Long-Term Potentiation

Long-term potentiation (LTP) is a persistent, activity-dependent strengthening of synaptic transmission between two neurons triggered by brief, high-frequency or coincident patterns of synaptic stimulation. First discovered in the rabbit hippocampus by Terje Lømo in 1966 and published by Timothy Bliss and Terje Lømo in 1973 at the perforant-path-to-dentate-gyrus synapse, LTP provides the primary cellular and molecular mechanism for Hebbian synaptic plasticity—the postulate formulated by Donald Hebb in 1949 that when a presynaptic neuron repeatedly participates in firing a postsynaptic neuron, the efficacy of the synapse connecting them is durably increased ("neurons that fire together, wire together").

At its canonical model synapse—the Schaffer collateral–CA1 pyramidal cell glutamatergic synapse in the mammalian hippocampus—LTP is induced by the biophysical coincidence-detection property of the NMDA (N-methyl-D-aspartate) glutamate receptor and expressed through CaMKII (Ca²⁺/calmodulin-dependent protein kinase II)-driven insertion and trapping of AMPA receptors within the postsynaptic density (PSD), coupled with actin-driven structural enlargement of the dendritic spine head. Conversely, prolonged low-frequency stimulation (1 Hz for 10–15 minutes) induces the inverse process, long-term depression (LTD), allowing neural circuits to bidirectionally store and update information.

Classical NMDA-receptor-dependent LTP at hippocampal CA1 synapses exhibits three defining computational properties that map directly onto the biophysics of dendritic spine signaling:

1. Input Specificity: When LTP is induced at one set of Schaffer collateral synapses on a CA1 pyramidal neuron, adjacent unstimulated synapses on the same dendrite (<1–2 μm away) are not potentiated. Input specificity arises because the narrow dendritic spine neck (~0.1–0.2 μm wide, 0.5–1.0 μm long) compartmentalizes the postsynaptic Ca²⁺ transient specifically within the activated spine head.
2. Cooperativity: A single weak presynaptic fiber activation cannot trigger LTP because the resulting small AMPA-mediated depolarization fails to expel the extracellular Mg²⁺ block from the NMDA receptor pore; simultaneous activation of a threshold number of converging presynaptic fibers is required to sum their depolarizations above −30 to −20 mV.
3. Associativity (Spike-Timing-Dependent Plasticity, STDP): A weak synaptic input that is insufficient to induce LTP on its own becomes strongly potentiated if it is paired within a +5 to +20 millisecond window with either a strong converging synaptic input or a back-propagating action potential (bAP) sweeping into the dendritic tree from the postsynaptic soma (whereas if the postsynaptic spike precedes the presynaptic EPSP by −5 to −30 ms, spike-timing-dependent LTD is induced).

1. The NMDA Receptor Voltage-Dependent Mg²⁺ Block

The heterotetrameric NMDA receptor (composed of two obligate GluN1 subunits binding the co-agonist glycine or D-serine and two GluN2A or GluN2B subunits binding L-glutamate) is markedly permeable to Ca²⁺ as well as Na⁺ and K⁺. However, as discovered by Philippe Ascher and Linda Nowak in 1984, at the resting membrane potential (−70 mV), hydrated extracellular Mg²⁺ ions (~1.2 mM) are drawn electrostatically into the outer vestibule of the NMDA channel pore and bind tightly to an asparagine ring (the N-site) at the tip of the M2 pore loop, mechanically plugging the channel even when glutamate is bound.

Only when two events coincide in time and space—(1) presynaptic glutamate release and (2) strong postsynaptic depolarization (driven by summated AMPA receptor currents or a back-propagating action potential)—is the positively charged Mg²⁺ ion electrostatically repelled out of the pore, permitting a concentrated pulse of Ca²⁺ to enter the dendritic spine.

2. The CaMKII Molecular Memory Switch (Thr286 Autophosphorylation)

Within the postsynaptic density, CaMKII (Ca²⁺/calmodulin-dependent protein kinase II) accounts for an extraordinary 1–2% of total brain protein. The CaMKII holoenzyme is assembled as a 12-subunit (dodecameric) double-ring rosette in which each subunit's catalytic kinase domain is held inactive at rest by an autoinhibitory regulatory segment:

1. When high-amplitude NMDA-receptor Ca²⁺ influx (>5–10 μM) saturates calmodulin (CaM, binding 4 Ca²⁺ ions), Ca²⁺/CaM binds the regulatory segments of adjacent subunits within the CaMKII ring, exposing their catalytic domains.
2. When two neighboring subunits in the same hexameric ring bind Ca²⁺/CaM simultaneously, one subunit phosphorylates its neighbor at Threonine-286 (Thr286) in the regulatory domain.
3. Thr286 autophosphorylation sterically prevents the autoinhibitory domain from folding back over the catalytic site even after intracellular Ca²⁺ falls back to baseline (100 nM) and calmodulin dissociates, converting CaMKII into an autonomous, Ca²⁺-independent active state that translocates onto the cytoplasmic C-tail of the NMDA receptor GluN2B subunit inside the PSD. Knock-in mice carrying a non-phosphorylatable CaMKIIα^T286A point mutation (generated by Alcino Silva and colleagues) show pronounced deficits in hippocampal CA1 LTP and spatial learning in the Morris water maze.

LTP progresses through two biochemically distinct temporal phases:

If synapses could only strengthen via LTP, neural networks would rapidly saturate all synaptic weights and descend into runaway epileptic excitation. Stability requires Long-Term Depression (LTD)—a durable reduction in synaptic strength induced at CA1 synapses by prolonged low-frequency stimulation (1 Hz for 900 pulses):

• Differential Ca²⁺ Affinity of Calcineurin vs. CaMKII: Low-frequency stimulation produces a modest, prolonged postsynaptic Ca²⁺ elevation (~0.5–1.5 μM) through partially unblocked NMDA receptors. Because the Ca²⁺/calmodulin-dependent serine/threonine protein phosphatase 2B (Calcineurin / PP2B) has a ~10-fold higher affinity for Ca²⁺/CaM (K_D ≈ 0.1–1 nM) than CaMKII, modest Ca²⁺ influx selectively activates Calcineurin → Inhibitor-1 dephosphorylation → Protein Phosphatase 1 (PP1).
• AMPA Receptor Dephosphorylation and Endocytosis: PP1 and Calcineurin dephosphorylate GluA1 (at Ser845) and TARP/Stargazin, untethering AMPA receptors from PSD-95 and triggering clathrin/dynamin-dependent endocytosis of AMPA receptors alongside cofilin-mediated F-actin severing and dendritic spine shrinkage.

(Note: A distinct, non-NMDA-receptor-dependent form of presynaptic LTP occurs at hippocampal mossy fiber–CA3 synapses and cerebellar parallel fibers, where presynaptic Ca²⁺ influx activates adenylyl cyclase 1/8 → cAMP → PKA → Rab3a / RIM1α, persistently increasing the probability of presynaptic vesicle release).

• Bliss, T. V., & 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. Google Scholar
• Malenka, R. C., & Bear, M. F. (2004). LTP and LTD: an embarrassment of riches. Neuron, 44(1), 5–21. Google Scholar
• Lisman, J., Yasuda, R., & Raghavachari, S. (2012). Mechanisms of CaMKII action in long-term potentiation. Nature Reviews Neuroscience, 13(3), 169–182. Google Scholar
• Nicoll, R. A. (2017). A brief history of long-term potentiation. Neuron, 93(2), 281–290. Google Scholar