Growth curves describe the change in body mass during ontogeny and are based on the histological growth record of the individual. The body mass of the individual is thus needed to establish its growth curve. The body mass at death of each individual was calculated using the equation for standard quadrupeds from Anderson et al. (1985). This equation estimates body mass from the circumference of the humerus and femur of the same individual. It is expected to reveal robust estimates of body mass in mammalian and non-avian reptilian taxa (Campione and Evans, 2012). Its applicability is untested for secondarily marine vertebrates, which are not standard quadrupeds due to, e.g., differences in body shape and gravity. However, body mass of the semiaquatic Alligator mississippiensis was tested in the study of Campione and Evans (2012), which has a dorsoventrally flat and anteroposteriorly elongated body form that is also seen in Simosaurus. Body mass calculations for Simosaurus are based on the most complete individual SMNS 14733. The specific equation that was used to estimate body mass (W) is

W = 0.078 × (7.9 + 6.1)^2.73 [Anderson et al., 1985]

It revealed 105 kg for SMNS 14733. The humerus length of SMNS 14733 is 262 mm, which is 76.5% of the length of the largest humerus known of Simosaurus (SMNS 17590; 342 mm). From the humeral length (262 mm) and the body mass (105 kg) of SMNS 14733 the body mass of the sampled individuals (M) was estimated from the length of the humeri (humerus length) applying the equation

M = 105 × humerus length/262.

For the femora SMNS 18038 and SMNS 18698 the corresponding humeral length was estimated assuming a humerus/femur length ratio of 1.29 in Simosaurus (Rieppel, 1994; own calculations). The femur length for the distal femur fragment SMNS 91983 (Table 1) was reconstructed from a regression line (R² = 0.8776) on femur length against distal width (data taken from Rieppel, 1994: Table 6). Resulting body masses are given in Table 1 for each individual.

To construct a mass-based growth curve for a single individual, the relationship between local bone apposition rate and body mass gain is taken into account. Minimal shaft circumference increases proportionally with bone length and the local bone apposition rate is closely tied to body mass gain. Thus, each growth mark laid down during ontogeny is linked to a respective body mass of the individual. First, to reconstruct respective body masses for growth marks an estimate of the body mass of the individual at death is needed as a reference (Table 1). Next, for each growth mark its distance from the center of the section is measured, and the proportion of this distance and the distance between the center of the section and the outer bone surface of the section is calculated. Then, for each of the growth marks the respective proportions are multiplied by the bone length at death yielding an estimate of the respective bone lengths for each growth mark. Finally, to estimate masses from bone length at each growth mark the equation given in Erickson and Tumanova (2000) was applied.

Mass (bone length) = (bone length³/bone length at death³) × body mass at death

The non-linear model fitting procedure used to establish growth curves for specimens is introduced and described in detail in Griebeler et al. (2013). Growth curves for four out of the five Simosaurus specimens were studied (Table 1). Usually, only samples taken exactly at midshaft were considered for the analysis of growth because they reveal the most complete growth record. However, the growth record of humerus SMNS 52095 that was sampled distal to midshaft is well preserved and seems to be fairly complete with only two or three annual growth cycles missing. Modelling this humerus revealed reasonable results. Thus, this sample was included into the analysis, but modelling results must be evaluated carefully. Femur SMNS 91983 was also sampled distally to the midshaft. Its annual growth record is, however, not very distinct, which means that growth cycles cannot be followed all around the cross section. Thus, annual cycles cannot unequivocally be distinguished from subcycles and this specimen was excluded from growth curve modelling.

