APP and β- and γ-secretases are found on both sides of the synapse, but no study has directly addressed whether both pre- and post-synaptic Aβ is detrimental, said Malinow. First author Wei Wei and colleagues used viruses to overexpress APP in just a few pre- or post-synaptic neurons in hippocampal slices and then examined how the overexpression affected synapses of infected neurons and nearby normal ones. They co-transfected the slices with enhanced green fluorescent protein (EGFP) to enable them to measure cell morphology by two-photon microscopy, and identified APP overexpressing cells by co-expressing the red fluorescent protein dsRed.

Wei and colleagues first looked at CA1 neurons, which receive projections from neurons in the CA3 region. Infecting the CA1 with low-titre APP viruses so that only a few cells would be affected, the researchers found that spine density dropped by about 40 percent in the dendrites of cells that overexpressed APP and also in neighboring cells that did not overexpress the precursor. The neighbor effect extended to cells within 10 microns of the cells overexpressing APP. The loss of spines seemed to depend on production of Aβ, since treating the hippocampal slices with a γ-secretase inhibitor prevented it.

Having established that dendritic Aβ could be toxic to spines, the researchers then turned to axonal Aβ. They expressed APP and dsRed in the CA3 projection cells and EGFP throughout the CA1. Looking at the latter, they again found a reduction in spine density (about 30 percent) in those dendrites adjacent to APP overexpressing axons, and again the losses could be mitigated by treatment with a γ-secretase inhibitor.

The work suggests that Aβ can be damaging to spines whether it is made in pre- or post-synaptic cells. Interestingly, there is increasing evidence suggesting that interstitial Aβ levels are linked to synaptic activity (see ARF related news story). This could be taken as a hint that Aβ secretion is tied to synaptic vesicle release, but the present work suggests that post-synaptic production of Aβ is equally detrimental, though it is not clear why. It is possible that Aβ produced in pre- and post-synaptic cells eventually ends up in the same place. “We know that it can be secreted, though whether the secreted form is the one responsible for the effects is not totally clear,” said Malinow. “That is the most likely scenario, but the target itself is still controversial,” he said.

Wei and colleagues did find that APP-induced spine loss was linked to synaptic activity, since it could be prevented by a variety of neurotransmitter blockers including the sodium channel antagonist, tetrodotoxin (TTX), the NMDA receptor antagonist AP5, and α-bungarotoxin, an inhibitor of α7-nicotinic acetylcholine receptors (a7nAChRs). The NMDA antagonist seems to work by blocking the action of Aβ since it protected spines against a preparation of Aβ oligomers added to the hippocampal slices. TTX and α-bungarotoxin failed to protect against this exogenous Aβ, but when incubated with hippocampal slices overnight, they did reduce secretion of endogenous Aβ. The findings support the emerging idea that Aβ release is tied into synaptic activity and that NMDA receptors mediate the toxic effects of the peptide (see ARF related news story). The effects of α-bungarotoxin are particularly germane to AD, since the most common treatment for the disease at present is to block acetylcholinesterase and increase levels of acetylcholine in the brain.

“Our results suggest that this may lead to increased Aβ secretion and would therefore be detrimental,” suggest the authors. Though acetylcholinesterase inhibitors clearly help prevent memory loss in AD patients, the researchers question whether these drugs remain beneficial over the long term, noting that they fail to slow the progression of the disease (see ARF related news story on Raschetti et al., 2007).

If pre- and post-synaptic Aβ both decrease spine density, then what of the spines that remain? To test if they function normally, Wei and colleagues turned to chemically induced long-term potentiation (cLTP), a measure of synaptic plasticity. This procedure causes spines to enlarge, which happened less extensively in APP-transfected cells. This cLTP effect extended to spines on nearby non-APP overexpressing cells, which also failed to enlarge. “It seems that the sphere of influence [of overexpressing APP] is fairly local, around 10 microns, which is important to know,” suggested Malinow.

The researchers were able to restore normal spine enlargement by treating with a γ-secretase inhibitor, the NMDAR antagonist AP5, or pre-treating with α-bungarotoxin, all of which fits with the scenario that synaptic activity-driven production of Aβ causes spinal deficits via an interaction with NMDA receptors. Interestingly, blocking Aβ production or NMDA receptors for only 30 minutes before LTP induction was sufficient to restore normal spine enlargement. “This suggests that there is a critical window or vulnerable period for neurons, but it also suggests that there is a continuous production of Aβ and that that is what is blocking plasticity,” said Malinow. If you could prevent that continuous production, then you might be able to restore plasticity, he added. —Tom Fagan

Reference:
Lei W, Nguyen LN, Kessels HW, Hagiwara H, Sisodia S, Malinow R. Amyloid beta from axons and dendrites reduces local spine number and plasticity. Nature Neuroscience 2009. Abstract