Selective suppression of oligodendrocyte-derived amyloid beta rescues neuronal dysfunction in Alzheimer's disease

Abstract

Reduction of amyloid beta (Aβ) has been shown to be effective in treating Alzheimer's disease (AD), but the underlying assumption that neurons are the main source of pathogenic Aβ is untested. Here, we challenge this prevailing belief by demonstrating that oligodendrocytes are an important source of Aβ in the human brain and play a key role in promoting abnormal neuronal hyperactivity in an AD knock-in mouse model. We show that selectively suppressing oligodendrocyte Aβ production improves AD brain pathology and restores neuronal function in the mouse model in vivo. Our findings suggest that targeting oligodendrocyte Aβ production could be a promising therapeutic strategy for treating AD.

Fig 1. Components required to produce Aβ are expressed at high levels in oligodendrocytes, but not other glial cells.

Heatmaps showing the log₂ (norm count) z-score of genes of interest across different cell types [Excitatory neurons (Ex), Inhibitory neurons (In), Astrocytes (Ast), Microglia (Mic), Oligodendrocyte Precursor Cells (OPC), and Oligodendrocytes (Oli)], from 4 publicly available human single nucleus RNA sequencing datasets. APP, BACE1, and all components of γ-secretase (PSEN1, PSENEN, NCSTN, APH1A, APH1B) with the exception of PSEN2 (which is interchangeable with PSEN1) are expressed at high levels in oligodendrocytes, many at higher levels than any other cell type. (a) Data from Zhou and colleagues [10] was generated using tissue from the motor cortex of 36 subjects including controls, AD patients, and those carrying TREM2 variants. (b) Data from Bakken and colleagues [11] was generated using tissue from the motor cortex of 5 control subjects. (c) Data from Lake and colleagues [12] was generated using tissue from the frontal cortex of 6 control subjects. (d) Data from Mathys and colleagues [5] was generated using tissue from the prefrontal cortex of 48 subjects with varying degrees of AD-related pathology. (e) Representative immunofluorescent images showing APP (green), oligodendroglial marker Olig2 (red), and DAPI (nuclei; blue) in the cortex of a 4-month-old wild-type mouse. Scale bar = 10 μm. (f) Representative immunofluorescent images showing BACE1 (green), oligodendroglial marker Olig2 (red), and DAPI (nuclei; blue) in the cortex of a 4-month-old wild-type mouse. Scale bar = 10 μm.

Fig 2. Human sporadic AD brains have more oligodendrocytes capable of producing Aβ compared to controls.

(a) Fluorescence images from Layers 5/6 of control (top) and sporadic AD (sAD; bottom) postmortem human prefrontal cortex labelled for MBP (oligodendrocyte-specific gene; green), BACE1 (yellow), APP (red), Aβ (identified by 6E10-antibody; white), and DAPI (nuclei; blue). Aβ-capable oligodendrocytes (MBP⁺ BACE1⁺ APP⁺ nuclei) are marked with white arrowheads. Scale bar = 25 μm. (b) Quantification showing that approximately 80% of oligodendrocytes are capable of producing Aβ in both control and sAD brains. (c) Quantification showing an increase in the number of oligodendrocytes in Layers 5/6 of sAD brains. (d) Quantification showing significantly more Aβ-capable oligodendrocytes in sAD brains than controls. In (b–d), each data point represents a single brain (n = 4 control brains, n = 5 sAD brains) with bars representing mean ± SEM; unpaired t test: t(7) = 0.568, 3.058, 2.581 in (b–d), respectively. Source data are available in S1 Data.

Fig 3. Human oligodendrocytes produce soluble Aβ and aggregates at higher levels than neurons.

(a) Fluorescent image of human iPSC-derived oligodendrocyte culture immunolabelled for MBP (green), OLIG2 (marker of all oligodendroglia; red), and DAPI (nuclei; blue). This example shows multiple mature oligodendrocytes extending MBP⁺ myelin processes, while the majority of other cells are OLIG2⁺ MBP^- OPCs. Scale bar = 25 μm. (b) Quantification by ELISA showing a significant reduction in the amount of Aβ₄₀ produced (as a % of the amount produced prior to treatment) by human oligodendrocytes when treated with BACE1 inhibitor (NB-360) compared to vehicle control (DMSO). (c) ELISA data showing more Aβ₄₀ produced by oligodendrocytes than neurons derived from the same fAD human-iPSC lines. (d) Quantification by ELISA showing higher Aβ₄₂/Aβ₄₀ ratio produced by oligodendrocytes compared to neurons derived from the same fAD human-iPSC lines. (e) DNA-PAINT super-resolution images showing more Aβ aggregates in media from oligodendrocytes (right) compared to neurons (left). Scale bar = 1 μm. (f) Quantification showing oligodendrocytes produce a higher proportion of Aβ as aggregates compared to neurons derived from the same fAD human-iPSC lines. (g) Violin plots showing the length of aggregates produced by neurons and oligodendrocytes (mean ± standard deviation: Neurons 79.8 nm ± 58.7 nm, Oligodendrocytes 77.7 nm ± 58.5 nm; n = 18,550 aggregates from 3 neuron lines [3 independent inductions per line] and 24,719 aggregates from 3 oligodendrocyte lines [3 independent inductions per line]). In (b), each data point represents the average of 2 independent inductions from each of a different cell line (n = 3 cell lines), with bars showing mean ± SEM; paired t test: t(2) = 9.613. In (c, d, f), each data point represents the average of 4 (for PSEN1 WT line neurons in d) or 3 independent inductions from each of a different cell line (n = 3 cell lines), showing mean ± SEM with each cell line shown in a different colour to highlight pairing (blue: PSEN1 WT; green: PSEN1 int4del; red: PSEN1 R278I); paired t test: t(2) = 4.435, 5.411 in (c, d), respectively; ratio paired t test: t(2) = 7.117 in (f). Un-pooled data for (b, c, d, f) are shown in S6 Fig. Source data are available in S1 Data. fAD, familial AD; iPSC, induced pluripotent stem cell; MBP, myelin basic protein; OPC, oligodendrocyte precursor cell.

