Ghosal K, Vogt DL, Liang M, Shen Y, Lamb BT, Pimplikar SW.
Alzheimer's disease-like pathological features in transgenic mice expressing the APP intracellular domain.
Proc Natl Acad Sci U S A. 2009 Oct 27;106(43):18367-72.
Please login to recommend the paper.
To make a comment you must login or register.
AICD as a Therapeutic Target?
AD research has produced more than 50,000 papers devoted to the clinical and molecular aspects of the disease, and a dominant theory, almost an ideology, named the amyloid cascade hypothesis. Failures of therapeutic trials based upon a modulation of Aβ secretion or Aβ polymers dissociation have raised some doubt about this hypothesis. However, the number of trials tested today is low compared to the different possibilities to control or destroy Aβ and its polymers.
This year we have observed several interesting papers showing that APP is not only the precursor of Aβ, but also a protein with powerful neurotrophic properties. Involved in these properties are different regions of the protein, at the N-terminal part (Nikolaev et al., 2009), close to the cytoplasmic membrane (Lourenco et al., 2009), and inside the cytosol, namely AICD (APP intracellular domain) (this paper). AICD is very intriguing in that the equivalent NICD from Notch has a clear function in the control of cell division, via nuclear signaling. According to some teams, AICD is a transcription factor (Cao and Sudhof, 2001) able to control gene expression (Pardossi-Piquard et al., 2005), but others have doubts (Hebert et al., 2006). Behind these discoveries is the concept of gain of function versus loss of function. Is neurodegeneration in AD due to a gain of toxic function linked to Aβ or a loss of function linked to the neurotrophic properties of APP?
The interesting aspects of the PNAS paper from Ghosal et al. is that they set up transgenic mice expressing AICD in order to decipher its possible function and role in the disease. This paper shows that overexpression of AICD is able to amplify or to generate some biochemical and cognitive parameters of AD. This is a first big step to understand better the physiological role of AICD.
Following these observations, my concern is the following: AICD is not overexpressed in cellular models with any of the mutations of PS1 that provokes degeneration in an Aβ- independent manner (Shen and Kelleher, 2007). Second, AICD may not be increased in AD. The published data are not convincing and have to be corroborated by independent teams. Opposite results have also been published, showing a collapse of all APP-CTFs (CarboxyTerminal Fragments) in AD (Sergeant et al., 2002).
At last, are these mice relevant to model the delicate physiology of AICD, i.e., with a precise targeting in the right cell and the right subcellular compartment? This is the weak point of transgenic mice in general. In fact there are several species of AICD generated from the α and β-secretase pathways. A recent paper has shown that AICD is targeted to the nucleus only if generated by β stubs from endosomes and not by α stubs from the cytoplasmic membrane (Goodger et al., 2009). This is a solid paper.
To conclude, the debate of LOF versus GOF is getting stronger, with more and more data in favor of LOF. We think that this new theory would be perfect if it allows therapeutic trials in a new spirit: not fighting the toxic, but stimulating or correcting the loss of APP neurotrophic function. AICD as a therapeutic target is certainly a good idea suggested in this paper.
The role and significance of the APP intracellular domain (AICD) has remained puzzling to many researchers. Eight years have passed since it was shown that the AICD domain could induce expression of heterologous reporter genes (Gao and Pimplikar, 2001; Cao and Südhof, 2001). Probably innumerable microarray gene expression screens have been performed in between. However, even today we do not have generally accepted target genes. Putative target genes such as KAI1, glycogen synthase kinase-3β (GSK3β) and neprilysin have been confirmed by some groups and vigorously disputed by others. Many groups have demonstrated nuclear translocation of AICD but this does not prove a function in transcription. Furthermore, the mechanism of transcriptional activation by AICD and the role of FE65 remain unclear, as well. AICD by itself does seem to be a weak transcriptional activator. On the other hand, it has been shown that FE65 alone can activate promoters of genes that have been claimed to be AICD target genes (Hebert et al., 2006). Will we ever identify undisputed target genes? My sense is that the evidence remains weak, particularly comparised Notch for which the signaling function of the intracellular domain and several target genes are well established.
