See also live discussion on this topic.
The Science Fiction channel occasionally runs a marathon of "Twilight Zone:" Back-to-back episodes all day long, each one related to the other through novel scripts with interesting ideas that border on fantasy. The satellite symposium on apolipoprotein E, held at the San Diego Convention Center the day before the full invasion of neuroscientists occurred, at times felt eerily like such a marathon. Twenty-eight speakers followed one another, each telling their own tale of ApoE (or ApoJ, as the case may be). For those of us who feel that ApoE hasn't gotten the attention it deserves, the daylong session was a welcome change. Even so, the ApoE story remains incomplete, and a variety of potential roles are just now beginning to be understood. The major plots of each episode are highlighted below, grouped roughly by functional themes.
Robert Mahley, Gladstone Institute, UCSF: The opening keynote presentation set the tone for the rest of the day by emphasizing that ApoE may exhibit isoform-specific effects based on aspects of protein structure that require more detailed studies. For example, interactions between N and C terminal domains (Arg61) render the structure of E4 more compact than that of E3. Transitional forms known as reactive molecular protein intermediates may interact with intracellular membranes in isoform-specific ways. β-VLDL or β-VLDL with E3 or E4 all cause increased production of Aβ, but E4 is more effective in this regard. Surprisingly, these effects appear to be attributable to the lipid-poor ApoE fractions. Is this due to differences in retention or clearance? Apparently not. Rather, there appears to be accelerated intracellular AβPP recycling, and E4 increases intracellular AβPP to a greater extent than E3. The mechanism is unknown but the isoform difference depends on the Arg61 site. Thus, overall conformation of ApoE is critical to its effect on Aβ production. Another effect of these intermediates may be to destabilize lysosomal membranes. Neuro2A cells exposed to E3 or E4 show greater lysosomal leakage in response to Aβ, with greater disruption in the presence of E4. E4 also potentiates Aβ-induced DNA fragmentation, whereas E3 has no effect. The hypothesis for these effects is that E4 is converted to a more reactive intermediate than E3. By interacting with membranes, this intermediate may cause lysosomal leakage, alter intracellular processes and disrupt the cytoskeleton.
Karl Weisgraber, Gladstone Institute, UCSF: Extending the long list of Gladstone contributions to understanding ApoE structure and function, Weisgraber provided more information on the idea of ApoE intermediate conformations (or molten globules). The induction of such forms, which appear to be the rule rather than the exception for many proteins, occurs with low pH. The molten globule protein retains secondary and tertiary structural features and has a partially exposed hydrophobic core. For the N-terminal domain of E4, this molten globule conformation forms more easily than for the other apoE isoforms. This results in increased sensitivity to pepsin digestion, and FTIR and DLS studies demonstrate a conformational change such that, in the absence of urea, there is greater alpha-helical content than in the presence of 3.75 M urea. The E4 molten globule is flexible and has an exposed hydrophobic core, which could affect its activity. Indeed, E4 binds and disrupts membranes more effectively than E3. The study of these intermediates may point to unique conformational differences that underlie the isoform-specific differences in ApoE activity.
Bruce Teter, VAMC, UCLA, reviewed evidence for isoform-specific effects of ApoE on neurite sprouting. He noted recent evidence by Pfrieger that astrocytes are implicated in delivering cholesterol to neurons via an ApoE-dependent pathway. Assuming this pathway is defective with E4, this could accelerate AD by a failure of regenerative events. He also reviewed studies suggesting that E4 has negative activity (human data from Arendt et al; mouse data from Buttini et al.) and summarized data from his own laboratory, indicating that mossy fiber sprouting depends on apoE and is stimulated by estrogen. E4 is defective in this regard; indeed, increased levels of E4 result in less sprouting.
