Date:Feb/25th/(thu)13:30-15:00
Place:Ew305
Speaker:Dr.Kosuke Hamaguchi(Dept. of Neurobiology, Duke University Medical Center)
Title:Synaptic level analysis of corollary discharge mechanisms in
neural pathways for vocal learning

Abstract:
Songbirds learn to sing by using auditory feedback to evaluate vocal
performance in reference to a memorized acoustic model, or song template.
Part of this process could harness corollary discharge of song motor
commands to generate predictions of auditory feedback associated with the
resultant vocalization.  Intriguingly, a single brain area, HVC, contains
one population of projection neurons (HVCRA) that transmit the song motor
commands through the song motor nucleus RA and a separate population of
projection neurons (HVCX) that transmit corollary discharge to an anterior
forebrain pathway (AFP) important to song learning.  The output of the AFP,
LMAN, is believed to generate trial-to-trial song variability necessary to
song learning by injecting gnoiseh into song motor commands at the level
of RA.
We are interested in how LMAN activity (and hence trial-to-trial
variability) is controlled by the motor signals emanating from HVCRA cells.
One attractive idea is that HVCX cells transmit precise corollary discharge
so that LMAN can control song variability as an exact function of time.
However, any corollary discharge presumably must propagate to the AFP from
HVCX cells, which integrate input from a variety of sources, including HVCRA
cells, local interneurons and auditory afferents.  This raises the question
of how activity in HVCRA and HVCX cells relates to activity in LMAN neurons.
To address this issue, we made simultaneous in vivo intracellular current
clamp recordings in HVC and LMAN neurons in urethane anesthetized zebra
finches.  We used covariance analysis and coherency analysis to quantify the
relationship between neurons in these two song areas. We found strong
correlations in the subthreshold activity of HVCRA and LMAN neurons, which
supports the idea that LMAN receives timing information about the activity
of cells that transmit song motor commands.  Interestingly, the subthreshold
activity of HVCX and LMAN neurons was on average more weakly correlated,
even though HVCX cells and not HVCRA cells provide direct input to the AFP.
Since HVCX neurons have strong inhibitory inputs from HVC interneurons
(HVCI), we investigated the possibility that local inhibition modulates
signals transmitted from HVCRA cells.  Consistent with this view, we found
that the membrane potential correlation between HVCX and LMAN neurons
increased significantly when we strongly hyperpolarized the HVCX neuron, a
manipulation that renders inhibitory potentials depolarizing.
One prediction of the corollary discharge model is that HVC activity drives
LMAN activity.  However, we found that most HVCRA-LMAN and HVCX-LMAN pairs
have correlation peaks around zero time-lag, suggestive of common excitatory
inputs to these two song areas.  Alternatively, HVC may drive LMAN, but
local circuit dynamics may obscure this interaction.  To distinguish between
these possibilities, we induced bursting activity in HVC with Bicuculline
(BMI), a GABAA receptor blocker. We reasoned that if the correlation is due
to common input, LMAN activity should not be affected.  Instead, when BMI
was applied in HVC, a strong HVC leading (~ 40 msec) peak could be detected
in the LMAN spike triggered average of HVC multiunit activity.  These
results support propagation of activity from HVC to LMAN, and also suggest
that local circuit dynamics normally obscure this feedforward relationship.

More broadly, these experiments support a model in which the information
that HVCX cells transmit to LMAN results from an interaction between motor
signals from HVCRA cells and inhibitory signals from HVCI cells.  Notably, a
recent report suggests that putative HVCI cells may encode real time
auditory feedback signals.  Over time, the interaction between motor and
feedback signals converging onto HVCX cells may precisely modify corollary
discharge transmitted to the AFP, thus affecting exactly when LMAN injects
noise into the song motor command.