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.



Journal of Neurochemistry


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



Samir Kumar-Singh


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.



Neurodegenerative Diseases


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?


  1. In the last years, several laboratories have shown experimentally that presenilin FAD mutations cause a loss of function in both γ-secretase-dependent (Chen et al., 2002; Marambaud et al., 2003; Georgakopoulos et al., 2006) and γ-secretase-independent (Kang et al., 1999; Baki et al., 2004; Tu et al., 2006) cellular pathways. Importantly, it has been shown that FAD mutations cause a loss of PS-dependent function in the cell survival PI3K/Akt/GSK-3 signaling. Based on this observation, we and others proposed that loss of function in this pathway is involved in the neurodegeneration and tau abnormalities of PS mutant-induced FAD (Baki et al., 2004; Kang et al., 2005). This review by Shen and Kelleher is a welcome addition to a growing literature suggesting that loss, rather than gain, of function in specific cellular pathways may be causally involved in the mechanism by which FAD mutations promote neurodegeneration.


    . Presenilin 1 mutations activate gamma 42-secretase but reciprocally inhibit epsilon-secretase cleavage of amyloid precursor protein (APP) and S3-cleavage of notch. J Biol Chem. 2002 Sep 27;277(39):36521-6. PubMed.

    . A CBP binding transcriptional repressor produced by the PS1/epsilon-cleavage of N-cadherin is inhibited by PS1 FAD mutations. Cell. 2003 Sep 5;114(5):635-45. PubMed.

    . Metalloproteinase/Presenilin1 processing of ephrinB regulates EphB-induced Src phosphorylation and signaling. EMBO J. 2006 Mar 22;25(6):1242-52. PubMed.

    . Presenilin 1 facilitates the constitutive turnover of beta-catenin: differential activity of Alzheimer's disease-linked PS1 mutants in the beta-catenin-signaling pathway. J Neurosci. 1999 Jun 1;19(11):4229-37. PubMed.

    . PS1 activates PI3K thus inhibiting GSK-3 activity and tau overphosphorylation: effects of FAD mutations. EMBO J. 2004 Jul 7;23(13):2586-96. PubMed.

    . Presenilins form ER Ca2+ leak channels, a function disrupted by familial Alzheimer's disease-linked mutations. Cell. 2006 Sep 8;126(5):981-93. PubMed.

    . Presenilins mediate phosphatidylinositol 3-kinase/AKT and ERK activation via select signaling receptors. Selectivity of PS2 in platelet-derived growth factor signaling. J Biol Chem. 2005 Sep 9;280(36):31537-47. PubMed.

  2. Undoubtedly, the PNAS paper by Shen and Kelleher provides alternative lines of thought for the primary cause of neurodegeneration in FAD cases with presenilin mutations due to a presenilin loss of function.

    My question to the authors is, what is their thinking regarding the vast majority of AD cases that are “sporadic”? Are any abnormalities seen in the expression of presenilin or the other γ-secretase components in sporadic AD, or is the presenilin loss-of-function hypothesis only true for FAD carrying presenilin and possibly APP mutations?
    Considering the major risk factor for late-onset AD, ApoE4, it is noteworthy that in a mouse model of AD, the Aβ pathology seen with an ApoE4 transgene expression was similar to that seen with mutant presenilin expression (Van Dooren et al., 2006). This suggests that Aβ may play a role, after all, in the non-familial cases if we can extrapolate results from mice to humans. That SorLA is another substrate for γ-secretase (Nyborg et al., 2006) is of particular interest, now that SorLA has been identified as another risk factor for late-onset AD (Rogaeva et al., 2007). When SorLA (SORL1) is underexpressed, more Aβ is produced, as well.


    . Sortilin, SorCS1b, and SorLA Vps10p sorting receptors, are novel gamma-secretase substrates. Mol Neurodegener. 2006;1:3. PubMed.

    . Neuronal or glial expression of human apolipoprotein e4 affects parenchymal and vascular amyloid pathology differentially in different brain regions of double- and triple-transgenic mice. Am J Pathol. 2006 Jan;168(1):245-60. PubMed.

    . The neuronal sortilin-related receptor SORL1 is genetically associated with Alzheimer disease. Nat Genet. 2007 Feb;39(2):168-77. PubMed.

  3. Shen and Kelleher present a compelling hypothesis that presenilin function is critical to the pathogenesis of AD. Alternatively, a loss of presenilin function may contribute more to the damage of calcium dynamics according to the calcium hypothesis of AD (Khachaturian, 1994).

    In the spirit of what other compelling data is not explained solely by loss of presenilin function, there's still that pesky aspect of aging. We must account for age even with human carriers of APP or presenilin mutations that appear to take decades for phenotypic expression (Brewer, 2000; Brewer et al., 2005).


    . Neuronal plasticity and stressor toxicity during aging. Exp Gerontol. 2000 Dec;35(9-10):1165-83. PubMed.

    . Age-related calcium changes, oxyradical damage, caspase activation and nuclear condensation in hippocampal neurons in response to glutamate and beta-amyloid. Exp Gerontol. 2005 May;40(5):426-37. PubMed.

    . Calcium hypothesis of Alzheimer's disease and brain aging. Ann N Y Acad Sci. 1994 Dec 15;747:1-11. PubMed.

  4. Shen and Kelleher raise some interesting points and provide a new view to the potential pathology of AD. Is AD caused by a loss of function of presenilin?

    Some presenilin mutations reduce NICD levels and thus could be pathogenic by reducing Notch function. It could well be that this contributes to the neuronal vulnerability, but there are many other presenilin mutations that have no influence on Notch cleavage yet still cause AD. The widely accepted mechanism of these mutations is that it changes the processing of APP towards higher levels of Aβ42, as do most of the mutations in APP. However, at this point we have no clear idea by which mechanism Aβ42 or its aggregates exert their pathologic actions, and why in most of the transgenic models, high Aβ42 levels do not lead to neuronal loss. Shen and Kelleher argue that high Aβ42 inhibits γ-secretase function, leading thereby to a loss of presenilin function. This is interesting but not really backed by experimental evidence.

    Furthermore, a complete and permanent knockout loss of presenilin function, as in the conditional presenilin knockout mice, could impair neuronal function and viability, but does this argue for a similar mechanism taking place in the AD brain?

    The pharmacology of γ-secretase inhibitors is complex. Some compounds are known that shut off γ-secretase activity completely in a dose-dependent manner, and there are other compounds that modify γ-secretase activity. This modification can be toward a decrease of Aβ42 and a parallel increase of Aβ37 or 38 or the inverse, an increase of Aβ42 and decrease Aβ38 or 37, with and without affecting Aβ40 processing. Many of these modifiers act as inhibitors at higher concentrations. This pharmacology clearly exceeds what is known from presenilin mutations and argues for a highly organized enzyme complex with a fragile specificity for the cleavage site and allosteric centers influencing the cleavage site.

    Although we have a good understanding of many pathological processes and mechanisms in AD, there are still many open questions. There is hope that some of these questions will be answered in the clinic by therapeutic approaches targeting different aspects of the disease process and finally confirming the amyloid cascade hypothesis, or disproving it.

