Neuron, Vol 40, 381-388, 9 October 2003
Review
Neurotransmitter Release at Central Synapses
Charles F. Stevens
Introduction
A PubMed Search on “presynaptic mechanisms” returns 17,362 references, and by restricting the search to the last 15 years—the time since Neuron first appeared—you get 13,220 references. Two recent review articles on subtopics covered here (short-term plasticity [Zucker and Regehr, 2002] and mechanisms of vesicle cycling [Murthy and De Camilli, 2003]) cited 360 and 220 references, respectively. Clearly, if my review on neurotransmitter release is to fit into this issue of Neuron, something has to be done.
My solution is to limit the size this review by identifying what, in 1988, I perceived—and thought the field as a whole thought—to be the most important questions in presynaptic function. I have described what we knew about the answers to these questions in 1988 and what we know now. My perception of central questions in presynaptic function is based on my own reading of the literature and on discussions with, I think, virtually all of the leaders in the field, from Bernhard Katz on, about where we want to go and how to get there.
I am mindful of the fact that friends come and go, but enemies accumulate, and I feel no need for fewer friends. By citing 104 references, I recognize that I have the potential to make the authors of at least 13,116 papers feel neglected. Very many of my favorite papers, some of the most beautiful and important ones, are not cited here because they did not fit into my selection scheme. I apologize to the neglected authors, especially the ones I admire and know personally, and hope they understand I was not making a value judgment about their contributions.
Because synapses are of such central importance to neurobiology, many reviews of synaptic structure and function have appeared over the past several years. The most comprehensive set of reviews appeared in a book, Synapses, two years ago, which includes chapters on synapse structure (De Camilli et al., 2001a), the physiology (Regher and Stevens, 2001) and molecular biology (Sudhof and Scheller, 2001) of neurotransmitter release, endocytosis (De Camilli et al., 2001b), and the synaptic cleft (and the extracellular matrix) (Sudhof, 2001). Other recent reviews cover topics such as the molecular biology of vesicle fusion (Jarousse and Kelly, 2001; Kavalali, 2002; Rizo and Sudhof, 2002), the molecular biology and biochemistry of synaptotagmin (Chapman, 2002), the cell biology of the presynaptic terminal (Murthy and De Camilli, 2003), important evidence from fly and worm about the SNARE hypothesis (Kidokoro, 2003; Richmond and Broadie, 2002), an evaluation of kiss-and-run as opposed to the Heuser-Reese model of vesicle cycling (Morgan et al., 2002), the chromifin cell as a model for studying fusion mechanisms (Rettig and Neher, 2002), possible functions of synapsins (Ferreira and Rapoport, 2002), and mechanisms of short-term synaptic plasticity (Zucker and Regehr, 2002).
Synapses in 1988
Katz and His Followers: Defining the Questions in Synaptic Function
The study of synapse function was dominated in the 20th century by Bernhard Katz, and his formulation of the problems in the field summarized the state of synaptic physiology in 1988. Among Katz's many contributions, two theories, called “hypotheses” by Katz, stand out in the current context: the quantal hypothesis and the calcium hypothesis.
Katz discovered that neurotransmitter is released in integer multiples of a packet, called a quantum, that corresponds to an individual synaptic vesicle (Katz, 1969). Implicit in this view is the idea that neurotransmitter is released by exocytosis and that vesicles must cycle so that their membrane is in some way reclaimed after use to keep the cell surface area from growing and to replace spent vesicles with new, competent ones. The classic work of Heuser and Reese (1973) concluded that vesicles fuse with the active zone (an active zone is a specialized presynaptic membrane where synaptic vesicles are docked and released) to release their contents and then are recovered as clathrin-coated vesicles that, in turn, fuse with an endosomal compartment; new vesicles then bud off these endosomal compartments to replace those that have been used. Although the Heuser-Reese model was the dominant one in 1988, Ceccareli and coworkers (Ceccarelli et al., 1973, 1979) pushed an alternative view—called kiss-and-run—in which vesicles opened a fusion pore in the presynaptic membrane to release their neurotransmitter and then reversed this process by closing the fusion pore, dissociating the vesicle from the membrane and refilling it with neurotransmitter; in this model, the vesicle maintained its identity throughout multiple rounds of release. The kiss-and-run model always had its adherents, but the Heuser-Reese work was so beautifully done and so compelling that most workers accepted this theory.
