Inhibitory and modulatory inputs to the vocal central pattern generator of a teleost fish

Abstract Vocalization is a behavioral feature that is shared among multiple vertebrate lineages, including fish. The temporal patterning of vocal communication signals is set, in part, by central pattern generators (CPGs). Toadfishes are well‐established models for CPG coding of vocalization at the hindbrain level. The vocal CPG comprises three topographically separate nuclei: pre‐pacemaker, pacemaker, motor. While the connectivity between these nuclei is well understood, their neurochemical profile remains largely unexplored. The highly vocal Gulf toadfish, Opsanus beta, has been the subject of previous behavioral, neuroanatomical and neurophysiological studies. Combining transneuronal neurobiotin‐labeling with immunohistochemistry, we map the distribution of inhibitory neurotransmitters and neuromodulators along with gap junctions in the vocal CPG of this species. Dense GABAergic and glycinergic label is found throughout the CPG, with labeled somata immediately adjacent to or within CPG nuclei, including a distinct subset of pacemaker neurons co‐labeled with neurobiotin and glycine. Neurobiotin‐labeled motor and pacemaker neurons are densely co‐labeled with the gap junction protein connexin 35/36, supporting the hypothesis that transneuronal neurobiotin‐labeling occurs, at least in part, via gap junction coupling. Serotonergic and catecholaminergic label is also robust within the entire vocal CPG, with additional cholinergic label in pacemaker and prepacemaker nuclei. Likely sources of these putative modulatory inputs are neurons within or immediately adjacent to vocal CPG neurons. Together with prior neurophysiological investigations, the results reveal potential mechanisms for generating multiple classes of social context‐dependent vocalizations with widely divergent temporal and spectral properties.

Among vocal species of fish, toadfishes (order Batrachoidiformes, family Batrachoididae) include species commonly known as toadfish and midshipman fish (Greenfield, Winterbottom, & Collette, 2008) that have provided tractable models for neurophysiological investigations of vocal CPGs (Ladich, Collin, Moller, & Kapoor, 2006). These fish produce several types of social context-dependent vocalizations (e.g., Figure 1a) by repetitively contracting a single pair of "superfast" vocal muscles attached to the walls of the swim bladder at frequencies of about 100-250 Hz depending on the ambient water temperature (Bass & Baker, 1991;Brantley & Bass, 1994;Cohen & Winn, 1967;Rome, 2006;Skoglund, 1961). These contractions are elicited by the activity of a hindbrain vocal motor nucleus (VMN) that innervates the vocal muscles via occipital motor nerve roots. These roots form a single vocal nerve (VN, Figure 1b, c) considered homologous to the hypoglossal nerve roots of other vertebrates (Bass, Gilland, & Baker, 2008). Transneuronal tracing via labeling of the cut end of one VN with neurobiotin or biocytin (Figure 1d, e) together with single cell electrophysiological recordings demonstrate three topographically separate vocal CPG nuclei that code for distinct call parameters: VMN for amplitude modulation, vocal pacemaker nucleus (VPN) for pulse repetition rate (corresponds to fundamental frequency) and vocal pre-pacemaker nucleus (VPP) for duration (Figure 1f-h;Bass & Baker, 1990;Bass, Marchaterre, & Baker, 1994;Chagnaud, Baker, & Bass, 2011;Chagnaud, Zee, Baker, & Bass, 2012).
While the organization of vocal CPGs has been investigated at network and single neuron levels for toadfishes and the African clawed frog Xenopus laevis (e.g., Bass & Baker, 1990;Chagnaud et al., 2011;Chagnaud et al., 2012;Kelley et al., 2017;Yamaguchi et al., 2000;Zornik & Yamaguchi, 2012), less is known about the neurochemicals that modulate their activity (e.g., Forlano, Kim, Krzyminska, and Sisneros, 2014;Yu & Yamaguchi, 2010;Zornik, Katzen, Rhodes, and Yamaguchi, 2010). The well-established neuroanatomical and neurophysiological characterization of the vocal CPG in species such as the Gulf toadfish, Opsanus beta, together with the ability to unambiguously identify each vocal CPG nucleus via transneuronal neurobiotin labeling, present a distinct opportunity to investigate the neurochemical profile of each component of an evolutionarily conserved vertebrate vocal CPG (Bass et al., 2008;Chagnaud & Bass, 2014). Here, we report the robust distribution of inhibitory neurotransmitters and neuromodulators within the Gulf toadfish vocal CPG that likely contribute to the ability of these and other species of toadfish to produce social context-dependent vocal behaviors with divergent temporal and spectral properties (Rice & Bass, 2009).

