LILRB2-mediated TREM2 signaling inhibition suppresses microglia functions

Abstract

Background: Microglia plays crucial roles in Alzheimer's disease (AD) development. Triggering receptor expressed on myeloid cells 2 (TREM2) in association with DAP12 mediates signaling affecting microglia function. Here we study the negative regulation of TREM2 functions by leukocyte immunoglobulin-like receptor subfamily B member 2 (LILRB2), an inhibitory receptor bearing ITIM motifs.

Methods: To specifically interrogate LILRB2-ligand (oAβ and PS) interactions and microglia functions, we generated potent antagonistic LILRB2 antibodies with sub-nanomolar level activities. The biological effects of LILRB2 antagonist antibody (Ab29) were studied in human induced pluripotent stem cell (iPSC)-derived microglia (hMGLs) for migration, oAβ phagocytosis, and upregulation of inflammatory cytokines. Effects of the LILRB2 antagonist antibody on microglial responses to amyloid plaques were further studied in vivo using stereotaxic grafted microglia in 5XFAD mice.

Results: We confirmed the expression of both LILRB2 and TREM2 in human brain microglia using immunofluorescence. Upon co-ligation of the LILRB2 and TREM2 by shared ligands oAβ or PS, TREM2 signaling was significantly inhibited. We identified a monoclonal antibody (Ab29) that blocks LILRB2/ligand interactions and prevents TREM2 signaling inhibition mediated by LILRB2. Further, Ab29 enhanced microglia phagocytosis, TREM2 signaling, migration, and cytokine responses to the oAβ-lipoprotein complex in hMGL and microglia cell line HMC3. In vivo studies showed significantly enhanced clustering of microglia around plaques with a prominent increase in microglial amyloid plaque phagocytosis when 5XFAD mice were treated with Ab29.

Conclusions: This study revealed for the first time the molecular mechanisms of LILRB2-mediated inhibition of TREM2 signaling in microglia and demonstrated a novel approach of enhancing TREM2-mediated microglia functions by blocking LILRB2-ligand interactions. Translationally, a LILRB2 antagonist antibody completely rescued the inhibition of TREM2 signaling by LILRB2, suggesting a novel therapeutic strategy for improving microglial functions.

Keywords: 5XFAD mice; Alzheimer’s disease; Amyloid; Antibody; ITAM; ITIM; LILRB2; Microglia; Phagocytosis; TREM2.

Fig. 1

LILRB2 and TREM2 are expressed on human microglia and they share ligands oAβ and PS. a. Immunofluorescence staining of AD patient brain tissue for LILRB2, IBA1 (microglia marker), amyloid plaques (6E10), GFAP (astrocyte marker), and TREM2. Scale bar = 20 μm. b. Titration curves of oAβ binding to LILRB2 and TREM2 Fc fusion proteins by ELISA. Data are presented as mean ± SD (n = 3 independent experiments). c. Titration curves of PS binding to LILRB2 and TREM2 Fc fusion proteins by ELISA. Data are presented as mean ± SD (n = 3 independent experiments). d. Representative immunostaining images of oAβ and PS binding to LILRB2 and TREM2 expressed on 293 T cell surface. Scale bar represents 5 μm. e-h. oAβ or PS binding to LILRB2 or TREM2 as measured by BLI. In the association stage, protein A sensor-captured Fc fusion proteins (LILRB2-Fc: e and f, TREM2-Fc: g and h) were incubated with oAβ (1 μM, e and g) or PS liposomes (1 mM, f and h), and the amount of oAβ (e and g) or PS (f and h) bound onto the sensors was presented as wavelength shift in nanometers (nm). The red dotted vertical line marks the transit from association stage to dissociation stage, where the sensors were dipped into kinetics buffer without ligands allowing free dissociation

