Acute knockdown of AMPA receptors reveals a trans-synaptic signal for presynaptic maturation

Abstract

Newly formed glutamatergic synapses often lack postsynaptic AMPA-type glutamate receptors (AMPARs). Aside from 'unsilencing' the postsynaptic site, however, the significance of postsynaptic AMPAR insertion during synapse maturation remains unclear. To investigate the role of AMPAR in synapse maturation, we used RNA interference (RNAi) to knockdown AMPARs in cultured hippocampal neurons. Surprisingly, loss of postsynaptic AMPARs increased the occurrence of presynaptically inactive synapses without changing the release probability of the remaining active synapses. Additionally, heterologous synapses formed between axons and AMPAR-expressing HEK cells develop significantly fewer inactive presynaptic terminals. The extracellular domain of the AMPAR subunit GluA2 was sufficient to reproduce this effect at heterologous synapses. Indeed, the retrograde signalling by AMPARs is independent of their channel function as RNAi-resistant AMPARs restore synaptic transmission in neurons lacking AMPARs despite chronic receptor antagonist treatment. Our findings suggest that postsynaptic AMPARs perform an organizational function at synapses that exceeds their standard role as ionotropic receptors by conveying a retrograde trans-synaptic signal that increases the transmission efficacy at a synapse.

Figure 1

GluA RNAi in cultured hippocampal neurons leads to a loss of functional AMPARs. (A) AMPAR knockdown efficiencies of GluA1–3 shRNAs were determined by co-expression with GFP-tagged GluA1–3 in HEK293 cells (n=3 experiments/group; ^*P<0.05; ^**P<0.005). (B) Immunostaining of all three AMPAR subunits (red) in neurons after 5 days expression of pSuper empty-vector or GluA RNAi. GluA RNAi=GluA1+GluA2+GluA3 RNAi. Scale bar: 20 μm. (C) Average intensity of GluA1, GluA2/3, and all three subunits GluA1, GluA2, and GluA3 in the soma of GluA RNAi neurons compared with pSuper controls (n=8–10; ^***P<1 × 10⁻⁴). The average immunofluorescence intensity was normalized to neighbouring untransfected neurons to account for variability in immunostaining. (D) Example traces and quantification of somatic outside-out patch recordings from control and GluA RNAi neurons. AMPAR currents were evoked with a 3-s application of AMPA (100 μm) in the presence of cyclothiazide (100 μm) (n=11 neurons/group; ^***P<1 × 10⁻⁵). Scale bars: 50 pA, 1 s.

Figure 2

Reduced miniature excitatory synaptic transmission in AMPAR knockdown neurons. (A) Synaptic AMPARs in neurons expressing pSuper or GluA-shRNAs were identified by co-localization of AMPAR puncta, comprised of total GluA1, GluA2, and GluA3 immunostaining (red), with VGluT1 (blue). Scale bar: 10 μm. (B) Quantification of the average synaptic AMPAR immunoreactivity. (n=9–10 cells/group; ^***P<1 × 10⁻⁵). (C) Percentage of AMPAR-lacking synapses identified by VGluT1 puncta devoid of AMPAR immunostaining (n=9–10 cells/group; ^**P<0.005). (D) The expression levels of GFP-tagged GluA1 and GluA2 rescue constructs were unaffected by the corresponding GluA-shRNAs in HEK293 cells. (E) Example traces of mEPSCs from dissociated cultured hippocampal neurons at 12 DIV. Scale bars: 10 pA, 200 ms. (F, G) Cumulative probability plot of mEPSC amplitude and frequency. (Inset) Average mEPSC amplitude and frequency (n=26–39 cells/group; ^***P<1 × 10⁻⁵).

Figure 3

Reduced NMDAR-mediated synaptic responses in neurons after AMPAR knockdown. (A) Representative traces (each an average of five eEPSCs) and quantification of AMPAR-mediated eEPSCs in cultured hippocampal neurons elicited by extracellular local field stimulation (n=12–28 cells/group; ^**P<0.005; ^***P<1 × 10⁻⁴). Scale bars: 200 pA, 50 ms. (B) Representative traces (each an average of five eEPSCs) and quantification of NMDAR-mediated eEPSCs recorded at −60 mV in external solution with CNQX (10 μM), glycine (20 μm), and zero magnesium (n=11–25 cells/group; ^*P<0.05; ^***P<1 × 10⁻⁴). Scale bars: 200 pA, 200 ms. (C) Representative traces and quantification of whole-cell AMPA-evoked (3 s, 100 μm) currents from cultured neurons in the presence of cyclothiazide (100 μm) (n=11 cells/group; ^***P<1 × 10⁻⁵). Scale bars: 1 nA, 1 s. (D) Representative traces and quantification of whole-cell NMDA-evoked (3 s, 1 mM) currents from cultured neurons (n=16 cells/group; P>0.7). Scale bars: 0.5 nA, 0.5 s.