Frank Pfrieger, Max-Planck/CNRS, Strasbourg, France, summarized data recently reported in Science in which his group identified a critical role for cholesterol in mediating the synaptogenic effect of astrocytes on cultured neurons. Their in-vitro system involves rat retinal ganglion cells purified by immunopanning and grown in the presence or absence of glia. These neurons normally die without glia but Pfrieger worked out conditions to keep them alive. With glia present there is much more synaptic activity, an effect that is not dependent on contact. Seeking to identify the soluble factor secreted by glia, the Pfrieger group identified apoE using mass spectroscopy. But ApoE by itself did not reproduce the effect-cholesterol did. Astrocyte-conditioned medium increases neuronal cholesterol content, suggesting that the neurons cannot produce sufficient cholesterol to support synapse formation. Glia provide the critical factors needed to get this cholesterol (see related news item).
Jonathan Smith, Rockefeller University, considered the interaction of ApoE and estrogen in mouse models. He noted that background strain may affect baselines such as Aβ burden. FVB mice, for example, have much higher levels than C57 APP mice. Studies of ApoE -/- mice expressing human E2, E3, or E4 and APP show that E3 and E4 mice have higher Aβ levels than E2 mice or apoE null mice. sAPP levels are lower in the E4 mice. What about estrogen? Women are more likely to develop AD than men. Estrogen users have a 29 percent lower incidence but only in E4-negative women. Estrogen specifically induces ApoE in the cortex and diencephalon, but apparently by different mechanisms. Induction in cortex is independent of ER-a, but these receptors are apparently involved in induction in diencephalon. 17β-estradiol led to a 24 percent decrease in Aβ. Estrogens were found to decrease the ratio of Aβ/sAPPa, indicating an effect on amyloidogenic processing.
Steven Paul, Eli-Lilly & Company, reviewed the role of ApoE in amyloid deposition. ApoE facilitates Aβ deposition and contributes to fibrillization into plaques. The role of inflammatory pathways in this process was examined by infusing LPS into PDAPP mice over two weeks. In addition to causing gliosis, LPS caused a dramatic upregulation of ApoE mRNA and a significant increase in plaques in three-month-old mice (comparable in amount to seven-month-old mice). Interestingly, animals lacking ApoE had enhanced mortality following LPS treatment but the surviving mice never developed amyloid deposits in response to LPS. A possible role for cholesterol is indicated by the fact that male PDAPP mice on a high-fat diet have much more Aβ deposits. However, serum cholesterol is very high in ApoE-/- mice, yet there is less Aβ deposition. So how to explain the cholesterol effect? In vitro studies indicate that mixed neuron-glial cultures stimulated with cytokines or NMDA release more ApoE, an effect that is inhibited by mevastatin. This provides a potential mechanism by which statins have anti-inflammatory effects independent of their role in cholesterol lowering. Paul suggested that drugs inhibiting glial ApoE expression may reduce Aβ deposition, amyloid burden and neuritic plaque formation.
Daniel Michaelson, Tel Aviv University, discussed the effect of environmental enrichment on cognitive performance in transgenic ApoE mice. On a T-maze, an enriched environment improves the performance of E3 but not E4 mice. In a working memory test, E4 and E3 mice both perform worse than controls but there is not much difference between E4 and E3 mice unless the effect of an enriched environment is considered, in which case E4 mice are much worse. Interestingly, NGF is elevated in the hippocampus of E3 but not E4 mice exposed to an enriched environment. Michaelson also described the effect of LPS on microglia and astrocyte activation. The astrocytic response does not occur in ApoE-deficient mice but the microglial response is intact. E3 can restore the astrocytic response, but E4 does not, suggesting the possibility of isoform-specific effects of ApoE on inflammation. In addition, using a head injury paradigm, E4 mice have a greater mortality and there is a greater production of APP in the E3 mice following head injury than in E4 mice.
Jacob Raber, Gladstone Institute, UCSF, reviewed several lines of evidence suggesting that female ApoE4 transgenic mice have impaired cognition. To understand the basis for this sex-specific effect, studies were carried out to examine the role of androgens. In fact, evidence suggest that androgens protect against ApoE4-induced cognitive defects, and performance depends on the balance between plasma androgen levels and cytosolic receptors. These results point to possible sex-based effects of ApoE that might contribute to differential vulnerability to AD.
