Skip to main content

Alzheimer's Disease

Gamma-Secretase Inhibitors: The Wrong Way Around?

Here’s an Alzheimer’s mystery, and that area was not lacking for them before. A paper in Neuron from Harvard/Mass General/Brigham & Women’s says that we may have been thinking about gamma-secretase the wrong way around. (Here’s a press release from Harvard, although I believe the paper is open-access as well).
That enzyme has been targeted for a long time now, since it’s thought to be involved in the proteolytic processing that generates beta-amyloid, a hallmark of Alzheimer’s. I myself was involved in a program to target the secretase enzymes (the other end is clipped by beta-secretase) even before anyone was sure what enzymes they were. (You can imagine how far that one got). But both of them have been clearly identified since back in the 1990s, and the amount of work that’s gone into finding inhibitors is truly mind-boggling.
Eli Lilly got a gamma-secretase inhibitor well into the clinic, before finding out that it was not only ineffective, but that it actually seemed to make the disease slightly worse. There have been several attempts to explain that result – the compound wasn’t potent enough, it didn’t to the right things to gamma-secretase, or it was the side effects for Notch signaling that caused the trouble. (When it was learned that gamma-secretase, as part of a complex called presenilin, also processes the Notch protein, a lot of people were understandably worried about what an inhibitor might do on the side). But this new paper suggests another possibility, and it’s one that I’ll bet not many people would have guessed.
The authors have been tracking presenilin mutations for some time. These are the most common cause of familial Alzheimer’s, a rare but biologically revealing form of the disease. Many people had assumed that such mutations somehow turned gamma-secretase loose to do its amyloid-forming work earlier and more thoroughly, but Jie Shen and Raymond Kelleher’s work had been pointing at the opposite conclusion: that these mutations inactivated gamma-secretase, and that this inactivation was a risk factor for Alzheimer’s. Now in this new paper, they’ve generated mice with two of the most common such mutations, and they find that (1) their gamma-secretase activity is indeed impaired, and (2) they show Alzheimer’s-type pathology in their brains – in fact, some of the most human-like pathology that has yet been seen in a rodent model. The total amount of amyloid protein is decreased, but the proportion of 42-amino-acid beta amyloid, long thought to be the dangerous species, is increased.
If all this does hang together, then a gamma-secretase inhibitor like Lilly’s is the last thing you’d want to give Alzheimer’s patients. And it might well make them worse, as was seen in the clinic. There are several other such compounds in development, as it happens, and this latest paper should give everyone involved a lot to think about. . .

39 comments on “Gamma-Secretase Inhibitors: The Wrong Way Around?”

  1. Ann o Mouse says:

    This is actually good news for the amyloid hypothesis, although clearly bad news for gamma secretase inhibitors. It is interesting that these mutations increased the ab42/ab40 ratio and increased deposition while decreasing overall gs activity. In principle one could identify a gamma secretase inhibitor or modulator that did the same, but the risk of failure for this approach just increased considerably.
    Bate set.

  2. Luigi says:

    Knowledge related to g-secretase advances backwards underlining Lilly’s extreme folly. In addition to the debate as to how g-secretase may be involved in AD, a paper by De Strooper (Lessons from a failed γ-secretase Alzheimer trial. Cell. 2014 Nov 6;159(4):721-6) reviewed what went wrong in the Phase III semgacestat trials noting that the preclinical evaluation of the compound was superficial, the “Knowledge gap in g-secretase biology, physiology, and structural biology [being] appallingly wide”. This echoes Karran and Hardy’s paper (A critique of the drug discovery and phase 3 clinical programs targeting the amyloid hypothesis for Alzheimer disease. Annal Neurol 2014. 76. 185-205) where review of 6 failed AD compounds came to the conclusion that there were major issues in the rigor of preclinical studies.

  3. a. nonymaus says:

    So, is there anything that increases gamma-secretase activity? Is there anything that degrades ab once it’s processed by the secretases? I guess the way to go is through altering expression levels of these enzymes since I’m guessing that delivering a recombinant protease through the bbb is a sticky wicket.

