Evolution and Diversification of Delphinid Skull Shapes

Summary

The diversity of the dolphin family was established during a short window of time. We investigated delphinid skull shape evolution, mapping shapes on an up-to-date nuclear phylogeny. In this model, the common ancestor was similar to Lagenorhynchus albirostris. Initial diversification occurred in three directions: toward specialized raptorial feeders of small prey with longer, narrower beaks, e.g., Delphinus; toward wider skulls with downward-oriented rostra and reduced temporal fossae, exemplified by suction feeders, e.g., Globicephala; and toward shorter and wider skulls/rostra and enlarged temporal fossae, e.g., Orcinus. Skull shape diversity was established early, the greatest later developments being adaptation of Steno to raptorial feeding on large prey and the convergence of Pseudorca toward Orcinus, related to handling large prey. Delphinid skull shapes are related to feeding mode and prey size, whereas adaptation to habitat is not marked. Over a short period, delphinid skulls have evolved a diversity eclipsing other extant odontocete clades.

Graphical Abstract

Highlights

• •Dolphin skull shape is highly dependent on phylogeny
• •Feeding mode and prey size are primary drivers of the evolution of dolphin skull shapes
• •Adaptive radiation of skull shapes has been followed by evolutionary stability

Biological Sciences; Evolutionary Biology; Evolutionary Processes; Phylogenetics

Introduction

In toothed whales, skull shape variation has been related to feeding and prey preferences (McCurry et al., 2017a; Werth, 2006a) as well as habitat (Galatius et al., 2011; Monteiro-Filho et al., 2002). Shape variation of skulls thus reflects the basic ecology of a species, making it a useful tool for studying macroevolutionary patterns. The family of oceanic dolphins (Delphinidae) consists of approximately 37 extant species that occur over a range of aquatic habitats from rivers to the open ocean and in climates ranging from the Arctic to the Tropics (Committee on Taxonomy, 2019; Jefferson and LeDuc, 2018). Prey preferences also show large variation, including benthic and pelagic fish in a large range of sizes, a diversity of cephalopods, and other marine mammals such as seals, dolphins, porpoises, and baleen whales (Slater et al., 2010). Unsurprisingly, this large ecological variation is reflected in a large diversity of skull shapes.

The odontocete family Delphinidae is an example of an explosive radiation with a rich diversity of species being established during a relatively short window of time approximately 10–15 mya (McGowen, 2011; McGowen et al., 2009, 2019; Steeman et al., 2009). This has been explained by physical restructuring of the oceans and temperature fluctuations during the late Miocene and early Pliocene epochs (Steeman et al., 2009). The rapid speciation and rich diversity also reflects a diversity of ecological niches where the basal, shared traits of delphinids such as large relative brain size, sophisticated echolocation, and sociality are presumed to have provided a competitive advantage (LeDuc, 2002). Some of these same traits are shared by the delphinoid relatives of delphinids to some degree, although, for example, sociality is low in phocoenids.

Several strategies for prey capture have evolved among aquatic tetrapods; usually these are divided into three or four classes: raptorial feeding, grip and tear feeding, suction feeding, and filter feeding (Kienle and Berta, 2016; McCurry et al., 2017b; Werth, 2000a). Except filter feeding, all of these strategies are seen in extant delphinids. Raptorial feeding is probably the plesiomorphic condition in odontocetes, as it requires few, if any, modifications relative to the feeding strategies of terrestrial ancestors (Werth, 2000a). Adaptations such as homodonty, polydonty, and longirostry have served to enhance raptorial capabilities (Werth, 2000a). Raptorial feeding, also called “pierce feeding,” simply involves grasping prey items in the jaws following movements of the head, neck, and/or the whole body to swallow prey whole. Several odontocetes have long, specialized rostra with many pointed teeth for grasping prey, such as river dolphins and, to a lesser extent, many species of delphinids. Grip and tear feeding may be seen as a subset of raptorial feeding, where prey is held, torn, and ripped using large, interlocking teeth as seen in delphinids such as Orcinus and Pseudorca (McCurry et al., 2017b; Werth, 2000a). In suction feeding, prey is drawn into the mouth by a vacuum created by depressing or retracting the tongue (Heyning and Mead, 1996; Werth, 2000b). To increase effectiveness, suction feeding odontocetes tend to have smaller gapes; this may be accomplished by shorter rostra or increased tissue covering the lateral margins of the gape (Werth, 2006b). The dentition of suction feeders is generally reduced or absent (Werth, 2000b, 2006b). Among odontocetes, suction feeding may have evolved several times as a strategy for capturing cephalopods (MacLeod et al., 2006). Many delphinids may use suction and/or raptorial feeding, depending on the targeted prey and the circumstances (Werth, 2000a, 2006a). An alternative classification of aquatic mammalian feeding with emphasis on the process has been proposed (Hocking et al., 2017). In the current study, the former framework will be used, as we are directly investigating morphology.

