IntroductionThe auditory system is responsible for processing sounds from the environment by decoding sound frequency, amplitude, and temporal features. It achieves remarkable precision in sound source localization and acoustic object discrimination. Inner hair cells (IHCs) perform the initial sensory transduction of sound stimuli and transmit information to the cochlear nucleus (CN) via auditory nerve fibers.1 The CN comprises diverse neuronal types that play a variety of roles in the processing of auditory signals. Some CN neurons are also involved in multisensory integration.2,3,4 The CN serves as the primary central relay in the auditory pathway and is the target for auditory brainstem restoration5 in patients who obtain little-to-no benefit from cochlear implants. Extensive efforts6,7,8,9,10,11,12,13,14,15 have been made to investigate the morphological, electrophysiological, and molecular properties of CN neurons. However, a comprehensive understanding of the changes, characteristics, and spatial distribution of CN cell types in response to auditory input underlying auditory processing is still lacking.Sensory experience is known to modify the structure and function of neural circuits through activity-dependent neuronal plasticity, including cell type specification, axon/dendritic arborization, and the formation of synaptic connections.11,16,17,18,19,20,21 Previous studies11,21,22,23,24,25,26 have demonstrated auditory activity-dependent development of the CN and plasticity of synapses between the auditory nerve and CN neurons. During postnatal development, CN neurons undergo physiological and morphological changes, including the development of calyceal synapses on bushy cells and conventional bouton synapses on stellate cells.27,28,29 Cochlear removal during the first postnatal week reduces the size and number of CN neurons in mice,30,31,32,33,34 and cochlear implantation yields better hearing and speech outcomes in hearing-impaired children younger than 12 months than in older children,35,36 indicating that auditory activity plays a significant role in CN development. However, which CN cell types and how the global expression of genes are affected by auditory input remain unknown.In this study, we identified CN subregions based on global gene expression as well as transcriptome-defined cell types and their spatial distribution within the CN using single-nucleus RNA sequencing (snRNA-seq) and spatially enhanced resolution omics sequencing (Stereo-seq).37 This atlas enabled us to identify the Spp1-expressing subtype of bushy cells as the critical CN cell type affected by congenital sensorineural hearing loss due to IHC malfunction. Spp1 expression was correlated with hearing onset in normal mice and was downregulated in hearing loss mice, and genetic deletion of Spp1 affected CN processing of auditory signals in mice. Thus, our study provides a valuable resource of the molecular atlas of CN, laying the foundation for understanding auditory processing physiology and pathophysiology within the mouse CN and contributing to the design of effective neuro-prostheses for hearing restoration.ResultsSpatial transcriptomic mapping of CN subregionsPrevious studies have shown the existence of subregions within the CN.6,38 We performed Stereo-seq analysis of the CN from postnatal day 45 (P45) wild-type (WT) mice. We obtained a total of 35 sagittal and coronal sections, including 15 consecutive sagittal sections (from one mouse, CN #1), 16 sagittal sections (at 60-μm intervals from four mice with different starting coordinates in the CN, CN #2–#5), and 4 coronal sections (at 120-μm intervals from one mouse, CN #6). The sections were laid onto Stereo-seq chips, where DNA nanoballs (DNBs; size, 220 nm) were docked onto the chip surface in a grid-patterned array (DNB center-to-center distance, 500 nm). These DNBs captured RNA transcripts, and the sequencing data were processed and integrated to generate a 2D spatial transcriptome map of each section (Fig. 1a). Unsupervised spatially constrained clustering (SCC) was used to group the samples into spatial clusters based on distinct expression profiles and spatial locations of DNBs (see “Materials and Methods”39). At a resolution of BIN50 (~25 μm × 25 μm, comprising 50 × 50 DNBs covering multiple adjacent cells), we observed clear CN subregions with distinct gene expression profiles (see Supplementary information, Fig. S1a for different BIN sizes). Typically, the average numbers of unique molecular identifiers (UMIs) and genes at BIN50 resolution were 5823 and 2031, respectively (Supplementary information, Fig. S1b). In total, we identified 13 CN subregions based on spatial transcriptomic patterns (Fig. 1b, c; Supplementary information, Fig. S1c). These transcriptome-defined subregions were highly consistent between adjacent sections and biological replicates (Fig. 1b; Supplementary information, Fig. S1d).Fig. 1: Clustering and annotation of molecularly defined CN subregions.Full size imagea Schematic of Stereo-seq and spatial clustering for the CN. b Spatial map of Stereo-seq-defined CN subregions for 15 consecutive sagittal sections (CN #1; S #1–#15) from lateral