Calvarial suture interdigitation in hadrosaurids (Ornithischia: Ornithopoda): Perspectives through ontogeny and evolution

Lambeosaurine hadrosaurids exhibited extreme modifications to the skull, where the premaxillae, nasals, and prefrontals were modified to form their iconic supracranial crests. This morphology contrasts with their sister group, Hadrosaurinae, which possessed the plesiomorphic arrangement of bones. Although studies have discussed differences between lambeosaurine and hadrosaurine skull morphology and ontogeny, there is little information detailing suture modifications through ontogeny and evolution. Suture morphology is of particular interest due to its correlation with the mechanical loading of the skull in extant vertebrates. We quantify and contrast the morphology of calvarial sutures in iguanodontians and ontogenetic series of Corythosaurus and Gryposaurus to test whether the evolution of lambeosaurine crests impacted the mechanical loading of the skull. We found that suture interdigitation (SI) increases through ontogeny in hadrosaurids, although this increase is more extreme in Corythosaurus than Gryposaurus, and overall suture complexity (i.e., overall shape) remained constant. Lambeosaurines also have higher SI than other iguanodontians, even in crestless juveniles, suggesting that increased sinuosity is unrelated to the structural support of the crest. Hadrosaurines and basal iguanodontians did not differ. Similarly, lambeosaurines have more complexly shaped sutures than hadrosaurines and basal iguanodontians, while the latter two groups do not differ. Taken together, these results suggest that lambeosaurine calvarial sutures are more interdigitated than other iguanodontians, and although suture sinuosity increased through ontogeny, the suture shape remained constant. These ontogenetic and evolutionary patterns suggest that increased suture complexity in lambeosaurines coincided with crest evolution, and corresponding modifications to their facial skeleton altered the distribution of stress while feeding.


| INTRODUCTION
Suture morphology has become increasingly relevant in the last few decades for studying cranial biomechanics, with numerous studies focusing on suture morphology to infer mechanical loading (e.g., Gruntmejer et al., 2019;Kammerer, 2021;Markey & Marshall, 2007;Porro et al., 2015). Cranial sutures are unossified syndesmotic joints that occur between the growth fronts of adjacent bones of the skull. Sutures have two main functions: (1) they are the primary regions of bone growth in the skull, and (2) they absorb mechanical stress by deforming, allowing the skull to respond to biomechanical loads. Suture growth and morphology are regulated by a complex set of biochemical and epigenetic factors (Herring, 1993;Jaslow, 1990). Due to their key role in cranial growth, suture patency in the calvarium (cranial vault) appears to be necessary in the early development of most vertebrates (Baer, 1954;Enlow, 1989;Jaslow, 1990;White et al., 2021). Sutures may close (i.e., fuse) once the period of rapid growth has ceased, but other factors such as biomechanical loading may result in the maintenance of patent (open) cranial sutures throughout the life span of an organism. This is because cranial suture morphology and patency are also influenced by mechanical stresses (Byron et al., 2018;Herring, 1993), and the biomechanical environment is an epigenetic factor thought to be the primary determinate of cranial suture morphology (Byron et al., 2018;Jaslow, 1990;Monteiro & Lessa, 2000;Moss, 1954Moss, , 1957Moss, , 1961Rafferty & Herring, 1999). Sutures may remain simple, increase in their sinuosity, or close through ossification across the synarthoses through ontogeny depending on changes in the type and magnitude of load applied to the suture (Jaslow, 1990).
