The dentate gyrus, the structure at the center of this research, acts as the main entry point for information entering the hippocampus. It receives excitatory input from the entorhinal cortex through two anatomically distinct pathways known as the medial and lateral perforant paths, or MPP and LPP. These two inputs carry functionally different kinds of information: the medial path conveys spatial and contextual signals, while the lateral path delivers content-related information about objects and cues. Granule cells, the principal neurons of the dentate gyrus, integrate both streams across their elaborate dendritic trees, and how strongly each input drives these cells shapes everything the hippocampus subsequently computes.

For decades, scientists studying synaptic plasticity, the activity-dependent strengthening and weakening of connections thought to underlie memory, have focused overwhelmingly on AMPA receptors, which mediate the fast excitatory component of synaptic transmission. Long-term potentiation, or LTP, of AMPAR-mediated responses is the canonical cellular model of learning. NMDA receptors, by contrast, are famous for their role as coincidence detectors that trigger plasticity, but whether NMDAR-mediated transmission itself can undergo heterosynaptic plasticity, meaning changes at neighboring unstimulated synapses, had never been systematically investigated. The Einstein team, led by Kaoutsar Nasrallah, Pablo Castillo, and Alma Rodenas-Ruano, set out to fill that gap using a stimulation protocol designed to mimic what these neurons actually experience in the living brain.

The protocol, known as burst-timing-dependent plasticity, or BTDP, pairs presynaptic bursts of six pulses at 50 hertz with postsynaptic bursts of five action potentials at 100 hertz, separated by a precise 10 millisecond interval and repeated 100 times. This timing relationship reflects the burst firing patterns observed in medial entorhinal cortex neurons in vivo and the high-frequency bursts of back-propagating action potentials that granule cells can generate. When the researchers applied this protocol to MPP synapses in acute hippocampal slices from rats, they observed robust homosynaptic NMDAR-LTP, meaning potentiation at the stimulated synapses themselves. But then came the surprise: the same induction also produced significant NMDAR-LTP at LPP synapses on the very same granule cells, synapses that had received no direct stimulation at all.

Crucially, the team ruled out stimulation crosstalk between the two pathways. When they applied the identical burst protocol directly to LPP synapses, no homosynaptic or heterosynaptic NMDAR plasticity appeared anywhere, confirming that LPP inputs simply cannot express this form of plasticity on their own. The heterosynaptic effect also showed remarkable spatial precision. When the LPP stimulation electrode was placed on the opposite side of the recorded cell, at least 250 micrometers away, activating LPP inputs onto different dendritic branches, the heterosynaptic potentiation vanished. This suggests that the underlying signaling is confined to the specific dendritic branches where homosynaptic LTP was induced, consistent with the idea that single dendritic branches act as fundamental computational units of a neuron. The mechanism was also conserved across species, appearing in both rat and mouse tissue.

The plot thickened when the researchers examined AMPA receptors under the same conditions. The identical burst protocol that produced homosynaptic AMPAR-LTP at MPP synapses simultaneously induced heterosynaptic AMPAR-LTD, long-term depression, at the distal LPP synapses. In other words, the two receptor types followed completely opposite heterosynaptic rules: NMDAR transmission was strengthened at unstimulated LPP synapses while AMPAR transmission was weakened there. This divergence implies that the brain can dynamically shift the relative contribution of slow, calcium-permeable NMDAR signaling versus fast AMPAR drive at different cortical inputs, potentially favoring burst-only LPP signals in what the authors describe as winner-take-all dynamics within the dentate gyrus.

Mechanistically, the two forms of NMDAR-LTP split along a critical fault line involving intracellular calcium. Both homosynaptic and heterosynaptic potentiation required co-activation of NMDARs and type 5 metabotropic glutamate receptors, and both were abolished when postsynaptic calcium was chelated with BAPTA. But when the team depleted internal calcium stores with cyclopiazonic acid, an inhibitor of the sarco-endoplasmic reticulum calcium ATPase, only the heterosynaptic form was blocked; homosynaptic MPP-LTP proceeded normally. Pharmacological dissection pinpointed the culprit: ryanodine, which blocks ryanodine receptor-mediated calcium release from the endoplasmic reticulum, abolished heterosynaptic LTP, while heparin, an IP3 receptor blocker, had no effect. The smooth ER extends deep into distal dendrites and spines, providing a physical substrate for this long-range signaling that compensates for the dendritic attenuation of back-propagating action potentials in granule cells.

The expression mechanism proved equally distinctive. Previous work had shown that homosynaptic NMDAR-LTP at MPP synapses involves the synaptic recruitment of GluN2D-containing NMDARs, an unusual receptor subtype normally associated with extrasynaptic locations and low sensitivity to the voltage-dependent magnesium block. Using the selective GluN2C/GluN2D antagonists DQP-1105 and QNZ46, the researchers found that these drugs, which leave basal LPP transmission untouched, significantly reduced LPP NMDAR responses only after LTP induction, exactly what would be expected if previously extrasynaptic GluN2D-containing receptors had been recruited to the synapse. Genetic confirmation followed: deleting the Grin2d gene selectively from granule cells using a Cre-lox conditional knockout strategy abolished heterosynaptic LTP entirely, while control virus injections left it intact. These receptors likely form triheteromeric GluN1/2B/2D assemblies, whose intermediate magnesium sensitivity allows current flow even near resting membrane potentials, precisely the property needed at distal synapses where back-propagating spikes may be too weak to fully unblock conventional NMDARs.

The functional payoff was demonstrated in current-clamp recordings. After inducing heterosynaptic NMDAR-LTP, LPP burst stimulation evoked significantly more action potentials in granule cells, and this firing enhancement was abolished by the GluN2D antagonist DQP-1105. The potentiation therefore translates directly into increased synaptic drive from the lateral entorhinal cortex, dynamically regulating the output of the dentate gyrus. Given that granule cell Grin2d deletion has been shown to impair dentate gyrus-dependent forms of memory, the findings suggest that heterosynaptic NMDAR plasticity could be a genuine contributor to hippocampal learning rather than a laboratory curiosity. The authors note that NMDAR potentiation might also serve as a metaplastic switch, adjusting the induction threshold for future AMPAR-dependent plasticity at these inputs.

Limitations remain, including the fact that potential sex-specific differences were not fully addressed, since data from male and female animals were pooled after no major differences were observed. Future in vivo calcium imaging combined with genetic strategies will be needed to clarify how ER calcium release and GluN2D recruitment converge during behavior. Nevertheless, the study establishes a new principle: functionally distinct entorhinal inputs do not compete in isolation but actively cooperate, with activity at one pathway exerting powerful, long-lasting control over NMDAR transmission at another. As the brain’s memory gateway processes the streams of context and content that construct our episodic memories, hidden calcium signals traveling along dendritic branches may be quietly rebalancing the conversation between cortex and hippocampus, one synapse at a time.

Subject of Research: Heterosynaptic NMDA receptor plasticity in hippocampal dentate granule cells

Article Title: Heterosynaptic NMDA receptor plasticity in hippocampal dentate granule cells

Article References: Heterosynaptic NMDA receptor plasticity in hippocampal dentate granule cells. (n.d.). https://doi.org/10.3389/fnsyn.2026.1825949

Image Credits: AI Generated

DOI: 10.3389/fnsyn.2026.1825949

Keywords: Heterosynaptic, NMDA, receptor, plasticity, hippocampal, dentate, granule, cells, scientific research