Synaptic Plasticity

Synaptic plasticity is the fundamental biological property whereby chemical and electrical synapses dynamically increase or decrease their transmission strength (w_ij) in response to patterns of presynaptic and postsynaptic action potential activity. First proposed conceptually as a cellular basis of associative memory by Santiago Ramón y Cajal (1894), formalized mathematically by Donald O. Hebb in The Organization of Behavior (1949), and demonstrated experimentally as long-term potentiation (LTP) in the rabbit hippocampus by Terje Lømo and Timothy Bliss in 1966–1973, synaptic plasticity converts transient electrical firing patterns into persistent structural, receptor-trafficking, and transcriptional modifications across the nervous system and its connectome.

In biophysical terms, the amplitude of a postsynaptic current (EPSC = N · p_r · q) is the product of three quantal parameters: the number of functional release sites (N), the presynaptic probability of vesicular neurotransmitter release (p_r), and the postsynaptic quantal response to a single vesicle (q, set primarily by the number and conductance of postsynaptic receptors). Different forms of synaptic plasticity modify p_r, q, or N over timescales spanning eight orders of magnitude—from 10 milliseconds to decades.

At nearly all central and neuromuscular synapses, pairs or trains of presynaptic action potentials alter neurotransmitter release probability (p_r) transiently without requiring protein synthesis or postsynaptic receptor insertion:

1. Paired-Pulse Facilitation (PPF, τ ≈ 20–500 ms) and Post-Tetanic Potentiation (PTP, τ ≈ 30 s–5 min): At synapses with a low initial release probability (p_r < 0.2) (such as hippocampal CA3→CA1 Schaffer collateral or cerebellar parallel fiber synapses), when two action potentials arrive within 20–100 ms, the second excitatory postsynaptic potential (EPSP₂) is up to 2 to 3 times larger than the first (paired-pulse ratio PPR = EPSP₂ / EPSP₁ > 1). As formulated by Bernard Katz and Ricardo Miledi (1968) in the residual calcium hypothesis, vesicular exocytosis bridged by synaptotagmin-1 scales with the fourth power of local intracellular calcium concentration (Release ∝ [Ca²⁺]_i⁴) and is aided by the high-affinity fast facilitation sensor synaptotagmin-7. Small amounts of residual Ca²⁺ left over from the first spike summate nonlinearly with Ca²⁺ influx during the second spike.
2. Short-Term Synaptic Depression (STD, τ ≈ 100 ms–10 s): At synapses with a high initial release probability (p_r > 0.5) (such as the calyx of Held or climbing fiber synapses), high-frequency stimulation rapidly depletes the Readily Releasable Pool (RRP) of docked synaptic vesicles faster than endocytosis and Munc13/RIM-mediated vesicle priming can replenish them (PPR < 1), acting as a dynamic low-pass temporal filter and cortical gain control.

Hebb's Postulate and the NMDA Receptor Coincidence Detector

In 1949, Donald Hebb postulated that when axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it, metabolic or growth changes take place such that A's efficiency in firing B is increased (summarized by Carla Shatz as "cells that fire together, wire together").

At canonical glutamatergic synapses in the hippocampal CA1 region and neocortex, the molecular machine that implements Hebbian coincidence detection is the N-methyl-D-aspartate (NMDA) subtype of ionotropic glutamate receptor (heterotetramers of two GluN1 and two GluN2A / GluN2B subunits), operating alongside AMPA receptors (AMPARs, GluA1–GluA4):

