Lack of MS4A6A/MS4a6d harms protective brain immune cells and increases inflammation in Alzheimer's disease model

All CSF samples in CABLE cohort were collected by lumbar puncture. The CSF was then processed or stored after instant centrifugation at 2000 × g for 10 min. The thaw/freezing cycle for each sample was limited twice. CSF sTREM2 concentrations were determined with the ab224881 Human TREM2 SimpleStep ELISA (Enzyme-linked immunosorbent assay) kit from Abcam. CSF Aβ concentrations were determined with INNOTEST® β-Amyloid (1–42) 81583 and INNOTEST® β-Amyloid (1–40) 80462 from FUJIREBIO. P-tau was measured with INNOTEST® PHOSPHO-TAU (181p) 81581 from FUJIREBIO. Total tau was measured with INNOTEST® hTAU-Ag 81579 from FUJIREBIO. Progranulin was detected using the RMEE103R Progranulin Human ELISA kit (Biovendor, Heidelberg, Germany). α-Synuclein was measured via LEGEND MAX™ Human α-Synuclein ELISA Kit with pre-coated plate (No:844101, BioLegend). All biomarker indexes were measured twice by skilled operators who were blinded to clinical data. For these detected proteins, intra-batch variance for technical replicates was < 5%. The inter-batch coefficients of variations were < 20%.

Samples were genotyped using the Infinium Asian Screening Array Kit on the Illumina BEADLAB platform. Standard quality control was then performed, as described in our previous study [37]. Variants located within the MS4A6A gene (chr11:59,939,080–59,952,139 according to GRCh37/hg19) were extracted using the PLINK 2.0 software [38]. An additive linear regression in PLINK 2.0 was applied to investigate the effects of genotypes on CSF biomarkers, adjusted with age, sex, the first two principal components as covariates. The suggestive and genome-wide significant thresholds were P < 1 × 10^–5 and P < 5 × 10^–8, respectively [39].

Three sources of genome-wide association studies (GWAS) data were used for this meta-GWAS analysis, including the UK Biobank database, and Kunkle stage1 [6], as well as data from the FinnGen consortium [36].

The imputed SNP genotype data employed in this study were based on the GRCh37/hg19 build [40]. In UK Biobank, we excluded participants who were not of European descent, those with a missing genotype rate exceeding 0.05, or a mismatch data on sex (Data field 31 and 22,001), those with chromosomal disorders, heterozygosity rate outliers, more than ten putative third-degree relatives, and those who are not utilized in computing the principal components. Additionally, multi-allelic variants and call rate < 0.95, a minor allele frequency < 0.01, Hardy–Weinberg P-value < 10^–6, or imputation quality score < 0.5 were filtered out. For subsequent analysis, we focused on variants specifically located in the MS4A6A gene. Sample whose age at disease onset or last follow-up was younger than 65 years old were removed. Following the quality control process, a total of 3,076 AD cases and 254,938 controls with available phenotype and covariates data were retained for further investigation. GWAS was performed to test the correlation between variants in MS4A6A and AD risk using logistic regression controlled for age, sex, batch, and the first three principal components in PLINK2.0 [38]. Furthermore, we obtained publicly accessible summary-level data from Kunkle stage 1 (21,982 AD cases and 41,944 controls) and FinnGen consortium (10,520 AD cases and 401,661 controls).

GWAMA software was used to combine summary statistics from the three studies [41].

Human microglial cells (HMC3) were maintained in MEM supplemented with 10% FBS at 37 °C with 5% CO₂. For transduction, cells were seeded at 5 × 10⁵ cells/well in 6-well plates and infected with MS4A6A overexpression lentivirus rLV-CMV-hMS4A6A-3xHA-P2A-EGFP-hPGK-Puro-WPRE at MOI 15 in the presence of 6 μg/mL Polybrene. After 24 h, the medium was replaced with fresh complete medium. Selection with 2 μg/mL puromycin commenced at 72 h post-infection and continued for 10 days until non-transfected controls showed complete mortality.

