagic forms, such as some marine turtles, mesosuchian marine crocodilians, ichthyosaurs, adult plesiosaurs, mosasaurs, and modern cetaceans [39,47]. It seems to be typical for active swimmers requiring high speed and maneuverability in open water habitats. However, although these groups also show BMD, it is reached in Nothosaurus in a higher degree and maybe by different processes involved.
To evaluate if a taxon shows BMI or BMD, the condition seen in its ancestor must be considered. However, the terrestrial ancestor of marine Sauropterygia is not known [1], and their relationships within higher diapsids are largely unresolved [25]. Nevertheless, terrestrial amniotes usually have a large free medullary cavity surrounded by a layer of compact cortex [51][52][53]. Long bones of Claudiosaurus, a Late Permian neodiapsid and a putative morphological precursor of Triassic eosauropterygians, has a small medullary cavity surrounded by a thick cortex, displaying osteosclerosis [35,39]. When compared to similar sized Nothosaurus spp., Simosaurus [16] and Ceresiosaurus [23] both show smaller medullary cavities surrounded by a relatively thicker, compact cortex and thus also display osteosclerosis. Microanatomy of European pachypleurosaurs is well known [30,31,35] and varies within a single taxon as well as between taxa: The pachypleurosaur A. heterodontus from the early to middle Anisian shows a comparable variable size range of the free medullary cavity when compared to Nothosaurus from the same locality but a generally higher vascular density, resulting in lower bone compactness [30] (Table 1). Pachypleurosaurs from the Ladinian of the Alpine Triassic (Neusticosaurus spp.; Serpianosaurus) have a small medullary cavity or show incomplete endochondral ossification [31,35,39]. These pachypleurosaurs document stronger osteosclerosis, when compared to A. heterodontus and Nothosaurus as do Placodontia [25,39] and Pistosauria [24]. Thus, Triassic Sauropterygia, except for some medium-sized to large-bodied individuals of Nothosaurus, principally display osteosclerosis and BMI. Modern semi-aquatic crocodiles share with largebodied Nothosaurus a similar dorsoventrally flat body shape and size range. Based on illustrations of thin sections of humeri and femora [54,55] crocodiles show, contrary to most largebodied Nothosaurus, osteosclerosis and therefore BMI. A distinct decrease in bone mass as observed in nothosaurs with thin-walled cortices may play an important role in adding positive buoyancy for surface swimmers. Surface swimming was also suggested for some small, marine squamates (dolichosaurids) that show much lightened vertebrae and ribs [56].
The microanatomical patterns seen in Nothosaurus spp. are subdivided in Table 1 into four categories. Category 1 refers to true BMI, with samples showing a small medullary cavity (i.e., reduced when compared to a hypothetical terrestrial ancestor) surrounded by a thick cortex of compact bone. Samples representing category 2 have a moderately sized medullary cavity surrounded by compact bone. Category 3 refers to samples with an enlarged medullary cavity surrounded by compact bone. This category is comparable to the microanatomy of humeri of modern varanids [57,58] and iguanines [59] but when compared to other Triassic Sauropterygia [1] this category documents a decrease in bone mass. Category 4 represents true BMD, because samples have an extremely thin-walled humeral cross section. These microanatomical categories were statistically tested and found significant.
Lower Muschelkalk samples. Compared to modern sauropsids and to the entire Nothosaurus sample, samples from the Lower Muschelkalk of Winterswijk generally show BMI (i.e., osteosclerosis; category 2, Table 1) and are comparable to Pachypleurosauria, Placodontia, Simosaurus, and Ceresiosaurus, although their medullary cavities are tendentially larger. An exception is the large humerus MB.R. 780 from the Lower Muschelkalk of Poland that shows an extreme reduction of the medullary cavity because it is completely filled by endosteal bone (BMI or category 1; Table 1). Bone compactness is here comparable to small pachypleurosaurs from the Alpine Triassic [31]. The small humerus MB.R. 782 from the Lower Muschelkalk of Poland has a reduced medullary cavity, and filling of the remaining cavity by endosteal bone was in progress when the animal died. Contrary to these samples, MB.R. 817.1 also from the Lower Muschelkalk of Poland shows high endosteal resorption rates, resulting in a spongy medullary region surrounded by a locally thick and compact cortex (category 2; Table 1). The various osteosclerotic microanatomies of these Lower Muschelkalk nothosaurs thus fit environmental conditions that would have been near-shore to shallow marine (S1 Text, S1 Table).
Middle  Table 1). The rest of the Middle Muschelkalk sample shows medium-sized to large free medullary cavities (category 3; Table 1) when compared to the nothosaurs from the Lower Muschelkalk. Comparing them to modern sauropsids, one could argue that these specimens are possibly closest to the condition of the unknown terrestrial ancestor. However, these specimens more likely represent a trend towards BMD, away from the osteosclerotic condition of the Lower Muschelkalk nothosaurs and that of placodonts, pachypleurosaurs, and pistosaurs. A third microanatomical category is represented by MB.R. 414 with its spongy periosteal tissue, possibly representing a different pathway towards BMD.