Figure 4

GluA RNAi does not alter the number of synaptic NMDARs. (A) Immunolabelling of the NMDAR subunit GluN1 (red) and VGluT1 (blue) on dendrites of neurons expressing either pSuper or GluA RNAi. Scale bar: 10 μm. (B, C) Synaptic GluN1 expression identified as GluN1 puncta co-localized with VGluT1. The mean intensity of synaptic GluN1 immunoreactivity (B), and the percentage of NMDAR-containing glutamatergic synapses (C), were analysed (n=10 cells/group; P>0.6). (D) Representative traces of dual component mEPSCs (grey trace). The AMPAR-only mEPSCs recorded from the same cells after APV perfusion is shown scaled to the peak of the dual component mEPSC (black trace). Scale bars: 4 pA, 20 ms. (E, F) Quantification of the mean AMPAR mEPSC amplitude (E) (n=18 cells/group; ^**P<0.002), and the mean NMDAR mEPSC amplitude (F). The scaled average AMPAR mEPSC was subtracted from the average dual component mEPSC to determine the NMDAR mEPSC amplitude (n=18 cells/group; P>0.3).

Figure 5

Reduction in postsynaptic AMPARs does not alter synapse density. (A) Immunostaining for VGluT1 (blue) and PSD-95 (red) at 12 DIV, 5 days after pSuper or GluA-shRNAs transfection. Scale bar: 10 μm. (B) Quantification of VGluT1 localization at synapses (n=12–15 cells/group; P>0.6). (C) Quantification of PSD-95 at synapses (n=12–15 cells/group; P>0.6). (B, C) The integrated puncta intensity is a measure of the sum of the pixel intensities within each puncta. The average puncta intensity is a measure of the average pixel intensity within each puncta. (D) Analysis of synaptic localization of several additional presynaptic and postsynaptic proteins (n=9–14 cells/group; ^*P<0.05; ^**P<0.005).

Figure 6

Loss of postsynaptic AMPARs does not alter the release probability at synapses. (A) Representative traces (each an average of five responses) of AMPAR EPSCs recorded from a postsynaptic neuron in paired recordings. The EPSCs were elicited by current injection to a synaptically connected presynaptic neuron in whole-cell current clamp recording mode. Scale bars: 20 pA, 20 ms. (B) Quantification of PPR from paired recordings at a 20- and 50-ms ISI (n=10–14 cells/group; P>0.8). (C) Example traces of NMDAR eEPSCs recorded at +40 mV in the presence of 10 μM MK-801, showing the progressive block of responses at each designated stimulus number. Scale bars: 200 pA, 100 ms. (D) Quantification of the NMDAR eEPSC amplitude in the presence of MK-801 at consecutive stimuli normalized to the amplitude of the first response. (Inset) Rate of response decay fitted with a double exponential equation (n=22–24 cells/group; P>0.2). (E) Example traces of the first five AMPAR eEPSCs in response to a 20-Hz stimulation (black traces). The response from the GluA RNAi neuron was scaled and superimposed onto the control response (grey trace). Each trace is an average of three individual 20 Hz trains of eEPSCs recorded from one neuron. Scale bars: 200 pA, 20 ms. (F) The amplitude of each successive response was normalized to the size of the first AMPAR eEPSC. (Inset) Time constants of the response decay fitted with a double exponential equation (n=13 cells/group; P>0.3).

Figure 7

AMPAR knockdown decreases the RRP size among all excitatory synapses. (A) Representative traces of NMDAR eEPSCs (left) and sucrose-evoked (3 s of 0.5 M sucrose) NMDAR responses (right). For each neuron, five NMDAR eEPSCs were recorded to generate an average response to extracellular stimulation, which was followed by a single application of sucrose to estimate the RRP size. (B) Average charge transfer of NMDAR eEPSCs elicited by extracellular field stimulation. (C) Average sucrose-evoked NMDAR responses from the same neurons in (B) (n=22–23 cells; ^**P<0.005). (D) The vesicular release probability estimated for each neuron by calculating the charge transfer of the average NMDAR eEPSC as a percentage of the total sucrose-evoked current (n=22–23 cells; P>0.7).

