Human postmortem brain specimens

8-μm-thick formalin-fixed paraffin-embedded (FFPE) tissue sections from the temporal [including Brodmann area (BA) 38], frontal (BA8/9), and occipital cortex (BA17/18), and cerebellum of n = 52 donors spanning the AD neuropathological continuum from normal aging brain (CTRL, “not AD/low ADNC burden”) to severe AD (high ADNC burden) were obtained from the Massachusetts Alzheimer’s Disease Research Center (MADRC) Neuropathology Core Brain Bank. FFPE sections from the hippocampus of selected donors were also used for illustration purposes. Table 1 depicts the demographic (age, sex) and neuropathological (Thal amyloid phase, CERAD neuritic plaque score, and Braak neurofibrillary tangle stage) characteristics, and the APOE genotype of the n = 52 study donors split by ADNC burden [32]. Frozen samples from the cerebellum were also available for n = 50 donors for DNA purification and MAOB SNP genotyping. We validated the main findings of this study in a separate sample of n = 51 donors with intermediate or high ADNC burden, for which quantitative neuropathological data was available [44]. Specifically, we used Aβ+ area fraction, cortical thickness, and stereology-based count data of pTau+ (PHF1) neurofibrillary tangles, GFAP+ astrocytes, and CD68+ microglia from the temporal association neocortex of this validation sample, which were obtained following procedures detailed elsewhere [44].

Table 1 Demographic and neuropathological characteristics, and APOE genotype of the study donors along the AD continuum split by the burden of AD neuropathological changes

In addition, 8-μm-thick FFPE sections from the frontal association cortex (BA8/9) of n = 70 donors with a primary neuropathological diagnosis of ADRD—comprising n = 30 LBD (including five with brainstem/limbic and 25 with neocortical Lewy body stage), n = 30 FTLD-Tau (including ten of each PiD, PSP, and CBD), and n = 10 FTLD-TDP—were also obtained from the MADRC Neuropathology Core Brain Bank (Table 2). Donors or their next-of-kin gave a written informed consent for the brain donation and this study was conducted under the MADRC Neuropathology Core Institutional Review Board.

Table 2 Demographic characteristics of the study donors with a primary neuropathological diagnosis of ADRDMultiplex fluorescent immunohistochemistry

To determine the cell type expressing MAO-B, we performed double fluorescent immunohistochemistry with antibodies for MAO-B and cell-type-specific markers on FFPE sections from the temporal association cortex of selected CTRL and AD donors followed by laser confocal microscopy. Briefly, FFPE sections were dewaxed in xylenes and rehydrated in decreasing concentrations of ethanol. Next, sections were microwaved in boiling citrate buffer (0.01 M, pH 6.0 with 0.05% Tween20) for 20 min at 95 °C for antigen retrieval. Subsequently, sections were blocked with 10% normal donkey serum (NDS) in Tris-buffered saline (TBS) for 1 h at room temperature (RT) prior to incubation with primary antibodies in 5%NDS/TBS at 4 °C overnight. On the following day, sections were washed with TBS before incubation with secondary antibodies at 1:200 in 5%NDS/TBS for 2 h at RT, washed with TBS, and counterstained with Thioflavin-S (ThioS, Sigma, T1892) by immersion in 0.05% ThioS in 50% ethanol for 8 min, followed by differentiation with 80% ethanol for 10 s. Finally, sections were incubated with Autofluorescence Eliminator Reagent (Millipore, 2160) for 5 min at RT to quench endogenous tissue autofluorescence, cleared with serial 70% ethanol washes of 20 s each, rehydrated in TBS, and coverslipped with Fluoromount-G mounting media with DAPI (Southern Biotech, 0100-01). The following primary antibodies were used: rabbit anti-MAO-B monoclonal antibody (raised against amino acids 1-100 of human MAO-B, clone EPR7102, Abcam, ab133270, 1:500), mouse anti-Aβ monoclonal antibody (clone 6E10, BioLegend, 803003, 1:100), mouse anti-ALDH1L1 monoclonal antibody (clone N103/39, Millipore, MABN495, 1:500), mouse anti-CD31 monoclonal antibody (clone 89C2, Cell Signaling Technology, #3528, 1:100), mouse anti-GFAP monoclonal antibody (clone G-A-5, Sigma, G3893, 1:1000), goat anti-IBA1 polyclonal antibody (Abcam, ab107159, 1:100), mouse anti-microtubule-associated protein 2 (MAP2) monoclonal antibody (clone SMI-52, Biolegend, 801801, 1:250), mouse anti-myelin basic protein (MBP) monoclonal antibody (clone 1, Millipore, MAB382, 1:500), mouse anti-platelet derived growth factor receptor-β (PDGFRβ) monoclonal antibody (clone D-6, Santa Cruz Biotechnology, sc-374573, 1:200). Secondary antibodies included a donkey anti-rabbit Cy3-conjugated antibody (Jackson ImmunoResearch) for MAO-B and species-appropriate donkey antibodies conjugated with Cy5 (ThermoFisher Scientific) for all other primary antibodies.

