A three-dimensional geometric morphometric analysis of the morphological transformation of Caiman lower jaw during post-hatching ontogeny

Introduction

Crocodylia is a clade represented by extinct and extant species that has evolved a huge spectrum of cranial morphological variation since its origin in the Late Cretaceous (e.g., Brochu, 1999, 2001, 2003; Jouve et al., 2008; Blanco et al., 2015; Bronzati, Montefeltro & Langer, 2015; Rio & Mannion, 2021). The group has a long and rich fossil record, spanning a wide range of skull shapes and body sizes (e.g., Gearty & Payne, 2020; Rio & Mannion, 2021; Stockdale & Benton, 2021; Stocker, Brochu & Kirk, 2021). Although extinct crocodylians vastly outnumber their modern relatives and the current morphological diversity is lower compared to that of the past (e.g., Brochu, 2001, 2003; Bronzati, Montefeltro & Langer, 2015; Mannion et al., 2015; Godoy, 2020; Rio & Mannion, 2021), extant species still exhibit a considerable variety of cranial shapes (Pierce, Angielczyk & Rayfield, 2008). In this sense, skull disparity has attracted the attention of researchers for many years, becoming the focus of several studies (e.g., Brochu, 1999, 2001; Busbey, 1995; McHenry et al., 2006; Pierce, Angielczyk & Rayfield, 2008, 2009a, 2009b; Piras et al., 2010, 2014; Pearcy & Wijtten, 2011; Fernandez Blanco, Cassini & Bona, 2014; Fernandez Blanco et al., 2015; Fernandez Blanco, Cassini & Bona, 2018; Escobedo-Galvan et al., 2015; Foth et al., 2017; Angulo-Bedoya, Correa & Benítez, 2019; Morris et al., 2019, 2021; Falcón Espitia & Jerez, 2021). It is not only the peculiar head configuration that makes the skull so broadly studied but also its greater presence in the fossil record and museum collections, compared to other components of the skeleton. Furthermore, the comparative study of the crocodylian skull is crucial since cranial features are essential for phylogenetic analyses, and therefore, for the evolutionary history of the group (e.g., Brochu, 1997, 1999, 2011; Hua & Jouve, 2004; Aguilera, Riff & Bocquentin-Villanueva, 2006; Bona, 2007; Jouve et al., 2008, 2015; Brochu & Storrs, 2012; Hastings et al., 2013; Hastings, Reisser & Scheyer, 2016; Scheyer et al., 2013; Fortier et al., 2014; Salas-Gismondi et al., 2015, 2016; Cidade et al., 2017; Massonne et al., 2019; Souza-Filho et al., 2019; Ristevski et al., 2021; Rio & Mannion, 2021).

Exploring the literature regarding crocodylian skull morphologies, several quantitative studies examine the patterns of cranial variation and their biomechanical, functional, and ecological implications (e.g., Iordansky, 1973; Busbey, 1989; Daniel & McHenry, 2001; McHenry et al., 2006; Pierce, Angielczyk & Rayfield, 2008, 2009a, 2009b; Sadleir & Makovicky, 2008; Piras et al., 2009, 2014; Erickson et al., 2012; Walmsley et al., 2013; Fernandez Blanco, Cassini & Bona, 2014; Fernandez Blanco et al., 2015; Fernandez Blanco, Cassini & Bona, 2018; Gignac & Erickson, 2016; McCurry et al., 2017; Ballell et al., 2019; Felice, Pol & Goswami, 2021). In all these articles, different ontogenetic stages after hatching of extinct and/or modern species are analyzed. However, it is worth pointing out that these researchers have paid attention only to the cranium and have neglected or only partially analyzed the lower jaw. This is striking since mandibles comprise meaningful elements to varying biological and paleobiological approaches (e.g., morphofunctional studies and feeding ecology; see Vizcaíno et al., 2016) as they may provide much anatomical information that can be extracted from bone correlates (such as crests and marks of muscle attachments, and nerves or blood vessels canals and foramina) (e.g., Iordansky, 1964, 2000, 2010; Schumacher, 1973, 1985; Bona & Desojo, 2011; Porro et al., 2011; Stubbs et al., 2013; Walmsley et al., 2013). Furthermore, mandibles are easy-to-handle elements (to measure and digitalize) in museum collections due to their sizes, and their study is not relatively hard since they are simpler structures than crania because they are made up of fewer elements. Despite this, there are few works dealing with crocodylian mandibles. Monteiro & Soares (1997) and Verdade (2000) analyzed the ontogenetic variation of the skull shape of three Caiman species (C. latirostris (Daudin, 1802), C. yacare (Daudin, 1802) and C. sclerops Schneider, 1801), and included few linear measurements of the lower jaw. Escobedo-Galvan et al. (2015) also used conventional metrics of the skull and mandibles, and evaluated the morphological position of C. crocodilus apaporiensis within the morphospace of C. crocodilus complex and C. yacare (ontogeny was not examined). Confirming and complementing the results of the latter, Angulo-Bedoya, Correa & Benítez (2019) applied a two-dimensional geometric morphometric approach over cranial and mandibular series of this C. crocodilus subspecies complex. On the other hand, Porro et al. (2011), Stubbs et al. (2013) and Walmsley et al. (2013) made the first comprehensive biomechanical analyses focused exclusively on different crocodylian mandibular metrics, and analyzed their relation to ecological diversity and feeding/mechanical behaviors. In sum, beyond these articles, there is no other study that merely emphasizes the quantitative analysis of the ontogenetic morphological transformation of the lower jaw in crocodylians or its morphological interspecific variation.

