On the Pathogenesis of Alzheimer's Disease: The MAM Hypothesis

Abstract

The pathogenesis of Alzheimer's disease (AD) is currently unclear and is the subject of much debate. The most widely accepted hypothesis designed to explain AD pathogenesis is the amyloid cascade, which invokes the accumulation of extracellular plaques and intracellular tangles as playing a fundamental role in the course and progression of the disease. However, besides plaques and tangles, other biochemical and morphological features are also present in AD, often manifesting early in the course of the disease before the accumulation of plaques and tangles. These include altered calcium, cholesterol, and phospholipid metabolism; altered mitochondrial dynamics; and reduced bioenergetic function. Notably, these other features of AD are associated with functions localized to a subdomain of the endoplasmic reticulum (ER), known as mitochondria-associated ER membranes (MAMs). The MAM region of the ER is a lipid raft-like domain closely apposed to mitochondria in such a way that the 2 organelles are able to communicate with each other, both physically and biochemically, thereby facilitating the functions of this region. We have found that MAM-localized functions are increased significantly in cellular and animal models of AD and in cells from patients with AD in a manner consistent with the biochemical findings noted above. Based on these and other observations, we propose that increased ER-mitochondrial apposition and perturbed MAM function lie at the heart of AD pathogenesis.—Area-Gomez, E., Schon, E. A. On the pathogenesis of Alzheimer's disease: the MAM hypothesis.

The main histopathological hallmarks of Alzheimer's disease (AD), a neurodegenerative disorder characterized by progressive neuronal loss in the cortex and hippocampus, are the accumulation of extracellular neuritic plaques and intracellular neurofibrillary tangles. The plaques are composed of numerous proteins, most prominent among them β-amyloid (Aβ). The chain of events leading to the formation of Aβ has been called the “amyloid cascade” (1) and is the most widely accepted hypothesis to explain the pathogenesis of AD. Whereas the amyloid cascade hypothesis explains the development of neuritic plaques, it does not help to explain a number of other features of the disease, some of which arise before the appearance of plaques and/or tangles (2, 3).

Among these other symptoms and features are the altered metabolism of fatty acids and phospholipids (4, 5), elevated levels of circulating cholesterol (6), accumulation of lipid droplets in cells (7–9), aberrant calcium regulation (10), reduced levels of brain glucose (11), and mitochondrial dysfunction (12, 13). We believe that these features of AD have received less attention than the plaques and tangles for 2 main reasons. First, plaques and tangles are physical entities that can observed pathologically and histopathologically—Alois Alzheimer himself apparently described them in his original report published >1 century ago (14, 15)—whereas biochemical and metabolic changes are more subtle and more difficult to detect and characterize. Second, as the accumulation of plaques and tangles was the initial stimulus for proposing the amyloid cascade hypothesis in the first place, a fortiori, there is no unifying conceptual framework within the amyloid cascade hypothesis that would explain the occurrence of other features that are ostensibly unrelated to plaques and tangles (16). Thus, these “other” features have been either ignored entirely or have been relegated to the position of being downstream events, secondary to plaque and tangle development. Of course, this latter view must deal with the fact, noted above, that some of these features are evident years before plaques and tangles become apparent.

When viewed globally, the data appear to indicate that mutations in presenilins (PSs) and amyloid precursor proteins (APPs) cause dysfunction in a process that lies upstream of all of the features described in AD—not just the plaques and tangles. From this point of view, all of these features—plaques, tangles, lipid dyshomeostasis, calcium dyshomeostasis, and mitochondrial dysfunction—would be considered to be important and critical symptoms of AD but not the primary cause of disease pathogenesis itself. What could that primary insult be?

As a laboratory focused on mitochondrial biology and disease, we were intrigued by the potential convergence of these apparently unrelated symptoms to a region of the cell that results from the interplay between mitochondria and the endoplasmic reticulum (ER), called the mitochondria-associated ER membrane (MAM). MAM is a specialized subdomain of the ER that connects mitochondria and the ER, both physically and biochemically (17). MAM, which has the characteristics of a lipid raft, is the regulatory locus within the cell for phospholipid, cholesterol ester, and fatty acid metabolism; for lipid droplet formation; for calcium homeostasis; for mitochondrial dynamics (17), and intriguingly, for Aβ production (18, 19). In agreement with our supposition that MAM is relevant to AD pathogenesis, the cleaved, active forms of the PSs and γ-secretase activity itself were found to be present predominantly in the MAM (20, 21). The localization of PSs to MAM could help explain the reported effects of PS mutations on mitochondrial function and the puzzling observation that PSs [which are localized in the ER (22)], γ-secretase activity, and Aβ have all been found at or near mitochondria (23–26). A localization of PSs and γ-secretase activity to MAM would also be in accordance with the finding that APP processing occurs in lipid rafts (27).