Cranial sutures have been repeatedly identified to exhibit distinct morphologies that are each best adapted to withstand specific types of mechanical loads (Herring & Mucci, 1991;Herring, 1972Herring, , 2008: lapping (scarf) joints are able to withstand highly variable loading regimes, including torsional forces; tongue-and-groove sutures are best able to withstand tension; similarly, butting sutures are able to handle tension; and finally, interdigitated sutures are best able to withstand compressive and tensional forces. These differences in suture mechanics are achieved due to the differing arrangement of the sutural ligament fibers, which absorb stress under tension and shear stress (Herring, 1972). For example, the interfingering of bones in an interdigitated suture allows the sutural ligament to be loaded with tension and shear forces along the interdigitations when the suture is under compression (Herring, 2008), but the increase in surface area works equally well when the suture is under tension (Herring & Teng, 2000). The calvarium of vertebrates is therefore often composed of interdigitated sutures to better withstand the complex compressive and tensional forces exerted by the masticatory muscles attaching to the temporal region (e.g., Anderson & Bolt, 2013;Herring & Mucci, 1991;Herring & Teng, 2000;Herring, 1972;Kathe, 1999;Markey & Marshall, 2007). The degree of interdigitation in the calvarium has been directly correlated with the magnitude of mechanical stress exhibited by bite force (e.g., Byron et al., 2018), and calvarial suture interdigitation is therefore a helpful tool to evaluate mechanical loading in the skull. At present, the majority of detailed work on suture morphology and related mechanics has been carried out on mammals (e.g., Byron et al., 2018;Herring & Mucci, 1991;Herring & Teng, 2000;Herring, 1972Herring, , 1993Jaslow, 1989Jaslow, , 1990Savoldi et al., 2019), though there have been some detailed studies on reptiles (Monteiro & Lessa, 2000), and fish (Markey & Marshall, 2007).
Given the strong link between form and function in cranial sutures, many studies have used suture morphology to predict loading patterns in fossil taxa (e.g., Gruntmejer et al., 2019;Kammerer, 2021;Kathe, 1999), with subsequent finite element analyses demonstrating accuracy in predicting overall patterns of stress distribution (e.g., Fortuny et al., 2015;Maloul et al., 2014;Rayfield, 2005). However, few studies have used suture shape as a proxy for mechanical loading of the skull in dinosaurs (e.g., Rayfield, 2004Rayfield, , 2005Weishampel, 1984), despite the diversity of cranial anatomy in the group (e.g., Brusatte et al., 2011;Felice et al., 2020;Foth et al., 2016). Potentially the most extreme modifications to the arrangement of the facial skeleton are observed in lambeosaurine hadrosaurs, where the premaxilla, nasals, and prefrontals were modified to form and support their supracranial crests. This morphology in lambeosaurines contrasts starkly with their sister group, Hadrosaurinae (=Saurolophinae), which possesses the plesiomorphic elongate skull condition and arrangement of bones. Although many studies have discussed anatomical differences between lambeosaurines and hadrosaurines, and changes through ontogeny (e.g., Brink et al., 2011;Evans, 2010;Evans et al., 2005;Prieto-Marquez, 2010;Weishampel, 1984), there is an absence of information detailing suture modification due to crest development. Discussing changes in suture morphology is significant since previous studies have suggested that lambeosaurines and hadrosaurines may have been specialized for processing different types of plants based on skull morphology, body proportions, and tooth wear (Carrano et al., 1999;Chapman & Brett-Surman, 1990;Dodson, 1975;Erickson et al., 2012;Mallon, 2019;Mallon & Anderson, 2013;Nabavizadeh, 2016). Additionally, patterns of ontogenetic suture development are rarely documented in extinct vertebrates, and as such the paleobiological implications of their evolution are rarely explored. Although supracranial crests are absent in juvenile lambeosaurines and rapidly develop into maturity, juvenile lambeosaurines still exhibit rearrangement of the premaxilla, nasal, and prefrontal, suggesting differences in stress distribution may exist at birth. Weishampel (1984) provided a detailed breakdown of cranial suture morphologies within Ornithopoda, but often treated Hadrosauridae with a single description, either not reporting on differences within the clade or attributing differences in suture morphology to support the weight of the lambeosaurine supracranial crest. However, changes in the distribution of mechanical stress throughout the skull while chewing is an alternate hypothesis that may pose a larger effect due to the extreme modifications to the bones of the snout in lambeosaurines. Here, we contrast the morphology of calvarial sutures between the lambeosaurine Corythosaurus and the contemporaneous hadrosaurine Gryposaurus notabilis through well-preserved ontogenetic series to test the following: (1) lambeosaurine calvarial suture interdigitation differed from other iguanodontians to support the weight of the supracranial crest; (2) lambeosaurine calvarial suture interdigitation differed from other iguanodontians due to differences in the distribution of stress through the skull during feeding. We then expand this comparison to include other lambeosaurines, hadrosaurines, and basal iguanodontians to provide a phylogenetic context for ontogenetic suture interdigitation patterns. Calvarial sutures were selected due to their close association with the masticatory muscles, and the well-documented relationship between calvarial suture complexity and mechanical stress.