Caimaninae is an alligatoroid crocodylian clade currently restricted to South and Central America (Simpson, 1937; Gasparini, 1996; Brochu, 1999, 2010, 2011; Bona, 2007; Pinheiro et al., 2012; Salas-Gismondi et al., 2015). Its fossil record is usually fragmentary and species diagnosis are mostly built over characters extracted from cranial and mandibular materials which are generally broken, incomplete, and/or deformed (e.g., Brochu, 1999, 2011; Aguilera, Riff & Bocquentin-Villanueva, 2006; Bona, 2007; Hastings et al., 2013; Hastings, Reisser & Scheyer, 2016; Scheyer et al., 2013; Fortier et al., 2014; Salas-Gismondi et al., 2015, 2016; Hastings, Reisser & Scheyer, 2016; Cidade et al., 2017; Souza-Filho et al., 2019; Rio & Mannion, 2021). Furthermore, some extinct caimanine species were identified and described only based on mandibular fragments (e.g., Eocaiman palaeocenicus: Bona, 2007, E. itaboraiensis: Pinheiro et al., 2012, Caiman tremembensis: Chiappe, 1988). Since limits currently used to delineate fossil species are based on ranges of variation of modern specimens (Brochu & Sumrall, 2020), any study analyzing cranial and lower jaw morphological variation of extant species (such as morphological transformations during ontogeny) would help us more accurately identify characters states used in systematics. In this sense, Fernandez Blanco, Cassini & Bona (2014, 2018) made the first morphogeometric contribution and quantified the ontogenetic morphological variation in the skull of two extant Caiman species, identifying cranial features that vary along ontogeny and are relevant in systematics. Given that Caimaninae is a clade that holds extant relatives, it comprises an excellent opportunity to study the development of bone features and to extrapolate this information with fossil specimens in order to reinforce paleobiological and paleontological taxonomic hypotheses.

The aim of this study is to analyze the ontogenetic morphological transformations that undergo the mandibles of two extant caimanine species, C. latirostris and C. yacare, and their interspecific variation, applying a three-dimensional geometric morphometric approach. Special attention will be paid to: 1-common patterns of ontogenetic change in both Caiman species and; 2-morphological features that remain stable during the post-hatching ontogeny. Given the principal role of jaws in food intake, we will also assess the degree to which diet and morphological disparity covary. Feeding ecology across different age categories of these crocodylian species has been analyzed in some previous quantitative studies such as Borteiro et al. (2008), Santos et al. (1996) and Fernandez Blanco, Cassini & Bona (2018). Although C. latirostris is considered as a more durophage species than C. yacare, juveniles of both species feed mainly on invertebrates such as insects, snails, crustaceans and spiders, and adults feed mainly on vertebrates. However, even though the size of the prey consumed by the two Caiman species increases throughout development as well as their frequency of consumption, predation on small prey never ceases (Santos et al., 1996; Melo, 2002; Borteiro et al., 2008). Thus, finally, we discuss about the relationship between the morphological variation of these crocodylian species to their habitat and feeding ecology.