In addition to prey preference and feeding strategy, habitat may also shape skull morphology. The most well-known examples are more ventrally inclined rostra and occipital condyles in bottom feeders (Galatius et al., 2011; Monteiro-Filho et al., 2002). Another aspect of habitat influencing skull shape may be the climate. Colder climates may favor stouter morphologies to facilitate a smaller surface-to-volume ratio. As an example, Lagenorhynchus cruciger, inhabiting Antarctic and subantarctic waters, has a more robust skull than the other members of the subfamily Lissodelphininae living in warmer waters (Galatius and Goodall, 2016).

In the current study, we investigate the diversity and radiation of delphinid skull shapes using the species richness of the North Atlantic, as these species cover all extremes of delphinid skull morphology. We use three-dimensional geometric morphometrics to describe skull shape across the 18 species occurring in the North Atlantic and map these shapes on a phylogenetic tree generated from nuclear genome data to reconstruct a model of the evolution of delphinid skull shapes and investigate the role of skull morphology with respect to habitat, niche partitioning, and feeding strategy. This study underlines the importance of investigating extant morphology and genome-based phylogenetic reconstruction in examining the origins and possible drivers of modern delphinid skull morphologies while supporting our model with evidence from the fossil record.

Results

Shape Is Dependent on Phylogeny

Skull shape was defined by a suite of 48 cranial landmarks (Figure S1; Table S2). A large proportion of the skull shape variation (62.6%) of the North Atlantic delphinids was represented by principal components 1 and 2 of the PCA at 45.6% and 17.0%, respectively. All subsequent components each accounted for less than 7% of the variance. PC1 describes a lengthening of the rostrum with a longer toothrow and a general lateral compression and dorsoventral expansion of the skull with increasing scores. PC2 describes a more ventral orientation of the rostrum and an anterior tilt of the foramen magnum with increasing scores. Furthermore, the temporal fossa is dramatically enlarged, whereas the orbit and its surrounding structures are displaced anteriorly and the braincase is expanded posteriorly.

Mapping of the phylogeny on the PCs (see Transparent Methods) shows two of the phylogenetically earliest-diverging species, Lagenorhynchus albirostris and Leucopleurus acutus, to maintain a shape proximate to the modeled ancestral shape at the root of the tree (Figure 1). Early branching shows the three subfamilies Orcininae, Globicephalinae, and Delphininae to evolve in different directions in terms of skull shape. The modeled evolution of Orcininae is toward a lower score along PC1 and a higher score along PC2. Globicephalinae have evolved toward lower scores along both PC1 and PC2, whereas Delphininae have evolved toward higher scores along PC1 and somewhat lower scores along PC2. The Delphininae species occupy the lower right part of the plot, with the exception of Tursiops truncatus, which retains a position close to the modeled ancestral shape of this subfamily. Among the Globicephalinae, Pseudorca crassidens is highly divergent from other members of the subfamily along PC2 and approaches convergence in skull shape with Orcinus. In our analysis, the affiliation of Steno with Globicephalinae entails the most dramatic adaptation of skull shape within Delphinidae, with a much higher PC1 score, and a higher PC2 score, than its closest relatives. A permutation test with 10,000 iterations of the null hypothesis of no phylogenetic signal in the skull shapes (Klingenberg and Gidaszewski, 2010) gave a significant result (P < 0.0001). Figure S2 shows the mean skull shape of each species compared with the grand mean shape of all species.

