TMEM59 regulates microglial Aβ phagocytosis in a TREM2‐dependent manner and acts as a potential therapeutic target for Aβ amyloidosis

Abstract

INTRODUCTION

Microglia‐mediated clearance of amyloid beta (Aβ) is crucial for mitigating Alzheimer's disease (AD) progression, yet the molecular regulators of microglial phagocytosis remain incompletely understood.

METHODS

We assessed AD‐related phenotypes in 5xFAD mice with microglia‐specific deletion of transmembrane protein 59 (TMEM59). Microglial transcriptomic profiling was performed using single‐cell RNA sequencing (scRNA‐seq). A chaperone‐mediated autophagy‐targeting chimera (CMATAC) peptide was developed to degrade TMEM59, and its therapeutic efficacy was evaluated.

RESULTS

Microglia‐specific TMEM59 ablation attenuated cognitive deficits, Aβ plaque burden, and synapse loss in 5xFAD mice. TMEM59 deficiency reprogrammed disease‐associated microglia toward a highly phagocytic state. Mechanistically, TMEM59 deficiency enhanced microglial Aβ phagocytosis in a triggering receptor expressed on myeloid cells 2 (TREM2)‐dependent manner, as it stabilized TREM2 protein, whereas loss of TREM2 abolished these protective effects. Treatment with the TMEM59‐degrading CMATAC peptide alleviated behavioral deficits and enhanced microglial Aβ phagocytosis in 5xFAD mice.

DISCUSSION

TMEM59 is a key regulator of microglial phagocytosis in AD and a novel target for amyloidosis intervention.

Highlights

• TMEM59 is a negative regulator of microglial Aβ clearance in AD.

• Microglial TMEM59 deficiency improves cognitive function and reduces Aβ pathology in 5xFAD mice.

• TMEM59 deficiency enhances microglial Aβ phagocytosis by stabilizing TREM2 through slowing its degradation.

• A CMATAC peptide that degrades TMEM59 attenuates AD‐associated phenotypes in 5xFAD mice.

1. BACKGROUND

Alzheimer's disease (AD) is the most common dementia, characterized by progressive cognitive decline accompanied by amyloid beta (Aβ) plaque accumulation, neurofibrillary tangles, and synapse loss. ¹ So far, several therapeutic drugs have been approved for AD intervention, including anti‐Aβ antibodies. ² However, these drugs only offer limited benefits and raise significant safety concerns, underscoring the urgent need for new therapeutic strategies.

As the resident immune cells of the central nervous system, microglia are responsible for inflammatory signaling, lipid metabolism, and phagocytic clearance of pathological substrates. ³ , ⁴ , ⁵ Upon Aβ plaque formation during AD progression, microglia are activated and assemble a compact barrier around Aβ plaques, limiting plaque‐associated toxicity and initiating phagocytic clearance. ⁴ , ⁶ Recent single‐cell RNA sequencing (scRNA‐seq) studies identified a novel microglial subtype surrounding Aβ plaques in AD, known as disease‐associated microglia (DAM). ⁷ , ⁸ , ⁹ , ¹⁰ The microglial transition to a DAM phenotype is a two‐step process: initial activation into an intermediate state (stage 1 DAM) in response to pathological stimuli, followed by triggering receptor expressed on myeloid cells 2 (TREM2)‐dependent induction into stage 2 DAM – a state involving lipid metabolism and phagocytosis. ⁷ , ¹¹ , ¹² However, in addition to TREM2, whether and how other factors regulate DAM phenotype transition remain incompletely known.

Transmembrane protein 59 (TMEM59) is an autophagy‐related protein that interacts with ATG16L1 to promote LC3 activation. ¹³ TMEM59 expression is significantly elevated in the brains of AD patients and AD mouse models. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ We previously found that TMEM59 haploinsufficiency ameliorated amyloid burden, synaptic dysfunction, and cognitive deficits in both Aβ‐ and tau‐driven AD models. ¹⁶ , ¹⁷ These findings establish TMEM59 as a promising therapeutic target across multiple AD pathologies. However, the mechanism underlying TMEM59 deficiency‐mediated reduction of Aβ burden remains unclear. ¹⁶ Since TMEM59 interacts with TREM2 and regulates microglial inflammatory responses, and downregulation of TMEM59 alleviates microglial dysfunction caused by TREM2 deficiency, ¹⁸ there is a possibility that TMEM59 participates in AD through regulating microglial function.

Targeted protein degradation (TPD) has become a powerful approach to modulating proteins previously considered “undruggable,” such as transcription factors, scaffolding proteins, and other intracellular non‐enzymatic proteins. ¹⁹ , ²⁰ , ²¹ Multiple TPD technologies have been proposed, including proteolysis‐targeting chimera (PROTAC), ²² lysosome‐targeting chimera (LYTAC), ²³ autophagy‐targeting chimera (AUTAC), ²⁴ autophagosome‐tethering compound (ATTEC), ²⁵ and chaperone‐mediated autophagy (CMA)‐targeting chimera (CMATAC). ²⁶ CMATAC represents a promising strategy that harnesses the CMA pathway to induce lysosomal degradation of specific substrates, ¹⁹ , ²⁶ , ²⁷ , ²⁸ providing new opportunities to modulate proteins such as TMEM59, a transmembrane protein mainly localized in the Golgi and endosomes, for which conventional inhibition strategies remain limited.

In this study, we investigated the effects of TMEM59 on microglial function under AD pathology. We demonstrated that loss of TMEM59 in microglia ameliorated cognitive impairment and amyloid pathology in the 5xFAD mouse model. TMEM59 deficiency enhanced microglial clustering and phagocytosis of Aβ by stabilizing TREM2. Furthermore, we demonstrated that CMATAC peptide‐directed degradation of TMEM59 attenuated behavioral deficits and enhanced microglial phagocytosis of Aβ in 5xFAD mice, indicating that targeting TMEM59 degradation using TPD technologies may offer a novel and effective therapeutic avenue for Aβ amyloidosis.

RESEARCH IN CONTEXT

1. Systematic review: Our previous study found that TMEM59 haploinsufficiency attenuated Aβ plaque deposition and cognitive deficits in 5xFAD mice, an amyloidosis model of AD. However, the underlying mechanism remains unknown.

2. Interpretation: Our findings identify TMEM59 as a novel negative regulator of microglial Aβ clearance. TMEM59 deficiency reprograms microglia toward a highly phagocytic state via stabilizing TREM2, thereby reducing Aβ plaque burden and attenuating cognitive deficits in 5xFAD mice. Promoting TMEM59 degradation using a chaperone‐mediated autophagy‐targeting chimera (CMATAC) peptide recapitulates these protective effects, demonstrating that TMEM59 is a novel therapeutic target for AD intervention.

3. Future directions: Future research includes (1) determining whether TMEM59 modulates other microglial activities beyond Aβ phagocytosis, (2) evaluating the pharmacological kinetics and long‐term safety and efficacy of TMEM59‐targeting CMATAC peptides, (3) investigating whether CMATAC‐mediated TMEM59 degradation also ameliorates tau pathology in AD, and (4) exploring the translational relevance of targeting TMEM59 in human AD patients.

2. METHODS

2.1. Animals

Microglia‐specific Tmem59 conditional knockout Tmem59^flox/flox;Lyz2^Cre/+ (cKO) mice were generated as described previously. ²⁹ 5xFAD (stock 008730) and Trem2 ^−/− (stock 027197) mice were purchased from the Jackson Laboratory. cKO mice were crossed with 5xFAD or Trem2 ^−/− mice to generate 5xFAD;cKO or cKO;Trem2 ^−/− mice (Figure 1A). All mice were maintained under a 12‐h light/dark cycle. All experimental procedures were approved by the Animal Ethics Committee of Xiamen University (approval numbers: XMULAC20170209, XMULAC20190074, and XMULAC20210096).

FIGURE 1.

