Conformation‐specific antibodies against multiple amyloid protofibril species from a single amyloid immunogen

Abstract We engineered and employed a chaperone‐like amyloid‐binding protein Nucleobindin 1 (NUCB1) to stabilize human islet amyloid polypeptide (hIAPP) protofibrils for use as immunogen in mice. We obtained multiple monoclonal antibody (mAb) clones that were reactive against hIAPP protofibrils. A secondary screen was carried out to identify clones that cross‐reacted with amyloid beta‐peptide (Aβ42) protofibrils, but not with Aβ40 monomers. These mAbs were further characterized in several in vitro assays, in immunohistological studies of a mouse model of Alzheimer's disease (AD) and in AD patient brain tissue. We show that mAbs obtained by immunizing mice with the NUCB1‐hIAPP complex cross‐react with Aβ42, specifically targeting protofibrils and inhibiting their further aggregation. In line with conformation‐specific binding, the mAbs appear to react with an intracellular antigen in diseased tissue, but not with amyloid plaques. We hypothesize that the mAbs we describe here recognize a secondary or quaternary structural epitope that is common to multiple amyloid protofibrils. In summary, we report a method to create mAbs that are conformation‐sensitive and sequence‐independent and can target more than one type of protofibril species.


| INTRODUCTION
Aggregation of proteins or peptides into amyloid fibrils is a characteristic pathological feature observed in many different diseases including type 2 diabetes mellitus (T2DM), Alzheimer's disease (AD), Parkinson's disease (PD), and Huntington's disease (HD). [1][2][3] Amyloid aggregates present in the brain are associated with a reduction in the efficiency of coordinated synaptic transmission, loss of synaptic plasticity and contribute to cognitive impairment 4,5 in AD [6][7][8] and other so-called tauopathies, 9,10 PD, 11 HD 12 and frontotemporal lobar degeneration (frontotemporal dementia, clinical amyotrophic lateral sclerosis and motor neuron disease). 13,14 Misfolded protein aggregates outside of the central nervous system result in amyloidosis syndromes, such as light-chain amyloidosis, familial amyloid cardiomyopathy, familial amyloid polyneuropathy 15 and T2DM. 16 Amyloid proteins or peptides polymerize to form a cross-β sheet structure and progressively self-aggregate into soluble protofibrils, insoluble fibrils and eventually deposit as amyloid plaques in tissue. 17 However, normally folded species of these proteins or peptides have important biological functions, the amyloid fibrils and their prefibrillar aggregates exhibit toxicity. 18,19 Antibodies that lead to clearance of the toxic forms of amyloid are likely more useful than those that target the monomeric amyloidogenic species. In fact, accumulating evidence suggests that prevention of aggregation of pathogenic amyloid species would prevent disease progression. 20 Clinical applications of antibodies that target amyloid conformations are primarily limited to AD. Passive immunization with monoclonal antibodies (mAbs) directed at amyloid beta-peptide (Aβ42) aggregates has shown interesting preliminary results. 21 BAN2401 (BioArtic Neuroscience AB, Eisai Co., Ltd., currently in phase II) was obtained by immunizing mice with Aβ42 (E22G) protofibrils and recognizes early aggregates with low affinity for fibrils or monomers. 22,23 Crenezumab (Genentech, Inc, currently in phase III) binds the middle domain and shows similar binding for Aβ42 monomeric, oligomeric and fibrillar species. 24 Finally, produced through a "reverse translational medicine" approach, aducanumab (Biogen, Inc, currently in phase III) was isolated from B-cells of healthy advancedage donors, who are hypothesized to harbour naturally developed antibodies against Aβ. Aducanumab selectively targets aggregates and dose-dependently reduces amyloid deposition. 25,26 We recently described a platform technology based on the use of the chaperone-like amyloid-binding protein (CLABP) NUCB1 to "cap," detoxify, and stabilize soluble intermediate protofibrils originating from various amyloidogenic proteins, such as Aβ42, α-synuclein, transthyretin and the human islet amyloid polypeptide (hIAPP). 27 Based on the hypothesis that there are common core protofibril conformations we tested the possibility that our technology could be used to develop antibody tools to detect these similarities. Here, we demonstrate that NUCB1-capped hIAPP protofibrils can be used as immunogen to produce mAbs against protofibrils derived from a different amyloid protein, Aβ42. We show that NUCB1-capped amyloid could serve as a platform technology for the discovery of therapeutic antibodies that bind elements unique to structured amyloid intermediates.

