Activity patterns govern synapse-specific AMPA receptor trafficking between deliverable and synaptic pools

Abstract

In single neurons, glutamatergic synapses receiving distinct afferent inputs may contain AMPA receptors (-Rs) with unique subunit compositions. However, the cellular mechanisms by which differential receptor transport achieves this synaptic diversity remain poorly understood. In lateral geniculate neurons, we show that retinogeniculate and corticogeniculate synapses have distinct AMPA-R subunit compositions. Under basal conditions at both synapses, GluR1-containing AMPA-Rs are transported from an anatomically defined reserve pool to a deliverable pool near the postsynaptic density (PSD), but further incorporate into the PSD or functional synaptic pool only at retinogeniculate synapses. Vision-dependent activity, stimulation mimicking retinal input, or activation of CaMKII or Ras signaling regulated forward GluR1 trafficking from the deliverable pool to the synaptic pool at both synapses, whereas Rap2 signals reverse GluR1 transport at retinogeniculate synapses. These findings suggest that synapse-specific AMPA-R delivery involves constitutive and activity-regulated transport steps between morphological pools, a mechanism that may extend to the site-specific delivery of other membrane protein complexes.

Fig. 1. GluR1 is predominantly expressed in retinogeniculate synapses

(A) Western blots of GluR1, GluR2L, GluR4 and GluR2/3/4c in the whole hippocampus (HP) and lateral geniculate nucleus (LGN) prepared from the same animals. Each pair of HP and LGN lanes was loaded with the same amount of protein (60–120 μg). (B) Amounts of GluR1 (n=8), GluR2L (n=9), GluR4 (n=8), and GluR2/3/4c (n=8) in LGN relative to the whole hippocampus. The relative values and standard errors were normalized to average amounts of GluR1, GluR2L, GluR4 and GluR2/3/4c from whole hippocampus. (C) GluR1 immunolabeling at synapses contacted by RL and RS terminals. (D) Percentages of GluR1 labeled retinogeniculate (RG) and corticogeniculate (CG) synapses relative to all RG or CG synapses (n=11; p<0.005). (E) GluR4 immunoperoxidase labeling at synapses contacted by RL and RS terminals. (F) Percentages of GluR4 labeled RG and CG synapses relative to all RG or CG synapses (n=15; p=0.09). (G) GluR2/3/4c immunoperoxidase labeling at synapses contacted by RL and RS terminals. Scale bar applies to C1–G3. Arrows indicate positive immunoperoxidase labeling associated with PSD postsynaptic to RL (red) and RS (blue) terminals. (H) Percentages of GluR2/3/4c labeled RG and CG synapses relative to all RG or CG synapses (n=10; p=0.06). See Supplemental Data for the values.

