The Synapse: What Crosses the Cleft, and What Decides Whether Anything Does

01The handover

The chemical synapse converts an electrical signal into a chemical one and back

A chemical synapse is built from three parts. The presynaptic terminal bouton stores and releases transmitter, the cleft separates the two cells, and the postsynaptic membrane carries the receptors that read what arrives. Exocytosis pushes the message into chemistry, and the receptors pull it back into current.

What crosses is a courier. Glutamate and GABA carry most of the fast traffic in the mammalian brain, and peptides released alongside them adjust the terms of the exchange. Whatever the molecule, the receiving cell answers in voltage, and that voltage fades as it spreads toward the trigger zone.

The handover costs time. At rat cerebellar synapses recorded at physiological temperature, postsynaptic responses begin 150 microseconds after the presynaptic action potential starts Sabatini 1996. Calcium-driven vesicle fusion lags calcium influx by 60 microseconds, and rapid calcium channel kinetics let calcium enter during the upstroke of the spike itself.

Delay is the price of a junction that can be adjusted

Those microseconds buy a chemical step, and the chemical step is what makes the junction adjustable. At a glutamatergic synapse the step runs in order. Transmitter binding opens the AMPA-kainate channel, sodium enters, and the local depolarization expels the magnesium plug from the N-methyl-D-aspartate receptor waiting beside it. Sodium and calcium then flow through the NMDA channel, and that calcium works as a second messenger rather than as charge alone.

The NMDA receptor therefore opens only when sender and receiver agree, which is coincidence detection at the molecular level. The measured width of the cleft and the receptor families that read what crosses it belong to The Neurotransmitters. The running balance between glutamate and GABA belongs to GABA, Glutamate and the Balance.

One synapse rarely decides anything alone. A neuron needs roughly 15 millivolts of depolarization from rest before it fires, and a single excitatory event delivers a fraction of that. Events arriving in quick succession stack by temporal summation, events arriving on separate dendrites add by spatial summation, and the axon hillock reads the total against threshold. That arithmetic is the subject of The Neuron and the Central Integrative State.

02Findings

What the research shows

Measured values for release timing, vesicle stock, transmission odds, and spinal gating.

03Packets, not a stream

Transmitter leaves the terminal in packets of fixed size, and the count varies

Release at a synapse is quantal. Fatt and Katz recorded spontaneous subthreshold activity at motor nerve endings in 1952 Fatt 1952. Two years later del Castillo and Katz showed that the evoked end plate potential is assembled from those same units del Castillo 1954. The unit is a vesicle, and the evoked response is a count of them.

The stock behind each count is small. Three-dimensional reconstructions of excitatory synapses in rodent hippocampus and in culture put about 10 docked vesicles at each active zone and about 200 vesicles in the whole bouton Schikorski 1997. Average active zone area came in near 0.04 square micrometers, and most boutons carry a single zone.

Each packet is crowded in molecules and short in reach. Simulation fitted to receptor mapping at the rat calyx of Held between postnatal days 12 and 16 put about 7,000 glutamate molecules in one vesicle Budisantoso 2013. A single vesicle does not saturate the receptors facing it, so a second packet still adds to the reply.

A spike does not guarantee a packet

Calcium is the trigger, and it is not a generous one. At the rat calyx of Held, comparing quantal content with the presynaptic calcium current showed that more than 60 calcium channels open for each vesicle released Borst 1996. Dialyzing the terminal with 1 millimolar EGTA already cut phasic release, which places the calcium sensor some distance from the channel pore.

That sensor has a name and a measured sensitivity. A point mutation in synaptotagmin I halving calcium affinity, knocked into the mouse gene, halved the calcium sensitivity of release Fernandez-Chacon 2001. Spontaneous release and the size of the readily releasable pool did not move. Evoked release carries its own adjustable setting, separate from the stock it draws on.

04Release probability

Release probability differs between terminals fed by the same axon

Release probability is the chance that one arriving spike releases a packet at one release site. It is not a constant of the nervous system. Measured with an irreversible open channel blocker at excitatory synapses on cultured hippocampal neurons, it ranged from 0.09 to 0.54 across terminals arising from a single axon Rosenmund 1993. Most sat at the low end.

Paired recordings show what that spread costs. Between thick tufted layer 5 pyramidal neurons in developing rat neocortex, unitary excitatory potentials averaged 1.3 millivolts and ranged from 0.15 to 5.5 millivolts Markram 1997. Single presynaptic spikes failed to evoke any response between 0 and 73 percent of the time, with a mean failure rate of 14 percent.

