Distribution of the cholinergic nuclei in the brain of the weakly electric fish, Apteronotus leptorhynchus: Implications for sensory processing

Acetylcholine acts as a neurotransmitter/neuromodulator of many central nervous system processes such as learning and memory, attention, motor control, and sensory processing. The present study describes the spatial distribution of cholinergic neurons throughout the brain of the weakly electric fish, Apteronotus leptorhynchus, using in situ hybridization of choline acetyltransferase mRNA. Distinct groups of cholinergic cells were observed in the telencephalon, diencephalon, mesencephalon, and hindbrain. These included cholinergic cell groups typically identified in other vertebrate brains, for example, motor neurons. Using both in vitro and ex vivo neuronal tracing methods, we identified two new cholinergic connections leading to novel hypotheses on their functional significance. Projections to the nucleus praeeminentialis (nP) arise from isthmic nuclei, possibly including the nucleus lateralis valvulae (nLV) and the isthmic nucleus (nI). The nP is a central component of all electrosensory feedback pathways to the electrosensory lateral line lobe (ELL). We have previously shown that some neurons in nP, TS, and tectum express muscarinic receptors. We hypothesize that, based on nLV/nI cell responses in other teleosts and isthmic connectivity in A. leptorhynchus, the isthmic connections to nP, TS, and tectum modulate responses to electrosensory and/or visual motion and, in particular, to looming/receding stimuli. In addition, we found that the octavolateral efferent (OE) nucleus is the likely source of cholinergic fibers innervating the ELL. In other teleosts, OE inhibits octavolateral hair cells during locomotion. In gymnotiform fish, OE may also act on the first central processing stage and, we hypothesize, implement corollary discharge modulation of electrosensory processing during locomotion.

in the nervous systems of invertebrates as well as vertebrates including mammals and it has attracted attention because it plays a pivotal role in fundamental brain processes from motor control to higher cognitive function (Hasselmo, 2006;Mooney et al., 2004;Sarter et al., 2005;Woolf & Butcher, 2011). ACh has also been shown to be a widespread modulator of information transmission in sensory systems (Fournier et al., 2004;Mooney et al., 2004;Woolf & Butcher, 2011). It controls receptive field properties and the gain of sensory responses, can enhance the signal-to-noise ratio of visual and auditory activity (Kimura, 2000;Sato et al., 1987;, and is involved in gustatory (Baldo et al., 2013;Hasegawa & Ogawa, 2007) and olfactory (Chan et al., 2017;de Almeida et al., 2016;Smith et al., 2015) processing. Importantly, the ACh projections from nucleus isthmi (nI) to tectum play a critical role in stimulus selection when multiple potential visual or auditory targets are present (Asadollahi et al., 2011;Maczko et al., 2006;Schmidt, 1995).
Choline acetyltransferase (ChAT) is the enzyme responsible for catalyzing the transfer of an acetyl group from acetyl-coenzyme A to choline, the step that forms ACh. As ChAT is exclusively expressed in ACh-synthesizing neurons, it has been used as a specific marker for studying the distribution and development of cholinergic neurons (see Oda, 1999, for review). The distribution of ChATcontaining cells has been described in several organisms from invertebrates to fish and mammals, including rats and humans (reviewed in Clemente et al., 2005;Ichikawa et al., 1997;Kasashima et al., 1998;Yasuyama & Salvaterra, 1999). These studies have contributed to the characterization of common organization and distribution of cholinergic neurons allowing the identification of homologous cell groups across taxa. Despite strong overall agreement between distribution patterns, some species-specific variations have been found Clemente et al., 2005;López et al., 2013;Morona et al., 2013).
Application of the cholinergic agonist carbachol led to increased excitability, burst firing, and shifts in frequency tuning of pyramidal cells in the first-order central processing stage of electrosensory inputs, the electrosensory lateral line lobe (ELL) of the hindbrain. These effects could be occluded by prior application of the muscarinic receptor antagonist atropine (Ellis et al., 2007). In a previous study, we mapped the distribution of muscarinic receptors in the brain of A.
The goal of the present study was to identify the sources of ACh in the brain of A. leptorhynchus, with a focus on the sources of ACh released in electrosensory areas. We mapped the distribution of ChAT messenger RNA and used retrograde labeling to identify cholinergic cell groups projecting to electrosensory nuclei previously shown to TOSCANO-M ARQUEZ ET AL. express muscarinic receptors (Toscano-Márquez, Dunn, & Krahe, 2013).

