A New View of Synaptic Transmission
An Interview with Terrence Sejnowski, Ph.D.


by Brenda Patoine

January, 2006

Q: You recently published research that bucks the conven­tional wisdom about nerve signal transmission at the synapse. Specifically, you have evidence for “ectopic release” of neurotransmitters, i.e., evidence that these chemical messengers are released not only at a single “active zone” at the synaptic cleft, but also that they diffuse more broadly at other sites beyond the recognized receiving site. Why is this work important?

Sejnowski: Our results provide evidence for a different concept of the synapse. The textbook view of the synapse describes it as a place where rifle-like volleys of neurotransmitter are launched from one defined region of the sending neuron (the “active zone”) to another defined target on the receiving neuron. In contrast, our data suggest that the synapse can act like a shotgun, firing buckshot-like bursts of neurotransmitter from multiple locations outside the active zones to reach receptors arrayed beyond the known receiving sites.

A study published a couple of years ago by Craig Jahr and colleagues at the Vellum Institute in Portland, Ore., showed ectopic release from Perkinje cells in the cerebellum onto a specific type of glial cell. While that’s not neuronal transmis­sion, it showed that ectopic release can happen between a neuronal structure and a glial structure in the central nervous system. We were able to show for the first time that it happens between two neurons, at a neuronal synapse.

Q: For this study, you used a computer model to reconstruct the architecture of the synapse and reproduce the act of neurotransmitter release at that synapse. Why did you choose this approach?

A: The computational modeling played a crucial role in achieving the result we found; it is really what makes this initiative unique. Without the model we couldn’t link together the anatomy and the physiology; it’s kind of the glue that allows those to be put together in the same framework.

Physiologists have for a long time been recording from neu­rons to track the electrical impulses that underlie the transmis­sion of nerve signals at the synapse. Meanwhile, anatomists, who examine the synapse from a purely structural point of view, see a great variety of geometry. In the case of the type of nerve junction we’ve studied, called a calyx synapse, the presy­naptic terminal appears as a cap on top of the surface of the cell body, or soma. Coming out of the soma are little fingers like a shag rug, with the cap on top of this shag rug. You can also see the vesicles that store neurotransmitter in the presy­naptic terminal in electron micrograph pictures.

Anatomical studies indicate that there are specialized regions at the synapse called active zones where you find pre- and post-synaptic specializations. These are the classic synapses you see in textbooks—a nice simple picture with the receptors in the post-synaptic density, the vesicles lined up ready to be released by an electrical signal moving down the nerve fiber. The question is, where are the neurotransmitters coming out, where are they being released? You can’t tell by looking at a static picture. So here you have anatomy on one hand giving you a very complex picture of that synapse, and on the other hand you have the physiologic recordings indicating the functional signaling, but how do you actually connect the two? That’s what we did. Mark Ellisman at the National Center for Microscopy and Imaging Research in La Jolla, Calif., reconstructed the anatomy of the synapse in exquisite three-dimensional detail, then we used the computer model to release vesicles from different locations throughout that area.

Q: What specifically did the computer model indicate?

A: We modeled minis—miniature synaptic potentials—which occur spontaneously in cells in various sizes and shapes. By measuring the minis from actual cells, then feeding those measures into the model, we were able to test our hypothesis that the majority of the neurotransmitter release actually occurred outside the active zone. Using the model, we varied the probability of release outside the active zone from 0% to 100%, with 0% modeling a scenario in which all release occurs in the active zone and 100% being all release outside the active zone. We found that if you restrict the release of neurotrans­mitter to those active zones where its release is traditionally expected, that you cannot reproduce the distribution of ampli­tude in these mini synaptic potentials that occurs in reality. The probability that matched the actual distribution occurs when 90% of the vesicles are released outside the active zone. That is very strong evidence for ectopic release.

This is surprising to a lot of people because no one really expected it, although there is a evidence in the literature point­ing in that direction. So now we actually have the smoking gun.

Q: Given this evidence for a new view of neurotransmitter release at the synapse, what are the next steps in this line of research?

