Conformation-sensitive Antibodies against Alzheimer Amyloid-β by Immunization with a Thioredoxin-constrained B-cell Epitope Peptide*

Immunotherapy against the amyloid-β (Aβ) peptide is a valuable potential treatment for Alzheimer disease (AD). An ideal antigen should be soluble and nontoxic, avoid the C-terminally located T-cell epitope of Aβ, and yet be capable of eliciting antibodies that recognize Aβ fibrils and neurotoxic Aβ oligomers but not the physiological monomeric species of Aβ. We have described here the construction and immunological characterization of a recombinant antigen with these features obtained by tandem multimerization of the immunodominant B-cell epitope peptide Aβ1-15 (Aβ15) within the active site loop of bacterial thioredoxin (Trx). Chimeric Trx(Aβ15)n polypeptides bearing one, four, or eight copies of Aβ15 were constructed and injected into mice in combination with alum, an adjuvant approved for human use. All three polypeptides were found to be immunogenic, yet eliciting antibodies with distinct recognition specificities. The anti-Trx(Aβ15)4 antibody, in particular, recognized Aβ42 fibrils and oligomers but not monomers and exhibited the same kind of conformational selectivity against transthyretin, an amyloidogenic protein unrelated in sequence to Aβ. We have also demonstrated that anti-Trx(Aβ15)4, which binds to human AD plaques, markedly reduces Aβ pathology in transgenic AD mice. The data indicate that a conformational epitope shared by oligomers and fibrils can be mimicked by a thioredoxin-constrained Aβ fragment repeat and identify Trx(Aβ15)4 as a promising new tool for AD immunotherapy.

Immunotherapy against the amyloid-␤ (A␤) peptide is a valuable potential treatment for Alzheimer disease (AD). An ideal antigen should be soluble and nontoxic, avoid the C-terminally located T-cell epitope of A␤, and yet be capable of eliciting antibodies that recognize A␤ fibrils and neurotoxic A␤ oligomers but not the physiological monomeric species of A␤. We have described here the construction and immunological characterization of a recombinant antigen with these features obtained by tandem multimerization of the immunodominant B-cell epitope peptide A␤1-15 (A␤15) within the active site loop of bacterial thioredoxin (Trx). Chimeric Trx(A␤15) n polypeptides bearing one, four, or eight copies of A␤15 were constructed and injected into mice in combination with alum, an adjuvant approved for human use. All three polypeptides were found to be immunogenic, yet eliciting antibodies with distinct recognition specificities. The anti-Trx(A␤15) 4 antibody, in particular, recognized A␤42 fibrils and oligomers but not monomers and exhibited the same kind of conformational selectivity against transthyretin, an amyloidogenic protein unrelated in sequence to A␤. We have also demonstrated that anti-Trx(A␤15) 4 , which binds to human AD plaques, markedly reduces A␤ pathology in transgenic AD mice. The data indicate that a conformational epitope shared by oligomers and fibrils can be mimicked by a thioredoxin-constrained A␤ fragment repeat and identify Trx(A␤15) 4

as a promising new tool for AD immunotherapy.
Antipeptide antibodies are valuable tools for probing structure-function relationships in proteins as well as for therapeutic and diagnostic applications. When immunotherapy is the ultimate goal, the sequence span of the peptide, and thus the nature of the epitope(s) associated with it, as well as the conformational selectivity of the resulting antibodies may also be dictated by safety reasons. This is the case of the amyloid-␤ (A␤) 2 peptide, a major neuropathological hallmark (1-3) and a promising immunotherapeutic target of Alzheimer disease (AD) (4 -7). Following the encouraging results obtained with A␤42 vaccination in transgenic mouse models of AD (8 -10), a Phase II clinical trial utilizing preaggregated A␤42 as an antigen and the QS-21 saponin as an immunoadjuvant (AN-1792 vaccine) was halted due to the occurrence of meningoencephalitis in ϳ6% of treated patients (11)(12)(13). T-cell-mediated autoimmunity produced by the C-terminal portion of A␤ (14) along with a predominantly pro-inflammatory (T helper 1) immunoresponse and cross-reactivity with the presumably physiological monomeric forms of A␤ (A␤38 -43 peptides plus the amyloid-␤ precursor protein have been hypothesized as the main causes of these adverse effects (15)(16)(17). Of further concern is the fact that not only plaques but also soluble A␤ oligomers should be bound by a therapeutically effective antibody (18 -20). The latter represent an intermediate conformation prior to fibril formation and are presently considered as the most proximate causative agents of AD (21)(22)(23)(24). Some of the above limitations have been overcome through the development of alternative immunogens relying on N-terminally located B-cell epitope-bearing fragments rather than on full-length A␤ (25)(26)(27)(28)(29)(30)(31)(32)(33). Also important in this regard is the availability of A␤ peptide formulations capable of eliciting antibodies that recognize specific assembly states of A␤42 (19, 34 -36). For example, antibodies binding to oligomeric and fibrillar (but not to monomeric) A␤, have been produced by using purified A␤42 oligomers (19) or nitrated A␤40 (35) as antigens, whereas antibodies selectively recognizing soluble A␤ oligomers have been obtained by immunization with A␤40 immobilized onto gold nanoparticles (36). Interestingly, the latter antibodies bind not only to A␤ oligomers but also to the oligomeric forms of various A␤-unrelated amyloidogenic polypeptides, suggesting that a common (as yet unidentified) conformation-dependent structure is shared by soluble oligomers regardless of their sequence (36).
