The role of NIf in the generation of sleep-related burst sequences in HVC

Richard Hahnloser, Alexay Kozhevnikov, and Michale Fee

Bell Labs Lucent Technologies

We have recently shown that premotor burst activity in nucleus RA of the zebra finch is directly driven by RA-projecting HVC neurons (HVC(RA)), each of which bursts at a single precise time in the song motif, and as a population burst throughout the song. This explicit representation of time has also been observed to underlie the generation of RA burst sequences during sleep. To understand the origin of the time representation in HVC, we are investigating the role of premotor inputs to HVC. Injections of lidocaine or muscimol into premotor nucleus Uvaeformis (Uva) in the sleeping bird resulted in a 2-3 fold increase in the rate of sleep-bursts in HVC. In contrast, injections into premotor Nucleus Interface (NIf) completely and reversibly abolished sleep-burst activity in HVC. Antidromically identified HVC-projecting NIf (NIf(HVC)) neurons generate dense patterns of bursts; during sleep a fraction of these bursts are highly correlated with subsequent bursts in HVC(RA) neurons. This result suggests that during singing, NIf(HVC) neurons will likewise generate dense patterns of bursts, some fraction of which will be tightly locked to the song vocalization. This prediction is being tested by recording identified NIf neurons in the singing bird. Furthermore, robust activation of the HVC neuron populations by electrical stimulation in RA or X has only a transient effect on song output (~20ms). Thus we suggest that the HVC time code originates from dynamics in NIf, is transmitted to HVC in a temporally dense code, and that the sparseness of the HVC time representation results from network or biophysical properties within HVC.