2 November 2002. It would seem intuitively appealing that the brain’s billions of minuscule information exchanges-the synapses-may be the main site of destruction in Alzheimer’s disease. After all, if no more transmitter buzzes across the narrow cleft and no more currents crackle on the receiving side, memories, learning, and other functions must surely disappear. Even so, surprisingly little is understood about what happens to synapses in the years preceding frank Alzheimer’s. This is beginning to change, however. Here in Orlando, a two-day satellite event to the main Neuroscience conference drew 140 researchers to focus attention on just this question. Featuring 22 talks and 45 posters, the symposium brought together basic synapse biologists with neurodegeneration researchers in an attempt to fire up some excitatory activity in the field and trigger new research projects. The journal Neurobiology of Aging sponsored the event, titled "Molecular and Cellular Basis of Synaptic Loss and Dysfunction in Alzheimer’s Disease."
Paul Coleman (University of Rochester), who co-chaired the conference, said that while certain populations of neurons clearly die in Alzheimer’s, they don’t disappear quickly, extinguishing well-functioning synapses overnight. Instead, he said, synapses probably fade away slowly long before the cell dies. The gradual dysfunction and structural disintegration of synapses deserves study in its own right independent of neuronal loss, which has received the bulk of attention to date.
To support his contention, Coleman reviewed studies in his and other labs showing that living neurons of people with AD are already losing synapses. Older studies showed that the dendrites of neurons in AD shrink first at the tips and then further towards the cell body while the neuron is still alive. Others showed that those regressing dendrites contain fewer synapses than do dendrites of normally aging brains. More recent work in Coleman’s lab showed that neurons that are filled with tangles in people with AD express much smaller amounts of certain synaptic markers than do neighboring neurons that have no tangles yet. Taking the same experiment a step earlier, Coleman said that even neurons from AD patients that have no tangles yet express lower levels of synaptic marker proteins than comparable neurons from normally aging people, suggesting that synapses decline in AD before classic pathology becomes visible.
Other work amplifying mRNA from individual neurons and analyzing it on microarrays identified a slew of synaptic proteins whose expression drops in still-living AD neurons, including dynamin, the adapter protein AP3MuB, and others. Interestingly, Coleman’s group found that genes involved in synaptic function decline even while structural synaptic markers remain normal, suggesting subtle synaptic dysfunction as an early sign of Alzheimer’s.
Most of the important questions remain wide open for study, said Coleman. This includes the mystery of what starts this decline, what its sequential steps are, and how these steps relate to pathology and cognitive function. Also unclear is why cholinergic neurons are supremely vulnerable while GABAergic neurons, for example, are relatively spared. Finally, it is also unclear whether synaptic loss precedes neuronal loss and symptoms in other neurodegenerative diseases.
Even so, one thing seems clear. "Cognitive decline is largely a function of the decline of synapses. The therapeutic notion that follows is, we should try to rescue the synapses, not just neurons. Once the synapses are lost, what’s the point in saving the cell body?" Coleman said.
What, then, should the intrigued AD investigator study? The remaining talks of the first day described different aspects of basic synapse biology that introduced some of the "whos" and "hows" of synaptic plasticity. Together, they offered numerous angles and candidate molecules to scrutinize in neurodegeneration studies. For example, co-chair David Bredt of the University of California, San Francisco, described the postsynaptic density (PSD) of the typical glutamatergic synapse. So named years ago because it appeared in electron microscopy as a stripe of fuzzy black dirt just below the nice clean line of the postsynaptic membrane, his dynamic organelle contains roughly 100 proteins. Especially prominent is a large, 30-protein signaling complex of glutamatergic AMPA and NMDA receptors, sodium and calcium channels, and anchoring and adaptor proteins for these channels called GRIP, homer, PSD-95, shank, as well as signaling proteins such as CaM kinase 2, immediate early genes such as arc, and c-fos, cell adhesion and cytoskeletal proteins (Kennedy, 2000).
Many of these proteins are highly dynamic; they respond to synaptic activity by turning on genes, recycling and trafficking receptors. For example, Bredt described studies of a protein underlying a spontaneous mutant mouse strain kept at Jackson Labs in Bar Harbor, Maine, that was dubbed stargazer because it stares upwards during epileptic seizures. Stargazin, it turns out, recruits intracellular AMPA receptors to the membrane and chaperones their glycosylation in the ER. All this happens in the dendrite, into which the ER extends. PSD-95 then recruits extrasynaptic receptors to the synapse, and both PSD-95 and stargazin activity changes with synaptic stimulation in ways that remain poorly understood, Bredt said. The stargazin variety in the unfortunate mouse strain is active in cerebellar granule cells, but one family member clusters AMPA receptors in glial cells, another in hippocampus, Bredt said.
