F0 Tyrobp^+/– mice were obtained from Cyagen Biosciences (Guangzhou, China). The mouse strain was constructed by microinjecting into fertilized eggs a transcription activator-like effector nuclease (TALEN) that removes 10 bases (GTACAGGCCC) from exon 2 of the TYROBP gene. F0 mice were bred with C57/BL6 mice to produce the F1 generation, and the mutant F1 generation was inbred to generate the F2 generation. Gene knockout was confirmed using Sanger sequencing and western blotting.

All experiments were carried out using male mice. Age-matched WT littermates were used as controls. APP^KM670/671NL/PSEN1^Δexon9 (APP/PS1) mice and C57/BL6J WT mice were purchased from Huafukang Bioscience Co., Ltd. (Beijing, China). Mice were housed in groups of four with ad libitum access to standard food pellets and water on a 12/12h light/dark cycle. Experiments were approved by the Ethics Committee for Animal Experiments at The Affiliated Hospital of Jining Medical University.

Each group in this test contained six mice aged 2, 6, and 9 months. Testing was conducted using a standard 5-day regimen with a circular pool filled with opaque water by handlers who were blinded to grouping. During training sessions, which were conducted once a day for the first 5 days, a platform was placed 1 cm below the surface, and the mice were placed into the water in different quadrants facing the pool wall. If the mouse failed to locate the platform within 60 s, it was guided to the platform and allowed to stay on the platform for 15 s. Animal trajectories were recorded using a video-based image tracking system and ANY maze software (Global Biotech, Mount Laurel, NJ, United States).

At 24 h after the Morris water maze testing, three mice per group were deeply anesthetized with 1% carbrital and perfused transcardially with 0.9% saline, followed by cold 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS, pH 7.4). Brains were dissected out and maintained overnight in 4% paraformaldehyde, cryopreserved in PBS containing 30% sucrose, then stored at −70°C until use. Brain sections (10 μm) were prepared and incubated for 12 h with one of the following antibodies: mouse monoclonal antibody against TYROBP (B-2, 1:100, cat# sc-166086, Santa Cruz Biotechnology, Dallas, TX, United States), mouse monoclonal antibody against tau (Tau5, 1:100, cat# ab80579, Abcam, Cambridge, MA, United States), mouse monoclonal antibody against Ser202/Thr205-phosphorylated tau (AT8, 1:100, cat# MN1020, Thermo Fisher Scientific, Waltham, MA, United States), mouse antibody against Aβ(1-16) (6E10, 1:100, cat# SIG-39320, Biolegend, San Diego, CA, United States), rabbit antibody against m6A (1:200, cat# A17924, ABclonal, Wuhan, Hubei, China) mixed with 2% BSA, 1 × DNase I Buffer (10 mM Tris-HCl, 2.5 mM MgCl2, 0.5 mM CaCl2), 25 U/mL DNase I (cat# 79254, Qiagen, Beverly, MA, United States), mouse monoclonal antibody against Iba1 (1:100, cat# ab283319, Abcam), rabbit monoclonal antibody against Iba1 (1:100, cat# ab178846, Abcam), mouse monoclonal antibody against GFAP (1:200, cat# CL488-60190, Proteintech Group, Chicago, IL, United States) and mouse monoclonal antibody against NeuN (1:200, cat# 66836-1-Ig, Proteintech Group). The sections were then incubated for 1 h with either Alexa Flour 488-conjugated goat anti-rabbit IgG (1:50, cat# SA00013-2, Proteintech Group) or Alexa Flour 568-conjugated goat anti-mouse IgG (1:1,000, cat# ab175473, Abcam). Images were acquired using an upright Zeiss microscope (Axio Imager.Z2, Carl Zeiss, Oberkochen, Germany), and analyzed using Image J (National Institutes of Health, Bethesda, MD, United States) (Schneider et al., 2012).

