According to the overflowing bathtub analogy, either a blocked drain (not enough clearance), an oversized faucet (too much production), or a combination, contributes to the accumulation of amyloid-β (Aβ) peptides in the Alzheimer disease (AD) brain. The solution to the problem, therefore, would be to unplug the drain or turn off the faucet. In this week’s Journal of Neuroscience, Jie Shen, Brigham and Women’s Hospital, Boston, and colleagues report that shutting off the spigot does help relieve symptoms in a mouse model of AD. Unfortunately, the relief seems only temporary.

The spigot in this case is γ-secretase, one of two enzymes necessary to clip Aβ from the larger Aβ precursor protein (APP). In 2001 (see ARF related news story), Shen and colleagues developed mice in which one of the essential components of γ-secretase, presenilin 1 (PS1), was conditionally “knocked out” (cKO). Because PS1 is essential for embryonic development and for the health and maintenance of adult non-neuronal tissues, Huakui Yu and colleagues designed these cKOs so that PS1 is ablated only postnatally and only in the pyramidal neurons of the cortex. Now, first author Carlos Saura at Shen’s lab, with help from colleagues at Gunnar Gouras’s lab at Cornell University, New York, and Richard Morris’s lab in Edinburgh, Scotland, has taken advantage of these mice to test how loss of PS1 helps mice suffering from AD-like pathology—he crossed the cKO PS1 mice with PDAPP transgenic mice that express human AβPP harboring the Swedish and Indiana mutations that cause familial AD (see Mucke et al., 2000.

When Saura and colleagues examined the PS1 cKO/PDAPP mice, they found that the levels of APP C-terminal fragments (CTFs) were dramatically higher than in the APP mice. This, plus failure to detect PS1 with anti-PS1 antibodies, confirmed the loss of the γ-secretase component, which helps to clear CTFs formed by the action of α- and β-secretases. The authors also found that the increase in CTFs, detected in mice as young as 2 months old, was age-dependent, getting worse as the animals got older, and that the fragments accumulated mostly in the presynaptic terminals.

In contrast, Saura and colleagues found that the age-dependent accumulation of Aβ normally seen in PDAPP animals was absent in mice that were also missing PS1. At 2, 6, and 17 months, Aβ42 was about 55, 90, and 99 percent lower in the PS1 cKO/PDAPP than in the PDAPP controls. When the authors tested for Aβ40, they found similar reductions.

These experiments confirmed that the spigot was indeed turned off—or at least down since the second presenilin, PS2, was presumably still active—and that the bathtub was no longer overflowing. But were the mice squeaky clean?

Cognitively, the mice lacking PS1 performed better than the control animals, suggesting that some of the neuronal deficits that retard PDAPP animals were absent. However, the improvement was only seen in very young mice. For example, when Saura and colleagues tested 3-month-old animals in a fear conditioning test, they found that the PDAPP animals performed significantly poorer than PS1 cKO/PDAPP or control wild-type mice. But by 6 months old, the PS1-negative mice, despite little accumulation of Aβ, were, in fact, performing worse than the PDAPP mice. A similar age-dependent rescue was found when the authors tested spatial memory using a water maze. At 3 months, the PS1 cKO/PDAPP mice performed as well as control wild-type animals, whereas the PDAPP mice were significantly weaker, but by 15-17 months, both PS1-negative and -positive PDAPP mice were performing the same and significantly poorer than wild-type animals.

To better understand the reasons for the memory impairments, the authors turned to electrophysiological measurements. When they measured neurotransmission in hippocampal slices isolated from the mice, they found that basal transmission, while normal at three months, was reduced by about 20 percent in both sets of transgenic animals by 6 months old. Also, long-term potentiation (LTP), or the ability of neurons to up the ante in response to repetitive stimuli and fire off stronger signals of their own, was compromised. In 6-month-old mice, LTP, which is required for learning and memory, was significantly reduced in PDAPP animals, which is to be expected given previous reports showing that Aβ can depress LTP (see ARF related news story). But Saura and colleagues also found that LTP was depressed to a similar degree in 6-month-old PS1 cKO/PDAPP mice, which do not accumulate the peptide.

