A TREM2-activating antibody with a blood-brain barrier transport vehicle enhances microglial metabolism in Alzheimer's disease models

Abstract

Loss-of-function variants of TREM2 are associated with increased risk of Alzheimer's disease (AD), suggesting that activation of this innate immune receptor may be a useful therapeutic strategy. Here we describe a high-affinity human TREM2-activating antibody engineered with a monovalent transferrin receptor (TfR) binding site, termed antibody transport vehicle (ATV), to facilitate blood-brain barrier transcytosis. Upon peripheral delivery in mice, ATV:TREM2 showed improved brain biodistribution and enhanced signaling compared to a standard anti-TREM2 antibody. In human induced pluripotent stem cell (iPSC)-derived microglia, ATV:TREM2 induced proliferation and improved mitochondrial metabolism. Single-cell RNA sequencing and morphometry revealed that ATV:TREM2 shifted microglia to metabolically responsive states, which were distinct from those induced by amyloid pathology. In an AD mouse model, ATV:TREM2 boosted brain microglial activity and glucose metabolism. Thus, ATV:TREM2 represents a promising approach to improve microglial function and treat brain hypometabolism found in patients with AD.

Fig. 1. ATV:4D9 induces temporally dynamic microglial states distinct from amyloid pathology by single-cell analysis.

a, Antibodies were detected by human IgG ELISA in whole brain lysates 1 day after IV dose of 10 mg kg⁻¹ ATV:4D9 or 4D9 (n = 5 mice). b, Designs for WT;TfR^mu/hu and App^SAA;TfR^mu/hu studies. Mice were injected with a single IV 10 mg kg⁻¹ dose of ATV:4D9 or ATV:ISO and sacrificed at indicated timepoints (n = 3 WT; TfR^mu/hu mice and n = 4 App^SAA;TfR^mu/hu mice). c, Integrated UMAP projection of 49,000 total cells from all mice in both studies. The WT;TfR^mu/hu dataset consisted of 102,043 cells, and the App^SAA;TfR^mu/hu dataset consisted of 74,758 cells (Extended Data Fig. 1a,b). Five distinct clusters of microglia were identified. d, Stacked bar plots of clusters distributed per group for both studies. Clusters are shown as percentages of the whole microglial compartment averaged for each biological replicate. e, UMAP projection split by group for cluster distribution. Data from WT;TfR^mu/hu mice are boxed with solid lines, and data from App^SAA;TfR^mu/hu mice are boxed with dashed lines. f, Heat map of average log₂FC in each cluster compared to the ‘homeostatic’ cluster. Pseudobulk expression was generated by summing every counts per gene. Each mouse was treated as a biological replicate, and differential expression (DE) was performed for each cluster versus ‘homeostatic’ using limma. g, Scatter plots of the comparison of versus ‘homeostatic’ log₂FC for every gene between clusters 3 and 2 (top) and clusters 3 and 5 (bottom). In the top-right quadrant, genes falling below the dashed line are more upregulated in cluster 3, whereas genes above the dashed line are further upregulated in cluster 2 (top) or cluster 5 (bottom). Canonical DAM genes and other genes of interest are highlighted in orange. h, Dot plot showing GSEA for each cluster using DE versus ‘homeostatic.’ Signatures were taken from the hallmark gene signatures collection. Dot size is inversely proportional to log₁₀(corrected P value), and color indicates direction. No dot indicates a non-significant result. FC, fold change. Source data

Fig. 2. ATV:TREM2 is a novel activating antibody that potentiates receptor clustering and endocytosis to enhance TREM2 signaling.

