Please see the following BMJ letter commenting on this and other related article: Alzheimer’s anti-amyloid vaccination and statins: two approaches, one dogma. The time for change. Alexei R. Koudinov, Natalia V. Koudinova BMJ 20 March 2002 [ Full Text ]

View all comments by Alexei R. Koudinov

Saido's team (Iwata et al.) infused labeled Aβ1-42 into rat hippocampus and found it was metabolized with a half-life of 39 minutes. They tried many classes of peptide inhibitors and found that neprilysin type were the only ones that prevented the breakdown. They then infused one inhibitor (thiorphan) into normal rat brain and produced rat amyloid deposits in 30 days. (That beats transgenic models.) We followed this up in our Neuroscience Letters paper (12 Jan 2001;297:97-100). We found low levels of neprilysin in plaque prone areas such as the hippocampus and temporal cortex, and high neprilysin levels in areas such as the striatum, cerebellum and peripheral organs, which never or rarely ever develop plaques. Moreover, we found Alzheimer patients had significantly lower neprilysin levels than normals in vulnerable areas.

Neprilysin is a membrane-bound enzyme highly expressed on axons and nerve endings of peptide-expressing neurons. Our data help explain...  Read more

Saido's team (Iwata et al.) infused labeled Aβ1-42 into rat hippocampus and found it was metabolized with a half-life of 39 minutes. They tried many classes of peptide inhibitors and found that neprilysin type were the only ones that prevented the breakdown. They then infused one inhibitor (thiorphan) into normal rat brain and produced rat amyloid deposits in 30 days. (That beats transgenic models.) We followed this up in our Neuroscience Letters paper (12 Jan 2001;297:97-100). We found low levels of neprilysin in plaque prone areas such as the hippocampus and temporal cortex, and high neprilysin levels in areas such as the striatum, cerebellum and peripheral organs, which never or rarely ever develop plaques. Moreover, we found Alzheimer patients had significantly lower neprilysin levels than normals in vulnerable areas.

Neprilysin is a membrane-bound enzyme highly expressed on axons and nerve endings of peptide-expressing neurons. Our data help explain why plaques only appear in certain brain areas, and probably why they only begin to appear in older people when transmitter-associated enzymes begin to wane. They suggest that there should be a major focus on the breakdown of A-beta since it should be metabolized within a few minutes of formation. The data also suggest that there may be a locus on chromosome 3 at the neprilysin site where polymorphisms might affect the expression and thus vulnerability to AD. They also suggest that insulin-degrading enzyme must play only a minor role since it does not compensate for thiorphan blockade of neprilysin in rats and does not show the same correlations with plaquedevelopment in humans.

View all comments by P.L. McGeer

I don't think this paper definitively rules IDE out as a LOAD gene. It simply says the authors could not find evidence for it in a case-control study of about 150 subjects with LOAD. It is not uncommon for different genetic methods to not confirm each other in the search for genes in complex, clinically and genetically heterogeneous disorders. So I would not say the paper negates the earlier work from Bertram et al., 2000.

View all comments by Dennis Selkoe

Authors' reply:
In our study of IDE, we detected five SNPs in IDE occurring at >5 percent frequency, which we tested for association in a minimum sample of over 200 AD cases and 200 controls. Where there was a suggestion of a significant result for two of these SNPs, we tested a further 90-100 cases and100 controls. Our main rationale for testing IDE was that it maps to a linkage region for late-onset AD on chromosome 10. Our hypothesis was that if IDE explained the linkage, we would expect to find a risk allele, or alleles, conferring a relative risk of >3, i.e., a gene with an effect size comparable to that of ApoE.

