|Home | About | Journals | Submit | Contact Us | Français|
We review attempts to treat Alzheimer disease with antibodies that bind amyloid β peptide (Aβ) and the feasibility of developing catalytic antibodies for this purpose. Naturally occurring immunoglobulin M (IgM) class antibodies that hydrolyze Aβ and inhibit Aβ aggregation were identified. The production of these antibodies increases as a function of age, ostensibly reflecting an attempt by the immune system to protect against the deleterious effect of Aβ accumulation in old age. A search for catalytic antibodies in a library of human immunoglobulins variable (IgV) domains yielded catalysts that hydrolyzed Aβ specifically at exceptionally rapid rates. The catalytic IgVs contained the light-chain variable domains within scaffolds that are structurally reminiscent of phylogenetically ancient antibodies. Inclusion of the heavy-chain variable domain in the IgV constructs resulted in reduced catalysis. We present our view that catalytic antibodies are likely to emerge as more efficacious and safer immunotherapy reagents compared to traditional Aβ-binding antibodies.
Alzheimer disease (AD) is the commonest age-induced dementia, with an estimated worldwide prevalence of approximately 26 million humans. Acetylcholinesterase inhibitors provide symptomatic relief for AD. However, no available AD therapy rectifies the dysfunctional processes underlying the disease. Accumulation of amyloid β (Aβ) peptide aggregates is thought to play a central role in the pathogenesis of AD.1 Even physiological aging may be associated with increased Aβ.2 The Aβ aggregates are composed of 39- to 43-residue peptides generated by proteolytic processing of the amyloid precursor protein (APP) by β- and γ-secretases.3 The predominant product of this processing pathway is the 40-amino-acid peptide corresponding to APP residues 597–636 (Aβ40), with the 42-amino-acid peptide corresponding to residues 597–638 (Aβ42) being the next most abundant product. Both peptides form toxic oligomeric aggregates and fibrillar aggregates found in amyloid plaques characteristic of the AD brain.3 Aβ42 tends to aggregate more rapidly and is the majority species in amyloid plaques.3 Aβ40 is the major species found in peripheral blood.4 At very low concentrations, Aβ can exert trophic effects on the cells.5, 6 However, Aβ overproduction occurs due to dysregulated neuronal metabolism, and there is no known physiological advantage of Aβ accumulation in the aged brain. Death of neurons cultured with synthetic Aβ aggregates has been reported.7 Soluble Aβ oligomers can induce neurodegenerative effects by several pathways, including altered expression of memory-related receptors8 and induction of aberrant neuronal responses to electrical stimulation.9 We review here the status of AD treatment with antibodies that bind Aβ and the potential of catalytic antibodies for inducing an improved therapeutic effect.
Removal of Aβ from the brain has been advanced as a potential treatment of AD. Studies in transgenic mice expressing mutant human amyloid precursor protein genes (APP-Tg mice) suggest that Aβ-binding antibodies clear brain Aβ deposits and correct the behavioral deficits evident in this animal model. The favorable effects were observed following peripheral administration of Aβ-binding monoclonal antibodies10,11 (passive immunotherapy) and after active immunization with Aβ itself (active immunotherapy), which induces the synthesis of Aβ-binding antibodies.12–14 The effects were evident when the antibodies were administered both prior to11 and after10 the appearance of Aβ plaques in the murine brain. These findings lead to clinical trials of active Aβ immunotherapy as a treatment for AD. Two important points emerged from the human trials.15 First, only about 20% of the recipients developed Aβ-binding antibodies, reflecting the limited immunogenicity of the Aβ vaccine formulation. Second, the trials were suspended because ≈5% of the immunized patients developed sterile meningoencephalitis, suggesting an inflammatory reaction. Patients who produced Aβ-binding antibodies displayed reduced decline of certain cognitive functions,16 but the therapeutic benefit has been subject to debate.17
Antibodies with Aβ-binding activity can cause undesirable side effects,18 and there is also the potential of harmful cell-mediated immunity after immunization with Aβ. The latter concern is eliminated if preformed Aβ-binding antibodies are employed for passive immunotherapy. A Phase II trial of Bapineuzumab, a humanized reversibly binding monoclonal Aβ-binding immunoglobulin G (IgG), administered intravenously to mild-to-moderate AD patients has been conducted.19 There was no indication of unacceptable inflammatory reactions, but a dose-limiting incidence of vasogenic edema was evident. This effect may be due to microbleeds caused by deposition of immune complexes in cerebral blood vessels (Fig. 1). AD patients homozygous for the apolipoprotein E4 allele are predisposed to increased Aβ accumulation and early development of AD.20 Administration of the Aβ-binding antibody to patients who were not homozygous for the apolipoprotein E4 allele resulted in significantly reduced cognitive decline and a lesser tendency toward vasogenic edema. The favorable effect on cognition did not reach statistical significance in the intent-to-treat population. From these results, there is cautious support for passive AD immunotherapy using Aβ-binding antibodies, but the safety and efficacy of the procedure leave room for improvement.