For SMNS 18038, SMNS 18689, SMNS 18698, and SMNS 52095 four standard sigmoidal growth models were tested, which basically differ in the position of the inflection point: von Bertalanffy (c. 30% of asymptotic mass), Gompertz (c. 38%), logistic (50%), and Chapman-Richards growth model (not predefined, variable). The number of growth marks preserved in all specimens is small, but the standard growth models have a comparatively large number of parameters. Therefore, different versions for each of the four sigmoidal growth models were tested. These model versions differ in the number of model parameters estimated, e.g., in the model equation the parameter body mass at the first growth mark was set to zero or the parameter asymptotic body mass was set to the body mass at death of the specimen. The fully-parameterized von Bertalanffy growth model and the Gompertz model have three model parameters (body mass at the first growth mark M ₀, asymptotic body mass A, and growth rate g), the logistic growth model has four parameters (body mass at the first growth mark M ₀, asymptotic body mass A, growth rate g, growth mark at which the inflection point is located i), and the Chapman–Richards model has five parameters (body mass at the first growth mark M ₀, asymptotic body mass A, growth rate g, parameters i and m set the position of the inflection point at the time and body mass axis). For each model version two model fitting techniques were used to estimate the number of missing growth cycles, to cope with the problem that in the inner part of the bone the growth record could be incomplete. For both techniques, a hatchling mass of 400 g was used based on histological data revealed from specimen SMNS 18698 (see Section 3). Under fitting technique 1 the number of missing cycles of a specimen was read off from the fitted growth curve at the age for which it predicts a mass of 400 g. Under fitting technique 2 the mass at cycle zero was set to a fixed hatchling weight of 400 g, and stepwise considered different numbers of missing cycles during model fitting (0, 1, …) until the residual standard error and the AIC value stabilized (change < 1). In this paper, contrary to Griebeler et al. (2013), a linear growth model was also considered for each specimen in order to test whether its growth record only covers the quasi-linear phase of growth. Finally, for each specimen the statistically best supported model out of the set of candidate models was identified. First, their biological plausibility were checked (hatchling mass ≥ 0 and 400 g, asymptotic mass ≥ 0 and larger than hatchling mass, and growth rate ≥ 0), and next their residual standard errors for absolute goodness of fit, as well as their AIC scores, ΔAIC values, and Akaike weights for relative goodness of fit were evaluated. Fitting of growth curves and calculation of AIC values was conducted in the free statistic software R (version 3.0.2) using the “nls”-function provided in the package “nls”.

The humerus of Simosaurus is a slender, curved bone with a rather simple morphology when compared to other Eosauropterygia (Fig. 1A, B) (Bickelmann and Sander, 2008, Klein, 2010 and Rieppel, 1994). The proximal head is smooth and round, but with a straight preaxial margin due to the prominent development of the ventrally running deltopectoral crest. The proximal head forms ventropreaxially a flat surface that extends down to the midshaft. The proximal end is more massive and longer than the distal end. The distal end is wider than the proximal one but not as much expanded as in other Sauropterygia. The midshaft margins are straight and the area is slightly constricted. The proximal end is angled/twisted when compared to the midshaft/distal end resulting in a curvature of the entire bone. Humerus SMNS 52095 is complete, except for reconstructed parts of the distal end. It has a shallow groove that can be interpreted as an entepicondylar foramen (contrary to Rieppel, 1994) (Fig. 1A). The entire bone is heavily striated. Sampling location of this bone is not directly at the midshaft but slightly distally due to the presence of plaster in the midshaft region. Humerus SMNS 18698 is also a complete bone with a few reconstructed areas. This humerus is heavily dorsoventrally flattened. The proximal end is much more angled (nearly 35°). It is disproportionate when compared to the midshaft and distal end of SMNS 52095. Distally, the humerus shows traces of oyster growth, indicating that the bone was for a while exposed before final burial/fossilization. Humeral cross sections are oval (Fig. 2).

The femur of Simosaurus is generally slender and only slightly curved (Fig. 1C, D, E). The midshaft region is clearly constricted. The prominent development of the trochanter results in a triangular cross section of the proximal head of the femur. The postaxial and dorsal margins of the proximal head are straight in dorsal view. The distal end is weakly broadened with the articular condyles being separated from each other by a shallow intercondylar fossa. This fossa separates the tibial and fibular articulations, which is not the case in other Eosauropterygia. SMNS 18689 is a complete femur that is ventrodorsally compressed. SMNS 91983 is the distal end of a femur with straight midshaft margins and two separated condyli. SMNS 18038 is a complete femur that resembles the morphology of SMNS 18689. Femur cross sections are round, except for SMNS 18038, where the cross section is oval (Fig. 2).

Due to different sampling locations along the midshaft the medulla of the sampled Simosaurus bones is variable. The distal sample of the largest humerus SMNS 52095 shows a medullary region surrounded by a spongeous perimedullary region (Fig. 2A, F). The midshaft sample of humerus SMNS 18698 has a prepostaxially elongated free medullary cavity surrounded by a narrow perimedullary region (Fig. 2B, G). The midshaft sample of femur SMNS 18689 and the midshaft sample of femur SMNS 18038 both share a central free cavity (Fig. 2C, E, H, J) and nearly no perimedullary region exists. Please note that the medullary cavity in SMNS 18038 is filled by crystallites (Fig. 2E). The distal sample of femur SMNS 91983 shows a large medullary region (Fig. 2D, I).

