ABSTRACT
Anthropogenic noise is becoming a major underwater pollutant because of rapidly increasing boat traffic worldwide. But its impact on aquatic organisms remains largely unknown. Previous studies have focused mainly on high-frequency and impulsive noises (i.e. sonar); however, boat noise is more pervasive, continuous, and its highest intensity and component frequencies overlap the auditory bandwidth of most fishes. We assessed the impacts of boat noise on saccular sensory hair cell density and hearing thresholds of a soniferous species, Atlantic croaker (Micropogonias undulatus). In two laboratory experiments, individuals were subjected to simulated boat noise: a single 15-min exposure and 3 days of intermittent noise (simulating passing vessels). Immediately after both experiments, fish were either (1) tested for hearing sensitivity with auditory evoked potential (AEP) tests or (2) euthanized for fluorescent phalloidin and TUNEL labeling for hair cell density counts. Relative to controls, no differences were observed in auditory thresholds nor hair cell density between individuals subjected to a single 15-min noise exposure. However, fish from the 3-day experiment showed decreased sensory hair cell density, increased apoptotic cells, and higher hearing thresholds than control fish at 300, 800 and 1000 Hz. Our results demonstrate that impacts from boat noise depend upon the duration and frequency of exposure. For a species reliant on vocalization for communication, these impacts may hinder spawning success, increase predation risks and significantly alter the ecosystem.
INTRODUCTION
Anthropogenic noise has been increasing and altering the acoustic characteristics of the underwater soundscape (Kight and Swaddle, 2011; Celi et al., 2016). These changes may have adverse effects on animals, especially those whose hearing range is within the frequency bandwidth of noise (Maruska and Mensinger, 2009; Tricas and Boyle, 2014; Putland et al., 2019). Anthropogenic noise can impact an individual physically, physiologically and behaviorally by causing temporary or permanent damage or potentially causing indirect or direct death (Kunc et al., 2016; De Jong et al., 2018). The most dominant noise source within an aquatic soundscape is boat noise, which has doubled between 1965 and 2003 and is expected to triple by 2025 (Hawkins et al., 2015; Putland et al., 2019).
Previous studies on anthropogenic noise in the aquatic environment have primarily focused on intermittent high-intensity noises, which predominately affect marine mammals and protected species (i.e. cetaceans and pinnipeds; Williams et al., 2015; Kunc et al., 2016). More recently, studies have started to investigate the impacts of more pervasive and continuous boat noises, the component frequencies and highest intensity of which overlaps the auditory bandwidth of most fishes (Jepson et al., 2003; Garrett et al., 2016; Southall et al., 2017). Boat noise can extend greater than 10 kHz, but is most intense in lower frequencies from 50 to 600 Hz (Southall et al., 2017). Low-level, chronic noise can lead to behavioral changes owing to decreased signal-to-noise ratios (Stanley et al., 2020). These changes can mask acoustic stimuli and signals that interfere with feeding, predator avoidance, group cohesion, settlement behavior and spawning success (Stanley et al., 2020). Research aimed at taxa in lower trophic levels, particularly those with large biomass, such as some fishes [e.g. Atlantic croaker (Micropogonias undulatus), menhaden (Brevoortia spp.), anchovies (Engraulidae)], is critically needed as impacts to species at these lower trophic levels can show a cascading effect and potentially disrupt entire ecosystems (Hawkins and Popper, 2017; Stanley et al., 2020).
Numerous studies have examined the impacts of noise pollution on fish hearing and behavior (Smith et al., 2006; Graham and Cooke, 2008; Casper et al., 2013; Mickle and Higgs, 2017; Vetter et al., 2018; Breitzler et al., 2020). There is variation among fish species in the ability to detect sound pressure stimuli in addition to sound particle motion, the primary stimulus of fish hearing (Popper and Lu, 2000), and sound pressure detection is more prevalent in species with accessory auditory structures and linkages between the swim bladder and inner ear (Hawkins, 1986; Ladich and Schulz-Mirbach, 2016). However, with such a diverse group of fishes (>26,000 spp.; Nelson et al., 2016) that live in a variety of habitats, and possess several different inner ear-swim bladder morphologies, it is imperative to studies species from different representative taxa, particularly those that produce sound for communication during reproduction (i.e. sciaenid species; Ramcharitar et al., 2001).
Fish inner ears contain three paired calcium carbonate otoliths: the saggitae, asteriscae and lapillae. These mass-loaded organs have greater inertia than a fish's body, allowing them to mechanically transduce the particle motion of a sound wave in water (Popper and Lu, 2000; Ramcharitar et al., 2006; Ladich and Schulz-Mirbach, 2016). All three otoliths have corresponding maculae that contain sensory hair cells with stereocilia bundles on the apical surface. These hair cells are the auditory and vestibular transduction sites, forming synaptic connections with ganglion cells in the eighth cranial nerve (Edds-Walton and Popper, 2000). Furthermore, these bundles have been described to have orientation patterns that are aligned in directions that allow for maximum excitation; adjacent areas typically have opposing (180 deg) orientations relative to each other (Edds-Walton and Popper, 1995; Ramcharitar et al., 2001). Exposure to intense noise stimuli or ototoxic chemicals has been shown to damage sensory hair cells, resulting in hearing loss in mammals (Hamernik and Qiu, 2000; Chen and Fechter, 2003), but few studies have observed hair cell changes in response to noise pollution in fishes (e.g. Smith et al., 2006, 2011), thus the need for increased studies on a variety of fish taxa, including Sciaenidae.