Eli Lilly has advanced its own humanized monoclonal IgG into Phase III trials. This IgG binds an epitope located in the middle region of Aβ, whereas Bapineuzumab binds the Aβ amino terminus.21 Detailed information about the beneficial and side effects of the Lilly monoclonal IgG has not been released, but the company has indicated that antibody administration induces an increase of Aβ levels in cerebrospinal fluid and peripheral blood.22 To the extent that the observed increase of peripheral Aβ is due to release from the brain peptide stores, the antibody may exert a favorable effect.
Interestingly, healthy humans and AD patients produce Aβ-binding autoantibodies spontaneously.23 Pooled human IgG marketed as intravenously administered immunoglobulins (IVIG) formulations contains small amounts of Aβ-binding autoantibodies. Early-stage clinical trials entailing intravenous administration of very large IVIG doses to AD patients (e.g., 1.2grams/kg over 3 days) were encouraging, but no definitive evidence for efficacy is available yet.24 No side effects have been reported.
Aβ-binding IgGs are proposed to reduce Aβ deposition in the brain by the following mechanisms (Fig. 1)25,26: (1) Small amounts of peripherally administered IgGs may cross the blood-brain barrier (BBB; ≈0.1% of injected IgG dose) and bind Aβ in the brain. Microglial cells then ingest the immune complexes via an Fcγ-receptor mediated process that results in Aβ clearance; (2) the IgGs can also bind the neonatal Fc receptor (FcRn) located on the abluminal (brain) side of the endothelial cells constituting the BBB, thereby facilitating Aβ efflux into the periphery; (3) Aβ binding to IgG may constrain the peptide into a nonaggregable conformation; and (4) according to the “peripheral sink” hypothesis,21 Aβ is cleared from the brain without IgG entry into the brain. In this hypothesis, Aβ binding by antibodies in peripheral blood perturbs the equilibrium between the peptide pools in the brain and periphery, thereby inducing Aβ release from the brain. In principle, these mechanisms are not mutually exclusive and may be triggered by the same antibody.
The feasibility of developing catalytic antibodies for immunotherapy is based on the following immunological and biochemical findings:
Michael Sierks at the University of Arizona observed that two recombinant antibody light-chain subunits with VIP-hydrolyzing activity also displayed cross-reactive Aβ-hydrolyzing activity.44 This prompted us to screen monoclonal and polyclonal preparations of intact antibodies for this activity. Of 10 monoclonal IgM antibodies from patients with Waldenström macroglobulinemia with promiscuous proteolytic activity, two also hydrolyzed Aβ40 and Aβ42.45 The catalytic IgMs did not bind Aβ40 according to enzyme-linked immunosorbent assay (ELISA) tests. By mass spectrometry of the product peptides, the Lys28–Gly29 and Lys16–Leu17 bonds were identified, respectively, as the major and minor hydrolysis sites. Active site titration using an electrophilic phosphonate diester designed to bind the catalytic site of serine protease enzymes irreversibly indicated the presence of 10.2 catalytic sites/IgM molecule, compared to the theoretical value of 10 antigen-combining sites.45 The catalytic activity was retained in the Fab fragments of the IgM, and the activity was maintained at constant levels following successive purification steps. These observations indicated that Aβ hydrolysis is attributable to the IgM.
