These are comfortable times for the amyloid hypothesis, it would seem. Every week brings more good news about some anti-amyloid intervention having “cured” mice from their “Alzheimer’s.” On the human front, we are eagerly awaiting such therapeutics to show their moxie in the clinic.
The debate has turned away from knowledge gaps in the amyloid hypothesis and toward issues of better trial design, antecedent markers, even prevention, to make the most of those hoped-for therapeutics.
But what if it’s wrong? Right now, all eggs seem to be in the anti-amyloid basket. What if they crack? Will the field be prepared to pull an alternative hypothesis out of its sleeve and push it into drug discovery, presto? In other words, if plan A fails, what ideas for plan B deserve serious attention now while the clinical trials play themselves out? Human trials proceed at a snail’s pace—are we using the intervening time as well as we could to groom a next generation of candidate treatments?
A leading scientific contender for such an alternative hypothesis is the presenilin loss-of-function hypothesis. It holds that problems with presenilin other than Aβ production account for neurodegeneration and dementia. Last year, an aspect of this hypothesis, dealing with the relative quantities of Aβ42 versus Aβ40, bubbled up into a lively Alzforum discussion when Peter Davies initiated a conversation about a paper from Bart de Strooper’s lab. But the origin of this alternative presenilin hypothesis goes back to another nettlesome thinker, Jie Shen, and to the surprising phenotype she observed in neuron-specific, conditional presenilin knockout mice (see 2003 ARF conference story). Now Shen, at Brigham and Women’s Hospital in Boston, formally presents her hypothesis in a sharply reasoned PNAS Perspectives article, written jointly with Raymond Kelleher of Massachusetts General Hospital. The Alzforum invites the community to consider anew the question of just how presenilin mutations cause AD.
Editor’s Note:—Updated 1 February 2007
By fortuitous timing, Bart de Strooper and Michael Wolfe have written their own analysis of the existing data on presenilin genetics, biochemistry, and AD, and John Hardy prefaces their new perspectives in a “talking point” review series in the February 2007 issue of EMBO reports that went online today. We include these articles here to expand the discussion. The Alzforum editors thank Nature Publishing Group for granting Alzforum readers free access to the full text of these essays until 31 March 2007.
- Hardy J. Putting presenilins centre stage. Introduction to the Talking Point on the role of presenilin mutations in Alzheimer disease EMBO reports 8, 2, 134–135 (2007) Full text
- Wolfe MS (2007) When loss is gain: reduced presenilin proteolytic function leads to increased Aβ42/Aβ40. EMBO reports 8: 136–140 Full text
- De Strooper B (2007) Loss of function presenilin mutations in Alzheimer disease. EMBO reports 8: 141–146 Full text
Editor's Note:—Updated 16 February 2007
Nikolaos K. Robakis
To reflect the fullness of recent data on this topic, we have arranged permission to give Alzforum readers free access to a paper published last month by Nikolaos Robakis's laboratory. In it, Junichi Shioi and colleagues found no consistent elevation of either Aβ42 production or the Aβ42/40 ratio in eight FAD mutations they analyzed in culture, and they argue that these mutations promote neurodegeneration by a different mechanism. The study injects new data into this discussion. Consider it, and send in your comment.
Our thanks to the Journal of Neurochemistry for giving us their permission to post the following paper:
Shioi J, Georgakopoulos A, Mehta P, Kouchi Z, Litterst CM, Baki L, Robakis NK. FAD mutants unable to increase neurotoxic Abeta 42 suggest that mutation effects on neurodegeneration may be independent of effects on Abeta. J Neurochem. 2007 Jan 24; [Epub ahead of print] See full text
Editor's Note:—Updated 28 February 2007
We have arranged permission to give Alzforum readers free access to a paper to be published in Neurodegenerative Diseases by Samir Kumar-Singh. This document reflects an uncorrected pre-print version of the article. Read it, ponder it, and send in your comments.
Our thanks to Karger and Neurodegenerative Diseases for giving us their permission to post the following paper:
Bianca Van Broeck, Christine Van Broeckhoven and Samir Kumar-Singh. Current insights into molecular mechanisms of Alzheimer disease and their implications for therapeutic approaches. Neurodegenerative Diseases. Accepted after revision: December 12, 2006 See full text. See Figure 1, Figure 2, Figure 3, and Figure 4 of full text.
