β- but not γ-secretase proteolysis of APP causes synaptic and memory deficits in a mouse model of dementia

A mutation in the BRI2/ITM2b gene causes loss of BRI2 protein leading to familial Danish dementia (FDD). BRI2 deficiency of FDD provokes an increase in amyloid-β precursor protein (APP) processing since BRI2 is an inhibitor of APP proteolysis, and APP mediates the synaptic/memory deficits in FDD. APP processing is linked to Alzheimer disease (AD) pathogenesis, which is consistent with a common mechanism involving toxic APP metabolites in both dementias. We show that inhibition of APP cleavage by β-secretase rescues synaptic/memory deficits in a mouse model of FDD. β-cleavage of APP yields amino-terminal-soluble APPβ (sAPPβ) and β-carboxyl-terminal fragments (β-CTF). Processing of β-CTF by γ-secretase releases amyloid-β (Aβ), which is assumed to cause AD. However, inhibition of γ-secretase did not ameliorate synaptic/memory deficits of FDD mice. These results suggest that sAPPβ and/or β-CTF, rather than Aβ, are the toxic species causing dementia, and indicate that reducing β-cleavage of APP is an appropriate therapeutic approach to treating human dementias. Our data and the failures of anti-Aβ therapies in humans advise against targeting γ-secretase cleavage of APP and/or Aβ.

Thank you for the submission of your manuscript to EMBO Molecular Medicine. We have now heard back from the three referees whom we asked to evaluate your manuscript.
You will see that they all find the topic of your manuscript potentially interesting but particularly referee #2 feels that the data need to be strengthened and makes constructive suggestions for that. Should you be able to address these criticisms in full, we would be willing to consider a revised manuscript.
Given these evaluations, I would welcome the submission of a revised version for further consideration and depending on the nature of the revisions, this may be sent back to the referees for another round of review. Please note that it is EMBO Molecular Medicine policy to allow only a single round of revision and that, as acceptance or rejection of the manuscript will depend on another round of review, your responses should be as complete as possible.
Revised manuscripts should be submitted within three months of a request for revision; they will otherwise be treated as new submissions, except under exceptional circumstances in which a short extension is obtained from the editor. Also, the length of the revised manuscript may not exceed 60,000 characters (including spaces) and, including figures, the paper must ultimately fit onto optimally ten pages of the journal. You may consider including any peripheral data (but not methods in their entirety) in the form of Supplementary information. I look forward to seeing a revised form of your manuscript as soon as possible.
This manuscript addresses how the amyloid precursor protein is involved in the modulation of synaptic and memory impairment observed in FDD knockin mice, a model of familial Danish dementia. FDD KI mice develop synaptic and memory deficits due to loss of BRI2, and these phenotypes are dependent on the presence of APP. BRI2 also binds APP and inhibits its processing. In this report, the authors mapped the APP interactive domain of BRI to amino acid residues 74-102 in the extracellular region, and further showed that BRI binds to the beta-cleavage site of APP, blocking BACE-mediated cleavage of APP. Inhibition of beta-but not gamma-cleavage of APP restores LTP and memory deficits in FDD KI mice, suggesting that BACE -cleaved products are involved in the mediation of memory and synaptic impairment. The current study is a natural extension of the authors' impressive work in the elucidation of the pathogenic mechanism underlying familial Danish dementia, and provide new mechanistic insight into how APP is involved in the regulation of BRI2-dependent memory and synaptic function.

