Monoclonal antibodies that target pathological assemblies of Aβ

Abstract

Amyloid beta (Aβ) immunotherapy for Alzheimer's disease has shown initial success in mouse models of Alzheimer's disease and in human patients. However, because of meningoencephalitis in clinical trials of active vaccination, approaches using therapeutic antibodies may be preferred. As a novel antigen to generate monoclonal antibodies, the current study has used Aβ oligomers (amyloid β-derived diffusible ligands, ADDLs), pathological assemblies known to accumulate in Alzheimer's disease brain. Clones were selected for the ability to discriminate Alzheimer's disease from control brains in extracts and tissue sections. These antibodies recognized Aβ oligomers and fibrils but not the physiologically prevalent Aβ monomer. Discrimination derived from an epitope found in assemblies of Aβ1–28 and ADDLs but not in other sequences, including Aβ1–40. Immunoneutralization experiments showed that toxicity and attachment of ADDLs to synapses in culture could be prevented. ADDL-induced reactive oxygen species (ROS) generation was also inhibited, establishing this response to be oligomer-dependent. Inhibition occurred whether ADDLs were prepared in vitro or obtained from Alzheimer's disease brain. As conformationally sensitive monoclonal antibodies that selectively immunoneutralize binding and function of pathological Aβ assemblies, these antibodies provide tools by which pathological Aβ assemblies from Alzheimer's disease brain might be isolated and evaluated, as well as offering a valuable prototype for new antibodies useful for Alzheimer's disease therapeutics.

Abbreviations used

Aβ

amyloid beta peptide

ADDL

Aβ-derived, diffusible ligand

DAB

3,3′ diamino-benzidine

DCF

dichlorofluorescein

DIV

days in vitro

GFAP

antiglial fibrillar acidic protein

LTP

long-term potentiation

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

NFDM

non-fat dry milk

PBS

phosphate-buffered saline

ROS

reactive oxygen species

tg

transgenic

SDS-PAGE

sodium dodecylsulfate–polyacrylamide gel electrophoresis

Immunization of transgenic (tg) mice models of Alzheimer's disease with fibrillar amyloid beta protein (Aβ) results in reduction of Aβ deposits in the brain and prevents the accumulation of this pathology when administered before its formation (Schenk et al. 1999; Schenk 2002). Learning and memory deficits produced in these mice are also reduced or prevented by similar active vaccination with preparations containing fibrillar Aβ (Janus et al. 2000; Morgan et al. 2000). Based on results from animal models, clinical trials were initiated and showed few adverse reactions in Phase 1. However, Phase 2 trials were halted when 6% of the patients developed meningoencephalitis (Birmingham and Frantz 2002; Schenk 2002; Hock et al. 2003; Orgogozo et al. 2003; Schenk et al. 2004). Reports of the clinical outcome of these trials revealed that after 1 year patients producing antibodies that targeted plaques had a slower rate of cognitive decline than those patients who did not produce antibodies (Hock et al. 2003). Postmortem results on three patients showed absent or sparse plaques in the neocortex, with reactive microglia suggesting an effective immune response (Nicoll et al. 2003; Ferrer et al. 2004; Masliah et al. 2005).

Alternative approaches to avoid inflammatory responses through the use of therapeutic antibodies are now under development (Gelinas et al. 2004; Morgan and Gitter 2004; Schenk et al. 2004; Agadjanyan et al. 2005). Most strikingly, it has been established that injections with Aβ-generated monoclonal antibodies produce cognitive improvement in tg mice models of Alzheimer's disease. Using an antibody whose epitope targets the center of the Aβ peptide, Dodart et al. (2002) showed that memory deficits can be reversed in PDAPP mice within 24 h after treatment. Reversal of memory deficits by monoclonal antibody injection has been confirmed independently in other laboratories (Kotilinek et al. 2002; Lee et al. 2006). Although passive vaccination previously had been shown to clear plaques from PDAPP and other tg mice models (Bard et al. 2000,2003; Bacskai et al. 2002; McLaurin et al. 2002), in some studies showing recovery from memory deficits, Aβ plaque burden was not decreased. A likely explanation for cognitive improvement without change in plaque burden is that these therapeutic antibodies immunoneutralize small, soluble oligomers of Aβ that have been implicated in Alzheimer's disease synapse failure (Lambert et al. 1998; Hardy and Selkoe 2002). Aβ oligomers form at low doses of Aβ1–42, block long-term potentiation (LTP), and specifically attach to synaptic terminals (Lambert et al. 1998; Wang et al. 2002; Lacor et al. 2004; Wang et al. 2004). These oligomers (referred to as amyloid β-derived diffusible ligands, ADDLs) are elevated in Alzheimer's disease brain and CSF and in tg mouse models (Chang et al. 2003; Gong et al. 2003; Georganopoulou et al. 2005). It previously had been predicted that if ADDLs were an important part of early Alzheimer's disease pathogenesis, then memory deficits might prove to be reversible (Lambert et al. 1998; Klein 2000), as subsequently proven by Dodart et al. (2002).

Given these considerations, oligomers may provide an optimum target for therapeutic antibodies. The current work therefore addresses the need to obtain monoclonal antibodies selective for oligomers. Our approach has used as antigen soluble Aβ oligomers because of their previously demonstrated utility in providing epitopes dependent on quaternary structure in the generation of polyclonal antibodies. This strategy has generated monoclonal antibodies that distinguish between Alzheimer's disease and control brains and that neutralize oligomers in solution, characteristics that will ultimately be essential for therapeutically useful antibodies.