a, Antibody schematic of ATV:TREM2 with human TREM2 Fab affinity and ATV binding site in the Fc domain and effectorless Fc mutations. b, Fluorescence polarization (FP)-based detection of TREM2 stalk peptide cleavage by recombinant ADAM17 (n = 4 independent experiments; two-tailed unpaired t-test, mean ± s.e.m.). c, ATV:TREM2 and PS liposome co-treatment enhanced pSyk in WT iMG (n = 3 independent experiments; Tukey’s multiple comparisons test, mean ± s.e.m.). d, Schematic illustrating ATV and TREM2 Fab valency effects on pSyk signaling. Antibodies include anti-TREM2 and ATV:TREM2 with MV and BV Fabs. e, hTREM2-DAP12 HEK293 cells treated with a dose response of antibodies for 5 minutes, followed by pSyk detection (n = 3 independent experiments; mean ± s.e.m.). f, pSyk is blocked by co-treatment of ATV:TREM2 and anti-TfR. hTREM2-DAP12 HEK293 cells were dosed with 100 nM TREM2 antibodies and a titration anti-TfR. pSyk was detetced 5 minutes after treatment (n = 3 independent experiments; mean ± s.e.m.). g, TfR and TREM2 co-IP. hTREM2-DAP12 HEK293 cells were treated with 100 nM per antibody for 5 minutes, followed by IP on cell lysates with anti-TREM2. TfR was detected by western blot. h, Co-IP quantification of western blot data in g (n = 6 independent experiments; Wilcoxon test for ISO versus anti-TREM2, two-tailed paired t-test for ATV:ISO versus ATV:TREM2, mean ± s.e.m.). i, Schematic of BioID TREM2 receptor clustering assay strategy. j, Representative western blot detection of biotinylated TREM2 after streptavidin IP. TREM2-BioID expression was induced 24 hours before the assay with 2 ng ml⁻¹ of Dox. Cells were treated with 100 nM antibody and 2 µM biotin. k, Quantification of western blot from j (n = 4 independent experiments; ratio of two-tailed paired t-test, mean ± s.e.m.). l, Representative immunofluorescence images of hTREM2-DAP12 HEK293 cells stained for IgG (green), pSyk (yellow) and EEA1 (red). Cells were treated with 10 nM per antibody for 10 min. m, Quantification of spot intensity for IgG and pSyk immunofluorescence per cell (n = 3 independent experiments with 3,000–5,000 cells per condition; Tukey’s multiple comparisons test, mean ± s.e.m.). n, Quantification of percent of IgG or pSyk spots localized within EEA1⁺ endosomes (n = 3 independent experiments with 3,000–5,000 cells per condition; Tukey’s multiple comparisons test, mean ± s.e.m.). Source data

Fig. 3. Microglial proliferation induced by ATV:TREM2 requires mTOR signaling and PLCG2.

a, Representative western blot images for p-mTOR (S2448), pAKT (S473), pGSK3b (S9) and pRPS6 (S235/236) in WT iMG treated with 100 nM ATV:TREM2 or isotype control. b–e, Quantification of mTOR-S2448 (b), AKT-S473 (c), RPS6-S235/236 (d) and GSK3b-S9 (e) phosphorylation levels were normalized to actin. Relative expression was calculated by normalizing to vehicle control (PBS) for each experiment (n = 10 independent experiments (b–d) and n = 9 independent experiments (e); two-tailed paired t-test, mean ± s.e.m.). f, ATV:TREM2 induces proliferation in WT iMG but not TREM2 KO iMG. WT or TREM2 KO iMG were treated with 100 nM ATV:TREM2 or isotype control for 96 hours. Forty-eight hours after dose, 20 µM EdU was added to media. The proliferation index was calculated as percentage of EdU⁺ cells normalized to vehicle control (PBS) (n = 3 independent experiments; two-tailed multiple-paired t-test, mean ± s.e.m.). g, Quantification of WT and PLCG2 KO iMG proliferation. iMG were treated with 100 nM ATV:TREM2 or ATV:ISO (n = 4 independent experiments (WT) and n = 3 independent experiments (TREM2 KO); two-tailed multiple-paired t-test, mean ± s.e.m.). h, Representative images of WT iMG proliferation. Cells were treated with vehicle (DMSO) or mTOR inhibitor AZD8055 (AZD). EdU⁺ iMG are marked with open arrow. i, Quantification of WT iMG proliferation treated with mTOR inhibitor AZD8055. 20 nM AZD8055 was co-dosed with 100 nM ATV:TREM2 for 96 hours (n = 5 independent experiments (DMSO) and n = 4 independent experiments (AZD); two-tailed multiple-paired t-test, mean ± s.e.m.). j, Nuclei quantification of iMG co-treated with ATV:TREM2 and AZD. Relative nuclei count was normalized to vehicle control (PBS) for each experiment (n = 5 independent experiments (DMSO) and n = 4 independent experiments (AZD); two-tailed unpaired t-test, mean ± s.e.m.). k, RNA-seq of iMG treated for 4 days with PBS, 100 nM ATV:ISO or ATV:TREM2 or 10 ng ml⁻¹ LPS, 20 ng ml⁻¹ TGFβ or 20 ng ml⁻¹ IFNγ. Relative expression (z-scores) of the top-most upregulated or downregulated genes selected from pathways of interest. Pathway definitions were taken from the hallmark MSigDB collection; genes shown are a subset of those found in the leading edge of the gene set for each category. Source data