The medullary regions and cavities are surrounded by a moderately vascularized compacta. In general, vascularization is highest in the inner cortex and decreases towards the outer cortex. Vascular canals appear as primary osteons, immature primary osteons (Klein, 2010 and Klein et al., 2015a), and simple vascular canals. Immature primary osteons are vascular canals, which are not completely lined by lamellar bone. Osteons are enlarged by erosion (secondary widening, see Klein et al., 2015a). Bone compactness of the cortex (without the medulla) revealed that vascular density is lower in femora than in humeri (Table 1). Bone compactness calculated with Bone Profiler (medulla included) revealed similar values for humeri and femora (Table 1) that result from the large medullary cavity/region in femora. Bone compactness ranges between 74.6% and 81.4% in humeri and between 73% and 88.5% in femora (Table 1). When the medulla is excluded from the calculation, bone compactness is around 90% in humeri and between 94.9% and 96.5% in femora (Table 1).

Femoral bone tissue is dominated by longitudinally organized canals, but radial canals can also occur. Femur SMNS 18689, for example, has a high amount of radial vascular canals that are mainly radially arranged along the cortex (Fig. 2C, H). In the humeri the organization and form of vascular canals are more variable than in the femora, but longitudinal canals dominate the tissue, as well. Most of the cortex of humerus SMNS 52095 is moderately vascularized by longitudinal canals, but locally the cortex is highly vascularized by a mixture of longitudinally, radially, and reticular arranged canals (Fig. 2A, F).

The extensive medullary region of humerus SMNS 52095 consists of large cavities surrounded by endosteal (secondary) trabeculae and endosteal bone, the latter are not arranged in a trabecular structure. The endosteal deposits comprise large amounts of calcified cartilage (Fig. 3A, B). The area is surrounded by the typical sharp line that is a thin layer of lamellar (endosteal) bone, which is the boundary between the endochondral and periosteal domain (Klein, 2010 and Klein et al., 2015a) (Fig. 3B, white arrow). Endochondral erosion has altered this boundary. Endosteal deposits reach into the periosteal domain. Secondary endosteal trabeculae comprise remains of primary bone obscuring the boundary between the endochondral and periosteal domain as well. The free medullary cavity of humerus SMNS 18698 is lined by a thick layer of endosteal bone (Fig. 3C, D) and endochondral remodeling has altered the inner cortex. The free cavities of femur SMNS 18689 and femur SMNS 18038 are both lined by a distinct layer of endosteal bone that locally reaches into the periosteal domain (Fig. 3E, F). The free medullary cavity of femur SMNS 18038 is additionally surrounded by a sharp line of endosteal bone comprising calcified cartilage as well as endosteal deposits. The large medullary region of femur SMNS 91983 is surrounded by a sharp line and consists of endosteal (secondary) trabeculae and of endosteal bone comprising large amounts of calcified cartilage (Fig. 3H, I).

In all samples, the cortex consists of a coarse parallel-fibred bone matrix with numerous osteocytes (Fig. 4A–D). Primary osteons are thickly lined by lamellar bone (Fig. 4A–E). Humerus SMNS 52095 shows an abrupt change in its outer cortex from coarse parallel-fibred bone to lamellar bone, resulting in an external fundamental system (EFS) (Fig. 4F). In femur SMNS 91983 an increase in tissue organization towards the outer cortex is visible, but no change to an EFS is observed. In femur SMNS 18689 and humerus SMNS 18698 (Fig. 4G) coarse parallel-fibred bone is interspersed with lamellar bone throughout the entire cortex resulting in a much higher tissue organization when compared to the other Simosaurus bones. The smallest femur SMNS 18038 also shows an EFS in its outer cortex (Fig. 4H). All femora have locally a funnel-shaped arrangement of the crystallites (only visible in polarized light) enclosing simple vascular canals (Fig. 5A, B) as was also described for nothosaur humeri by Klein (2010). Femur SMNS 91983 has a thick ring of fibers that surround the entire medullary region and immediately follow the sharp line (Fig. 5). The innermost part is made of a layer of angled fibers that is followed by a layer of longitudinally (rhombic appearance) arranged fibers and ends in a layer of smaller longitudinal fibers (Fig. 5D). The fibers are short and oblique, very dense and had replaced the primary periosteal bone tissue. Thus, no primary bone tissue is visible between these fibers.

Remodelling occurs in all samples and is reflected in the form of erosion of primary osteons (secondary widening). Mature secondary osteons i.e. largely infilled secondary osteons (Currey, 2002) are rare and intermixed with primary osteons and secondarily widened osteons, in which the lamellar bone still surrounds a large canal (Fig. 4).