| Animals
Seventeen Gulf toadfish, Opsanus beta, (12 males, four females, sex not reported for one fish; 3.5-18.7 cm in standard length; median 5 10.5 cm; interquartile range 5 10.0) were obtained from a commercial source (Gulf Specimen, Panacea, FL) and housed in aquaria at 228C in an environmental control room on a 12:12 dark:light cycle.
All surgical methods and collection of tissues were approved by the Cornell University Institutional Animal Care and Use Committee.

| Vocal CPG labeling
Neurobiotin crystals (Vector Laboratories, Burlingame, CA) were applied to the cut end of one vocal nerve at the level of the swim bladder. A detailed description of the methods can be found in Bass et al. (1994). After a survival time of 1.5-7 days following neurobiotin application, fish were deeply anaesthetized by immersion in aquarium water with benzocaine (0.025%; Sigma Aldrich, St. Louis, MO) and then trans-

| Immunohistochemistry (IHC)
One day prior to sectioning, fixed brains were cryoprotected in 30% sucrose in 0.1 M PB (Carl Roth GmbH 1 Co. KG, Karlsruhe, Germany) at 48C overnight. Brains were sectioned in the transverse plane with a cryostat at 25 mm (Leica microsystems, Wetzlar, Germany) and directly mounted onto microscope slides (Superfrost Ultra Plus Adhesion Slides; Thermo Fisher Scientific Inc., Braunschweig, Germany). Slides were subsequently left at room temperature for 1 hr to allow the sections to dry and then either processed for IHC immediately or stored at 2808C. Each brain was sectioned into four complete series, each of which was stained with a single antibody.
For IHC, the slides were immersed in 0.1 M PB-saline (PBS) for 30 min for rehydration. Glutaraldehyde fixed brains were additionally washed in 0.001% sodium borohydride (Sigma Aldrich Chemie GmbH, Munich, Germany) in 1 ml 0.1 M PBS for 5 min to reduce glutaraldehyde background. A washing series (four times, 5 min each) in 0.5% Triton 100 (Sigma Aldrich Chemie GmbH, Munich, Germany) in 0.1 M PBS (PBS-T) followed. Subsequently, each slide was washed in 10% normal donkey serum (Jackson Immunoresearch Europe Ltd., Suffolk, United Kingdom) in PBS-T for 1 hr, before incubation with primary antibody (Table 1) overnight. Slides were then washed four times in PBS-T before incubation with secondary antibody and anti-biotin antibody (Table 1) for 4 hr. Slides were again washed four times with PBS-T, dried and then coverslipped using a fluorescent mounting medium (Vectashield, Vector Labs Inc., Peterborough, United Kingdom) containing 4 0 ,6-diamidino-2-phenyindole (DAPI). All incubation and washing steps were performed at room temperature.
Images of brain sections were taken on a confocal laser microscope (Leica microsystems, Wetzlar, Germany) and at an epifluorescence microscope (ECLIPSE Ni, Nikon GmbH, D€ usseldorf, Germany).
Acquired confocal images were stacked and converted to maximum zprojections using the free software Fiji (Schindelin, Arganda-Carreras, & Frise, 2012). Maximum z-projections were cropped, resized and contrast and brightness were optimized for the entire image using Adobe