Microglial TMEM59 deficiency ameliorates cognitive impairment and Aβ burden in 5xFAD mice. (A) The scheme for the generation of 5xFAD with microglial Tmem59 conditional knockout (5xFAD;cKO) mice and the experimental procedure. Both male and female mice were subjected to behavioral tests at 7 months of age, followed by Aβ plaque analysis. All other analyses were performed using samples from male mice. (B–I) 7‐month‐old Ctrl, cKO, 5xFAD, and 5xFAD;cKO mice were subjected to various behavioral tests. Spontaneous alternation (%) (B) and total distance traveled (C) were assessed in the Y‐maze test. Freezing behavior was measured in the contextual (D) and cued (E) fear conditioning tests. The escape latency during the training phase (F) and the number of platform crossings (G), the first entrance to platform (H), and swimming speed (I) during the probe trial were studied in the Morris water maze test. n = 22 mice for Ctrl, n = 21 mice for cKO, n = 22 mice for 5xFAD, and n = 20 mice for 5xFAD;cKO. (J–O) Representative images of ThioS staining (J) and quantification of ThioS⁺ plaque areas in the hippocampus (K) and cortex (L) of 7‐month‐old 5xFAD and 5xFAD;cKO mice. Insets (1) and (2) show magnified views of the hippocampal and cortical regions, respectively. The images were contrast‐reversed for better illustration. n = 10 mice per group. Scale bar: 500 µm, 100 µm for inset. (M–O) Representative images of 6E10 immunostaining (M) and quantification of 6E10⁺ plaque areas in hippocampus (N) and cortex (O) of 7‐month‐old 5xFAD and 5xFAD;cKO mice. Insets (1) and (2) show magnified views of the hippocampal and cortical regions, respectively. Images were contrast‐reversed for better illustration. n = 10 mice per group. Scale bar: 500 µm, 100 µm for inset. Open circle: male, closed circle: female. Data presented as mean ± SEM. One‐way ANOVA with Tukey's multiple‐comparisons test for (B–E) and (G–I). Two‐way ANOVA with Sidak's multiple‐comparisons test for (F). Unpaired Student's t test for (K), (L), (N), and (O).

2.2. Primary microglial cultures

Primary microglial cells were prepared from the cerebral cortex and hippocampus of neonatal mice at postnatal days 1 to 3. Tissues were dissociated and plated onto poly‐L‐lysine (Sigma‐Aldrich, 26124‐78‐7)‐coated flasks. Cells were maintained in DMEM (Gibco, 11965092) containing 10% heat‐inactivated fetal bovine serum (ExCell, FSP500) and 25 ng/mL GM‐CSF (R&D Systems, 415‐ML). After culturing for 10 days in vitro, microglial cells were collected by gentle shaking for 20 min for subsequent experiments.

2.3. Intravenous injection

The control and CMATAC peptides were synthesized by Sangon Biotech. Peptides were delivered by intravenous injection via the tail vein. To assess blood–brain barrier (BBB) permeability, wild‐type (WT) mice received a single injection of 200 µg fluorescien isothiocyanate (FITC)‐labeled CMATAC or control peptide. To assess the efficacy and safety, 2‐month‐old WT mice were intravenously injected with unlabeled CMATAC or control peptides at a dose of 5 mg/kg body weight once every other day for a total of 20 days. For behavioral evaluation, 6‐month‐old 5xFAD mice and age‐matched WT littermates were subjected to the same dosing protocol. The peptide administration was maintained consistently throughout the behavioral test period. After behavioral tests, mice were sacrificed, and brain tissues were collected for subsequent analyses.

2.4. Behavioral tests

Mice were allowed to acclimate to the testing room for 30 min prior to each test. All behavioral assays were conducted by investigators blinded to genotype. Data were recorded and analyzed using TopScanLite software (Clever Sys. Inc.) for the Y maze and Morris water maze tests and FreezeScan software (Clever Sys. Inc.) for the fear conditioning test.

2.4.1. Open field

Each mouse was placed in a square open field (40 × 40 × 40 cm) and allowed to move freely for 10 min. Locomotor activity was assessed by measuring the total distance traveled, and anxiety‐like behavior was evaluated by measuring the time spent in the center zone.

2.4.2. Y‐maze

Spontaneous alternation behavior was assessed using a Y‐shaped maze. ³⁰ Mice were placed at the center of the maze and allowed to freely explore all three arms for 5 min. An alternation was defined as consecutive entries into three different arms. The spontaneous alternation percentage was calculated as the ratio of actual alternations to the maximum possible alternations (total arm entries: 1).

2.4.3. Fear conditioning test

The fear conditioning test was performed as previously described with minor modifications. ³¹ , ³² During the training phase, mice were placed individually into the conditioning chamber with a metal grid floor and allowed to explore for 2 min to establish baseline freezing levels. An auditory white noise (80 dB, 30 s) was used as the conditioned stimulus (CS), and a foot shock (0.8 mA, 2 s) was delivered during the final 2 s of the CS as the unconditioned stimulus (US). Three CS‐US pairings were administered, separated by 30‐s intertrial intervals. Mice were kept in the chamber for an additional 1 min. After 24 h, mice were re‐exposed to the original context for 5 min without any stimulus, and freezing behavior was recorded for the contextual test. The percentage of freezing behavior was calculated by subtracting the baseline freezing percentage established during the training phase from the freezing percentage during the 5‐min re‐exposure. For the cued test, mice were placed into a novel testing chamber. After a 3‐min habituation period, the CS was presented for 3 min. The percentage of freezing behavior was calculated by subtracting the freezing percentage during the initial 3‐min habituation period from that during the 3‐min CS presentation.

2.4.4. Morris water maze

Spatial ,  ,  ,   learning and memory were evaluated using the Morris water maze. ³³ The apparatus consisted of a circular pool (120 cm in diameter) filled with opaque water, with a hidden platform 1 cm below the water surface. Mice underwent two training trials per day for six consecutive days, with varying starting positions across trials. Escape latency, defined as the time required reaching and mounting the platform, was recorded. If a mouse failed to locate the platform within 60 s, it was gently guided to it and left there for 10 s. On day 7, a probe trial was conducted for 60 s after platform removal. The time spent for the first entry into the platform region and the crossing numbers over the platform region were recorded.

FIGURE 2.

TMEM59 deficiency reprograms the transcriptional profile of DAM in 5xFAD mice. (A) UMAP visualization of unsupervised clustering of cells from 7‐month‐old Ctrl, 5xFAD, and 5xFAD;cKO mice. n = 3 mice per group. (B) Dot plot showing representative marker genes defining homeostatic microglia, stage 1 DAM, and stage 2 DAM. (C) tSEN plots illustrating distribution and relative proportions of homeostatic microglia, stage 1 DAM, and stage 2 DAM across the indicated genotypes. (D) Volcano plot displaying the DEGs between 5xFAD and 5xFAD;cKO in stage 1 DAM. (E) KEGG pathway enrichment analysis of upregulated DEGs between 5xFAD;cKO and 5xFAD in stage 1 DAM. (F) Volcano plot displaying DEGs between 5xFAD and 5xFAD;cKO in stage 2 DAM. (G) KEGG pathway enrichment analysis of upregulated DEGs between 5xFAD;cKO and 5xFAD in stage 2 DAM.

FIGURE 3.

