Intrinsic properties in HVC differ by projection classes and individual bird.

A: Diagram of song circuit, illustrating two main projection targets of HVC (Area X and Nucleus RA). B: Illustration of our three tutoring paradigms. Natural live tutoring occurred in a family/sibling context; in controlled live tutoring juveniles interacted solely with a single adult male; in instrumental tutoring juveniles engaged a device (pulling a string) to elicit tutor song playback. C: An example motif sung by one bird, with harmonic stacks highlighted in blue and the longest stack marked by a white asterisk. D: Example traces from an HVCRA (top panel in blue) and an HVCX (middle panel in green), evoked by the current shown in black (bottom panel). Inserts show a zoomed in view of the hyperpolarized portion of the traces with a dotted line denoting the pre-inhibition baseline. For the HVCX neuron, a post-inhibitory rebound response can be seen (red arrow). E, F: Histograms showing distributions of membrane capacitance and resistance, for 147 HVCX (green), and 42 HVCRA (blue). Solid lines show the kernel density estimation. G: Bootstrapped distributions for firing frequency, sag ratios, and membrane resistance for the mean within-bird variances (95% confidence intervals in black dashed line) compared to real mean within-bird variance (red dashed line).

Internal temporal structure in zebra finch song.

A: Durations of longest harmonic stacks in a song plotted against the remaining song motif (motif minus the longest harmonic stack) for songs in our study (green, R2: 0.47) and additional songs from other labs (yellow, R2: 0.24) and corresponding linear regression (dashed line). B: Histogram showing the percentage of song motifs that occurs before the longest harmonic stack. C,D: Scatter plots for only the songs from this study (green) and lines of best fit for the number of syllables versus motif duration (R2: 0.55) and longest syllable versus motif remainder (motif minus longest syllable, R2: 0.37).

Intrinsic properties associated with rebound excitation are related to specific timing features of zebra finch song.

A: Example voltage traces resulting from our standard current injections used to estimate cells’ post-inhibitory rebound excitability (black dotted box). Firing frequency was measured as the number of spikes over the duration of the spike train (dotted line), evoked by a 100pA depolarizing pulse (darker upper trace). Sag ratio was measured as the ratio between the minimum membrane potential (black triangle) evoked by a −100pA hyperpolarizing pulse (lighter lower trace), and the membrane potential before the release from the hyperpolarizing current injection (black square). B: Two example spectrograms showing representative song motifs of two birds. Blue boxes denote the longest duration harmonic stack in each song. C: Scatter plots of mean analyzed parameters for all HVCX for each bird, against features of song duration (error bars are standard error of the mean).

Instrumental manipulation of song learning changes intrinsic properties.

A: Diagram of the instrumental and live tutoring paradigms that yielded birds singing very similar songs. Birds were tutored with either Song A or Song B. B: Example spectrograms from one bird from each song group. Song B included an added harmonic stack (orange asterisk). C: Distributions (kernel density estimations) for all neurons grouped by song type for evoked firing frequency, sag ratio, and membrane capacitance. Asterisks represent p < 0.05 for a KS test. D: An example voltage trace for a neuron from each song group (orange for modified song B, and blue for unmodified song A). Both neurons received the same input current of +100 pA for 300 ms followed by −100 pA for 300 ms. E: Average evoked currents for varying hyperpolarized voltage steps for four birds.

Hodgkin-Huxley network model leverages rebound excitation to produce in vivo bursting properties and sequence sensitivity.

A: Experimentally recorded traces from neurons receiving hyperpolarizing and depolarizing currents at various delays (example current injection shown in bottom traces). The upper traces show voltage for rebound and direct depolarization timepoints (blue and green arrows, respectively) and highlights their area above resting potential (blue and green, respectively). Left panel shows example subthreshold responses, while right panel shows a suprathreshold response to a short delay (note overlap in blue and green areas). B: Traces from two HVCX neurons with different sag and rebound responses to the same protocol in A. One neuron shows low rebound (blue) and another shows high rebound (red). Spike time distributions relative to delay for 6 neurons are shown on the right, including the example blue and red neurons shown on the left panel. C: Model diagram and Hodgkin-Huxley model traces for the basic sequence selectivity module (start of song depicted by dashed line with bird icon). The module utilizes inhibition release (blue arrow, bottom trace) and rebound, resulting in a depolarization ‘window’ (blue highlighted area, bottom trace), and requires a second, depolarizing event during the rebound window (green bracket and arrow) to produce a spike (no spike shown). An HVCRA neuron (top, black trace) excites a phasic interneuron (yellow trace), which inhibits a tonic interneuron (orange trace), resulting in disinhibition of the HVCX neuron (bottom, green trace). Blue arrows represent excitatory synapses, yellow and orange arrows represent inhibitory synapses. D: Using the basic module from C, a backbone sequence of HVCRA neurons define timepoints (vertical lines in spectrogram) in a two-syllable song segment. The timepoints define intervals that are encoded by spikes in HVCX (green circles) which result from precise timing between disinhibition and excitation. The left inset shows a detailed view of the di-synaptic inhibition (red lines) in a portion of the greater circuit covering the entire two-syllable segment (small, dashed box). Spike waveform colors inside HVCX circles correspond to colored intervals in song. E: Multiple modeled traces from neurons in this network, participating in interval representation for the corresponding colored intervals and numbered neurons in D.

Intrinsic properties are unrelated to fundamental frequency of longest harmonic stack.

Scatter plots of mean analyzed parameters for all HVCX for birds singing natural songs, against the fundamental frequency of the longest harmonic stack. Each point is the mean value for each bird, and error bars represent standard error of the mean.

Additional correlations between intrinsic properties and temporal song structure.

Scatter plots of mean analyzed parameters for all HVCX for birds singing natural songs, against longest syllable and the sum of additional harmonic elements beyond the longest harmonic. Error bars represent standard error of the mean.

Single and multi-bursting model HVCX neurons.

A: Voltage traces from multiple neurons modeled and wired as described in Figure 5, illustrating the time dependence for the sequence sensitivity of the network model. An interneuron (orange trace) inhibits an HVCX (green trace). One HVCRA (first grey spike) di-synaptically inhibits the orange interneuron while a second, later-bursting HVCRA (later black spike) excites the green HVCX neuron. The top panel shows the outcome where the second spike arrives too late, resulting in no spike in the HVCX. The bottom panel shows a well-timed second HVCRA spike producing a spike in the HVCX (green star). B: Model circuit diagram depicting nested intervals leading to one HVCX neuron (bottom dark green circle) that bursts twice.

Time window encoded by HVCX.

Hodgkin-Huxley model module with an HVCX neuron with no voltage-gated sodium channels and varying timing of excitatory inputs. A: Overlayed traces from three interneurons synapsing onto one HVCX. B: Multiple HVCRA voltage traces overlayed (each HVCRA is depicted by a different color). C: The voltage traces of the same HVCX arising from inputs from interneurons in A, and each individual HVCRA in B. Peak voltages are shown by black dots. D: Peak amplitudes from C, and their relative timing from inhibition release (black dashed line). Baseline amplitude (blue dashed line) was taken from excitatory input that occurred before, and does not overlap with, release from inhibition.

Intrinsic property homogeneity promotes spike time homogeneity in modeled neurons.

Hodgkin-Huxley model neurons receiving identical inputs (top panel) and producing differently timed spike responses (one example model trace, middle panel). Adjusting percent variance among five modeled ionic conductances (gNa, gK, gH, gSK, and gCa-T) between 0 and 50% produced different ranges spike times (bottom panel). Each row in the bottom panel represents all spike times for 100 modeled neurons at a given range of IP variance.