CBA/CaH mice (n = 20) and GAD67-EGFP (n = 4) of either sex and aged 2.5–5 months were used in this study. CBA/CaH mice have long been used in hearing research because they exhibit stable auditory brainstem response thresholds and minimal cochlear pathology over the first 2 years of life (Henry and Chole, 1980; Sergeyenko et al., 2013; Ohlemiller et al., 2016). GAD67–EGFP mice express enhanced green fluorescent protein (EGFP) under the glutamic acid decarboxylase 67 (GAD67) promoter in a C57Bl/6 background and are used to label neurons containing gamma-aminobutyric acid (GABA, Tamamaki et al., 2003). Animals were housed in the Biological Testing Facility of the Garvan Institute of Medical Research on a 12 h light/dark cycle and had ad libitum access to food and water. All procedures followed the animal care guidelines of the National Health and Medical Research Council and were approved by the Animal Ethics Committee of the Garvan Institute and St. Vincent’s Hospital, UNSW Australia.

An auditory brainstem response (ABR) was recorded from each animal prior to surgery to verify normal hearing (Stamataki et al., 2006). Each mouse was anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). When the animal was unresponsive to a toe pinch, it was placed in an electrically shielded, double-walled, sound-attenuated chamber padded with acoustic foam (Sonora Technology Co., Gotenba, Japan). ABR testing followed standard procedures of the lab as previously published (Connelly et al., 2017; Suthakar and Ryugo, 2017; Muniak et al., 2018). Only mice with click thresholds better than 30–40 dB SPL were included in these experiments.

Each animal was secured in a stereotaxic frame (Stoelting, Wood Dale, IL, USA) using ear bars and a bite bar. Body temperature was maintained at 37°C using an infrared heating pad. Anesthesia was maintained using isoflurane (1.5–2.0% in ∼600 cc/min O₂). The skull landmarks bregma and lambda were surgically exposed, and a custom-made steel head post was cemented to the skull just rostral to bregma to stabilize the animal for later physiological recordings (Muniak et al., 2012) and a tungsten ground-pin was inserted into the skull nearby. A small craniotomy was made directly over the target structure (DCN or VCN), which was stereotaxically guided by a mouse brain atlas (Paxinos and Franklin, 2019). The craniotomy was covered with bone wax. The mouse given 1 cc of saline subcutaneously for rehydration and left to recover for 1 day prior to electrophysiological recordings and injection.

Recordings were performed in our above-mentioned sound-attenuated chamber. The mouse was lightly sedated with an intraperitoneal injection of acepromazine (0.07 mg/kg), placed in a plastic tube that restricted body movement, and secured by affixing the head post to a custom-built apparatus mounted within a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA). Bone wax was removed from the craniotomy just prior to recording. Quartz glass micropipette electrodes filled with 10% neuronal tracer dye [Alexa Fluor 488, Alexa Fluor 555, or biotinylated dextran amine (BDA) in a solution of 0.05M Tris Buffer and 0.15 KCl, pH 7.3 (Supplementary Table 1)], were used for multiunit recordings (inner tip diameter: 15–20 μm).

Stimulus delivery and neural recordings were controlled via custom software. Acoustic stimuli were generated digitally (DAP5016a, Microstar Laboratories, Bellevue, WA, USA), anti-aliased (Model 3202 Krohn-Hite, Brockton, MA, USA), amplified (Halo A23, Parasound, San Francisco, CA, USA), attenuated (PA5, TDT), and delivered by a calibrated free-field speaker (EMIT High Energy; Infinity, La Crescent, MN, USA) placed 10 cm from the mouse and 25° off the midline. Neural signals were amplified and filtered (2400A; Dagan, Minneapolis, MN, USA), passed through a spike signal enhancer (40-46-1; FHC, Bowdoinham, ME), and digitized for analysis (DAP5016a; Microstar Laboratories). 200 ms broadband or sinusoidal tone search stimuli (4/sec) were delivered as the recording electrode was advanced into the brain using a motorized hydraulic micromanipulator (Model 2650; David Kopf Instruments).

