Introduction

What have all these years of research contributed to our understanding of the role of apolipoprotein E (apoE) polymorphisms in Alzheimer's disease (AD)?[1-3]

The risk of developing AD at an earlier age than average is, in fact, correlated with the APOE allele carried by an individual, with the risk associated with APOE4 > E3 > E2. In a recent study, the mean age at AD onset was 68 years in patients with 2 E4 alleles, 75 years with 1 E4 allele, and 84 years in individuals with no E4 alleles. The importance of these associations is underscored by the fact that APOE4 was found linked to about 50% of AD cases analyzed.[1,2]

When the results are analyzed in terms of relative risk, as done by Farrer and colleagues,[4] the APOE4 allele was found associated with an increased risk of AD, whereas the E2 allele appeared to have a protective effect. The relative risk of AD for individuals E4/E4 was in fact 14.9; for E3/E4, 3.2; for E2/E4, 2.6; and for E2/2, .6.

Beyond genetic analysis, more data on the structure/functions of apoE proteins and their mechanisms of action that may explain their association with AD have recently become available, and were discussed at the 9th International Conference on Alzheimer's Disease and Related Disorders by leading investigators. We report here on 2 such presentations that give an idea as to how widely researchers are casting their nets to understand the biological function of apoE and its putative role in the development of minimal cognitive impairment (MCI) and AD. Data are being accumulated and novel hypotheses put forward.

If indeed apoE substantially contributes to MCI and AD, and 1-2 of its isoforms have a protective/therapeutic effect, then new roads for intervention in AD patients may open up in the future.

Structure of ApoE

As illustrated by Dr. Karl Weisgraber,[5] of the University of California, San Francisco, from a structural point of view, E3 is more similar to E4 than E2, as it carries only 1 amino acid substitution: Cys112 in E3 instead of Arg112 in E4. Conversely, E2 has 2 amino acid differences with E4: Cys112 instead of Arg112 and Cys113 instead of Arg113, as in E4 and E3.

In the brain, the apoE proteins have been found associated with the characteristic AD plaques and with the tau protein in neurofibrillary tangles.[6] In vitro, apoE4 can increase phosphorylation of tau. Overexpression of human apoE4 or E3 in transgenic mice led to an increase in plaque formation, with the association of E4 with human amyloid precursor protein (hAPP), and with Abeta peptides in the plaques, a decrease in presynaptic terminals, and an increase in tau phosphorylation and in surrounding gliosis. The effects mediated by E4 followed a dominant pattern. Such morphologic changes were accompanied in the experimental animals by impairments in spatial learning and memory.[7]

Bypassing, however, the apoE amyloid hypothesis, apoE may function, independently of Abeta, in a parallel pathway that, in the presence of increased APP, increased Abeta and changes in tau, contributing over time to cognitive decline and dementia. If this hypothesis proposed by Dr. Weisgraber is correct, then what are the mechanisms involved? How could we move from biomolecular mechanisms to the development of new therapeutic agents?[5]

Characterization of the three-dimensional structure of apoE shows 4 helix bundles in between the amino and carboxy terminal of the molecule. As protein folding and stability are of critical importance in many neurodegenerative diseases, do the amino acid changes Arg/Cys112 and Arg/Cys113 have an effect on apoE function?

Indeed among the 3 allelic forms, apoE4 folds in the least stable; E2 folds in the most stable conformation; and E3 shows an intermediate stability.[6] Consistently, the molten globule conformation linked to greater flexibility is acquired most often by E4 than E3 or E2. In other words, E4 has the highest tendency among these 3 proteins to form molten globules providing a clue as to its activity in vivo. Molten globule conformations may, in fact, lead to[5]:

• Increased degradation;
• Alterations in cell signaling;
• Increased binding to lipids;
• Modifications in protein-protein interactions;
• Increased membrane binding;
• Changes in transport through membranes; and
• Modified interactivity with cellular receptors.

The structural changes occurring in E4 seem to be related to an interaction between Arg112 and Arg61 with Glu255 (salt bridge) that does not occur in E3 owing to the presence of a cysteine residue at position 112.[7]

Hence came the idea to find a small molecule able to bind E4 near the Arg61 residue to inhibit these domain interactions and induce more "spreading" in the molecule, and thus obtain an apoE3-like conformation. More than 200,000 compounds have been screened with the crystal structure of E4 and 115 identified that bound E4, but not E3. After subtraction of close analogs and those mediating charge complementation, 65 compounds presented suitable pharmacologic properties.

Only 5 or 6 compounds at the end of this long search were able to modify the structure of E4, thus providing proof of principle that conformational changes in E4 may affect its biological activity. Wild-type E4, in fact, differently from E3, seems to be associated with higher Abeta production, more extensive disruption of the cytoskeleton, and increased lysosomal cleavage.[5,6]

Activity of ApoE

In vitro studies showed that Neuro 2a cells transfected with apoE4 showed more lysosomal leakage after exposure to Abeta1-42 than did apoE3 transfectants, as well as more pronounced cell death and DNA fragmentation.[8] Thus, the highly acidic environment of lysosomes would favor the molten structure of E4, giving origin to what is now called the "lysosomal leakage" hypothesis of E4-mediated cell damage.

