Live CT imaging of sound reception anatomy and hearing measurements in the pygmy killer whale, Feresa attenuata

Human-generated noise in the oceans has increased over the last century, which has raised concern about its impact on marine mammals (Richardson et al., 1995; Miller et al., 2000; National Research Council, 2005; Di Iorio and Clark, 2010) (reviewed by Malakoff, 2010). Some cetaceans may be particularly sensitive to changes in their acoustic habitat because they rely on vocalizations and a well-tuned auditory system for communication, detection of prey and predators, and orientation. Anthropogenic ocean noise may cause acoustic masking of individuals and populations attempting to perform these tasks, and these effects could be exacerbated if animals have hearing deficits. These deficits could be caused by chronic noise, age-related hearing loss, congenital hearing impairment, parasites and ototoxic drug treatment for animals under human care (reviewed by Houser and Finneran, 2006a). In addition to these factors, one extremely important, unexplored mechanism in marine mammals may be the impacts of chemical pollution on the development of hearing. In fact, odontocetes (toothed whales, dolphins and porpoises) have been shown to accumulate extremely high levels of polychlorinated biphenyls (PCBs) (Kannan et al., 1993; Muir et al., 1996; Ross et al., 2000; Hansen et al., 2004) in their blubber and have been shown to pass this burden on to their offspring (Wells et al., 2005). Studies with rats have shown that developmental exposure to PCBs results in severe hearing loss (Goldey et al., 1995; Goldey and Crofton, 1998; Crofton et al., 2000).

Thus, it is important to learn as much as we can about the hearing of coastal and pelagic cetaceans at a structural and functional level. Post-mortem investigations and anatomical dissections have revealed important characteristics of sound reception in delphinids, which include fusion of the middle and inner ear into the tympanoperiotic complex, the separation of the bullar complex from the skull, air-filled sinuses surrounding the bulla (Ketten, 2000), and mandibles that are hollow but contain fats that channel sound to the middle and inner ears (e.g. Norris, 1964; McCormick et al., 1970; Varanasi and Malins, 1970a; Varanasi and Malins, 1970b; Brill et al., 1988). X-ray computed tomography (CT) has added to this body of knowledge by providing a means of investigating in situ auditory anatomy without anatomical dissection that can be used on postmortem specimens and live animals (e.g. Ketten and Wartzok, 1990; Houser et al., 2004; Cranford et al., 2008).

For many decades, the standard approach for examining hearing sensitivity in marine mammals has been behavioral testing (reviewed by Nachtigall et al., 2000). Behavioral audiograms have been determined for the bottlenose dolphin, Tursiops truncatus (Johnson, 1967), harbor porpoise, Phocoena phocoena (Anderson, 1970), common dolphin, Delphinus delphis (Belkovich and Solntseva, 1970), killer whale, Orcinus orca (Hall and Johnson, 1972), Amazon river dolphin, Inia geoffrensis (Jacobs and Hall, 1972), beluga whale, Delphinapterus leucas (Awbrey et al., 1988), false killer whale, Pseudorca crassidens (Thomas et al., 1988), and tucuxi, Sotalia fluviatilis guianensis (Sauerland and Dehnhardt, 1998). This approach is not practical when assessing hearing in stranded, hospitalized or newly captured dolphins and small whales because it requires training to test the hearing of a naive animal. However, auditory evoked potential (AEP) techniques provide a means to study hearing sensitivity quickly by exposing the animal to a calibrated sound stimulus and then measuring the evoked potential on the surface of the skin with suction cup electrodes positioned on the head (e.g. Cook et al., 2006). The use of AEP techniques to measure hearing sensitivity of odontocetes has become increasingly popular (e.g. Popov and Supin, 1990a; Popov and Supin, 1990b; Szymanksi et al., 1999; Andre et al., 2003; Nachtigall et al., 2004; Nachtigall et al., 2005; Yuen et al., 2005; Cook et al., 2006; Houser and Finneran, 2006a; Mooney et al., 2008; Nachtigall et al., 2008; Mooney et al., 2009). Thus, a powerful tool to learn more about the hearing of live-stranded cetaceans, particularly those that are rare or endangered, would be to combine CT imaging and AEP techniques.

The pygmy killer whale, Feresa attenuata, is a pelagic species found in subtropical and tropical waters of the Atlantic, Indian and Pacific Oceans (Ross and Leatherwood, 1994). These whales are seldom seen and are one of the most poorly understood toothed whales (McSweeney et al., 2009). Much of what is known about the ecology and biology of pygmy killer whales comes from stranded specimens (reviewed by Mignucci-Giannoni et al., 1999). However, no studies have focused on hearing of this rarely observed pelagic species.

On 16 June 2008, two pygmy killer whales were stranded alive separately near Boca Grande, FL, USA, and were taken into rehabilitation. We used this opportunity to learn more about the structural anatomy and hearing sensitivity of these rare toothed whales. The specific objectives were to: (1) perform CT imaging of a live pygmy killer whale to create three-dimensional (3-D) reconstructions of the peripheral hearing apparatus; (2) measure the modulation rate transfer function (MRTF) for these two animals and estimate AEP hearing thresholds; (3) characterize the auditory brainstem response (ABR) in this species; (4) examine how varying the electrode placement on the head surface affected the amplitude of the ABR waves; and (5) provide information on the neuroanatomical sources of the ABR waves by integrating the CT and electrode mapping data.