Not only did we fail to find any highly probable functional SNPs, but also, examination of our odds ratios (ORs) reveals that for all the SNPs we did detect, substantially smaller ORs can be excluded with 95 percent confidence from contributing even a fairly small genetic effect [IDE1 (203 cases/247 controls): OR 0.97 (95 percent CI 0.64-1.4); IDE2 (220 cases/245 controls): OR 0.94 (95 percent CI 0.69-1.3); IDE3 (224 cases/247 controls): OR 1.05 (95...  Read more

Authors' reply:
In our study of IDE, we detected five SNPs in IDE occurring at >5 percent frequency, which we tested for association in a minimum sample of over 200 AD cases and 200 controls. Where there was a suggestion of a significant result for two of these SNPs, we tested a further 90-100 cases and100 controls. Our main rationale for testing IDE was that it maps to a linkage region for late-onset AD on chromosome 10. Our hypothesis was that if IDE explained the linkage, we would expect to find a risk allele, or alleles, conferring a relative risk of >3, i.e., a gene with an effect size comparable to that of ApoE.

Not only did we fail to find any highly probable functional SNPs, but also, examination of our odds ratios (ORs) reveals that for all the SNPs we did detect, substantially smaller ORs can be excluded with 95 percent confidence from contributing even a fairly small genetic effect [IDE1 (203 cases/247 controls): OR 0.97 (95 percent CI 0.64-1.4); IDE2 (220 cases/245 controls): OR 0.94 (95 percent CI 0.69-1.3); IDE3 (224 cases/247 controls): OR 1.05 (95 percent CI 0.96-1.6); IDE4 (222 cases/248 controls): OR 0.97 (95 percent CI 0.970.73- 1.3); IDE5 (210cases/230 controls): OR 0.96 (95 percent CI 0.64-1.4)].

We therefore conclude that it is extremely unlikely that variation within IDE is responsible for the linkage findings. We stated this conclusion in our paper "…our results indicate that this activity (of IDE) is unlikely to be important in influencing susceptibility to late-onset AD. As the approach taken has an 80 percent power to detect alleles of >3 percent frequency, we can conclude that coding sequence variants in IDE cannot be a significant risk factor for LOAD. Rare coding polymorphisms would not account for LOAD in more than a small proportion of the population and could not therefore account for our linkage data on chromosome 10…. ".

Further, recognizing the strength of the functional data, we concluded that whatever these data say about IDE being involved in amyloid degradation, our findings suggest that genetic variation in the IDE gene does not greatly influence susceptibility to LOAD. "It remains possible that one or more of the SNPs we identified itself alters the expression of IDE or is linked to other variants that do, but that this has no influence on LOAD." This situation is directly analagous to that of BACE, where the functional activity of the b-secretase product against AβPP is clear, yet no genetic association has been established despite extensive analysis.

We agree with Dr. Selkoe in so far as no study in science can ever exclude the possibility of a hypothesis, and that small genetic effects remain possible. However, in this case, our data overwhelmingly fail to reject the null hypothesis. The evidence published in support of IDE being a susceptibility gene for late-onset AD, which is based upon a microsatellite and was not confirmed in our dataset using the same marker, is comparatively weak. The balance of evidence is strongly in favor of the belief that IDE is not a susceptibility gene for AD.

View all comments by Lesley Jones

The newly published paper by Miller et al., well summarized by Tom Fagan above, both confirms and complements our recent study, in which we also utilized IDE gene-disrupted mice to demonstrate that IDE regulates the levels of cerebral Aβ and the AβPP intracellular domain (AICD, CTF in vivo: see Farris et al., 2003). A major strength of this latest work is that by breeding heterozygous IDE knockout mice (IDE+/-) to each other, these authors generated single litters containing all three possible IDE genotypes (IDE+/+, IDE+/-, and IDE-/-). The advantages of this breeding strategy are that it enables one to look for a "gene-dosage" effect on the phenotype of interest while providing littermates of different genotypes for comparison. The disadvantage of this method is that relatively few mice of a given genotype can be generated within one litter. Although the numbers of age-matched animals of a given phenotype in this study were relatively small (n = 2-4), the brain Aβ ELISA data were "tight" enough...  Read more

