Cell type specializations of the vocal-motor cortex in songbirds

Abstract

Identifying molecular specializations in cortical circuitry supporting complex behaviors, like learned vocalizations, requires understanding of the neuroanatomical context from which these circuits arise. In songbirds, the robust arcopallial nucleus (RA) provides descending cortical projections for fine vocal-motor control. Using single-nuclei transcriptomics and spatial gene expression mapping in zebra finches, we have defined cell types and molecular specializations that distinguish RA from adjacent regions involved in non-vocal motor and sensory processing. We describe an RA-specific projection neuron, differential inhibitory subtypes, and glia specializations and have probed predicted GABAergic interneuron subtypes electrophysiologically within RA. Several cell-specific markers arise developmentally in a sex-dependent manner. Our interactive apps integrate cellular data with developmental and spatial distribution data from the gene expression brain atlas ZEBrA. Users can explore molecular specializations of vocal-motor neurons and support cells that likely reflect adaptations key to the physiology and evolution of vocal control circuits and refined motor skills.

Keywords: CP: Neuroscience; arcopallium; birdsong; cell types; evolution; glia; interneurons; pyramidal neurons; snRNA-seq; upper motor neuron; vocal learning.

Figure 1.. Cell classes of the intermediate arcopallium (AI)

(A) Workflow to generate single-nuclei transcriptomes (snRNA-seq) of AI from adult male zebra finches. Nucleus RA receives input from LMAN and HVC and projects to vocal and respiratory centers in the brainstem. Sections of the caudal telencephalon were prepared and the AI region (green) containing RA, AId, and AIv was microdissected, followed by nuclei isolation, droplet-based 103 Genomics profiling, and Illumina sequencing. (B) Uniform manifold approximation and projection plot visualization of AI cell classes (n = 2 animals; nuclei = 1,504). (C) Dot plot for cell type-defining genes used for cluster identity. SNAP25, neuronal; SLC17A6, excitatory; GAD2, inhibitory; FGFR3, astrocyte; PLP1, oligodendrocyte; CSF1R, microglia; CFD, endothelial. On this and similar plots in all figures, color indicates normalized expression levels and size indicates percentage of labeled cells (D) Proportions of AI cell types from snRNA-seq: astrocytes (33.2%), oligodendrocytes (9.6%), microglia (3.2%), endothelial cells (2.5%), excitatory neurons (33.6%), and inhibitory neurons (17.8%). (E) Nissl staining (upper left) and in situ hybridization images for cell type-defining genes: excitatory (SLC17A6), inhibitory (GAD2), astrocytes (ASS1), oligodendrocytes (PLP1), and microglia (RGS10). Unless otherwise indicated, in situ images in all other figures are from this same region of frontal sections. Scale bars, 400 μm. (F) Proportion of each cell class across AI subdivisions from images in (E). AA, anterior arcopallium; AD, dorsal arcopallium; AM, medial arcopallium; AId, dorsal AI; nAId, neck of the AI; DM, dorsomedial nucleus of the intercollicular complex; HVC, proper name; LMAN, lateral magnocellular nucleus of the anterior nidopallium; nXIIts, tracheosyringeal division of the hypoglossal nucleus; RA, robust nucleus of the arcopallium; AIv, ventral AI. See also Figure S1 and Table S1.

Figure 2.. Molecular specializations of AI excitatory neurons

(A) Dot plot of top defining markers of cluster Excit_1. (B) In situ hybridization shows that SRD5A2, a highly differential Excit_1 marker, is exclusively expressed in RA. Scale bar, 400 μm. (C) Age and sex differences in Excit_1 marker expression; bulk RNA-seq plots show that many defining markers of this RA-specific cell type increase in males only during the period of song learning. (D) Dot plot of the top defining markers of clusters Excit_2-3. (E) In situ hybridization shows that ADCYAP1, one of the most differential markers of clusters Excit_2-3, is highly expressed in the AI but not in RA. Scale bar, 400 μm. (F) Age and sex differences in Excit_2-3 marker expression; bulk RNA-seq plots show that many defining markers of Excit_2-3 that are low or absent in adult male RA decrease in males only during the period of song learning. (G) Dot plot shows that growth factor genes developmentally upregulated in male RA are markers of the RA-specific Excit_1 cluster. (H) Genomic maps of chromosome 1 in chicken and zebra finch, exhibiting the LRRC32 duplication in the latter. (I) Age and sex differences in LRRC32 genes; bulk RNA-seq plots show that only LRRC32-2 is developmentally regulated in male RA. Bulk RNA-seq plots in (C), (F), and (I) represent values of transcript abundance and are false discovery rate (FDR) < 0.01 for the male 20-50 comparison and FDR > 0.01 for the female 20-50 comparison, with the exception of LRRC32-1, which is FDR > 0.01 for both comparisons. Age values in (C), (F), and (I) are in days post hatch (dph). See also Figures S2 and S3.