Oswald Steward of UC Irvine outlined work into the poorly understood question of how the synapse communicates with the nucleus as a synaptic activity is being consolidated into a lasting memory. "There must be signaling from the synapse to the nucleus to turn on gene expression. Newly synthesized gene products must then be delivered back to the active synapses and changes take place over time that create an enduring alteration in synaptic efficacy," Steward said.
To get at this problem, Steward first pointed out that the base of dendritic spines (synapses ring the top of these spines) contains membranous polyribosomes that translate mRNAs. James Eberwine of the University of Pennsylvania reported that up to 600 different mRNAs get transported from the nucleus to these spines, and Steward finds about 20 of those in high abundance. These include cytoskeletal proteins, kinases, MaP2, calmodulin, G-protein, BDNF, and the many other components of the PSD signaling complex.
Consider the immediate early gene arc, for example. Synaptic activity induces its mRNA expression; in the hippocampus, a single large stimulus involving NMDA receptor activation strongly turns on its expression, said Steward. Interestingly, arc mRNA then is transported only to those synapses that were active, and accumulates there within two hours, about the time period during which consolidation takes place. In addition, Steward said, arc mRNA that is already present in the dendrite at the time of stimulation gets targeted to active synapses within minutes.
How does the call for arc expression reach the nucleus? Steward’s research favors the MAP kinase pathway in such a way that NMDA activation would trigger Erk phosphorylation (Erk is a downstream member of this pathway and translocates to the nucleus when activated.) The targeting of the resulting mRNA back to active synapses, however, works through a different signal, Steward said, perhaps CAM kinase 2, and further stimulation of the requisite NMDA receptors then trigger docking of arc protein to the postsynaptic density. He suggested this sequence as a possible pathway for how repeated stimulation can dynamically change a given set of synapses, but he added that it remains unknown whether and how these signaling mechanisms change in age-related memory disorders.
Eberwine expanded on Steward’s reference to mRNA transport and translation in the dendrites. Dendrites make up 90 percent of the cell surface of many neurons, he said, so it is not surprising that incipient neurodegeneration would begin there. Eberwine reviewed the clever methods his lab developed to remove individual dendrites from cultured primary hippocampal neurons, amplify their mRNA content, and screen microarrays to identify them.
To show that these mRNAs are actually translated into protein locally in the dendrite, his group sprayed tagged mRNA directly into dendrites and then detected its protein with antibodies. They developed a GFP marker for local translation in the dendrites, as well. Moreover, they found that regions in the mRNAs and in the local ribosomes control the particular characteristics of dendritic translation, which differ from those in the soma. Which mRNAs are translated in dendrites? One example is glutamate receptor mRNA, which is locally made and then inserted into dendritic membranes.
How do the mRNAs get into the dendrites? Signal transduction is not enough, since nucleic acids do not float unaccompanied through the cytoplasm. Eberwine showed the mRNAs travel with RNA-binding proteins, such as translin or PABP. However, it remains unknown which RNA-binding proteins bind which cargo RNAs, or how this binding controls transport. While this area is in its infancy, Eberwine stressed that RNA-binding proteins’ ability to modulate the transport and distribution of particular mRNAs will make RNA-binding proteins important in understanding neurodegenerative diseases. One example exists in their role in Fragile-X, the most common form of mental retardation, and Eberwine recommended that researchers look for differences in the functioning of these proteins in Alzheimer’s.
Paul Worley of Johns Hopkins School of Medicine in Baltimore, Maryland, reviewed the function of three immediate early genes that function at the excitatory synapse. Narp (neuronal activity-regulated protein) is made and secreted into the excitatory hippocampal synapse, peaking at about 16 hours after strong stimulation (typically, immediate early proteins peak at two hours). Narp expressed in one cell can cluster AMPA receptors on a contacted cell. In this mechanism of synaptic plasticity, narp can act as a presynaptic aggregator for postsynaptic receptors. In this sense, narp is a cousin of agrin, a molecule well-studied for its key role in the clustering acetylcholine receptors on the developing neuromuscular junction. Similarly, NP1 is another immediate early gene that induces AMPA receptor clusters. NP1 and narp can dimerize and form mixed hexamers that can cluster AMPA receptors more stably.