Manuel Buttini, Gladstone Institute/Elan Pharmaceuticals, summarized results from NSE-ApoE mice (see Raber) and from double-transgenic mice expressing ApoE along with AβPP. Aβ levels are similar in the different ApoE genotypes but synaptophysin staining indicated greater synapse loss in E4 mice. In double transgenics, the isoform difference is smaller, i.e., there is more synapse loss in both the E3 and E4 mice, suggesting that E3 is less protective against amyloid-induced changes. Interestingly, the loss of synapses appears to precede amyloid deposition.
Eliezer Masliah, UCSD, reported that ApoE4 increases death of primary cortical neurons and that this effect may involve calcium influx. Others had previously reported similar findings. The toxicity is dose-dependent and blocked by EGTA and RAP. E3 is toxic only at higher concentrations. The correlation between increased intracellular calcium and cell death appears to be good. Ionomycin toxicity is also enhanced by E4. Thapsigargin does not block, but EGTA and cobalt both inhibit E4 toxicity. Masliah proposed that E4 may not be directly neurotoxic but could be increasing vulnerability of the neuron to other insults.
Mary Jo LaDu, Northwestern University, who has previously documented isoform-specific effects of ApoE on amyloid structure and function, outlined a plan for pursuing the mechanism of ApoE-Aβ interactions. Taking advantage of recent evidence indicating that the structural state of Aβ may be critical for toxicity, her group has developed a means to reliably produce both oligomeric and fibrillar preparations of Ab42. The oligomeric form was found to be approximately 10-fold more toxic than the fibrillar form and was more effective in activating astrocytes. The availability of such preparations now makes it possible to study the effect of ApoE on Aβ structure and function in a more mechanistic way.
Yadong Huang, Gladstone Institute, UCSF, summarized studies supporting the hypothesis that ApoE proteolysis may play a role in Alzheimer's. Beyond published data indicating the presence of ApoE fragments in AD brain tissue, Huang reported preliminary data suggesting that such fragments also appear in transgenic mice but vary with the type of promoter. NSE-ApoE mice show more evidence of ApoE fragments than GFAP-ApoE mice. Also, there appears to be some isoform difference in the types of fragments generated. 14-20 kDa fragments are more abundant in E4 mice, whereas 29 kDa fragments occur in E3 mice, suggesting neuron-specific processing with greater susceptibility of E4. There is also an apparent age-dependent accumulation of the apoE fragments. Interestingly, phosphorylated-tau increases in NSE-E4 mice but not NSE-E3 mice and intraneuronal p-tau inclusions appear in the hippocampus where they colocalize with ApoE. Since ApoE expression in neurons is normally thought to be limited, Dr. Huang suggested that aging, stress, or Aβ may contribute to increased ApoE expression in neurons.
Keith Crutcher, University of Cincinnati: Yours truly returned to the theme of ApoE proteolysis. This overview included data supporting isoform-specific neurotoxic effects of ApoE, and a role for LRP and NMDA receptors in this effect, as well as evidence for increased ApoE fragments in AD brain as compared with control tissue. The N- and C-terminal fragments appear to be differentially associated with tangles and plaques, respectively, based on immunohistochemical data obtained with monoclonal antibodies that recognize distinct epitopes. I proposed that proteolysis of ApoE may result in the generation of N-terminal fragments that contribute to toxicity and tangle formation and C-terminal fragments that bind to Aβ and contribute to plaque formation.
Robert Brendza, Washington University, St. Louis, provided an ultrastructural view of amyloid deposits in transgenic mice using quick-freeze deep-etch technology. Strikingly, the morphology of the deposits depended on apoE. In its presence, the plaques had a highly organized fibrillar structure but in its absence, Aβ deposits are disorganized. ApoE appears to affect plaque structure even though it is not absolutely required amyloid deposition, since 37 percent of the ApoE null mice still have at least some thioflavin-positive plaques.
Anne Fagan, Washington University, St. Louis, reviewed evidence from transgenic mice implicating ApoE in amyloid deposition (see Brendza). Human E3 appears to suppress amyloid deposition but E4 mice eventually catch up to AbPP mice expressing mouse apoE. E2 is not able to overcome the ApoE KO phenotype. The pattern of amyloid deposition in E4 mice is also more like the pattern in the presence of mouse ApoE than in mice carrying E3. Attempts to identify the interactions that might be relevant include studies of lipid rafts, where evidence suggests that ApoE may influence the association of Aβ with the lipid rafts (DIGS).