Target cells were harvested. Total RNA was isolated with FreeZol Reagent (Vazyme). Libraries were prepared using the Illumina TruSeq Stranded mRNA Kit and subsequently used for RNA-seq analysis. High-quality reads were aligned to the mouse reference genome using Bowtie2. Expression levels for each of the genes were normalized to fragments per kilobase of exon model per million mapped reads using RNA-seq by Expectation Maximization. DEGs (differentially expressed genes) between samples were identified and performed clustering analysis and functional annotation. Differential expression analysis was conducted in DESeq2 (P < 0.05, |log2FC|> 0.5). Gene ontology (GO) and KEGG pathway enrichment were analyzed using ClusterProfiler v4.2.2. Three biologically independent replicates were included for each experimental group.

We generated the base editing plasmid used for MS4A6A related SNP mutation. For rs646924: T > C (A > G) mutation, the adenine deaminase/cytidine deaminase MU148 described previously was inserted into the internal 1,249 site of the nickase Cas9 variant SpG [42]. sgRNA was cloned into the pU6-sgRNA-EF1α-UGI-T2A-PuroR plasmid linearized with BsaI; paired oligonucleotides were synthesized, annealed and inserted to construct the sgRNA expression plasmid. The sgRNA sequence used for the rs646924: A > G mutation was 5’-ggcaagggaatagtgagatg-3’.

Human microglial cells were seeded at 600,000 cells per well on 6-well plates. The following day, 1500 ng of the base editor expression plasmid was co-transfected with 500 ng the sgRNA expression plasmid into human microglial cells using the jetPRIME® transfection reagent (Polyplus-transfection) following the manufacturer’s instructions. 24 h post-transfection, the complete medium was replaced with a fresh complete medium containing 2 μg/ml puromycin every day for 4 days. Next, cells were trypsinized and resuspended in complete medium and then sorted into a 96-well plate containing complete medium. Surviving single-cell colonies were transferred into the 12-well plates and cultured for another 9–10 days. After cell expansion, the genome DNA of each single-cell clone was extracted for targeted amplification to examine the mutation status. Cell clones containing the expected SNP mutations were used for functional studies.

Total RNA was isolated from edited cultured microglia and unedited control using FreeZol Reagent (Vazyme) according to the manufacturer’s protocol. RNA concentration and purity were determined spectrophotometrically (NanoDrop). After DNase I treatment, 1 μg RNA was reverse-transcribed into cDNA (HiScript III RT SuperMix for qPCR, Vazyme) with oligo(dT) primers. qPCR reactions were performed in triplicate on QuantStudio 5 (Applied Biosystems) using ChamQ Universal SYBR qPCR Master Mix (Vazyme) in 20 μL total volume. Relative expression was normalized to GAPDH reference genes and calculated using the 2^(-ΔΔCt) method. Primer sequences for target genes (e.g., MS4A6A: F 5′-TGTGGCATGATGGTATTGAGC-3′, R 5′-AGGGTCCTATGAATGGGTAAGC −3′; GAPDH: F 5′-GAAGGGCATCTTGGGCTACAC-3′ R 5′-GTTGTCATTGAGAGCAATGCCA-3′).

The Ms4a6d^−/− mice were generated by Cyagen Bioscience using the CRISPR/Cas9 technique. Exon 2–7 of Ms4a6d were knocked out. APP/PS1 mice were purchased from Aniphe Biolaboratory and crossed with Ms4a6d^−/− mice to develop Ms4a6d^−/− under APP/PS1 mice background (termed throughout the paper APP/PS1: Ms4a6d −/−, APP/PS1: Ms4a6d +/−, and APP/PS1: Ms4a6d +/+). The genotype of mice was confirmed by Sangon Biotech. Age-matched wild type mice were bought from Vital River Laboratory Animal Technology. In the animal center of Huashan hospital, mice were housed in rooms with temperature and humidity controls (21 ± 1.5℃ and 50 ± 10%) under conventional 12-h light/dark cycle settings. The mice were kept in an environment free of pathogens. All of the experiments comply with the animal study protocol approved by the Fudan University.