Fig. 2

The shared ligands oAβ and PS induce co-ligation of LILRB2 and TREM2 and inhibit TREM2 signaling. a. Schematic illustration of the cell line LILRB2-TREM2 showing the design of BiFc assay with LILRB2 and TREM2 fusion constructs. b-c. oAβ or PS-induced co-ligation of TREM2 and LILRB2 as measured by the BiFc assay. HEK293T cells co-expressing LILRB2-N173 Venus and TREM2-C155 Venus were incubated with plate-coated oAβ (b) or PS (c). Y-axis shows the MFI signals from complemented Venus. Data are presented as mean ± SD (n = 3 independent experiments). d. Representative immunofluorescence images showing oAβ-lipid-induced co-ligation of TREM2 and LILRB2. Scale bar represents 5 μm. e. Schematic diagram showing co-ligation of LILRB2 and TREM2 by oAβ or PS inhibits TREM2 signaling in reporter cell assays. The ITIM motifs of LILRB2 attenuate signaling generated by ITAM motifs of TREM2 upon cross-linking by shared ligands oAβ or PS. f-g. oAβ or PS-induced TREM2 signaling of reporter cells co-expressing LILRB2 and TREM2. Reporter cells expressing corresponding receptors (x-axis) were incubated with oAβ (f) or PS (g). Y-axis is TREM2 signaling of treatment groups normalized based on the percentage of GFP⁺ reporter cells expressing only TREM2 (set to 100%). Data are presented as mean ± SD (n = 4 independent experiments). h. oAβ-lipoprotein complex phagocytosis profile of BV2 cells expressing the LILRB2 transgenes (x-axis). BV2 cells were incubated with FAM-oAβ-lipoprotein complex. Phagocytosis was quantified by flow cytometry, with the MFI shown. Negative phagocytosis control included 10 μM CytoD together with indicated treatments. Data are presented as mean ± SD (n = 4 independent experiments)

Fig. 3

LILRB2 targeting antibody Ab29 showed potent blocking activity, high affinity, and specificity. a-b. Titration of blocking the activity of Ab29 against oAβ or PS-LILRB2 interactions. Plate-coated oAβ (a) or PS (b) was incubated with LILRB2-chimeric reporter cells under the presence of increasing concentrations of Ab29 or control IgG. The activation of LILRB2-chimeric reporter cells was observed as a percentage of GFP⁺ cells. Data are presented as mean ± SD (n = 3 independent experiments). c. Representative immunostaining images showing Ab29 blocks oAβ and PS from binding to LILRB2 expressed on the cell surface. Scale bar represents 5 μm. d. Ab29 blocks LILRB2 binding with oAβ as measured by BLI. LILRB2 was loaded onto protein A sensors via binding with sensor-captured Ab29. The LILRB2-loaded sensors were then incubated with biotinylated oAβ (1 μM) and the binding signals were amplified with streptavidin. The amount of oAβ bound onto the sensors was presented as wavelength shift in nanometers (nm). The red dotted vertical line marks the transit from incubation with biotinylated oAβ to signal amplification by streptavidin. e. Ab29 blocks LILRB2 binding with PS liposomes as measured by BLI. LILRB2 was loaded onto protein A sensors via binding with sensor-captured Ab29. The LILRB2-loaded sensors were incubated with PS liposomes (1 mM) (association stage) and then the sensors were dipped into kinetics buffer without PS allowing the bound PS to freely dissociate (dissociation stage). The amount of PS bound onto the sensors was presented as wavelength shift in nanometers (nm). The red dotted vertical line marks the transit from association stage to dissociation stage. f. Cross-reactivity of Ab29 against other LILRB and LILRA family receptors as measured by ELISA. Plate-coated LILRB and LILRA family receptors (shown in the x-axis) were incubated with Ab29 (4 nM). The amount of bound antibody was detected by anti-human F(ab)2 HRP, and the values are shown as OD₄₅₀. LILRB2-Fc was included as the positive control. Data are presented as mean ± SD (n = 4 independent experiments)

