17 January 2003. While the hunt for new Alzheimer's disease genes continues apace, researchers are still struggling to understand the exact modus operandi of the one prize they have had in hand for years: the lipid carrier protein ApoE. Aiding in the maintenance and repair of neurons, this 34 kD protein on chromosome 19 comes in three common alleles: E2, E3, and E4. The last allele is known to contribute strongly to the great majority of AD cases by bringing on earlier onset and accelerating the cognitive decline. It also is less potent at supporting the brain’s attempt to regenerate from injury. However, just how ApoE4 functions in AD remains hotly contested. The hypothesis that ApoE exerts its main effect through amyloid plaque deposition is perhaps a leading contender that continues to evolve (see below). At the same time, other ideas abound and no explanation is as yet broadly accepted across the field. This story highlights some recent developments in an attempt to stimulate new discussion. (See Keith Crutcher's meeting summary and live discussion for prior coverage).
ApoE3 Protects Against Synaptic Damage Even before Plaques Form
In the December 15 Journal of Neuroscience, Manuel Buttini, working with Lennart Mucke and colleagues at the Gladstone Institute of Neurological Diseases, University of California, San Francisco, and elsewhere, describe provocative findings suggesting that some of apoE4's detrimental effects may have little to do with plaques (Buttini et al., 2002). A large body of prior work has demonstrated that apoE4 binds Aβ avidly and promotes fibril formation and plaque deposition (see, for example, Sanan et al., 1994; Rebeck, 1993; Bales et al., 1997; and Holtzman et al., 2000). And yet, Jacob Raber et al. have found that mice that are double transgenic for human APP/ApoE4-but not mice expressing hAPP and ApoE3-show a spatial memory deficit already at a young adult age when they do not yet contain plaques (Raber et al., 2000; see also Hartman et al., 2001). Mucke's group is working to test their hypothesis that ApoE3 protects against damage that nondeposited Aβ wreaks on synapses and cholinergic pathways early on in the disease. This is part of the lab's broader interest in the pathogenic mechanisms causing early cognitive deficits in AD.
In this study, Buttini et al. studied the interplay of human Aβ with different human ApoE isoforms in aging mice. To do so, they created transgenic mice that express mutant human APP, plus neuronally expressed human ApoE (either 3 or 4), but no mouse ApoE. The researchers then analyzed immunoreactivity for synaptophysin-a marker for presynaptic terminals-in the hippocampus and neocortex, as well as choline acetyltransferase activity in the medial septum, and immunohistochemistry of cholinergic fibers. These measures, they write, correlate better with cognitive dysfunction in AD than does plaque distribution (see, for example, Terry et al., 1991).
Buttini et al. conclude that ApoE3, but not E4, can delay synaptic damage wrought by a combination of increasing age and Aβ. They found a decline in the number of synaptophysin-positive terminals and fibers, which was associated with deficits in synaptic transmission strength in six different genotypes tested. In both neocortex and hippocampus, human APP expression caused a synaptic deficit between six/seven and 12/15 months of age. Importantly, this began prior to plaque deposition. Concurrent human ApoE3 expression delayed this deficit until around 19 months of age, but ApoE4 did not.
In very old mice, ApoE3 no longer protected against synaptic deficits. At this age, the hAPP/ApoE3 transgenics had similar synaptic and cholinergic deficits as the hAPP/ApoE4 mice, even though they had fewer plaques. In the neocortex, the synaptic and cholinergic loss occurred even though few plaques formed there. All these findings do not correlate synaptic damage with plaques very well. Rather, they imply that the different abilities of ApoE3 and E4 to protect synapses play out early and independently of plaque deposition. ApoE4’s effect on plaque deposition, and its attendant neuritic dystrophy, may occur later, these authors suggest.
"This pattern is reminiscent of what happens in humans, where ApoE3 and ApoE4 have their greatest impact on Alzheimer risk between the ages of 45 and 65, with less of an impact seen in older persons," said Mucke. "Although we confirmed previous reports that ApoE4 promotes plaque formation more than ApoE3, our study suggests that this is not all there is to the ApoE4-Alzheimer link. Plaque-independent mechanisms are likely involved as well, or are even more important."
How Does ApoE3 Protect Synapses?