A member of a rapidly growing family, the immediate early homer plays a role in the calcium dynamics and G-protein signaling of the metabotropic glutamate receptor (mGluR). Together with arc, homer is the most dynamic of all known immediate early genes in neurons, Worley said. Homer is both a scaffolding protein and a catalyst of receptor function. It has an EVH1 domain, much like mena, Wasp, and other proteins known to be able to change the assembly of the actin cytoskeleton that runs just underneath the dendritic membrane. Homer also is able to crosslink the mGluR receptor with the endoplasmic reticulum, meaning that this organelle comes into very close contact with the postsynaptic density, extending right to the top of the dendritic spines.
Interestingly for Alzheimer’s, perhaps, introducing homer into dendrites of cultured neurons produced a receptor clustering that dramatically increased calcium influx in response to stimulation, Worley said. Intracellular calcium signaling and homeostasis are known to play a role in APP processing (see further below).
Regarding the immediate early gene arc, Worley said that it truly is a postsynaptic density protein that may be a chaperone or scaffold for other proteins bearing an SH3 domain. In addition to the ERK-mediated transcription Steward described, Worley said his lab detects arc transcription as early as three minutes after a mouse is beginning to explore a new cage.
While in general, immediate early genes are thought to enhance synaptic transmission, there are also indications that they can dampen transmission following a large stimulus. How these roles relate to neurodegeneration is still unclear, Worley added.
Howard Federoff of the University of Rochester, New York, presented the newest results of his ongoing studies of NGF. Federoff said NGF is much more than a developmental survival factor for neurons. He believes it is a key regulator of synaptic plasticity, in part because it is present on the postsynaptic side, its expression is regulated by activity, and its release from hippocampal neurons and dendrites is controlled by depolarization.
Federoff’s lab created inducible NGF transgenics. The mice develop normally while the transgene is dormant; then the researchers can express it in any brain region by injecting a Herpes simplex-derived activator. Federoff’s group turned the NGF on in the hippocampus of young adult mice to study its effect on cholinergic input into the hippocampus from the basal forebrain. This septohippocampal projection in the mouse corresponds to the human perforant pathway going from layer two of the entorhinal cortex to the molecular layer of the dentate gyrus in the hippocampus, which is involved in learning and degenerates early in Alzheimer’s. It turned out that mice with gain of NGF function in their hippocampus are better at special learning than are controls. (Incidentally, spatial learning is the behavioral category that generally shows the most robust results, i.e., decline, in APP-transgenic mice, as well.) The mice had extensive spatial reorganization of the cholinergic fibers into the hippocampus, and the extra NGF further strengthened this effect.
Federoff also reported that his group ran microarray tests of gene activation in the NGF-enhanced hippocampus compared to the control side, and of the septum area that projects to the hippocampus. (The NGF released by hippocampal neurons is taken up and transported back to the septal neurons.) He is currently analyzing genes whose expression changed; these include signaling molecules, synaptic proteins, ion channels, and transcription factors.
Finally, Federoff described some unpublished work on NGF expression in cultured primary neurons. Almost all of the NGF they saw was in the postsynaptic density. Apparently, the neuron maintains two pools of NGF. A constitutive pool in the soma may be the one that maintains survival, and a regulated pool in the dendrites may be the one that responds to activity by the presynapse. When NGF levels rise in the neuron, it re-apportions the relative amounts in those pools.
Mark Bear of Brown University, Rhode Island, presented work on long-term depression (LTD), a negative form of synaptic plasticity usually induced by stimulation of NMDA receptors. In LTD, the expression of AMPA receptors and their number in the synaptic membrane declines. Protein kinase A, the phosphatase 2B, and the anchoring protein AKAP 150 mediate this loss and could be checked for changes in Alzheimer’s, Bear said.