Entry into the VCN or DCN was marked by the unambiguous presence of sound-evoked spike discharges. A frequency response area was obtained using tone bursts at 5dB variable intensities swept through the mouse’s audible range (∼4–100 kHz). A site was chosen that gave a strong evoked response as determined audiovisually by manually adjusting the tone burst frequency and attenuation. Characteristic frequency (CF) for the site was confirmed with an automated tuning curve protocol (MATLAB, MathWorks, Natick, MA, USA) that measured responses to a 4-octave (oct.) sweep centered on the test frequency at 20 dB above threshold, sampling every 1/25-oct (Muniak et al., 2013). With these frequency data, the neuronal tracer was deposited iontophoretically using a high voltage, constant current source (CS 3; Midgard/Stoelting) set at 5 μA and 50% duty cycle for 5–10 min into the site giving the best frequency-evoked response. On several occasions, two injections were attempted at 2 different frequency sites in the same nucleus. After a rest period of 5 min the electrode was withdrawn, the craniotomy covered with bone wax, and the mouse returned to its cage. A survival period of 10–18 days ensured adequate filling of neuronal tracers.

Animals were deeply anesthetized via a lethal dose of sodium pentobarbitone (100 mg/kg, intraperitoneally), and perfused transcardially with 5 ml 1% sodium nitrite prewash in 0.1M phosphate buffered saline followed immediately by 60 ml 4% paraformaldehyde/0.1% glutaraldehyde in 0.1M phosphate buffer. Heads were postfixed overnight after which the brain was dissected from the skull, embedded in gelatin-albumin, and cut in the transverse plane at 50–60 μm using a vibrating microtome (VT1200S; Leica Systems, Nussloch, Germany). All reagents and rinses were made up in 0.12M Tris-buffered saline.

Visualization of Alexa Fluor 488 or 555 did not require any further tissue processing. Coronal sections were mounted and coverslipped using Vectashield (H-1400; Vector Labs) and viewed with a fluorescent microscope. To visualize BDA labeling for brightfield and electron microscopy, sections were incubated for 10 min in 1% H2O2, rinsed 3×, permeabilized for 1 h in 0.5% Photo-Flo (Kodak, Rochester, NY, USA) or 0.5% Triton X-100 (brightfield only), and then incubated in ABC (Vectastain Elite ABC Kit, PK-6100; Vector Labs) with 0.5% Photo-Flo/Triton for 1 h. The tissue was rinsed 3× again and BDA labeling was developed using nickel-intensified diaminobenzidine (DAB; Sigma-Aldrich). Sections were mounted, dehydrated, and coverslipped with Permount (Fisher Scientific, Pittsburgh, PA, USA).

Photomicrographs were collected from a Zeiss Axioplan microscope equipped for brightfield and fluorescent imaging with Plan Neofluar Objectives (20×/0.50, 40×/0.75, and 100×/1.25 oil), and standard light microscopic methods were applied to both brightfield and fluorescent images (Muniak et al., 2013; Muniak and Ryugo, 2014; Williams et al., 2022). Labeled fibers and terminals were manually traced from digitized photomontages with a graphics tablet (Cintiq 22HD; Wacom, Portland, OR, USA) to optimize details.

Serial coronal sections of the entire left CN were photographed using a 10× objective and photomontages were produced by manually aligning digitized sections using Adobe Photoshop 2021. The granule cell domain (GCD) was used to separate the DCN from the VCN (Mugnaini et al., 1980a; Muniak et al., 2013). The DCN and VCN borders and the injection site were outlined in consecutive photographs using a graphic tablet and Photoshop 2021 for each mouse. The area of the injection site was determined by outlining the densest accumulation of DAB reaction product or border of fluorescence at medium illumination. The outlines were mapped as an image-stack to identify the approximate geometric center of the injection site, which was then photographed. A single injection site was recovered in the VCN of 6 mice and in the DCN of 13 other mice; 2 injections were recovered in one DCN of a single mouse. Data from these mice form the basis of this report.