Obviously, this does not exclude other important functions of the AICD domain aside from transcriptional control. Something like this seems to at work in the AICD-transgenic mice described in the new PNAS paper by Ghosal, Pimplikar and colleagues (PNAS Early Edition, October 12, 2009). Previously, the same group had shown early activation of GSK3β and increased phosphorylation of CRMP2, a GSK3β target, in these mice. GSK3β activation was not caused by transcriptional induction, as GSK3β transcript levels were unchanged. They now show that these mice further display abnormal tau phosphorylation and age-dependent accumulation of insoluble tau, memory deficits and neurodegeneration in the hippocampus. The observed behavioral deficits in the AICD-transgenic mice as compared to wild-type mice are small, but the neurodegeneration reaches 50 percent in the CA3 region of the hippocampus; this is substantial compared with other models.
Overall, the sequence of events proposed in the paper (AICD expression, GSK3β activation, tau phosphorylation, memory deficits, neurodegeneration) convincingly explains the phenotype in these mice. A major unresolved issue is the mechanism of GSK3β activation. In addition, it is important to note that the authors are analyzing the combined effects of AICD and FE65 transgene expression, not effects of AICD overexpression alone. It would be interesting to cross the FE65 single-transgenic mice to conventional APP-transgenic mouse models to see whether they replicate findings from the AICD/FE65 mice. One group has performed this cross and observed reduced amyloid pathology in the double-transgenic mice (Santiard-Baron et al., 2005).
Finally, the most important question is whether AICD has any role in AD pathogenesis. There is no clear genetic evidence that would support a role of the AICD domain. One could argue that Down syndrome or APP gene duplications mimic some aspects of the AICD-transgenic mice, but in humans this coincides with increased Aβ levels, which can equally well or better explain the disease. In fact, other genetic evidence might argue against a role of AICD in AD. A number of eFAD presenilin mutations have clearly been shown to reduce AICD (and NICD) formation. This is most likely due to reduced gamma-secretase activity conferred by these mutants. Formal proof that AICD levels are reduced in eFAD patients with presenilin mutations is lacking but it is certainly possible. On the contrary, the new study by Ghosal et al. finds increased AICD levels in brains of Alzheimer patients as compared to controls. However, this data should be interpreted with caution. Normalization to GAPDH or other highly expressed control proteins is not a very accurate way to compare candidate protein levels between brain samples (Aldridge et al., 2008). This may be particularly true for AICD, which is highly unstable and present in very small amounts in brain. In this case, differential post-mortem delays of the brain samples could easily lead to non-homogenous degradation of AICD without much effect on the "control" protein. Hopefully, this intriguing new paper by Ghosal et al. will stimulate rapid follow-up studies to confirm increased AICD levels in AD, and to further define the physiological function of this still mysterious protein domain.
The Debate Is (Still) On: Is AICD-mediated Signaling Relevant to AD?
Since the initial publication from Cao and Sudhof almost nine years ago (Cao and Sudhof, 2001), the AICD field has been filled with hopes, questions and contradictions.
Sanjay Pimplikar’s group provides an interesting follow-up study addressing the role of AICD in vivo. Here, the authors take advantage of their Fe65/AICD C59 transgenic mice that express, under control of the α-CamKII promoter, approximately two- 10-fold more AICD (depending on cellular fractions). They previously showed that by 2 months of age, these mice display increased GSK3β phosphorylation and Kai1 expression. Notably, no changes in GSK3β mRNA and protein levels were observed in these mice, arguing against a direct role of AICD in the gene transcription regulation of GSK3-beta.
