Alzheimer's disease associated isoforms of human CD33 distinctively modulate microglial cell responses in 5XFAD mice

Abstract

Microglia play diverse pathophysiological roles in Alzheimer's disease (AD), with genetic susceptibility factors skewing microglial cell function to influence AD risk. CD33 is an immunomodulatory receptor associated with AD susceptibility through a single nucleotide polymorphism that modulates mRNA splicing, skewing protein expression from a long protein isoform (CD33M) to a short isoform (CD33m). Understanding how human CD33 isoforms differentially impact microglial cell function in vivo has been challenging due to functional divergence of CD33 between mice and humans. We address this challenge by studying transgenic mice expressing either of the human CD33 isoforms crossed with the 5XFAD mouse model of amyloidosis and find that human CD33 isoforms have opposing effects on the response of microglia to amyloid-β (Aβ) deposition. Mice expressing CD33M have increased Aβ levels, more diffuse plaques, fewer disease-associated microglia, and more dystrophic neurites compared to 5XFAD control mice. Conversely, CD33m promotes plaque compaction and microglia-plaque contacts, and minimizes neuritic plaque pathology, highlighting an AD protective role for this isoform. Protective phenotypes driven by CD33m are detected at an earlier timepoint compared to the more aggressive pathology in CD33M mice that appears at a later timepoint, suggesting that CD33m has a more prominent impact on microglia cell function at earlier stages of disease progression. In addition to divergent roles in modulating phagocytosis, scRNAseq and proteomics analyses demonstrate that CD33m⁺ microglia upregulate nestin, an intermediate filament involved in cell migration, at plaque contact sites. Overall, our work provides new functional insights into how CD33, as a top genetic susceptibility factor for AD, modulates microglial cell function.

Keywords: Alzheimer’s disease; Amyloid-beta; CD33; Microglia; Plaque compaction; Siglec.

Fig. 1

hCD33 isoforms have opposing effects on Aβ accumulation in 5XFAD mice. a Representative epifluorescent images of Aβ deposition in 5XFAD mice at 4 months by IF imaging with the anti-Aβ antibody (clone MOAB2, white) and Hoechst (blue). b Quantification of total Aβ levels at 4 months in pooled male (squares; n = 20, 18, and 12 for control, CD33M, and CD33m, respectively) and female (circles; n = 20, 15, and 15 for control, CD33M, and CD33m, respectively) mice. c Representative epifluorescent images of Aβ deposition in 5XFAD mice at 8 months. d Quantification of total Aβ levels at 8 months in pooled male (squares; n = 19, 11, and 10 for control, CD33M, and CD33m, respectively) and female (circles; n = 22, 14, and 13 for control, CD33M, and CD33m, respectively) mice. Scale bar = 1000 µm

Fig. 2

Altered Aβ plaque composition by hCD33 isoforms. a Representative confocal fluorescent images showing the different degrees of plaque compaction by co-staining for total Aβ and ThioS. A diffuse plaque without ThioS staining at the far left, a plaque with small ThioS core in the middle, and a highly compact plaque at the far right are presented. Scale bar = 20 µm (b-e). Quantification of Aβ deposits containing a ThioS.⁺ core. Representative epifluorescent images of Aβ deposits in the dorsal subiculum of 5XFAD mice at (b) 4 and (d) 8 months, with co-staining with anti-Aβ antibody (white) and ThioS (blue). Scale bar = 50 µm. Quantification for pooled males (squares; n = 5 per genotype) and female (circles; n = 5/genotype) mice at (c) 4 and (e) 8 months. f Representative confocal fluorescent images of plaque composition in the dorsal subiculum of 8 months old CD33M and CD33m mice. Scale bar = 20 µm (g,h) Quantification of plaque composition (ratio of Aβ over ThioS levels) in the dorsal subiculum of 5XFAD mice at (g) 4 and (h) 8 months. A total of 300 plaques per mouse, from 5 males and 5 females, were quantified for each group. i Biochemical characterization of PK-sensitivity of insoluble Aβ_1-42 from 5XFAD mice. j Quantification of the percentage of insoluble Aβ_1-42 levels after PK digestion compared to not treated with PK. k Quantification of the ratio of ApoE to Aβ_1-42 in the insoluble fraction from 5XFAD mice (n = 9, 10, and 9 mice of control, CD33M, and CD33m genotypes, respectively)

Fig. 3

hCD33 isoforms modulate differential response of microglia to Aβ. a Representative confocal fluorescent images of CD68 staining of control, CD33M, or CD33m 5XFAD mice at 8 months. IF images are co-stained with anti-Aβ antibody (white), anti-Iba1 (red), and Hoechst (blue). Scale bar = 20 µm. b Quantification of total CD68 area measured as the area of CD68 within Iba1 per FOV. A total of 8 FOV, were quantified for each group (n = 5/genotype). c Representative confocal fluorescent images of plaque associated microglia in control, CD33M, and CD33m mice. IF images are co-stained with anti-Aβ antibody (white), anti-Iba1 (red), and Hoechst (blue). Scale bar = 20 µm. d,e Quantification of plaque associated microglia normalized to µm² of plaque area at 8 month. A total of 150 plaques in cortex (d) and 300 plaques in subiculum (e), of 10 mice (5 males and 5 females per group) were analyzed. f Representative confocal fluorescent images of microglia-ThioS core interface in control, CD33M, and CD33m mice. IF images are co-stained with anti-Iba1 (red), and ThioS (blue). Scale bar = 20 µm. g,h Quantification of plaque-microglia interface in 5XFAD mice at 8 months, measured as the percentage area of ThioS perimeter with overlapping Iba1 signal. A total of 150 plaques in cortex (g) and 300 plaques in subiculum (h), of 10 mice (5 males and 5 females per genotype) were analyzed. i Representative confocal fluorescent images of Ki-67 staining of microglia in 8 months-old control, CD33m and CD33m mice. IF images are co-stained with anti-Ki-67 (yellow), anti-Iba1 (red), and Hoechst (blue). Scale bar = 100 µm. j Quantification of Ki-67⁺ microglia per FOV measured as the number of macroglia with clear Ki-67 signal in the nuclei. A total of 5 confocal images of subiculum or cortex per mouse were analyzed (n = 5 mice/genotype). k Number of Ki-67 + microglia per FOV normalized to average PAM density of subiculum or cortex for each genotype

