Case Notes 3

The business end of the mammalian brain is its huge mess of interconnected neurons, all constantly sending signals to one another in a massively-connected network. Each neuron is a single cell with a rather complicated structure. The exact morphology varies according to location and function, but there's generally a sprawling 3-dimensional collection of branched input channels, called dendrites because of their tree-like appearance; a main body or soma that contains the nucleus and organelles; and a long output channel called the axon. The axon from each neuron intersects the dendrites from many others at synapses. The neurons are not exactly joined at the synapses, but they are very close together.

Neurons fire by sending an electrochemical action potential down the axon, which triggers the release at the synapses of small packets (vesicles) of chemicals called neurotransmitters. Some of these released molecules will bind to receptors on the other side of the synapse, which then allow passage of particular ions into the interior of that neuron, which in turn affect whether or not it fires.

There are a two basic kinds of neurotransmitter: inhibitory, which reduce the likelihood of firing in the receiving neuron, and excitatory, which increase it. At any moment, the firing rate of each neuron acts as a kind of summation of all the excitatory and inhibitory signals it is receiving from the hundreds or thousands of neurons tickling its dendrites.

There are a bunch of different neurotransmitters at work in different neurons, but the most common are glutamate, which is excitatory, and gamma-aminobutyric acid (GABA), which is inhibitory. A given neuron will generally have receptors for both, but will release only one or the other at all its synapses; in other words, a neuron takes both positive and negative inputs, but always gives only one kind of output.

Neurotransmitter receptors are ligand-gated ion channels -- protein complexes that float around in the cell membrane, ready to admit ions when the appropriate chemical turns them on. For synapses -- and hence the whole neural apparatus -- to function, the correct receptors must be present in the membrane of the receiving cell in the area where the neurotransmitter gets released. However, cell membranes are dynamic -- seething seas of lipids and embedded proteins that get shoved this way and that by all sorts of random forces. While some kind of order is, on average, maintained, it is not at all clear how.

Membrane proteins are pretty small, making it rather difficult to see just what's going on down there. However, it is possible to get some idea by using confocal microscopy with single-photon fluorescence. If you can bind an appropriate fluorophore to the receptors of interest, you can image them as little specks of light in the wine-dark membrane.

There are a number of ways to do this, none of them perfect. Candidate fluorophores include quantum dots, which are artificial semiconductor beads whose emission spectrum depends on their size, and GFP, a naturally-occurring jellyfish protein which has become one of the great workhorses of biological imaging by virtue of being a pretty stable fluorophore that can be manufactured by the genetic machinery of the cell itself. In the former case, you need to attach the quantum dot to specific antibodies that will seek out and bind the proteins of the receptor. In the latter, you genetically engineer the cell so that when it expresses the receptor proteins it adds the GFP to them. (This is a potentially risky process, since protein function is very dependent on shape and having a huge stack of alien peptides plonked on the end could easily change the target's behaviour, but it seems to work in a surprising number of cases. Presumably the failures tend not to wind up in the literature: scientific journals almost exclusively publish positive results rather than papers concluding "Um, who knows?")

Both these approaches have been used to image receptor movement in the post-synaptic membrane, but the problem doesn't end there. A video sequence of the cell membrane obtained this way simply shows a pattern of bright specks at successive time steps. It's quite difficult to go from such an image sequence to a clear model of how the receptors are actually moving.

Even if the data were ideal -- each image providing a perfect identification of the positions of all the receptors at each instant -- it would be non-trivial to correctly identify the movements of the receptors, because there are a lot of them and they all look the same; it's a giant shell game. And -- of course -- the data are not ideal.

Experimental measurements are always prone to instrumental noise, but in this case there are also other confounding factors. The field of view is limited, so the receptors may move in and out: those visible at one instant may not be the same ones visible at the next. The resolving power is also limited, so if two receptors are close together they may be misread as one. Worse, the fluorophores are imperfect. GFPs degrade and quantum dots blink -- getting trapped in quantum states from which it isn't possible to drop to ground -- so only an indeterminate subset of receptors may be visible in each frame. New receptors may be made and added to the membrane, and ones already there may be removed by invagination. Finally, there's the very effect we want to measure: the motility of the receptors may vary across the field of view, but we don't know where.

Fortunately, analytic methods exist to deal with such difficulties. Unfortunately they can be very computationally expensive (as bad as O(n!); problems can get harder than that, but not much). Heuristics are available to help us along, but there's an element of lossiness. Certainty -- as ever -- is the province of the dull. The domains of science are weakness and imperfection.

It's early days yet, but individual receptors do indeed seem to wander around stochastically while statistically maintaining sufficient concentrations at the synapses for the neuron to function. One theory for how this works involves scaffold proteins inside the cell, anchored to the cytoskeleton, which encourage -- in a thermodynamic sense, which is to say: make energetically favourable -- the localization of receptors in the nearby membrane.
Posted by matt at December 3, 2006 11:05 AM