TMEM59 deficiency enhances microglial clustering and phagocytosis of Aβ both in vivo and in vitro. (A and B) Representative images of Iba1 (in green) and 6E10 (in magenta) co‐immunostaining in the hippocampus of 7‐month‐old 5xFAD and 5xFAD;cKO mice (A) and quantification of Iba1⁺ microglia surrounding Aβ plaques (B). n = 6 mice per group. Scale bar: 20 µm. (C and D) Representative images of Iba1 (in green), 6E10 (in gray), and CLEC7A (in red) co‐immunostaining in the hippocampus of 7‐month‐old 5xFAD and 5xFAD;cKO mice (C) and quantification of CLEC7A⁺ microglial areas surrounding plaques (D). n = 6 mice per group. Scale bar: 20 µm. (E and F) Representative images and 3D reconstructions of Iba1 (in green), CD68 (in cyan), and 6E10 (in magenta) co‐immunostaining in the hippocampus of 7‐month‐old 5xFAD and 5xFAD;cKO mice (E) and quantification of Aβ signals within CD68⁺ microglial lysosomes (F). n = 6 mice per group. Scale bar: 20 µm. (G and H) Representative FACS plots showing methoxy‐X04⁺ microglia isolated from 7‐month‐old 5xFAD and 5xFAD;cKO mice 3 h after methoxy‐X04 injection (G) and quantification of methoxy‐X04⁺ microglial proportions (H). n = 5 mice per group. (I and J) Representative FACS plots of primary microglia from Ctrl and cKO mice following incubation with FAM‐labeled Aβ_1–42 oligomers for 1 h (I) and quantification of primary microglia containing Aβ_1–42 oligomers (J). n = 5 independent experiments. Data are presented as mean ± SEM. Unpaired Student's t‐test for (B), (D), (F), (H), and (J).

FIGURE 4.

TMEM59 regulates microglial phagocytosis by modulating TREM2 protein stability. (A and B) Representative images of Iba1 (in green), 6E10 (in gray), and TREM2 (in red) co‐immunostaining in the hippocampus of 7‐month‐old 5xFAD and 5xFAD;cKO mice (A) and quantification of TREM2⁺ microglial areas surrounding plaques (B). n = 6 mice per group. Scale bar: 20 µm. (C) qRT‐PCR analysis of Trem2 and Tmem59 mRNA levels in primary microglia from Ctrl and cKO mice. n = 6 independent experiments. (D and E) Representative immunoblots (D) and quantification (E) of TREM2 and TMEM59 protein levels in HEK293T cells transfected with TREM2‐myc and TMEM59 siRNA (si59) or control siRNA (NC). n = 6 independent experiments. (F and G) Representative immunoblots (F) and quantification (G) of TREM2 degradation in HEK293T cells transfected with TREM2‐myc and si59 or NC following CHX (100 µM) treatment for the indicated time period. n = 6 independent experiments. (H and I) Representative FACS plots of primary microglia from Ctrl, cKO, and cKO;Trem2‐KO (cKO;T2^−/−) mice following incubation with FAM‐labeled Aβ_1–42 oligomers for 1 h (H) and quantification of primary microglia containing Aβ_1–42 oligomers (I). n = 5 independent experiments. (J and K) Representative images of Iba1 staining (in red) in Ctrl, cKO, and cKO;T2^−/− mice 16 h after hippocampal injection of FAM‐labeled Aβ_1–42 oligomers (in green (J) and quantification of Iba1⁺ microglial areas surrounding Aβ_1–42 oligomers (K). n = 5 mice per group. Scale bar: 50 µm. Data are presented as mean ± SEM. Unpaired Student's t‐test for (B). Two‐way ANOVA with Sidak's multiple comparisons test for (C, E, and G). One‐way ANOVA with Tukey's multiple‐comparisons test for (I) and (K).

FIGURE 5.

Construction of CMATAC for TMEM59 degradation. (A) Scheme of seven WD‐repeat domains of ATG16L1. The three truncated WD7 domains, D1 (aa 575–590), D2 (aa 582–598), and D3 (aa 591–607), are also shown. (B–D) Representative co‐IP (B) and quantification (C and D) of interactions between exogenously expressed TMEM59‐CTF‐myc and GFP‐WD7‐D1, GFP‐WD7‐D2, or GFP‐WD7‐D3 in HEK293T cells. (E) Scheme of TMEM59‐targeting CMATAC and control peptides. PBD, protein‐binding domain; CTM, CMA‐targeting motif; CPP, cell‐penetrating peptide. (F) Dose–response curve showing direct binding between TMEM59 and the CMATAC peptide assessed by MST. n = 3 replicates. (G and H) Representative two‐photon images (G) and quantification (H) of BBB permeability of control and CMATAC peptides with tail‐vein injection in 6‐month‐old mice. n = 4 distinct brain regions of interest (ROIs) per peptide. Data are presented as mean ± SEM. One‐way ANOVA with Tukey's multiple‐comparisons test for (C) and (D).

2.5. Stereotaxic surgery

To assess microglial migration, stereotactic injection of Aβ_1–42 oligomers was performed. Oligomeric Aβ_1–42 was prepared as previously described. ³⁴ , ³⁵ Fluorescein amidite (FAM)‐labeled Aβ_1–42 (AnaSpec, AS‐23526‐01) or Fluor 555‐labeled Aβ_1–42 (AnaSpec, AS‐60480‐01) was dissolved in DMSO (1 mM) and subsequently diluted with phosphate buffered saline (PBS). The solution was sonicated for 20 min and incubated at 22°C for 16 h, followed by an additional 24 h at 4°C to obtain Aβ_1–42 oligomers. Mice were deeply anesthetized and head‐fixed in a stereotaxic apparatus (RWD Life Science). Aβ_1–42 oligomers (1.5 µg in 3 µL) and vehicle control were delivered into the right (anterior posterior, AP: −2.0 mm from bregma, medial lateral, ML: −1.5 mm from midline, dorsal ventral, DV: −2.0 mm from dura) and left (AP: −2.0 mm from bregma, ML: +1.5 mm from midline, DV: −2.0 mm from dura) hippocampus, respectively. Brains were harvested 16 h after surgery.

For sparse neuronal labeling, AAV2/9‐hSyn‐FLP‐WPRE‐pA (5.53 × 10^{1 2} v.g./mL; BrainVTA, PT‐0341) was diluted at 1:20,000 with sterile PBS and mixed at a 1:1 volume ratio with AAV2/9‐nEF1a‐FDIO‐EYFP‐EYFP‐WPRE‐pA (5.35 × 10^{1 2} v.g./mL; BrainVTA, PT‐1210). The viral mixture was bilaterally injected into hippocampus (AP: −2.0 mm from bregma, ML: ± 1.5 mm from midline, DV: −1.5 mm from dura) and cortex (AP: −2.0 mm from bregma, ML: ± 0.75 mm from midline, DV: −0.8 mm from dura). A total volume of 200 nL per site was delivered at a rate of 1 nL/s.

2.6. scRNA‐seq and data analysis

The cerebral cortex and hippocampus were isolated from 7‐month‐old WT, 5xFAD, and 5xFAD;cKO mice (n = 3 male mice per group) and mixed for scRNA‐seq at BGI Genomics, using the DNBelab C4 platform. Briefly, fresh tissues were enzymatically dissociated at 37°C until digestion was completed. The cell suspension was filtered through a 40‐µm strainer, followed by centrifugation and red blood cell lysis. Cells were washed with PBS containing 0.04% bovine serum albumin (BSA). Cell viability and concentration were assessed by trypan blue prior to library preparation. Single‐cell suspensions that passed quality control were combined with oil and barcoded beads and processed to generate water‐in‐oil droplets using the DNBelab C‐TaiM system. Cell lysis and subsequent mRNA capture occurred within the droplets. Reverse transcription was performed to generate cDNA, followed by demulsification, cDNA enrichment, and amplification. Sequencing libraries were constructed through cDNA fragmentation, end repair, A‐tailing, adaptor ligation, and polymerase chain reaction (PCR) amplification according to the standard protocol. Final cDNA and oligo libraries were circularized to generate single‐stranded DNA and sequenced using combinatorial probe‐anchor synthesis technology.

