Metaplasticity at single glutamatergic synapses

Abstract

Optimal function of neuronal networks requires interplay between rapid forms of Hebbian plasticity and homeostatic mechanisms that adjust the threshold for plasticity, termed metaplasticity. Numerous forms of rapid synapse plasticity have been examined in detail. However, the rules that govern synaptic metaplasticity are much less clear. Here, we demonstrate a local subunit-specific switch in NMDA receptors that alternately primes or prevents potentiation at single synapses. Prolonged suppression of neurotransmitter release enhances NMDA receptor currents, increases the number of functional NMDA receptors containing NR2B, and augments calcium transients at single dendritic spines. This local switch in NMDA receptors requires spontaneous glutamate release but is independent of action potentials. Moreover, single inactivated synapses exhibit a lower induction threshold for both long-term synaptic potentiation and plasticity-induced spine growth. Thus, spontaneous glutamate release adjusts plasticity threshold at single synapses by local regulation of NMDA receptors, providing a novel spatially delimited form of synaptic metaplasticity.

Figure 1. Local Inactivation Enhances NMDAR Currents at Single Synapses

(A) Illustration of the experimental system. Hippocampal neurons were sparsely infected with a lentivirus expressing synaptophysin-GFP-IRES-tetanus toxin light chain (SphGFP-IRES-TeNT) at DIV11-12. Individual postsynaptic neurons were visualized by subsequent mCherry transfection at DIV17-19. Left panel, 7–10 days after infection, a small fraction of the spines on mCherry-filled neurons contacted presynaptic boutons that were visible as green due to the expression of synaptophysin-GFP and silenced due to the co-expression of tetanus toxin light chain (TeNT) (arrows). The large majority of synapses originated from uninfected neurons, which were spontaneously active. Upper right, a magnified view from the boxed region showing a silenced synapse on a dendritic spine with a contacting green bouton (arrowhead) and neighboring active synapses. Lower right, schematic diagram of single-synapse inactivation. S, silenced; A, active. Scale bars: 50 μm left, 1 μm right. (B) Paired recordings of synaptic currents at neighboring active and silenced synapses. Uncaging-evoked excitatory postsynaptic currents (uEPSCs) were recorded upon focal two-photon glutamate uncaging at nearby active (A) and silenced (S) synapses on dendritic spines. For each pair, average uEPSCs were sequentially recorded from active (uEPSC-A) and silenced synapses (uEPSC-S) at holding potentials of −70 mV (bottom trace) and +40 mV (top trace). Scale bar, 1 μm. (C) Enhanced NMDAR-mediated uEPSCs at silenced synapses. Upper panel, average uEPSCs of active (A, black) and silenced (S, green) synapses from 19 A-S pairs recorded at +40 mV. Bottom panel, means ± SEM of the NMDAR-mediated component of the uEPSC measured 45 ms after the peak. Active (A), 8.5 ± 1.3 pA; silenced (S), 14.0 ± 1.9 pA; **p < 0.01, paired t-test; n = 19 A-S pairs from 16 neurons. (D) Across pairs of active (A) and silenced (S) synapses recorded, silenced synapses consistently showed higher NMDAR/AMPAR uEPSC ratios than neighboring active synapses. A, 0.7 ± 0.1; S, 1.1 ± 0.1; paired t-test, **p < 0.01).