The midbrain tissue used in this study was collected from an 89 years old female sporadic Parkinson's disease patient. This patient was autopsied after death and pathologically diagnosed with Lewy body pathology found in the substantia nigra region. The formalin-fixed, paraffin-embedded specimen was a gift from Body Organ Donation Center of Dalian Medical University. The consent for body and organ donation (code number: 00303) was carefully signed by the donor. Her next of kin have provided patient's basic information and clinical history.

A total of nine C57BL6/J mice (three for each age group) were initially housed at Jiangsu Cavens Biolaboratory Inc., and maintained at two weeks younger than their designated sacrifice ages: postnatal 2 months (8 weeks), 6 months (24 weeks), and 12 months (48 weeks). Subsequently, they were transferred to Daoke Pharmacology Inc. Facility for a minimum of one week to acclimate prior to sacrifice. Mice were group-housed and, prior to trans-cardial perfusion, anesthetized using 4% isoflurane via face mask. The collected whole brains were washed with cold phosphate-buffered saline, dried, and then embedded in optimal cutting temperature (O.C.T.) compound. Brains were cryosectioned at −20°C on a cryostat and immediately snap-frozen at −80 °C. Stereo-seq (SpaTial Enhanced REsolution Omics sequencing) for the spatial trasncriptomic analysis were prepared as described before [43]. In total, three forebrain sections (interaural 3.94 mm, bregma 0.14 mm) and three hindbrain sections (interaural 1.50 mm, bregma −2.30 mm) of 10 μm thickness were collected at each age group, and these sections were promptly adhered to the surface of Stereo-seq chips at −20°C. Tissue sections were fixed in cold methanol for at least 40 min at −20°C and washed three times in 1 × saline sodium citrate before proceeding with Stereo-seq library preparation. All of the procedures were conducted with the approval of the Institutional Animal Care and Use Committee of Fudan University (approval number: JS-256).

Microdialysis probes with a 2 mm, 100 kDa molecular weight cutoff membrane (CMA 8 microdialysis probes) were stereotaxically implanted into following the coordinates for left hippocampus (bregma: −3.0 mm, 2.5 mm lateral to midline, and 4.0 mm below dura). Perfusion buffer (0.15% bovine serum albumin in artificial cerebrospinal fluid (in mM: 1.3 CaCl₂, 1.2 MgSO₄, 3 KCl, 0.4 KH₂PO₄, 25 NaHCO₃, and 122 NaCl, pH7.35) was perfused at a 1-µl/min flow rate with a syringe pump (CMA). Microdialysates were collected every 60 min at 4 °C using a refrigerated fraction collector. Mice was anesthetized with isoflurane inhalation during the process.

Interstitial fluid (ISF) human Aβ_1-x concentrations were quantified hourly via ELISA (No.27729, IBL). Baseline Aβ_1-x levels were defined as the mean concentration from hours 3–5 after probe insertion. Then ISF was obtained for 8 h after intraperitoneally injecting each mouse with 6.7 mg/kg of the γ-secretase inhibitor Compound E (Med Chem Express). Aβ_1-x values at each time point were expressed as percentage of baseline mean (% baseline).

All animals received transcardial perfusion with ice-cold phosphate-buffered saline (PBS) after being anesthetized with isoflurane inhalation. The 4% paraformaldehyde fix solution was utilized to fix the right hemispheres for a whole night at 4 °C. Fixed mice brain tissues were sectioned at 45 μm-thick on a vibratomes (Leica, Leica VT1000S). The left hemisphere's cortex and hippocampus were carefully dissected out, then quickly frozen in liquid nitrogen for further biochemical testing.