A similar conformational selectivity has not been achieved so far with a T-cell epitope lacking a fragment of A␤. Despite the lack of detailed information on the molecular basis of the conformational mimics generated by chemical modification or surface immobilization of the A␤ peptide, it is clear that some kind of structural constraining is involved. We thus reasoned that a similar conformational effect, besides an enhanced peptide stabilization and immunogenicity (37,38), might be obtained with the use of a scaffold protein, such as thioredoxin, with the ability to constrain the structure of short peptides inserted within its surface-exposed active site loop (38 -40) and to stimulate T-cell proliferation (41,42). This kind of recombinant antigen would be highly desirable because of its expected safety, ease of construction, and large scale production in a chemically homogeneous form.
We have shown here that antibodies recognizing A␤42 oligomers and fibrils (but not monomers) are produced by immunization with a 4-fold repeat of the A␤15 peptide bearing an interposed three-amino-acid linker and arranged in tandem within the display site of bacterial thioredoxin (Trx). As revealed by comparison with Trx(A␤15) n antigens bearing one or eight copies of A␤15, conformational selectivity critically depends on A␤15 multiplicity and on the use of multipeptide insertions fitting the structural constraining capacity of Trx. We have further documented the ability of the anti-Trx(A␤15) 4 antibody to bind A␤ aggregates in human brain and to ameliorate AD pathology in APP transgenic mice.

EXPERIMENTAL PROCEDURES
TrxA␤ Constructs-Chimeric Trx polypeptides bearing the A␤15 or the A␤42 peptide (TrxA␤) were constructed by using a modified pET28 plasmid (Novagen), designated as pT7Kan-Trx, harboring the sequence for an N-and C-terminally His 6tagged version of bacterial thioredoxin along with a kanamycin resistance marker. The unique CpoI site present within the Trx coding sequence (nucleotide positions 99 -105, corresponding to amino acid residues 34 -35) was used as the cloning site. The phosphorylated oligonucleotides 5Ј-gtccgatggatgcagaattccgacatgactcaggatatgaagttcatcatcaaggcg-3Ј (forward) and 5Ј-gaccgccttgatgatgaacttcatatcctgagtcatgtcggaattctgcatccatcg-3Ј (reverse), both bearing 5Ј-protruding CpoI sequences (underlined), were annealed and ligated to CpoI-digested pT7Kan-Trx at a 1:10 vector:insert molar ratio. The resulting construct, pT7Kan-TrxA␤15, encodes a polypeptide, Trx-(1-33)gPMDAEFRHDSGYEVHHQGGpTrx (36 -109) in which A␤15 (underlined) is preceded by a Met residue at the N terminus and is followed by the Gly-Gly-Pro linker at the C terminus (with the Gly and Pro residues provided by the inserted ds-oligonucleotide and by the Trx scaffold, respectively). Constructs bearing multiple copies of A␤15 were generated in a similar way but at a 1:100 vector:insert molar ratio using the sequence-verified insert excised from pT7-Kan-TrxA␤15 as starting material. Recombinant clones were screened by restriction analysis, and two of them, bearing four or eight copies of A␤15, were selected. Essentially identical experimental conditions were used to construct TrxA␤42, which was cloned in both pT7Kan-Trx as well as in a compan-ion vector (pT7Amp-Trx) carrying an ampicillin rather than a kanamycin antibiotic resistance marker.