Fig. 2. GluR1 selectively mediates retinogeniculate transmission

(A) Upper schematic drawing shows the setting for in vivo viral delivery of recombinant proteins into LGN. Lower schematic drawing illustrates the stimulating and recording electrode locations in the in vitro LGN preparation. IC: internal capsule; LGN: dorsal lateral geniculate nucleus; OT: optic tract; ST: striatum; TRN: thalamic reticular nucleus; vLGN: ventral lateral geniculate nucleus. (B) Simultaneous recordings, made under transmitted light illumination (lower panel), from pairs of a recombinant protein expressing neuron, identified by GFP fluorescence (upper panel), and a neighboring non-expressing control neuron. Recording traces show AMPA-R-mediated EPSCs evoked by electrical stimulation of retinogeniculate (RG) and corticogeniculate (CG) afferents at −60 mV. Note the paired-pulse depression of RG responses and facilitation of CG responses of both control non-expressing and expressing neurons. (C) Upper, evoked AMPA-R-mediated responses at RG and CG synapses from non-expressing (Ctrl) and GluR1-GFP expressing neurons recorded at −60 mV and +40 mV. Lower left, AMPA responses in neurons expressing GluR1-GFP at RG (n=16; p=0.55) and CG synapses (n=20; p=0.11) relative to neighboring control neurons. Lower right, rectification of GluR1-GFP expressing neurons at RG (n=16; p<0.005) and CG synapses (n=20; p=0.88) relative to neighboring control cells. Rectification is defined as the ratio of responses at −60 mV and +40 mV. (D) Upper, evoked AMPA-R- and NMDA-R-mediated responses at RG and CG synapses from non-expressing (Ctrl) and GluR1ct-GFP expressing neurons recorded at −60 mV and +40 mV. Lower left, AMPA responses in GluR1ct-GFP expressing neurons from rats at RG (n=16; p<0.05) and CG synapses (n=15; p=0.61), from wild type (WT) mice (n=24; p<0.005), and from GluR1 knockout (KO) mice (n=24; p=0.95) at RG synapses relative to neighboring control neurons. Lower right, NMDA responses in GluR1ct-GFP expressing neurons at RG (n=16; p=0.73) and CG synapses (n=15; p=0.87), from wild type mice (n=24; p=0.75), and GluR1 knockout mice (n=24; p=0.75) at RG synapses relative to neighboring control cells. Note that GluR1 knockout mice had increased ratio of NMDA and AMPA responses compared to WT mice (n=24; p<0.05; Mann-Whitney Rank Sum test). AMPA-R and NMDA-R mediated current amplitude and standard errors were normalized to average values from control cells. Asterisks indicate statistical significance (Wilcoxon test). See Supplemental Data for the values.

Fig. 3. Time courses of evoked retinogeniculate and corticogeniculate events at synaptic sites

(A–B) Evoked EPSCs in retinogeniculate (RG) and corticogeniculate (CG) pathways recorded at the soma of a thalamocortical neuron in LGN. (C–D) Additional synaptic currents due to hyperpolarizing somatic voltage jumps made relative to EPSC onset (~−2—−3 ms to 12—14 ms, 0.4 ms interval). Scale bars apply to A–D. (E–F) Charge recovery curves obtained from integration of voltage jump-induced synaptic currents in C and D. (G) Decay time constant (τ) of evoked EPSCs in RG and CG pathways at somatic (n=11; p<0.05) and synaptic (n=11; p<0.01) sites. Asterisks indicate statistical significance (Wilcoxon test). See Supplemental Data for the values.

Fig. 4. Ras controls forward GluR1 trafficking from deliverable to synaptic pools

(A–C) GluR1 immunogold labeling at synapses in normal control LGN (A1–5), LGN expressing Ras(ca)-GFP (B1–4), and LGN expressing Ras(dn)-GFP (C1–4). Arrows point to silver-enhanced gold particles associated with PSDs postsynaptic to RL (red arrows) or RS (blue arrows) terminals. Scale bar applies to A–C. (D) Relative distributions of GluR1 at synapses contacted by RL (red) and RS (blue) terminals in normal LGN (D1: n=710 for RL; n=1,572 for RS), LGN expressing Ras(ca)-GFP (D2: n=625 for RL; n=1,544 for RS), CaMKII-IRES-GFP (D3: n=616 for RL; n=1,380 for RS), and Ras(dn)-GFP (D4; n=594 for RL; n=1,549 for RS). (E) Left, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in normal LGN (n=12), LGN expressing Ras(ca)-GFP (n=12, p<0.05), CaMKII-IRES-GFP (n=9, p<0.05), or Ras(dn)-GFP (n=10, p<0.001). Right, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in normal LGN (n=12), LGN expressing Ras(ca)-GFP (n=12, p<0.001), CaMKII-IRES-GFP (n=9, p<0.001), or Ras(dn)-GFP (n=10, p>0.05). Note (not shown) that average percentages of GluR1 silver-gold particles in the residual pool at synapses contacted by RL (n=9–12; p>0.05) and RS (n=9–12; p>0.05) terminals were the same as that in normal LGN. Asterisks indicate statistical significance relative to normal control LGN (Mann-Whitney Rank Sum test). See Supplemental Data for the values.