The localities from which all these Middle Muschelkalk samples originate were close to the center of the basin with seemingly deeper water, but the Middle Muschelkalk is characterized by sea level regression (S1 Text). The presence of islands, fluctuating sea level, and taphonomic processes such as redeposition and accumulation of bones from different habitats as well as migration of marine reptiles into the Germanic Basin from more pelagic environments (i.e., from the Tethys through the temporarily open gates) as well as possible developmental plasticity or a sexual dimorphism might explain this strong variation in microanatomy.
Upper Muschelkalk samples. Several samples from the Upper Muschelkalk document BMI in different degrees of intensity (category 1, 2; Table 1) when compared to the Nothosaurus sample in general as well as to other Sauropterygia. This was achieved by different processes: remodelling inhibition, incomplete ossification/presence of calcified cartilage at midshaft, filling of the medullary region by endosteal bone, and excessive secondary bone deposits. In addition to the samples exhibiting BMI, many large humeri show BMD (category 3, 4; Table 1). As already noted, there are two ways of achieving BMD in the larger Nothosaurus humeri from the Upper Muschelkalk: a general reduction in compactness by resorption vs. an increase in the size of the free medullary cavity. As described in S1 Text, the Upper Muschelkalk was a phase of transgression with a sea level high stand. The increase in microanatomical diversity is accompanied by an increase in taxonomical diversity [1,9]. However, developmental plasticity and sexual dimorphism might be considered as well.

Taxonomic implications
A taxonomic classification of nothosaur humeri based on morphology remains inconclusive as discussed above, raising the question whether histology or microanatomy allow a more precise taxonomic assignment.
Two Nothosaurus species are known from the Lower Muschelkalk of Winterswijk: the diminutive N. winkelhorsti, and the small-bodied N. marchicus [14,60,61]. Recently, the small nothosaurid Lariosaurus vosseveldensis was also described from the locality of Winterswijk [62]. Despite their small size, humeri Wijk13-259 and SMNS 80154 do not represent early ontogenetic stages based on bone tissue organization and vascularization pattern. They thus seem to represent a smaller-bodied taxon than N. marchicus; likely N. winkelhorsti because the humeri are not pachyostotic as should be the case for Lariosaurus. The rest of the humeri from Winterswijk might belong to N. marchicus, but show morphological, histological, and microanatomical variability. The presence of a third Nothosaurus species at this locality is also conceivable [14,15]. SMNS 54317 represents an early ontogenetic stage based on bone tissue and vascularization. This sample and MB.R. 782, however, fit the histological and microanatomical variability of what is thought to represent N. marchicus. MB.R. 817.1 and MB.R. 780 from the Lower Muschelkalk of Upper Silesia are both much larger than humeri from Winterswijk (Table 1). They differ distinctly in microanatomy (and morphology) when compared to each other as well as to the Winterswijk sample, most likely representing two different taxa, one of which may be Germanosaurus. Rieppel and Wild [9] noted that Nothosaurus mirabilis may extend down into the early Anisian (upper Lower Muschelkalk; mu2; [9]).
According to Rieppel and Wild [9] and Rieppel [1], a small-bodied nothosaur (cf. N. marchicus) as well as a larger-bodied nothosaur (cf. N. mirabilis) occur in the Middle Muschelkalk. Klein [12] described humeri from the Middle Muschelkalk of Freyburg (River Unstrut), which differs in morphology, microanatomy, and histology from N. marchicus and might represent N. mirabilis. However, sexual dimorphism for both taxa cannot be excluded as an explanation of these histological and microanatomical differences.
Except for a tendency towards a larger medullary cavity, samples IGWH-25 and IGWH-28 are morphologically and histologically (bone tissue and vascularization) similar to what is thought to represent N. marchicus. The microanatomy and bone tissue of MB.R. 162.4, MB.R. 477, and MB.R. 174.2 are comparable to N. marchicus but are morphologically different. Again, morphological differences might be related to developmental plasticity or sexual dimorphism. Histology and microanatomy of MHI 1193 is similar to the Lower Muschelkalk samples SMNS 80154 and Wijk 13-259 and might indicate the presence of diminutive taxon in the Middle Muschelkalk. Most of the Middle Muschelkalk samples are larger than the known size range of N. marchicus and have a tendency towards a decrease in bone mass. These humeri differ also morphologically from N. marchicus and have been assigned before to another humeral morphotype of unknown affinity [12]. Sample MB.R. 414 with its spongy periosteal tissue clearly differs from all other Middle Muschelkalk samples and may indicate the presence of yet another nothosaur taxon if other explanations can be excluded. MB.R 539 from the Middle Muschelkalk of Rüdersdorf has a very distinct cross section, with a characteristically pointed preaxial margin and a broad but also pointed postaxial margin. Its bone tissue is characterized by an untypical high amount of lamellar bone. Its overall humeral morphology, however, is that of a typical Nothosaurus.
