Beyond the limits: identifying the high-frequency detectors in the anuran ear

Abstract

Despite the predominance of low-frequency hearing in anuran amphibians, a few frog species have evolved high-frequency communication within certain environmental contexts. Huia cavitympanum is the most remarkable anuran with regard to upper frequency limits; it is the first frog species known to emit exclusively ultrasonic signals. Characteristics of the Distortion Product Otoacoustic Emissions from the amphibian papilla and the basilar papilla were analysed to gain insight into the structures responsible for high-frequency/ultrasound sensitivity. Our results confirm the matching of vocalization spectra and inner ear tuning in this species. Compared to most anurans, H. cavitympanum has a hyperextended hearing range spanning from audible to ultrasonic frequencies, far above the previously established ‘spectral limits’ for the amphibian ear. The exceptional high-frequency sensitivity in the inner ear of H. cavitympanum illustrates the remarkable plasticity of the auditory system and the extent to which evolution can modify a sensory system to adapt it to its environment.

1. Introduction

A fascinating example of convergent evolution being driven by ambient environmental noise emerged with the discovery of some frog species that communicate using very high signal frequencies [1–3]. Odorrana tormota, Huia cavitympanum and O. graminea are all southeast Asian species that inhabit riverine environments dominated by high levels of low-frequency noise (approx. 58–70 dB SPL, 0.05–4 kHz; [4]). Despite the prevalence of low-frequency vocalizations in anuran amphibians, usually below 5−8 kHz [5], these frog species have converged on the ability to emit calls of high frequencies that extend into the ultrasound (i.e. greater than 20 kHz) [2–4]. Huia cavitympanum is most remarkable with regard to upper frequency limits; it is the first frog species known to emit exclusively ultrasonic signals [2]. Behavioural, morphological, electrophysiological and vibrometric evidence all point toward high-frequency auditory sensitivity in H. cavitympanum spanning an exceptionally broad spectral range. An important piece of the puzzle is still missing: how does this extraordinary auditory system recognize very high-frequency signals of biological significance such as species-specific vocalizations?

One approach to identifying the inner ear end-organ(s) mediating high-frequency detection in H. cavitympanum is to determine the site(s) of origin of otoacoustic emissions in the inner ear. Distortion Product Otoacoustic Emissions (DPOAEs) are considered a by-product of nonlinearities in inner ear processing that appear when two simultaneous pure tones are presented to the ear [6]. DPOAE recording has proven to be a reliable tool for hearing assessment in many animal species; maximum DPOAE amplitude and lowest threshold generally correlate to the frequencies of high auditory sensitivity [7–9]. Anuran DP-grams typically show a bimodal dependence on the stimulus frequency, reflecting DPOAE generation in the two sound-sensitive end-organs, the amphibian papilla (AP) and the basilar papilla (BP) [10]. Across species, the AP is most sensitive to the low- and mid-frequencies (0.1−1.25 kHz; [11,12]) and the BP is most sensitive to the highest frequencies within the frog hearing range (up to 8.2 kHz; [13,14]). Therefore, it has been assumed that lower frequency DPOAEs originate from the AP, while higher frequency DPOAEs are generated in the BP [7,10].

In this study, we recorded DPOAEs to examine the specific contributions of the auditory end-organs to the spectral recognition of calls of H. cavitympanum. To gain insight into the structures responsible for ultrasound sensitivity in frogs, characteristics of the DPOAE responses from the AP and the BP are analysed.

2. Material and methods

(a). Subjects

Adult males of Huia cavitympanum (n = 5) were collected along the banks of the Nyipa River in Gunung Mulu National Park, Sarawak, Malaysia (4°3' N, 114°51' E) between 23 and 26 February 2019. Animals were anaesthetized with an intramuscular injection of a pentobarbital sodium solution (Nembutal, Ovation Pharmaceuticals, Inc., 50 mg ml⁻¹: approx. 1−1.2 µl g⁻¹ body mass) in one of the hind limbs. Anesthetized frogs were covered with wet gauze to facilitate cutaneous respiration.

