Organization of projection from brainstem auditory nuclei to the inferior colliculus of Japanese house bat (Pipistrellus abramus)

Abstract Objectives Echolocating bats show remarkable specialization which is related to analysis of echoes of biosonars in subcortical auditory brainstem pathways. The inferior colliculus (IC) receives inputs from all lower brainstem auditory nuclei, i.e., cochlear nuclei, nuclei of the lateral lemniscus, and superior olivary complex, and create de novo responses to sound, which is considered crucial for echolocation. Inside the central nucleus of the IC (ICC), small domains which receive specific combination of extrinsic inputs are the basis of integration of sound information. In addition to extrinsic inputs, each domain is interconnected by local IC neurons but the cell types related to the interconnection are not well‐understood. The primary objective of the current study is to examine whether the ascending inputs are reorganized and terminate in microdomains inside the ICC. Methods We made injection of a retrograde tracer into different parts of the ICC, and analyzed distribution of retrogradely labeled cells in the auditory brainstem of Japanese house bat (Pipistrellus abramus). Results Pattern of ascending projections from brainstem nuclei was similar to other bat species. Percentages of labeled cells in several nuclei were correlated each other. Furthermore, within the IC, we identified that large GABAergic (LG) and glutamatergic neurons made long‐range connection. Conclusions Synaptic organization of IC of Japanese house bat shows specialization which is likely to relate for echolocation. Input nuclei to the IC make clusters which terminate in specific part of the ICC, implying the presence of microdomains. LG neurons have roles for binding IC microdomains.


| INTRODUC TI ON
Echolocating bats emit sound pulses with specific acoustical features to targets, and measure the distance to the target according to the echoes of the sound pulses. To analyze the echoes and pulses themselves, auditory nuclei of bats are remarkably hypertrophied (Zook & Casseday, 1982a), and show specialization related to echolocation (Covey & Casseday, 1995), which is obvious in the medial superior olive (MSO) of the superior olivary complex (SOC) (Casseday, Covey, & Vater, 1988;Covey, Vater, & Casseday, 1991;Grothe, Park, & Schuller, 1997;Grothe, Schweizer, Pollak, Schuller, & Rosemann, 1994) and monaural nuclei of the lateral lemniscus, namely intermediate nucleus (ILL) and compact and multipolar cell divisions of ventral nucleus (VLLc and VLLm) . Lower brainstem nuclei, namely, cochlear nuclei (CN), SOC, and nuclei of the lateral lemniscus (NLL), analyze specific features of sound and send the analyzed information to the inferior colliculus (IC), which is the largest nuclei in entire subcortical auditory system (Ito, Bishop, & Oliver, 2016). Since all the sound information is converged, the IC is considered to integrate sound information and create coding of complex sound information: In bats, analyses related to echolocation such as tunings of sound duration (Casseday, Ehrlich, & Covey, 1994), and delay coding (Portfors & Wenstrup, 1999) are found in the IC.
It has been proposed that IC local circuit is composed of multiple small domains, i.e., "synaptic domains", which receive specific combination of extrinsic inputs (Cant & Benson, 2006;Casseday, Fremouw, & Covey, 2002;Loftus, Bishop, & Oliver, 2010;Oliver, 2005), and contain three cell types, i.e., large GABAergic (LG), small GABAergic (SG), and glutamatergic (GLU) cells, that have different input and output patterns (Ito & Oliver, 2012). Neurons in a synaptic domain likely to make influence on different synaptic domains since most IC neurons shown to have local collaterals (Oliver, Kuwada, Yin, Haberly, & Henkel, 1991;Wallace, Shackleton, & Palmer, 2012), although the knowledge about cell types that influence on other synaptic domains is poorly known. As the three cell types organize different patterns of neuronal connection, we hypothesized that specific cell types are involved in connecting different synaptic domains in the IC.
In this study, we made injection of fluorogold (FG), a retrograde tracer, into different parts of the ICC, and analyzed distribution of retrogradely labeled cells in the entire auditory brainstem pathways of Japanese house bat (Pipistrellus abramus), which uses frequency modulated (FM) pulses for echolocation, to examine whether the ascending inputs are reorganized and terminate in small domains inside the ICC. Furthermore, by analyzing the distribution of retrogradely labeled cells inside the IC, we identified cell types that make long-range intrinsic connection inside the IC. Although this species have been used frequently on biosonar studies (Fujioka et al., 2014;Sumiya, Fujioka, Motoi, Kondo, & Hiryu, 2017;Takahashi et al., 2014;Yamada, Hiryu, & Watanabe, 2016), the information about the organization of auditory pathways of this species has not been reported except for a single unit recording study from the IC (Goto, Hiryu, & Riquimaroux, 2010). Therefore, our study will give new neuroethological insights on biosonar.