Figure 2. Single Silenced Synapses Accumulate NR1 and NR2B but not NR2A

(A) Immunocytochemistry for NR1, NR2A and NR2B at silenced and active synapses. Silenced synapses (arrowheads) contain more NR1 and NR2B compared to neighboring active synapses on dendritic spines. The dashed white line indicates the dendritic outline. Scale bar, 1 μm. (B) Subunit-specific accumulation of NMDARs at silenced synapses. The synaptic content of NR1, NR2A and NR2B detected by immunocytochemistry was quantified at silenced synapses and normalized to mean values at active synapses (S/ Ā ; see Experimental Procedures for details). The cumulative plot (B1) shows overlapping right-shifted NR1 and NR2B distributions (p > 0.05, Kolmogorov-Smirnov test), which significantly deviate from the NR2A trace (p < 0.05 relative to both NR1 and NR2B, Kolmogorov-Smirnov test). Panels B2–B4 show frequency distributions of S/ Ā values for NR1, NR2A and NR2B subunits. S/ Ā >1 indicates accumulation at silenced synapses. (C) Across neurons, silenced synapses contained nearly 2-fold more NR1 and NR2B but similar levels of NR2A compared to active synapses. **p < 0.01. (D) Homosynaptic accumulation of NR1 at silenced synapses. Shown is the spatial profile of NR1 accumulation at the nearest proximal (P) and distal (D) synapses in a 30 μm long stretch of dendrite centered by a single silenced synapse. NR1 accumulation is spatially confined to single silenced synapses. S/ Ā : silenced, 1.8 ± 0.2; P1, 1.1 ± 0.1; P2, 1.1 ± 0.1; P3, 1.2 ± 0.1; D1, 1.0 ± 0.1; D2, 0.8 ± 0.1; D3, 1.1 ± 0.1; n = 10 neurons. p_S-Ā < 0.01 for comparisons between single silenced synapse and its neighboring active synapses; paired t-test. (E) Inactivity-induced accumulation of NR2B is homosynaptic. Analysis as in (D). S/ Ā : silenced, 2.1 ± 0.2; P1, 1.1 ± 0.2; P2, 0.9 ± 0.3; P3, 1.1 ± 0.3; D1, 1.1 ± 0.2; D2, 0.9 ± 0.1; < 0.01 for comparisons between single silenced synapse D3, 1.1 ± 0.4; n = 18 neurons. p_S-Ā and its neighboring active synapses; paired t-test.

Figure 3. Local Spontaneous Release Determines NMDAR Content

(A) Schematic diagram indicating how specific manipulations affect action potential (AP)-evoked and spontaneous (or miniature) glutamate release at single active (A) and TeNT-silenced (S) synapses. (B) Spontaneous glutamate release is sufficient to reduce NMDAR content at active synapses. Postsynaptic neurons contacted by active and TeNT-silenced (arrowheads) boutons were kept in the presence of TTX, AP5, or control solution for several days prior to fixation and immunolabeling for NR1 (see Experimental Procedures for details. Scale bar, 1 μm. (C–E) Quantitative analyses of NMDAR accumulation at TeNT-silenced synapses. The relative accumulation of NR1 (C), NR2B (D), and NR2A (E) at silenced synapses (S/ Ā ) was measured as in Figure 2 after prolonged treatment with TTX, AP5, or control (Ctrl) solution. Since both 3 and 7 days of pharmacological blockade resulted in similar effects on synaptic NMDAR accumulation (Figure S2C), pooled data are represented here. **p < 0.01. n.s., not significant. (F) The time course of NR1/NR2B accumulation was monitored upon AP5 washout. Postsynaptic neurons contacted by active and TeNT-silenced boutons were incubated in TTX/AP5 medium for 3 days to equalize NMDAR content and then switched to TTX only medium for 6, 12, or 24 hours prior to fixation and immunostaining for NR1 and NR2B. The relative abundance of NR1 (black) and NR2B (grey) at silenced synapses relative to active synapses (S/ Ā ) was measured over 24 hours. Data points are far right (squared) indicate S/ Ā values under chronic TTX treatments derived from (C) and (D).

Figure 4. NR2B-Mediated Currents Dominate at Single Inactive Synapses

(A) Left, application of the selective NR2B antagonist ifenprodil allows estimation of the fractional contribution of NR2B-containing receptors to NMDAR currents elicited by focal two photon glutamate uncaging (orange spot) at single spines. Right, representative trace of NMDAR-uEPSCs measured before (control) and after ifenprodil application, indicating partial block. (B) NMDAR currents at silenced synapses are more sensitive to ifenprodil. Shown are representative uEPSC traces recorded at −70 mV and + 40 mV in the presence of ifenprodil (3 μM) from one active synapse (A, left) and one silenced synapse (S, right). At the active synapse, uEPSC_+40mV contained both a fast ifenprodil-resistant AMPAR-uEPSC and a slower ifenprodil-resistant NMDAR component (arrow). The silenced synapse contained an intact AMPAR-uEPSC but little ifenprodil-resistant NMDAR component (arrow). (C) Average uEPSC_+40mV traces from active (top) and silenced (bottom) synapses recorded with (Ifen.) or without (Ctrl) ifenprodil. (D) Averaged ifenprodil-resistant uEPSC_+40mV traces from active (A) and silenced (S) synapses. (E) Data represent means ± SEM of NMDAR-uEPSCs at active (A) and silenced (S) synapses in the absence (Ctrl) or presence (Ifen.) of 3 μM ifenprodil. n = 12–26 synapses from 6–16 neurons for each condition. *p < 0.05, **p < 0.001. (F) Percent ifenprodil sensitivity of NMDAR-uEPSCs recorded at active (A) and silenced (S) synapses. *p_A-S < 0.05. (G) A second NR2B selective antagonist, Ro25-69821 (Ro25), has similar effects. Shown is the percent Ro25 sensitivity of NMDAR-uEPSCs recorded at active (A) and silenced (S) synapses. n = 7–26 synapses on 7–16 neurons for each condition. *p_A-S < 0.05.