To assess the potential of physiological Aβ clearance by catalytic antibodies, we studied polyclonal antibody preparations purified from human sera.45 Most IgM preparations from nondemented elderly humans displayed detectable Aβ40 hydrolytic activity. The magnitude of Aβ hydrolysis was variable over two log orders in a group of 25 study subjects, suggesting a polymorphic distribution of the catalytic antibodies. IgG antibodies displayed lesser levels of activity. Antibodies from comparatively young blood donors (<35 years) displayed significantly lower levels of Aβ-hydrolytic activity compared to the elderly donors (>70 years), indicating an aging-induced synthesis of the catalytic antibodies. No hydrolysis of irrelevant polypeptides by the IgMs was evident. It may be concluded that Aβ40-specific IgMs are responsible for the catalytic activity.
The concentrations of Aβ40 and Aβ42 in blood are ≈0.2 and ≈0.05ng/mL, respectively.4 If IgM present in peripheral blood is assumed to express catalytic activity in vivo at the average observed level, >90% of blood-borne Aβ40 will be hydrolyzed after 5 days (corresponding to the half-life of IgM in blood; see ref. 45 for details). At concentrations of the IgM displaying readily detectable Aβ hydrolysis, no Aβ binding was evident. This suggest that the IgM autoantibodies with catalytic activity hold superior potential for inactivating Aβ compared to any Aβ-binding IgMs present in the blood. Peripheral and brain Aβ peptide levels are maintained in a state of equilibrium by means of Aβ transport mechanisms operative at the BBB (Fig 1). Catalytic antibodies can be expected to clear large amounts of peripheral Aβ. According to the “peripheral drainage hypothesis,” autoantibodies that catalyze hydrolysis of peripheral Aβ may induce depletion of the brain peptide stores.
IgM-class antibodies are large molecules and cross the BBB only in small amounts. However, the brain is an immunocompetent organ, and B cells within the brain hold the potential of producing catalytic antibodies. IgM concentrations in the cerebrospinal fluid (CSF) of nondemented humans and AD patients are in the μg/mL range.46 If the IgMs found in the brain express sufficient catalytic activity, they can directly hydrolyze and clear Aβ plaques and toxic oligomers within the brain. The catalytic IgMs described above protected a neuronal cell line from the toxic effect of oligomeric Aβ. Formation of Aβ oligomers and fibrillar structures was inhibited potently by a catalytic IgM antibody.45 If the catalytic IgMs are present in the brain in sufficient amounts, therefore, they may play a physiological role in clearing Aβ and alleviating its harmful effects on neuronal function.
We interpret the aging-induced production of catalytic antibodies to Aβ as a compensatory effort mounted by the immune system that counters the pathology associated with Aβ accumulation in the brain. Understanding homeostatic processes has historically provided valuable clues to development of engineered drugs for intractable diseases. The following section describes our attempt to identify catalytic antibodies for therapy of AD.
Polyclonal antibody preparations are mixtures of individual antibodies with diverse levels of catalytic activity. The expressed antibody repertoire in humans can be as large as 1010–11 molecules, and modern screening and selection techniques permit identification of rare antibodies from libraries consisting of human immunoglobulin V domains (IgVs). We searched for Aβ-hydrolyzing recombinant IgVs in a library composed of ≈107 clones from humans with systemic lupus erythematosus, an autoimmune disease associated with enhanced production of catalytic antibodies.47 The majority of the clones in the library are single chain Fv (scFv) molecules with a scaffold that mimics the V domain organization of physiological human antibodies responsible for specific antigen recognition (VL-Li-VH-t, where Li denotes a 16-residue linker peptide that joins the VL domain carboxyl terminus to the VH domain amino terminus, and t denotes a 26-residue carboxy-terminal peptide containing the c-myc peptide and His6 tags that facilitate identification and purification of the scFv molecules).35 A minority of clones possess unusual IgV structures generated by DNA manipulation errors that inevitably accompany repeated nucleic acid replication and cloning cycles over the course of library construction.