The newly published paper by Miller et al., well summarized by Tom Fagan above, both confirms and complements our recent study, in which we also utilized IDE gene-disrupted mice to demonstrate that IDE regulates the levels of cerebral Aβ and the AβPP intracellular domain (AICD, CTF in vivo: see Farris et al., 2003). A major strength of this latest work is that by breeding heterozygous IDE knockout mice (IDE+/-) to each other, these authors generated single litters containing all three possible IDE genotypes (IDE+/+, IDE+/-, and IDE-/-). The advantages of this breeding strategy are that it enables one to look for a "gene-dosage" effect on the phenotype of interest while providing littermates of different genotypes for comparison. The disadvantage of this method is that relatively few mice of a given genotype can be generated within one litter. Although the numbers of age-matched animals of a given phenotype in this study were relatively small (n = 2-4), the brain Aβ ELISA data were "tight" enough that statistically significant differences were found.

First, the authors did a thorough characterization of the three genotypes (standard and real-time RT-PCR, immunoblotting and endorphin degradation assays to measure IDE activity on liver and brain) to demonstrate that the IDE+/- mice do, in fact, have an amount of IDE transcript, protein, and activity that is roughly half that of the IDE+/+ animals. They then convincingly demonstrate a gene-dose response of cerebral Aβ40 and Aβ42. Compared to wild-type mice, those with one copy of the IDE gene had, by my calculations, ~35 percent increase in Aβ40 and ~20 percent increase in Aβ42, while those without any IDE genes had ~55 percent increase in Aβ40 and ~35 percent increase in Aβ42. In addition to strengthening the hypothesis that IDE is responsible for the elevated cerebral Aβ, this gene-dose analysis allows the authors to conclude that IDE activity, rather than substrate availability, is the rate-limiting step in the degradation of Aβ by IDE, and that there is not significant compensatory upregulation of IDE expression by the normal allele when the other is dysfunctional (or, as in this model, absent).

Of note, although it is inferred that the elevated cerebral Aβ in the IDE+/- and IDE-/- mice is secondary to a deficiency of IDE-mediated Aβ degradation, it was not demonstrated directly in this study that these mice actually have an Aβ degrading deficit. One could hypothesize that the elevated AICD levels in these mice, demonstrated in both studies, could lead to upregulation of AβPP levels. If this increased AβPP was then cleaved by BACE and the γ-secretase complex instead of accumulating in the membrane, Aβ levels could increase without a measurable increase in AβPP—all via a mechanism independent of IDE’s effect on Aβ degradation. To exclude this possibility, we demonstrated a major (>50 percent) Aβ degrading deficit in brain membrane and soluble fractions and intact primary neurons from IDE-/- compared to IDE+/+ mice. We confirmed that the levels of soluble AβPP (AβPPs), released by BACE or γ-secretase cleavage, were indistinguishable between the IDE-/- and IDE+/+ animals, arguing against an increase of Aβ production in the IDE-/- as an explanation for the observed increased levels of cerebral Aβ in these two studies. Additionally, we found that the levels of AβPP and its other proteolytic derivatives (C99, C89, C83), presenilin NTF and CTF, BACE, ADAM 10 and neprilysin were all identical by immunoblotting in the IDE-/- and IDE+/+ mice.