Peroxidase-DAB immunohistochemistry

For quantitative analyses, we immunostained FFPE sections from all donors with the peroxidase-3,3′-diaminobenzidine (DAB) method using the BOND Polymer Refine Detection kit (Leica Biosystems, DS9800) in a Leica BOND RX Fully Automated Research Stainer (Leica Biosystems). Specifically, the following primary antibodies were used: rabbit anti-MAO-B monoclonal antibody (clone EPR7102, Abcam, ab133270, 1:1000), rabbit anti-Aβ monoclonal antibody (clone D52D4, Cell Signaling Technology, 1:2000), mouse anti-CD68 monoclonal antibody (clone KP-1, Dako, M0814, 1:500), mouse anti-GFAP monoclonal antibody (clone G-A-5, Sigma-Aldrich, G3893, 1:20,000), and mouse anti-pTauSer396/404 monoclonal antibody (clone PHF1, kind gift from Dr. Peter Davies, 1:500). MAO-B immunohistochemistry was performed in all four regions (temporal, frontal, occipital, and cerebellum) of all donors in Table 1 except for one donor for which cerebellum was not available, in the hippocampus of selected donors in Table 1, as well as in frontal sections from all donors in Table 2. Aβ, CD68, GFAP, and pTau immunostains were only performed in temporal cortex sections from all donors in Table 1. The automated immunohistochemistry protocol consists of: (1) baking at 80 °C for 60 min; (2) dewaxing using BOND Dewax solution (Leica Biosystems) followed by rehydration; (3) heat-induced epitope retrieval (HIER) with citrate (ER1) at 100℃ for either 20 min (MAO-B, Aβ, and pTau) or 30 min (CD68) except for GFAP, which did not require this step; (4) blocking of endogenous peroxidase activity with a 4% (v/v) hydrogen peroxide solution in methanol for 20 min; (5) primary antibody incubation at 37℃ for 35 min except for anti-CD68 antibody, which was incubated for 60 min; (6) secondary antibody incubation with post-primary solution containing 10 g/mL Poly-HRP IgG in 0.01% 2-methylisothiazol-3(2H) for 10 min followed by polymer solution containing 25 g/mL Poly-HRP IgG in 0.01% 2-methylisothiazol-3(2H) for 8 min; (7) DAB substrate-peroxidase reaction for 1 min; (8) counterstaining with 0.1% hematoxylin for 15 min; and (9) washes with wash buffer and double-distilled water. Immunostained sections were dehydrated in increasing concentrations of ethanol and coverslipped with Permount mounting media (Fisher Scientific, SP15-500).

MicroscopyLaser confocal microscopy

To study the expression of MAO-B in various brain cell types, selected sections subjected to multiplex fluorescent immunohistochemistry for MAO-B and cell-type-specific markers were imaged in an Olympus FV3000 confocal laser scanning inverted microscope (Olympus, Tokyo, Japan), which is equipped with 405, 488, 561, and 640 nm lasers as well as two regular and two high-sensitivity spectral detectors. Briefly, we used a 40 × /0.95 air objective and imaged z-stacks with the sequential mode of acquisition and different detectors for MAO-B and the cell-type-specific marker of interest to minimize the risk of bleeding through, and then obtained maximum intensity projections of the z-stacks.

Whole-slide bright-field microscopy

Sections immunostained for MAO-B, Aβ, CD68, GFAP, and pTau with the peroxidase-DAB method were scanned in a VS120 Olympus Virtual Slide Microscope to obtain whole-slide images under a 40 × /0.95 air objective.