Sciaenids produce vocalizations that arise from the contraction of sonic muscles on their swim bladders (Ladich and Fine, 2006; Lechner and Ladich, 2008). As sound producers, Atlantic croakers have known sounds that are associated with courtship or spawning, disturbance and distress calls (Ramcharitar et al., 2006; Gannon, 2007; Montie et al., 2017). Salient signals of sound producers need to be heard by conspecifics and others for successful communication and, ultimately, survival. Previous studies have shown that Atlantic croaker from the east coast of North America have a hearing range of 100–1200 Hz (Ramcharitar and Popper, 2004; Horodysky et al., 2008), which is well within the range of boat noise (Jepson et al., 2003; Garrett et al., 2016; Southall et al., 2017). Because vessel noise is increasing globally and presents a potential threat to fishes that utilize sound production for communication, we aimed to test noise impacts on hearing in Atlantic croaker.
We performed two laboratory experiments. The first experiment assessed whether a short-term single event exposure affected hearing sensitivity and caused epithelia damage. The second experiment assessed whether a prolonged (3-day) repeated event exposure that is more likely to occur in the natural environment caused potential damage to hearing sensitivity and epithelia damage.
MATERIALS AND METHODS
Fish acquisition and maintenance
All animal collections, maintenance and experimental procedures were approved by the University of New Orleans Institutional Animal Care and Use Committee (protocols 20-001 and 20-002). Atlantic croaker [Micropogonias undulatus (Linnaeus 1766)] were collected via hook and line along the south shore of Lake Pontchartrain, LA, USA, an area that experiences minimal small watercraft presence, and estuarine habitats along Grand Isle, LA, an area inaccessible by motorized boats. Though we avoided potential collection sites we suspected would have high vessel traffic, we do not know the past locations and level of vessel noise exposure for individual fish. Salinity levels (18±1 ppt) and temperature levels (22±2°C) were kept constant for holding tanks, experimental tanks and auditory evoked potential (AEP) setup to eliminate any potential effects on hearing sensitivity. Individuals collected from Lake Pontchartrain were acquired from 0.5–3.0 ppt waters. These individuals were initially maintained in a filtered (Marineland Magniflow canister filter) 8±1 ppt, 568-liter oval stock tank for 2 days prior to being placed in filtered 18±1 ppt, 568-liter oval stock tanks to allow for acclimation to experimental salinity. Individuals collected from Grand Isle were acquired from 13–25 ppt waters and were maintained in filtered 18±1 ppt, 568-liter oval stock tanks prior to boat noise exposure experiments. Fish experienced natural light cycles, were always held in quiet conditions (filter outflow submerged in water; no aerator stones) relative to ambient levels measured in situ (Fig. S1), and were kept in the holding tanks for a minimum of 5 days prior to being used in a trial. Fish were fed three days a week until satiated with shrimp or squid. Overall, 63 fish collected during October to December 2020 and March to September 2021 were used for this study (Table 1).
Boat noise
Tank setup
Noise exposure experiments were conducted in two tanks [3-sided stone, 1-sided glass, 318-liter (1.14×0.52×0.53 m, length×width×water level)]. Tanks were randomly selected as exposure (Fig. 1A) or control (Fig. 1B) tank prior to start for both the 15-min and 3-day exposure experimental trials to eliminate any bias between tank setups. An underwater speaker (UW-30; Lubell Labs Inc., Columbus, OH, USA) was suspended by PVC at 0.3 m below the water surface (center of speaker), centrally along the side of the tank, behind a barricade. A barricade was present to ensure fish were positioned directly within the path of the noise stimulus, preventing them from seeking refuge behind the sound source. In the noise exposure tank, the UW-30 speaker was connected to an amplifier (Bogen CC4021, Bogen Communications Inc., Memphis, TN, USA) with input fed from a Tascam DR-05X recorder for boat noise playback. The control tank (Fig. 1B) had the same setup apart from having a disconnected UW-30.
Field recordings and simulated vessel noise
Recordings were opportunistically made once every month during 2018 near the channel entrance for the Dauphin Island Ferry in Dauphin Island, AL, USA. A calibrated hydrophone (High Tech, Inc., Gulfport, MS, USA; sensitivity −163.9 dB re. V µPa−1; frequency response 2 Hz–30 kHz) was suspended from a dock (∼1 m below the surface) with occasional concurrent visual observations to determine boat types associated with recorded noise. All recordings were made using a Tascam recorder (DR-05, linear PCM) at a sampling rate of 48 kHz. Number of boats, time between boats, and length of boat passing were quantified for each sampling month (Table S1).
Sound files (.wav format) were constructed from vessel recordings for playback: one 15-min file for the first experiment and eight files (one ferry, one offshore crew boat and six unidentified boats) for the 3-day experiment. Sound files were edited in Adobe Audition v. 3.0 to reduce the bias and non-flat response of the underwater speaker to improve the similarity of the simulated boat noise to the original field recording (Fig. 2C).