Fig. 4. ATV:TREM2 demonstrates improved brain exposure and pharmacodynamic microglial responses compared to anti-TREM2.

a–e, Four-day single-dose study. a, ELISA detection of antibodies in whole brain lysates for mice dosed with ATV:TREM2 (1, 3, 10 or 30 mg kg⁻¹) and anti-TREM2 (30 mg kg⁻¹) 1 day after dose (n = 5 mice per group). b, Microglia detected by IBA1 staining (purple), and proliferative cells were detected by EdU labeling (green) 4 days after dose. c, Quantification of IBA1/EdU staining (n = 5 mice per group; Dunnett’s multiple comparisons test, mean ± s.e.m.). d, CSF1R detected by ELISA in whole brain lysate (n = 5 mice per group; Kruskal–Wallis test for day 1 and Dunnett’s multiple comparisons test for day 4). e, CSF1R detected by ELISA in CSF (n = 5 mice per group for day 1 (30 mg kg⁻¹ ATV:ISO); day 1 (3 mg kg⁻¹ ATV:TREM2); day 1 (10 mg kg⁻¹ ATV:TREM2); day 1 (30 mg kg⁻¹ ATV:TREM2); day 4 (30 mg kg⁻¹ ATV:ISO). n = 4 mice per group for rest and Dunnett’s multiple comparisons test for day 1; Kruskal–Wallis test for day 4, mean ± s.e.m.). f,g, Ex vivo microglial phagocytosis for different substrates. Myelin debris (f) (n = 8 mice per group; Dunnett’s multiple comparisons test) and Aβ (g) (n = 8 mice per group; two-tailed unpaired t-test, mean ± s.e.m.). a–g, Circle represents male mice, and triangle represents female mice.

Fig. 5. ATV:TREM2 drives metabolism through mitochondrial fatty acid oxidation and PLCG2-dependent respiration in microglia.

