Evidence for convergent evolution of ultrasonic hearing in toothed whales (Cetacea: Odontoceti)

Abstract

Toothed whales (Cetacea: Odontoceti) are the most diverse group of modern cetaceans, originating during the Eocene/Oligocene transition approximately 38 Ma. All extant odontocetes echolocate; a single origin for this behaviour is supported by a unique facial source for ultrasonic vocalizations and a cochlea adapted for hearing the corresponding echoes. The craniofacial and inner ear morphology of Oligocene odontocetes support a rapid (less than 5 Myr) early evolution of echolocation. Although some cranial features in the stem odontocetes Simocetus and Olympicetus suggest an ability to generate ultrasonic sound, until now, the bony labyrinths of taxa of this grade have not been investigated. Here, we use µCT to examine a petrosal of a taxon with clear similarities to Olympicetus avitus. Measurements of the bony labyrinth, when added to an extensive dataset of cetartiodactyls, resulted in this specimen sharing a morphospace with stem whales, suggesting a transitional inner ear. This discovery implies that either the lineage leading to this Olympicetus­-like taxon lost the ability to hear ultrasonic sound, or adaptations for ultrasonic hearing evolved twice, once in xenorophids and again on the stem of the odontocete crown group. We favour the latter interpretation as it matches a well-documented convergence of craniofacial morphology between xenorophids and extant odontocetes.

1. Background

Odontocetes (toothed whales, dolphins and porpoises) are a highly successful group of marine mammals, with 75 species [1] currently distributed across Earth's oceans, from the Arctic, to the tropics, to depths of 3000 m [2]. Undoubtedly one of the key adaptations that underlies their success is echolocation: the use of sound to hunt or navigate in low visibility waters. Compelling evidence for a single origin of echolocation using ultrasonic sound in Odontoceti is present (electronic supplementary material, S1), but determining when this adaptation evolved requires detailed study of the fossil record. Based on osteological correlates (electronic supplementary material, S2), previous work has concluded that the phonic lips and dorsal bursae complex (MLDB), or a homologous sound-producing structure, was already present in the face of the oldest known odontocetes (ca 34 Ma) [3]. This interpretation was largely driven by the presence of these features in the Xenorophidae, which are typically placed as the first, or one of the first, clades to branch off of the odontocete stem [3–7]. Furthermore, the three studies on xenorophid cochlear morphology that indicate high-frequency hearing in the group all sampled conspecific or congeneric taxa [6,8–10] (electronic supplementary material, S3), and thus it is crucial that inferences on the origin of echolocation based on these small samples be tested with data from other fossil odontocetes.

The specimen examined here, CCNHM 1000, is a small odontocete skull that lacks key synapomorphies of Xenorophidae [5], and otherwise closely resembles Olympicetus avitus [11] from the same locality. Phylogenetic analyses of Olympicetus avitus reconstruct it in a polytomy with Xenorophidae and the clade including Simocetus + crown Odontoceti [11], or more closely related to crown Odontoceti (this study). The new material we describe here tests the hypothesis that echolocation involving ultrasonic frequencies evolved early in Odontoceti.

2. Material and methods

The petrosal of CCNHM 1000 (detailed descriptive data including specimen locality and age provided in electronic supplementary material, S4) was µCT scanned using the North Star Imaging µCT scanner of the Department of Earth and Environmental Sciences at Vanderbilt University (TN, USA). A high-resolution scan focused on the bony labyrinth resulted in a voxel size of 17.2 µm. Extraction of a digital endocast from the close-up CT scan data was performed in VGStudioMax v. 2.2, following previous work [12–14].

Recent studies [14–16] have shown that nine measurements of cetartiodactyl cochleae enable terrestrial, ultrasonic, infrasonic and specialized narrow-band high-frequency hearers to be separated in a principal components analysis (PCA). To infer the hearing sensitivity of CCNHM 1000, we built on these analyses by acquiring these nine measurements for CCNHM 1000 and adding them to the PCA performed by Racicot et al. [14], which includes 25 specimens representing 13 taxa added to the analysis performed by Mourlam & Orliac [15] (electronic supplementary material, Data table S1; S5). Rotational head movement sensitivity of CCNHM 1000 was estimated using ‘90var’ [17,18] (electronic supplementary material, S6). The ontogenetic status of CCNHM 1000 should not impact inner ear measurements because cetaceans exhibit exceptionally precocial development of the inner ear [19].

The phylogenetic position of CCNHM 1000 was reconstructed based on a revised supermatrix of morphological and molecular data [20]. We added CCNHM 1000, Olympicetus avitus [11] and Inermorostrum xenops [7] to this matrix, resulting in 114 taxa coded for 61 177 characters. A maximum parsimony phylogenetic analysis was performed in the application TNT [21], using default settings under a ‘New Technology Search’ except for setting the minimum length to be found 1000 times. Support for clades was assessed using bootstrap with replacement in TNT, with absolute frequencies reported and 50 random taxon addition sequences for each bootstrap replicate (electronic supplementary material, S7). Five characters derived from measurement data (following [15]) were coded for the specimens from our dataset (electronic supplementary material, S8). Ancestral character state reconstructions of all characters were performed on the agreement subtrees that include all taxa for which these five characters could be coded. We reconstructed ancestral states using Mesquite v. 3.03, with the parsimony option for ‘Trace Character History’ and with the parsimony model ‘Ordered’ chosen (electronic supplementary material, S8).