Anatomy did not account for the spread. Each connection used four to eight contacts, averaging 5.5, and amplitude correlated only weakly with contact number or distance from the soma. The 20-fold range in efficacy between connections was attributed mainly to differences in transmitter release probability at the projecting cell.

The stock behind a terminal sets its odds

Release probability is not free-floating. Hypertonic solution applied to hippocampal synapses empties a defined pool, and evoked release and hypertonic release change together as that pool changes, which places release probability downstream of readily releasable pool size Rosenmund 1996. A terminal holding more vesicles docked speaks with more confidence.

This is a state, and it is held on the sending side. Two axons carrying identical spike trains deliver different messages when their terminals hold different pools. Nothing in the spike carries that information, because the spike is all or none. The difference belongs to the junction.

05History sets the filter

A synapse answers a train of spikes according to what it just finished doing

A synapse rewrites its own strength on the strength of its recent work, over seconds. Facilitation raises the second response and depression lowers it. Which one a terminal shows follows from its release probability, because a bouton that spends its docked stock on the first spike has little left for the second. This is short-term plasticity.

That setting decides which code survives the crossing. Paired recordings from neocortical pyramidal neurons combined with theory showed that the rate of synaptic depression follows release probability Tsodyks 1997. Where depression is slow, the postsynaptic cell reports presynaptic firing rate. Where depression is fast, it reports temporal coherence instead.

Depression also works as automatic gain control. Cortical neurons receive afferents firing from under 1 hertz to over 200 hertz. Modeling built on experimental measurements showed that short-term depression lets equal percentage rate changes on fast and slow afferents produce equal postsynaptic responses Abbott 1997. The adjustment is input-specific, so a quiet afferent is not buried by a loud neighbor.

The filter setting is visible in the electron microscope

Serial reconstruction in mouse brain found olfactory cortical synapses about 2.5 times larger than their hippocampal counterparts Schikorski 1999. Layer Ia synapses held fewer vesicles than layer Ib but matched the smaller hippocampal synapses in docked vesicle number. Layer Ia facilitates and layer Ib does not.

Two populations in one cortical region therefore filter their inputs in opposite directions, and the difference is structural. A synapse is a tuned device rather than a wire, and its tuning is readable in the count of vesicles waiting at the membrane.

The Unified Model of Tone makes its claim here, in the synapse's own units. Tone at a chemical synapse is release probability together with the docked stock standing behind it. That pair decides whether an arriving train is amplified, passed through, or filtered out before the next cell hears anything.

A synapse holding that pair inside its working range can facilitate when input is sparse and depress when input is heavy, which is what keeps a circuit legible across a 200-fold range of firing rates. A synapse pinned high empties on the first spike and reports nothing after it. A synapse pinned low drops the message. Both are illness at the scale of a junction, because a circuit that cannot move its filter cannot answer a body that keeps changing.

06Brakes on brakes

Disinhibition drives circuits by interrupting a brake that never stops running

Inhibition at a synapse is an active sculpting force, and disinhibition is its sharpest use. Spontaneously active neurons are held quiet by GABAergic cells. Silence those cells with still others of their kind and the target is released. A brake placed on a brake produces excitation with no excitatory transmitter anywhere in the chain.

The basal ganglia run on that arrangement. Output neurons of the internal globus pallidus and substantia nigra pars reticulata discharge tonically at 50 to 100 spikes per second in monkeys Elias 2007. Only 6 percent of those output cells were classified as pausers, against 56 percent of external pallidal cells. The brake on movement is close to permanently on.

Lifting it is how movement starts. The basal ganglia arouse executive motor centers through that disinhibitory mechanism, using their own output as a template for which motor elements to engage Chevalier 1990. The reticular system applies the same double-negative logic to motor control. The circuit is followed further in The Basal Ganglia.

Presynaptic inhibition edits a message before it is sent

The cleft can also be governed from upstream, on the sending side. GABAergic interneurons form axo-axonic synapses on the terminal arborizations of sensory fibers in the spinal cord, producing primary afferent depolarization and cutting transmitter release Rudomin 1999. Intraspinal sensory terminals are dynamic systems whose traffic central mechanisms can address to selected targets.

The editing runs branch by branch. Cable modeling with experimental verification put primary afferent depolarization near 10 millivolts on some distal branches of mammalian group I collaterals Lamotte d'Incamps 1999. The electrotonic structure of those collaterals kept it from spreading through the whole arbor. One fiber can be silenced at one target and heard clearly at another.

Descending pathways from the brain reach these same terminals, so what a segment sends upward is already an edited report. A change in what a person feels can come from a brake lifting as readily as from a stimulus growing. The receptor feeding these gated terminals is the subject of Muscle Spindles and Proprioception.