The scRNA‐seq data were processed using the DNBelab_C4scRNA pipeline (version 1.0.1) and aligned to the mouse reference genome (GRCm39) to generate a count matrix. This matrix was directly imported into Seurat (version 3.2.0) for downstream analysis. Cells were retained if they expressed 300 to 10,000 genes, contained fewer than 10,000 UMIs, and showed <10% mitochondrial reads. Data were log‐normalized, highly variable genes were identified, and scaled expression values were used for dimensionality reduction by principal component analysis (PCA). Cell clustering was performed using the Louvain algorithm based on the top principal components (PCs), and low‐dimensional visualization was generated using uniform manifold approximation projection (UMAP) or t‐distributed stochastic neighbor embedding (tSNE). To identify cluster‐defining marker genes, the FindAllMarkers function (logFC > 0.25, minPct > 0.1, and Padj ≤ 0.05) in Seurat was employed with the Wilcoxon rank‐sum test. Initial cell type annotation was performed based on canonical marker gene expression. Functional enrichment analyses were conducted in R using Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases, with p value ≤ 0.05 considered significant. The dataset was subsequently refined through an iterative process of subsetting, re‐dimension reduction, and re‐clustering. Specifically, for subcluster analysis within annotated populations, the top 2000 variable features were recalculated from the subset, followed by scaling, PCA (using the top 20 PCs for Louvain clustering), and tSNE visualization (using the top 50 PCs).

2.7. Immunofluorescence staining

Mice were deeply anesthetized and perfused with ice‐cold PBS. Brains were collected, fixed overnight in 4% paraformaldehyde at 4°C, and subsequently dehydrated in 30% sucrose for 24 h. Coronal brain sections (30 µm thick) were prepared using a freezing microtome (FS800, RWD Life Science). Brain sections were blocked in PBS containing 3% BSA and 0.2% Triton X‐100 for 1 h at room temperature, followed by incubation with primary antibodies overnight at 4°C. After washing, sections were incubated with appropriate fluorescence‐conjugated secondary antibodies (Invitrogen; A11005, A11008, A21247, and A21448; 1:500) for 1 h at room temperature. Primary antibodies used in this study are listed below: anti‐Iba1 (Cell Signaling Technology, 17198S, 1:300), anti‐CD68 (Biolegend, 137001, 1:300), anti‐6E10 (Biolegend, 803014, 1:400), anti‐CLEC7A (InvivoGen, mabg‐mdect, 1:200), anti‐TREM2 (R&D Systems, AF1729, 1:200), anti‐PSD‐95 (Millipore, MAB1596, 1:200), and anti‐Synapsin I (Proteintech, 20258‐1‐AP, 1:300). Confocal images were captured with an FV4000 or FV1000MPE‐B confocal microscope (Olympus).

2.8. Thioflavin S staining

Thioflavin S staining was carried out by incubating brain sections in 1 mM Thioflavin S (Sigma, T1892‐25G) prepared in 50% ethanol for 8 min. After staining, brain sections were washed with 80% ethanol and PBS and then observed under a confocal microscope.

2.9. Hematoxylin and eosin (H&E) staining

The heart, liver, spleen, lung, and kidney of treated mice were collected, fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned at 10 µm thickness using a rotary microtome (RM2245, Leica). Tissue sections were stained with a H&E stain kit (Solarbio, G1121) according to the manufacturer's instructions. Images were acquired with a VS200 microscope (Olympus).

2.10. Confocal image analysis

Confocal images were analyzed using Imaris software (version 10.0.0). Synaptic puncta were identified using the Spots module with a 0.3‐µm diameter. Colocalization was defined as spots located within a distance of less than 0.6 µm. For Aβ engulfment by microglia, the 6E10 signals localized within CD68‐positive lysosomal compartments were quantified. The Iba1⁺ microglial cells, CD68⁺ lysosomes, and 6E10⁺ plaques were reconstructed using the Surface module. Next, 6E10 signals masked within CD68 structures were reconstructed as a separate channel, and the ratio of engulfed Aβ volume within CD68 to total Aβ plaque volume was quantified.

2.11. Cranial window surgery and in vivo multi‐photon imaging

Cranial window surgery and multi‐photon imaging were carried out as described previously. ³³ Briefly, mice were anesthetized and head‐fixed on a stereotaxic frame. After removal and cleaning of the scalp, a circular craniotomy (4 mm in diameter) was carefully created above the cerebral cortex using a drill. The exposed cortical surface was covered with a glass coverslip, and a custom‐designed head plate was fixed to the skull to stabilize the animal during imaging. Multi‐photon fluorescence imaging was performed using an FVMPE‐RS multiphoton microscope (Olympus). Mice were injected intravenously with Texas Red‐conjugated dextran (70 kDa; Thermo Fisher Scientific, D1830) to visualize the cerebral vasculature, followed by administration of FITC‐labeled control or CMATAC peptide to evaluate their BBB permeability. Fluorescence excitation was achieved using a 920‐nm‐wavelength laser, and emitted signals were collected by GaAsP photomultiplier tubes with 495‐ to 540‐nm (for FITC detection) or 575‐ to 645‐nm (for Texas Red detection) filters. Image sequences were processed using ImageJ software. Regions of interest (ROIs) were selected in extravascular areas to measure fluorescence intensity over time. BBB permeability was quantified as the change in extravascular fluorescence intensity, ΔF = (F_n −F ₀)/F ₀, where F_n is the fluorescence intensity at a given time point, and F ₀ is the baseline intensity from the initial frame.

2.12. Engulfment assays by flow cytometry

For Aβ phagocytosis by microglia in vivo, 5xFAD and 5xFAD;cKO mice were intraperitoneally injected with methoxy‐X04 (10 mg/kg body weight, Tocris, 4920) at a 1:9 ratio of DMSO and PBS (pH 12.0). Cerebral cortical and hippocampal tissues were dissected 3 h after injection, and microglial cells were isolated as previously described. ²⁹ , ³⁶ The tissues were homogenized in Hanks' Balanced Salt Solution using a Dounce homogenizer, filtered, and subjected to centrifugation using 30% Percoll solution (Cytiva, 17089102). Following centrifugation at 800 × g for 30 min at 4°C without braking, the pellets were collected and resuspended in PBS. Cells were incubated with anti‐CD16/32 antibody (Biolegend, 101301, 1:200) for Fc receptor blocking, followed by staining with anti‐CD11b‐APC (eBioscience, 17‐0112‐82, 1:200) and anti‐CD45‐FITC antibodies (eBioscience, 11‐0451‐85, 1:200) in the dark for 30 min. After washing, cells were resuspended in 1% FBS/PBS and analyzed using a Cytoflex LX flow cytometer (Beckman Coulter).

For Aβ engulfment in vitro, primary microglia were incubated with FAM‐labeled Aβ oligomers for 1 h. After removing non‐engulfed Aβ by washing with PBS, microglia with Aβ uptake were quantified by flow cytometry.

2.13. Transfection

HEK293T cells were transfected with plasmids using Liposomal 2000 Transfection Reagent (Yeasen, 40802ES03). For siRNA transfection, siRNAs synthesized by GenePharma were introduced into the cells with the Lipofectamine 2000 reagent (Thermo Fisher Scientific, 11668019). All steps followed the manufacturers’ recommended guidelines. siRNA sequences used in this study were as follows: negative control (NC), 5’‐ UUCUCCGAACGUGUCACGU ‐3’; Tmem59 siRNA‐1, 5’‐ GCACAGAGCUUCAUAACCU ‐3’; Tmem59 siRNA‐2, 5’‐ GGACCAAGCUGG AAUGUGA ‐3’; TMEM59 siRNA, 5’‐ ACCUCUUCAUGGACUUUUUAU ‐3’.

2.14. Co‐immunoprecipitation (Co‐IP)

Transfected HEK293T cells were lysed in ice‐cold IP lysis buffer (Beyotime Biotechnology, P0037), supplemented with protease inhibitor cocktail (MedChemExpress, HY‐K0010). Cell lysates were subjected to centrifugation. Equal amounts of supernatants were incubated with Protein A/G magnetic beads (MedChemExpress, HY‐K0202) and the indicated primary antibodies at 4°C overnight. The precipitated proteins were analyzed by Western blotting.