Fig. 4

LILRB2 blocking antibody Ab29 completely rescued LILRB2-mediated TREM2 signaling inhibition and promoted microglial phagocytosis by blocking LILRB2/TREM2 co-ligation. a. Schematic diagram showing the design of a BiFc assay studying the effect of LILRB2 and TREM2 blocking antibodies in preventing the ligand-mediated receptor co-ligation. b and c. LILRB2 blocking antibodies abolish oAβ or PS-induced co-ligation of LILRB2 and TREM2. Antibodies at 10 μg/mL were incubated with HEK293T cells co-expressing LILRB2-N173 Venus and TREM2-C155 Venus plus plate-coated oAβ (b) or PS (c). The MFI signals from complemented Venus are shown. Data are presented as mean ± SD (n = 3 independent experiments). d. Representative immunofluorescence images showing Ab29 blocks oAβ-lipid from inducing LILRB2-TREM2 co-ligation. Scale bar represents 5 μm. e. schematic diagram showing LILRB2 blocking antibody rescues TREM2 signaling inhibition by blocking LILRB2-ligand interactions. The assay is depicted as an NFAT-GFP reporter assay. The ITIM motifs of LILRB2 block signaling generated by ITAM motifs of TREM2 upon cross-linking of the two receptors by shared ligands oAβ or PS. f-g. Titration of Ab29 in rescuing of oAβ or PS-LILRB2-induced inhibition of TREM2 signaling. Plate-coated oAβ (f) or PS (g) was incubated with LILRB2/TREM2 reporter cells under the presence of increasing concentrations of Ab29 or control IgG. The activation of LILRB2/TREM2 reporter cells was observed as percentages of GFP⁺ cells. TREM2 signaling of treatment groups was normalized based on the percentage of GFP⁺ reporter cells expressing only TREM2 (set to 100%). Data are presented as mean ± SD (n = 4 independent experiments). h. Ab29 promotes oAβ-lipoprotein complex phagocytosis by HMC3 human microglial cell line. HMC3 cells were incubated with FAM-oAβ-lipoprotein complex and indicated antibodies (treatment table). Phagocytosis was quantified by flow cytometry, with the MFI shown. Negative phagocytosis control included 10 μM CytoD together with indicated treatments. Data are presented as mean ± SD (n = 4 independent experiments). i. Ab29 improves TREM2 pathway signaling as indicated by the increased pSYK level. HMC3 cells were incubated with oAβ-lipoprotein complex with indicated treatments for 1 hour. Immunoblot of phosphorylated SYK (pSYK), total SYK of HMC3 upon treatment, with β-actin as the loading control. Data are presented as mean ± SD (n = 3 independent experiments). j. Ab29 blocks LILRB2 pathway signaling as indicated by the decreased pSHP1 level. Immunoblot of phosphorylated SHP1 (pSHP1), SHP1 of HMC3 upon incubation with oAβ-lipoprotein complex with indicated treatments for 1 hour. Results are from anti-LILRB2 immunoprecipitation with LILRB2 as the loading control. Data are presented as mean ± SD (n = 3 independent experiments)

Fig. 5

LILRB2 blocking antibody Ab29 promoted hMGL migration, phagocytosis, and cytokine responses. a. Surface expression of TREM2 and LILRB2 on hMGL cells. The fluorescent signals (x-axis) were plotted as a histogram with normalized cell percentage in the y-axis. b. MFI of hMGL cells labeled by indicated antibodies (x-axis) as shown in a. Data are presented as mean ± SD (n = 3 independent experiments). c. Ab29 promotes hMGL cell migration. hMGL cells were plated onto transwell chamber inserts, and hMGL cell migration toward the oAβ-lipoprotein complex in the receiving chamber was measured with treatments shown. For quantification, crystal violet-stained hMGL cells were quantified by eluting crystal violet in 33% acetic acid and measured the absorbance at 590 nm using a plate reader. Percentages of migrated cells over total are plotted. Data are presented as mean ± SD (N = 3 independent experiments). d. Representative immunostaining images showing Ab29 promotes hMGL migration induced by oAβ-lipid. The hMGLs were pre-labeled with CFSE to visualize the cells under the fluorescence microscope. Scale bar = 100 μm. e. Ab29 promotes oAβ-lipoprotein complex phagocytosis by hMGLs. hMGL cells were incubated with FAM-oAβ-lipoprotein complex and indicated antibodies. Phagocytosis was quantified by flow cytometry, with the MFI shown. Negative phagocytosis control included 10 μM CytoD together with indicated treatments. Data are presented as mean ± SD (n = 3 independent experiments). f. Representative immunostaining images showing Ab29 promotes hMGL phagocytosis of oAβ-lipid. Scale bar represents 5 μm. g. PLA assay indicates that oAβ-lipid induces clustering of LILRB2 and TREM2, while Ab29 blocks the clustering. hMGLs were incubated with oAβ-lipid (100 nM) and antibodies (Ctrl IgG or Ab29, 10 μg/mL), and the LILRB2 and TREM2 clustering were detected by PLA assay. The red color indicates the PLA signals detected, and the blue color represents the cell nucleus. Scale bar = 20 μm. h-k. Ab29 promotes oAβ-lipoprotein complex-induced inflammatory cytokine responses in hMGLs. hMGL cells were incubated with oAβ-lipoprotein complex with indicated treatments. The relative mRNA levels of TNF-α (g), IL-6 (h), CCL3 (i), and IL-1β (j) were determined using qRT-PCR with GAPDH as the internal control. Data are presented as mean ± SD (n = 3 independent experiments)

