Monoclonal antibody with conformational specificity for a toxic conformer of amyloid β42 and its application toward the Alzheimer’s disease diagnosis

Amyloid β-protein (Aβ42) oligomerization is an early event in Alzheimer’s disease (AD). Current diagnostic methods using sequence-specific antibodies against less toxic fibrillar and monomeric Aβ42 run the risk of overdiagnosis. Hence, conformation-specific antibodies against neurotoxic Aβ42 oligomers have garnered much attention for developing more accurate diagnostics. Antibody 24B3, highly specific for the toxic Aβ42 conformer that has a turn at Glu22 and Asp23, recognizes a putative Aβ42 dimer, which forms stable and neurotoxic oligomers more potently than the monomer. 24B3 significantly rescues Aβ42-induced neurotoxicity, whereas sequence-specific antibodies such as 4G8 and 82E1, which recognizes the N-terminus, do not. The ratio of toxic to total Aβ42 in the cerebrospinal fluid of AD patients is significantly higher than in control subjects as measured by sandwich ELISA using antibodies 24B3 and 82E1. Thus, 24B3 may be useful for AD diagnosis and therapy.

Scientific RepoRts | 6:29038 | DOI: 10.1038/srep29038 Results Development of antibody 24B3 that specifically recognizes the toxic conformer of Aβ42. To generate the antibody that specifically recognizes the "toxic turn" at Glu22 and Asp23 of Aβ 42, the seven monoclones previously selected using E22P-Aβ 10-35 13 containing the toxic turn were re-evaluated in detail based on their ability to react with various Aβ 42 mutants with a proline replacement mainly at C-terminal region 8 . In order to precisely evaluate the selectivity for toxic conformer of Aβ 42, the concentration of clones used in this re-evaluation (15~120 ng/mL) was lower than that of the previous study 13 (125~1,000 ng/mL). Enzyme immunoassay (EIA) identified a unique clone, named 24B3, as showing the strongest immunoreactivity with E22P-Aβ 42 among the other mutants in a dose-dependent manner (Fig. 1c). The specificity of 24B3 for E22P-Aβ 42 was much higher than that of 11A1. Moreover, 24B3 barely bound to E22V-Aβ 42, in which valine is known as a turn breaker, while 11A1 reacted weakly with E22V-Aβ 42 (Fig. 1c). These results suggest that 24B3 is more conformation-specific for the toxic Aβ 42 conformer compared with 11A1 since the proportion of toxic conformer of Aβ 42 could be enhanced by E22P mutation.

Prevention of Aβ42-induced neurotoxicity by 24B3 occurs through sequestration of Aβ42 oligomers.
To determine the effects of 24B3 on Aβ 42-induced neurotoxicity, we performed the MTT assay on SH-SY5Y human neuroblastoma cells, a neuronal cell culture model. Not only 24B3 but also 11A1 significantly rescued the neurotoxicity induced by both Aβ 42 and E22P-Aβ 42 as a toxic conformer surrogate 8 , whereas neither 4G8 nor 82E1 (anti-N-terminus of Aβ 42) could rescue (Fig. 2a,b). Notably, the protective effects of 24B3 were slightly higher than that of 11A1. All of these antibodies did not affect cell viability (Fig. 2c). Because we previously reported the neuroprotective potential of 11A1 on rat primary neurons 21 , the additional experiments were carried out. We then obtained similar results using primary neuronal cultures originated from rat ( Supplementary Fig. 1). The slight inhibition of cytotoxicity by 11A1 was also observed with PC12 cells 13 . These observations are consistent with previous findings that A11 rescues Aβ 42-induced toxicity in SH-SY5Y cells, whereas 6E10 does not 22 , raising the possibility that 24B3 may target the oligomeric species of Aβ responsible for neurotoxicity in a manner similar to A11 although the haptens of these antibodies are different from each other. These results suggest that conformation-specific antibodies may be more beneficial in treating AD pathology than sequence-specific antibodies, and that Aβ aggregates sequestered by 24B3 may contain oligomers formed by the toxic Aβ 42 conformer.