Why the improvement in learning and memory deficits is not maintained throughout the lives of the PS1-negative transgenic mice is not clear, but the authors do suggest that “any beneficial effect on learning that results may be more than offset by other biochemical changes taking place in these older mice.” One obvious change in these animals is the accumulation of the AβPP C-terminal fragments at synapses. The failure to clear these fragments might turn out to be as bad for neurons as generation of Aβ.

So where does this leave γ-secretase inhibition as a therapeutic approach to treating AD? The data suggest that closing the spigot by inhibiting PS1 could, at the very least, provide temporary relief. Of course, in human AD cases, AβPP is not being overproduced as it is in the PDAPP animals, so the potential complication of a flood of CTFs might be less of an issue.—Tom Fagan


  1. It is very satisfying to see a totally independent confirmation of our work, especially when important conclusions are directly attached to it.

    After we identified PS1 as essential for γ-secretase activity (De Strooper et al., 1998) we all hoped it would be a—if not the—major therapeutic target in AD.

    But in 2002 we had to report that the neuron-specific knockout of PS1 did not rescue the cognitive defects of APP mice, despite the nearly complete elimination of plaque and vascular amyloid pathology in old APPxPS1(n-/-) mice (Dewachter et al., 2002). The outcome was a complete and major surprise for us, difficult to explain and impossible to get past the referees of more than one major journal…and a major blow to the therapeutic potential of γ-secretase inhibitors in AD.

    We believe that, despite the criticism on the non-physiological "total KO problem," the outcome of the paper of Saura et al., and of our 2002 paper, is as relevant now as it was then—and for more than one reason.

    Inhibition of PS1—or "modulation" if so preferred—will result in accumulation of CTF of APP and of a bunch of other transmembrane proteins. I asked in one of my previous comments on this site: Who is keeping tally on the substrates of γ-secretase? At least for the β-CTF (C99) of APP, we know they are potentially as neurotoxic as the amyloid peptides, and probably even more, since they remain attached to, and concentrated in the neurons in which they are produced. We do not know much about the (non-) physiological repercussions of remnants of other substrates of γ-secretase, but their accumulation can be safely predicted to be "not good for your brain."

    The prevention of formation of AICD by inhibition of γ-secretase can or should be added to the drawbacks, now even more than in 2002, given the most recent evidence that AICD actually regulates expression of neprilysin (Pardossi-Piquard et al., 2005), that, as we all know, is a major Aβ killer!

    Does that imply that the γ-secretase complex is off-bounds as a therapeutic target for AD as we advocated before (Dewachter and Van Leuven, 2002) based on our 2002 data? I believe so, but not being clairvoyant, I cannot but leave the question open. Nevertheless, the structural and functional complexity of γ-secretase, the inherent and not understood control of its activity and specificity, combined with the disparity of its substrates in neurons and in many (most?) other cell types in our body, is so overwhelming that finding weak spots or leap holes in its armor is a daunting task.


    . Neuronal deficiency of presenilin 1 inhibits amyloid plaque formation and corrects hippocampal long-term potentiation but not a cognitive defect of amyloid precursor protein [V717I] transgenic mice. J Neurosci. 2002 May 1;22(9):3445-53. PubMed.

    . Deficiency of presenilin-1 inhibits the normal cleavage of amyloid precursor protein. Nature. 1998 Jan 22;391(6665):387-90. PubMed.

    . Secretases as targets for the treatment of Alzheimer's disease: the prospects. Lancet Neurol. 2002 Nov;1(7):409-16. PubMed.

    . Presenilin-dependent transcriptional control of the Abeta-degrading enzyme neprilysin by intracellular domains of betaAPP and APLP. Neuron. 2005 May 19;46(4):541-54. PubMed.

