Analyzing microglial-associated Aβ in Alzheimer’s disease transgenic mice with a novel mid-domain Aβ-antibody

The mechanisms of amyloid-β (Aβ)-degradation and clearance in Alzheimer’s disease (AD) pathogenesis have been relatively little studied. Short Aβ-fragments form by enzymatic cleavage and alternate amyloid-beta precursor protein (APP)-processing. Here we characterized a novel polyclonal Aβ-antibody raised against an Aβ mid-domain and used it to investigate microglial Aβ-uptake in situ by microscopy at the light- and ultrastructural levels. The rabbit Aβ-mid-domain antibody (ab338), raised against the mid-domain amino acids 21–34 (Aβ21–34), was characterized with biochemical and histological techniques. To identify the epitope in Aβ recognized by ab338, solid phase and solution binding data were compared with peptide folding scores as calculated with the Tango software. The ab338 antibody displayed high average affinity (KD: 6.2 × 10−10 M) and showed preference for C-terminal truncated Aβ-peptides ending at amino acid 34 and Aβ-mid domain peptides with high scores of β-turn structure. In transgenic APP-mouse brain, ab338 labelled amyloid plaques and detected Aβ-fragments in microglia at the ultra- and light microscopic levels. This reinforces a role of microglia/macrophages in Aβ-clearance in vivo. The ab338 antibody might be a valuable tool to study Aβ-clearance by microglial uptake and Aβ-mid-domain peptides generated by enzymatic degradation and alternate production.

Microglial localization of ab338-labelled Aβ-peptides in situ in brain with light microscopy. Given the preferential binding of ab338 to a β-turn in Aβ we postulated ab338 to display a distinct immunolabelling and be suitable to study in vivo microglial Aβ localization. This was studied in situ in tgArc-Swe mice at 12 months, an age when the amyloid plaques are abundant but not extensive in these mice enabling analysis of individual deposits. Ab338 labelled amyloid plaques, while tomato lectin, a surface marker of macrophages and vessels was used to label the surrounding microglia (Fig. 3A,B and Supplementary Fig. S2).
Omission of the primary antibody, served as the negative control, ruling out unspecific binding of the secondary antibody ( Supplementary Fig. S2). In addition, ab338 labelled cerebrovascular deposits ( Supplementary Fig. S3).
In microglia nearby plaques, z-stacking supported intracellular Aβ labelling, that appeared located in the ramified processes of microglia (Fig. 3C,D). Phagocytosis is one of several cell uptake mechanisms. Cells of monocyte lineage, like microglia, highly express CD68 which localizes to endosomes and lysosomes 43 and can serve as a microglial phagocytosis marker 52 ( Supplementary Fig. S4). To add evidence of microglial Aβ-uptake by phagocytosis while concomitantly further characterizing ab338 Aβ-labelling, brain sections of another APP-transgenic model, tgSwe, were stained with ab338 with and without a CD68-antibody. We found spots of co-localization of ab338 and CD68 suggesting www.nature.com/scientificreports/ Aβ-uptake by phagocytosis in tgSwe mouse brain ( Fig. 4A-C). Tissue treatment with fluorophore-conjugated secondary antibodies alone showed that the signal was due to binding of primary antibodies (data not shown).
Aβ-immunolabelling in microglia with transmission electron microscopy. Further examining ab338 labelling and microglial Aβ-uptake in situ in tgArcSwe mice at the ultrastructural level, ab338-immunogold particles labelled Aβ-fibrillar structures. The microglial processes were directed toward ab338-labelled Aβ-fibrils seemingly enconvuluting them. Intriguingly ab338-labelling was also present in the cell soma, indicating Aβ internalization by microglia in tgArcSwe mice ( Fig. 5A-C). The distinct microglial anatomical structure was recognizable at the ultrastructural level, and confirmed by Iba-1 labelling of a microglia infiltrating an amyloid plaque ( Supplementary Fig. S5).