| Animals
We used weakly electric fish, A. leptorhynchus, of both sexes for this study (n = 20). The animals were obtained from a local supplier and maintained on a 12 h light cycle in tanks at a temperature of 26-28 C. All procedures were approved by the animal care committees of McGill University and the University of Ottawa. The ex vivo DiI tracing was approved by the Landesamt für Gesundheit und Soziales of Berlin, Germany (# T 0087/19).
As initial attempts to use commercial antibodies against ChAT failed, we decided to use an in situ hybridization approach to map the expression of ChAT-positive cell bodies.

| Cloning and sequencing of ChAT mRNA
Two fish were anesthetized with 500 ppm buffered 3-aminobenzoic acid ethyl ester (MS222; Sigma, St. Louis, MO) in cold water and ice.
The brains were rapidly removed and homogenized in TRI reagent solution (Ambion, Austin, TX) to obtain total RNA. cDNAs were generated using SuperScript III kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The resulting cDNA was used to clone a fragment of the ChAT sequence using a forward primer (5 0 -GTG TCV ACC TAY GAG AGY GC-3 0 ) and reverse primer (5 0 -GCR CTC TCR TAG GTB GAC AC-3 0 ) that produced a 730 bp-long product.
The ChAT sequence was isolated using a pair of primers designed to start and end the amplification in a conserved region of the ChAT

| In situ hybridization
We performed in situ hybridization as described in Toscano-Márquez, Dunn, and Krahe (2013

| Dextran-conjugated dye injection and microscopy
Brain slices were prepared as described in Harvey-Girard et al. (2010. Four fish were anesthetized with 0.2% MS222 in oxygenated water. Their brains were cut in a transverse manner in the rostral part of mesencephalon (level T 18-T 20; Maler et al., 1991), and the electrosensory nerves were sectioned without pulling brain tissue to avoid structural damage. The caudal brain portion was immersed in ice-cold artificial cerebrospinal fluid (aCSF containing in mM: 124 NaCl, 3 KCl, 0.75 KH 2 PO 4 , 2 CaCl 2 , 1.5 MgSO 4 , 24 NaHCO 3 , and 10 D-glucose), and embedded in 2.5% agarose/aCSF suspension.
Total of 500 μm thick transverse slices of rhombencephalon containing the ELL and the octavolateral efferent (OE) nucleus (sections T −4 to T −5 in the A. leptorhynchus brain atlas by Maler et al. (1991)), and of mesencephalon containing nucleus praeeminentialis (nP; sections T 4 to T 10 in Maler et al. (1991)) were sectioned on a vibratome and placed in an electrophysiology chamber at RT and perfused with oxygenated aCSF (95% O 2 and 5% CO 2 ).
The injection pipette was connected to a stimulation unit driven by a TTL pulse from an Axon Axoclamp 900A Amplifier (Molecular Devices, Sunnyvale, CA) using pClamp 10 (Molecular Devices).
Dextran-conjugated tetramethylrhodamine (10 mg/ml; Mini-Ruby, Life Technologies, Waltham, MA) in aCSF was electrically injected (pulses of 100 V lasting 400 ms) in ELL or nP using several pulses at several locations to maximize the injection volume in ELL and nP. Brain slices were perfused in oxygenated aCSF for 4 h to allow dextran transport in neurons. Slices were then fixed overnight in 4% PFA in PBS (pH 7.2). After rinsing twice for 10 min in PBS, the SeeDB clearing protocol (Ke et al., 2013)  Z-stack images of the thick brain slices were taken using AxioObserver.Z1 microscope (Zeiss). The fluorescent and bright-field images were acquired using AxioVision (Zeiss) on Windows 7. Fluorescent images were processed with the Fiji suite to reduce background.
Bright-field and fluorescent images had minor brightness adjustments to facilitate viewing. Z-stack images were transformed to 2D images using the maximum intensity z-projection method from the Image-J

| Ex vivo DiI tracing and 3D reconstructions
Brains from A. leptorhynchus were perfused with 4% PFA and collected as described above (N = 22). After 2-4 weeks postfixation at 4 C, the target brain regions where the dye was placed (i.e., ELL and nP) were exposed through vibratome sectioning (brains were sectioned until either target region was reached).
DiI (1,1 0 -dioctadecyl-3,3,3 0 ,3 0 -tetramethylindocarbocyanine perchlorate; product #42364, Sigma-Aldrich) and NeuroVue (NeuroVue F I G U R E 1 Legend on next page. (ca. 20 C) to allow the passive diffusion of the dye through fiber tracts, from the application site to the rest of the tissue. The 3D reconstruction of DiI-stained tracts and brain areas was carried out 2-4 weeks later, following previously described methods (Oboti et al., 2018). Briefly, dye-stained brains were cut using a vibratome in serial coronal sections (50 μm) and mounted on glass slides. Images from these sections were digitally acquired, postprocessed, and uploaded into ImageJ for alignment and registration (