Materials and methods

Monoclonal antibody development

Immunization and fusion for this project were done at the Northwestern University Core Antibody Facility and at Immuno-Precise Antibodies, Ltd, Victoria, B.C., Canada. Growth, screening, and subcloning (when necessary) were performed in the Klein laboratory.

At Northwestern, ADDLs in F12 medium, prepared from Aβ1–42 as previously described (Lambert et al. 1998; Klein 2002), were mixed 1 : 1 with complete Freund's adjuvant (first and second vaccination) or incomplete Freund's adjuvant (all subsequent vaccinations) and injected subcutaneously (first two vaccinations) or intraperitoneally into three mice in a total volume of ∼ 1 mL/mouse. Each injection consisted of ADDLs equivalent to 194 ± 25 μg total protein. Mice were injected approximately every 3 weeks. After six injections, one mouse died and its spleen was frozen for future projects. The spleen from the mouse with the highest titer serum was then fused with SP2.0 myeloma cells and plated out into six 96-well plates. The last mouse was injected a seventh time 2 months later and fused as before. These hybridoma cells were plated into 20 96-well plates and grown at 37°C, 5% CO₂. Growth medium was Dulbecco's minimal essential medium/F12, supplemented with HAT (1×, Sigma, St. Louis, MO, USA), glutamine (2 mm), hybridoma enhancing supplement (10%, Sigma), OPI (oxaloacetate, purine, and insulin, 0.1%, Sigma), phorbol monoamine acetate, 0.6 μg/L, (Sigma), fetal calf serum (0.15%), and Fetal Clone II serum (15%, HyClone). The hybridoma plates were transferred to the Klein lab, where the culture medium supernates were screened as below. Selected hybridomas were expanded and frozen for later subcloning. Two hybridomas were subcloned without freezing.

The screening process eventually employed five assays: a dot immunoblot and western immunoblot, described previously (Lambert et al. 2001), and native immunoblot, described below, using synthetic ADDLs, and a dot immunoblot and Western blot using endogenous fibrils obtained from human tissue, described below. These assays tested the binding of antibodies to ADDLs (dot immunoblot) and affinity to specific oligomer species (Western). Supernatants were initially tested by dot immunoblot using 5 pmol ADDLs or 22.5 ng/dot (576 supernates in the first fusion and 1920 supernates in the second). Those clones that tested positive were further screened using Western blot at 10–20 pmol ADDLs/lane. The screen was repeated to identify low positives or false positives. A total of 45 clones were expanded and then frozen or subcloned.

Immuno-Precise Antibodies, Ltd. was employed to produce clones using their proprietary Rapid-Prime method. The animals were immunized with ∼ 0.25–0.35 mL ADDL5 over a period of 2 weeks at an average of 0.505 mg/mL total protein. Clone supernates (288/project) were then sent to the Klein laboratory for screening as above.

Preparation of synthetic and endogenous ADDLs and fibrils

Synthetic ADDLs were prepared in F12 as described previously (Lambert et al. 2001; Klein 2002). Synthetic fibrils were prepared as described (Stine et al. 2003). Endogenous ADDLs were extracted from Alzheimer's disease brain tissue as described (Gong et al. 2003) omitting the concentration step. Endogenous fibrils were obtained using the same method of preparation, except the pellet was then incubated for 1 h with 2% sodium dodecyl sulfate (SDS) in F12 and centrifuged at 220 000 × g for 1 h. The pellet, which contained fibrils, was then suspended in F12 and sonicated for 30–60 s to suspend material for determination of protein concentration.

Dot immunoblot and peptide competition

Initial screening by dot blot was performed as described previously (Lambert et al. 2001) applying either ADDLs (22.5 ng = 5 pmol/dot) or fibrils (1 μg/dot) to the nitrocellulose. For competition dot blots, ADDLs were applied to dry nitrocellulose in duplicate at various pmolar concentrations in 0.5 μL volume using a template derived from the SurfBlot apparatus. Blots were then dried for 15 min, blocked for 1 h, and inserted into the SurfBlot apparatus. Antibody plus or minus peptide was incubated in the wells for 1.5 h, removed from the dot blot apparatus, the wells washed with blocking buffer, and the membrane removed from the apparatus. The nitrocellulose was then washed, treated with secondary antibody, and visualized as cited below.

Immunoblotting

Sodium dodecyl sulfate polyacrylamide gel electrophoresis [SDS-polyacrylamide gel electrophoresis (PAGE)] was performed as previously described (Lambert et al. 2001) except that 10–20% Tris-Tricine gels (Bio-Rad, Hercules, CA, USA) were used and the separation was performed at 120 V. The secondary antibody was routinely used at 1 : 40 000 dilution.

For initial screening, a preparative 10–20% gel was used with 2.7 μg ADDLs in the sample well, which is equivalent to ∼ 16–20 pmol/lane. Electrophoresis and transfer were as above. Using the tracking dye as a guide, the blocked nitrocellulose was placed into the Surfblot apparatus and 200 μL of hybridoma supernate mixed with blocking buffer [5% non-fat dry milk (NFDM) in TBST – 20 mm Tris, 0.2 m NaCl, pH 7.6, with 0.1% Tween 20] were added to each of 20–21 wells. After incubation at room temperature (22–24°C) for 1.5 h, the supernates were removed and the wells washed with buffer. The membrane was then removed from the Surfblot and washed 3 × 15 min in TBST. The secondary antibody was then incubated with the membrane for 1 h at room temperature. After washing, the oligomers were visualized with half strength SuperSignal (Pierce, Rockford, IL, USA). The Western immunoblot using human fibrils was performed in the same manner using approximately 64 μg of prepared material for each gel.

