Non-ionotropic NMDA receptor signaling gates bidirectional structural plasticity of dendritic spines

Abstract

Experience-dependent refinement of neuronal connections is critically important for brain development and learning. Here, we show that ion-flow-independent NMDA receptor (NMDAR) signaling is required for the long-term dendritic spine growth that is a vital component of brain circuit plasticity. We find that inhibition of p38 mitogen-activated protein kinase (p38 MAPK), which is downstream of non-ionotropic NMDAR signaling in long-term depression (LTD) and spine shrinkage, blocks long-term potentiation (LTP)-induced spine growth but not LTP. We hypothesize that non-ionotropic NMDAR signaling drives the cytoskeletal changes that support bidirectional spine structural plasticity. Indeed, we find that key signaling components downstream of non-ionotropic NMDAR function in LTD-induced spine shrinkage are also necessary for LTP-induced spine growth. Furthermore, NMDAR conformational signaling with coincident Ca²⁺ influx is sufficient to drive CaMKII-dependent long-term spine growth, even when Ca²⁺ is artificially driven through voltage-gated Ca²⁺ channels. Our results support a model in which non-ionotropic NMDAR signaling gates the bidirectional spine structural changes vital for brain plasticity.

Keywords: CaMKII; LTD; LTP; NMDA receptor; dendritic spine; non-ionotropic; p38 MAPK; structural plasticity.

Figure 1.. p38 MAPK activity is required for LTP-induced spine growth, but not for synaptic strengthening

(A) Images of dendrites from DIV14-18 EGFP-expressing CA1 neurons from rat organotypic slices before and after HFU stimulation (yellow crosses) at individual dendritic spines (yellow arrowheads) with and without p38 MAPK inhibitor SB203580 (SB, 2 μM). (B and C) Inhibition of p38 MAPK with SB (black-filled circles/bar; 11 spines/11 cells) reduced HFU-induced spine growth compared with vehicle (red-filled circles/bar; 11 spines/11 cells). Volume of unstimulated neighbors (open circles/bars) was unaffected. (D and E) Top, average traces of uEPSCs from a target spine and an unstimulated neighbor on CA1 neurons from mouse organotypic slices before (gray) and 25 min after HFU stimulation during vehicle conditions (target, red; neighbor, dark gray) or in the presence of SB (target, black; neighbor, dark gray). Bottom, HFU-induced increases in uEPSC amplitude in vehicle (red; 8 spines/8 cells) were unaffected by p38 MAPK inhibition with SB (black; 9 spines/9 cells). uEPSC amplitudes of unstimulated neighboring spines (gray) were unaffected. (F) HFU induced a long-lasting uEPSC amplitude increase (red filled bar) compared with baseline, which is unaffected by p38 MAPK inhibition (black-filled bar). Two-way ANOVA with Tukey’s test used in (C) and (F) and two-way repeated-measures (RMs) ANOVA with Dunnett’s test to baseline used in (B), (D), and (E). Data are represented as means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 2.. Non-ionotropic NMDAR signaling pathway is required for LTP-induced spine growth

(A, D, and G) Images of dendrites from EGFP-expressing CA1 neurons from rat (A and D) and mice (G) organotypic slices at DIV14-18 before and after HFU stimulation (yellow crosses) of individual spines (yellow arrowheads) in the presence of L-TAT-GESV (1 μM) and the L-TAT-GASA control peptide (1 μM), during vehicle conditions and in the presence of MK2 inhibitor III (10 μM) or the NOS inhibitor, L-NNA (100 μM). (B and C) Disruption of the NOS1AP/nNOS interaction with L-TAT-GESV (black-filled circles/bar; 9 spines/9 cells), but not application of the inactive L-TAT-GASA control peptide (red-filled circles/bar; 9 spines/9 cells) inhibited persistent spine growth after LTP induction. Volume of the unstimulated neighbors (open circles/bars) was unchanged. (E and F) Inhibition of MK2 activity (black-filled circles/bar; 11 spines/11 cells) prevented HFU-induced persistent spine enlargement (red-filled circles/bar; 12 spines/12 cells). Volume of the unstimulated neighbors did not change (open circles/bars). (H and I) Inhibition of NOS activity (black-filled circles/bar; 11 spines/11 cells) prevented HFU-induced persistent spine enlargement (red-filled circles/bar; 12 spines/12 cells). Volume of the unstimulated neighbors did not change (open circles/bars). Two-way ANOVA with Tukey’s test used in (C), (F), and (I) and two-way RMs ANOVA with Dunnett’s test to baseline used in B, E, and H. Data are represented as means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 3.. Non-ionotropic NMDAR signaling with Ca²⁺-influx through voltage-gated calcium channels is sufficient to drive LTP-induced spine growth