Two male pygmy killer whales, Feresa attenuata Gray 1875 (MML0802 and MML0803), were rescued on 16 June 2008 near Boca Grande, FL, USA. The whales were transported to Mote Marine Laboratory and Aquarium's Dolphin and Whale Hospital (DWH), where they were maintained in an aboveground tank (9 m diameter and 3 m deep). The body lengths and weights of MML0802 and MML0803 were 211 and 213 cm, and 110 and 107.5 kg, respectively. These whales presented with severe weakness, disorientation and multiple systemic bacterial and parasitic infections.

On 16 June (the day of arrival at DWH), AEP measurements were completed on MML0802 and MML0803. During rehabilitation, the two whales received several antibiotics and antifungals, including amikacin sulfate, which was administered to MML0802 from 14 to 22 July 2008 and from 15 to 24 September 2008, every other day, intramuscularly at a dose of 21 mg kg^–1. CT imaging of MML0802 was completed on 10 July 2008. On 29 September 2008, AEP measurements were performed again on MML0802 to determine whether antibiotics might have caused hearing loss. Hearing tests and CT imaging were performed under the direct supervision of Charlie Manire, DVM.

MML0802 died in captivity on 23 November 2008. The necropsy revealed hemorrhaging in the stomach, lungs, and parietal and visceral pleura, as well as rib fractures. Post-mortem magnetic resonance imaging (MRI) of the head was completed, and the images were interpreted by a veterinary and a human radiologist (Sophie Dennison and Jerome Barakos, respectively). The MRI revealed a large parenchymal hemorrhage in the right cerebral hemisphere with associated extension of blood into the ventricular system. Given the lack of significant edema and lack of obstructive hydrocephalus, this hemorrhage occurred immediately before death. MML0803 died in captivity on 3 September 2008. The necropsy revealed a parasitic infestation of the lungs and cerebellum. Histological examination noted large aggregates of trematode ova and fragments (likely Nasitrema spp.) in the cerebellum and associated macrophages and white matter loss. Death was attributed to parasitic encephalitis and acute pneumonia.

MML0802 was transported to Axcess Diagnostics in Sarasota, FL, located ∼10 miles from DWH. The whale was given midazolam (0.05 mg kg^–1 intramuscularly) to reduce anxiety during transport and movement during scanning. A scan of the head and thorax was completed using a Philips Brilliance CT 64-channel scanner (Philips Healthcare, Andover, MA, USA). Images were acquired in the transaxial plane (i.e. at right angles to the long axis of the body) and helically by rotating an X-ray source of 140 kV at 399 mA. A total of 250 transverse slices at 3 mm thickness were collected, with a matrix size of 512χ512 and a field of view of 48.7χ48.7 cm. These parameters yielded voxel dimensions of 0.9χ0.9χ3.0 mm.

Segmentations (i.e. assigning pixels to particular structures), 3-D reconstructions and volume calculations were conducted using the software program AMIRA 5.2.2 (Mercury Computer Systems, San Diego, CA, USA). Anatomical structures were identified using the head atlas of the bottlenose dolphin (Houser et al., 2004) and beaked whale, Ziphius cavirostris (Cranford et al., 2008). Briefly, 3-D reconstructions were completed for the skull, mandibles, brain, tympanoperiotic complex, outer core of the mandibular fat body, inner core of the mandibular fat body and cranial air spaces, which consisted of the nasal passages and laryngeal air, pterygoid sinus and peribullary sinus. Segmentations were completed by applying a threshold of tissue density values [represented by Hounsfield units (HU)] that defined each anatomical structure. For example, the values for the inner core of the mandibular fat body ranged from –139 to –91 HU whereas values for the outer core of the mandibular fat body ranged from –90 to 10 HU. This threshold procedure was followed by visual inspection and manual editing to ensure that structures were properly defined. Spatial relationships of anatomical structures were visualized by making some structures shaded and others transparent.

Background noise in the aboveground tank at DWH was recorded using a RESON TC4013 hydrophone (RESON, Inc., Goleta, CA, USA; sensitivity –211 dB re. 1 V μPa^–1 and frequency response up to 170 kHz). The hydrophone was connected to a TDT AEP Workstation (Tucker-Davis Technologies, Alachua, FL, USA) and laptop computer. The TDT RX6 sample rate was 260,416.67 Hz. Background noise was calculated as spectrum level (dB re. 1 μPa² Hz^–1), and these data were included in audiograms (see Figs 4 and 5).

Tests were conducted in the water, at a depth of ∼1.5 m, with the whale held so that the top of its head was above the water and the lower jaw was submerged. Stimulus presentation and data acquisition was controlled from a TDT AEP Workstation attached to a laptop computer. On 16 June 2008, the first set of tests comprised MRTF measurements for MML0802 and MML0803. Stimuli consisted of sinusoidal amplitude modulated (SAM) tones. The MRTF was measured by presenting a 40 kHz carrier tone at 117±3 dB re. 1 μPa_r.m.s. (mean ± s.d. of three calibrations) with modulation frequencies from 200 to 2000 Hz in 100 Hz steps. The MRTF results were used to determine the modulation frequency that yielded a strong and distinct AEP response. This frequency was used as the modulation frequency to measure AEP hearing thresholds.