The potential connection between the symptoms of AD and MAM function prompted us to measure various aspects of MAM function in PS-mutant cells and in cells from patients with AD (28). Overall, we found a significant up-regulation of MAM behavior in these cells vs. controls, which correlated with a significantly higher degree of apposition between ER and mitochondria. For example, it has long been known that acyl CoA:cholesterol acyltransferase 1 (ACAT1; gene sterol O-acyltransferase 1), which converts free cholesterol to cholesteryl esters that are deposited in lipid droplets, is enriched in MAM and can be used as a measure of MAM activity (29). We found significantly higher ACAT1 activity in these cells compared with controls (28), and they contained significantly more lipid droplets, as well (28). Not only do these observations help explain the elevated, circulating cholesterol and lipid droplets seen in tissues from patients with AD (8), but they are also consistent with the remarkable observation that ACAT1 activity is necessary for the production of Aβ (18, 19).

Besides cholesteryl ester synthesis, another key function of MAM is the synthesis of phospholipids. In broad view, phospholipids are synthesized in 2 ways: de novo, in the Kennedy pathway; and salvaged and recycled, in a pathway requiring MAM. In the latter pathway, phosphatidylserine (PtdSer) is synthesized in the MAM and then is translocated to mitochondria, where it is converted to phosphatidylethanolamine (PtdEtn) by the mitochondrially localized enzyme, PtdSer decarboxylase (30); this reaction is a well-recognized measure of ER-mitochondrial communication (30). Accordingly, we measured the synthesis of these phospholipids in our AD cell models and found that the levels of both PtdSer and PtdEtn were increased significantly in both the PS-mutant cells and in AD fibroblasts (28), indicating increased ER-mitochondrial cross-talk in these cells. These results could help explain the altered phospholipid profiles seen in AD (5).

Whether the increased cholesterol and phospholipid metabolism seen in AD is a consequence of increased apposition between ER and mitochondria (28), then relaxing the communication between the 2 organelles by decreasing apposition should reverse these phenotypes. Accordingly, we took advantage of previous reports showing that mitofusin-2 (MFN2), a protein involved in mitochondrial fusion, localizes to MAM and that its ablation decreases ER-mitochondrial connectivity (31). Therefore, we repeated our assays in Mfn2-knockout (KO) cells and confirmed that loss of Mfn2 results in reduced apposition between ER and mitochondria and that this decreased connectivity correlated with the down-regulation of MAM activity, as measured by reduced phospholipid synthesis, reduced cholesterol esterification, and reduced cellular lipid droplet content (28). Moreover, we demonstrated that PSs and MFN2 have opposite effects on MAM regulation and ER-mitochondrial apposition, as we were able to mitigate the up-regulation in MAM activity in PS-KO cells by knocking down MFN2 in those cells and vice versa (28).

Based on these and other data, we have proposed that altered MAM function and increased connections between ER and mitochondria are causative and early events in the pathogenesis of AD (16, 32, 33). Whereas the plaques and tangles found in brains from patients certainly exacerbate the course and progression of AD, we believe that they are later events and are not fundamental causes of the disease. Instead, we propose that it is the increased communication between ER and mitochondria and the concomitant increase in MAM functionality that lie at the heart of AD pathogenesis (32, 33).

What is the underlying cause of this increased MAM functionality? In the case of familial AD (FAD), it is a result, either directly or indirectly, of the mutations in PSs and in APP, but that begs the question of how altered processing of APP by γ-secretase can cause increased ER-mitochondrial communication. The question becomes even more acute when considering the fact that the alterations in MAM function and topology that we observed in cells from patients with FAD were also found in those from patients with sporadic AD (SAD), in which the PS and APP genes are normal. Thus, if SAD and FAD are essentially the same disorder, clinically, biochemically, and morphologically (28), then it would be difficult to propose that aberrant γ-secretase processing per se causes the disease, as that activity is presumably normal in SAD, at least qualitatively (the quantitative aspect of γ-secretase processing in SAD, on the other hand, is a matter of some debate, especially given that patients with Down syndrome are prone to developing AD, presumably because they have 3 copies of the APP gene rather than 2).