| RESULTS
We partially cloned a segment of the ChAT gene from A.
leptorhynchus whole-brain cDNA using degenerate primers targeting a region that we found to be conserved in various teleosts. The PCR When compared by ClustalW alignment, the Apteronotus ChAT (aptChAT) amino acid sequence showed high homology with several known ChAT protein sequences. The predicted protein sequence had an amino acid similarity of 86.6% with its electric eel ortholog, 65.4% with zebrafish ChAT, and 60.7% with its murine ortholog (Figure 1(a)).
Amino acids known to lower enzyme activity and substrate specificity when mutated and residues that are thought to be important in the catalytic action (Cronin et al., 2000;Govindasamy et al., 2004;Ohno et al., 2001) were found to be perfectly conserved in aptChAT ( Figure 1(a); black triangles and asterisks). These results suggest conserved functional activity of ChAT between teleosts and mammals.
Phylogenetic analysis confirmed that the sequenced cDNA fragment is ChAT when we compared it with other ChAT homologs from a wide range of species (Figure 1(b)).

| Distribution of aptChAT transcripts
The expression of aptChAT transcripts in the brain of A. leptorhynchus was analyzed by in situ hybridization in transverse sections of the brain. The nomenclature used is based on Maler et al. (1991).
Intense positive signals showing expression of ChAT mRNA were restricted to the perikaryon of cells in various nuclei throughout the fish brain. The ChAT sense probe did not label any cells in spinal cord or brain (Figure 3(a)). We divide the description of the ChAT-positive nuclei in motor and nonmotor nuclei, and further divide the nonmotor nuclei according to their locations in the brain.

| Motor nuclei
Sections from the midbrain (tectal and isthmic areas), hindbrain, and medulla of A. leptorhynchus were examined. Labeled cells were found in several motor nuclei from rostral to more caudal levels.
The oculomotor nucleus (nIII) appears as an intensely labeled group of neurons in the most medial part of the mesencephalon ( Figure 2(a)). This nucleus shows similar appearance and location to that reported in other fishes, in which it has been described as cholinergic (Adrio et al., 2000;Anadón et al., 2000;Clemente et al., 2004;Mueller et al., 2004;Pérez et al., 2000). The trochlear nucleus (nIV), located close to the midline of the rhombencephalon, next to the ventricle, also showed strong staining (Figure 2(b)). The abducens nucleus was, for unknown reasons, not clearly detected in our samples.
Positive cells were also seen in the facial motor nucleus (nVIIm; Figure 3(a)) and the motor trigeminal nucleus (nVm; Figure 3(b)). Some ChAT-labeled cells with large soma were scattered in the paramedian reticular formation (PRF) area, lateral to the medial longitudinal fasciculus (MLF; Figure 3(b,c)). From here toward more caudal positions, the spinal motor column (smc) is seen throughout the medulla oblongata, as well as expression of ChAT transcripts in cranial nerve motor nuclei (Maler et al., 1991). The vagal motor nucleus (nXm) and possibly the glossopharyngeal nucleus (nIXm) appear continuous, as part of the spino-occipital motor neurons. They are located periventricular at different levels of the rhombencephalon in the transverse sections F I G U R E 1 Sequence alignment and phylogenetic tree of choline acetyltransferase. (a) Predicted amino acid sequence of Apteronotus leptorhynchus choline acetyltransferase (aptChAT) aligned against electric eel, zebrafish and mouse ChAT orthologs. Identical amino acids to aptChAT are shown as dots. Capitalized letters display the sequence location of the in situ RNA probe. The secondary structural elements are based on the structure of rat ChAT taken from Govindasamy et al. (2004). The sheets and helices are identified by solid and dashed horizontal arrows, respectively. Conserved amino acid residues, where deleterious point mutations have been identified for the rat ChAT, are identified by asterisks. Residues that are thought to be important in the catalytic action are identified by black triangles (Cronin et al., 2000;Govindasamy et al., 2004;Ohno et al., 2001). (b) Phylogenetic tree of the partial aptChAT protein sequence compared to protein sequences from 21 vertebrate species and common fruit fly as an outgroup. Numbers in the phylogeny tree provide confidence level for each branch. Species were chosen to include representatives of each major vertebrate clade. Teleostei are over-represented to classify aptChAT better  Figure 4(a,c)). Dorsolateral to this cell cluster, small cell bodies of the medial olfactory terminal field (MOTf; Sas et al., 1993) showed ChAT mRNA expression (Figure 4(a,b)).
In the preoptic region, we identified two groups of ChAT-positive cells. A first group was located in the posterior subdivision of the preoptic nucleus (PPp), which lies dorsal to the suprachiasmatic nucleus (Sc; Figure 4(d,e)). We also found positive staining for ChAT mRNA transcripts in a more posterior nucleus, dorsal to the Sc and ventral to the anterior hypothalamic nucleus (Ha), denoted as nucleus anterior periventricularis (Figure 4(f,g)). Interestingly, although in most other vertebrates the habenula has been observed to contain ChATpositive neurons, we did not find ChAT-positive cells in the habenula of A. leptorhynchus, as reported for other teleost species (Casini et al., 2018;Clemente et al., 2005;Ekström, 1987).