A: The question now is why: what is the purpose of ectopic neurotransmitter release? Also, how widespread is it? Is this something that happens outside this particular synapse? We suspect so, for a couple of reasons. The synapse we studied is in the chick ciliary ganglion, where neurons innervate the pupils of the eye, a part of the autonomic nervous system. Humans have these same synapses, even though the neuron structure is different.

I suspect that ectopic release is much more common than we have realized. My hunch is based on the fact that there are many extrasynaptic receptors outside the active zone; they’ve been known to exist for a long time. Certain types of recep­tors—GABA receptors for example—are more prevalent outside than inside the active zone. No one has ever been able to pin down the reason for this. There have been suggestions that maybe there is spillover, that some neurotransmitter spills outside of the cleft during release, and if it gets out far enough from the cleft it can activate these extrasynaptic receptors, but it’s very difficult to actually prove that. But if I’m right, if our result holds for some of the other synapses in the central nervous system, then it could be that those extrasynaptic receptors are directly activated by neurotrans­mitter release from vesicles that are outside the active zone.

Q: If, as you suggest, this type of ectopic release is common, why hasn’t it been seen before?

A: It’s very difficult to get direct evidence of ectopic release— or even for traditional neurotransmitter release at active zones, for that matter. In the neuromuscular junction, which is the only synapse where there is direct evidence, scientists stimulated the nerve and froze it at the moment it released the vesicle. By doing so at the precise instant of release, they were able to capture “omega figures,” partially fused vesicles, at the point where they’ve joined the membrane but haven’t collapsed yet after releasing their neurotransmitter load. These “figures” were in exactly the space they were expected to be, at the active zones where the vesicles are lined up. This was a heroic experiment, done back in the 1970s. It’s extremely difficult to do that in a central synapse or in the middle of the brain: there is just no access. You can’t freeze the brain quickly enough.

That’s why we’ve turned to indirect methods, such as the computer modeling. This approach enables us to create very accurate, very detailed computer simulations that incorporate all of the quantitative data from the kinetics of receptors, the locations of the receptors, the detailed geometry of the synap­tic cleft, the space between neurons, and so on. All of that data has been painstakingly gathered over many years, from many different laboratories—it’s all in the literature. The model is the essential glue that holds it all together. Without the glue to hold the data together, you don’t know what they imply.

Q: Does this type of complex computational modeling repre­sent our best hope for ever truly understanding the essential dynamics of brain function, given the innate complexities?

A: This is the future of neuroscience. Neuroscientists are really good at being reductionists, at taking things apart. We’re really good at identifying the molecules and all the different pieces of the puzzle, and even, to some extent, at determining how they interact with one another. However, in order to understand brain function, you have to be able to synthesize a tremendous amount of data. You need to take all those pieces and put them together and reconstitute the function so that you know if you have all the right pieces put together in the right way.

This is the future of neuroscience...in order to understand brain function you have to be able to synthesize a tremendous amount of data.

We worked on just one synapse; tackling the more complicat­ed brain is going to be much more difficult. Still, I think the very same approach can be used inside the brain to try to gather all the knowledge that has been painstakingly gained in all the neuroscience labs around the world, and to integrate all that data and all those quantitative measurements into the sort of models that we’re building. Then we can see if the model behaves the same way as real tissue. If it does, we’re on the right track. If not, then we’re missing something, and that may be just as interesting, or even more interesting, because we can use it as a discovery tool to figure out what needs to be added or what’s missing or what’s wrong.

My lab is now collaborating with Mary Kennedy at Caltech, Karl Svoboda at Cold Spring Harbor Laboratory, and Richard Weinberg at the University of North Carolina to take the next step and model synapses at the dendritic spines of pyramidal cells in the hippocampus, which are thought to be important for learning and memory. Our goal is to try to reconstruct what goes on inside the dendritic spine, and see if we can understand what occurs when calcium enters through the NMDA receptor and triggers a cascade of biochemical reac­tions, leading to a change in synaptic strength. So this is the synapse we’ll be tackling next with the computer modeling approach, with the goal of a better understanding of how the brain accomplishes the feat of learning at the synaptic level.