Here, candidate mechanisms include the promotion of neurite outgrowth. Researchers have shown years ago that ApoE3 promotes neurite outgrowth better than E4 in cultured embryonic neurons and cell lines (Nathan et al., 1994, and others since then). Last February, Britto Nathan extended his original research to cortical neurons cultured from adult mice by using a method developed by Greg Brewer at Southern Illinois University School of Medicine. They showed that neurons cultured from adult ApoE knockout mice have shorter neurites than neurons from adult wild-type mice, and that adding human ApoE3 increased neurite outgrowth while adding ApoE4 decreased it. They suggest that this effect might involve the low-density lipoprotein receptor-related protein LRP (Nathan et al., 2002). Two months later, Bruce Teter working with Greg Cole at University of California, Los Angeles, published results of organotypic slice cultures of the hippocampus from ApoE3- or E4-transgenic mice. Experiments changing the expression levels of E3 and E4 suggest to these researchers that E4's outgrowth inhibition results not just from the lack of E3's growth-promoting effect, but from a negative gain of function in E4 (Teter et al., 2002).
Plaques Behind Conspiracy of Head Injury and ApoE
If all this implies to you that plaques are a mere sideshow in the ApoE-AD connection, read on. David Holtzman’s lab at Washington University in St. Louis, Missouri, has bred a series of double-transgenic mice similar to those of the Mucke lab, except that they express human ApoE in astrocytes, not in neurons. (In the human brain, glial cells are the main cell type in brain that synthesizes apoE, notes Holtzman.) His lab has focused its analysis of these mice more on Aβ metabolism and the process of plaque formation, which begins around 15 months of age in these animals, as well as on the relationship of plaques to dystrophic neurites (see ARF related news story and scroll to Holtzman). In the December 1 Journal of Neuroscience, first author Richard Hartman et al. report on the relationship between ApoE4 and traumatic brain injury Hartman et al., 2002. Both these factors increase the risk of developing AD in humans. There is evidence that they act synergistically; for example, ApoE4 carriers are 10 times more likely to develop AD after sustaining a traumatic brain injury, while ApoE4 alone “only” doubles their AD risk (Tang et al. 1996).
To study whether this interaction occurs via Aβ or via other effects on cell death and tissue injury, Hartmann et al. subjected PDAPP/hApoE double-transgenic mice to a head injury model originally developed in Tracy McIntosh'sgroup. They did so when the mice were 9-10 months old and then analyzed the mice’s brains for Aβ deposition, tissue shrinkage, and cell loss three months later, still prior to this strain’s typical age of Aβ deposition. The brain injury, it turned out, accelerated Aβ deposition more strongly in PDAPP/ApoE4 than in E3 mice. Fifty-six percent of the E4 mice had diffuse Aβ deposits and 44 percent had fibrillar plaques, whereas only 20 percent of the E3 mice had deposited diffuse Aβ, and none had mature plaques. By contrast, volume measurements and neuronal counts in the hippocampus and the cortex were similar between the genotypes. This suggests that the higher dementia risk of ApoE4 carriers after TBI is due, in part, to interactions between ApoE and Aβ, the authors write. They caution, however, that this relationship is complex, as ApoE is also involved in Aβ clearance (see Fagan et al., 2002). In both Mucke’s and Holtzman’s labs, hAPP-transgenic mice that do not also express human ApoE deposit more Aβ than double-transgenics of either the ApoE3 or 4 genotype. Hartmann et al. note that while ApoE isoforms clearly contribute to premature earlier amyloid deposition, this alone is probably not sufficient to cause dementia. They suspect that neuritic dystrophy associated with plaques, as well as other factors including oligomer formation, tangle formation, and synaptic loss, contribute to cognitive dysfunction.
Leaking Lysosomes: Molten Globule Gets the Blame
Let’s switch gears and turn to a more reductionist, biophysical view of ApoE. Last June, Zhong-Sheng Ji, working with Robert Mahley, Karl Weisgraber, and others at UCSF’s Gladstone Institute, added another candidate mechanism for how apoE isoforms might differ in their function (Ji et al., 2002). They analyzed the interaction of different ApoE isoforms with Aβ in cultured neuronal cells transfected to produce low levels of ApoE. They found that, together with Aβ42, ApoE4 made these cells’ lysosomes prone to leakage. This potentiated DNA fragmentation and death, probably via links to lysosomal proteases and their ability to activate caspases, as well as release of cytochrome C from mitochondria. ApoE3 was not protective in this assay. Researchers led by Charles Glabe had previously shown that Aβ accumulates in lysosomes and is slowly degraded there, but also disrupts their membranes (Burdick et al., 1997; Yang et al., 1998). ApoE shuttles in and out of neurons, but Ji et al. saw the lysosomal leakage and death only when ApoE was inside the neurons. The authors speculate that ApoE4 potentiates this lysosomal toxicity because it is more prone than ApoE3 to form a particular intermediate state, namely, the partially unfolded conformation referred to as a molten globule (Dobson, 2001). Some researchers believe that these reactive species represent a distinct thermodynamic state a protein can assume. Molten globules retain many features of the native state’s secondary structure, but do not adopt its tertiary structure, and they have more internal mobility than the fully folded protein. Molten globules bind to phospholipids, transfer through membranes, and interfere with membrane processes, the authors write. (For background read, for example, Bychkova and Ptitsyn, 1995.) Ji et al. speculate that in AD, ApoE4 might form such a structure, which then disrupts lysosomal membranes together with Aβ42. At the lysosomal pH4, ApoE4 becomes more unstable and reactive than ApoE3, Ji et al. write.