Referee #2
In this manuscript Tamayev and colleagues provide data to show that inhibiting β-secretase cleavage of APP (using either beta secretase inhibitor IV or MoBA, the peptide derived from the BRI2 protein) rescues synaptic/memory deficits in a mouse model of FDD, while inhibition of -secretase cleavage did not ameliorate the synaptic/memory deficits of FDD mice. Therefore, the authors conclude that sAPP β and/or β-CTF, rather than Aβ, are the toxic species of APP causing dementia. They also suggest that reducing the beta-cleavage of APP is an appropriate therapeutic approach to treating human dementias.
The finding is interesting and provocative. Although the results are potentially important and the MoBA peptide used to inhibit beta-cleavage of APP could be a candidate for the therapeutic treatment of dementia, there are some major concerns that the authors should address: Major concerns: 1.The authors need to be cautious in formulating their main conclusion. Inhibition of -secretase will lead to accumulation of β-CTF, as shown in Fig.3C. Since sAPPβ and β-CTF are toxic, the failure of γ-secretase inhibitor to rescue the LTP and memory deficits of FDD mice could be due to the increased levels of β-CTF. Therefore, the conclusion that Aβ is not the primary toxic species causing dementia needs to be better justified.
2.Most of the cellular experiments were performed in HEK or Hela cells, which do not fully represent the physiological events of neurons/astrocytes/glial cells in the brain.
3.In figure 1D, blots for the internal control (such as a-tubulin) and sAPP are necessary to make the data more comprehensive. In figure 2A and B, the alpha-/beta-CTF blots are lacking as well. Showing this additional data may help us to understand if both sAPP-beta and beta-CTF contribute similarly to dementia in FDD KI mice or if sAPP-beta is the dominant factor. 4.Although electrophysiological/behavioral tests were performed, direct data covering how those treatments were performed in vivo needs to be shown. Considering that the inhibitors were injected only 1h before the training/testing for the behavioral tests or perfused for 6min for LTP measurements, it's hard to imagine how much they could have affected the proteolytic products of APP and, in turn, downstream synaptic events.
2. It would be better to show the Aβ level upon γ-secretase inhibitor treatment in Fig.3C.
3. The y-axis label in figure 1E has a problem.
4. The detailed sequences of peptides N1 to N8 need to be included. 5. The manuscript should be carefully proof-read as there are many errors.
Referee #3 (Comments on Novelty/Model System): Authors reported their findging that a small peptide derived from Br12 inhibits BACE1 processing of APP. This is a quite novel and this inhibitory peptide appears to inhbit BACE1 processing of APP specifically. More interestingly, authors showed that BACE1-cleaved products had more impairment on LTP and memory function in their animal models, emphasizing the precaution of BACE1 inhibition for AD therapy. The overall study was well done and results are convincing. The publication of this manuscript will attract great attentions from AD researchers.

Referee #3 (Other Remarks):
The manuscript by Tamayev et al. described their findings that a small region with BR12 binds to APP surrounding its BACE1 cleavage site. More interestingly, peptide named N3 or N8 can inhibit BACE1 cleavage of APP, and this inhibition is binding-dependent. Their further mutagenesis studies showed that the mutation of first two residues in N3 enhanced its inhibition, and they renamed N3-2A to MoBA for Modulator of b-cleavage of APP. In their LTP test, MoBA showed strong reversal of LTP deficits seen in FDDki samples while gamma-secretase inhibitor compound E showed no reversal of LTP deficits. Similarly, 9-month-old FDDki mice showed deficits in the novel object recognition test, and MoBA treatment reversed this deficit. Based on these results, authors concluded that inhibition of BACE1 should be practiced with caution. Overall, this study is well done and results are convincing.