The second set of tests involved the envelope following response (EFR) procedure, which was used to estimate AEP hearing thresholds. To determine frequency-specific threshold estimates, SAM tones were played to the whales at specific frequencies and levels. These SAM tones consisted of 14 ms tone bursts modulated at 1000 Hz (i.e. the modulation frequency that resulted in a strong and distinct response). The carrier frequencies tested were 5, 10, 20, 30, 40, 60, 80, 100 and 120 kHz. Sound pressure levels (SPLs) of the SAM tones were attenuated in 6 dB steps.

The third set of tests recorded ABRs (in response to click stimuli). The ABR allows the viewing of the electrical activity of the different brain structures in response to a click stimulus. To determine the ABR, a click stimulus of 384 μs in duration was presented to the whales (supplementary material Fig. S1A). The SPL of the frequency spectrum of the click is presented in supplementary material Fig. S1B. The peak frequency was 98 kHz, and the –10 dB bandwidth was 90–120 kHz. SPLs of the clicks were also attenuated in 6 dB steps.

On 29 September 2008, the EFR procedure was repeated for MML0802 to determine whether AEP hearing thresholds changed following treatment with the potentially ototoxic antibiotic, amikacin sulfate. In addition, we examined how changing the electrode placement on the head surface of MML0802 affected the amplitude of the ABR. From these data, we tested whether there was a relationship between the distance of electrodes to auditory brain structures and amplitudes of the specific ABR waves.

All sound stimuli were presented to the whales using a jawphone. The jawphone was composed of an ITC-1042 transducer (International Transducer Corporation, Santa Barbara, CA, USA) embedded in a suction cup with impedance similar to that of water (constructed from VI-SIL V-1062 silicone rubber; Rhodia, Inc.; Cranbury, NJ, USA). The acoustic signals were transmitted from the jawphone after amplification by a Hafler P1000 amplifier (Tempe, AZ, USA). SPLs were controlled by the computer using a programmable attenuator (TDT PA5, Tucker-Davis Technologies). The jawphone was placed on the lower left jaw ∼10 cm from the tympanoperiotic complex (supplementary material Fig. S2). Acoustic stimuli were calibrated by placing the jawphone 30 cm below the water surface and measuring the sound level 10 cm away with a hydrophone (RESON TC4013, RESON, Inc.; sensitivity –211 dB re. 1 V μPa^–1 and frequency response up to 140 kHz).

Evoked potentials were measured with suction cup electrodes that were composed of standard 8 mm gold-plated electrodes (Rochester Electro-Medical, Tampa, FL, USA) embedded in silicone rubber (VI-SIL V-1062; Rhodia, Inc.). The body was dried with gauze pads before attachment. Redux® electrolyte paste (Parker Laboratories, Inc., Prescott, AZ, USA) was placed on the underside of the suction cup to provide a connection between the electrodes and the whale's skin. The recording electrode 1 (defined as ‘Center’) was placed ∼9 cm caudal to the blowhole, and recording electrode 2 (defined as ‘Rec2’) was placed ∼15 cm caudal to the blowhole. Both electrodes were centered on the dorsal midline. The reference electrode was positioned ∼15 cm caudal to Rec2, and a ground electrode was placed in the water. All electrodes were attached to an HS4 Fiber Optic Bioamp Headstage (Tucker-Davis Technologies), which was connected via a fiber-optic cable to the TDT system for data acquisition. The TDT system was connected to a laptop; BioSig software (Tucker-Davis Technologies) was used to collect the evoked potential data at a sample rate of 24.4 kHz. AEP signals were low-pass filtered at 3 kHz, high-pass filtered at 50 Hz and notch-filtered at 60 and 120 Hz prior to signal averaging. Large electrical artifacts originating from whale breathing and movement were removed by artifact rejection in BioSig, which did not affect the AEP signal. For each stimulus, 500 to 1000 evoked response epochs were recorded. Once an evoked potential was recognized, averaging at that test level was terminated and the next level was tested. The mean biological background noise level was 8 nV.

Data analysis was conducted using MATLAB and Excel. To determine the MRTF, the EFR amplitudes were measured by taking the root mean square (r.m.s.) of the voltages and then plotted with respect to modulation frequency. To estimate AEP hearing thresholds, a fast Fourier transform (FFT) was calculated from the AEP response only, consisting of 440 points with a frequency resolution of 55.5 Hz. Then, the peak FFT amplitude at the 1000 Hz modulation frequency was determined, as well as the mean noise level three bins below and above this signal. From these values, the signal-to-noise ratio was calculated by dividing the peak FFT amplitude by the mean noise level. An F-test was used to determine whether the AEP signal was significantly different from the noise based on α0.05 (Dobie and Wilson, 1996). If the calculated signal-to-noise ratio was greater than the critical F-value, then an AEP signal existed (Dobie and Wilson, 1996). Frequency-specific threshold estimates were then determined by reporting the lowest SPL in which the AEP signal was significantly different from the noise. The F-test is preferable to t-tests and comparable to magnitude-squared coherence to determine AEP signals amidst background noise (Dobie and Wilson, 1996). Input–output functions of evoked potential level were plotted for each frequency, and frequency-specific threshold estimates were plotted relative to tank noise. For ABRs, the response was high-pass filtered at 450 Hz. ABR waves were identified based on previously documented studies in the bottlenose and common dolphins (Ridgway et al., 1981). Peak amplitudes and latency of ABR waves II, III, IV and VI were determined and plotted as a function of SPL. Linear regression was used to determine the relationship between sound level and latency.