According to Katz (Katz, 1969), release can occur only at specific presynaptic locations, called release sites, N of which are present at a synapse. When a nerve impulse arrives, neurotransmitter release occurs with a probability p at each site, and the sites function independently of one another. An aside: discussions of neurotransmitter release often use two distinct probabilities, the probability that a quantum is released at a Katzian release site (p), and the probability that a nerve impulse arrival at a synapse will result in neurotransmitter release (this is called the release probability). There is the possibility of confusion because, according to some, a single synapse is the same as a Katzian release site. Because, as will be discussed below, the identity of a Katzian release site is not settled, I will distinguish between these two probabilities, calling the first (p) a Katzian probability and the second simply release probability.
Release, according to the Katz picture, follows binomial statistics, just like coin flipping, so that the average number of quanta released by a nerve impulse is Np; this is the average number of heads you would get by flipping N coins (one for each release site), each with a probability p of heads. Katz's quantal theory was based almost exclusively on studies of the neuromuscular junction, which is, in fact, many simultaneously activated synapses that one axon makes onto a postsynaptic muscle cell (Steinbach and Stevens, 1976).
Because the Katz view was derived from the study of many synapses activated by a single nerve impulse, the meaning of release site was unclear: is a release site a single fusion-ready vesicle, a single active zone, or a single synapse with, perhaps, multiple active zones? Several workers had proposed that a single synapse can release only one vesicle per nerve impulse (Korn and Faber, 1991; Redman, 1990; Zucker, 1973a), in which case a release site would correspond to a synapse (or possibly a single active zone), but this view was always controversial, and the question of what constitutes a Katz release site was unsettled in 1988. And because the physical correlate of a release site is unclear, so is the correlate of the Katzian probability p. If a release site is a single active zone, for example, then p would be the probability that one of the multiple available vesicles will be released, but if a release site is an individual release-ready vesicle, then p would be the fusion probability for that vesicle.
The Katzian probability p, whatever its physical referent, was known to be increased by increasing the extracellular calcium concentration and decreased by increasing the extracellular magnesium concentration, and this observation led to Katz's “calcium hypothesis.”
According to the calcium hypothesis, a nerve impulse causes calcium ions to enter the presynaptic terminal where they bind to a calcium sensor, the calcium-liganded form of which causes an increase in p so that vesicle fusion occurs. The quantitative relation between the average quantity of neurotransmitter released by a single nerve impulse and divalent ion concentrations was established by Dodge and Rahamimoff (1967) when they were postdoctoral fellows with Katz. They derived an equation describing the calcium dependence of neurotransmitter release (the Dodge-Rahamimoff equation) from the assumption that a calcium-sensing molecule can bind up to four calcium ions (the binding sites were assumed to be independent and identical), and that the Katzian probability p is proportional to the concentration of sensors that are fully occupied (i.e., with four bound calcium ions) (Dodge and Rahamimoff, 1967). Katz, working with his long-term collaborator Miledi, further showed that a brief, transient calcium influx produced by a nerve impulse was directly required for release (Katz, 1969). These combined observations constitute the theory (the calcium hypothesis) that release is produced directly by calcium interacting with a molecular calcium sensor (identity unknown) that enables vesicular exocytosis.
In addition to the intraterminal calcium concentration, the past history of synaptic use was well known in 1988 to alter the Katzian probability p. Mallart and Martin (1967, 1968) had characterized facilitation (now often called “paired-pulse facilitation” and abbreviated PPF), an increase in p that follows a nerve impulse arrival and lasts for several hundred milliseconds. A very long-lasting increase in the Katzian probability after extensive synapses use, posttetanic potentiation (PTP), also had long been known. Starting in 1975, Magleby developed a quantitative description of four kinetically distinct processes that increase neurotransmitter release—two phases of facilita