In a demonstration of what "Use It or Lose It" means in molecular terms, Bear also reported on work with monocular deprivation. Traditionally, shutting off sensory input from one eye has been used to study synaptic pruning and activity-based refining of connections in the visual cortex during a particular sensitive period in postnatal development. Temporarily covering one eye is also used to help correct certain eye asymmetries in children. Here, Bear asks if the known molecular mechanisms of LTD underlie changes at postnatal synapses that have fallen silent. When Bear analyzed synaptic currents in the visual cortex from rats kept with a closed eye, he found a signature typical of LTD. Not only were there 20 percent fewer synaptic AMPA receptors in slices cultured from rats after only 24 hours of deprivation, but also, the GluR1 subunit of AMPA receptors had a telltale dephosphorylation at serine 845. Previously, Bear had reported dephosphorylation of this site by cAMP-dependent PKA in in-vitro experiments with pharmacologically induced LTD.
What this might mean for Alzheimer’s is unclear. Bear said that while in postnatal life, LTD is the effect of insufficient activity, in aging it might occur in response to inappropriate activity. Coleman replied that the rapid time course seen in Bear’s study does not jibe with the presumably gradual decline of synapses in the degenerating brain; however, the mechanisms of LTD deserve exploration because it is especially prominent during two periods of life, namely postnatal development and aging.
Edward Ziff of New York University reported that palmitoylation of AMPA receptor-binding proteins is a newly appreciated mechanism by which these proteins control the trafficking of AMPA receptors between the synapse and other intracellular vesicle-like structures. He introduced PICK as an AMPA receptor-binding protein that may be involved in LTD by dissociating the Glur2 AMPA receptor from the membrane; a protein called NSF antagonizes this action. Ziff described other work with protein complexes called SNARE and SNAP, which determine the mobility of AMPA receptors.
Switching gears, Mike Ehlers of Duke University in Ithaca, New York, described recent research into how groups of proteins of the postsynaptic density change in response to synaptic activity. The talk visibly fascinated the audience with its implication of the proteasome as a force behind these changes, and not the lysosomal system, which has previously been implicated in synaptic protein turnover.
Ehlers said that while most of the components of the PSD are known, their relative abundance was not, nor was there a global understanding of the dynamic changes within them. He said the PSD is so highly plastic that many components reorganize within minutes of activity; lasting changes in activity alter the strength of the synapse, the number of receptors in it, as well as the size-and even appearance and disappearance-of the PSD.
To assess multiprotein changes, Ehlers developed a method to isolate, from cultured neurons, PSDs that were in a similar state of activity, and analyzed their protein content with gel density measurements. Following increased synaptic activity, he saw discrete groups of proteins go up and down in concentration, and the mirror image of that occurred following blocking. These changes reached a plateau after one day and reversed when the activating or blocking drugs were washed out. Overall protein content did not change.
To understand further some of the underlying mechanisms-in addition to the translational and transcriptional findings as described by Steward and Eberwine-Ehlers looked at protein turnover with pulse-chase experiments. He found that the proteins of the PSD are normally turned over every five hours; again, synaptic activity sped this up and blockade slowed it down.
What might be controlling this protein turnover? One candidate is the ubiquitin-proteasome system. It is not clear how it affects synapses, but it recently has been implicated in LTP and synapse development, Ehlers said. With fractionation experiments, he found that he could detect ubiquitin conjugation (the step marking proteins for degradation) in fractions containing synaptic proteins, and this ubiquitin labeling increased with synaptic activity.
These data, while preliminary, may open a new aspect for the role of the proteasome in neurodegenerative diseases. What if the proteasome slows down too much? Could that impair the ability of the postsynaptic side to meet the demands posed by signals coming from the presynaptic side?
If indeed proteasome activity declines in the run-up to AD, one protein that might accumulate is tau. Eva-Maria Mandelkow of the Max-Planck Institute in Hamburg, Germany, updated her and her husband’s hypothesis that tau’s role in AD pathogenesis might derive from its blockade of axonal transport, starving the synapse of vital ingredients including APP (see previous news story). Even very minor elevations of tau concentration, over time, might cause neurodegeneration in this way, Mandelkow said.
Jesus Avila of University of Madrid, Spain, described a conditional transgenic mouse in which overexpression of GSK3, one of numerous kinases known to phosphorylate tau, can be induced with tetracycline. In these mice, tau was hyperphosphorylated, but tangles did not form. Avila suggested that the cytoskeleton would change as hyperphosphorylated tau is no longer able to stabilize microtubuli.
This concludes the first day of presentations delivered at the Neuroscience Conference in Orlando. Please see Day 2 summary.-Gabrielle Strobel.