| Materials
Images were taken of relevant specimens in dorsal view to capture the size and shape of the interfrontal and frontoparietal sutures, with a scale bar for reference (see Supporting Information: Table S1 for a list of specimens). Specimens that could not be evaluated in person were supplemented with dorsal braincase images from the literature (see Supporting Information: Table S1 for a list of specimens and their source). Although a twodimensional view of the suture surface does not capture complete information on the three-dimensional suture face, isolated calvarial bones in iguanodontians show that they are fairly consistent in suture interdigitation along these sutural contacts, and studying the suture surface is, therefore, a good approximation of suture shape. Gryposaurus specimens were identified as species following Lowi-Merri and Evans (2020) and Mallon et al. (2022). Isolated lambeosaurine braincases from Dinosaur Park were tentatively identified by us as Lambeosaurus if the frontal-nasal suture formed a raised ridge, and as Corythosaurus if this ridge was absent.
The following clade definitions are used when discussing phylogenetic positions of taxa in this study: Iguanodontia, all ornithopods more closely related to Parasaurolophus walkeri than to Hypsilophodon foxii or Thescelosaurus neglectus (Sereno, 2005); Hadrosauridae, the most recent common ancestor of Saurolophus osborni and P. walkeri and all descendants (Sereno, 1998); Hadrosaurinae, all hadrosaurids more closely related to S. osborni than to P. walker (Sereno, 1998); Lambeosaurinae, all hadrosaurids more closely related to P. walkeri than to S. osborni (Sereno, 1998). We follow Xing et al. (2022) in using the term Hadrosaurinae over Saurolophinae given that these clades are functionally the same except for the inclusion of Hadrosaurus in Hadrosaurinae.
Specimens were categorized into one of three ontogenetic age bins: juvenile, subadult, and adult. These bins were based on skull length, following Evans (2010): juvenile, less than 50% maximum skull length; subadult, 50%-85% maximum skull length; adult, above 85% maximum skull length. When complete skulls were not preserved, specimen size was estimated based on the size of preserved bones relative to other specimens. If only a small number of skulls were known for a taxon, making relative skull length inaccurate, ontogenetic age was assigned using the literature. It should be noted that this method of binning artificially partitions continuous growth and variation into discrete bins, but it is a useful metric for discussing ontogenetic skull growth in hadrosaurids because they represent the three distinct stages of crest development: juvenile, crest-less; subadult, rapid crest growth; adult, fully derived crests with little growth.