By random screening and covalent phage-IgV selection using an Aβ40 analog containing electrophilic phosphonates at the side chains of Lys16 and Lys25 (Bt-E-Aβ40; Fig. 2A), we isolated rare IgV clones capable of hydrolyzing Aβ40 rapidly (Fig. 2B). To rule out sample preparation artifacts, the phagemid DNA encoding one of the high-activity IgV clones (2E6) was prepared and re-expressed in 15 individual bacterial colonies. IgV 2E6 from all 15 individual colonies expressed robust Aβ40-hydrolyzing activity. No activity was observed in empty vector control extract devoid of an IgV insert.
To determine their structure, the cDNAs of four IgVs distinguished by high-level Aβ40-hydrolyzing activity were sequenced. None of the catalytic IgVs contains the archetypal VL–VH paired structure of physiological Igs. IgV 2E6 is a heterodimer of two different VL domains with the intervening linker peptide and the expected carboxy-terminal tag (designated IgVL2-t; Fig. 2C). The remaining catalytic IgVs are single-domain VL clones with unexpected carboxy-terminal polypeptide segments (designated IgVL-t′ clones). The carboxy-terminal polypeptide region t′ consists of (1) the expected linker peptide, (2) a 15- to 28-residue aberrant peptide in place of the customary VH domain composed of ≈115 residues, and (3) the expected 26-residue peptide containing c-myc and His6 tags (Fig. 2C). Sequencing of 26 randomly picked clones from the library indicated that only a minority are IgVL2–t (12.5%) or IgVL–t′ (4.2%) structures. The structure of the IgVs was also confirmed by mass determination of the highly purified proteins, tryptic digestion and mass spectroscopic analysis of the peptides.47
The cumulative probability of identifying four high-activity Aβ40-hydrolyzing clones from the library with the rare IgV structures in Fig. 2C by random chance alone is very small (p=0.92×10−5). We generated four repaired versions of a highly catalytic IgVL–t′clone (clone 5D3) by insertion of the missing VH residues 8–115 into the molecule (Kabat numbering). This yielded scFv clones mimicking the combining sites of physiological Igs. All of the scFv mimetics expressed Aβ40-hydrolyzing activity lower than the parent IgVL–t′ by ≈2 orders of magnitude. Partnering of the single VL domain of the IgVL–t′ with full-length VH domains, therefore, suppresses the catalytic activity. The rate of Aβ hydrolysis by the IgVs was 3–4 orders of magnitude greater than the naturally occurring IgMs described in the preceding section. Taken together, these observations indicate that the nonphysiological IgV clones identified serendipitously in our studies can express Aβ-hydrolyzing activity beyond the range observed for natural Igs.
Mass spectrometry of the Aβ fragments produced by the catalytic IgVs indicated that the major hydrolysis site was the His14–Gln15 bond, with smaller levels of hydrolysis occurring at the Gln15–Lys16, Phe20–Ala21, Lys27–Gly28, and Gly28–Ala29 bonds (Fig. 3A).47 The IgVs did not hydrolyze irrelevant proteins and model protease substrates customarily employed to monitor catalyst specificity, including proteins containing the His–Gln bond. There is no evidence, therefore, for IgV-catalyzed hydrolysis of proteins other than Aβ, suggesting the feasibility of specific Aβ clearance with little or no damage to other proteins. Synthetic Aβ fragments corresponding to residues 29–40 inhibited 2E6-catalyzed 125I-Aβ40 hydrolysis as potently as full-length Aβ40, while Aβ1–17 encompassing the major His14–Gln15 scissile bond did not. This indicates that IgV specificity derives mainly from noncovalent recognition of the carboxy-terminal Aβ region remote from the major scissile bond. Recognition of His14–Gln15 dipeptide unit by the IgV in the transition state of the reaction is necessary for catalytic acceleration of peptide bond hydrolysis. However, as most other scissile bonds of Aβ do not contain His and Gln residues, the recognition of unique amino acid side chains at the peptide bond transition state must be a lesser contributory factor.