Discussion
APP-metabolites and Aβ-peptides are challenging to assay in tissues and body fluids due to their relatively low abundance, aggregation propensity and the complexity with many types of Aβ-species in tissues. In the current study, we characterized a novel antibody raised against the Aβ-mid-region Aβ [21][22][23][24][25][26][27][28][29][30][31][32][33][34] . The binding site was identified by biochemical and computational methods. The affinity-purified polyclonal antibody ab338 showed preferential binding to Aβ-peptides with C-terminal aa 34 (e.g. Aβ 1-34 ). Such Aβ x-34 fragments could be detected by competitive ELISA at pM-concentrations. In the indirect ELISA, the only peptide with relevant binding except the Aβ 21-34 peptide was Aβ [15][16][17][18][19][20][21][22][23][24][25][26][27][28] . This was consistent with ab338 binding to spot-synthesized peptides having a high β-turn score. Mapping the Aβ-epitope of ab338 by ELISAs and spot-synthesized Aβ-peptides gave overall consistent results with some discrepancies. This may be due to the spot-synthesized peptides presumably being more accessible for antibody binding as each spot had much more peptide compared to the indirect ELISA, although this in part was accounted for by using a lower ab338 concentration. As according to the analyzes of membrane spot-synthesized consecutive Aβ-peptide sequences, ab338 binding to multiple staggered sequences was observed and the signal did not very much depend on a specific aa, as the antibody bound from Aβ 19-28 until Aβ [25][26][27][28][29][30][31][32][33][34] . We therefore asked if ab338 could be detecting a folding structure. Aβ-monomers aggregate into fibrils with a β-sheet structure composed of β-strands segments of Aβ wherein the peptide backbones are connected as the peptide flip over in a β-turn region (loop or reverse turn) 51 . The aggregation propensity of Aβ-peptides can be calculated by the Tango software, which has shown good prediction of peptide folding structure and aggregationproperties 53 . Calculating the β-turn conformational score with Tango aligned well with the findings of in vitro epitope-mapping suggesting that ab338 recognized a β-turn structure in Aβ. Preference of ab338 towards the β-turn in Aβ was supported by a negative outcome when instead trying to align α-helix and β-strand scores of the spot-synthesized Aβ-peptides with binding data. The Tango β-turn data also aligned well with structural analysis of Aβ 1-40 fibrils with aa12-24 and aa30-40 forming β-sheets separated by a turning sequence at aa 25-29 54 . www.nature.com/scientificreports/ Thus the ab338 antibody seemed to bind Aβ in the β-turn region, preferentially β-turn and a C-terminal aa34. Still, given the high prevalence of full-length Aβ in brains of aged APP-transgenic mouse, we do not exclude that immunofluorescent staining to some extent is due to ab338 also recognizing longer Aβ-peptides that contain the Aβ 21-34 domain e.g. Aβ 1-40 and Aβ 1-42 or other Aβ-species harboring this domain. Existing Aβ antibodies mostly recognize the Aβ terminals either the N-terminal fragment of Aβ e.g. 6E10 55 or a C-terminal epitope. Such antibodies are commonly combined to create sandwich Aβ-ELISAs to determine full-length peptides e.g. Aβx-42 56 . In contrast the 4G8 antibody recognize an epitope close to the α-cleavage site in the Aβ-domain 57 . Still there are few antibodies targeting the Aβ-mid-domain sequences and which thereby enable specific detection of mid-domain peptides. Such Aβ-mid-domain fragments may be released directly from APP or result from enzymatic cleavage of full-length Aβ-peptides 17,25 . Alternate APP-processing include a second BACE1 cleavage suceeding the initial BACE1/γ-secretase cleavage generating Aβ 1-34 peptides 58,59 . Consistent with such cleavage the Aβ 1-34 level in CSF is reduced following BACE-1 inhibiton and γ-secretase modulation 60,61 . Developing AD therapeutics like β-/γ-secretase inhibitiors is challenging 62,63 and assaying a biomarker like CSF Aβ 1-34 might aid in adjusting dosage and monitoring target engagement. ELISAs with Aβ C-terminal aa34 specificity would be advantageous compared to mass spectrometry to quantify such Aβ-species for logistical reasons. Aβ-mid-domain fragments could be a useful measure of Aβ-degradation by enzymatic breakdown, with neprilysin as the most well established degrading enzyme 64 . Neprilysin typically cuts the Aβ-peptides at the N-and C-terminals but cleavage sites also include the Aβ-mid-region at aa 19-20 and 33-34 65 . Additional enzymes may catabolize Aβ-peptides in the extracellular space. Examples are angiotensin converting enzyme (ACE) 66 and matrix metalloproteinases (MMPs) 67 with cleavage sites at aa 20/21, aa 33/34 and 34/35 65 generating Aβ 21-x and Aβ x-34 peptides. In the intracellular compartments other enzymes like IDE, Cathepsin D and pre-sequence protease (PreP) complement the extracellular catabolic enzymes of which the latter two are suggested to generate Aβ [21][22][23][24][25][26][27][28][29][30][31][32][33][34] and Aβ x-34 peptides 68,69 . Despite their low aggregation potential, reports indicate retained neurotoxicity of these shorter mid-domain peptides and the need for endolysosomal degradation and detoxification 51,[69][70][71] .