2.15. Western blotting

Cells or brain tissues were lysed in RIPA buffer containing protease inhibitor cocktail (MedChemExpress, HY‐K0010) and phosphatase inhibitor cocktail (TargetMol, C0002, C0003). Lysates were centrifuged at 12,000 rpm for 15 min at 4°C to remove the debris. Total protein concentrations were measured using a bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, 23225). Samples were analyzed by SDS‐PAGE and then transferred to PVDF membranes for immunoblotting. The primary antibodies used in this study were as follows: anti‐TREM2 (CST, 91068S, 1:1000), anti‐TMEM59 (ABclonal, WG‐03224D, 1:1000), anti‐GFP (Proteintech, 50430‐2‐AP, 1:2000), anti‐myc (CST, 2276S, 1:1000), anti‐myc (CST, 2278S, 1:1000), anti‐HA (Abmart, M20003L, 1:2000), anti‐β‐actin (CST, 8457S, 1:2000), and anti‐α‐tubulin (Millipore, MABT205, 1:10000). The following horseradish peroxidase‐conjugated secondary antibodies were used: Goat anti‐Mouse IgG (LABLEAD, S0100, 1:5000) and Goat anti‐Rabbit IgG (LABLEAD, S0101, 1:5000).

2.16. Aβ ELISA

N2a695 cells were transfected with TMEM59‐myc and control vector or with TMEM59 siRNAs and control siRNA. Cells were then cultured in fresh serum‐free DMEM for 1 h prior to harvest. Both cells and conditioned media were collected. Cell lysates were prepared by lysis followed by centrifugation, and protein concentrations were determined using a BCA assay. Conditioned media were centrifuged at 3000 rpm for 5 min at 4°C to remove debris. Levels of Aβ40 and Aβ42 in the media were measured using human Aβ40 (Thermo Fisher Scientific, KHB3481) and Aβ42 (Thermo Fisher Scientific, KHB3441) ELISA kits according to the manufacturer's instructions. The measured Aβ levels were normalized to the protein concentration of the corresponding cell lysates.

2.17. RNA isolation and quantitative real‐time PCR (qRT‐PCR)

Total RNAs were extracted from primary microglial cells or the brain of treated mice using TRIzol reagent (Invitrogen, 15596018CN) and reverse‐transcribed into cDNA using HiScript II Q RT SuperMix (Vazyme, R222‐01). qRT‐PCR was performed using Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Q712‐03). Trem2 and Tmem59 expression levels were normalized to those of β‐actin for comparison. Gene‐specific primer sequences were as follows: Trem2‐Forward: TGCTGGCAAAGGAAAGGTG, Trem2‐Reverse: GTTGAGGGCTTGGGACAGG; Tmem59‐Forward: AACTCACAAGCAC ACAGGAACT, Tmem59‐Reverse: CAACACCATCACCGAGAGGA; β‐actin‐Forward: AGCCATGTACGTAGCCATCCA, β‐actin‐Reverse: TCTCCGGAGTCCA TCACAATG.

2.18. Microscale thermophoresis (MST)

MST experiments were performed as previously described. ³⁷ Briefly, His‐tagged recombinant TMEM59 protein (IPODIX Biotechnology, PA2000‐1829) was labeled with RED‐tris‐NTA fluorescent dye (NanoTemper Technologies, MO‐L018) following the manufacturer's instructions. The CMATAC peptide was dissolved in MST buffer (saline supplemented with 0.05% Tween‐20) to a stock concentration of 1 mM and then serially diluted to generate a concentration gradient. An equal volume of labeled TMEM59 protein was added to each peptide dilution, resulting in a final TMEM59 concentration of 400 nM. The samples were incubated and subsequently analyzed by MST using standard protocols provided by NanoTemper Technologies.

2.19. Statistics

All data are presented as mean ± SEM. Statistical analyses were performed using GraphPad Prism 10.4.1. Unpaired two‐tailed Student's t‐test was applied for comparisons between two groups, and one‐way or two‐way ANOVA was applied for multiple‐group comparisons as appropriate. Statistical significance was defined as p < 0.05.

3. RESULTS

3.1. Microglial TMEM59 deletion ameliorates cognitive impairment in 5xFAD mice

We previously demonstrated that TMEM59 haploinsufficiency alleviated Aβ plaque burden in 5xFAD mice without affecting amyloid precursor protein processing. ¹⁶ Here, we further found that neither knockdown (Figure S1A–C) nor overexpression (Figure S1D,E) of TMEM59 altered Aβ production in N2a695 cells. To explore whether TMEM59 modulated Aβ through regulating microglial responses and clearance, we injected Aβ_1–42 oligomers (oAβ) into the hippocampus of control mice (Ctrl) and Tmem59^flox/flox;Lyz2^Cre/+ (cKO) mice, ²⁹ in which Tmem59 was specifically deleted in microglia. We observed an increased number of microglia clustering around the injected Aβ in both Ctrl and cKO mice. Notably, microglia clustering was more dramatic in cKO mice than in Ctrl mice (Figure S1F,G). Next, we generated 5xFAD mice with microglia‐specific deletion of Tmem59 (5xFAD;cKO) by crossing cKO mice with 5xFAD mice to investigate the effects of microglial TMEM59 deficiency on AD‐related behavioral deficits (Figure 1A). We conducted behavioral tests in both male and female mice at 7 months of age to evaluate cognitive function. When data from both sexes were pooled, we observed that 5xFAD mice exhibited reduced spontaneous alternation in the Y‐maze task without a concomitant change in the total distance traveled, suggesting that they had impaired working memory, whereas 5xFAD;cKO mice displayed markedly improved performance relative to the 5xFAD group (Figure 1B,C). In the fear conditioning test, 5xFAD mice displayed reduced freezing responses in both contextual and cued tests (Figure 1D,E), suggesting deficits in fear memory. In contrast, microglial TMEM59 deficiency largely increased the percentage of freezing in both test conditions in 5xFAD mice (Figure 1D,E). In the Morris water maze test, 5xFAD mice spent more time finding the hidden platform during the training phase and had decreased platform crossings and increased latency to reach the platform zone during the probe test (Figure 1F–H). While 5xFAD;cKO mice spent less time finding the hidden platform and exhibited more platform crossings and reduced time to reach the platform zone compared to 5xFAD mice (Figure 1F–H). There were no significant differences in swimming speed among the tested groups (Figure 1I). These results suggest that impaired spatial learning and memory in 5xFAD mice can be rescued by microglial TMEM59 deletion.

We also analyzed the mice separately by sex. Both male (Figure S2A–H) and female (Figure S2I–P) 5xFAD mice had marked cognitive deficits in behavioral tests, except that female 5xFAD mice showed comparable spontaneous alternation in the Y‐maze test (Figure S2I). Importantly, we found that male 5xFAD;cKO mice exhibited significant amelioration of cognitive deficits compared to 5xFAD mice, including working memory in the Y‐maze test (Figure S2A), fear memory in the contextual fear conditioning test (Figure S2C), and spatial learning and memory in the Morris water maze test (Figure S2E–G). Similarly, female 5xFAD;cKO mice showed improved performance in working memory and spatial learning and memory compared to controls (Figure S2I and M–O), whereas contextual fear memory was not significantly different (Figure S2K). Together, these findings demonstrate that microglial TMEM59 deficiency alleviates most cognitive impairments in 5xFAD mice independent of sex. Therefore, we used both male and female mice for the following studies.

3.2. Microglial TMEM59 deletion attenuates Aβ pathology and synapse loss in 5xFAD mice

We next studied whether microglial TMEM59 deletion affected AD‐related pathologies. Thioflavin S (ThioS) staining revealed robust Aβ plaque accumulation in the hippocampus and cortex of 5xFAD mice. Notably, microglial TMEM59 deletion resulted in approximately a 50% reduction in Aβ plaque accumulation in both the hippocampal and cortical regions (Figure 1J–L). Consistent with this finding, immunostaining with the 6E10 antibody against Aβ showed a substantial reduction in total Aβ deposition in 5xFAD;cKO mice compared to 5xFAD controls (Figure 1M–O). These results indicate that microglial TMEM59 deletion markedly reduces Aβ burden in 5xFAD mice.

