Song system overview and general intrinsic and network properties of HVC neurons in vivo and in vitro.
A. Schematic diagram showing a sagittal view of the male zebra finch song system. The vocal motor pathway (VMP, red color) contains circuits that directly pattern song output. The anterior forebrain loop (AFP, blue color) pathway contains circuits that are important for song learning and plasticity. B. HVC includes multiple classes of neurons; HVCX neurons that project to area X (blue), HVCRA neurons that project to nucleus RA (red) and HVC interneurons (HVCINT, black). HVCX and HVCRA excite HVCINT via AMPA and NMDA synapses (green arrow), while HVCINT neurons inhibit both classes of projecting neurons via GABA synapses (brown arrows with circle heads). Each class of HVC neurons is characterized with its own family of ionic currents (Daou et al., 2013). C. HVCRA neurons exhibit a very sparse activity during singing eliciting a single 4-6 ms burst at a single and exact moment in time during each rendition of the song. On the contrary, HVC interneurons burst densely throughout the song (Adopted from (Hahnloser et al., 2002)). D. Similar to HVCRA, HVCX neurons generate 1-4 bursts that are time-locked and highly stereotyped from one rendition of the song to another (Adopted from (Kozhevnikov & Fee, 2007)).
Box plots showing the ranges of ionic and synaptic currents that were allowed to vary while maintaining robust network propagation and biologically realistic in vivo behavior of all HVC neuronal classes.
A. The ionic conductances that were varied are gA of HVCRA, gSK of HVCRA and HVCX, gh and gCaT of HVCINT and HVCX. The shown ranges reflect values whereby each neuron class is able to maintain realism in terms of electrophysiological behavior and network properties. B. Ranges of values of synaptic conductances that connect two classes of HVC neurons while conserving sequential activity propagation and the general network activity.
DC-evoked action potentials in HVCRA neurons trigger iPSPs in HVCX neurons (A1) as well as fast dPSPs in HVCINT neurons (B1). Brief (∼10 ms) depolarizing current pulses (0.5 nA) applied to model HVCRA neuron (same values used as by Mooney and Prather (2005)) evokes similar responses in the corresponding model HVCX (A2) and model HVCINT (B2) neurons. DC-evoked action potentials in HVCINT neurons generate fast iPSPs in HVCX neurons (C1, D1). Similar responses were elicited in model HVCX neurons when model HVCINT neuron was stimulated by brief (10 ms) depolarizing pulses (0.5nA) (C2) or when it was given a DC-pulse of 0.5 nA for 500 ms (D2). Finally, HVCINT neurons elicit fast dPSPs when HVCX neurons are injected with 10 ms pulses of 0.5 nA current (E1), which was simulated in the model (E2). In this and subsequent figures (unless otherwise specified), HVCX neurons’ traces are represented in blue, HVCRA neurons in red, and HVCINT neurons in black. Panels in the left column adopted from Mooney and Prather (2005).
Cartoon diagram illustrating the network architecture configuration.
A. Each grey oval represents a microcircuit encoding for a sub-syllabic segment (SSS). The number of microcircuits is envisioned to be equal to the number of SSSs representing the song. Each microcircuit contains a number of HVCRA, HVCX and HVCINT neurons selected randomly from the total pool of neurons (see text). B. Within each microcircuit, HVCRA neurons are connected to each other in a chain-like mode and they send excitatory afferents to HVCINT neurons in the same and other microcircuits, selected randomly. HVCINT neurons send GABAergic synapses onto HVCX neurons in the same microcircuit only as well as to HVCRA neurons in any other random microcircuit except the microcircuit they belong to. Finally, HVCX neurons send excitatory afferents to HVCINT neurons in the same microcircuit. Activity starts by a small DC pulse to the first HVCRA neuron in the first microcircuit. Activity propagates from one microcircuit to another by excitatory coupling between the last HVCRA neuron in microcircuit i and the first HVCRA neuron in microcircuit i+1. C. During singing, the propagation of activity unfolds across the chain of microcircuits, such that neurons belonging to microcircuit x gets activated and encode for SSSx.
Spiking patterns of 120 HVCRA neurons (labeled with numbers) showing the propagation of sequential activity.
The neural traces are aligned by the acoustic elements of a spectrogram from an exemplar bird’s song illustrating the firing of HVCRA neurons with respect to ongoing part of a song. The inset shows a zoomed version of two subsequent HVCRA neurons firing patterns illustrating the delay between their individual bursts.
Effects of the AMPA synaptic conductance and A-type K+ conductance on the delay between two successive HVCRA bursts.