| Peptide preparation
The hIAPP (Phoenix Pharmaceutics) was solubilized in HFIP at 1 µg/ µl, dried and stored at −80°C. On the day of the experiment, hIAPP was solubilized in 20 mmol/L sodium phosphate buffer, pH 7.6 to a final concentration of 10 µmol/L, incubated at 25°C and tested at different time points. Aβ40 or Aβ42 (American Peptide) synthetic peptide was solubilized in HFIP at 1 µg/µl, dried and stored at −80°C. On the day of the experiment, Aβ42 was reconstituted in 2 mmol/L NaOH to 1 µg/µl, dried and diluted in 20 mmol/L sodium phosphate buffer, pH 8.0 to a final concentration of 10 µmol/L. The peptide was incubated at 37°C and tested at different time points.

| Immunogen preparation and immunization strategy
The mtNUCB1-capped hIAPP complex was prepared by co-incubating mtNUCB1 (10 µmol/L) and hIAPP (33 µmol/L) peptide at 37°C for 3 hours while stirred. The capped-protofibril containing solution was then applied to a Superdex200 26/60 PG size exclusion chromatography (SEC) column (GE Healthcare, Piscataway, NJ) equilibrated with buffer (20 mmol/L sodium phosphate, pH 7.6, 150 mmol/L NaCl). The main peak was collected, characterized 27 and used as immunogen. ImmunoMax mice (n = 3) were immunized with consecutive boosts of NUCB1-hIAPP complex and produced robust titer to the immunogen.
The blots were developed using the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific).

| Hybridoma supernatant screening
Hybridomas were created using the HTP Hybridoma production pro-

| Conformation-sensitive screening
To test the conformation-sensitive property of the mAbs, their binding to Aβ40 monomers and Aβ42 protofibrils was measured through sandwich ELISA where mAbs were coated on Maxisorp 96-well plates (NUNC) overnight at 4°C. After rinsing five times with ELISA wash buffer (0.1 M sodium phosphate, pH 7.5, 0.15 M NaCl, 0.05% Tween 20) the wells were treated for 2 hours with Protein Free Blocking Buffer. Antigen was added in blocking solution and the plates were treated as above.

| Antibody production and purification
The hybridoma lines were cultured in Hi-Growth Hybridoma Media and stored in 1x DPBS, 10% glycerol, 0.02% NaN3. All mAbs were purified by SEC to remove protein aggregates and immediately used for the assays, in order to avoid potential artefacts due to avidity effect. Temporal autocorrelation curves were fitted using the theoretically derived autocorrelation function for a system that consists of a single molecular species with one triplet state when the molecules are undergoing free three-dimensional (3D) diffusion: To allow direct graphical comparison of characteristic decay times, the temporal autocorrelation curves were normalized to the same amplitude, G n (τ) = 1 at = 10 µs.

| Surface plasmon resonance
Surface plasmon resonance (SPR) studies were carried out with the ProteOn XPR36 protein interaction array system (Bio-Rad) based on SPR technology.
The antibodies were immobilized in the vertical direction on GLM sensor chips (Bio-Rad) using amine-coupling chemistry, as described previously, 29  | 2105 performed at 25°C and the data were normalized by interspot and by buffer.

| Protofibril staining in human frontal cortex
All brain materials were obtained from the Huddinge Brain Bank at Karolinska Institutet Alzheimer Disease Research Center. All familial AD subjects met the criteria for definitive AD according to the Consortium to Establish a Registry for AD (CERAD). 30 Serial 10 µm thick sections were exposed to antigen retrieval with DIVA Decloaker

| Immunization campaign and hybridoma screen for the discovery of tractable mAbs
The engineered Ca 2+ -free NUCB1 inhibits hIAPP aggregation (Figure 1A) by binding to and stabilizing short protofibrils ( Figure S1), as previously reported. 27 The hIAPP-NUCB1 complex was purified using SEC ( Figure 1B), characterized by atomic force microscopy (AFM) and ELISA ( Figure S1) and injected into three mice that showed robust titre to the immunogen ( Figure S2A). The immunization campaign ( Figure 1C) resulted in 752 screened fusion wells that were subjected to a pre-subclone screen by ELISA for their reactivity to the immunogen complex NUCB1-hIAPP as well as to NUCB1 (Figure S2B Figure 1D).