| Mesencephalon and rhombencephalon
Within the mesencephalon and the rhombencephalon, cells expressing ChAT transcripts are present in the optic tectum (TeO), the isthmic area, and the region of the OE nucleus.
In the TeO, ChAT-positive perikarya were found exclusively in the stratum periventriculare (SPV; Figure 2(a 1 ) and insert), where they were arranged in small clusters of three to four cells. The SPV of Apteronotus (Sas & Maler, 1986b) and other teleosts (Meek & Schellart, 1978) is densely packed with small neurons, and it is therefore clear that only a relatively limited subset was ChAT-positive. (a3) aptChAT mRNA labeling is densely present in the oculomotor nucleus (nIII). (b1) in a more caudal section, the strongly labeled trochlear nucleus (nIV) can be seen periventricular and medial to the lateral lemniscus (LL). Ventral to nIV, the superior reticular nucleus (SRn), a sub-nucleus of the reticular formation, is also labeled. In addition to the moderate number of cells labeled in nucleus isthmi (nI), two relatively densely stained nuclei are present in the isthmic area. The more medial and dorsal one corresponds to the nucleus lateralis valvulae (nLV), the second one may correspond to the medial subdivision of the perilemniscal nucleus (PLm), or a second branch of nLV; based on in situ hybridization only, we are unable to determine their exact identity. (b2) labeling in the isthmic area is also seen in a slightly more belong to the nucleus lateralis valvulae (nLV; Figure 2(b 1 )). Ventral to these cells lies a string of stained neurons that may be part of a second branch of nLV or the medial paralemniscal nucleus (PLm; shown to be cholinergic in zebrafish; Clemente et al., 2004;Mueller et al., 2004). Since our approach did not permit fiber staining, we cannot use connectivity as a criterion for a more conclusive identification of these cell clusters.
The last region in the midbrain showing ChAT mRNA expression consisted of a dense cluster of cells ventromedial of the lateral lemniscus (LL; Figure 2(b 1 )). It is likely that these cells belong to the superior reticular nucleus (SRn), which lies just ventromedial to the LL in zebrafish and has been shown to be cholinergic (Clemente et al., 2004;Wullimann et al., 1996). SRn has been suggested to be homologous to the cholinergic laterodorsal tegmental area in mammals .
In the rhombencephalon, cells expressing aptChAT transcript were found lateral to the MLF just under the ventricle, in the area where the OE has been described in other teleosts ( Figure 5(a); Danielson et al., 1988;Clemente et al., 2004). Labeling was also seen in large cells of the PRF at this level.