The 15-min playback sound was amplified to be louder than the original field recording (Fig. 2A,C) during the experiment to approximate a higher boat noise amplitude, as boat noise can reach up to 188 dB (McKenna et al., 2013). Our boat playback is more variable over the entirety of sound but still reflects the sound level of the original field recording. The sound pressure for our playback was highest around 500 to 800 and 2000 Hz (120–130 dB re. 1 µPa; Fig. 2C). For the 3-day boat noise exposure, fish were exposed to a 24-h track loop. During the loop, the ferry recording was incorporated following the current Mobile Bay Ferry summer schedule at the time of the experiment (Table S1). The remaining boat recordings were randomly incorporated in the loop (Table S1) based on the average number of boats passing per hour that were previously quantified from the monthly recordings.
Particle acceleration
Particle acceleration level (PAL) and sound pressure are related but not directly proportional. PAL decays more rapidly than sound pressure and falls off at a higher rate as frequency increases (Nedelec et al., 2016). We estimated PAL based on the assumption of a monopole sound source. Laboratory measurements occurred within one wavelength for all frequencies (0–5 kHz) in which PAL was estimated and were thus assumed to be in the acoustic near field at a distance of 0.275 m from the source for AEP measurements and 0.470 m for noise exposure in the aquarium. Field vessel noise measurements were made at several distances to the source and incorporated estimates of PAL from sound sources within and greater than one wavelength.
(5) Each value calculated for each FFT frequency was calculated by averaging over all 10 s PAL magnitude measurements. Near-field frequency range cut-offs were determined by taking the sound speed (m s−1) divided by source distance (m): speaker to the center of the experimental tank=0.470 m, and fish distance to the speaker during AEP tests=0.275 m. To show variation in PAL over time, possible distance scenarios based on in situ calculated distances 30, 85, 145 and 205 m were used (Fig. 2B).
Field versus tank recordings
In 2021, we measured simultaneous sound pressure level (SPL) and PAL at the same Dauphin Island location where recordings were made with the GeoSpectrum M20-40. These recordings allowed us to estimate the relative contribution of SPL and PAL, as PAL attenuates more rapidly with distance.
Noise exposure experiments
Fish were euthanized with an overdose of 2-phenoxyethanol (0.006 mol l−1, Sigma-Aldrich, St Louis, MO, USA) after noise exposure or control trials, or after AEP tests for fish tested for hearing thresholds. Body size (standard length) was determined for each fish to the nearest mm and fish were dissected and sex determined by examination of gonads under a stereomicroscope (Leica M165 FC).
Fifteen-minute exposure
Fish were randomly assigned to either a control or noise exposure treatment and placed individually in each of their respective treatment tanks (Fig. 1A,B; modified for N=1 per tank) 18 h prior to noise exposure. The noise exposed fish was subjected to a 15-min single boat event noise playback (Fig. 2A,C) while the control fish was held simultaneously in quiet conditions. Immediately after a trial, one fish was randomly selected to be immediately euthanized to quantify hair cell density on the left saccule (Fig. 1C) while hearing thresholds from auditory evoked potentials (AEPs) were tested on the fish from the other treatment (Fig. 1D). Overall, there were 12 noise exposure trials for this experiment (Table 1).
Three-day repeated exposures
Fish were randomly assigned to a control tank (N=2 per trial) or noise exposed treatment tank (N=2 per trial) 18 h prior to noise exposure (Fig. 1A,B). The noise exposed fish was subjected to a 3-day repeated boat noise exposure while the control was simultaneously held in quiet conditions. Immediately after each trial, one fish from each treatment (one fish from the control treatment and one fish from the noise exposed treatment) was immediately euthanized to quantify hair cell density on the left saccule (Fig. 1C) and to quantify apoptotic cells on the right saccule (Fig. 1C). The remaining two individuals (one fish from the control treatment and one fish from the noise exposed treatment) received AEPs to assess hearing thresholds. The order of fish tested first was randomly selected and the second AEP was conducted at the conclusion of the first AEP (Fig. 1D).
Histology
After euthanasia, heads with cracked otic capsules were fixed in 4% methanol-free paraformaldehyde for 3 to 5 days before excision of saccular maculae for phalloidin and Terminal dUTP Nick-End-Labeling (TUNEL). Cell counts were conservative; images with artifacts (overexposure, tears, or regions covered by excess tissue that obscured hair cells) were re-imaged if possible or not included in cell counts.
Hair cell density assessed with phalloidin labeling
To visualize sensory hair cells of the left saccular macula, filamentous actin (F-actin) was labeled with Alexa Fluor 488 phalloidin (Invitrogen, Molecular Probes Inc., Eugene, OR, USA) according to the manufacturer’s protocol. Saccules were excised and rinsed in pre-warmed phosphate buffered saline (PBS) three times for 2 min each. Isolated saccules were placed in 4% methanol-free formaldehyde for 15 min at room temperature. A volume of 0.5 µl of 400× phalloidin stock solution was diluted in 200 µl of PBS and applied to saccules on glass slides and allowed to incubate in a humidified chamber for 60 min. After incubation, saccules were rinsed in PBS two times for 5 min each before being mounted on glass slides and cover slipped with CytoSeal 60 (Electron Microscopy Sciences, Hatfield, PA, USA). Sensory hair cells were viewed under a 40× objective using a green fluorescent protein (GFP; excitation 450–490 nm, emission 500–550 nm) filter of a Leica DM 2500 LED microscope and imaged using a Leica MC170 HD camera. Six 200×200 µm designated regions of the saccule were imaged (Fig. 3A, pink squares). Regions selected (H1, H2, H3 and H4 located in the ‘head’ region, and T1 and T2 in the ‘tail’ region, respectively; Fig. 3) were based on previously described ciliary bundle orientation patterns (Ramcharitar et al., 2001). This histology protocol was completed on the left saccular macula for both 15-min and 3-day exposure experiments. Region images were counted manually using the ‘cell counter’ plugin in Fiji-ImageJ version 2.3.1 software.
