A synapse-specific refractory period for plasticity at individual dendritic spines

The study addresses how synapses can remain both plastic to encode new memories and stable to retain old memories. A leading hypothesis proposes that recently strengthened synapses enter a temporary refractory period during which they are resistant to further potentiation (plasticity saturation). Evidence from behavioral and circuit-level studies suggests that nonoverlapping populations of synapses encode distinct tasks learned close in time and that prior potentiation can limit subsequent LTP induction. However, the cellular and molecular mechanisms, and the spatial and temporal scales at which a refractory period operates at individual synapses, remain unclear. The authors test whether prior potentiation at single dendritic spines of hippocampal CA1 neurons induces a synapse-specific refractory period, examine its time course, determine whether it is postsynaptically initiated, and probe the role of CaMKII signaling and PSD scaffolding proteins (PSD95 vs. PSD93) in establishing and releasing this refractory period.

Prior work indicates that learning engages changes in synaptic strength and structure, with stability and plasticity jointly required for memory retention and updating. Circuit-level studies in hippocampus showed that recently potentiated circuits are unable to undergo further potentiation within hours (e.g., tetanic, theta-burst, and pairing protocols), and in vivo saturation of potentiation impairs learning. Behavioral evidence supports that separate tasks learned close in time recruit nonoverlapping synapse populations. Ultrastructural studies reported delayed PSD enlargement (30 min to 2 h) after LTP, implicating postsynaptic structural constraints in limiting further plasticity. Molecular imaging showed delayed accumulation of PSD scaffolding proteins (including PSD95) despite rapid spine volume increases. CaMKII is central to LTP/LTD; metaplastic regulation could involve altered calcium influx (e.g., NMDAR–SK2 feedback, ER remodeling), NMDAR subunit switching (GluN2B→GluN2A), or endogenous CaMKII inhibitors (CaMK2N1) upregulated after LTP. These findings frame expectations that a refractory period could be synapse-local, time-limited, and regulated by postsynaptic signaling and scaffolding protein availability.

Organotypic hippocampal slice cultures (300–400 µm) were prepared from P6–P8 C57BL/6 mice (both sexes) per UC Davis IACUC approval. Slices were transfected via biolistic gene transfer (180–210 psi) with constructs depending on experiment: EGFP-N1; tDimer-dsRed + SEP-GluA2; green-Camui-α (CaMKII activity reporter) + CyRFP1; DsRedExpress ± GFP-tagged PSD95α or PSD93α. Transfection occurred 24 h to 4 d prior to imaging depending on constructs. Two-photon time-lapse imaging of transfected CA1 pyramidal neurons (10–50 µm depth; 11–17 DIV) used a custom microscope controlled by ScanImage. Image stacks (512×512 px; 0.02 µm/px; 1 µm z-steps) were collected every 5 min at 29 °C in recirculating ACSF (127 NaCl, 25 NaHCO3, 1.2 NaH2PO4, 2.5 KCl, 25 D-glucose; 95% O2/5% CO2; 310 mOsm; pH 7.2) containing 1 µM TTX, 0 Mg2+, and 2 mM Ca2+. MNI-glutamate (2.5 mM) was added ≥15 min before uncaging. High-frequency uncaging (HFU) of glutamate delivered 60 pulses (720 nm; ~7.5–9.5 mW) at 2 Hz with pulse durations of 2 ms (standard), 4 ms (HFU+), or 5–6 ms (HFU++). ACSF for uncaging contained 2 mM Ca2+, 0 Mg2+, 2.5 mM MNI-glutamate, and 1 µM TTX. Target spines were chosen from the 25–50% quartile of initial sizes. The uncaging beam was parked ~0.5–1 µm from the spine head away from the dendrite. Cells without noticeable transients to HFU++ were excluded. Image analysis used custom MATLAB software: 3×3 median filtering, background subtraction, and integrated fluorescence quantification from boxed regions around spine heads. Spine volume estimates used the cell fill fluorescence (EGFP, tDimer-dsRed, or DsRedExpress). Bar graphs summarized averages from specified post-HFU windows: HFU0 (20–30 min), HFU30 (50–60 min), HFU45 (60–70 min), HFU60 (65–75 min), HFU90 (80–90 min). Relative spine sizes were normalized to mean fluorescence of all spines on the same dendrite using the two pre-HFU time points. Two-photon FLIM measured CaMKII activity using time-correlated single-photon counting with GaAs(P) detectors. Images (24 frame scans; 128×128; 0.7 µm/px) were analyzed with custom MATLAB (Yasuda lab). CaMKII activity was quantified as ΔLifetime from baseline (average of 6 min pre-HFU) at each time point. Statistics: Data are mean ± SEM. Statistical tests (GraphPad Prism) included paired/unpaired Student’s t-tests for two-group comparisons and one/two-way ANOVAs with Bonferroni or Tukey post hoc tests for multiple comparisons (α = 0.05). P values are reported in text and n values in figure legends. Data availability: All study data are in the article and SI Appendix; raw image and analysis files are deposited at Dryad (DOI: 10.5061/dryad.ghx3ffc0b).

