Specialized neuropeptide pathway expression in the song system.

(a) Schematic overview of the song system. HVC, proper name; RA, robust nucleus of the arcopallium; LMAN, lateral magnocellular nucleus of the nidopallium; Av, avalanche; DLM, medial portion of the dorsolateral thalamic nucleus; D, dorsal; P, posterior. (b) Circuit diagram of the song system. Arrowheads and closed circles indicate excitatory and inhibitory connections, respectively. NC, caudal nidopallium; Arco., arcopallium; NR, rostral nidopallium; Stri., striatum. (c) Left: Experimental overview of SLCR-seq on hearing and deaf birds, from Colquitt et al. 2023. After a baseline period of song recording, birds were either deafened through bilateral cochlear removal or underwent a sham surgery. After 4, 9, or 14 days post-surgery, birds were euthanized and SLCR-seq libraries were prepared from HVC, NC, RA, Arco., LMAN, NR, Area X, and Stri. Right: Uniform Manifold Approximation and Projection (UMAP) plot of SLCR-seq data colored by section position. Each point reflects the gene expression profile of a single SLCR-seq sample. Samples show segregation by broad anatomical area — striatal (Area X), nidopallial (HVC, NC, LMAN, NR), arcopallial (RA, Arco.) — and song system nuclei from surrounding areas. (d) Volcano plots showing differentially expressed genes between each song and paired non-song region across all genes. Numbers in the top left and right indicate the number of genes with adjusted p-values < 0.1 and an absolute log2 fold-change greater than 1. (e) Volcano plots as in (d) subsetted for neuropeptide-associated genes (ligands, receptors, and modulators; see Methods for definition). (f) Patterns of neuropeptide gene expression enrichment and depletion in song regions relative to surround regions. (g) Normalized log gene expression data of an example gene SSTR3 whose expression is enriched in RA relative to Arco, and depleted in HVC and LMAN relative to their surrounding regions. Each point indicates gene expression in a single SLCR-seq sample. (h) Coronal anatomical atlas representation of the expression of SSTR3. Left: Each region shown with color scheme as in A. Right: log gene expression value. D, dorsal; L, lateral. (i) Analysis of differential expression between the three pallial (cortical) song nuclei, HVC, RA, and LMAN and their surrounding regions, NC, Arco, and NR.

Patterns of cell type specific expression of neuropeptide systems in the song motor pathway.

(a,b) Uniform manifold approximate projection plots of single-nucleus RNA-sequencing from two song regions (RA and HVC) adapted from (Colquitt et al., 2021). (a) Points colored and labeled by cluster and (b) points colored and split by region. (c) Schematic of identified clusters in each region. GABAergic neuron naming convention follows that used in (Colquitt et al., 2021). (d) Expression of neuropeptide ligands and receptors across HVC and RA cell types. (e) CellChat analysis of overall neuropeptide cell-cell communication probability in HVC and RA. (f) Predicted relative communication probability of neuropeptide ligand-receptor pairs in each region (see Methods). Communication probabilities were summed across cell types for each ligand-receptor pair, then normalized by the total communication probability. (g) Communication probabilities for ADCYAP1-ADCYAP1R1 and SST-SSTR1, the top two ligand-receptor pairs in HVC and RA, respectively. Arrow widths are scaled to communication probabilities. The top 1% of interactions are shown.

Song destabilization results in reduced expression of several neuropeptide-associated genes in the motor output regions RA and HVC