We suggest these questions for discussion:
- What new arguments do these five articles add to the debate?
- What incontrovertible data do they ignore?
- If Aβ lowering disappoints in the clinic, what other angles of presenilin and APP offer targets for therapeutic approaches?
- What evidence is there that presenilin dysfunction is at the root of late-onset AD, not just FAD?
- The goal is not debate itself; it’s truth. Can both the amyloid and the presenilin hypothesis be right? How?
- Presenilin acts outside of γ-secretase. What does this mean for AD?
- In this debate, what’s semantics, what’s real?
- Do these reviews approach a consensus interpretation of the data?
- What are the remaining key differences?
- What's the way forward? What experiments can resolve the issue?
Presenilin is central to AD research, and understanding its role in pathogenesis is of huge importance. You are cordially invited to type your thoughts into the comment box below, or e-mail them to us via our Contact us Form. The Alzforum will post your comments, as well as author replies.
Summary of Shen and Kelleher, 2007
By Gabrielle Strobel
Shen and Kelleher question the link between amyloid deposition, neurodegeneration, and clinical dementia. They argue that the synaptic loss that correlates so well with clinical symptoms may be caused by something other than amyloid alone. They remind us of the stubborn fact that mouse models of APP overexpression generate amyloid but no neurodegeneration. By contrast, mice that lose presenilin function at a young age reproduce AD quite nicely. At first, synaptic plasticity, NMDA receptor-mediated synaptic responses, signaling cascades, and gene expression break down, then memory performance drops off, and soon after tau becomes hyperphosphorylated and neurons begin dying in droves in an environment of gliosis (Saura et al., 2004). These mice have no amyloid, but they model AD more faithfully than do APP transgenics, Shen and Kelleher write. Besides resurrecting the old criticism that amyloid plays a bit part in AD, these data suggest that FAD mutations cause AD primarily by perturbing other aspects of presenilin function that are essential for a healthy brain, yet remain woefully understudied.
Nice idea, you might say, but is there anything to it? After all, a full-bore genetic inactivation of presenilin is not what afflicts people with AD, as some scientists have pointed out. Here are further sets of evidence Shen and Kelleher discuss:
1. FAD Presenilin Mutations Do More Than Upping Aβ42
Presenilin mutations causing AD are known to increase relative Aβ42 levels, and that is frequently cited as evidence for the amyloid hypothesis. But it is a myopic argument in that γ-secretase, the complex intramembrane protease assembled from presenilin and three other proteins, does much else besides generate Aβ. Notch is a physiologically important substrate, in adulthood as well as in development. Many FAD mutations reduce Notch cleavage. How does this affect Notch signaling, and what are the downstream biological effects in the aging human brain? Likewise, many FAD PS1 mutations reduce generation of APP’s intracellular domain, AICD. But despite intensive research, both the biological role of AICD in human brain and any consequences of AICD reduction by FAD mutations remain elusive. In any event, in FAD mutants NICD levels are down, AICD levels are down, and whether this indicates a general loss of function by γ-secretase represents a legitimate question that needs to be followed up, the authors argue. In addition, many AD PS mutations impair the initial, internal cut that presenilin performs on itself, and researchers need to sort out if this reduces the functional ability of γ-secretase.
Moreover, cadherins are γ-secretase substrates, and some FAD PS mutations suppress the cleavage of N-cadherin, a trans-synaptic stabilizing protein. What are the biological consequences in the human brain of suppressing N-cadherin cleavage? Finally, presenilin acts outside of γ-secretase, as well. For example, it tamps down Wnt signaling and some FAD PS mutations interfere with that process. Some recent data indicate that presenilin, when lodged in the ER, might act as a calcium channel of sorts; that function, too, is separate from Aβ generation. Taken together, the authors argue, pathogenic PS mutations weaken presenilin function inside and outside of γ-secretase, and the effect of this weakening on AD pathogenesis deserves more scrutiny. The authors tabulated data from 15 published studies analyzing the effect of various FAD mutations on levels of Aβ42, Aβ40, NICD, and AICD, as well as their ability to rescue knockout of the worm PS homolog sel12. The table suggests that whatever product one measures, partial γ-secretase inhibition by FAD PS mutations is the rule, not the exception. The Aβ42 increase represents an aberration from that rule. In other words, everything goes down, Aβ42 goes up. The question is what causes AD, the former or the latter?