  5. Loss of presenilin function and abnormal calcium signaling
    The papers discussed here, by Jie Shen and Raymond Kelleher, Bart De Strooper, and Michael Wolfe, offer a very interesting analysis of FAD-linked mutations in presenilins. In particular, Shen and Kelleher make a strong point that loss of presenilin function may be responsible for the AD phenotype. The strongest experimental evidence in support of this idea comes from previous analysis of PS cDKO mice performed by Shen’s laboratory [1].

    Presenilins form a catalytic core of the γ-secretase, and most of the discussion in all three papers focuses on mutations in presenilins that cause a loss of γ-secretase function. I am not an expert in γ-secretase and therefore not qualified to comment on the specific points raised by these authors. I would, however, like to point out that our previous analysis revealed that the M146V mutation in PS1 and N141I mutations in PS2 resulted in the loss of a ER Ca2+ leak function mediated by presenilins [2]. Moreover, by coexpressing M146V and wild-type PS1 constructs, we demonstrated that the M146V mutant is able to “kill” the ER Ca2+ leak channel activity of wild-type PS1, and this is consistent with the dominant-negative mechanism exerted by these mutations [2]. In our more recent studies, we evaluated the effects of six randomly selected FAD mutations in PS1, and discovered that five of them completely abolish the ER Ca2+ leak function of PS1 [3]. In contrast, we found that three known FTD-associated mutations in PS1 (L113P, G183V, and Rins352) did not appear to affect the ER Ca2+ leak function of PS1 in our experiments. From these results, we concluded that either these mutations are not pathogenic or that defects in the ER Ca2+ leak pathway are not involved in FTD pathogenesis. Consistent with the former possibility, the insR352 mutation in PS1 is likely not pathogenic, and FTD in the affected family results from mutation in the progranulin gene [4].

    Obviously, the connection between deranged Ca2+ signaling and AD pathogenesis is a wide-open question that is far from being resolved. But in the context of the current discussion regarding the loss of presenilin function in AD, our results agree “in spirit” with ideas expressed by Shen, Kelleher, De Strooper and Wolfe, although the mechanistic basis for our conclusions would be quite different.


    . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.

    . Presenilins form ER Ca2+ leak channels, a function disrupted by familial Alzheimer's disease-linked mutations. Cell. 2006 Sep 8;126(5):981-93. PubMed.

    . Familial Alzheimer disease-linked mutations specifically disrupt Ca2+ leak function of presenilin 1. J Clin Invest. 2007 May;117(5):1230-9. Epub 2007 Apr 12 PubMed.

    . Frontotemporal dementia and parkinsonism associated with the IVS1+1G->A mutation in progranulin: a clinicopathologic study. Brain. 2006 Nov;129(Pt 11):3103-14. PubMed.

  6. Shen and Kelleher's PNAS paper raises interesting questions, and it compiles data in favor of provocative answers. But what incontrovertible data did the authors ignore?

    I agree with the comments made by Christian Czech. The pharmacology of γ-secretase inhibitors can no longer be summarized as depicted in Figure 3 of the PNAS Perspective. This figure may have been valid for active site-directed transition state analogues in 2003. But there are now several straight and inverse modulators (Leuchtenberger et al., 2006, Narlawar et al., 2006), several binding sites, and non-active site directed inhibitors. And a minority thereof displays the profile as in Figure 3. The authors do not take into account the activity of the Torrey Pines γ-secretase modulator (presented in March 2005), which is highly potent and selectively reduces Aβ42. Yet the authors arrive at the dimer model with allosteric regulation (see below).

    Dirk Beher/Mark Shearman et al. published such contradictory data for allosteric inhibitors (Beher et al., 2004) and presented a model (Clarke et al., 2006) for their observation. The model is based on two PS1 active sites and a dimeric substrate. This model provides a rationale for the modulation by potent active site inhibitors at low concentrations.


    . Selective modulation of Abeta42 production in Alzheimer's disease: non-steroidal anti-inflammatory drugs and beyond. Curr Pharm Des. 2006;12(33):4337-55. PubMed.

    . Selected non-steroidal anti-inflammatory drugs and their derivatives target gamma-secretase at a novel site. Evidence for an allosteric mechanism. J Biol Chem. 2004 Oct 15;279(42):43419-26. Epub 2004 Aug 10 PubMed.

    . Intra- or intercomplex binding to the gamma-secretase enzyme. A model to differentiate inhibitor classes. J Biol Chem. 2006 Oct 20;281(42):31279-89. PubMed.

    . Scaffold of the cyclooxygenase-2 (COX-2) inhibitor carprofen provides Alzheimer gamma-secretase modulators. J Med Chem. 2006 Dec 28;49(26):7588-91. PubMed.

  7. I first heard Jie present her presenilin hypothesis of Alzheimer disease at a Keystone meeting in Breckenridge in February 2006. At that time, I had read her seminal papers on the relation between the loss of function of presenilin and neurodegeneration (especially the crucial paper Saura et al., 2004), but did not fully grasp the extent of her ideas suggesting a general role for the loss of function of presenilin in driving neurodegeneration. Her presentation at the Keystone meeting really was a signature event for me in that it clearly spelled out her ideas, as now published in her PNAS Perspective. It is amazing how fertile a ground Jie's ideas have found as judged by the wave of recent opinion pieces echoing these ideas.

    Although this is not really my field, I think that at this point, we have a major opportunity to reevaluate our approach to neurodegeneration. To me, as an outsider in the Alzheimer field, there appears to be little doubt that neurodegeneration is as complex as the brain that affects it, and any simple cause/effect relationship is unlikely. Clearly, multiple pathways feed into the gradual and slow decline of neuronal function observed in Alzheimer disease. The idea that a loss of function of presenilins is a major contributor to this process is overwhelmingly attractive in view of the many presenilin mutations in familial Alzheimer disease, and Jie and Ray make a beautiful case for this idea in their PNAS Perspective. As they themselves state, however, this does not mean that Aβ is irrelevant, or that ApoE does not play a major role.

    My personal view is that the major challenge now is to determine the precise nature of neuronal cell death in Alzheimer disease, the molecular identity of the final pathway that leads to neuronal cell death downstream of presenilins, and how events such as a loss of function of presenilins initiate neuronal cell death. Is it a lack of synaptic function, directly or indirectly caused by a loss of presenilin function? It seems likely to me that understanding the normal functions of the proteins involved, beyond the very important identification of presenilins as likely catalytic γ-secretase subunits, will be important for meeting the challenge of understanding neurodegeneration, for example, the role of ApoE in presenilin function, if any, and the role of APP cleavage in cell physiology.


    . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.