Materials and Methods

We examined two post-hatching ontogenetic series of mandibles of two modern species of Caiman that inhabit Argentina, C. latirostris and C. yacare. Samples belonged to wild animals coming mostly from Argentina (Chaco and Corrientes Provinces; Table 1) and sex, age, and sexual or somatic maturity are not determined. The material is housed at the Museo de La Plata (MLP) and Museo Argentino de Ciencias Naturales “Bernardino Rivadavia” (MACN).

A three-dimensional geometric morphometric approach was implemented to evaluate and visualize the morphological transformations that mandibles undergo from juvenile to adult individuals. For this purpose, jaws were assigned to three different age categories (juveniles, subadults and adults) following the same criteria used in Fernandez Blanco, Cassini & Bona (2018) (Table 1), which defined these classes by the snout-vent length (see Santos et al. (1996) and Borteiro et al. (2008)). The C. latirostirs sample consisted of three juveniles, six subadults and six adults, whereas the C. yacare sample consisted of seven juveniles, 18 subadults and 23 adults. A total of 77 landmarks (Type I, II, and semilandmarks; Fig. 1; Table 2) were digitized on 63 jaws (48 jaws of C. yacare and 15 jaws of C. latirostris) using a Microscribe G2L digitizer.

There are 20/48 specimens of C. yacare with missing landmarks. Fourteen of them have less than five missing landmarks and six have more than seven. The regions of the mandible more frequently broken include the bone contacts in the external and internal mandibular fenestrae, coronoid, medial glenoid cavity, and less frequently, the retroarticular process. In C. latirostris 9/15 specimens have missing landmarks. Five of them have less than six missing landmarks, and four have more than eight. The regions of the mandible more frequently broken in this species include the bone contacts in the internal mandibular fenestrae, coronoid and the retroarticular process. The missing landmarks in broken mandibles were estimated to maximize the number of specimens in the sample. For a detailed procedure, see Fernandez Blanco, Cassini & Bona (2018) and the Supplemental Material 1 therein. Semilandmarks were equispaced using the Botton-Divet et al. (2016) Supplemental 2 for curve re-sampling script, and then landmarks configurations were superimposed using a generalized procrustes analysis (GPA) to remove the effects of orientation, positioning, and scaling. A principal component analysis (PCA) was performed to explore morphological changes. These analyses were implemented in procSym function of Morpho v2.5.1 R package (Schlager, 2017). The method proposed by Bookstein (2014) implemented in the function getMeaningfulPCs from same package was applied to determine whether a PC is eligible to interpret. In addition, Pearson correlation tests were performed to assess the extent and significance of the association between centroid size and pc scores.

Ontogenetic allometry deals with covariation among metric characters and corresponds to shape changes during growth of an individual (Klingenberg, 1996). Allometry could be expressed at the interspecific level (e.g., functional changes from an evolutionary perspective) or at the intraspecific level (e.g., allometry of growth) (see Klingenberg & Zimmermann, 1992; Klingenberg, 1996; Cassini, Flores & Vizcaíno, 2015 and references there in). In the GPA, shape and size were decomposed where centroid size was stored and used as a proxy for size (Goodall, 1991; Dryden & Mardia, 1998). Size was removed during the GPA, but the allometric component was not (Mitteroecker et al., 2013). In the absence of allometry, centroid size does not correlate with shape (Bookstein, 1986; Kendall, 1986). Therefore those components of shape that increase in size without correlation with size are isometric. To explore how shape variation is associated with size, and shape describes the relative size increase or decrease of specific parts (see Segura et al., 2017), we performed a multivariate regression (lm base function and RegScore from Morpho; see Schlager, 2022) on the Procrustes coordinates against the log-transformed centroid size for each species, separately (Klingenberg, 2016). Then the predicted component was expressed as a percentage of the total variation (i.e., coefficient of determination, R²) to quantify the shape variation explained by size in the dataset (Drake & Klingenberg, 2008; Cassini, 2013; Segura, Prevosti & Cassini, 2013; Segura et al., 2017). We also used the two-block partial least squares analysis (PLS: pls2B function from Morpho) to explore the patterns of covariation between mandibular morphology and diet through ontogeny of both caiman species (Mitteroecker & Bookstein, 2008). All information regarding feeding ecology was obtained from literature and analyzed following the same criteria used in Fernandez Blanco, Cassini & Bona (2018). We defined Block-1 as mandibular shape (i.e., landmark configurations) and Block-2 as continuous diet characters (i.e., the logit transformed diet proportions matrix; see Olsen, 2017; Fernandez Blanco, Cassini & Bona, 2018; Cassini & Toledo, 2021). PCA, PLS and Regressions produce vectors in shape space in different directions which were compared (angular comparison; see Segura, Prevosti & Cassini, 2013; Cassini, Muñoz & Vizcaino, 2017; Segura et al., 2017). The angles between these vectors were computed and compared using the angleTest function of Morpho package (see Li, 2011). When these angles are close to zero, this means that both analyses are similar, and consequently, share a similar shape change (Drake & Klingenberg, 2008; Klingenberg & Marugán-Lobón, 2013). All analyses were performed using the R v4.2.2 environment (R Core Team, 2022). The visualization and graphics to see the colour pattern associated with shape changes were made following Muñoz et al. (2017) and using the Morpho R package 2.5.1 (Schlager, 2017). Four specimens (one juvenile and adult species) were chosen and used to design the template average mesh for shape change visualizations. Landmarks and semi-landmarks were manually added to these meshes in the Landmark Editor. Semilandmarks were equispaced as it was described above. In each analysis, the corresponding meshes were transformed to consensus and then to the target shape (tps3d function) to finally compute a colored mesh representing the distances (as percent of centroid size) between the vertices of target and consensus meshes. The color keys are interpreted as follows: gray means no relative changes; red means relative increase in size and blue relative decrease in size. Additionally, a slide of thin plate spline gridlines was computed using the deformGrid3d function.