Apoptotic cells assessed with TUNEL labeling
Right saccular maculae of 3-day experiment individuals were processed for TUNEL according to manufacturer protocols (Uribe et al., 2013; ApopTag Fluorescein In Situ Apoptosis Detection kit S7110; CHEMICON/Millipore, Billerica, MA, USA). Right saccular maculae were excised and permeabilized with proteinase K (10 µg ml−1) for 25 min at room temperature. Maculae were then refixed in 4% methanol-free paraformaldehyde for 20 min and then rinsed in PBS three times for 5 min each. Saccules were properly oriented on glass slides, counterstained with DAPI (MBD0020, Sigma-Aldrich) and mounted under glass coverslips with CytoSeal. TUNEL-labeled cells were viewed and imaged under a 40× objective first using a GFP filter and second using a DAPI filter (DAPI; excitation 325–375 nm, emission 435–485 nm). Two images were taken per region: one under GFP and one under DAPI. These images were then merged and composited in Fiji-ImageJ and apoptotic cells were manually counted. Selected saccular regions (200×200 µm) are the same as those described above (Fig. 3B).
Auditory evoked potentials
The auditory responses were determined from AEP tests (Kenyon et al., 1998). Experimental fish were first deeply anesthetized with a 2-phenoxyethanol (0.004 mol l−1) saltwater solution, before being lightly placed in a mesh harness suspended from a wooden frame, positioned 10 cm below the water surface of a 40 l steel chamber coated with marine paint (0.3×0.52 m, diameter×height), and ventilated with a 0.002 mol l−1 2-phenoxyethanol seawater solution to keep the fish anesthetized. The chamber sat on top of a vibration isolation air table, surrounded by a faraday cage and acoustic foam baffling (Fig. 1D). Stainless steel electrodes (Rochester Electro-Medical, Inc., Tampa, FL, USA; 13 mm needle) were sealed on the ends with glue and nail polish so that ∼1 mm of metal was exposed at the tip. The recording electrode (pink line in Fig. 1D) was inserted subcutaneously above the medulla along the dorsal midline. The reference electrode (blue line in Fig. 1D) was inserted subcutaneously laterally near the trunk of the tail.
Sound stimuli were generated and AEPs were recorded with a Cambridge Electronics Design (CED, Cambridge, UK) Micro 1401 controlled by Spike 2 version 10.19 software and a CED 3505 attenuator. Signals were amplified through a Bogen CC4021 and sent to a UW-30 speaker placed at the bottom of the chamber and approximately 40 cm below the fish. For the 15-min experiment, a total of 12 stimulus frequencies from 100 to 1200 Hz were tested for each subject. Frequencies were selected based on prior studies examining Atlantic croaker hearing thresholds (Ramcharitar and Popper, 2004; Horodysky et al., 2008). For the 3-day experiment, a total of eight stimulus frequencies from 100 to 1800 Hz were tested for each subject. Higher frequencies were tested in this experiment as they were opportunistically tested and resulted in an AEP response. Acoustic pips at test frequencies ≥200 Hz consisted of 2000 pure tone 40 ms pulses (10 ms plateau with rise and fall times of 5 ms). Test frequencies at 100 Hz pulse consisted of 100 ms pulse, 20 ms plateau, and 10 ms rise and fall times. Stimulus artifacts in AEP recordings were minimized by sequential alternation of the pip phase. Each trial began at a suprathreshold intensity and was decreased in 5 dB steps. Threshold was determined for each frequency before moving on to the next test frequency. Sound pressure levels (SPLs) of the tone pip stimuli were calibrated with a Reson hydrophone (Teledyne Marine Model EC6081 mk2 VP2000 preamplifier in conjunction with TC4013 hydrophone; sensitivity −211.3 dB re. 1 V Pa−1; frequency response 1 Hz–170 kHz; Seafloor Systems Inc., Shingle Springs, CA, USA) placed in the chamber at the position of the fish head. For calibration, pips were presented without alternation of phases and voltage levels of sounds at all frequencies and intensities were measured. The signals were averaged in Spike 2 to determine actual sound pressure levels in dB rms re. 1 µPa. Furthermore, the chamber was calibrated for PAL with a GeoSpectrum M20-40. For each frequency, PAL measurements were made for each SPL throughout the attenuation range. PAL measurements were determined for each directional axis and magnitude of PAL was used for assessing thresholds. AEP waveforms were differentially amplified (10,000×) and band-pass filtered (1–10,000 Hz) with a differential amplifier (AM Systems Model 1700, AM Systems Inc., Carlsborg, WA, USA), digitized on a CED Micro 1401 running Spike 2 software. Responses to 2000 tone-pip stimuli were averaged for each frequency and intensity and then (1) visually assessed for a visible response following stimulus onset (Fig. 4A) and (2) analyzed using power spectra (FFT, 1024 or 2048 bin) to identify peaks at double the stimulus frequency (Fig. 4B) that results from the stimulation of sensory hair cells oriented in opposing directions. Threshold was considered the lowest sound level with (1) an averaged evoked potential detectable above background noise, (2) obtained repeatedly, and (3) and with a peak at double the stimulus frequency on the power spectrum. To test for recording artifacts in our setup, a freshly euthanized fish was tested in the same manner.