Figure 1.

Modeled Evolution of Delphinid Skull Shapes

PC1 (x axis) versus PC2 (y axis) scores of mean skull shapes of the 18 delphinid species of the North Atlantic and the reconstructed scores for nodes and root of the mitogenome phylogeny. Shape changes along the PCs are illustrated at the margins of the plot. Above and below the plot are skull shapes (dorsal and lateral views) representing the highest and lowest scores among the species along PC2, with PC1 scores kept neutral (at zero). Right and left are skull shapes (dorsal and lateral views) representing the highest and lowest scores among the species along PC1, with PC2 scores kept neutral (at zero). See also Figure S2 for species-specific shapes.

Feeding Mode and Prey Size Are Primary Drivers of Skull Shape Evolution

Feeding mode appears to be a strong driver of skull shape variation in delphinids (Figure 2). Along the first two PCs, the feeding modes were well separated: grip and tear feeders occupied the upper left quadrant, suction feeders occupied the lower left quadrant, and raptorial feeders mostly occupied the lower right quadrant and extended into the lower left with more robust species such as Lagenorhynchus albirostris, Peponocephala electra, and Feresa attenuata (Figure 2A). In terms of maximum prey size (see Transparent Methods), the species with “small” maximum prey size (<0.2% of body weight) were mostly isolated in the lower right quadrant. Species with “medium” maximum prey size (>0.2% < 1% of body weight) were distributed in a band of the morphospace running from the lower left quadrant to the upper right quadrant. Species with “large” maximum prey size (>1% of body weight) were positioned in the upper and lower left quadrants (Figure 2B). Phylogenetic generalized least squares regressions of predator-prey size ratios on PCs 1–3 for all species did not yield significant results. Following the example of McCurry et al. (2017b), who removed suction feeders from their analysis, we removed Globicephalinae from the analyses and achieved a significant association for PC2 (see Table 1 for results of PGLS analyses). In terms of climate, species occurring in the “warm temperate-tropical” zone were found throughout the morphospace. Species from the “arctic-cold temperate” zone were only found in the left half of the morphospace along PC1 (Figure 2C). In terms of habitat, species occurring in the “continental slope-oceanic” habitat were found in most of the morphospace. Species occurring in the “continental shelf” habitat were not found in the lower left quadrant. Species occurring in the “coastal” habitat were only found in a narrow band running from the upper left to the lower right quadrant (Figure 2D). To further analyze habitat use, we conducted PGLS regressions of dive depth on PCs 1–3. These did not yield significant results (Table 1). To investigate relationships of shape with size, we conducted PGLS regressions of centroid size on PCs 1–3. These analyses yielded a significant result for PC1, indicating that species relying on suction or grip and tear feeding tend to be larger than raptorial species, irrespective of phylogeny.

Figure 2.

Skull Shape Morphospace in Relation to Feeding Mode, Prey Size, Climate, and Habitat

Plots of average shapes of species along PCs 1 and 2 with polygons defining sub-morphospaces for categories based on feeding strategy (A), maximum prey size (B), occurrence in climate zones (C) and occurrence in habitat type (D).

Table 1.

Explorations of the Relationships of Prey Size and Dive Depth with Shape

	Coefficient	Standard Error	t Value	p Value
Predator-prey size ratio
PC1	−238.89	251.52	−0.95	0.36
PC2	551.98	347.12	1.59	0.13
PC3	−157.67	456.62	−0.35	0.73
Predator-prey size ratio, without Globicephalinae
PC1	−297.76	80.69	1.13	0.28
PC2	1,812.12	751.34	2.41	0.03
PC3	−206.38	534.81	−0.39	0.71
Dive depth
PC1	−1,306.97	902.51	−1.45	0.17
PC2	137.25	1,386.51	0.10	0.92
PC3	1,358.20	1,667.50	0.81	0.42
Centroid size
PC1	−1,782.17	734.22	−2.43	0.03
PC2	562.62	1,232.98	0.46	0.65
PC3	99.27	1,522.44	0.07	0.95

Coefficients, standard errors, and t and p values of phylogenetic generalized least squares regressions of predator-prey size ratios and dive depths on principal components 1–3. significant p values are given with bold numerals.