| Subjects
Japanese house bats (Pipistrellus abramus) were used in this study.
P. abramus is a member of the Vespertilionidae and is commonly found in Japan. The bats were captured from a large colony roosting on bridge girders near the campus of Doshisha University, Kyotanabe, Japan. The bats were kept in a humidity-and temperature-controlled rearing cage, and were allowed free access to food (mealworms) and water. Experiments were conducted on seven adult bats of both sexes. All experiments were conducted in accordance with institutional guidelines at the University of Fukui and the Doshisha University. All efforts were made to minimize the number of animals used and their suffering.

| Surgery
Bats were anesthetized with isoflurane mixed with oxygen for the duration of the surgery. Skin covering the skull was cut at the midline. A small screw was fixed at Bregma with instant glue, and the animal head was fixed by holding the screw. A craniotomy opened the right parietal bone to expose the IC. Glass micropipettes (4-7 μm tip) were filled with 3% FG (Fluorochrome, Denver, CO, USA) diluted in saline. FG was iontophoretically injected from the recording pipette with a 2-5 μA positive current for durations of 2-5 min with a 50% duty cycle (7 s on/7 s off) using a constant current injector (BCC-3621; Bio-Medica, Osaka, Japan). Twenty-four hours after the surgery, the bats were deeply anesthetized with pentobarbital sodium (64.8 mg/kg, i.p.) and perfused transcardially with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The brains were dissected out, and immersed in the same fixative overnight. After cryoprotection with 30% sucrose in 0.1 M phosphate buffer (pH 7.4) at 4°C for 1-2 days, brains were cut at the level of superior colliculus (SC), and 30 μm-thick serial coronal sections of the brainstem were cut with a freezing microtome. Every 5th section was used for histology. Three sets of every 5th section was used and processed for Nissl staining, bright-field immunostaining for FG, and fluorescent immunostaining for FG, glutamic acid decarboxylase 67 (GAD67), and vesicular glutamate transporter 2 (VGLUT2). We found that the angle of sectioning differed among animals, and we estimated the angle from the 3D reconstruction of the brainstem (shown in top of Figures 3-5).

| Antibodies
In this study, mouse monoclonal antibody for GAD67 (antiserum MAB5406; Merck, Darmstadt, Germany, RRID: AB_2278725) and rabbit polyclonal antibodies for VGLUT2 (Fujiyama, Furuta, & Kaneko, 2001) and FG (AB153; Merck, RRID: AB_2632408) were used. Specificity of the antibody for FG was confirmed by absence of staining in bat brain which did not receive injection of FG. The immunogen used for anti-GAD67 antibody was a recombinant whole protein of rat GAD67. We tested homology of rat GAD67 by using Protein BLAST (https://blast.ncbi.nlm.nih.gov/), and found that the amino acid sequence of rat GAD67 protein (Accession No: NP_058703.1) shared 96% identity to predicted sequence of GAD67 of big brown bat (Eptesicus fuscus) (Accession: XP_008136804.1), which is another member of the Vespertilionidae. In a previous study (Ito et al., 2007), we confirmed the specificity of the anti-GAD67 by western blot and an absorption test. In the western blot, a single band around 67 kDa was detected (which is also described in manufacturer's sheet). No signal was detected in rat brain sections incubated with the antibody (1:3,000) preabsorbed with recombinant rat GAD67 protein (180 μg/ml). Furthermore, the staining pattern of the antibody was consistent with that in a previous study (Vater, Kossl, & Horn, 1992).
The immunogen used for anti-VGLUT2 antibody was a synthetic peptide to C-terminal amino acid residues CWPNGWEKKEEFVQESAQDAYSYKDRDDYS of rat VGLUT2 (aa. 554-582), which has 79% similarity in Protein BLAST to the corresponding part of predicted sequence of VGLUT2 of big brown bat (Accession: XP_008147186.1). Specificity for the antibody was characterized by western blot and an absorption test (Fujiyama et al., 2001). In the western blot, a single band around 62 kDa was detected. No signal was detected in rat brain sections incubated with antibody preabsorbed with 10,000-fold (in mol) excess amounts of the synthetic peptide used for the immunization.

