Roles of stargazin and phosphorylation in the control of AMPA receptor subcellular distribution

Overexpressed GluR1 accumulates in cell bodies, not in dendrites

The total levels of GluR1 in GFP-GluR1–infected neurons relative to nearby uninfected neurons were measured by immunostaining with antibodies to the C terminus of GluR1, which recognizes both endogenous and recombinant GluR1 (Fig. 1a and Online Methods). After 2.5 d of expression, the intensity of GluR1 staining in infected cell bodies was 8.1 ± 0.6-fold greater than that of nearby uninfected cell bodies (n = 99, P < 0.0001; Fig. 1b). In contrast with cell bodies, dendritic levels of GluR1 in infected cells were only moderately higher than those in neighboring uninfected neurons (1.64 ± 0.07-fold, n = 149, P < 0.0001; Fig. 1d). This large discrepancy in the expression level of recombinant protein between cell body and dendritic regions was not observed for viral expression of GFP-tagged GluR1 cytoplasmic C-terminal tail (C tail; Fig. 1b,d), suggesting that the level of full-length AMPARs at dendritic regions is under tight control. Overproduction of the related receptor GluR2 or the unrelated NMDA receptor 2B subunit (NR2B) did not decrease GluR1 levels (Supplementary Fig. 1), indicating that competition with machinery synthesizing membrane-bound proteins was not responsible for the observed somatic and dendritic levels of recombinant GluR1.

Overexpression of GluR1 was not accompanied by an increase of GluR2 in dendrites (0.95 ± 0.03, n = 120, P = 0.01), which is consistent with the finding that extra dendritic GluR1 subunits predominantly represent a pool of homomeric receptors²³. To estimate the effect of the overproduction of recombinant GFP-GluR1 on endogenous GluR1 levels, we infected CA1 pyramidal neurons with a mutant GluR1 construct that lacks seven C-terminal amino acids (GFP-GluR1Δ7) and is therefore not recognized by our antibody to GluR1 (data not shown). Infection with GFP-GluR1Δ7 led to a 42 ± 3% (n = 100, P < 0.0001) reduction of endogenous GluR1 levels in CA1 dendritic regions (Fig. 1d). Because the total dendritic levels of GluR1 were on average increased by 64% by GluR1 infection, the ratio of recombinant to endogenous GluR1 in dendrites was approximately 1.8 to 1 after 2–3 d of infection.

Phosphorylation drives GluR1 accumulation in dendrites

Newly synthesized homomeric GluR1s were transported to, but did not accumulate in, dendritic areas. What processes could control the amount of AMPAR in the dendrites? Phosphorylation events on the C tail of GluR1 have been shown to influence their synaptic trafficking; here, we examined whether they also affect the total dendritic levels of GluR1. GluR1 can be phosphorylated on serine 818 (S818) by protein kinase C (PKC)²⁴, on S831 by CaMKII or PKC, and on S845 by protein kinase A^25–27. All of these activity-dependent phosphorylation events are most strongly mimicked by substituting residues S816, S818, S831 and S845 with aspartates (GluR1_4D)²⁴. Viral expression of GluR1_4D resulted in a significant increase in dendritic GluR1 levels compared with wild-type GluR1 (2.96 ± 0.12 (n = 67) versus 1.64 ± 0.07 (n = 149), respectively, P < 0.0001; Fig. 2a). These data suggest that phosphorylation events at the GluR1 C tail control dendritic GluR1 levels. By examining individual phosphorylation sites, we found that mimicking phosphorylation at S831 or S845 was sufficient to increase the dendritic levels of GluR1 (Fig. 2a). Although GluR1 phosphorylation led to increases in dendritic levels, this was not accompanied by a decreased GluR1 accumulation in cell bodies (Fig. 2b), suggesting that GluR1 phosphorylation events do not control the relocation of AMPARs from cell body regions to dendrites.