a, Representative microscopy images from iMG treated with 10 µM oleic acid and then 100 nM ATV:ISO or ATV:TREM2 for 48 hours. BODIPY fluorescence (shown in green) was quantified (n = 5 independent experiments; two-tailed paired t-test, mean ± s.e.m.). b, Heat map of LC–MS analysis for TG species and acyl carnitines modulated by ATV:TREM2 in iMG treated with myelin. Plotted values are log₂-transformed raw counts and scaled by row. c,d, ATV:TREM2 reduces TGs and increases acyl carnitines in WT iMG treated with myelin (c) (n = 6 independent experiments (TG) and n = 3 independent experiments (acyl carnitines); two-tailed paired t-test, mean ± s.e.m.) but not PLCG2 KO iMG treated with myelin (d) (n = 3 independent experiments; two-tailed paired t-test, mean ± s.e.m.). e, Seahorse fatty acid oxidation OCR respiration measurements in TREM2 KO and PLCG2 KO iMG (n = 3 independent experiments; mean ± s.e.m.). f, ATV:TREM2 increases maximal respiration in WT iMG detected with Seahorse fatty acid oxidation OCR measurements (n = 7 independent experiments; two-tailed paired t-test, mean ± s.e.m.). The CPT-1 inhibitor etomoxir blocks the effect of ATV:TREM2 on respiration (n = 5 independent experiments; two-tailed paired t-test, mean ± s.e.m.). g,h, Seahorse analysis for glucose oxidation. ATV:TREM2 was treated at 100 nM for 3 days. The ATV:TREM2 effect is blocked by an MPC inhibitor, UK5099 (n = 4 independent experiments; two-tailed paired t-test, mean ± s.e.m.). i, ATV:TREM2 increases average TMRE intensity in iMG after 3 days with 100 nM antibody (n = 6 independent experiments; two-tailed paired t-test, mean ± s.e.m.). j, Representative images of super-resolution microscopy of TMRE staining in iMG. Mitochondria were segmented into networked and punctate morphologies. k, Morphometric analysis of the prevalence of networked mitochondria (n = 3 independent experiments, two-tailed paired t-test). l, Volcano plots of RNA-seq analysis of microglia isolated from mice dosed with 10 mg kg⁻¹ of ATV:ISO or ATV:TREM2 for 1 day, 4 days or 7 days. Red or blue indicate significantly upregulated or downregulated genes, respectively. The x axis represents log₂ fold change in expression compared to vehicle-treated mice, and the y axis represents –log₁₀ adjusted P value. m, Relative expression (z-scores) of the top-most upregulated or downregulated genes at day 1 after dose selected from oxidative phosphorylation and glycolysis pathways. OCR, oxygen consumption rate.

Fig. 6. ATV:TREM2 increases brain microglial activity and glucose metabolism in an AD model.

a, Coronal and axial slices show cold scaled group average images of TSPO-PET (SUV_H) projected upon a standard MRI T1-weighted atlas from 5×FAD;hTREM2 tg;TfR^mu/hu mice (top row) or WT;hTREM2 tg;TfR^mu/hu mice (bottom row) mice dosed with 10 mg kg⁻¹ of antibody. b,c, Quantification of TSPO-PET 1, 4, and 8 days after dose in 5×FAD;hTREM2 tg;TfR^mu/hu mice (b) and WT;hTREM2 tg;TfR^mu/hu mice (c). Scatter plot of individual TSPO-PET (SUV_H) values. Dotted lines represent linear associations between interval after antibody dosing and TSPO-PET quantification per group and with a 95% confidence interval (n = 6 mice per group; two-tailed unpaired t-test for each timepoint, except for day 8, which used the two-tailed unpaired t-test with Welch’s correction). d, Coronal and axial slices show cold scaled group average images of FDG (SUV) projected upon a standard MRI T1-weighted atlas from 5×FAD;hTREM2 tg;TfR^mu/hu mice (top row) or WT;hTREM2 tg;TfR^mu/hu mice (bottom row) after 10 mg kg⁻¹ of antibody. e,f, Quantification of cortical glucose uptake measured by FDG-PET 1, 4, and 8 days after dose of ATV:ISO or ATV:TREM2 for 5×FAD;hTREM2 tg;TfR^mu/hu mice (e) and WT;hTREM2 tg;TfR^mu/hu mice (f). Scatter plot of individual FDG (SUV) values. Dotted lines represent linear associations between interval after antibody dosing and FDG-PET quantification per group with a 95% confidence interval (n = 6 mice per group; two-tailed unpaired t-test for each timepoint). g,h, Regional correlation of biomarker alterations (5×FAD;hTREM2 tg;TfR^mu/hu versus WT;hTREM2 tg;TfR^mu/hu mice) between FBB-PET at 5 months and TSPO-PET (SUV_H) (g) and FDG-PET (SUV) (h) at the group level.

Extended Data Fig. 1. ATV:4D9 induces temporally dynamic microglial states distinct from amyloid pathology.