The cortex of Simosaurus is stratified by zones and annuli. Lines of arrested growth (LAGs) only occur within an EFS. Zones are identified as broad layers of vascularized and poorly organized coarse parallel-fibred bone tissue (Fig. 4D, G). Annuli are thin layers of highly organized and avascular bone tissue, consisting mainly of lamellar bone (Fig. 4D, G). The growth record in femora SMNS 18689 and SMNS 91983 is only locally preserved and not completely traceable along the cross section. Subcycles stratify the broad zones and resemble annuli (thin layers of avascular lamellar bone).

The growth record on which growth curve modelling was based, was measured on the ventral side of humeri SMNS 52095 and SMNS 18698 (Fig. 6A, B). After the fifth visible annulus, the tissue changes in SMNS 52095 to an EFS that contains five LAGs (Fig. 4C, F).

In femur SMNS 18689 the growth record is best preserved at the dorsopreaxial bone side (Fig. 6C). Annulus 3 and 4 each contain several resting lines (Fig. 6C, E, F). From annulus 6 onwards the annuli are regularly spaced. In femur SMNS 91983 the growth record is diffuse. Around five, more or less regularly spaced annuli are countable, but not traceable all around the cross section. They are additionally obscured by numerous subcycles throughout the entire cortex. The smallest sampled femur SMNS 18038 has the most distinct annual growth cycle record, which is preserved at its postaxial bone side (Fig. 6D). The growth cycle starts here with a broad zone that ends in an annulus. Between the second and the fourth growth cycle the zones and annuli are nearly equally broad (Fig. 6G, H). Growth cycle distance narrows down after the third visible growth cycle, possibly indicating the onset of sexual maturity. The zone of the fifth growth cycle consists only of a single row of vascular canals. The sixth and seventh growth cycles are embedded in an EFS and only recognizable under the microscope using polarized light (Fig. 6H).

For all specimens, the logistic, von Bertalanffy, and Gompertz growth models were successfully fitted to the growth record, but not all models were biologically realistic. The Gompertz models generally had negative growth rates, and the von Bertalanffy models obtained for SMNS 18689 and SMNS 18698 had negative growth rates and negative asymptotic masses (Table 2). For the Chapman-Richards growth model fitting always failed, presumably because of the large number of parameters of this model compared to the small number of annual growth cycles preserved.

The histological examination of humerus SMNS 52095 revealed ten visible annual growth cycles (Table 1). After the fifth visible annual growth cycle an EFS is observed. Two or three annual growth cycles are lost in the inner part of the bone, which suggests that the individual was between 12 and 13 years old when it died. As no considerable decrease in growth rate is preserved in the mid cortex of this humerus, the onset of maturity is not documented. Growth in SMNS 52095 was best explained by logistic growth models that basically differ in the number of growth marks assumed that are lost in the growth center (Table 1 and Table 2; Figs. 6A, 7A). The logistic growth model assuming that no (one, two, and three) growth marks are lost had a low residual standard error of 6.4 (5.5, 5.4, 5.4) and a low AIC value of 70.0 (73.3, 72.8, 72.6). The best von Bertalanffy growth model provided a distinctly poorer fit (res. s.e. = 9.9 and AIC = 79.4). Statistically, none of the four logistic models is preferred over any other (ΔAIC ≈ 2, Burnham and Anderson, 2002). Hatchling mass predicted by the logistic model that assumes that no growth cycle is missing was 8.68 ± 0.1 kg (s.e., standard error), which is unrealistic high compared to the hatchling mass of c. 400 g suggested by specimen SMNS 18698 (see below). It was zero and thus much closer to this hatchling mass under the other three logistic models. Estimated asymptotic mass did only slightly differ between fitted models, and ranged from 117.1 ± 4.1 kg (no cycle is missing) to 119.3 ± 3.8 kg (three cycles are missing). The ages at death predicted by the models ranged between 10 and 13 years, and the ages at which asymptotic mass is reached between 10 and 13 years, too, which is consistent with the presence of an EFS in the histological record. All models located the inflection point between the third and fourth growth mark, which would coincide with an onset of reproductive maturity between 3–4 years up to 6–7 years. The maximum growth rate for this bone was within its fourth year of life and was between 64 and 65 g per day (23.2 and 25.2 kg per year).