Total RNA was isolated from each group of nine mice aged 6 months, then purified using TRIzol reagent (cat# 15596018, Invitrogen, Carlsbad, CA, United States). The amount and quality of the purified RNA were examined using the ND-1000 system (NanoDrop, Wilmington, DE, United States). Only RNA giving an absorbance ratio A₂₆₀/A₂₈₀ of 1.8–2.0 was used in further experiments. An aliquot of mRNA (1 μg per sample) was reverse-transcribed into cDNA using the SuperScript III First-StrandKit (cat# 18080051, Invitrogen), and 1 μL of cDNA (diluted 1:2) was used as template in quantitative PCR in the ChamQ™ Universal SYBR qPCR Master Mix (cat# Q711-02, Vazyme, Nanjing, Jiangsu, China). β-actin served as the internal control. Primer sequences were designed using the online Primer Blast tool (¹ Supplementary Table 1). Levels of mRNA were expressed using the 2^–ΔΔCt method (Lv et al., 2016). Only genes associated with transcript Ct ≤ 30 were considered to be expressed.

Total protein was isolated from frozen hippocampi from groups of six mice aged 6 months using RIPA lysis buffer (Beyotime Biotechnology, Nanjing, China) containing PMSF (Beyotime Biotechnology). Lysates were left standing for 30 min, then centrifuged at 12,000 × g for 20 min at 4°C. Protein concentration was estimated using bicinchoninic acid (Beyotime Biotechnology), and equal amounts (30 μg) were separated by electrophoresis on precast 10% Bis-Tris gels (Bio-Rad Laboratories, Hercules, CA, United States), transferred to polyvinylidene difluoride membranes, and incubated with one of the following primary antibodies: rabbit antibody against TYROBP (B-2, 1:200, cat# sc-166086, Santa Cruz Biotechnology), rabbit antibody against METTL3 (1:1,000, cat# ab195352, Abcam), rabbit antibody against METTL14 (1:1,000, cat# A8530, ABclonal), rabbit antibody against WTAP (1:1,000, cat# 56501, Cell Signaling Technology, Danvers, MA, United States) and mouse antibody against GAPDH (1:50,000, cat# AC033, ABclonal). Secondary antibodies included horseradish peroxidase-conjugated goat anti-rabbit secondary IgG (1:5,000, cat# AS014, ABclonal) and goat anti-mouse IgG (1:5,000, cat# AS003, ABclonal). Antibody binding was visualized using enhanced chemiluminescence (cat# 32106, Thermo Fisher Scientific) and a Tanon 5200 imaging analysis system (Tanon Technology, Shanghai, China). Band intensities were analyzed using Image J.

RIPA-soluble protein was isolated from the brains of groups of three to six mice aged 2, 6, and 9 months. Aβ levels were quantified using commercial ELISAs against Aβ40 (cat# MU30299, BIOSWAMP, Wuhan, Hubei, China) and Aβ42 (cat# MU30114, BIOSWAMP) according to the manufacturer’s protocols.

Levels of m6A methylation in total hippocampal RNA from groups of six mice aged 6 months were measured using a commercial kit (cat# ab185912, Abcam) according to the manufacturer’s instructions. Each sample contained 1,000 ng of total RNA. Absorbance was measured at 450 nm and converted to m6A levels using a standard curve.