  8. Jie Shen and Ron Kelleher’s PNAS Perspective summarizes experimental data from several laboratories suggesting a presenilin loss-of-function mechanism in Alzheimer disease. This hypothesis is strongly supported by the following experimental evidence:

    1. FAD-linked PS mutations reduce γ-secretase activity (Bentahir et al., 2006; Kumar-Singh et al., 2006).

    2. Some FAD-linked PS1 mutations alter γ-secretase-independent PS functions (PI3K, β-catenin, Ca2+…).

    3. Some PS1 mutations (G183V and L113P) cause frontotemporal dementia characterized by the presence of tauopathy but not amyloid plaques (Dermaut et al., 2004; Hutton, 2004).

    4. PS cDKO mice with targeted disruption of presenilins in the cerebral cortex have memory deficits and neurodegeneration (Saura et al., 2004).

    However, it is still unclear whether loss of PS function can explain the majority of sporadic cases of AD. First, not all FAD-linked PS mutations increase Aβ42 or decrease Aβ40 levels or other γ-secretase cleavages to the same degree (Bentahir et al., 2006; Kumar-Singh et al., 2006), indicating that other γ-secretase-dependent or -independent PS functions besides APP processing should be affected in this disease. Second, the fact that abnormal levels of Aβ42 are found in the AD brain supports the prevailing view that changes in the generation and/or degradation of Aβ are key factors in AD pathogenesis.

    Whether Aβ peptides may cause AD by interfering with PS function and/or expression by acting as γ-secretase inhibitors, as suggested by the authors in their PNAS Perspective, is an attractive but provocative idea that has not been proved. If that is the case, then a reduction of PS/γ-secretase activity should be present in AD brains. Furthermore, if abnormal Aβ levels and loss of PS function coordinately cause neurodegeneration and memory loss in AD, then both events should impair a common set of cellular pathways essential for neuronal survival and memory.

    I totally agree with Tom Sudhof that understanding the normal functions of the proteins involved in the disease will be important for understanding neurodegeneration. Indeed, our investigations of PS cDKO mice were precisely aimed to gain insight in the normal function of PS. From those studies it is clear that PS regulates several signaling pathways (NMDA, CREB…) that are essential for synaptic function and plasticity and neuronal survival, and that loss of PS function disrupts memory and causes neurodegeneration. It is surprising to me that after several decades of research on the molecular basis of AD, the normal function of Aβ is still unknown. In my opinion, it will be important to continue searching for the normal function of PS and Aβ and the molecular mechanisms that they share in order to fully understand the etiology of this complex disorder.


    . Presenilin clinical mutations can affect gamma-secretase activity by different mechanisms. J Neurochem. 2006 Feb;96(3):732-42. PubMed.

    . A novel presenilin 1 mutation associated with Pick's disease but not beta-amyloid plaques. Ann Neurol. 2004 May;55(5):617-26. PubMed.

    . Presenilin mutations associated with fronto-temporal dementia. Ann Neurol. 2004 May;55(5):604-6. PubMed.

    . Mean age-of-onset of familial alzheimer disease caused by presenilin mutations correlates with both increased Abeta42 and decreased Abeta40. Hum Mutat. 2006 Jul;27(7):686-95. PubMed.

    . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.

    . The presenilin hypothesis of Alzheimer's disease: evidence for a loss-of-function pathogenic mechanism. Proc Natl Acad Sci U S A. 2007 Jan 9;104(2):403-9. PubMed.

  9. A Ca2+ Channel Model for Presenilin (PS)
    Shen and Kelleher raised a critical question for presenilin (PS) mutations by suggesting a loss-of-function model. I think this model is long overdue. We know most gene mutations cause human diseases by loss of function. This debate may be more exciting if we also ask, what is presenilin’s function?

    Based on the initial report that the PS structure highly resembles those of ion channels, and that PS1 has amino acid sequence homology to a Ca2+ channel [1], I proposed in 1998 that PS most likely acts as a calcium channel in vivo and that pathogenic mutations did their damage by reducing the channel’s function [2].

    By today, numerous studies have strongly linked PS to “calcium imbalance” [3, review]. Recently, Tu et al. [4] have explicitly suggested that PS acts as a “calcium leak channel.” But how would FAD mutations change the channel’s function? These studies have concluded that the mutations increase intracellular calcium levels, a finding that fits with the prevailing theory of “calcium overload” [5], which posits that calcium levels increase with age, leading to excessive calcium activities and cell death in sporadic AD. This notion has been widely publicized, so is the problem solved? I submit that it is not. This notion, in fact, is incompatible with the common observation that Ca2+-regulated activities in the body all decrease with age, including fertilization, cell division, growth and secretion, muscle contraction, and synaptic transmission [6].

    Where is the common ground between the two notions? The PS studies mentioned above are based on a time-honored assumption that Ca2+ functions through steady-state “level” changes, like water in a swimming pool. This view, however, has missed a recent conceptual revolution in Ca2+ research: the ion actually functions through rapid spikes, like radio waves, whose frequency and amplitude change within a 1 millisecond range [6, and references therein]. Such events can go undetected by commonly used techniques, which usually measure Ca2+ on second or minute time scales. Thus, these data do not represent actual signaling activities in the brain [Fig. 1].



    Figure 1. Schematic illustration of two different “calcium hypotheses” for AD
    (A) The current "calcium overload" hypothesis assumes that Ca2+ functions in a steady-state change, which is measured in second or minute time scales. It suggests that Ca2+ "levels" increase with age, leading to excessive signal activity and cell death. (B) Based on a new oscillation concept, we propose that Ca2+ spike frequency declines with age, resulting in a reduced signal potency. Such subtle changes can only be detected in sub-millisecond time scales. Notably, the reduced spike frequency means a Ca2+ “overstay” in the cytosol, so appears as "higher Ca2+ levels" if measured in longer time scales. Note that these two hypotheses would predict opposing intervention strategies for AD: inhibiting Ca2+ or boosting signaling potency.

    The rapidly oscillating nature of Ca2+ also implies that the age-related deterioration will affect the spikes on a similar time scale. Thus, for instance, a spike that lasts 1/1,000 of a second in a young cell may last 2/1,000 or 3/1,000 of a second in an old cell. This can be a result of energy levels, which drive the waves, declining in aging [6]. This would be a significant decrease in the spike frequency, but can be measured on longer time scales as a “higher Ca2+ level” or “Ca2+overstay in the cytosol” instead [Fig 1]. In other words, it is possible that higher Ca2+ “levels” actually mean a lower signaling “potency.”

    If PS is a Ca2+ channel, then any mutations on it will disrupt its function, most likely by reducing the speed of Ca2+ release. This would decrease the frequency and potency of the signals, leading to insufficient neurotransmission in the PS mutant hosts [7]. But again, a change as subtle as this may have to be measured in millisecond time scales.

    Ca2+ is a central regulator in neurotransmission and cognition, and AD is characterized by insufficient neurotransmission. This and other considerations led me to propose a “Ca2+ signaling deficit” model for sporadic AD [7,8]. By this model, autosomal-dominant and sporadic AD share a similar defect, but are fundamentally different in etiology, as one is due to a mutated channel, the other to age-related deterioration of the Ca2+ signaling system. As direct measurement of Ca2+ waves in the human brain has not been possible today, this model has yet to be directly tested.