We examined how changing the electrode placement on the head surface of MML0802 affected the ABR (see supplementary material Fig. S1 for the power spectrum of the click stimulus). This test was completed by changing the Center electrode position to various locations on the left and right side of the head (i.e. to positions Left 1C, Left 1D, Left 1F, Right 1B, Right 1C, Right 1D, Right 1E and Right 1F) but keeping Rec2, reference and ground electrodes in the same position. Left 1C was positioned 9 cm ventral to Rec 2 (which was placed ∼15 cm caudal to the blowhole), Left 1D was 15 cm ventral and 3 cm caudal to Rec 2, and Left 1F was 15 cm ventral to Rec 2. Right 1B was positioned 9 cm ventral and 12 cm rostral to Rec 2, Right 1C was 9 cm ventral and 6 cm rostral to Rec 2, Right 1D was 9 cm ventral to Rec 2, Right 1E was 15 cm ventral and 12 cm rostral to Rec 2, and Right 1F was 15 cm ventral and 6 cm rostral to Rec 2. For each electrode location, the ABR was recorded in response to click stimuli presented at five SPLs (164, 158, 152, 146 and 122 dB re. 1 μPa). From CT images, 3-D models were created of the body surface, electrodes at various positions, brain and tympanoperiotic complex. Spatial relationships of the electrode sites relative to the brain and tympanoperiotic complex were visualized by making the body surface transparent but the other structures shaded. Locations of electrodes on the 3-D model were mapped from photographs of electrodes on the head surface.

We integrated the CT and electrode mapping data to provide information on the neuroanatomical sources of the ABR waves. Based on past research (Allen and Starr, 1978; Ridgway et al., 1981) and on some speculation, we hypothesized that wave I arose from the auditory nerve, wave II from the cochlear nucleus, wave III from the superior olive, wave IV from the inferior colliculus and wave VI from the medial geniculate body. To test these hypotheses, we determined the relationship of the distances of electrode positions to the auditory nerve, cochlear nucleus, superior olive, inferior colliculus and medial geniculate body and the amplitude of the respective ABR waves. A negative relationship between the distance and amplitude would provide support that the hypothesized neuroanatomical structure was the actual source of the respective ABR wave. Amplitudes for each wave were determined by dividing the peak amplitude of the evoked potential response to the five different SPLs (164, 158, 152, 146 and 122 dB re. 1 μPa) by the corresponding peak amplitude of the evoked potential response for Rec2. These ratios were then averaged for all SPLs. For each ABR wave, these ratios were then plotted as a function of electrode distance from the auditory nerve, cochlear nucleus, superior olive, inferior colliculus and medial geniculate body. These distances were determined from the CT images and 3-D reconstructions of the electrodes and brain, with anatomical guidance from an MRI-based atlas of the brain of an Atlantic white-sided dolphin (Lagenorhynchus acutus) (Montie et al., 2007). Linear regression was used to determine whether there was a significant negative relationship between the distances and mean ratios.

CT imaging and 3-D reconstructions of the pygmy killer whale revealed a peripheral auditory system that was similar to the system of other odontocetes (Figs 1 and 2; see supplementary material Movie 1) (e.g. Cranford et al., 2008). The mandibles were hollow, lacked a bony lamina medial to the pan bone and contained mandibular fat bodies. This fat was not homogenous but contained an inner core (defined by HU that ranged from –139 to –91) surrounded by a denser outer core (defined by HU that ranged from –90 to 10). The inner core of the mandibular fat body continued caudally and abutted the tympanoperiotic complex in two locations. One branch attached along the lateral aspect of the bulla whereas the other branch attached to the ventral aspect of the bulla (Fig. 2E). The volumes of the mandibles, tympanoperiotic complex and mandibular fat bodies were determined (supplementary material Table S1).

The spatial relationship of the cranial air spaces to the tympanoperiotic complex was also investigated (Fig. 2; see supplementary material Movie 1). Identification and labeling of sinuses followed previously published studies (Fraser and Purves, 1960; Houser et al., 2004). CT imaging and 3-D reconstructions revealed three separate, major pockets of air: the nasal passages and laryngeal air, the left pterygoid sinus attached to the left peribullary sinus, and the right pterygoid sinus attached to the right peribullary sinus. The pterygoid sinus was composed of the primary sinus and the optic lobe. The peribullary sinus partially surrounded the tympanoperiotic complex, forming a bone–air interface. The volumes of the pterygoid sinus and the peribullary sinus were determined (supplementary material Table S1).