Anatomical abbreviations—aa, anterior ampulla; asc, anterior semicircular canal; cc, canaliculus cochleae for membranous perilymphatic duct; co, cochlea; cr, common crus; er, elliptical recess of vestibule; fc, fenestra cochleae; fv, fenestra vestibuli; la, lateral ampulla; lsc, lateral semicircular canal; pl, primary bony lamina; psc, posterior semicircular canal; sl, secondary bony lamina; sr, spherical recess.

3. Results

(a). Specimen identification

CCNHM 1000 closely resembles Olympicetus avitus in a suite of characters (electronic supplementary material, S9). The phylogenetic analysis found 16 most parsimonious trees, each 4 173 236 steps in length (CI: 0.554, RI: 0.480), resulting in CCNHM 1000 placed in a clade of all odontocetes except xenorophids, Archaeodelphis patrius and Ashleycetus planicapitis (electronic supplementary material, S7). This clade is supported by five synapomorphies in all most parsimonious trees: base of rostrum slightly bowed (character 12, 0 > 1), occipital bears sagittal crest (156, 0 > 1), petrosal lacks groove for capsuloparietal vein (207, 0 > 1), deep anterior bullar facet on petrosal (210, 1 > 3) and posterior process of bulla has few laminations (264, 1 > 0).

In all most parsimonious trees, CCNHM 1000 is outside a clade that includes Mirocetus, Olympicetus, and crown odontocetes, among other taxa. Thus, one or two internal branches separate CCNHM 1000 and Olympicetus avitus. The clade that includes Olympicetus but excludes CCNHM 1000 is supported by two putative synapomorphies: temporal crest far behind orbit and roofs over temporal fossa (132, 1 > 0), and long zygomatic process (189, 2 > 3). The key question is whether these differences could be because the type and referred specimens of Olympicetus are of a more advanced ontogenetic stage than CCNHM 1000. We suggest that CCNHM 1000 represents neonate remains based on the unfused intersphenoidal and sphenoccipital synchondroses, persistent interparietal, unfused median frontal suture, limited dental development (enamel and minor mineralization of distal dentine shell of the roots), and small size (approx. 75% of the neonatal specimen LACM 126010).

(b). Bony labyrinth morphology

Much like the rest of the skull of CCNHM 1000 (electronic supplementary material, S4), the bony labyrinth displays a mixture of derived and plesiomorphic features (figure 1). Among the most striking features related to hearing sensitivity is the number of turns, which at 2.39 is well above the typical number of turns for odontocetes (less than or equal to two turns) and similar to the number of turns in archaeocetes (2.1–2.5) or mysticetes (2–3.3). Compared to cf. Cynthiacetus, Echovenator sandersi and Squalodon calvertensis [6] the apical turns of CCNHM 1000 are more loosely coiled (electronic supplementary material, S10).

Figure 1.

Morphology of the bony labyrinth of CCNHM 1000. Digital endocast of the left bony labyrinth rendered from CT data in (a) anterior, (b) lateral, (c) dorsal and (d) vestibular view. Anatomical abbreviations listed in the methods section.

(c). Inferred hearing sensitivity and head rotational sensitivity

The specimen CCNHM 1000 occupies a position in the PCA among protocetids, close to Carolinacetus sp. [15] (figure 2a), suggesting a hearing sensitivity little modified from those of terrestrial artiodactyls and far different from modern odontocetes. A variables factor map of the analysis, which presents a view of the observed variables projected into the plane defined by the first two principal components, illustrates that the primary driver of this position is the number of cochlear turns (‘T’, inset figure 2a). The 90var value for CCNHM 1000 (5.7) is low, suggesting high sensitivity to head movements (electronic supplementary material, S6).

Figure 2.

(a) Principle components analysis (PCA) of nine cochlear measurements with variables factor map inset. Dataset modified from [14] to include CCNHM 1000 from this study and a revised number of cochlear turns estimate for Orcinus orca (electronic supplementary material, Data table S1). Morphospaces for taxonomic groupings (and corresponding hearing types) are highlighted in different colours; the position of CCNHM 1000 is indicated with a blue star. (b) Schematic tree based on our phylogenetic analysis (electronic supplementary material, figure S4) showing alternative evolutionary histories for ultrasonic hearing, as indicated by the morphospaces implied by the PCA. Ancestral state reconstructions for two cochlear features support parallel evolution (electronic supplementary material, figures S5–S6, S13–S14) whereas one supports loss of ultrasonic hearing (electronic supplementary material, figures S9–S10).

