Long-Term Potentiation

What long-term potentiation is

Push a memory into your brain and something physical changes. The most likely thing that changes is the strength of specific synapses — and long-term potentiation is the best-understood way that synapses get stronger. The recipe is almost absurdly simple: stimulate an excitatory synapse with a brief, intense burst — classically a 1-second train at 100 Hz, called a tetanus — and the synapse's response to a single test pulse afterward is permanently larger. Not 5% larger for a minute, but 50–200% larger, and it stays that way for hours in a brain slice and for days to over a year in a living animal.

The deep idea underneath LTP is older than the experiment. In 1949 the psychologist Donald Hebb proposed that when one neuron repeatedly helps fire another, "some growth process or metabolic change" strengthens the connection — usually paraphrased as "cells that fire together wire together." LTP is Hebb's postulate caught in the act. The whole mechanism is built around one molecule, the NMDA receptor, that physically refuses to open unless the presynaptic and postsynaptic neurons are active at the same moment. That coincidence requirement is what lets a brain associate a face with a name, a bell with food, or a place with danger.

How LTP works, step by step

The canonical site is the Schaffer-collateral synapse onto a CA1 pyramidal neuron in the hippocampus. Here is the molecular sequence:

1. Glutamate release. The presynaptic terminal fires and dumps glutamate into the synaptic cleft, which is about 20 nm wide. Glutamate binds two kinds of receptor on the postsynaptic spine: AMPA receptors and NMDA receptors.
2. AMPA receptors carry the everyday signal. AMPA receptors open immediately, pass Na+ and K+, and depolarize the spine. This is ordinary fast excitatory transmission — it happens every time, with or without LTP.
3. The NMDA receptor stays plugged — until coincidence. At the resting potential of about -70 mV, a Mg2+ ion sits in the NMDA channel pore and blocks it even though glutamate is bound. Only if the membrane is already depolarized — to roughly -40 mV or higher, which requires strong or repetitive input summing many AMPA responses — does the positive charge electrostatically eject the Mg2+. So the NMDA receptor opens only when "glutamate is here" AND "the cell is already firing" are true together.
4. Calcium floods the spine. Unlike AMPA receptors, the open NMDA receptor is highly permeable to Ca2+. A sharp pulse of Ca2+ enters the spine head, raising local concentration from a resting ~100 nM to several micromolar within milliseconds. The spine's tiny volume (~0.01–0.1 femtoliter) means even a modest current produces a large concentration jump.
5. CaMKII flips into a self-sustaining state. A large, fast Ca2+ rise binds calmodulin, which activates calcium/calmodulin-dependent protein kinase II (CaMKII). CaMKII then autophosphorylates at threonine-286, locking it into an active conformation that persists even after Ca2+ falls back to baseline — a molecular memory switch.
6. More AMPA receptors, bigger response. Active CaMKII phosphorylates the GluA1 AMPA-receptor subunit (at serine-831), raising single-channel conductance, and it drives exocytosis and lateral diffusion of additional AMPA receptors into the postsynaptic density. The same glutamate pulse now produces a much larger excitatory postsynaptic current. This is early-LTP, and it needs no new genes.
7. The spine grows; new proteins lock it in. Over minutes to hours the spine head enlarges, actin filaments reorganize, and the postsynaptic density expands. For late-LTP, signaling through PKA and MAPK reaches the nucleus and activates the transcription factor CREB, triggering new mRNA and protein synthesis that structurally consolidate the change for the long haul.

The molecular cast and the conditions

• Glutamate. The main excitatory neurotransmitter in the vertebrate brain and the agonist for both receptor types here. The NMDA receptor also needs a co-agonist — glycine or D-serine — bound at a separate site to open.
• AMPA receptors. Fast ionotropic glutamate receptors (subunits GluA1–GluA4) that do the everyday work of excitatory transmission. The number and conductance of AMPA receptors in the postsynaptic membrane is essentially the synapse's "weight," and LTP works by increasing it.
• NMDA receptors. The coincidence detector and Ca2+ gateway (subunits GluN1 plus GluN2A/GluN2B). They trigger LTP but carry little of the steady-state current. The competitive antagonist AP5 (APV) blocks them and blocks LTP — the single most important pharmacological proof of the mechanism.
• Mg2+ block. The voltage-dependent magnesium plug is the physical basis of coincidence detection. Remove extracellular Mg2+ in a dish and NMDA receptors open at rest, defeating the requirement.
• CaMKII. Strikingly abundant — roughly 1–2% of total protein in the forebrain and a dominant component of the postsynaptic density; its T286 autophosphorylation is the leading candidate for a persistent molecular switch.
• CREB and new proteins. The transcription factor whose activation is necessary for converting early-LTP into the protein-synthesis-dependent late phase, and for converting short-term memory into long-term memory.
• Conditions to induce it. Classic high-frequency stimulation is a 100 Hz, 1 s tetanus; theta-burst stimulation mimicking natural hippocampal rhythms works at lower intensity; and spike-timing-dependent plasticity shows that the presynaptic spike must precede the postsynaptic spike by a few to tens of milliseconds for potentiation — reverse that order and you get depression.