Thomas Wisniewski, New York University, updated the audience on his hypothesis, proposed when a role of ApoE in AD was just being considered, that ApoE is a pathologic chaperone. ApoE mice show better clearance of injected fibrillar Aβ, and ApoE enhances Aβ42uptake by astrocytes, with E4 improving uptake of Aβ40 more than of Aβ42. These effects are partly inhibited by RAP, suggesting a role for LRP.
In order to compare morphology of neurons developing under different transgenic conditions, gene gun labeling studies of dendritic spines of dentate gyrus granule cells were carried out in hippocampal slices from E3, E4 or apoE KO mice. Spine density is greater in E3 neurons as compared with knockout and E4 neurons. The ApoE genotype had no effect on dendritic length or dendritic number.
Linda Van Eldik, Northwestern University, Evanston, Il provided an update on the role of ApoE in modulating Aβ effects on glial neuroinflammatory responses. ApoE blocks Aβ- but not cAMP-induced glial activation. There is no apparent isoform difference. Furthermore, when Ab was aged in the presence of ApoE it did not cause activation, and ApoE was also able to reverse the "activated" phenotype. In addition, cultured glia activated by Aβ demonstrate increased ApoE production. Eldik proposed that the increase in ApoE may actually attenuate inflammatory processes.
Daniel Laskowitz, Duke University, Durham, NC, described animal studies in which peptides derived from the receptor-binding domain of ApoE were used in a closed-head injury model to determine if they would modify the outcome. The rationale stems from studies showing that ApoE suppresses glial TNF-a secretion in an isoform-specific manner and downregulates microglia. E3 is more effective than E4 in this regard. ApoE-/- mice show greater upregulation of TNF-a and more brain swelling (as visualized on MRI) in response to closed-head injury compared to wildtype mice. To determine whether ApoE mimetic peptides might also show protection, a variety of peptides were injected into the tail vein after the head injury. Peptides made up of amino acids 130-149 was as protective as the whole protein, whereas peptide 130-146 had no effect. Peptide 136-149 had intermediate activity but peptide 139-146 had no activity. Peptide 133-149 improved survival after closed-head injury, prevented weight loss, and improved rotarod and Morris water maze performance. These results suggest that ApoE peptides may have potential therapeutic uses.
Huntington Potter, University of South Florida, Tampa, gave a brief history of the amyloid cascade hypothesis and asked "What exactly is causing AD?" Aβ is clearly important but which form matters most? Is it Aβ or amyloid? He noted that IL-1 polymorphisms are also risk factors and suggested that there may be ways to link these factors. ApoE and the chaperone ACT are amyloid promoters. Several mouse lines were studied to determine the impact of these two promoters of amyloid. APP/ACT mice have plaques that contain ACT, and having both ACT and apoE causes tremendous amounts of amyloid deposition. If you drop out ACT you get less, with ACT only you get even less. Dropping out both ACT and apoE leads to almost no amyloid. ACT and ApoE are both pathological chaperones that work together, but ApoE seems to have the stronger influence. ACT promotes congophilic amyloid in the presence of ApoE but doesn't cause it alone. In behavioral tests, cognitive impairment depends on ApoE and ACT-catalyzed amyloid formation. Mice without apoE or ACT have no deficits. So amyloid is necessary for the effect. There is a good correlation between amyloid load and a cognitive deficit (escape latency). Potter cautioned that he didn't know the process or the product of amyloid formation was important, but he proposed that the interaction between ApoE and Aβ may be a good drug target.