  4. Luysii says:

    There is a large caveat about all work of this sort. The following is a (slightly modified) direct quote from the editorial on this paper in the same issue of Neuron (vol. 85 pp. 893 – 895 ’15)
    “In order to see the neurodegenerative phenotype resulting from knocking in the L435F PS1 (Presenilin1) mutation, the authors eliminated ALL wild-type PS (Presenilin) alleles (PS1 and PS2).
    Human carriers of FAD (Familial Alzheimer Disease)-linked PS1 mutations have one intact copy of PS1 in addition to two (intact) copies of PS2 (Presenilin 2).
    However, such “genetic exaggeration” is routinely done in modeling human diseases in mice; for example, APP (Amyloid Precursor Protein) transgenic mice frequently overexpress several-fold human APP containing one or more FAD mutations. It is possible that such genetic exaggeration represents an acceleration of phenomena that in humans result from the accumulation of small insults over years or decades, but this is hard to validate in humans.”

  5. Ann O Mouse says:

    @3. a. nonymaus
    Blindly increasing GS activity is probably as ill-advised as blindly decreasing it. With all the caveats stated by @4, the model is very interesting, and I would look carefully (although very cautiously) at the AB42/Ab40 ratio.
    Second worm.

  6. Anonymous says:

    In before Lane comes flying in.

  7. bank says:

    @ Ann o Mouse,
    With regards to the ratio of Ab42 increasing, it is worth noting that in late onset AD, both Ab42 and Ab40 go *down* with disease progression, it is just that Ab40 goes down more then Ab42. So I do not think this data supports the amyloid hypothesis.

  8. Lane Simonian says:

    #6 If I were up earlier (Pacific Daylight Time), I would have been here even earlier. I was hoping Derek would post this study.
    Presenilin have been tabbed as the catalytic subunit of the gamma secretase assuming that they were part of gamma secretases themselves, but they actually work solely or primarily upstream of the gamma secretase. Presenilins activate g protein-coupled receptors which results in intracellular calcium release which leads to the activation of gamma secretases. (see last paragraph in section plot thickens).
    They also activate protein kinase C which can lead to phosphatidylinositol 3-kinase/Akt activation–which is a neuroprotective pathway, but when this pathway is cut off, the result is the production of peroxynitrites.
    Most genetic mutations tied to Alzheimer’s disease overactivate g protein-coupled receptors (amyloid precursor protein mutations and presenilin-2 mutations). On the other hand, presenilin-1 mutations due to a leucine substitution out decreases g protein-coupled receptor activation which would lead to a decrease in gamma secretase activity. On the other hand, the leucine substitution out also prevents the activation of the phosphatidylinositol-3 kinase/Akt pathway. This would increase amyloid production. Peroxynitrites via nitration also cut off the phosphatidylinositol-3 kinase/Akt pathway.
    Gamma secretase inhibitors actually make things worse because they prevent the conversion of the c terminal fragment of the amyloid precursor protein into amyloid oligomers. The c terminal fragment initially increases peroxynitrite production via activation of g-protein coupled receptors. Amyloid oligomers on the other hand increase the production of hydrogen peroxide. The latter is a less destructive oxidant in Alzheimer’s disease than the former and its production ends with the formation of amyloid plaques.

  9. annon too says:

    Science trudges on, often in misdirections, but then self-corrects.

  10. Lane Simonian says:

    The other big news (albeit old news) of last week was that uric acid–a peroxynitrite scavenger–decreased the risk for Alzheimer’s disease by 24 percent. Here is the effect of uric acid on the damage resulting from presenilin-1 gene mutations.
    It just happens that the pathways leading to the formation of peroxynitrites and amyloid overlap, but not completely as one can have peroxynitrite formation without intracellular calcium release whereas the same is not true of amyloid (as a side point peroxynitrites increase both beta and gamma secretase activity). So it is possible to have Alzheimer’s disease with little or no amyloid.
    More importantly, none of the clinical trials targeting either amyloid plaques or oligomers have worked so far whereas all the trials using highly effective peroxynitrite scavengers have worked. The evidence that Alzheimer’s disease is caused by peroxynitrites and not by amyloid still is short of being overwhelming, but the evidence keeps moving in that direction.