Here, the authors propose a model where increased AICD (and Fe65) expression, through sustained GSK3β phosphorylation, causes abnormal tau hyperphosphorylation, aggregation and, ultimately, cell death. The GSK3β-tau connection per se is well documented in the field. Notably, Pimplikar and colleagues show that kainic acid accelerates, whereas lithium rescues, respective neurodegeneration and behavioral deficits in vivo, suggesting an important role for kinases (GSK3β?) in these events. Interestingly, the salient pathological effects on tau seem independent of A-beta levels and plaque formation. Last, the data is suggestive for increased AICD levels in AD patients, providing clinical relevance for the hypothesis that AICD-mediated signaling could contribute significantly to AD pathology.
In my humble opinion, this study brings more questions than answers, particularly with regard to the physiological relevance of these findings. Because of its “binding promiscuity” (Reinhard et al., 2005), it is likely that exogenous AICD modulates a series of signaling cascades. Of course, working with this small and very labile metabolite is technically challenging, and the use of mouse models is an interesting alternative to cell or in-vitro models.
I would like to comment on some issues with regards to mechanistic, physiological, and technical aspects of this work. First, it is very surprising that the authors are able to detect sarcosyl-insoluble (endogenous) tau in wild-type mice (Fig. 1B).
Second, it is curious why the authors compare mutant FeCgamma25 with wildtype BL6 mice in the biochemical studies, and not the Fe.27 mice (that express Fe65 alone). The authors have previously shown that Fe.27 mice have increased “nuclear” AICD (when compared to cytosolic and membrane-bound pools) (Ryan and Pimplikar, 2005). In addition, other reports have implicated Fe65 in APP metabolism and cell death.
Third, the Western blotting experiments in Fig. 7 raise some concerns. Is normalization of AICD levels to GAPDH a good idea? Recent data suggest that AICD levels correlates with APP C-terminal fragments (CTFs) (e.g., see Waldron et al., 2008). Thus, it would have been very informative to quantify in parallel endogenous APP levels (full-length and CTFs) in control and AD patients (Fig 7A). An interesting possibility is that increased AICD correlates with elevated BACE1 levels in a subgroup of sporadic AD patients (Hebert et al., 2008). The very recent study by Uwe Konietzko’s group suggesting that AICD formation is mediated by BACE1 is in line with this hypothesis (Goodger et al., 2009). Of mention, it remains unclear why the authors used a One-way ANOVA statistical test to analyze patient data.
Fourth, an interesting question is whether GSK3β phosphorylation is sustained in older (4-18 months) FeCgamma25 mice. This would strengthen significantly the pathological role for this kinase in this mouse model. In this regard, it would have been informative to validate GSK3β dephosphorylation in the lithium-treated mice.
Overall, this study provides nonetheless strong evidence that prolonged overexpression of Fe65/AICD results in tau pathology and neurodegeneration, at least in the mouse model studied here. Whether cell loss is due to apoptosis requires further investigation. With respect to the patient material, further characterization is needed to assess whether AICD “alone” is affected in AD patients, and to what degree these changes reflect changes in, for instance, Fe65 levels and GSK3β phosphorylation. Of course, independent validation is necessary (it should be noted that, in my hands, I could not detect changes in AICD levels in AD patients when compared to age-matched controls [Hebert SS, unpublished data]). Perhaps these discrepancies are due to technical issues. Unfortunately, the authors did not provide detailed information with regard to tissue homogenization and protein extraction for Western blot (lysis buffer composition, etc.). These questions and others are likely under investigation, as indicated by the number of unpublished “preliminary data” in the manuscript.
In retrospect, it appears to me that the AICD field is missing consistency. In this regard, ectopic expression of AICD alone (C59 or C57) in the mouse or in cells does not correlate with changes in the expression of AICD candidate genes (Hebert et al., 2006; Giliberto et al., 2008; Waldron et al., 2008). It is likely that novel models will be necessary to address, at the individual or global levels, the proposed function(s) of AICD (gene transcription, calcium homeostasis, neurodegeneration, “docking” molecule, catabolite, etc.).