| Sources of ACh released in electrosensory nuclei
In previous work, we localized the expression of muscarinic receptors (mAChR) in A. leptorhynchus brain. The receptors were expressed in many nuclei throughout the brain including dorsal telencephalon and hypothalamic areas. Interestingly, various nuclei devoted to electrosensory information processing were positive for mAChR including the ELL, nP, TS, and TeO (see Figure 11 for summary; Toscano-Márquez, Dunn, & Krahe, 2013).
We sought to identify the source of cholinergic innervation arriving at two key electrosensory nuclei, nP and ELL. ELL is the first stage of central nervous processing of electrosensory information. It projects to nP, which provides feedback to ELL via several routes (Bell & Maler, 2005). Eurydendroid cells located at the ventral limit of the eminentia granularis posterior (EGp) had been suggested to be cholinergic (Maler et al., 1981;Sas & Maler, 1987) and to provide modulatory input to ELL via vertical fibers (Berman & Maler, 1999).
Interestingly, these eurydendroid cells turned out not to be labeled by our ChAT antisense probe ( Figure 5(b)), indicating that these cells are not cholinergic and that the cholinergic inputs to ELL come from another source, contrarily to what had previously been hypothesized (Maler et al., 1981). To determine the source of ACh to ELL we performed injections of dextran-conjugated tetramethylrhodamine (mini-Ruby 10.000 MW) in thick brain slices, because in vivo injections in ELL risk leakage into the dorsally adjoining EGp, which is heavily connected to many parts of the brain (Sas & Maler, 1987). The transverse slices included the sections T −4/−5 of the rhombencephalon as defined in the atlas of the A. leptorhynchus brain (Maler et al., 1991) using the rostral aspect of the pacemaker nucleus (not shown) to visualize the level containing OE. We found a bundle of labeled axons that started in OE and crossed the midline immediately after leaving OE and then followed a vertical trajectory to the site of injection in the contralateral ELL ( Figure 6(a-c)). Unfortunately, clear soma labeling in OE was not achieved. The lack of full soma labeling could be due to difficulties in keeping the slices alive long enough for the dye to fill the neurons sufficiently.
nP is a central feedback nucleus of the electrosensory system and showed a high density of muscarinic receptor mRNA expression (Toscano-Márquez, Dunn, & Krahe, 2013). Given the close proximity of nP and the isthmic region, which showed high expression of ChAT mRNA ( Figure 2(b 1,2 )), we suspected that it could be the source of  Tracing from injections in nP (Figure 8(a)) resulted in very intense staining in the areas surrounding the application site (Figure 8(b,c)), including nLV, reticular formation (RF), and LL, suggesting the presence of diffuse interconnections (possibly traced both anterogradely and retrogradely) between the nP and both the ipsilateral and contralateral mesencephalon. Several strongly labeled axons (empty arrowheads in Figure 8(d)) were found departing from the nP and directed toward the dLL, but no stained cell bodies were found ( Figure 8(d,e)). Due to the lesser section thickness (50 μm) allowing a clearer spatial separation of the isthmic nuclei, we were able to localize labeled somata more precisely along the rostrocaudal axis, approximately 700-800 μm from the targeted site (Figure 8(f)). Because the isthmic nuclei were localized in close proximity to the often strongly labeled injection sites positioned at this level (T 7-T 8), we performed an additional set of injections at more posterior levels (T 4, Figure 9(a-c)). To further limit the spread of the dye around the injected area, we used a different dye with lower local diffusion and faster tracing times (NeuroVue). In these experiments, dye-labeled cells were found again in the isthmic region, in correspondence of the nLV/PLm area (Figure 9(d-f)). Further, 3D brain reconstruction allowed to trace dye-stained axons across several consecutive and adjacent tissue sections ( Figure 10). This allowed to validate our results based on the presence, absence, and coherence of dye-stained tissue when belonging to the same pathway or brain region.
Only few labeled cell bodies were found, which were located in proximity of nI (lateral to the LL and ventral to the TSd; Figure 8

| DISCUSSION
The aim of this study was to determine the distribution of cholinergic cells in the brain of the weakly electric fish, A. leptorhynchus. We first identified ChAT mRNA of A. leptorhynchus and then used an in situ hybridization approach to label the cholinergic cells in transverse sections of the brain. This approach combined with injections of retrograde tracer in electrosensory nuclei allowed us to track some of the sources of ACh released in electrosensory nuclei.
As expected, we found ChAT expression in cells belonging to the spinal motor column and the cranial nerve motor nuclei (Figures 2 and   3). In addition, we identified previously undescribed projections from the OE to the ELL, the sole recipient of electrosensory afferents in the hindbrain. We were able to retrogradely label from the nP putative following, we will first briefly compare the sequence data for aptChAT with ChAT sequences known from other species and then place our anatomical findings in their larger context of brain connectivity and possible function.

| Comparison of aptChAT with ChAT sequences from other species
The amino acid sequence of aptChAT is highly conserved between mammals, other teleosts, and A. leptorhynchus. Functional residues identified based on the mouse ChAT (mChAT) are known to be part of the acetyl-CoA binding site (Govindasamy et al., 2004). Specifically, point mutations in mChAT Leu102, Pro103, Ile197, and Arg312, which lead to decreased affinity to acetyl-CoA, and point mutations at Glu441, which lack catalytic activity, have been linked to congenital myasthenic syndrome with fatal episodes of apnea in humans (Ohno et al., 2001). All of these identified point mutation sites are conserved in aptChAT as well as the other teleosts included in our comparison (Figure 1(a)). The His334 residue that has been shown to be the catalytic site of ChAT (Govindasamy et al., 2004) was apparently lost in apteronotid and zebrafish orthologs, but is present in electric eel.
Other residues, such as Arg250, Asn311, Arg312, and Trp313 that are hypothesized to interact with the catalytic site of ChAT, were found to be conserved in aptChAT. Finally, four amino acid residues involved in the choline/carnitine binding specificity, VDN453-455 and E441, are also conserved (Cronin et al., 2000). All these results suggest a strong conservation of functional features of ChAT activity in all vertebrates.