  11. Lane Simonian says:

    One last thought for now presenilin-1 gene mutations decrease gamma secretase activity by reducing g protein-coupled receptor activity but they increase the substrate (phosphatidylinositol 4,5 triphosphate) for the enzyme that likely triggers Alzheimer’s disease (phospholipase C). So it is possible to have a decrease in gamma secretase activity but an increase in amyloid42 production.

  12. Anon says:

    So gamma secretase modulators could still be worth while? Instead of inhibiting the activity of gamma secretase, these modulate the place at which it ‘snips’ making it favor producing Abeta 40 or 38 rather than the longer (and more prone to aggregation) Abeta 42.
    I worked on a gamma secreatse modulator programme about 10 years ago, I left that company and haven’t followed the AD field, did any modulators ever make it to the clinic?

  13. Ann o mouse says:

    Bait taken, like shootin fish in a barrel
    But beyond that,
    @7 bank. Point taken, and that seems to be what was observed in this model as well. The interesting thing is that there was increased plaque deposition and, most importantly, an AD-like phenotype in these mice, unlike earlier models, despite the decrease in overall abeta levels. Also, I suspect that you are referring to CSF abeta levels which, if there is a defect in clearance or excess deposition rather than clearance, would be expected to go down. As always, the data can be interpreted ( or misinterpreted) according to your bias.
    Fresh bait.

  14. bank says:

    @Ann o mouse,
    Agreed that the levels of Abeta are best known in the CSF, however, if decreased clearance of Abeta was responsible for decreased CSF Abeta, one would expect it to increase in the brain parenchyma. However, it does not, instead it plateaus.
    To reconcile these, yet keep the amyloid hypothesis, one could propose that decreased Abeta levels were due to a decreased number of neurons, which, somehow, also change the ratio of Ab40 to 42 they produce.
    In the present paper, though, I was disappointed to not find any measurements of tau phosphorylation or localization. In the absence of a tau phenotype I would hesitate to claim that it faithfully recapitulated AD.

  15. Kazoo Chemist says:

    @10 Is there any data on the incidence of Alzheimer’s in people with gout?

  16. Ann o mouse says:

    14 bank
    You are absolutely correct, it is still not an accurate model of AD, but it is progress. As for your other point, at the risk of showing my ignorance, would the extracellular location of plaques ( as opposed to more soluble abeta) qualify as parenchyma?
    Agreed, a tau measurement would have been interesting.

  17. Cellsman says:

    Used to work on a GS modulator program too, we never managed to get rid of severe preclinical tox and it never made the clinic.
    @Lane,if presenilin (plus Nicastrin plus Pen2)is not the gamma-secretase, how do you explain the data where people reconstituted the complex biochemically and demonstrated activity? Also, these mysterious GPCRs you talk about, these beasts are activated by extracellular ligands. So what are these GPCRs and what are their ligands. Also, you confuse me with the C-terminal APP fragment – this never leaves the cell, at one point was thought to function as transcription factor akin to Notch.
    I left this field for more fertile grounds 10 years ago, and looking from the outside there seems not to have been much progress at all.

  18. Anon says:

    @10. ” …all the trials using highly effective peroxynitrite scavengers have worked”
    I am having trouble finding verifiable support for this statement. All I come up with is your various posts repeating the point but no references to real clininal trials. In general, antioxidant therapy has not shown benefit in clinical trials for anything that I can find, although there is some epidemiological support. Could you help me out with a couple of references ( not a lecture…I get your point, just looking for references, preferably more than one).

  19. YetaB says:

    @3 – Is there anything that degrades ab once it’s processed by the secretases? Obviously, or the world would be made up of amyloid peptide by now 🙂
    Neprilysin, IDE…
    The key might well be – as my old biochemistry teachers kept saying, and I wish I had listened more – understanding the flux and control points in the pathway that produces whatever the toxic species is. It seems at least plausible that the evidence for BACE as a control point for Ab42 production remains strong.

