The Origin of Filter Feeding in Whales

Open ArchivePublished:June 29, 2017DOI:https://doi.org/10.1016/j.cub.2017.06.003

Highlights

•
A new species of 30 million year old whale has been found near Charleston, South Carolina
•
This new species is a relative of modern baleen-bearing whales but retains teeth
•
Its molars are large, multi-cusped, and overlapping and were used for filter feeding
•
Filter feeding evolved before baleen; early whales had teeth and baleen

Summary

As the largest known vertebrates of all time, mysticetes depend on keratinous sieves called baleen to capture enough small prey to sustain their enormous size [

]. The origins of baleen are controversial: one hypothesis suggests that teeth were lost during a suction-feeding stage of mysticete evolution and that baleen evolved thereafter [

], whereas another suggests that baleen evolved before teeth were lost [

]. Here we report a new species of toothed mysticete, Coronodon havensteini, from the Oligocene of South Carolina that is transitional between raptorial archaeocete whales and modern mysticetes. Although the morphology and wear on its anterior teeth indicate that it captured large prey, its broad, imbricated, multi-cusped lower molars frame narrow slots that were likely used for filter feeding. Coronodon havensteini is a basal, if not the most basal, mysticete, and our analysis suggests that it is representative of an initial stage of mysticete evolution in which teeth were functional analogs to baleen. In later lineages, the diastema between teeth increased—in some cases, markedly so [

]—and may mark a stage at which the balance of the oral fissure shifted from mostly teeth to mostly baleen. When placed in a phylogenetic context, our new taxon indicates that filter feeding was preceded by raptorial feeding and that suction feeding evolved separately within a clade removed from modern baleen whales.

Keywords

Results

Systematics

Order Cetacea; Suborder Mysticeti; Coronodon havensteini gen. et sp. nov.

Holotype

CCNHM 108. Nearly complete, 1.0-m-long skull, mandibles, 14 vertebrae, and partial ribs (Figures 1, 2, and 3; Figures S1–S3; Tables S1 and S2).

Figure thumbnail gr1 — Figure 1Cranium and Upper Dentition of *Coronodon havensteini* sp. et gen. nov.

Show full caption

View Large Image

Download Hi-res image

Download (PPT)

Figure thumbnail gr2 — Figure 2Filter Feeding in *Coronodon havensteini* and Associated Morphology

Show full caption

View Large Image

Download Hi-res image

Download (PPT)

Figure thumbnail gr3 — Figure 3Cranium of *Coronodon havensteini* in Anterior View

Show full caption

View Large Image

Download Hi-res image

Download (PPT)

Etymology

Coronodon havensteini. Genus is Greek for “crown tooth,” referring to the multi-cusped molars. The species name recognizes Mark Havenstein, who discovered the holotype.

Locality and Age

Wando River near Highway 41 Bridge, South Carolina, Berkeley County. Ashley Formation, Oligocene, uppermost Rupelian [

]. Additional locality information available upon request.

Diagnosis

Coronodon has the following mysticete synapomorphies: supraoccipital level with temporal fossa (character 25: state 1), broad basioccipital crests (39: 2), all cusps of posterior teeth subequal (99: 1), upturned antorbital process of maxilla (100: 1), and splayed basal cusps on posterior teeth (206: 1) (Figures 1, 2, S1, and S2; Data S1). Like some archaeocetes [

], its rostrum is twisted counterclockwise in anterior view (Figure 3). Coronodon havensteini is unique in having anterior lower molars labially overlapping posterior lower molars (Figure 2).

Feeding Behavior

Toothed mysticetes evolved from archaeocetes, a paraphyletic group ancestral to all extant cetaceans. Archaeocetes are interpreted as raptorial feeders: they hunted and caught prey with their teeth, one at a time. This inference is supported by fossilized stomach contents [

] and bite marks on small archaeocetes [

]. Raptorial feeding is also indicated in Coronodon by the caniniform incisors and the truncated and worn crown of the right P2 (Figure S1). Similar wear has been interpreted as being created by abrasion during feeding by the skeletons or other hard parts of prey [