| ChAT-positive elements in the forebrain of gymnotiform weakly electric fish
In the light of recent studies that revealed several homologies between pallial and subpallial areas in fish and mammals (Briscoe & Ragsdale, 2019;Ganz et al., 2012;Giassi, Duarte, et al., 2012;Giassi, Harvey-Girard, et al., 2012;Harvey-Girard et al., 2012;Nieuwenhuys, 2009;, it is worth mentioning that, in areas ascribed to the dorsal telencephalon (Giassi, Duarte, et al., 2012;Giassi, Harvey-Girard, et al., 2012;Harvey-Girard et al., 2012), no ChATpositive cells were detected. This situation is similar to that described for the pallium of most other fish species (Lopez et al., 2012;López et al., 2013;Morona et al., 2013;Mueller et al., 2004). As in these studies, cholinergic cells were restricted to the ventral telencephalic areas of A. leptorhynchus, where we found small groups of ChATpositive cells only in the VVv and the MOTf area (Figures 4 and 11). The VVv has been identified as a septal region (Ganz et al., 2012;Mueller et al., 2008;Wong, 1997). connected with the preoptic area and hypothalamic nuclei (Wong, 1997). Moreover, in the gymnotiform weakly electric fish, Eigenmannia virescens, VVv connects with nucleus electrosensorius (nE), a nucleus known to be critical to processing of electrocommunication signals (Heiligenberg et al., 1991;Wong, 1997). All the known targets of VVv express at least one type of muscarinic receptor in A. leptorhynchus (Figure 11; Toscano-Márquez, Dunn, & Krahe, 2013). Thus, our present results confirm VVv as a septal region and suggest that it modulates detection of, and responses to, electrocommunication signals via muscarinic receptors.

| Cholinergic neurons of the isthmic region
The most prominent cholinergic population in the brain of A.
leptorhynchus was located in the mesencephalic isthmic region, ventral to TSd and medial to nP and LL (Figure 2(b 1 ,b 2 )). This area has been found to contain very high ChAT levels also in other teleosts (Clemente et al., 2004;Contestabile et al., 1979;Mueller et al., 2004;Pérez et al., 2000). It comprises several nuclei including the nI, nLV, and the medial paralemniscal nucleus (PLm). The cholinergic part of nI in fish and birds is homologous to the parabigeminal nucleus of mammals (Diamond et al., 1992;Wang et al., 2006;Xue et al., 2001). It is located ventral to the TSd and dorsomedial to the nP (Sas & Maler, 1986a), but varies greatly in size in different species (Pérez et al., 2000;Zottoli et al., 1988). The nI is considered to be a visual center in non-electroreceptive teleosts (Northmore, 1991;Wang, 2003). It is interconnected with the tectal formation by feedback loops, which can involve both the ipsi-and the contralateral tectum, depending on the species. The nI has been functionally implicated in the detection of looming stimuli in goldfish and other species (Gallagher & Northmore, 2006), while studies con- The nLV has been considered to be a relay of sensory information to the cerebellum (Yang et al., 2004). The combination of our in situ hybridization and tracing experiments suggests the presence of cholinergic afferents to the nP originating in the nLV (Figures 2(b) and 9(a,   b)). In teleost fishes, nLV receives lateral line inputs as well as retinotopic and somatotopic inputs from trigeminal nuclei (Ito & Yoshimoto, 1990;Xue et al., 2005). It relays sensory information mainly to the valvula and corpus cerebelli including eurydendroid cells in tilapia, but reciprocal projections to the inferior lobe have also been shown (Ahrens & Wullimann, 2002). Although not much is known about the function of nLV, it has been suggested to be homologous to some of the pontomesencephalic nuclei in mammals based on its position and connections (Imura et al., 2003;Oda, 1999). These nuclei play an important role in arousal as part of the reticular activating system, cardiovascular and respiratory control, and modulation mechanisms controlling gait and posture (Jones, 2003;Martinez-Gonzalez et al., 2011).
However, recent studies on the connectivity and neurochemical phenotype of the cells of the isthmic region have suggested that the identity of nLV as a cholinergic nucleus should be revisited, and the cholinergic cells of that region should be considered instead part of the secondary gustatory and visceral nucleus (SG/V), which has previously been shown to be cholinergic (Castro et al., 2006;Mueller et al., 2004;Yáñez et al., 2017). The SG/V area is very prominent in teleosts and consistently found to send cholinergic projections coursing through the ventral midbrain to the RF, thalamus, and hypothalamus Pérez et al., 2000;Yáñez et al., 2017).
Nonetheless, both, because the exact location of the SG/V in the brain of A. leptorhynchus is still not clear, but also due to the debated presence of cholinergic neurons in the nLV (Castro et al., 2006;Clemente et al., 2004;Mueller et al., 2004), we have decided to assign While most of the stained axons seem to progress rostrally toward the nI (a-c), several isolated DiI-stained processes were found in proximity to the injection site directed toward the dLL (arrowheads in d). However, no DiI-stained somata were found at this level (T 4, d) or at the level of the tLT (T 7, e), whereas a few DiI-stained cells were visible in the nI (T 12, f). The panels on the upper right show the localization of dye application (orange). The arrows indicate the location of the application sites and the direction of dye transport. The areas in which labeled cells were found are shown in red. Scale bars: 500 μm (diagram), 100 μm (a-f), 15 μm (details of double positive cell in f). Since image acquisition was set to detect the minimum labeling intensity in the specimen and kept constant across all section, the dye application sites appear overexposed the neurons labeled in our tracing experiments to the nLV area, designated according to published anatomical references (Maler et al., 1991). In addition, we cannot exclude that in our tracing experiments targeting the nP we have retrogradely labeled fibers of passage belonging to any of these circuits, and therefore not projecting to nP. Future tracing studies will be required to further subdivide the isthmic nuclei in the light of these comparative observations, in order to precisely delineate their boundaries.