(a) Example spectrograms from one hearing (sham) and one matched (sibling) deaf (bilateral cochlea removal) bird. Songs are shown from before the procedure and 14 days following the procedure. Labels below each spectrogram correspond to discrete categories of song units (‘syllables’). kHz, kiloHertz. (b) Uniform Manifold Approximation and Projection (UMAP) representation of syllable spectrograms (see Methods) across the entire recording period for the deafened bird (4 days before to 14 days after the procedure). Data are split into ‘pre’-procedure (blue: 4 to 1 day before surgery) and ‘post’-procedure (red: 14 days after surgery) subsets. For reference, gray points in each plot correspond to data across the entire recording period. Example syllable spectrograms are placed adjacent to their position in UMAP space. (c) Relative spectral distance between syllables pre- and post-procedure, represented as the mean Kullback-Leibler (KL) distance between Gaussian mixture models or ‘Song DKL’ (see Methods for calculation). Song DKL trends higher with increasing days from deafening. Significance calculated using a two-sided Wilcoxon rank-sum test. (d) Gene set enrichment analysis (GSEA) of song destabilization-associated genes. Shown are the Gene Ontology (GO) terms that are significant in at least one song or non-song region (adjusted p-value < 0.1, see Methods). Dot color indicates the normalized enrichment score, and dot size indicates the number of leading edge genes (core genes most responsible for the enrichment associated with the GO term). Terms are ordered by hierarchical clustering (Euclidean distance, Ward squared method). Highlighted is the term Hormone Activity (GO:0005179). (e) Log2 fold-change expression of the leading edge genes in term Hormone Activity across song and surround regions. (f) Volcano plot of destabilization-associated differential expression, subsetted for neuropeptide-associated genes.

CRHBP expression in the song motor pathway increases during song acquisition and decreases following deafening

(a) Developmental timeline of birdsong learning. dph, days post hatch. (b) Representative images of in situ hybridizations against CRHBP in RA and HVC (outlined) from birds at different developmental stages or adult birds that are hearing or deaf. Scale bar is 200 μm. (c) Quantification of signal intensities across conditions and brain regions. Each point is the CRHBP signal intensity in each CRHBP-positive cell. P-values are derived from linear mixed models (see Methods).

Modulation of neuropeptide system expression in the song system by singing

(a) Analysis of singing-dependent gene expression across two different time scales: total number of songs sung during the two hours before euthanasia and total number of songs sung during the two days before euthanasia. (b) Total number of songs sung by the birds in the SLCR-seq dataset in the two hours (left) or two days (right) before euthanasia. Birds are divided into low, mid, and high groups for differential gene expression analysis. (c) Heatmap of genes showing significantly increased expression (adjusted p<0.1) across pallial song nuclei (RA, HVC, and LMAN) between 2 day high and low singers. (d) Gene set enrichment analysis of singing-rate associated expression responses that are shared across the pallial song system (for both two-hours and two-days measures). Enriched are several groups of terms including those related to neuropeptides, activity-dependent gene regulation, neurotransmission, and cellular respiration. Listed are the five top leading edge genes for selected gene sets related to neuropeptides (“Hormone activity”, “Dense core granule”) and regulation by neural activity (“Delayed PRG (primary response genes)”). (e) Singing-modulated expression of CRHBP and SST across the song system and surrounding regions for the two-days measure.

Distributed expression of the CRH neuromodulatory pathway in the song motor pathway

(a, b) Expression of CRH pathway components, CRHBP, CRH, CRHR1, CRHR2, in the song system and non-song surround regions. (a) Atlas representation of SLCR-seq data showing expression in coronal brain sections. (b) Point representation of SLCR-seq data. Each point is the average gene expression in a given region in a single bird. Error bars are standard error of the mean across 2-6 sections/region/bird. Colored dots represent mean expression across birds for a given region. Mean differences were calculated between each song and adjacent non-song region, and adjusted p-values (Student’s t-test, Benjamini-Hochberg adjustment) are shown for comparison with p < 0.05. (c) Expression of CRH pathway components in HVC and RA cell types. A previously generated single-cell/single-nucleus RNA-seq (scRNA/snRNA-seq) dataset (Colquitt et al., 2021) was queried for CRH, CRHBP, CRHR1, and CRHR2 expression, represented here as a dotplot. Dot intensity reflects the average expression of a given gene in a given cluster (cell type), and dot size indicates the percentage of cells in a cluster that have detectable expression of a given gene. (d, e) In situ hybridization (ISH) validation of CRH, CRHBP, and CRHR2 expression in the song system. (d) CRH and CRHBP show non-overlapping expression patterns in HVC and RA. PVALB is used here to demarcate HVC and RA which show elevated levels of PVALB in glutamatergic neurons (Wild et al., 2001). PVALB is also a general marker for GABA-2/3/4 (Sst-class and Pvalb-class) interneuron subtypes. (e) ISHs of the CRH receptor CRHR2 in HVC and RA indicate broad expression in glutamatergic neurons (indicated by SLC17A6). Scale bar is 100 μm. (f) Expression of CRH pathway components CRHBP, CRH, CRHR1, and CRHR2 in neurons of the mouse neocortex (Tasic et al., 2018) and turtle pallium (Tosches et al., 2018). Expression is scaled to the minimum and maximum of each gene. Cell cluster annotations are labels from the original publications. Crhr2 in mouse and CRHR1 are not shown due to low expression. (g) Model of the local CRH neuromodulatory pathway in HVC and RA. Arrows indicate positive interactions, blocked lines indicate negative interactions. CRH, released by Vip-class GABA-5 (and to a lesser extent Lamp5-class GABA-6), activates CRHR2 receptors expressed by glutamatergic neurons, GABA-7 interneurons, and GABA-5 interneurons (autoregulatory). CRHBP is released from GABA-4 Pvalb-class interneurons and binds to CRH, thereby reducing activation of CRHR2.