Raw data, meshes and supporting information are supplied on CONICET institutional repository (https://ri.conicet.gov.ar/handle/11336/184901).

Results

The PCA resulted in ten principal components that account for the total cumulative variance. Accordingly to the getMeaningfulPCs function report, only the two first axes were meaningful and used to describe the total variation. The PC1 explains the 33.08% of the total variance whereas PC2 describes the 16.36%. Only PC2 correlates significantly and positively with log-transformed centroid size (r = 0.6546, p-value < 0.0001) and has a vector angle of 34.544° with the pooled within-group (species) regression of shape coordinates against log-transformed centroid size (p-value < 0.00001). That means that morphological changes along the PC2 are mostly associated with an age gradient, and partially represent the shared shape variation explained by size (see below).

Species were segregated along the PC1 with C. yacare located mainly on the positive values (from −0.01 to 0.03 approximately) and C. latirostris on the negative ones (from −0.10 to −0.01) (Fig. 2). Both species slightly overlie near the value of −0.01 of the PC1. Shape changes associated with negative values of PC1 shows tall and wide mandibles (essentially in medial section) with small internal and external mandibular fenestrae, a large Meckelian fossa (area between both fenestrae), a dorsally curved dorsal area of the surangular bone, a massive articular bone with a wide articular area and a postero-ventrally inclined retroarticular process, a high and straight symphyseal area, not procumbent teeth, and a curved postero-ventral area of the angular bone (Fig. 3A). Instead, positive values show low and narrow mandibles (essentially in medial section) with large internal and external mandibular fenestrae, a small Meckelian fossa, a straight dorsal area of the surangular bone, a slender articular bone with a narrow articular area and a horizontal retroarticular process, a low and curved symphyseal area, procumbent teeth, and a straight postero-ventral area of the angular bone (Fig. 3B).

Age categories of C. latirostris were segregated along the PC2 with juveniles and adults lying on the extreme negative and positive values, respectively (Fig. 2). The three age classes of C. yacare are located sequentially from negative to positive values, but there is a great overlapping among them. Therefore, shape changes associated to PC2 (Figs. 3C and 3D), among other components, seem related to the age gradient.