Discussion

Skull Shapes Radiated from a Central Position of the Morphospace

According to our reconstruction of the radiation of delphinid skull shapes, the ancestral shape of delphinid skulls was similar to those of the extant species Lagenorhynchus albirostris, Leucopleurus acutus, or Tursiops truncatus. This entailed a moderately long and robust rostrum, a temporal fossa of intermediate size, a toothrow running most of the length of the rostrum, and a general shape that is intermediate relative to the extremes shown by the current variation. The earliest known delphinid fossil is Eodelphinus kabatensis, dating to the late Miocene, 13–8.5 mya (Murakami et al., 2014a, 2014b). The width of its premaxillae at the base of the rostrum and relatively robust teeth point to a rostrum of intermediate length and width, similar to our estimated ancestor.

When compared with the range of skull shapes within extant and extinct odontocetes, the diversity within Delphinidae is rather low. The earliest known odontocetes (Oligocene, 33.9–24 mya) possessed the greatest disparity and diversity in facial region morphologies associated with telescoping of the skull, compared with any later lineage including modern clades (Churchill et al., 2018). Diversity in rostral length peaked in the early Miocene (~20–16 mya) (Boessenecker et al., 2017); however, bizarre forms that greatly deviated from a generalized odontocete rostrum existed into the Pliocene (5.3–2.6 mya), such as the walrus-like delphinoid Odobenocetops and the strange porpoise with an elongate mandible, Semirostrum ceruttii (Benites-Palomino et al., 2020; Boessenecker et al., 2017; Churchill et al., 2018; de Muizon, 1993; Lambert, 2005; McCurry and Pyenson, 2019; Racicot et al., 2014). Thus, the relatively low diversity of modern and fossil delphinid skull morphology may result from the relative age of the taxon; however, facial asymmetry, telescoping of the skull, and loss of tooth replacement, all of which emerged in the Oligocene, undoubtedly constrain facial morphologies (Boessenecker et al., 2017; Churchill et al., 2018, 2019).

Although "kentriodontids" (Odontoceti: Kentriodontidae), an extinct paraphyletic group of small odontocetes with relatively symmetrical skulls that were present from the late Oligocene to late Miocene (Ichishima et al., 1994), have been suggested as potential ancestors to modern delphinoids, only a clade of six kentriodontids have been reconstructed phylogenetically as sister to Albireonidae + Iniodea + Delphinoidea (Lambert et al., 2017; Racicot, 2018). The group thus requires further study in the context of relationships with Delphinoidea before we can make inferences regarding delphinid ancestry.

From the ancestral condition reconstructed in our analysis, our model shows a development in two opposite directions: (1) toward blunter and wider rostra with shorter toothrows, accompanied by a generally wider and dorsoventrally compressed skull, as seen in the subfamilies Orcininae and Globicephalinae, or (2) toward the extended, narrow rostrum with longer toothrow, and laterally compressed skull of the Delphininae and Steno. These two opposite trends are each further divided on both sides of the described spectrum into (1) forms with enlarged temporal fossae and an anteriorly displaced orbit (O. orca and P. crassidens) or a reduced postorbital process (S. bredanensis) and (2) an opposite trend toward diminished temporal fossae, exemplified by Globicephalinae other than P. crassidens and Delphininae other than T. truncatus.

