A Vesicle Protein Helps Synapses Turn Down Release After Rapid Firing

Neurons adjust their activity to keep circuits stable. Several forms of this self-correction are known, but how a sending terminal detects activity and changes transmitter release has been poorly understood. ^[1]

The finding

A vesicle protein controls later release

Homeostatic plasticity is the set of mechanisms that keep neural circuits stable while learning-related changes reshape them. The paper's introduction describes two known forms: postsynaptic changes in glutamate receptors, and presynaptic increases in transmitter release. ^[1]

The authors report a presynaptic form that does not depend on postsynaptic receptors. Their claim is that high-frequency stimulation changes how a terminal releases transmitter afterward, and that VAMP7 does the work. VAMP7 is a vesicular SNARE protein, a member of the family of proteins best known for fusing vesicles with the cell membrane. ^[1]

Fast suppression with slow recovery

If the result holds, it gives a concrete molecular candidate for how a terminal senses recent activity: surface delivery of VAMP7 that quickly tightens calcium entry, with the authors noting that clearing it by endocytosis takes hours. That asymmetry could let a synapse respond fast to overactivity and recover slowly. ^[1]

The evidence is from mouse tissue and cultured neurons, mostly with whole-animal knockouts. The authors saw similar effects at four excitatory synapse types (CA1 stratum radiatum, perforant path, corticothalamic and cerebellar parallel fibre) and propose the mechanism acts throughout the nervous system; that is an extrapolation. No disease claim is made.

The evidence

Measuring calcium and release in mouse neurons

The experiments combine cultured mouse hippocampal neurons, which are nearly all glutamate-releasing, with acute hippocampal slices from mice aged 21 to 25 days. Slice recordings were made at 34 to 35 °C. ^[1], Methods

To mimic low activity, the authors silenced cultures for 24 hours with drugs that block glutamate receptors, or with tetrodotoxin. Several readouts followed. A fluorescent tag on a glutamate-vesicle protein lights up when a vesicle fuses, reporting exocytosis. A low-affinity calcium dye reports calcium entry into terminals. A glutamate sensor reports release at individual synapses. In slices, a drug that blocks NMDA receptors only when they are open (MK-801) acts as a ruler: the faster the response declines, the higher the release probability. ^[1], Figures 1 to 3

The causal tests used knockout mice and toxins. Mice lacking VAMP7 were compared with normal mice. Toxins cleaved specific SNARE proteins (botulinum toxin C cuts syntaxin, botulinum toxin A cuts SNAP25, tetanus toxin cuts VAMP2), and selective blockers separated P/Q-type, N-type and L-type calcium channels. ^[1], Figures 2 to 5

VAMP7 dependence across firing conditions

Silence raises calcium entry, and activity resets it. After 24 hours of silencing, terminals showed a much larger calcium response to 5 Hz and 25 Hz trains, and more synapses released glutamate at 0.2 Hz. Two to three 25 Hz bursts brought calcium entry in silenced neurons back to baseline. ^[1], Figure 1

High-frequency firing lowers later release, and VAMP7 is required. In slices, 100 action potentials at 50 Hz, compared with 5 Hz, increased facilitation and reduced later depression within the train in normal mice. In Vamp7-knockout slices, depression within the train was larger and the 50 Hz train did not change it. Loss of VAMP7 raised initial release probability, and raising extracellular calcium to 4 mM masked that difference. ^[1], Figures 2 and 3; Discussion

VAMP7 acts on P/Q-type channels. Cultured knockout neurons had higher calcium transients at every frequency tested, 1 to 50 Hz. The P/Q blocker ω-agatoxin IVA removed most of the excess. The N-type blocker had a modest effect that was similar in both genotypes. The L-type blocker nimodipine, however, substantially reduced calcium entry in knockout neurons stimulated at 25 Hz, and the authors report that VAMP7 also inhibits L-type channels at high frequency. So P/Q-type channels are the main target, not the only one. ^[1], Figure 3

VAMP2 does not do this, and the linker matters. The authors conclude that VAMP7, but not the canonical VAMP2, inhibits calcium entry. A linker segment of VAMP7 is required for binding and inhibiting P/Q channels, and placing that linker into VAMP2 gave VAMP2 the inhibition. ^[1], Figures 3 and 4

Syntaxin is required for the inhibition; SNAP25 is not. Cleaving syntaxin removed VAMP7's effect on calcium entry, and the authors state that for inhibition of P/Q-type entry by surface VAMP7, the role needs syntaxin rather than SNAP25. SNAP25 matters at a different step: cleaving it mostly prevents the burst-induced suppression after silencing, because it blocks the vesicle exocytosis that delivers VAMP7, and SNAP25 also regulates calcium entry in the knockout. VAMP7 and syntaxin co-precipitated with the P/Q channel from synaptosomes, and knockout neurons showed less colocalization of syntaxin 1 with the channel. ^[1], Figure 4; Discussion

Activity delivers VAMP7 to the surface. Silenced neurons had less VAMP7 on the cell surface. Repeated 50 Hz bursts progressively reduced responses in normal neurons, and loss of VAMP7 greatly reduced that decline, in cultures and in slices. This is a different comparison from the within-train depression above: knockout slices showed more depression inside a single train, yet less of the gradual decline across successive bursts. In hippocampus, loss of VAMP7 raised the number of FOS-positive neurons two- to fivefold. FOS is a marker of recently active neurons. ^[1], Figure 5

Temperature matters for one slice measurement. At Schaffer-collateral synapses, the genotype effect on 5 Hz and 25 Hz trains and on paired-pulse ratio was absent at room temperature and seen at 34 to 35 °C, which the authors say accounts for an earlier null result. The direct comparison was made for this synapse; a repeat for the other readouts is not reported in the main text. ^[1], Methods; Discussion

Conditions to keep in mind: the sampling unit varies. Culture experiments report fields or cells from independent preparations, while slice experiments report cells and mice, and at least one Extended Data panel reports n = 3 mice. The authors state that whole-cell recording analysis was not blind to genotype or condition, and that no sex-specific results are reported. ^[1], Methods

Future directions

The authors say it is unclear how the mechanism affects information transfer. They suggest that inactive synapses, with higher release probability, may favor timing information, while strong stimulation shifts that balance. ^[1]

Does the effect show up in behavior or learning? The paper reports no behavioral tests. The authors suggest VAMP7 may help maintain sparse connectivity, which is a hypothesis. Deleting VAMP7 only in selected cell types would help separate a presynaptic effect from broader circuit changes, since the knockout here affects every cell.

How direct the contact between VAMP7 and the channel is remains open. The authors suggest the linker may interact directly with the channel; co-precipitation alone does not establish that. The cultures were almost entirely glutamatergic, so whether inhibitory synapses use the same brake is untested in this study.