Fig. 5. Rap2 controls reverse GluR1 trafficking from synaptic to deliverable pools

(A–B) GluR1 immunogold labeling at synapses in LGN expressing Rap2(ca)-GFP (A1–4) and LGN expressing Rap2(dn)-GFP (B1–4). Red arrows point to silver-enhanced gold particles associated with PSDs postsynaptic to RL terminals. Scale bar applies to A-B. (C) Relative distributions of GluR1 at synapses contacted by RL (red) and RS (blue) terminals in LGN expressing Rap2(ca)-GFP (C1: n=551 for RL; n=1,335 for RS), Rap2(dn)-GFP (C2: n=462 for RL; n=1,243 for RS), Rap1(ca)-GFP (n=534 for RL; n=1,238 for RS), and Rap1(dn)-GFP (n=709 for RL; n=1,387 for RS). (D) Left, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in LGN expressing Rap2(ca)-GFP (n=10, p<0.001), Rap2(dn)-GFP (n=10, p<0.05), Rap1(ca)-GFP (n=10, p>0.05), or Rap1(dn)-GFP (n=10, p>0.05) relative to normal control LGNs presented in figure 2E. Right, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in LGN expressing Rap2(ca)-GFP (n=10, p>0.05), Rap2(dn)-GFP (n=10, p>0.05), Rap1(ca)-GFP (n=10, p>0.05), or Rap1(dn)-GFP (n=10, p>0.05) relative to normal control LGNs presented in figure 2E. Note (not shown) that average percentages of GluR1 silver-gold particles in the residual pool at synapses contacted by RL (n=10, p>0.05) and RS (n=10, p>0.05) terminals were the same as that in normal LGN. Asterisks indicate statistical significance relative to normal control LGN (Mann-Whitney Rank Sum test). See Supplemental Data for the values.