The MRTF was measured in the pygmy killer whales by presenting a 40 kHz carrier tone and varying the modulation frequency from 200 to 2000 Hz at 100 Hz steps (supplementary material Fig. S3). This approach was used to determine the amplitude of the evoked potential at the different modulation frequencies (Fig. 3). In both whales, evoked potentials were detected in response to modulation frequencies ranging from 200 to 1500 Hz. Strong potentials were observed at 500 and 1000 Hz whereas distinct troughs were observed at 800 Hz (Fig. 3). The strongest response occurred at 500 Hz. For subsequent measurements to determine frequency-specific thresholds, we selected 1000 Hz as the modulation frequency rather than 500 Hz, because of the lower noise floor near 1000 Hz.

For MML0802 and MML0803, input–output functions were plotted for each frequency to compare the evoked potential and SPLs (supplementary material Figs S4 and S5). In general, for all frequencies tested, whales presented with high SPLs exhibited a strong evoked potential whereas whales presented with low SPLs displayed a minimal response. However, the input–output functions were non-linear. Thus, we used signal-to-noise ratios and the F-test to determine whether the evoked potential for each carrier frequency and level was significantly different from the noise. Using these data, we estimated AEP frequency-specific thresholds (supplementary material Table S2). Threshold estimates from hearing tests on 19 June 2008 were shown as an audiogram relative to tank noise (Fig. 4). Maximum hearing sensitivity occurred between 20 and 60 kHz, with the best sensitivity at 40 kHz.

Some drug therapies, such as aminoglycoside antibiotics, can be ototoxic. Thus, hearing sensitivity of MML0802 was retested on 29 September 2008 after treatment with amikacin sulfate (see Materials and methods for dose and rate of administration). Threshold estimates from hearing tests on 29 September 2008 were compared with estimates obtained on 19 June 2008 (Fig. 5; supplementary material Fig. S5 and Table S2). These comparisons showed a possible 20 dB re. 1 μPa increase in threshold at 5 kHz and a possible 34 dB re. 1 μPa increase in threshold at 10 kHz.

ABRs in response to clicks were determined for the two stranded pygmy killer whales. The duration of the click-evoked potential was ∼6 ms (Fig. 6A). The ABR consisted of seven waves, of which waves II, III, IV and VI were highest in amplitude (Fig. 6A). Waves I, V and VII were often not detectable (Fig. 6A). In general, the ABR waves showed a decrease in amplitude as the click stimuli were attenuated to below threshold levels (Fig. 6B). The peak amplitudes of the waves as a function of SPL were linear for low-to-moderate SPLs but reached an asymptote at sound levels of 130 to 140 dB re. 1 μPa. The absolute latencies were determined for ABR waves II, III, IV and VI and plotted with respect to SPL (Fig. 6C). Linear regression was used to determine how latency of the ABR waves changed with sound intensity. For every 1 dB increase in stimulus intensity, the latency of wave II decreased by 2.2 μs, wave III decreased by 4.3 μs, wave IV decreased by 3.7 μs and wave VI decreased by 1.7 μs (Fig. 6C). The mean absolute latencies (mean of all SPLs) of MML0802 for ABR waves II, III, IV and VI were 2.71, 3.43, 4.34 and 5.95 ms, respectively. MML0803 exhibited similar latencies.

We examined how changing electrode location affected the ABR for MML0802 (Fig. 7; supplementary material Fig. S6). For all electrode placements, the responses to five SPLs were superimposed and waves were identified (Fig. 7; supplementary material Fig. S6). Three-dimensional models were included to display the spatial relationship of the electrode sites relative to the brain and tympanoperiotic complex. Changing the Center electrode position to various locations (i.e. Left 1C, Left 1D, Left 1F, Right 1B, Right 1C, Right 1D, Right 1E and Right 1F) affected the morphology of the click-evoked potential (Fig. 7; supplementary material Fig. S6). When the Center electrode was moved to Left 1D and Left 1F, wave I was more pronounced but waves II, III, IV, VI and VII decreased in amplitude (Fig. 7). When the Center electrode was moved to Right 1D, waves II, III, IV, VI and VII decreased in amplitude (supplementary material Fig. S6).

Using the CT and electrode mapping data, we determined the relationship of the distances of electrode positions to the auditory nerve, cochlear nucleus, superior olive, inferior colliculus and medial geniculate body and the amplitude of their hypothesized waves (supplementary material Fig. S7). A significant negative relationship existed between the distance and amplitude for the inferior colliculus (P=0.009) and the medial geniculate body (P=0.002) but not the auditory nerve (P=0.057), cochlear nucleus (P=0.114) or superior olive (P=0.168) (supplementary material Fig. S7).

In odontocetes, evidence supports the hypothesis that the mandibular fat body in the lower jaw is one acoustic pathway that channels sound to the middle and inner ears (e.g. Norris, 1964; McCormick et al., 1970; Varanasi and Malins, 1970a; Varanasi and Malins, 1970b; Brill et al., 1988; Cranford et al., 2008). In the present study, CT imaging and 3-D reconstructions of the pygmy killer whale revealed mandibles that were hollow, lacked a bony lamina medial to the pan bone and contained mandibular fat bodies (Figs 1 and 2). The mandibular fat bodies were not homogeneous and consisted of an inner and outer core of which the inner core intimately contacted the tympanoperiotic complex at two locations. These attachments have also been observed in CT images of the beaked whale (Cranford et al., 2008). Based on Hounsfield units, in the pygmy killer whale, the fat in the outer core was denser than the fat in the inner core. A previous study has found that the mandibular fat is not homogeneous. Koopman et al. discovered that the shortest and branched-chain fatty acids were concentrated in the inner core of the mandibular fat bodies in several odontocete species (Koopman et al., 2006). The authors hypothesized that this configuration would bend sound inward, channeling it to the middle and inner ears, because sounds travel slower through shorter branched-chain fatty acids than through longer straight-chain fatty acids (Koopman et al., 2006).