Digital imaging facilitated the mapping of axons as they ascended the lateral lemniscus (LL) to enter the contralateral IC. Using a digital drawing tablet, axons leaving the injection site were traced through serial sections of the dorsal, intermediate, and ventral acoustic striae (DAS, IAS, and VAS, respectively) of the CN and across the brainstem to the contralateral LL (Figure 1). The principle is similar to camera lucida style drawings using a microscope drawing tube attachment. These axons did not stay as a tight bundle but scattered with an overall trajectory that is consistent with textbook illustrations (Carpenter, 1978).

The boundary of the IC and the approximate subdivisions are shown in Supplementary Figure 1 (criteria of Lesicko et al., 2016; Milinkeviciute et al., 2017; Weakley et al., 2022). Photographic tile sets that encompassed the brainstem, LL, and IC were collected using 10× and/or 20× objectives and were manually montaged (Supplementary Figure 2). The 10× images were resampled at 800 dpi and then digitally magnified so that individually labeled fibers could be manually drawn.

Drawings were also conducted on images collected using 10×, 20×, and 40× objectives through the terminal fields of CN projections (Supplementary Figure 3). The digitized drawings collected using different magnification objections were compared and it was determined that images collected at 400 dpi and with a 10× objective suffered no significant loss of accuracy and greatly accelerated progress.

Manual drawing of labeled fibers and terminals was also performed on fluorescent images to illustrate patterns of terminations without interference from background fluorescence (Supplementary Figure 4). Using Adobe Photoshop 2021, we traced individual fibers on a different image layer on the graphics tablet at high magnification in high resolution images and then transferred just the drawing and IC outline to a new image (Supplementary Figure 4, panel 2C). This drawing method was key for showing global patterns of projections with low magnification images.

Photographic z-stacks were collected through the middle of the terminal fields in 2 adjacent sections for analysis of the terminal endings. A rectangular box was placed over the hypothetical isofrequency laminar line in the CNIC and parallel to the laminar axis (see Supplementary Figure 5A, arrows). The image was digitally magnified such that 10 μm measured 10 mm on the digital drawing tablet. All distinct swellings (overlapping swellings were not included) were drawn using a digital pen having a 2 pixel tip (on a 400 dpi image). As previously noted in cats and rats (Malmierca et al., 2005), large and small terminals dominated the terminal field. Drawing and measuring terminal silhouettes were performed from digital images and analyzed. The size distinction could be qualitatively resolved during drawing, given different colors, and placed on separate layers. Drawn terminals were placed on a lined grid, with straight lines separated by 10 μm. For each terminal, the size (μm²) and perpendicular distance to the central laminar line could be assessed (Supplementary Figure 5B).

Terminals were also analyzed with respect to their location in the CNIC, DC, LC, or VLC. Boxes 170 μm x 200 μm were placed over the CNIC, whereas boxes 100 μm x 100 μm were placed over DC, LC, and VL to obtain an estimate of terminal patterning. All subdivisions of the IC exhibited a skewed distribution with respect to terminal silhouette area: there were many more small terminal endings than large ones. Terminal counts and density were not calculated because they are subject to uncontrollable factors including size and placement of the dye injection, dye transport efficacy, histological processing, and individual mouse differences.

Ascending projections of the CN exit through well-known acoustic striae (Adams and Warr, 1976; Cant and Benson, 2003) but within a few hundred micrometers, begin to disperse (Figure 1). Axons leaving the injection site in the VCN traveled ventrally and medially to enter the ventral acoustic stria (VAS), also known as the trapezoid body (TB), along the medial border of the nucleus. The main branch of fibers continued in the TB toward the LSO. Since some VCN neurons project to the DCN and also have collaterals traveling through the TB, DCN injections will result in axonal labeling in the TB. The TB forks into dorsal and ventral streams at the lateral edge of the LSO, whereas some fibers continue straight to enter the nucleus. Their termination in the LSO defined a laminar sheet that was tonotopic and topographically related to the injection site (Doucet and Ryugo, 2003; Gómez-Álvarez and Saldaña, 2016; Williams et al., 2022). Fibers from the dorsal stream entered the ipsilateral superior paraolivary nucleus (SPN) and formed vertical sheets. Fibers from the ventral stream terminated around the ipsilateral medial superior olive (MSO) between the LSO and SPN. Past the midline, fibers from each stream entered and terminated in the contralateral MNTB. The terminations were tonotopic with high frequency fibers distributed medially and lower frequencies located more laterally. A dorsal stream passed over the top of the MNTB, gave off a few collaterals into the periolivary region, and entered the LL. The ventral stream continued as the TB and sent collateral branches and terminals into the contralateral MSO, the contralateral VNTB, and the contralateral VNLL where some fibers gave rise to calycine terminations. The remaining fibers entered the LL.

