Ontogenetic, dietary, and environmental shifts in Mesosauridae

Material investigated

We examined 270 articulated mesosaur specimens from collections around the world. Of these, 115 were previously attributed to Mesosaurus tenuidens, 96 to “Stereosternum tumidum”, 13 to “Brazilosaurus sanpauloensis”, and 46 were unidentified (Table S1). Of the identified specimens, 47 had a preserved skull and teeth, with 18 attributed to Mesosaurus, 29 to “Stereosternum”, and only one to “Brazilosaurus”. Earlier studies mentioned the presence of teeth in “Brazilosaurus” (Rossmann, 2002; Modesto, 2006; Silva et al., 2017), all of them relying on the average tooth length of 2 mm provided by Rossmann (2002). We find this value contentious in different regards. First, Rossmann (2002) did not provide the raw measurements for his calculation. Second, we reexamined the material described by Rossmann (2000); Rossmann (2002) as bearing teeth and noticed that elements he describes as teeth are in fact misidentified broken tooth sockets from the maxilla or the dentary (Fig. S1). Finally, out of the twelve specimens assigned to “Brazilosaurus” studied here, only ROM 28496 had some visible complete teeth. In all other specimens, teeth are either missing or completely broken, in any case not in a state of preservation allowing their measurement. Consequently, we must dismiss the average tooth length for “Brazilosaurus” of two mm given by Rossmann (2002). For this reason, no specimens previously attributed to “Brazilosaurus” are included in our statistical analysis of teeth.

Measurements

For each specimen, we measured the length of most preserved long bones: humerus, radius, ulna and metacarpals in the forelimb; femur, tibia, fibula, and metatarsals in the hind limbs. When possible, we also measured autopod length, from the proximal edge of the intermedium to the distal tip of digit III for the forelimb, and from the proximal edge of the astragalus to the distal tip of digit V in the hind limb, as digit III and V are the longest digits in the anterior and posterior autopod, respectively. In juveniles where carpals and tarsals were not ossified, the distalmost portions of the ulna or fibula were taken as most proximal extent of the autopodium. To account for cranial proportions, we measured the total skull length as well as antorbital and postorbital length.

Likely due to dental replacement, mesosaur teeth can greatly vary in size within a single specimen (Modesto, 1999; Modesto, 2006; Piñeiro et al., 2021). To better account for the dimensions of mesosaur teeth, we therefore chose not to use average tooth length and diameter, unlike previous works (Piñeiro et al., 2021). Instead, we measured the length of the longest visible tooth and the diameter of this tooth at its base.

Our specimens were preserved in various states of completeness, with some showing only the anterior or the posterior half of the body and others missing limbs or neck while others were mostly complete. Unlike Núñez Demarco, Ferigolo & Piñeiro (2022) who studied ontogenetic variation within bones, we wanted a shared proxy for specimen size to be able to model and compare variations in all bones independently. Thus, we calculated the average centrum length of dorsal vertebrae for each specimen, as this measurement has been shown to be a reliable estimate for body length (Núñez Demarco et al., 2018).

All measurements were taken using a caliper directly on specimens or Fiji 1.53c on distortion-corrected high-resolution photographs.

Facies

Mesosaur specimens can be found in various deposits that represent different sections of the carbonate ramp boarding the Irati-Whitehill Sea (Xavier et al., 2018; Ng, Vega & Maranhão, 2019). To account for size distribution across deposit environments and potential niche partitioning of ontogenetic stages in mesosaurs, we categorized our specimens into five facies: (i) yellow limestone, (ii) white calcarenite, (iii) grey siltstone, (iv) grey mudstone, and (v) bituminous black shales. These facies represent a gradient in depth, energy level, and oxygenation of the water layers.

Statistics

According to Huxley (1932), simple allometry is described by the equation ${\displaystyle y=k{x}^{\alpha }}$ or in its logarithmic form ${\displaystyle logy=\alpha logx+logk.}$

The allometric coefficient α characterizes the slope of the linear relationship between the logged valued of y and x and log k is the intercept of the linear relationship. Thus, an isometric growth (no changes in proportion to size during ontogeny) is characterized by an allometry coefficient of α = 1 while values of α >1 and α <1 respectively reflect positive and negative allometry.