Native PAGE was performed as described (Chromy et al. 2003) at 120 V.

Immunocytochemistry

Immunocytochemistry was performed as described (Lambert et al. 2001; Lacor et al. 2004), except cultures were 21 days in vitro (DIV) and AlexaFluor 488 (Molecular Probes, Eugene, OR, USA) was used to visualize mouse or rabbit IgG. In the blocking experiment, antibodies and ADDLs were preincubated for 1 h at room temperature, using a molar ratio of 4 : 1 antibody : ADDL (calculated using monomer molecular weight), before application to the hippocampal cell cultures.

Immunohistochemistry

Immunohistochemistry was performed as described (Lacor et al. 2004). Briefly, lightly fixed frozen sections (10% formalin at 4°C for 30 h, then cryoprotected in sucrose, 40 μm) from Alzheimer's disease and control middle frontal gyrus were incubated with monoclonal antibodies (1 : 1000) overnight at 4°C and then incubated with HRP-conjugated secondary antibody (1 : 500) for 2 h at room temperature. Staining was revealed with the Vector ABC peroxidase standard kit and DAB (0.05% or 1 : 500 in Tris buffer for 2 min). Sections were counterstained with hematoxylin, mounted with Permaslip, and imaged on a Nikon Eclipse E600 light microscope with a Spot Insight digital video camera (v. 3.2). For fluorescent imaging, sections were blocked with 5% horse serum in TBST for 40 min, followed by 1% NFDM in TBST (blocking buffer) for 30 min, then incubated with NU-1, -2, or -4 (1 : 1000 in blocking solution) overnight at 4°C, washed, and then incubated with AlexaFluor488-conjugated anti-mouse IgG (1 : 500) for 2 h at room temperature. Imaging was done on a Leica confocal microscope. For NU-6, sections were blocked with 10% normal goat serum + 0.25% Triton in phosphate-buffered saline (PBS) and then the hybridoma supernatant containing the IgM was incubated directly with the section overnight at 4°C. AlexaFluor488 anti-IgM was used as secondary antibody. Double label experiments used a rabbit polyclonal anti-glial fibrillar acidic protein (GFAP) antibody (Chemicon) and NU-6. Sections were mounted using ProLong antifade medium.

Size exclusion chromatography

Size exclusion chromatography (SEC) was performed as described (Chromy et al. 2003), utilizing a Superdex 75 HR 10/30 column equilibrated with F12 (0.45 mL/min). Absorbance was monitored at 280 and 215 nm and 0.3 mL fractions were collected.

MTT assay

Primary hippocampal cells were isolated and grown as described previously (Lambert et al. 2001). For the assay, 15 000 cells/well were plated on poly-l-lysine coated 96-well plates. Cultures were used at 11–12 days in vitro. ADDLs, plus/minus antibody, were diluted in conditioned medium (from previous feedings diluted 1 : 1 with fresh medium) and preincubated for 1 h before addition to cultures. After overnight incubation, the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay (Roche Cell Proliferation Kit, Roche, Mannheim, Germany) was performed according to the manufacturer's directions.

Measurement of reactive oxygen species (ROS)

Primary hippocampal cultures were prepared and grown as described previously (Lambert et al. 2001). After 19 DIV, the cultures were incubated with vehicle or with 500 nm ADDLs with or without 500 nm antibody. The antibody, when present, was added just prior to addition of ADDLs. After 3 h, the ROS probe CMH2-DCFDA [5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetyl ester, Molecular Probes] was added at 2 μm for 40 min at 37°C. Cells were rinsed three times with warm Neurobasal medium and then maintained in Neurobasal medium without phenol red during observation, which was carried out immediately. Randomly chosen fields were examined on a Nikon Eclipse TE 2000-U fluorescence microscope and images were digitally acquired using metamorph software (Meta Image Series, Universal Imaging Corporation, West Chester, PA, USA). All images were acquired using a fixed exposure time. Nine independent fields were analyzed for each experimental condition, which were carried out in triplicate. Quantitative analysis of the dichlorofluorescein (DCF) fluorescence data was carried out by histogram analysis of the fluorescence intensity at each pixel across the images using Image J [NIH; Windows version (Abramoff et al. 2004)] and normalized for cell number.

Results

Oligomer-specific monoclonal screening strategy

To identify potentially useful clones, supernatants were screened initially by dot immunoblots using 5 pmol of synthetic ADDLs or 1 μg Alzheimer's disease brain pellets enriched in fibrils (Fig. 1). Positives from this screen (∼ 30%) were then tested by Western immunoblots to further define binding species. Of the antibodies tested by western blots, ∼2% were found to bind oligomers and not monomer at low ADDL concentrations, a relatively high yield. Various interesting properties were noted. For example, one clone, which ultimately became senescent, was found to produce antibodies that bound oligomers but not fibrils (NU-7). Two clones were found that produced antibodies that bound fibrils but not oligomers (NU-6, NU-9). Antibodies from one clone (NU-10) bound the higher multiple oligomers (12–24mer) better than the trimer/tetramer species. One antibody (NU-8) was found that bound ADDLs only under native conditions. Overall, 38 clones were saved for future expansion and use, while the properties of seven were investigated in the current study.