We examined whether the mechanisms controlling dendritic GluR1 levels could involve regulated targeting to lysosomes, which has been shown to affect AMPAR levels^10,12. When we blocked lysosomal degradation in GFP-GluR1–infected hippocampal slices (chloroquin and leupeptin) for 4 h, the amount of GluR1 in the dendrites was increased (2.76 ± 0.14, n = 79) and approached that of dendrites infected with GFP-GluR1_4D (3.43 ± 0.22, n = 52) (Fig. 2c). These data suggest that overexpressed GluR1-containing receptors can reach dendritic areas, but are they are targeted to lysosomes unless they are phosphorylated at their C tail.

To examine the effects of increased dendritic levels of GluR1 on receptor surface expression, we carried out immunohistochemistry on nonpermeabilized hippocampal slices using an antibody to the extracellular region of GluR1. Although recombinant GFP-GluR1s did reach the surface of CA1 neurons²⁸, surface GluR1 levels were only slightly increased on overexpression of GluR1 or GluR1_4D (Fig. 2d), indicating that overexpressed GluR1 mainly accumulates intracellularly.

STG interaction drives GluR1 accumulation in dendrites

Although GluR1 phosphorylation affected the levels of GluR1 in dendrites, a large proportion of overexpressed GluR1 appeared to be retained in cell body regions. We examined individual GFP-GluR1–expressing neurons and found that the amounts of GluR1 in dendrites minimally correlated to somatic levels of GluR1 (Supplementary Fig. 1), suggesting that there is a limiting factor controlling the amount of GluR1 in dendrites. Previous studies have indicated that hippocampal neurons that express reduced numbers of TARPs have a loss of AMPARs mainly in dendrites, with most AMPARs residing in neuronal cell bodies²⁹, suggesting that TARPs help maintain dendritic AMPAR levels. To test whether increased amounts of TARPs would enhance GluR1 levels in dendritic compartments, we generated Sindbis viruses that expressed STG with either GFP or GFP-GluR1. Cells overexpressing both GFP-GluR1 and STG had a significant decrease in somatic levels of GluR1 (4.00 ± 0.30 (n = 69) with STG versus 8.08 ± 0.54 (n = 99) without, P < 0.0001; Fig. 3a), whereas dendritic GluR1 levels increased compared with cells overexpressing GluR1 alone (2.71 ± 0.07 (n = 168) versus 1.64 ± 0.07 (n = 149), respectively, P < 0.0001; Fig. 3b). Overexpression of STG with GFP did not alter the amounts of endogenous GluR1 present in dendrites (Fig. 3b), suggesting that the expression levels of endogenous GluR1 subunits and TARPs are balanced under physiological conditions.

STG can be phosphorylated at nine different serine residues in its cytoplasmic tail by PKC and/or CaMKII, and the majority of endogenous TARPs are phosphorylated under basal conditions³⁰. Because overexpressed recombinant STG remains mainly nonphosphorylated³⁰, we evaluated whether STG phosphorylation affects dendritic GluR1 levels by expressing a phospho-mimetic STG (STG_9D). Coexpression of GluR1 with STG_9D produced similar results as expressing GluR1 with wild-type STG (Fig. 3b). Because coexpression of GluR1_4D and STG_9D did not give an additive effect to the total dendritic GluR1 levels compared with expression of GluR1_4D or coexpression of GluR1 and STG_9D (Fig. 3b), we reasoned that both the GluR1-STG interaction and GluR1 phosphorylation use overlapping mechanisms to increase dendritic GluR1 levels. Indeed, binding STG also protected GluR1 from lysosomal targeting (Fig. 3c).