A. Presynaptic model HVCRA neuron (top, black) is connected to a postsynaptic model HVCRA neuron and the corresponding AMPA excitatory conductance (gAMPA) was increased from 10 nS (bottom, red) to 12 nS (bottom, blue), while keeping all other parameters fixed. Increasing gAMPA reduce the delay between the peaks of the pre– and post-HVCRA’s first spikes and increase the number of spikes in the postsynaptic neuron. B. Larger magnitudes of the A-type K+ conductance (gA) leads to longer delays to spiking. While keeping all intrinsic and synaptic parameters fixed (gAMPA = 10), increasing gA from 10 nS (bottom, red) to 13 nS (bottom, blue) delayed the onset to spiking and reduced the number of spikes. Bars on the top shows the duration in ms between the peak of first action potential in the presynaptic neuron to the peak of the first action potential in the postsynaptic neuron.
Intrinsic changes in HVCRA halts the propagation of sequential activity.
Up-regulating the A-type K+ current (A) or the Ca2+ – dependent K+ current (B) in exemplar neurons (, A) or (, B), by increasing gA 10-fold (A), or gSK from 15-fold nS (B), reduces the excitability of corresponding HVCRA neuron markedly, eliminating its corresponding burst and breaking the sequence.
Activity patterns for 10 HVCINT and 10 HVCX neurons are illustrated.
A. HVC interneurons displays dense spiking and bursting throughout the song, due to the dense HVCRA – HVCINT and HVCX – HVCINT excitatory coupling (Fig. 4). B. HVCX neurons display 2-4 rebound bursts that vary in their strength and duration due to HVCINT – HVCX inhibitory coupling as well as intrinsic properties (Fig. 9).
Patterned activity of HVC interneurons illustrated for an exemplar interneuron ().
For this neuron, were selected randomly from the pool of HVCRA’s and HVCX’s to form excitatory coupling. The number of bursts in is controlled by the number of bursts that each of the HVCRA and HVCX neurons that connect to it exhibit. The strength of each of the bursts depend on the magnitude of gAMPA from the corresponding neuron(s) they cause it as well as the simultaneous bursting of any of the projecting neurons. For example, the asterisk (*) shows a region of dense firing in because neurons elicit their spikes at similar times causing a potentiated response in . HVCX neurons exhibit multiple sags and rebounds because they’re receiving inhibition from several interneurons (not shown here).
Intrinsic changes in HVCINT halts the propagation of sequential activity.
Up-regulating the T-type Ca2+ current conductance (A) or the hyperpolarization-activated inward current conductance (B) in exemplar HVCINT neurons eliminates sequence propagation. Increasing gCaT in HVC23 10-fold results in dense bursting and firing in , which in its turn blocks the bursting of due to the inhibitory GABA coupling between and (A). Similarly, increasing gH in 10-fold results in increased firing in , which in its turn blocks the bursting of due to the inhibitory GABA coupling between and (B). Sequence of HVCRA bursts truncated at the level of for better visualization purposes.
Activity patterns illustrating the interplay between HVC interneurons and X-projecting neurons. (blue) is an exemplar projecting neuron that receives inhibition from (black) and (green) due to the random inhibitory coupling
(A). is inhibited whenever and are firing, eventually escaping inhibition at some intervals and eliciting rebound bursts due to the activation of IH and ICaT. HVC interneurons can inhibit multiple HVCX neurons. is an exemplar from the network that inhibits and (orange) (C). Bursts in elicit subsequent bursts in and unless silenced by other HVCINT neurons that connect to them. Zoomed versions of (A) and (C) are show in (B) and (D).
Intrinsic changes in HVCX halts the propagation of sequential activity.
A. Up-regulating the hyperpolarization-activated inward current conductance in a sample HVCX neuron (, by increasing its g 10-fold) leads to increased firing in all HVCINT neurons it connects to (for example, ), which in its turn inhibits all HVCRA neurons it connects to (for example, , being first in the pool that it inhibits) breaking the sequence at the level of . B. Up-regulating the T-type Ca2+ current conductance in a sample HVCX neuron (, by increasing its gCaT 15-fold) leads to stronger rebound bursts in which leads to increased firing in all HVCINT neurons it connects to (for example, ), which in its turn inhibits all HVCRA neurons it connects to (for example, , being first in the pool that it inhibits) breaking the sequence at the level of . C. Finally, down-regulating the Ca2+ – dependent K+ current conductance in a sample HVCX neuron (, by setting its gSK to zero) leads to stronger rebound bursts in which leads to increased firing in all HVCINT neurons it connects to (for example, ), which in its turn inhibits all HVCRA neurons it connects to (for example, , being first in the pool that it inhibits) breaking the sequence at the level of . Sequence of HVCRA bursts truncated at the level of for better visualization purposes.
Altering the intrinsic properties of HVCx neurons disrupt network activity in 59% of the cases (out of 100 simulations) where the maximal conductances (gCaT, gSK and gH) of HVCx neurons are randomly varied within their allowed ranges (Figure 2A).
As a result, some HVC interneurons (B) and X-projecting neurons (C) generated non-biological realistic firing patterns, halting the propagation of sequential activity in RA-projectors (A).
