This past August, Alzheimer disease researchers met with colleagues from other fields and with foundation and NIH representatives in Bar Harbor, Maine, at the sixth annual workshop on Enabling Technologies for Alzheimer Disease Research. The participants were charged with identifying knowledge gaps that hold back progress toward developing new therapeutic strategies, and they discussed opportunities for bridging these gaps. One session dealt with the influence of the immune system on synapses in brain development, animal models, and in aging and AD. The second session emphasized synaptic function in aging and AD; the third highlighted some high-throughput approaches that can be applied to problems of brain aging and neurodegenerative disease. The report below offers a summary of two days of presentations and discussions, followed by a list of recommendations the group compiled. Readers will find a broad description of how the big questions in AD research are changing, as well as plenty of ideas to update their own studies. Comments are always welcome.
Clearly, the functions of the immune system in the brain must be better understood. Open questions fall into three areas: 1) What normal roles do immune molecules play in the brain irrespective of inflammation or disease? 2) When immune cells become activated in the aging, degenerating brain, is that good or bad, and how so? 3) What challenges and opportunities does the immune system pose for AD immunotherapy?
Current interest in immune molecules in the brain focuses on the CNS synapse. Mechanisms of synapse maintenance are drawing increasing attention in neurodegeneration research. First, some words on the structure of the CNS synapse. Its protein scaffold is surprisingly stable and is beginning to be better understood in molecular terms. Adhesion molecules are anchored to the cytoskeleton beneath the presynaptic and postsynaptic membranes; they protrude into the synaptic cleft and hold both sides together by homophilic binding. A major class of synaptic adhesion molecules are cadherins, and the brain expresses all 80 known human forms of it. It is thought that the binding specificity of different cadherins helps specify neural connections. N-cadherin clamps excitatory CNS synapses together. At immature synapses it is aligned across the active zone and postsynaptic density, but in highly active mature synapses, it is arranged like the rim of a drum around a central cadherin-free core, which gives the appearance of being "perforated" as the synapse matures. In addition to being tissue glue, cadherins transduct signals and suppress tumors by holding cells in place. Cleavage of their prodomain activates cadherins for their adhesive function, and dysfunction of this process is implicated in metastasis. Prodomain cleavage may also play a role in synapse development; its role in synapse aging, repair, or loss, if any, is not known (Dustin and Colman, 2002).
The field of neuroimmunology is undergoing a shift as a growing list of proteins formerly thought of as strictly immunological is found to occur in the nervous system and to be required for brain function. The complement cascade represents one such set. Current research implicates the complement as a risk factor for age-related macular degeneration (see ARF related news story). More study is needed to explore their role in other forms of neurodegeneration, with a particular focus on glial cells.
The complement cascade removes infected cells and cellular debris in the periphery by lysing cells or by coating bacteria and waste materials to trigger phagocytosis by macrophages (opsonization). Complement deficiency leads to autoimmune disorders and impairs tissue regeneration; excessive activation leads to tissue damage. In neurodegeneration, the complement is activated but its precise role in AD remains unclear. New research indicates that the complement cascade might serve a fundamental role in brain development. Specifically, complement proteins would create a permissive window of time, in the early postnatal period, during which activity-dependent synaptic pruning in the CNS can occur to refine an approximate projection pattern established in utero. Astrocytes begin releasing the initiating complement protein, C1q, just before pruning begins, and stop producing C1q after it has ended. Both C1q and C3 knockout mice are impaired in desegregating initially overlapping projection patterns of visual neurons coming from either eye. It is unclear whether the complement merely helps remove synapses once they are defunct or plays a more causal role, and how that would relate to synaptic activity. Early AD features both synapse loss and reactive astrocytes, which are known to re-express C1q. Genetic complement defects in humans are known and could be examined for consequences on cognition.
Complement proteins do not appear to hold promise as a CSF biomarker, but they have been reported in the vicinity of amyloid plaques in AD brain tissue. Both Aβ and tau pathology are thought to activate complement. In genetic studies crossing either APP or P301L tau transgenic mice with complement-inhibited mice, complement deficiency increases amyloid deposition, but also reduces tau pathology and neuron number. Taken together, the new mouse genetics would suggest that the complement in AD brain clears amyloid through the microglial/macrophage opsonization function, and that it can also promote tau phosphorylation and eliminate synapses and neurons through mechanisms still unknown.