| Suture sinuosity
To quantify suture interdigitation, we employed the suture sinuosity index (SI), which is a measure of the sinuous length of a suture following the curves relative to its absolute length in a straight line from start to end (Supporting Information: Figure S1A,B; White et al., 2020). The closer the SI is to 1, the closer the suture is to a straight line, whereas higher SI values indicate increased sinuosity. Images were loaded into ImageJ v1.53k (Schneider et al., 2012) and size calibrated using the "set scale" function. Suture lengths, defined as the sinuous length of the suture from end to end, were then measured for the interfrontal and the left and right frontoparietal sutures by tracing the sutures manually using the "lines" function set to "freehand." Suture distances, defined as a straight line from one end of the suture to the other, were then measured using the "lines" function set to "straight." The start of the interfrontal suture was defined as the posterior-most junction of the two frontal bones, and the end of the interfrontal suture was defined as the anterior-most ectocranial junction of the two frontals. For the frontoparietal suture, the suture started at the midline of the interfrontal process (the same start-point for the interfrontal suture) and ended at the lateral-most contact between the frontal and parietal. The suture SI was then calculated as the suture length divided by the suture distance, producing a unitless value to describe the sinuosity of the suture. Left and right frontoparietal suture SI values were averaged for subsequent analyses.
We subjected the interfrontal and average frontoparietal SI for Corythosaurus and G. notabilis to an analysis of covariance (ANCOVA) to evaluate whether there were significant differences in sinuosity between these taxa when ontogeny is considered as a covariate. This was done in RStudio v1.4.1106 using the "aov" and "Anova" functions from the package stats v4.0.4 (R Core Team, 2021), with "type" set to III. A second ANCOVA was then run on the full data set, comparing lambeosaurines to hadrosaurines and basal iguanodontians, with ontogeny as a covariate.
To visualize these data, we plotted the logtransformed suture lengths against their corresponding suture distances to visualize potential allometric differences between Corythosaurus and Gryposaurus. We then constructed two sets of boxplots using the "boxplot" function from the package graphics v4.0.4 (R Core Team, 2021). The first set was the SI for Corythosaurus and Gryposaurus at each ontogenetic stage. The second set was the SI for all specimens of lambeosaurines, hadrosaurines, and basal iguanodontians, binned according to their genus and ontogenetic stage (see Supporting Information: Table S1 for specimens and identifications).

| Suture complexity
To evaluate suture complexity (i.e., the shape of the suture), we implemented a windowed short-time Fourier transformation (STFT) with a power spectrum density (PSD) estimate. Fourier analyses compare contours by generating coefficients defining the sine/cosine waves that together form the original curve (Allen, 2006). PSD estimates use the STFT coefficients to describe how the variance (i.e., power) is distributed within frequency intervals of the curve (Stoica & Moses, 1997), and is therefore used to indicate patterns among waves and curves (Allen, 2006). STFT with PSD has been used to describe suture complexity in invertebrates (Supporting Information: Figure S1C,D; Allen, 2006), but recent comparative studies have demonstrated their utility and accuracy in vertebrate suture complexity (White et al., 2020). PSD is useful in addition to SI because it quantifies the differences in suture shape that SI does not detect. Although White et al. (2020) used the total PSD to compare suture shape, we have chosen to bin the harmonics according to Allen (2006) to better illustrate potential differences between taxa.
Suture curves were digitized in R (R Core Team, 2021) using the "digitizeImages" function in the Stereo-Morph package (Olsen & Westneat, 2014). A single curve was drawn for each suture in the digitizeImages window, starting at the junction between the left and right frontoparietal sutures and the interfrontal suture, and ending at the lateral-/anterior-most extent of the sutures. Sutures were scaled using the scaling feature in the digitizeImages window. Suture curves were then resampled using the "readland.Shapes" function in the geomorph package Baken et al., 2021) to produce 500 evenly spaced landmarks along each suture. The landmarks for each suture were separately scaled and aligned using a general Procrustes alignment (GPA) through the geomorph function "gpagen," with the start and end landmarks set as fixed landmarks, and the remaining 498 landmarks set as sliding semilandmarks that were allowed to slide along their tangent plane.