(a) UMAP projections of individually processed data sets for WT; TfR^mu/hu timecourse and TfR^mu/hu; App^SAA studies. Microglia are color coded according to their experimental group. (b) Combined UMAP of integrated data by study. Microglia are color coded by unbiased cluster assignment. (c) Stacked barplots showing the proportion of microglia per biological replicate by cluster. Plots are grouped by experimental group, and each bar represents a biological replicate within that group. Barplot color scheme is consistent with clusters in b. (d) Feature plots showing expression of selected individual genes. Microglia are color coded according to log normalized expression of each gene. (e) Antibody concentrations detected in whole brain lysate from either WT; TfR^mu/hu or TfR^mu/hu; App^SAA mice dosed with 10 mg kg⁻¹ ATV:ISO or ATV:4D9 (n = 8 mice/group, except for ATV:4D9 WT;TfR^mu/hu and 4D9 APP^SAA; TfR^mu/hu n = 4 mice/group).

Extended Data Fig. 2. ATV:4D9 induces temporally dynamic microglial morphology and marker expression.

(a) Representative morphometric images of microglia in the cortex 1 day post dose with ATV:ISO or ATV:4D9 (10 mg kg⁻¹) at day 1-, 7-, 14-, and 28-days post dose. (b) UMAP plot of all microglia over time. (c) Percentage of microglia in the responsive cluster as a proportion of all segmented microglia over time (n = 3 male mice per group, two-tailed unpaired t-test between ATV:ISO and ATV:4D9 at day 1, mean ± SEM). (d) Volcano plot showing the top 6 differentially high and low normalized features comparing the responsive cluster to the homeostatic cluster. (e) Heatmap of normalized features for all segmented microglia (1,143 total cells) over time, measured across 65 morphometric and immunohistochemical features, grouped by treatment (rows) with features hierarchically clustered (columns). (f) Representative images of cortical brain sections co-stained with Iba1 and CD74 at 1-, 7-, 14-, 28-days days post 10 mg kg⁻¹ dose of ATV:ISO or ATV:4D9. CD74⁺ microglia are noted with white arrows. (g) Mean intensity of CD74 staining within segmented IBA1⁺ microglia normalized to background CD74 intensity at each timepoint (n = 5 mice/group, two-tailed unpaired t-test between ATV:ISO and ATV:4D9 at day 1, mean ± SEM). (h) Representative images of cortical brain sections stained for IBA1 and AXL 1 day post dose of ATV:ISO or ATV:4D9. Double positive microglia are noted with white arrows. (i) Quantification of mean intensity of AXL staining within segmented IBA1⁺ microglia normalized to background (n = 5 mice/group, two-tailed unpaired t-test, mean ± SEM).

Extended Data Fig. 3. ATV:4D9 and ATV:TREM2 demonstrate similar mechanisms of action with high affinity stalk binding epitopes and cellular function.

(a) Antibody schematic comparing human specific ATV:TREM2 and mouse specific ATV:4D9 with high affinity TREM2 binding. (b) Epitope map of overlapping stalk binding regions for ATV:TREM2 and ATV:4D9 Fabs (space filled model of TREM2 ECD). The binding epitope of ATV:4D9 antibody is located 12-amino acids N-terminal of the ADAM cleavage site at His157. (c) FACS analysis of cell binding of ATV:TREM2 to hTREM2-DAP12 HEK293 or parental cells. Endogenous TfR expression on HEK293 cells drives weak binding observed for ATV:ISO and ATV:TREM2 (n = 3 independent experiment, mean ± SEM). (d) FACS detection of ATV:TREM2 (100 nM) binding to WT and TREM2 KO iMG with isotype control (ATV:ISO). (e) Soluble TREM2 measured in the supernatant of hTREM2-DAP12 HEK293 cells dosed with ATV:TREM2 for 24 h shows dose-dependent reduction of sTREM2 to levels similar to 1 uM GM6001 (n = 3 independent experiment, mean ± SEM). (f) ATV:TREM2 and lipid ligands induce pSyk signaling in iMG 24 h post antibody exposure (n = 3 independent experiments, Tukey’s multiple comparisons test, mean ± SEM). (g) Human monocytes cultured in limited M-CSF with plate coated ATV:TREM2 or ATV:ISO shows dose-dependent activity of ATV:TREM2 (EC50 0.95 +/− 0.45 nM). Representative data from one out of four human donors are shown.

