Channel activation was similar in all four types of receptors (20%–80% rise times of ∼0.3 ms), but the time courses of deactivation ( Figure 5A) and desensitization ( Figure 5C) were markedly different, strongly depending upon the subunit composition of the AMPARs. While GSG1-l alone caused a moderate slowing of both kinds of channel closure very similar to TARP γ-2 (p < 0.001, Wilcoxon rank test for ± GSG1-l; Figures 5B and 5D),

it largely reversed the pronounced effects of CNIH-2 on the time constants of deactivation and desensitization when coassembled into the same AMPARs (p < 0.001, Wilcoxon rank test for CNIH-2 versus CNIH-2+GSG1-l; Figures 5B and 5D). Moreover, receptor channels assembled from GluA1, GluA2, CNIH-2, and GSG1-l no longer exhibited the marked nondesensitizing steady-state current (Iss) observed with receptors composed

of the GluA1, GluA2, and CNIH subunits learn more alone (Iss of 25% ± 10% [mean ± SD, n = 20] and 6% ± 3% [n = 12] for GluA1+A2+CNIH-2 and GluA1+A2+CNIH-2+GSG1-l channels, respectively). In contrast to the PS 341 moderate slowing of desensitization, GSG1-l decelerated the reverse process, recovery from desensitization, by almost 10-fold, and a pronounced slowing was still present upon addition of CNIH-2, albeit to a lesser extent (p < 0.001, Wilcoxon rank test for ± GSG1-l and ± GSG1-l+CNIH-2; Figures 5E and 5F). Interestingly, the dominant effects of GSG1-l over CNIH-2 in AMPAR gating were not recapitulated in receptors where CNIH-2 was replaced by TARP γ-2 (p > 0.7, Wilcoxon rank test for TARP γ-2 versus TARP γ-2+GSG1-l; Figures 5B, 5D, and 5F). Conversely, the CNIH-2 effects on gating were only moderately affected by coassembly of the TARP γ-8 subunit(s) (p > 0.7 and p < 0.001 Wilcoxon rank test for τdesens and τrecovery, respectively; Figure S5A). Together, these results demonstrated that coassembly of various auxiliary subunits generates AMPARs

with quite distinct functional properties. The Levetiracetam particular effects of GSG1-l may modulate the gating of AMPARs in various regions of the brain including the hippocampal CA3 region, where postembedding immunogold electron microscopy colocalized this protein with GluA2- and/or GluA4-containing AMPARs in dendritic spines of pyramidal cells (Figures 5G and S5B). Next, we used comparison of protein amounts obtained in anti-GluA APs from WT and GluA1 or GluA2 knockout mice and quantitative data from BN-PAGE separations (as in Figures 2 and 3, see Experimental Procedures) to probe whether the identified AMPAR constituents are preferentially associated with one of the two most abundant GluA subunits. Figure 6A summarizes the respective results together with the topology of the complex constituents suggested by public databases.