Figure 1.  Schematic representation of screening process used to develop oligomer-selective monoclonal antibodies against Aβ1–42. As detailed in Materials and methods, mice were inoculated with ADDLs (∼194 μg/injection/mouse) every 3 weeks for six inoculations. Hybridomas produced from the fusion of these mouse spleens with SP2 cells were plated in 96-well plates and screened in two types of dot blots as shown. Promising hybridomas were then validated using various assays, e.g. Western blots, dot blots using soluble extracts from Alzheimer's disease brain (containing endogenous ADDLs), and immunocyto/histochemistry. Positive hybridomas were then subcloned by limiting dilution. Selected antibodies were collected from culture medium and further purified using Protein G Sepharose. AD, Alzheimer's disease.

Monoclonals specifically distinguish Alzheimer's disease from control brain

An essential test was to determine whether the antibodies could distinguish between soluble extracts of human Alzheimer's disease brain (containing endogenous ADDLs) and extracts of control brain. Synthetic ADDLs (1 pmol) and three human brain extracts (0.5 μg, staged for disease severity) were assayed by dot immunoblot. A representative result is shown in Fig. 2. 6E10 (a commercial antibody with epitope within Aβ3–8), NU-2, and NU-3 showed weak binding to Alzheimer's disease brain extract but still discriminated between Alzheimer's disease and control brain. However, two monoclonals, NU-1 and NU-4, along with polyclonal M71/2, distinguished between Alzheimer's disease and control brain with a high degree of specificity.

Figure 2.  Selected antibodies discriminate between Alzheimer's disease and control brain tissue. Synthetic ADDLs (1 pmol) and human brain extracts in F12 (from human Alzheimer's disease and control brain; 0.5 μg, see Materials and methods) were spotted onto nitrocellulose in duplicate. After peroxidase treatment (3% H₂O₂, 20 min) and blocking, each vertical lane was then probed with the indicated monoclonals or rabbit polyclonal (M71/2) for 1 h at room temperature. After washing, the nitrocellulose was incubated with the appropriate HRP-conjugated secondary antibody; bound secondary was then visualized with chemiluminescence. Note that NU-1, NU-4, and M71/2 bind strongly to Alzheimer's disease brain samples but not to control brain samples.

Monoclonal immunoreactivity in Alzheimer's disease brain

NU-1, NU-2, NU-4, and NU-6 next were tested for their ability to detect antigen in human brain sections. Fixed Alzheimer's disease and control brain sections were exposed to antibody and visualized with a fluorescent or HRP-conjugated secondary and counterstained with hematoxylin. All four antibodies showed strong immunoreactivity in Alzheimer's disease brain with little to no staining in control brain (Fig. 3a). With NU-1 and NU-4, the labeling was varied, consisting of diffuse and dense-core deposits as well as vascular elements. As shown, staining density increased with severity of disease. Immunoreactivity of NU-1 and NU-4 was prominent in the hippocampus, entorhinal cortex, and middle frontal gyrus of Alzheimer's disease brain (data not shown). Most intriguingly, with these antibodies a pattern of diffuse labeling was found surrounding individual cells and cell clusters (pericellular staining, Fig. 3b), confirming results previously seen with polyclonal antibodies (Lacor et al. 2004). This pattern was readily evident following analysis by confocal microscopy. When the stacked image of an apparent diffuse plaque (3c, left) was examined as individual Z-plane images (3c, right), immunoreactivity was found to be surrounding the cell or cell cluster but not intracellularly. The pericellular labeling resembles binding of oligomers to dendritic sites seen in culture (Lacor et al. 2004).

Figure 3.  Selected antibodies identify Aβ oligomers in Alzheimer's disease brain; some labeling is pericellular. (a) Middle frontal gyrus sections (40 μm) prepared from control (CERAD 0, Braak 0–I), mild Alzheimer's (CERAD A-B, Braak I–III), and severe Alzheimer's (CERAD B-C, Braak V–VI) brain were incubated with the monoclonals NU-1, NU-4, and NU-6 overnight at 4°C. Bound antibodies were then visualized with anti-mouse secondaries conjugated to either HRP or AlexaFluor. Samples revealed with DAB were also counterstained with hematoxylin. Immunoreactivity is not present in control brain, is found in variable amounts in mild Alzheimer's disease cases, and is found extensively in severe Alzheimer's disease brain (20×). (b) Higher magnification (100×) shows a distinct pericellular pattern. (c) Two areas of diffuse NU-4 staining in a mild Alzheimer's disease brain section were examined by confocal analysis. On the left, the stacked image suggests typical diffuse deposits are stained by NU-4, whereas on the right, examination of a single Z-plane shows that immunoreactivity instead surrounds individual cell bodies. Arrows indicate cell bodies. (d) A section from severe Alzheimer's disease brain was incubated with monoclonal NU-6 and polyclonal GFAP overnight at 4°C and then visualized with AlexaFluor 488-conjugated anti-mouse IgM (left) and AlexaFluor 555-conjugated anti-rabbit (middle), respectively. The overlap is shown in yellow. Clearly, NU-6 staining is present in astrocytes.

Staining with NU-6, which bound only aggregate material in immunoblots (not shown), had a significantly different pattern, with immunoreactivity found in cells of frontal cortex sections that presented morphology compatible with reactive astrocytes (Fig. 3a). Labeling of the entire cell (cell body and processes) was evident in severe Alzheimer's disease cases whereas in mild Alzheimer's disease cases only the cell body was labeled; no labeling was seen in non-demented control cases, nor were amyloid plaques labeled in any sections. To verify that NU-6 labeling was astrocytic, brain sections from a severe Alzheimer's disease case were double labeled with NU-6 and a polyclonal antibody against GFAP (Fig. 3d). As shown, cells positive for GFAP also were labeled with NU-6, confirming that NU-6 immunoreactivity is found in reactive astrocytes.