STG interaction does not drive GluR1 surface expression

To examine whether interaction with STG stabilized dendritic GluR1 by translocating AMPARs to the cell surface, we measured the amount of surface GluR1 in STG-infected CA1 neurons relative to uninfected neighbors. We were surprised to find that expression of STG did not elevate endogenous GluR1 levels on the surface of dendrites (1.05 ± 0.04, n = 46, P = 0.65; Fig. 3d) or cell bodies (Supplementary Fig. 1), as antibody staining and biotinylation assays have previously shown STG to increase recombinant AMPAR levels on the surface of heterologous cells^7,31–33. Furthermore, overexpression of both STG and GFP-GluR1 led to a modest increase in GluR1 subunits on the cell surface, which was indistinguishable to expression of GFP-GluR1 alone (1.20 ± 0.04 (n = 65) versus 1.24 ± 0.05 (n = 55), respectively, P = 0.81; Fig. 3d). Additional phosphorylation events (that is, coexpression of GluR1_4D and STG_9D) also had no effect on GluR1 surface expression (Fig. 3d).

To further examine surface AMPAR levels, we obtained paired whole-cell recordings from neighboring infected and uninfected neurons and measured the currents induced by bath application of AMPA. As has been previously shown, AMPA-induced currents were substantially increased in neurons overexpressing STG³⁴ (Fig. 3e). To determine the proportion of these increased currents that results from STG-mediated inhibition of AMPAR channel desensitization^32,33,35, we repeated these experiments in the presence of drugs that block AMPAR channel desensitization for both the flip (cyclothiazide) and flop (4-[2-(phenylsulfonylamino)ethylthio]-2,6-difluoro-phenoxyacetamide, PEPA) AMPAR isoforms (the flop isoform is dominant in CA1 neurons³⁶). In the presence of these drugs, AMPA-evoked currents in infected cells were very similar to those in neighboring uninfected cells (Fig. 3f), confirming that STG primarily enhances bath-applied AMPA responses in hippocampal neurons by reducing AMPAR desensitization³⁵. Furthermore, AMPA-evoked responses were also left unchanged when we overexpressed both GFP-GluR1 and STG (Fig. 3f), suggesting that even with an excess of GluR1, the STG interaction is not sufficient to drive GluR1 to the surface. These data indicate that enlarging the dendritic intracellular GluR1 pool does not lead to an enhancement of surface GluR1.

Increasing GluR1 levels does not produce synaptic changes

We next examined whether increasing total dendritic GluR1 levels affected synaptic AMPAR content. Increased synthesis of the recombinant GluR1 subunit led to the formation of homomeric GluR1 receptors, which showed rectification: little outward current at positive membrane potentials compared with endogenous heteromeric AMPARs. Synaptic incorporation of recombinant homomeric GluR1 receptors can therefore be detected as an increase in rectification of AMPAR-mediated synaptic responses^21,22. It has previously been shown that neurons that overexpress GFP-GluR1 show neither an increase in rectification values, nor differences in synaptic strength compared with uninfected cells²¹. We found that increasing the levels of dendritic GluR1 through coexpression of STG also failed to produce either synaptic incorporation of GluR1 (Fig. 4a,b) or changes in synaptic strength (Fig. 4c). Adding phospho-mimetic mutations on STG (STG_9D) or on both STG and GluR1 (GFP-GluR1_4D + STG_9D) did not induce synaptic insertion of GluR1 homomeric receptors (Fig. 4b) or alterations in synaptic strength (Fig. 4c).

Together, our data indicate that increasing the total numbers of GluR1 receptors in the dendrite under basal conditions is not sufficient to modify synaptic transmission. No correlation could be found between total dendritic GluR1 levels and either synaptic incorporation of GluR1 or synaptic strength (Fig. 4d). Even when comparing the two extremes, neurons that express either no GluR1 with neurons that overexpress GluR1 (that is, overexpression of GFP-GluR1 for 2 d in slices from GluR1-deficient (Glur1^−/−, also known as Gria1) mice), synaptic strength remained the same (Fig. 4e). These experiments demonstrate that increasing the availability of GluR1-containing AMPARs does not affect synaptic currents.

