Novel multivalent design of a monoclonal antibody improves binding strength to soluble aggregates of amyloid beta

The heavy and light chains of the different antibodies were cloned into two separate pcDNA3.4 vectors (GeneArt, Regensburg, Germany). The recombinant proteins were expressed as described previously [37]. Briefly, Expi293 cells were transfected with 70% light chain plasmids and 30% heavy chain plasmids using polyethyleneimine as a transfection agent. Seven to 12 days after transfection, the antibodies were purified using an Äkta start system with protein G columns (Cytiva, Uppsala, Sweden). Acetic acid at 0.7% was used as an elution buffer, and absorbance at 280 nm was used to measure the concentrations of purified antibodies. Buffer was exchanged to phosphate-buffered saline (PBS) using Zeba desalting columns (Pierce biotech, Rockford, IL) and the antibodies were stored  at -80 °C until further application.

SDS-PAGE analysis was performed to confirm the purity and the size of the purified proteins. The antibodies were mixed with LDS sample buffer (Invitrogen, Waltham, MA) and loaded onto 4%–12% Bis–Tris protein gels (Invitrogen, Waltham, MA) without adding reducing agents. The gel was then stained with PAGE blue protein solution (Thermo Scientific, Waltham, MA) using a pre-stained protein marker (LI-COR biosciences, Bad Homburg, Germany) as a molecular weight standard.

The structural stability of the generated antibodies under thermal stress was evaluated using Tycho nt.6 instrument (NanoTemper Technologies, München, Germany). Equimolar concentrations of the antibodies were heated in a glass capillary where the temperature increased linearly from 35 °C to 95 °C. Fluorescence intensities at 330 nm and 350 nm were recorded, corresponding to tryptophan fluorescence. The assay was performed under three conditions: immediately after thawing the antibodies, following  2-h incubation at 37 °C and following  72-h incubation at 37 °C.

The recombinant antibodies were labelled with iodine-125 using Chloramine-T as described previously [38]. Briefly, equal amounts of RmAb158 (molecular weight [MW] 150 kDa), Tetra-RmAb158 (MW 200 kDa) and Hexa-RmAb158 (MW 250 kDa) were mixed with the iodine-125 stock solution (Perkin Elmer Inc, Waltham, MA) and 1 mg/ml of Chloramine-T (Sigma Aldrich, Stockholm, Sweden) in PBS. After 90 s, the reaction was stopped by addition of 1 mg/ml sodium meta-bisulphite (Sigma Aldrich, Stockholm, Sweden). Free and unbound iodine was removed from the labelled antibodies using Zeba mini desalting columns having a MW cut-off of 7 kDa (Pierce biotech, Rockford, IL) and the antibodies were eluted in PBS.

The binding properties of the antibodies to protofibrils were studied using LigandTracer Grey (Ridgeview Instruments AB, Uppsala, Sweden). Briefly, a Petri dish was coated with protofibrils in a defined (target) area and placed on a tilted, rotating support. The area opposite to the target area was defined as the background area. The LigandTracer Grey has a low-energy gamma detector mounted above the upper part of the dish. When the buffer containing radio-labeled antibodies was added to the dish, the inclination ensures that the liquid was mainly in the lower part of the dish outside the detection area. During each full rotation, the decay corrected signals from the target and the background areas were recorded for 30 s each. Each recording was delayed for 5 s to allow the buffer to drain from the area being detected. The background signal was subtracted from the target signal to represent specific binding of labeled antibodies to protofibrils. To this end, Petri-dishes (Corning Inc., Corning, NY) were coated overnight at 4 °C with 300 μl of 500 nM Aβ1-42 protofibrils, prepared as described previously [31]. The dishes were blocked with 5% BSA in 1 × PBS for 2 h at room temperature (RT). To establish the affinity and kinetics of the interaction processes, protofibrils were incubated with two antibody concentrations (0.3 and 1 nM in 2 ml of 0.1% BSA/PBS) for 2.5 h and 3 h, respectively. After a total incubation of 5.5 h, the dissociation measurement was initiated by removing the antibody-containing solution and adding 3 ml of 0.1% BSA/PBS buffer to the dish. Data evaluation was performed using TraceDrawer software (Ridgeview Instruments AB, Uppsala, Sweden). The labelling procedure was performed roughly two hours before starting the first incubation.