Fig. 6. Multiple kinases regulate GluR1 interpool trafficking

(A–B) GluR1 immunogold labeling at synapses contacted by RL terminals in rats with LGN infusion of KN-93 (A1–3) and SP600125 (B1–3). Scale bar applies to A–B. (C) Relative distributions of GluR1 silver-gold particles at synapses contacted by RL (red) and RS (blue) terminals in rats with LGN infusion of KN-93 (C1: n=464 for RL; n=1,024 for RS), SP600125 (C2: n=469 for RL; n=1,156 for RS), and PKI (n=551 for RL; n=1,335 for RS). Relative distributions of GluR1 silver-gold particles at synapses contacted by RL and RS terminals in rats with LGN infusion of PD98059 (n=479 for RL; n=1,068 for RS), LY294002 (n=528 for RL; n=1,074 for RS), SB203580 (n=556 for RL; n=1,196 for RS), and Go6850 (n=475 for RL; n=976 for RS) were not shown. (D) Left, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in rats with LGN infusion of 50 μM KN-93 (n=10, p<0.001), 50 μM SP600125 (n=9, p<0.001), 200 μM PD98059 (n=10, p<0.001), 100 μM LY294002 (n=10, p<0.001), 20 μM SB203580 (n=10, p>0.05), 200 μM PKI 14–22 amide (n=10, p<0.005), or 100 nM biosindolylmaleimide (Go6850, n=10, p<0.05), relative to normal control LGNs presented in figure 2E. Right, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in rats with LGN infusion of KN-93 (n=10, p>0.05), SP600125 (n=10, p>0.05), 200 μM PD98059 (n=10, p>0.05), LY294002 (n=10, p>0.05), SB203580 (n=10, p>0.05), PKI (n=10, p>0.05), or Go6850 (n=10, p>0.05) were unchanged compared to normal control LGN. Note (not shown) that average percentages of GluR1 silver-gold particles in the residual pool at synapses contacted by RL (n=9–12; p>0.05) and RS (n=9–12; p>0.05) terminals were the same as that in normal LGN. (E) Upper, evoked AMPA-R-mediated responses at retinogeniculate (RG) synapses from non-expressing (Ctrl) and GluR1-GFP expressing neurons in rats with LGN infusion of 50 μM KN-93 and 50 μM SP600125 recorded at −60 mV and +40 mV. Lower left, RG AMPA responses in neurons expressing GluR1-GFP in rats with LGN infusion of KN-93 (n=12; p=0.58), or SP600125 (n=14; p=0.64) relative to neighboring control neurons. Lower right, rectification of RG AMPA responses in GluR1-GFP expressing neurons in rats with LGN infusion of KN-93 (n=12; p=0.39), or SP600125 (n=14; p<0.005) relative to neighboring control cells. Note that GluR1ct-GFP expressing neurons in rats with LGN infusion of SP600125 had more enhanced rectification compared to GluR1ct-GFP expressing neurons in control rats (Ctrl: n=16; SP: n=16; p<0.005; Mann-Whitney Rank Sum test; cf. Fig. 2C). (F) Upper, evoked AMPA-R- and NMDA-R-mediated responses at RG synapses from non expressing (Ctrl) and GluR1ct-GFP expressing neurons in rats with LGN infusion of 50 μM KN-93 and 50 μM SP600125 recorded at −60 mV and +40 mV. Lower left, RG AMPA responses in neurons expressing GluR1ct-GFP in rats with LGN infusion of KN-93 (n=17; p=0.69), or SP600125 (n=14; p<0.005) relative to neighboring control neurons. Note that GluR1ct-GFP expressing neurons in rats with LGN infusion of SP600125 had more significantly reduced AMPA responses compared to GluR1ct-GFP expressing neurons in control rats (Ctrl: n=16; SP: n=14; p<0.05; Mann-Whitney Rank Sum test; cf. Fig. 2D). Lower right, RG NMDA responses in neurons expressing GluR1ct-GFP in rats with LGN infusion of KN-93 (n=17; p=0.38), or SP600125 (n=14; p=0.64) relative to neighboring control cells. AMPA-R and NMDA-R mediated current amplitude and standard errors were normalized to average values from control cells. Asterisks indicate statistical significance (Mann-Whitney Rank Sum or Wilcoxon test). See Supplemental Data for the values.

Fig. 7. Vision-dependent activity drives GluR1 insertion at retinogeniculate synapses