| Immunohistochemistry
For bright-field microscopy, sections were incubated with the rabbit anti-FG antibody (1:8,000) diluted in 0.3% Triton X-100 with 1% normal goat serum in 0.05 M phosphate-buffered saline (PBS-XG), followed by an incubation in a biotinylated goat anti-rabbit sec-

| Imaging
Bright field micrographs were acquired with an upright microscope (AX80; Olympus, Tokyo, Japan) equipped with ×4 and ×10 lenses (NA = 0.16 and 0.4, respectively) and a digital camera (DMC-GH3; Panasonic, Osaka, Japan). To compensate non-uniform illumination, a blank image was acquired, and subtracted from original microscopic images.

| Data analysis
We plotted the FG-positive (+) cells in the entire auditory brainstem nuclei using a bright-field microscope (BX50; Olympus) equipped with a motorized stage (Bioprecision, Ludl Electronic Products, Hawthorne, NY, USA) that is controlled by Neurolucida (MBF Bioscience, Williston, VT, USA). Fluorescent montage images of the IC were imported to Neurolucida and the distribution of FG+ GABAergic neurons were analyzed. Injection sites were defined as the location where the FG immunoreactivity was present in extracellular space, and the center of injection was defined as the site of lesion caused by the current injection. The boundary of auditory nuclei was identified by cytoarchitecture revealed by Nissl counterstain or differential interference contrast optics.
The Neurolucida traces were exported and further analyzed with a custom-made script working on MATLAB. Shrinkage correction was not performed.
To make quantitative analysis of distribution of FG+ GABAergic and non-GABAergic neurons around the injection site and ipsilateral IC, distance from the center of injection was calculated for every cell, and histograms of the distance with a bin width of 100 μm were created. In each bin, the ratio of SG and LG neurons to total labeled cells was calculated, and the relationship between distance and ratio was compared between LG and SG cells using 2-way analysis of variance (ANOVA).