| Data analysis
For three similar-sized fish (OB-15-05 (ID code): 5.3 cm (standard length); OB-15-06: 5.1 cm; OB-16-10: 6.1 cm), we evaluated several dimensions of the neurobiotin-labeled vocal CPG nuclei (VMN, VPN, VPP): rostral-caudal extent of each nucleus, neuron number for each nucleus, individual neuron size reported as diameter, and individual neuron shape evaluated by the shape factor. Rostral-caudal extent was determined by counting all sections of each brain where neurons of the respective nuclei were labeled and then multiplied by 25 mm, that is, the section thickness. Neuron number was estimated using the following criteria: First, in one series (out of four, see above) of each brain all neurobiotin labeled neurons were counted in which the nucleus of a given neuron could be identified with DAPI. This value was multiplied by four to account for neurons in the four brain series (see section 2.3), and then corrected using the Abercrombie correction factor (Abercrombie, 1946). For each fish the diameter of the nucleus was measured in 10 neurons of each cell population (VMN, VPN, VPP), and the average for each population was used for the Abercrombie correction for each respective population. Neuronal soma diameter was calculated as the average of minor and major axis measurements in each fish as has been done previously in O. beta (Chagnaud & Bass, 2014). Shape factor was calculated as the aspect ratio of a neuron's smallest dimension in a single plane (minor axis) divided by the neuron's largest dimension (major axis). A shape factor of one represents a perfect circle with decreasing values indicating more elongated shapes. Minor and major axes were measured with the ruler tool in Adobe Photoshop CS6 software.
As we report, four groups of labeled neurons were recognized in the premotor VPN region: neurobiotin-only, GABA-only, glycine-only, and neurobiotin-glycine co-labeled. A non-neurobiotin labeled neuron (i.e., glycinergic or GABAergic) was attributed to a vocal CPG nucleus if it was surrounded by neurobiotin-labeled neurons of that nucleus or if it was located adjacent to such neurons. Due to the heterogeneity of VPN neurons, neuron number and diameter were assessed separately for those groups along with the neuron shape factor. In order to avoid size effect between fishes, neuron diameter and shape factor were compared between the VPN groups in three fish (OB-15-04: 10.5 cm; OB-16-09: 8.5 cm; OB-16-10: 6.1 cm) that exhibited neurobiotin, GABA and glycine label. For each fish, we separately tested if neuron diameter and shape factor were significantly different between the four neuron groups. We visually assessed a normal distribution for each neuron group within each fish using normal quantile-quantile plots (Supporting Information Figure S1). We used a we marked each neuron type with an asterisk or star (see Figure 5a, b).
Next, we formed a polygon by connecting neurobiotin-positive neurons with lines that enclose all other neurobiotin-positive neurons in one section. We restricted this step to neurobiotin-positive neurons because they form the principal neurons of VPN. For the resulting polygon, we calculated the centroid using the python script "Centroid.

| Glycine
A polyclonal antibody (1015GE; MoBiTec, G€ ottingen, Germany; RRID:AB_2560949) raised in rabbit was used to identify the amino acid glycine. Antibody specificity was tested by the manufacturer with a glycine-glutaraldehyde-protein in an ELISA test by crossreactivity experiments with b-alanine, aspartate, GABA, glutamate, and taurine. This showed low cross-reactivity. The antibody used here was previously used to identify glycinergic neurons in the auditory and vestibular system in guinea pig, Cavia porcellus (Peyret, Campistron, Geffard, & Aran, 1987) and in the brain of rats, Rattus rattus (Campistron, Buijs, & Geffard, 1986)   Beach, CA). The antibody manufacturer reports two bands on western blots of hybrid bass whole brain (exact species not provided on Millipore website): one between 25 and 37 kDalton (kDa) and one between 50 and 75 kDa. We found similar results in Gulf toadfish whole brain homogenates on western blots ( Figure 2). An additional band >250

| Connexin
kDa is likely the result of six connexin protein subunits coming together to form a single connexin hemi-channel.
According to the manufacturer, this antibody is specific toward its target in a variety of vertebrate species and has been used in studies of the Atlantic salmon, Salmon salar, (Sandbakken, Ebbesson, Stefansson, & Helvik, 2012), and sea lamprey (Villar-Cerviño et al., 2006). Additionally, the manufacturer states that the antibody serum does not react with 5-hydroxytryptophan, 5-hydroxyinodole-3-acetic acid and dopamine in cross-reactivity experiments.

| Tyrosine hydroxylase
We used a monoclonal antibody that was raised in mouse against tyro- For all primary and secondary antibodies, we tested for nonspecific binding of the antibody to the tissue by carrying out the staining protocol as above but omitting the secondary or primary antibody, respectively. All tests revealed no staining.

| Identification of the vocal CPG nuclei
Confirming the results of a prior study of Gulf toadfish (Chagnaud & Bass, 2014), transneuronal neurobiotin transport after neurobiotin appli- columns. VPP appears bilaterally as ventrolateral columns just rostral to VMN and VPN. Based on measurements in three animals, the rostralcaudal extent, number, diameter, and shape factor of neurobiotinlabeled neurons in the three CPG nuclei (Table 2) is consistent with values reported in a previous study (Chagnaud & Bass, 2014).