to medial and 4 coronal sections (CN #6; C #1–#4) from anterior to posterior. Regions are colored on the basis of their spatial transcriptome patterns (A anterior, P posterior, V ventral, D dorsal, L lateral, and M medial; #: numbered sagittal and coronal CN sections). c Heatmap showing the normalized expression of marker genes for the indicated CN subregions. d Specific colored and labeled subregions of the CN, shown against the background of the other regions in gray (top). Spatial expression of the region-specific marker genes Gabra6, Car8, Penk, Sst, Spp1, Calb2, and Cabp7 (middle). Immunostaining (GABRA6, CA-VIII, SST, SPP1, and calretinin (CR)) and smFISH (Penk, Cabp7) images of the identified spatial markers of the CN (bottom).We annotated the transcriptome-defined CN subregions that were consistent with histology-based regions using the same names.6 The previously described “granular cell region” (GCR) was marked by the granular cell marker gene Gabra6 (refs.40,41; Fig. 1b, c). The CN can be broadly divided into dorsal (DCN) and ventral (VCN) regions on the basis of its cytoarchitecture,6 and these regions are believed to perform different CN functions. A previous subdivision of the CN based on cytoarchitectural features showed that the DCN exhibits layered structures.3 Our Stereo-seq data enabled molecular identification of these layers. The outermost layer that covers the CN surface, known as the ependymal layer,42 expressed high levels of the ependymal cell marker gene Foxj1 (Supplementary information, Fig. S1e). The molecular layer, which is enriched with many unmyelinated “parallel” fibers and sparsely distributed inhibitory stellate cells,43,44 exhibited low expression of the myelin marker Mbp and high expression of the GABAergic cell marker Gad1 (Supplementary information, Fig. S1f). Transcriptomic patterns enabled us to identify subregions within some of these histology-defined regions. The fusiform cell layer was previously defined as the region containing fusiform cells, but it also includes a group of non-fusiform cells known as cartwheel cells, which are Purkinje cell-like inhibitory neurons45,46 that express Car8. Using a high level of Car8 expression to define the fusiform cell layer in our transcriptomic map, we found that the gene Fam19a1 was highly expressed in a subregion of this layer, implying the existence of a further subdivision of the conventional fusiform layer. As shown later, Fam19a1 was the primary marker gene for the annotation of fusiform cells (see Fig. 2a for cell typing), and the previously histology-defined fusiform layer could be redefined into the “superficial fusiform cell layer” and the “deep fusiform cell layer”. The latter annotation was based on a higher density of fusiform cells than that in the superficial layer (as shown later in Fig. 2g) and is consistent with the finding from physiological recordings that cartwheel cells are encountered before fusiform cells.47 Our spatial transcriptome profiles of glutamatergic and GABAergic/glycinergic neurons also support this layer definition based on cartwheel cells (GABAergic/glycinergic) and fusiform cells (glutamatergic) (Fig. 1d; Supplementary information, Fig. S1g). Our spatial gene expression profiles also supported the molecularly defined DCN deep layer, which is known to have a spatially mixed distribution of excitatory neurons (such as unipolar brush cells)48 and inhibitory neurons (such as vertical cells)49 (Fig. 1d; Supplementary information, Fig. S1h).Fig. 2: Stereo-seq-defined CN cell types and spatially resolved cell atlas of the mouse CN.Full size imagea, b Uniform manifold approximation and projection (UMAP) visualization of segmented cells from all groups of mouse CN sections colored according to their cell type annotations (a). Spatial visualization of Stereo-seq-defined cell types in sagittal and coronal sections colored according to their annotations (b). c Spatial distribution of Spp1+ and Sst+ bushy cells (top left), co-immunostaining of SPP1 and SST (bottom left), patch-clamp recordings (top right), and co-immunostaining of cell type-specific marker proteins with biocytin staining of the recorded cell (bottom right). d, e Spatial distribution of T-stellate cells (d) and fusiform cells (e), immunostaining of their marker genes, patch-clamp recordings, and immunostaining of the marker proteins, with biocytin staining of the recorded cell. f Overview of the compositional diversity of cell types in each Stereo-seq-defined CN subregion (with the exception of the nerve fiber-enriched area) in a representative sagittal section (contour lines indicate CN subregion boundaries). Cells are colored according to their cell-type identities (a). g Composition of different cell types (indicated by the percentage of a specific cell type in each layer or subregion) in each Stereo-seq-defined CN subregion. h Mean nearest-cell interaction scores of Stereo-seq-defined cell types in the aAVCN and pAVCN. The rectangle indicates the interaction scores between Spp1+/Sst+ bushy cells and other cell types. UBCs unipolar brush cells, OPCs oligodendrocyte precursor cells, Oligo oligodendrocytes, VECs vascular