By contrast, other features suggest that Coronodon was less effective at raptorial feeding than archaeocetes. The latter resemble raptorial odontocetes in having a long, narrow rostrum, which likely allowed prey to be caught with only a turn of the head and minimal drag [

]. The rostrum of Coronodon is wider, as indicated by the straight sides and shorter mandibular symphysis (Figures 1 and 2). In archaeocetes, the symphysis extends to p3, whereas the symphysis of Coronodon terminates anterior to the canine. Importantly, a wider rostrum in extant mysticetes is associated with a larger oral cavity, which is a critical adaptation for filter feeding [

]. Extant mysticetes also adjust the size of their oral cavity when feeding [

], and some have suggested that loose rostral sutures may facilitate this [

]. Somewhat surprisingly, Coronodon has simple and open rostral sutures too, unlike the sutures of many other toothed mysticetes [

The premolars and molars of Coronodon differ from those of Basilosauridae (Figures 2A and 2G), the closest relatives of mysticetes among archaeocetes. The sides of the p4 in the latter are steeper: lines connecting the apices of three central cusps form an angle of 82° or 98° in Cynthiacetus (MNHN.F.PRU 102) and Dorudon (UM 10122), respectively, as compared to 155° in Coronodon. A smaller angle is more effective at puncturing prey because it concentrates and sustains the bite force on the central cusp. Even greater differences are seen in the molars. In Coronodon, the first two molars are subequal to the p4 and resemble it in having mesial and distal accessory cusps. By contrast, the molars of basilosaurids are much smaller (e.g., p4/m1 = 1.59 for Dorudon), and the lower molars lack mesial cusps [

]. Large molars are often indicative of greater mastication, but the pattern of wear makes this interpretation unlikely. Each lower molar has a labial wear facet that extends apically onto the base of the crown but remains far removed from the carinae (Figure 2C). As a result, the scissor-like shearing between upper and lower molars, as seen in protocetids like Georgiacetus, is absent. Although the posterior molars could have been used to impale prey, this behavior seems uncommon given the small size of most apical wear facets (1.6–4 mm; Figures 2B and 2E) and the fact that the molars had reduced support from alveolar bone (Figure 2A). For each double-rooted tooth, only the distal half of each root is surrounded by alveolar bone, suggesting that the high occlusal pressures associated with macrophagy were rarely encountered.

Early studies speculated that toothed mysticetes used their teeth to filter feed [

], an idea later described as the “dental filtration hypothesis.” However, the teeth of previously described toothed mysticetes are too few, too small, too simplified, or too worn to be effective in filtering [

]. This led to a spate of recent studies that have developed a new hypothesis: that filter-feeding mysticetes evolved from edentulous, suction-feeding whales that lacked baleen [

]. The molars of Coronodon are far larger than those of other toothed mysticetes and hearken back to the dental filtration hypothesis. Unlike archaeocetes and most neocetes, its upper teeth widely overlap its lower teeth instead of interdigitating with them. As a result, when the mouth is opened, the posterior teeth enclose diamond-shaped gaps (∼15 × 35 mm at m1 and m2) that could filter out prey of varying size. When the mouth is closed, the gaps in Coronodon would have been closed off by the crown of the opposing dentition. Even so, narrow slots (0.5–3 mm wide) between the imbricated lower molars remained open (Figure 2B), allowing even smaller prey to be filtered. The serrate borders of these slots are formed by small accessory cusps that point distally from the tooth preceding the slot and mesially from the tooth following the slot. In archaeocetes, the basal cusps are directed more apically, instead of mesially or distally (Figure 2G). Many of the basal cusps in Coronodon have minor, but distinct, apical wear, indicating that they were exposed and not covered in gingiva (Table S2). The wear on the fairly sheltered, mesially directed cusps is unexpected and may have been formed by prey that accumulated along the slots during filtering. Such passive wear is quite common in marine mammals, with substrate being the best-documented culprit, particularly in porpoises [

]. Modern beaked whales use suction feeding to capture prey [

], which impact their tusks upon entering the oral fissure. This can result in strong wear on the mesial side of the tooth, specifically that portion exposed during typical gape employed during feeding (Figure 2H). Interestingly, the wear of baleen in extant mysticetes seems driven by the intraoral flow of water and the prey and sediment carried with it [

]. Apical wear also occurs in some aetiocetids [

], but we have come to a different interpretation because, in Coronodon, that wear also occurs on cusps sheltered by the preceding tooth. One unnamed aetiocetid (NMV P252567) has mesodistal grooves and large patches of wear on the lingual sides of its crowns [

]. We agree with the interpretation that this wear reflects suction feeding from the benthos [

], as it mirrors the only extant mammal, Odobenus, that does this as a primary feeding mode [

]. Importantly, no comparable wear or grooves exist in the only known specimen of Coronodon.