Since both Aβ 42 and E22P-Aβ 42 induced neurotoxicity in a time-dependent manner 21 , we determined by dot blotting how 24B3 detects the toxic conformer of Aβ 42. Regarding the immunoreactivity of 24B3, the relative intensity of E22P-Aβ 42 gradually increased to a maximum after 4 h of incubation, whereas Aβ 42 reacted more slowly than E22P-Aβ 42 (Fig. 2d,e). It should be noted that the relative quantification of reactivity of 24B3 against Aβ 42 was carried out under the longer exposure of blots due to its moderate affinity. It is noteworthy that the potent immunosignal of E22P-Aβ 42 against 24B3 immediately after dissolution means more rapid formation of the toxic conformer by E22P-Aβ 42 compared with Aβ 42. In contrast, the relative intensity by 4G8 against Aβ 42 was potent even after 0 h of incubation, and it remained almost constant up to 24 h (Fig. 2d,e). Under the longer exposure in the case of E22P-Aβ 42 treated with 4G8 because of its weak affinity, the similar results to Aβ 42 treated with 4G8 were obtained (Fig. 2d,e). These results suggest that the Aβ 42 oligomers responsible for neurotoxicity can be attributed to those detected by 24B3, but not those detected by 4G8.
Synthesis and characterization of the E22P-Aβ42 dimer as a toxic oligomer surrogate. We recently validated optically active L,L-diaminopimelic acid (DAP) 23 as a useful linker near the intermolecular β -sheet region (Ala30) of E22P-Aβ 40 24 . This strategy was applied to the synthesis of the E22P-Aβ 42 dimer as a toxic oligomer model for Aβ 42. The E22P mutation not only enhanced the neurotoxicity of Aβ 42 ~10-fold 25 , but also increased the ratio of toxic conformation of Aβ 42 7 . According to our proposed dimer model (Fig. 1b), Val40 in the C-terminal hydrophobic core was replaced with the linker. Solid-phase synthesis using Fmoc-L,L-DAP as a substitute for Val40 provided a sufficient amount of E22P-Aβ 42 dimer (Fig. 1b) with high purity (6.0% yield, > 98% purity, Supplementary Fig. 2).
The aggregative ability of the E22P-Aβ 42 dimer with a covalent linker at Val40 was evaluated using thioflavin-T (Th-T), a reagent that fluoresces when bound to Aβ aggregates, and transmission electron microscopy (TEM). E22P-Aβ 42 aggregated with a lag time of ~4 h and a maximum fluorescence value after being incubated for 96 h (Fig. 3a), and its velocity for aggregation was higher than that of Aβ 42, as previously reported 25 . In contrast, the fluorescence of the E22P-Aβ 42 dimer remained almost unchanged even after a 96 h incubation (Fig. 3a). Given the moderate increase in the fluorescence after incubation for 168 and 336 h, these time-point samples were subjected to TEM analysis. As shown in Fig. 3b, we observed short and globular aggregates (oligomers) predominantly in the E22P-Aβ 42 dimer, unlike the typical fibrils found in E22P-Aβ 42.
Moreover, circular dichroism (CD) spectrometry was measured to analyze the secondary structure of the E22P-Aβ 42 dimer. A positive peak at ~190 nm and a negative peak at ~210 nm began to increase gradually after dissolution, suggesting that a random structure transformed into a β -sheet architecture in the E22P-Aβ 42 dimer in a manner similar to E22P-Aβ 42 (Fig. 3c).
