Figures and data
Song degradation and transcriptomic changes after chronic inactivation of inhibitory neurons in HVC by viral expression of tetanus toxin (TeNT).
a (left) Examples of spectrograms of songs produced by an animal treated with TeNT before, and 5, 20 days post-injection (dpi). a (right) 2D principal component analysis (PCA) visualization of syllables aka song segments produced by the animal on the same day as the spectrogram example on the left. Note: even when the songs are degraded, we continue to refer to an uninterrupted length of vocalization as a ‘syllable or song segment’. Individual dots indicate single syllable renditions. In the unperturbed song (before virus injection), syllables cluster into distinct groups (each differently colored circle represents a distinct syllable in the PCA space). b Log-fold change in total number of cells per cell type between TeNT-treated and control animals. c Volcano plot showing statistical significance over magnitude of change of differentially expressed genes between TeNT-treated and control animals across all cell types. Dotted lines indicate fold change = 1.5 and p value = Bonferroni corrected alpha of 0.05. d Violin plots of normalized counts of major histocompatibility complex 1 α chain-like (MHC1) (ENSTGUG00000017273.2) and beta 2 microglobulin-like (B2M) (ENSTGUG00000004607.2) genes in control (n=2, blue) and TeNT-treated (n=2, red) animals per example cell cluster. A star indicates a significant increase in gene expression in TeNT-treated animals compared to control (p < 0.05 and fold change > 1.5).
Behavioral and chronic electrophysiology recordings in HVC after chronic inactivation of inhibitory neurons by viral expression of TeNT.
a Examples of spectrograms of songs produced by an animal treated with TeNT before, and 60 days post-injection (dpi). Data from this animal (V648) are also shown in c. b 2D principal component analysis (PCA) visualization of syllables aka song segments produced by the animal on the same day as the spectrogram example in a. c Different features of the vocalizations averaged/grouped before and after stereotaxic injection of control (GFP-expressing) or TeNT virus. The grouped mean frequency and goodness of pitch measures represent averages over song segments. Control animals (Or295 and PK31) are labeled in blue (average of n=2); TeNT-treated animals (V648, B183, B138 and Or144) are labeled in orange (average of n=4). Traces indicate averages over syllables sung within the 3 closest days of recorded vocalizations. Error bars indicate standard error. Values are normalized relative to distributions 5 days before perturbation (see Methods). d (top) Average spectrogram of 5 vocalizations in a control animal d (bottom) Average raw deflection traces aligned to the vocalizations. e (top)) Average spectrogram of 5 vocalizations of a TeNT-treated animal d (bottom) Average raw deflection traces aligned to the vocalizations. f Raw electrophysiology traces during night time (lights off period) from TeNT-treated and control animals 5 days post-injection (dpi). Note: electrodes were implanted on the same day of viral injection. Insets show examples of stereotypical deflections that at 5 dpi were approx. 50 ms in duration, and - 530 µV in amplitude, in TeNT-treated animals, compared to 96 ms and -140 µV in control animals. g (left) Quantification of song quality of chronically implanted animals over time. Control animals (PK31 and Or195) are depicted in shades of blue. TeNT-treated animals (B138 and Or296) are depicted in shades of orange and red. g (right) Rate of deflection events during night time. Control animals (PK31 and Or195) are depicted in shades of blue. TeNT-treated animals (B138 and Or296) are depicted in shades of orange and red. The highest rate of sleep voltage deflections in TeNT-treated animals occurred during the periods when the songs were most degraded, as shown on the left.
Acute electrophysiology recordings after chronic inactivation of inhibitory neurons in HVC by viral expression of TeNT.