    There is hope. Thanks to renewed interest in the research field on the loss-of-function model and the fact that rapid Ca2+ waves can be measured in cultured cells [9], it is possible to directly determine the subtle Ca2+ wave changes in cultured primary neurons derived from PS mutant transgenic animals. Also, age-related changes in Ca2+ spike shape may be determined in the nerve cells of primitive animals [10]. Such studies may eventually help answer the central question of whether the higher Ca2+ “levels” really mean a greater signaling activity.

    See also:

    Chen, M. "Ca2+ signaling deficit hypothesis" for Alzheimer’s disease. (2001) Full Text


    . Candidate gene for the chromosome 1 familial Alzheimer's disease locus. Science. 1995 Aug 18;269(5226):973-7. PubMed.

    . The Alzheimer's plaques, tangles and memory deficits may have a common origin. Part II: therapeutic rationale. Front Biosci. 1998 Jun 8;3:A32-7. PubMed.

    . Presenilin mutations and calcium signaling defects in the nervous and immune systems. Bioessays. 2001 Aug;23(8):733-44. PubMed.

    . Presenilins form ER Ca2+ leak channels, a function disrupted by familial Alzheimer's disease-linked mutations. Cell. 2006 Sep 8;126(5):981-93. PubMed.

    . Calcium hypothesis of Alzheimer's disease and brain aging. Ann N Y Acad Sci. 1994 Dec 15;747:1-11. PubMed.

    . Ca2+ signaling down-regulation in ageing and Alzheimer's disease: why is Ca2+ so difficult to measure?. Cell Calcium. 1999 Sep-Oct;26(3-4):149-54. PubMed.

    . The Alzheimer's plaques, tangles and memory deficits may have a common origin; part I; a calcium deficit hypothesis. Front Biosci. 1998 May 11;3:a27-31. PubMed.

    . Sensitivity of CaM kinase II to the frequency of Ca2+ oscillations. Science. 1998 Jan 9;279(5348):227-30. PubMed.

    . Presynaptic calcium currents in squid giant synapse. Biophys J. 1981 Mar;33(3):289-321. PubMed.

  10. This discussion around the Shen and Kelleher paper is probably the beginning of the end of the amyloid cascade hypothesis. There is also room for the loss of function of APP, and apparently the only possible common point for a global hypothesis is the physiological role of AICD.

  11. Shen and Kelleher offer a very interesting hypothesis. I have reported that AICDε (C50) is preferentially produced from the α-secretase product C83, not from C89 and C99, indicating that γ-site cleavage and ε-site cleavage are regulated differently (Kume and Kametani, 2006). Thus, based on our results, an increase in C89 and C99 by BACE1 induces a reduction in C83, thereby reducing AICDε.

    Elevated BACE1 expression and activity in sporadic AD (Holsinger, et al., 2002; Yang, et al., 2003) and the Swedish FAD mutation of APP (Citron, et al., 1992) may induce an increase in C89 and C99, and the reduction in C83 and AICDε. In addition, FAD mutations of APP clustering near the γ-secretase cleavage site decreased the generation of AICDε, suggesting impaired γ-secretase cleavage (Wiley, et al., 2005). These suggest that impaired AICDε generation (i.e., impaired ε-site cleavage of APP) is a common feature in AD (Kume and Kametani, 2006).

    Interestingly, abnormal accumulation of APP C-terminal fragments in neurons of the cerebral cortex, especially in the synaptic terminals, contributes to memory impairment in older mutant APP mice, in which PS1 is inactivated (Saura, et al., 2004; Saura, et al., 2005). In PDAPP mice with an aspartate to alanine mutation at position 664 of APP (D664A), the APP cytoplasmic domain plays a critical role in the generation of AD-related pathophysiological and behavioral changes, and Aβ production and deposition do not affect these changes (Galvan, et al., 2006).

    In addition, it has been reported that increased expression of APP or the APP C-terminal fragment causes abnormal axonal transport and neurodegeneration, independent of Aβ production (Gunawardena and Goldstein, 2001; Salehi, et al., 2006). These findings indicate that the APP cytoplasmic domain is responsible for AD-related pathophysiological changes. I think that understanding the normal functions of APP cytoplasmic domain is important.


    . Abeta 11-40/42 production without gamma-secretase epsilon-site cleavage. Biochem Biophys Res Commun. 2006 Nov 3;349(4):1356-60. PubMed.

    . Increased expression of the amyloid precursor beta-secretase in Alzheimer's disease. Ann Neurol. 2002 Jun;51(6):783-6. PubMed.

    . Elevated beta-secretase expression and enzymatic activity detected in sporadic Alzheimer disease. Nat Med. 2003 Jan;9(1):3-4. PubMed.

    . Mutation of the beta-amyloid precursor protein in familial Alzheimer's disease increases beta-protein production. Nature. 1992 Dec 17;360(6405):672-4. PubMed.

    . Familial Alzheimer's disease mutations inhibit gamma-secretase-mediated liberation of beta-amyloid precursor protein carboxy-terminal fragment. J Neurochem. 2005 Sep;94(5):1189-201. PubMed.

    . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. 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.

    . Reversal of Alzheimer's-like pathology and behavior in human APP transgenic mice by mutation of Asp664. Proc Natl Acad Sci U S A. 2006 May 2;103(18):7130-5. PubMed.

    . Disruption of axonal transport and neuronal viability by amyloid precursor protein mutations in Drosophila. Neuron. 2001 Nov 8;32(3):389-401. PubMed.

    . Increased App expression in a mouse model of Down's syndrome disrupts NGF transport and causes cholinergic neuron degeneration. Neuron. 2006 Jul 6;51(1):29-42. PubMed.

  12. I wholeheartedly agree with Ilya Bezprozvanny. Calcium leakage is an important consequence of presenilin activity. Those ion channels cannot be overlooked, their structure and function are of direct mechanistic importance as important catalysts of actual AD symptoms. Amyloid is a piece of the much bigger puzzle.

  13. In the field of genetics, it is a rare and precious opportunity that we are afforded as scientists, when we can assess the effects of multiple mutations in multiple genes leading to common clinical and pathological disease phenotypes. We have this privilege in AD genetics with over 150 early onset FAD-causing mutations in APP and the presenilin genes, PSEN1 and PSEN2. As a result, we have the luxury of asking: What are the molecular and biochemical phenotypes that these disparate mutations share in COMMON?

    To ask this question, first we need to think about how the presenilins and APP functionally interact. We know that presenilins are required for the cleavage and likely, the trafficking of APP. Second, we need to ask what functions are commonly affected by the FAD mutations in these genes. The COMMON event altered by FAD mutations in APP and the presenilins is cleavage of APP. While mutations in each of these molecules also lead to other molecular phenotypes, e.g. PS mutations leading to calcium dyshomeostasis, Occam's razor dictates that with regard to AD phenotype, we would best focus on the COMMON consequences of the fully penetrant FAD mutations in THREE DIFFERENT genes all causing FAD. This leaves us with APP cleavage (and perhaps APP trafficking as it relates to APP processing).