  2. This paper is one of the most interesting contributions of the year, and may well be one of the most informative animal models of AD yet published. To fully appreciate it, readers should first read two prior papers from this same group in which they systematically analyze the consequences of conditionally knocking out PS1 activity under different conditions. If PS1 is knocked out in postnatal neurons, PS2 can compensate, unless the APP load is excessive, as is the case when the PS1 KO is generated in animals bearing mutant forms of APP. The big surprise is that animals with such combinations do not generate large amounts of amyloid material, yet they eventually become as mentally disabled as those who do have large Aβ deposits. Predictably, these animals also generate large amounts of the APP C-terminal peptide, C-99, the consequence of an almost total lack of γ-secretase activity. Why neuronal dysfunction follows is the big question, since the secreted form of Aβ should not be a factor. The authors believe that the accumulation of C-99 may be responsible for the memory deterioration, and they offer microscopic evidence that these peptides concentrate in synaptic terminals, but several questions remain to be answered. In which subcellular compartment do they accumulate? Do they exist as dimers that are still embedded within lipid raft domains of membranes? If so, why don’t the variant secretases (δ, ε) and the membrane-associated ADAM proteases digest them into smaller hydrophobic peptides? These could conceivably remain within the membrane interior for long periods and behave as toxic elements. In addition to these APP related questions, having compromised amounts of PS1 raises another set of problems for such animals, given the wide range of physiological intramembranous cleavages (RIPs) that are known to require PS1 collaboration.

    I am impressed that these animals develop neurological problems, reminiscent of clinical AD, without the accumulation of vast amounts of extracellular deposits of Aβ peptides. One has to wonder whether the earliest forms of the human disease share these characteristics, with the intracellular accumulation of as yet unidentified toxic APP products preceding the development of amyloidosis.

  3. Interesting piece of work. The more sophisticated these models are, the more you can conclude one way or another. Is this model suggesting to kill presenilin to cure AD? Of course not, but data are in favor of that.

    What if the explanation of the beneficial effect is linked to the prevention of the inflammatory response?

    View all comments by Andre Delacourte
  4. Another Disconnect between Amyloid and Cognition
    Saura and colleagues (2005), like Van Leuven before (Dewachter et al., 2002), demonstrate a clear disconnect between amyloid-β and cognitive decline. As such, while it is clear that mutations in APP cause disease, the mechanism(s) by which mutations cause the disease is far from clear. The fact that cognitive deficits are apparent in PS1 cKO/PDAPP mice indicates that amyloid-β is unlikely to be involved, and that the worsening of cognition with age points to other mechanisms (Nunomura et al., 2004). Notably, the fact that PS1 cKO/PDAPP lacking amyloid-β fare worse than PDAPP animals with amyloid-β might even indicate that amyloid-β is beneficial in certain circumstances as we previously indicated (Nunomura et al., 2001; Rottkamp et al., 2001; Lee et al., 2004).

    Hyoung-gon Lee, Xiongwei Zhu, George Perry, Mark Smith


    . Neuronal deficiency of presenilin 1 inhibits amyloid plaque formation and corrects hippocampal long-term potentiation but not a cognitive defect of amyloid precursor protein [V717I] transgenic mice. J Neurosci. 2002 May 1;22(9):3445-53. PubMed.

    . Challenging the amyloid cascade hypothesis: senile plaques and amyloid-beta as protective adaptations to Alzheimer disease. Ann N Y Acad Sci. 2004 Jun;1019:1-4. PubMed.

    . Oxidative damage is the earliest event in Alzheimer disease. J Neuropathol Exp Neurol. 2001 Aug;60(8):759-67. PubMed.

    . Neuronal RNA oxidation is a prominent feature of familial Alzheimer's disease. Neurobiol Dis. 2004 Oct;17(1):108-13. PubMed.

    . The state versus amyloid-beta: the trial of the most wanted criminal in Alzheimer disease. Peptides. 2002 Jul;23(7):1333-41. PubMed.

    . Conditional inactivation of presenilin 1 prevents amyloid accumulation and temporarily rescues contextual and spatial working memory impairments in amyloid precursor protein transgenic mice. J Neurosci. 2005 Jul 20;25(29):6755-64. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Mouse Presenilin-1 Knockouts Have Lowered Levels of Aβ Peptides
  2. Earliest Amyloid Aggregates Fingered As Culprits, Disrupt Synapse Function in Rats

Paper Citations

  1. . High-level neuronal expression of abeta 1-42 in wild-type human amyloid protein precursor transgenic mice: synaptotoxicity without plaque formation. J Neurosci. 2000 Jun 1;20(11):4050-8. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Conditional inactivation of presenilin 1 prevents amyloid accumulation and temporarily rescues contextual and spatial working memory impairments in amyloid precursor protein transgenic mice. J Neurosci. 2005 Jul 20;25(29):6755-64. PubMed.