The ability of the E22P-Aβ 42 dimer to form oligomers was further studied using size exclusion chromatography. As a control reference, the peaks corresponding to the monomer of E22P-Aβ 42 and its 6~8-mer were observed immediately after dissolution, but these peaks disappeared almost completely after 1 h incubation (Fig. 3d). This observation suggests the formation of insoluble fibrillar aggregates of E22P-Aβ 42, as observed in Fig. 3b. In contrast, the E22P-Aβ 42 dimer formed stable oligomers (comprising 6~8-mers) during incubation for ~24 h even after dissolution (Fig. 3d). It should be noted that the peak of E22P-Aβ 42 dimer did not appear to be observed even at the initial time point (Fig. 3d). Actually, we confirmed the presence of E22P-Aβ 42 dimer in reverse-phase HPLC on ODS column ( Supplementary Fig. 2b). Once dissolved in the PBS solution, the E22P-Aβ 42 dimer could intrinsically form stable oligomers during elution in size exclusion chromatography. Alternatively, in the size exclusion chromatography using Superdex75 10/300GL column, E22P-Aβ 42 dimer itself might not be able to give a sharp peak because of adopting various conformations. Similar phenomena were also observed in Aβ 42 monomer using reverse-phase HPLC on ODS column under the acidic condition; Aβ 42 was observed only as a broad peak 26 .
The neurotoxicity of the E22P-Aβ 42 dimer was measured using the MTT assay on human neuroblastoma SH-SY5Y cells. After being incubated for 16 h when the stable oligomer of the E22P-Aβ 42 dimer is suspected to be predominant, the viability of cells treated with the E22P-Aβ 42 dimer was lower than that of E22P-Aβ 42 at 1~10 μ M, indicating that the E22P-Aβ 42 dimer is more neurotoxic than E22P-Aβ 42 (Fig. 3e). Overall, the E22P-Aβ 42 dimer can form stable oligomers with a β -sheet structure, suggesting that the E22P-Aβ 42 dimer is a suitable model for studying toxic Aβ 42 oligomers. However, additional oligomer models such as trimer or tetramer should be examined to conclude which oligomeric species are toxic.
Binding affinity of 24B3 for the E22P-Aβ42 dimer. To investigate the potential of 24B3 to recognize toxic Aβ 42 oligomers, we performed an EIA test. 24B3 showed stronger immunoreactivity against the E22P-Aβ 42 dimer than E22P-Aβ 42; its specificity for the E22P-Aβ 42 dimer was much higher than that of 11A1 (Fig. 4a,b). In contrast, the relative affinity of 4G8 for the E22P-Aβ 42 dimer was weak (Fig. 4c).
Surface plasmon resonance (SPR) was employed to evaluate the binding constant of 24B3 for Aβ , in which Aβ was immobilized on a sensor chip based on a biotin-streptavidin interaction. The biotinylation of Aβ 42, E22P-Aβ 42, and the E22P-Aβ 42 dimer did not affect their aggregation profiles and neurotoxicity, respectively (Fig. 3a, Supplementary Fig. 3). 24B3 bound 6.5-fold more strongly to the E22P-Aβ 42 dimer (K D = 1.8 nM) than 11A1 (K D = 10 nM) ( Table 1, Supplementary Fig. 4). In addition, consistent with the EIA test ( Fig. 4a,b), the binding affinity of 24B3 for E22P-Aβ 42 (K D = 3.1 nM) was ~8-fold higher than that of 11A1 for E22P-Aβ 42 (K D = 24 nM). Taken together with the lower binding affinity of 24B3 than 11A1 for Aβ 42 (24B3: K D > 100 nM, 11A1: K D 15 nM), these results strongly suggest that 24B3 specifically recognizes toxic oligomers of Aβ rather than monomeric Aβ .   