The GluR2 C tail allows AMPAR surface expression

GluR1 and GluR2 subunits have different roles in mediating synaptic plasticity as a result of their structurally distinct C tails. Receptors containing long GluR1 C tails are only incorporated into synapses with plasticity-inducing activity, whereas AMPARs containing only short GluR2 C tails constitutively replace AMPARs in the synapse independently of synaptic activity²². We next investigated whether these different functions of AMPAR subunits are reflected in the differences in the regulation of their subcellular distribution. We measured the effect of GluR2 overexpression on the total and surface levels of GluR2 by immunostaining. As with GluR1, CA1 pyramidal cells that were virally infected with GFP-GluR2 retained elevated levels of GluR2 in their cell bodies (3.57 ± 0.50, n = 45, P < 0.0001; Fig. 5a). Although recombinant receptors could be detected in dendrites (by their GFP expression), this led to only a small increase in total dendritic GluR2 content (1.16 ± 0.03, n = 76, P < 0.001; Fig. 5b,c) and no change in GluR2 surface levels (Fig. 5d).

Dendritic GluR2 accumulation may be restricted by the fact that GluR2 is edited at a single residue at the channel-lining pore, resulting in a positive charge in a transmembrane region (GluR2_R). Overexpression of GFP-GluR2_R mainly results in the production of homomeric GluR2 receptors that are largely retained in the endoplasmic reticulum in neurons³⁷. To study the effect of RNA editing on the subcellular distribution of homomeric GluR2, we used unedited GluR2 (GluR2_Q). Overexpression of GFP-GluR2_Q did not result in a significant increase in total dendritic GluR2 levels compared with overexpression of GFP-GluR2_R (1.22 ± 0.03 (n = 89) versus 1.16 ± 0.03 (n = 76), respectively, P = 0.16; Fig. 5c), but did lead to a nearly 60% increase in GluR2 surface expression (Fig. 5d). This increase in AMPAR surface expression, which was not observed in cells overexpressing GluR1, was regulated through the GluR2 C tail, as we did not observe an increase in surface expression when we expressed a chimeric GluR2 in which a large part of the C tail was swapped for that of GluR1 (GluR2_QR1t) (Fig. 5d). Stable AMPAR surface expression has been shown to depend on NSF binding to the GluR2 C tail³⁸, which we confirmed by expression of an N-ethylmaleimide–sensitive fusion protein (NSF) interaction mutant, GFP-GluR2_Q(N839A-P840A) (Fig. 5d).

STG does not protect GluR2 from degradation

Previous studies have shown that AMPARs bind STG and are delivered to the surface of heterologous cells by STG independent of subunit composition^31,39. However, although STG coexpression led to GluR1 accumulation in dendrites and reduced the GluR2 accumulation in cell bodies (Supplementary Fig. 2 online), dendritic levels of GluR2 were not significantly increased when STG was coexpressed with either edited GluR2_R (1.09 ± 0.04 with (n = 64) versus 1.16 ± 0.03 without (n = 76), P = 0.10) or unedited GluR2_Q (1.31 ± 0.04 with STG (n = 124) versus 1.22 ± 0.03 without (n = 89), P = 0.28) (Fig. 5e). These data indicate that STG controls the somato-dendritic distribution of AMPARs in a subunit-specific manner. To study the mechanism behind the differential distribution of AMPAR subunits, we used GluR2R1t. Coexpressing STG with GFP-GluR2_RR1t (in which the edited GluR2_R protein contains the GluR1 C tail) did not further increase the amount of GluR2 present in the dendrites compared with expressing GluR2_R (1.16 ± 0.04 (n = 76) versus 1.16 ± 0.03 (n = 76), respectively, P = 0.7; Supplementary Fig. 2), indicating that adding a GluR1 C tail and/or coexpressing STG is not sufficient to cause dendritic accumulation of homomeric GluR2_R receptors. However, adding the GluR1 C tail to unedited GluR2 (GFP-GluR2_QR1t) by itself partially increased dendritic GluR2 levels (Fig. 5c), and coexpression of unedited GFP-GluR2_QR1t with STG increased dendritic GluR2 levels twofold compared with GFP-GluR2_Q and STG (1.59 ± 0.03 (n = 153) versus 1.31 ± 0.04 (n = 124), respectively, P < 0.0001; Fig. 5e). Expression of GluR2_Q lacking a C tail (GFP-GluR2_QΔt) in conjunction with STG resulted in dendritic GluR2 levels that were similar to those observed when we expressed GFP-GluR2_QR1t and STG (1.56 ± 0.05 (n = 68) versus 1.59 ± 0.03 (n = 153), respectively, P = 0.54), indicating that it is not the GluR1 C tail that allows dendritic accumulation, but rather the GluR2 C tail that prevents dendritic accumulation.