    The third question concerns the mechanism and, particularly, whether we are looking at gain- or loss-of-function mutations. The fact that these are autosomal-dominant mutations a priori predisposes us toward a gain-of-function mechanism. Yet, given the dissemination of FAD mutations throughout the presenilins, it makes most sense to think about loss-of-normal function as a common consequence. Of course, loss-of-normal function can simultaneously lead to gain of a novel toxic event. So, we need to look closely at the molecular consequences to ask the question of gain-, or loss-of-function of FAD mutations. With regard to APP, the FAD mutations cluster roughly around the cleavage sites in the Abeta portion of APP. The Swedish mutation leads to gain-of-function: more beta-secretase cleavage leading to more Abeta, more APP-C99, and more sAPP-beta. The 717 mutations alter gamma secretase cleavage leading to an increase in the ratio of Abeta42:other Abeta species. Other FAD mutations within the Abeta domain of APP, e.g. Dutch, increase the aggregation properties of Abeta. Thus, the mutations in APP all lead to gain-of-function in the forms of increased Abeta, increased Abeta aggregation, or increased Abeta42:other Abeta species ratio. Meanwhile, the vast majority of PS FAD mutations also increase the ratio of Abeta42:other Abeta species. Since the increase in ratio due to PS mutations involves both increases in Abeta42 generation and/or decreases in the more common Abeta40 form of the peptide, one must surmise that the loss-of-normal function engendered by FAD mutations in presenilins translates into a gain-of-function at the level of APP: increased ratio of Abeta42:other Abeta species. Thus, the functional intersection of FAD mutations in APP and the presenilins so far point at only ONE COMMON biochemical gain-of-function: an increased ratio of Abeta42:other Abeta species. Of course, these mutations also lead to several other molecular phenotypes, some of which may contribute to AD pathology, and some which may not. But, once again, it is important to emphasize that we are extremely fortunate in this field to have the luxury of comparing fully penetrant, autosomal dominant, FAD-causing mutations in THREE DIFFERENT genes to identify the COMMON molecular phenotype. So far, ONLY increased Abeta42:other Abeta species qualifies as a gain-of-function phenotype that is shared in COMMON among the vast majority of FAD mutations in these three genes. And, gain-of-function makes most sense with the autosomal-dominant (and possibly additive) inheritance models of the FAD mutations. In conclusion, UNTIL we can find another COMMON molecular phenotype of the FAD-causing mutations in these three genes, we would do best to focus on trying to understand how the increase in the ratio of Abeta42:other Abeta species causes AD. A similar approach with HDL and LDL in cardiovascular disease has met with very impressive success in lowering the incidence of heart disease.

  14. The emerging evidence suggests that the role of PS FAD-associated mutations involves not only a gain of negative function via APP processing but also a loss of function. Although there is vast evidence to suggest that APP processing and clearance plays a critical role in AD, AD is likely a syndrome with more than one mechanism being responsible for a common pathological and clinical phenotype.

    Hence in our development of potential therapeutic approaches, it is important to not only target amyloid. Efforts and resources should also be spent to target other factors such as apolipoprotein E and non-APP dependent PS functions as potential therapies for AD. It is likely that individually tailored, multi-modality interventions will be the most effective.

  15. Much effort has been expended in pursuit of an AD treatment based on the theory that Aβ is the primary cause of this disease. The paper by Shioi et al., featured above challenges one of the most important assumptions supporting this theory namely that all FAD mutations increase neurotoxic Aβ. In short, the data in the paper raise the possibility that Aβ-based toxicity may not be able to explain all the neuropathology of AD and that we may need to consider additional alternative mechanisms by which FAD mutants promote neuropathology.

    Whether more than one mechanism are in play remain an open question. An example of such mechanisms may be the effects of FAD mutants on the production of signaling peptides as it has been shown for Notch, N-cadherin and even for APP.

  16. The Relevance of APP Mutations
    Like Rudy Tanzi, I am impressed by the story that mutations can tell. Yet I think lessons learned from the study of APP mutants may be just as instructive as those of the presenilins. Some APP mutants increase the generation of Aβ peptides, such as those close to the β- and γ-secretase sites, but a cluster of mutations at what might be called the "middle" segment of the Aβ domain generates peptides that are more prone to aggregation than native forms and are potentially more toxic (1). The relevant sequences of three of these mutant forms are shown
    below. Families with Dutch, Iowa, and Arctic (and other) FAD mutations have these changes.

    Sequences of Normal and Mutant Forms of Aβ


    The changes in the D/I/A forms involve a shift of one of the two negatively charged amino acids, glutamic (E) or aspartic (D), to a non-charged one. These charged residues are believed to stabilize the turn regions of the β "hairpins" that are thought to characterize the conformation of Aβ peptides after their release from membranes. How they might contribute to increased aggregability is not clear. Their toxicity does not appear to be due to increased production, as with the Swedish mutation, or modified Aβ42/40 ratios, as with the PS1 mutations. Many investigators have shown that these A/I/D type Aβ peptides have three properties that may enhance their toxicity: (i) they are resistant to proteolytic digestion, (ii) they collect around blood vessels, and (iii) they can damage vessel walls leading to amyloid angiopathy, which over time induces inflammatory and thrombotic complications (2,3). It has also been noted by many observers that amyloid angiopathy of small vessels (capillaries and venules) is commonly found in brains of sporadic AD patients (4).

    But is it fair to ask how relevant these rare mutations are to the pathogenesis of the most prevalent forms of sporadic AD? A close look at the amino acid sequences of these peptides provides grounds for some interesting speculation. Each of the changes shown above can be generated by a single base change in the coding sequence involving either a G-C or a G-A switch. If some of the guanine bases of the RNA are modified, rather than those in the DNA, which is modified in the FAD families, such modified mRNAs could direct the synthesis of abnormal APPs without a corresponding change in nuclear DNA (5).

    How could this happen? It has been known for many years that large amounts of oxidatively modified RNA are present in neurons of AD patients (6,7). Guanine is reported to be the most sensitive of the bases to oxidative damage, and 8-OHG is produced, both as the oxidized triphosphonucleotide (8-OH-GTP) and in mRNA itself. Oxidized mRNA that contains 8-OHG has the potential to code for modified forms of APP. A recent study shows that oxidized mRNA can direct the synthesis of polypeptide chains of varying length and changes in amino acid sequence (8). If mRNAs that code for APP have oxidized guanines at appropriate sites, possibly at exposed loops, they could in principle code for APP molecules that contain one of the A/I/D changes in their transmembrane segments.

    One of the obvious objections to such a proposal is the very low probability that enough mRNA that codes for APP will be modified at levels high enough to make a difference. However, because of their resistance to degradation, small amounts of modified peptides could accumulate around blood vessels, possibly over decades, leading to destruction of small blood vessels and localized ischemic changes. Over time such ischemic changes might lead to the localized activation of β-secretase as recently described (9). To test this idea, mRNA samples from AD brains should be analyzed for the presence of abnormal APP messages. If they do indeed exist, they would represent only a small fraction of the total, so some method for enriching the oxidized fraction will be needed. Polynucleotide phosphorylase may be one way to select for oxidized forms of RNA (10).