Minor suggestions: 1. The y-axis in figure 1E should be labeled clearly. 2. Numbers corresponding to N1 to N8 in figure 2A should be specified. 3. ADan should be refereed clearly in the introduction. 4. The description of figure 1E can be extended to emphasize that residue between 102-134 retained inhibitory effects. Thank you for orchestrating the revision of this manuscript. We also want to thank the reviewers for the encouraging comments and helpful suggestions. To facilitate the reviewers' task, the text that has been modified or inserted in response to the reviewers' suggestions is highlighted using red characters.
The first reviewer finds no issues with the work and underlines that this "study is a natural extension of the authors' impressive work in the elucidation of the pathogenic mechanism underlying familial Danish dementia, and provide new mechanistic insight into how APP is involved in the regulation of BRI2-dependent memory and synaptic function." We thank the reviewer for the insightful and positive comments.
The second reviewer finds the work "interesting and provocative" and the results "potentially important". He/She also underlines that "the MoBA peptide used to inhibit beta-cleavage of APP could be a candidate for the therapeutic treatment of dementia". MoBA, I may add, is "ideologically neutral", because it would be effective also in preventing Ab-mediated symptoms, and not only those triggered by sAPPb/b-CTF. This concept is now expressed at the end of the Discussion. However, the reviewer also raises several (major and minor) points. All the criticisms are explicitly motivated and well articulated. Following, is a point-by-point response to this reviewer's comments.
Major concerns: Question1. The authors need to be cautious in formulating their main conclusion. Inhibition of g-secretase will lead to accumulation of b-CTF, as shown in Fig.3C. Since sAPPb and b-CTF are toxic, the failure of g-secretase inhibitor to rescue the LTP and memory deficits of FDD mice could be due to the increased levels of b-CTF. Therefore, the conclusion that Ab is not the primary toxic species causing dementia needs to be better justified.
Response1. We understand the reviewer's viewpoint and have addressed his/her helpful criticism as follows: Abstract: the sentence "This result demonstrates that sAPPb and/or b-CTF, rather than Ab, are the toxic species causing dementia" has been substituted by the sentence "These results suggest that sAPPb and/or b-CTF, rather than Ab, are the toxic species causing dementia".
Discussion: We have substituted the sentence: "In addition, they indicate that metabolites derived from g-cleavage of APP, such as Ab, P3 and AID/AICD, are not involved in these pathogenic processes (Fig. 4C)" with the following text: "The failure of GSI to rescue the deficits of FDDKI mice suggests that metabolites derived from g-cleavage of APP, such as Ab, P3 and AID, are not involved in these pathogenic processes (Fig. 4C). Alternatively, the GSI may have a beneficial effect on LTP/memory deficits, possibly due to a reduction in g-secretase-derived APP fragments, that is counterbalanced and masked by increased levels of toxic b-CTF caused by inhibition of g-cleavage of APP (Fig. 3C). Regardless, the data imply that inhibition of g-processing of APP, even using gsecretase modulators that inhibit APP cleavage sparing g-processing of other substrates, may be therapeutically ineffective in correcting memory deficits and, per chance, harmful on account of the increase in b-CTF." We have substituted the sentence "The inference that Aβ does not cause synaptic and memory dysfunction in FDDKI mice" with "The inference that Aβ may not cause synaptic and memory dysfunction in FDDKI mice".
We also added the following texts: "Besides, although our data provide no evidence supporting a role for amyloid peptides in synaptic plasticity and memory deficits, amyloid plaques and/or "toxic amyloid conformers" may set off other clinical symptoms of FDD/AD patients", and "It is worth noting that loss of function induced by PSEN1 and PSEN2 FAD mutations may cause, perhaps transiently during synaptic transmission and memory acquisition, an increase in the levels of APP-CTFs including b-CTF."