Described delphinid fossils besides Eodelphinus further support our model of morphospace evolution from a generalized morphology similar to Lagenorhynchus albirostris and Tursiops. As expected from a rapid radiation, they are preserved from the Pliocene (5.3–2.6 mya) onward and include globicephalines such as Protoglobicephala mexicana (Aguirre-Fernández et al., 2009; Boessenecker et al., 2015), possible relatives of Orcinus such as Orcinus citonensis (Bianucci, 1996), and delphinines such as Septidelphis morii (Bianucci, 2013) and Hemisyntrachelus, which share skull shape similarities and possible transitional morphologies among their extant relatives. Protoglobicephala has a slightly longer rostrum than many of the extant globicephalines and was described as “intermediate” in morphology between Tursiops and extant globicephalines (Aguirre-Fernández et al., 2009), a transitional morphology that is supported by our model. Atadelphis gastaldi is described as sharing affinities with Steno bredanensis, while also possessing plesiomorphic characters similar to kentriodontids (Bianucci, 1996). Armidelphis sorbinii shares skull shape similarities with Peponocephala and Feresa, with antorbital processes similar to Orcinus orca, and probably had a strong bite for seizing large prey (Bianucci, 2005). The skull proportions of Stenella giulli are within the range of extant Stenella species, with a relative rostral length between that of S. coeruleoalba and S. longirostris, and with a relatively antero-posteriorly elongated neurocranium distinguishing it from the short and broad neurocranium of S. clymene (Bianucci, 1996). Septidelphis morii is reconstructed as sister to delphinines in a phylogenetic analysis and has an elongate, narrow rostrum similar to Stenella species and Astadelphis (Bianucci, 2013). Hemisyntrachelus is interpreted as having intermediate features between Tursiops and Pseudorca and Orcinus; for example, the skull size of Hemisyntrachelus cortesii is much larger (60 cm condylobasal length) than that of Tursiops, but the premaxillae do not narrow at the apical portion of the rostrum and fewer (14–15), larger teeth are present, which is more similar to Orcinus and Pseudorca (Bianucci, 1996). Similarly, Hemisyntrachelus oligodon has a lower tooth count (11–12) than is typical for extant Tursiops and is even larger than Hemisyntrachelus cortesii (62.5 cm condylobasal length). Dorsal views of the skull of Hemisyntrachelus oligodon show transversely narrow neurocranium and preorbital region compared with extant Tursiops (Pilleri and Silber, 1989). The vertex of the skull forms a more concave surface than that of Hemisyntrachelus cortesii, which has a shallow angle reaching the vertex (Pilleri and Silber, 1989). The rostra of both Hemisyntrachelus species are more elongate than in Orcinus and Pseudorca, reflecting their similarities with Tursiops, and they have deeper and narrower antorbital notches. Tursiops osennae possesses intermediate features between Hemisyntrachelus and extant Tursiops (Bianucci, 1996). Orcinus citonensis is distinguishable from extant Orcinus orca in its smaller size and larger number of smaller teeth (Bianucci, 1996), which seems intermediate between delphinines and globicephalines or Orcinus. All of these extinct delphinids appear to be transitional among extant species and are thus supported within the framework of our model of morphospace evolution. Some more bizarre extinct delphinids existed, however, including the toothless ziphiid-like Australodelphis mirus and a “hammerhead” globicephaline Platalearostrum hoekmani (Fordyce et al., 2002; Post and Kompanje, 2010). If the phylogenetic relationships of the extinct taxa were reconstructed, we might have a better understanding and ability to include them in similar morphometric analyses. One difficulty is the convergence in skull shapes between Orcinus and Pseudorca leading to possible confusion on the affinities of certain extinct taxa. Inclusion of the fossils with our extant dataset could not only inform on possible feeding mode and prey type, but also provide data on the transitional forms leading to the more extreme or convergent morphologies.

Feeding Mode Drives Skull Shape Evolution

Feeding mode is a major driver of delphinid skull radiation, as raptorial feeders, suction feeders, and grip and tear feeders each form discrete sub-spaces within the larger morphospace (Figure 2A). This is in agreement with the overall trends in skull diversity of odontocetes, which encompasses long narrow beaks in archaeocetes, river dolphins, and some delphinids and blunt beaks in some delphinids, phocoenids, and monodontids (Boessenecker et al., 2017; Norris and Møhl, 1983). Second, the teeth of ancient forms were typically numerous and prominent, whereas many modern genera show marked tooth reduction or outright loss. Norris and Møhl (1983) related the blunt-beaked forms with reduced dentition to suction feeding, whereas longer, narrower rostra and a long tooth row were related to feeding by grasping prey with the interlocking teeth. This axis of variation has been further investigated by Werth (2006a) and Werth (2006b), who found that wider and shorter beaks in delphinids provided superior suction. Thus, there is no sharp distinction between suction feeders and raptorial feeders, but our study underlines and quantifies this axis as the source of most of the skull shape variation among Delphinidae (Figure 2A).