By transforming the simple allometry equation to express the ratio y/x rather than y as a function of x, we obtain ${\displaystyle \frac{y}{x}=k{x}^{\alpha -1}}$ which in turns gives ${\displaystyle log\frac{y}{x}=\left(\alpha -1\right)logx+logk.}$

In this equation we can define another allometry coefficient β = α − 1. Thus, for the same values of x and y, isometry is defined by β = 0 while negative and positive allometry are respectively characterized by negative and positive values of β, improving readability both graphically and numerically (Klingenberg, 1998). Therefore, we prefer β to α in the present study.

For each metric, we computed a linear regression model of the logged ratio of the metric’s values over size proxy value ( $log\frac{y}{x}$) as a function of logged size proxy values (log x) and calculated the allometry coefficient β to test for proportion changes throughout ontogeny. Similarity between species was also tested for each measurement. Linear modelling and statistical analyses were carried out in R 4.0.5 (R. Core Team, 2021).

Three-species hypothesis

Skull

Skull measurements behave differently with respect to size in the three formerly recognized mesosaur “species”. In “Brazilosaurus”, all three skull measurements show negative allometry, albeit not significant (Table 1). In Mesosaurus, postorbital length is the only skull metric to display a significant negative allometry, while total length and antorbital length both show a non-significant positive allometry (Table 1). In “Stereosternum”, all three skull measurements show significant negative allometry but each with different allometry coefficients (Fig. 2). The negative allometry is less marked for antorbital length (β =  − 0.090, p = 0.022) than for total length (β =  − 0.204, p < 0.001), itself less marked than postorbital length (β =  − 0.341, p < 0.001).

For all three species, skull measurements rank similarly in terms of allometry coefficients with antorbital length having the highest, followed by total length and finally postorbital length last. This shows how skull elements vary differently in proportion throughout ontogeny. To better account for these variations, we measured the effect of allometry in function of total skull length instead of body size. In this configuration, we obtain very concordant results between the three species: antorbital length shows significant positive allometry, whereas postorbital length shows significant negative allometry (Table 2). Thus, over the course of ontogeny, the postorbital length decreases while the snout increases relative to overall skull length (Fig. 3).

Table 1:

Allometry coefficient of logged size ratios for teeth and skull measurements.

	Allometry coefficient β	SE	t-value	p-value
“B. sanpauloensis”
Tooth length	–	–	–	–
Tooth diameter	–	–	–	–
Skull total length	−0.462	0.290	−1.596	0.149
Skull antorbital length	−0.509	0.362	−1.405	0.198
Skull postorbital length	−0.586	0.491	−1.193	0.267
M. tenuidens
Tooth length	0.286	0.417	0.685	0.504
Tooth diameter	0.466	0.338	1.380	0.188
Skull total length	0.094	0.140	0.268	0.792
Skull antorbital length	0.204	0.160	1.279	0.222
Skull postorbital length	−0.511	0.275	−2.308	0.037
“S.” tumidum
Tooth length	−0.381	0.115	−3.310	0.003
Tooth diameter	−0.787	0.096	−8.179	<0.001
Skull total length	−0.256	0.050	−5.131	<0.001
Skull antorbital length	−0.177	0.064	−2.770	0.008
Skull postorbital length	−0.376	0.074	−5.076	<0.001
Single-species hypothesis
Tooth length	0.029	0.149	0.191	0.849
Tooth diameter	−0.653	0.106	−6.190	<0.001
Skull total length	−0.203	0.050	−4.090	<0.001
Skull antorbital length	−0.122	0.066	−1.835	0.070
Skull postorbital length	−0.392	0.062	−6.296	<0.001

DOI: 10.7717/peerj.13866/table-1

Notes:

Significant values (p < 0.05) are indicated in bold.