Data analysis
Two-way repeated-measures ANOVA tested by generalized linear mixed-effect models (GLMMs) were conducted to test for differences among auditory thresholds, hair cell density changes and apoptosis occurrence between fish from control treatments and noise-exposed treatments. Two GLMMs were performed for the 15-min noise exposure (AEP and phalloidin hair cell density), and three GLMMs for the 3-day noise exposure (AEP, phalloidin hair cell density and apoptosis). Each dataset was visually and statistically checked for normality. Where normality was slightly violated, models were kept as Gaussian as there is evidence that mixed-effect models are robust even if the residual distributional assumptions are objectively violated (e.g. Schielzeth et al., 2020; Knief and Forstmeier, 2021). For AEP statistical analyses, because multiple frequencies were tested within each individual, subject was a random factor, frequency and treatment (control or noise exposed) were fixed factors, and auditory threshold response was the dependent variable. For histology statistical analyses, because multiple regions on the saccule within an individual were counted, individuals were used as a random subject factor, region and treatment as fixed factors, with counts as dependent variable. To determine the minimum adequate model, Akaike's information criterion with correction for small sample sizes (AICc) was used and residual versus fitted value plots were checked. All final minimum adequate models included the interaction terms (i.e. treatment×frequency and treatment×region) and were fit by restricted (or residual) maximum likelihood (REML).
We adjusted for potential inflation of type I error that could arise from examination of hair cell density and apoptosis that were conducted in ears from the same individual fish from the 3-day noise exposure experiment. We used a sequential Bonferroni (Rice, 1989) procedure to adjust type I error for the two GLMMs (hair cell density, apoptosis) and subsequent post hoc tests conducted on data from the same set of fish. No other type I error adjustments were made because we considered results from the 15-min and 3-day noise exposure experiments, which were conducted separately to be independent. Further, because hearing thresholds and examination of saccular epithelia were conducted on different fish, we did not make any adjustments for type I error.
When differences were observed among treatments or treatment×factor interactions (α≤0.05) in GLMMs, we conducted custom a priori Bonferroni post hoc comparisons. These custom contrasts allowed us to test for threshold differences at specific stimulus frequencies (i.e. threshold of 100 Hz control versus 100 Hz noise exposed) and differences at specific anatomical regions (i.e. H1 control versus H1 noise exposed) on the saccular macula and to maintain statistical power by avoiding nonsensical contrasts (e.g. threshold of 100 Hz control versus 1000 Hz noise exposed. In one case (see ‘Results, Hair cell apoptosis following 3-day repeated vessel noise exposure’), support for the model including an interaction term was much stronger as indicated by AICc (Burnham and Anderson, 2002) despite a P-value >0.05 for this term. In this case, we explored potential differences at each region with custom a priori Bonferroni post hoc comparisons. All analyses were computed using RStudio (http://www.rstudio.com/) version 4.0.3. with ‘nlme’ (https://CRAN.R-project.org/package=nlme) and ‘emmeans’ (https://CRAN.R-project.org/package=emmeans) packages.
RESULTS
Experiment 1: 15-min single boat noise exposure
Hair cell bundle density following 15-min vessel noise exposure
In tests for differences among regions between experimental treatment fish and controls, the minimum adequate model (Table 2) with interactions (saturated) was the best fit (AICc 771) compared with the models without interaction and simplified pooled models (AICc 819, 827 and 855, respectively). The GLMM analysis revealed that there was a significant interaction between treatment and region (P=0.032; Table 2) on hair cell density counts. To test for differences in saccular hair cell density, a priori post hoc pairwise comparisons for specific regions between treatments (i.e. H1 control versus H1 noise exposed) were conducted and found no differences between treatments for any region (P>0.05; Table S2; Fig. 5A). Pairwise comparisons indicated that differences among average hair cell densities observed in five of six regions (H1, H2, H3, H4 and T1) among control and noise-exposed fish were not greater than expected by chance (P>0.05; Table S2).
Hearing sensitivity in Atlantic croaker following 15-min vessel noise exposure
Responses to stimuli were obtained at all tested frequencies in both treatments. Atlantic croaker had the highest sensitivity at 100, 200 and 300 Hz and lowest sensitivity at 600, 1100 and 1200 Hz (Fig. 5B). In tests for differences among frequencies between experimental treatment fish and controls, the minimum adequate model (Table 2) with interactions (saturated) was the best-fit model (AICc 857) compared with the models without interaction and simplified pooled models (AICc 883, 886 and 1100, respectively). Exposure to 15 min of vessel noise did not result in increased hearing thresholds: two-way ANOVA, treatment (P>0.05; Table 2, Fig. 5B). The GLMM analysis revealed that there was a significant difference in frequency (P<0.001; Table 2). To test for differences in hearing thresholds, a priori post hoc pairwise comparisons for specific frequencies between treatments (i.e. 100 Hz control versus 100 Hz noise exposed) were conducted, and there were no differences between treatments for any frequency tested (Table S3; Fig. 5B). Audiograms for particle acceleration thresholds (Fig. 5B) were of a similar shape to the sound pressure audiograms.
Experiment 2: 3-day repeated boat noise exposure
GLMM analyses on hair cell density and apoptosis (Table 2) from the 3-day exposure experiment both had P-values (interaction and treatment, respectively) that were less than the initial sequential Bonferroni adjustment cut-off (P=0.025) and thus adjusted alpha was considered as P≤0.05.
