Synaptic AMPA receptor subunit trafficking is independent of the C terminus in the GluR2-lacking mouse

GluR2^−/− KO Characterization.

To test the validity of GluR2^−/− as a system for determining the rules for subunit trafficking, we first fully characterized the synaptic responses in this mouse. AMPA receptor excitatory postsynaptic currents (AMPA EPSCs), evoked by stimulating Schaffer collateral axons, were recorded with patch pipettes and isolated pharmacologically (see Methods). AMPA EPSCs from WT hippocampal CA1 pyramidal cells in acute slices, have linear I/Vs (Fig. 1 A and B). In contrast, AMPA EPSCs from the GluR2^−/− mouse are inwardly rectifying at positive potentials because of pore block by internal polyamines (Fig. 1 A and B) (12, 15).

Because the current/voltage relationship is linear in the WT mouse, this suggests that most, if not all, AMPA receptors contain the GluR2 subunit in the WT background (16, 17). We therefore checked to see whether there was a reduction in the overall AMPA signaling in GluR2^−/−. To accomplish this, we compared the size of the AMPA receptor-mediated EPSC to the NMDA receptor-mediated EPSC at the synapse. In the presence of the GABA_A receptor blocker picrotoxin, AMPA and NMDA EPSCs were measured at −70 and +40 mV, respectively. To ensure that the NMDA-mediated EPSC at +40 mV was free of AMPA receptor contamination, the current was measured 100 ms after stimulation. In WT cells, the AMPA/NMDA ratio was 2.37 ± 0.13, compared with 1.12 ± 0.08 in the KO (Fig. 1C). These values are in agreement with those reported for GluR2^−/− (12).

The above finding suggests that the loss of the GluR2 subunit results in either a decrease in the relative number of AMPA receptors to NMDA receptors at these synapses or in a greater number of synapses containing only NMDA receptors (silent synapses) in the KO. To determine whether AMPA receptor number at the synapse is reduced in GluR2^−/−, we recorded miniature EPSCs (mEPSCs). mEPSCs recorded in the presence of the sodium channel blocker tetrodotoxin were modestly, but significantly, decreased in GluR2^−/− compared with the WT mouse (mean mEPSC amplitude: GluR2^−/−, 8.6 ± 0.4 pA, n = 9; WT, 9.8 ± 0.3 pA, n = 15; P < 0.05; Fig. 1D). In addition, the frequency of mEPSCs, an indicator of functional synapses, was substantially reduced in GluR2^−/− (mean mEPSC frequency: GluR2^−/−, 0.10 ± 0.01 Hz, n = 9; WT, 0.41 ± 0.06 Hz, n = 15; P < 0.05; Fig. 1E), suggesting that, in addition to a reduction in the number of AMPA receptors at synapses, a greater proportion of synapses in GluR2^−/− are silent. Because the AMPA receptor composition is fundamentally different in GluR2^−/− compared with the WT mouse, we also examined the kinetics of the mEPSCs. Surprisingly, the decay kinetics of mEPSCs was not different in the two mice, suggesting that the GluR2 subunit is not the determining factor of mEPSC decay (Fig. 1F).

A reduction in the AMPA/NMDA ratio concomitant with a reduction in both the amplitude and the frequency of AMPA receptor-mediated mEPSCs indicate that hippocampal basal synaptic transmission is diminished in GluR2^−/−. To further characterize AMPA-dependent synaptic transmission, we performed extracellular field recordings in acute slices of GluR2^−/− and WT mice. The presynaptic fiber volley amplitude was used as a measure of input strength and plotted against the size of the AMPA field EPSP. We found a large reduction in the AMPA field potentials of GluR2^−/− compared with WT littermates (≈50%, Fig. 2A), consistent with previous studies (13, 15). We also checked the NMDA field EPSP (recorded in reduced extracellular Mg²⁺) and found that NMDA receptor transmission was also reduced in GluR2^−/− (≈25%), although to a lesser extent than AMPA receptor transmission (Fig. 2B). The loss of CA3 pyramidal cells in the GluR2 KO mouse (15) could contribute substantially to the diminished synaptic transmission onto CA1 pyramidal cells.

Finally, we compared plasticity in both animals using a pairing protocol to induce LTP in whole-cell recordings from CA1 pyramidal cells. As has been reported, the magnitude of LTP was not diminished in GluR2^−/− (12, 14, 15), suggesting that AMPA receptor trafficking mechanisms in the context of synaptic plasticity are not perturbed in the KO (Fig. 2C).