| GABAergic and glycinergic label in vocal CPG
Neurophysiological evidence supports the role of GABA in determining the temporal properties of the vocal CPG output (Chagnaud et al.,FIG URE 2 Connexin 35/36 antibody labeling in western blot of whole brain homogenate of Gulf toadfish. Bands are between 25 and 37 kilo Dalton (kDa) ladder marks, as well as between the 50 and 75 kDa ladder marks. Results are identical to the antibody manufacturer's results in hybrid bass (exact species not provided on Millipore website for antibody MAB3045), a different fish species than the one toward which the antibody was raised. The heavy band above the 250 kDa ladder mark likely represents the six connexin protein subunits binding together to form a connexin hemichannel    The results are summarized in Table 3. Table 3a presents a summary of the results for neuron diameter and shape in each fish; Table 3b,c (Table 3b; p < .0001 for all comparisons), except for Toadfish 3 that shows no significant difference in diameter between GABA-only labeled and neurobiotin-glycine co-labeled neurons (Table 3b). The shape factor of GABA-only neurons is significantly different from neurobiotin-only and glycine-only neurons, but not from neurobiotin-glycine co-labeled neurons in only the largest of the three toadfish examined (Table 3c).
While this might suggest an effect of fish size on neuronal dimensions, a much larger sample size would be needed to assess this possibility.
To determine whether the four VPN neuron groups display a topographic organization, we assessed their distributions throughout the rostral-caudal extent of VPN in one fish (Figure 5a

| Gap junction label in vocal CPG
Anatomical and neurophysiological evidence has led to the proposal that the robust transneuronal transport of neurobiotin throughout the vocal CPG depends, in whole or in part, on electrotonic coupling within the VPP-VPN-VMN network Bass et al., 1994;Chagnaud et al., 2011;Chagnaud et al., 2012). We investigated the distribution of gap junction proteins throughout the vocal CPG

| Serotonergic label in vocal CPG
Serotonin modulates the activity of a variety of motor systems including those responsible for vocal behavior (e.g., Wood, Lovell, Mello, and Perkel, 2011;Yu & Yamaguchi, 2010). Serotonergic innervation occurs throughout the vocal CPG (Figure 7). Label in the VMN is minimal (Figure 7a), except for the most rostral pole where it is prominent ( Figure   7b). Serotonergic label is abundant lateral to the VMN, amongst the VMN and VPN dendrites (e.g., arrow Figure 7c; see Figure 1f for orientation). In contrast to most of the VMN, label is found throughout VPN and VPP; in some cases, the label is directly over somata (insets, Figure   7f, g), although it is most prominent in adjacent regions (Figure 7f, g).

| Cholinergic label in vocal CPG nuclei
In addition to serotonin and catecholamines, we investigated choliner- Ellipsoid-shaped cholinergic somata were, however, within and in close proximity to larger neurobiotin-labeled VPN neurons (arrow, Figure 9g,

| Vocal CPG-auditory hindbrain pathway
A prominent neuroanatomical and neurophysiological feature of the vocal CPG is its link to rostral hindbrain auditory nuclei, in particular a medial division of the descending octovolateralis nucleus (DON, Figure   10a) that is a part of the ascending auditory system (Bass, Bodnar, & Marchaterre, 2000;Bass et al., 1994). Although transneuronally labeled DON neurons are not co-labeled with any of the neurotransmitters and modulators studied here, weakly labeled glycinergic neurons are immediately adjacent to transneuronally labeled DON neurons (arrows, Figure 10b-d).
Unlike prior studies in midshipman fish and toadfish Chagnaud & Bass, 2014), transneuronally labeled somata also lie within the rostral hindbrain inferior reticular formation (  ChAT label is also present in the vocal pacemaker (VPN, g-i) and vocal pre-pacemaker (VPP, j-l) nuclei. Asterisks in (i) and (l) indicate neuron shown in respective insets that highlight punctate-like label. Small, ellipsoid-shaped ChAT positive somata are adjacent to VPN neurons (arrow in g, i). Cells adjacent to VPP show weak label (asterisk and arrow in j, l; inset in j highlights labeled cell adjacent to asterisk). The scale bar represents 50 mm in (c) for (ac), 10 mm in (f) for (d-f), 10 mm in (i) for (g-i), 10 mm in (l) for (j-l), and 5 mm in insets in (i), (j), and (l) This is consistent with extensive gap junction protein (connexin) labeling in VPN. Gap junctions are also abundant in the motor neuron population (VMN), supporting the hypothesis that gap junction coupling is prominent in the VPN-VMN circuit ; see also Bass et al., 1994;Chagnaud et al., 2011;Chagnaud et al., 2012).
The weaker connexin label on premotor VPP somata suggests that transneuronal labeling of VPP is due to gap junction coupling between VPP axons and the somata and/or dendrites of VMN and/or VPN neurons. As discussed below, the evidence for catecholaminergic, serotonergic, and cholinergic label suggests a role for these neuromodulators along with inhibitory neurotransmitters and gap junctions in establishing the rhythmic, oscillatory-like output of the vocal CPG that is translated directly into the temporal features of vocal behavior.