Hair cell bundle density following 3-day repeated vessel noise exposure
In tests for differences among regions between experimental treatment fish and controls, the minimum adequate model (Table 2) with interactions (saturated) was the best-fit model (AICc 942) compared with the models without interaction and simplified pooled models (AICc 990, 1000 and 1027, respectively). Further, the GLMM analysis revealed that there was a significant interaction between treatment and region (P=0.005; Table 2) on hair cell density counts. We examined this interaction with a priori post hoc pairwise comparisons for specific regions between treatments (i.e. H1 control versus H1 noise exposed). These post hoc analyses indicated that fewer hair cell bundles were present in the T1 region of noise exposed fish compared with controls (P=0.016; Table S2; Fig. 6A). No differences in hair cell density were observed between noise-exposed and control treatment fish among the other saccule regions (P=0.016; Table S2; Fig. 6A).
Hair cell apoptosis following 3-day repeated vessel noise exposure
In tests for differences among regions between experimental treatment fish and controls, the minimum adequate model (Table 2) with interactions (saturated) was the best-fit model (AICc 745) compared with the models without interactions and simplified pooled models (AICc 767, 780 and 785, respectively). The GLMM analysis revealed that there was a significant effect of treatment on number of apoptotic cells (P<0.001; Table 2). Because of strong model support for inclusion of the treatment×region interaction (ΔAICc=20; Table 2), we examined differences in apoptotic cells with a priori post hoc pairwise comparisons for specific regions between treatments (i.e. H1 control versus H1 noise exposed). Apoptosis occurred at a higher rate in H1 of noise-exposed fish relative to controls (P=0.002; Table S2; Fig. 6A). No other differences in apoptosis rates were observed among regions between noise treatment and control fish (Table S2; Fig. 6A).
AEP hearing thresholds following 3-day repeated vessel noise exposure
Responses were elicited from all fish from frequencies 100 to 1400 Hz. At 1800 Hz, 80% of control and 90% of noise exposed fish exhibited a response. Atlantic croaker had the highest sensitivity at 100 and 1000 Hz, and the lowest sensitivity was observed at 1200 and 1800 Hz (Fig. 6B). In tests for differences among frequencies between experimental treatment fish and controls, the minimum adequate model (Table 2) with interactions (saturated) was the best-fit model (AICc 926) compared with the models without interaction and simplified pooled models (AICc 956, 968 and 1207, respectively). The GLMM analysis revealed that there was a significant interaction between treatment and frequency (P=0.003; Table 2) on hearing thresholds. Driving the interaction were significant differences in treatment and frequency (P=0.002 and <0.001, respectively; Fig. 6B, Table 2). We examined this interaction with a priori post hoc pairwise comparisons of thresholds between treatments at specific frequencies (i.e. 100 Hz control versus 100 Hz noise exposed) and found differences between treatments at 300, 800 and 1000 Hz (P=0.023, 0.001 and 0.004, respectively; Fig. 6B; Table S3).
DISCUSSION
Our study aimed to determine whether boat noise affects the hearing sensitivity of the soniferous Atlantic croaker through sensory hair cell density changes on the saccular macula and auditory threshold shifts by making comparisons between control and noise-exposed fish. We found that individuals exposed to a single 15-min boat noise exposure event had no change in sensory hair cell density and no change in hearing thresholds immediately after exposure. Furthermore, individuals exposed to a longer-term (3-day repeated exposure) experiment had a significant decrease in sensory hair cell density at one region of the saccule, an increase in apoptotic activity, and higher auditory thresholds (decreased hearing abilities). These results indicate that prolonged exposure, longer time since exposure or both are associated with damage to the ear for Atlantic croaker.
High-intensity, low-frequency anthropogenic noise is a major concern for fishes as there is potential for negative consequences after being exposed to artificial noise (Hildebrand, 2009; Simpson et al., 2016; Cox et al., 2018). This is especially important for species that produce sound and rely on salient signals and hearing to monitor their surroundings (Radford et al., 2014). The inability to accurately hear and respond to biologically relevant acoustic stimuli owing to shifts in thresholds and hair cell damage could lead to the inability of individuals to survive in their environment.
Three-day repeated exposure resulted in a difference in one localized region, the tail region, T1. Previous studies have seen localized damage on the caudal region of the saccule in zebrafish and goldfish (Schuck and Smith, 2009; Smith et al., 2011; Sun et al., 2011) as well as in hybrid striped bass (white bass, Morone chrysops×striped bass, M. saxatilis) and Mozambique tilapia (Oreochromis mossambicus) in response to acoustic stimuli (Casper et al., 2013). This could be due to hair cells being tonotopically organized; however, to our knowledge, this has currently only been observed in goldfish (Smith et al., 2011). We cannot deduce from our study whether tonotopic organization is present in this species; however, our T1 region, located on the caudal region, did show a difference after the 3-day repeated experiment, which is consistent with findings from Smith et al. (2011). However, this should be explored further, primarily because the sensory hair cells are known to exhibit specific orientation patterns. Most studies assessing sensory hair cell density have conducted counts at 5%, 25%, 50%, 75% and 95% of the distance along the rostral–caudal axis of the saccule (Sun et al., 2011; Uribe et al., 2013; Casper et al., 2013); however, Atlantic croaker and other sciaenids exhibit rostral expansion of the saccular maculae, thereby examining hair cells centrally along the rostral–caudal axis in this species may not be the most representative (Ramcharitar et al., 2001; Popper et al., 2005; Tuset et al., 2016). Decreases in hair cell density indicate hair cell kinocilia and stereocilia were sheared from the saccular epithelium, which is shown to decrease hearing abilities and cause vestibular impairment.