Ratio of toxic conformer to total Aβ42 as a potential biomarker for AD pathology in CSF. It is becoming important to identify and validate biomarkers in biological fluids for AD diagnosis, especially for prediction of patients with mild cognitive impairment (MCI) who will convert to AD. Given the potential of 24B3 as a specific probe for toxic Aβ 42 oligomers, we developed a novel sandwich ELISA using a combination of anti-N-terminus antibody (82E1) for capture and 24B3 conjugated with horseradish peroxidase for detection. To determine if the ratio of toxic conformer to total Aβ 42 is correlated with the AD pathology in the brain 7 , we performed ELISA on human CSF samples from 13 patients with AD/MCI and 12 age-matched controls ( Table 2); we combined the patients with MCI and those with AD to compare with the age-matched controls, since all the MCI patients in this study were those with "MCI due to AD" 27 . Accordingly, the ratio of toxic conformer to total Aβ 42 in AD/MCI patients was significantly higher than that in the age-matched controls (p = 0.0135, Fig. 5a). In contrast, the difference in total Aβ 42 as one of the conventional biomarkers 28 was not significant (p = 0.0863, Fig. 5c). Intriguingly, the absence of a significant difference in the amount of toxic Aβ 42 conformer between AD/ MCI and control groups (p = 0.4620, Fig. 5b) might be due to the disappearance of toxic oligomers by moving onto fibrillization of Aβ 42 monomer. Indeed, the reduction of Aβ 42 levels in CSF are originated from plaque formation by enhanced aggregation in the cerebral parenchyma as well as the disturbed clearance of Aβ from the cerebral parenchyma into CSF 28 . These results signify the proportion of toxic conformer of Aβ 42 rather than its amounts. Considering its potential inaccuracy when to judge the diagnosis based on only the amount of total Aβ 42 17,18 , the ratio of toxic conformer of Aβ 42 to total Aβ 42 could be a better substitute for precise diagnosis of AD.

Discussion
Immunotherapy-mediated removal of Aβ aggregates is one of the first disease-modifying therapies for AD. However, several anti-Aβ antibodies recently tested in clinical trials failed 29 , possibly because they were evaluated in later stage AD patients, when neuronal cells have been irreversibly damaged. In this stage, the elimination of Aβ 42 oligomers is no longer beneficial. Several ongoing immunotherapy trials involving patients diagnosed with preclinical AD or who carry a genetic mutation that induces AD symptoms may be more successful.
Our ideas are based on the concept that selective removal of toxic Aβ oligomers is important in immunotherapy since non-toxic Aβ plays a physiologically important role in synapse communication 30 and regulation of glucose metabolism 31 . We developed the conformation-specific monoclonal antibody 24B3, which specifically recognizes the toxic conformer of Aβ 42 in toxic oligomers (Fig. 4), and demonstrated a correlation between the ratio of toxic conformer to total Aβ 42 and the pathology of AD (Fig. 5). Moreover, 24B3 suppressed significantly the Aβ -induced cytotoxicity in vitro, while sequence-specific antibodies such as 4G8 with weaker affinity for the E22P-Aβ 42 dimer model did not (Figs 2 and 4). Several failures of passive immunization in clinical trials might be due to the insufficient affinity of sequence-specific antibodies 32 for toxic Aβ 42 conformers. Recent unexpected findings on the neuronal hyperactivity induced from immunotherapy using a sequence-specific antibody (3D6) reaffirm the significance of confirmation-specific antibody 33 . Otherwise, they might be ascribed to the unintended elimination of physiologically necessary non-toxic conformers of Aβ 42 7,34,35 , though other reasons such as inadequate timing and period to treat with anti-Aβ antibodies have already been proposed 36 .
