Characteristics of sound production and associated pharyngeal jaws in the tomtate grunt Haemulon aurolineatum (Cuvier, 1830) in Caribbean reefs

The ability to produce sounds for acoustic communication is well known in different grunt species (Haemulidae). However, most of the sounds have not been described and the sound-producing mechanism of very few grunt species has been deeply studied. Additional data is needed to search for synapomorphy in the sonic mechanism. This study describes acoustic features and branchial anatomy in Haemulon aurolineatum. Correlations were found between some acoustic features and standard length, showing the largest specimens produced shorter, lower-pitched grunts of higher intensity. Examinations of acoustic features and branchial anatomy show that H. aurolineatum uses the same stridulatory mechanism described previously in H. flavolineatum. The unusual feature of Haemulon species concerns the fourth ceratobranchials. These appear to be part of the lower pharyngeal jaws since they possess firmly attached teeth that face the upper pharyngeal jaws. The stridulation results from the rubbing of both pharyngeal and fourth ceratobranchial teeth. This mechanism is probably common to the 23 Haemulon species, but additional information is needed regarding the mechanism of other Haemulinae species to produce stridulatory sounds. Fourth ceratobranchials could constitute a key element of Haemulinae ability to produce sounds providing an eventual synapomorphic aspect of the mechanism in the family.


Introduction
In teleosts, acoustic signals are used during competitive feeding (aMoriM & hawkinS 2005), distress or alarm situations (knight & laDich 2014), territorial interactions (longrie et al. 2013;Picciulin et al. 2006), conspecific identification, as well as during courtship and agonistic displays (colleye & ParMentier 2012;ParMentier et al. 2010a), mate choice (aMoriM et al. 2015), mate quality assessment (aMoriM et al. 2015) or coordination of gamete release (loBel 1992). Descriptions of these sounds are based on various acoustic characteristics, including different temporal features (e.g., number of pulses, period of pulses), dominant frequencies and amplitude. Detailed analyses suggest that many of these sounds could be species-specific (colleye et al. 2011;geralD 1971;hawkinS & raSMuSSen 1978;MalavaSi et al. 2008). In some taxa, acoustic features can also provide information about the individual, such as the age, sex or size of the sender (colleye et al. 2009;loBel & Mann 1995;MyrBerg et al. 1993;PeDroSo et al. 2013;tellechea et al. 2010). A relationship between spectral and individual features can be found in all species of teleosts, but the quantitative value of this relationship will depend on the mechanisms of sound production involved (ParMentier & Fine 2016).
The high diversity of mechanisms that fish have developed suggests that morphological characteristics promoting acoustic communication evolved independently in distant phylogenetic taxa (ParMentier et al. 2017a;SchneiDer 1967). Moreover, morphological mechanisms used to produce sounds in fish could result from independent modification of existing structures with other primary functions. In different taxa, existing anatomical structures, such as teeth, bones and swim-bladders, were likely first used in non-voluntary sound production and then selected for signal production (ParMentier et al. 2017a). Therefore, sound mechanisms in fish could be examples of exaptations, which refers to a functional character previously acquired for a particular function being later co-opted for a new use (goulD & vrBa 1982).
Haemulidae are commonly known as grunts due to their ability to produce sounds in distress situations (e.g., when handheld or netted) and during competitive feeding. It is not known whether sound production in other social contexts takes place, given that no experiments have been performed. Sounds have been reported, but not fully described, when fish were netted or during manual stimulations in H. macrostomum, H. album, H. parrai, H. carbonarium, H. melanurum, H. plumieri, H. aurolineatum, H. striatum, Anisostremus virginicus, Conodon nobilis, Pomadasys corvinaeformis, Orthopristis chrysopterus (BurkenroaD 1930;cuMMingS et al. 1966;FiSh & MowBray 1970;Moulton 1958). More precise descriptions of acoustic features have been given in H. flavolineatum (Bertucci et al. 2014). In H. plumieri and H. flavolineatum, sound production mechanisms have been experimentally associated with stridulation resulting from the rubbing of pharyngeal teeth at the level of the branchial basket (Bertucci et al. 2014). In the French grunt (H. flavolineatum), movements of pharyngeal jaws during sound production appeared to correspond to movements described during food processing (Bertucci et al. 2014;wainwright 1989a, 1989b, supporting that this stridulatory mechanism could be an exaptation of the feeding mechanism (ParMentier et al. 2017a). Sound production and acoustic features related to stridulation of the pharyngeal jaws could be common to the entire family (FiSh & MowBray 1970;Moulton 1958;tavolga & woDinSky 1965). However, additional morphological data on other species are needed to confirm this hypothesis.
The present work aims to describe acoustic features and the sound-producing apparatus of the tomtate grunt, Haemulon aurolineatum Cuvier, 1830. This species is found west of the Atlantic Ocean, between Cape Cod (Massachusetts) and Brazil, including the Caribbean, Gulf of Mexico and Central American coasts (Manooch & BaranS 1982). They occur from shallow water to offshore reefs, and are schooling fish whose adults feed on benthic invertebrates (Darcy 1983). The tomtate grunt is one of the most common deep-water reef species at depths greater than 30 metres in the Eastern Gulf of Mexico, and can reproduce all year round (Darcy 1983).