The presence of interdental, filter-feeding slots, as we propose, is consistent with the pattern of dental erosion on the crowns of the posterior lower teeth. The p3-m2 (right) and p3-m1 (left) teeth have ovoid pockets of dental erosion that emanate from the deep notches between the apical-most three cusps (Figures 2C–2E), similar to caries that form in the carnassial notch of dogs [

]. The remaining three to four notches between accessory cusps lack dental erosion, even though they are much further from the apex of the tooth. This pattern is expected if, as we suggest, the filter-feeding slots flushed this area with seawater, small prey, and/or other particulate matter that could inhibit plaque formation or abrade it off the tooth. The closest extant analog for the feeding behavior we reconstruct for Coronodon is the leopard seal (Hydrurga leptonyx), which uses its anterior teeth to secure prey and its postcanine dentition to filter out smaller prey, mostly krill [

]. The crabeater seal (Lobodon carcinophagus) also uses its posterior dentition for filter feeding [

], and these two pinnipeds are distinguished from other phocids in having a longer rostrum [

], larger teeth, small diastema, and subequal molars and premolars [

]. These traits characterize Coronodon as well, but filter-feeding seals primarily use gaps between cusps of the same tooth to filter out crustaceans [

], whereas we interpret that Coronodon utilized gaps between teeth.

Evolution of Baleen and Filter Feeding

There is little evidence for baleen in Coronodon. The only osteological correlates for baleen are laterally positioned palatal foramina, which, in the gray whale and presumably all other extant mysticetes, convey branches of the superior alveolar artery to supply the baleen-bearing, oral epithelium [

]. In Coronodon, there are only three to four minute, palatal foramina, most of which are clustered around the P3. This contrasts with the six widely distributed foramina and sulci in Aetiocetus weltoni, which indicates the presence of baleen in this taxon [

]. Although Coronodon probably lacked baleen, its morphology is suggestive of the following hypothesis. The first mysticetes, and their descendants like Coronodon, filter fed by funneling water through interdental slots whose dorsal margins were rimmed by thickened gingiva. If proto-baleen evolved from the gingiva at the lateral ends of these slots, then proto-baleen would not have been disturbed by the lower dentition (contra [

]) and selection could have favored smaller teeth and larger, baleen-filled diastema. In fact, there is evidence of thick gingiva in Coronodon. Broad zones of dental erosion, which typically form in gingival pockets, occur on the labial side of P4 and M2 (Figures 1D and 1F) and suggest a maximal gingival thickness of 5 cm (distance from alveolus to apical edge of erosion).

We tested this scenario by tracing the evolution of tooth size, morphology, and spacing on the most parsimonious tree for a modified mysticete supermatrix [

] (Figures 4 and S4; Table S3). Coronodon and the unnamed taxon ChM PV5720 from the Charleston area are the most basal mysticetes, followed by Metasqualodon symmetricus from Japan. This topology broadly supports our hypothesis that Coronodon is representative of a pre-baleen dental stage of filter feeding. Molars having mesially oriented, basal denticles, which encroach into the interdental slots, are optimized as evolving at the base of Mysticeti and then persisting until molars are lost in mysticetes (Figure 4). Apical orientation of basal cusps in the clade including aetiocetids and mammalodontids is interpreted as a reversal of the archaeocete condition. The evolution of molar diastema is complicated, but the base of Mysticeti is characterized by two or more successive widenings of diastema that culminate in the exceptionally broad diastema of Llanocetus denticrenatus, followed by the loss of posterior teeth in eomysticetids (Figure 4) [

]. Like a previous study [

], we view the palatal foramina in A. weltoni as indicative of baleen. Thus, under our topology, either mammalodontids lost baleen [

] or else baleen persisted in this family despite the absence of its osteological correlate.