To test whether dendritic GluR2 accumulation is prevented through lysosomal degradation, we subjected hippocampal slices expressing either GFP-GluR2_Q or GFP-GluR2_Q and STG to a 4-h lysosomal degradation blockade. In this protocol, coexpression of GFP-GluR2_Q and STG led to an accumulation of GluR2 in dendrites that was significantly higher than when GFP-GluR2_Q was expressed alone (1.54 ± 0.06 (n = 68) versus 1.18 ± 0.04 (n = 64), respectively, P < 0.0001; Fig. 5f). These experiments suggest that STG interaction promotes GluR2-containing receptors to reach dendrite regions, but does not mask a putative lysosomal targeting signal located in the GluR2 C tail. This lysosomal targeting motif likely resides in the NSF-binding site within the GluR2 C tail, since expression of the NSF interaction mutant GFP-GluR2_Q(N839A-P840A) plus STG led to a significant increase in dendritic GluR2 accumulation compared with GFP-GluR2_Q plus STG (1.54 ± 0.05 (n = 37) versus 1.31 ± 0.04 (n = 124), respectively, P < 0.0001; Fig. 5e). We conclude that STG works in an AMPAR-subunit specific manner by selectively preventing GluR1-containing AMPARs from lysosomal degradation.

Synaptic GluR2 insertion correlates with surface levels

We next examined whether phosphorylation events control the subcellular distribution of GluR2-containing AMPARs. Although mimicking STG phosphorylation (GluR2_Q and STG_9D, 1.32 ± 0.06, n = 84) or mimicking STG phosphorylation and PKC phosphorylation on the GluR2 C tail by changing serine880 to glutamate⁴⁰ (GluR2_QE and STG_9D: 1.26 ± 0.05, n = 60) had no effect on total dendritic GluR2 levels (Fig. 6a), these phosphorylation events did control surface and synaptic levels. Specifically, although wild-type STG (which remains largely nonphosphorylated on overexpression³⁰) prevented GluR2 from appearing on the surface or in synapses, coexpressing STG_9D rescued surface and synaptic insertion (Fig. 6b,c). Although STG_9D expression has been shown to potentiate synaptic strength³⁰, we found that expressing STG_9D (Supplementary Fig. 3 online) or coexpressing STG_9D and GFP-GluR1 (Fig. 4b) produced no potentiation, and coexpressing GFP-GluR2_Q and STG_9D only modestly increased AMPAR excitatory postsynaptic currents (EPSCs) (132 ± 13%, n = 27, P = 0.05; Fig. 6d). Consistent with previous data indicating that GluR2 phosphorylation by PKC leads to GluR2 endocytosis^40–42, expression of GFP-GluR2_QE with STG_9D failed to increase synaptic and surface levels of GluR2 (Fig. 6b,c).

In general, we observed that increases in total dendritic GluR2 levels failed to correlate with synaptic levels or synaptic strength (Fig. 6e). In contrast, the increase in surface GluR2 levels correlated strongly with their synaptic incorporation and surface levels of GluR2 tended to correlate with synaptic strength (Fig. 6f). Therefore, a critical determinant of changes in synaptic GluR2 content is not an increase in the production or degradation of intracellular GluR2, but is instead an alteration of its insertion into or removal from the cell surface.