a Histological section example of a brain slice from a control animal shows the track of the probe (NPIX coated with DiI) insertion (red) and GFP from the viral injection (green). The raw electrophysiology traces are color-coded by brain area (HVC in red, NCL in black, RA in blue). b Spectral decomposition of LFPs of the average normalized deflections during lights-off acute recordings in control animals and TeNT-treated animals at 3-6, 20, and 70 dpi. c Average firing rates of neurons within HVC and RA in control animals (n=3) and TeNT-treated animals at 3 (n=1), 4-6 (n=4), 20 (n=4) and 70 (n=4) dpi. In HVC, the increase in average firing rate at 3 dpi after TeNT injection significantly differs from control (p-value from rank sum test: 0.4*10-6), as well as the decrease in average firing rate after TeNT injection at 20 dpi (p = 0.002) and 70 dpi (p = 0.57*10-11). In RA, the decrease in average firing rate at 3 dpi significantly differs from control (p = 0.43*10-6), as well as the increase in average firing rate at 20 dpi (p = 0.09*10-9) and 70 dpi (p = 1.45*10-13). The stars above the bar plots indicate statistical significance (* = p < 0.005, ** = p< 0.01, *** = p < 0.001). d Model of HVC-RA circuit dynamics before and after TeNT-perturbation in HVC based on c. e (top) Example traces of the firing rate of three individual neurons during deflection events that lock or do not lock to a specific phase of the alpha oscillation (shown in red in bottom plots) in HVC of control animals and TeNT-treated animals at 3-6, 20, and 70 dpi. e (bottom) Examples of neuronal firing (indicated by black dots) for one neuron per condition in HVC. f Probability distribution of the phase of firing events relative to the alpha oscillation of the LFP in control animals and TeNT-treated animals at 3-6 dpi (n=4), 20 dpi (n=4), and 70 dpi (n=2).
Song degradation and recovery after interneuron perturbation without LMAN a brain area necessary for song learning
a Schematic of the experimental design. We injected adult male animals n=3 with ibotenic acid to lesion LMAN 7-12 days before the TeNT-virus injection into HVC and recorded the song before lesion, after lesion and after interneuronal manipulation. b Spectrogram example of the song of an adult male zebra finch showing the behavioral readout of the schematic presented in a. N=3 adult males degraded and recovered their song after interneuron perturbation in HVC without LMAN (LMAN lesion was confirmed by CGRP immunohistochemistry shown in Supplementary Figure 3-5 along with spectrogram examples)
Examples of abnormally long syllable lengths after injection of interneuron muting virus in two animals.
a Syllable length durations for the length of song degradation and recovery. The Y axis depicts the length of the syllables in milliseconds plotted over days post-injection (dpi) of either TeNT virus. TeNT-treated animals displayed a short period during which some vocalizations were of length not observed in normal animals and eventually became highly variable and shorter (shifts to shorter length sounds). The green rectangle highlights the day post-injection portrayed in B for each animal. B Histogram of syllable durations (blue trace is before injection of virus, orange trace is after injection of virus). The green rectangle highlights vocalizations of abnormal length.
Example spectrograms of a TeNT-treated animal (B138) during song degradation and recovery.
Vocalizations between 15 dpi and 30 dpi were much shorter than the first long syllables shown in Figure 1 A at 5 dpi.
Song degradation and recovery after chronic removal of inhibition in an animal without LMAN.
Spectrograms are showing the song of the animal before and after LMAN lesion at 5 and 40 days post viral injection (dpi). The histology image shows the amount of LMAN left (based on CGRP staining) in the right hemisphere (RH).
Song degradation and recovery after chronic removal of inhibition in an animal without LMAN.
Spectrograms are showing the song of the animal before and after LMAN lesion at 5 and 60 days post viral injection (dpi). The histology images indicate the amount of LMAN left (based on CGRP staining) in the left (LH) and right (RH) hemispheres.
Song degradation and recovery after chronic removal of inhibition in an animal without LMAN.
Spectrograms are showing the song of the animal before and after LMAN lesion at 5 and 55 days after viral injection (dpi). The histology images indicate the amount of LMAN left (based on CGRP staining) in the left (LH) and right (RH) hemispheres.
Example electrophysiology traces averaged over 5 instances of normal or 5 instances of degraded vocalizations and averages of sleep voltage deflections.
a Averaged spectrogram of degraded vocalization (n=5) 5 days post-electrode-implantation in a chronically recorded TeNT-treated animal. The plot below the spectrogram shows the raw averaged trace of extracellular recording. Below the raw trace is the averaged continuous wavelet transform of the local field potentials (LFP, 1-300Hz). The plots show a large deflection event (similar to those seen during lights-off in Figures 2 and 3) right before the onset of the vocalization in the TeNT-treated animal. B Averaged song spectrogram (n=5) 5 days post-electrode-implantation from a chronically recorded control animal. The plot below the spectrogram shows the raw averaged trace of extracellular recordings. Below the raw trace is the averaged continuous wavelet transform of the local field potentials (LFP, 1-300 Hz). The averaged control song shows more and smaller amplitude deflections mostly during the vocalization compared to the TeNT-treated vocalization. C-D Spectral decomposition of local field potentials (LFPs) of the averaged deflections during night time at 5, 20, and 60 dpi in one control (C) and one TeNT-treated animal (D). The vertical red line depicts the trough of the raw deflection trace. The % increase is the relative increase compared to non-deflection events in the same recording timeframe of the same animal (Methods). Deflections of TeNT-treated animals in the 15-30 Hz range were approx. 27 times larger than those in control animals at 5 dpi. These differences are statistically significant between control and TeNT-treated groups, but not within the control group (p=0.7631 between controls, p<10-35 between all other pairs). The deflections across all animals and frequencies become more similar by 60 dpi (15-30 Hz: p=0.1371 between controls, p<10- 6 between all other pairs; 30-70 Hz: p=0.7493 between controls, p<10-2 between all other pairs). For details on statistics see Methods.