2. Many γ-secretase Inhibitors Enhance Aβ42 Production
Enhance? Yes, you read correctly (at least at certain doses). A slew of small-molecule inhibitors of this intramembrane aspartic protease complex have been studied, and some appear to be doing fine in the clinic so far. But paradoxically, some of these compounds have turned out to do the opposite of what scientists expected. They increase generation of Aβ42 even while reducing that of other γ-secretase cleavage products, such as Aβ40. In that, these γ-secretase inhibitors mimic some FAD PS1 mutations, Shen and Kelleher caution, and this would cast doubt on their therapeutic potential. Other scientists have echoed this concern. The authors take this inhibitor data to support their argument that those FAD mutations cause AD by impairing certain γ-secretase functions. Increased Aβ42 production is but a symptom of a generally “sick” γ-secretase, they write. There is also a puzzling biphasic dose effect that needs a mechanistic explanation, whereby low doses of γ-secretase inhibitors boost Aβ42 production and higher doses decrease it.
3. Some PS1 Mutations Cause Dementia Without Amyloid
The argument here is that a few families are known to science who have what appear to be pathogenic mutations in PS1 but no amyloid, and who clinically have frontotemporal dementia more than AD. The mutations are L113P (Raux et al., 2000), G183V (Dermaut et al., 2004), M233L (Mendez and McMurtray, 2006), and an insertion called insR352. The last was called into question while Shen and Kelleher’s article was in press. An affected family upon later analysis proved also to carry a mutation in the progranulin gene, which is now blamed for the family’s symptoms (Boeve et al., 2006). This discussion welcomes comments on whether the field still considers the partial inhibition of γ-secretase reported earlier for this PS1 insertion to be relevant to the family’s disease (e.g., Amtul et al., 2002; AD/FTD mutation database).
4. Mutations All Across PS1 Imply Loss of Function
Unlike pathogenic APP mutations, which cluster around its cleavage sites (see APP diagram), the more than 150 pathogenic PS1 mutations known to date are widely distributed. They tend to cluster in all its nine transmembrane regions, as well as its endoproteolytic region (see new PS1 diagram). This pattern emphasizes the general importance of presenilin’s normal physiological function, rather than pointing to a specific, toxic gain of function with regard to Aβ42 production, the authors argue. They suspect that mutant, impaired PS1 protein acts like a “dominant-negative,” meaning that it somehow interferes with the proper activities of the normal allele that is also present in people with FAD. The authors discuss an allosteric mechanism that would allow mutant PS1 to both effect an increase in Aβ42 production and an inhibition of other γ-secretase functions.
5. What About APP Mutations? They Cause AD Through More Aβ!
True enough, the authors readily acknowledge, but they add that, as yet, nobody is quite sure how. Shen and Kelleher propose that all that extra Aβ generated off the mutant APP might interfere with presenilin function and in this way lead to a partial loss of function. A negative feedback loop between excess product, i.e., Aβ, and the generating enzyme might be at play. This, in effect, would turn excess Aβ into a sort of γ-secretase inhibitor that would sit in the active site for prolonged periods of time and hold up the enzyme’s other important cleavages. There is no shortage of ideas in this Perspective. Another suggests that excess Aβ might reduce expression of presenilin genes, and the authors point out that increased Aβ and impaired γ-secretase function may well converge at downstream signaling steps to reduce synaptic NMDA receptors and gene expression.
All things considered, the authors note that their model leaves ample room for elevated Aβ42 levels to contribute to AD. They do not view the presenilin loss-of-function hypothesis to be mutually exclusive with a role for Aβ42. They propose, however, that Aβ wreaks its damage by interfering with other, needed γ-secretase activities. In this view, loss of γ-secretase function would occur downstream of Aβ42 accumulation in FAD, and independently of Aβ in frontotemporal dementia. Shen and Kelleher stress that the “presenilin hypothesis reconciles important discrepancies in our current understanding of AD, thereby uniting a fragmented set of observations.” If true, the presenilin hypothesis would bode ill for some therapeutic attempts to inhibit γ-secretase. It would instead point to alternative strategies of inhibiting opposing pathways or even boosting certain physiologically important presenilin functions. Will selective presenilin agonists be up next? Esteemed colleague, what do you think?