Aligned and scaled coordinates were subjected to STFT using the function "stft" from the e1071 package (Meyer et al., 2021). The PSD was then calculated for each frequency as the sum of the squared STFT coefficients. PSD values were then summed in bins to describe the variance of very-low-(VLF), low-(LF), medium-(MF), and high-frequency (HF) waves in the sutures. These bins were defined following Allen (2006): VLF, 1-4 (broad and shallow lobes); LF, 5-12 (narrow lobes); MF, 12-32 (highly multilobate sutures); HF, 32< (serrations and other HF subdivisions). Although previous studies have focused on the total PSD value for a suture (i.e., PSD summed across all frequencies, without binning), we instead chose to bin the PSD frequencies to better illustrate differences in extremes (Allen, 2006). Lastly, we determined the proportion of each bin by dividing the bin power value by the total power for the F I G U R E 1 (See caption on next page) DUDGEON and EVANS | 213 suture and multiplying by 100, allowing suture complexity to be compared between individuals.
We subjected the interfrontal and average frontoparietal PSD percentages of each suture for Corythosaurus and G. notabilis to an ANCOVA to evaluate whether there were significant differences in suture complexity between these taxa when ontogeny is considered as a covariate. This was done in RStudio v1.4.1106 using the "aov" and "Anova" functions from the package stats v4.0.4 (R Core Team, 2021), with "type" set to III. A second ANCOVA was then run on the full data set, comparing lambeosaurines to hadrosaurines and basal iguanodontians, with ontogeny as a covariate.
Finally, we constructed two sets of bar graphs using the "barplot" function from the package graphics v4.0.4 (R Core Team, 2021) to visualize suture complexity. The first set was the PSD bin percentages for Corythosaurus and Gryposaurus at each ontogenetic stage. The second set was the PSD bin percentages for all specimens of lambeosaurines, hadrosaurines, and basal iguanodontians, binned according to their genus and ontogenetic stage.

| Suture sinuosity
Suture sinuosity clearly increases through ontogeny in both Corythosaurus and Gryposaurus (Figures 1, 2, and 3), supporting previous observations by Weishampel (1984   (1.20-1.83) and and 1.8 for the frontoparietal (1.52-2.06) sutures. Mature Corythosaurus had the highest SI among these two genera, averaging approximately 1.7 for the interfrontal suture (1.40-2.60) and 1.9 for the frontoparietal suture (1.53-2.71). The interfrontal suture is much shorter on average in all ontogenetic stages of Corythosaurus than Gryposaurus (Figure 2), likely due to the shorter frontal platform in lambeosaurines. The frontoparietal suture, however, is similar in size between Corythosaurus and Gryposaurus, but with higher interdigitation in the former taxon (Figures 2 and 3). Overall, SI tends to be similar between the frontoparietal and interfrontal sutures of Gryposaurus, but the frontoparietal sutures tend to be more sinuous than the interfrontal suture in Corythosaurus. This is an interesting observation, given that most Gryposaurus specimens observed here, in addition to specimens described by Weishampel (1984), tend to have interfrontal sinuosity F I G U R E 3 Sinuosity index for (a) the interfrontal suture, and (b) the average frontoparietal suture for juvenile, subadult, and adult Corythosaurus and Gryposaurus. Thick bars represent the mean, boxes represent the upper and lower quartiles, and whiskers represent the maximum and minimum values. concentrated in the posterior half of the suture (i.e., the posterior half of the interfrontal suture is more interdigitated than the anterior half). These observations suggest that increased sinuosity is common in the posterior calvarium of hadrosaurines and may reflect increased loading of forces in the posterior portion of the braincase, similar to the observations of Byron et al. (2018) in mice.
The ANCOVA found significant differences in both interfrontal and frontoparietal sinuosity between Corythosaurus and Gryposaurus, but did not find significant differences between ontogenetic ages (Table 1). Visually assessing sinuosity in the ontogenetic stages reveals that Corythosaurus is more sinuous than Gryposaurus at each stage, including the juvenile stage, and the only groups with similar sinuosity are juvenile Corythosaurus and mature Gryposaurus.