(A–B) GluR1 immunogold labeling at synapses contacted by RL terminals in rats with eyelids stitched (ES, A1–3), and rats with LGN infusion of TTX (B1–3). Scale bar applies to A–B. (C) Relative distributions of GluR1 silver-gold particles at synapses contacted by RL (red) and RS (blue) terminals in rats with eyelids stitched (C1: n=524 for RL; n=1,334 for RS) and with LGN infusion of TTX (C2: n=620 for RL; n=1,594 for RS). (D) Left, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in rats with eyelids stitched (n=10, p<0.05), or with LGN infusion of TTX (n=10, p<0.001) relative to normal control LGNs presented in figure 2E. Right, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in rats with eyelids stitched (n=10, p>0.05), or with LGN infusion of TTX (n=10, p>0.05) were unchanged compared to normal control LGN. Note (not shown) that average percentages of GluR1 silver-gold particles in the residual pool at synapses contacted by RL (n=10–12; p>0.05) and RS (n=10–12; p>0.05) terminals were the same as that in normal LGN. (E) Western blots of GTP-bound active Ras, total Ras, phosphorylated CaMKII and total CaMKII in LGN from normal control rats and rats with eyelids stitched. For each set of cell lysates, 35 μg protein was used to purify and blot GTP-bound Ras, 7.5 μg protein was used to directly blot total Ras, and 45 μg protein was used to blot phos-CaMKII and total CaMKII. (F) Relative amounts of Ras-GTP (n=12; p<0.05), total Ras (n=12; p=0.40), phos-CaMKII (n=16; p<0.01), and total CaMKII (n=16; p=0.47) in LGN from normal control rats and rats with eyelids stitched. The relative values and standard errors were normalized to average amounts of Ras-GTP, total Ras, phos-CaMKII or total CaMKII in LGN from normal control rats. (G) Upper, evoked AMPA-R-mediated responses at retinogeniculate (RG) synapses from non-expressing (Ctrl) and GluR1-GFP expressing neurons in rats with eyelids stitched and rats with LGN infusion of TTX recorded at −60 mV and +40 mV. Lower left, RG AMPA responses in neurons expressing GluR1-GFP in rats with eyelids stitched (n=22; p=0.88), or with LGN infusion of TTX (n=15; p=0.91) relative to neighboring control neurons. Lower right, rectification of RG AMPA responses in GluR1-GFP expressing neurons in rats with eyelids stitched (n=22; p=0.57), or with LGN infusion of TTX (n=15; p=0.14) relative to neighboring control cells. (H) Upper, evoked AMPA-R- and NMDA-R-mediated responses at RG synapses from non-expressing (Ctrl) and GluR1ct-GFP expressing neurons in rats with eyelids stitched and with LGN infusion of TTX recorded at −60 mV and +40 mV. Lower left, RG AMPA responses in neurons expressing GluR1ct-GFP in rats with eyelids stitched (n=13; p=0.65), in rats with LGN infusion of TTX (n=14; p=0.64), in wild type (WT) mice with eyelids stitched (n=17; p=0.80), or in GluR1 knockout (KO) mice with eyelids stitched (n=17; p=0.72) relative to neighboring control neurons. Lower right, RG NMDA responses in neurons expressing GluR1ct-GFP in rats with eyelids stitched (n=13; p=0.38), in rats with LGN infusion of TTX (n=14; p=0.25), in wild type (WT) mice with eyelids stitched (n=17; p=0.69), or in GluR1 knockout (KO) mice with eyelids stitched (n=17; p=0.44) relative to neighboring control cells. AMPA-R and NMDA-R mediated current amplitude and standard errors were normalized to average values from control cells. Asterisks indicate statistical significance (Mann-Whitney Rank Sum or Wilcoxon test). See Supplemental Data for the values.

Fig. 8. Vision-dependent activity drives synaptic insertion of endogenous and recombinant GluR1

(A–B) GluR1 immunogold labeling at synapses contacted by RL terminals in GluR1ct-GFP expressing neurons from control rats (Ctrl, A1–2) and rats with eyelids stitched (ES, A3–4), and GFP immunogold labeling at synapses contacted by RL terminals in GluR1-GFP expressing neurons from control rats (Ctrl, B1–2) and rats with eyelids stitched (ES, B3–4). Red arrows point to silver-enhanced gold particles associated with PSDs postsynaptic to RL terminals. Scale bar applies to A–B. (C) Relative distributions of GluR1 silver-gold particles at synapses contacted by RL (red) and RS (blue) terminals in GluR1ct-GFP expressing neurons from control rats (C1: n=499 for RL; n=1,154 for RS) and rats with eyelids stitched (C2: n=508 for RL; n=1,196 for RS), and GFP silver-gold particles at synapses contacted by RL (red) and RS (blue) terminals in GluR1-GFP expressing neurons from control rats (C3: n=487 for RL; n=1,161 for RS) and rats with eyelids stitched (C4: n=524 for RL; n=1,168 for RS). (D) Left, average percentages of GluR1 or GFP silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in GluR1ct-GFP neurons from control rats (n=10, p<0.001) and rats with eyelids stitched (n=10, p<0.005), and in GluR1-GFP neurons from control rats (n=10, p>0.05) and rats with eyelids stitched (n=10, p>0.05) relative to normal control LGNs presented in figure 2E. Note no significant differences for average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in GluR1ct-GFP neurons from control rats and those from rats with eyelids stitched (p>0.05), but significant differences for average percentages of GFP silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in GluR1-GFP neurons from control rats and those from rats with eyelids stitched (p<0.005). Right, average percentages of GluR1 or GFP silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in GluR1ct-GFP neurons from control rats (n=10, p>0.05) and rats with eyelids stitched (n=10, p>0.05), and in GluR1-GFP neurons from control rats (n=10, p>0.05) and rats with eyelids stitched (n=10, p>0.05) were unchanged compared to normal control LGN. Note (not shown) that average percentages of GluR1 or GFP silver-gold particles in the residual pool at synapses contacted by RL (n=9–12; p>0.05) and RS (n=9–12; p>0.05) terminals were the same as that in normal LGN. Asterisks indicate statistical significance (Mann-Whitney Rank Sum test). See Supplemental Data for the values.