| Tectal cholinergic neurons
As in most vertebrates, the TeO plays a significant role in the integration of visual and other sensory inputs in the teleost mesencephalon (Bastian, 1982;Heiligenberg & Rose, 1987;Perez-Perez et al., 2003).
These cholinergic neurons extend their dendrites and axons throughout several tectal layers, innervating cells in the outer layer of the stratum opticum and, to a lesser extent, cells in deeper layers of the TeO . In addition to these intrinsic cholinergic neurons, injections in the preglomerular nucleus (PG) of A. leptorhynchus revealed the presence of extrinsic efferents, retrogradely labeled in the SPV (Giassi, Duarte, et al., 2012). The morphology of these cells is similar to that of neurons immunolabeled with ChAT antibodies in two holostean fish species . Thus, although in other teleosts cholinergic SPV neurons have only been found to form intrinsic circuitry within TeO, our results together with those of Giassi, Duarte, et al. (2012) suggest that they also send cholinergic projections to PG, which we previously found to express muscarinic receptor mRNA densely ( Figure 11; Toscano-Márquez, Dunn, & Krahe, 2013).
SPV (Heiligenberg & Rose, 1987) and PG (Wallach et al., 2018) neurons are known to respond strongly to electrosensory and visual object motion. Direct studies of the role of ACh in modulation of tectal and PG responses to visual and electrosensory motion stimuli will be required for a deeper understanding of how this transmitter regulates the encoding of motion stimuli (see also Bastian, 1982).
In our previous study, we found expression of mAChR2 and F I G U R E 1 1 Sagittal brain representation summarizing the nonmotor elements of the cholinergic system in Apteronotus leptorhynchus brain. The nonmotor ChAT-containing nuclei that we found in this study are shown in red. Muscarinic receptor mRNA-containing nuclei are shown in different shades of blue (Toscano-Márquez, Dunn, & Krahe, 2013). Dark blue: muscarinic receptor 3 (M3); blue: muscarinic receptor 2 (M2); light blue: muscarinic receptor 4 (M4). Red arrows show the putative connections from the cholinergic nuclei to the muscarinic receptor-containing nuclei inferred from previous publications (Ahrens & Wullimann, 2002;Asadollahi et al., 2010;Sas & Maler, 1987 (e,f)), demonstrating a hitherto unknown input to the nP. Since the nP is tightly interconnected with the first central processing station of the electrosensory system, the ELL (Bastian et al., 2004;Maler et al., 1982;Sas & Maler, 1983), this cholinergic input could be acting indirectly on ELL as part of a previously hypothesized sensory searchlight mechanism (Bastian, 1999) and enhance the responses to object motion (Clarke & Maler, 2017).
Further physiological experiments will be critical to understanding the role of this cholinergic pathway on electrosensory processing. Injection of muscarinic agonists and antagonists into nP while recording from ELL pyramidal neurons would help elucidate the possible role of this modulatory input on the detection and processing of electrolocation stimuli.