GluR2 Synaptic Insertion in GluR2^−/− Pyramidal Cells.

To assess whether introducing GluR2 constructs into GluR2^−/− neurons was a valid strategy for determining regions of the GluR2 C-tail required for constitutive trafficking, we introduced the full-length GluR2 subunit into hippocampal pyramidal cells of GluR2^−/− mice. Using biolistics as a method of gene delivery, we shot cultured organotypic hippocampal slice preparations made from GluR2^−/− with microprojectiles coated with DNA encoding GFP-tagged GluR2. Simultaneous dual whole-cell recordings were then performed on a cell transfected with GFP-GluR2 and a neighboring untransfected cell. GluR2 introduction into GluR2^−/− pyramidal cells did not change the amplitude of the NMDA EPSC (recorded at +40 mV) or the AMPA EPSC (recorded at −70 mV) compared with untransfected neighboring cells (Figs. 3 A and B). However, GluR2-containing AMPA receptors clearly were incorporated into the synapse, appearing to completely replace the GluR2-lacking receptors, because the synaptic I/Vs of the GluR2-transfected KO cells were linear (Fig. 3C). Because the single-channel conductance of GluR2-lacking receptors is different from GluR2-containing receptors, it is difficult to determine in what ratio the receptors were swapped. Regardless, the linear I/V indicates that these synapses contain only GluR2-containing receptors.

Synaptic incorporation of introduced GluR2 in GluR2^−/− cells confirmed that GluR2-trafficking mechanisms are still in place in the KO. Therefore, to test whether the GluR2 tail is required for synaptic incorporation of AMPA receptors, we deleted the entire region of the GluR2 C-tail, from amino acid V827 onwards (GluR2ΔC, Fig. 4A). Similar to full-length GluR2, GluR2ΔC expression did not result in a change in the amplitude of either AMPA or NMDA receptor transmission (Fig. 4 A and B). Surprisingly, GluR2ΔC did traffic to the synapse just as well as the full-length GluR2, because the synaptic I/Vs of cells transfected with this construct were also linear (Fig. 4C). One might argue that because all AMPA receptors in the GluR2^−/− mouse are calcium-permeable, the calcium flux through these receptors is sufficient to engage the activity-dependent mechanism responsible for inserting the introduced GluR2 C-tail-lacking receptors into the synapse. To address this possibility, we added kynurenic acid, an antagonist of both AMPA and NMDA receptors, immediately after transfection. Although this treatment blocks all excitatory transmission (18), it did not prevent the linearization of the I/V (Fig. 4D).

Finally, we made a GluR2 construct in which the C-tail of GluR2 was replaced with that of GluR1 (GluR2/R1). Specifically, this GFP-GluR2/R1 chimera contains amino acids 823–889 of GluR1 fused to the body of GluR2 (amino acids 1–826, Fig. 5A). Based on the lack of GFP fluorescence, Shi et al. (9) have reported that this construct is excluded from spines. Moreover, when the critical arginine at position 607 was mutated to glutamine in this construct [GluR2(Q)], which creates a rectifying homomeric receptor, no change in rectification was observed at the synapse, further suggesting that this construct was not delivered to the synapse (9). When we transfected GluR2/R1 into GluR2^−/− pyramidal cells, AMPA and NMDA receptor transmission appeared to be unaltered when compared with untransfected neighboring cells (Fig. 5 A and B). However, cells expressing GluR2/R1, similar to both GluR2 full-length and GluR2ΔC constructs, exhibited synaptic currents with linear I/V relationships, indicating that it trafficked and incorporated into the synapse (Fig. 5C). In two cells cultured in the presence of kynurenic acid, GluR2/R1 still linearized the I/V (RI = 0.71 ± 0.05), suggesting an activity-independent mechanism of synaptic incorporation.

Molecular Biology.

The GluR2 subunit and GluR2 subunit constructs were tagged with enhanced GFP after amino acid S24 in the N-terminal domain, after the putative signal peptide sequence of GluR2. These constructs were then cloned into the pCI-neo mammalian expression vector (Promega) between the XbaI (5′) and NotI (3′) sites. Amino acids V827–I862 of GluR2 were deleted to make GFP-GluR2ΔC by using a PCR-based mutation strategy. Similarly, an overlap PCR-based strategy was used to fuse the carboxyl tail of GluR1 (amino acids G824–L892) to the end of the fourth transmembrane domain of GluR2 (after amino acid K826) to make GFP-GluR2/R1. All constructs made by using PCR were validated by DNA sequencing (ElimBio).