Scientists need to clarify how well current amyloidosis and AD mouse strains model the human brain with regard to its immunological component. Scientists are beginning to realize that while APP transgenic mice do upregulate and initiate the complement cascade, at least some strains fail to complete it. This raises the question of whether the absence of the complement's downstream effectors might explain why many APP/PS models do not show the neurodegeneration that marks AD. The role of the complement in immunotherapy remains unclear.
MHC class I proteins are a distinct family of immune proteins that are unexpectedly expressed in the brain. Well-known for their role as ubiquitous antigen-presenting proteins, they have also recently been found to shape synaptic plasticity. Contrary to better-known enhancers of synaptic plasticity, such as BDNF or CREB, MHC class I proteins are emerging as negative regulators of plasticity. The mouse brain during development and in adulthood expresses these proteins in specific patterns, particularly in regions undergoing activity-dependent synaptic plasticity and remodeling, and endogenous activity upregulates MHC class I proteins in vivo. They are required for normal activity-dependent remodeling of initial synaptic circuits, as well as for normal synaptic plasticity in the adult hippocampus. MHC class I proteins appear to influence synaptic physiology by limiting long-term potentiation and favoring long-term depression, which may in turn affect certain forms of learning and memory. These and other recent findings together weaken the old view of the brain as an immune-privileged organ that lacks MHC class I expression, to one where many different MHC class I genes—and indeed other components of the immune system—are expressed and perform key neuronal functions, although their immune functions are actively suppressed.
In AD, what MHC I does is still speculative. Aging changes MHC class I levels such that they decrease in non-neural tissues, but may paradoxically increase in the brain. Due to the novel neuronal functions of MHC class I, this age-related increase in MHC class I could promote functional and anatomical synaptic weakening, loss of neuronal connectivity, and impaired learning and memory. Since MHC class I in the brain functions to limit plasticity, eventual therapeutic implications could be that drugs lifting this negative regulation might boost synaptic plasticity in remaining circuits. In addition, scientists know that activated T cells (whose receptors bind MHC class I complexes) enter the brain in great numbers, particularly where amyloid has damaged the blood-brain barrier. Activated T cells bind MHC class I, and can kill neurons in vitro. This raises the question of whether neuronal expression of MHC class I in specific brain regions, perhaps affected by age or changes in synaptic activity, could render these neurons selectively vulnerable to T cell-mediated damage. Like many proteins involved in AD, neural MHC class I proteins are upregulated after injury, indicating that a secondary wave of neuronal and/or immunological MHC class I-mediated synaptic weakening, remodeling, and damage could occur in regions undergoing active degeneration during later stages of AD.
The molecular details of how MHC class I signals in the brain are still being worked out. One putative neuronal MHC class I receptor has recently been identified (Syken et al., 2006). Part of the innate immune system, paired immunoglobulin-like receptor B (PIRB) binds MHC class I proteins on some neurons and restricts the extent of plasticity that is normally induced by changes in visual input. However, mice genetically deficient for PIRB fail to recapitulate the developmental remodeling defects seen in mice lacking MHC class I, indicating that MHC class I uses multiple binding partners to achieve its neuronal functions. Outside the brain, MHC class I binds dozens of receptors, and ongoing efforts are aimed at identifying additional binding partners for MHC class I in neurons.
Research opportunities in understanding the role of these proteins in AD include identification of the receptors that mediate the effects of MHC class I on normal developmental synaptic remodeling; characterization of expression of MHC class I in normal aging brain, AD patients, and APP/PS transgenic mice; characterization of LTP, LTD, synaptic remodeling, and learning and memory in MHC class I-deficient and overexpressing mice; and crossbreeding of AD mouse models onto MHC class I transgenic backgrounds. More broadly, MHC class I genes exist in many different alleles and are expressed in specific patterns in the brain, offering opportunities for studies in other instances of neurodegenerative or learning paradigms in which other brain regions are affected. Importantly, the peptides that different MHC class I proteins display in different brain areas and normal/disease states are completely unknown.
On the second question of what consequences immune cell activation has in the AD brain, the prevailing view that it is destructive is gradually giving way to a more nuanced realization that aspects of immune/glial cell activation can be protective. Evidence that the immune system contributes to AD pathogenesis includes complement activation, the apparent protection afforded by NSAIDs, cytokine expression, activated microglia, and the presence of autoantibodies to Aβ, cholinergic neurons, and neurofibrillary tangles. Evidence for a therapeutic role exists in the immunotherapy results thus far. Different activation states of immune cells need to be characterized and understood functionally. For example, microglia can be activated to be phagocytic, but also to be neurotoxic. Understanding the differences will provide a basis for interfering therapeutically to elicit a helpful, safe immune response.