The trends observed between Corythosaurus and Gryposaurus are also true more broadly in Iguanodontia, where lambeosaurines consistently have more sinuous calvarial sutures than hadrosaurines (Figure 4). ANCO-VA found significant differences in both interfrontal and frontoparietal SI (Table 2), and a post hoc Tukey's test found that this difference is significant between Lambeosaurinae and Hadrosaurinae, and Lambeosaurinae and basal Iguanodontia (Supporting Information: Table S2). No significant differences were found between Hadrosaurinae and basal Iguanodontia. Suture sinuosity also showed no significant differences between ontogenetic groups. Interestingly, the most extreme SI observed here was for the frontoparietal suture of the juvenile Parasaurolophus sp. (CMN 8502;Evans et al., 2007), a notable observation given that juveniles usually have lower sinuosity than mature individuals.

| Suture complexity
Differences in suture complexity as indicated by PSD bins are much more subtle, and ontogenetic trends are not clear for either genus ( Figure 5). Interestingly, interfrontal PSD values differ little between Corythosaurus and Gryposaurus, with significant differences present only in LF and HF features (Table 3). Frontoparietal PSD values are overall quite different between these two genera, where Corythosaurus has significantly more MF and HF features, and Gryposaurus has significantly more VLF features. There are no significant differences between these genera in LF features. No significant differences were found between ontogenetic age groups in either suture.
ANCOVA of PSD bins between Lambeosaurinae, Hadrosaurinae, and basal Iguanodontia reveals similar trends across this larger data set ( Figure 6). Significant differences in the interfrontal suture occur in VLF, MF, and HF features, where lambeosaurines have significantly less VLF features than basal iguanodontians, more MF features than basal iguanodontians, and more HF features than hadrosaurines and basal iguanodontians (Table 4). We also found significant differences in the LF features across ontogenetic ages, though post hoc Tukey's test found no significant differences between groups (subadult-adult approach significance). Significant differences also occur in the frontoparietal sutures in VLF, MF, and HF features. Post hoc Tukey's tests indicate that all significant differences between phylogenetic groups are between Lambeosaurinae and Hadrosaurinae, and Lambeosaurinae and basal Iguanodontia, where Lambeosaurinae has more MF and HF features, and less VLF features, than Hadrosaurinae and basal Iguanodontia (Supporting Information: Table S3). The ANCOVA also recovered significant differences in VLF, MF, and HF features between ontogenetic stages in the frontoparietal suture, and post hoc Tukey's test reveals that these differences are only significant between the juvenile and adult groups (Supporting Information: Table S3).

| DISCUSSION
We found that the SI of Corythosaurus was significantly higher than Gryposaurus, a trend that was also true for lambeosaurines and hadrosaurines more broadly, supporting the hypothesis that suture sinuosity differs between lambeosaurines and hadrosaurines. Both taxa demonstrate an increase in sinuosity through ontogeny, an expected phenomenon given that bite force and jaw lever advantage increase with body size in hadrosaurids (Mallon & Anderson, 2015), along with suture sinuosity in ornithopods more generally (Weishampel, 1984). Sinuosity appears to be an ontogenetically rapid phenomenon in Corythosaurus, where sutures are simple or slightly sinuous in juvenile individuals, and dramatically increase in sinuosity leading into maturity, particularly for the interfrontal suture. Interestingly, skull size continues to increase after sinuosity has increased in subadults, and the degree of sinuosity does not change dramatically over the large size range from subadult to adult, the time when crests are rapidly growing. Although ontogenetic sinuosity of the calvarial bones does occur in Gryposaurus, it is not comparable to the change in depth of sinuosity observed in Corythosaurus (Figure 3). The interfrontal suture, which is highly sinuous in subadult and larger lambeosaurines, is virtually straight in the dorsal view in most hadrosaurine and basal iguanodontian taxa. PSD values also differ significantly between Corythosaurus and Gryposaurus calvarial sutures, though the interfrontal suture only showed differences in HF features. The frontoparietal suture showed notable differences between groups, suggesting a more dramatic difference in suture complexity between these taxa than the interfrontal. Comparisons between taxonomic groups within Iguanodontia show a similar trend, where the complexity of the interfrontal suture differs little between lambeosaurines, hadrosaurines, and basal iguanodontians (though lambeosaurines have more HF features than the latter two groups), and the frontoparietal suture T A B L E 2 ANCOVA of calvarial suture sinuosity index between Lambeosaurinae, Hadrosaurinae, and basal iguanodontia.