Fibers from the posterodorsal VCN and DCN fibers from the low frequency region form the intermediate acoustic stria (IAS). This bundle had a shallow arc that passed upward and medially to penetrate the inferior cerebellar peduncle (ICP), and then bent downward over the internal part of the solitary nucleus. These fibers crossed the midline in the central region of the brainstem and scattered while still maintaining a lateral and now upward arc. Some IAS fibers passed over and around the SOC where they dropped collaterals off into the resident SOC nuclei and the VNLL; the remaining fibers curved upward to enter the LL. Some of the fibers passed through the LL to enter the contralateral CN.

Projecting fibers from pyramidal and giant cells of the DCN entered the DAS from the fiber layer that formed the deep border of the nucleus. The DAS travels dorsally, medially, and anteriorly, looping over the ICP and curving through the medial vestibular nucleus (MVN). Fibers start to scatter while still maintaining the general trajectory to cross the midline through the central region of the brainstem, passing through the facial nerve root after the genu and the medial longitudinal fasciculus. Once past the midline, some fibers arc backward and upward into the contralateral DAS to enter the CN, whereas other continue their downward and anterior trajectory over the SOC to enter the LL.

Ventrotubercular axons of VCN planar stellate neurons were retrogradely labeled following dye injections in the DCN (Doucet and Ryugo, 1997; Doucet et al., 1999). These axons descended along the medial border of the CN to a branch point near the ventromedial base of the CN. One branch traveled laterally to the parent planar stellate neuron, whereas the other branch continued ventromedially in the VAS to join the trapezoid body (Doucet and Ryugo, 2003; Darrow et al., 2012). It is important to remember that these axons arise from the VCN, so their terminations are not part of the DCN system.

As the LL ascends from the SOC to the IC, it forms a gentle curve that bends 1 mm anteriorly as it moves dorsolaterally and then back to enter the IC along its ventrolateral edge. Several features of the LL are distinguished: the ascending fibers from the contralateral DCN and VCN maintain relatively separate positions: the VCN fibers ascend in a lateral position, whereas the DCN fibers maintain a medial position (Figure 2). These fibers do not appear to establish a tonotopic position as they ascended in their respective corridors. The ascending fibers from both the VCN and DCN emit horizontal collaterals that travel to the different nuclei of the lateral lemniscus. There is a very minor ipsilateral projection from the CN to the IC: we observed 0–3 fibers from the ipsilateral VCN entering the IC per animal and none were observed from the ipsilateral DCN.

The systematic arrangement between frequency and place was observed decades ago showing an orderly relationship between the cochlea and parts of the central auditory system (Rose et al., 1963; Boudreau and Tsuchitani, 1973). The IC has been a focus of such studies using physiological recording methods to establish place-frequency maps (Merzenich and Reid, 1974; FitzPatrick, 1975; Portfors and Roberts, 2014). The substrate for this frequency organization is initiated in the cochlea (Müller et al., 2005), established in the CN (Muniak et al., 2013), and passed along to the components of the SOC and higher centers. The distinct yet separate contributions to this organization in the IC is evident for the VCN and DCN (Figure 8). The projection pattern shows both convergence and divergence, a reminder of the involvement of frequency in processing multiple parameters of sound.