Specificity of anti-ADDL antibodies: Western and dot immunoblots

We then tested the ability of NU-1, NU-2, and NU-4 to bind Aβ1–40, ADDLs, and ADDL fractions separated by size exclusion chromatography (SEC) using a Superdex 75 column. As previously reported, the SEC provided two major fractions, Peak 1, which elutes shortly after the void volume, and Peak 2, which elutes near a 13-kDa marker (Chromy et al. 2003; Lacor et al. 2004). For comparison we also tested the binding of 6E10. SDS-PAGE was followed by silver stain (not shown) or immunoblotting (Fig. 4a). Only 6E10 recognized Aβ1–40, which migrated exclusively as a monomer. 6E10 also gave prominent staining of monomeric Aβ1–42 in all three fractions. The low molecular weight Peak 2 was primarily monomer. NU-1, NU-2 and NU-4, in contrast to 6E10, showed only faint staining of monomer in Peak 2 and little to no staining of monomers in Peak 1 or unfractionated ADDLs. All three antibodies showed reaction with trimer, tetramer, and 12–24mer found in ADDL preparations and Peak 1. Western blots thus indicate that NU-1, NU-2, and NU-4 have strong preferential binding for oligomers over monomers.

Figure 4.  NU-1, NU-2, and NU-4 show minimal monomer detection of Aβ peptides. Aβ1–40, ADDLs, and the two ADDL fractions obtained by SEC (Chromy et al. 2003) were separated by SDS-PAGE using 10–20% Tris-Tricine gels. Samples were transferred to nitrocellulose and then probed with NU-1, NU-2, NU-4, and 6E10. (a) On silver stain (not shown), Aβ1–40 has only a heavy monomer band. ADDLs and the separated ADDL peaks all have monomer, trimer and tetramer bands of various intensities. In Western analyses, 6E10 identified monomer bands in all four samples, with ADDLs and Peak 1 showing heavy trimer, tetramer, and 12–24mer bands, and Peak 2 primarily monomer with light trimer and tetramer. Monoclonals NU-1, NU-2, and NU-4, on the other hand, showed minimal or no monomer staining with ADDLs and Peak 1, but heavy trimer, tetramer, and 12–24mer that was similar or possibly more intense than 6E10. Peak 2 showed light monomer, trimer, and tetramer staining with all NU antibodies, but no 12–24mer. Aβ 1–40 was not recognized by NU-1, NU-2, or NU-4. (b) The aggregation of monomerized Aβ1–42 peptide on ice was monitored by dot blot, comparing NU-1, NU-2, and NU-4. Equal aliquots (20 fmol) of the solution were spotted on nitrocellulose at the times indicated and then probed with antibody. All antibodies detected time-dependent ADDL formation from Aβ1–42. (c) The relative intensity of the dot blots is shown as a function of time. In this assay, NU-1 and NU-4 have higher affinity for the oligomers than NU-2. These data are additional evidence that the NU monoclonals are oligomer-specific.

Validation of selectivity for oligomers

Two additional tests of specificity were carried out. First, we measured the time-dependent formation of oligomers from 10 nm monomeric Aβ1–42 (Figs 4b,c) by dot immunoblot. At the outset, when monomers would predominate, staining was light. Over the next 10 min, oligomerization was relatively rapid, and immunoreactivity increased 150-fold. In harmony with Western blots, these assays of undenatured molecules bound to nitrocellulose indicate NU-1, NU-2, and NU-4 are selective for oligomers.

Next, the affinity of NU-4 for ADDLs or Aβ1–40 in solution was assessed using a competition assay. NU-4 was preincubated with increasing amounts of ADDLs or fresh Aβ1–40 and then tested for the ability to bind to ADDLs immobilized on nitrocellulose (Fig. 5a). ADDLs in solution effectively blocked binding to immobilized ADDLs, with half maximal inhibition at ∼ 50 nm (Fig. 5b). Monomeric Aβ1–40 was ineffective at blocking binding, showing a slight linear decrease indicative of a nonspecific association.

Figure 5.  Soluble ADDLs block NU-4 binding to low concentration immobilized ADDLs better than monomeric Aβ1–40 does. (a) ADDLs were applied to nitrocellulose at a concentration of 10 pmol/0.5 μL. After blocking, each lane was incubated for 1.5 h with NU-4 that had been preincubated with either ADDLs or Aβ1–40 at various concentrations as indicated. Binding of antibody was then determined by incubation with anti-mouse-HRP and chemiluminescence. The Kodak Image Station 440 was used to quantify blot intensity. These data were then plotted as a function of peptide concentration using sigmaplot (b). ADDLs in solution effectively block NU-4 binding to immobilized ADDLs (half maximal blocking ∼ 50 nm), whereas monomeric Aβ1–40 blocks binding only nonspecifically.