Fig. 6

Ab29 increases microglial responses to amyloid plaques in vivo. a. schematic diagram showing the stereotaxic injection of hMGLs and Ab29 in 5XFAD mice. The hMGLs were injected into the lateral ventricles through stereotaxic injections together with antibodies. The 5XFAD mice were sacrificed 96 hours after the injection and brains were harvested for analysis. b. Antibody concentrations in perfused 5XFAD at the experiment endpoint as described in a. n = 5 independent mice. c. Representative amyloid plaque-microglia co-localization immunofluorescence staining of 5-month-old 5XFAD mice cortex as treated in a. Scale bar = 20 μm. IBA1, microglia marker; 6E10, amyloid plaque marker; the nucleus is labeled by TO-PRO-3. d. Quantification of IBA1 area within 30 μm of amyloid plaques in the cortex of mice treated as described in a. n = 5 independent mice. e. Representative amyloid plaque-CD68 co-localization immunofluorescence staining of the cortex of 5XFAD mice treated as described in a. CD68, microglia phagocytic marker; the nucleus is labeled by TO-PRO-3. Scale bar = 20 μm. f. Quantification of CD68 area within 30 μm of amyloid plaques in the cortex of mice treated as described in a. n = 5 independent mice. g. Quantification of Aβ co-localized with CD68 per plaque in the cortex of mice treated as described in a. n = 5 independent mice. For all the data presented, bar graphs with error bars represent mean ± SD. For the statistical analysis, *** P < 0.001, two-tailed Student t-test

Fig. 7

Key domains and amino acid residues of LILRB2 for ligands and antibody binding. a. Schematic diagram showing the design of domain-swapping mutants of LILRB2. B1D1 and B1D2 mutants were generated by replacing the D1 and D2 domains of LILRB2 with their corresponding domains from LILRB1, respectively. b-d. oAβ, PS, and Ab29 binding to LILRB2 mutants as measured on chimeric reporter cells. Chimeric NFAT-GFP reporter cells expressing individual mutants of LILRB2 (listed in x-axis) were incubated with oAβ (b) PS (c), and Ab29 (d). The activation of reporter cells is shown as the percentage of GFP⁺ cells. Data are presented as mean ± SD (n = 3 independent experiments). e. Alignment of D1 and D2 domains of LILRB2 and LILRB1. Amino acid residues are numbered according to Uniprot entry Q8N423. Identical residues are shown in white text with red background. Similar residues are shown in black text with a yellow background. Non-conserved residues are in white background. Individual domains of LILRB2 are also labeled. The red arrows indicate the single-amino acid mutations designed for testing. f. Crystal structure of LILRB2 (2gw5) visualized in UCSF Chimera showing the D1 and D2 domains as ribbons. Selected residues for mutation studies are shown in ball-and-stick with side chains displayed. Three-letter code names and residue numbers are also labeled following Uniprot entry Q8N423. g. Schematic diagram showing the distribution of single-amino acid mutations in LILRB2 D1 and D2 domains