Table 2:

Allometry coefficients of logged tooth and skull measurements with respect to total skull length.

	Allometry coefficient β	SE	t-value	p-value
“B. sanpauloensis”
Tooth length	–	–	–	–
Tooth diameter	–	–	–	–
Skull antorbital length	0.134	0.086	1.563	0.157
Skull postorbital length	−0.925	0.522	−1.771	0.114
M. tenuidens
Tooth length	−0.051	0.276	−0.186	0.855
Tooth diameter	−0.103	0.260	−0.398	0.696
Skull antorbital length	0.187	0.058	3.206	0.005
Skull postorbital length	−0.590	0.156	−3.785	0.001
“S. tumidum”
Tooth length	−0.091	0.097	−0.936	0.357
Tooth diameter	−0.678	0.107	−6.336	<0.001
Skull antorbital length	0.121	0.021	5.761	<0.001
Skull postorbital length	−0.172	0.076	−2.249	0.029
Single-species hypothesis
Tooth length	0.282	0.115	2.456	0.018
Tooth diameter	−0.558	0.092	−6.082	<0.001
Skull antorbital length	0.147	0.017	8.506	<0.001
Skull postorbital length	−0.315	0.059	−5.314	<0.001

DOI: 10.7717/peerj.13866/table-2

Notes:

Significant values (p < 0.05) are indicated in bold.

Teeth

Several authors have noted an apparent change in tooth morphology between mesosaur species: “Brazilosaurus” and “Stereosternum” are described as bearing short, straight conical teeth whereas Mesosaurus is defined by thin, curved, and elongated teeth (Oelofsen & Araújo, 1987; Rossmann, 2002; Modesto, 2006). Indeed, for the same body size, adult Mesosaurus teeth are systematically longer than those of “Stereosternum” albeit similar in diameter (Figs. 2 and 3).

Tooth length and diameter in Mesosaurus both display positive allometry, although the linear relationship is not significant (Table 1). In “Stereosternum”, both measurements show a significant negative allometry, with tooth diameter showing the lowest coefficient (β =  − 0.622, p < 0.001).

Since our results highlight variation in cranial dimensions throughout ontogeny, we also calculated allometry coefficients for tooth measurements with respect to skull length. Surprising, this shows non-significant negative allometry for tooth length in both species, whereas tooth diameter exhibits negative allometry for both species, although it is only significant in “Stereosternum” (Table 2). Thus, mesosaur teeth grow in length at the same rate as the skull, meaning that they do not become proportionally longer throughout ontogeny. However, these teeth keep their juvenile diameter and grow only in length in “Stereosternum” whereas both length and diameter grow isometrically to the skull in Mesosaurus (Fig. 4).

Single-species hypothesis

In a recent study, Piñeiro et al. (2021) proposed Mesosaurus tenuidens to be the only valid species within the clade by statistically rejecting most characters used to discriminate between the three previously recognized mesosaur species. To account for this new hypothesis, we reproduced our analysis with all specimens (re)assigned to Mesosaurus.

Limbs

There is a strong trend to hind limb reduction when considering a single mesosaur taxon (Fig. 5). All eight long bone measurements display negative allometry coefficients in relation to body size and this relationship is significant for seven of them (Table 3D). Only metatarsal I length is not significantly isometric in growth. Posterior autopod length also shows a significant negative allometry.

In contrast, forelimb long bones do not show such a reduction in size. Only metacarpals display significant negative allometry, with the exception of metacarpal I that has a non-significant positive allometry coefficient. Thus, while stylopod and zeugopod do not vary in proportions, it appears the manus is reduced during ontogeny.

At first glance, it could seem that these results do not match those of Núñez Demarco, Ferigolo & Piñeiro (2022), who report isometric growth in mesosaur limbs. Nevertheless, our findings are compatible with theirs. What Núñez Demarco, Ferigolo & Piñeiro (2022) measure is that long bones grow isometrically with respect to their respective stylopod, humerus length for fore limbs and femur for hind limbs. What we measure is the allometry of these bones with respect to size. Here we show that femur length displays a significant negative allometry with respect to size, and so do most other limb bones. In sum, the isometric growth measured by Núñez Demarco, Ferigolo & Piñeiro (2022), and by extension their other results for limbs, is likely an effect of using different reference frames.