Conformational nature of antibody epitope

To investigate the nature of the epitope for NU-1, NU-2 and NU-4, competition dot immunoblots first were carried out using short, non-overlapping peptides that encompassed the entire length of Aβ1–42 (1–12, 12–28, 25–35, 35–42). None of the peptides had any effect on binding of NU-1, NU-2, or NU-4 to ADDLs (Fig. 6a and data not shown), whereas ADDL controls completely blocked binding (data not shown). Next we tested the longer peptides Aβ1–28 and Aβ17–42. Aβ1–28 completely inhibited binding of NU-1 and NU-4, but not NU-2, to ADDLs (Fig. 6a and data not shown), suggesting that the monoclonal antibodies possess different conformational epitopes. Shown is an Aβ1–28 competition plot for NU-4. Half maximal inhibition for NU-4 was < 50 nm (Fig. 6c), slightly lower than with ADDLs. As ADDLs are a mixed solution that includes monomers, this difference is not surprising. Aβ17–42 had no impact nor did any combination of short peptides. Preincubation of NU-4 with Aβ1–28 also blocks binding in Western blots (data not shown). The conformational epitope of Aβ1–28 could derive from tertiary or quaternary structure. It previously has been suggested (Li and Fenselau 2004) that Aβ1–28 forms a dimer and when analyzed by SDS-PAGE (Fig. 6b), our preparation migrates as a single band consistent with this suggestion. This dimer however, was not recognized by NU-4 in Western blots (data not shown).

Figure 6.  The soluble peptide Aβ1–28 blocks NU-4 binding to ADDLs. Studies to investigate the epitope of NU-2 and NU-4 used incubation with four short linear peptides of 8–17 amino acids covering the entire length of Aβ1–42 to block binding of the antibody to immobilized ADDLs. None of the tested peptides blocked binding, even at 100-fold excess over antibody (a). Two longer peptides were then tested in a similar assay, with the result that Aβ1–28 showed excellent blocking ability for NU-4 (a, right) but not NU-2 (a, left). Blocking occurred either alone or in combination with other peptides (data not shown). When analyzed using SDS-PAGE and silver stain, Aβ1–28 appeared to run as a dimer (b). The blocking effectiveness of Aβ1–28 was similar to ADDLs, with an estimated half-maximal blocking near 50 nm(c).

Immunolocalization and blockade of ADDL binding to cells

ADDLs, whether formed in vitro or obtained from Alzheimer's disease brain, attach to synapses in hippocampal cultures, localizing to dendritic spines (Lacor et al. 2004). This dendritic binding, punctate in appearance, was readily apparent when probed using the indicated ADDL-generated monoclonal antibodies (Fig. 7a) and was comparable to detection using previously introduced polyclonals (M94 –Lambert et al. 2001; Lacor et al. 2004). Vehicle and no-secondary antibody controls showed no staining. The same punctate pattern was evident using a monoclonal that bound oligomers but not fibrils in Western blots (NU-7) (Chromy et al. 2003), verifying that binding was not attributable to large protofibrillar structures. When antibodies were incubated with ADDLs prior to addition to cultures, binding was effectively blocked (∼ 80%, Fig. 7b graph); significantly, this block occurred not only for synthetic ADDLs (data not shown) but also for ADDLs present in Alzheimer's disease brain extracts (Fig. 7b).

Figure 7.  NU-1 neutralizes binding of endogenous and synthetic ADDLs to cultured cells. Three-week-old cultured hippocampal cells were incubated with synthetic ADDLs for 60 min, washed, and fixed (see Materials and methods). Bound ADDLs were identified with various antibodies and visualized using the appropriate fluorescently conjugated secondary antibodies (a). Punctate labeling, found primarily on neurites and surrounding the cell body (pericellular), was similar for NU-1, NU-2, NU-4, and NU-7 and recapitulated that found with polyclonal antibodies reported previously (M94) (Lambert et al. 2001; Lacor et al. 2004). ADDLs have been shown previously to attach to dendritic spines in culture (Lacor et al. 2004); these patterns therefore reflect dendritic labeling. To test for immunoneutralization, ADDLs were preincubated with NU-1 for 1 h at 37°C prior to incubation with cultured hippocampal cells. After washing, anti-mouse AlexaFluor 488 was used to identify bound antibody. NU-1 blocked binding of both synthetic ADDLs (data not shown) and soluble extract from Alzheimer's disease brain containing endogenous ADDLs (b). Control plates in which fresh NU-1 was added to cells before AlexaFluor 488 also showed no ADDL binding. Quantification of hot spot binding with (+) and without (–) NU-1 preincubation is shown on the right panel. p = 0.005.

ADDLs cause a dose-dependent toxicity in cultured hippocampal cells, losing ∼ 50% of their reactivity in the MTT assay after overnight incubation with 1 μm ADDL (data not shown). This toxicity can be completely prevented by preincubation of NU-1 with ADDLs for 1 h before addition of the solution to the cells (three separate experiments with six replicates/data point), equivalent to the protection previously observed with polyclonal ADDL antibodies (Lambert et al. 2001).

Immunoblock of ADDL-induced reactive oxygen species (ROS) generation

Oxidative damage is a feature of many age-related neurodegenerative diseases, including Alzheimer's disease. To test whether blockade of binding by anti-ADDL antibodies was preventing the downstream impact of ADDLs on neurons, we investigated the possible changes in ROS generation by ADDLs. Addition of 500 nm ADDLs to the culture medium for 3 h induced a marked increase in the intracellular levels of ROS in neurons compared to vehicle-treated cultures (Fig. 8). After 3 h incubation, the increase in ROS is approximately 3.5-fold. The addition of anti-ADDL antibodies NU-1, NU-2 and NU-4 to the culture medium completely blocked the ADDL-induced intracellular ROS generation, as indicated by a lower intensity of neuronal fluorescence (Fig. 8). Thus, the monoclonals not only prevent binding of ADDLs to cells, they also can prevent a pathological response to ADDL treatment.