MRIs of several odontocete species have shown that there are multiple lobes of fatty tissues in and around the lower jaw (Ketten, 2000). Ketten identified three discrete lobes of fat oriented along different axes that are connected to the tympanoperiotic complex (Ketten, 2000). These structures included a fat body in the lower jaw, a fat body on the external surface of the mandible and a fat body postero-laterally. As suggested by Ketten, these fat bodies may have separate tuning properties, capturing different frequency sounds and delivering these signals to the middle and inner ears (Ketten, 2000). However, the CT images of the pygmy killer whale did not indicate three discrete mandibular fat bodies.

CT images and 3-D reconstructions provided data on the spatial relationship of the cranial air spaces to the tympanoperiotic complex. In the pygmy killer whale, the pterygoid sinus was attached to the peribullary sinus that surrounded the tympanoperiotic complex (Fig. 2; see supplementary material Movie 1). This configuration has been observed in other CT imaging studies of odontocete species such as the bottlenose dolphin (Houser et al., 2004) and the beaked whale (Cranford et al., 2008). The general thought is that the presence of air around the tympanoperiotic complex forms a sound-reflective barrier that acoustically isolates each ear (see Houser et al., 2004; Cranford et al., 2008). It has been suggested that the dorsal-medial coverage of the tympanoperiotic complex with air acts as a barrier and protects the ear from self-made sounds originating in the nasal passages (see Cranford et al., 2008). In addition, the air spaces may aid in hearing directionality by contributing to the animal's ability in timing sound arrival differences between the ear, as the air spaces would impede sound conduction through soft tissues that exist between the ears (Houser et al., 2004).

The MRTFs measured in pygmy killer whales were similar in shape to those MRTFs observed in other odontocetes such as the bottlenose dolphin (Dolphin et al., 1995; Supin and Popov, 1995), beluga whale (Dolphin et al., 1995; Klishin et al., 2000), killer whale (Szymanski et al., 1998), beaked whale (Cook et al., 2006), Risso's dolphin (Grampus griseus) (Mooney et al., 2006) and white-beaked dolphin (Lagenorhynchus albirostris) (Mooney et al., 2009). In the pygmy killer whales MML0802 and MML0803, the MRTF showed maximum peaks at modulation frequencies of 500 and 1000 Hz (Fig. 3; supplementary material Fig. S3). Compared with pinnipeds and manatees, odontocetes have MRTFs that are broader in bandwidth and contain maximum responses at higher frequencies, which suggests higher temporal resolution (Mann et al., 2005; Mulsow and Reichmuth, 2007) (see Mooney et al., 2009). The temporal resolution of the auditory system in odontocetes is orders of magnitude higher than the resolution observed in some terrestrial mammals (i.e. humans and gerbils) (Dolphin and Mountain, 1992; Purcell et al., 2004) (see Mooney et al., 2009). The high temporal resolution in odontocetes is likely a function of the necessity to process and discriminate clicks and echoes quickly and efficiently during echolocation (see Mooney et al., 2006).

All sound stimuli were presented to the whales using a jawphone. This approach has been used by researchers to deliver acoustic stimuli to the lower jaw of bottlenose dolphins and a beaked whale (e.g. Cook et al., 2006; Houser and Finneran, 2006a) because it is generally accepted that the mandibular fats and foramen act as acoustic wave guides (e.g. Norris, 1964; McCormick et al., 1970; Brill et al., 1988; Cranford et al., 2008). The advantage of using a jawphone over a speaker is that the acoustic effects of movement of the animal during testing are reduced. The sound field can change drastically over small distances, and using a jawphone maintains precise timing of sound presentation and evoked potential averaging. However, there are disadvantages of using jawphones as the sound transducer. Using a jawphone delivers sound to a single point, in this case, the lower left jaw, which may not be analogous to presenting sounds via headphones to a human. In addition, during calibration, it is assumed that acoustic energy passing through tissue and through water are equivalent, which may not be the case. Future studies should compare AEP hearing thresholds using a jawphone versus free-field presentation with a speaker.