Glial uptake of Aβ-peptides from the extracellular space, has until recent years been controversial in vivo 37 and mainly supported by experimental in vitro data. The ab338 antibody displayed good immunolabelling, presumbaly due to preferential binding to a strech of amino acids forming a β-turn in Aβ. This region is likely more accessible for antibody binding since the peptides are interlinked in the β-strand regions. We have previously described astroglial responses to brain amyloid pathology 72 . Here we show ab338 Aβ-immunstaining located to microglia in situ at the ultrastructural level, with amyloid fibrills seemingly engulfed by microglia. www.nature.com/scientificreports/ Confocal microscopy with z-stacking at the light microscopic level was consistent with transmission electron microscopy data and added evidence of in vivo microglial Aβ-phagocytosis. Plaque formation attracts microglia and these plaque-attractated microglia aquire a disease associated microglial (DAM) phenotype 45 also refered to as microglial neurodegenerative (MnDG) phenotype 73 . The DAM microglial phenotype is among others characterized by upregulation of Axl and TREM2 74 . These are phagocytic microglial receptors upregulated with neurodegeneration 75 including in AD in the vicinity of senile plaques 44 . Both receptors constitute potential therapeutic targets and their role in Aβ-amyloid phagocytosis deserves further attention.

conclusions
Here we describe a novel Aβ mid-domain antibody, ab338 that preferentially binds a β-turn region in Aβ-peptides and a C-terminal aa 34. By ultrastructural-and light-microscopy histological techniques, we demonstrate microglial Aβ-uptake in situ supporting the more recently acknowledged role of microglia to Aβ-phagocytosis. Aβ-mid-domain antibodies like ab338 may complement existing laboratory tools when investigating APP processing and Aβ-pathology in biological samples and may be useful for theragnostic purposes.
Peptide aggregation and β-turn. Protein folding of peptides, including total β-turn score, α-helical score and β-strand score of the various Aβ-peptides was calculated by the Tango software 53 .
Transgenic mice. Transgenic mice overexpressing human APP harboring the Arctic (E693G) and the Swedish (K670M/N671L) double mutation (tgArcSwe) or the Swedish double mutation only (tgSwe) were used to investigate localization of Aβ to microglia or brain macrophages, hereafter referred to as microglia. Homozygous transgenic mouse expressing CX3CR1-GFP (stock #005582, Jackson Laboratory, Bar Harbor, Maine, USA) 76 , was bred with hemizygous tgSwe to generate double transgenic tgSwe x CX3CR1-GFP mouse. Use and handling of animals were approved by the Biological Research Ethics Committee in Norway (Norwegian Animal Research Authority (NARA) permits id: 5693, 6006 and 7240. The mice were fed ad libitum and housed under standard conditions with a 12 h light/dark cycle. DNA from ear cartilage was used to detect the transgene by polymerase chain reaction (PCR) as previously described 77 . TgArcSwe-and tgSwe mice were sacrificed at the age of 12 and 18 months respectively for investigation of microglial Aβ-localization. Briefly the mice were given a ZRF-cocktail (zolazepam; 18.7 mg/ml, tiletamine; 18.7 mg/ ml xylazine; 0.45 mg/ml; fentanyl; 2.6 µg/ml, 0.75 ml/g body weight) or a mixture of ketamine (300 mg/kg body weight) and medetomidine (4 mg/kg body weight) for anesthesia. When the mice lacked pain reflexes they were transcardially perfused, decapitated and the brains quickly dissected. For light microscopic analyses the mice were perfused with 0.9% (w/v) saline and the brains fixed in 4% paraformaldehyde (PFA) in Sørenson's phosphate buffer (SPB; 23 mM KH 2 PO 4 , 70 mM Na 2 HPO 4 *2H 2 0, 5 mM NaN 3 , pH7.4) o.n. at 4 °C. For ultrastructural analyses, the mice were perfused with a mixture of 4% PFA and 0.1% glutaraldehyde in 0.1 M phosphate buffer for 15 min subsequent to a flush with 2% dextran sulfate in phosphate buffer.

Microscopy and immunostaining. Immunohistochemistry at the light microscopic level.