- A single LTP-inducing HFU stimulus at an individual dendritic spine (HFU0) produced robust long-term spine growth (e.g., 210% ± 20%; P < 0.001) and increased surface SEP-GluA2 (142% ± 9%; P < 0.001). - An identical second HFU at the same spine 30 min later (HFU30) failed to induce additional long-term spine growth (102% ± 12%; P > 0.99) or further SEP-GluA2 insertion (102% ± 4%; P > 0.99), while size-matched control spines on different dendrites grew/increased SEP-GluA2 (volume: 170% ± 18%; P < 0.01; SEP-GluA2: 139% ± 11%; P < 0.001). This establishes a refractory period at the previously potentiated spine. - The refractory period is synapse-specific: neighboring spines (≤10 µm; avg 2.9 ± 0.6 µm) on the same dendrite grew in response to HFU40 (171% ± 21%; P = 0.004), comparable to control spines (171% ± 21%; P = 0.048), while the original target remained saturated at 45 min. - CaMKII activation (Camui-α FLIM) was robust during HFU0 (peak ΔLifetime: 109 ± 27 ps) but significantly reduced at the same spine during HFU30 (~60% reduction; 44 ± 9 ps). Size-matched control spines at HFU30 exhibited peak ΔLifetime comparable to initial (90 ± 14 ps; P = 0.54 vs. HFU0 target). - Increasing stimulus strength at 30 min partially overcame structural saturation: HFU30+ (4 ms) yielded 136% ± 6% (P = 0.17) and HFU30++ (5–6 ms) 144% ± 12% (P = 0.06) spine volume, still below typical control HFU30-induced growth (from Fig. 1E). - Stronger HFU30+ fully restored CaMKII peak activation to levels indistinguishable from HFU0 target and control HFU30 (HFU0 target: 98.0 ± 21 ps; HFU30 target with strong stimulus: 106 ± 21 ps; HFU30 control: 127 ± 27 ps; P = 0.61 and P = 0.99), indicating that restoring CaMKII activation alone is insufficient to fully recover long-term spine growth. - The refractory period is released within 60 min: HFU60 at the same target spine produced additional long-term growth (135% ± 11%; P < 0.01) while the HFU0-induced growth was maintained over the interval (20–30 min: 171% ± 13%; 55–60 min: 166% ± 17%; P = 0.8). Control spines showed comparable growth to HFU60 (137% ± 9%; P < 0.01). - Elevating PSD95-GFP expression shortened the refractory period: target spines exhibited additional growth to HFU45 (136% ± 10%; P < 0.05), comparable to control spines (140% ± 13%; P < 0.05), even with AMPAR currents blocked by NBQX (target HFU45: 148% ± 14%; P < 0.05). - Elevating PSD93-GFP did not release the refractory period: target spines showed no significant growth to HFU45 (109% ± 7%; P > 0.99), while control spines grew robustly (164% ± 20%; P < 0.05). PSD95-GFP and PSD93-GFP expression levels and initial spine sizes were comparable. Overall, prior potentiation induces a synapse-specific, postsynaptically initiated refractory period lasting ~45–60 min, associated with reduced CaMKII signaling; recovery of CaMKII alone does not restore plasticity, but increased PSD95 (not PSD93) or waiting 60 min reinstates the capacity for potentiation.