The small medulla in humerus 18698 suggests that the growth record is complete and no annual growth cycle is missing in its innermost cortex. Seven annual growth cycles are preserved suggesting that the individual died in its eighth year of life. As no EFS is documented in the bone, this individual was not in the asymptotic phase of growth when it died. The considerable decrease in growth rate after the second annual growth cycle could indicate sexual maturity. Growth in the humerus SMNS 18698 was clearly in terms of absolute and relative goodness of fit best described by a logistic model assuming that no growth mark is missing (Table 1, 2; Fig. 7A). This model estimated a hatchling mass of 570 ± 100 g, and an asymptotic mass of 125.7 ± 20.7 kg. The age at death of the individual is seven years and thus considerably lower than the age at which asymptotic mass is reached, which is 16 years. The inflection point of this logistic growth model is located between the fifth and sixth year of life, and thus does not coincide with the considerable decrease in growth observed after the second annual growth cycle. Maximum growth rate was within its sixth year of life and was 56 g per day (20.4 kg per year).

The histological examination of femur SMNS 18689 revealed ten annual growth cycles. Extrapolation suggested that three annual growth cycles are missing in the inner part of the bone. Thus, the individual was 13 years old when it died. After visible annulus 6 the growth marks are regularly spaced, but no EFS is visible. Four logistic models differing in the number of missing cycles assumed (0 up to 3 cycles) are well preferred over the linear model (quasi-linear phase of growth) due to their lower residual standard errors (Table 2). Contrary, the AIC based model selection indicated a very small support of the linear model over the logistic models in general and especially over the logistic model assuming that three growth cycles are missing (ΔAIC = 0.3; Table 2). The latter logistic model assumed a hatchling mass of zero, which is actually wrong from a biological point of view, but we think still tolerable when modelling growth for two reasons. First, the hatchling mass in Simosaurus is c. 400 g (as suggested by SMNS 18698 in which the first annual growth cycle is preserved) and thus small. Second, the estimated asymptotic masses of all specimens are two orders of magnitude larger than the hatchling mass. The estimated asymptotic mass of the SMNS 18689 individual derived from this logistic growth model is 129.2 ± 6.7 kg and its estimated age at death is 12 years. Its age at which asymptotic mass is reached is 20 years and thus consistent with bone histology as asymptotic age is considerably higher than the age at death. The inflection point (= onset of sexual maturity) of this logistic growth model is located between the sixth and seventh year. This suggests that the individual reached sexual maturity within its ninth year of life. The maximum growth rate derived from the logistic model is within its ninth year of life and is 44 g per day (15.9 kg per year). In total, for SMNS 18689 a best growth model was not unequivocally identified. Sigmoidal models assume a phase of acceleration of growth followed by a phase of deceleration of growth during the life of an individual. None of these phases is clearly documented in the growth record of SMNS 18698 as growth marks are missing in the inner part of the bone and the individual died before it reached the phase of growth deceleration.

The histological examination of femur SMNS 18038 revealed seven annual growth cycles. Extrapolation suggested that one cycle is missing in the inner part of the bone, and thus estimates that the individual died during its ninth year of life. After the fifth growth mark, an EFS is documented in the bone. Contrary to the other three bones, for which the logistic growth model was more or less preferred over the von Bertalanffy model, growth in SMNS 18038 was clearly best explained by a von Bertalanffy growth model in terms of absolute and relative goodness of fit (Table 2; Fig. 7A). This is unexpected as the number of growth marks documented in the bone is small (seven), and the AIC based model selection approach favors models with fewer parameters (two for the linear model) over models with more parameters (three for the von Bertalanffy model). The von Bertalanffy model predicts that no annual growth cycle is missing in the inner part of the bone, but estimates an unrealistic high hatchling mass of 10.5 ± 1.5 kg compared to c. 400 g suggested by SMNS 18698. The model predicts an age at death of seven years. The estimated asymptotic mass of the individual is 109.1 ± 15.4 kg, which is reached by the individual after 20 years of life. The inflection point of the growth model is located between the first and second year, which contradicts the histological examination, and is inconsistent with the ages at sexual maturity found for the other specimens (Table 1). However, this reflects the general earlier sexual maturity under a von Bertalanffy (age at which 30% of the asymptotic mass is reached) than under a logistic growth model (age at which 50% of the asymptotic mass is reached). The maximum growth rate for this bone was within its second year of life and was 36 g per day (13.1 kg per year). That SMNS 18038 follows the von Bertalanffy growth model could be related to its earlier ontogenetic stage and the absence of a sufficient number of growth marks documenting the asymptotic phase of growth. However, different individuals from a single species can follow different growth models (Frazer and Ehrhart, 1985, Halliday and Verrell, 1988, Magnusson and Sanaiotti, 1995 and Ritz et al., 2010). This reflects developmental plasticity or variability in growth due to differences in exogenous (climate, food or water availability) and endogenous factors (stress, diseases) experienced by individuals. Although all Simosaurus bones sampled originate from bone beds and were thus not found together with diagnostic skull material, an incorrect taxonomical assignment of SMNS 18038 can be excluded. Long bone morphology of Simosaurus differs distinctly from that of other Euauropterygia found at the same localities (Rieppel, 1994 and Rieppel, 2000).