Example traces of raw deflections in the acute Neuropixel recordings during lights-off periods.
Control animals barely showed any visible deflection events, while TeNT-treated animals (example traces shown at 3-6, 20, and 70 dpi) displayed large amplitude voltage deflection events.
Relationship between alpha or gamma oscillations during lights-off voltage deflection events and local neuronal firing in HVC and RA from the acute NPIX recordings
a The normalized probability distribution of neurons locally within HVC fire during a specific phase (angle) of the gamma (30-40 Hz) oscillations extracted from the LFP signal of the averaged deflection events at 3-6 dpi (n=4 animals), 20 dpi (n=4 animals), 70 dpi (n =2 animals that recovered their song by then). There was a slight change in local neuronal firing to the angle and locking precision to gamma oscillations that resembled control by 70 dpi. B Normalized probability distribution of neurons locally within RA fire during a specific phase (angle) of the alpha (1-10 Hz) oscillations in HVC extracted from the LFP signal of the averaged deflection events at 3-6 dpi, 20, 70 dpi. No change in RA spontaneous neuronal firing to alpha oscillations in HVC during the deflection events over the course of the manipulation. C Normalized probability distribution of neurons locally within RA fire during a specific phase (angle) of the gamma (30-40 Hz) oscillations in HVC extracted from the LFP signal of the averaged deflection events at 3-6, 20, and 70 dpi. We observed no change in RA spontaneous neuronal firing to the gamma oscillations in HVC during the deflection events over the course of the manipulation.
Quantification of the angle relationship between alpha (1-10 Hz) and gamma (30-40 Hz) frequencies during deflection events in control and TeNT-treated animals during acute head-fixed recordings.
A: The average difference in power (at alpha, 1-10 Hz, and low gamma 30-40 Hz frequency ranges) between voltage deflection and non-deflection events in control, 3-6, 20, 70 dpi. The power content in the alpha range increased in a statistically significant way (Wilcoxon, rank sum test) between control and 3-6 (p=2.9*10^-36), 20 (p=8.6*10^-167), 70 (p=5.4*10^-4) dpi animals. However, the increase in power between control and 3-6 (p=2.2*10^-37), 20 (p=6.3*10^-27) dpi is statistically significant but returns to control level by 70 (p=0.37) dpi. The stars above the bar plots (*) indicate statistical significance (*: p < 0.005, **: p< 0.01, ***: p < 0.001). B Examples of neuronal activity in a control animal, TeNT-treated animals at 3 and 20 dpi. The red arrows highlight the “superbursts” or extreme firing levels within HVC and RA which we observed in 3 animals (two animals at 3-4 dpi and one at 20 dpi) in a total of seven instances.C: The polar histograms of the angle of the alpha oscillations (1-10Hz) at the maximum amplitude of the gamma oscillation (30-40 Hz) during deflection events. The red distribution represents a randomly shuffled dataset, while the blue is the true distribution of angles in control (n=3), and TeNT-treated animals at 3-6 dpi (n=4), 20 dpi (n=4) and 70 dpi (n=4) during deflection events. D: Relationship of alpha and low gamma oscillations during deflection events in control (n=3), 3-6 (n=4), 20 (n=4), and 70 (n=2 animals) dpi animals (over animals and conditions). The probability distribution of a specific angle of the low-frequency oscillation at the maximum amplitude of the gamma oscillation. E: The results of the Kolgomorov-Smirnov test on the cumulative density function (CDF) to assess if the change in probability distribution shown in C is statistically significant from control distributions at 3-6, 20, and 70 dpi. The purple (20 dpi) and orange (3-7 dpi) distributions differ significantly from the blue control and the 70 dpi green distributions. The 70 dpi population is not significantly different from the control group.