It is encouraging that the results of the Miller et al. and Farris et al. studies are in quite good agreement regarding IDE’s role in regulating cerebral Aβ and AICD levels in vivo. (It probably should be noted that the cerebral Aβ was extracted by the same DEA protocol and measured using the same ELISA system.) One discrepancy between the papers should be mentioned. A major point of the Farris et al. paper was that selective deletion of one gene, IDE, was not only able to elevate Aβ levels in the brain, but also to recapitulate some of the phenotypic hallmarks of type 2 diabetes, namely hyperinsulinemia and glucose intolerance. We found this particularly exciting in light of the epidemiological evidence suggesting that subjects with type 2 diabetes have approximately a doubling in their risk of Alzheimer’s disease (independent of vascular risk factors), and the emerging genetic evidence for both an Alzheimer’s and type 2 diabetes gene in the IDE region of chromosome 10q (please see paper for more detailed discussion and references). The Miller et al. paper only mentions IDE’s role in insulin/glucose metabolism in one sentence in the Materials and Methods, where they state that they found no difference in blood glucose and insulin levels in fed IDE-/- and IDE+/+ mice, and then suggest that insulin catabolism and activity are not perturbed by the lack of IDE. We demonstrated marked deficiency in insulin degradation by liver fractions, and a significant 2.8-fold elevation in fasting insulin in IDE-/- mice compared to wild-types. The number and age of mice examined and the method of glucose and insulin quantification by Miller et al. is not mentioned, so it is difficult to directly compare experiments, but there may be less of a difference between fed insulin levels in the IDE-/- and IDE+/+ mice, as measured by Miller et al., and fasting insulin levels, as measured by us. We are currently measuring insulin levels in fed animals to see if this may be the case. We found no difference in fasting blood glucose levels between IDE-/- and IDE+/+ mice, but when they were challenged with a glucose load (glucose tolerance test), the IDE-/- animals revealed a significantly delayed drop in blood glucose at 60, 90, and 120 minutes compared to the IDE+/+ mice, showing that the IDE-/- mice are glucose intolerant. Thus, we believe that IDE does have a role in insulin catabolism and glucose tolerance in vivo.

Taken together, two separate groups have now shown that IDE regulates cerebral levels of Aβ in vivo, making pharmacologic manipulation of protease, either directly or via blocking a natural IDE inhibitor, a potential therapeutic target. The new findings that an abnormally low expression of IDE from one allele is not sufficiently compensated for by the normal allele, and that this partial hypofunction of IDE is sufficient to elevate Aβ levels in the brain, are important. We have recently reported that diabetic GK rats, with their two naturally occurring missense mutations in IDE, show ~30 percent deficit in Aβ degradation in brain fractions and primary neurons, and have markedly elevated levels of endogenously produced Aβ in their neuronal conditioned media (Farris et al., SFN Meeting Abstract, 2002, submitted). Together with the emerging genetic evidence, these findings suggest that IDE mutations could underlie some forms of AD.

View all comments by Wesley Farris

Re: Causes of premature death in APP transgenic mice…and how to alleviate them.

The premature death of APP Tg mice is a practical problem—as anyone knows who has talked to biotech or pharma companies about licensing a strain of mice that lives “unpredictably.” But even more than that, it is a major scientific problem that was puzzling us in the mid 90’s, soon after entering this field.

Contrary to finding out what the underlying reasons were, the “executive summary” of the outcome reads rather simply: Premature death of APP Tg mice is caused by excitotoxicity, as shown by massive neuronal death in the hippocampus. I refer the interested readers to four of our publications (spear-headed by D. Moechars) listed below, directly or indirectly addressing the problem.

To a large extent, premature death was caused by environmental “stress” defined in its broadest sense, i.e., occasional or persistent infections, high background noise levels, poor training of animal caretakers and researchers entering the rooms, some room cohabitants (other strains of Tg mice). ...  Read more

Re: Causes of premature death in APP transgenic mice…and how to alleviate them.

The premature death of APP Tg mice is a practical problem—as anyone knows who has talked to biotech or pharma companies about licensing a strain of mice that lives “unpredictably.” But even more than that, it is a major scientific problem that was puzzling us in the mid 90’s, soon after entering this field.

Contrary to finding out what the underlying reasons were, the “executive summary” of the outcome reads rather simply: Premature death of APP Tg mice is caused by excitotoxicity, as shown by massive neuronal death in the hippocampus. I refer the interested readers to four of our publications (spear-headed by D. Moechars) listed below, directly or indirectly addressing the problem.

