The Svoboda paper in the November issue of PLOS Biology (to differentiate from the Svoboda article in the December issue) forces us to really take seriously the idea that the synapse is dynamic. It is easiest to imagine a synapse sitting with say 10 AMPA receptors and then following learning or potentiation 20 more receptors are inserted. There the synapse sits with 30 receptors, newly strengthened. The truth is that AMPA receptors are constantly being removed and replaced as are other members of the post-synaptic density. Synaptic potentiation probably takes the form of something like a brief moment when the rate of receptor insertion is greater than the rate of receptor removal followed by a return to equal rates with a larger number of receptors in the turnover pool. The post-synaptic density (PSD) is a structure containing neurotransmitter receptors of different sorts, signaling proteins, actin cytoskeletal components, and scaffolding proteins that hold the whole shoot'n'match together. PSD-95 is a scaffolding protein. It binds multiple PSD constituents including Stargazin which binds AMPA receptors.

One theory holds that PSD-95 marks the spot that AMPA receptors can cycle in and out of, so synaptic strength is set by the number of 'slots' provided by PSD-95 rather than a fixed population of AMPA receptors. The implication is that PSD-95 levels in the synapse will be more stable and could provide the mechanism for memory maintenance since high AMPAR turnover rates obviously can't do the job. But we run into a sort of homunculus problem. We have a little man that can tell the AMPARs where to go, but who tells the little man where to go. This problem is easiest if PSD-95 is planted at a synapse once and stays there without degrading. This isn't the case, though as Gray et al. show.

And how they show it is way nice. They have a form of GFP that is photactivatable (paGFP). They fuse it to PSD-95 in an expression plasmid and put it in mice's heads using in utero electroporation. I've never heard of it before either. They open up a pregnant mouse mom, pull out the uterine horn, grab fetal mouse heads in between a special pair of electric tweezers, inject their plasmids into the lateral ventricle, and give a little stimulation to disrupt cell membranes to let the plasmid in. Amazingly, it works. Although they don't mention how many fetuses they trashed in the process. One insane technique per paper isn't enough though. Once the injected fetuses grow up, a small segment of their skull is removed and replaced with a glass window so the dendrites containing paGFP can be visualized over many days in a living animal.

When you look at the dendrites you don't see anything until you blast the GFP with a laser to turn it on. Since lasers are very precise (and this technique requires the meeting in space of two different laser pulses) they are able to activate GFP in individual spines. They proceeded to activate PSD95-paGFP and measure how long it takes to disappear from the spine, thus gauging the retention time of PSD95 at a given synapse. It takes longer than GFP by itself suggesting that binding interactions help retain PSD-95 in the PSD, but it still takes less than an hour. Where does it go when it leaves? It's not time to be degraded yet. PSD-95 has a half-life of ~36 hours. Instead, it goes barhopping. If the dendrite is Bourbon St, then PSD-95 is a reveler stopping in this spine for a sec, that hotel for a minute, and on and on until the cops finally catch it and toss it in the proteasome.

If our master controller architectural component, PSD-95, isn't there directing traffic, how does the synapse maintain its strength? Well, we have a glimpse at a solution because Gray et al. studied the movement as a function of spine size. Larger spines capture diffusing PSD 95 better and they hold onto it longer. Now here is an area where my understanding is going to be a little fuzzy. In my reading, we know that spine shape doesn't seem to be much of a barrier to diffusion because GFP on its own diffused out of synapses with a half-second median retention time. However, Gray et al. go to great pains and even produce a stimulation to suggest that spine geometry can influence retention time. In one of the papers they cited (Bloodgood and Sabatini, 2006), a spine was considered diffusionally isolated if the equlibration constant between spine and dendrite was greater than two seconds. Two seconds seems negligible to me compared to the apparent retention time of PSD-95, but to be honest I don't want to take the time to deal with all of their models and equations. If anyone else wants to take a crack at it and tell me what you find, please be my guest.

So now we're back at the homunculus problem. Who is gonna tell PSD-95 which spines it should hang out at the most? We can go searching for another protein that is retained in spines longer and has a half-life of months or years. I can even think of a couple good candidates, but I think the quest is in vain. I predict that we will not find a single protein that can account for lifelong changes in synaptic strength and therefore childhood memories. I know just enough about dynamic systems now to make an ass of myself on the internet, but my feeling is that this sort of analysis is going to have to be brought to bear on the PSD. We don't need a master controller, we need a stable system-state between several constituents. I've cycled through a number of metaphors in my head, but I think maybe a decent image at least is a table with hundreds of legs cemented in the ground. You can remove and replace all of the legs over some period of time as long as you leave three or four. You could remove the table top and replace it with a new one. You could even dissolve the cement and replace it as long as you don't do all of these things at once. PSD-95 has many binding partners in the PSD. If one is removed, the structural integrity could be maintained. If you hold your friend's place in line they can go pee. If you change the guard, he can get some relief and the palace is still protected. We don't need an immortal homunculus to achieve lifelong memories.