We sparsely labeled neurons in the hippocampus and cortex using AAV‐nEF1α‐FDIO‐EYFP and AAV‐hSyn‐FLP. We found that 5xFAD mice had significantly decreased dendritic spine density in both the hippocampus and cortex compared to control mice, and such a reduction was attenuated in 5xFAD;cKO mice (Figure S3A–D). Colocalization between the presynaptic marker Synapsin I (SynI) and the postsynaptic marker postsynaptic density protein 95 (PSD‐95) can reflect the synapse formation. Our immunofluorescence staining revealed a dramatic reduction of SynI/PSD‐95 colocalization in both the hippocampus and cortex of 5xFAD mice, suggesting a loss of synapses (Figure S3E–H). In contrast, 5xFAD;cKO mice showed significantly increased numbers of colocalized synaptic puncta compared to 5xFAD mice (Figure S3E–H). These results suggest that TMEM59 deficiency in microglia not only mitigates amyloid pathology but also alleviates synaptic degeneration, highlighting a protective role of microglial TMEM59 loss in Aβ‐driven pathology.

3.3. TMEM59 deficiency reprograms DAM functional states in AD pathology

To elucidate the molecular mechanisms by which microglial TMEM59 deficiency influences AD pathology, we performed scRNA‐seq on hippocampal and cortical tissues from 7‐month‐old Ctrl, 5xFAD, and 5xFAD;cKO mice. UMAP clustering identified 12 distinct cell populations across the three genotypes (Figure 2A and Figure S4A). Within the microglial population, homeostatic microglia and DAM were clearly separated, with stage 1 and stage 2 DAM exhibiting distinct gene expression signatures (Figure 2B). t‐SNE visualization revealed that microglia were predominantly in homeostatic status in Ctrl mice, whereas a large amount of microglia transited into DAM phenotypes in 5xFAD and 5xFAD;cKO mice (Figure 2C). There were no apparent proportion differences of stage 1 and stage 2 DAM between the 5xFAD and 5xFAD;cKO groups (Figure 2C). We then performed KEGG pathway enrichment analysis for the differentially expressed genes (DEGs) and found that upregulated DEGs in stage 1 DAM (such as Tnf, Ifnb1, and Cxcl2) from 5xFAD;cKO mice compared to 5xFAD mice were strongly associated with proinflammatory signaling cascades, including IL‐17, NF‐κB, and tumor necrosis factor pathways (Figure 2D,E). In contrast, upregulated DEGs in stage 2 DAM (such as Lamp1, Ctsd, and Actg1) from the 5xFAD;cKO group were largely enriched for lysosomal function and phagosome‐related pathways (Figure 2F,G). On the other hand, the downregulated DEGs in both stage 1 and stage 2 DAM (such as Hspa1a, H2‐Ab1, and Cd74) from 5xFAD;cKO mice compared to those from 5xFAD mice were enriched in pathways such as antigen processing and presentation, toxoplasmosis, and protein processing in endoplasmic reticulum (Figure 2D,F and Figure S4B,C). Collectively, these findings highlight a critical regulatory role for TMEM59 in shaping the dynamics of microglial state in AD.

3.4. TMEM59 deficiency promotes microglial phagocytic capacity of Aβ

Since scRNA‐seq results revealed that TMEM59 deficiency predominantly modulated the phagocytic function of stage 2 DAM, we studied whether TMEM59 is involved in Aβ phagocytosis and clearance. Immunostaining analysis revealed that microglia numbers surrounding Aβ plaques significantly increased in the hippocampus (Figure 3A,B) and cortex (Figure S5A,B) of 5xFAD;cKO mice compared to those of 5xFAD mice. We also measured the expression of CLEC7A, a stage 2 DAM marker gene. ⁷ Immunofluorescence revealed higher expression of CLEC7A in microglia surrounding Aβ plaques in 5xFAD;cKO mice than that in 5xFAD mice (Figure 3C,D and S5C,D), suggesting that the increased microglia surrounding Aβ plaques are stage 2 DAM. To investigate whether increased DAM enhances microglial phagocytosis of Aβ in vivo, we studied Aβ engulfment by microglia using immunostaining and 3D reconstruction. We observed that 5xFAD;cKO microglia engulfed more Aβ than 5xFAD microglia in both the hippocampus (Figure 3E,F) and cortex (Figure S5E,F). To further confirm the effects of TMEM59 deficiency on microglial phagocytosis, we intraperitoneally injected methoxy‐X04 to stain Aβ in 7‐month‐old 5xFAD and 5xFAD;cKO mice for 3 h and then isolated microglia for flow cytometry analysis. The results showed that microglia from 5xFAD;cKO mice contained more methoxy‐X04 than those from 5xFAD mice (Figure 3G,H), suggesting that TMEM59‐deficient microglia have enhanced Aβ engulfment in vivo. When primary microglia were isolated from Ctrl and cKO mice and incubated with FAM‐labeled oAβ, TMEM59‐deficient microglia also exhibited elevated Aβ phagocytosis in vitro (Figure 3I,J). Together, these results demonstrate that TMEM59 deficiency promotes microglial clustering to plaques and phagocytosis of Aβ.

3.5. Microglial TMEM59 regulates Aβ phagocytosis in a TREM2‐dependent manner

TREM2  signaling is required for the activation of stage 2 DAM to acquire full phagocytic capacity. ¹¹ , ³⁸ Emerging evidence indicates that upregulation of TREM2 enhances microglial phagocytosis of Aβ plaques. ³⁹ , ⁴⁰ , ⁴¹ Our previous study found that TMEM59 interacted with TREM2 and knockdown of TMEM59 alleviated the deficits in microglial survival, proliferation, migration, and phagocytosis resulting from TREM2 deficiency. ¹⁸ Therefore, we studied whether TREM2 played a role in microglial TMEM59‐mediated Aβ phagocytosis. We first examined and indeed found that TREM2 expression in microglia surrounding Aβ plaques significantly increased in TMEM59‐deficient microglia compared to controls (Figure 4A,B). However, the mRNA levels of Trem2 were comparable in primary microglia from Ctrl and cKO mice (Figure 4C), suggesting that TMEM59 deficiency post‐transcriptionally regulates TREM2. To directly test whether TMEM59 influenced TREM2 protein stability, we knocked down TMEM59 using siRNA in HEK293T cells that overexpress TREM2. Immunoblotting confirmed that TMEM59 knockdown significantly increased TREM2 protein levels (Figure 4D,E). Furthermore, we found that TMEM59 knockdown significantly slowed the degradation rate of TREM2 when using cycloheximide to inhibit protein synthesis (Figure 4F,G), indicating that TMEM59 negatively regulates TREM2 protein stability.

FIGURE 6.

Peptide‐directed degradation of TMEM59 attenuates cognitive impairment and Aβ plaque burden in 5xFAD mice. (A) Scheme of experimental procedure. 6‐month‐old WT and 5xFAD mice were intravenously injected with CMATAC or control peptide at 5 mg/kg body weight every other day for 20 days. Mice were then subjected to various behavioral tests for an additional 10 days, during which the peptides were kept given. After behavioral test, mice were sacrificed for biochemical and histological analyses. (B–D) Representative immunoblots (B) and quantification of TMEM59 protein levels in the hippocampus (C) and cortex (D) of WT and 5xFAD mice treated with control or CMATAC peptide. (E–L) Treated mice were subjected to various behavioral tests. Spontaneous alternation (%) (E) and total distance traveled (F) were assessed in the Y‐maze test. Freezing behavior was measured in the contextual (G) and cued (H) fear conditioning tests. The escape latency during the training phase (I) and the number of platform crossings (J), the first entrance to platform (K), and swimming speed (L) during the probe trial were studied in the Morris water maze test. n = 12 mice for WT‐Control, n = 11 mice for 5xFAD‐Control, and n = 11 mice for 5xFAD‐CMATAC. (M–O) Representative images of ThioS staining (M) and quantification of ThioS⁺ plaque areas in hippocampus (N) and cortex (O) of 5xFAD‐Control and 5xFAD‐CMATAC mice. Insets (1) and (2) show magnified views of hippocampal and cortical regions, respectively. Images were contrast‐reversed for better illustration. n = 5 mice per group. Scale bar: 500 µm, and 100 µm for inset. Data are presented as mean ± SEM. One‐way ANOVA with Tukey's multiple‐comparisons test for (C–H) and (J–L). Two‐way ANOVA with Sidak's multiple‐comparisons test for (I). Unpaired Student's t‐test for (N) and (O).