Quantitative neuropathological methodsMeasurement of immunoreactive area fractions

We used the open access software QuPath (version 0.3.2) for all quantitative neuropathological analyses. To account for the variable distribution of the signal across the entire section, we outlined the cortex (and for MAO-B also the white matter) of the whole temporal, frontal, and occipital sections, and converted those outlines into annotations. For cerebellar sections, we only annotated the cortex and white matter of a single folium because one was representative of the entire section. We chose the QuPath pixel classifier tool because this outperformed the thresholder tool in our pilot studies. Specifically, the pixel classifier parameters were set at full resolution (0.17 μm/pixel), default multiscale features (scale 1.0, Gaussian filter, no local normalization), and smoothing sigma 0.25 for all markers. A separate pixel classifier was trained for each marker quantification by providing the artificial neural network (ANN-MLP) with 25 positive annotations (i.e., DAB-positive pixels) and 25 negative annotations (i.e., background and hematoxylin-positive pixels) of the temporal, frontal, and occipital sections of five CTRL and five AD donors (total training sample per classifier: 1500, 750 positive and 750 negative annotations). Next, we processed the images in project batches using automated scripts for each pixel classifier. The area fraction was defined as the % area occupied by DAB-positive pixels with respect to the area of the entire cortical or white matter annotation.

Measurement of cortical thickness

To obtain the temporal lobe cortical thickness, we applied the QuPath line annotation tool to GFAP-immunostained whole-section images and measured the shortest distance between the pial surface and the white matter (i.e., line perpendicular to both) in 20 random sites distributed throughout the cortical ribbon. The cortical thickness was defined as the average of those 20 measurements, as described before [44]. The GFAP-immunostained sections were chosen for cortical thickness measurements because the high GFAP expression by the glia limitans serves as fiduciary landmark of the cortical surface and the higher GFAP expression in white matter versus cortex facilitates the gray-white matter differentiation.

Measurement of plaque-centered area fractions

To determine whether the expression of MAO-B is spatially associated with dense-core Aβ plaques, we performed plaque-centered quantitative analyses in FFPE sections from the temporal cortex of n = 10 high ADNC (AD) donors and n = 10 not AD/low ADNC (CTRL) donors fluorescently immunostained for MAO-B and GFAP (as positive control) and counsterstained with ThioS. Briefly, in each AD donor, n = 50 ThioS+ dense-core Aβ plaques were randomly selected and outlined in the green channel of the whole-slide image with the appropriate tool in QuPath ensuring that all layers of the cortical ribbon were equally represented, while being blind-folded to the expression level of MAO-B in the TRITC channel and of GFAP in the Cy5 channel. Next, a 50 μm concentric halo was added to each selected plaque and n = 50 regions-of-interest of similar size but far (> 50 μm) from any ThioS+ plaque were overlaid on the whole-slide image with the appropriate tools in QuPath software. For each CTRL donor, n = 50 regions-of-interest of similar size to the AD plaques were added onto the cortical ribbon as sham plaques, together with their 50-μm halo, and n = 50 more regions-of-interest far (> 50 μm) from those sham plaques. The MAO-B+ and GFAP+ area fractions were measured in the three types of regions-of-interest, i.e., ThioS+ or sham plaques, peri-plaque halo (≤ 50 μm), and areas distant (> 50 μm) from ThioS+ or sham plaques.

MAOB rs1799836 SNP genotyping

To genotype the MAOB rs1799836 SNP, we used a commercially available Taqman PCR assay on genomic DNA extracted from frozen cerebellar samples. Briefly, we purified genomic DNA from ≈ 25 mg of frozen cerebellum using the PureLink Genomic DNA Extraction Mini Kit (ThermoFisher Scientific, K182002) following manufacturer’s instructions. We measured the DNA concentration in a DS-11 spectrophotometer (DeNovix Inc) and prepared 1.8 ng/μL working dilutions for the Taqman PCR assay. The latter reaction volume was 25 μL comprising 1.25 μL of 20 × TaqMan MAOB rs1799836 genotyping assay (ThermoFisher Scientific, Assay ID C—8878790_10), 12.50 μL of 2 × TaqMan Fast Universal PCR Master Mix, no AmpErase UNG (Thermo Scientific, 4352042), and 11.25 μL of DNA sample (20 ng). DNA samples were run in duplicates in 96-well plates (Bio-Rad) in a Bio-Rad CFX96 Touch Real-Time PCR Detection System with the following protocol: 95 °C × 10 min (ramp 1 °C/s), 95 °C × 15 s, and 60 °C × 1 min, for 45 cycles. Principal component analysis of VIC versus FAM fluorescence [corresponding to base A (major allele) vs. G (minor allele), respectively] enabled allele discrimination and genotype assignment (AA, AG, or GG).

