, 2008; Mizuseki et al., 2009, 2011). Spike sorting was carried out offline from the digitally high-pass filtered (0.8–5 kHz) data using an automatic clustering Alpelisib price algorithm (http://klustakwik.sourceforge.net) (Mizuseki et al., 2009). Principal cells and interneurons were separated on the basis of their autocorrelograms, combination of trough-to-peak latency,

and the asymmetry index of the filtered spike waveform, bursting properties, and mean firing rates (Supplemental Experimental Procedures). Two major comparisons of firing patterns and LFP were used. First, changes across sleep were defined as differences between the first and the last non-REM episodes in a sleep session. Second, the duration of REM and non-REM episodes was normalized (100% each epoch) and divided into equal normalized thirds. Changes within the episodes were then analyzed by comparing the first and the last thirds of each REM and non-REM episode. Third, changes within REM episodes refer to differences between the first and the last thirds of each REM. Ripple events were detected during nontheta periods from the band-pass filtered (120–250 Hz) trace by defining periods during which ripple power is continuously greater than mean 2 SD, and peak of power in the periods was greater than mean 3 SD of ripple power. Three approaches were used

to characterize firing patterns associated with ripples. (1) Within-ripple firing rate: all www.selleckchem.com/products/at13387.html ripples within a given epoch E were concatenated. Within-ripple firing rate for almost each cell C is defined as the number of spikes detected in the concatenated ripple epochs divided by the total concatenated time ( Figure 1Bvii). (2) Ripple participation: for each cell C, ripple participation is the percentage of ripples in

E in which C fired at least one spike ( Figures 1Bix and 2B, bottom, left). (3) Ripple participant firing rate: for each pyramidal cell C, only those ripples in a given epoch E in which C fired at least one spike were concatenated. Ripples in which C fired no spikes were excluded. Ripple participant firing rate was the total number of spikes divided by the total time in the concatenated subset of ripples ( Figure 1B). These methods were designed to disambiguate ripple-related firing rate changes due to increased spiking of the neuron C in the same number of ripples in different sleep episodes from increased participation of the neuron in more ripple events without changing the firing rate within individual ripple participation events. The relationship between LFP and firing patterns were examined using spectral methods. Further details about the experimental techniques are available in the Supplemental Experimental Procedures. We thank Mariano Belluscio, Adrien Peryache, and Richard W. Tsien for comments on the manuscript. This work was supported by the International Human Frontiers Science Program Organization, the U.S.