The MeRIP-Seq required at least 100 μg RNA in each sample; therefore, the RNAs of three mouse hippocampi (either WT or Tyrobp^–/–) were pooled as one sample for MeRIP-Seq. The RNA was isolated as described above, and its integrity was assessed using a Bioanalyzer 2100 (Agilent, CA, United States) and denaturing agarose gel electrophoresis. RNA was used only if the RNA integrity number > 7.0. Poly(A) RNA was purified from 50 μg total RNA using oligo(dT)₂₅ Dynabeads (cat# 61005, Thermo Fisher Scientific), and fragmented into small pieces at 86°C for 7 min using a Magnesium RNA Fragmentation Module (cat# e6150, New England Biolabs, Ipswich, MA, United States). The cleaved RNA fragments were incubated at 4°C for 2 h with an antibody against m6A (cat# 202003, Synaptic Systems, Göttingen, Niedersachsen, Germany) in 50 mM Tris-HCl, 750 mM NaCl and 0.5% Igepal CA-630. Immunoprecipitated RNA was reverse-transcribed into cDNA using SuperScript™ II Reverse Transcriptase (cat# 1896649, Invitrogen), which was then used as template to synthesize U-labeled second-strand DNA using E. coli DNA polymerase I (cat# m0209, New England Biolabs), RNase H (cat# m0297, New England Biolabs) and dUTP (cat# R0133, Thermo Fisher Scientific). The blunt ends of strands were extended with A bases for ligation to indexed adapters. Each adapter contained a T-base overhang to allow it to be ligated to the A-tailed DNA. Single- or dual-index adapters were ligated to the fragments, which were selected by size using AMPureXP beads. The ligated products were treated with a heat-labile UDG enzyme (cat# m0280, New England Biolabs), then amplified by PCR under the following conditions: initial denaturation at 95°C for 3 min; eight cycles of denaturation at 98°C for 15 s, annealing at 60°C for 15 s, and extension at 72°C for 30 s; then final extension at 72°C for 5 min. The average insert size for the final cDNA library was 300 ± 50 bp. The library was subjected to 2 × 150-bp paired-end sequencing (PE 150) on an illumineNovaseq™ 6000 (Illumina, San Diego, CA, United States).

Fastp software² (Chen et al., 2018), with its default parameters, was used to remove adapter contamination and low-quality reads, defined as Q ≤ 10. Fastp was also used to verify sequence quality of the input and immunoprecipitated samples. We used HISAT2³ (Kim et al., 2015) to map the reads to the Mus musculus genome (version: v96). Mapped reads of immunoprecipitated and input libraries were analyzed using the exomePeak package in R⁴ (Meng et al., 2014), which identified m6A peaks using the bed or bigwig format. Output was visualized using IGV software⁵ (Robinson et al., 2011). MEME⁶ (Bailey et al., 2009) and HOMER⁷ were used to identify de novo and known motifs, followed by localization of the motif with respect to the peak summit.

Peaks were annotated based on intersection with gene architecture using the ChIPseeker package in R⁸ (Yu et al., 2015). The expression levels of all mRNAs in input libraries were assessed using StringTie⁹. FPKM was calculated as total exon fragments/mapped reads (millions) × exon length (kB). The mRNAs differentially expressed between Tyrobp^–/– and WT mice were defined as those showing fold change ≥ 2 or ≤ −2 and P < 0.05 based on the edgeR package in R¹⁰ (Robinson et al., 2010).

All statistical analyses were conducted using GraphPad Prism (version 8.0, Graphpad, San Diego, CA, United States). Data were presented as mean ± SEM. Pairwise comparisons were assessed for significance using Student’s t test for independent samples. Differences in the Morris water maze test were assessed using two-way ANOVA for repeated measures, followed by Tukey’s post hoc test. Differences in gene expression profiles were assessed in terms of fold change. P < 0.05 were considered statistically significant.

Tyrobp^–/– mice showed extremely low expression of TYROBP, whereas WT animals showed abundant protein, especially in microglia (Figures 1A–C).

On day 1 in the Morris water maze, escape latency was higher for Tyrobp^–/– mice than for WT animals aged 2 months (20.95 ± 3.31 vs. 10.51 ± 3.21 s, P = 0.008), 6 months (24.27 ± 4.80 vs. 12.54 ± 2.78 s, P = 0.019) or 9 months (39.63 ± 9.33 vs. 23.90 ± 3.86 s, P = 0.036; Figures 2A–C). Similar results were observed on day 5 among animals aged 2 months (10.65 ± 0.61 vs. 8.97 ± 0.81 s, P = 0.032), 6 months (22.38 ± 0.77 vs. 13.27 ± 0.99 s, P = 0.0001) or 9 months (33.39 ± 1.93 vs. 16.58 ± 1.02 s, P = 0.002; Figures 2A–C). In contrast, Tyrobp^–/– and WT mice did not differ significantly in swimming speed (data not shown). Tyrobp^–/– mice at all three ages showed higher levels of soluble Aβ40 and Aβ42 than WT mice in the hippocampus, cortex and cerebellum (Figures 2D–I).

