Reply by Dennis Selkoe
Dear John:
In response to your News and Views in the 8 April issue of Nature, I thought it would be helpful to indicate several points about your comments regarding the role of presenilins in APP processing with which my colleagues and I disagree. While your central conclusion that our experiments do not prove that presenilins (PS) are γ-secretases is true, your statement that "the case is far from nailed down" and your citing of evidence that all of the available data could readily be explained by a role of PS in membrane trafficking ignores or misunderstands some of the data from our and others' work, as follows.

1. We took pains to show that there was no evidence of an alteration of subcellular distribution of either holoPS and its fragments or of APP, C99 and C83 in cells expressing asp-mutant vs wt PS1 (Wolfe et al, Nature, 1999). Moreover, we have published that there is no change in the subcellular distribution of APP, C99 and C83 in cells that entirely lack PS1 (Xia et al, Biochem. 1998). These data...  Read more

Reply by Dennis Selkoe
Dear John:
In response to your News and Views in the 8 April issue of Nature, I thought it would be helpful to indicate several points about your comments regarding the role of presenilins in APP processing with which my colleagues and I disagree. While your central conclusion that our experiments do not prove that presenilins (PS) are γ-secretases is true, your statement that "the case is far from nailed down" and your citing of evidence that all of the available data could readily be explained by a role of PS in membrane trafficking ignores or misunderstands some of the data from our and others' work, as follows.

1. We took pains to show that there was no evidence of an alteration of subcellular distribution of either holoPS and its fragments or of APP, C99 and C83 in cells expressing asp-mutant vs wt PS1 (Wolfe et al, Nature, 1999). Moreover, we have published that there is no change in the subcellular distribution of APP, C99 and C83 in cells that entirely lack PS1 (Xia et al, Biochem. 1998). These data are consistent with the findings of Struhl and Greenwald and Ye et al that Notch subcellular distribution is not detectably changed in PS(-) flies. Rather, all 4 Nature papers suggest that proteolytic processing events are either mediated or regulated by PS. In my opinion, none of the new data provides evidence for a principal role of PS in trafficking. If PS functioned in protein trafficking, one might expect a potentially widespread disturbance of trafficking of multiple proteins, not just an effect on the proteolytic release of the intracellular domain of Notch to the nucleus (and the cleavage of APP within the membrane). Parenthetically, I was surprised to see you say "alteration of substrate trafficking is the interpretation favored by Ye and colleagues", as I did not get this impression from their paper. They do state that their results "make it improbable that PS is required specifically for the cleavage that occurs within or just C-terminal to the transmembrane domain that translocates to the nucleus," but I did not see evidence in their paper that PS functions to regulate protein trafficking.

2. You cite an older hypothesis about a possible trafficking role for presenilins (based principally on their homology to spe4), but I don't know of any compelling data indicating such a role. Randy Nixon stated at Keystone that he observed differences in endosomal and lysosomal size and immunostaining for Aß and certain hydrolases in PS-mutant vs sporadic AD brains, but he provided no evidence that I heard that "presenilin and APP mutations have different effects on the vesicular trafficking of APP", as you state in your News and Views. I wonder whether the evidence of different endosomal/lysosomal morphological patterns in APP-mutant vs PS-mutant AD brain tissue which Randy described may represent later effects of the disease process; in any event, I do not think these neuropathological data speak to issues of APP and PS protein trafficking.

3. In citing your data that ApoE genotype doesn't influence age of onset of AD in PS mutation carriers, you suggest that APP-717 mutations affect γ-secretase cleavage (certainly true) but that PS mutations act differently somehow. But numerous labs have produced strong evidence that it is the γ-secretase cleavage of APP that is also being altered by the PS mutations. So, there is no compelling evidence that these two genetic forms of FAD involve fundamentally distinct biochemical processes. I assume that the apparent lack of an ApoE4 effect on PS-mutant disease onset (assuming that this is confirmed when many additional PS1 and PS2 FAD patients and their brains are examined) is because the Aß42-elevating effect of PS mutations is so substantial (and life-long) that any decreased clearance of Aß40 that may result from expression of the ApoE4 protein does not contribute sufficiently to disease progression to be clinically quantifiable as an even earlier disease onset. In any event, I don't think these clinical onset data can be cited as evidence that APP and PS mutations work through fundamentally distinct pathogenic processes. Instead, I suspect that the APP mutations make the substrate of the γ-secretase reaction mutant, whereas the PS mutations make the protease (or an essential co-factor thereof) of this reaction mutant.