Expression and Purification of the TrxA␤ Polypeptides-Expression was induced by adding 1 mM isopropyl-␤-D-thiogalactopyranoside to Escherichia coli BL21Star (DE3) cells (Invitrogen) transformed with each of the above constructs and allowed to proceed for 2 h at 37°C. A different E. coli strain (Origami-DE3; Novagen) and modified expression conditions (pT7Amp-Trx vector; 5 h at 30°C) were used for TrxA␤42, which was otherwise completely insoluble. Following cell lysis, His 6tagged TrxA␤ polypeptides were bound to a metal affinity resin (Talon; Clontech), purified according to the manufacturer's instructions, and extensively dialyzed against phosphate-buffered saline (PBS). Protein concentration was determined with the Coomassie dye method (Bio-Rad) and by UV light absorbance. The composition and purity of individual polypeptide preparations was assessed by gel electrophoresis on 11% polyacrylamide-SDS gels.
Immunization Protocols-TrxA␤ polypeptides (2 mg/ml in PBS) were filter-sterilized, and an aliquot of each (10 nmol) was mixed with 1 mg of alum (Sigma-Aldrich) in a final volume of 400 l immediately before use. A␤42 (Sigma-Aldrich) was dissolved in PBS (2 mg/ml) and aggregated overnight at 37°C prior to immunization. Five randomly assorted groups of onemonth-old male BALB/c mice (10 animals each) (Charles River Laboratories) were injected subcutaneously with the above antigens at days 1, 15, 30, and 60, as specified in the legend to Fig. 2A. The same treatment was applied to two negative control groups that were injected with PBS and with aggregated A␤42, both without alum. Sera were collected two weeks after the last boost and randomly pooled in pairs. A standard immunization protocol (three doses of Trx(A␤15) 4 antigen in Freund adjuvant administered to one animal over a period of two months) was used to generate rabbit anti-Trx(A␤15) 4

antibodies (SeqLab).
Detection of Anti-A␤42 Antibodies-Total anti-A␤42 antibodies were detected by enzyme-linked immunosorbent assay (ELISA) at a fixed 1:200 dilution, using aggregated A␤42 (0.5 g/well) as the target antigen (43). Following incubation, washing, and the addition of horseradish peroxidase-conjugated anti-mouse immunoglobulins (1/5000; Sigma-Aldrich) and the chromogenic substrate o-phenylendiamine (Sigma-Aldrich), plates were read spectrophotometrically at 450 nm. Immunoglobulin isotype determination was conducted at a fixed 1:200 dilution using rat anti-mouse Ig subclass-specific, horseradish peroxidase-conjugated secondary antibodies (TechniPharm). ELISAs were conducted in triplicate on the five-paired sera from each group; only a subset of sera from the three top responders in Groups 3, 4, 6, and 7 (see Fig. 2A) was utilized for isotype determination. Comparisons between groups were conducted by one-way analysis of variance using Analyze-it software.
Immunohistochemistry-Sera from mice immunized with each of the three Trx(A␤15) n polypeptides were screened for their ability to bind A␤ plaques in human brain sections from a 68-year-old patient with neuropathological and clinical symptoms typical of severe AD. Various dilutions (1:100 -1:1000) of pooled sera from the three top responders in Groups 5, 6, and 7 were analyzed. Sera were added to serial 8-m brain sections of formalin-fixed, temporal cortical tissue, pretreated with formic acid (80%, 15 min). Sera from mock-treated animals (PBS; Group 1) and a commercial anti-A␤40 polyclonal antibody (Anti-Pan ␤-Amyloid, BIOSOURCE) were used as negative and positive controls, respectively. Immunolabeling was revealed with the EnVision Plus/horseradish peroxidase system (Dako), using 3-3Ј-diaminobenzidine as the chromogenic substrate, according to the manufacturer's instructions. Images were captured with a digital camera at magnifications ranging from 50 to 400ϫ.