| Cholinergic input to the ELL: The OE nucleus
Muscarinic ACh receptors are also present in the ELL of A. leptorhynchus ( Figure 11; Phan & Maler, 1983;Toscano-Márquez, Dunn, & Krahe, 2013). Previous work suggested that the corresponding cholinergic input might arise from vertical fibers emanating from eurydendroid cells of the cerebellum (Maler et al., 1981). Even though some studies have found evidence for ChAT-positive cells in cerebellum in fish  and mammals (Delacalle et al., 1993;Ikeda et al., 1991), our study is in agreement with most other work on teleosts and amphibians (see Morona et al., 2013, for a review) that showed the cerebellum to be devoid of cholinergic cells ( Figure 5(b)). The cholinergic innervation in these cases has been suspected to derive from the reticular nucleus and the octavolateral area . The presence of cholinergic cells in both areas has been consistently reported as a conserved feature in fish and other vertebrates (Clemente et al., 2004;López et al., 2013;Morona et al., 2013). In this study, we found a small population of ChAT-positive cells in the RF close to the MLF as well as a prominent paramedian population on either side of the MLF in the octavolateral area corresponding to the OE in other fish (Figures 5(a) and 6(a); Clemente et al., 2004;Tomchik & Lu, 2006).
The injections of dextran-conjugated tetramethylrhodamine into the molecular layer of the ELL terminated in the contralateral OE area ( Figure 6(c,d)). As no projections from ELL toward OE have ever been found (Krahe & Maler, 2014), we hypothesize that the fibers observed here provide cholinergic input from OE to ELL. This is confirmed by our DiI tracing experiments, showing retrogradely labeled cell bodies in the OE (Figure 7(e)). Projections from OE to EGp have already been described (Sas & Maler, 1987), which raises the possibility of the connections to ELL being collaterals of the projection to EGp. This latter connection is further supported by the presence of M3 receptors in eurydendroid cells in EGp (Figure 11; Toscano-Márquez, Dunn, & Krahe, 2013). Therefore, we hypothesize that OE is involved in direct cholinergic modulation of ELL processing as well as in indirect effects via modulation of cerebellar feedback to ELL.
Besides the well characterized role of the OE in the modulation of sensory organ activity (Katz et al., 2011;Köppl, 2011), to our knowledge, this may constitute the first example of OE modulation acting on a central nucleus. Further anatomical and physiological characterization of this cholinergic nucleus is required to confirm this hypothesis and further understand the functional and evolutionary implications of this finding. An additional interesting difference to previously described OE action is that the cholinergic effects in ELL are mediated by muscarinic ACh receptors and not by nicotinic receptors as in the other octavolateral systems (Katz et al., 2011;Köppl, 2011).

| CONCLUDING REMARKS
Overall, the spatial distribution of cholinergic neurons in the brain of the brown ghost knifefish, A. leptorhynchus, agrees with earlier descriptions in other teleosts and tetrapods. Connections between previously described muscarinic receptor-containing nuclei and cholinergic nuclei have been characterized in other studies in electric fish (Berman & Maler, 1999;Sas & Maler, 1987) or can be inferred from previous cholinergic pathways in other teleosts based on homology ( Figure 11; Ahrens & Wullimann, 2002;Asadollahi et al., 2010). In addition, the present study describes a new cholinergic input from the isthmic region (nI, nLV, and/or SG/V) to nP and from OE to ELL. These results suggest hitherto unrecognized additional levels of multisensory integration and control of electrosensory processing ( Figure 11). Further targeted physiological experiments and connectivity analyses are required to explore the role of these pathways in behaving fish and to determine if these mechanisms evolved uniquely in gymnotiform weakly electric fishes or whether they are a general feature of the control of sensory information processing.

ACKNOWLEDGMENTS
Thanks to Anh-Tuan Trinh for his help with SeeDB processing of the retrogradely labeled brain sections, and to William Ellis, Sofia Ibarraran, and Ina Seuffert for technical assistance. Martin Cuddy helped with the partial cloning of aptChAT. The authors are grateful to Ehab Abouheif and François Fagotto for access to their lab equipment. This work was supported by Natural Sciences and Engineering Research Council of Canada (NSERC);