Slice Culture Transfection and Electrophysiology.

Organotypic hippocampal slice cultures made from P6-P9 GluR2^−/− mice were made as described (30). Methods conform to the University of California Institutional Animal Care and Use Committee guidelines. For experiments in which 10 mM kynurenic acid was added to the slice culture media, the pH of the media was brought back to 7.3 after addition of kynurenic acid. Biolistic transfection (Bio-Rad) was used to introduce GFP-GluR2 constructs into pyramidal cells of GluR2^−/− slice cultures as described (31). Whole-cell recordings from slice cultures were performed 3–7 days after transfection. For whole-cell electrophysiological recordings, slices were perfused in artificial cerebrospinal fluid (aCSF) containing the following: 119 mM NaCl, 2.5 mM KCl, 4 mM CaCl_2, 4 mM MgSO₄, 1 mM NaH₂PO₄, 26.2 mM NaHCO₃, 11 mM d-glucose, 0.1 mM picrotoxin, 0.02 mM bicuculline methiodide, and 0.01 mM 2-chloroadenosine (to suppress epileptiform activity), and bubbled with 5% CO₂/95% O₂. aCSF osmolarity was 297–305 mOsm. To isolate AMPA receptor-mediated currents, 100 μM D-APV was added to the perfusate. Patch electrodes (3–5 MΩ) were filled with the following internal solution: 110 mM CsMeSO₄, 15 mM CsCl, 10 mM Hepes, 2.5 mM MgCl₂, 4 mM Na₂ATP, 0.4 mM Na₃GTP, 10 mM EGTA, 0.1 mM spermine tetrahydrochloride, and 5 mM QX-314 chloride (pH = 7.3), ≈290 mOsm. Simultaneous whole-cell recordings were made by first patching a transfected cell, followed by an untransfected neighboring cell (identifying cells with UV fluorescence and guided by 40× DIC optics). Cells were broken into immediately after achieving GΩ patch resistances on both cells. Synaptic stimulation was induced by placing a monopolar electrode filled with aCSF in the stratum radiatum to stimulate Schaffer collateral axons (0.2-Hz stimulation frequency) by delivering a 100-μs current injection of varying strength. Twenty to 50 synaptic responses were recorded and averaged at membrane potentials of −70, 0, and +40 mV. Stimulus artifacts have been blanked for clarity. The rectification index was then calculated by using the following formula: RI = [(I₊₄₀ − I₀)/(I₀ − I₋₇₀)] × 7/4.

Acute Slice Electrophysiology.

Transverse hippocampal slices were prepared for whole-cell (300 μm) or field potential (400 μm) recordings from GluR2^−/− and WT mice between postnatal day (P)14 and P28 as described (32). Whole-cell recordings were performed by using the same aCSF and internal solutions as described above. Current/voltage relationships for synaptic AMPA currents were done in the presence of 100 μM D-APV (Fig. 1 A and B). The AMPA component of the AMPA/NMDA ratios was taken at −70 mV. The NMDA component was determined by taking the current at +40 mV 100 ms after the stimulation artifact. mEPSCs were recorded in the presence of 500 nM tetrodotoxin. LTP experiments were also performed in the same aCSF and internal solutions described above, although the internal solution lacked 10 mM EGTA. fEPSP recordings were made by placing a stimulating electrode filled with aCSF in stratum radiatum ≈150–200 μm away from a recording electrode (≈3–5 MΩ) filled with aCSF also in the stratum radiatum. Care was taken to ensure that the latitude of stimulation and recording within the radiatum was similar between WT and GluR2^−/− slices. Input–output relationships were generated in this way for both AMPA receptor-mediated fields and NMDA receptor-mediated fields, the latter in the presence of 10 μM CNQX. Stimulation frequency for fEPSP was 0.1 Hz. Input–output relationships were determined in similar aCSF as described above, except CaCl₂ concentration was 2.5 mM, and MgSO₄ concentration was 1.3 mM (AMPA fEPSP experiments) or 0.1 mM (NMDA fEPSP experiments). All data in this study are reported as mean ± SEM. Statistics performed in this study were done by using Student's t test (except for mEPSC analysis in which the Kolmogorov–Smirnov test was used), and statistical significance was determined as P < 0.05.