Small humeri collected from the Upper Muschelkalk of Crailsheim (GPIT/RE/sample, Table 1) are unified by morphology, histology, and microanatomy that clearly differ from those in N. marchicus. The morphology of these humeri differs also from that of N. jagisteus [10]. No humeri are available for comparison for N. edingerae and N. juvenilis, the other smallbodied nothosaur taxa described from the Upper Muschelkalk (see S1 Text and S1 Table) [1,10]. Early ontogenetic stages of larger-bodied nothosaur taxa can be excluded for these smaller Upper Muschelkalk specimens due to morphology and histology. The morphology of humerus SMNS 53012 is obscured by compression, but histology and microanatomy fit well with the above mentioned GPIT/RE samples indicating similar taxonomical affinities.
SMNS 2557 most likely represents a juvenile of the larger-bodied nothosaur exhibiting BMI (SMNS 84851, SMNS 7175) due to its small size, the absence of a free medullary cavity, and the presence of calcified cartilage at midshaft. The majority of large humeri from the Upper Muschelkalk show some degree of BMD (Table 1). These most likely belong either to N. mirabilis or to N. giganteus. As already noted, Simosaurus can be excluded from consideration due to different morphology, microanatomy, and histology [16]. Thus, either N. mirabilis or N. giganteus show in general BMI or BMD, respectively. The taxonomic assignment of StIPB R 40, StIPB R 45, and StIPB R 53 by Krahl et al. [24] to N. giganteus and of StIPB R 54/2 to N. mirabilis was solely based on the size of the humeri and cannot be verified. The microanatomy of these samples is very similar and taxonomic distinction based on bone histology is not possible.
N. mirabilis and N. giganteus differ in the length of their neural spines [9] which is also related to swimming style. N. giganteus has short neural spines, which is a plesiomorphic feature for Nothosaurus, whereas N. mirabilis has elongated neural spines [9]. N. marchicus has short neural spines and its humerus is osteosclerotic. In N. jagisteus the length of the neural spines is intermediate but microanatomy of the humerus is not known [10]. Based on these differences, we speculate that N. giganteus is represented by the plesiomorphic histological pattern of retaining BMI, in keeping with its short neural spines. N. mirabilis could be represented in our sample by the humeri with the very thin cortex, uniting the derived states of BMD and high neural spines, reflecting evolution towards an active swimming style.

Conclusions
The four major groups of Triassic Sauropterygia (Placodontia, Pachypleurosauria, Nothosauroidea, Pistosauroidea) show a general tendency for osteosclerosis and bone mass increase [12,16,[23][24][25]30,31,36,[39][40][41]. However, among Nothosaurus spp. from the latest Anisian and early Ladinian (Middle Triassic; Upper Muschelkalk deposits) occur some large-bodied individuals (or taxa) that are unique among Sauropterygia in documenting a distinct decrease in bone mass. This has resulted in a very thin-walled cortex in some individuals. Another peculiar pattern observed in some other individuals of Nothosaurus from the latest Anisian and early Ladinian is a heterogeneously spongious periosteal cortex. Both patterns are unique among amniotes. The difference to the thin-walled bones of birds is that the former are pneumatized, which certainly was not the case for nothosaurs.
Quantitative analyses reveal that size, indicated by the midshaft width proxy, has an influence on bone compactness. Smaller taxa have generally higher bone compactness than larger taxa throughout the evolutionary history of the group. They also showed that there is a significant difference between the four microanatomical categories documented among the nothosaur sample, and also that there is a significant difference in the distribution of the quantitative microanatomical parameters following the stratigraphy (i.e., between specimens from the Lower, Middle and Upper Muschelkalk, respectively) reflecting specificities linked to each stratigraphic level, such as that Lower Muschelkalk specimens generally all show a compact organization, whereas Middle Muschelkalk and Upper Muschelkalk specimens show a high diversity in the types of patterns observed.
Small bodied taxa with their compact organization are assumed to have been bound throughout their evolution to near coastal environments, whereas large bodied Nothosaurus spp. diversified from the middle Anisian onwards and settled near coastal environments as well as in open water. Microanatomical diversity observed in Nothosaurus spp. may reflect developmental plasticity, sexual dimorphism, but also taxonomical diversity.
Supporting Information S1 Table. Table on Table. Bone profiler parameters. C is the global bone compactness for the whole sectional area. S is the reciprocal of the slope at the inflection point and generally reflects the width of the transition zone between the cortical bone and the medullary region. P is the relative distance from the center of the section to the point of inflection, i.e. where the most abrupt change in compactness is observed. P is thus proportional to the size of the medullary cavity. C is the global bone compactness for the whole sectional area. S is the reciprocal of the slope at the inflection point and generally reflects the width of the transition zone between the cortical bone and the medullary region. P is the relative distance from the center of the section to the point of inflection, i.e., where the most abrupt change in compactness is observed. P is thus proportional to the size of the medullary cavity. (DOC) S1 Text. Sea level and distribution of Nothosaurus spp. in the Germanic Basin (Muschelkalk and Keuper deposits) during the Anisian and Ladinian. (DOC)