endothelial cells, VLMCs vascular and leptomeningeal cells.Our global spatial transcriptome enabled us to subdivide the VCN into two primary regions based on molecular profiles: the anteroventral CN (AVCN) and the posteroventral CN (PVCN). The AVCN could be further subdivided into anterior (aAVCN) and posterior (pAVCN) regions, which showed high expression of the somatostatin gene Sst and the osteopontin (OPN) gene Spp1, respectively (Fig. 1b, d). Within the PVCN, we also identified marker genes such as Cabp7 and Calb2 for the previously histology-defined multipolar cell region and octopus cell region, respectively (Fig. 1b), and confirmed their protein expression with immunostaining (Fig. 1d). We also identified a subregion corresponding to the previously described “auditory nerve root” region,50 which contains mostly glutamatergic neurons (such as bushy cells). A complete list of identified marker genes for CN subregions is provided in Supplementary information, Table S1. In summary, our Stereo-seq data provide a comprehensive molecular fingerprint for more refined definitions of CN subregions.Spatial transcriptome of CN cells at single-cell resolutionWe next examined the spatial transcriptome map of the CN at single-cell resolution. We performed cell segmentation of the Stereo-seq data based on single-stranded DNA staining that highlighted the nucleus (Supplementary information, Fig. S2a). The watershed algorithm and Gaussian blur algorithm were then used to identify the outlines of individual cells (Supplementary information, Fig. S2a, b). After excluding poorly captured cells, we obtained 219,457 segmented cells with an average of 1861 UMIs and 817 genes per cell (Supplementary information, Fig. S2c). Using Seurat’s unsupervised clustering analysis, we identified 20 major cell types (Fig. 2a; Supplementary information, Fig. S2d; the complete list of differentially expressed genes (DEGs) is available in Supplementary information, Table S2) on the basis of their spatial locations and the expression of known marker genes.15,51 The spatial distributions of these 20 cell types are shown in the composite map in Fig. 2b.Based on our Stereo-seq map, bushy cells could be divided into two types, annotated as Sst+ (high Sst and low Spp1 expression) and Spp1+ (high Spp1 expression and low Sst expression), and this was supported by the finding that very few cells were co-immunostained for both SPP1 and SST (Fig. 2c). This division of SST+ and SPP1+ subtypes largely overlapped with our snRNA-seq-defined subtypes of bushy cells (Supplementary information, Figs. S4g, S6c) — the spherical bushy cells (SBCs) and globular bushy cells (GBCs), as defined by their respective marker genes15Atoh7 and Hhip (Fig. 3a). Co-immunostaining experiments for SPP1 and HHIP, as well as for SST and ATOH7 (Supplementary information, Fig. S2e), confirmed the high-level co-localization of these corresponding proteins. Furthermore, both SPP1+ and SST+ neurons showed electrophysiological properties characteristic of bushy cells (Fig. 2c). We found that Sst was highly expressed in the Sst+ subtype of bushy cells (known to be excitatory; Supplementary information, Fig. S2f), consistent with previous findings,15 despite Sst being a well-known marker gene for GABAergic neurons in the brain.52 We also identified candidate marker genes for T-stellate cells (C1ql1) and fusiform cells (Fam19a1), which were confirmed on the basis of cells that could express the corresponding marker proteins and exhibit electrophysiological properties13 known to be distinct for these cell types (Fig. 2d, e).Fig. 3: Role of auditory activity-dependent CN cell gene expression and spatial organization.Full size imagea UMAP of CN cells from all groups using snRNA-seq data. Cells are colored according to their cell-type annotations. b Heatmap showing the expression of conserved and mouse group-enriched marker genes for each cell type from the snRNA-seq data. c UpSet plots showing the numbers of shared and divergent marker genes for GBCs from snRNA-seq data across different mouse groups. d GO terms associated with DEGs highly expressed in GBCs from different mouse groups. e Spatial maps (top) of Stereo-seq-defined CN cell types from P1, P7, P14, P45 (WT), Vglut3–/– and Vglut3–/–+GT. Cells are colored according to their cell-type identity (Fig. 2a). Spatial maps of cells, colored according to their spatial module identities (bottom), are shown in representative sagittal sections from different mouse groups. f Fraction of cells in three representative spatial modules across different mouse groups. g Distribution of mean cell–cell distances within the same cell type (“to self”, red) or between different cell types (“to other”, blue) determined from the Stereo-seq data. False discovery rate (FDR): P-value determined by the one-sided Wilcoxon rank-sum test and adjusted to FDR by the Benjamini–Hochberg (BH) procedure. h Mean intra-type distances of the two types of bushy cells and the density of glial cells in the two spatial modules in which bushy cells are mainly localized. Statistical analysis was performed using one-way ANOVA followed by a Bonferroni post hoc test. ***P