By contrast, if narrow

spikes reflected passing axons, no significant correlation is expected because passing nRT axons cannot interact with TC cells. The data showed that both under urethane anesthesia and drug-free conditions, the activity of TC cells and nRT axon was not random. The two cell types fired phase-locked to the thalamic OTX015 spindles within a shank at characteristically different phases (Figures 4A and 4B). When considering local spindles only (from urethane anesthetized recordings), cross-correlograms revealed strong correlation between TC cells and nRT axons recorded on the same shank (Figure 4C). This correlation was weaker at 200 μm and was not present between shanks 400 μm apart (Mann-Whitney test). Because the spatial extent of TC-nRT correlation was compatible with the size of nRT axon terminal arbor in VB (Pinault and Deschênes, 1998), we conclude that narrow spikes are generated by the axon terminals, not by passing fibers of nRT cells. The fact that axon terminals

produced signals large enough to detect extracellularly, most probably resulted from the occurrence of strings of extremely closely spaced nRT boutons (Figure S3). Simultaneous recording of the somata of TC cells and the axon terminals of reciprocally connected nRT neurons allowed us to quantitatively investigate the structure of population from activity during sleep spindles in a cycle-by-cycle basis in freely sleeping animals (Figures 5A and S4). According to one hypothesis (see Introduction), spindles terminate due to disruption of thalamic Veliparib molecular weight synchrony by cortical input (Bonjean et al., 2011 and Timofeev et al., 2001). This model predicts that the precision of TC-nRT interaction should be degraded as the spindle progresses. To test this, we computed cross-correlograms between the two cell populations for short (six cycles, n = 5,579) and long (14 cycles, n = 3,159) spindles for each consecutive

cycles (Figure 5B). The cross-correlograms showed no marked difference in timing between spindles of different lengths and no change from cycle to cycle, indicating a constant latency of nRT activation by TC cells in every cycle of the spindles. We next assessed the jitter of TC-nRT synchrony by computing the SD of spike times relative to spindle peaks for every cycle in the same data set. This measure also showed no change with spindle progression (Figure 5C). Repeating the same two analyses for each cycle of every spindle length, in both freely sleeping and anesthetized animals, yielded identical results (Figure S5). None of the groups showed significant slope (Spearman’s rank correlation p > 0.1). We therefore conclude that decreased TC-nRT efficacy and increased jitter among thalamic cells is not a major factor in spindle termination.