Consistent with the serine protease-like mechanism of previously described catalytic antibodies, the IgVs were bound irreversibly by the biotinylated electrophilic phosphonate diester Bt-E-hapten-1 (Fig. 3B). This probe reacts covalently with the nucleophilic antibody sites responsible for attacking the electrophilic carbonyl of peptide bonds in the initial step of proteolysis.35 The poorly electrophilic control hapten-2 did not form detectable covalent adducts with the IgVs. E-hapten-1 also inhibited 125I-Aβ40 hydrolysis by the IgVs (Fig. 3B). A molecular model of catalytic IgVL2–t 2E6 identified the potential catalytic site composed of the Ser76–Arg18–Asp17 residues47 (Fig. 3C).
The limited efficacy and side effects of a reversibly binding anti-Aβ monoclonal IgG in patients with mild-to-moderate AD has highlighted the importance of searching for safer and more effective immunotherapeutic reagents.19 Large quantities of stochiometrically binding monoclonal IgGs are usually required for immunotherapy. Catalytic Igs hold the potential of clearing Aβ efficiently by virtue of the specific Aβ-degrading activity. For example, from its turnover number determined at excess Aβ concentration, a single IgVL–t′ 5D3 molecule will degrade 4320Aβ molecules in 3 days.47 The rate of catalysis is comparable to that of neprilysin, an enzyme that has received attention as a potential Aβ-clearing drug.48 In unpublished studies, we observed that minimizing conformational perturbations of the IgVs during purification improves the catalysis rate by another order of magnitude. The IgVs are also unique by virtue of their specificity for Aβ. Neprilysin and other Aβ-hydrolyzing enzymes, in contrast, hydrolyze irrelevant polypeptides.
Aβ-binding IgGs are described to induce inflammation and vascular microhemorrhages.25,26 The underlying mechanisms are, respectively, immune complex-stimulated release of microglial inflammatory mediators and IgG-stimulated Aβ deposition in cerebral blood vessels. The IgVs failed to bind immobilized Aβ40 in ELISA tests, indicating their inability to form stable immune complexes. Moreover, they degrade Aβ permanently, minimizing the risk of Aβ redeposition in the vascular walls. Taken together, the observed properties of the IgVs support the idea that they can be developed to specifically and efficiently clear Aβ with minimal side effects.
Modern adaptive immunity is characterized by high-affinity antibody binding to antigens. The VL and VH domains of higher organisms express substantial sequence identity with each other. They have likely evolved by duplication and sequence diversification of a common primordial gene encoding the Ig fold. The structure of the best antibody catalysts is reminiscent of primordial antibodies. Jawed fish contain the earliest known immune system.49 Their immune system produces a functional homolog of the single-domain catalytic IgVs, i.e., a single VH domain thought to bind antigen in its unpaired state. Jawed fish also produce IgMs, but not other classes of antibodies. In higher organisms, IgMs are the first class of antibodies produced by B cells, with other antibody classes appearing at later stages of the adaptive immune response. IgMs from humans and mice consistently express catalytic activities superior to class-switched IgG antibodies.50 These arguments suggest that catalysis is a primitive function that may have developed prior to the evolution of binding activity (Fig. 4). For the antibody engineer, these ideas present novel opportunities, because primordial antibodies may be a rich source of efficient catalysts.
This work was supported by the U.S. National Institutes of Health (1R01AG025304). We thank coauthors listed in our previous publications for their collaborations.