In the December 27 Journal or Biological Chemistry, Julie Morrow of Weisgraber’s group, and colleagues, follow up with experiments suggesting that ApoE4 indeed forms such a molten globule more readily than E3 or E2 Morrow et al., 2002. Morrow et al. first analyzed the process by which ApoE isoforms denature when treated with urea, and then used structural tools to characterize the different states. At pH4.0, ApoE4 denatured in a three-step fashion, with an intermediate plateau populated by a monomeric version of the protein that resembles a molten globule. In this conformation, the four-helix bundle of ApoE’s amino-terminal domain is opened to expose an otherwise hidden hydrophobic core. ApoE2 never assumed this intermediate state, and ApoE3 did so only under more harshly denaturing conditions.
Researchers are increasingly beginning to detect molten globules and realize that they can play important roles in physiological processes such as protein trafficking and translocation across membranes, Morrow et al. write. For example, human apolipoprotein A-1 can occur as a molten globule (Gursky and Atkinson, 1996). Mutations affecting the formation or stability of the molten globule state can interfere with their normal physiological roles. In Alzheimer’s, this conformation of ApoE4 could alter lipid transport across membranes and in this way help explain why ApoE4 is less able to support neurite outgrowth and neuronal repair after injury, Morrow et al. speculate. Researchers should also consider whether this conformation might disrupt lysosomal membranes, and whether its exposed β, structure might explain ApoE4’s proposed action as a “pathological chaperone” promoting β-sheet and fibril formation of Aβ (Gallo et al., 1994).
IDE: New Link Connecting ApoE and Aβ Degradation to Insulin
On a different front, a paper in this month’s American Journal of Pathology joins together ApoE with Aβ-degrading enzymes, introducing a potential new mechanism by which ApoE could affect AD risk Cook et al., 2002. David Cook, working with Suzanne Craft, Jerry Schellenberg, and others at the Veteran Affairs Puget Sound Health Care System and the University of Washington in, Seattle, report that levels of insulin-degrading enzyme (IDE), one of the handful of proteases implicated in Aβ degradation, are down by about half in AD patients of the ApoE4 genotype compared with ApoE2- and ApoE3-carriers. The researchers quantified protein and mRNA levels of this enzyme in different hippocampal subregions of postmortem brain slices from 26 patients with LOAD and 15 normal adults.
Lower IDE levels in ApoE4 carriers could presumably reduce Aβ degradation, especially intracellularly, leaving more of it around to aggregate. But how would ApoE4 reduce IDE levels? "At this point, we have no solid leads as to how or whether ApoE4 causes reduced IDE, or whether both are related to some other pathophysiological process," said Craft.
Craft et al. had previously shown that AD patients had changes in their insulin metabolism that varied depending on ApoE status. Specifically, they had shown that AD patients of the ApoE2 and 3 genotype had insulin resistance and an altered CSF/plasma insulin ratio, while AD patients with the ApoE4 genotype had normal peripheral insulin levels and metabolism. At first blush, this might seem contradictory; however, Craft et al. combine these findings into a proposed model for further testing. It goes like this: ApoE3/3 AD patients are more likely to have high levels of peripheral insulin, which disrupts the plasma/CSF insulin ratio and causes insulin resistance. These high levels of insulin act as a competitive inhibitor of Aβ degradation by IDE, even though these patients have normal IDE levels. Conversely, ApoE4 patients have low IDE levels, so that even though they may have normal insulin levels, there is insufficient IDE to degrade enough Aβ. "Thus, you can get to the same endpoint in two ways: by having high insulin levels that inhibit IDE activity, which we believe is more common for ApoE3/3 AD patients, or by having low IDE levels, which we believe is more common for ApoE4 patients," said Craft.