Sound collection
Nineteen specimens of Haemulon aurolineatum (Standard Length, SL: 48-115 mm ) were caught during daytime at the Anse du Mancenillier, on the east side of Grande-Terre in Guadeloupe (French West Indies), in January 2020.
Specimens were gently handheld and maintained within a submerged landing net. A hydrophone HTI-96-Min (sensitivity: -163.9 dB re 1 V μPa −1 ; flat frequency response range 2 Hz-30 kHz; High Tech, Inc. Long Beach, ms, USA) was placed in the centre of the net and connected to a TASCAM DR-05 portable audio recorder (sampling frequency: 44.1 kHz, 16-bit resolution; TEAC, Wiesbaden, Germany). These recordings were performed at sea (1.2 m depth), which provides the advantage of not deforming sounds, as can occur in aquaria . Specimens were positioned about 3 cm from the hydrophone until sounds were obtained. Recordings of each specimen never lasted more than three minutes and were stopped when at least 40 sounds were recorded.

Sound analysis
A minimum of 30 sounds/specimen was analysed. Sounds were analysed with Avisoft -SASLab Pro 5.2.13 software (Avisoft Bioacoustics, Glienicke, Germany). All recordings were sub-sampled at 11,025 Hz, with a high-pass filter then applied to remove frequencies below 50 Hz, corresponding to the majority of background noise.
Several acoustic features were measured for each sound (grunt) (Figs 1-2): grunt duration (ms), number of pulses in the grunt, pulse period (peak-to-peak interval between two consecutive pulses, ms), dominant frequency (Hz) and the amplitude of the dominant frequency peak (dB SPL, Sound Pressure Level). Because the hydrophone was not calibrated, all the fish were recorded the same day, at the same location with the same tools to allow comparison of data between specimens. As the recording loop was not calibrated, SPLs are not absolute values but allow comparison of the relative intensity between all the recorded fish.
When sounds were produced in series, other acoustic features were measured ( Fig. 1): series duration (s), number of sounds per series, grunt intervals (interval between the end of a grunt and the beginning of the next grunt in a series, ms) and grunt period (interval between two successive repetitions of the same pattern between two grunts following each other in a series, ms).

Morphology
After recording, seven specimens were euthanised with an overdose of tricaine methanesulfonate (MS-222). They were fixed in a 5% formaldehyde solution and, after two weeks, they were transferred to a 70% ethanol solution. The 12 remaining specimens were freed.
Specimens were dissected under a binocular microscope (Leica Wild M10, Leica, Solms, Germany) coupled to a camera lucida to observe the branchial basket morphology.
Pharyngeal jaws of the biggest specimen (SL: 115 mm ) were prepared for a classical surface examination by scanning electron microscopy (SEM). Samples were critical point dried with CO 2 , directly mounted on glass slides with carbon tape, and sputter-coated with a 20 nm Ag in a Balzers SCD 030 unit (Oerlikon Balzers Coating, Balzers, Liechtenstein). Samples were then examined in a FEI ESEM Quanta 600 for classical secondary electron (SE) imaging.

Statistical analyses
Mean and standard deviation were calculated for each acoustic feature of grunts. Analysis concerned 556 grunts from 17 specimens (Table 1). A Shapiro-Wilk test was used to test for normal distribution of acoustic data. A Bartlett test was applied to control homoscedasticity. As the assumptions of normal distribution and homoscedasticity were not met, non-parametric tests were used. A Spearman's correlation matrix was performed between grunt duration (ms), dominant frequency amplitude (dB), mean of pulse periods per grunt (ms), number of pulses per grunt and standard length (mm).
Relationships between amplitude of dominant frequency (dB) and standard length (mm), dominant frequency (Hz) and standard length (mm), and number of pulses per grunt and standard length (mm) were estimated using a linear model (R package "nlme"). Specimens were maintained as a random effect to take into account that each specimen produced a different number of sounds. All statistical analyses were achieved with R studio version 1.1.463 software (RCore Team 2018).

Ethics
All the experiments were carried out under the approval of the Ethics Committee of the University of Liège (Case 1759).