The relationship of microglial activation to plaque formation needs more study. Recent genetic data indicate that, in CRND8 APP transgenic mice, microglia require the adaptive immune system to clear amyloid. Without functioning T and B cells, microglia failed to become activated, and the mice's plaque and Aβ burden increased. Genetic ablation of natural killer cells of the innate immune system further increased Aβ load. This demonstrates a connection between T and B cells and CNS microglia, providing yet another indication that there is no firm blood-brain barrier to separate cellular immune responses. The finding that amyloid pathology increased without microglial activation would suggest that, in future immunotherapy approaches, an Aβ-specific Th2 response might still be necessary to promote microglial plaque clearance. Because the meningoencephalitis that halted Elan/Wyeth's AN1792 trial probably arose from Aβ-specific T cells, some current immunotherapy efforts try to avoid an Aβ-specific T cell response. More effort needs to focus on how to distinguish a microglial clearance function from an inflammatory one. Research also should address the question of whether microglia by themselves can have a role in synaptic function or memory.
The complications encountered in the AN1792 trial have created the need for safer vaccines. An active vaccine is viewed as less expensive and more likely to become widely available than a passive one. Efforts to nudge the immune system toward a humoral response but away from an Aβ-specific T cell response focus on finding the right combination of adjuvant, route of immunization, and the right immunogen. B cell and T cell epitopes on Aβ are largely known. Novel immunogens include, for example, a dendrimeric construct that displays 16 copies of Aβ1-15 on a lysine tree (Seabrook et al., 2006), or constructs that place Aβ1-15 on either side of RGD or lysine spacers (Maier et al., 2006). In general, these immunogens achieve higher antibody titers in mice than full-length Aβ, and stimulate T cells against the synthetic immunizing peptide but not endogenous Aβ. Intranasal immunization combined with E. colienterotoxin-based adjuvants boosts the immune response in wild-type and APP-transgenic mice. Some of these novel immunogens reduce cerebral Aβ levels and pathology and improve spatial memory performance in the mice. They are being tested in Caribbean vervet monkeys, but none have entered clinical trials yet. Immunotherapy studies are limited by the lack of an AD-like mouse model that shows the peripheral T cell infiltration following immunization that occurred in the human trial; however, immunizing mice with both full-length Aβ and a combination of several adjuvants partially recapitulates this situation.
While second-generation active and passive immunotherapy approaches are being studied in phase 1 and 2 trials, questions about its mechanisms remain open. The human data show patchy clearance consistent with T cell activation of microglia and microglial phagocytosis, but mouse data on efficacy and Aβ clearance vary among models. Some mouse data support microglial phagocytosis and complement opsonization but others do not. It is also possible that antibodies neutralize toxic aggregates, disrupt toxic oligomers, inhibit their aggregation, or simply cap their growth, and that multiple mechanisms are at play. According to the peripheral sink hypothesis, peripheral antibody binding would enhance Aβ efflux from the brain. This hypothesis has been complicated by recent data suggesting that Aβ binding to antibody in the blood stabilizes Aβ long past its normal half-life, and that peripheral antibody injections need to be given for extended periods of time before they begin to decrease brain Aβ levels. To address this issue, Aβ levels and turnover rates are beginning to be quantified in different body compartments, both in mice and humans.
A novel opportunity in immunotherapy research is emerging around microglia. Independent lines of investigation have reported that the copolymer glatiramer acetate activates microglia to generate an inflammatory state in which they clear amyloid pathology independently of antibodies (Frenkel et al., 2005; Butovsky et al., 2006). Efforts to separate clearance from the attendant inflammation led to intranasal administration of protollin, a proteasome-based mucosal adjuvant that has been tested in humans for shigella and other infections. Protollin alone, without glatiramer acetate, appears to reduce brain Aβ levels in APP-transgenic mice. Soluble and insoluble brain Aβ levels, as well as astrocytosis, decreased in a long-term nasal protollin prevention study. Different time courses of prevention and treatment suggest that protollin might boost a transient microglial activation that removes amyloid and then recedes. Of several proposed microglial amyloid clearance pathways, the one triggered by protollin appears to involve Toll-like receptors. The different possible activation profiles of microglia in human brain need to be characterized. Going forward, the challenge of any deliberate microglial activation lies in restraining it to avoid damage to the surrounding milieu (for a recent review, see Weiner and Frenkel, 2006).