(b). Instrumentation and signal analysis

A custom-built probe with an overall tip diameter of approximately 4 mm was placed about 1−1.5 mm from the frog's tympanic membrane in an open sound system configuration. Two multi-field speakers (MF1, Tucker-Davis Technologies) were coupled to the probe through 0.5 cm pieces of tygon tubing. To record the ear responses, the probe included a microphone (MK 301, Microtech Gefell GmbH) that was connected to an amplifier (MN920, Microtech Gefell GmbH) with a gain of 32 dB. Stimuli were generated and responses acquired using an RME Fireface UC audio interface (128 kHz, 24 bit). Stimulus generation, data collection and analyses were performed using Matlab R2016b (The MathWorks, Inc.).

DPOAEs were evoked with two simultaneous pure tones (primaries) of 4.16 s duration (1 ms rise/fall time) with an inter-stimulus pause of 500 ms. The frequencies and levels of the primaries are denoted as f₁, f₂, L₁ and L₂, respectively. An FFT on the time domain-averaged (n = 65) signal was performed to obtain the amplitude of the distortion products at the frequencies 2f₁-f₂ (termed DP2f₁-f₂) and 2f₂-f₁ (termed DP2f₂-f₁), which are typically the most prominent DPOAEs measurable in the frog inner ear [7,15]. Background noise level was calculated as the mean amplitude of 5 FFT bins on either side of and 25 Hz from the 2f₁-f₂ and 2f₂-f₁ frequencies. To assess system-generated distortions, experiments were repeated with the probe against a hard, inanimate surface.

(c). Distortion product otoacoustic emission recordings

DP-grams (DPOAE amplitude as a function of stimulus frequency) were recorded for f₂ frequencies from 2 to 42 kHz (in 1 kHz steps) and stimulus levels equal to 84 dB SPL. No distortion products were detected in exploratory recordings using stimulus frequencies between 0.4−2 kHz. A fixed f₂/f₁ ratio of 1.1 and equal primary tone levels (L₁ = L₂) were used since these stimulus parameters are optimal to evoke large DPOAE levels in frogs [10,16]. To qualify as a valid DPOAE recording, the signal-to-noise ratio (SNR) at each frequency had to be ≥6 dB. To evaluate the DPOAE threshold curve, a matrix of frequency-level tone combinations was presented pseudo-randomly with f₂ values from 2−42 kHz (in 1 kHz steps), and primary tone level values from 50 to 84 dB SPL (in 2 dB steps). The DPOAE threshold curve was defined by the interpolated stimulus levels necessary to elicit a DP2f₁-f₂ amplitude equal to 0 dB SPL for all frequencies tested.

DPOAE growth functions (DPOAE amplitude as a function of the stimulus level) were recorded by increasing stimulus levels from 50 to 84 dB SPL (in 2 dB steps) for the stimulus frequencies that elicited local maxima of 2f₁-f₂ emissions in the DP-grams. The slopes of the growth functions, the detection thresholds (stimulus levels sufficient to evoke DP2f₁-f₂ amplitudes greater than 6 dB above the background noise) and the saturation thresholds (second point along the DPOAE amplitude plateau going from low to high levels) were assessed.

3. Results

DP-grams revealed an inner ear spectral sensitivity that extends from 6.7 to 32.7 kHz and exhibited three relative amplitude maxima in the low-, mid- and high-frequency regions (LF-DPs, MF-DPs, HF-DPs) (figure 1). Characteristics of DPOAEs for each response region are summarized in table 1.

Figure 1.

Hyperextended hearing range of Huia cavitympanum. (a) Adult male of H. cavitympanum. Inset: Three spectra of the microphone signal showing DPOAE spectral peaks (indicated by arrows). Cubic distortion products are labelled by their frequency: 2f₁-f₂ and 2f₂-f₁. (b) DPOAE level as a function of the stimulus frequency (f₁ for DP2f₁-f₂, f₂ for DP2f₂-f₁) reveals sensitivity peaks at low-, mid- and high-frequency regions (LF-DPs, MF-DPs, HF-DPs). Light lines represent individual DP-grams and bold lines represent median DP-grams. L₁ = L₂ = 84 dB SPL; f₁/f₂ = 1.1.

Table 1.

Characteristics of DPOAEs in H. cavitympanum. The stimulus frequency corresponding to maximum DPOAE level (best frequency, BF), and the low- and high-side cutoff frequencies for which the DPOAE level was 6 dB above the noise floor, were evaluated in the DP-grams for the low-, mid- and high-frequency ranges. The stimulus frequency that requires the lowest stimulus level for which DP2f₁-f₂ amplitude ≥0 dB SPL (characteristic frequency, CF) was evaluated in the individual DPOAE matrices. Median and interquartile range are specified. Measurements of DP2f₁-f₂ and DP2f₂-f₁ are expressed as a function of f₁ and f₂ (kHz), respectively_.