Figure 3.. Inhibitory subtypes of the AI

(A) Markers that differentiate the four GABAergic neuron subtypes. (B) Spatial distribution of inhibitory subtypes based on mapping of in situ hybridizations for VWC2 (Inhib_1), NPY (Inhib_2), COLEC12 (Inhib_3), and SLIT2 (Inhib_4); corresponding images in Figure S4B. (C) Quantification of spatial distribution of inhibitory subtypes; plotted are the proportions of cells from each subtype per AI subdivision. (D) GAD2 in situ hybridization of RA (white border) in juveniles (20 dph) and adult males. Open arrow denotes example of large soma GAD2⁺ morphotype; closed arrow denotes example of small soma GAD2⁺ morphotype. Scale bar, 400 μm. (E) Quantification of GAD2⁺ soma size in RA and in the caudal arcopallium outside of RA (arco) in juvenile (20 dph) and adult males. Adult RA vs. 20 dph RA, p < 0.0001; adult RA vs. adult arco, p < 0.0001; 20 dph RA vs. 20 dph arco, p = 0.9596. (F) Counts of GAD2⁺ cells of different soma sizes in juvenile (20 dph) and adult male RA. The larger somata are notably absent in juveniles. Adult vs. 20 dph RA, p < 0.0001. (G) In situ hybridization for NPY reveals two morphotypes in RA but not in AId. Scale bar, 100 μm. (H) Quantification of soma size of inhibitory subtypes in RA and AId. VWC2 RA vs. AId, p = 0.8679; NPY RA vs. AId, p = 0.0814; COLEC12 RA vs. AId, p = 0.9804; SLIT2 RA vs. AId, p = 0.9804. Shown in (E) and (H) are the median, 25th and 75th percentiles (box), and maximum and minimum values; shown in (F) are average values. ***p < 0.0001, *p < 0.1 following one-way ANOVA multiple comparisons. n = 3–4 animals per condition. See also Figure S4.

Figure 4.. Electrophysiological characterization of GABAergic interneurons in RA of adult males

(A) GABAergic cells in RA slices. AAV-mDlx-eGFP was injected into RA (top left); after adequate survival, targeted recordings of GFP⁺ cells (n = 24 total) were performed (top right). Image of slice from injected brain (bottom) shows representative examples of small soma (open arrowhead) and large soma (closed arrowhead) neurons in RA. Scale bar, 20 μm. (B) Representative recordings of responses to negative current injections (top) in cells with high (black traces, bottom) or low (red traces, middle) input resistance. Shown are five overlaid example traces with different levels of current injected used to calculate capacitance (C_m). Insets of highlighted (yellow) regions in the traces show that larger mini PSPs are detected during current injections in cells with high (black traces) vs. low (red traces) input resistance. (C) Graph plotting the input resistance vs. the time constant of recorded GFP⁺ cells; dot color indicates the C_m values of the individual neurons. (D) Frequency distribution histogram of calculated C_m for all GFP⁺ cells recorded; bins = 10 pF. Histogram was fit by two Gaussian curves revealing distinct populations of neurons as determined by the calculated C_m. (E) Principal component (PC) analysis of passive and active membrane properties of recorded cells. PC1 accounted for 31.7% of the variance, PC2 accounted for 23.0%, and PC3 accounted for 13.1%. (F) Proportions of large vs. small cells that were spontaneously active. (G) Overlays of average action potential waveforms (left) and phase plots (right) for all large and small GFP⁺ cells recorded during a 1 s +100 pA current injection; solid red and black traces are the average waveforms and phase plots for each cell subtype. (H) Overlay of recordings in large (red trace) and small (black trace) neurons during the first 24 ms of a +100 pA current injection. (I) Cumulative distributions of firing frequencies from large (red) and small (black) cells. Kolmogorov-Smirnov test, D = 0.4126, n = 289 large cell events and 1,044 small cell events. In (E)–(I), large soma/high C_m cells are labeled in red, and small soma/low C_m cells are labeled in black. See also Figure S5.