The binding strength of the recombinant antibodies to Aβ protofibrils and monomers was further investigated by SPR using Biacore 8 K instrument (Cytiva, Uppsala, Sweden). Soluble Aβ1-42 protofibrils, prepared as described previously [31], were immobilized by amine coupling (NHS/EDC, GE kit #BR100633) on a biacore CM5 sensor chip (Cytiva, Uppsala, Sweden). Three-fold dilution series of the recombinant antibodies (ranging from 100 to 1.23 nM) in  PBS-P+ buffer (pH 7.4) (Cytiva, Uppsala, Sweden) were injected using a single cycle kinetic method at a flow rate of 30 μl/min with an association phase of 120 s, followed by a dissociation phase of 7200 s. Injection of buffer was used as a blank control. The binding data were fitted to a 1:1 kinetic model using Biacore Insight Evaluation. The dissociation rate constant was calculated as an average value from 2–5 measurements. Surfaces were regenerated using 3 M MgCl₂ between cycles. For binding to monomers, the recombinant antibodies were immobilized on the mentioned chips at a fixed concentration of 1 μg/ml. A single cycle kinetic method was used to inject a two-fold dilution series of Aβ1-40 monomers (ranging from 4000 to 250 nM) (Bachem, H-1194, Bubendorf, Switzerland). The steady-state kinetic model was used for analysis.

Ninety-six-well half-area plates (Corning Inc., Corning, NY) were coated overnight with Aβ1-42 protofibrils (45 ng/well), prepared as described previously [31], at 4 °C. The plates were blocked the next day with 1% BSA in PBS for 2 h at RT, followed by incubation for an additional 2 h at RT with serial dilutions of RmAb158 (with a starting concentration of 10 nM) and Fab fragment of RmAb158 (with a starting concentration of 1.25 μM). A polyclonal horse-radish peroxidase (HRP)-conjugated anti-mouse IgG (Sigma-Aldrich, Stockholm, Sweden) was added to the plates and incubated for 1 h at RT, followed by signal development with K-blue aqueous TMB (Neogen Corp, Lexington, KY). Absorbance was measured at 450 nm using a FLUOstar Omega microplate reader (BMG Labtech, Ortenberg, Germany).