Suture
Variable F I G U R E 5 Proportion of total power for very low-frequency (dark blue), low-frequency (light blue), middle-frequency (yellow), and high-frequency (red) features for (a) the interfrontal suture, and (b) the average frontoparietal suture for the three ontogenetic ages of Corythosaurus casuarius and Gryposaurus sp.
is more complex in lambeosaurines than hadrosaurines and basal iguanodontians. Taken together, the SI and PSD results suggest that although the interfrontal suture was more sinuous in lambeosaurines than in other iguanodontians, the complexity or overall shape of the suture differed little. This is contrasted with the frontoparietal suture, which was both more sinuous and more complex in lambeosaurines, meaning lambeosaurines have more lobes and serrations (i.e., HF features) in the suture than other iguanodontians. The combination of differences in sinuosity and complexity through phylogeny and ontogeny in Iguanodontia suggests that the increased sinuosity observed in lambeosaurines is associated with supracranial crest modifications to the skull. This suggests that the rearrangement of the cranial bones associated with the development of these crests forced significant changes in the physical loading of the skull, resulting in changes in suture morphology (Figure 7).
The similar pattern of suture sinuosity and complexity in the calvaria of lambeosaurines and hadrosaurines (i.e., increased sinuosity and complexity posteriorly) suggests that the skull roof of these groups likely experienced a similar pattern of loading when chewing, but the forces were more intense in lambeosaurines than hadrosaurines. It is well known that lambeosaurines have smaller adult skull sizes than hadrosaurines (Mallon & Anderson, 2013), suggesting that lambeosaurines would have lower bite forces given that jaw lever advantage is isometric in hadrosaurids (Mallon & Anderson, 2015). It is therefore unlikely that the increased sinuosity observed in lambeosaurines is due to high bite forces, and instead suggests that mechanical stress was more concentrated in the calvaria of lambeosaurines compared to hadrosaurines and basal iguanodontians. Additionally, there is little evidence for broad dietary differences between lambeosaurines and hadrosaurines; lambeosaurines likely preferred low browse, and hadrosaurines likely preferred high browse, with significant niche overlap between these groups (Carrano et al., 1999;Mallon, 2017). It is, therefore, unlikely that the large differences in the calvarial suture pattern observed here are due to dietary differences. Weishampel (1984) suggested that differences in suture morphology between lambeosaurines and hadrosaurines may be due to the imparted strain on the calvarium from the large supracranial crests, but associated changes in the mechanical performance of the skull while chewing may have a greater influence due to the extreme modifications to the bones of the snout in lambeosaurines. The results presented here support the hypothesis that suture interdigitation is associated with feeding mechanics and not crest support, given that the juvenile lambeosaurines exhibit increased interdigitation above that of other iguanodontians before the crest has begun to develop. Although it is possible that the greater calvarial suture interdigitation in juvenile lambeosaurines is a genetically predisposed condition caused by the development of crests later in life, this is unlikely given the strong dependence of suture shape on mechanical loading of the skull through ontogeny in extant taxa (e.g., Byron et al., 2018;Herring, 1993). Interdigitation due to the structural support of the lambeosaurine supracranial crest also appears unlikely because these crests were formed by thin sheets of bone from the premaxillae and nasals that primarily housed air-filled expansions of the nasal passages and were therefore light. Additionally, the crested hadrosaurine Saurolophus does not exhibit increased interdigitation compared to other hadrosaurines and basal iguanodontians. This provides further evidence that increased suture interdigitation is not related to crest support, but instead the rearrangement of the facial skeleton in lambeosaurines to form the crest. Jaslow's (1989Jaslow's ( , 1990) studies of sexual differences in suture morphology in wild sheep (Ovis orientalis) found that increased suture interdigitation in head-butting adult males was best correlated with horn size, providing an analogue for other supracranial structures.