Werth (2006a) further found that dentition was reduced as rostra became wider and shorter. Our study reveals two exceptions to this rule: O. orca and P. crassidens. Although their teeth are numerically reduced, they are very large and the relative extent of the toothrow is not shorter than in species with long, narrow rostra (Figure S1). As mentioned, these two species diverge on the second axis of delphinid skull shape variation, characterized by a large temporal fossa. Among the species included here, only O. orca, P. crassidens, S. bredanensis, and T. truncatus have evolved in this direction from the estimated shape root of the delphinid tree. All other species have evolved in the opposite direction, toward a smaller temporal fossa and a more posteriorly placed orbit.

As previously alluded to, delphinid feeding strategies and their associated morphological adaptations likely form a continuum, so delineations between feeding strategies are somewhat arbitrary. It could be argued that all delphinid species rely on some combination of raptorial and suction feeding, seeing that even long-beaked forms have suction capability, albeit at a much lower level than short-beaked forms (Johnston and Berta, 2011; Werth, 2006a).

Species with low PC1 scores, i.e., suction and/or grip and tear feeders were significantly larger than species tending to the raptorial side of the shape spectrum. Grip and tear feeders may need to be large to cope with attractive prey. The delphinid suction feeders included here tend to dive more deeply and for longer durations, for which larger size is an advantage (Noren and Williams, 2000). On the other hand, we did not see a significant relationship between dive depth and PC1. It should be noted that a relationship between suction feeding and size is not a general trend across Odontoceti, as phocoenids seem to largely rely on suction feeding.

Prey Size Drives Skull Shape Evolution

The first axis of variation is associated with suction feeding versus raptorial feeding, whereas the second axis seems to be associated with prey size. The two grip and tear feeders, O. orca and P. crassidens, display the two highest scores along this axis and there is an association between maximum prey size and this axis, particularly for species on the raptorial side of the suction-raptorial spectrum (Figure 3B). The development of larger temporalis muscle mass indicated by the larger temporal fossa associated with this axis is a logical adaptation for larger prey items, and the association of a larger temporal fossa with larger prey has been proposed previously (Perrin, 1975). Other authors have suggested that the size of the temporal fossa (in Delphinus delphis) is largely constrained by asymmetry of the facial region of the skull, an ancient feature of odontocetes (Churchill et al., 2019). The ability of species with shorter and wider rostra to handle larger prey than those with longer and narrower rostra is unsurprising; McCurry et al. (2017c) have demonstrated the intuitive consequence of less mechanical strain with shorter and wider rostra. Thus, the evolution of grip and tear feeding in O. orca and P. crassidens may have progressed from the ancestral delphinid skull shape via an initial adaptation toward suction feeding. This is almost certainly the case for P. crassidens, phylogenetically nested deep within Globicephalinae, a subfamily otherwise consisting of members with extreme to moderate morphological adaptations for suction feeding. O. orca is also capable of suction when feeding on, e.g., schooling fish such as herring and mackerel (Werth, 2000a). Interestingly, for F. attenuata, the closest morphospace neighbor of O. orca and P. crassidens among the Globicephalinae, there are reports of killing and eating of D. delphis and Stenella spp. in relation to tuna fisheries in the eastern tropical Pacific (Perryman and Foster, 1980), providing further evidence that the short, wide rostrum related to suction feeding among delphinids may be a preadaptation for handling large prey items. Compared with Globicephala, however, F. attenuata is more adapted to large prey in terms of skull shape.