| Sites of injection
In all seven cases, the injection of FG was made inside the IC, and we analyzed the distribution of FG+ in the local circuit of the hemilateral IC that received injection (referred as "ipsilateral IC" in the following). Since two cases showed leakage of FG injection to the periaqueductal gray, we excluded the cases from the analysis of ascending input to the IC from lower auditory brainstem nuclei. Remaining five cases showed no leakage of FG injection outside the IC (Figure 1a-e), and we analyzed these cases for further analysis about the ascending projection to the IC.
Three-dimensional reconstruction of the ipsilateral IC (dark gray area in Figure 1f-j) revealed that the injection site (blue area), which is defined as the diffusion of FG in the extracellular space, was restricted in the central nucleus of the IC (ICC; light gray area) in 4 cases (Pa119, Pa125, Pa114, and Pa128). In the ipsilateral IC of these cases, majority of FG+ IC cell bodies (red dots in To obtain objective classification of injection sites, we normalized the location of injection sites, since amount of shrinkage during histology seemed to be different among cases. We measured medio-lateral distance between medial edge of the IC and injection center and distance between medial and lateral edges of the IC, and calculated the medio-lateral ratio between them.
Since the angle of sectioning is different among cases (top drawings in Figures 3-5), we used three-dimensional reconstruction to measure dorso-medial and rostro-caudal ratios. After the 3D models were rotated, ratio of distance between dorsal surface of the IC and injection center to distance between dorsal surface and ventral edge of the IC was calculated as dorso-ventral ratio. The dorso-ventral ratios of Pa119, Pa125, Pa131, Pa114, and Pa 128 were 0.600, 0.450, 0.203, 0.282, and 0.331, respectively. In a single unit recording study from Japanese house bat IC (Goto et al., 2010), dorso-ventral length of the IC was around 2.0 mm, and region shallower than 1.0 mm (dorso-ventral ratio <0.5) showed best frequency lower than 40 kHz, which is the quasi constant frequency of the sonar pulse, and region deeper than 1.0 mm (dorsoventral ratio >0.5) did not contain unit with best frequency lower than 40 kHz. Therefore, we assigned Pa119 as high frequency injection case, and other cases as low frequency injection cases. To calculate rostro-caudal ratio, distance between rostral end of the IC to the injection center was divided by distance between rostral and caudal ends of the IC. The rostro-caudal ratios of Pa119, Labeled cells that were found close to the injection site were likely to take up the tracer directly from the dendrites and cell bodies as well as axonal terminals, and therefore we cannot distinguish retrogradely labeled cells and direct labeled cells. On the other hand, labeled cells of which distal end of dendritic trees was outside the diffusion of tracer are considered as retrogradely labeled cells. In all the seven cases, the radius of tracer diffusion was ranged from 88.1 to 177.9 μm. In Japanese house bats, the maximal length of a dendritic tree was approximately 170 μm, which was measured from Golgi-stained materials (not shown). In rats, the maximal width of dendritic arborization is smaller than 400 μm (Malmierca, Blackstad, Osen, Karagulle, & Molowny, 1993). Therefore, if the cell bodies of labeled neurons were more than 200 μm distant from the edge of tracer diffusion, they are regarded as retrogradely labeled cells.
Finally, in the current samples, labeled cells which were more than 400 μm distant from the center of the injection were regarded as retrogradely labeled. A few cells were found in a nearly symmetrical manner in the contralateral IC (Figures 3-5). More labeled cells were found in the medial injection case (Pa128). In the dorsal nucleus of the lateral lemniscus (DLL), FG+ cells were found bilaterally (Figures 2a and   3-5). In the large injection case (Pa119) (Figure 3a), FG+ cells were distributed throughout the nucleus, while the other dorsal injection cases, FG+ cells were distributed in the lateral part of the nucleus (Figures 3-5). In monaural nuclei of the lateral lemniscus, i.e., ILL, VLLc, and VLLm, labeling was found only in the ipsilateral side (Figures 2b and 3-5). In Pa119, FG+ cells were found mainly in the center of the nuclei (Figure 3a), while in the other cases, FG+ cells were found in the periphery of the nuclei (Figures 3-5).

| Distribution of retrogradely labeled cells in brainstem auditory nuclei
In the SOC, labeled cells were found bilaterally. Inside the SOC

| Distribution of FG+ cells inside the ipsilateral IC
In the ipsilateral IC, FG+ cells were found not only around the injec-

| D ISCUSS I ON
In this study, we showed the synaptic organization of ascending and local inputs to the IC of Japanese house bat. Distribution of retrogradely labeled cells inside auditory nuclei was related to the dorso-ventral location of tracer injection, suggesting tonotopic organization in many brainstem auditory nuclei. We also found that composition of brainstem input sources is related to the location inside the ICC. These findings suggest that ICC is further subdivided into small domains which are specialized for certain functions. Furthermore, analysis of local connection revealed that the small domains are interconnected by LG and GLU cells.

| Suggested tonotopic organization
Although we did not perform physiological recording to confirm the tonotopicity of the IC, according to the previous study which determined the tonotopic organization of the IC of Japanese house bat (Goto et al., 2010), the injection site of Pa119 was higher frequency region (>20 kHz), and those of the other cases were lower frequency.
Based on the tonotopy of the IC and tonotopic organization of other mammalian species including bats, in the following we suggest the tonotopic organization of auditory brainstem nuclei of the Japanese house bat. Note that the number of samples was small and intersample variations, it is difficult to determine the tonotopic organization in some nuclei.
In mustache bat (Pteronotus parnelli), which uses echolocation pulses composing of constant frequency followed by FM (CF-FM), origins of ascending inputs to the IC were described in detail (Zook & Casseday, 1982b). In the study, the authors showed topographical organization of inputs to the IC. From the projection pattern to IC region with identified best frequency, tonotopic organization of brainstem nuclei is suggested, and consistent with unit recording study on cats (Guinan, Norris, & Guinan, 1972;Spirou, May, Wright, & Ryugo, 1993): CN shows ventral-to-dorsal frequency gradient. The LSO has a frequency gradient which increases from lateral to medial.
The MSO has dorsal-to-ventral tonotopicity. Such relationships are also found in Japanese house bat.
In the dorsal injection cases, FG+ cells were mainly found in the periphery of the NLL. However, it has been shown that tonotopic organization of monaural nuclei of the lateral lemniscus in big brown bat (Epitesicus fuscus) is complicated , and it is difficult to conclude about the tonotopic organization in the NLL of Japanese house bat. Finer injection with best frequency identification is necessary to solve the issue.