  20. bank says:

    Ann o mouse,
    In the context of AD, parenchymal amyloid usually refers to “typical” insoluble amyloid plaques, in contrast to insoluble amyloid associated with the vasculature (as in cerebral amyloid angiopathy). At least that is what I intended it to mean…
    PiB scans on AD brains show an increase in signal (~amyloid) in the prodromal phases of the disease, i.e. before significant cognitive decline. However, the signal does not keep increasing as cognition worsens, yet Abeta in the CSF declines.

  21. Anon says:

    We should all hope for a clean and unequivocal result from BACE inhibitors now in clinical trials, whatever that result may be.

  22. Ann O Mouse says:

    Ah, yes, I believe I have heard this before. Thank you for jump starting my memory in an area well outside my expertise. I could argue that abeta production should go up in the prodromal period then decline as neurons die and cognition declines, but then I would really be hanging over the edge of my knowledge. I’ll leave it at that, and say thanks for the discussion.

  23. YetaB says:

    @20 – The evidence is that PIB (or other PET ligand) accumulation and CSF Ab42 decline are well correlated. They would seem to be effectively equivalent measures, and both stabilise at around the same time in disease progression.

  24. Worth pointing out that the behavioral effects in fig 5 (a key part of the paper) are pretty small.

  25. Lane Simonian says:

    @15 Unfortunately this is the best answer I can provide at this point:
    The researchers identified 309 new cases of Alzheimer’s disease among 59,224 patients with gout and 1,942 cases among 238,805 people in the comparison group of patients who did not suffer from gout.
    I believe this includes gout with co-morbidities, so the incidence of Alzheimer’s disease in people with gout but without co-morbidities could be considerably less.
    # 17 The question is whether presenilin’s directly or indirectly activate gamma secretases or most likely both. The link above poses this question.
    However, it is the link between γ-secretase cleavage and intracellular Ca2+ stores which is the most intriguing aspect. Cheung et al. [50] found that PS mutant-induced enhancement of Aβ secretion can be abolished by IP3 receptor knockout, indicating that γ-secretase activity is controlled downstream of PS by IP3. How this finding relates to the hypothesis that PS is the catalytic subunit of the γ-secretase needs to be explored further.
    In either case, presenilin-1 gene mutations do not represent a complete loss of function or no amyloid would be produced in people with the presenilin-1 gene mutations.
    I may have misunderstood the mechanism regarding leucine and g protein-coupled receptors. This research suggests that leucine plays a role in keeping g protein-coupled receptors in an inactive state.
    Here is the title for the one linking the c terminal fragment to the activation of g-proteins, but I am still looking for the exact mechanism.
    Amyloid precursor protein carboxy-terminal fragments modulate G-proteins and adenylate cyclase activity in Alzheimer’s disease brain.
    I am going to answer #18 separately because it is the really critical question.

  26. Lane Simonian says:

    The list of clinical trials in which peroxynitrite scavengers have partially reversed Alzheimer’s disease (the first three trials involve polyphenolic antioxidants; the rest involve what are the best peroxynitrite scavengers–methoxyphenols).
    Melissa officinalis extract in the treatment of patients with mild to moderate Alzheimer’s disease: a double blind, randomised, placebo controlled trial
    Salvia officinalis extract in the treatment of patients with mild to moderate Alzheimer’s disease: a double blind, randomized and placebo-controlled trial.
    Saffron in the treatment of patients with mild to moderate Alzheimer’s disease: a 16-week, randomized and placebo-controlled trial.
    Effect of ferulic acid and Angelica archangelica extract on behavioral and psychological symptoms of dementia in frontotemporal lobar degeneration and dementia with Lewy bodies.
    Effects of ferulic acid and Angelica archangelica extract (Feruguard) in patients with Alzheimer’s disease
    Effect of aromatherapy on patients with Alzheimer’s disease
    Improvement of Cognitive Deficit in Alzheimer’s Disease Patients by Long Term Treatment with Korean Red Ginseng
    Heat-processed ginseng enhances the cognitive function in patients with moderately severe Alzheimer’s disease
    I will try not to lecture, but I will try to avoid the standard criticisms: the trials were not well-designed (some of them were well-designed and some were not), they were too small (they were all small but each produced similar results), it was the placebo effect (hard to make this argument in double-blinded studies), and I am cherry picking (yes, I picked the clinical trials in which there was a positive effect).
    Once I see an effective treatment for Alzheimer’s disease that does not involve a peroxynitrite scavenger than I can alter this hypothesis.