Sound production
Of the 19 specimens collected, one did not produce sounds and recordings of another one could not be analysed because the signal-to-noise ratio was too low. A total of 556 grunts from 17 specimens were analysed. Grunts were isolated or produced in series, lasting between 23.5 and 96.0 ms (mean ± SD: 53.4 ± 11.7 ms). Their dominant frequency ranged between 303 and 1384 Hz (686 ± 172 Hz), with a relative amplitude between 104 and 135 dB (123 ± 6 dB). Grunts were made of 3 to 12 pulses (6 ± 2 pulses) with a pulse period extending from 1.9 to 37.5 ms (9.3 ± 4.5 ms). Twelve specimens produced grunts in series (n series = 77). The four smallest specimens (48, 52, 56 and 57 mm SL) did not produce any series of grunts. Series lasted from 0.2 to 3.9 s (0.7 ± 0.6; n = 150) and were composed of 2 to 10 grunts (3 ± 2; n = 77; n = 150). Within series, grunt intervals lasted between 10 and 790 ms (250 ± 14 ms) and the grunt period was between 110 and 860 ms (310 ± 140 ms).
There was a positive correlation (Spearman's correlation matrix, r = 0.69; p < 0.001; n = 556) between the relative amplitude and the standard length, and negative correlations between the dominant frequency (r = -0.23; p < 0.001) and the number of pulses per grunt (r = -0.21; p < 0.001) and the standard length. Therefore, longer specimens produced grunts with fewer pulses but with higher amplitude and a lower frequency than smaller specimens. Relationships between these acoustic features were also estimated using linear mixed effect models (Fig. 3).

Morphology
Upper pharyngeal jaws covered the lower pharyngeal jaws and anterior parts of both ceratobranchials 4 (Fig. 4A). Teeth were present on upper pharyngeal jaws and lower pharyngeal jaws (Fig. 4B-4D).
Teeth of the upper pharyngeal jaw were uniformly distributed on tooth plates. Thin and elongated teeth were found on pharyngobranchials (PBR) 2 and 4, whereas teeth of PBR3 were stout (Fig. 4B). All these teeth possessed a tightening before a sharp tip, bent towards the back of the branchial basket (Fig. 4C). Teeth possessing a tightening before sharp tips were also found on the lower pharyngeal jaw (Fig. 4D). Medial teeth were the largest and their size decreased from the centre to the periphery. At the anterior part, some lateral teeth of lower pharyngeal jaws were more elongated and directed laterally.
Ceratobranchials (CBR) 4 had teeth on their dorsal anterior half. These teeth were not evenly distributed, forming five triangular groups of 12-14 teeth (Fig. 4E). Smaller teeth were found at the tip of triangles, whereas longer teeth were found at the base, close to the elongated teeth of the lower pharyngeal jaw (Fig. 4F).
Teeth tips showed an orange colouration (Fig. 4G). Enamel of the most external and thin teeth in upper and lower pharyngeal jaws, and even in ceratobranchials 4, showed erosion (Fig. 4F).  Pharyngeal muscles were similar to pharyngeal muscles illustrated in H. flavolineatum (Bertucci et al. 2014) and in 9 different haemulid species (wainwright 1989b). Observations of the inner mouth during sound production (in air) confirms movements were done at the level of the pharyngeal jaws during sound production.