Another novel approach toward generating high-affinity Aβ-binding agents might be to view Aβ aggregates as a thymus-independent type 2 antigen. These atypical antigens have a polymeric, repetitive structure and activate B cells to produce IgG and IgM antibodies without the help of T cells, which are typically needed to complete B cell activation and antibody production. Single-chain variable fragment (scFv) antibodies against amyloid intermediate structures can be generated using ordered, heterologous antigens that have no sequence homology to Aβ, and thus preclude autoimmune reactions against Aβ or APP.
In summary, participants agreed that immunotherapy in mice improves cognition, both through acute amyloid-independent effects and through slower effects secondary to amyloid removal. The questions have shifted from whether Aβ removal will have cognitive benefits to how oligomers relate to plaques in terms of affecting cognition, and whether oligomers are important in human AD. Mouse models differ importantly from AD. Mice develop a detectable cognitive impairment before plaques, and impairment reverses rapidly with passive immunization. Humans remain asymptomatic until after they have developed a nearly full load of plaques and become symptomatic only once neurodegeneration is well underway. Possible explanations include that dementia represents an end-organ failure that APP/PS1 mouse models do not model, that humans compensate better than mice, and that cognitive tests are too crude to pick up subtle deficits in people.
This issue spawned discussion on diagnostic AD testing. Current tests are inadequate, especially the MMSE, which regrettably remains in wide use. The reason has to do with brain reserve. Family members tend to notice a deficit first, and a CDR-based clinical diagnosis supported by informant interviews is generally viewed as more sensitive than neuropsychological measures. Highly educated patients report they cannot do what they used to anymore, but even if they do have early AD, they still score in the ninety-eighth percentile of available cognitive tests. At this early stage, cognitive tests pick up a deficit only by the second or third measurement, when comparison to the person's baseline can be made. The key indicator for the AD diagnosis is the change from baseline, and huge baseline variability has stymied development of more sensitive tests despite intensive efforts in academia and industry. Moreover, the disease moves slowly; it has occurred silently for at least four years by the time the patient or relative suspects a problem.
That said, a person's hippocampus and entorhinal cortex have suffered significant cell loss in particular sub-areas by the time MCI or CDR 0.5 can be diagnosed; hence new cognitive tests that specifically call on those pathways could be explored. Spatial memory tests such as maze navigation offer a promising lead.
Another diagnostic option to explore is a neurotransmitter-based pharmacological stress test. It would use a single dose of, for example, an anti-cholinergic drug. In people who are particularly vulnerable to an added decrement in transmitter level, this could unmask an underlying synaptic dysfunction. In some elderly people, anti-cholinergic drugs prescribed for urinary incontinence are known to produce a cognitive deficit that resolves upon discontinuation. Similar tests are available for diabetes (insulin challenge) and depression (serotonin antagonist challenge).
- Dustin ML, Colman DR. Neural and immunological synaptic relations. Science. 2002 Oct 25;298(5594):785-9. PubMed.
- Syken J, GrandPre T, Kanold PO, Shatz CJ. PirB restricts ocular-dominance plasticity in visual cortex. Science. 2006 Sep 22;313(5794):1795-800. PubMed.
- Seabrook TJ, Thomas K, Jiang L, Bloom J, Spooner E, Maier M, Bitan G, Lemere CA. Dendrimeric Abeta1-15 is an effective immunogen in wildtype and APP-tg mice. Neurobiol Aging. 2007 Jun;28(6):813-23. PubMed.
- Maier M, Seabrook TJ, Lazo ND, Jiang L, Das P, Janus C, Lemere CA. Short amyloid-beta (Abeta) immunogens reduce cerebral Abeta load and learning deficits in an Alzheimer's disease mouse model in the absence of an Abeta-specific cellular immune response. J Neurosci. 2006 May 3;26(18):4717-28. PubMed.
- Frenkel D, Maron R, Burt DS, Weiner HL. Nasal vaccination with a proteosome-based adjuvant and glatiramer acetate clears beta-amyloid in a mouse model of Alzheimer disease. J Clin Invest. 2005 Sep;115(9):2423-33. PubMed.
- Butovsky O, Koronyo-Hamaoui M, Kunis G, Ophir E, Landa G, Cohen H, Schwartz M. Glatiramer acetate fights against Alzheimer's disease by inducing dendritic-like microglia expressing insulin-like growth factor 1. Proc Natl Acad Sci U S A. 2006 Aug 1;103(31):11784-9. PubMed.
No Available Further Reading