Synthetic Aβ1-42 (Bachem, Bubendorf, Switzerland) was dissolved in 10 mM NaOH supplemented with 0.005% Tween-20 to a final concentration of 100 µM and stored in aliquots at − 80 °C. For generation of Aβ1-42 aggregates, the stock dilution was diluted in 2 × PBS, pH 7.4, to a final concentration of 50 µM. The 50 µM solution was incubated without shaking for 15 min at 37 °C to generate Aβ1-42 aggregates. For stabilization of the metastable Aβ1-42 aggregates, the aggregates were stabilized covalently by the Photo-Induced Cross-linking of Unmodified Proteins (PICUP) method according to a protocol described previously [39]. Mechanistically, PICUP involves photo-oxidation of Ru3 + in a tris(bipyridyl)Ru (II) complex (RuBpy) to Ru3 + by irradiation with visible light in the presence of an electron acceptor. Briefly, a typical PICUP reaction was performed in a 50 µl reaction volume. To the 50 µM Aβ1-42 aggregate solution, 5 µl of RuBpy (2.5 mM dissolved in water) followed by 5 µl of ammonium persulfate (APS, 10% (w/v) in water) was added by pipetting (final concentrations of RuBpy and APS in the reaction mixture were 0.25 mM and 1%, respectively). The solution was quickly irradiated for 5 s under a general light bulb on the lab bench. Immediately after irradiation, the reaction was quenched by separating the reaction mixture with Zeba spin desalting column 7 k MWCO equilibrated with PBS supplemented with 0.005% Tween-20 (Thermofisher, Waltham, MA). The PICUP stabilized Aβ1-42 aggregates were heat-treated for 5 min at 95 °C prior to separation of Aβ1-42 aggregates by SEC. A Superdex 200 Increase 3.2/300 column (Cytiva, Uppsala, Sweden) was used for size separation of the first batch of Aβ1-42 aggregates (Fig. 6) on a Merck Hitachi D-700 HPLC LaChrom system. The second batch of Aβ1-42 aggregates (Additional file 1: Fig. S3a) was separated using a Superdex 75 column (Cytiva, Uppsala, Sweden). The samples were eluted with PBS-Tween, pH 7.4 (50 mM sodium phosphate, 0.15 M NaCl, 0.1% Tween-20, pH 7.4) at a flow rate of 0.08 ml/min and data obtained at 214 nm. Before sample injection, the quenched reaction mixture was mixed 1:1 with a 2 × mobile-phase buffer (100 mM sodium phosphate, 0.3 M NaCl, 0.2% Tween-20, pH 7.4). Fractions were collected every two minutes and stored at − 20 °C for further analysis. To estimate the size of the separated PICUP Aβ1-42 fractions, western blot analysis was performed. Samples were mixed with Laemlli sample buffer containing reducing agent and heated for 5 min at 95 °C. The denaturated samples were loaded on NuPAGE 12% Bis–Tris gels (Thermofisher, Waltham, MA), and transferred onto nitrocellulose membranes (Bio-Rad, Hercules, CA). The membranes were blocked with 5% dry milk (Bio-Rad, Hercules, CA) in TBS-Tween buffer, followed by incubation with rabbit anti-Aβ42 antibody (Bioarctic AB, Stockholm, Sweden).

The generated antibodies were tested with sandwich ELISA to establish their binding strength to different sizes of the generated cross-linked Aβ1-42 aggregates using the N-terminal specific mouse monoclonal antibody 82E1 (IBL/Tecan Trading AG, Mannedorf, Switzerland) as a positive control. Ninety-six-well half-area plates (Corning Inc., Corning, NY) were coated overnight with 1 μg/ml of the rabbit polyclonal C-terminal-specific Aβ1-42 antibody (Invitrogen, Waltham, MA) at 4 °C. On the next day, the plates were blocked with 1% BSA in PBS for 2 h at RT and incubated with 250 pM of the generated cross-linked Aβ1-42 aggregates of different sizes for another 2 h. Serial dilutions of the recombinant antibodies were added to the plates and incubated for 2 h at RT, followed by detection for 1 h at RT with HRP-conjugated anti-mouse IgG antibody (Sigma Aldrich, Stockholm, Sweden). Signals were developed with K-blue aqueous TMB (Neogen Corp, Lexington, KY) and the absorbance was measured at 450 nm using a FLUOstar Omega microplate reader (BMG Labtech, Ortenberg, Germany). The wells were washed with ELISA washing buffer (PBS with 0.05% Tween-20) between each step. All serial dilutions were made with ELISA incubation buffer (PBS with 0.1% BSA and 0.05% Tween-20). The absorbance of the blank (no addition of recombinant antibodies) was subtracted from the absorbance of the samples. For α-synuclein sandwich ELISA, plates were coated overnight with 1 μg/ml of antibody MJFR-14–6-4–2 (Abcam, Cambridge, United Kingdom) at 4 °C. Following block with 1% BSA in PBS for 2 h at RT, α-synuclein oligomers, prepared as described previously [40], were added at a final concentration of 10 nM and the plates were further incubated for 2 h at RT. Serial dilutions of the recombinant antibodies (RmAb158, Tetra-RmAb158 and Hexa-RmAb158) were added to the plate and incubated for 2 h at RT. The α-synuclein antibody (clone SynO2, recombinantly produced in the lab) was used as a positive control [41]. The signals were developed and detected as described above.