To a large extent, premature death was caused by environmental “stress” defined in its broadest sense, i.e., occasional or persistent infections, high background noise levels, poor training of animal caretakers and researchers entering the rooms, some room cohabitants (other strains of Tg mice). Premature death was and is “normalized” to that of other Tg (non-APP) strains and (nearly) to that of wild-type non-Tg strains by improving these conditions in our animal house, i.e., SPF or IVC cages, silence or soft background music, no “dilettante” but well-instructed animal caretakers and researchers handling the mice, strict application of sanitary rules, inverted day-night cycle, etc. This set of measures (not tested individually) appears to alleviate the environmental stress and thereby prevent premature death, but we occasionally see it reappear in the conventional animal rooms when the above conditions are not applied rigorously.

The data support a direct relation of APP-metabolites Aβ and β-CTF to augmented excitability of APP Tg mice and to excitotoxicity as the underlying cause for premature death. We have observed premature death only in APP Tg mice and not in the 50+ other Tg strains expressing different transgenes that we have generated and studied over the last 15 years. It is perfectly in line with the experimental demonstration that mutant presenilins increase Aβ levels and effectively decrease the threshold for excitotoxicity (Schneider et al., 2001).

That brings us full circle, since a decrease in Aβ-levels then must decrease premature death, as demonstrated by Leissring and coworkers. It is safe to predict that their double APP (x IDE or NEP) Tg mice also will have decreased excitotoxicity relative to the parent APP single Tg mice.

Intriguingly, neuronal deficiency of PS1 effectively decreases Aβ levels, but contributes negatively to neuronal excitability by increasing intracellular calcium store. This also implicates β-CTF, which are increased in PS1(n-/-) mice (Dewachter et al., 2002). One wonders what happens to the β-CTF in the double-Tg mice: Are they also chewed up by the overexpressed proteinases or not?

And I do want to quote from an old comment by Chris Exley (Submitted 16 November 2001) referred to in and below this story: “It would seem that the 'fog' has already begun to clear. It just went, apparently, unnoticed.” This notion pertains to many discoveries in the AD field these days!

References:
Moechars D, Lorent K, Dewachter I, Baekelandt V, De Strooper B, Van Leuven F. Transgenic mice expressing an alpha-secretion mutant of the amyloid precursor protein in the brain develop a progressive CNS disorder. Behav Brain Res 1998;95:55-64. Abstract

Moechars D, Dewachter I, Lorent K, Reverse D, Baekelandt V, Naidu A, Tesseur I, Spittaels K, Haute CV, Checler F, Godaux E, Cordell B, Van Leuven F. Early phenotypic changes in transgenic mice that overexpress different mutants of amyloid precursor protein in brain. J Biol Chem 1999;274:6488-6492. Abstract

Moechars D, Lorent K, De Strooper B, Dewachter I, Van Leuven F. Expression in brain of amyloid precursor protein mutated in the alpha-secretase site causes disturbed behavior, neuronal degeneration and premature death in transgenic mice. EMBO J 1996;15:1265-1274. Abstract

Moechars D, Lorent K, Van Leuven F. Premature death in transgenic mice that overexpress a mutant amyloid precursor protein is preceded by severe neurodegeneration and apoptosis. Neuroscience 1999;91:819-830. Abstract

Schneider I, Reverse D, Dewachter I, Ris L, Caluwaets N, Kuiperi C, Gilis M, Geerts H, Kretzschmar H, Godaux E, Moechars D, Van Leuven F, Herms J. Mutant presenilins disturb neuronal calcium homeostasis in the brain of transgenic mice, decreasing the threshold for excitotoxicity and facilitating long-term potentiation. J Biol Chem 2001;276:11539-11544. Abstract

Dewachter I, Reverse D, Caluwaerts N, Ris L, Kuiperi C, Van den Haute C, Spittaels K, Umans L, Serneels L, Thiry E, Moechars D, Mercken M, Godaux E, Van Leuven F. Neuronal deficiency of presenilin 1 inhibits amyloid plaque formation and corrects hippocampal long-term potentiation but not a cognitive defect of amyloid precursor protein [V717I] transgenic mice. J Neurosci 2002;22:3445-3453. Abstract

View all comments by Fred Van Leuven