Similar to our results, McCurry et al. (2017b) also found an association between prey size and skull shape in odontocetes. However, McCurry et al. (2017b) only included landmarks on the mandible and the anterior skull and did not detect a clear pattern in the analysis of prey size in relation to skull shape among odontocetes and had to exclude suction feeders from the analysis to obtain a signal. Similarly, we only obtained significant results when we excluded Globicephalinae from the analysis. This subfamily contained some outliers relative to the other species; for example, G. griseus and P. electra had much larger maximum prey sizes than expected from the analyses. Thus, the blunter rostra of these species may allow for larger prey sizes despite a reduced temporal fossa. The analysis of McCurry et al. (2017b) only included aspects of shape related to the rostrum and mandible, thus leaving out information found to be relevant to prey size in our analyses, namely, the size of the temporal fossa and the associated displacement of the orbit. Of the species with high PC2 scores, S. bredanensis is noteworthy as being the only species in this study to be a raptorial feeder on larger prey, as S. bredanensis is known to take large prey items, such as Mahi-Mahi (Coryphaena hippurus) (Pitman and Stinchcomb, 2002). The most specialized suction feeders, G. griseus, G. melas, and G. macrorhynchus, all have small temporal fossae, indicating a weaker temporalis muscle, an adaptation opposed to that of the dedicated grip and tear feeders, O. orca and P. crassidens. Given the reduced temporal fossa and teeth and the limited maximum gape (Werth, 2000b), morphological evidence indicates specialization toward small to moderate prey sizes for both Globicephala species. G. griseus is morphologically very similar to Globicephala, but much smaller, and has been recorded to take relatively large Octopus prey up to about 7 kg (Cockcroft et al., 1993).

Habitat and Climate Are Less Important Drivers of Evolution

Delphinids are found in a wide range of habitats, ranging from rivers to the open ocean, and in another odontocete family (Phocoenidae), the primary axis of shape variation among extant species was related to habitat (Galatius et al., 2011). In the current study, however, no clear links of morphology to habitat were detected. The current study focuses on the macroevolutionary trends of delphinid skull shape evolution, and it is very probable that adaptations for specific habitats do occur in this family. Such adaptations have been reported for the genus Sotalia (Monteiro-Filho et al., 2002) and in the subfamily Lissodelphininae (Galatius and Goodall, 2016). In the current dataset, detection of this is hampered by the fact that the four species included in the coastal category, O. orca, D. delphis, S. frontalis, and T. truncatus all are found in shelf habitats as well, and most also in oceanic habitats.

Species spanning the entire morphospace occur in tropical and warm temperate waters, but in the temperate and arctic zones, there is a conspicuous lack of longirostrine forms. These colder environments also have small pelagic prey suitable for longirostrine forms, evidenced by large colonies of sea birds foraging in pelagic waters of these climate zones. Thus, thermoregulation could explain the lack of longirostrine delphinids in colder waters. Alternatively, McCurry et al. (2017a) suggested that potentially faster swimming speeds in fish (von Herbing, 2002) could have initiated predator-prey escalation, driving evolution of longer rostra in warmer climates.

Adaptive Radiation Followed by Evolutionary Stability

Our model indicates a phylogenetically stable morphospace distribution with most taxa remaining within a narrow range of shapes after the initial adaptive radiation. The primary exception to this rule is S. bredanensis, which is phylogenetically affiliated with the globicephalines but has made the most dramatic adaptation from an ancestral shape recorded in our analysis to become a raptorial feeder of large prey. However, the phylogenetic relationship of the genus Steno is unclear with some analyses based on mitogenomes or a small set of nuclear genes indicate affiliation with Delphininae (Steeman et al., 2009; Galatius et al., 2019), whereas several other analyses based on nuclear DNA—including the most comprehensive genomic analysis conducted to date—indicate affiliation with Globicephalinae (McGowen, 2011; McGowen et al., 2009, 2019). Thus, future work is needed to address the evolution of S. bredanensis. A more minor exception is P. crassidens, which is phylogenetically nested deep within the Globicephalinae but shows convergent evolution with O. orca, having a similar skull shape to this species. Another species that has adapted its skull morphology after the initial radiation is L. hosei with adaptation to a greater reliance on suction than other members of the subfamily Delphininae. This subfamily is otherwise specialized in raptorial feeding on small prey items, with the exception of T. truncatus, which has retained a morphology closer to the ancestral shape of the subfamily, apparently with greater reliance on suction and adaptation for larger prey items. The early-diverging Lagenorhynchus and Leucopleurus lineages have most likely occupied morphospace spheres close to the ancestral shape throughout their evolutionary history.