The size, shape, density, and distribution of the terminals from VCN and DCN projections into the IC were reliably stable. CN terminals in the IC could be qualitatively categorized as small or large by visual inspection (Figure 9). The pattern of axon branching and distribution of synaptic boutons appeared characteristic for each IC subdivision. The fibers were thick or thin and the endings were large or small. The thick fibers branched roughly every 100 μm, alternating with a long and a short branch. The short branches were 2–5 μm in length, terminating as a single bouton 3–4 μm in diameter. The thin fibers branched more frequently but irregularly; they also gave rise to en passant swellings, similar in size and shape as those arising from the thick fibers (3–4 μm in diameter). Secondary branches varied in length and exhibited large and small terminal and en passant swellings. While the terminal patterns were consistent within the different IC subdivisions, the branching density and number of terminals seemed related to the size of the injection site.

The average size of all terminals in the CNIC from the VCN and DCN projections (Figure 9, red boxes labeled A and A’) was not statistically different (VCN, 3.3.08 ± 0.62 μm²; DCN, 3.16 ± 0.54 μm²; Mann Whitney Test, 2-tailed, p = 0.914). The skewed distribution emphasized that there were many more small terminal endings compared to larger ones (Figure 10). When comparing the average size of terminals sampled from the IC (Figure 11, red boxes A, A’, B, C, D), there was no statistical difference in their means (CNIC, 2.60 ± 1.32 μm²; DCIC, 2.47 ± 1.51 μm²; LCIC, 2.46 ± 1.26 μm²; and VLC, 2.96 ± 1.81 μm²; Mann Whitney Test, 2 tailed, p = 0.12.

The pattern of terminations from VCN and DCN projections were qualitatively the same to the CNIC and DC: both projected strongly to the contralateral CNIC with some fibers extending into the DC. Very few fibers (0–3) were observed going to the ipsilateral IC in any mouse. In addition, however, the DCN projected contralaterally to the LC and VLC, with some fibers continuing to the ipsilateral medial geniculate nucleus. For the DCN, the CNIC was the main target with the LC a prominent secondary target; projections to the DC and VLC were smaller but present in all cases.

It has been previously reported that projections from the CN were organized into a laminar and paralaminar pattern (Malmierca et al., 2005). In cats and rats, thick fibers and large terminal endings were distributed primarily in the middle of the terminal field that define the fibrodendritic and isofrequency lamina; in contrast, thin fibers and small terminals were distributed throughout the terminal field and noticeably also along the edges of the terminal field where large endings were absent. In our CBA/CaH mice, however, we did not observe this pattern. Large and small terminal endings appeared uniformly mixed within the CNIC terminal fields of VCN and DCN projections (Figure 12). This distribution was quantified by parsing the terminal field into parallel strips, 10 μm wide, extending away from the central axis of the isofrequency field (Figure 12, labeled 0). The number of large and small endings in each strip was plotted with respect to the distance from the central axis of the lamina. Although there were many more small endings than large endings, counts of each type co-varied together according to position with endings densest near the central axis of the lamina and declining gradually and progressively toward the edge of the terminal field (Figure 13).

We summarize the VCN and DCN projection patterns to the IC as revealed using anterograde tracers in the CBA/CaH mouse (Figure 14). It has been long known that these structures project to the IC. Retrograde labeling following relatively large dye injections into the IC provided further answers as to what regions (Brunso-Bechtold et al., 1981; Frisina et al., 1998) and what types of neurons were projecting (Beyerl, 1978; Adams, 1979; Cant, 1982). Our anterograde projection data complement retrograde evidence by showing the differential terminations by the VCN and DCN in the IC (Ryugo et al., 1981).

The IC is a hub for ascending and descending auditory pathways with proposed subdivisions serving as informational switchboards (Winer et al., 2005). The VCN with its confined connection to the CNIC implies a purely auditory role in the perception of sound. The CNIC contains neurons with tufted dendrites (Rockel and Jones, 1973, Morest and Oliver, 1984) and sharp frequency tuning (Aitkin et al., 1975; Portfors et al., 2011), consistent with the hypothesized, modality-specific “lemniscal” pathways of the brain (Lorente de Nó, 1938; Ramon-Moliner, 1962; Graybiel, 1973). In contrast, the DCN contributes to both a lemniscal and non-lemniscal system. LCIC is a multimodal convergence zone (Robards, 1979; Bácskai et al., 2002; Dillingham et al., 2017; Liu et al., 2023) with physiological response properties that seem to deal more with the physical context of sound (Aitkin et al., 1978). The different CN projections to the IC have a practical implication: the effectiveness of auditory brainstem implants in the CN as reflected by IC responses favors the VCN over the DCN as the preferred target for stimulation (McInturff et al., 2022, 2023).