Mouse sections were rinsed with PBS three times, then blocked with 5% bovine serum albumin in PBS and permeabilized with 0.5% Triton-X 100 in blocking solution. To ensure adequate antigen retrieval, the human specimen was heated at 95 °C in the 10 mM citrate buffer (pH 6.0) for 40 min and washed for 5 min three times after cooling down at room temperature.

Tissues were first incubated with corresponding primary antibodies overnight: Iba1 (1:100, ab5076, Abcam) or (1:500, 019–19741, Wako), Lamp1 (1:200, 1D4B, DSHB), Phospho-Jak2 (Tyr1007/1008) (1:100, 3771S, Cell Signaling Technology), 4G8 (1:1000, 800,701, BioLegend), Phospho-NF-κB p65 (Ser536) (1:1000, 3033S, Cell Signaling Technology), GFAP (1:500, ab4674, Abcam), CD68 (1:200, MCA1957, Bio-rad), MS4A6A-N-terminal (1:50, ab189983, Abcam). Secondary antibodies were chosen according to primary antibodies host species: anti-goat IgG Alexa-Fluor555 (1:500, Invitrogen), anti-mouse IgG Alexa Fluor 488 (1:500, Invitrogen), anti-rabbit IgG Alexa Fluor 647 (1:500, Invitrogen), anti-rat IgG Alexa Fluor 555 (1:500, Invitrogen), anti-chicken IgG Alexa-Fluor 488 (1:500, Invitrogen) for 1 h at room temperature at dark. Thioflavin S (Sigma-Aldrich, 2 mg/10 mL) was used to stain plaques for 25 min. Then sections were washed for 5 min three times before mounting on slides with Fluoromount-G® mounting medium or DAPI Fluoromount-G® mounting medium (SouthernBiotech).

Adult C57BL/6 J mice were anesthetized with isoflurane. Brains were harvested within 5 min post-sacrifice and maintained on ice. For embedding, tissues were briefly blotted on sterile paper towels, oriented in cryomolds, and covered with O.C.T. compound. Molds were rapidly frozen on a liquid nitrogen platform for 10 min until O.C.T. solidified. Continuous coronal sections were made at −20 °C using a constant cooling microtome (Leica) with a thickness of 20 μm. The sections were adhered to slides treated, dried at room temperature for 30 min, and then stored at −80 °C for long-term preservation.

RNA in situ hybridization was processed according to the manufacturer's protocol for the RNA-scope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics). The sections were dehydrated in a gradient of 70%, 50%, and 100% ethanol, then incubated with hydrogen peroxide at room temperature for 10 min, and washed twice. Target-specific probes were hybridized sequentially: mouse Aif1-C2 (433,091, Advanced Cell Diagnostics) and Ms4a6d (314,591, Advanced Cell Diagnostics) probes were applied and incubated at 40 °C for 2 h in a HybEZ™ hybridization oven and wash twice. Post-hybridization signal amplification was performed through 2-step amplifier reactions (AMP1-2) as per kit instructions. Tyramide Signal Amplification (TSA) was subsequently conducted using the TSA Plus Cyanine Evaluation Kit, with fluorophore-conjugated tyramide (1:750, Advanced Cell Diagnostics) incubated at 40 °C for 30 min protected from light. Then sections were mounted on slides with DAPI Fluoromount-G® mounting medium (SouthernBiotech).

All confocal images were collected through an Olympus system. More than 20 plaques from each mouse were taken for all genotypes. For each staining set, the setting of laser and detector were constant. Using a 60 × oil lens, graphs were captured at 1024 × 1024 pixels with 4 × zoom and a 1 μm z-step. Images were processed with ImageJ or Matlab (Mathworks). Maximum intensity projection of the z-stack for each graph was used for morphological analysis. During analysis, the mice and group data were kept blind.