Environmental distribution

Figure 6 shows the distribution of size classes in the main facies of the Irati-Whitehill Sea. In our sample, small mesosaurs (i.e., juveniles) are overrepresented in limestone, although they also occur in siltstone and black shales. Mid-sized mesosaurs are well represented in all facies despite frequency variations due to sample size differences. Contrary to what has been hypothesized (Núñez Demarco et al., 2018), we do not find that large specimens are only found in the marginal regions of the basin. Instead, in our dataset, very large specimens (vertebral length ≥ 10 mm) are exclusively found in mudstone and black shales, facies that correspond to deeper, more pelagic, and more anoxic environments (Xavier et al., 2018; Bastos et al., 2021).

Missing juveniles and single-species hypothesis

One of the most significant results of our study is not found with regards to size and proportional changes in mesosaur anatomy but instead in the distribution of body size itself. Despite the 270 specimens examined here, there are a number of surprising gaps in the size range covered by our sample (Fig. 7). First, the few “Brazilosaurus” specimens studied here are concentrated in the adult size range of “Stereosternum”, but no juveniles or subadults are preserved. Second, there is a clear divide between the youngest “Stereosternum” specimens on the one hand and subadults and adults of that species on the other hand. Third, and more importantly, only two specimens of Mesosaurus are the size of juvenile “Stereosternum”, whereas all other specimens range from the size of “Stereosternum” subadults to larger sizes than the largest “Stereosternum” specimens. These gaps in size distribution have a direct effect on the interpretation of allometric measurements as they alter the results of linear regressions, making them less representative of ontogenetic changes.

Piñeiro et al. (2021) highlighted that most features employed to define mesosaur species are not statistically supported or are practically unusable for identification. Characters like the shape of the interclavicle, tooth length, or presacral vertebral count are highly preservation-dependent, whereas others such as the degree of pachyostosis of ribs or the skull-neck length ratio are subjective and ontogenetically variable (for pachyostosis, see also Klein et al., 2019). This lack of reliable characters led the authors to suggest that both “Stereosternum tumidum” and “Brazilosaurus sanpauloensis” are junior synonyms of Mesosaurus tenuidens.

The ontogenetic variability of supposedly taxonomic characters led to an age partitioning of mesosaur specimens into the different “species”. On the one hand, juvenile and subadult specimens tended to be identified as “Stereosternum”, because they bear shorter conical teeth and are usually preserved complete, displaying the maximum number of presacral vertebrae. Both features were formally admitted as prominent characters for mesosaur identification.

On the other hand, the largest mesosaur specimens bear longer, thinner teeth and have more markedly pachyostotic ribs, and they tended historically to be attributed to Mesosaurus. As for “Brazilosaurus” specimens, they were usually identified on the basis of their cervical vertebral count, which is however a misidentification as the result of a displaced rib cage (A. Verrière, 2022, pers. obs.). In fact, even the original description of the species (Shikama & Ozaki, 1966) miscounted the number of cervicals due to the posterior displacement of dorsal ribs (A. Verrière, 2022, pers. obs.), which led to the erroneous erection of this feature as a defining character. To summarize, there are several reasons as to why mesosaur invalidated “species” occupy different regions of the size range.

Rather than a sampling bias, the size distribution gaps in our dataset more likely reflect a species-identification bias. Our data shows consequent overlap in the morphometrics of all three species. “Brazilosaurus” is virtually indistinguishable from “Stereosternum”, whereas the old definition of Mesosaurus seems to have mostly representatives on the higher side of the size spectrum (Fig. 7). Furthermore, the low significance of allometry coefficients (Tables 1–3) is also likely an effect of the small number of juveniles of the Mesosaurus type in our dataset, although more juvenile specimens have been described that we could not include (Piñeiro et al., 2012a).