Simosaurus humeri and femora show exactly at midshaft a free medullary cavity. Distally or proximally to the midshaft the bones have a medullary region dominated by secondary trabeculae, which is also seen in placodonts (Klein et al., 2015a). The medullary cavities are lined by a thick layer of endosteal bone and are surrounded by a thick compacta, both resulting in bone mass increase, and more specifically osteosclerosis (Houssaye, 2009), when compared to terrestrial tetrapods. Osteosclerosis is also evident due to the presence of extensive calcified cartilage (incomplete endochondral ossification) in Simosaurus along the midshaft (Ricqlès and Buffrénil, 2001), (Fig. 3, Fig. 4 and Fig. 5A, B). Although our sample size is limited, there is evidence that the medullary cavity in Simosaurus is larger in femora than in humeri, a tendency that is also documented in placodonts (Klein et al., 2015a and Klein et al., 2015b). Medullary cavities in humeri and femora are equally large in the pachypleurosaur Anarosaurus heterodontus (Klein, 2010 and Klein, 2012), but always larger in humeri of the Upper Muschelkalk nothosaurs than in their femora (Klein, 2010 and Krahl et al., 2013), which indicates different swimming styles. Microanatomical differences between the humeri and femora in Simosaurus indicate different functional requirements for both elements. This can be expressed for example, in that the humerus is more intensively used during swimming (e.g., for propulsion) than the femur.

A Principal Component Analyses (PCA) plot, which is based on Bone Profiler values and which included three Simosaurus midshaft samples (humerus SMNS 18698, femur SMNS 18038, femur SMNS 18689) as well as other aquatic extinct and extant taxa revealed different microanatomical clusters (Klein et al., 2015a: figure 8). Herein the Simosaurus humerus SMNS 18698 plots close to the marine reptile Horaffia, Placodontia indet aff. Cyamodus, Pistosaurus, and the pachypleurosaur Anarosaurus heterodontus. This group again is clearly separated from the Upper Muschelkalk nothosaurs as well as from the other aquatic taxa included that are highly or poorly efficient swimmers. Due to the difference in the size of their medullary cavities, both femora of Simosaurus generally have a large distance in this plot. Additionally, their distance to the femora of nothosaurs from the Upper Muschelkalk is large. Thus, microanatomical analysis suggests that Simosaurus has had certain active swimming skills as presumed for the above mentioned taxa, but was less efficient when compared to contemporaneously living and similar sized nothosaurs from the Upper Muschelkalk. If it was “capable of sustained swimming” (Rieppel, 1994: 1) was neither confirmed nor contradicted by the analysis of Klein et al. (2015a).

The most distinct feature of Simosaurus long bone histology is the coarse parallel-fibred bone tissue that is well interspersed with numerous and thick osteocytes (Fig. 4). Coarse parallel-fibred bone was recently described for some placodonts (Klein et al., 2015a) and for the temnospondyl Plagiosaurus (Konietzko-Meier and Schmitt, 2013). Houssaye et al. (2013) described an unusual parallel-fibred bone for mosasaurs. In placodonts, this coarse parallel-fibred tissue is intermixed with woven bone in a fibro-lamellar bone tissue. Plagiosaurus shows in its deepest cortex woven bone and immature primary osteons forming incipient fibro-lamellar bone (Klein, 2010 and Konietzko-Meier and Schmitt, 2013), whereas neither mosasaurs (Houssaye et al., 2013) nor Simosaurus show fibro-lamellar bone. Houssaye et al. (2013) related the unusual parallel-fibred bone of mosasaurs to an increase in growth rate, an interpretation that is also realistic for Simosaurus. The presence of modified parallel-fibred bone tissue in mosasaurs and Simosaurus thus could represent a kind of pre-stage in the formation of fibro-lamellar bone. Coarse parallel-fibred bone is so far restricted to aquatic tetrapods. Some terrestrial taxa show a modification of the fibro-lamellar complex by parallel-fibred bone. This was described for early pseudosuchian archosaurs (Ricqlès et al., 2003), the ornithopod dinosaur Gasparinisaura (Cerda and Chinsamy, 2012), and the titanosaur dinosaur Ampelosaurus (Klein et al., 2012a). In all these taxa the woven bone component of the fibro-lamellar complex was replaced or modified (intermixed) by parallel-fibred bone, resulting in a lower growth rate.