(A) Left: proposed model in which activity-induced spine growth requires both non-ionotropic and ionotropic NMDAR signaling. Middle: schematic of experiment to test proposed model. Non-ionotropic NMDAR signaling is activated with glutamate, whereas calcium influx through the NMDAR is blocked with L-689. Instead, calcium influx is driven through VGCCs with the stronger HFU+ conditions in the presence of Bay K to favor opening of VGCCs. Right: control experiments block non-ionotropic NMDAR signaling by blocking glutamate binding to the NMDAR with CPP. (B) Images of dendrites from CA1 neurons of acute slices from P16-20 GFP-M mice before and after HFU+ stimulation (yellow crosses) of individual spines (yellow arrowheads) in the presence of L-689 (10 μM) and Bay K (10 μM) or in combination with CPP (50 μM). (C and D) HFU+ stimulation drives spine growth in the presence of Bay K, even when ion flow through the NMDAR is blocked with L-689 (red-filled circles/bar; 9 spines/9 cells), but not when non-ionotropic NMDAR signaling is blocked with CPP (black-filled circles/bar; 10 spines/10 cells). Volume of the unstimulated neighbors (open circles/bars) was unchanged. (E) Images of dendrites under experimental conditions in (B)–(D), except blocking the influx of calcium through VGCCs by removing Bay K and adding NBQX (50 μM). (F and G) HFU+ stimulation in the presence of L-689 and NBQX (blue-filled circles/bar; 7 spines/7 cells) led to spine shrinkage, which was blocked when non-ionotropic NMDAR signaling was inhibited with the presence of CPP (black-filled circles/bar; 6 spines/6 cells). Volume of the unstimulated neighbors (open circles/bars) did not change. Two-way ANOVA with Tukey’s test used in (D) and (G) and two-way RMs ANOVA with Dunnett’s test to baseline used in C and F. Data are represented as means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S2.

Figure 4.. CaMKII activity is required for LTP-induced spine growth, independent of calcium source

(A) Images of dendrites from CA1 neurons of acute slices from P16-20 GFP-M mice before and after HFU⁺ stimulation (yellow crosses) of individual spines (yellow arrowheads) in the presence of L-689 (10 μM) and Bay K (10 μM) or in combination with TAT-SCR (5 μM), TAT-CN21 (5 μM), or KN-62 (10 μM). (B and C) HFU+-induced spine growth in the presence of Bay K and L-689 is blocked by TAT-CN21 peptide (black-filled circles/bar; 6 spines/ 6 cells) or KN-62 (gray-filled circles/bars; 6 spines/6 cells), but not in the presence of control TAT-SCR peptide (red-filled circles/bar; 6 spines/ 6 cells). Volume of the unstimulated neighbors (open circles/bars) was unchanged. Two-way ANOVA with Tukey’s test used in (C) and two-way RMs ANOVA with Dunnett’s test to baseline used in B. Data are represented as means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. (D) Proposed model. Plasticity-inducing glutamatergic stimulation activates non-ionotropic NMDAR signaling, driving cofilin-dependent severing of the actin cytoskeleton, which, in the absence of strong Ca²⁺ influx, leads to spine shrinkage. On the other hand, during the strong Ca²⁺ influx associated with LTP induction, severed actin filaments serve as new starting points for actin filament nucleation and branching by the Ca²⁺- and CaMKII-dependent actin-modifying proteins to expand the F-actin cytoskeleton and drive spine growth.