Immunodot Blot Assays and Atomic Force Microscopy Imaging-A stock solution of trifluoroacetic acid-pretreated, monomeric A␤42 peptide (1 mM) was prepared as described previously (44) and diluted to the required A␤42 final concentration (2-50 M) with 0.1% trifluoroacetic acid prior to immunodot blot analysis. A␤42 dissolved in 2 M dimethyl sulfoxide (1 mM final concentration) was utilized for the preparation of aggregated A␤42 species (19,21,45). A 10-fold dilution of the above stock solution into cold Ham's F-12 K medium (phenol red-free; BIOSOURCE) followed by incubation at 4°C for 24 h was used to produce soluble oligomers; the same stock solution diluted into 10 mM HCl at a final concentration of 100 M and incubated for 24 h at 37°C was used to generate A␤ fibrils. A␤ oligomer and fibril formation, as well as the absence of fibrillar aggregates in soluble oligomer preparations, were verified by atomic force microscopy. To this end, samples of the abovedescribed A␤42 preparations were diluted 10-fold in 20 l of deposition buffer (4 mM HEPES, pH 7.4, 10 mM NaCl, 7 mM MgCl 2 ) and immediately deposited onto freshly cleaved ruby mica at room temperature. After 5 min, mica disks were rinsed with milli-Q grade water and gently dried under a stream of nitrogen. Images were collected with a Nanoscope III microscope (Digital Instruments) operated in tapping mode using commercial diving board silicon cantilevers (MikroMasch). Oligomer formation was also checked by electrophoretic analysis on Tris-Tricine polyacrylamide gels followed by silver staining as described in Ref. 19. Samples of recombinant Ile 84 3 Ser transthyretin (Ser 84 -TTR) enriched in soluble oligomers or higher order fibrillar aggregates were prepared by incubating the protein (2 mg/ml) for either 2 or 96 h at 37°C in 100 mM sodium-acetate buffer (pH 4.0) as described in Refs. 46 and 47; control samples of monomeric Ser 84 -TTR were obtained from parallel incubations in 50 mM K-phosphate buffer (pH 7.6). For dot blot analysis, the various A␤42 or Ser 84 -TTR species were hand-spotted onto nitrocellulose membranes (GE Healthcare) pre-wetted with TBS (20 mM Tris-HCl, pH 7.5, 0.8% NaCl) (19). Antisera for dot blot analysis were affinity-purified on protein-A minicolumns (Diatheva) according to the manufacturer's instructions. Following the determination of total immunoglobulin concentration with the Coomassie dye method, purified immunoglobulins were used for dot blot assays at a final concentration of 0.75 g/ml. After blocking at room temperature with 5% nonfat dry milk in TBS supplemented with 0.05% Tween 20 (TBST), the blots were incubated for 1.5 h with each of the three anti-Trx(A␤15) n antibodies in dry milk-TBST, washed 3 ϫ 10 min with TBST, followed by mouse immunoglobulin detection with the SuperSignal West Femto kit (Pierce), as specified by the manufacturer. To distinguish between monomer and oligomer binding, affinity-purified anti-monomeric TrxA␤15 antibodies (0.75 g/ml) were preincubated (2 h at 4°C under stirring conditions) with a 50-fold molar excess of the monomeric A␤42 peptide in TBS prior to hybridization. At least two replicates were carried out for each of the 15 pools of antisera from the various Trx(A␤15) n -treated groups (see Fig.  2A, . Dot blot images were quantified with the Quantity One program (Bio-Rad). Signal volume intensities for individual A␤42 species within each membrane were averaged, and the resulting data were plotted with Sigma Plot (SPSS).