As is typical of most mammalian audiograms, the AEP-derived audiograms of the pygmy killer whales were U-shaped (Figs 4 and 5). Of the carrier frequencies tested, the lowest hearing thresholds occurred between 20 and 60 kHz, with the best hearing sensitivity at 40 kHz. In general, the auditory sensitivity of the stranded pygmy killer whales in the present study resembled the sensitivity of other odontocete species (supplementary material Table S3) (Szymanski et al., 1999; Kastelein et al., 2002; Kastelein et al., 2003; Nachtigall et al., 2005; Houser and Finneran, 2006a; Nachtigall et al., 2008). The audiogram of the pygmy killer whale was most similar to the audiogram of the killer whale (supplementary material Table S3) (Szymanski et al., 1999). The stranded pygmy killer whales did not hear well at higher frequencies (i.e. thresholds of 96 and 90 dB re. 1 μPa at 100 kHz for MML0802 and MML0803, respectively), which was unlike the pattern observed in the striped dolphin (Kastelein et al., 2003), white-beaked dolphin (Nachtigall et al., 2008) and harbor porpoise (Kastelein et al., 2002), which all exhibited thresholds below 60 dB re. 1 μPa at ∼120 kHz (supplementary material Table S3). However, caution should be applied when comparing frequency-specific thresholds of the present study to thresholds obtained from other hearing studies. Differences often exist in methodology, including whether the test was completed in water or air, how the stimulus was delivered (e.g. transducer attached to the lower jaw or free in the water), how the test was performed (i.e. behavioral versus AEP) and how thresholds were determined.

During rehabilitation at the Mote Marine Laboratory DWH, the pygmy killer whale MML0802 received amikacin sulfate, as well as numerous other antibiotics and antifungals, for treatment of systemic infections. From 14 to 22 July 2008 and from 5 to 24 September 2008, MML0802 received a dose of 21 mg kg^–1 amikacin sulfate intramuscularly every other day. In humans, the aminoglycoside antibiotic amikacin has been shown to be ototoxic (Black et al., 1976). Amikacin has also been implicated in hearing loss of a beluga whale, where this antibiotic was used to treat a Nocardia spp. infection (Finneran et al., 2005). Thus, we compared AEP frequency-specific thresholds obtained from MML0802 on 16 June 2008 with thresholds determined on 29 September 2008 (Fig. 5; supplementary material Fig. S5). These comparisons suggested the possibility of a 20 dB hearing loss at 5 kHz and a 34 dB loss at 10 kHz. However, there was no indication of hearing loss at higher frequencies at the amikacin dose and rate administered.

These results suggest that amikacin treatment may have caused low-frequency hearing loss, which is unexpected. In clinical studies with humans that exhibited hearing impairments caused by amikacin therapy, hearing loss occurred most frequently at higher frequencies (Black et al., 1976). In a beluga whale that showed deficits in auditory sensitivity associated with amikacin treatment, hearing loss also occurred at higher frequencies (Finneran et al., 2005). Thresholds were normal below 37 kHz; however, at 50 kHz and greater, the whale exhibited a hearing loss of ∼90 dB (Finneran et al., 2005). Hearing loss at higher frequencies may occur because amikacin penetrates outer hair cells at the base of the cochlea (which detect high frequencies) to a greater degree than outer hair cells at the apex (which detect low frequencies) (Aran et al., 1995). Thus, amikacin treatment was mostly likely not responsible for the increase in thresholds at 5 and 10 kHz for MML0802.

We investigated the possibility that any changes observed in thresholds were due to inherent variability in hearing measurements. Unfortunately, multiple hearings tests on the same whale on the same day were not completed because the whales were in rehabilitation and prolonged, repeated handling of the whales was discouraged. Houser and Finneran (Houser and Finneran, 2006b) have shown that variability of hearing tests can be as high as 11 dB in a population of captive bottlenose dolphins. There are various sources for this error, including jawphone calibration, jawphone placement, electrode placement, electrical noise on that specific day, determination of threshold estimates and the brain's response to the stimulus on that specific day. In addition, AEP signals are usually less robust and prone to more variability at lower frequencies because we often lack the ability to generate sound levels very high above threshold. These factors may have contributed to the differences observed in hearing thresholds on the different dates.

Another explanation for the possible increase in threshold estimates at 5 and 10 kHz could be a disturbance in auditory processing related to a trematode (i.e. Nasitrema spp.) infestation in the brain. Histological examination of the brain of MML0803, who stranded with MML0802, revealed large aggregates of trematode ova and fragments in the cerebellum, as well as associated macrophages and white matter loss. Although the brain of MML0802 was not examined at the histological level, it is possible that the brain contained a trematode infestation and associated lesions, which could affect auditory processing. However, CT imaging of MML0802 on 10 July 2008 did not reveal any gross brain pathologies. Post-mortem MRI did reveal a large parenchymal hemorrhage in the right cerebral hemisphere, but this was shown to occur immediately before death.

The morphology of the pygmy killer whale ABR resembled that of the bottlenose and common dolphins (Ridgway et al., 1981). The pygmy killer whale ABR was composed of seven waves (Fig. 6A). Peak amplitudes of ABR waves matched the very high amplitudes that characterize cetaceans (Ridgway et al., 1981; Dolphin, 2000). These increased amplitudes are due to the large neural tracts and abundant neurons involved in processing auditory information (Ridgway et al., 1981; Dolphin, 2000). Despite the large brain in the pygmy killer whale (i.e. 918 cm³) and, hence, longer conduction pathways, the response latency of the ABR waves were short, a characteristic of cetaceans (Ridgway et al., 1981; Dolphin, 2000). The reduced latencies in cetacean ABRs are most likely due to high axonal conduction velocities that are typical of large-diameter axons found in the cetacean auditory pathway (Ridgway et al., 1981; Dolphin, 2000). Functions of ABR wave amplitudes and sound intensity in the pygmy killer whale (Fig. 6B) were typical of cetaceans (Dolphin, 2000). The latency/sound intensity functions of the ABR waves were flat in pygmy killer whales (Fig. 6C), another characteristic of cetaceans (Dolphin, 2000). The latency/sound intensity relationships for the ABR waves in the pygmy killer whale were –1.7 to –4.3 μs dB^–1, typical of cetaceans and much smaller than other mammals (Dolphin, 2000). In summary, these data provide additional support that temporal processing by cetaceans is quick and not sensitive to variation in the intensity of the acoustic signal (Dolphin, 2000).

