The conventional assumption is that the amyloid plaques cause the problems in amyloidoses. However, just because they are the visible correlate of the disease process, they need not be the real problem. It could be that they are simply a way for the cell to dispose of oligomeric beta-helix proteins (plaque precursors), which are the real culprit. Indeed, the failure of clinical tests of antibodies against amyloid plaques would point in this direction: dissolving the plaques makes the disease worse.

Another obvious problem with your approach is that you need to ensure specificity of the protease, it wouldn't do to randomly cleave brain proteins!

I have reviewed amyloidoses in doi:10.1007/978-3-319-19920-7_10

Thanks for the prompt answer! Precisely based on the failure of antibodies I stay with that conventional wisdom about the role of the plaques. My take: antibodies bind the amyloid peptide (be it monomeric, oligomeric or further aggregated), but in fact fail to solubilize the already existing, insoluble aggregates - if I remember correctly, this was the conclusion from a recent, failed clinical trial of an antibody - it didn't manage to reduce the plaque load.

So, back to the idea of solubilizing the existing plaques, rather than just preventing them from forming: it's either devising hydrophobic molecules capable of permeating the aggregates and disrupting their structure, or something else. This something else could be a "protease" cutting the amyloid peptide's backbone (within the aggregate!), making the hydrophobic properties of the fragments insufficient to keep the structure (PDB 2MXU, Xiao et al.) organized.

Specificity is a derivative of structure-based de novo design. I already fairly routinely design specific ligands. The tempting, logical next step is to design ligands that also have properly arranged catalytic groups. The amyloid peptide is not the easiest target, but definitely an important one.

One idea that has been discussed is to design small molecules that prevent the association of beta-strands into the helix and hence oligomerisation of amyloid proteins. Such compounds would have to form hydrogen bonds exactly where the interactions between strands are.

Alternatively, small molecules might stabilise the native of the prion-state (pharmacoperons). However, I am not aware of any great progress with either method.

In the specific case of Alzheimer disease, one might also think of inhibiting secretase beta, and thus preventing amyloidogenic peptides from being formed in the first place. This, of course, would not work with the other protein folding diseases.

All these things can prevent the plaques from being formed, but not necessarily deal with existing ones. There is an uncertainty in regard to aducanumab's ability to actually clear the plaque. Indisputably, it failed to achieve a meaningful clinical outcome, but are we certain about the mechanism of the failure? Dr. Golde elsewhere makes a convincing point about the validity of the correlation of the PET signal with actual reduction of the plaque. https://www.alzforum.org/news/research-news/biogeneisai-halt-phase-3-aducanumab-trials .

It is possible that all these antibodies fail because they after all fail to clear mature, insoluble aggregates. The bottom line: we have reasonable prospects of developing therapies that slow the progress of AD, but curing advanced stages requires a completely different approach.

In general, it is far better to prevent things from happening in the first place than to wipe the blood from the floor afterwards. This is particularly true for neuronal diseases, even if the regenerative ability of the brain may be larger than I was taught it to be when I was a student.

And another thought: soluble precursors of amyloid plaques and tangles have a much larger chance to interact with something important (and break it) than the insoluble precipitates. If I had to bet, I'd put my stake on them, just like Alzheimer did in 1911 (English translation in doi:10.1177/0957154X9100200505).