Figure 5. Prolonged Inactivation Primes Single Synapses for Potentiation

(A) Silenced synapses exhibit an elevated Ca²⁺ transient compared to their active neighbors. Upper panel, shown are representative kymographs of Fluo-4FF signal upon local glutamate uncaging (arrowhead) at nearby active (A) and silenced (S) synapses on dendritic spines. Lower panel, plotted is the average change in Ca²⁺ concentration over time following local glutamate uncaging. See Experimental Procedures for details. (B) Data show means ± SEM of the peak change in Ca²⁺ concentration at active (A) and silenced (S) synapses under control conditions (Ctrl) or in the presence of ifenprodil (Ifen) or AP5. *p < 0.05; n = 8, 7, 4 pairs from 6, 6, 4 neurons respectively; paired t-test; n.s. not significant. (C) Pairing-induced LTP at single synapses. By pairing 30 focal glutamate uncaging pulses of 4 ms duration with postsynaptic depolarization, LTP of AMPAR-uEPSCs was reliably induced at the stimulated spine (orange arrowheads, left traces), but not at the non-stimulated neighboring spine (NS, right traces). Current traces before (black) and after (red) conditioning stimuli are shown. (D) Quantification of uEPSC potentiation following LTP induction. The potentiation of AMPAR-uEPSCs following a 30 pulse pairing LTP protocol persisted for more than 20 min (LTP, 193 ± 14% at 20 min post LTP induction, n = 13). The inset shows that the potentiation of active (A, black) and silenced (S, green) synapses was undistinguishable using the 30 pulse protocol (p > 0.05 for the last three time points shown). NS, non-stimulated neighboring spine. *p < 0.05. (E) Weak stimuli potentiate silenced but not active synapses. A weaker stimulation protocol (subthreshold protocol: 20 pulses of 1 ms duration paired with 0 mV depolarization), failed to potentiate active synapses (left traces), but triggered a sustained potentiation at silenced synapses (right traces). Current traces before (black) and after (red) conditioning stimuli are shown. (F) Quantification of uEPSC potentiation following subthreshold stimulation at active (Sub-A) and silenced (Sub-S) synapses compared to non-stimulated neighboring spines (NS). Active, 90 ± 13% at 20 min post-induction; silenced 235 ± 40%; n = 6, 9 respectively; NS, 106 ± 10%. *p < 0.05.

Figure 6. Subthreshold Stimuli Elicit Plasticity-Induced Spine Growth at Silenced Synapses

(A) LTP induction by two-photon glutamate uncaging (LTP, 30 pulses of 4 ms duration at 0.5 Hz in 0 Mg²⁺) triggers sustained spine growth at active synapses. For (A)–(C), time is indicated in minutes. NS, non-stimulated neighboring spine. Scale bar, 1 μm. (B) A weaker subthreshold stimulus (Sub) consisting of 20 glutamate uncaging pulses of 1 ms duration at 0.5 Hz in 0 Mg²⁺ does not elicit spine growth at active synapses. (C) The normally subthreshold protocol of 20 focal glutamate uncaging pulses of 1 ms duration (Sub, arrowhead) elicits robust spine growth at silenced (green bouton) synapses. (D) Quantitative analysis of spine volume change following a strong LTP protocol (30 pulse, 4 ms) or a subthreshold stimulus (20 pulses, 1 ms). All trials are shown. Bars indicate means. A, active synapse; S, silenced synapse; NS, non-stimulated synapse. 30 pulses: *p_A-NS < 0.01, *p_SNS < 0.01, p_A-S > 0.1; 20 pulses: p_A-NS > 0.1, *p_S-NS < 0.01, *p_A-S < 0.05. (F) Spine growth at silenced synapses elicited by subthreshold stimuli requires NR2B-containing NMDARs. In the presence of ifenprodil (Ifen) or AP5, the subthreshold protocol failed to induce spine enlargement at silenced synapses. Ctrl, 151 ± 11% at 20 min post-induction; Ifen, 101 ± 3%; AP5, 104 ± 2%; n = 6, 7, 4 respectively; *p < 0.05. n.s., not significant.