Caleb Finch, University of Southern California, Los Angeles, strayed from the apoE theme to highlight the possible role of ApoJ expression in normal aging brain and in Alzheimer's disease. He noted that inflammatory markers increase in brain aging and in AD. ApoJ is activated more than apoE, and aging increases apoJ mRNA and protein in B6 mouse brain at least 2fold. Cultured rat microglia are activated by apoJ (400 nM, 48 hr), but not ApoE, based on morphological criteria and NO production. ApoJ also induced TNFa release, and infusing a diffusible neurotoxin causes activation of microglia in white and gray matter. Furthermore, there is a correlation between apoJ levels and ApoE genotype. Finch also pointed out the need to carefully examine cellular interactions. For example, monotypic astrocyte cultures yield different results than mixed glial cultures. Aβ induces ApoE and ApoJ in mixed glia but decreases ApoE message in isolated astrocytes. Since astrocyte always contact microglia, such heterotypic interactions may be important to model in vitro. Finch proposed the hypothesis that brain regions expressing higher levels of ApoJ are more at risk for neurodegeneration, either through activation of microglia or interaction with Aβ aggregation to favor forming soluble aggregates.
William Klein, Northwestern University, continued the theme, showing that ApoJ can enhance Aβ toxicity. Aβ may be innocuous at first but become toxic with aging; the first formed oligomers may be toxic. Added at sub-stoichiometric concentrations, ApoJ can enhance Aβ toxicity, suggesting that apoJ may chaperone the assembly of globular molecules (ADDLs), although apoJ is not required to make fibril-free oligomers. ADDLs rapidly form in vitro at nanomolar doses and change with temperature. They also rapidly inhibit LTP although they have no impact on EPSPs or LTP. Klein proposed that this effect of ADDLs could help explain the vaccination data of Morgan et al., who showed a beneficial effect of vaccination without loss of plaques. ADDLs show selectivity in that they kill hippocampal neurons readily than cerebellar ones, consistent with the fact that ligand blot assays show more "receptors" in hippocampus than in cortex and none in cerebellum. ADDLs from AD patients are comparable in size, isoelectric point and conformation and are much more abundant in AD brain compared to controls.
Ron DeMattos, Washington University, described studies of apoJ KO/APP mice to determine whether apoJ might participate in the amyloid cascade. No difference in total Aβ levels was found but there was sixfold less fibrillar Aβ and far fewer dystrophic neurites surrounded amyloid deposits. Biochemical studies show that oligomers may be present but there was no difference in the total amount present in carbonate lysates. Non-denaturing ELISAs, however, show less overall Aβ in the ApoJ KO mice, suggesting that the oligomers are not being detected. When oligomers are detected, there is no difference in the levels. Acid-urea gel analysis (85 percent formic acid extraction) shows good separation between Aβ40 and Aβ42. Thus, the presence of ApoJ may affect the type of amyloid formed.
Joachim Herz, University of Texas Southwestern Medical Center, Dallas, updated the audience on rapidly expanding data relating to ApoE receptors, which now are suspected to be involved in signaling pathways to a much greater extent than previously thought. He noted that cardiovascular and neurodegenerative diseases share common mechanisms, both pathobiochemical and genetic, that are hard to understand in light of traditional views of ApoE function. Some of these commonalities might best be understood if ApoE is involved in signaling functions. The family of ApoE receptors continues to grow. VLDLR and ApoER2 are both implicated in neuronal migration during brain development. Reelin binding to these receptors leads to phosphorylation of disabled adaptor protein (dab1), which eventually causes cell shape changes. KO of dab1 or these receptors leads to the same phenotype as the reeler mutant. Other gene defects that cause neuronal migration disorders include cdk5 KO (cytoplasmic protein kinase), p35 KO (activator of cdk5) and LIS1 deficiency (cytoplasmic PAFAH). Tau phosphorylation is increased in reeler, ApoER2 KO and VLDLR KO mice. So cdk5 may be downstream of these effects since it is involved in tau phosphorylation. P35 KO does not have a major effect on hippocampal anatomy. But p35 KO with receptor KO results in cortical lamination defects but no problems in cerebellum, unlike reeler. None of these mouse models have yet been crossed with AD mouse models.