Type I stellate cells in the core of the VCN and small stellate cells in the VCN margins project to the IC (Roth et al., 1978; Adams, 1979; Cant, 1982; Doucet and Ryugo, 2006). Data from the cat and rat yielded similar but not identical conclusions about CN projections in mice. For instance, single injections into physiologically defined regions of the DCN and/or VCN labeled two prominent bands of terminals; a main band in the CNIC and a shorter lateral band in the LC (Malmierca et al., 2005; Loftus et al., 2008). This dual terminal field is distinctly absent from the VCN projections of mice (Figures 3, 8). Moreover, in an exhaustive study of the gerbil (Cant and Benson, 2006), cells in the VCN, PVCN and DCN were labeled in 72 of the 74 injections into various parts of the IC. We cannot explain these differences, but the most plausible interpretation involves species variations and their different natural histories. With respect to cats and rats, mice occupy a different position in the food chain—mice are prey for cats and rats (Yang et al., 2004; Campos et al., 2013; Lahger and Laska, 2018). Gerbils are social, desert animals that live in multibranched burrows (Ågren, 1976; Scheibler et al., 2005), whereas mice inhabit forests, grasslands, or urban areas and live in simple burrows (Fisher and Llewellyn, 1978; Bradford, 2014). It is argued that different lifestyles exerted selective survival pressures on certain brain features of individual species. In spite of these variations, it should be stressed that the fundamental design of the central auditory system is similar across mammals.

Natural selection requires adaptation of different animals to their habitat, so brain specializations and variations should be expected. The differences in hearing range (gerbils, 0.1–60 kHz; mice, 4–90 kHz) and structure of certain auditory brainstem nuclei offer potential clues into what features of the sound stimulus are providing information necessary for predator avoidance and social communication (Fay, 1988). Both animals have small heads, which limit the utility of head shadow effects and impose physical limits on interaural time differences (Grothe, 2000). The task, then, is to collect neural evidence to explain how interaural disparity cues are transformed into spatial awareness by the separate pathways.

The majority of pathway tracing studies have used retrograde tracing methods (Cant and Benson, 2003, 2006). Even when an injection of a retrograde tracer is placed precisely on target, the dye marking the injection site will obscure important details about the anatomy of the projection, such as the branching structure of individual fibers, the organizational pattern of terminals, and the target neurons. Thus, the use of anterograde tracing methods provides data that are independent yet complementary to those revealed by retrograde studies. When applying pathway tracing methods, it is also important to be aware of technical differences and limitations (Mesulam et al., 1980; Saleeba et al., 2019, 2020), the impact of age (LeVay et al., 1980; Ryugo and Fekete, 1982; Willott, 1984; Frisina and Walton, 2001), and the importance of species (Irving and Harrison, 1967; Moore and Moore, 1971; Merzenich et al., 1973; Masterton et al., 1975). These variables have an important role in the interpretation of the results.