  27. Anonymous says:

    You are likely recollecting the study by Bateman et al, “Clinical and Biomarker Changes in Dominantly Inherited Alzheimer’s Disease”, NEJM v.367 p 780.
    Here is the summary figure: i(dot)imgur(dot)com/knOOeex.png
    Where indeed amyloid and CSF ab42 stabilize. However, these cases are early onset AD, with mutations in PS1, PS2 or APP.
    My recollection is that CSF Ab42 continues to decline after onset of late onset AD. After 30 mins of searching, I can’t find a longitudinal study with a decent N-number though — do you know of one?

  28. Anon says:

    Thank you for this list. While it would have been nice if you had actually provided references rather than titles, I was able to track these down. I will try not to lecture, too, and simply say that all the criticisms of these studies you cite are absolutely valid. I say that as one who sees merit in your position, but these studies are just an awful way to make your point. It may sound good on general AD advocacy sites, but that and a $1.50 will buy you a cup of coffee (and not particularly good coffee). Find someone who is willing to do a proper trial, and please don’t give me the excuse that no one will do this for a “supplement”…that is bullshit. If the science is strong someone will do the trial. In the meantime, continue to advocate but be open about what you know and what you just believe. OK, I’m lecturing. Good luck.

  29. Lane Simonian says:

    @28 I am working on it, but even people interested in this avenue have studies lined up before the ones that I have suggested. Despite the limitations of these clinical trials, I would still state without much hesitation that peroxynitrite scavengers are the only compounds that have effectively treated Alzheimer’s disease.

  30. Kazoo Chemist says:

    Thanks for the response. It just seemed like a great opportunity for meta-analysis of the data. Do high levels of uric acid protect?

  31. YetaB says:

    @27 – you are correct, almost all the data is cross-sectional. There are even fewer longitudinal studies with tau, if I recall.

  32. anony-mous(e) says:

    Should we be thinking about beta-secretase differently in view of the (gamma secretase) hypothesis? Sorry, not an AD expert

  33. kk says:
    “Via aromatherapy the methoxyphenols are inhaled directly into the hippocampus and over a period of months and years can reverse at least part of the damage caused by peroxynitrites”
    Is that believable?

  34. Ann O Mouse says:

    @32- In answer to your question about BACE and whether we should think about it differently, the answer is YES if you believe this study calls into question the entire amyloid hypothesis and NO if you think this study specifically relates to gamma secretase and the impact of presenilin mutations on amyloid protein and plaque production. I fall in the latter category, but I think we really need to have a good result from a BACE inhibitor in the clinic (or maybe in the animal model from this study) to know for sure.

  35. Anonymous says:

    @30. You are welcome. I agree that this presents an opportunity.
    @33. I based this statement on a variety of studies, including several small-scale clinical trials and on the eight years that my sisters and I took care of our mother with Alzheimer’s disease (five years using aromatherapy).
    @32 and Ann O Mouse (one of the best names here and also someone who gives gracious replies even when disagreeing with another poster). You don’t have to believe in the amyloid hypothesis to think that BACE inhibitors may be of some benefit. Gamma secretase inhibitors may have made the disease worse because they led to higher levels of the c terminal fragment which is probably more toxic than amyloid oligomers. Some researchers believe in a reverse amyloid cascade where amyloid plaques (which produce neither peroxynitrites nor hydrogen peroxide) are less toxic than amyloid oligomers (that help produce hydrogen peroxide) are less toxic than the c terminal fragments of the amyloid precursor protein which increases peroxynitrite production. Gamma secretase inhibiton is almost dead for the treatment of Alzheimer’s disease, amyloid plaque antibodies are even closer to being dead, and amyloid oligomers are on life support. What remains left of the amyloid hypothesis is BACE inhibitors.
    The first BACE inhibitors have been withdrawn on the basis of side effects (liver, skin, and eye damage–the last two are likely the direct result of BACE inhibition).
    BACE inhibitors may help early on because they inhibit the c terminal fragment of the amyloid precursor protein which would limit peroxynitrite production. My guess at this point is that BACE inhibition may delay the initiation of Alzheimer’s disease and slow down it progression early on.