Inhibition ELISA was performed as described previously [34]. Briefly, 96-well half-area plates (Corning Inc., Corning, NY) were coated overnight with 45 ng/well of Aβ1-42 protofibrils at 4 °C followed by blocking with 1% BSA in PBS for 2 h at RT. Serial dilutions of sonicated Aβ1-42 insoluble fibrils, prepared as described previously [31], large cross-linked Aβ1-42 protofibrils (fraction 1 in SEC chromatogram, Fig. 6), cross-linked Aβ1-42 oligomers (fractions 3 and 4 in SEC chromatogram, Fig. 6) and Aβ1-40 monomers (Bachem, Bubendorf, Switzerland) were pre-incubated in a non-binding 96-well plate with HexaRmAb158 at a fixed concentration of 250 pM. After 1.5-h incubation, the mixtures were transferred to the Aβ protofibril-coated plates and further incubated for 15 min at RT. The plates were washed between the steps and the signal was developed and detected as described above.

The mouse neuroblastoma Neuro2a cell line was obtained from ATCC (Manassas, VA) and grown in Minimum Essential Medium (Gibco, Carlsbad, CA) supplemented with 10% fetal bovine serum (Gibco, Carlsbad, CA) at 37 °C with 5% CO₂. The cells were plated in a 96-well plate (Sarstedt cells + , Numbrecht, Germany) at a density of 5000 cells in 100 µl complete medium per well. To eliminate the effect of serum, the cells were starved 24 h prior to the addition of Aβ/antibody mixtures by changing to a serum-free medium. Cells were then treated with a heterogenous mixture of cross-linked Aβ1-42 aggregates (the sample before being passed through SEC, final concentration 500 nM) either alone or in combination with the antibodies (final concentration 150 nM). Cells incubated at standard conditions for 24 h with PBS were used as a negative control and with 0.005% H₂O₂ as a positive control. After the mentioned incubation time, the treatment medium was discarded and 50 μl serum-free medium plus 50 μl MTT reagent (Abcam, Cambridge, United Kingdom) was added in each well and incubated for 3 h at standard conditions, which lead to the formation of formazin proportional to how many cells that are alive. To solubilize the formazin product, MTT solvent (Abcam, Cambridge, United Kingdom) was added and the plates were incubated in the dark for 15 min while shaking. The absorbance was measured at 590 nm using a FLUOstar Omega microplate reader (BMG Labtech, Ortenberg, Germany). Data were obtained from two repetitive experiments where five to six replicates were used for each condition and the absorbance of the blank (cell medium only) was subtracted from the absorbance of the samples.

A scFv, consisting of the heavy variable domain attached through a glycine-serine rich linker [42, 43] to the light variable domain of RmAb158 (Fig. 1a), was attached to the N-terminal end of the RmAb158 heavy chain to create Tetra-RmAb158 (Fig. 1c), or the N-terminal end of both heavy and light chains of RmAb158 to create Hexa-RmAb158 (Fig. 1d). The scFv was attached with an in-house designed linker where prolines were added to the linker to ensure no alpha helices were formed, glycines added to provide flexibility and serines added to give more hydrophilicity [44]. In addition, a dual variable domain (DVD) antibody was designed, where an additional heavy chain variable domain was attached to the N-terminal end of the heavy chain and an additional light chain variable domain was attached to the N-terminal end of the light chain (Fig. 1b).

The four antibodies were expressed in Expi293 cells [37], giving rise to ~ 20 mg of RmAb158 or 2–5 mg of DVD, Tetra or Hexa-RmAb158 per liter of transfected cell cultures. After purification with affinity column chromatography, SDS-PAGE analysis showed bands at 150 kDa (RmAb158), 200 kDa (DVD and Tetra-RmAb158) and 250 kDa (Hexa-RmAb158), confirming the size and purity of the produced recombinant proteins (Fig. 1e).