Q2. Most of the cellular experiments were performed in HEK or Hela cells, which do not fully represent the physiological events of neurons/astrocytes/glial cells in the brain.
R2. We agree with the reviewer that testing the effects of peptides on processing of endogenous APP in primary (neurons/astrocytes) cultures would have been the ideal approach. However, we are unable to detect production of endogenous sAPPa and sAPPb in the supernatant of primary neuronal cultures (we have tested all the commercially available anti-sAPPa and anti-sAPPb antibodies). Therefore unfortunately, we cannot use this system to analyze the inhibitory effects of BRI2-derived peptides on b/a-processing of APP. It is our opinion however that the data showing that MoBA binds endogenous APP from brain cells (Fig. 2G) and that MoBA has a "therapeutic" effect on both synaptic plasticity (hippocampal slices, in vitro) and memory deficits (in vivo) similarly to bsecretase-inhibitor IV (Figs. 3 and 4), support the data obtained with HeLa and HEK293 cells. We have also emphasized more the experiment shown in Fig. 2G by adding the following text: "To determine whether these biological properties of N3-2A, described in cancer cell lines over-expressing APP, are conserved when using brain cells expressing endogenous APP, we tested if N3-2A-F binds endogenous APP from freshly isolated murine brain cells. As shown in Fig. 2G, N3-2A-F/endogenous APP complexes are readily detectable." Q3. In figure 1D, blots for the internal control (such as a-tubulin) and sAPP are necessary to make the data more comprehensive. In figure 2A and B, the alpha-/beta-CTF blots are lacking as well. Showing this additional data may help us to understand if both sAPP-beta and beta-CTF contribute similarly to dementia in FDD KI mice or if sAPP-beta is the dominant factor.
R3. We have provided the data requested for Fig. 1D. The Figure legend has been modified accordingly and the results are described in the section entitled "The BRI2 domain that binds APP and inhibits APP processing maps to amino acids 74-102" as follows: "The levels of a-Tubulin were similar in all transfected cells and a-Tubulin was not precipitated by anti-myc, further underlying the specificity of the mBRI2/mAPP interaction. In addition, the BRI2 deletion constructs that bind APP also reduced the levels of both sAPPb and sAPPa (Fig. 1D)." To address the second part of the question I have performed new experiments, since the samples shown in Fig. 2A and B are not available any longer. The results of these experiments are shown in Fig. 2C and are explained in the text as follows: "Since N3-2A has a strong and specific inhibitory activity on b-cleavage of APP, we further analyze the effects of N3-2A on APP processing. We measured the levels of APP-CTFs in HEK293-APP cells treated with N3-2A, control peptide N1 or media alone. APP-CTFs are rapidly turned over in cells by g-secretase. Thus, to reduce the confounding effects of APP-CTFs' turnover, we performed a parallel experiment in which cells were treated with the g-secretase inhibitor (GSI) compound-E. As shown in Fig. 2C, N3-2A reduces the levels of b-CTF as compared to cells treated with either media alone or N1, which is consistent with the inhibition of b-cleavage of APP. The a-secretase-derived APP-CTF, a-CTF, is not altered by N3-2A, again consistent with the fact that N3-2A does not change the levels of sAPPa (Fig. 2C). Similar data where obtained in HeLa-APP cells (Fig. 2C). Q4. Although electrophysiological/behavioural tests were performed, direct data covering how those treatments were performed in vivo needs to be shown. Considering that the inhibitors were injected only 1h before the training/testing for the behavioural tests or perfused for 6min for LTP measurements, it's hard to imagine how much they could have affected the proteolytic products of APP and, in turn, downstream synaptic events.
R4. The surprise of the reviewer concerning the acute therapeutic effect of b-secretase-Inhibitor IV and MoBA, is comprehensible (although the hippocampal slices were perfused for 60 min and not 6 min as stated by the reviewer). In the first part of this criticism, the reviewer asks to show "data covering how those treatments were performed in vivo". We have interpreted this as a request to clearly describe the behavioural experiment. To this end, we have extensively re-written the result section entitled "Inhibiting b-, but not g-, cleavage of APP rescues the memory deficit of FDD KI mice." In addition, we have extended the section entitled "Inhibiting b-, but not g-, cleavage of APP inhibitor rescues the LTP deficit of FDD KI mice" to include more experimental details. Moreover, we have extended the MATERIAL AND METHODS section to describe the LTP, brain cannulation and drug delivery, and NOR experiments in greater detail. Hopefully these corrections address the reviewer's criticism.
In the second part of this criticism the reviewer expresses some scepticism concerning the impact of short treatments with APP-processing inhibitors on the levels of APP metabolites. In the original manuscript we addressed this issue in the following way: Results, page 6: "These findings indicate that b-cleavage of APP during LTP prompts the synaptic plasticity deficits of FDD KI mice and suggest that de novo produced sAPPb and/or b-CTF and not Ab, are the synapto-toxic APP species"; in the legend of Fig 4C: "Two inhibitors of b-cleavage of APP (Inhibitor IV and MoBA), but not a GSI, rescue the LTP/memory deficits, suggesting that newly synthesized sAPPb and/or b-CTF, but not Ab/P3/AID cause these deficits in FDD KI mice (+ and in black)"; and in the discussion: "Our findings demonstrate that the synaptic plasticity and memory deficits in FDD are mediated through production of sAPPb and/or b-CTF during LTP and memory acquisition." Perhaps, these arguments were too weak and not properly articulated. To directly tackle the reviewer's question, we have added the following text to the discussion. We have taken the liberty to use part of the reviewer's comment that very clearly stated the conundrum. "Considering that MoBA and b-secretase-inhibitor IV were injected only 1h before the training/testing for the behavioral tests or perfused for 60 min prior to LTP measurements, it's hard to imagine that they could have manifestly affected the total hippocampal levels of proteolytic products of APP. A logical interpretation of these results is that MoBA and b-secretase-inhibitor IV exert a therapeutic effect by reducing b-cleavage of APP, perhaps in the synaptic cleft, during synaptic events leading to LTP and memory acquisition. The temporarily therapeutic efficacy of bsecretase-inhibitor IV and MoBA in this longitudinal study supports this hypothesis, and suggests that, after the drugs are cleared and APP processing by b-secretase is restored at pathological levels, memory acquisition is impaired and the mice relapse into amnesia." Minor concerns: Q1. N3 reduced both b-and a-cleavages of APP ( Fig.2A), however, N3-2A reduced b-cleavage but promoted a-cleavage of APP (Fig.2B); the authors should comment on this difference.
R1. The peptide that seems to have this effect (increase in sAPPa and decrease in sAPPb) is N3-1A and not N3-2A. To comment on this difference, we have inserted the following text at page 4: "Notably, a-cleavage of APP is inhibited by N3, unaffected by N3-2A and, probably, increased by N3-1A. It is possible that these 3 peptides bind APP differently thereby reducing (N3), unaffecting (N3-2A) or increasing (N3-1A) access of a-secretase to the APP-docking/cleavage site. However, the mechanism underlying this potentially useful difference remains to be investigated." Fig.3C.

