Monday, October 09, 2006

Dendritic translation and memory   posted by Coffee Mug @ 10/09/2006 08:15:00 PM

The recent Science paper by Raab-Graham et al. that rosko linked shows dendritic synthesis of a protein modulated by various pharmacological agents. A review by Sutton and Schuman came out in Cell on the same day and serves as a good guide to the importance of dendritic protein synthesis in synaptic plasticity and memory. Raab-Graham et al. speaks to one of the issues in the S&S review, namely, how to identify proteins that are actually being synthesized in dendrites. What follows is the Readers' Digest guide to key issues identified by S&S and an interjection showing where the Science paper fits in.

Why are we concerned with local/dendritic protein synthesis? Well, first, why are we concerned with protein synthesis at all? Memories appear to require protein synthesis for permanent storage. People have studied this in numerous learning paradigms by injecting protein synthesis inhibitors into slugs or rodents at different times after training sessions. Without going into specifics and caveats, you can generally take it that these inhibitors are only effective in disrupting the memory if they are administered within the first couple or three hours after training. The same goes for cellular models of learning and memory (i.e. long-term potentiation, LTP). So in order to stabilize a nascent memory new proteins must be synthesized.

The end result of all of this protein synthesis must be the modification of synapses. We know that not all of the synapses on a neuron are potentiated in response to stimulation, so we need the new proteins to act only at certain synapses, presumably synapses that are in some way related to the ones that drove the protein synthesis. The proteins could either be generated in the soma and trafficked to the proper portions of the dendrite or they could be synthesized from preexisting mRNAs in the dendrites. You can see some advantages to the latter proposal as all of the action can happen in one little area without expending energy to carry around all those proteins and figure out a zipcode system to make sure they end up at the right inputs. In ways it is similar to the contrast between using snailmail to ask for a reprint and receiving the pdf via email so you can print it off yourself next to your desk.

Not to mention, as S&S lay it out, polyribosomes (protein printers) were shown in dendrites by Steward and Levy using electron microscopy. It remains possible that they are there performing a housekeeping role and that plasticity-related synthesis is carried out in the soma. We do know that dendritic synthesis can do the job of maintaining LTP into the protein synthesis-dependent phase. Some of the best demonstrations involve microsurgical dissection where the dendrites are physically separated from the cell body prior to LTP induction. Others have shown that restricting protein synthesis inhibition to the dendrites still blocks late-phase LTP.

It has been more difficult to show the necessity of dendritic protein synthesis in live, behaving animals. The closest anyone has come was a transgenic mouse carrying a mutated version of the alpha subunit of calcium/calmodulin-dependent protein kinase II (CaMKII), one of the leading candidates for the plasticity-induced, synapse-altering protein. The mouse lacked a chunk of the CaMKII mRNA that doesn't code for protein, but instead carries a dendritic localization signal. So there can be no local synthesis of this really important synaptic protein because the mRNA isn't there locally to synthesize. These mouse had impaired memory. But alas, this mouse and all the other mice that have been used to study protein synthesis and memory are second-generation transgenics, meaning that the genetic manipulation is not tightly restricted in the temporal domain. Compensation is the rule not the exception, so if we screwed up the system, I expect lots of things besides CaMKII mRNA localization were altered in this mouse. In memory research, timing is everything. Not only could the mouse be just globally weird, but the manipulation could be affecting the acquisition, storage, maintenance, or retrieval of the memory.

S&S suggest a number of areas where basic knowledge related to the issue of local protein synthesis is lacking.

1) What mRNAs are in dendrites? Should we be identifying each one individually or will microarrays using samples that ostensibly carry only dendritic or synaptic mRNAs do the trick? It's a classic quantity vs. quality issue in my mind. It would be nice if you could do a neuronal cell line array like people have been doing with yeast arrays . I can imagine a library of mammalian cells (either a neuronal line or a stem cell-line treated with neuronal differentiation cues) containing MSH2 tags in each mRNA and an MSH2-binding protein-GFP hybrid. Sorry, I'm indulging myself. There is probably a more efficient solution than what immediately pops into my head.

2) Which proteins are actually synthesized in dendrites? This is the part that Raab-Graham et al is getting at. It is not good enough to show which proteins can be found in dendrites, you have to show that they appear there in response to synaptic activity. The CaMKII mouse mentioned above provides one way to examine the issue. Restricting the mRNA to the soma led to a large decrease in synaptic CaMKII. This is reasonable evidence that synaptic CaMKII comes from dendritic CaMKII mRNA. Also, the amount of CaMKII in dendrites far from the cell body jumps up following stimulation, so fast that it can't have been transported from the soma. This protein, and now Kv1.1, are the only ones that have undergone this type of analysis. The CaMKII studies didn't require any photoconvertible protein hybrid though, so I'm a little uncertain as to why Raab-Graham et al had to go through all that technical trouble. Also, if my understanding is correct, they could have reached the same conclusions using FRAP (fluorescence recovery after photobleaching). Rather than looking for a new green signal after converting everything to red, they could have simply looked for a new green signal after bleaching out all existing fluorescence. Paging Dan Dright.

3) How specific is LTP? While many use the terminology, S&S note that we haven't shown that LTP is synapse-specific. When ribosomes are found in the dendrites they are found near synapses, at the bases of dendritic spines, not in synapses. Synaptic tagging experiments have shown that proteins synthesized due to strong stimulation at one input can be 'captured' by other synapses receiving weak stimulation. So the locally synthesized proteins aren't all that faithful to the inputs that led to their creation. Some have used two-photon glutamate uncaging to try to stimulate individual synapses, but still it is not clear that the changes are restricted to just those receiving stimulation. Perhaps multiple nearby synapses are affected at once. There is some evidence that synapses near to each other might at lease share input modality, so you could potentially still be encoding a specific association between stimuli using clustered plasticity.

4) One last issue, and I'll keep it short. Protein synthesis inhibitors usually only have effects in the first few hours after training at best, while memory persists for weeks, months, or years. Proteins turn over though. We really just don't know how fast most synaptic proteins are degraded. They could last long enough to do the job of maintaining memory. If not, perhaps the memory can be maintained for the amount of time it takes for the inhibitors to wear off. You can't really address this question by extended protein synthesis inhibition because you will eventually end up harming the cells irreversibly. So we need to measure some protein half-lives.

A few of these questions seem like they would be trivial for people outside of the neuroscience community, people who have expertise in high-throughput biology. It's not clear to me why their wild ever-shifting attention hasn't landed on synaptic plasticity yet, but if any of you are reading this maybe you could just get some physiologist to hand you tissue (preferably synapse-enriched) on a time course after LTP. Run a microarray. Do some mass spec proteomics. Tell us which proteins there are more or less of before and after. That's not that big a deal right? I'm pretty sure I can find you the tissue if you are having a hard time.