sequence-independent mAbs
To test our hypothesis that NUCB1-hIAPP originated mAbs would detect the common quaternary amyloid protofibril structure independently from the primary sequence, we analysed whether they cross-react with structured protofibrils originating from the Aβ42 peptide.
Sandwich ELISA and FCS were used to determine whether mAbs bind Aβ monomers or protofibrils. Sandwich ELISA shows that each mAb specifically binds to the protofibril enriched pools of Aβ42 antigen and not to Aβ40 monomers (Figure 2A-D, left column). These data suggest that the conformation-sensitive anti-hIAPP mAbs specifically recognize the protofibril conformation in a sequence-independent way.  We observed a strong, concentration-dependent binding to Aβ42 protofibrils for all mAbs ( Figure 2I-L, right column), as well as for the positive control, anti-Aβ antibody 6E10 ( Figure S4A,C), but not the F I G U R E 2 Conformation-sensitive mAbs bind Aβ42 protofibrils but not unstructured Aβ40 monomers. A-D) A sandwich enzyme-linked immunosorbent assay (ELISA) was used to test the binding of A) 4A8.E11, B) 4B1.H9, C) 3F2.E10 and D) 5C9.A2 mAbs to Aβ42 protofibrils or Aβ40 monomers. The antibodies were coated on the plate and used to capture either freshly prepared Aβ40 (monomers) or Aβ42 that had been incubated at 37°C for 30 min to enrich protofibril species. Data are normalized to the lowest antibody concentration. E-L) The SPR was also used to test the mAbs binding to E-H) Aβ40 monomers and I-L) Aβ42 protofibrils. Freshly solubilized Aβ40 monomers (10 µmol/L) and Aβ42 protofibrils (10 µmol/L) were flown at different concentrations for 60 s over each antibody (300 nmol/L) previously immobilized on the chip (RL = 6500). Data are normalized by interspot and buffer and presented as mean ± standard error of the mean (SEM) negative control 1D4 ( Figure S4B,D). Notably, the shape of the dissociation curve indicates a strong association of the binding clones to the protofibrils that does not spontaneously resolve during the washing phase with NaCl (740 seconds). Furthermore, we show that the anti-Aβ antibody 6E10 ( Figure S4A, K d = 4.66 nmol/L), but none of our mAbs ( Figure 2E-H, centre column) or the negative control 1D4 ( Figure S4B), concentration-dependently binds Aβ40 monomers.
The SPR assay provides accurate, label-free measurement of the binding kinetics. However, in the case of antibody-amyloid binding, the data analysis requires particular care. In fact, the correct immobilization of the antibody on the chip surface is a critical step to guarantee good quality data. The ideal immobilization level (R L ) is calculated according to the equation: where Ligand is the antibody and Analyte is the amyloid

| Conformation-sensitive mAbs inhibit hIAPP and Aβ42 aggregation
In order to determine if the antibody binding to amyloid aggregates has a functional effect on fibrillization, a ThT assay was used to measure aggregation kinetics. The assay is based on the alteration that the ThT fluorescence spectrum encounters upon binding to amyloid and therefore the fluorescent signal is considered a measure of protein aggregation.
Notably, the minimal decrease in fluorescence exerted by the negative control is a commonly reported effect of the protein mass, also observed when hIAPP aggregates in the presence of BSA (Figure 1).
When the data were plotted together, it appeared that the antibodies have a comparable inhibitory effect on the two amyloid proteins ( Figure 3I-L), except for 3F2.E10 that acts more potently on Aβ42 aggregation ( Figure 3K). We hypothesized that the sequence-independent inhibition of amyloid aggregation may occur through functional binding of the antibodies to early pre-fibrillar aggregates, preventing them from maturing to the fibril state.

| Protofibril staining in APP 102 /TTA brain tissue
To test the ability of these mAbs to detect protofibrils in tissue, we tested target engagement in ex vivo tissue from a transgenic mouse model of AD. The cortex of APP 102 /TTA mice was sectioned and costained with polyclonal anti-Aβ antibody, our mAbs (4A8.E11, 4B1.H9, 3F2.E10, and 5C9.A2), as well as the lysosome membrane marker LAMP2 and the nuclear marker DAPI (Figure 4). We observed that the mAbs staining was present in proximity, but not inside of the Aβ plaques stained by the total anti-Aβ antibody in the transgenic mice ( Figure 4), but not in wild-type (WT) tissue or regions with no plaques (not shown). Furthermore, we observed co-localization of the mAbs staining with LAMP2 as well as DAPI (Figure 4). This staining pattern suggests that our mAbs do not bind to the Aβ species that deposit in plaques but only to intracellular species localized in the lysosome.

| Intraneuronal staining in human AD frontal cortex
Frontal cortex brain tissue obtained from a familial AD patient was serially sectioned and adjacent slices were stained with the positive control anti-Aβ42 antibody or our conformation-sensitive mAbs to test the signal localization using immunohistochemistry. We observed that while the anti-Aβ42 antibody stained both plaques and pyramidal neurons in the AD frontal cortex, none of our mAbs detected plaques, but specifically and strongly stained pyramidal cells ( Figure 5). These results suggest that the pyramidal neurons contain Aβ protofibril morphotypes, whereas the Aβ-dense plaques do not contain these structures, therefore strengthening the results obtained in animal tissue and confirming the conformation-sensitive feature of our mAbs.