E22P-Aβ 42 as a toxic conformer surrogate readily forms toxic oligomers to inhibit long term potentiation (LTP) 37 . Provided that Aβ 42 oligomerization involves the formation of various transient or intransient assemblies, which are on-or off-pathway aggregation products, it is indispensable to develop an "off-pathway" Aβ 42 oligomer model for developing anti-Aβ drugs with few side effects. To the best of our knowledge, this is the first report on the generation and characterization of an Aβ 42 dimer model, which is connected at the C-terminal hydrophobic region. Practical synthesis of Aβ dimers has thus far been limited to Aβ 40 dimers 4,23,38 , partly due to the intrinsic and potent ability of Aβ 42 to aggregate during synthesis and preparation. Cross-linkage within this region has never been achieved in spite of its significance in oligomerization [10][11][12] . Although one study has reported the synthesis of a dityrosine cross-linked Aβ 42 dimer at Tyr-10 38 , its biological activity was not tested due to insufficient yield. Other oligomer models with high molecular weight such as ADDLs (~24-mer) 39 are also considered off-pathway aggregates. Dimers, trimers, and tetramers of Aβ 42, as prepared by photo-induced cross-linking of unmodified protein (PICUP) technology using 2,2′ -bipyridyl-dichlororuthenium(II) hexahydrate as a catalyst, were mainly bound covalently at Tyr10 40 , although it seems difficult to isolate these oligomers in a pure form. Recently, unique synthesis of N-terminal-tethered triple Aβ fragment (Aβ 25-35) has been reported 41 as a trimeric model, which may be another minimal unit of toxic oligomers (2 or 3 × n-mer) of Aβ 42. When we take into account these models as candidates of toxic oligomers, further investigation will be required to clarify the significance in vivo, though the pathology of the genetically modified mice to produce Aβ dimer cross-linked at Ser8 in N-terminal region was recently reported 42 .
We identified 24B3 based on the ability to bind the conformer possessing the turn structure at residues 22 and 23 using some proline-substituted mainly at C-terminal core region mutants of Aβ 42 (Fig. 1). Our findings indicate that 24B3 could be more favorable to the toxic conformer of Aβ 42 with the turn at residues 22 and 23 than to 11A1. It should be noted that 11A1 may bind various conformers with turns at other residues, even though 11A1 is not proline-specific like 24B3. The enhanced reactivity of 24B3 with the E22P-Aβ 42 dimer compared with E22P-Aβ 42 (Fig. 4) suggests the stabilization of the toxic Aβ 42 oligomer by formation of the C-terminal core and intermolecular β -sheet. Furthermore, the time-dependent reactivity of 24B3 towards Aβ 42 in solution (Fig. 2) reflects the amount of toxic Aβ 42 oligomers formed during incubation.
In previously reported sandwich ELISAs for Aβ oligomers 43 , the same anti-Aβ antibodies were used for both capture and detection based on the idea that oligomers have multiple reactive sites. However, it is likely to be difficult to discriminate toxic Aβ oligomers from less-toxic fibrillar Aβ aggregates using such ELISAs, and their validity has thus been questioned 44 . It is reasonable to use the conformation-specific antibody for detection after capturing total Aβ using 82E1 in sandwich ELISA. Decreased Aβ 42, increased total tau, and increased phosphorylation of tau, are currently the most accepted biomarkers for diagnosing AD (probable, possible, or definite AD) 28 . There have also been intensive studies on other biomarkers such as Aβ -related or tau-related molecules in order to increase diagnostic validity by modifying sensitivity and specificity. However, outliers largely affect the validity in these conventional biomarkers, and there have been no studies on the qualitative difference between various Aβ conformers in spite of accumulating structural studies of Aβ . Our group proposed the significance of the ratio of the toxic conformer to total Aβ 42 in AD pathogenesis 7 using solid-state NMR analysis, which demonstrated that a ~2-fold increase in the ratio of toxic conformer of E22K-Aβ 42 to wild-type Aβ 42 can in part explain the extensive aggregative ability and neurotoxicity of E22K-Aβ 42 25 . Although the ratio of Aβ 42 to Aβ 40 has been suggested as another biomarker 45 , such a biomarker might correlate more preferably with the amount of senile plaque containing less-toxic fibrils rather than with the amount of toxic oligomers.
In summary, the target of a novel conformation-specific monoclonal antibody 24B3 could be oligomers that include the toxic conformer of Aβ 42, and the ratio of toxic conformer to total Aβ 42 could be an alternative evaluation criterion toward accurate diagnosis of AD. Although the analysis with a larger number of CSF samples is indispensable for further validation, the development of a less invasive test using plasma is an attractive goal for the future. 24B3 could be promising as a diagnostic tool of AD because of its superior affinity for toxic oligomers consisting of toxic conformation of Aβ 42, which is likely to be formed at an earlier stage before AD symptoms. Moreover, it could be applied to therapeutics during the early stages of AD.