Structural stability of the generated antibodies under thermal stress was evaluated using the Tycho nt.6 system (NanoTemper Technologies, München, Germany). Inflection temperature which refers to the temperature at which the proteins unfold was determined by measuring fluorescence intensities at 350 and 330 nm. RmAb158 yielded an infection temperature of 79 °C (Fig. 2a). Two inflection temperatures were detected for Tetra-RmAb158 (72.6 °C and 79.4 °C) (Fig. 2a). Hexa-RmAb158 displayed a single inflection temperature of 72.2 °C (Fig. 2a). The slightly lower inflection temperatures detected for Tetra- and Hexa-RmAb158 antibodies can be attributed to the presence of scFvs on the N-terminal ends of these antibodies, as scFvs are more susceptible to thermal shifts than full IgG antibodies. The inflection temperatures of the three antibodies remained constant after incubation at 37 °C for 2 h (Fig. 2b) and 72 h (Fig. 2c), further suggesting the structural stability of the generated antibodies.

To demonstrate that RmAb158 utilizes the avidity effect to ensure strong binding to Aβ protofibrils but not monomers, the Fab fragments of the antibody were generated (Fig. 3a). Using an indirect ELISA setup (Fig. 3b), the binding strength to Aβ1-42 protofibrils was examined using the full RmAb158 antibody (MW 150 kDa), where both binding sites are available, and the Fab fragment of RmAb158 (MW 50 kDa), where only one binding site is available. A much stronger binding to Aβ protofibrils was observed for RmAb158 as compared to its Fab fragment alone (Fig. 3c). Hence, the strong binding that RmAb158 has to the protofibrils requires the binding of both arms to the same protofibril. This implicates the importance of avidity as a decisive factor in RmAb158’s binding selectivity to Aβ aggregates.

The first question on our multivalent antibodies was if they could bind with high avidity to the rather large protofibrils (> 100 kDa in size). The affinity of RmAb158 to Aβ protofibrils has previously been estimated to be approximately 70 pM [35, 45], and the affinity of antibody formats with higher valency was expected to be even higher with a lower rate of dissociation. To verify this, the binding strength of the recombinant antibodies to Aβ1-42 protofibrils was determined with real-time interaction analysis using LigandTracer. For these experiments, iodine-125-labeled antibodies were added to dishes coated with Aβ1-42 protofibrils. The association and dissociation rate constants were obtained by monitoring binding during consecutive incubation steps with two concentrations (300 pM and 1 nM) of the antibodies for 2.5 h and 3 h respectively, followed by a phase where the antibodies were removed to measure dissociation rate. Among the three tested antibodies, Hexa-RmAb158 displayed a much slower dissociation with an almost straight dissociation curve (Fig. 4). Kinetic evaluation was done using a 1:1 curve fit. Here, the association event was detected when the first binding site of an individual antibody binds, while for the appearance of signal reduction for the dissociation phase, all binding interactions of the individual antibody need to be dissociated at the same time and hence the antibody not attached at all. Kinetic evaluation of the binding curve for RmAb158 showed an association rate constant (k_a) of 1.5 × 10⁵ M⁻¹s⁻¹, and a dissociation rate constant (k_d) of 6.5 × 10^–6 s⁻¹. The affinity, K_D, of the antibody to Aβ protofibrils was determined to be 43 pM. Regarding Tetra-RmAb158 binding to Aβ protofibrils, the antibody displayed an association rate constant of 1.9 × 10⁵ M⁻¹s⁻¹ and a dissociation rate constant of 4.5 × 10^–6 s⁻¹, with an affinity K_D of 24 pM (Fig. 4). Kinetic evaluation of Hexa-RmAb158 binding to Aβ protofibrils showed an association rate constant of 2.0 × 10⁵ M⁻¹s⁻¹. For Hexa-Rmab158, the dissociation rate was very low (almost a straight line, see Fig. 4) and for an accurate estimate of the off-rate and affinity, the fitting was performed in two steps where first the dissociation rate was established followed by fitting the association rate constant while using the dissociation rate from the first step. This resulted in a k_d value of 2.1 × 10^–7 s⁻¹, and an affinity K_D of 1 pM (Fig. 4), which were ~ 40 times lower than the k_d and K_D values of RmAb158. These results indicate that after incubating the antibodies for several hours with a protofibril-coated layer, additional scFv units lead to a more stable and higher-affinity binding of the antibodies.