In the multivariate regressions, a slightly common pattern of ontogenetic morphological change, shared by the two species, can be detected (Figs. 4 and 5). The angular comparison between these two regression vectors (in their own shape space) was 65.368° (p-value < 0.0001; i.e., non-orthogonal). In addition, these two regression vectors are orthogonal to all the PCs except PC1 and PC2. The angular comparison between these vectors (i.e., regressions and PCs) has a higher vector angle for PC1 (i.e., >72°) than PC2, which was 34.01° for C. latirostris and 50.71° for C. yacare (all p-values < 0.0001) which means that, although morphological changes of PC2 could be interpreted as a shared pattern of morphological change along ontogeny (see above), they resemble more the ontogeny of C. latirostris than C. yacare (see below). Morphological transformations from juveniles to adults can be observed in the thin plate spline gridlines. Allometry accounts for 19.01% in C. yacare and 22.31% in C. latirostris proportions of the total shape variation. The juveniles of both species have low and narrow mandibles with a small internal mandibular fenestra, a wide Meckelian fossa, a ventrally curved dorsal area of the surangular bone, a small and less curved (less concave) retroarticular process, a low and straight symphyseal area (the anterior portion is closer to the sagittal plane), procumbent teeth, and a straight postero-ventral area of the angular bone (Figs. 4B and 5B). Morphological changes towards adults show high and wide mandibles with a large internal mandibular fenestra, an elongated Meckelian fossa, a dorsal curved dorsal area of the surangular bone, a large and more curved (more concave) retroarticular process, a high and curved symphyseal area (the anterior portion is further to the sagittal plane), not procumbent teeth, and a curved postero-ventral area of the angular bone (Figs. 4C and 5C). In adults of C. yacare the symphyseal area becomes longer, narrower and more pointed. Although both caiman species share this pattern of morphological change along ontogeny, C. latirostris has higher Procrustes distances and lower centroid size amplitude between all specimens in the sample (i.e., 1.097 and 767.33, respectively) than C. yacare (i.e., 0.905 and 837.01; see Supplementary Information, Figs. 1 and 2). Consequently, some features are more evident (more colored) from early ontogenetic stages in C. latirostris. In addition, juveniles of C. yacare has slender mandibles with more procumbent teeth, lower symphyseal areas, high articular bones, larger external mandibular fenestrae, and a slender and more elongated Meckelian fossa than C. latirostris.

The PLS analysis on both species (Figs. 6 and 7) show a significant relationship between shape and diet (Table 3). The PLS analysis on C. latirostris shows that the first pair of PLS explains about 98.41% of covariation. The block-2 PLS coefficients of the five diet categories segregate spiders, crustaceans and insects items to negative values (ca −0.766, −0.36 and −0.267 respectively), and vertebrates and snails items to positive values (~0.454 and 0.079 respectively) (Fig. 6A). The PLS1 scores show a high and significant correlation between blocks. While juveniles lie on the double negative quadrant (with highly negative values on block-2), sub-adults lie near the zero values (but slightly displaced on negative values of block-1) and adults are on the double positive quadrant.

The PLS analysis on C. yacare shows that the first pair of PLS explains about 94.85% of covariation. The block-2 PLS coefficients of the five diet categories segregate spiders, snails and insects items (ca −0.831, −0.242 and −0.235, respectively) to negative values, and vertebrate items to positive values (~0.442) (Fig. 7A). The PLS1 scores show a high and significant correlation between blocks. While juveniles lie on the double negative quadrant (with highly negatives values on block-2), sub-adults lie near the zero values (but slightly displaced on negatives values of block-1) and adults are located on the double positive quadrant (but slightly displaced on negatives values of block-1).

Morphological changes of the PLS1 vector of block-1 in both species are visualized as surface plus thin plate spline gridline deformations in Figs. 6B, 6C, 7B and 7C. Transformations match with those of the regression analysis (see above) and show an angle between regression and PLS shape change vectors of 17.902° in C. latirostris and 19.599° in C. yacare (p-value < 0.00001; i.e., non-orthogonal).

Discussion

Conclusion

This is the first three-dimensional geometric morphometric study that quantifies the morphological variation in the mandibles of caimans along the ontogeny. Each caiman species show a clearly distinguishable morphotype and both species exhibit a common pattern of ontogenetic change in the lower jaw. When comparing the range of morphological variation along ontogeny (as Procrustes distances amplitude) in both species, we observe that this spectrum is greater in C. latirostris, and some mandibular transformations are more conspicuous from the beginning of the post-hatching development of this species. As a consequence, grown specimens of C. yacare retain some juvenile mandibular features, resulting in a slender morphotype, depicting a probable case of heterochrony.

Common ontogenetic morphological changes in caiman mandibles may probably be related to the common shift in the diet that crocodylians undergo during growth, whereas interespecific differences could be related to item food hardness and the variation in the percentage of each item consumed by each species during different ontogenetic stages. Furthermore, we propose that part of the lower jaw morphological disparity may be related to the differential use of the habitat by both caiman species as well. However, all these hypotheses should be tested through a morpho-functional approach. Although there are several works on the biomechanical and hydrodynamic properties of the crocodylian skull, further work is needed on mandibles. In this sense, a study of the associated morphological changes of the cranium and mandible in crocodylians would be relevant to achieve a complete understanding of craniomandibular mechanics and how it operates throughout the complete ontogeny.

Supplemental Information