The auditory brainstem response is a result of firing of neurons as excitation moves through the auditory pathway in the brain. Each ABR wave reflects the excitation of a mass of neurons in a specific region of the brain. In cetaceans, the neuroanatomical sources of these ABR waves are still a matter of debate (Dolphin, 2000). On the basis of the prevailing views (Allen and Starr, 1978; Ridgway et al., 1981) and on some speculation, we hypothesized that wave I arose from the auditory nerve, wave II arose from the cochlear nucleus, wave III arose from the superior olive, wave IV arose from the inferior colliculus and wave VI arose from the medial geniculate body.

Studies with humans have shown that electrode placement affects the amplitude of the ABR (King and Sininger, 1992). In bottlenose dolphins, Popov and Supin (Popov and Supin, 1990b) showed that the best ABR recording occurred when the electrode was positioned 6–9 cm caudal to the blowhole. Similarly, in the present study, ABR waves had the highest amplitudes when the recording electrode was positioned ∼9 cm posterior to the blowhole (i.e. the Center position) (Fig. 7; supplementary material Fig. S6). Interestingly, when we moved the Center electrode to positions Left 1D and Left 1F, ABR wave I increased in amplitude (Fig. 7). In cetaceans, it is speculated that the neuroanatomical source of ABR wave I originates in the auditory nerve (Ridgway et al., 1981). When we measured the distance of the Center, Left 1D and Left 1F electrodes from the location where the auditory nerve entered the brain, the distances were 15.47, 12.26 and 11.46 cm, respectively. Thus, it appears that the closer the electrode is to the neuroanatomical source, the higher the amplitude of the respective wave.

Using the CT imaging and electrode mapping data, we determined whether a negative relationship existed between the distances of electrode positions from auditory brain structures and the amplitudes of their hypothesized waves (supplementary material Fig. S7). These data provided some evidence of the anatomical source of each ABR wave. The data suggested that the origination of ABR waves I, IV and VI were the auditory nerve (i.e. P-value was close to statistical significance for regression line), inferior colliculus and medial geniculate body, respectively. These findings support previous speculation on the origination of ABR waves in bottlenose and common dolphins (Ridgway et al., 1981). However, more research is needed to determine the sources of ABR waves. As demonstrated in the present study, one very powerful approach is to combine structural imaging and AEP techniques. Future research could expand imaging modalities and AEP techniques to include MRI and an electrode montage.

AEPs are extremely powerful in determining hearing sensitivity in stranded, hospitalized and newly captured marine mammals. Of the at least 84 cetacean species, we now have measured audiograms for at least 13 of those species (Nachtigall et al., 2007). As illustrated in the present study, the combination of biomedical imaging and AEP techniques is a powerful tool to learn more about the hearing of live-stranded cetaceans, particularly those that are rare or endangered.

It is important to continue to measure the hearing of stranded cetaceans even if an audiogram is already known, primarily for three reasons. First, an audiogram is not a reliable method to estimate a species' hearing capacity from thresholds obtained from one individual. Second, if possible, we recommend hearing tests of stranded cetaceans in rehabilitation prior to release, especially when behavior of the animal indicates hearing loss. The release of a wild animal with severe hearing loss would be inhumane, as these aquatic mammals depend upon a functional auditory system to detect prey and predators, navigate and communicate. Third, and most importantly, many anthropogenic factors – such as acute and chronic noise, ototoxic drugs and chemical pollution – may play a role in hearing loss of cetaceans. These potentially interactive factors need to be investigated to determine their impact on the cetacean auditory system.

 
• 3-D

  three-dimensional

 
• ABR

  auditory brainstem response

 
• AEP

  auditory evoked potential

 
• CT

  computed tomography

 
• EFR

  envelope following response

 
• FFT

  fast Fourier transform

 
• HU

  Hounsfield units

 
• MRI

  magnetic resonance imaging

 
• MRTF

  modulation rate transfer function

 
• r.m.s.

  root mean square

 
• SAM

  sinusoidal amplitude modulated

 
• SPL

  sound pressure level

We thank all the staff and volunteers, especially Lynne Byrd and David Smith, at the Dolphin and Whale Hospital for their assistance during CT imaging and hearing tests; Axcess Diagnostics, Sarasota, FL, USA, for the use of their CT scanner at no cost; and Danielle Greenhow, Kelly Martin, Christin Murphy and Jessica Powell for technical assistance during hearing tests. We also recognize Jerome Barakos (MD) and Sophie Dennison (DVM) for interpretation of the magnetic resonance images of MML0802, and Katie Colegrove (DVM) for histological examination of MML0803. Earlier versions of this manuscript were improved with the assistance of Misty Montie and Peter Simard.