The binding strength of the recombinant antibodies to Aβ1-42 protofibrils was also determined using SPR (Fig. 5a-d). With SPR, it is more difficult to measure the kinetic properties for as long times as with LigandTracer, which is desired in the case of antibodies with slow rate of dissociation. The curve fit to the dissociation phase was good and there was a large difference in the dissociation rate. RmAb158 demonstrated a k_d of 3.86 × 10^–3 s⁻¹ (Fig. 5a). DVD-RmAb158 demonstrated an almost 10 times slower dissociation with a k_d value of around 2.64 × 10^–4 s⁻¹ (Fig. 5b). Binding of Tetra-RmAb158 and Hexa-RmAb158 to Aβ1-42 protofibrils displayed a k_d value of 2.90 × 10^–5 s⁻¹ and 2.56 × 10^–5 s⁻¹, respectively. These k_d values were more than 100 times lower than that of RmAb158 (Fig. 5c, d).

The binding strength of the recombinant antibodies to Aβ1-40 monomers was determined with SPR using the steady-state affinity model. The four antibodies displayed weak binding with a K_D value in the range of 2–4 μM (Fig. 5e-h).

The second part of our study aimed at comparing the binding of the different antibody formats to different sizes of soluble Aβ aggregates.

Stable Aβ1-42 aggregates of different sizes were generated using the PICUP method followed by fractionation through SEC using a Superdex 200 column (GE Healthcare). The HPLC-SEC chromatogram confirmed the presence of aggregates of different sizes collected in different fractions, with fraction-1 having the largest size and fraction-5 having the lowest MW (Fig. 6). Synthetic monomers of Aβ had a size of ~ 4.5 kDa. What is defined as Aβ oligomer and protofibril varies among studies. In the current study, Aβ protofibrils were defined as structures with a size of more than 100 kDa that remain soluble after centrifugation at 16,000 × g, and oligomers as soluble aggregates smaller than this. Therefore, fractions 1 and 2 in the SEC chromatogram were defined as large and small protofibrils respectively. Fraction 3 was defined as large oligomers, fraction 4 as medium-sized oligomers and fraction 5 as small oligomers (Fig. 6).

Using a sandwich ELISA sparsely coated with an antibody that binds the C-terminal end of Aβ, we evaluated the binding strength of the multivalent antibodies to Aβ aggregates that vary in size. We used this type of sandwich ELISA with two-hour-long incubations to detect binding to differently sized cross-linked Aβ1-42 aggregates (Fig. 7a). Results showed that all antibody formats showed a similar response to the large cross-linked aggregates (protofibrils, fractions 1 and 2, > 100 kDa), regardless of the valency (Fig. 7b, c). The method was not sensitive enough to detect differences between very strong binders. However, antibody binding to the smaller aggregates was expected to be substantially reduced and hence differences should be detectable. With such sandwich ELISA setup, we also evaluated the binding strength of the different antibodies to the cross-linked Aβ1-42 fractions 3 and 4 containing the medium-to-large oligomers of around 60–100 kDa. Results showed that the Hexa-RmAb158 bound stronger than the other recombinant antibodies to fractions 3 and 4 (Fig. 7d, e). The Hexa-RmAb158 also displayed some binding to the small cross-linked oligomers at nanomolar concentrations, while the other recombinant antibodies displayed weaker binding (Fig. 7f). The antibody 82E1 has been reported to bind to Aβ of different sizes with similar binding strengths [46, 47] and was used as a control here. The DVD-RmAb158 antibody showed a very weak binding strength to almost all the cross-linked Aβ1-42 fractions in this experiment. This antibody format was more prone to aggregate and required very sensitive handling and was therefore not included in all our experiments. The binding strength of each individual antibody to the different fractions of Aβ from the same experiment is illustrated in Additional file 1: Fig. S2.