    This idea that oxidized RNA may generate toxic forms of Aβ and other proteins has many interesting implications. It identifies a specific target for oxidative damage as a cause of dementia, long postulated by others (11), and raises the possibility that attempts to reduce levels of oxidized nucleotides through MutH-like enzymes (12,13), which hydrolyze 8-OHG, and other approaches, might be pursued as new therapeutic initiatives.


    . The 'Arctic' APP mutation (E693G) causes Alzheimer's disease by enhanced Abeta protofibril formation. Nat Neurosci. 2001 Sep;4(9):887-93. PubMed.

    . Pathogenic effects of D23N Iowa mutant amyloid beta -protein. J Biol Chem. 2001 Aug 31;276(35):32860-6. Epub 2001 Jul 5 PubMed.

    . Deficient cerebral clearance of vasculotropic mutant Dutch/Iowa Double A beta in human A betaPP transgenic mice. Neurobiol Aging. 2006 Jul;27(7):946-54. Epub 2005 Aug 18 PubMed.

    . Only cerebral capillary amyloid angiopathy correlates with Alzheimer pathology--a pilot study. Acta Neuropathol. 2004 Feb;107(2):83-90. PubMed.

    . Mutagenic properties of oxidized GTP and ATP in in vitro transcription-reverse transcription. Nucleic Acids Symp Ser (Oxf). 2006;(50):99-100. PubMed.

    . RNA oxidation is a prominent feature of vulnerable neurons in Alzheimer's disease. J Neurosci. 1999 Mar 15;19(6):1959-64. PubMed.

    . The identification and characterization of oxidized RNAs in Alzheimer's disease. J Neurosci. 2003 Jun 15;23(12):4913-21. PubMed.

    . Oxidized messenger RNA induces translation errors. Proc Natl Acad Sci U S A. 2007 Jan 2;104(1):66-71. PubMed.

    . Hypoxia facilitates Alzheimer's disease pathogenesis by up-regulating BACE1 gene expression. Proc Natl Acad Sci U S A. 2006 Dec 5;103(49):18727-32. PubMed.

    . Specific binding of 8-oxoguanine-containing RNA to polynucleotide phosphorylase protein. Biochemistry. 2001 Aug 21;40(33):9977-82. PubMed.

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

    . Mammalian enzymes for preventing transcriptional errors caused by oxidative damage. Nucleic Acids Res. 2005;33(12):3779-84. PubMed.

    . MTH1, an oxidized purine nucleoside triphosphatase, prevents the cytotoxicity and neurotoxicity of oxidized purine nucleotides. DNA Repair (Amst). 2006 Jul 13;5(7):761-72. PubMed.

  17. The recent opinion articles (1-4) on the role of presenilin loss of function in AD are very interesting. The presented presenilin hypothesis (1) is provocative and based on two important recent observations. First, there is indeed strong evidence that presenilin mutations are intrinsic partial loss-of-function alleles (e.g., 5,6) that lead to a dysfunctional γ-secretase (most likely through a dominant-negative mechanism) and result in amyloid-positive and tau-positive neurodegeneration AD. Second, complete conditional presenilin knockout in the mouse forebrain leads to amyloid-negative and tau-positive neurodegeneration (7). The presenilin hypothesis is interesting and original since it suggests that presenilin-mediated neurodegeneration might occur in the absence of Aβ amyloid deposits, placing presenilin and not amyloid in a central position in AD pathogenesis.

    Part of the presenilin hypothesis is also based on data suggesting a role of presenilin mutations in amyloid-negative FTD. Basically, I share the view as outlined in an excellent recent clinical review on the topic (8). Although there is no doubt that several PS mutations can clinically result in FTD presentations, the underlying pathology in such cases is mostly AD (amyloid and tau pathology). Our own PS1 G183V mutation in a patient with pure tau-positive Pick pathology (9) remains a rare (but interesting!) exception to this rule but should be interpreted with caution. In contrast to the FTD patient with the PS1 insR352 mutation, which appears to be a polymorphism since a loss-of-function mutation in the progranulin gene was detected in this case (10), our patient has no other known genetic FTD defects (tau/progranulin). Therefore, the PS1 G183V mutation might indeed be causative, although the possibility of an extremely rare benign polymorphism or hypomorphic modifier mutation can still not be excluded. More confirmatory data on this topic are clearly needed. The currently available evidence on the role of clinical presenilin mutations in amyloid-negative FTD is interesting but, in my opinion, presently not the strongest part of the presenilin hypothesis.

    The major challenge is to reconcile the presenilin and amyloid hypotheses. The amyloid hypothesis is of course supported by the hard genetic fact that APP mutations (including duplications) cause amyloid-positive and tau-positive neurodegeneration. An oversimplified way to look at the whole problem is perhaps to see AD as an imbalance in the ratio between enzyme and substrate caused by either decreased enzyme activity (loss of PS function) or too much/altered substrate (APP duplications/mutations). Since neurodegeneration in PS null brains occurs without Aβ deposition, it seems tempting to speculate that accumulation of insufficiently degraded APP C-terminal fragments (and not β amyloid peptide) might be the unifying harmful event.

    Finally, most striking throughout this whole neurodegenerative spectrum (clinical APP mutations, clinical PS mutations, and complete PS knockout) is the presence of tau pathology (11). A major question remains how PS/APP imbalances could lead to downstream tau pathology, and here the presenilin hypothesis provides no new answers. In this respect, and based on our own work in a Drosophila lysosomal storage disease model (12), I have wondered many times if AD might be a kind of storage disease in which intracellular accumulations of APP derivatives clog the cellular degradative machinery, resulting in the accumulation of tau proteins? This would fit with recent studies showing that a defective intracellular protein degradation machinery is one of the most important upstream modifiers for neurodegenerative protein toxicity throughout a wider range of neurodegenerative diseases (reviewed in 13) and with cell biological studies showing accumulation of degradative organelles in PS1 null cells (14,15).


    . The presenilin hypothesis of Alzheimer's disease: evidence for a loss-of-function pathogenic mechanism. Proc Natl Acad Sci U S A. 2007 Jan 9;104(2):403-9. PubMed.

    . Putting presenilins centre stage. Introduction to the Talking Point on the role of presenilin mutations in Alzheimer disease. EMBO Rep. 2007 Feb;8(2):134-5. PubMed.

    . Loss-of-function presenilin mutations in Alzheimer disease. Talking Point on the role of presenilin mutations in Alzheimer disease. EMBO Rep. 2007 Feb;8(2):141-6. PubMed.

    . When loss is gain: reduced presenilin proteolytic function leads to increased Abeta42/Abeta40. Talking Point on the role of presenilin mutations in Alzheimer disease. EMBO Rep. 2007 Feb;8(2):136-40. PubMed.

    . Presenilin clinical mutations can affect gamma-secretase activity by different mechanisms. J Neurochem. 2006 Feb;96(3):732-42. PubMed.