4. I would speculate that PS probably will be shown to be at the site of both Notch and APP intramembranous cleavage events, including at the plasma membrane and in endosomes. The apparent ER/Golgi localization of PS in transfected non-neural cells reported to date is unlikely to be the final word, given the evidence for an apparent plasma membrane locus in fly.

5. Your citing of a membrane trafficking role ignores the fact that PS NTF/CTF heterodimers, APP, C99, C83, Aß40 and Aß42 have all been found together in isolated ER- and Golgi-rich vesicles of transfected cells (e.g., Xia et al, Biochem. 1998). In vitro incubation of these vesicles at 37oC generates new Aß (Wolfe et al, Nature, 1999). Moreover, we can co-immunoprecipitate APP (both N- and N+O-glycosylated) from these same vesicles with PS antibodies (Xia et al, Biochem. 1998). While I agree that co-ip of the two proteins is difficult to obtain, this does not mean they never interact, and we have consistent data that small amounts of holoAPP can be co-ip'd with PS in transfected cells. We are now attempting to capture C99/C83 with PS in a co-ip experiment. We believe the co-ip of PS and APP (published by at least 3 labs now), taken together with the clear colocalization of the Aß-generating components (PS NTF/CTF and C99) within the same purified vesicles, are more consistent with a role for PS in catalytic complex formation with APP than as a regulator of the trafficking of APP or γ-secretase to each other.

6. Our data showing in vitro generation of Aß in microsomes only at mildly acidic, not neutral, pH and only with wt PS, not when either transmembrane aspartate is mutated independently to alanine, are much more reasonably explained as an effect on a proteolytic reaction (probably as an aspartyl protease) than as an effect on membrane trafficking. Furthermore, the evidence that difluroketone and difluoroalcohol peptidomimetic compounds made to mimic the APP γ-secretase cleavage site inhibit both this cleavage (Wolfe et al, J Med Chem 1988; Wolfe et al, Biochem, in press) and that of Notch (DeStrooper et al, Nature, 1999) clearly suggests an aspartyl protease mechanism for the responsible protease(s), consistent with our in vitro results. We feel that the latter data fit much better with a proteolytic process than a membrane trafficking process. We are now attempting to show that such designed APP γ-secretase inhibitors bind to presenilin, but only when it is wt, not when a single TM asp is mutated. Finally, the inhibition of both PS endoproteolysis and γ-secretase cleavage by substituting glu for asp is also more consistent with an aspartyl protease activity than a trafficking role for PS. While PS could turn out to be a very unusual intramembranous diaspartyl co-factor for γ-secretase rather than the actual protease (we feel this is unlikely), we don't see any positive evidence for a third possibility: that it acts principally in membrane trafficking.

7. Finally, our evidence that mutating either TM asp of PS1 [or of PS2 (C. Haass et al, Taos NM, 3/99)] blocks PS endoproteolysis means that these 2 asps are essential for two independent proteolytic cleavages: C99 and holoPS. To these we can probably add the cleavage of Notch (asp-mutant experiments underway) and perhaps those of APLP1 and APLP2. We believe the most parsimonious explanation for the involvement of PS in the cleavage of arguably 5 different substrates is as an intramembranous aspartyl protease. In my view, the most substantive concern about our hypothesis is that it is without clear precedence in biology.

Because these issues are important both biologically and medically, I thought I would provide details of why I don't concur with some of your conclusions in your News and Views. In the interest of opening up a further scientific dialogue on these and related issues, I agree with your suggestion that this letter be posted on the Alzheimer Research Forum website. I look forward to further discussion of these and other unresolved questions.