| GABA
In line with a previous study of midshipman fish that used a GABA antibody shown to be specific in oyster toadfish, Opsanus tau (Holstein et al., 2004), GABAergic neurons are positioned lateral to VMN in the Gulf toadfish. These neurons are the likely source of the strong GABAergic input to VMN (Chagnaud et al., 2012). GABAergic neurons are also found within VPN and adjacent to VPP. None of the GABAergic neurons are co-labeled with neurobiotin, suggesting a lack of gap junction coupling to other vocal CPG neurons. Neurophysiological studies in midshipman fish are consistent with these findings. Injection of bicuculline, a competitive GABA A receptor antagonist, into VPP leads to an increase in call duration, the vocal parameter coded by VPP (Chagnaud et al., 2011). Intracellular recordings of VMN neurons together with local bicuculline injections into VMN show that GABAergic action at first distorts and then eliminates VMN activity, revealing that GABAergic inhibition is essential to generate vocal signals (Chagnaud et al., 2012).
Activation of GABAergic neurons that inhibit VMN might originate from within or outside of the vocal CPG. VPP is a well-suited candidate from within the vocal CPG, as VPP neurons fire just before and for the duration of the vocal behavior (Chagnaud et al., 2011). A candidate from outside of the vocal CPG would be the midbrain periaqueductal gray (PAG). The PAG activates the vocal CPG via direct input to the duration coding VPP neurons (Chagnaud et al., 2011;Goodson & Bass, 2002;Kittelberger, Land, & Bass, 2006) and itself may influence call duration (Kittelberger et al., 2006 (Bass & Baker, 1990;Chagnaud et al., 2011).
The oscillatory properties of VPN neurons would then be a network property of the neurobiotin-only labeled VPN population rather than an intrinsic property of these neurons. Electrophysiological recordings from VPN in midshipman fish strengthen this hypothesis as current injection into VPN neurons does not induce the characteristic firing frequency displayed during vocal activity (Chagnaud et al., 2012).  (Matsunaga, Kohsaka, & Nose, 2017), as well as in the Xenopus vocal pattern generator (Lawton, Perry, Yamaguchi, & Zornik, 2017).
Serotonergic input to all three vocal CPG nuclei suggests that serotonin can modulate one or more of the three main vocal parametersamplitude, frequency and duration, all of which remains to be tested using electrophysiology and behavioral assays. Serotonin is known to initiate vocal motor patterns in an isolated brain preparation of Xenopus laevis (Rhodes, Yu, & Yamaguchi, 2007). In contrast, systemically administered serotonin agonists terminate territorial calling in the Puerto Rican coquí frog, Eleutherodactylus coqui (Ten Eyck, 2008). While the influence of serotonin on vocalization could be species-dependent, these contrasting effects of serotonin in frogs might reflect methodological differences using bath application of serotonin in an isolated brain preparation versus a systemic injection in intact animals.
Serotonergic projections to vocal populations have also been observed for cat respiratory motor neurons involved in controlling airflow for vocalization (Holtman, 1988) as well as to the nucleus of the arcopallium in zebra finch (Wood et al., 2011). Similarly, X. laevis laryngeal pre-motor and motor neurons receive serotonergic input (Yu & Yamaguchi, 2010). Thus, serotoninergic input to vocal motor areas appears to be a conserved pattern among vocal vertebrates.