Pyramidal and giant cells are sources of DCN input to the IC (Osen, 1972; Beyerl, 1978; Ryugo et al., 1981; Oliver, 1984; Ryugo and Willard, 1985). These cells exhibit narrow frequency tuning (Godfrey et al., 1975; Rhode et al., 1983; Rhode and Smith, 1986) but they differ in terms of their dendritic orientation and cell body location in the DCN (Brawer et al., 1974; Schweitzer and Cant, 1985). Pyramidal cells are the dominant neuron, located in layer II and exhibiting a planar dendritic field that form frequency layers across the long axis of the nucleus (Blackstad et al., 1984; Ryugo and Willard, 1985; Spirou et al., 1993). The dendritic structure of pyramidal cells, aligned with the incoming auditory nerve fibers, is consistent with their sharp tuning. In contrast, giant cells represent a much smaller numerical population, are located in the deep layers of the DCN and exhibit widely branching dendrites, consistent with their intercepting of a variety of diverse inputs (Ryugo and Willard, 1985). Because both cells have sharp tuning, the planar dendrites of pyramidal cells intercept a narrow frequency band of input. The giant cells, with their wide dendritic branching, might receive frequency-specific auditory input exclusively onto their cell bodies but accept diverse inputs onto their dendrites. In this way, both cell types could exhibit relatively narrow tuning and frequency-specific projections but the giant cells would be capable of conveying a richer palate of information further upstream.

We noted three tonotopic projection arms into the IC from the DCN. There were the previously reported main CNIC projection (that included a lateral part of the DC) and a second projection arm into layer III of the LC, plus the newly detected short projection branch into the VLC. The two main branches were inferred to endow the IC with its tonotopic organization, where the frequency axes of the two projection arms were orthogonal to each other yet conformed to the frequency features recorded in cats (Aitkin et al., 1975; Loftus et al., 2008).

The three branches of the DCN projection were distinct and present in all our mice. The CN projections in cat and rat were shown to stack (Loftus et al., 2008), a relationship verified by us for the mouse. When sections at the level the IC commissure are selected from the different mice, superimposed, and scaled, the result is an imperfect but credible stacking of DCN frequency projections. This tonotopic stacking, where one frequency band “nests” into the one below, implies that there is a proportional relationship to frequency bands, with the largest band resting at the ventral edge of the CNIC, and progressively smaller bands stacked up to the dorsal extremity. Moreover, this stacking includes the three arms of the DCN input to the IC (Figure 14). This relationship implies that there is a gradual but progressive increase in tissue devoted to processing higher frequency sounds.

The apparent expansion of tissue representing higher frequencies is noteworthy because the highest frequency response we’ve recorded in the mouse auditory brainstem is 56 kHz, whereas the hearing range of the mouse extends beyond 100 kHz. Our lab data are consistent with the results of Portfors and Roberts (2014) who suggest that the mid-high frequency regions of the IC are processing the lower frequency harmonics of high frequencies. Such an adaptation would be a clever way of accessing high frequency auditory stimuli without having to devote new brain tissue for the expanded sensitivity.

The tuning characteristics of IC units in the three IC divisions have been described; all units exhibited frequency tuning curves, but DC and LC units were more broadly tuned and “habituated” after a few stimulus presentations (Aitkin et al., 1975). Moreover, in recordings free of anesthesia effects, IC responses were shown to resemble those of the DCN, excited by low levels of stimulation and silenced by higher levels (Davis et al., 2003); these neurons were also sensitive to spectral troughs created by pinna cues (Rice et al., 1992). The spectral changes caused by head-related transfer functions means that a particular sound, composed of its own unique spectrum of frequencies, will change depending on its location in space by virtue of its altered spectrum. Synthesized waveforms played through headphones that simulate pinna transformations mimicked free-field listening and generally permitted source location in space (Wightman and Kistler, 1989a,b). For sounds to maintain their identity when subject to wide ranging variations in spectral waveforms depending on environmental conditions and/or source location, there must be real-time, ongoing auditory connections to memory circuits.

The terminal endings of the CN projections are, in a simplistic way, large or small. There are an estimated 2–3× as many small endings as large ones, and the two ending types appear uniformly distributed within the terminal fields. We tested the hypothesis, proposed by Malmierca et al. (2005), that the large terminals would define the laminar axis, whereas the small terminals would lie on the “paralaminar” periphery (Supplementary Figure 5). The seemingly uniform mixture of large and small terminals over the terminal field for the CBA/CaH mouse is not consistent with that plan. Moreover, we did not observe large endings arising from thick fibers and small endings arising from thin fibers; rather, endings of either size could arise of fibers of either thickness. Fiber thickness is not always a reliable anatomical feature because myelinated fibers often appear thinner as they pass through the depth of the tissue section. This artifact was frequently observed for myelinated auditory nerve fibers and was attributed to inadequate penetration of the chromogenic reagents through the tissue and especially the myelin (Fekete et al., 1984). The endings in our material were overly stained so no internal details of the endings were visible when viewed in an electron microscope. It was evident, however, that the labeled endings from the CN were apposed entirely against dendrites in the CNIC, in agreement with what was observed in cats (Oliver, 1985).