Surgery and Neuropathological Evaluation of Transgenic AD Mice-Female transgenic AD (Tg2576) mice expressing the Swedish mutation of human APP (48) were obtained from the Boston University Alzheimer Disease Center mouse colony. Founders for this colony were provided by Dr. Karen Hsiao-Ashe (Department of Neurology, University of Minnesota Medical School). The APP Tg2576 mice developed behavioral abnormalities and exhibited histological evidence of brain A␤ deposits as plaques along with associated astrogliosis from as early as 8 months. Mice were genotyped using a standardized PCR assay on tail DNA and were housed four in each cage under standard conditions with ad libitum access to food and water. Six 14-month-old APP mice (32-34 g each), placed on a 12-h light schedule, were used for surgeries. Mice were anesthetized with ketamine HCl/xylazine intraperitoneal injection (100 mg/kg ketamine and 10 mg/kg xylazine; 100 l/10 g body weight) and were positioned in a stereotaxic apparatus (Koph) with a mouse head adaptor. Thermoregulation was maintained at 37°C using a warming pad with respiratory monitoring throughout the procedure. The scalp was incised in the midline to expose the sagittal suture, and stereotaxic coordinates in both hemispheres were determined (49). The bregma was used as a reference point (Ϫ2.0 mm), and holes were drilled in the calvarium at the junction of the left and right lateral coordinates (1.75 mm). Affinity-purified anti-Trx(A␤15) 4 antibodies along with mock immunoglobulins from PBS-treated mice (2 l each) were stereotaxically injected into the left and right hippocampi (Ϫ2.0 mm ventral), respectively, using a blunt-tipped 10-l syringe (Hamilton). Upon syringe placement, there was a 2-min dwell time followed by a 4-min injection time and an additional 2-min dwell time prior to removal of the syringe. A topical antiseptic was applied as the incision was closed using a 9-mm autoclip. Mice were kept on a warming pad until full recovery. All animal experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and both the Department of Veterans Affairs and Boston University Animal Care committees. Seven days post-injection, the mice were deeply anesthetized and transcardially perfused with 2% buffered paraformaldehyde (100 ml). Brains were post-fixed for 2 h, cryoprotected in a graded series of glycerol, and subsequently frozen-sectioned (50 m). Serially cut mouse tissue sections were stained for Nissl substance, immunostained with anti-A␤42 (catalog number 344; BIOSOURCE International), anti-A␤ oligomer (A11; BIOSOURCE International), and glial fibrillary antigen protein (Dako) antibodies and silver-stained using the Campbell-Switzer method for identification of mature A␤ plaques. Immunostained coronal tissue sections, serially cut within the hippocampus beginning from interaural (1.68 mm)/bregma (Ϫ2.12 mm) to interaural (2.16 mm)/ bregma (Ϫ1.64 mm), were analyzed. A␤-positive plaques were quantified from high resolution images of the same brain areas within the anti-Trx(A␤15) 4 -treated hemisphere and the contralateral mock-treated hemisphere using BioVision (50) and Neurolucida software programs (MicroBrightField, Williston, VT). BioVision differentiates and counts plaques from the background neuropil, whereas Neurolucida extracts the data from the BioVision images, exporting it to Excel (Microsoft, Redmond, WA) for statistical analysis.

Design and Construction of the Trx(A␤15) n Polypeptides-
The choice of A␤15 as a peptide epitope was based on the lack of T-cell reactivity and good solubility previously ascribed to N-terminal peptides encompassing the first 15 residues of A␤ (4,14,27,30,43,51). Trx was chosen because of its small size (109 amino acids) and solubility, structural rigidity, ability to act as a nontoxic immunoenhancer capable of stimulating murine T-cell proliferation, and presence of a convenient restriction site (CpoI) within its solvent-accessible active site loop (40 -42). Insertions into this loop are usually well tolerated, and grafting both ends of the inserted peptide provides a conformational constraint that are neither achievable with Nor C-terminal fusions nor with random chemical cross-linking of peptides to a protein carrier (38 -40). To mimic different assembly states of A␤ and to overcome the poor immunogenicity of short A␤ fragments (25,27), Trx(A␤15) n polypeptides bearing a single copy or multiple tandemly arranged copies of A␤15 were constructed. A cloning strategy relying on the use of an excess of the A␤15 DNA insert with respect to a modified recipient vector bearing the Trx coding sequence under the control of a phage T7 promoter (see "Experimental Procedures") was utilized for Trx(A␤15) n construction (Fig. 1A). Constructs bearing one, four, or eight copies of A␤15 were used to express the corresponding polypeptides, which were then purified by metal affinity chromatography (Fig. 1B). Instrumental to the production of properly assembled A␤15 multimers were the directionality and in-frame fusion capability of the unique CpoI site present within the Trx active site loop region as well as the incorporation into A␤15 DNA of a terminal sequence coding for an intervening Gly-Gly-Pro linker, thus also preventing the formation of junctional epitopes (52). A fourth construct (TrxA␤42) bearing a single copy of the fulllength A␤42 peptide was prepared in a similar way (see "Experimental Procedures" for details). Although all Trx(A␤15) n polypeptides were completely soluble, regardless of A␤15 multiplicity, most of the TrxA␤42 protein ended up in inclusion bodies in an insoluble form (not shown). Thus, A␤42 appears to be poorly soluble, even when fused to Trx in the heterologous context of bacterial cells.