07Electrical junctions

Electrical synapses trade the ability to be modulated for the ability to synchronize

Not every synapse uses a chemical courier. Furshpan and Potter described direct electrical transmission at the giant motor synapses of the crayfish in 1959 Furshpan 1959. Electrical synapses are gap junctions, also called nexuses, between the dendrites or somas of contiguous neurons, built from channels roughly 1.5 nanometers wide.

Cytoplasm runs continuous through those channels, so no transmitter is involved and there is no synaptic delay. Voltage changes pass straight from one cell into the next. Coupling that tight cannot be weighted or gated, so an electrical synapse cannot learn. Its work is making neighbors fire together.

In mammalian cortex the wiring is selective. Paired recordings found frequent electrical coupling among fast-spiking GABAergic interneurons, none among pyramidal neurons, and none between fast-spiking cells and other cortical types Galarreta 1999. Coupling drove the linked interneurons toward synchronous spiking.

One protein carries the synchrony

Mice carrying a reporter in place of connexin 36 lost almost all electrical synapses between cortical interneurons, and the rhythmic inhibitory activity that survived was weaker and more spatially restricted Deans 2001. A single gap junction protein sets how far a cortical rhythm can spread through the tissue.

The reach is short and strictly local. In rat somatosensory cortex layer 4, coupling probability and coupling coefficient both fall to zero beyond 200 micrometers of intersomatic distance, and each interneuron holds measurable coupling with 20 to 50 others Amitai 2002. The two synaptic designs split the labor between fidelity in timing and the capacity to compute.

08Spinal input at the gate

An adjustment reaches this machinery as a volley into a gate that is already set

The gate on spinal afferent terminals moves with posture alone. Going from lying supine to standing lowered the soleus H-reflex in 10 adults from 68.7 to 54.8 percent of the maximal motor response. The conditioned reflex fell with it, from 30.7 to 17.5 percent Mynark 1997. The protocol reads presynaptic inhibition of the Ia pathway indirectly, and both measures fell on standing.

The same input can reverse its sign with the task. Stimulating a cutaneous nerve in the opposite foot facilitated the soleus H-reflex during standing and suppressed it during the early stance phase of walking Suzuki 2014. Background soleus activity was matched between the two conditions. Nothing about the stimulus changed, and the reply inverted.

What the manipulation studies measure

Spinal manipulation is measured against this same pathway. A side-posture thrust at L5 to S1 attenuated the reflex by 28.4 percent, against 15.3 percent for side-posture positioning alone Dishman 2005. Assisted manipulation attenuated it 18.2 percent against 9.5 percent unassisted, and joint preload without thrust produced 8.5 and 7.5 percent.

The timing is tight. Across 66 asymptomatic adults and 45 patients with subacute low back pain, inhibition was greatest 10 seconds after the thrust, and the effect ran similar in both groups Dishman 2018. A thrust moves a gate setting on the timescale of neural signaling rather than tissue repair.

Null results belong in the same account. In 75 people with previous knee injury and current quadriceps inhibition, lumbopelvic manipulation and patellar mobilization changed neither the H-reflex nor presynaptic nor postsynaptic excitability across 90 minutes Grindstaff 2014. In this model that input reached a gate the knee injury already held down, and found nothing left to move.

This is the ground the practice stands on. We are the practice organized around the nervous system, and an adjustment is a mechanical event whose first measurable consequence is a change in presynaptic gating. The model stakes something specific on that reading. Presynaptic inhibition at afferent terminals moving with reflex gain and with the postural state that sets both is what confirms this account.

Instrument values, outcome evidence, and the history of these measures sit on the neurophysiology page. The receptor densities behind spinal afferent traffic sit on input quality.

09Tone

How this system expresses tone

Read one junction closely and the foundations of tone are already written in its numbers.

The rest of tone is legible at the same terminal. Oscillation: whether a burst is passed as a firing rate or as a coincidence depends on how fast that synapse depresses. Prediction: layer Ia olfactory synapses facilitate and layer Ib synapses do not, so each population is already committed to the input pattern it expects. Load: about 200 vesicles per bouton and about 10 docked at the zone put a hard ceiling on what a single message can spend. Constraint: more than 60 calcium channels must open per vesicle released, so calcium entry limits what the terminal is able to say. Set point: standing lowered the soleus reflex from 68.7 to 54.8 percent of maximum, a gate reset by posture with no change in the afferent. Time course: fusion lags calcium by 60 microseconds, facilitation and depression run over seconds, and a manipulated gate reads deepest at 10 seconds.

10Across the library

How this page relates to the rest of the library

Where this junction is read across the rest of the library.

11Frequently asked

Questions about this topic

12The sources

References

Sources: primary literature, linked inline.