	DP2f1-f2				DP2f2-f1
	BF	low cutoff	high cutoff	CF	BF	low cutoff	high cutoff
LF-DPs (n = 5)	9.09	6.36	11.81	7.27	7	6	11
	7.27–10.91	5.45–7.27	10.91–15.45	6.36–8.19	7–9	6–7	10–12
MF-DPs (n = 5)	18.19	13.64	21.81	18.19	18	13	23
	13.64–20.00	11.81–18.19	16.36–23.64	13.64–18.19	17–20	12–16	21–23
HF-DPs (n = 4)	28.64	26.36	30	27.73		(n = 2)
	28.19–29.09	26.36–27.27	30–32.73	27.26–29.09	29, 30	26, 27	30

Growth functions were classified as monotonic or nonmonotonic (figure 2a). Generally, growth functions from either LF-DPs or MF-DPs showed nonmonotonic growth patterns. They exhibited a compressive nonlinearity with an extended plateau in which the 2f₁-f₂ amplitude levelled off and often decreased. Growth functions for HF-DPs showed monotonic patterns and were fit with straight lines using linear regression. A mean r² of 0.97 indicates that the amplitude growth for HF-DPs was well-characterized by a linear fit.

Figure 2.

Comparison of DPOAE growth functions. (a) DPOAE growth functions measured at the best frequencies of LF-DPs, MF-DPs and HF-DPs. (b) Histograms of slopes of DPOAE growth registered with low- and intermediate-stimulus levels in LF-DPs and MF-DPs. No significant differences were found in slopes' distribution between frequency ranges. (c) Matrix of DPOAE response of one male of H. cavitympanum. This heat map was constructed with the DP2f₁-f₂ amplitude for each frequency-level combination. DPOAE threshold curve (DP2f₁-f₂ amplitude = 0 dB SPL) depicted as a black line. Data extracted for further analysis of growth functions are delimited by white rectangles. (d) Comparison of detection thresholds (bottom) and saturation thresholds (top) corresponding to the growth functions extracted from the individual matrices of DPOAE response. The plots depict the median values (horizontal line), the first and third quartiles (box) and the 10th and 90th percentiles (whiskers); **p < 0.01.

Slopes of the LF-DPs and MF-DPs growth functions were calculated for each data point of the curves by fitting a straight line through its two neighbours. For each growth function, the slopes were divided into two segments: (i) slopes below the stimulus level at which the amplitude plateau changed towards steeper increasing slope values (low and intermediate-stimulus level segment) and (ii) those above it (high-stimulus level segment). The average slope registered with low- and intermediate-stimulus levels was 0.34 dB dB⁻¹ for the LF-DPs and 0.67 dB dB⁻¹ for the MF-DPs. A two-sample Kolmogorov–Smirnov test was used to compare slopes' distributions, and no significant differences were found (k = 0.2136 p = 0.1884; figure 2b). At the high-stimulus level segment, growth functions have a much steeper slope: 4.37 and 4.02 dB dB⁻¹ for LF-DPs and MF-DPs, respectively.

The comparative analysis of DPOAE growth functions also included data extracted from the individual DPOAE response matrices (figure 2c,d). The selected analysis frequencies comprised the characteristic frequency (frequency of highest sensitivity; CF) and the frequencies at +/– 1 kHz and +/– 2 kHz around the CF. MF-DPs showed a trend toward slightly elevated detection thresholds that were significantly different from the detection thresholds of LF-DPs (U = 143 p < 0.01). There were no statistically significant differences in saturation threshold between frequency ranges (U = 61 p = 0.23).