LTP vs long-term depression

Property	Long-term potentiation (LTP)	Long-term depression (LTD)
Direction of change	Synapse strengthens	Synapse weakens
Induction stimulus	High frequency (e.g. 100 Hz tetanus)	Low frequency (e.g. 1 Hz for ~15 min)
Ca2+ signal	Large, fast Ca2+ rise	Small, prolonged Ca2+ rise
Main enzyme activated	CaMKII (kinase)	Calcineurin / PP1 (phosphatases)
AMPA receptor fate	Inserted into membrane, conductance up	Removed by endocytosis, conductance down
Spike-timing rule	Pre before post (≈ +10 ms)	Post before pre (≈ −10 ms)
Spine structure	Head enlarges	Head shrinks, spine may be lost
Functional role	Stores associations / strengthens	Prunes, resets, sharpens, forgets

Sizes, times, and concentrations

Quantity	Typical value	Note
Resting membrane potential	≈ −70 mV	Where Mg2+ plugs the NMDA pore
Depolarization to relieve Mg2+ block	≈ −40 mV and above	Needs summed/strong input
Inducing tetanus	100 Hz for 1 s	Classic high-frequency protocol
Synaptic cleft width	≈ 20 nm	Glutamate diffuses across in ~µs
Resting spine Ca2+	≈ 100 nM	Rises to several µM during induction
Dendritic spine volume	≈ 0.01–0.1 fL	Tiny volume amplifies Ca2+ signal
Potentiation magnitude	+50% to +200% EPSP	Above pre-tetanus baseline
Early-LTP duration	1–3 hours	No new protein synthesis
Late-LTP / memory trace	Hours to > 1 year	Needs transcription & translation
STDP timing window	± ~20 ms	Pre→post potentiates, post→pre depresses

Where LTP shows up — learning, disease, and drugs

• Spatial memory in rodents. Blocking NMDA receptors with AP5 infused into the hippocampus prevents rats from learning the Morris water maze — a place they would otherwise remember — directly linking LTP machinery to a behavioral memory. Richard Morris demonstrated this in 1986.
• "Doogie" mice learn faster. Joe Tsien's mice engineered to overexpress the NMDA-receptor subunit NR2B (GluN2B) showed enhanced LTP and outperformed normal mice on novel-object recognition, fear conditioning, and maze learning — a gain-of-function argument that more LTP means more learning.
• Fear conditioning in the amygdala. LTP at thalamic and cortical inputs to the lateral amygdala underlies Pavlovian fear learning, the model system for studying PTSD and its potential reversal.
• Engram manipulation. Susumu Tonegawa's lab used optogenetics to tag the specific neurons activated during a memory and then artificially reactivate or strengthen them, even implanting a false memory and reversing apparent amnesia — strong evidence that strengthened synapses store the memory itself.
• Alzheimer's disease. Soluble amyloid-beta oligomers impair LTP and enhance LTD at hippocampal synapses well before plaques and cell death appear, offering a synaptic explanation for the earliest memory failures of Alzheimer's.
• Drugs and addiction. Drugs of abuse such as cocaine induce LTP-like strengthening at synapses onto dopamine neurons of the ventral tegmental area, hijacking the same machinery that normally encodes reward learning. NMDA antagonists like ketamine and memantine, and the proposed cognitive enhancers called ampakines that boost AMPA receptors, all act on this pathway.

Common misconceptions

• "LTP is the memory." LTP is a cellular mechanism that almost certainly contributes to memory, but a single synapse strengthened for an hour in a dish is not a memory. Real memories are distributed across cell assemblies and also involve changes in intrinsic excitability and inhibition. LTP is a building block, not the whole edifice.
• "The presynaptic neuron does all the work." The trigger for NMDA-dependent LTP is overwhelmingly postsynaptic: Ca2+ entry, CaMKII activation, and AMPA-receptor insertion happen on the receiving side. Presynaptic release probability can change too, but the classic switch is in the dendritic spine.
• "Any stimulation strengthens a synapse." Direction depends on the Ca2+ signal. Strong, brief input gives a big fast Ca2+ rise and LTP; weak, prolonged input gives a small slow rise and LTD. The exact same synapse can be strengthened or weakened depending on the pattern — this is the BCM/sliding-threshold idea.
• "AMPA and NMDA receptors are basically the same." They are both glutamate receptors but play opposite roles here. AMPA carries the moment-to-moment current; NMDA carries almost none at rest but acts as the Ca2+-gated trigger. Confusing them inverts the mechanism.
• "Late-LTP is just more of early-LTP." Early-LTP runs on existing proteins and survives a protein-synthesis blocker; late-LTP requires new transcription via CREB and is abolished by anisomycin. They are mechanistically distinct phases, which is why a memory can be disrupted hours later by blocking consolidation.
• "LTP lasts forever once induced." In a brain slice LTP decays over hours as the slice deteriorates; even in vivo it requires maintenance and reconsolidation. Some forms last over a year, but persistence is an active, ongoing process, not a one-time permanent stamp.

Famous experiments

• Bliss & Lømo 1973 — the discovery. Terje Lømo first saw potentiation in 1966 in the anesthetized rabbit dentate gyrus; with Tim Bliss he published the definitive paper in the Journal of Physiology in 1973, showing that brief high-frequency stimulation of the perforant path produced a strengthening of synaptic transmission lasting hours. This is the founding observation of LTP.
• Collingridge, Kehl & McLennan 1983 — the NMDA receptor. They showed that the NMDA-receptor antagonist AP5 blocks the induction of LTP at the Schaffer-collateral/CA1 synapse without affecting baseline transmission, pinning the trigger on the NMDA receptor and launching the modern molecular era of LTP research.
• Morris 1986 — LTP and behavior. Richard Morris showed that infusing AP5 into the rat hippocampus blocks spatial learning in the water maze, providing the first direct link between the NMDA-receptor-dependent LTP mechanism and an actual memory.
• Tsien (Tang et al.) 1999 — Doogie mice. Overexpressing the NR2B subunit enhanced LTP and improved learning and memory across multiple tasks, the landmark gain-of-function demonstration that more LTP capacity yields better memory.
• Tonegawa engram studies 2012–2014. Using optogenetics to label and reactivate the exact neurons engaged during learning, the Tonegawa lab artificially triggered, implanted, and reversed memories, providing causal evidence that the strengthened synapses of an engram store the memory.