Importantly, juveniles of both sexes, and mature female sheep, did not differ significantly in the level of suture interdigitation (Jaslow 1989), despite many female sheep possessing horns. These horns are largest in adult males, but most importantly, only males engage in head-to-head butting behavior. The increased interdigitation of cranial sutures in males is, therefore, an indication of the compensatory development of cranial sutures to cope with the added forces of head butting (Jaslow, 1989) and is not due to the presence and support of horns. This further suggests that increased interdigitation in lambeosaurines is not due to the support of supracranial structures, and instead suggests that they may be due to other mechanics affecting the skull. Lambeosaurine crests were obviously far too delicate to engage in physical confrontations like head butting (Hopson, 1975), indicating that increased stress during feeding is the likely mechanism for increased interdigitation in lambeosaurines.
The rearrangement of the facial skeleton in lambeosaurines also caused changes in the arrangement of cranial sutures, reflecting the distribution of mechanical stress. The expansion of the premaxilla onto and over the skull roof in lambeosaurines resulted in the snout being nearly devoid of sutures, save the premaxilla-maxilla suture, serving a stark contrast to the snout of hadrosaurines that possess two premaxilla-nasal sutures and the premaxilla-maxilla suture. The absence of sutures in the snout of lambeosaurines suggests that mechanical stress is not dissipated (i.e., absorbed) efficiently in the snout, and may instead be transferred posterodorsally through the solid premaxilla. The transfer of stress posterodorsally could, therefore, result in greater loading of more posterior sutures, and may be an explanation for increased calvarial sutures in lambeosaurines. If true, this may also explain why the premaxilla forms an ossified septum between the anterior nasal passages in lambeosaurines while the septum remains unossified in hadrosaurines. Additionally, lambeosaurines possess a tongue-and-groove lacrimal-maxilla suture, and an interdigitated lacrimalprefrontal suture, both of which are more complex than the simple scarf joints in hadrosaurines (Weishampel, 1984). Lambeosaurines are further unique in having a contact between the prefrontal and postorbital, which is also interdigitated. This combination of complex sutures F I G U R E 7 Suture interdigitation (SI) for the interfrontal (a) and frontoparietal (b) sutures mapped on a phylogeny of included taxa. posterior to the snout is consistent with the hypothesis that lambeosaurines experienced increased loading in this region. Greater posterior transferal of mechanical stress within the skull of lambeosaurines than hadrosaurines and basal iguanodontians would likely require the sutures in the posterior region of the skull to be better adapted to absorb this stress. These complex joints in lambeosaurines, combined with increased interdigitation of the calvarial sutures in these animals, suggest that lambeosaurines experienced significantly different loading regimes than other iguanodontians. Our work presented here provides new evidence that feeding mechanics differ between lambeosaurines and hadrosaurines due to the development of the prominent lambeosaurine supracranial crest. At present, feeding has only been completely digitally modeled in hadrosaurines (Bell et al., 2009;Rybczynski et al., 2008). In the future, we will test the stress distribution pattern outlined here across Hadrosauridae using finite element analysis.
AUTHOR CONTRIBUTIONS Thomas W. Dudgeon and David C. Evans devised the project and determined methodology. Thomas W. Dudgeon collected the data, conducted the analyses, constructed the tables and figures, and wrote the manuscript. David C. Evans supervised the project and edited the manuscript.