Immunological Characterization of the Trx(A␤15) n Polypeptides-Five groups of 10 male BALB/c mice were treated with 10 nmol of the above-described Trx(A␤15) n polypeptides or with equivalent amounts of preaggregated synthetic A␤42 or TrxA␤42, all supplemented with alum, an immunoadjuvant approved for human use (Fig. 2A). Two additional groups injected with buffer alone (PBS) or with alum-free A␤42 served as negative controls. No adverse side effect was observed in animals injected with four doses of the various Trx(A␤15) n polypeptides over a period of two months. Sera were collected two weeks after the fourth injection and randomly pooled in pairs, and the five resulting pools were analyzed with ELISA using aggregated fibrillar A␤42 as the target antigen. As shown in Fig. 2A, mean anti-A␤ antibody levels elicited by Trx(A␤15) 4 and Trx(A␤15) 8 (but not monomeric TrxA␤15) were significantly higher (p Յ 0.05) than those of mock-treated controls and similar to those of the A␤42-treated groups, where TrxA␤42 performed as well as free A␤42. An anti-inflammatory T helper 2-polarized response, typical of the alum adjuvant (14,25,53), was revealed by isotype profiling (Fig. 2B). Although a prevalence of IgG1 was observed with all antigens, the IgG1:IgG2a ratio, an indicator of T helper 2 polarization, was reproducibly higher (p Յ 0.05) for multimeric Trx(A␤15) n and TrxA␤42 immunoconjugates than for unconjugated A␤42.
Detection of A␤ Aggregates in Human Brain by Anti-Trx(A␤15) n Antibodies-The ability of antisera generated in response to Trx(A␤15) n to bind amyloid plaques was investigated next. This property, which is considered the best prognostic indication of in vivo anti-A␤ antibody efficacy (26), is not shared by all previously described anti-A␤ antibodies (e.g. m266 and other antibodies targeting the C-terminal portion of A␤42 (26)). As shown in Fig. 3, sera from mice immunized with the tetrameric (Fig. 3A) or the octameric (Fig. 3B) form of Trx(A␤15) n bound to amyloid plaques up to a dilution of 1:1000. Large neuritic plaques, as well as mature and immature plaques, were labeled by anti-multimeric Trx(A␤15) n antibodies. A broader immunostaining, especially within senile plaque cores, was observed with the positive control anti-Pan ␤-amyloid antiserum generated in rabbits using A␤40 as the antigen (not shown). By comparison, no plaques were detected either with sera from mock-treated animals (not shown) or with sera from mice immunized with monomeric TrxA␤15 (Fig. 3C).
Differential Recognition of Distinct A␤ Assembly States by Anti-Trx(A␤15) n Antibodies-Immunodot blot assays were used then to assess the recognition selectivity of the various anti-Trx(A␤15) n antibodies toward different assembly states of A␤42 (monomers, oligomers, and fibrils) generated in vitro under previously determined conditions (21,45). Fibril formation, the lack of A␤ aggregates in the monomer solution, as well as the lack of fibrils in the oligomer preparation were verified by atomic force microscopy (Fig. 4A). Silver-stained polyacrylamide gels were used to evaluate oligomer preparations, which in keeping with previous reports (19,21,45) and despite prolonged incubation in F-12 medium, were found to contain ϳ50% residual monomer, with trimers and tetramers as the most represented oligomeric species (Fig. 4B). As shown in Fig.  4C, antibodies from mice immunized with monomeric A, anti-A␤42 reactivity of sera from mice immunized with the indicated Trx(A␤15) n polypeptides (Groups 5-7), with preaggregated synthetic A␤42 (Group 3), or TrxA␤42 (Group 4), all adjuvanted with alum, or mock-immunized with either adjuvant-free A␤42 (Group 2) or PBS (Group 1) (see "Experimental Procedures" for details). ELISA data for five randomly paired pools of sera from each group were obtained using aggregated synthetic A␤42 as target antigen and are presented as a scatter plot; the A␤ binding activity of the 35 individual pools as well as the mean binding activity of each of the seven groups (bars) are shown. B, immunoglobulin isotypes of antisera from the three top responders in each immunopositive group (3, 4, 6, and 7). TrxA␤15 do not