Figure thumbnail gr4 — Figure 4Phylogenetic Position of *Coronodon havensteini* and Evolution of Key Features Associated with Filter Feeding

Show full caption

View Large Image

Download Hi-res image

Download (PPT)

Discussion

In contrast to other studies [

], we infer that filter feeding evolved shortly after odontocetes and mysticetes diverged. Across extant vertebrates, filter feeding is associated with a large body size [

]; thus, one way to test our hypothesis is to reconstruct body size evolution in mysticetes. There has been substantial work in this area [

], but several questions remain. A recent study inferred that the most recent common ancestor of all extant cetaceans was 167 kg [

], whereas another suggested that this taxon was about 2.5 m in length [

]. Using equations that relate body mass to length [

], a cetacean this long should be about 175 kg. Coronodon is much larger: an equation that estimates body length from width across the zygomatic process results in a length of 4.9 m [

]. This length corresponds to a mass of 1,150 kg, very similar to the mass estimated for the archaeocete Dorudon atrox [

]. Determining whether Coronodon simply retained the body size of archaeocetes or represents a dramatic increase over a small ancestral neocete will be difficult. Llanocetus was undoubtedly very large, whereas Metasqualodon was much smaller; however, the skull of neither taxon is well known or fully described.

Although our findings support the dental filtration hypothesis, are they also at odds with the suction-feeding hypothesis? The two are not mutually exclusive, because filter feeding can coexist with suction feeding, as demonstrated by leopard and crabeater seals [

]. Among extant odontocetes, the mandibular bluntness index, or MBI (i.e., ratio of the posterior width to the oblique length of the mandible) [

], is significantly correlated with suction feeding [

]. The MBI for Coronodon is 0.41, in line with extant raptorial feeders; however, there are exceptions. Many odontocetes with long, narrow rostra occasionally suction feed, and others, like ziphiids, rely almost exclusively on this behavior [

]. In fact, the only study to test associations of suction-feeding traits in a phylogenetic context inferred that basal neocetes used a combination of teeth and suction for capturing and ingesting prey [

]. Given the importance of suction feeding in discussions on the origin of baleen [

], it is critical to develop more methods to distinguish degrees of suction feeding in fossil taxa. Otherwise, the suction-feeding hypothesis for the origin of baleen will remain a speculative scenario, instead of a hypothesis corroborated by testing.

In reconstructing the behavior of Coronodon, we take a more conservative view by suggesting that it primarily employed ram feeding, whereby an aquatic predator opens its mouth and then thrusts its body onto prey. Ram feeding is one of the simplest forms of aquatic predation [

] and is commonly used among odontocetes, and specialized forms are used by all but one species of extant mysticetes [

]. If our interpretation is correct, then suction feeding in aetiocetids and mammalodontids is not representative of an early stage through which the ancestor of all extant mysticetes passed, but instead evolved after their ancestor diverged from other mysticetes. Ram feeding is associated with positive allometry of the skull, mandibles, and buccal cavity in rorquals, and similar allometric relationships apply to other extant mysticetes as well [

]. These differences are likely the result of positive feedback in filter feeders: larger mouths can capture more prey, and more prey can sustain a larger body size. Thus, we predict that future studies on skeletal proportions of the most basal mysticetes will find that they had proportionally larger mandibles and rostra than archaeocetes. This prediction, as well as others we have made involving body size, tooth size, and dental morphology, provide clear direction for further testing of the dental filtration hypothesis. In the meantime, we encourage those with differing views on the origin of baleen to do the same.

STAR★Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited Data
Morphological partition of dataset for phylogenetic analysis	This paper; http://morphobank.org	P2442
Molecular partition of dataset for phylogenetic analysis	[ 5 ]	N/A
Supermatrix of morphological and molecular data used for phylogenetic analysis	This paper; http://morphobank.org	P2442
Trees found from all phylogenetic analyses	This paper; http://morphobank.org	P2442
Diastema size as discrete character with tree	This paper; http://morphobank.org	P2442
Diastema size as continuous character with tree	This paper; http://morphobank.org	P2442
Software and Algorithms
TNT, tree analysis using new technology	[ 40 ] http://www.lillo.org.ar/phylogeny/	N/A
Mesquite 3.2	[41] http://mesquiteproject.org	N/A
Amira 5.4.3		N/A

Open table in a new tab

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Jonathan Geisler (jgeisler@nyit.edu).