Subunit-specific lysosomal targeting of endogenous AMPARs

We analyzed the effects of increased production of an AMPAR subunit on AMPAR levels in cell bodies, dendrites, synapses and on the cell surface. We found that the trajectory of newly synthesized GluR1 was different from that of GluR2, dependent on differential regulation by STG and phosphorylation events (Supplementary Fig. 4 online). However, overexpression experiments mainly produce homomeric AMPARs, whereas the endogenous AMPAR pool mainly consists of heteromeric GluR1/GluR2 and GluR2/GluR3 (ref. 8). To assess whether endogenous AMPARs abide by the rules that we determined for recombinant receptors, we measured the relative amounts endogenous GluR1, GluR2 and GluR3 with and without lysosomal degradation blockade. To determine the relative amounts of endogenous AMPAR subunits, we expressed a GluR1/2 (GFP-GluR2R1t) or a GluR1/3 (GFP-GluR3R1t) chimeric protein that contains the N-terminal region of GluR2 and GluR3, respectively, with the C tail of GluR1. Immunohistochemistry with antibodies to the GluR1 C tail and GluR2 N terminus allowed us to compare the relative levels of endogenous GluR1 and endogenous GluR2 present in dendrites with those of recombinant GFP-GluR2R1t and to derive the ratio of endogenous GluR1 to GluR2 that was present in uninfected CA1 neurons. We used the same method to derive the ratio of endogenous GluR1 to GluR3 (see Online Methods).

The ratio of endogenous GluR1 to GluR2 to GluR3 in CA1 neurons of hippocampal slices under basal conditions was approximately 1 to 3 to 1 in cell bodies (Fig. 7a) and 1 to 2 to 1 in dendrites (Fig. 7b). These data predict that dendrites of uninfected CA1 neurons contain equivalent amounts of GluR1/GluR2 and GluR2/GluR3 heteromers, which is consistent with previous findings⁸. After blocking lysosomal degradation for 4 h, the ratio between GluR1, GluR2 and GluR3 in dendrites changed to approximately 1 to 7 to 6 (Fig. 7d), which suggests that dendrites contain sixfold more GluR2/GluR3 than GluR1/GluR2 receptors. This indicates that endogenous GluR1-containing AMPARs (GluR1/2 heteromers) are less subject to lysosomal degradation under basal conditions than those lacking GluR1 (GluR2/3 heteromers). These observations are consistent with a subunit-selective regulation of AMPAR targeting to lysosomal compartments.