Figure 8.  NU-1, NU-2 and NU-4 block ADDL-induced ROS generation in cultured cells. Cultured E-18 hippocampal neurons (19 days in vitro) were incubated with vehicle, ADDLs (500 nm), or ADDLs plus the indicated antibody at equal molar concentrations for 3 h at 37°C. When present, anti-ADDL antibodies were added immediately before ADDLs. Shown here are representative fluorescence images of CM-H₂DCFDA staining (loaded for 40 min), which indicate the intracellular ROS levels. The right bottom panel shows the quantitative analysis of DCF fluorescence for treated neurons obtained in three separate experiments (∼ 200 cells analyzed per experimental condition) using Image J software (NIH Windows version). ADDL antibodies NU-1, NU-2 and NU-4 all blocked the ADDL-induced increase in ROS.

Discussion

ADDLs were used to generate monoclonal antibodies that targeted pathological Aβ assemblies. Three types of antibodies of interest were generated: those capable of binding oligomers and fibrils; those capable of binding fibrils but not oligomers; and those capable of binding oligomers but not fibrils. The first class was the most common, and several clones of this class were expanded and characterized in depth. These clones distinguished Alzheimer's disease from control brains in extracts and in tissue sections. In some cases, discrimination derived from a 3D epitope common to ADDLs and assemblies of Aβ1–28 that were absent from Aβ1–40 and other linear sequences. The antibodies blocked binding of ADDLs to synapses in hippocampal cultures, inhibited ADDL-induced increases in ROS, and prevented ADDL toxicity in the MTT assay. This selective immunoneutralization of pathological Aβ assemblies indicates these antibodies provide a useful prototype for therapeutic strategies based on monoclonal antibodies.

Immunization with various forms of Aβ fragments and assemblies has previously generated antibodies with a variety of useful properties. Short N-terminal peptides, for example, were used to generate the 6E10 antibody (Kim et al. 1990) used in the current study which binds to virtually all forms of Aβ. C-terminal antigens produced monoclonals that distinguish Aβ1–40 from Aβ1–42, making them useful for selective ELISAs (Savage et al. 1998; Jensen et al. 2000). Kayed et al. (2003) recently used Aβ1–40 coupled to gold colloids to generate antibodies that bind oligomers (but not monomers or fibrils) of several amyloidogenic proteins, including amyloid beta. In human brain sections the antibodies react with diffuse, early stage plaques but not thioflavin-positive dense-core plaques. Polyclonal antibodies generated by vaccination with ADDL preparations (Lambert et al. 2001) have been used to distinguish Alzheimer's disease brain tissue from control in both slices and dot immunoblots, to identify the 12mer found in Alzheimer's disease brain, to recognize ADDLs bound to synapses on cultured cells, and to identify lead compounds for anti-ADDL drug discovery (Lambert et al. 2001; Chang et al. 2003; Gong et al. 2003; Lacor et al. 2004).

The NU-1 and NU-4 monoclonals resemble ADDL-generated polyclonal antibodies in recognizing pathogenic assemblies but not monomers in Western and dot immunoblot paradigms. In Alzheimer's disease brain sections, the expected Alzheimer's disease-type immunoreactivity was present, coupled with an intriguing pericellular immunoreactivity. The pericellular immunoreactivity has been attributed hypothetically to oligomer attachment to synapses (Lacor et al. 2004), thought to occur early in Alzheimer's disease or even preclinically. With NU-6, which also was generated using ADDLs as antigen, the staining was strikingly different. Immunoreactivity was intracellular and present in cells whose morphology and GFAP-staining were compatible with reactive astrocytes. Appearance of reactive astrocytes and activated microglia in Alzheimer's disease pathology has been extensively described (Itagaki et al. 1989). The pattern seen here is interesting because, while NU-6 binds human aggregates but not oligomeric Aβ in western blots, it does not stain compact plaques. Thus the intracellular astrocyte species might represent a novel intermediate in fibril pathology. How these species originate is unknown, but astrocytes have been shown to participate in clearance and degradation of Aβ in tg mice (Wyss-Coray et al. 2003). Hypothetically, this capacity in Alzheimer's disease becomes overwhelmed, perhaps in a manner related to the intracellular accumulation and processing of ADDLs.

The species recognized by NU-4 in western blots included prominent higher order oligomers (12–24mers), even though no material was detected by silver stain. Aβ1–40 monomer was not recognized in immunoblots by NU-4, nor was Aβ1–42 monomer recognized when SDS was present in transfer buffer. Without SDS, however, NU-4 showed immunoreactivity at the Aβ 1–42 monomer position. Transfer of monomer was unaffected by SDS, verified in streptavidin-HRP assays for biotinylated Aβ1–42 (data not shown). We hypothesize that without SDS, monomers combine during transfer and produce oligomers, consistent with the rapid oligomerization seen even at very low Aβ concentrations (Fig. 4). Competition experiments verified that NU-4 had minimal affinity for monomers in solution, circumventing the ambiguities associated with SDS-PAGE analysis.