Quality control of the single-cell RNA sequencing HVC datasets from control and TeTN-treated animals at 25 days post-injection (dpi).
a “Knee plots” showing the set of barcodes (top row) and number of genes detected (bottom row) over UMI counts. The dashed lines depict the quality filtering cutoff. B-C Barplot depicting the fraction of cells from each replicate per cluster for control (B) and TeNT (C), normalized (by dividing) to the total number of cells in each replicate. Control and TeNT datasets were clustered separately using the Leiden algorithm. The equal distribution of replicates across the clusters suggests that technical effects do not dominate the clusters. Thus, we did not perform batch correction. The numbers on top of the bars indicate the total number of cells in each cluster. D Barplot depicting the fraction of cells from each dataset in the cell type clusters obtained after jointly clustering the control and TeNT datasets. The numbers on top of the bars indicate the total number of cells in each cluster.
Heatmap of top 5 differentially expressed genes per annotated cell type/cluster obtained by single-cell RNA sequencing of HVC from control and TeNT-treated birds at 25 dpi.
Differentially expressed genes between clusters were identified using Scanpy’s rank_genes_groups (p values were computed using a t-test and were adjusted with the Bonferroni method for multiple testing. They were then confirmed by comparison to p values generated with the nonparametric Wilcoxon test with Bonferroni correction). The heatmap depicts the min-max scaled expression for each gene.
In situ hybridization of microglia marker gene RGS10 in adult male control, TeNT-treated and juvenile male HVC & MHC1 gene in adult male control and TeNT-treated HVC.
a Histological sections of HVC (in control and TeNT-treated animals at 25 and 90 dpi) after in situ hybridization of RNA probes for RGS10 (a gene marker for microglia). B Histological sections of HVC in naive juvenile males (at 20, 50, and 75 days post-hatching (dph)) after in situ hybridization of RNA probes for RGS10. C Histological sections of HVC (from control and TeNT-treated animals at 25 and 90 dpi) after in situ hybridization of RNA probes for MHC1. Black/darker dots indicate enzyme reactions resulting in successful probe localization and suggest target gene expression.
Quantification of the in situ hybridization against microglia marker gene RGS10 in adult male control, TeNT-treated and juvenile male HVC; and MHC1 in adult male control, TeNT-treated animals.
A-B Quantification of the in situ hybridization for RGS10 between control (n=4 animals) and TeNT-treated animals at 25 dpi (n=4) and 90 dpi (n=4). C-D Quantification of the in situ hybridization for RGS10 between juvenile males at 20, 50, and 70 days post-hatching (dph) (n=4). E-F Quantification of the in situ hybridization for MHC1 between control (n=4) and TeNT-treated animals at 25 (n=4) and 90 dpi (n=4). Error bars represent standard deviation.
Histology of electrode array location in HVC in the chronically implanted animals.
The white dotted line outlines HVC. Some sections display missing tissue due to the removal of the electrodes after perfusion of the animals. The stronger cyan signal indicates glial scar formation around the electrode array, which provides an approximation of the location of the electrodes. Electrodes located closer to the bottom of HVC close to the shelf were not used for analysis.
Histology to confirm the high-density silicone electrode location in the acute head-fixed animal recordings.
The red trace represents the electrode location. The green trace represents the second electrode location in animals that were recorded twice, 40 days apart. The white labels represent the animal IDs. “LH” and “RH” stands for left and right hemisphere, respectively.
Comparison of different pre-processing methods for the HVC single-cell RNA sequencing datasets and marker genes.
a Heatmap showing min-max scaled expression of cell type marker genes for each cell type (data from both control and TeNT-treated animals). B Number of cells retained after quality control for each dataset and alignment method. C Mean UMI counts per cell for each dataset and pre-processing method. D Percentage of reads confidently mapped to transcriptome for each pre-processing method.
Amplitudes and durations of the chronic voltage deflections measured throughout the recording.
A Mean duration (calculated as the distance from the onset to the half-width point of the event) in ms of voltage deflection events with standard deviation, each row represents an event from one control (Or 295, PK31) or TeNT-treated (B138, Or296) animal. The data was sampled at 3-5, 15, 30, 45, 60, and 75 dpi. B Mean amplitudes (in µV) of voltage deflection events with standard deviation.