Those who have both low IDE and hyperinsulinemia would suffer a double whammy. Last April, Lenore Launer and colleagues reported a greatly increased AD risk in diabetic adults who carry ApoE4 among adults in the Honolulu-Asia Aging Study (Peila et al., 2002; ARF Ab degradation chat; ARF insulin-related chat; Craft ARF news story).
Sphingolipids and APP Processing: ApoE Meddles Here, Too
A separate and fascinating new line of investigation just became even more interesting, thanks to an added ApoE angle. This research looks at the role of sphingolipids and their metabolism (such as sulfatide and its hydrolytic product, ceramide) in APP processing. Available online in manuscript form at the Journal of Biological Chemistry, a new paper by Xianlin Han et al. from Washington University, St. Louis, Missouri, reports that ApoE strongly affects the content of one particular type of sphingolipid, namely sulfatide (Han et al., 2002). This ester of galactocerebroside, made primarily by oligodendrocytes, is a major component of myelin. It has long been known that accumulating sulfatide causes a form of leukodystrophy. But Han, with John Morris and others at Washington University, had reported last August that in people who had died after being diagnosed with very early AD, sulfatide of the gray matter was depleted by up to 93 percent, while the levels of most other lipid classes were normal. This appeared to result not from a defect in sulfatide biosynthesis, but from increased degradation, as these patients' ceramide content was up threefold (Han et al., 2002). While linking a sulfatide decrease and ceramide increase to mild dementia, this study did not implicate ApoE. Han's present JBC paper does that.
The researchers used electrospray ionization mass spectrometry to compare the levels of sulfatide (and a range of other phospholipids, sphingolipids, and cholesterol) in the brains of ApoE knockout mice, David Holtzman’s human ApoE3- and E4-transgenic mice, and CSF from people with different ApoE isoforms. ApoE affected none of the other lipids, but appeared to lower sulfatide content strongly in an isoform-dependent fashion. Further studies examining the carriers of these sulfatides suggest that it associates specifically with ApoE-containing high-density lipoproteins (HDL), which presumably pick up oligodendrocyte-derived sulfatide after their release from astrocytes, Han et al. speculate. The authors also write that their paper is the first to show that ApoE, indeed, as was proposed years ago by Weisgraber and others, regulates lipid transport and metabolism not only in the plasma, as is well-known, but also in the brain.
This new proposed role for ApoE is all the more intriguing, as other ongoing work is implicating ceramide in APP processing. At last year’s Neuroscience conference in Orlando, both Dora Kovacs' group at the Massachusetts General Hospital in Charlestown, and Mark Mattson’s group at the NIA in Baltimore, Maryland, presented data suggesting that rising ceramide levels in neurons correlate with increased Ab generation in an age-related fashion. Mattson presented a hypothesis integrating accumulating sphingomyelins with oxidative stress, which would lead to more ceramide and ACAT activity (see ARF related news story and scroll down to Kovacs). Luigi Puglielli, working with Kovacs, did not observe changes in ACAT activity in response to changing ceramide levels, but rather proposed an effect of ceramide on BACE stabilization. If confirmed and expanded, such work would point to a molecular pathway for how age-related changes in membrane lipid composition, perhaps compounded by ApoE isoform, could lead over time to excessive Ab generation in cases of LOAD. Finally, for a current study on ApoE isoform effects on HDL metabolism, see Hopkins et al., 2002.
Lipids, of course, are also implicated in Ab toxicity via their peroxidation. The three different isoforms of ApoE differ in cysteine residues. Ward Pedersen and Mark Mattson have published data showing that the ApoE2 and ApoE3 isoforms are much more effective in binding to a toxic product of lipid peroxidation called 4-hydroxynonenal (Pedersen et al., 2000). Mattson suggests that ApoE2 and ApoE3 may protect neurons against being damaged and killed by Ab by binding 4-hydroxynonenal with its cysteine residues. (ApoE4 lacks the cysteines, ApoE3 has one cysteine and ApoE2 has two.) Mattson and colleagues had previously demonstrated a key role for membrane lipid peroxidation in the neurotoxic action of Ab.
This ApoE update does not presume to be comprehensive, and the writer invites everyone to fill gaping holes by e-mailing Gabrielle@alzforum.org about other recent work that appears particularly relevant. Yet, even this partial summary shows plainly that the manifold aspects of ApoE’s functions form a complex kaleidoscope of clues and glimpses. Is the challenge now to find the one most important to the development of Alzheimer's? More likely, the truth involves several overlapping and synergistic functions, making the challenge one of assembling the pieces into a big picture that enlightens everyone. Brilliant minds out there, step up to the plate!-Gabrielle Strobel.