Such differences were, however, also accompanied by a higher sensitivity of apoE4 (vs E3) to proteolytic cleavage by a chymotrypsin-like serine protease. Transgenic mice, generated to express the carboxy-terminal fragment of apoE4 (delta 272-299), died at 2-4 months of age, with neurodegeneration, phosphorylated tau, and preneurofibrillary tangles scattered throughout their brains. A decrease in the amount of truncated apoE4 expressed led to a better survival (up to 6-7 months), but these apoE4 transgenic mice showed significant learning and memory deficits.[9]

Hence came the "proteolytic processing" hypothesis in relation to the flexible structure of E4. As the apoE4 fragmentation occurs in humans and in NSE-apoE4 transgenic mice, but not in green fluorescent protein-apoE mice, the cleavage seems to be neuron-specific. If differential proteolytic processing underlies the pro-AD activity of E4, then targeting those proteolytic enzymes that "activate" E4 could suppress its undesired functions and offer, in principle, therapeutic benefit.[5]

A third hypothesis is being evaluated in relation to the possible mechanisms of action of E4 vs E3 and E2, which takes into account the "increased Abeta production" effect seen after E4 overexpression. In vitro transfection of E4 vs E3 led to larger deposits of Abeta, with accumulation of up to 620-800 pg of amyloid/mg of cellular proteins.[5]

Thus, different targets might be evaluated to inhibit E4: modification of protein stability and proteolytic processing, or domain-domain interactions within E4 to prevent the molten globule structure and the undesired effector functions of "molten" E4.

Murine E4 does not share these structural properties with human E4. It has a human E3-like behavior (less molten globule structures) with a sequence encompassing Thr61, Arg112, and Glu255 vs Arg61, Arg112, and Glu255 in human E4. The salt bridge between Thr61 and Glu255 would thus be missing in murine E4 as in the human E3 molecule. Human E4-like molecules can be introduced in mice, through targeted mutations, to generate "designer" transgenic mice that can be used to screen and test for potentially therapeutic compounds targeted at apoE4.[5]

ApoE and Clearance of Amyloid Beta

Specific APOE genotypes have thus been associated with an increased risk of AD, decreased age at onset, and an increased risk of conversion from MCI to AD. In particular, elderly E4 carriers have a higher amyloid burden in their brains, even before the appearance of dementia.

The finding that expression of the APOE3 allele in a transgenic model decreased the amyloid beta load in a dose-dependent form in PDAPP mice at 12-15 months of age raised the question of whether apoE molecules are involved in the clearance of brain amyloid by astrocytes. And indeed experimental evidence has suggested that there is an apoE-dependent degradation of Abeta with brain slice preparations in vitro.[10]

As illustrated by Dr. Steven Paul,[11] of Eli Lilly and Co., Indianapolis, Indiana, astrocytes migrated and aggregated in response to deposition of Abeta in apoE-transgenic mice. Such a process was apoE-dependent, as it did not occur with apoE-negative astrocytes. Of note, astrocyte aggregation was associated with in situ degradation of Abeta. Astrocytes underwent characteristic morphologic changes in response to Abeta with flattening and extension of cellular processes to engulf and internalize Abeta.[12]

It is not yet known whether the different isoforms of apoE have different clearance activities, and, of course, one of the expectations is that apoE4 may have a significantly reduced clearance activity when compared with the other apoE isoforms.[10] On the other hand, apoE3 does mediate Abeta clearance in vivo, as it could restore Abeta degradation in apoE-knockout mice with extensive Abeta deposits.[12]

In vitro assays further showed that Abeta induced, in a dose-dependent way, astrocyte chemotaxis and aggregation in transwell assays. A similar activity was exerted on astrocytes by Abeta peptides, with Abeta1-42 displaying a substantially higher chemotactic/aggregatory potential than Abeta42-1. Both chemotaxis and aggregation were blocked by an excess of exogenous Abeta and Abeta peptides, and they were absent in astrocytes obtained from apoE-knockout mice.[13]

As antibodies are being generated to target Abeta, it is not clear whether their interaction with Abeta would mask the sites of interaction with apoE-expressing astrocytes and thus prevent clearance from the brain.[11] Different antibodies binding to different determinants of Abeta may, however, have, in principle, different interference activities.

In conclusion, astrocytes appear to play a critical role in the clearance of Abeta in the brain following migration to areas of the brain rich in neurotoxic deposits. A receptor-specific uptake seems to mediate internalization and degradation. Defects in any of these steps (eg, apoE) may impair clearance, thus favoring further accumulation of Abeta and the appearance of neuropathologic damage and AD.[11]

Enhanced Expression of ApoE and Abeta Modulation

Further experiments were conducted by expressing apoE with lentiviral vectors in PDAPP/ApoE-knockout mice that develop Abeta deposits with aging. Of note, expression of apoE4 led to increased deposition of amyloid in these animals. ApoE2, on the other hand, induced a marked decrease in Abeta accumulation, as shown by immunohistochemistry and enzyme-linked immunosorbent assay (ELISA).[11]

ApoE2 thus appeared to have both a protective and therapeutic effect in this animal model. Dr. Paul was, however, quick in adding that he was not advocating for a gene therapy approach of this kind in AD. The lentiviral vectors used, in fact, were toxic for a specific set of neurons in the hippocampus of treated mice.[11]

If not through gene delivery, an alternative way to enhance the expression of apoE would be to use compounds that can induce the upregulation of expression in vivo. As recently shown by Liang and colleagues,[14] LXR/RXR ligands induced the expression of apoE heterodimers. An approximately 2-3-fold increase was achieved in primary mouse astrocytes. Although promising, these results are quite preliminary.

As summarized by Dr. Paul, now we know that:

• Astrocytes are important for the degradation of Abeta;
• Aging, specific apoE isoforms, and deficits in Abeta clearance may lead to AD; and
• Upregulation of specific apoE alleles (eg, E3) led to clearance and prevention of Abeta deposition in vitro and in experimental animal models.

More in-depth studies will clarify whether enhancers of expression of specific apoE isoforms may indeed represent a new preventive/therapeutic approach for diseases associated with toxic accumulation of amyloid and Abeta peptides in humans.

References