Dudley Strickland, American Red Cross, Maryland, who has contributed extensively to our knowledge of LRP, examined this enigmatic receptor's possible role in regulating protease activity in the brain. LRP has at last 30 ligands; most of these are proteases or protease complexes. TPA, one of the 30, has an inhibitor called neuroserpin (NS, a serine proteinase inhibitor), which has been reported to be associated with dementia (Davis et al, 1999.) Is this due to formation of inclusion bodies or a deficiency of neuroserpin? Serpins form a stable covalent complex with inhibitors but dissociate slowly over a period of days. LRP clears these complexes from the circulation. The NS:tPA complex is relatively unstable. Does tPA modulate NS function? Only the NS:tPA complex strongly binds LRP in vitro, NS does not. However, cells internalize NS and the complex. Cleaved NS is not a ligand and is not internalized. This depends on LRP. NS internalization is not mediated by tPA, so a cell-surface co-factor is probably involved. Fast turnover of the complex suggests that tPA may modulate NS function. Another protease, MMP-9, is upregulated in AD neurons and elevated in ischemia. It facilitates growth of oligodenrocytes. MMP-9 is degraded by LRP+ cells but not LRP- cells. Cortical neurons accumulate MMP-9 in media, an effect increased by RAP. Both MMP-9 and TIMP-1 bind LRP. In summary, both the tPA and MMP-9 systems may be important ligands for LRP.
Brad Hyman, Massachusetts General Hospital, Boston, described studies of APP-LRP interactions using FRET. LRP and APP co-immunoprecipitate and colocalize in cells in culture. The use of FRET reveals close interactions between two fluorophores (20 nm), thus enabling demonstration of relevant interactions in living cells at relatively high resolution. Tagged intracellular domains of LRP or APP with myc or GFP were used to study possible interactions, as revealed by a change in FRET ratio. APP770 and LRP showed high interaction that is decreased by RAP. APP695 and LRP also show low-level interaction not reduced by RAP. This suggested another site of interaction, perhaps in cytoplasmic domains. Fe65 shows interaction with APP770. These studies suggest that the extracellular domains of APP770 and LRP are closely linked (see ARF news story.)
William Rebeck, Massachusetts General Hospital, continued the theme of apoE receptors and noted that apoE lipid-binding and receptor-binding domains are present on Aβ deposits, suggesting the possibility of a defect in Aβ clearance. He asked whether the presence of these domains in plaques might contribute to long-term effects on signaling pathways in adjacent cells. LRP and ApoE partially co-localize in Aβ deposits and colocalize with astrocytic processes surrounding plaques in mice. Immunohistochemistry indicates that Fe65 is also in a similar location around plaques. Activated α2 macroglobulin (A2M) alters calcium homeostasis in primary neurons by decreasing resting calcium and attenuating the increase normally caused by NMDA. E3 has a similar effect. E4, however, causes an increase in resting calcium and in the response to NMDA. This is blocked by RAP, suggesting a role for LRP. These effects point to the possibility of long-term signaling effects of LRP ligands present in plaques.
Guojun Bu, Washington University, discussed the endocytic trafficking of LRP and other lipoprotein receptors. He noted that the cytoplasmic domain of the LDL receptor family is small relative to the ectodomain. RAP is an ER molecular chaperone for LRP that also serves as a useful antagonist for blockade of ligands to LRP. Membrane-containing LRP minireceptors were generated and expressed in LRP-deficient cells to study binding and degradation of LRP ligands and signals for endocytosis. There are four putative signals in the 100 aa cytoplasmic tail of LRP, which has an endocytosis half-time of less than 30 seconds. VLDLR and apoER2 have much slower endocytosis rate, perhaps relating to their signaling functions.
So there you have it, perhaps more than you ever wanted to know about ApoE. My hat comes off to organizer Lennart Mucke, moderator David Holtzman, and all presenters for such a stimulating symposium!—Keith Crutcher
- Davis RL, Shrimpton AE, Holohan PD, Bradshaw C, Feiglin D, Collins GH, Sonderegger P, Kinter J, Becker LM, Lacbawan F, Krasnewich D, Muenke M, Lawrence DA, Yerby MS, Shaw CM, Gooptu B, Elliott PR, Finch JT, Carrell RW, Lomas DA. Familial dementia caused by polymerization of mutant neuroserpin. Nature. 1999 Sep 23;401(6751):376-9. PubMed.