The NMDA receptor (NMDAR)-dependent regulation of AMPA receptor (AMPAR) membrane trafficking is a key molecular mechanism in forms of synaptic plasticity where NMDAR activation alters the membrane density of AMPARs in parallel to the changes in synaptic strength (1). The signalling that underlies the bidirectional trait of the NMDAR-dependent plasticity is directed by the kinetics of the NMDAR-generated Ca²⁺ transients and also depends on the NR2 subunit composition of the NMDAR complexes (2,3). However, the early determinants of NMDAR signals leading to AMPAR internalisation are largely unknown. Here, we investigated the requirements for NMDAR subtypes and NMDAR-mediated Ca²⁺ influx in GluR2 AMPARs endocytosis in primary murine hippocampal neurons (4). Using a live antibody feeding assay and immunofluorescence imaging we found that activation of NMDARs (20 µM NMDA + 10 µM glycine, 2 min) enhanced the internalisation of endogenous GluR2-containing AMPAR as early as 5 min after agonist washout (2.52 ± 0.38, n=17; mean ± s.e.m. of internalised fraction normalised to unstimulated controls, p<0.001, two-tailed U test). The effect was comparable for recombinant AMPARs containing N-terminally EGFP-tagged GluR2 with short but not long C-termini, whether being expressed in cultures from wild-type or GluR2 knock-out mice. This indicates that NMDAR activation selectively triggered the internalisation of AMPAR containing GluR2 splice variants with short cytoplasmic domains. The pharmacological block of NR2B-containing NMDARs with 10 µM ifenprodil completely inhibited the NMDAR-induced endocytosis of GluR2-AMPAR (1.16 ± 0.19, n=19, p=0.5 compared to unstimulated controls). In contrast, the NR2A-preferring antagonist NVP-AAM077 (1 µM) (3) did not affect NMDAR-induced GluR2 endocytosis (3.21 ± 0.5, n=20, p<0.001 compared to unstimulated controls). However, we found that at the stage when the internalisation assays were performed (14-21 days in vitro), NR2B-containing NMDARs contribute approximatively 25% of the NMDA-induced whole-cell currents. Our findings provide direct evidence that NR2B-containing NMDARs are selectively linked with the endocytosis of GluR2 AMPARs, although they make up a minor subpopulation of NMDARs. Furthermore, in neurons from mice expressing the mutant NR1(N598R) subunit (5) which renders NMDARs Ca²⁺-impermeable we found that NMDAR activation failed to increase GluR2-AMPAR endocytosis. These findings indicate the requirement of Ca²⁺ influx directly through the activated NMDAR channels for the NMDA-dependent GluR2-AMPAR endocytosis. In summary, our findings suggest that the NMDAR activation selectively induces the internalisation of short C-terminal GluR2-containing AMPARs through a mechanism that requires a NMDAR channel-carried Ca²⁺ signal processed in an NMDAR subtype-selective manner.