    . Mean age-of-onset of familial alzheimer disease caused by presenilin mutations correlates with both increased Abeta42 and decreased Abeta40. Hum Mutat. 2006 Jul;27(7):686-95. PubMed.

    . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.

    . Frontotemporal dementia-like phenotypes associated with presenilin-1 mutations. Am J Alzheimers Dis Other Demen. 2006 Aug-Sep;21(4):281-6. PubMed.

    . A novel presenilin 1 mutation associated with Pick's disease but not beta-amyloid plaques. Ann Neurol. 2004 May;55(5):617-26. PubMed.

    . Frontotemporal dementia and parkinsonism associated with the IVS1+1G->A mutation in progranulin: a clinicopathologic study. Brain. 2006 Nov;129(Pt 11):3103-14. PubMed.

    . Tau is central in the genetic Alzheimer-frontotemporal dementia spectrum. Trends Genet. 2005 Dec;21(12):664-72. PubMed.

    . Aberrant lysosomal carbohydrate storage accompanies endocytic defects and neurodegeneration in Drosophila benchwarmer. J Cell Biol. 2005 Jul 4;170(1):127-39. PubMed.

    . The roles of intracellular protein-degradation pathways in neurodegeneration. Nature. 2006 Oct 19;443(7113):780-6. PubMed.

    . Presenilin 1 mediates the turnover of telencephalin in hippocampal neurons via an autophagic degradative pathway. J Cell Biol. 2004 Sep 27;166(7):1041-54. PubMed.

    . Degradative organelles containing mislocalized alpha-and beta-synuclein proliferate in presenilin-1 null neurons. J Cell Biol. 2004 May 10;165(3):335-46. PubMed.

  18. It would seem reasonable to extrapolate from the current data that loss of appropriate protein function is due to partial loss of, or change of, intermediate RNA editing instructions due to genetic mutation. More than one gene is affected in this case of research into presenilin I and II. A question worth asking is whether domain splicing mechanisms are involved here, and whether they are relevant to finding gene-based therapeutic targets.

  19. How Might Presenilin-1 (PS1) Mutations Induce Early Onset Alzheimer Disease?
    Shen and Kelleher have suggested that presenilin dysfunction can cause AD phenotypes, because they found that presenilin double knockout mice exhibited synaptic dysfunction, tau hyperphosphorylation, neuronal loss, and memory impairment without Aβ deposition (1). Their hypothesis may be plausible, especially since FAD mutations of PS1 were recently reported to reduce the normal function of PS1 (2-4). However, the presenilin double knockout mouse shows no Aβ deposition, which is required in AD, so it is not an AD model. Moreover, Shen’s laboratory previously found that presenilin double knockout mice had reduced Aβ deposition and rescued an Aβ-induced behavioral deficit (5,6), and other groups reported that mutant PS1 accelerates Aβ deposition and memory impairment (7,8). Therefore, even while they both appear to downregulate PS function, the gap between PS deficiency and PS1 mutation with regard to Aβ-induced phenotype is large.

    Even if a downregulation of PS function is involved in AD, differences in PS levels could not be found when age-matched normal and AD brains were compared. In fact, the total level of PS1 tends to increase with aging (9). Therefore, the claim that PS dysfunction is involved in the development of familial and sporadic AD may be difficult to accept. Many PS1 mutations do increase the ratio of Aβ42 and 40, and a PS1 mutation was found in an FTD patient showing tau pathology without Aβ deposition (10); this suggests that PS1 mutations might contribute to Aβ and tau pathology, albeit independently. PS1 mutant knock-in mice exhibited increased Aβ42/40 and detergent-insoluble tau filaments without Aβ deposition (because of less aggregate formation with mouse Aβ) (11). We found that the Aβ42/40 ratio commonly increased with 28 different PS1 mutations, in a mutation-specific manner (12), and that PS1 mutations reduce Aβ40 production, which increases the ratio of Aβ42/40 (13). The increase in Aβ42/40 ratio enhances Aβ aggregation and toxicity. These in-vivo, and in-vitro findings do indicate that PS1 mutations affect both Aβ processing and NFT formation. Therefore, mutations in PS1 may cause Aβ and tau to contribute to the development of AD via two different processes.

    Consider the results of the Braak and Braak study carefully (14,15). They show that NFT formation in entorhinal cortex precedes Aβ deposition, with NFTs appearing in the limbic- and neo-cortices only after Aβ deposition. Based on this analysis, one could argue that NFT formation in entorhinal cortex occurs during the course of normal brain aging, and only spreads into limbic and neocortical areas in AD (16). Thus, the chronology of the pathological changes appears to start with the appearance of NFTs in the entorhinal cortex in the absence of Aβ, which is then followed by Aβ deposition in other regions that triggers the spread of NFTs into the limbic and neo- cortices. This suggests that brain aging can be characterized by NFTs in entorhinal cortex and that Aβ accelerates this brain aging to cause AD. PS1 mutations might affect NFT formation in entorhinal cortex independently of Aβ deposition and accelerate the rate of brain aging even as a PS1 mutation facilitates the formation of toxic Aβ aggregates, thereby inducing early onset Alzheimer disease.


    . APP processing and synaptic plasticity in presenilin-1 conditional knockout mice. Neuron. 2001 Sep 13;31(5):713-26. PubMed.

    . Presenilin clinical mutations can affect gamma-secretase activity by different mechanisms. J Neurochem. 2006 Feb;96(3):732-42. PubMed.

    . Loss-of-function presenilin mutations in Alzheimer disease. Talking Point on the role of presenilin mutations in Alzheimer disease. EMBO Rep. 2007 Feb;8(2):141-6. PubMed.

    . FAD mutants unable to increase neurotoxic Abeta 42 suggest that mutation effects on neurodegeneration may be independent of effects on Abeta. J Neurochem. 2007 May;101(3):674-81. Epub 2007 Jan 24 PubMed.

    . Reduced beta-amyloid production and increased inflammatory responses in presenilin conditional knock-out mice. J Biol Chem. 2004 Nov 5;279(45):46907-14. Epub 2004 Sep 1 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.

    . Accelerated Alzheimer-type phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes. Nat Med. 1998 Jan;4(1):97-100. PubMed.

    . Behavioral changes in transgenic mice expressing both amyloid precursor protein and presenilin-1 mutations: lack of association with amyloid deposits. Behav Genet. 1999 May;29(3):177-85. PubMed.

    . Presenilin 1 associates with glycogen synthase kinase-3beta and its substrate tau. Proc Natl Acad Sci U S A. 1998 Aug 4;95(16):9637-41. PubMed.

    . A novel presenilin 1 mutation associated with Pick's disease but not beta-amyloid plaques. Ann Neurol. 2004 May;55(5):617-26. PubMed.

    . Formation of tau inclusions in knock-in mice with familial Alzheimer disease (FAD) mutation of presenilin 1 (PS1). J Biol Chem. 2006 Feb 24;281(8):5037-41. PubMed.

    . Enhancement of amyloid beta 42 secretion by 28 different presenilin 1 mutations of familial Alzheimer's disease. Neurosci Lett. 1999 Apr 9;265(1):61-3. PubMed.

