What Are Synapses?

Synapses are specialized junctions between neurons, or between a neuron and an effector cell, across which nerve impulses are transmitted by chemical or electrical signaling. The chemical synapse, by far the more prevalent type in the vertebrate nervous system, converts an electrical signal (the action potential) in the presynaptic neuron into a chemical message carried by neurotransmitter molecules, and then reconverts that chemical message back into an electrical signal in the postsynaptic cell. This transduction mechanism allows synapses to amplify, filter, integrate, and modulate neural information in ways that a purely electrical junction cannot. The synapse is, consequently, the fundamental site of computation and plasticity in the nervous system.

Synaptic Structure and Chemical Transmission

A chemical synapse consists of three anatomical elements: the presynaptic terminal (bouton), the synaptic cleft, and the postsynaptic density. The presynaptic terminal contains synaptic vesicles filled with neurotransmitter, along with the molecular machinery required for vesicle docking and fusion. When an action potential depolarizes the terminal membrane, voltage-gated calcium channels open, and the resulting calcium influx triggers synaptotagmin to facilitate fusion of vesicles with the plasma membrane through the SNARE complex. Neurotransmitter molecules diffuse across the 20 to 40 nanometer synaptic cleft and bind to receptor proteins in the postsynaptic density, producing either direct channel opening (ionotropic receptors) or second-messenger cascades (metabotropic receptors). The entire sequence from vesicle fusion to postsynaptic current occurs within milliseconds. A thorough description of these molecular events is provided in the NIH NCBI chapter on synaptic transmission from Basic Neurochemistry.

Synaptic Plasticity

The efficiency of synaptic transmission is not fixed; it changes as a function of neural activity history, a property called synaptic plasticity. Short-term plasticity operates over timescales of milliseconds to minutes through facilitation, in which residual presynaptic calcium from prior activity augments the next release event, and through depression, in which vesicle depletion reduces subsequent responses. Long-term potentiation (LTP) and long-term depression (LTD) are activity-dependent changes that persist for hours to weeks and are considered the cellular mechanisms underlying learning and memory. LTP at glutamatergic synapses involves NMDA receptor activation, calcium entry, and subsequent insertion of additional AMPA receptors into the postsynaptic membrane. These processes follow a rule first articulated by Donald Hebb in 1949: synapses strengthen when presynaptic and postsynaptic activity are correlated. The role of plasticity in learning, memory, and neural circuit development is reviewed in the NIH PMC article on synaptic signaling in learning and memory.

Artificial and Neuromorphic Synapses

The computational properties of biological synapses have motivated the development of artificial synaptic devices for neuromorphic computing. These devices emulate the analog weight-update behavior of a biological synapse in hardware, enabling neural network computations to be performed with far greater energy efficiency than conventional digital processors running equivalent algorithms in software. Memristors, organic electrochemical transistors, and phase-change memory cells have all been demonstrated as artificial synapses with multiple programmable conductance states, each state corresponding to a stored synaptic weight. Organic electrochemical neuromorphic devices operating below one volt and capable of more than 500 distinct conductance states have been demonstrated at laboratory scale. Photonic synapses, which use optical rather than electrical state changes, offer switching speeds below one nanosecond, compatible with high-throughput inference hardware. IEEE Spectrum reported on a low-power flexible organic artificial synapse designed for potential brain-computer interface integration. An overview of device types and their neuromorphic applications appears in the IEEE Germany EDS chapter article on synaptic devices and artificial neurons.

Applications

Synapses have relevance across a wide range of disciplines, including:

• Neurology and psychiatry, where synaptic dysfunction underlies disorders including Alzheimer's disease, schizophrenia, and major depression
• Pharmacology, where most psychoactive drugs act by modulating neurotransmitter release, reuptake, or receptor binding at synapses
• Neuromorphic engineering, where artificial synaptic devices implement energy-efficient pattern recognition and inference hardware
• Brain-computer interfaces that record and stimulate at the synaptic and single-neuron level
• Computational neuroscience and machine learning, where biologically inspired learning rules such as spike-timing-dependent plasticity inform network training algorithms