Pyramidal and perhaps giant cells are also innervated at the apical ends of their dendrites by parallel fibers of granule cells (Mugnaini et al., 1980a,b; Manis, 1989; Shore and Moore, 1998; Rubio et al., 2008). Granule cells are known for their integration of multimodal inputs (Shore and Moore, 1998; Ryugo et al., 2003). These inputs include projections from the somatosensory system that would provide information from proprioceptors from the head, neck, and pinna (Robards, 1979; Weinberg and Rustioni, 1987; Wright and Ryugo, 1996; Haenggeli et al., 2005; Zhou and Shore, 2006). Information from the vestibular system would inform the animal about head and spine position with respect to gravity, movement, posture, and 3D space (Künzle, 1973; Flumerfelt et al., 1982; Shokunbi et al., 1985; Walberg et al., 1985; Burian and Gstoettner, 1988; Kevetter and Perachio, 1989; Bácskai et al., 2002; Bukowska, 2002; Newlands and Perachio, 2003). Such information is interacting with sound processing early in the ascending auditory pathway (Kanold and Young, 2001; Kanold et al., 2011). Multimodal information from the cerebral cortex by way of the pons (Ohlrogge et al., 2001), from the cerebellum via the lateral reticular nucleus (Zhan and Ryugo, 2007), and from the visual system (Qvist and Dietrichs, 1985, 1986; McCrea et al., 1987) is also being delivered. The output of the DCN has already been subject to sophisticated processing before it gets delivered to the IC (Young et al., 1992; Nelken and Young, 1996; Davis et al., 2003). The IC exhibits both convergence and compartmentalization of its diverse inputs (Lamb-Echegaray et al., 2019; Weakley et al., 2022) so we are reminded that navigating in our acoustic environment requires more than just the transduction of vibrations in air.

Our cognitive perception of the world is richly multimodal. Stereognosis, the ability to identify the 3D shape of an object manually or orally is linked to auditory discrimination (Lewis and Kelly, 1974). The ventriloquism effect demonstrates that sound localization is influenced by vision (Spence and Driver, 2000), and facial movements can affect the perception of speech (McGurk and MacDonald, 1976) as well as non-speech stimuli (Saldaña and Rosenblum, 1993; Laeng et al., 2021). There is also the sound-induced flash illusion where the observer misinterprets the number of visual flashes due to the simultaneous presentation of a different number of clicks (Virsu et al., 2008). Audiotactile interactions are dramatically illustrated by the parchment-skin illusion where tactile sensation is influenced by sound (Jousmäki and Hari, 1998) in sighted but not blind individuals (Champoux et al., 2011). These illusions or misinterpretations of our sensory experience nonetheless depend on the essential parameters of sound – frequency, timing and experience. No illusion would occur if the sound of a trumpet came from a moving mouth or if speech was produced by rubbing hands together.

Species differences, habitat ecology, and hearing behavior provide insight for understanding circuits. The VCN and the DCN exhibit differential projections to the IC. The idea of parallel processing is a little misleading because ascending circuits convey different kinds of information and synaptic mechanisms at the targets are variable. The VCN primarily deals with frequency and timing of auditory information, whereas the DCN integrates auditory and non-auditory inputs. They both project to the IC while maintaining separate domains. Multimodal integration at the levels of the DCN and lateral cortex of the IC executes processes that help integrate sound and context for sorting concurrent sound streams. These and other circuits are presumed to enable sound recognition under a variety of different conditions, where the spectral composition changes but our recognition does not. It is anticipated that identifying species-differences in auditory circuits will contribute to a better understanding of central mechanisms of hearing.