| Catecholamines
In accordance with previous studies in midshipman fish using different antibodies (Forlano et al., 2014;Goebrecht et al., 2014), we show dense catecholaminergic input to the vocal CPG in Gulf toadfish. Like midshipman fish (Forlano et al., 2014), neurons located dorsolateral to the VMN and within the area postrema likely provide catecholaminergic input to VMN neurons in Gulf toadfish. Additional catecholaminergic inputs from other neurons within the central nervous system, for example, spinal projecting dopaminergic neurons in the diencephalon as suggested for midshipman fish (Forlano et al., 2014), can, however, not be excluded.
Consistent with the anatomical evidence for catecholaminergic input to VMN, dopamine receptor subtypes are upregulated in midshipman VMN compared to the surrounding hindbrain (Feng, Fergus, & Bass, 2015). Catecholamine innervation of VMN is also denser in sneaker type II male compared to highly vocal, advertisement calling type I male midshipman fish, suggesting an inhibitory role of catecholamines Goebrecht et al., 2014; for behavior see also Brantley & Bass, 1994). A preliminary report shows that injection of dopamine into the PAG of midshipman suppresses vocal motor output, while dopamine receptor antagonists partially block this inhibition (Heisler & Kittelberger, 2012). Similarly, dopamine-like receptor activity negatively influences advertisement calling in green tree frogs, Hyla cinerea, after intraperitoneal injection of dopamine agonists (Creighton et al., 2013). Noradrenaline reduces activity in a brain slice preparation of the robust nucleus of the arcopallium, the premotor song control nucleus in songbirds, after bath application (Solis & Perkel, 2006

| Acetylcholine
The pattern of ChAT label observed for the Gulf toadfish matches that previously shown for the VMN of midshipman fish using a different antibody (Brantley & Bass, 1988). Like serotonin and catecholamines, there is cholinergic input to VPN and VPP. While injections of cholinergic antagonists into vocal midbrain sites of squirrel monkeys, Saimiri sciureus show no effect on vocal production (J€ urgens & Lu, 1993), in vivo and in vitro application of acetylcholine and its agonists to respiratory pacemaker-like neurons of the pre-B€ otzinger complex show increased duration, frequency and amplitude of spontaneous inspiratory bursts (Burton, Johnson, & Kazemi, 1997;Monteau, Morin, & Hilaire, 1990;Murakoshi, Suzue, & Tamai, 1985;Shao & Feldman, 2000). Perhaps acetylcholine plays a more widespread role in enhancing premotor activity in hindbrain pattern generators, including the vocal CPG of toadfish.

| Vocal CPG projections to brainstem areas
Like previous studies in midshipman fish and Gulf toadfish Chagnaud & Bass, 2014;Weeg, Land, & Bass, 2005), the vocal CPG is connected to auditory hindbrain neurons with evidence for nearby glycinergic neurons. Although previous electrophysiological experiments reveal that inhibition is important in determining vocal CPG output (earlier section), it is also known to play an important role in brainstem mechanisms of vocal-acoustic integration in mammals (Smotherman, 2007). Unexpectedly, neurobiotin-labeled cells are also located in the iRF in close proximity to glycinergic neurons. After carefully revisiting the material used by Chagnaud and Bass (2014), neurobiotin-labeled iRF neurons are also observed in that material, although the iRF labeling in that study was quite weak compared to the robust labeling that we observe here and so was easily missed.
GABAergic neurons are also found in the iRF of the African cichlid fish, Astatotilapia burtoni, that is also sonic (Maruska, Butler, Field, & Porter, 2017). Studies of squirrel monkeys demonstrate the reticular formation's involvement in vocal production (J€ urgens & Hage, 2007). The PAG and other vocal midbrain sites in midshipman are also connected to the reticular formation (Goodson & Bass, 2002;Kittelberger & Bass, 2013). While suggestive, the involvement of the reticular formation in vocal patterning or production in toadfishes awaits neurophysiological investigation.

| CON CL U DI N G R EM A R KS
In this study, we show prominent inhibitory and neuromodulatory input at all levels of the toadfish vocal CPG that suggest a suite of neurophysiological mechanisms to achieve a variety of motor programs (Gjorgjieva, Drion, & Marder, 2016;Harris-Warrick & Marder, 1991), with a single set of topographically separate nuclei-VMN, VPN, VPPresulting in context-dependent vocal signals. The presence of inhibitory transmitters such as GABA and glycine, along with gap junction coupling within each of the CPG nuclei, likely contribute to determining two predominant features of the vocal CPG-extreme temporal precision and synchrony (Chagnaud et al., 2012). How serotonin, catecholamines and acetylcholine interact with these transmitters to shape vocal production remains to be shown. The proposed evolutionarily conserved organization of vocal CPGs (Bass et al., 2008;Bass et al., 1994), together with the available neurophysiological evidence in toadfishes and other sonic species of vertebrates, suggests that comparable neurochemically dependent mechanisms are present in other vertebrate vocal CPGs.

ACKNOWLEDGMENTS
We thank Mario Wullimann for helpful comments on a previous ver-