Q2. It would be better to show the Ab level upon g-secretase inhibitor treatment in
R2. In principle we agree. However there are technical problems in performing endogenous Ab measurements in hippocampal slices. The hippocampal slices treated with compounds and used for LTP are constantly perfused with ACSF. For example, during the course of 3 hrs the slices are perfused with ~ 300 ml of ACSF. We presume that a large fraction of secreted proteins (such as Ab and sAPPs) will diffuse from the interstitial fluid (ISF) to the perfusing ACSF and only a (small?) fraction may remain trapped in the ISF inside the slices. Since it is not feasible to recover Ab/sAPPs from the perfusing ACSF, we analyzed APP-CTFs, which are membrane-bound products of APP processing and accumulate when g-secretase is inhibited. It is also worth noting that Compound E is one of the best-characterized and widely used GSI. figure 1E has a problem.

Q3. The y-axis label in
R3. Thank you for pointing this out. The labeling has been corrected.
Q4. The detailed sequences of peptides N1 to N8 need to be included.
R4. The sequences are now included.
Q5. The manuscript should be carefully proof-read as there are many errors.
R5. Thank you for pointing this out. Errors have been corrected.
The third reviewer comments that "the overall study was well done and results are convincing. The publication of this manuscript will attract great attentions from AD researchers", and makes several minor suggestions, which are addressed below. figure 1E should be labeled clearly.

Q1. The y-axis in
R1. Thank you for pointing this out. The labeling has been corrected. figure 2A should be specified.

Q2. Numbers corresponding to N1 to N8 in
R2. We have included the peptides' sequences.

Q3. ADan should be referred clearly in the introduction.
R3. To address this point we have substituted in the Introduction the sentence: "The FDD plaques contain Aβ and ADan, which derives from processing of mutant BRI2 by convertases (Vidal et al, 2000;Choi et al, 2004)." With the following text: "In normal individuals the immature BRI2 precursor (imBRI2) is cleaved by convertases in the Golgi into mature BRI2 (mBRI2), which is transported to the plasma membrane, and a C-terminal 23 amino acid peptide (Bri23), which is secreted. In the Danish kindred, the presence of a 10-nt duplication one codon before the normal stop codon produces a frame-shift in the BRI2 sequence generating a larger-than-normal precursor protein called BRI2ADan. Cleavage by convertases releases the amyloid subunit that comprises the last 34 COOH-terminal amino acids (ADan) and mBRI2. ADan accumulates into amyloid plaques, which contain both Aβ and ADan (Choi et al, 2004;Vidal et al, 2000)." Q4. The description of figure 1E can be extended to emphasize that residue between 102-134 retained inhibitory effects.
R4. Thank you. The following sentence has been added to the figure legend; "Overall, our analysis shows that BRI2 residues comprised between amino acids 102 and 134 retained APP-binding properties and inhibitory effects on APP processing." 2nd Editorial Decision 30 November 2011 Thank you for the submission of your revised manuscript to EMBO Molecular Medicine. We have now received the enclosed report from the referee who was asked to re-assess it. As you will see the reviewer is now globally supportive and I am pleased to inform you that we will be able to accept your manuscript pending the following final amendments: -The referee noted mis-labelled bands in Fig 1D and 2C. Please check and correct.
-Could you please make sure that all western blots are not too contrasted so we can see the contour and surround them with a black line to delineate the limits of the blots. In case several panels are juxtaposed, please make sure that this is clearly visible.
-The Paper Explained: Could you please rewrite the "Impact" section to address the readers, particularly the 1st sentence.
-Could you please amend the Material and Methods section to add an ethical statement regarding the use and well fare of mice used in this study.
Please submit your revised manuscript within two weeks. I look forward to seeing a revised form of your manuscript as soon as possible.
I look forward to reading a new revised version of your manuscript as soon as possible.