The bone tissue type of Simosaurus can be categorized as lamellar-zonal, although it is not typical lamellar-zonal when compared to modern amphibians and reptiles. This is mainly due to its moderately high vascular density and due to the presence of coarse parallel-fibred bone. Although similar to nothosaurs (Klein, 2010; NK pers. obs.) Simosaurus has a unique combination of bone tissues as well as vascular density and organization when compared to other Sauropterygia (e.g., Hugi, 2011, Hugi et al., 2011, Klein, 2010, Klein, 2012, Klein et al., 2015a, Klein et al., 2015b, Krahl et al., 2013, Ricqlès and Buffrénil, 2001, Sander, 1990 and Wiffen et al., 1995).

The thick inner ring composed of three successive layers of fibers (angled, large longitudinal, small longitudinal) in femur SMNS 91983 is very peculiar. None of the other Simosaurus samples shows a comparable fiber deposition. These fibers do not represent Sharpey's fibers, which are long, run through different bone layers and represent muscle and tendon attachments of gross morphology. In temnospondyls, similar fibers as described above for SMNS 91983 are interpreted as attachment fibers of the periost (Konietzko-Meier and Klein, 2013 and Witzmann, 2009) or as ossified tendons in the cervical ribs in sauropod dinosaurs (Klein et al., 2012b). However, these explanations are not satisfying here. A pathological reason such as a false posture or a deficiency in early ontogeny of this individual could be a more likely explanation. The femur itself, however, does not show any abnormalities such as a healed break that would be indicated by a callus.

In general, annual growth cycles in Simosaurus consist of broad zones alternating with thin annuli. Two bones show an EFS in their outer cortex. However, a certain variability and developmental plasticity in the sequence and spacing of growth marks of individuals are obvious, which is similar to the variation observed in placodonts (Klein et al., 2005b). In femur SMNS 18689 two annuli contain several resting lines indicating several cessations of growth during these two slow annual growth phases. These additional cessations of growth during an annual growth cycle can be the result of exogenous (climate, food or water availability) or endogenous factors (stress, diseases). Non-annual growth marks are known from many modern amphibians and reptiles (e.g., Castanet, 1994 and Castanet et al., 1993). Additional non-annual resting lines within a single annulus are described in extinct taxa. They were observed in placodonts (Klein et al., 2015b) and in the temnospondyl Metoposaurus (Konietzko-Meier and Klein, 2013).

Hugi and Sánchez-Villagra (2012) report that marine iguana Amblyrhynchus cristatus had thicker zones than annuli before sexual maturity was reached, whereas zones were equally broad after the onset of sexual maturity, documenting that growth mark expressions can also depend on endogenous factors such as reproduction. A similar change in the proportion between zones and annuli is documented in femur SMNS 18038. However, the possibility that this change in proportion reflects the onset of sexual maturity remains questionable here (see above; Table 1).

Two approaches are applied to compare growth rates of Simosaurus to that of other marine and extant reptiles. First, bone tissue type as well as vascular density and organization of Simosaurus were compared to those of other Sauropterygia and marine reptiles, which both being a proxy for growth rates. Low tissue organization and high vascular density as well as the presence of radial vascular canals and a radial vascular canal organization indicate high growth rates in Simosaurus (e.g., Margerie et al., 2004 and Ricqlès, 1976). These histological characters suggest that the growth rate of Simosaurus was intermediary between those of other sauropterygians and marine reptiles. Compared to nothosaurs, the tissue organization was lower due to the presence of coarse parallel-fibered bone and vascular density was increased in Simosaurus (Table 1; Hugi, 2011, Klein, 2010 and Krahl et al., 2013; NK pers. obs.). Compared to the pachypleurosaurs Neusticosaurus spp., (Hugi et al., 2011 and Sander, 1990) that have parallel-fibred bone and show a mainly longitudinal vascular canal organization, Simosaurus had higher growth rates. The bone tissues of Simosaurus were more organized and vascular density was lower when compared to the pachypleurosaur Anarosaurus, because the latter grew with incipient fibro-lamellar bone and show a high amount of radial vascular canals (Table 1; Klein, 2010 and Klein, 2012). Bone tissues and vascular canal organization of Simosaurus were similar to mosasaurs (Houssaye et al., 2013), but much lower when compared to placodonts (Buffrénil and Mazin, 1992 and Klein et al., 2015a) or ichthyosaurs (Buffrénil and Mazin, 1990; Houssaye et al., 2014) both taxa having highly vascularized fibro-lamellar bone.