Extended Data Fig. 4. ATV promotes TfR-TREM2 receptor complex formation and internalization and endosomal TREM2 signaling.

(a) Representative Western blot of co-IP of TREM2 with TfR. hTREM2-DAP12 HEK293 cells were treated with 100 nM ATV:TREM2, anti-TREM2, or isotype controls for 5 min at 37 °C. (b) Co-IP quantification of Western blot from (A) (n = 6 independent experiments; two-tailed paired t-test for ISO vs anti-TREM2; two tailed Wilcoxon test ATV:ISO vs ATV:TREM2, mean ± SEM). (c) Schematic illustration of cis- and trans-activation models that could mediate pSyk enhancement by ATV:TREM2. (d) Western blot validation of TfR knockdown in the TfR^RNAi cell line. (e) Cell based cis/trans assay indicates ATV:TREM2 enhances pSyk activity in cis. Relative pSYK signal is expressed as raw pSYK AlphaLisa value normalized to ATV:TREM2 treated control (n = 3 independent experiment, mean ± SEM). (f) Normalized pSyk signal measured by AlphaLisa assay. TfR^RNAi cells were treated with 10 nM anti-TREM2 or ATV:TREM2 pre-incubated with a dose titration of recombinant TfR protein for 5 min at 37 °C (n = 3 independent experiment, mean ± SEM). (g) Normalized pSyk signal detected by AlphaLisa. TfR^RNAi cells were treated with 10 nM anti-TREM2 or ATV:TREM2 pre-incubated with a dose titration of a secondary anti-human IgG Fc antibody for 5 min at 37 °C (n = 3 independent experiment, mean ± SEM) (h) Immunofluorescence microscopy of hTREM2-DAP12 HEK293 cells demonstrates reduction of surface TREM2 levels with ATV:TREM2 vs anti-TREM2, no changes in total TREM2 levels, consistent with re-distribution of the receptor from the plasma membrane to endosomes (n = 4 independent experiments except for anti-TREM2 MV (n = 3), Tukey’s multiple comparisons test, mean ± SEM). (i) hTREM2-DAP12 HEK293 cells dosed with antibody for 10 minutes shows that at similar amounts of bound antibody detected by anti-IgG (representing 5 nM of ATV:TREM2 and 10 nM of anti-TREM2, n = 4 independent experiments, Tukey’s multiple comparisons test, mean ± SEM(j) Images depicting masking algorithm used to identify whether TfR-Alexa-647 labeled recycling endosomes (rainbow spots in middle images) either contain (green spots in right-most images) or do not contain (red spots in right-most images) IgG spots (white spots in left-most image) upon dosing with anti-TREM2 (top row) or ATV:TREM2 (bottom row) for 10 minutes with 10 nM antibody. (k) Representative images for hTREM2-DAP HEK293 cells dosed with 10 nM antibody for 10 minutes including 20 ug mL⁻¹ TfR-Alexa-647, fixed, permeabilized, and stained with anti-IgG and anti-pSyk. IF shows ATV increased colocalization of antibody with pSyk in early endosomes. (l) Quantification of percent of IgG or pSyk spots localized within Tf-positive endosomes (n = 3 independent experiments, Tukey’s multiple comparisons test, mean ± SEM).

Extended Data Fig. 5. ATV:TREM2 does not promote ERK1/2 phosphorylation or a proinflammatory signature in microglia.