Competition experiments (Fig. 6) further showed that binding of ADDLs to some antibodies depended on conformational determinants, consistent with previous suggestions of Aβ1–42 antigenicity (Cribbs et al. 2003). Preincubation of NU-1 and NU-4 with solutions containing ADDLs or Aβ1–28 blocked subsequent binding to ADDLs in dot immunoblots. NU-2 was blocked only by ADDLs. Other peptides, including Aβ1–40, were without effect. Analysis of Aβ1–28 by electrospray mass spectrometry indicates it forms dimers and possibly other higher order oligomers in the gas phase (Li and Fenselau 2004). Two dimer structures were hypothesized to fit the data, with core overlaps of either Aβ17–23 or Aβ17–28. We therefore suggest that the NU-1 and NU-4 epitope is three-dimensional, in which aggregation of Aβ1–28 to form a dimer is necessary to align correct amino acids in the binding pocket, perhaps several from each single peptide. Denatured Aβ1–28 migrates at a position consistent with dimer, as seen by silver stain (Fig. 6). Alternatively, the single peptide itself could fold to align amino acids in the proper sequence, whereas the shorter peptides, as well as full-length Aβ1–40, could not.

Earlier, antibodies with epitopes against the N-terminal of Aβ were suggested to be more effective than C-terminal or central epitopes at plaque clearance in tg mice (Bard et al. 2003) and at inhibition of fibrillogenesis and cytotoxicity (McLaurin et al. 2002). For example, a substituted single chain antibody that relies on the Aβ3–6 sequence (EFRH) suppressed in vitro formation of Aβ aggregates and caused a loss of brain amyloid burden in tg mice (Frenkel et al. 2000; Solomon 2003). However, more recent results have shown that antibodies developed against other parts of the Aβ sequence, as well as against cross-linked Aβ1–40 oligomers (Lee et al. 2006), can be effective at reversing memory deficits. In most cases, this reversal is seen without change in amyloid burden. One antibody, M266, generated against the central domain Aβ13–28, caused reversal of memory deficits after only 24 h with no change in brain Aβ plaques (Dodart et al. 2002). As this antibody alters clearance of CNS and plasma Aβ (Seubert et al. 1992), the authors suggest that the antibody is acting as a sink for Aβ, leading to CNS clearance (DeMattos et al. 2001). Another N-terminal antibody, BAM-10, caused reversal of memory impairment in tg mice with no decrease in amyloid plaques (Kotilinek et al. 2002). The antibody was hypothesized to act by neutralizing soluble Aβ assemblies in the brain that cause cognitive impairment. Soluble Aβ assemblies accumulate in tg mice (Chang et al. 2003) as well as in Alzheimer's disease brain, where the major species is a 12mer (Gong et al. 2003). Confirmation that the 12mer is associated with learning deficits has recently been reported in Tg 2576 mice by Lesne et al. (2006). In other studies with tg mice, oligomer-selective antibodies have recently been shown to reduce levels of oligomers, plaques, and Alzheimer's disease-type tau phosphorylation (Ma et al. 2006; Oddo et al. 2006).

Success of active and passive vaccinations in mouse Alzheimer's disease models led to clinical trials for an active vaccine in Alzheimer's disease patients (Schenk et al. 1999, 2004). Phase 2 trials were stopped due to the appearance of meningoencephalitis in 6% of the patients (Hock et al. 2003; Orgogozo et al. 2003), but early clinical data from immunotherapy have been encouraging and efforts are underway to develop alternative Aβ immunotherapies (Gelinas et al. 2004). Active vaccines are being developed using immunoconjugates composed of a fragment of Aβ, usually from the N-terminal or central region, linked to a carrier protein (Schenk et al. 2004). These antigens contain no epitopes from Aβ1–42 that will initiate a T-cell response, which may eliminate autoimmune reactions. Passive vaccines are also being developed, which present advantages in avoiding poor immune responses in older patients as well as eliminating the T-cell response to Aβ. Oral immunization in tg animals has also been reported (Youm et al. 2005). We previously suggested that development of therapeutic antibodies that target ADDLs may optimize the efficacy of this approach (Klein 2000). The current study establishes that relatively potent monoclonal antibodies can be generated using synthetic ADDLs as immunogens, and that the resultant antibodies are selective for pathological assemblies of Aβ. The antibodies, moreover, immunoneutralize ADDLs obtained from Alzheimer's disease brain, blocking their toxicity and attachment to synapses in cell culture assays and blocking ADDL-induced generation of ROS. The antibodies thus hold promise for use in future behavioral studies of ADDL neutralization in tg-mice Alzheimer's disease models and provide a step toward antibodies of clinical value.

ADDL-selective monoclonal antibodies are also expected to be of value for Alzheimer's disease diagnostics. In recent studies, ADDL-selective antibodies were combined with developments in nanotechnology to develop assays capable of detecting ADDLs in human CSF (Georganopoulou et al. 2005; Haes et al. 2005). The new assays are orders of magnitude more sensitive than the best available ELISAs, capable of detecting atomolar ADDL concentrations. Results showing a 10-fold difference in CSF ADDL levels between Alzheimer's disease and control subjects (Georganopoulou et al. 2005) suggest that nanotechnology-based assays in combination with ADDL-selective monoclonal antibodies could provide the first clinical diagnostics for Alzheimer's disease.

Conflict of interest statement

W. L. Klein and G. A. Krafft are cofounders of Acumen Pharmaceuticals, Inc., which has an exclusive license from Northwestern University and the University of Southern California to develop Aβ-derived diffusible ligands (ADDLs) for Alzheimer's disease diagnostics and therapeutics.

Acknowledgements

This work was supported by grants from the Alzheimer's Disease Research Fund; National Institutes of Health-National Institute on Aging Grants R01-AG18877, AG22547, and R01-AG11385 (WLK); and P30AG13854 (Northwestern Alzheimer's Disease Center); the Boothroyd, Feiger, and French Foundations; and by a gift from Acumen Pharmaceuticals to Northwestern University. FGF is a Human Frontier Science Program (HFSP) Fellow.