Second, maximum growth rates were estimated from the inflection point of fitted growth curves (Fig. 7A). These growth rates are compared to those of extant reptile species (Fig. 7B, regression line and species-specific maximum growth rates, data from Werner and Griebeler, 2014). The marine reptile Simosaurus exhibits a higher maximum growth rate when compared to modern average reptiles (Fig. 7B; Werner and Griebeler, 2014), which is also seen in other extinct marine reptiles (e.g. Buffrénil and Mazin, 1990, Buffrénil and Mazin, 1992, Houssaye, 2013, Klein et al., 2015a and Klein et al., 2015b). In modern mammals, marine species also have higher growth rates than terrestrial species (White, 2011). The increased growth rate in Simosaurus, however, is still consistent with the variability in rates seen in extant reptiles (Fig. 7B). Growth rates in extant ectothermic animals reflect ambient temperature conditions with tropical species usually having higher rates than temperate species. Simosaurus lived in a shallow sea in a warm subtropical climate presumably allowing higher growth rates.

Logistic growth models best described growth in the Simosaurus specimens studied (for SMNS 18698 the linear model was similar well supported, Table 2), except for the femur SMNS 18038, in which growth follows the von Bertalanffy model. Nevertheless, both models were applicable to all specimens (Table 2). Both sigmoidal standard models have also been successfully used to describe growth in extant reptiles. Moreover, both models were applicable to two different individuals from a single species (Halliday and Verrell, 1988, Magnusson and Sanaiotti, 1995 and Ritz et al., 2010). The von Bertalanffy model estimated for SMNS 18038 a hatchling mass of 10.5 ± 1.5 kg. However, this mass is unrealistic high compared to SMNS 18698 (400 g), and could strongly question the biological reliability of this von Bertalanffy model. Alternatively, the high hatchling masses estimated by the growth models for SMNS 18689, but also for SMNS 18038 and SMNS 52095 (Table 2), could indicate ovoviviparity in Simosaurus and a large mass of the offspring. The von Bertalanffy growth model predicted for SMNS 18038 that the individual died during its eight year of life, whereas the growth record suggested that it lived at least one year longer. The logistic models found for SMNS 52095, SMNS 18698, and SMNS 18689 estimated ages at death of individuals between 7 and 13 years, which all agree with the histological record. The observed low longevities in Simosaurus specimens could indicate a stronger predation pressure on this taxon than seen in similar-sized extant reptiles. However, longevities are expectancies at the population level that cannot be reliable estimated from few individuals. The predicted asymptotic masses of specimens ranged between 113 and 129 kg, and the derived ages at which asymptotic masses were reached ranged between 10 and 20 years. The estimated ages at which the asymptotic mass is reached are considerably lower than those seen in large extant species (Alligator mississippiensis: 150 kg, 73 years in captivity, Caretta caretta: 156.5 kg, 37 years, Chelonia myda: 160 kg, 75 years, Varanus komodoensis: 87 kg, 62 years; data from database AnAge, Tacutu et al., 2013). However, all estimates on asymptotic masses and ages in Simosaurus are not well constrained as none of the specimens studied well covers the asymptotic phase of growth. Especially, estimated asymptotic masses of SMNS 18698 and SMNS 18038 have high standard errors (Table 2). An analysis of the growth record of fully-grown individuals could reveal more precise asymptotic masses and ages in Simosaurus than the individuals could that were only available for this study. This would also allow to proof whether longevities and asymptotic ages are indeed lower than in extant similar-sized reptiles. The predicted onsets of sexual maturity (location of the inflection point) of specimens were between 3 and 9 years. These estimated ages at the onset of sexual maturity in Simosaurus specimens are more or less consistent with those of similar-sized (body mass) extant aquatic and semi-aquatic reptiles (Alligator mississippiensis: 150 kg, 12 years; Caretta caretta: 156.5 kg, 10 years; Chelonia mydas: 160 kg, 10 years; Varanus komodoensis: 87 kg, 5 years; data from database AnAge, Tacutu et al., 2013). However, given that asymptotic ages in Simosaurus could have been underestimated in the current study they could contrary indicate an earlier maturity in Simosaurus than in similar-sized extant reptiles or a sexual dimorphism as was already suggested based on morphology (Rieppel, 1994). The current observations on the life history of Simosaurus questions whether extant aquatic and semi-aquatic reptiles are good models for Simosaurus and other Sauropterygia (Klein et al., 2015b). It could indicate differing life history strategies in extinct marine reptiles and modern forms. However, this hypothesis has to be tested in the future based on much more samples of Simosaurus.