Slice culture, Sindbis infection and immunofluorescence

GluR1, GluR1Δ7 (GluR1(1–882)), GluR1_4D (S816D-S818D-S831D-S845D), GluR1(S816D-S818D), GluR1(S831D), GluR1(S845D), GluR1 C tail (GluR1(809–889)), GluR2_R, GluR2_Q (R586Q), GluR2R1t (GluR2(1–826)-GluR1(823–889)), GluR2Δt (GluR2(1–826)), GluR2_QE (R586Q-S880E), GluR2_Q(N839A-P840A), GluR3R1t (GluR3(1–821)-GluR1(823–889)) and NR2B constructs, all tagged with enhanced GFP (Clontech) were created as described previously^22–24. All GluR1 and GluR2 constructs were in the flop isoform. These constructs were expressed in CA1 neurons in rat hippocampal slices using Sindbis virus²³. The double promoter Sindbis (SinEGdsp) constructs were cloned from an original vector kindly provided by H. Nawa (Niigata University). The eGFP-tagged AMPAR subunits were expressed under the control of the first promoter and untagged human STG or STG_9D (generated as described in ref. 30) were under the control of the second subgenomic promoter. Hippocampal slices were stained in parallel with a monoclonal antibody to STG (Upstate). Approximately 80% of cells that overexpressed STG also expressed GFP. Organotypic hippocampal slices were prepared from postnatal 6- or 7-d-old rats and infected with Sindbis virus after 6–8 d in culture. In experiments in which lysosomal degradation was blocked, we added 200 μM chloroquin and 100 μg ml⁻¹ leupeptin to the culture medium during the last 4 h of culture. Slices were incubated in fixative (4% paraformaldehyde/4% sucrose (wt/vol), pH 7.4, in phosphate-buffered saline (PBS)) 2.5 days after viral infection for 60 min at 4°C and washed three times with PBS for 20 min. If the total AMPAR levels were measured, the fixed slices were permeabilized in 0.3% Triton X-100 (vol/vol) in PBS for 45 min at 4°C and washed three times in PBS for 20 min. Slices were incubated in blocking solution (10% donor horse serum (vol/vol) in PBS) for 1 h at 23–25°C, and subsequently exposed to polyclonal rabbit antibody to GluR1 (Chemicon) at 1 μg ml⁻¹ for total GluR1 staining, polyclonal rabbit antibody to GluR1 (Calbiochem) at 10 μg ml⁻¹ for surface GluR1 staining, monoclonal mouse antibody to GluR2 (Chemicon) at 10 μg ml⁻¹ or monoclonal mouse antibody to GluR3 (Chemicon) at 10 μg ml⁻¹ in blocking solution overnight at 4°C, followed by three 20-min washes with PBS at 23–25°C. The slices were then incubated in Texas Red–conjugated goat antibody to rabbit or goat antibody to mouse IgG (Molecular Probes) at 6 μg ml⁻¹ in PBS overnight at 4°C. The slices were washed three times with PBS for 20 min and stored in PBS at 4°C.

Two-photon laser scanning microscopy and image analysis

Three-dimensional images were collected on a custom-built two-photon microscope based on a Fluoview laser-scanning microscope (Olympus America). The light source was a mode-locked Ti:sapphire laser (Chameleon, Coherent) tuned to 910 nm. We used a 60× objective with a 0.9 numerical aperture. Optical sections were captured every 0.5 μm (Fluoview software, Olympus). Images were taken from both infected and nearby uninfected CA1 cell bodies (3× digital zoom) and dendrites near the point of bifurcation of primary to secondary dendrites, ~300 mm from cell bodies (5× digital zoom). Projections of three stacked sections were analyzed using custom-written software in MatLab (Mathworks). To measure AMPAR subunit densities in dendrites or cell bodies independent of their width, we obtained maximum peak intensities by drawing a profile (consisting often lines spread over 2 μm) that included uninfected and infected cell bodies or dendrites. Because of limited antibody penetration, fluorescence intensity of antibody staining varied at different imaging depths. Therefore, peak values for Texas red fluorescence were normalized. Each peak value was divided by the average of the peak fluorescence values of the uninfected (GFP negative) cell bodies or dendrites in the same projection.