There was a noticeable difference in the number of apoptotic hair cells after noise exposure in region H1 compared with control treatments from the 3-day experiment. Previous studies have shown that it takes approximately 1 day for acoustically damaged hair cells to become non-responsive, begin the process of apoptosis, and be ejected from the saccular epithelium (Smith et al., 2006; Schuck and Smith, 2009; Sun et al., 2011). Because of this, apoptotic activity was not tested in our first experiment as there was less than 1 h between noise exposure, euthanasia and preservation. Our regional TUNEL results (H1 differences) do not mirror what was seen in the phalloidin assay (T1 difference) of the same 3-day experiment (opposite ears). Therefore, acoustic damage could be occurring randomly and not at specific macula regions, as observed in Smith et al. (2011). These hair cells are specialized mechanosensory receptors that, when deflected by the relative motion of the sagitta otolith, convert mechanical stimuli into neural signals (Monroe et al., 2015). Acoustic overstimulation has been shown to decrease hair cell bundles, cause lesions on the epithelium where previous hair cells were, and cause hair cell bundles to be thinner and shorter, all of which can lead to decreased hearing and vestibular function (Monroe et al., 2015). Although sensory hair cells have been shown to regenerate after ototoxic and sound exposure, regeneration times vary depending on noise source, intensity and species (Lombarte et al., 1993; Smith et al., 2006; Monroe et al., 2015; Liu et al., 2013). Some research has shown that hair cells can take days to weeks to regenerate (Hastings et al., 1996; McCauley et al., 2003; Scholik and Yan, 2002; Monroe et al., 2015). We are unaware of any studies assessing hair cell generation time in Sciaenidae, and further study on hair cell regeneration is warranted as our study demonstrated detrimental effects from noise in this family that uses sound in reproductive behavior.
In our study, the 15-min exposure experiment was conducted during peak reproductive season. In another vocal fish species, plainfin midshipman (Porichthys notatus), mature females have increased hair cell density and auditory sensitivity when in spawning condition (Coffin et al., 2012; Lozier and Sisneros, 2019), which may be adaptive for mate localization and detection. We did not determine the age or spawning condition of fish utilized in the present study; however, most individuals were females within the size range when fish reach maturity at year 1 (Barger, 1985; Hales and Reitz, 1992; Anderson et al., 2018).
Hair cell bundle damage and hearing threshold increases were not evident after 15 min of noise exposure but were present after 3 days of repeated exposure. Previous studies observed minimal hearing loss immediately after sound exposure, whereas maximal hearing threshold shifts may occur 24 h after exposure (Smith et al., 2004, 2006; Uribe et al., 2013; Liu et al., 2013; Nissen et al., 2019). Our observation following repeated noise exposure is consistent with a previous study on fathead minnow (Pimephales promelas) that explicitly assessed hearing threshold changes in response to boat noise exposure. Their results indicated decreases in hearing thresholds at some of the fishes' most sensitive hearing frequencies (Scholik and Yan, 2001). Other studies focusing on high frequency (e.g. McCauley et al., 2003; Halvorsen et al., 2012a,b; Casper et al., 2013) and Gaussian white noise sources (e.g. Smith et al., 2004; Wysocki and Ladich, 2005; Breitzler et al., 2020) also observed diminished hearing post noise exposure or during masking events.
Particle motion is an essential component of sound detection for most fish species and aquatic invertebrates and thus it is critical to characterize particle motion amplitudes in studies on noise exposure and hearing thresholds (Popper and Hawkins, 2018). In most studies to date, the focus has been on determining the effect of sound pressure impacts on organisms, even though there is evidence for particle motion within the sound field to be a major cause for effects. Because non-otophysan (e.g. Atlantic croaker) species lack Weberian ossicles and respond primarily to particle motion, it is important to characterize particle motion in ambient and noise conditions, both in situ and in laboratory studies. However, mimicking aquatic acoustics in a laboratory setup is complicated as many factors (i.e. tank size and material) can alter acoustic characteristics, which have been previously explored extensively (see Rogers et al., 2016; Duncan et al., 2016; Popper and Hawkins, 2018). This study recorded particle motion in situ of original field recordings, within the experimental exposure tanks, and the AEP setup. Particle acceleration is frequency-dependent and highest in the nearfield. As in other studies, our AEP setup and experimental exposure tanks were in the acoustic near field. Thus, most of the boat playback (frequencies below 4800 Hz) was within the acoustic near field, resulting in high PAL. Therefore, PAL from high-frequency exposures does not spatially occupy as much of the available aquatic environment compared with intense, low-frequency boat noise. To our knowledge, hearing damage occurring from PAL versus SPLs has not been explored. However, we expect high PAL levels to be more likely to cause hair cell shearing in fish without accessory auditory structures that transduce sound pressure because the fish's head would be moved with greater force relative to the otolith's inertia.