bind to A␤ fibrils, in accordance with their inability to recognize higher order A␤42 aggregates in ELISAs or A␤ fibrils in AD plaques (see Figs. 2A and 3C), yet they appear to react with both monomers and oligomers. However, because of the presence of sizeable amounts of monomeric A␤ in the oligomer preparation (see Fig. 4B), it is likely that A␤ monomers, rather than oligomers, are the actual targets of antimonomeric TrxA␤15 antibodies. In fact, a loss of immunoreactivity was observed upon preincubation of these antibodies with an excess of the monomeric A␤42 peptide (not shown). A different recognition pattern was observed with antibodies raised against Trx(A␤15) 4 , which recognized A␤42 oligomers and fibrils but not monomers (Fig. 4C). In contrast but similar to the results obtained with the commercial 4G8 monoclonal antibody (not shown), all three A␤42 species were bound by antibodies raised against Trx(A␤15) 8 (Fig. 4C), thus indicating that excessive multimerization of the A␤15 peptide leads to a loss of conformational selectivity. This also appears to be the case for a recently developed DNA immunogen composed of an unscaffolded 11-fold repeat of A␤1-6 without an intervening spacer, which yielded antibodies that indiscriminately recognized monomeric and oligomeric A␤42 species yet barely bound to higher order fibrillar aggregates (28). Importantly, as shown in Fig. 4D, the same differential recognition profile was shared by the different, randomly assorted pools of antisera raised against monomeric, tetrameric and octameric TrxA␤15. As further shown in Fig. 5, the same conformational selectivity along with the capacity to recognize human A␤ plaques were observed with a rabbit antiserum generated using Freund-adjuvanted Trx(A␤15) 4 as antigen, thus suggesting that this property is neither influenced by the animal host nor the type of adjuvant utilized for immunization.
Additional evidence as to the conformational selectivity of anti-Trx(A␤15) 4 antibodies was provided by an immunodot blot experiment testing their ability to recognize different assembly states of an amyloidogenic polypeptide unrelated in sequence to A␤42. The Ile 84 3 Ser variant of human transthyretin, a strongly amyloidogenic protein (54), was used for this purpose. As shown in Fig. 6, although with a reduced immunoreactivity, the same differential recognition pattern observed with A␤42 (i.e. the binding of higher order fibrillar aggregates and oligomers but not monomers) was observed with anti-Trx(A␤15) 4 antibodies when challenged with different assembly states of Ser 84 -TTR. As expected, no reactivity toward aggregated Ser 84 -TTR was observed under the same conditions with antibodies raised against monomeric TrxA␤15 (not FIGURE 4. Differential recognition of distinct A␤ assembly states by antibodies raised against different Trx(A␤15) n polypeptides. A, representative 2 ϫ 2 m atomic force microscopy images of the indicated A␤42 species; total z-ranges (left to right) are 2, 12, and 13 nm. B, silver-stained SDS-PAGE of a typical A␤42 oligomer preparation; the migration positions of individual A␤42 species and of molecular mass markers are indicated. C, representative results obtained from immunodot blot assays carried out at a fixed concentration (0.75 g/ml) of the indicated affinity-purified anti-Trx(A␤15) n antibodies and 10 pmol of the various A␤42 species (M, monomers; O, oligomers; F, fibrils). D, differential recognition profiles of antibodies from the five pools of antisera in each of the Trx(A␤15) n -treated groups (Group 5, anti-Trx(A␤15); Group 6, anti-Trx(A␤15) 4 ; Group 7, anti-Trx(A␤15) 8 ; see Fig. 2A). Dot blot assays (at least two replicates for each pool) were conducted as described for C and quantified; signal volume intensities for individual A␤42 species within each membrane were averaged. Cumulative data for the three groups are expressed as relative values (Ϯ S.D.) for each A␤42 species, normalized with respect to the sum of the average hybridization signal intensities measured for all A␤42 species. shown), thus further strengthening the notion that these antibodies selectively recognize A␤ monomers.