Modulation of the CRH pathway has bidirectional effects on song variability

(a) Transient siRNA knockdown of CRHBP and CRH in the song motor output nucleus RA. (b) Example calculation of KL distance between syllable power spectral densities. Top example presents a syllable that is relatively unchanged and yields a small KL distance. The syllable in the bottom example is more divergent, yielding a higher KL distance. (c) Power spectral density (PSD) Kiebler-Lullbeck (KL) distances over time following (left) cochlea removal or sham surgery or (right) injection of CRHBP, CRH, or control siRNAs into RA. The PSD KL distance for each syllable was calculated relative to the average PSD for the same syllable in the pre-procedure baseline period. Values were then z-scored relative to the mean and standard deviation of PSD KL distances in the baseline period. * p < 0.01, linear mixed effects model comparing the difference from baseline. (d) Influence of CRHBP knockdown on syllable variability. Left, Spectrograms of two syllables sung during the pre-injection period, one from a bird injected with a control siRNA and one from a bird injected with a siRNA targeting CRHBP. Spectrograms were averaged along the time axis within red boxed regions shown on the examples and plotted over time. Arrowheads indicate the fundamental frequency. Vertical dashed red line indicates the day after siRNA knockdown. Shaded gray box highlights a period of high syllable variability post CRHBP knockdown. Tick marks indicate the beginning of each day. Shown is 20% of the total number of syllables. (e) Fundamental frequency variability over time for the example syllables shown in panel (D). Each point is the FF coefficient of variation calculated across a window including the fundamental frequencies of a given syllable and the 10 syllables preceding it. Vertical dashed red line indicates the time of siRNA knockdown. Shaded gray box indicates the period used to calculate post-knockdown summary statistics in panel (F). (f) Influence of CRH pathway knockdowns on fundamental frequency (FF) across-rendition variability. For each bird, syllable, and period (pre vs. post-knockdown), the across-rendition coefficient of variation was computed. Values were then normalized within each bird and syllable to give a percent change relative to pre-knockdown values. These normalized values were then averaged across birds and syllables. Shown here are the normalized post-knockdown values. Pre-knockdown data includes at least 2 days of song prior to injection. Post-knockdown data includes days 2-7 following injection. Error bars are standard errors across birds. A linear mixed-effects model (see Methods) was fit to the data to obtain p-values. (g) Chronic reverse microdialysis of CRH and the NMDA receptor inhibitor AP5 to the song motor output nucleus RA to assess the real-time influence of CRH on song variability. (h) Influence of AP5 and CRH on across-rendition fundamental frequency (FF) coefficient of variation (CV). Depicted are three example days of reverse microdialysis of either PBS, AP5, or CRH. Each point is the FF CV calculated across the 5 syllables flanking each syllable on either side. Gray region indicates the period of drug delivery. For the PBS control, a syringe with fresh PBS was swapped in for a 4 hour period to mimic the syringe change during drug delivery. (i) Summary of FF CV changes across birds and syllables. Shown are measures for syllables with easily computed FF (harmonic stacks) that had consistent reductions in CV following AP5 treatment. Each dot represents the average FF CV within the treatment period for a given syllable, normalized to the average FF CV in the pre-treatment period (from 2 hours before drug onset to drug onset). A linear mixed-effects model (see Methods) was fit to the data to obtain Benjamini-Hochberg adjusted p-values.