γ-secretase inhibitors (GSIs) are a class of drug candidates that have been tested (and largely failed) as Alzheimer’s therapeutics. Steve DeArmond, of UCSF, has also spent some time researching their potential as prion disease drugs, though surprisingly, their mechanism may different in prion disease from that in Alzheimer’s. So far, the results don’t suggest that this class of drugs is likely to ever reach the clinic for Creutzfeldt-Jakob disease, but I still wanted to understand the science behind this idea, so I did a bit of reading.

γ-secretase is a proteolytic complex made of PSEN1, PSEN2 and a few other membrane proteins, that cleaves a number of different proteins including APP. The cleavage of APP by γ-secretase and then by β-secretase generates amyloid beta, and both of these events have been seen as potential drug targets for Alzheimer’s [De Strooper 2010] – if you could inhibit either of them, you’d expect to reduce the amount of amyloid beta being produced, and in so doing, hopefully slow down the amyloid cascade. As of 2010, there were four clinical trials in progress of γ-secretase inhibitors for Alzheimer’s [Table 1]. One of the earliest GSIs, Lilly’s semagacestat (formerly known as LY450139) not only showed no effect on cognitive decline, but also led to a number of adverse events, leading to the clinical trial’s early termination [Doody 2013 (ft)].

One reason for the adverse events is that γ-secretase has more targets than just APP. Among them are the Notch proteins (NOTCH1-4) – γ-secretase cleavage of these generates the Notch intracellular domain (NICD), which moves into the nucleus and controls transcription of a number of genes, helping to determine cell fate. Such is the Notch signalling pathway. Inhibiting γ-secretase altogether prevents the formation of NICD, which alters the fate of intestinal stem cells [van Es 2005 (ft)], which may be partly responsible for the adverse effects. In terms of Alzheimer’s therapeutics, this motivated a search for more specific GSIs which would prevent γ-cleavage of APP while leaving Notch untouched, with some success [Kounnas & Danks 2010].

With that background, here’s the relevance to prion disease. In the early 2000s, DeArmond and his colleagues in the Prusiner lab were looking to figure out why neurons undergo dendritic atrophy in prion disease. Because β-catenin is pro-dendrite growth and NICD is anti-dendrite growth, they wondered if the proximate molecular cause of dendritic atrophy might be a decrease in β-catenin or an increase in NICD. As it turns out, they found ample evidence for the latter [Ishikura & Clever 2005]: NICD was increased by 2-3x in the prion-infected mouse brain, and a similar increase could be seen even in ScN2a cells, which also have fewer and shorter “processes” than uninfected N2a cells. Treating ScN2a cells with siRNA against Notch1 rescued that phenotype and made them look more like their uninfected counterparts, growing longer and more numerous neurites.

So it seemed that prion infection somehow accelerated Notch-1 cleavage, leading to an elevation of NICD and promoting dendritic atrophy. This would make Notch a potential drug target in prion disease – one might expect targeting Notch to influence the downstream neurotoxicity, but not to influence prion propagation itself.

But when DeArmond and colleagues got around to doing an in vivo treatment study with GSIs, the picture was more complicated [Spilman 2008]. When they treated mice with a GSI called LY411575, starting at 50 days post-infection and then examining the mouse brains at 93 dpi, they found that the treatment reduced NICD levels but didn’t rescue the dendritic atrophy – the treated mice had about the same number of dendrites as untreated mice. Surprisingly, though, it did reduce PrPSc levels – not in the thalamus, where the mice were originally inoculated, but rather in the neocortex and thalamus. They then combined LY411575 with quinacrine, which can, at high enough concentrations, transiently reduce PrPSc levels before giving rise to drug-resistant prions [Ghaemmaghami & Ahn 2009]. LY411575 plus quinacrine had a much more pronounced effect: after 43 days of combined treatment, the level of PrPSc in the thalamus was reduced by about 50%, and the levels in the neocortex and hippocampus were reduced by 95% compared to untreated mice. On that 43-day timescale, the dendritic atrophy phenotype of prion-infected mice was completely rescued too: the treated mice had as many dendrites as uninfected mice. This effect resulted from some sort of synergy between quinacrine and the GSI: neither drug alone achieved such dramatic effects. One possible explanation is that the GSI inhibited the axonal transport of PrPSc, slowing the spread of prions throughout the brain. The relatively minor in vivo anti-prion effects of quinacrine were then magnified by this slowed spreading of the infection.

If it’s true that the GSI slowed down axonal transport of PrPSc then it’s not clear whether the molecular mechanism for this would stem from the inhibition of Notch-1 cleavage, or from preventing cleavage of some other target of γ-secretase. If Notch-1 is indeed the important target here, then it will be hard to figure GSIs into a drug cocktail for prion disease. The Spilman study didn’t evaluate the effects of the GSI + quinacrine cocktail on survival time, because the cocktail was so toxic that the mice had to be sacrificed before they died of prion disease. Much of that toxicity probably derives from Notch-1 inhibition. In the Alzheimer’s arena there has been some success in finding GSIs specific to APP cleavage that spare Notch-1 cleavage [Kounnas & Danks 2010], but that wouldn’t help for prion disease if the target actually is Notch-1 itself.

The situation may be analogous to last year’s report about PERK inhibitors for prion disease. In that study, the PERK inhibitor had a dramatic therapeutic effect for a short time window, but was is too toxic to even evaluate its effects on survival, and the toxicity and therapeutic benefit probably both derive from the drug hitting the same target. In other words, the problem isn’t a side effect, it’s a main effect.

The mechanism of GSI action in vivo in prion disease is not clear yet, so it could yet turn out to be something other than Notch-1. But even if GSIs never become part of a drug cocktail, they may prove a valuable tool to learn more about prion trafficking and neurotoxic mechanisms.