Fig 4. Genetic suppression of oligodendrocyte Aβ production reduces Aβ plaques in the App^NL-G-F mouse model of AD.

(a) Immunofluorescent images showing Aβ (6E10 antibody; green) and DAPI in the retrosplenial cortex of App^NL-G-F control mice (left), App^NL-G-F mice with BACE1 knocked out (KO) specifically in oligodendrocytes (middle; Oligo-KO), and App^NL-G-F mice with BACE1 KO specifically in neurons (right; Neuron-KO). Images show all layers of the cortex, with Layer 6 at the bottom. Scale bar = 100 μm. (b) Quantification of the number of Aβ⁺ plaques across the visual, retrosplenial, and motor cortical areas, showing a 25% reduction in Oligo-KO mice compared to App^NL-G-F, and an elimination of plaques in Neuron-KO mice. (c) Quantification of the total area of Aβ⁺ plaques across the visual, retrosplenial, and somatomotor cortical areas, showing approximately 25% reduction in Oligo-KO mice compared to App^NL-G-F and an elimination of plaques in Neuron-KO mice. In (b and c), data points represent individual mice (n = 8 App^NL-G-F, 9 Oligo-KO, 4 Neuron-KO) with bars showing mean ± SEM. One-way ANOVA with Dunnet’s post hoc tests: F(2,18) = 25.38(b), 13.24(c); p < 0.0001(b), p = 0.0003(c). Source data are available in S1 Data.

Fig 5. Genetic suppression of oligodendrocyte Aβ production rescues neuronal dysfunction in the App^NL-G-F mouse model of AD, while oligodendrocyte-derived media promotes neuronal dysfunction in vivo.

(a) Raster plots from Neuropixels recordings showing spontaneous neuronal firing in 20 randomly selected cortical neurons/units from 3-month-old awake WT (top), App^NL-G-F (middle) and oligodendrocyte BACE1 KO (bottom) mice, illustrating rescue of hyperactivity phenotype in oligodendrocyte BACE1 KO mice to WT levels. Units are sorted from high MFRs to low (top to bottom) and plots show a 6s resting state period. (b) Quantification of the MFR showing significantly reduced activity in cortical neurons of both Oligo-KO and Neuron-KO mice compared to App^NL-G-F mice, with firing rates returning to levels observed in WT controls (WT vs. Oligo-KO: p > 0.9999; WT vs. Neuron-KO: p > 0.9999). In (b), data points represent individual neurons/units (n = 82 WT, 134 App^NL-G-F, 165 Oligo-KO, 75 Neuron-KO) across 4 (WT) or 3 mice per group with the shaded area representing smoothed distribution. Median is shown by a thick black line, and quartiles indicated with grey lines. Kruskal–Wallis test (H(3, n = 456) = 15.33, p = 0.0016) with Dunn’s post hoc tests. (c) Raster plots showing neuronal firing in the same 20 cortical neurons/units at baseline (left) and during injection of oligodendrocyte conditioned media containing soluble Aβ aggregates (right) in a 4-month-old WT mouse, illustrating the strong increase in neuronal activity upon exposure to oligodendrocyte-derived Aβ. (d) Quantification showing an increase in neuronal firing rates upon local injection of oligodendrocyte conditioned media (Aβ₄₀ concentration by ELISA: 133 pM) into retrosplenial cortex, compared to injection of either the same media which had been immunodepleted of Aβ (Aβ₄₀ concentration by ELISA: 24 pM) or media from BACE1 inhibitor-treated oligodendrocytes (Aβ₄₀ concentration by ELISA: 18 pM). In (d), data points represent individual mice (n = 5 oligodendrocyte conditioned media, 4 other groups) with mean ± SEM shown. One-way ANOVA (F(2,10) = 11.23, p = 0.0028) with Tukey’s post hoc tests. Source data are available in S1 Data. AD, Alzheimer’s disease; KO, knockout; MFR, mean firing rate.