Thioflavin-T (Th-T) assay.
The aggregative ability of each Aβ was evaluated with a previously described thioflavin-T (Th-T; Sigma) fluorescence assay 25 . Aβ was dissolved in 0.1% NH 4 OH at 250 μ M, followed by 10-fold dilution with phosphate buffered saline (PBS; 50 mM sodium phosphate, and 100 mM NaCl, pH 7.4) to a final concentration of 25 μ M. After incubating at 37 °C for the desired period, 2.5 μ L of the reaction solution was added to 250 μ L of 5.0 μ M Th-T in 5.0 mM Gly-NaOH (pH 8.5), followed by the measurement of fluorescence at 430 nm excitation and 485 nm emission using a microplate reader (Fluoroskan Ascent; Thermo Scientific).

Transmission electron microscopy (TEM). The Aβ aggregates after incubation for the desired period in
the Th-T assay were examined under a H-7650 electron microscope (Hitachi). The experimental procedure has been described elsewhere 25 . After the supernatant was removed from the pellets, the resultant aggregates were then suspended in water (100 μ L) by gentle vortex mixing, and centrifuged at 6,600 rpm for 1 min. These suspensions were applied to a 200 mesh Formvar-coated copper grid (Nissin EM), and allowed to dry in air for 5 min after being negatively stained for several seconds with 2% uranyl acetate and subsequently subjected to microscopy. Circular dichroism (CD) spectrometry. The secondary structure of the Aβ dimer was estimated by CD spectrometry (J-805; JASCO) using a 0.1 mm quartz cell, as described elsewhere 37 . The Aβ solution (25 μ M) Scientific RepoRts | 6:29038 | DOI: 10.1038/srep29038 prepared above was incubated at 37 °C. An aliquot was loaded into the quartz cell, and CD spectra were recorded at 190-260 nm. The spectra of Aβ are shown after subtraction of the spectrum for the vehicle alone.
Size exclusion chromatography. The Aβ solution (25 μ M) was incubated at 37 °C. After the solution was collected periodically and centrifuged at 17,860 × g at 4 °C for 10 min, the supernatant was analyzed by size exclusion chromatography on the Superdex75 10/300GL column (10 mm i.d. × 300 mm; GE Healthcare) with elution at 0.6 mL/min by filtered-and degassed-PBS (pH 7.4), attached to a Waters LC system with a 2489 UV/Visible detector and 1525 binary HPLC pump controlled by Empower TM 3 software (Waters), as described elsewhere 24 . The peptide was detected by absorbance at 220 nm. Calibration curves of size exclusion columns were constructed using dextran standards (Mp: mean peak molecular weight, 43500, 21400, 9890, 4440) (Sigma) together with Blur dextran 2000 (GE Healthcare) as an indicator of the void volume (V 0 ).
MTT assay on SH-SY5Y cells. Human neuroblastoma, SH-SY5Y cells, maintained in a 1:1 mixture of Eagle's minimum essential medium (Wako) and Ham's F12 medium (Wako) containing 10% fetal bovine serum (Biological Industries), were used as a neuronal cell model to estimate the neurotoxicity of each Aβ with a slight modification to the previously described method 24 . In brief, each Aβ was dissolved in 0.1% NH 4 OH to generate a 10X stock solution. The resultant peptide solution (10 μ L) was diluted with 0.1% NH 4 OH to appropriate final concentrations in medium before being added to 100 μ L of the culture medium of near-confluent cells (10 4 cells/well) after one or two overnight incubation. In the case to test the effect of antibodies on the cells, the culture medium was replaced with fresh medium containing pre-incubated (30 min) Aβ solution with antibodies. After being treated at 37 °C for 16 or 48 h, 10 μ L of 5 mg/mL MTT (Sigma) was added to cells, followed by incubation for 4 h at 37 °C. After removing the medium, 100 μ L cell lysis buffer (10% SDS, 0.01 M NH 4 Cl) was subsequently added to the cells. The resulting cell lysate was subsequently incubated overnight in the dark at room temperature before absorbance measurements were made at 595 nm with a microplate reader (MultiScan JX; Thermo Scientific). Absorbance obtained by the addition of vehicle (0.1% NH 4 OH) was taken as 100%.