Electrophysiology

Sindbis virus was injected into 5–7 d in vitro slice cultures and cells were maintained for 48–60 h. A cut was made between CA3 and CA1 to prevent stimulus-induced bursting. Simultaneous whole-cell recordings were obtained from pairs of neighboring control and infected CA1 pyramidal neurons under visual guidance using differential interference contrast and fluorescence microscopy. Two stimulating electrodes, two-contact Pt/Ir cluster electrode (Frederick Haer), were placed between 100 and 300 μm down the apical dendrite, 100 μm apart, and 200 mm laterally in opposite directions. Whole-cell recordings were obtained with Axopatch-1D amplifiers (Molecular Devices) using 3–5-MΩ glass pipettes with an internal solution containing 115 mM cesium methanesulfonate, 20 mM CsCl, 10 mM HEPES, 2.5 mM MgCl₂, 4 mM Na₂ATP, 0.4 mM Na₃GTP, 10 mM sodium phosphocreatine, 0.6 mM EGTA and 0.1 mM spermine (Sigma), at pH 7.25. External perfusion consisted of artificial cerebrospinal fluid at 27°C (see above) with 4 mM MgCl₂, 4 mM CaCl₂, 4 μM 2-chloroadenosine, 100 μM picrotoxin and 100 μM D(−)-2-amino-5-phosphonovaleric acid (Sigma). All recordings were carried out by stimulating two independent synaptic inputs 1.5 s apart; results from two pathways were averaged and counted as n = 1. The AMPA-mediated EPSC was measured as the peak inward current at −60 mV. Rectification was calculated as the ratio of the peak AMPA current at −60 and +40 mV, corrected by the current at 0 mV. EPSC amplitudes were obtained from an average of 80–120 sweeps at each holding potential. Data were acquired and analyzed using custom software written in Igor Pro (Wavemetrics). AMPA-mediated whole-cell currents were obtained from neighboring infected and uninfected cell simultaneously by bath perfusion of 1 μM AMPA. External artificial cerebrospinal fluid perfusion was carried out as described above with 1 μM tetrodotoxin (Calbiochem) and did or did not include 100 μM cyclothiazide and 80 μM PEPA (Sigma).

Measuring relative endogenous levels of GluR1, GluR2 and GluR3

Hippocampal slices were infected with SindBis virus expressing the chimeric GFP-GluR2R1t for 2.5 d. Slices were fixed and permeabilized, and we compared immunostaining intensities between infected and neighboring uninfected neurons. To calculate the endogenous GluR1 and GluR2 levels, we used the following equations.

where R1^u is the amount of endogenous GluR1 in the uninfected cell, R1ⁱ is the amount of endogenous GluR1 in the infected cell, X is the total amount of GluR1 in the infected cell relative to the uninfected cell, R₁₂ is the amount of recombinant GFP-GluR2R1t, R2 is the amount of endogenous GluR2 in the uninfected cell, R2ⁱ is the amount of endogenous GluR2 in the infected cell and Y is the total amount of GluR2 in the infected cell relative to the uninfected cell. Equation (3) is derived by combining equations (1) and (2).

To measure X, we use an antibody to GluR1 C tail that recognized GluR1 and GFP-GluR2R1t and compared the amount of GluR1 present in infected to uninfected cells. To measure Y, we used an antibody to N-terminal GluR2 that recognized GluR2 and GFP-GluR2R1t and compared the amount of GluR2 present in infected to uninfected cells. We measured $\frac{R{\mathit{1}}^i}{R\mathit{1}}$ with an antibody to the GluR1 N terminus that recognized GluR1 and not GFP-GluR2R1t; the amount of staining in cells infected with GFP-GluR2R1t, relative to staining in uninfected cells, is a measure of $\frac{R{\mathit{1}}^i}{R\mathit{1}}$. Because $\frac{R{\mathit{1}}^i}{R\mathit{1}}=96\%$ and $\frac{R{\mathit{3}}^i}{R\mathit{3}}$ (R₃ is the amount of endogenous GluR3 in the uninfected cell, R3ⁱ is the amount of endogenous GluR3 in the infected cell) = 96%, assuming that most AMPARs are heteromers of either GluR1/2 or GluR2/3, $\frac{R{\mathit{2}}^i}{R\mathit{2}}$ was estimated to be 96%. Similar experiments (with expression of GFP-GluR3R1t) and equations revealed $\frac{R\mathit{1}}{R\mathit{3}}$. For both cell bodies and dendrites, $\frac{R\mathit{1}}{R\mathit{2}}$ and $\frac{R\mathit{1}}{R\mathit{3}}$ were calculated with and without 4-h lysosomal degradation block.

Statistical analysis

Statistical comparisons (P) were performed using two-tailed Student’s t test of log-transformed data, nonpaired for antibody staining and rectification data, and paired for EPSC measurements.