Given that Tyrobp^–/– mice showed abnormal behavior and elevated levels of soluble Aβ40 and Aβ42 at 2, 6, and 9 months, we used six-month-old animals in subsequent experiments. Tyrobp^–/– mice showed significantly higher hippocampal levels of total tau (5.23 ± 0.24 vs. 3.06 ± 0.20%, P < 0.0001), Ser202/Thr205-phosphorylated tau (13.82 ± 0.64 vs. 6.20 ± 0.37%, P < 0.0001) and Aβ (17.40 ± 0.56 vs. 5.24 ± 0.28%, P < 0.0001; Figure 3). Similar results were observed in the cortex: total tau, 8.82 ± 0.45 vs. 4.82 ± 0.32%, P < 0.0001; Ser202/Thr205-phosphorylated tau, 12.75 ± 0.48 vs. 3.71 ± 0.45%, P < 0.0001; and Aβ, 15.39 ± 1.92 vs. 5.09 ± 1.10%, P < 0.0001 (Supplementary Figure 1). Similar results were also observed in the cerebellum: total tau, 7.91 ± 0.31 vs. 4.76 ± 0.27%, P < 0.0001; Ser202/Thr205-phosphorylated tau, 7.98 ± 0.53 vs. 3.68 ± 0.28%, P < 0.0001; and Aβ, 7.27 ± 0.43 vs. 3.32 ± 0.24%, P < 0.0001 (Supplementary Figure 2).

To benchmark the phenotype of Tyrobp^–/– mice against an AD phenotype, we compared hippocampal levels of total tau and Ser202/Thr205-phosphorylated tau between Tyrobp^–/– and APP/PS1 mice, all 15 months old. Tyrobp^–/– mice showed higher levels of Ser202/Thr205-phosphorylated tau (0.45 ± 0.10 vs. 0.88 ± 0.07, P = 0.043) and total tau (0.79 ± 0.22 vs. 1.51 ± 0.07, P = 0.047) than WT animals. However, Tyrobp^–/– mice showed lower levels of Ser202/Thr205-phosphorylated tau (0.79 ± 0.22 vs. 1.49 ± 0.10, P = 0.015) and total tau (1.51 ± 0.07 vs. 1.81 ± 0.02, P = 0.012) than the APP/PS1 mice (Figure 4).

Tyrobp^–/– mice contained significantly lower hippocampal levels of Mettl3, Mettl14, and Wtap mRNAs, which encode methyltransferases (P < 0.001, Figure 5A). These mRNA results were verified at the protein level by western blotting (Figures 5B,C). In contrast, the two types of animals did not differ significantly in expression of the Fto or Alkbh5 genes encoding demethylases.

The global m6A RNA methylation level in hippocampal was significantly lower in six-month-old Tyrobp^–/– mice than in age-matched WT animals (0.0357 ± 0.00008 vs. 0.0529 ± 0.00005%, P = 0.005), (Supplementary Figure 3A). Similarly, based on immunofluorescence staining, the m6A RNA methylation were significantly lower in six-month-old Tyrobp^–/– mice than in age-matched WT animals (12.12 ± 1.69 vs. 5.54 ± 0.81, P = 0.024) (Supplementary Figure 3B), and the decrease in m6A methylation occurred in microglia (1.14 ± 0.12 vs. 0.63 ± 0.06, P = 0.020) (Supplementary Figures 3C,D), astrocytes (4.27 ± 0.39 vs. 2.76 ± 0.30, P = 0.037) (Supplementary Figures 3E,F) and neurons (10.17 ± 1.20 vs. 4.78 ± 0.55, P = 0.015) (Supplementary Figures 3G,H).