MTT assay on rat primary neurons. Animals were treated in accordance with guidelines by the Kyoto University Animal Experimentation Committee and guidelines by The Japanese Pharmacological Society. This study was approved by Kyoto University Animal Experimentation Committee. Neuronal cultures were obtained from the cerebral cortices of fetal Wistar rats (Nihon SLC) at 17-19 days of gestation as described previously 21 . Cultures were maintained in Neurobasal medium with 2% B-27 supplement, 25 μ M sodium glutamate, and 0.5 mM L-glutamine at 37 °C in a humidified atmosphere of 5% CO 2 . After 4 days in culture, medium was replaced with sodium glutamate-free Neurobasal medium. Only mature cultures (8~12 days in vitro) were used for the experiments. In all experiments, B-27 supplement without antioxidants was utilized during the treatment of Aβ 42 as described previously 21 .
Neurotoxicity was assessed by MTT assay according to the previously reported protocol 21 . After 30 min of pre-incubation on ice for Aβ 42 solution (10 μ M) in 0.1% NH 4 OH, followed by 10-fold dilution with Neurobasal medium, the medium containing Aβ was added to the cell culture for replacement. After incubation of Aβ 42 at 37 °C for 96 h, the culture medium was replaced with medium containing 0.5 mg/mL MTT, and cells were incubated for 30 min at 37 °C. 2-Propanol was added to lyse the cells, and absorbance was measured at 595 nm with an absorption spectrometer (microplate reader model 680, Bio-rad). The medium of vehicle treatment for each experiment contained 0.01% NH 4 OH. The absorbance obtained by the addition of vehicle (0.1% NH 4 OH) was taken as 100%.

Enzyme immunoassay (EIA).
A 96-well Maxisorp plate (Nunc) was incubated with each Aβ (2.5 μ g/well) dissolved in 50 mM sodium carbonate for 2 h at room temperature, followed by treatment for blocking with 5% bovine serum albumin at 4 °C overnight, as described previously 13 . Briefly, after incubation with each clone obtained in the previous work 13 for 1 h at room temperature, the plate was treated with a horseradish peroxidase-coupled anti-mouse IgG antibody (IBL), and quantified using o-phenylenediamine dihydrochloride substrate (Sigma) before measurements at 492 nm with a microplate reader (MultiScan JX; Thermo Scientific).

Surface plasmon resonance (SPR).
Binding affinity tests were performed using a BIAcore X100 biosensor (GE Healthcare), as previously described 13 . In brief, sensor chip SA, to which streptavidin was anchored, was preconditioned by running HBS-EP buffer (GE Healthcare). Each HFIP-treated biotinylated Aβ dissolved in HBS-EP buffer (1 nM) was immobilized on the chip according to the manufacturer's protocol. The antibody was dissolved in HBS-EP buffer and injected over the chip-immobilized Aβ at a flow rate of 5 μ L/min. Either 20 mM Gly-HCl buffer (pH 2.0) or 5 M guanidine hydrochloride was used as the regeneration buffer. The values of response units (RU) obtained from a sample cell minus the RU obtained from a reference cell were used for analysis. Association and dissociation data were collected with flowing running buffer for 180 s and 360 s, respectively. The association and dissociation rate constants, and dissociation constant (k a , k d , and K D ) were calculated using a serial dilution series of antibody concentrations according to BIAevaluation 3.1 software (GE Healthcare). Dot blotting. One microliter of each Aβ solution (25 μ M) was applied to a nitrocellulose membrane after incubation at 37 °C (0.2 μ m pore size; Bio-rad) as previously described 21 . After blocking in 5% non-fat milk dissolved in Tris-buffered saline containing 0.1% Tween-20 overnight at 4 °C, the membrane was treated with 24B3 or 4G8 (0.5 μ g/mL) for 1 h at room temperature before being incubated with secondary antibody. Development was performed with enhanced chemiluminescence and quantified using LAS-4000 (Fujifilm). ImageJ 1.42 (NIH) software was used to quantify the blots.