Discussion
Stridulation sounds are often composed of a series of irregular transient pulses, containing a wide range of frequencies and great variation in temporal parameters (Fine & ParMentier 2015;hawkinS 1993;laDich 1997). Similar to other vocal haemulids, the sounds of H. flavolineatum consist of rapidly produced grunts that show important variability in both pulse period and sound duration with broadband frequency ranging from 303 to 1384 Hz. Sound features, movements at the level of pharyngeal jaws during sound production and common features in the branchial basket anatomy suggest a common morphological mechanism to produce sounds in these haemulids. Most sounds resulting from movements of pharyngeal teeth are thought to be by-products of feeding activities (ParMentier et al. 2013). In the present study, specimens did not feed and produced sounds, whether isolated or in series, when being handheld. The smallest specimens produced only isolated grunts. The production of grunts, whether single or in series, could suggest different meanings or motivations of the sound producer. Although the sex or sexual maturity of specimens were not determined, the range of the size of the 19 specimens probably encompassed both sexes. This means the ability to produce sounds is not related to sex or age.
A wide dominant frequency range (between 303 and 1384 Hz) and a high variability of pulse period could be a characteristic of sounds produced by stridulation (DeMSki et al.1973;Fine & ParMentier 2015;tavolga 1971). Three main kinds of stridulating sound-production mechanisms have been described in teleost species: friction of pectoral spine against the pectoral girdle in different catfish species (Fine et al. 1996;ParMentier et al. 2010b), articulation between two bones in the back of the head in seahorses (liM et al. 2015), and rubbing of the pharyngeal teeth in haemulid grunts (Bertucci et al. 2014;BurkenroaD 1930) andgobies (takeMura 1984). In teleosts, the lower pharyngeal jaws are only composed of ceratobranchials 5 (vanDewalle et al. 2000;wainwright 2005). Ceratobranchials 4 are usually not part of lower pharyngeal jaws and they do not possess teeth that could help in food processing. However, fourth ceratobranchials commonly possess movable gill rakers, as it is the case for ceratobranchials 1, 2 and 3. In teleosts, these skeletal pieces are at least used to protect gills during water flow (vanDewalle et al. 2000). Nevertheless, the case of Haemulon species seems to be more complex. They do not possess movable gill rakers on the dorsal margin of fourth ceratobranchials, but rather teeth that are clearly implanted in triangular bony plates and that face teeth of the upper pharyngeal jaws. Moreover, the synchronous activities of pharyngeal muscles ventral to ceratobranchials 4 and 5 confirm that all these skeletal elements could act as a single unit (wainwright 1989b). This characteristic is also found in the French grunt (H. flavolineatum), with both species possessing the same pharyngeal muscles. All these morphological and acoustic similarities suggest sounds are produced in the same way. Movements of pharyngeal jaws have been described for the French grunt through high-speed X-ray video recordings during sound production (Bertucci et al. 2014). These observations have shown that sounds were effectively the result of scratching the teeth of the upper and lower pharyngeal jaws. Moreover, erosion of teeth of the upper and lower pharyngeal jaws could result from a friction between these two pharyngeal units.
Although data from the literature do not allow deep comparisons, it seems that the same kind of sounds have been recorded in other haemulid species (BurkenroaD 1930;FiSh & MowBray 1970;Moulton 1958), meaning they should also use the same kind of mechanism. Haemulon plumieri, H. flavolineatum and H. aurolineatum, for example, have common anatomy at the level of ceratobranchials 4. Although unusual within teleosts, the organisation of teeth in triangular patches on the ceratobranchials 4 is not referred to as a synapomorphy in the haemulid literature (JohnSon 1981). However, we hypothesise that this feature coincides with the ability to make sounds. This could be true for all 23 valid Haemulon species (www.fishbase.se 2019). As sounds were also recorded (but not described) in haemulid specimens from the Anisotremus, Conodon, Pomadasys and Orthopristis species (FiSh & MowBray 1970), we expect that the same kind of anatomy can be found in these species. However, such sound features and anatomy remain to be studied. Moreover, all these genera belong to the sub-family Haemulinae, meaning additional data are required to understand the acoustic abilities of the second subfamily, the Plectorhinchinae.
In H. aurolineatum, there was a negative relation between dominant frequency and size of the sound producer. Decreasing dominant sound frequencies with increasing body size is a general phenomenon in animals (henglMüller & laDich 1999). This negative relation was found in several teleost species (aMoriM et al. 2003;Bertucci et al. 2012;De Jong et al. 2007;MyrBerg & lugli 2006;PruzSinSzky & laDich 1998). This supports the notion that sounds could provide information on the size of the sound producer (colleye et al. 2009;loBel & Mann 1995). In addition, one study revealed that this relationship could provide information about the morphological mechanism responsible for sound production (ParMentier & Fine 2016). In species producing sounds with fast-contracting muscles, the relationship between the dominant frequency and fish size is weak, since the slopes are between 1 and 10°. When the soundproducing mechanism is not based on direct muscle contractions, slopes were steeper, between 60 and 97° (ParMentier & Fine 2016). In the tomtate grunt, the steep slope (85°) suggests that the sound production mechanism is not based on fast-contracting muscles, reinforcing our hypothesis on the stridulatory mechanism. A positive relation was also found between amplitude and fish size: large specimens produce louder sounds than small specimens. This relationship was found in different studies (Boyle & tricaS 2011;connaughton et al. 2000;knight & laDich 2014;linDStröM & lugli 2000;ParMentier et al. 2017b;takeMura 1984) but there is no satisfactory explanation for this positive relation to date. Relationships were found between three acoustic features (dominant frequency, amplitude, and number of pulses per grunt) and size, which could allow a precise estimation of the size of a specimen just by listening to sounds.

Conclusion
Current studies in marine ecology are trying to develop monitoring of underwater areas based on the identification of species-specific sounds. For this, describing the sounds and understanding in which contexts they are produced by closely related species appears particularly necessary. This study showed that Haemulon aurolineatum is able to produce sounds using a stridulatory mechanism. This could be related to a family feature since, unlike other teleost species, ceratobranchials 4 appear to be part of lower pharyngeal jaws. This feature could be related to the ability to produce loud sounds with teeth of the branchial basket. Additional studies in other Haemulidae species are required to further describe the ceratobranchials 4 morphology.