We next assessed whether the enhanced microglial phagocytosis caused by TMEM59 deficiency depended on TREM2. For this purpose, we knocked out Trem2 in mice with microglia‐specific deletion of TMEM59 (cKO;T2^−/−) by crossing TMEM59 cKO mice with Trem2 knockout mice. Primary microglia isolated from each genotype were incubated with FAM‐labeled oAβ and then subjected to flow cytometry analysis. The results showed that TMEM59‐deficient microglia exhibited markedly increased Aβ uptake compared with Ctrl microglia. Notably, loss of TREM2 abolished the enhanced phagocytic activity of TMEM59‐deficient microglia (Figure 4H,I). Consistently, in the hippocampal region of mice with FAM‐labeled oAβ injection, we found increased microglia surrounding Aβ in cKO mice, whereas TREM2 deletion reversed such an increase in cKO mice (Figure 4J,K). Collectively, these findings indicate that TMEM59 deficiency promotes stage 2 DAM clustering and enhances Aβ clearance through stabilizing TREM2 protein.

3.6. CMATAC peptide construction for TMEM59 degradation

To develop a therapeutic strategy targeting TMEM59 for its downregulation in AD, we designed a peptide capable of selectively degrading TMEM59 using the CMATAC technique. Because the C‐terminal fragment (CTF) of TMEM59 was reported to interact with the tryptophan‐aspartic acid (WD) repeat domain of ATG16L1, ¹³ which has seven WD repeats (Figure 5A), we first validated this interaction using co‐IP (Figure S6A). To identify the precise binding region, we generated a series of ATG16L1 WD1‐WD7 deletion mutants. However, deletion of any single WD domain of ATG16L1 had no significant effect on its interaction with TMEM59‐CTF (Figure S6B). We then co‐expressed the individual ATG16L1 WD domain with TMEM59‐CTF and found that WD7 exhibited the strongest binding affinity for TMEM59 (Figure S6C,D). To define the dominant minimal region within WD7 required for TMEM59 binding, we truncated the WD7 domain of ATG16L1 into three smaller fragments (D1 to D3) (Figure 5A) and found that the D3 fragment displayed the most robust interaction with TMEM59‐CTF (Figure 5B–D). We further confirmed that the D3 fragment also interacted with full‐length murine TMEM59 (Figure S6E).

We then synthesized a targeting peptide (termed CMATAC) for the degradation of TMEM59. The CMATAC peptide consisted of the D3 fragment as a protein‐binding domain (PBD; D3), a flexible linker GSGS, ⁴² a CMA‐targeting motif (CTM), ²⁶ a cell‐penetrating peptide (CPP; TAT), ²⁶ , ⁴² and a FITC label for visualization (Figure 5E). A control peptide lacking the CTM was synthesized in parallel (Figure 5E). Microscale thermophoresis (MST) assay confirmed that CMATAC directly binds to recombinant TMEM59 protein in vitro with a K_d of 2.64 ± 0.33 µM (Figure 5F). When CMATAC and control peptides were injected into mice through the tail vein, we found that both peptides diffused into the brain parenchyma within 60 s (Figure 5G,H), suggesting that they can rapidly cross the BBB. Moreover, immunofluorescence staining showed that the two peptides efficiently entered neurons and microglia, with little uptake in astrocytes (Figure S6F).

3.7. Peptide‐directed degradation of TMEM59 attenuates behavioral deficits and enhances microglial phagocytosis of Aβ in 5xFAD mice

To evaluate the efficacy and safety of the CMATAC peptide both in vitro and in vivo, we added the CMATAC and control peptides to the culture medium of HEK293T cells. Treatment with the CMATAC peptide resulted in reduced TMEM59 protein levels in these cells (Figure S7A,B). Subsequently, we intravenously injected the peptides into WT mice at a dose of 5 mg/kg body weight every other day for 20 days (Figure S7C). In the brain tissues of CMATAC‐treated mice, TMEM59 protein levels were also significantly decreased (Figure S7D,E), whereas Tmem59 mRNA levels remained unchanged (Figure S7F). H&E staining revealed no obvious histopathological changes in the heart, liver, spleen, lung, and kidney tissues of CMATAC‐treated mice compared to those of controls (Figure S7G). These results indicate that the CMATAC peptide degrades TMEM59 protein without inducing obvious toxicity in peripheral organs. We next assessed the therapeutic potential of the CMATAC peptide in 5xFAD mice. The CMATAC and control peptides were administered intravenously to WT or 5xFAD mice at 6 months of age. Mice were then subjected to behavioral tests for 10 days and sacrificed for biochemical and histological analyses (Figure 6A). Immunoblotting of brain lysates revealed that CMATAC treatment significantly reduced the elevated TMEM59 protein levels in both the hippocampus and cortex of 5xFAD mice (Figure 6B‐D ). Behavioral tests revealed that the CMATAC treatment attenuated working memory deficits in 5xFAD mice during the Y‐maze test (Figure 6E,F) and contextual fear deficits during the fear conditioning test (Figure 6G,H), but had no effects on anxiety‐like behavior or locomotor activity in the open field test (Figure S8A,B). Moreover, during the Morris water maze test, we found that the CMATAC treatment reduced the escape latency during the training phase (Figure 6I) and increased platform crossings, while also shortening the latency to the first platform visit without affecting swimming speed during the probe trial in 5xFAD mice (Figure 6J–L).

Consistent with the behavioral improvements, the CMATAC treatment markedly reduced Aβ plaque burden in the hippocampus and cortex of 5xFAD mice, as shown by staining with ThioS (Figure 6 M–O) and immunofluorescence staining with the 6E10 antibody (Figure S8C–E). The CMATAC treatment also mitigated synaptic loss, characterized by an increase in puncta of colocalized SynI and PSD‐95 in 5xFAD mice (Figure S8F,G). To determine whether the therapeutic benefits of the CMATAC treatment were mediated through alterations in microglial function, we examined plaque‐associated microglial responses. CMATAC‐treated 5xFAD mice displayed a pronounced increase in microglial clustering around Aβ plaques (Figure 7A,B), accompanied by elevated numbers of CLEC7A‐positive DAM (Figure 7C,D). Moreover, the CMATAC treatment increased TREM2 expression in plaque‐associated microglia (Figure 7E,F). Importantly, microglial phagocytosis of Aβ plaques was significantly enhanced in 5xFAD mice with the CMATAC treatment (Figure 7G,H).

FIGURE 7.

