One Synapse, Many Memories: How the Brain Builds LTP in Different Ways

For half a century, long-term potentiation, or LTP, has been the leading candidate for the physical basis of memory. The idea is seductively simple: when two neurons fire together in the right pattern, the connection between them strengthens rapidly, holds that strength for months, and affects only the synapses that were active. That combination of speed, persistence, and synapse specificity is exactly what a memory system would need. Yet behind this tidy story lies a surprisingly messy reality. A new review published in Frontiers in Synaptic Neuroscience argues that much of the decades-long confusion over how LTP works stems from a hidden assumption that there is only one kind of potentiation. In fact, the evidence now points to multiple mechanistically distinct forms of LTP, even within a single brain region, and even at the same population of synapses.

The review, written by researchers who have spent decades dissecting the hippocampus, focuses on two of the most intensively studied connections in the forebrain: the Schaffer-commissural (SC) pathway running from field CA3 to CA1, and the perforant path projections from entorhinal cortex into the dentate gyrus. These two connections represent the major output and input stations of the hippocampal circuit, and comparing them reveals a striking diversity of plasticity mechanisms. The authors argue that many of the classic disputes in the field, particularly the long-running argument over whether LTP is expressed presynaptically or postsynaptically, can be explained by the simple fact that different investigators were studying different phenomena.

The story begins with theta burst stimulation, or TBS, a pattern of short high-frequency bursts separated by the period of the brain’s theta rhythm. This pattern mimics what cortical neurons naturally do during learning, and it turns out to be nearly optimal for potentiating the SC projection. When a conventional train of ten theta bursts is delivered, the potentiation develops within seconds, is restricted to the synapses directly activated, and can persist for hours in slices and weeks in living animals. Crucially, it also has a memory-like consolidation window lasting roughly ten to thirty minutes, during which the change becomes progressively harder to reverse.

For the SC system, the induction of this potentiation is postsynaptic. The theta burst interval allows inhibitory interneuron terminals to engage GABA-B autoreceptors, transiently reducing shunting inhibition. This means the second burst in a train produces a much larger depolarization than the first, activating voltage-sensitive NMDA receptors and triggering calcium influx into the spine. The expression of potentiation then involves an expansion of the dendritic spine, the postsynaptic density, and the associated pool of AMPA receptors, the molecular workhorses of fast excitatory transmission. Super-resolution microscopy has revealed that LTP may also increase the alignment of subsynaptic domains into nanocolumns, placing points of neurotransmitter release directly adjacent to concentrations of AMPA receptors.

But induction is only the first step. The newly potentiated state must be stabilized, and this is where the actin cytoskeleton comes in. Within minutes of TBS, a small subset of spines shows a dramatic increase in filamentous actin, driven by a cascade of small GTPases including RhoA, Rac, Cdc42, and Ras. RhoA signaling through ROCK, LIMK, and cofilin triggers actin polymerization, while Rac and Ras signaling through Arp2/3 stabilizes the branched network. Integrin adhesion receptors and the neurotrophin BDNF acting through its TrkB receptor are both critical for this process, and blocking either one prevents LTP consolidation. Remarkably, the beta-1 integrins exhibit an activation-deactivation cycle: after one TBS train, there is a one-hour refractory period during which a second train cannot activate them or induce further potentiation.

That refractory period may explain one of the most intriguing findings in the field: spaced stimulation effects that mirror the spaced-trials advantage in learning. A second TBS train applied one hour after the first doubles the magnitude of SC LTP, whereas the same train applied ten, twenty, or forty minutes after the first has no additional effect. Single-spine experiments using glutamate uncaging confirmed that individual spines are refractory to further potentiation for roughly an hour, with the refractory state associated with reduced CaMKII activation and an absence of membrane GluA2 insertion. Blocking beta-1 integrins at the one-hour mark erases the initial potentiation, revealing a second, delayed consolidation stage that also appears to operate during memory encoding in living animals.

Recent work has forced a further revision to the standard model. The prevailing view held that calcium influx through NMDA receptors was essential for setting the actin remodeling machinery in motion. But new experiments show that blocking NMDAR-mediated ion flux with the channel blocker MK-801 eliminates LTP without interfering with TBS-induced actin polymerization or spine enlargement. This points to ion flux-independent, or metabotropic, NMDA receptor signaling, largely mediated by the GluN2B subunit, as the critical driver of cytoskeletal consolidation in males. Pharmacological blockade of GluN2B disrupts actin signaling, spine F-actin increases, and LTP consolidation in male hippocampal slices, and also blocks CA1-dependent episodic cue-location memory in adult males, demonstrating that these metabotropic functions matter in vivo.

The most surprising twist involves sex. Females do not simply add an estrogen step to the male mechanism; they replace it. Inhibiting P450 aromatase, the rate-limiting enzyme in estrogen synthesis, abolishes TBS-induced CA1 LTP in female mice but has no effect in males. The estrogen effect is mediated by membrane-associated estrogen receptor alpha, which drives the same RhoA-dependent actin polymerization cascade that GluN2B drives in males. Blocking ER alpha disrupts potentiation in females only, and transgenic mice with mutations preventing ER alpha trafficking to the membrane show severely impaired LTP in females but not males. These differences have behavioral consequences: near-threshold stimulation elicits robust LTP in males but not females, and males outperform females on episodic

Subject of Research: Mechanistic diversity of hippocampal long-term potentiation across brain regions and sexes

Article Title: The how and why for multiple forms of hippocampal LTP

Article References: Gall, C. M., Le, A. A., & Lynch, G. (2026). The how and why for multiple forms of hippocampal LTP. Frontiers in Synaptic Neuroscience, 18, Article 1855059. https://doi.org/10.3389/fnsyn.2026.1855059

Image Credits: AI Generated

DOI: 10.3389/fnsyn.2026.1855059

Keywords: long-term potentiation, hippocampus, synaptic plasticity, theta burst stimulation, NMDA receptor, estrogen, actin cytoskeleton, endocannabinoid, dentate gyrus, memory encoding, sex differences, perforant path