Fig. 9. Synaptic stimulation drives GluR1 insertion at corticogeniculate synapses

(A–B) GluR1 immunogold labeling at synapses contacted by RS terminals in LGN after synaptic stimulation of CG pathway in normal bath solution (A1–3), and bath solution with additional 100 μM DL-APV (B1–2). Scale bar applies to A–B. (C) Relative distributions of GluR1 silver-gold particles at synapses contacted by RL (red) and RS (blue) terminals in LGN after synaptic stimulation of CG pathway in normal bath solution (C1: n=575 for RL; n=1,381 for RS), and bath solution with 100 μM DL-APV (C2: n=506 for RL; n=1,236 for RS). (D) Left, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RS terminals in LGN after synaptic stimulation of CG pathway in normal bath solution (n=10, p<0.001), or bath solution with 100 μM DL-APV (n=10, p>0.05) relative to normal control LGNs presented in figure 2E. Right, average percentages of GluR1 silver-gold particles in synaptic and deliverable pools at synapses contacted by RL terminals in LGN after synaptic stimulation of CG pathway in normal bath solution (n=10, p>0.05), or bath solution with 100 μM DL-APV (n=10, p>0.05) were unchanged compared to normal control LGN. Note (not shown) that average percentages of GluR1 silver-gold particles in the residual pool at synapses contacted by RL (n=10–12; p>0.05) and RS (n=10–12; p>0.05) terminals were the same as that in normal LGN. (E) Upper, evoked AMPA-R-mediated responses at corticogeniculate (CG) synapses from non-expressing (Ctrl) and GluR1-GFP expressing neurons in rats after synaptic stimulation of CG pathway in normal bath solution and bath solution with DL-APV recorded at −60 mV and +40 mV. Lower left, CG AMPA responses in neurons expressing GluR1-GFP in rats after synaptic stimulation of CG pathway in normal bath solution (n=19; p=0.72) or bath solution with 100 μM DL-APV (n=22; p=0.76) relative to neighboring control neurons. Lower right, rectification of CG AMPA responses in GluR1-GFP expressing neurons in rats after synaptic stimulation of CG pathway in normal bath solution (n=19; p<0.05) or bath solution with 100 μM DL-APV (n=22; p=0.66) relative to neighboring control cells. (F) Upper, evoked AMPA-R- and NMDA-R-mediated responses at CG synapses from non-expressing (Ctrl) and GluR1ct-GFP expressing neurons in rats after synaptic stimulation of CG pathway in normal bath solution and bath solution with DL-APV recorded at −60 mV and +40 mV. Lower left, CG AMPA responses in neurons expressing GluR1ct-GFP in rats after synaptic stimulation of CG pathway in normal bath solution (n=16; p<0.05) or bath solution with 100 μM DL-APV (n=16; p=0.84) relative to neighboring control neurons. Lower right, CG NMDA responses in neurons expressing GluR1ct-GFP in rats after synaptic stimulation of CG pathway in normal bath solution (n=16; p=0.96) or bath solution with 100 μM DL-APV (n=16; p=0.33) relative to neighboring control cells. AMPA-R and NMDA-R mediated current amplitude and standard errors were normalized to average values from control cells. Asterisks indicate statistical significance (Mann-Whitney Rank Sum or Wilcoxon test). See Supplemental Data for the values. (G) Model for triple AMPA-R pools at synapses.