Our results therefore corroborates the single-species hypothesis of Piñeiro et al. (2021) and their discarding of the former characters as ontogenetically variable. In fact, the three previously identified “species” more probably represented three artificial types: the “Brazilosaurus”-type is a poorly preserved mesosaur with displaced ribs, the “Stereosternum”-type is a size class encompassing juveniles to young adults, and the “Mesosaurus”-type represents adults with more extreme sizes and morphologies. Consequently, hereafter we will only consider our results in the light of the single-species hypothesis.

Environmental partitioning

Even when considering Mesosaurus as the only valid mesosaur taxon, one puzzling difference between the previously distinct types cannot be explained as an effect of ontogenetic changes or dubious characters alone. Some authors pointed out that the three types were not equally distributed across the facies of the Irati-Whitehill Sea. Oelofsen & Araújo (1983) and Rossmann (2002) noted that “Brazilosaurus” and “Stereosternum” were more commonly found in limestone deposits, whereas Mesosaurus was more frequent in oily black shales. Since these publications however, mesosaur specimens have been described in other facies and the former species have been synonymized, meaning their observations must be considered with caution.

Nevertheless, our study demonstrate that size classes are not homogeneously distributed across lithological facies: juveniles are overrepresented in limestone, and very large/old specimens in black shales (Fig. 6). As argued above, flaws in the formerly admitted diagnostic characters have led to juveniles and young adults being preferentially assigned to “Stereosternum” and old adults to Mesosaurus (Fig. 7). Because we look at articulated specimens, it is also fair to assume that the specimens were buried relatively close to their life environment. Therefore, what Oelofsen & Araújo (1983) and Rossmann (2002) noticed were likely the signs of ontogenetic stage partitioning rather than species partitioning.

The overrepresentation of juveniles and large adult in some facies suggests that mesosaurs preferentially occupied some environments at different stages of their life. Juveniles were mostly present in the shallower, more coastal sections of the carbonate ramp, while medium-sized mesosaurs were present in all areas, and the largest and oldest individuals were restricted to pelagic and anoxic regions (Fig. 6). Thus, if mesosaurs were not confined to a specific region of their environment, it seems that, overall, they gradually moved away from the coast during their growth.

The pelagic migration of mesosaurs with growth raises the question of their reproduction. Associated articulated juveniles and adults, even gravid individuals, have been found in limestone, siltstone, and mudstone (Piñeiro et al., 2012a; Bickelmann & Tsuji, 2018), but the greater frequency of juveniles in limestone suggests that mesosaurs were mostly born near the coast. If they were oviparous like most reptiles, this would have been easily explained by the necessity to lay eggs on land. However, evidence suggests instead that mesosaurs were viviparous, and possibly even cared for their young (Piñeiro et al., 2012a). Mesosaurs could therefore have given birth anywhere in the sea, making the causes of a partitioning less clear. To try to explain this, we must look at our morphometric analyses.

Ontogenetic changes and diet shifts

Our results show that mesosaur growth is marked by several morphological changes: an elongation of the teeth relative to the skull (Fig. 3), an elongation of the snout (Fig. 4), and a reduction of limbs (Fig. 5). Limb reduction throughout ontogeny hints at a difference in locomotory habits between juveniles and adults. Mesosaurs are swimming reptiles, and the degree of involvement of the limbs in locomotion is still debated. Some authors support the idea of mainly tail-driven mesosaurs (Braun & Reif, 1985; Villamil et al., 2016) while others suggested a potentially larger role of limbs in propulsion (Da Silva & Sedor, 2017; MacDougall et al., 2020). Yet, the ontogenetic limb reduction we expose seems incompatible with the conception that limbs were crucial for locomotion. Instead, our results provide evidence for a decreasing involvement of limbs in mesosaur swimming and weigh in favor of a tail-driven model of mesosaur swimming throughout their ontogeny.