    . Decreased Abeta secretion by cells expressing familial Alzheimer's disease-linked mutant presenilin 1. Neurosci Res. 2007 Mar;57(3):446-53. PubMed.

    . Frequency of stages of Alzheimer-related lesions in different age categories. Neurobiol Aging. 1997 Jul-Aug;18(4):351-7. PubMed.

    . Staging of Alzheimer-related cortical destruction. Int Psychogeriatr. 1997;9 Suppl 1:257-61; discussion 269-72. PubMed.

  20. Distilling the general comments and feedback from colleagues, most agree (as do I) that this new perspective is needed and appreciated, and does an amicable job of reconciling the unsettled role of β amyloid with presenilin loss of function. Taking a step back, however, and doing some more polling, it is apparent that not enough is really known about the functional roles of PS in neurons to draw definitive conclusions about its contributions to AD once it is impaired. Exposing my bias, there is a wealth of data about presenilin function and its substrates, but largely in non-neuronal cell types. Yet, AD affects neurons specifically. The model cell systems are invaluable for uncovering many of presenilin’s actions, but there still must be something different going on within neurons, and even subpopulations of neurons.

    One aspect of cell signaling that has eccentricities unique to neurons is calcium regulation. In line with Ilya Bezprozvanny’s and Ming Chen’s comments, there is strong evidence that PS is directly linked to calcium regulation in neurons, although this was regrettably not discussed in detail in the PNAS Perspective. And, calcium dysregulation can be linked to the major predictors, markers, and symptoms of AD (reviewed in Stutzmann, 2007), including acceleration of amyloid fibril deposition (Isaacs et al., 2006), hyperphosphorylation of tau (Avila et al., 2004), synaptic and plasticity deficits, cell death (Boehning et al., 2004), ApoE4-mediated pathology (Veinbergs et al., 2002), and, of course, PS mutations (Stutzmann et al., 2006). Although every AD case may not immediately involve loss of PS function, perhaps a common denominator may be at work.

    Perhaps a next step in this well-received new perspective is a detailed look at neuronal PS function, and how it differs from its role in cells unaffected in AD. Identifying and isolating systems that are unique to neurons may be a good place to start.


    . Tau in neurodegenerative diseases: tau phosphorylation and assembly. Neurotox Res. 2004;6(6):477-82. PubMed.

    . Apoptosis and calcium: new roles for cytochrome c and inositol 1,4,5-trisphosphate. Cell Cycle. 2004 Mar;3(3):252-4. PubMed.

    . Acceleration of amyloid beta-peptide aggregation by physiological concentrations of calcium. J Biol Chem. 2006 Sep 22;281(38):27916-23. PubMed.

    . Enhanced ryanodine receptor recruitment contributes to Ca2+ disruptions in young, adult, and aged Alzheimer's disease mice. J Neurosci. 2006 May 10;26(19):5180-9. PubMed.

    . The pathogenesis of Alzheimers disease is it a lifelong "calciumopathy"?. Neuroscientist. 2007 Oct;13(5):546-59. PubMed.

    . Neurotoxic effects of apolipoprotein E4 are mediated via dysregulation of calcium homeostasis. J Neurosci Res. 2002 Feb 1;67(3):379-87. PubMed.

  21. Closing Comment
    I think the reviews and commentaries posted on this Forum Discussion reflect a nice balance. They should prompt us, without excluding Aβ42, to consider Aβ-independent mechanisms that might be coactive or act synergistically with Aβ to cause AD. To me, Aβ is so central to all observable AD-related events that to try and exclude it or move it to the periphery in any single, alternate cascade makes the alternative hypothesis even more incomplete than the Aβ cascade.

    In the Aβ versus PSEN debate, two simple examples of amyloidogenesis-mediated degeneration are cited that should not involve PSEN. One is of non-Aβ-mediated progressive dementia. The other example is of Aβ-mediated amyloidogenesis due to APP mutations at the α-secretase site, also discussed here by Vincent Marchesi. I won't go into the details of this hypothesis by Vincent, but it is certainly interesting, addressing the oxidized RNA-mediated mutant peptide production as suggested by the G¬C or G¬A switch in some of these α-secretase site APP mutations (Flemish and Arctic APP do not involve this switch). The altered aggregation properties of mutant Aβ itself provide further proof that Aβ-led degeneration is independent of PSEN unless we presume that these mutations, too, affect γ-secretase processing.

    On the other hand, Aβ is not monolithic in causing the disease. This was clearly suggested by the early clinico-pathologic-genetic observations that an average PSEN mutation is more drastic than an average APP mutation (with comparable alterations in Aβ or Aβ subspecies). It is therefore good to see that all these γ-secretase dependent or independent pathways are now neatly fitting into the common pathways either directly or as sub-branches. I see a consensus that AD might be a syndrome where more than one mechanism is responsible for a common pathological and clinical phenotype; and all this occurs in the foreground of aging which progressively weakens our anti-disease reserves.

    Thus, I believe that if we continue to stay open-minded without giving up our specific molecular or pathway loyalties, many of the pathways proposed here will eventually turn out to be correct, even if they may be not eventually carry the same weight as we assign them now. Reviews, commentaries, and excellent moderation are testimony to this effort.

Make a Comment

To make a comment you must login or register.


Book page Citations

  1. Contact Us

Webinar Citations

  1. Gain or Loss of Function—Time to Shake up Assumptions on γ-Secretase in Alzheimer Disease?

News Citations

  1. New Orleans: Symposium Probes Why Synapses Are Suffering

Paper Citations

  1. . Loss of presenilin function causes impairments of memory and synaptic plasticity followed by age-dependent neurodegeneration. Neuron. 2004 Apr 8;42(1):23-36. PubMed.
  2. . Dementia with prominent frontotemporal features associated with L113P presenilin 1 mutation. Neurology. 2000 Nov 28;55(10):1577-8. PubMed.
  3. . A novel presenilin 1 mutation associated with Pick's disease but not beta-amyloid plaques. Ann Neurol. 2004 May;55(5):617-26. PubMed.
  4. . Frontotemporal dementia-like phenotypes associated with presenilin-1 mutations. Am J Alzheimers Dis Other Demen. 2006 Aug-Sep;21(4):281-6. PubMed.
  5. . Frontotemporal dementia and parkinsonism associated with the IVS1+1G->A mutation in progranulin: a clinicopathologic study. Brain. 2006 Nov;129(Pt 11):3103-14. PubMed.
  6. . A presenilin 1 mutation associated with familial frontotemporal dementia inhibits gamma-secretase cleavage of APP and notch. Neurobiol Dis. 2002 Mar;9(2):269-73. PubMed.

Other Citations

  1. document

External Citations

  1. Full text
  2. Full text
  3. Full text
  4. Journal of Neurochemistry
  5. See full text
  6. Neurodegenerative Diseases
  7. AD/FTD mutation database
  8. EMBO reports

Further Reading


  1. . A stop-codon mutation in the BRI gene associated with familial British dementia. Nature. 1999 Jun 24;399(6738):776-81. PubMed.