Best regards,

Dennis

View all comments by Dennis Selkoe

Question from Ratan Bhat
There have been suggestions that an increase in beta-catenin could mask some the mutation sites within PS1 (Tanzi and Finch in their Science comment), thereby stabilizing PS1, decreasing γ-secretase activity and subsequently β amyloid levels. Is this a valid hypothesis? If so, what is the evidence?

View all comments by Ratan Bhat

Reply to Ratan Bhat by Rudy Tanzi:
The main problem with this hypothesis is that since the time of publication of the commentary by Tuck Finch and I in Science, we now know that PS2 does not bind β- or delta-catenin (Tesco et al., J. Biol Chem., 1998). Yet, when PS2 carries FAD mutations it has the same molecular phenotype as mutant PS1 i.e. drives up relative levels of AβX-42. One could argue that perhaps other loop-binding proteins that interact with PS2 (e.g. Sorcin; Kim et al., 1999) take the place of β-catenin with similar effects according to the hypothesis under consideration. The problem with both of these potential scenarios is that Gopal Thinakaran has recently shown that most of the hydrophylic portion of the loop between TM6 and TM7 of PS1 can be removed (including the β-catenin binding site) with no apparent effect on the ability of FAD mutations (introduced into the same construct) to increase relative levels of AβX-42. Thus, while I can believe that stabilization of PS1 (and γ-secretase activity) may contribute to the increased production of AβX-42 and...  Read more

Reply to Ratan Bhat by Rudy Tanzi:
The main problem with this hypothesis is that since the time of publication of the commentary by Tuck Finch and I in Science, we now know that PS2 does not bind β- or delta-catenin (Tesco et al., J. Biol Chem., 1998). Yet, when PS2 carries FAD mutations it has the same molecular phenotype as mutant PS1 i.e. drives up relative levels of AβX-42. One could argue that perhaps other loop-binding proteins that interact with PS2 (e.g. Sorcin; Kim et al., 1999) take the place of β-catenin with similar effects according to the hypothesis under consideration. The problem with both of these potential scenarios is that Gopal Thinakaran has recently shown that most of the hydrophylic portion of the loop between TM6 and TM7 of PS1 can be removed (including the β-catenin binding site) with no apparent effect on the ability of FAD mutations (introduced into the same construct) to increase relative levels of AβX-42. Thus, while I can believe that stabilization of PS1 (and γ-secretase activity) may contribute to the increased production of AβX-42 and that FAD mutations may somehow enhance this, it would not appear that loop-binding partners (e.g., catenins and sorcin) are necessarily required for stabilization. In addition, more direct experimental evidence is needed to make the case for increased stabilization of presenilins carrying FAD mutations.

One final note. It would be very helpful to identify the protein(s) that stabilize the limited pool of full-length presenilin that is ultimately targeted for endoproteolysis. Once it is identified, one could ask whether FAD mutations affects this interaction. The most likely bet for the interacting protein(s) that may stabilize full-length presenilins, thereby leading to a saturable pool of endoproteolytic fragments, is that they bind the common C-terminal regions of the presenilins. And, it is interesting deletions on the C-terminus have been shown by Dr. Iwatsubo to abrogate effects on Aβ production. We will just have to wait for the data to emerge. Luckily, in this field, it shouldn't be too long a wait.

View all comments by Rudy Tanzi

Reply by E. Preddie and J. Bergmann
1. Considering what was known before, the nature of the experiments done by Wolfe and colleagues and the results obtained, it appears that the conclusion expressed by 'Ye' et al from their results is equally valid for the results of Wolfe et al..

2. It has been fairly well established that physical contact occurs between APP and PS1 during APP "processing". Such physical contact is consistent with an enzyme substrate reaction but it is also consistent with other types of interactions which occur during vesicular transport, e.g., mRNA encoding membrane associated proteins interacting with membrane associated mRNA binding protein complexes.

3. The four other 'substrates' for PS1/ PS2 not withstanding, we suggest that the experimental results obtained by Wolfe et al .indicate that they may have discovered, within membrane associated PS1, 'core components' of a binding/interacting site for APP or APP associated molecules.