The association of limb reduction with cranial and dental modifications strongly suggest the existence of locomotion and dietary shift in mesosaur. Such shifts are not rare in reptiles but are usually associated with dietary partitioning between juveniles and adults, meaning they exploit different food sources at different stages of their growth (Arthur, Boyle & Limpus, 2008; Gignac & Erickson, 2015; Dick, Schweigert & Maxwell, 2016; Wang et al., 2017). In the case of mesosaurs, a change in diet could explain the environmental migration we observe in size classes (Fig. 6). As mesosaurs grew, they would have shifted from an abundant food source in the coastal environment to another one more prevalent in pelagic regions or other water layers. Preying on different food sources in different water layers would have been a way to avoid competition between ontogenetic stages in a sea where resources were scarce. Different feeding habits would also explain the presence of juveniles in pelagic environments, as young mesosaurs could have been born anywhere in the sea but would have preferentially migrated to the coastal region looking for food. Yet, there remains the question of the alternative food source.

Traditionally, mesosaurs are considered filter-feeders that preyed on pygocephalomorph crustaceans, based on stratigraphic occurrences (Oelofsen & Araújo, 1983; Piñeiro et al., 2012b; Xavier et al., 2018; Ng, Vega & Maranhão, 2019) as well as on gastric contents and coprolites (Piñeiro et al., 2012b; Ramos, 2015; Silva et al., 2017). While this applies to adults, it is unclear what the diet of juveniles could have been. Mesosaur-bearing strata of the Irati-Whitehill Sea deposits are rich in mesosaurs and pygocephalomorphs remains but scarce in other types of fossils. Only rare invertebrate traces, bivalves, and paleonisciform fish scales have been described (Oelofsen & Araújo, 1983; Piñeiro et al., 2012b; Matos et al., 2017; Xavier et al., 2018), although recent works have revised the abundance of paleonisciform fishes upwards (Xavier et al., 2018). This raises a puzzling question: if juvenile mesosaurs did not have the same diet as adults, what could they possibly have been eating?

Our morphometric analysis shows that young mesosaurs would have had longer limbs, shorter snouts and shorter teeth than adults, a morphology better suited to active hunting than that of adults. Among the known Irati-Whitehill fauna, only pygocephalomorph crustaceans and paleonisciform fishes could have been the subject of active hunting. Thus, juvenile mesosaurs could have actively preyed on small fish or large pygocephalomorphs, while mesosaurs would have slowly filtered waters to catch smaller pygocephalomorphs or even larvae. The coastal environment would have been better oxygenated and richer in nutrients, fostering the growth of larger preys, whereas smaller crustaceans could have occupied the harsher pelagic waters. Unfortunately, the ecology of pygocephalomorph crustaceans or paleonisciform fish is very poorly known. One way of testing this hypothesis would be to examine size distribution across facies for fish and crustaceans and see if there is a trend, similar to what we did for mesosaurs.

Another possible answer is that juvenile mesosaurs fed on animals that did not preserve well under the conditions where mesosaurs were buried. High energy conditions in the mid-coastal environments as well as acidic anoxic waters in the black shale environment might have hampered the preservation of some animals. Moreover, young mesosaurs could have eaten jellyfish or soft invertebrates with little to no mineralized tissues, which greatly reduces the chances at fossilizing the latter. Although the sea’s hypersaline conditions would not have favored such animals (Piñeiro et al., 2012b; Bastos et al., 2021), there are occurrences of soft invertebrates in hypersaline contexts (Parker, 1959; Tanner, Glenn & Moore, 1999; Gilabert, 2001). Yet, there are many cases of outstanding preservation in the Irati-Whitehill sediments producing complete articulated fossils (Modesto, 1999; Modesto, 2010; Piñeiro et al., 2012b; Núñez Demarco et al., 2018), sometimes even showing soft tissue preservation (Piñeiro et al., 2012b; Piñeiro et al., 2012a; MacDougall et al., 2020). With such conditions of preservation, it seems very unlikely that no trace of other invertebrates would remain. It is possible, however, that soft animal fossils are yet to be discovered, and that further study of the mesosaur-bearing Irati-Whitehill deposits will reveal them.