  36. Lane Simonian says:

    Some more evidence that the pathways to gamma and beta secretases are via g protein-coupled receptors and receptor tyrosine kinases (whether some of these receptors are more important than others remains to be seen).
    Cell Res. 2010 Feb;20(2):138-53. doi: 10.1038/cr.2010.3. Epub 2010 Jan 12.
    A GPCR/secretase complex regulates beta- and gamma-secretase specificity for Abeta production and contributes to AD pathogenesis.
    J Biol Chem. 2003 Mar 14;278(11):9290-7.
    Platelet-derived growth factor induces the beta-gamma-secretase-mediated cleavage of Alzheimer’s amyloid precursor protein through a Src-Rac-dependent pathway.
    Gianni D1, Zambrano N, Bimonte M, Minopoli G, Mercken L, Talamo F, Scaloni A, Russo T.
    But the main problem is not that these receptors increase beta and gamma secretases. The main problem is that they increase the formation of peroxynitrites via phospholipase C, protein kinase C, and p38 MAPK.
    When you inhibit the gamma secretase via presenilin inhibition you cut off the neuroprotective phosphatidylinositol 3-kinase and you increase the levels of c terminal fragments of the amyloid precursor protein which in turn increases g protein activity, the production of peroxynitrites, the activation of caspases, the activation of beta secretase.
    Like everyone else, I am waiting for the results from the ongoing trials with beta secretase inhibitors. My guess is that they are too far down the line to do much good, other than possibly slightly delaying the onset and the speed of the progression of the disease. I would certainly be pleased if the results were better than this.

  37. Douglas Kell says:

    There is a literature A MILE long (going back to Alois Alzheimer) to the effect that Abeta, tau, neurofibrillary tangles and the rest are simply NOT correlated with cognitive decline (which is what matters, though harder to measure than a protease activity). The first paper in the Lancet showing what *is* the case was in 1999: Crapper McLachlan, D. R., Dalton, A. J., Kruck, T. P., Bell, M. Y., Smith, W. L., Kalow, W. & Andrews, D. F. (1991). Intramuscular desferrioxamine in patients with Alzheimer’s disease. Lancet 337, 1304-8. Better to look under a lamppost that has some light and electricity (IMHO, of course). Others at

  38. Lane Simonian says:

    From Douglas Kell’s citation:
    We conclude that sustained administration of desferrioxamine may slow the clinical progression of the dementia associated with AD.
    And you should be able to guess it by now, desferrioxamine is a peroxynitrite scavenger:
    “The iron chelator deferoxamine is also scavenger of peroxynitrite.”
    A person has to be very wedded to a particular hypothesis (amyloid plaques/oligomers or tau tangles) to ignore all evidence to the contrary.

  39. Miguel Rodriguez Manotas says:

    You can see at your own page a comment that I posted 3 years ago:
    “33. Rolling Stones on January 9, 2012 4:41 AM writes…
    It can not be as simple as: “amyloid-beta² being the central part of the AD mechanism”.
    The inhibitors and modulators of the secretases should work as using cyanide to inhibit the respiratory chain at mitochondria to avoid the production of CO2 that is accumulated in excess in patients with chronic obstructive pulmonary disease. How about the normal role of the enzymes that we are trying to inhibit, in order to decrease the Abeta production?”
    Now the paradigm has changed (fortunatelly) and is time to search for drugs that should enhance the enzyme activity.

Comments are closed.