(a) Representative Western blot images of phosphorylation of 4EBP1 (T37/46) and ERK1/2 (T202/Y204) in WT iMGs treated for 96 h with 100 nM ATV:TREM2 or an isotype control. (b) Quantification of p4EBP1 (T37/46) and pERK1/2 (T202/Y204) normalized to actin. Relative expression was calculated by normalizing to PBS vehicle control for each experiment (n = 4 independent experiments, two-tailed paired t-test, mean ± SEM). (c) Representative Western blot images of total protein levels of mTOR and AKT in WT iMG treated for 96 h with 100 nM ATV:TREM2 or an isotype control. (d) Quantification of total mTOR and AKT protein normalized to actin. Relative expression was calculated by normalizing to PBS control for each experiment (n = 4 independent experiments, two-tailed paired t-test, mean ± SEM). e) Representative Western blot images for p-mTOR (S2448), pAKT (S473), pGSK3b (S9) pRPS6 (S235/236), p4EBP1 (T37/46) and pERK1/2 (T202/Y204) showing inhibition of mTOR pathway activation in WT iMG co-treated with 20 nM AZD8055 and 100 nM ATV:TREM2 after 96 h. (f) Quantification of phosphorylation targets shown in (E). Phosphorylation signals were normalized to actin. Relative expression was calculated by normalizing to PBS vehicle control for each experiment (n = 4 independent experiments, two-tailed paired t-test, mean ± SEM)). (g) Heatmap of human cytokine profiling in supernatant from WT iMG treated with 100 nM ATV:TREM2 for 96 h. Media collected from iMG treated with 10 ng mL⁻¹ LPS for 24 h was used for comparison.

Extended Data Fig. 6. ATV:TREM2 increases phagocytosis and CSF1R in human TREM2 tg;TfR^mu/hu mice.

(a) Antibody levels detected in brain lysates by ELISA shows increased brain exposure of ATV:TREM2 (30 mg kg⁻¹) compared anti-TREM2 (30 mg kg⁻¹) and comparable brain exposure for ATV:TREM2 at 10 mg kg^-1 and anti-TREM2 at 30 mg kg⁻¹ at day 4 post-dose. (n = 5 mice/group except for ATV:TREM2 3 mg kg⁻¹ (n = 4)). (b) ATV:TREM2 increases CSF1R in the brain compared to brain exposure matched anti-TREM2 day 2 post dose (n = [8, 9, 8, 9, 10]mice/group, Kruskal-Wallis test, mean ± SEM). (c) CSF1R analysis in CSF same as in B. (n = [8, 9, 8, 9, 10]mice/group, Tukey’s multiple comparisons test, mean ± SEM) (b-c) Circles represents male mice and triangle represents female mice. (d) Detection of antibody concentrations show matched brain exposure of three ATV:TREM2 molecules with different ATV affinities 1-day post-dose (n = 5 mice/group). (e-f) High and mid-affinity ATV:TREM2 molecules induce comparable increase of CSF1R in the brain (e) n = 5 mice/group except for Vehicle (Veh) and day 4 10 mg kg⁻¹ 8,000 nM (n = 4, Kruskal-Wallis test, compared to vehicle group) and CSF (f) (n = 5 mice/group, except for Veh and day 4 5 mg kg⁻¹ 110 nM n = 4), Dunnett’s multiple comparisons test for day 1, Kruskal-Wallis test for day 4) day 1 and 4 post dose whereas low-affinity ATV:TREM2 induces a weak elevation of CSF1R in CSF at day 4 (n = 4-5 mice/group, mean ± SEM). (g) Schematic of experimental approach to evaluate in vivo dosed antibody impact to microglial phagocytosis ex vivo. A single dose of ATV:TREM2, anti-TREM2, or ATV:ISO was administrated to human TREM2 tg; TfR^mu/hu KI mice and brain microglia were isolated 2 days post dose for ex vivo myelin phagocytosis analysis. The same method was used to assess amyloid phagocytosis. (h) FACS gating strategy to quantify pHrodo-myelin positive microglia. After treatment with pHrodo-green myelin and staining, cell suspensions containing microglia were analyzed using BD FACS Aria III. Single cells were separated from debris by FSC and SSC characteristics. Live microglia were identified as a population of CD11b⁺ and propidium iodide^negative cells. pHrodo-myelin uptake was then quantified in 20,000 microglia recorded from each sample. (i) Antibody concentrations were detected in brain lysates by ELISA which shows comparable levels of ATV:TREM2 and anti-TREM2 day 2 post dose (n = [9, 10, 10, 9, 10] mice/group). (j) ATV:TREM2 induces transcription of genes associated with phagocytosis. Axl, Itgax, Lgals3 mRNA levels were detected in isolated brain microglia after peripheral administration of 10 mg kg⁻¹ ATV:TREM2 1-, 4-, and 7-days post dose. Graphs represent bulk mRNA measured by RNA-seq. Data shown as log₂ counts per million (with a pseudocount of 1 added) in each biological replicate (n = 8 mice/group, each group compared to ATV:ISO group, Kruskal-Wallis test for Axl, Dunnett’s multiple comparisons for Itgax and Lgals3, mean ± SEM).