For immunofluorescence, tgArcSwe brains were embedded in paraffin and coronal sections (6 µm) prepared. Prior to immunostaining, the sections were deparaffinized by serial immersion in xylene, 96% ethanol, 70% ethanol and finally water. All steps were repeated twice for 3 min. Immunostaining was performed as previously described 77 with modifications. For Aβ-and microglial visualization in tgArcSwe brain, sections were immersed in PBS prior antigen retrieval by microwave treatment in citrate buffer (25 mM) and 70% (v/v) formic acid (5 min). The tissue was made permeable by immersion in 0.4% Triton X-100 in PBS (v/v) for 5 min. Sections were blocked with Dako protein block (#X0909, Dako, Glostrup, Denmark) and thereafter incubated with the primary antibody (0.5 µg/ml ab338, rabbit polyclonal) in PBS-T o.n. at 4 °C. In negative control experiments, PBS-T alone was used. Next day, the sections were rinsed in PBS prior to incubation for 30 min at RT in the dark with goat antirabbit antibody conjugated to Alexa-Fluor 488 (2 µg/ml, #A-11034, Thermo Fischer, Waltham, USA) alone or in combination with DyLight 594 labeled Lycopersicon Esculentum Tomato Lectin (2 µg/ml, #DL-1177, Vector laboratories, Burlingame, CA, USA) in PBS-T. The sections were mounted in SlowFade Gold Antifade Reagent with DAPI (#S36942 Molecular Probes, ThermoFischer, Waltham,USA) and sealed by use of nail polish.
Confocal images were obtained using a LSM 510 Meta confocal microscope (Zeiss) and 40 × or 63 × oil immersion objectives. Z-stacking were done by use of the 63 × objective at 0.4 µm and 1.0 µM intervals for tgArcSwe and tgSwe brain sections respectively. The fluorophores were excited at 405 nm, 488 nm and 561 nm wavelengths at equal pinhole size. Detection of the DAPI, Alexa Fluor 488, Alexa Fluor 594 and Dylight 594 fluorophores were done sequentially and images merged as to outline fluorophore localization. A Carl Zeiss inverted microscope (Axio Observer Z1) equipped with a Hamamatsu ORCA Flash 4.0 camera were applied to obtain additional conventional fluorescence images. A 5 × fluar or a 40 × neofluar oil-immersion objective was used to obtain the images with DAPI, 38 HE Alexa and 63 HE Red Fluorescence filter cubes with set exposure times for the different filters of comparable sections. All subsequent image processing was performed in the ZEN Blue software (ZEN 2012, Carl Zeiss Microscopy, Germany).
Embedding and immunocytochemistry for electron microscopy. Pieces from mouse cerebral cortex (1.0 × 0.5 × 0.5 mm 3 ) were dissected from 500 μm thick sections and embedded in Lowicryl HM 20 as previously described 72,78 . Cryoprotection and cryosubstitution were the two main steps of tissue preparation. Cryoprotection was undertaken by immersing tissues in phosphate buffered glucose, followed by increasing glycerol concentrations (10, 20, 30% (v/v)) before inserting the tissue specimens into liquid propane at − 190 °C in a liquid nitrogen cooled Scientific RepoRtS | (2020) 10:10590 | https://doi.org/10.1038/s41598-020-67419-2 www.nature.com/scientificreports/ unit KF80 (Reichert, Vienna, Austria). Cryosubstitution was done in 0.5% uranyl acetate in anhydrous methanol at − 90 °C for 24 h in a cryosubstitution unit (AFS, Reichert). The temperature was gradually increased to − 45 °C and Lowicryl HM20 stepwise substituted by methanol. The specimens were polymerized under UV light for 48 h at − 45 °C. Ultrathin sections (90 nm) were cut and transferred onto formvar-coated single hole grids and post-embedding immunogold labelling carried out. Briefly, the sections were incubated in 50 mM glycine in Tris buffered saline with 0.1% Trition X-100 (TBS-T) followed by 2% human serum albumin (HSA) in TBS-T (w/v). The primary antibodies diluted in 2% HSA TBS-T (ab338; 1:2000, Iba-1; 1:500) were applied to the sections for 2 h. The sections were rinsed twice with TBS-T before incubation with goat-anti-rabbit antibodies coupled to 15 nm gold particles in 2% HSA TBS-T for 1 h. For enhancing the contrast, uranyl acetate (Fluorochem) and lead citrate were used successively. The micrographs were obtained digitally by a transmission electron microscope (Technai 12, Hillsboro, Oregon, USA).
Statistical analyses. The GraphPad Prism software (ver. 7.4, Graph Pad Software, La Jolla, USA) was applied for statistical analyses and to create graphs. Estimation of the equilibrium dissociation constant (K D ) and the half inhibitory concentration (IC 50 ) were done by the one site total function and curve fitting by adapting the results to sigmoidal dose-response function respectively.
Use of experimental animals.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/