We have developed a reactive tag system using newly developed short amino acid sequences for the EM tag combined with the synthetic chemical probe, which makes the covalent link with the tag (Tabata et al., 2019). When we expressed a tagged receptor in HEK cells, a 1.4nm gold-conjugated chemical probe specifically labeled surface receptors with much higher sensitivity than the FLAG tag located in an adjacent sequence. The resolution of this labeling is a few nanometers, which is more than 4 times higher than conventional immunolabeling using primary and secondary antibodies (20 nm). We have generated knock-in mice with EM tag sequences inserted in AMPA-type and NMDA-type glutamate receptors to enable us to examine the subunit composition of single ionotropic receptors in situ. To visualize multiple subunits in single channels, we have developed a second chemical labeling method using a different tag-probe pair (Zenmyo et al., 2018). However, because it showed cross-reactivity with the first one, we tested the third tag-probe pair, which makes a covalent linkage between the His tag and a newly synthesized modified NTA probe. For asking physiological roles of glutamate receptors in hippocampal pathways involved in emotional memory formation, we identified a new glutamatergic ventro-dorsal hippocampal projection mediated by hilar mossy cells (Fredes et al, 2020). We found that activation of ventral mossy cells is necessary and sufficient to activate dorsal granule cells, which otherwise are rarely activated by other glutamatergic pathways. This indicates unexpected ventro-dorsal interaction of the hippocampus critical for linking emotional inputs to context memory formation in the dorsal hippocampus. Using the newly generated knock-in mice, we further examine which subunit composition of glutamate receptors is involved in this emotion-facilitated context memory formation. In naive mice, we found that GluA1 and GluA2 subunits of the AMPA receptors have relatively homogeneous distribution patterns in the hippocampal pyramidal cell synapses and high extrasynaptic density, whereas GluA3 has a different pattern showing less density in the extrasynaptic sites and concentration in the center compared with the periphery of the postsynaptic density (Jevtic et al, under preparation). However, contextual fear learning has induced a shift of GluA1 to a more centered position with a significant increase of density specific to the CA1 area close to the CA3 area. This implies a compartment-specific change in the subunit composition of AMPA receptors by physiological learning for the first time. We are examining if this is the case using the newly developed chemical probe labeling for AMPA receptors (Islam et al., in preparation).