1. Baseline Low-Frequency Transmission (Mg²⁺ Block): When a single presynaptic action potential releases glutamate onto a resting dendritic spine (V_m ≈ −70 mV), glutamate binds both AMPA and NMDA receptors. AMPA receptors open immediately, conducting Na⁺ inward to generate the fast excitatory postsynaptic potential (EPSP). However, at −70 mV, the pore of the NMDA receptor is physically plugged from the outside by an extracellular magnesium ion (Mg²⁺), preventing ion flow even though glutamate is bound.
2. Coincident Presynaptic Release + Postsynaptic Depolarization (LTP Induction): When high-frequency presynaptic stimulation (e.g., 100 Hz tetanus or theta-burst stimulation) or a back-propagating postsynaptic action potential depolarizes the dendritic spine above −35 to −20 mV at the exact same time that presynaptic glutamate is bound to the NMDA receptor, positive intracellular voltage electrostatically expels the Mg²⁺ plug from the NMDAR pore. Unblocked NMDA receptors exhibit high permeability to Ca²⁺ (P_Ca/P_Na ≈ 10:1), driving a large, rapid surge of [Ca²⁺] (>5–10 μM) inside the dendritic spine head.
3. CaMKIIα Autophosphorylation and AMPA Receptor Slot Trapping (LTP Expression): High spine Ca²⁺ binds calmodulin (CaM), which activates the dodecameric holoenzyme Ca²⁺/calmodulin-dependent protein kinase II (CaMKIIα)—the most abundant protein in the postsynaptic density. Adjacent activated subunits within the 12-subunit CaMKII ring trans-autophosphorylate one another at Thr286, locking CaMKII into an autonomous active state even after Ca²⁺ falls. Activated CaMKII (1) phosphorylates Ser831 on the GluA1 subunit of AMPA receptors (increasing single-channel conductance γ from ~9 pS to ~15–28 pS), (2) phosphorylates transmembrane AMPA receptor regulatory proteins (TARPs / stargazin) so laterally diffusing extrasynaptic AMPA receptors are captured and immobilized by PSD-95 scaffolding slots in the postsynaptic density, and (3) activates small Rho GTPases (Rac1, Cdc42) to polymerize branched actin filaments, physically enlarging the dendritic spine head by 50–200%.

The BCM Calcium-Threshold Sliding Rule for LTP vs. LTD

Why does high-frequency stimulation (100 Hz) strengthen a synapse (LTP), whereas prolonged low-frequency stimulation (1 Hz for 10–15 minutes) weakens the exact same synapse (Long-Term Depression, LTD)? As formulated theoretically by Elie Bienenstock, Leon Cooper, and Paul Munro (the BCM theory, 1982) and verified biochemically by John Lisman and Mark Bear, the direction of synaptic modification is governed by the amplitude and kinetics of the postsynaptic NMDAR Ca²⁺ signal:

• High, Brief [Ca²⁺] Surge (>5–10 μM during 100 Hz tetanus or pre-before-post spikes): Activates lower-affinity protein kinases (CaMKII, K_d ≈ 1–2 μM Ca²⁺/CaM, and PKC), driving AMPAR phosphorylation, exocytic insertion, and LTP.
• Modest, Prolonged [Ca²⁺] Elevation (~0.5–1.5 μM during 1 Hz stimulation or post-before-pre spikes): Fails to reach the threshold for CaMKII, but selectively activates the high-affinity Ca²⁺/calmodulin-dependent protein phosphatase calcineurin (PP-2B, K_d ≈ 0.1 μM), which dephosphorylates Inhibitor-1 to activate Protein Phosphatase 1 (PP-1). PP-1 dephosphorylates Ser845 on GluA1, triggering clathrin- and dynamin-mediated endocytosis of AMPA receptors out of the synapse, actin depolymerization via cofilin, dendritic spine shrinkage, and LTD.

Millisecond Causal Timing (STDP)

Discovered in cortical and hippocampal pyramidal neurons by Henry Markram, Guo-qiang Bi, and Mu-ming Poo (1997–1998), Spike-Timing-Dependent Plasticity (STDP) reveals that synapses measure the exact millisecond order between a single presynaptic spike (t_pre) and a back-propagating postsynaptic action potential (t_post), Δ*t* = t_post − t_pre:

• Causal Pre-Before-Post Timing (0 < Δ*t* < +20 ms): When the presynaptic EPSP arrives 5 to 20 milliseconds before the postsynaptic neuron fires an action potential (meaning the presynaptic input helped cause the postsynaptic spike), the back-propagating dendritic spike arrives while glutamate is still bound to NMDA receptors, expelling Mg²⁺ to produce a supralinear Ca²⁺ spike and robust LTP.
• Anti-Causal Post-Before-Pre Timing (−40 ms < Δ*t* < 0): When the presynaptic input arrives after the postsynaptic neuron has already fired (during the afterhyperpolarization and Ca_V channel inactivation), NMDAR Ca²⁺ influx is blunted into the calcineurin window, inducing LTD.

Why Pure Hebbian Networks Explode Without Homeostatic Synaptic Scaling

Because Hebbian LTP is a positive feedback loop—strengthening a synapse makes it more likely to drive postsynaptic spikes, which triggers even more LTP—any neural network governed solely by Hebbian LTP rapidly suffers runaway epileptic excitation (or complete silencing under LTD) while erasing previously stored relative weight patterns.