Anti-Trx(A␤15) 4 Antibody Clears A␤ Pathology in APP Transgenic AD Mice-The immunotherapeutic potential of anti-Trx(A␤15) 4 was evaluated next by stereotaxically injecting this antibody into the hippocampus of 14-month-old APP transgenic AD (Tg2576) mice. Mock immunoglobulins injected into the contralateral hemisphere from mice treated with PBS only served as an internal control for this experiment. Seven days post-injection, histopathological examination and plaque quantitation revealed a significant (p Ͻ 0.01) reduction of A␤ immunostaining in the anti-Trx(A␤15) 4 -injected hemispheres (0.97 ϫ 10 3 Ϯ 0.27 plaques) compared with the mockinjected hemispheres (3.34 ϫ 10 3 Ϯ 0.58 plaques) (Figs. 7 and 8, A-C). A␤-positive plaques were not only absent at the injection site but significantly diminished within the injection penumbra (2 mm anterior/posterior to the injection site) (Figs. 7 and 8, A-C). As shown in Fig. 8D, similar results were obtained with an antibody that selectively reacts with soluble high molecular weight A␤40 and A␤42 oligomers (36). This indicates that the clearing effect of anti-Trx(A␤15) 4 does not depend on the particular primary antibody utilized for immunodetection and suggests that also higher order oligomers (23) are targeted in vivo by the anti-Trx(A␤15) 4 antibody. To verify that these findings were not the result of a competition between anti-Trx(A␤15) 4 and the primary anti-A␤ antibody, we performed alternative histopathological analyses using glial fibrillary antigen protein immunostaining and Campbell-Switzer silver staining to detect astrogliosis and A␤ plaques. An amyloid plaque-associated astroglial response, increasing with the severity of disease progression, has been reported in APP mice (55). As shown in Fig. 9, A and B, astrogliosis and glia-associated plaques were markedly reduced within the anti-Trx(A␤15) 4 antibody injection penumbra compared with the contralateral mock-injected hemisphere. In addition, as revealed by Campbell-Switzer silver staining, there were far fewer plaques in the anti-Trx(A␤15) 4 -injected hemisphere compared with the mock-injected hemisphere (cf. the left and the right panels in Fig. 9C). Both observations are consistent with the immunostaining data obtained with anti-A␤ antibody detection.

DISCUSSION
As revealed by this study, Trx(A␤15) 4 is a soluble derivative of A␤ with good immunogenicity even when formulated with a moderate strength adjuvant, such as alum, that promotes an anti-inflammatory T helper 2-polarized immunoresponse. Even more significant, Trx(A␤15) 4 is the first B-cell epitope A␤ antigen that is shown to be capable of generating antibodies that bind to A␤ fibrils and synaptotoxic A␤ oligomers but not to the presumably physiological monomeric species of A␤. In this regard, anti-Trx(A␤15) 4 antibodies resemble assembly state selective antibodies previously developed by using preassembled A␤42 oligomers (19) or the nitrated A␤40 peptide (35) as antigens. The main difference here is that a selected A␤ fragment, rather than full-length A␤, was used as a recombinant antigen and that antibodies with differential recognition properties were generated simply by varying the copy number of the A␤15 peptide within the display site of thioredoxin. The ability of anti-Trx(A␤15) 4 antibodies to recognize oligomeric and fibrillar aggregates of a different amyloidogenic protein such as TTR, is reminiscent of the "pan-amyloid reactivity" previously observed with antibodies raised against sonicated A␤40 fibrils (34) or A␤40 immobilized on gold nanoparticles (36) as structurally constrained antigens. Various hypotheses have been proposed to explain this cross-reactivity. For example, regardless of their amino acid sequence, all amyloids may share a cross-␤ pattern structure or other common features, such as a unique array of H-bond donors and acceptors (34). Apart from the, as yet, largely unknown structural determinants of this cross-reactivity, what our data suggest is that, when grafted to the active site loop of thioredoxin, the (A␤15) 4 polypeptide is constrained into a conformation that mimics a structural epitope shared by both fibrils and neurotoxic soluble oligomers but not A␤ monomers. Conformational selectivity, coupled with enhanced immunogenicity, may increase the safety and efficacy of anti-Trx(A␤15) 4 as an A␤ aggregate scavenger. This  expectation is supported by the immunotherapeutic performance of the anti-Trx(A␤15) 4 antibodies, revealed by the in vivo experiments in APP transgenic AD (Tg2576) mice. These mice recapitulate many aspects of human AD and provide a means to investigate the temporal and spatial progression of several important neuropathological events, such as amyloid deposition and astrogliosis. Our experimental studies show significant amelioration of the neuropathological phenotype displayed by this AD mouse model following administration of the anti-Trx(A␤15) 4 antibody. Because of the above-mentioned favorable features, Trx(A␤15) 4 lends itself as a prototype B-cell epitope vaccine for AD. Because of its recombinant nature and chemically defined molecular composition, Trx(A␤15) 4 is also a promising candidate for a DNA vaccination approach for the production of monoclonal antibodies for passive immunization purposes as well as for structural analysis of its amyloid-like epitope.