Scientific RepoRts | 6:29038 | DOI: 10.1038/srep29038 Subjects and collection of CSF samples. This study was conducted in accordance with the principles of Helsinki Declaration. The study was approved by the University Ethics Committee of Kyoto Prefectural University of Medicine. All subjects provided written informed consent to participate in the study. We collected CSF samples from 13 patients [aged 68~84 (mean ± SD, 77.1 ± 5.6) yr] with clinically diagnosed AD (n = 8) or MCI (n = 5), and 12 age-matched control subjects [aged 61~84 (mean ± SD, 72.3 ± 5.8) yr; see Table 2 for characteristics of study participants]. At the time of diagnosis, a full clinical history was taken, and physical and neurological examinations, Mini-Mental State Examination (MMSE), routine blood analyses, and magnetic resonance imaging (MRI) of the brain were performed for all subjects. The patients with AD met the criteria for probable AD defined in the diagnostic and research criteria established by the National Institute on Aging (NIA) and the Alzheimer's Association (AA) 46 . All MCI patients were non-demented, not fulfilling the criteria for probable AD dementia 46 , but met the criteria of "MCI due to AD" defined by NIA and AA 27 . There were no significant differences in age between the AD/MCI group and the control group. None of the control subjects had memory complaints or any other cognitive symptoms.
Fresh CSF samples were obtained from the enrolled subjects and then immediately stored at − 80 °C until used for immunoassays. All lumbar punctures were performed in the early morning to exclude the effects of daily fluctuation in the levels of Aβ in CSF 47 .
Enzyme-linked immunosorbent assay (ELISA). Microtiter plates (96 wells) were coated with 100 μ L/well of 50 mM sodium carbonate containing 82E1 (IBL) and allowed to adhere overnight at 4 °C. Plates were washed with PBS and blocked for overnight at 4 °C with 200 μ L/well of 1% (w/v) bovine serum albumin in PBS containing 0.05% NaN 3 . After two washes with PBS containing 0.02% Tween-20 (PBS-T), 100 μ L of CSF was serially diluted in 1% bovine serum albumin in PBS-T before being added in triplicate to wells before incubation overnight at 4 °C. After four washes with PBS-T, each well was treated with 100 μ L of horseradish peroxidase-conjugated 24B3 for 1 h at 4 °C. ELISA signals were detected by chemiluminescence using an enhanced chemiluminescent substrate (SuperSignal ELISA Femto Maximum Sensitivity, Thermo Scientific), and then measured with a microplate luminometer (SpectraMax Pro, Molecular Devices).
Synthetic E22P-Aβ 40 dimer 24 was used as a standard protein because the E22P-Aβ 40 dimer also containing the toxic turn at Glu22 and Asp23 was more stable than the E22P-Aβ 42 dimer. The concentration of the E22P-Aβ 40 dimer was determined using the Bradford assay (Bio-Rad). The amount of total Aβ 42 in CSF was determined by sandwich ELISA with a human β Amyloid ELISA Kit of Aβ 42 [Cat# 27711, human amyloid β (1-42) including (X-42)] (IBL) according to the manufacturer's protocols.
Statistical analysis. All data are presented as mean ± s.e.m. The differences were analyzed with one-way analysis of variance (ANOVA) followed by Bonferroni's test or unpaired Student's t-test. These tests were implemented within GraphPad Prism software (version 5.0d). p values < 0.05 were considered significant.