One clue that can help us think about this issue comes from genetics. It has long been known that the single greatest genetic risk factor for developing SAD is the ε4 allele of apolipoprotein E (ApoE4). ApoE is a component of lipoproteins that ferry lipids and in particular, cholesterol and cholesteryl esters, throughout the circulation (in the brain, astrocytes produce ApoE, but neurons normally do not). For unknown reasons, an amino acid variant at 1 position in the ApoE4 protein (i.e., Arg-112) confers this increased risk compared with that conferred by ApoE3 (i.e., Cys-112), the most common allele in the population. As we considered altered MAM function to be a key element in AD pathogenesis, we hypothesized that ApoE4 might affect MAM function differentially compared with ApoE3, and in fact, that is exactly what we found: both MAM-mediated phospholipid transport and ACAT1-mediated cholesteryl ester synthesis and lipid droplet formation were significantly increased in fibroblasts and in explanted neurons that were treated with astrocyte-conditioned medium containing ApoE4 vs. ApoE3 lipoproteins (34). Notably, these results were obtained only when ApoE4 was a component of lipoproteins; there was no effect when ApoE was added as the free, unlipidated protein, implying that the role of ApoE4 as a risk factor in AD is likely related to its lipid-transport function as a component of lipoproteins.

The ApoE result has a number of implications. First, it is consistent with the idea that MAM plays an underlying role in both FAD and SAD. Second, it supports the view that FAD and SAD are, at bottom, the same disorder, differing only in age of onset. Third, and perhaps most important, it implies that in some currently unclear way, ApoE function (via its role as a component of lipoproteins) and APP processing are related events that drive AD pathogenesis. As lipoproteins are designed to transport cholesterol (and cholesteryl esters that are converted into cholesterol following lipoprotein import) and as intracellular cholesterol is poorly recycled in ApoE4-containing cells (35), one possible connection between ApoE and APP processing is the regulation of cholesterol homeostasis (36). This supposition is also consistent with the fact that γ-secretase resides in MAM (20), a sphingomyelin- and cholesterol-rich lipid raft, and that APP contains a cholesterol-binding domain in its C terminus (37) that may act as a cholesterol sensor (38). Thus, it is possible that altered cholesterol levels—either via aberrant ApoE4-mediated cholesterol trafficking in the case of SAD or via aberrant cholesterol sensing/homeostasis in the case of FAD—somehow provoke an increase in ER-mitochondrial communication that gives rise to the phenotypes seen in AD. This possibility is currently under investigation.

In summary, we believe that the MAM hypothesis offers a unified explanation for the pathogenesis of both SAD and FAD and could potentially take AD research in a new and fruitful direction. We emphasize that we do not deny the role of plaques and tangles in the pathogenesis of AD; they are clearly detrimental and accelerate the course of the disease. However, we feel that they are consequences, not causes, as the amyloid cascade hypothesis does not explain many aspects of the disease that are seemingly unrelated to plaques and tangles.

The MAM hypothesis also opens the door to new ways of thinking about diagnosis and treatment. For example, given that up-regulated MAM causes increased phospholipid and cholesteryl ester synthesis, it should be possible to use these alterations as markers to diagnose AD in relatively accessible cells. Moreover, it might be more useful to devise a treatment aimed at restoring normal MAM function, rather than dealing with the accumulation of plaques and tangles, which we consider to be downstream consequences of increased MAM function and increased ER-mitochondrial connectivity.

Glossary

Aβ

β-amyloid

ACAT1

acyl CoA:cholesterol acyltransferase 1

AD

Alzheimer's disease

ApoE3/4

ε3/4 allele of apolipoprotein E

APP

amyloid precursor protein

ER

endoplasmic reticulum

FAD

familial Alzheimer's disease

KO

knockout

MAM

mitochondria-associated endoplasmic reticulum membrane

MFN2

mitofusin-2

PS

presenilin

PtdEtn

phosphatidylethanolamine

PtdSer

phosphatidylserine

SAD

sporadic Alzheimer's disease

AUTHOR CONTRIBUTIONS

E. Area-Gomez and E. A. Schon wrote the paper.