4. The same results might also indicate that: (i) interaction between APP and PS1 is a very early event...  Read more

Reply by E. Preddie and J. Bergmann
1. Considering what was known before, the nature of the experiments done by Wolfe and colleagues and the results obtained, it appears that the conclusion expressed by 'Ye' et al from their results is equally valid for the results of Wolfe et al..

2. It has been fairly well established that physical contact occurs between APP and PS1 during APP "processing". Such physical contact is consistent with an enzyme substrate reaction but it is also consistent with other types of interactions which occur during vesicular transport, e.g., mRNA encoding membrane associated proteins interacting with membrane associated mRNA binding protein complexes.

3. The four other 'substrates' for PS1/ PS2 not withstanding, we suggest that the experimental results obtained by Wolfe et al .indicate that they may have discovered, within membrane associated PS1, 'core components' of a binding/interacting site for APP or APP associated molecules.

4. The same results might also indicate that: (i) interaction between APP and PS1 is a very early event in expression of APP and Aß40; this may involve APP mRNA transport, APP mRNA anchoring or APP mRNA translation, and (ii) that either alterations within or in the vicinity of the membrane spanning segments of PS1 protein or point mutations in APP mRNA exon 16 or 17 can alter the size of the translated Aß component of APP mRNA, (abolishment of interaction with ligand due to mutations within or close to transmembrane signals in seven- transmembrane helix proteins is a known phenomenon).

5. Two suggestions are inferred in #4 above: (i) PS1 is involved in a membrane associated RNA binding protein complex which controls translation of APP mRNA, and (ii) Aß is produced at the level of translation of APP mRNA.

6. We know of no data, so far, that shows PS1 is associated with a membrane associated RNA binding protein complex; (but also there is no evidence that it is not.). However, there is very good structural evidence, relationship evidence and published evidence, that a potential bi-functional translational control element we have discovered within the APP mRNA can control, independently, expression of ~100 K APP protein and Aß, all lengths. Firmly establishing the biological relevance of this translational control element will obviate the need to postulate γ secretase.

View all comments by Johanna Bergmann

Reply by Steven Barger
Regardless of whether presenilins really are γ-secretases (for which the data seem to tipping the scales), the data seem quite clear that they are somehow involved in proteolytic processing of both APP and Notch. To me, one of the most intriguing implications is that this may hint at functional connections between APP and Notch (and APLPs?). It's difficult to imagine that evolution would have parsed these into the same processing path (apparently exclusive of many other proteins) unless there was a good reason. Of course, there is independent evidence that both are involved in cell attachment and the connection of cell attachment/recognition to intracellular events. But the potential similarities may warrant a little closer look at the intracellular fate/effects of the cytosolic tail of APP. PS mutations could obviously influence the generation of an intracellular fragment. It is notable that the transcriptional role of the Notch intracellular fragment went undetected for some time because the exceedingly small amounts necessary for...  Read more

Reply by Steven Barger
Regardless of whether presenilins really are γ-secretases (for which the data seem to tipping the scales), the data seem quite clear that they are somehow involved in proteolytic processing of both APP and Notch. To me, one of the most intriguing implications is that this may hint at functional connections between APP and Notch (and APLPs?). It's difficult to imagine that evolution would have parsed these into the same processing path (apparently exclusive of many other proteins) unless there was a good reason. Of course, there is independent evidence that both are involved in cell attachment and the connection of cell attachment/recognition to intracellular events. But the potential similarities may warrant a little closer look at the intracellular fate/effects of the cytosolic tail of APP. PS mutations could obviously influence the generation of an intracellular fragment. It is notable that the transcriptional role of the Notch intracellular fragment went undetected for some time because the exceedingly small amounts necessary for biological activity were below the detection limits of the method of localizing it (immunofluorescence). Perhaps, more emphasis should be placed on the studies of Fe65, X11, N-PAK, and other proteins that interact with the APP C-terminus.

View all comments by Steve Barger