Recent meta-analyses suggest that continuous and irregular sounds such as boat noise may have the most pronounced effect on various factors such as stress, masking and hearing loss (de Jong et al., 2018). Our hearing thresholds showed shifts within the frequency range of Atlantic croaker sound production (300–1000 Hz). Atlantic croaker produce sounds (as described by Ramcharitar et al., 2006) that are shown to be associated with disturbance (croak: 540–650 Hz), reproduction (drumming: 300 Hz) and unknown behavioral context (knock: 600–1600 Hz). As a species reliant on vocalization for spawning, damage to hearing may hinder successful spawning events and vestibular function. Increased noise disturbances may also make it harder for spawning aggregations to form, as signal masking (more so if individuals are exhibiting hearing loss) could impact spawning aggregation localization (Collin et al., 2003), and subsequently decrease spawning success and lower population abundance.
This study represents the first record of audiograms for Atlantic croaker within the northern Gulf of Mexico population (compared with Ramcharitar and Popper, 2004 and Horodysky et al., 2008 descriptions of Atlantic croaker from the Northeast United States). In the present study, Atlantic croaker hearing threshold estimates measured by sound pressure were higher from 100 to 700 Hz than thresholds reported in two previous studies (Ramcharitar et al., 2006; Horodysky et al., 2008), but all three studies show a general decrease in sensitivity by 400 Hz. Thresholds expressed as PAL were determined in Horodysky et al. (2008) but not Ramcharitar et al. (2006) and were lower than, though generally similar to, the present study except at 200 and 300 Hz. Further, fish sex was not reported in Ramcharitar et al. (2006) nor Horodysky et al. (2008), but nearly all fish in our study were female, potentially mature, but not large adults. The difference in audiograms established in this study compared with previous published audiograms of Atlantic croaker could be due to a variety of factors. AEP setups vary across studies, such as the use of metal chambers, plexiglass tanks and plastic buckets as the holding chamber, water depth and position of an individual, and speaker placement (i.e. under or above water); these differences can alter the relative contribution of particle motion and sound pressure within a setup. These differences may cause sound acoustics to vary per setup, both within a laboratory and across laboratories, which have been extensively discussed in Higgs (2002) and Ladich and Fay (2013). Further, water parameters (i.e. temperature and salinity level during acclimation periods and trials) may have varied among these studies, and in several other fishes, hearing sensitivity increases with temperature (Wysocki et al., 2009; Ladich, 2018). Our study, however, compares responses to sound stimuli between noise-exposed and control fish tested under identical conditions. The observed >10 dB threshold increases at three frequencies for fish exposed to repeated 3-day noise suggest that sound at these frequencies needs to be at least three times the amplitude regardless of the absolute intensity to evoke a similar response. Furthermore, differences in age and seasonality (i.e. spawning season) may alter hearing sensitivities, which has been seen in other species (Alderks and Sisneros, 2011; Coffin et al., 2012).
The integrative approach of this study is one of the few that investigates boat noise impacts on sensory hair cell damage and auditory thresholds within an ecologically important estuarine fish. Recent reviews on information gaps and setting sound exposure criteria (Hawkins et al., 2015, 2020) have addressed many future directions and questions, including the need to select appropriate species for continued sound studies. Prioritization should be given to studies representing species that exhibit various anatomical structures (i.e. have structures that enhance hearing), have different ecological associations, and can serve as representatives of vulnerable and threatened species (Hawkins et al., 2015, 2020). Our study indicates that a common coastal fish species in the western North Atlantic shows noticeable impacts to the auditory system after modest levels of noise exposure that may be expected in harbors and navigation channels. Future work should evaluate boat-noise-induced changes by assessing hearing thresholds and hair cell density at multiple days post exposure, recovery time and hair cell regeneration time, as well as behavioral responses both experimentally and in situ. Altogether, this study established important baseline data for noise impacts on an ecologically and commercially valuable species which may be representative for other important fishes, as well as provided valuable insight into potential impacts of boat noise on fish that will be beneficial for fisheries management and the monitoring of the underwater soundscape.
Acknowledgements
We acknowledge Dr Bernard Rees, Dr Martin O'Connell and Dr Simon Lailvaux for their comments and suggestions throughout the duration of experiments. We also acknowledge the support from all members of the Fish Morphology and Behavior Lab at the University of New Orleans; Bennett Price and the army of undergraduates (Ariel Alonso, Olivia Bergeron, Diamond Burrows, Aubrey Byrd, Dakota Brunetti, Jade Cockrell, Amanda Demesia, Hannah Tanib and Kaitlyn Tillman) for their help in fish collection, husbandry, histology processing and hair cell counting throughout the experiments. We also thank the University of New Orleans for use of its facilities.
Footnotes
Author contributions
Conceptualization: G.A.B., K.S.B.; Methodology: G.A.B., K.S.B.; Validation: K.S.B.; Formal analysis: G.A.B.; Investigation: G.A.B.; Resources: G.A.B., K.S.B.; Data curation: G.A.B.; Writing - original draft: G.A.B.; Writing - review & editing: G.A.B., K.S.B.; Visualization: G.A.B., K.S.B.; Supervision: K.S.B.; Project administration: K.S.B.; Funding acquisition: K.S.B.
Funding
This study was funded by the Louisiana Board of Regents [LEQSF(2020-23)-RD-A-30, to K.S.B.] and the Louisiana Environmental Education Commission through the Louisiana Department of Education (to G.A.B.).
Data availability
All relevant data can be found within the article and its supplementary information.
References
Competing interests
The authors declare no competing or financial interests.