The original data, which were deposited in the GEO database under accession number GSE179827, were of generally high quality, with > 97% of reads meeting the Q20 criterion and > 92% of reads exceeding the Q30 criterion (Supplementary Table 2). Tyrobp^–/– mice showed 810 m6A peaks differing significantly from WT animals (Supplementary Table 3), of which 498 peaks were significantly higher and 312 significantly lower in Tyrobp^–/– mice (Figure 6A). In Tyrobp^–/– and WT mice, the m6A peaks were enriched mainly near stop codons and 3′-untranslated regions (Figure 6B), and the peaks differing significantly between the two types of mice occurred most often in the 3′-untranslated region (53.26%), followed by other exons (18.62%), 5′-untranslated regions (17.93%), and first exons (10.2%) (Figure 6C). Compared to WT animals, Tyrobp^–/– mice showed higher proportions of m6A peaks in 3′-untranslated regions (49.32 vs. 47.22%) and first exons (11.63 vs. 11.47%), but lower proportions in 5′-untranslated regions (17.34 vs. 18.88%) and other exons (21.71 vs. 22.43%) (Figures 6D,E).

In Gene Ontology (GO) analysis, the upregulated peaks were significantly associated with the following biological processes: regulation of DNA-templated transcription, positive regulation of transcription by RNA polymerase II and signal transduction. The upregulated peaks were associated with the cellular components of cytoplasm, membrane and nucleus; and they were associated with the molecular functions of protein binding, metal ion binding and nucleotide binding (Figure 6F).

The downregulated peaks, in contrast, were associated mainly with the biological processes of regulation of DNA-templated transcription, signal transduction, and positive regulation of transcription by RNA polymerase II (Figure 6G). Downregulated peaks were also associated with the cellular components of membrane, cytoplasm and nucleus; and they were associated with the molecular functions of protein binding, metal ion binding and DNA binding.

RNA sequencing data showed that 86 genes were upregulated in Tyrobp^–/– mice relative to WT controls, while 85 genes were down-regulated (Supplementary Table 4). The top five upregulated genes were Fam177a, Pmch, Pcdhgb4, Hcrt, and Tmem181c-ps, and the top five downregulated genes were Mgam, Pcdhga2, Gpr176, Pcdhga9, and Proz (Figures 7A,B). The altered expression of three upregulated genes (Fam177a, Pcdhgb4 and Tmem181c-ps) and three downregulated genes (Pcdhga2, Gpr176 and Slc16a7) was verified using quantitative real-time PCR (Figure 7C).

The major GO terms and KEGG pathways involving upregulated genes are shown in Figures 7D,E. Upregulated KEGG pathways included the cAMP signaling pathway, axon guidance and MAPK signaling pathway (Figure 7E). The major GO terms and KEGG pathways involving downregulated genes are shown in Figures 7F,G. Downregulated KEGG pathways included the AMPK signaling pathway, PI3K-Akt signaling pathway, and human papillomavirus infection (Figure 7G).

We identified genes whose m6A methylation at the RNA level and whose gene expression were altered (Supplementary Table 5), leading to four groups (Figure 8A): hypermethylation and upregulation, 38 genes; hypomethylation and downregulation, 16 genes; hypomethylation and upregulation, 43 genes; and hypermethylation and downregulation, 84 genes.

In GO analysis, the top three biological processes were regulation of DNA-templated transcription, negative regulation of ERK1 and ERK2 cascades, and chemical synaptic transmission. The top three cellular components were nucleus, cytoplasm, and membrane, while the top three molecular functions were protein binding, metal ion binding, and nucleic acid binding (Figure 8B). The KEGG analysis enriched for the overlap genes were cell adhesion molecules, NOD-like receptor (NLR) signaling and folate biosynthesis (Figure 8C).

Cell adhesion molecules of interest included Cldn19, H2-M5 and Alcam. In Tyrobp^–/– mice, these genes were m6A-hypomethylated and upregulated. The NLR signaling pathway includes Nlrp6 and Pstpip1. In Tyrobp^–/– mice, the Nlrp6 gene was m6A-hypermethylated and downregulated, while the Pstpip1 gene was m6A-hypomethylated and upregulated.