As discovered by Gina Turrigiano and colleagues (1998), neurons stabilize their global firing rate around a cell-type-specific homeostatic set-point (e.g., ~1–5 Hz) via homeostatic synaptic scaling. Over 24 to 48 hours of chronic network hyperactivity or sensory deprivation, the neuron senses its average somatic firing rate via nuclear Ca²⁺/CaMKIV signaling and multiplicatively scales all of its thousands of excitatory postsynaptic weights up or down by a single uniform factor α (*w*'_i = α w_i). Because every synapse is multiplied by the same constant α, total firing rate returns to the homeostatic set-point while strictly preserving the relative ratios (w₁ : w₂ : w₃) encoded by Hebbian learning!

How can a memory last 50 years if individual AMPA receptor and CaMKII protein molecules turn over every few days?

This question—originally framed by Francis Crick (1984)—is resolved by the transition from Early-LTP (E-LTP, lasting 1–3 hours) to Late-LTP (L-LTP) and synaptic tagging and capture (formulated by Uwe Frey and Richard Morris, 1997):

1. During strong LTP induction, local actin polymerization leaves a temporary "synaptic tag" at the potentiated dendritic spine while sending signaling cascades (cAMP–PKA, MAPK/ERK, and CaMKIV) to the cell nucleus to phosphorylate CREB at Ser133, activating transcription of immediate-early and plasticity-related genes (c-Fos, Zif268/Egr1, Arc, BDNF, PKMζ / PKCι/λ).
2. Newly transcribed mRNAs are transported into dendrites (or translated locally at polyribosomes at the base of the spine) and captured specifically by tagged spines, permanently remodeling the post-synaptic density (PSD-95, SHANK3, Homer) and trans-synaptic neurexin–neuroligin adhesion complexes into a larger structural configuration that self-templates across protein turnover.

Why does cerebellar motor learning rely on Long-Term Depression (LTD) rather than LTP?

In the cerebellar cortex (following the Marr–Albus–Ito theory of motor learning, verified by Masao Ito), Purkinje cells receive two distinct excitatory inputs: ~150,000 parallel fibers (carrying context and motor-command efference copy from granule cells) and a single powerful climbing fiber from the inferior olive (which fires a complex spike whenever a movement error occurs, such as missing a target). When a parallel fiber fires simultaneously with a climbing-fiber error signal, metabotropic glutamate receptor mGluR1 activation at the parallel fiber synapse (producing IP₃ and DAG) coincides with massive climbing-fiber voltage-gated Ca²⁺ influx. In cerebellar Purkinje cells, this high Ca²⁺ + DAG combination activates Protein Kinase Cα (PKCα), which phosphorylates Ser880 on GluA2, triggering AMPA receptor endocytosis and Long-Term Depression (PF-PC LTD). Because Purkinje cells are inhibitory (GABAergic), depressing the parallel-fiber excitation of a Purkinje cell disinhibits the deep cerebellar nuclei to correct the motor trajectory.

How did Aplysia californica reveal the molecular basis of short-term and long-term memory?

In work awarded the 2000 Nobel Prize in Physiology or Medicine, Eric Kandel studied the gill-withdrawal reflex of the marine sea hare Aplysia californica, whose central nervous system contains only ~20,000 large, reproducibly identifiable neurons. Kandel demonstrated that short-term sensitization (induced by a single tail shock) releases serotonin (5-HT) onto the presynaptic sensory neuron terminal, activating Gα_s → adenylyl cyclase → cAMP → PKA, which phosphorylates and closes S-type K⁺ channels—broadening the presynaptic action potential, increasing Ca²⁺ influx, and enhancing glutamate release for minutes without new protein synthesis. In contrast, long-term sensitization (induced by 4–5 spaced tail shocks) causes the PKA catalytic subunit and MAPK to translocate to the sensory neuron nucleus, phosphorylating CREB-1 (and relieving CREB-2 repression) to induce ubiquitin hydrolase and structural growth of twice as many presynaptic varicosities lasting weeks.

1. Bliss, T. V. P., & Lømo, T. (1973). "Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path." The Journal of Physiology, 232(2), 331–356. https://doi.org/10.1113/jphysiol.1973.sp010273
2. Bi, G.-Q., & Poo, M.-M. (1998). "Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type." The Journal of Neuroscience, 18(24), 10464–10472. https://doi.org/10.1523/JNEUROSCI.18-24-10464.1998
3. Turrigiano, G. G., Leslie, K. R., Desai, N. S., Rutherford, L. C., & Nelson, S. B. (1998). "Activity-dependent scaling of quantal amplitude in neocortical neurons." Nature, 391(6670), 892–896. https://doi.org/10.1038/36103
4. Kandel, E. R. (2001). "The molecular biology of memory storage: a dialogue between genes and synapses." Science, 294(5544), 1030–1038. https://doi.org/10.1126/science.1067020