We got the first half of the synaptic activity a couple weeks ago. Recall that when one neuron is inspired to talk to another one it releases neurotransmitters into the synapse. Sites of neurotransmitter release can be very precisely lined up with clusters of receptors on the other side of the divide. The receptors are highly organized with other binding and signaling proteins into a large architecture referred to as the post-synaptic density. Neurotransmitters float across the relatively short and confined chasm to the receiving neuron where they bind to neurotransmitter receptors. Binding of neurotransmitter to the receptor produces two general classes of events: 1) It causes the receptor to change shape such that it forms a channel allowing certain ions to flow through (ionotropic receptors) or 2) it causes the receptor to change shape such that enzymes on the intracellular side of the post-synaptic membrane are activated (metabotropic receptors). One transmitter, such as glutamate, can have both ionotropic and metabotropic effects. The effect is a property of the receptor, not of the transmitter. Glutamate has, for instance, three classes of ionotropic receptors with different ion permeabilities: AMPA-type, NMDA-type, and kainate-type, and three groups of metabotropic receptors: I-III.

Glutamate is the main transmitter for fast excitatory transmission, the sort of thing you measure in EEGs, the sort of speed you expect to handle information processing. The majority of this transmission occurs via AMPA receptors located in the post-synaptic density. Post-synaptic densities occur very often at the heads of dendritic spines. A neuron is a polarized cell having two types of processes: axons and dendrites. Major excitatory neurons in the hippocampus and the neocortex usually have one axon (that can branch many times to allow multiple synaptic partners) that serves as an output and many dendrites that serve as inputs. The dendrites are covered in dendritic spines. So in the image graciously provided by wikipedia below, you can see what the structure of a neuron on the post-synaptic side looks like. An axon from another neuron would likely be passing nearby and produce a presynaptic specialized structure called a bouton adjacent to the spine head. Spines are dynamic structures. You can imagine them wiggling around, shrinking and extending, widening and thinning, etc..

There have been a number of recent investigations into the function of dendritic spines. A fairly widely accepted account is of the spine as a biochemical compartment. If neurotransmitter action causes calcium to flow into a spine head you might want to keep it right there next to the synapse that triggered the influx so the inevitable enzymatic signaling functions are performed in an input-specific manner. Calcium can turn on and off a lot of enzymes, in case that's not clear. You wouldn't necessarily want it floating up and down the dendrite all willy-nilly. Kasai and colleagues have done beautiful work showing that spines of different shape and size can have different calcium compartmentalization features. They suggest that spines which restrict diffusion are more likely to undergo biochemical signaling leading to synaptic plasticity, producing subsets of learning vs memory spines. Others have called into question this role for spines, pointing out that calcium diffusion can be restricted in neurons that don't even have spines because of a high-density of calcium binding proteins and a general molecular crowding in dendrites.

The Yuste group in particular has repeatedly made this point and in a series of papers this year suggested a role for spines in electrical isolation of synaptic inputs. They adopted the use of a recently synthesized novel class of chemicals that can insert into cellular membranes and, upon precise laser stimulation, fluoresce in a voltage-dependent manner. They can thus visualize the electrical signals at spine heads and have made the suggestion, predicted by much older theories of spine function, the spine necks act as resistors reducing charge flow into the main dendritic shaft. For reasons I won't go into, isolating the electrical signals in individual spines changes the nature of excitation in neurons, allowing all excitatory signals to be simply summed instead of added up in a complex sublinear fashion. It is not clear how the electrical function works. The resistor property is neck-length dependent, but the effect on summation is not.

I have recently come across two other attributes of spines. One is that spines cause anomalous diffusion along the main dendritic shaft. I quote so as not to pretend to know too much, "Anomalous diffusion arises when the random walk performed by molecules is influenced by their previous positions in space." At face value, this is confusing because you would expect any movement to be dependent on the prior position, else I could teleport if I didn't disintegrate first, but I have been beaten into submission by fancy Monte Carlo simulations and graphical solutions of diffusion equations and I will believe whatever Santamaria et al. tell me. What I take away from it is that as important signaling molecules drift up and down the dendrite they sometimes take a sidetrip into a dendritic spine and get trapped there for a while, especially if the spine head is pretty big and the neck is pretty small. This generally makes the flow of molecules up and down the dendrite seem more erratic. A similar trapping idea comes across in a recent Svoboda lab paper. I think I want to discuss this paper separately because they do some crazy things to get the images they are working with and it can take us down into the nitty grit of postsynaptic density constituents. Related to the current discussion, they suggest that spines with larger postsynaptic densities contain more receptors and thus stronger synapses and that one way to maintain the difference between strong and weak synapses over time in the face of degradation of constituents is for larger spine heads to trap important post synaptic density molecules better. This structural method is only one possibility of how the trapping could occur suggested in the paper. It's PLOS, so go read it if you can't take the suspense.

All this thinking about dendritic spine functions got me thinking about how they must've evolved in the first place. Which function was selected for or is it possible to select for multiple functions at once? Of course, the solutions aren't mutually exclusive. I can't seem to discover what the most distant organism that has these structures is. I found one abstract that says that planarians have them. Before spines develop their characteristic shape they resemble and are called filopodia. Single-celled organisms even have filopodia, so I think that's going too far back. When did filopodia start to specialize? Where is the first synapse? On a spine or no? Does anybody out there know? If not, maybe we could reconstruct it by doing molecular phylogeny on proteins known to induce spine formation. One of these is called Shank. Perhaps there are others and a compelling story can be told about the order in which the constituents were initially produced.