As the primary axes of variation relate to foraging strategy and prey size, it is most likely that adaptation in these regards has been the major evolutionary driver of diversification of delphinid species and skull shapes. This is not a new idea but something that has been suggested for Odontoceti in general (Boessenecker et al., 2017; McCurry et al., 2017a; Norris and Møhl, 1983; Werth, 2006a). However, it is interesting in light of the fact that other extant odontocete clades have much less variation now than some of their extinct relatives. This is obvious for the modern oligotypic clades Pontoporiidae, Physeteridae, Kogiidae, Lipotidae, Platanistidae, Iniidae, and Monodontidae but is also true for the more speciose extant Phocoenidae and Ziphiidae. Extinct members of all modern odontocete clades were more morphologically diverse throughout the Miocene and Pliocene than they are in the present day, with the exception of the Delphinidae, which (in terms of species richness) currently rival that of the entire remaining Odontoceti. In this context, an interesting question is: did diversity within non-delphinid odontocete clades decline as a result of global environmental change toward the end of the Pliocene? In this scenario, evolutionary radiation in the Delphinidae may have been in response to the availability of resources and the opening of free ecospace. Alternatively, did the explosive radiation of delphinids in part drive this diversity decline via resource competition and evolutionary replacement? Future analyses integrating geochronological, paleoenvironmental, and palaeontological (including paleoecological and phylogenetic) data, combined with an analysis of past and current morphospace, could be used to test these hypotheses.

Conclusions

Delphinids are the result of a rapid radiation that has resulted in the most species-rich and morphologically diverse extant cetacean family in terms of skull shape. This diversity is primarily related to feeding strategy, where delphinids are unique among cetaceans in encompassing specialized raptorial, suction, and grip and tear feeders. Secondary features of the skull are associated with feeding on prey items of different size. Climate may have also shaped the phylogeography of delphinids as none of the species specialized for raptorial feeding are found in colder climates. This remarkable radiation of skull shapes seems to have occurred in a simple radiation, mostly driven by feeding strategy and prey size, with few major events after initial diversification.

Limitations of the Study

The results of this study come with a few caveats. By focusing on the delphinid species inhabiting the North Atlantic, we do not comprehensively cover this family's ~37 species. The bulk of the species not covered are represented by the subfamily Lissodelphininae, which occurs in the Southern Hemisphere and the North Pacific. Galatius and Goodall (2016) found a primary axis of variation within Lissodelphininae very similar to the one found here for the Delphinidae of the North Atlantic. All 10 lissodelphinine species were positioned between L. albirostris and D. delphis (included for comparative reasons) along this axis, indicating that we are not missing major variations of shape by the exclusion of this taxon. The same can be said for the missing species of the mostly tropical genera Sousa and Sotalia, which, from a visual inspection, have similar skull shapes to S. bredanensis, including large temporal fossae. Another species not included in the current study, Orcaella brevirostris, is indicated to be sister to the Globicephalinae subfamily (McGowen et al., 2019), in line with its gross skull morphology. Another part of delphinid diversity that we did not include was fossils. These were omitted because of the lack of firm knowledge regarding phylogenetic relationships and the rather few skulls that are sufficiently complete to record most landmarks. Given rigorous phylogenetic analysis of delphinid fossils, they can be included in a future study.

Resource Availability

Lead Contact

Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Anders Galatius (agj@bios.au.dk).

Materials Availability

No materials were newly generated for this paper.

Data and Code Availability

The specimens included in the study are listed in Table S1. Variable values of size, feeding mode, max weight, max prey weight, climate, habitat, and dive depth are in Table S3. Morphometric data used for this paper have been deposited to Mendeley Data: https://doi.org/10.17632/x4kfyfzyc6.1.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: October 23, 2020