Figure 5.. Non-neuronal specializations in song nuclei of adult male zebra finches

(A) Dot plot with top differentially expressed genes in the oligodendrocyte cluster and putative oligodendrocyte markers on ZEBrA based on expression patterns. (B) Top: drawing of parasagittal section with forebrain song nuclei and fiber tracts (adapted from ZEBrA: www.zebrafinchatlas.org and from); areas in colored rectangles are shown in (C). Bottom: in situ hybridization for UGT8 (from ZEBrA) shows cellular expression pattern in RA consistent with oligodendrocytes. (C) In situ hybridization for UGT8 (from ZEBrA) shows oligodendrocyte enrichment in pallial song nuclei HVC (red), LMAN (green), and RA (blue) compared to surrounds. Bottom panels show views of the fiber tract from HVC to RA (black) and the occipitomesencephalic tract (OM) (gray); rectangle locations shown in (B). (D) Dot plot with top differentially expressed genes in the astrocyte clusters. (E) In situ hybridization for CRISPLD1, an Astrocyte_2 enriched gene, shows higher density of labeled cells in RA (top) compared to AId (bottom). (F) Dot plot for CLIC4, a gene enriched in endothelial cells. (G) In situ hybridization image for CLIC4 in frontal section shows higher density of labeled cells in RA than elsewhere in the AI. Arrowheads indicate expression in ventricle. Scale bars, 100 μm in (B) (bottom left); 400 μm in (C), (E), and (G). See also Figure S6.

Figure 6.. Developmental changes in non-neuronal composition in RA

(A) Most genes that increase in RA in both males and females are highly enriched in the oligodendrocyte cluster. (B) Developmental gene expression plots for four highly specific oligodendrocyte markers from (A). (C) In situ hybridization for PLP1 showed increased developmental expression in RA in both sexes; labels represent age in dph. Scale bar, 250 mm. (D) Endothelial cell markers show a male-specific developmental increase in expression. (E) Dot plot for endothelial cell markers shown in (D). Bulk RNA-seq plots in (B) and (D) are FDR < 0.01 for the male 20–50 comparison and FDR > 0.01 for the female 20–50 comparison.

Figure 7.. Integration of spatial distribution and cell expression data and summary of RA cell types

(A) Cell type expression data provide insights into gene expression patterns seen by in situ hybridization; in the examples shown, cell class distributions (left) help explain differential expression of positive and negative RA markers (right; in situ hybridization data from ZEBrA: www.zebrafinchatlas.org). (B) Integration of cell type and developmental datasets reveals sex- and age-dependent regulation of specific RA cell type markers. Plotted is the log₂ fold change for the top cluster-defining genes in each comparison; for age comparison, 20:50 dph ratios; for sex comparison, male/female at 50 dph ratios. Scale bar, 500 μm. (C) R Shiny applications to interact with the datasets presented in this study are linked on the homepage of ZEBrA; image adapted from. (D) Examples of using the applications depicted in (C) for assessment of cell type (top) and sex/age developmental (bottom) profiles in RA. (E) RA excitatory neurons had a specialized gene expression profile that differed from those in AId. Similar GABAergic subtypes were present in RA and AId, but in RA Inhib_1 had a lower density, and a subset of Inhib_2 had large soma. Non-neuronal differences included greater numbers of oligodendrocytes in RA and an RA-specific astrocyte subtype. (F) Summary of major cell types and subtypes found in RA and AId, and respective sets of genes linked to cell excitability. (G) RA is hypothesized to have evolved as a specialization of preexisting AId, RA projection neurons (RAPNs) arising via specialization of existing AIdPNs. (H) Changes in RA cell type markers during vocal development. With an enlargement of RA in males, we observe increases in some RA excitatory markers and non-neuronal markers and a decrease in GABAergic markers. These specializations may contribute to the physiology of producing learned vocalizations. Note that schematics are not to scale.