Extended Data Fig. 7. Single photon emission computed tomography imaging demonstrates increased ATV:TREM2 biodistribution and catabolism in brain.

(a) Representative autoradiography images of sagittal brain sections from WT; hTREM2 tg; TfR^mu/hu mice 96 h after administration of ATV:TREM2, anti-TREM2, and ATV:ISO radiolabeled with [¹¹¹In]DOTA or [¹²⁵I]SIB. (b,c) Longitudinal SPECT/CT imaging quantification of whole-brain uptake of single dose (100 µCi, 200 µL, 1.5 mg kg⁻¹, IV) ATV:TREM2, anti-TREM2, and ATV:ISO, radiolabeled with [¹²⁵I]SIB (b) or [¹¹¹In]DOTA (c) in WT; hTREM2 tg; TfR^mu/hu mice. Whole brain %ID/g was corrected for contribution from cerebral blood volume. Data are represented as mean +/− SEM (n = 3 mice/group). (d) Percent of catabolized ATV:TREM2 in several brain regions 96 h after single dose exceeds that of ATV:ISO control. [¹¹¹In]DOTA and [¹²⁵I]SIB signal was quantified by ex vivo gamma counting in resected brain regions, and percent of catabolized antibody was estimated by subtracting %ID/g for [¹²⁵I]SIB signal from that of [¹¹¹In]DOTA signal (n = 4 mice/group, one-way ANOVA with Dunnett’s multiple comparison test for each region, except for cerebellum which applied Brown-Forsythe and Welch ANOVA tests, mean ± SEM).

Extended Data Fig. 8. ATV:TREM2 increases microglial metabolism in a TREM2 and PLCG2-dependent manner.

(a) Additional species of triglycerides (TG) and short chain carnitines modulated in iMG with ATV:TREM2 treatment (n = 3-5 independent experiment, two-tailed paired t-test, mean ± SEM). (b) ATV:TREM2 does not modulate TG in PLCG2 KO iMG (n = 3 independent experiment, two-tailed paired t-test, mean ± SEM). (c) Maximal respiration measured by Seahorse fatty acid oxidation kit is reduced in both TREM2 KO and PLCG2 KO iMG (n = 3-4 independent experiment, two-tailed paired t-test, mean ± SEM). (d) ATV:TREM2 increased spare capacity measured by Seahorse fatty acid oxidation kit (n = 8 independent experiment, two-tailed paired t-test, mean ± SEM). (e) ATV:TREM2 does not modulate maximal respiration or spare capacity in TREM2 KO iMG (n = 4 independent experiment, two-tailed paired t-test, mean ± SEM). (f) ATV:TREM2 does not modulate maximal respiration and spare capacity PLCG2 KO iMG (n = 4 independent experiment, two-tailed paired t-test, mean ± SEM). (g) RNAseq analysis of brain microglia isolated from hTREM2 tg; TfR^mu/hu mice dosed with 10 mg kg⁻¹ ATV:TREM2. GSEA for top pathways based on a p-value cutoff of 0.05 for up- or downregulated gene sets 1 day post ATV:TREM2 dose.