SDS-PAGE and western blot

To validate the specificity of the anti-MAO-B antibody used in the immunohistochemical studies, we performed western blot in MAOB-overexpressing and knock-down human cell line lysates and human recombinant MAO-A and MAO-B proteins. The human recombinant MAO-A protein was purchased from LS-Bio (LS-G3233), whereas the human recombinant MAO-B protein was purchased from Abcam (ab82944), both purified from Escherichia coli with a 6xHis-tag. The human cell line lysates were kindly provided by Dr. Boyang (Jason) Wu (Washington State University, Spokane, WA) and consisted of lysates from PrSC human prostate stroma cells stably expressing either a control plasmid or a MAOB plasmid, and lysates from HepG2 human liver carcinoma cells stably expressing either a control shRNA or one of two anti-MAOB shRNAs [50]. Briefly, sample protein concentration was measured with the Pierce BCA Protein Assay Kit (ThermoFisher Scientific, 23225). After boiling at 95 °C for 5 min with 10X sample reducing agent (NuPAGE, ThermoFisher Scientific, NP0009) and 4X protein sample loading buffer (LI-COR Biosciences, P/N: 928-40004) for protein denaturalization, samples (4 μg for recombinant proteins, 10 μg for cell lysates) were loaded onto a 4–12% Bis–Tris gradient gel (NuPAGE, ThermoFisher Scientific, NP0343BOX) together with Precision Plus Dual Color Protein Standards (Bio-Rad, 1610374). SDS-PAGE was run in MOPS SDS running buffer (NuPAGE, ThermoFisher Scientific, NP0001) at 130 V for ~ 2 h. The gel proteins were transferred to a nitrocellulose membrane using a wet method at 310 mA for 2 h in ice. The membrane was blocked with Intercept Blocking Buffer (TBS) (LI-COR Biosciences, P/N 927-50000) for 1 h at RT, followed by incubation with primary antibodies in the Intercept Blocking Buffer (TBS) overnight at 4 °C. We used a mouse monoclonal anti-β-actin clone AC-15 (Sigma-Aldrich, A5541, 1:3000) as housekeeping loading control. Next day, the membrane was washed with TBS-Tween20 and incubated with the appropriate near-infrared fluorescent secondary antibodies (LI-COR Biosciences, 1:5000) for 1 h at RT. Finally, the membrane was thoroughly washed with TBS-Tween20 followed by TBS and scanned in an Odyssey CLx Imager (LI-COR Biosciences).

Statistical analyses

We classified donors in three groups corresponding to their ADNC burden [32]. Normality of all datasets was determined by D’Agostino-Pearson Omnibus test. We ran one-way ANOVA with Tukey’s multiple comparison test if all datasets were normally distributed and Kruskal–Wallis ANOVA with Dunn’s multiple comparison test if one or more datasets were non-normally distributed. For correlation analyses between MAO-B area fraction and other neuropathological measures, we used Spearman’s rank correlation tests (since MAO-B area fraction was non-normally distributed) and simple linear regression. To investigate the effects of sex and MAOB rs1799836 SNP genotype on MAO-B area fraction, we used two-way ANOVA with ADNC burden and either sex or MAOB rs1799836 SNP genotype as factors and the sex × ADNC burden or genotype × ADNC burden interaction terms. To examine the independent associations between MAO-B expression and all other neuropathological measures, we built multiple linear regression models with temporal cortex MAO-B area fraction as outcome variable; temporal cortex Aβ, pTau, GFAP, and CD68 measures, and cortical thickness as independent variables; and age at death as co-variate (to account for possible influences of aging on cortical thickness, GFAP immunoreactivity, and diffuse Aβ deposits). Lastly, for the plaque-centered analyses, we used mixed effects models with MAO-B or GFAP area fraction as outcome variable, diagnosis (CTRL vs. AD) and location (sham or ThioS+ plaque, ≤ 50 μm peri-plaque halo, or > 50 μm), with or without a diagnosis × location interaction term, as fixed effects, and controlling for donor ID (random effect). The significance level was set at a two-sided p < 0.05 in all statistical analyses. All statistical analyses and graphs were performed with GraphPad Prism version 9.4.1 (GraphPad Inc., La Jolla, CA) except mixed-effect models, which were run in STATA version 15.0 (StataCorp, LLC., College Station, TX).