| DISCUSSION
In the past decades, the pathologic role of toxic protofibrils in the development of amyloid diseases has been increasingly recognized, 32 Aside from the biologics currently in clinical trials, the polyclonal antibody A11, described more than 10 years ago, has been a useful tool for researchers for detecting the oligomeric species of amyloid.
This polyclonal antibody was produced using the C-terminal thioester Aβ40 monomers tethered to gold colloid nanoparticles as an immunogen. 18 Alternatively, the monoclonal WO1 and WO2 antibodies were created using sonicated fibrils and react to the general structure of fibrils from many amyloid sources. 34 The murine version of BAN2401, mAb158, was produced using protofibrils obtained from a mutated Aβ42 peptide (E22G, Arctic mutation). 35 Clone 13C3 was created using Aβ42 fibrils followed by screening for protofibril reactivity. 36 Recently, an anti-Aβ42 polyclonal antibody has been obtained by immunizing animals with isolated Aβ42 protofibrils. 37 In contrast to these approaches, we have created a platform technology using the novel NUCB1-capping method to produce stable amyloid protofibrils 27,34 that can be used as immunogen to create panels of conformation-sensitive mAbs. We hypothesized that our technology could be used to produce conformation-sensitive, sequence-independent mAbs that detect the quaternary structure of the protofibrils and bind to early amyloid aggregates independently from the primary structure.
In this work we show that immunizing animals with hIAPP protofibrils stabilized by the CLABP, NUCB1, can produce mAbs that bind both hIAPP and Aβ protofibrils and we describe the methodology used to screen for mAbs with this conformation-sensitive feature. Binding assays, such as dot blot ( Figure 1D These novel anti-protofibril mAbs have shown, in a disease-specific animal model of AD, staining around, but not inside the Aβ plaques in the prefrontal cortex. This staining pattern appears to be intracellular and granular. There has been debate as to whether F I G U R E 4 mAbs reveal punctate staining around plaques but don't associate with the plaques in an AD mouse model cortex. The brain cortex obtained from APP102/TTA mice was stained to analyze the localization of our mAbs-binding signal. Immunohistochemistry reveals that mAbs 4A8.E11, 4B1.H9, 3F2.E10, and 5C9.A2 (green) detect small, subdiffraction-limited spot size regions surrounding the total α-Aβ antibodypositive plaques (grey). mAbs signal co-localizes with the lysosome marker LAMP2 (red) localized in the vicinity of cell nuclei stained by the neuronal marker DAPI (dark cyan), but not with the plaques. Arrowheads indicate spots where co-localization was observed; scale bar = 20 µm intracellular or extracellular amyloid aggregation is a major driver of plaque formation. 38 When co-stained with the lysosomal marker LAMP2 the mAbs display a co-localization pattern suggesting that protofibrils accumulate within lysosomes. These studies indicate that our mAbs co-stain intracellular lysosomal vesicles that contain Aβ and protofibril conformations, supporting the model of intracellular protofibril formation or accumulation.
We further studied these mAbs using human familial AD frontal cortex samples. The AD tissue displayed Aβ-positive plaques and Aβpositive intraneuronal staining pattern. Using serial sections and imaging near the same area, we found that our mAbs do not display a plaque-like pattern, but rather showed a specific intracellular pyramidal neuron staining. These data support the mouse tissue staining pattern and therefore we conclude that our mAbs do not react with Aβ-positive plaque, but specifically bind to an intraneuronal protofibril conformation.
In summary, this work shows that our platform technology of NUCB1-capped hIAPP protofibrils is a suitable tool for discovering mAbs with different reactivity profiles. We provide a detailed screening platform to assess functional binding and target engagement in in vitro and ex vivo models. These anti-protofibril antibodies The Magnus Bergvall's Foundation. We thank the Electron Microscopy Resource Facility at Rockefeller University and in particular Dr. Devrim Acehan for guidance in the use of the microscopes. We also thank Dr. Joanna Jankowsky for providing tissue, reviewing the MS, and giving comments.

CONFLI CT OF INTEREST
The authors declare that they have no competing interests.

AVAILABILITY OF DATA AND MATERIALS
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. F I G U R E 5 Conformation-specific mAbs stain pyramidal cells in AD human frontal cortex. The brain cortex obtained from familial AD patients was stained to analyze the localization of our mAbs-binding signal. DAB immunohistochemistry reveals that while the α-Aβ42 antibody stains plaques, 4A8.E11, 4B1.H9, 3F2.E10, and 5C9.A2 do not bind to plaques but specifically stain pyramidal cells. Scale bar = 500 µm in A and 250 µm in B