Figure 8

GluA RNAi increases the number of functionally inactive glutamatergic presynaptic terminals. (A) The Syt1 antibody uptake (red) and postfixation immunostaining of VGluT1 (blue) were performed. Active glutamatergic presynaptic terminals exhibit Syt1 immunostaining (arrows), whereas inactive terminals do not (arrowheads). (B) Quantification of the number of functionally inactive glutamatergic synapses on transfected neurons (n=18–30 neurons/group; ^***P<0.001). (C) Quantification of the mean intensity of Syt1 immunostaining at active glutamatergic synapses (n=18–30 neurons/group; P>0.1). (D) Representative traces and quantification of NMDAR eEPSCs recorded from neurons treated with CNQX (10 μM) throughout the 5 days of construct expression (n=16–33 cells/group; ^***P<0.001). Scale bars: 200pA, 200 ms.

Figure 9

Postsynaptic AMPARs participate directly in trans-synaptic retrograde signalling to influence glutamate release at a subset of presynaptic terminals. (A) Images of heterologous synapse formation between HEK293 cells and neurons. HEK293 cells were transfected with NL1 alone, or NL1 with either a GFP-tagged AMPAR subunit, GluA1 or GluA2, or a GFP-tagged kainate receptor subunit, GluK2 and co-plated with hippocampal neurons. Glutamatergic presynaptic terminals were identified by VGluT1 puncta (blue). Synaptic vesicle cycling at each terminal was measured by Syt1 antibody uptake (red). Functionally inactive presynaptic terminals were identified as VGluT1 puncta that lack Syt1 immunofluorescence (arrows). Scale bar: 10 μm. (B) Quantification of functionally inactive presynaptic terminals. The fraction of inactive glutamatergic terminals on each HEK293 cell was calculated and normalized to the fraction of inactive terminals at neighbouring neuronal synapses (n=24–27 cells; ^**P<0.005). (C) Mean Syt1 uptake intensity at heterologous synapses (normalized to the mean Syt1 uptake intensity at neighbouring neuronal synapses; n=24–27 cells/group; P>0.05). (D) Mean VGluT1 puncta intensity at heterologous synapses (normalized to the VGluT1 intensity at neighbouring neuronal synapses, n=24–27 cells/group; P>0.05). (E) Density of glutamatergic synaptic contacts made onto HEK293 cells (n=24–27 cells/group; P>0.1). (F) Image of heterologous synapse formation on a HEK293 cell co-expressing NL1 and GluA2 ecto. Similar to (A), arrows indicate functionally inactive presynaptic terminals. Scale bar: 10 μm. (G) Quantification of functionally inactive presynaptic terminals formed on HEK293 cells expressing NL1 alone, NL1+GluA2, or NL1+GluA2 ecto (n=27–31; ^**P<0.005; ^***P<0.001).

Figure 10

N-cadherin is not required for AMPAR retrograde signalling in presynaptic maturation. (A) Immunoblot of endogenous N-cadherin from lysate of 14 DIV hippocampal culture and HEK293 cells. (B) N-cadherin RNAi delivered with lentivirus had a 94% knockdown efficiency in hippocampal cultures (n=3 experiments). (C) Representative images of N-cadherin knockdown in neuronal dendrites (red: N-cadherin; green: MAP2). Scale bar: 10 μm. (D) Quantification of the number of inactive glutamatergic synapses formed on HEK293 cells normalized to neighbouring neuronal synapses (n=13–15 cells/group; ^*P<0.05; ^**P<0.01; ^***P<0.005). (E) Representative images of 12 DIV neurons labelled with Syt1 antibody uptake (red) and postfixation VGluT1 immunostaining (blue). Neuronal dendrites were identified by MAP2 immunostaining (green). Arrows indicate active glutamate release sites. Scale bar: 10 μm. (F) Quantification of the mean Syt1 puncta intensity at active synapses and the percent of inactive presynaptic terminals (n=18 cells/group; P>0.2).