Peptide‐directed degradation of TMEM59 enhances microglial phagocytosis of Aβ in 5xFAD mice. (A and B) Representative images of Iba1 (in green) and 6E10 (in magenta) co‐immunostaining in the hippocampus of 5xFAD‐Control and 5xFAD‐CMATAC mice (A) and quantification of Iba1⁺ microglia surrounding Aβ plaques (B). n = 5 mice per group. Scale bar: 20 µm. (C and D) Representative images of Iba1 (in green), 6E10 (in blue), and CLEC7A (in red) co‐immunostaining in the hippocampus of 5xFAD‐Control and 5xFAD‐CMATAC mice (C), and quantification of CLEC7A⁺ microglial areas surrounding plaques (B). n = 5 mice per group. Scale bar: 10 µm. (E and F) Representative images of Iba1 (in green), 6E10 (in blue), and TREM2 (in red) co‐immunostaining in the hippocampus of 5xFAD‐Control and 5xFAD‐CMATAC mice (E), and quantification of TREM2⁺ microglial areas surrounding plaques (F). n = 5 mice per group. Scale bar: 10 µm. (G and H) Representative images and 3D reconstructions of Iba1 (in green), CD68 (in cyan), and 6E10 (in magenta) co‐immunostaining in the hippocampus of 5xFAD‐Control and 5xFAD‐CMATAC mice (G) and quantification of Aβ signals within CD68⁺ microglial lysosomes (H). n = 5 mice per group. Scale bar: 20 µm. Data are presented as mean ± SEM. Unpaired Student's t‐test for (B), (D), (F), and (H).

Together, these results demonstrate that CMATAC‐based degradation of TMEM59 effectively decreases TMEM59 protein levels in the brain, reduces amyloid deposition, promotes DAM accumulation and microglial phagocytosis, and ultimately improves cognitive functions in 5xFAD mice.

4. DISCUSSION

TMEM59 levels are abnormally elevated in patients with AD. ¹⁴ , ¹⁵ , ¹⁷ Our recent study demonstrated that TMEM59 deficiency attenuated behavioral deficits and tau‐related pathologies in PS19 mice by regulating CMA. ¹⁷ Moreover, we and others have shown that either haploinsufficiency or antibody‐mediated blockade of TMEM59 reduces plaque burden and ameliorates cognitive function in various Aβ‐driven AD mouse models. ¹⁶ , ⁴³ Despite these observations, the mechanism by which TMEM59 deficiency reduces Aβ burden remains elusive. In this study, we demonstrated that TMEM59 deficiency ameliorated cognitive impairment and Aβ pathology in 5xFAD mice through enhancing microglial phagocytosis of Aβ plaques.

DAM represent a specialized subtype of activated microglia in AD, characterized by the upregulation of genes involved in lysosomal, phagocytic, and lipid metabolism pathways, including some AD‐risk genes (e.g., Trem2, Apoe, Spp1, Clec7a, Axl, Cst7, and Lpl), and the downregulation of homeostatic markers (e.g., P2ry12, P2ry13, Cx3cr1, Tmem119, and Hexb). ⁷ , ¹¹ , ⁴⁴ Among the pathways governing DAM induction, TREM2 signaling has emerged as a central regulator for the transition from stage 1 DAM to stage 2 DAM. ⁷ , ¹¹ , ³⁸ In 5xFAD mice, stage 1 DAM are a minor and transient population, while the majority of microglia rapidly progress to execute the stage 2 DAM transcriptional program. However, TREM2 deficiency resulted in a complete loss of stage 2 DAM and a substantial accumulation of cells arrested at stage 1. ⁷ Emerging evidence suggests that TREM2‐mediated DAM transition plays protective roles in neurodegeneration: TREM2 expression positively correlates with plaque‐associated microglial clustering and Aβ phagocytic efficiency. ⁴¹ , ⁴⁵ Loss of TREM2 impairs microglial engulfment of plaques and accelerates AD pathology, ⁴⁶ whereas enhancing TREM2 activity promotes microglial clearance of Aβ plaques and ameliorates disease progression. ³⁹ , ⁴⁰ , ⁴¹ The protein tyrosine kinase SYK transmits the intracellular signaling of TREM2. 5xFAD mice with microglial SYK deletion showed reduced DAM clustering and exacerbated Aβ deposition and memory deficits. ⁴⁷ , ⁴⁸ TMEM119 is a homeostatic marker of microglia, and its overexpression or agonists promoted DAM accumulation, ameliorating cognitive impairments and Aβ pathology in AD mice. ⁴⁹ Here, our results revealed that TMEM59 deficiency reshaped stage 2 DAM toward phagocytic states for Aβ clearance, indicating that TMEM59 acts as another essential regulator of microglial responses to AD‐related Aβ pathology. Moreover, we found that TMEM59 deficiency stabilized TREM2 protein levels by inhibiting its degradation, whereas TREM2 deletion abolished the increased Aβ phagocytosis by TMEM59‐deficient microglia, suggesting that TMEM59 regulates microglial phagocytosis in a TREM2‐dependent manner.

TPD technologies have provided novel approaches to developing therapeutics for neurodegenerative diseases. PROTAC harnesses the ubiquitin‐proteasome system to achieve selective protein degradation and is one of the most common TPD strategies used. ⁵⁰ Previous studies successfully developed PROTAC peptides capable of degrading tau protein in tau transgenic (3xTg) mice ⁵¹ and neurons derived from patients with frontotemporal dementia. ⁵² Similarly, PROTAC peptides targeting α‐synuclein markedly reduced α‐synuclein expression and rescued mitochondrial dysfunction in neuronal cells overexpressing α‐synuclein. ⁵³ Targeting lysosomal pathways offer an alternative strategy for TPD. For example, the AUTAC strategy has been shown to efficiently degrade tau protein in AD mouse models. ⁵⁴ CMATAC peptides have also been demonstrated to have therapeutic efficacy in multiple disease models, such as reducing mutant huntingtin in Huntington's disease mice, ²⁷ decreasing DAPK1 levels in ischemic stroke mice, ²⁶ and degrading Aβ oligomers in induced pluripotent stem cell (iPSC)‐derived neurons from AD patients. ⁵⁵ Herein, we report the application of a novel CMATAC peptide for Aβ amyloidosis therapy: The CMATAC peptide targeting TMEM59 degradation effectively reduced TMEM59 levels in vivo, exhibited good BBB permeability, and led to significant improvements in cognitive function, microglial phagocytosis of Aβ, and synapse loss in 5xFAD mice.

In summary, our study identifies TMEM59 as an essential regulator that modulates microglial phagocytosis through a TREM2‐dependent mechanism and contributes to AD pathology. Genetic ablation or peptide‐directed degradation of TMEM59 enhances microglial clearance of Aβ and ameliorates cognitive deficits in AD mice. These findings highlight TMEM59 as a potential therapeutic target and suggest that CMATAC‐based degradation could represent an effective strategy for restoring microglial function in Aβ amyloidosis.

Our study has several limitations. First, the study was conducted using the 5xFAD mouse model, which primarily recapitulates Aβ‐driven pathology and does not fully model tau pathology or the neuronal loss characteristic of late‐stage AD. Accordingly, future investigations in tauopathy mouse models are warranted to determine whether the regulatory effects of TMEM59 deficiency on microglial states and phagocytosis extend beyond Aβ‐driven pathology. Second, while CMATAC peptide‐directed TMEM59 degradation demonstrated therapeutic efficacy, additional studies are needed to assess kinetics, long‐term safety, and potential effects on the peripheral immune system of this peptide. Third, the Lyz2‐Cre driver is active in peripheral myeloid cells, but our experiments only focused on microglia, leaving the potential contribution of peripheral cell populations to the observed phenotypes unresolved. Future studies using alternative, more microglia‐specific drivers, such as the Tmem119‐CreERT2 system, would be better suited to delineate the cell‐autonomous contribution of microglia.

CONFLICT OF INTEREST STATEMENT

Y.‐w. Zhang, J. Meng, Y. Wang, J. Wang, and Z. Zhao are co‐applicants of a Chinese patent related to this study (patent no. ZL202411765927.4). The other authors have declared that no conflict of interest exists. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Consent was not necessary for the authors.

Supporting information

Supporting Information

Supporting Information

DATA AVAILABILITY STATEMENT

scRNA‐seq data (GEO: GSE324258) have been deposited in the Gene Expression Omnibus (GEO) repositoryhttps://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc . All the other data are present in the article and supplementary information.