4. Discussion

The general spectral range for which DPOAEs were detected in the present study (6.7−32.7 kHz) is largely consistent with the frequency selectivity previously reported for the middle ear and auditory midbrain of H. cavitympanum [17]. Compared to the ears of most anurans, the inner ear of this species operates not only at exceptionally high frequencies, but also over an unusually large frequency range. At the same time, it resembles the ears of most frogs that are tuned to signals of biological significance [13]. DP-grams exhibited three relative amplitude maxima in the low-, mid- and high-frequency regions suggesting that the peripheral auditory system has a frequency response that matches the species-specific vocalization spectra. Two relative maxima in DPOAE amplitude were identified for stimulus frequencies around 9 and 18 kHz, slightly below the two clusters of dominant frequencies (11.7 and 20.2 kHz) previously found in Huia calls [2]. A third relative maximum in DPOAE amplitude (approx. 28 kHz) corroborates that this inner ear is also sensitive to ultrasonic frequencies. Such minor mismatches do not imply a diminished capability in species-specific vocal repertoire recognition. In H. cavitympanum, call dominant frequencies show large intra-individual variability (greater than 15 kHz) and downward frequency modulation over a wide frequency range is a common call motif [2,17].

All DPOAEs recorded in H. cavitympanum appeared at frequencies above the previously accepted ‘spectral limit' for the AP (1.25 kHz; [18]). Similarly, more than 85% of recorded DPOAEs were at frequencies above those found to be detected by the frog's BP (8.2 kHz; [5]). The frequency range of auditory sensitivity in this species deviates considerably from those of other anurans, thus precluding direct comparison to previously studied frog species for the identification of the end-organs generating DPOAEs. Features other than spectral sensitivity are useful for distinguishing DPOAE's origin. DPOAE growth functions from the BP grow linearly at a rate greater than 1 dB dB⁻¹ for low- and high-stimulus levels, whereas DPOAEs from the AP usually grow gradually for low- and mid-level stimuli (less than 1 dB dB⁻¹) but steeply (greater than 1 dB dB⁻¹) for high-level stimuli [15]. Higher stimulus levels are required to evoke measurable DPOAEs from the BP [19,20]. DPOAE generation in the AP is thought to involve an active amplification mechanism [16,21]. Otherwise, DPOAEs from the BP originate from passive nonlinearities [16].

In the present study, the growth functions corresponding to HF-DPs are monotonically increasing with mean slopes of 1.52 dB dB⁻¹, a signature of passive DPOAE generation. Also, DPOAE response at the highest frequencies suggests lower sensitivity. Taken together, these data support our hypothesis that HF-DPs originate from the BP, the end-organ consistently associated with high-frequency detection across frog species. The morphological characteristics found in the BP of H. cavitympanum are consistent with this high-frequency specialization [22].

By contrast, growth functions corresponding to LF-DPs and MF-DPs exhibited a compressive nonlinearity with an extended plateau typical of DPOAE growth functions from the frog's AP. Mean growth function slopes were shallow for low- and intermediate-stimulus levels and steeper for high-stimulus levels. Distributions of slopes and saturation thresholds were also similar between LF-DPs and MF-DPs. Lower frequency sensitivity peaks in the anuran DP-grams are presumed to originate from the AP [7,10]. The parallels between LF-DPs and MF-DPs suggest that these DPOAEs originate from the two functionally distinct regions of the papilla [11]: the low-frequency rostral region generates LF-DPs and the mid-frequency caudal region generates MF-DPs. Thus, based on the comparison of DPOAE data, we suggest that in H. cavitympanum (i) the AP is exceptional in its sensitivity to a broad range of frequencies exhibiting two DPOAE amplitude peaks and (ii) the BP is sensitive to ultrasound, corresponding to the highest spectral peaks in the call. Our results do not preclude the possibility of interference between DPOAEs from both papillae and from distinct regions of the AP, thus resulting in amplitude notches that predisposed our estimation of ranges of frequency sensitivity. Further neurophysiological research could reveal the tuning mechanism determining this shift in the inner ear spectral sensitivity. Huia cavitympanum is not only an illustrative example of adaptive signal and sensory system coevolution, but also provides clear evidence that there is no inherent high-frequency limit to anuran hearing.

Ethics

The experimental protocol adhered to the ABS guidelines for animal use in research and was approved by the UCLA Animal Research Committee (protocol no. 1994-086-81).

Data accessibility

Data are available from the Dryad Digital Repository: https://doi.org/10.5068/D1X95F [23].

Authors' contributions

P.M.N. conceived the project. T.U.G. collected the animals and facilitated every aspect of this study. A.C.-C. carried out the experiments, analysed the data and drafted the manuscript with support from P.M.N. and T.U.G. All authors provided edits and approved the final manuscript, and agree to be held accountable for the work therein.

Competing interests

We declare we have no competing interests.

Funding

This study was supported by the National Science Foundation (grant no. 1555734) and by Dan Hurley funds to P.M.N.