To estimate the size of Aβ oligomers that can best generate an avidity effect by binding Hexa-RmAb158, we generated another batch of cross-linked Aβ1-42 oligomers (Additional file 1: Fig. S3a-b) where we could collect fractions containing even smaller oligomers than those generated before (Fig. 6). In the same sandwich ELISA setup as described previously, Hexa-RmAb158 displayed the best binding strength to aggregates with a size range of 50–200 kDa (Additional file 1: Fig. S3c). Nonetheless, the avidity of Hexa-RmAb158 to Aβ1-42 decreased in fractions with size < 50 kDa (Additional file 1: Fig. S3d–f). These results indicate that oligomers of ~ 50 kDa and above are the species of Aβ that can be best detected with Hexa-RmAb158. As expected from the previous results, the other tested recombinant antibodies also bound weaker to small aggregates in the concentration range used (Additional file 1: Fig. S3d–f).

To investigate if Hexa-RmAb158 bound stronger to oligomers and protofibrils  than to insoluble fibrils and monomers, an inhibition ELISA was performed. In this setup, a pre-incubated Aβ aggregate and Hexa-RmAb158 mixture was added to a protofibril-coated ELISA plate (Fig. 8a). The same setup has previously been used to display strong binding of mAb158 to Aβ protofibrils, while having intermediate and weak bindings to fibrils and monomers respectively [31, 35, 36]. In this ELISA setup, the ability of the different preincubated Aβ aggregates to inhibit Hexa-RmAb158 binding to the protofibrils bound to the plate should be related to the antibodies’ binding strength. The Aβ species used in this ELISA setup were sonicated insoluble Aβ1-42 fibrils prepared as described previously [31], cross-linked Aβ1-42 protofibrils (fraction 1 in the SEC chromatogram, Fig. 6), two smaller sizes of cross-linked Aβ1-42 oligomers (fractions 3 and 4 in the SEC chromatogram, Fig. 6) and Aβ1-40 monomers. Because Aβ1-42 is more prone to aggregate, Aβ1-40 was used to assess monomer binding as it is more stable as a soluble monomer [48]. Hexa-Rmab158 displayed a similar strong binding to the three fractions of cross-linked Aβ1-42 (IC50 of 0.9, 1.0 and 2.0 nM, respectively) (Fig. 8b). Importantly, Hexa-RmAb158 exhibited moderate binding strength to sonicated Aβ1-42 fibrils (IC50 227 nM), and weak binding to Aβ1-40 monomers (IC50 834 nM) (Fig. 8b).

To display that Hexa-RmAb158 does not recognize aggregates of other amyloidogenic proteins, a sandwich ELISA was applied (Fig. 9a) detecting α-synuclein oligomers, the pathological aggregates in Parkinson’s disease. The anti α-synuclein antibody (SynO2) was used as a positive control [41]. Hexa-RmAb158, Tetra-RmAb158 and RmAb158 showed no binding to α-synuclein aggregates (Fig. 9b).

The efficacy of the different antibodies in vitor was studied by investigating their ability to protect mouse neuroblastoma Neuro2a cells from Aβ-induced effects on cell metabolism using MTT assay. Since the concentration of the separated fractions of cross-linked Aβ42 aggregates (Fig. 6) was too low to induce reduction in MTT signal, we performed this experiment with a heterogenous mixture of cross-linked Aβ1-42 aggregates, which also included some proportion of monomers. After over 24 h of incubation, 500 nM of Aβ1-42 mixture resulted in a reduction of MTT signal by ~ 53% (MTT signal 47% relative to 100% of PBS-treated cells) (Fig. 10a). The reduced MTT signal was significantly reversed to ~ 65% by RmAb158 (P = 0.002), to ~ 70% by Tetra-RmAb158 (P = 0.003), and to 74% by Hexa-RmAb158 (P < 0.001) (Fig. 10a). The control antibody 82E1 that has a high affinity to the non-toxic monomers as well as the aggregates did not significantly reduce the Aβ1-42 induced toxic effects (Fig. 10a).