Method Details

Internal Anatomy

The internal cranial morphology of Coronodon was studied with CT scans and 3D visualizations of that data. CT scans were acquired using a Siemens SOMATOM sensation 64 at the Medical University of South Carolina, with voxels 0.9765 × 0.9765 × 0.6 mm in size. The rostrum, mandibles, and braincase were scanned separately and then articulated in virtual space using the program Amira 5.4.3.

Phylogenetic Analyses

We based our phylogenetic analysis on a modified version of a published supermatrix [

]. To that matrix we added 109 characters from other studies, 17 taxa, and made several changes to characters and codings (Methods S1). Codings for Llanocetus denticrenatus were taken from published matrices [

] as well as our own observations of the described mandibular fragment. The supermatrix was analyzed using unweighted parsimony, with implied weighting (k = 2 −10), and with molecular data excluded (k = 3) using the application TNT [

]. A “New Technology” search was conducted using default values, except searches were terminated after the best tree was found 1000 times. Gaps in sequence data were read as missing data.

Character Evolution

Two different optimizations of diastema size were conducted on the tree derived from our k = 3 analysis. First the variation was divided into 4 equal states and modeled using likelihood with a Mk1 model and equal branch lengths in Mesquite [

]. Next the variation was mapped as a continuous ordered character in TNT [

]. The fragment ZMT-62 was previously interpreted as including p2-p4, but here we reinterpret this specimen as including p4-m2. An undescribed specimen, ChM PV4745, which was included in some previous studies [

] was not included here because it is clearly a juvenile specimen, and may be conspecific with Coronodon havensteini.

Quantification and Statistical Analysis

Relative Size of Diastema

Diastema length was standardized by the length of p4 (or closest tooth) (see Table S3).

Estimates of Body Size

Body Length

Body length estimates were calculated using the following equation [

], where TL is total body length and BIZYG is maximum width across the zygomatic processes of the skull.

Body Mass

Body mass in kg (BM) was estimated from body length in cm with this equation [

Author Contributions

Conceptualization, J.H.G. and B.L.B.; Formal Analysis, J.H.G.; Investigation, all authors; Resources, M.B.; Writing, J.H.G., B.L.B., and R.W.B; Funding Acquisition, J.H.G., B.L.B., and M.B.

Acknowledgments

During this project we benefitted from discussions with M. Churchill, M. Mihlbachler, and A. Sanders. We thank S. Boessenecker (Mace Brown Museum of Natural History), M. Gibson, J. McCormick, and A. Sanders (The Charleston Museum) for access to specimens. We acknowledge the Department of Radiology, Medical University of South Carolina, for CT scans of the holotype. Use of the application TNT was provided by the Willi Hennig Society. This research was supported by the National Science Foundation (NSF EAR-1349607 to J.H.G. and B.L.B.).

Supplemental Information

Download .pdf (.59 MB)

Help with pdf files

Document S1. Figures S1–S4 and Tables S1–S3

Download .doc (.14 MB)

Help with doc files

Data S1. Details on Systematics, Morphology, and Paleobiological Interpretations of Coronodon havensteini gen. et sp. nov., Related to Figures 1, 2, and 3

Download .doc (.09 MB)

Help with doc files

Methods S1. Character List Used for Phylogenetic Analysis, Related to Figure 4 and STAR Methods

References

Goldbogen J.A.
Calambokidis J.
Croll D.A.
McKenna M.F.
Oleson E.
Potvin J.
Pyenson N.D.
Schorr G.
Shadwick R.E.
Tershy B.R.

Scaling of lunge-feeding performance in rorqual whales: mass-specific energy expenditure increases with body size and progressively limits diving capacity.

Funct. Ecol. 2012; 26: 216-226

sexta-feira, 19 de outubro de 2018