Fig. 4

Quantitative proteomics reveal changes in protein abundance in the brain of 5XFAD mice expressing hCD33 isoforms. a Schematic representation of the mass spectrometry workflow used for the assessment of global proteome changes in the brain of 5XFAD mice expressing CD33M or CD33m compared to control at 8 months. A total of 5 male mice were used per genotype. On-column digestion and data-independent acquisition mass spectrometry (DIA-MS) was used to quantify protein levels. b Correlation of the proteomic changes in the mouse brain for all proteins identified by DIA-MS. The triplot shows protein abundance changes calculated from each pairwise comparison (CD33M vs control, CD33m vs control, and CD33m vs CD33M). Identifications closer to the vertex correspond to proteins with higher association to that genotype, while identifications closer to the center of the triangle indicate similar levels of that protein in all comparisons. Proteins with significant changes (fold change ≥ 1.5 and p-value ≤ 0.05) are highlighted. The proteins increased in mouse brains expressing CD33m are colored in blue, while the proteins increased in the CD33M isoform are colored in red. c Quantification of nestin in each individual sample as measured by mass spectrometry. d Representative confocal fluorescent images of nestin staining of PAM in 5XFAD mice at 8 months in control, CD33m, and CD33m mice. IF images are co-stained with anti-nestin (yellow), anti-Iba1 (red), and ThioS (blue). Scale bar = 20 µm. e Quantification of nestin levels in PAM of control, CD33M and CD33m mice. Approximately 30 plaques from 3 mice were quantified per genotype

Fig. 5

Single-cell RNA sequencing reveals Ccl3 and Trem2 DAM enriched in the CD33m genotype. a Unsupervised and iterative machine-learning based clustering of 15,200 microglia (Hexb⁺Fcrls⁺Tmem119⁺Sall1⁺) and BAM (Ms4a7⁺Mrc1⁺Lyve1⁺Timd4⁺) collected from 5XFAD control, CD33M 5XFAD, CD33m 5XFAD, and control non-5XFAD mice. Microglia from three homeostatic (HM1-3), two transitioning (TM1-2), two RNA binding protein (RBM1-2) subpopulations along with the disease-enriched interferon responsive (IRM), myelin transcript enriched (MTEM), and disease associated (DAM) clusters.  b-e UMAPs of individual samples showing (b) 5XFAD control, c CD33m 5XFAD, d CD33M 5XFAD, and e control non-5xFAD. f,g Separation of each cluster by (f) proportion of cells and (g) absolute number of cells belonging to each genotype. DAM, specifically Trem2 DAM, are enriched in the CD33m group and reduced in the CD33M group. h Differential gene expression per cluster used to define microglial subpopulations. HM are characterized by homeostatic genes (P2ry12, Tmem119), RBM by genes related to RNA binding (Son, Fus), TM by a combination of homeostatic (P2ry12, Tmem119), complement (C1qa, C1qb), and proliferative (MKi-67, Top2a) genes. DAM are defined by Clec7a and Apoe expression and further delineated by expression of Ccl3 and Trem2. i UMAP showing enriched Nes expression within the Ccl3 DAM cluster. j-l Violin plots showing the upregulation of (j) Nes,  k Jun, and l Fos in the CD33m⁺ 5XFAD group relative to 5XFAD control and CD33M⁺ 5XFAD

Fig. 6

hCD33 isoforms impact neuronal health and animal behavior. a Representative immunofluorescent confocal images of dystrophic neurites stained with lysosomal marker LAMP1 (yellow) in the dorsal subiculum of control, CD33M, and CD33m mice at 8 months. Scale bar = 50 µm. b Quantification of total area of DNs in the dorsal subiculum. A total of 5 male and 5 female mice per genotype were used. c Higher magnification immunofluorescent confocal images of individual DNs in the dorsal subiculum of control, CD33M and CD33m mice at 8 months. Scale bar = 50 µm. d Quantification of DN area in individual neuritic plaques of 5XFAD control, CD33M, and CD33m mice at 8 months. Individual neuritic plaques within the subiculum and parts of frontal cortex adjacent to the subiculum were analyzed. A total of 300 plaques from 5 males and 5 females per genotype were quantified. e Quantification of time spent in light from a Light–Dark box assay carried out on 5XFAD and CD33m⁺ mice at 8 months (n = 13 female and 7 male mice for the 5XFAD control group, and n = 11 female and 10 male mice for the CD33m⁺ 5XFAD group were used). f Quantification of alternation index from a spontaneous Y-maze assay carried out on 5XFAD control and CD33m⁺ 5XFAD mice at 8 months (n = 13 female and 7 male mice for the 5XFAD control group, and n = 11 female and 10 male mice for the CD33m⁺ 5XFAD group were used). g Quantification of maximum alternation from the spontaneous Y-maze carried out on 5XFAD control and CD33m⁺ 5XFAD mice at 8 months