(d). Phylogenetic position and evolution of high-frequency hearing in toothed whales

In the strict consensus of the 16 most parsimonious trees, CCNHM 1000 is positioned as the second lineage along the odontocete stem, with the Xenorophidae as sister to this group (figure 2; electronic supplementary material, S7). Although bootstrap values support exclusion of CCNHM 1000 from the Xenorophidae, the relative positions of Olympicetus, xenorophids, and other archaic cetaceans along the odontocete stem received low support. If we take the tightly circumscribed morphospaces on the PCA plot as indicative of hearing capabilities (figure 2a), then it is equally parsimonious to infer (1) that CCNHM 1000 lost the ability to hear ultrasonic sound, or (2) parallel evolution of ultrasonic hearing sensitivity in Xenorophidae and crown odontocetes (figure 2b). Both scenarios receive support from ancestral state reconstructions of individual features that contributed to the PCA: parallel evolution occurs in the basal ratio and length of the secondary bony lamina whereas a reversal occurs for inter-turn distance (electronic supplementary material, S8 and figures S5–S14).

4. Discussion

Comparative cochlear measurements, PCA, and reconstructions of ancestral character states reveal unforeseen complexity in the auditory evolution in odontocetes. Previous analyses of xenorophids, the sister group to the clade including most if not all other odontocetes, show that this group possessed high-frequency hearing and likely were capable of echolocation [6,9,10]. CCNHM 1000 possessed relatively unspecialized hearing sensitivity, which, based on the PCA, had a maximum hearing threshold comparable to mysticetes or some terrestrial artiodactyls. Our findings imply one of two equally parsimonious scenarios: either the ability to hear higher frequencies evolved twice, once in xenorophids and again on the stem to the crown group, or that CCNHM 1000 lost this ability.

Although we cannot rule out a loss of ultrasonic hearing in CCNHM 1000, the balance of evidence points to parallel evolution of ultrasonic hearing in odontocetes. Although posterior expansion of the nasal process of the maxilla is an odontocete synapomorphy, it is fairly small in some taxa, most notably Archaeodelphis patrius [22]. When the size of this process, as well as posterior migration of the external nares, are mapped on odontocete phylogeny, there is clear convergence in these characters between xenorophids and the odontocete crown group [3,23]. If correlations between facial muscles and the size of the process, as seen in extant odontocetes, apply to extinct species, then this implies that some aspects of the MLDB complex and associated muscles evolved in parallel in both clades alongside the previously noted parallel evolution of retrograde cranial telescoping [23]. Furthermore, fossae for air sinuses in xenorophids suggest that their soft tissue morphology was quite different from those of extant odontocetes, including a deep rostral basin, a narrow premaxillary fossa, and a postnarial fossa [3]. The hypothesis of parallel evolution of echolocation was briefly mentioned, but then dismissed, by [3], but the homoplasy implied by CCNHM 1000 indicates it may, in fact, be correct. The simplest explanation that summarizes all of these observations is that the MLDB complex did evolve first, but initially was used to generate high, but not ultrasonic, frequencies. Once established in odontocetes, we further hypothesize the MLDB complex further specialized separately in xenorophids and along the stem to crown Odontoceti to be able to produce ultrasonic sound, and this occurred simultaneously with parallel evolution of ultrasonic hearing.

We conceive of two important ways to test the hypothesis we have outlined here. First, it is possible that we have not identified more subtle morphological features indicative of ultrasonic hearing in odontocete cochleae. Expanding the application of three-dimensional geometric morphometrics to a much larger sample of extinct, stem odontocetes has already provided important insights into cochlear evolution across all Mysticeti [10,24], and correlations between habitat and cochlear shape in crown odontocetes [25]. Another avenue would be to revisit the laminar gap, which approximates the width of the basilar membrane, and is much narrower in taxa with ultrasonic hearing [26]. Preservation of bony lamina in some fossil specimens is on par with what is seen in extant osteological specimens (e.g. [12,13,16]), however, and we encourage a second look at the pioneering work of Fleischer [26], but with the aid of a much-improved fossil record and detailed investigation of these fossils using µCT technology.

In summary, our findings indicate that one of the earliest odontocetes along the stem to modern (crown) odontocetes did not have ultrasonic hearing sensitivity. Clear convergence in the evolution of facial morphology and cochlear morphology between a clade of extinct archaic dolphins (Xenorophidae) and the clade including modern (crown) dolphins indicates that echolocation evolved independently in each group, an exciting and unexpected result revealing that pinpointing the possible origin(s) of echolocation in the extant clade of odontocetes will require further comprehensive analyses.

Data accessibility

CT data files and three-dimensional model are available at Dryad: https://doi.org/10.5061/dryad.9hg52sm [27]. The electronic supplementary material includes text, figures, code, and phylogenetic and measurement datasets.

Authors' contributions

R.A.R. designed the research, analysed data from the µCT scan and performed ancestral character state reconstructions. R.A.R. and S.A.F.D. µCT scanned the specimen. S.A.F.D. performed statistical analyses. R.W.B. identified, described and prepared the specimen. J.H.G. coded characters and implemented phylogenetic analyses. All authors wrote the manuscript. All authors approved the final version of this manuscript and agree to be held accountable for the content.

Competing interests

We have no competing interests.

Funding

This work was supported by Vanderbilt University and grant no. NSF-EAR 1349607 to J.H.G.