| Projection of MSO
In Japanese house bat, MSO projects to bilateral IC, although contralateral projection is one-third weaker than ipsilateral one. In other bat species, projection of MSO of mustache and horseshoe bats is virtually restricted to ipsilateral side (Casseday et al., 1988; F I G U R E 5 Line drawing showing the traces of auditory nuclei and distribution of labeled cell bodies (red dots) of a dorso-medial injection case, Pa128. See legend of Figure 3 for details Zook & Casseday, 1982b). On the other hand, the MSO of Mexican free-tailed bat (Tadarida brasiliensis) sends axons to the IC bilaterally (Grothe et al., 1994). It is possible that the difference reflects the phylogeny, as Mexican free-tailed bat is phylogenically closer to Japanese house bat than mustache or horseshoe bat. It is also possible that the difference arise from type of echolocation calls as mustache and horseshoe bats uses calls composed of CF-FM, while free-tailed and Japanese house bats use FM calls (Fenton, 1995).
The different projection patterns of MSO may also relate to composition of cell types. In the MSO of horseshoe and mustache bats, GABAergic neurons are sparsely present (Vater et al., 1992), while they are virtually absent in Japanese house bat (Ito et al., unpublished observation). Together, in horseshoe and mustache bats, MSO makes both excitatory and inhibitory influence on the IC in the same side, while in Japanese house bat it make excitatory influence on the bilateral IC. Both patterns are very different from common mammalian plan, and imply its relationship to special auditory function of echolocating bats.

| Functional considerations
Correlation of the ratio of input nuclei (Figure 6b) suggests that the composition of brainstem inputs to the ICC is related to the location inside the ICC. Therefore, similar to other non-echolocating mammals (Cant & Benson, 2006;Loftus et al., 2010) and other bat species (Zook & Casseday, 1987), IC of Japanese house bat is likely to be It is also noted that synaptic domains are interconnected by LG neurons as well as glutamatergic interneurons (Ito & Oliver, 2014).
The computation made in each synaptic domain would be modified LG and SG to all labeled cells in each bin of (e). In the distant part, the percentage of LG increased, while that of SG decreased, and the distribution of LG and SG cells were significantly different (p < 0.001, two-way ANOVA, N = 7). (g) Same analysis as (e) was made for ICC cells. (h) Percentage of LG and SG to all labeled cells in each bin of (g). In the distant part, the percentage of LG increased, while that of SG decreased, and the distribution of LG and SG cells were significantly different (p = 0.019, twoway ANOVA, N = 7). Error bars indicate standard error of mean by the interconnection of LG and GLU cells. Furthermore, information conveyed by interconnecting LG and GLU cells is likely to be different since both cell types receive different combination of extrinsic inputs (Chen, Cheng, Ito, & Song, 2018) and LG cells receive inputs from GLU cells in different synaptic domains (Ito & Oliver, 2014). Based on studies on rodents, LG neurons integrate information from multiple brainstem nuclei, while GLU cells integrate inputs related to their synaptic domains (Chen et al., 2018;Ito & Oliver, 2014;Ito et al., 2015). Therefore, it is suggested that LG cells control the entire IC circuit according to global state of brainstem nuclei, while GLU cells control according to the state of specific pathways.
LG cells are shown to send long-range projection to thalamus to elicit inhibitory response faster than excitation (Peruzzi, Bartlett, Smith, & Oliver, 1997). The long-range connection inside the IC may also cause fast inhibition, which is important for coding complex information such as duration (Casseday et al., 1994). To understand the roles of local neurons on the synaptic-domain-based information coding, cell-type specific manipulation of local circuit is required.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
TI designed the study, and performed statistical analyses. TI, TF, and KH performed experiments. TF, KH, KIK, and SH prepared the animals. All authors analyzed the data, and wrote the paper.