The findings demonstrate that a single glutamatergic LTP-inducing event at an individual CA1 spine is sufficient to trigger a local, synapse-specific refractory period that prevents further structural and functional potentiation for approximately 45–60 minutes. This refractory period is initiated postsynaptically, as the uncaging stimulus bypasses presynaptic release, and correlates with reduced CaMKII activation. Although increasing stimulation strength restores CaMKII signaling, it does not fully recover long-term spine growth, indicating additional mechanisms beyond CaMKII govern the refractory state. The temporal release of this refractory period aligns with delayed enrichment and PSD enlargement reported by ultrastructural and live imaging studies. Directly increasing PSD95 levels—without similar effects from PSD93—restores plasticity within the refractory window, implicating PSD95-mediated PSD expansion and/or signaling regulation (e.g., effects on STEP61 and GluN2B stabilization) in terminating the refractory state. These results resolve key aspects of spatial and temporal regulation of metaplasticity at single synapses and suggest synaptic scaffolding availability acts as a gatekeeper for subsequent potentiation. Implications include safeguarding newly formed memory traces by excluding recently potentiated synapses from subsequent encoding, preventing runaway potentiation, and providing a mechanistic basis for the benefits of spaced learning, wherein inter-session intervals allow synapses to recover their plasticity capacity. Targeting molecular pathways that modulate the refractory period (e.g., PSD95 dynamics) could offer strategies to improve learning under disease conditions.

This study establishes that potentiation at a single hippocampal CA1 spine induces a synapse-specific refractory period for further potentiation that lasts ~45–60 min, is initiated postsynaptically, and is marked by reduced CaMKII activation. While stronger stimulation restores CaMKII signaling, full recovery of plasticity requires time-dependent postsynaptic changes, notably the replenishment of PSD components. Increasing PSD95 levels, but not PSD93, accelerates release from the refractory period, enabling renewed potentiation within 45 min. These results identify PSD95 availability as a critical regulator of single-spine metaplasticity and provide a cellular mechanism underpinning the efficacy of spaced learning and the preservation of newly formed memories. Future research should dissect the precise molecular pathways linking PSD95 dynamics to refractory period termination, evaluate the roles of endogenous CaMKII inhibitors, NMDAR–SK2 feedback, ER remodeling, and NMDAR subunit switching at single synapses, and test whether similar synapse-specific refractory mechanisms operate in vivo across different activity patterns and brain regions.

- Experiments were performed in organotypic hippocampal slice cultures using two-photon glutamate uncaging, which bypasses presynaptic vesicle release; presynaptic contributions to refractory mechanisms were not assessed. - The ability to increase uncaging pulse duration was limited (beyond 6 ms caused cell health issues), constraining tests of stimulus-strength rescue. - Findings pertain to CA1 pyramidal neuron spines under specific ACSF conditions (0 Mg2+, 2 mM Ca2+, TTX) and defined developmental stages (DIV 11–17); generalization to in vivo conditions and other synapse types requires further validation. - Although recovery at individual synapses was demonstrated, the study does not exclude contributions from recruitment of additional naïve synapses or circuit-level processes under different stimulation patterns. - Proposed molecular mechanisms (e.g., CaMK2N1 upregulation, NMDAR subunit switching, SK2 feedback) are discussed but not directly tested here.