Plaque shapes were identified using a threshold fluorescence intensity on the Thioflavin S channel that was two standard deviations above the average fluorescence intensity of the image. Activated microglia processes tightly encased amyloid plaques, with Iba1 (Ionized calcium binding adapter molecule 1) expression higher than baseline microglial processes, were characterized as the microglia barrier surrounding compact plaques. The Iba1 fluorescence intensity in the cell body was used as the cutoff value to threshold the Iba1 channel. The percentage of the plaque perimeter covered by the microglia barrier was ascertained by calculating the angle between contact points from the plaque center.

For analysis of the staining of Iba1, CD68, 4G8, phosphorylated JAK2 (Janus kinases 2) and phosphorylated p65, a z-projection of maximal fluorescence intensities across 10 optical slices through the center of the plaque core was made. The intended fluorescent channel was thresholded using “Default” preset in FIJI and fluorescent colocalization analysis were performed using “AND” function for colocalization.

For measuring amyloid plaque distribution, the Thioflavin S fluorescent channel was thresholded using “Triangle” present in FIJI and those plaques were measured using “Analyze Particle” function. The size and intensity of each particle was recorded and analyzed together. The total plaque area was calculated according to the percentage in each region (cortex, hippocampus, thalamus, basolateral amygdala).

Axonal spheroids areas were calculated by thresholding the Lamp1 (Lysosome associated membrane protein 1) channel image with 150% average fluorescence intensity of the background after subtracting the area of plaque in a 10-slice z-projection through the center of plaque core.

Multichannel fluorescence images were acquired on an Olympus confocal system equipped with a 60 × oil-immersion objective. Z-stack images (1 μm step) were processed with maximum intensity projection in ImageJ. Assays were typically performed in parallel with positive and negative controls, to ensure interpretable results. To quantify the RNA expression signal, the fluorescent dots of Aif1 and Ms4a6d within the DAPI-stained cell nuclei in the field of view were recorded. The presence of Aif1 expression indicated the microglia. The proportion of microglia among all the cells expressing Ms4a6d was calculated.

Brain tissues were lysate by buffers with Protease Inhibitor Cocktail (Beyotime) and boiled with 5 × SDS loading buffer (Beyotime). The protein level of each sample was measured by BCA kit (Beyotime) and each sample was then diluted into the same level before loading. 25ug per lane was the standard protein amount for western blotting in this study. Antibodies and dilutions used in this study include: NLRP3 (1:1000, AG-20B-0014, Adipogen), β-tubulin (1:2000, 10,068–1-AP, Proteintech), PSD95 (1:2000, ab238135, Abcam), Synaptophysin (1:2000, 36406S, Cell Signaling Technology), β-actin (1:2000, 20,536–1-AP, Proteintech). Quantitative analyses were performed on ImageJ software.

The concentrations of mouse IL-1β (Interleukin-1β) in cortical lysates were tested with ELISA kits (FineTest) under standard procedures. Total protein level of each sample was measured by BCA kits (Beyotime). Samples were measured in duplicates.

The novel object recognition task is carried out in an open field arena using two different types of objects, cylinder and cuboid. During habituation, the animals were free to explore an empty arena for 5 min. Next day, the animals were brought back to the same arena where mice could investigate two identical objects placed at equal distances from the edge for 10 min. Mouse who had a preference of one side was excluded from further testing. 24 h later, mice were then tested for long-term identification recall by allowing them to wander the same field with the familiar object and a novel object in 10 min. The objects and arena were thoroughly cleaned between each mouse experiments. Mice whose total exploration time was less than 20 s were excluded. The time spent exploring each object, as well as the investigation ratio (novel object exploration time/(novel object exploration time + familial object exploration time)) were recorded. Trajectory of each mouse was tracked with DeepLabCut^TM and analyzed with MATLAB (MathWorks).

The data were shown as mean ± S.E.M. (Standard Error of Mean). For comparisons between two groups, two-tail unpaired Student’s t test was used. For multiple group comparisons, one-way ANOVA tests was employed with post hoc Tukey tests. All statistical analysis were completed on Prism (GraphPad). P values less than 0.05 were regarded significant in this study. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns = not significant.

In order to further investigate the mechanisms underlying these associations, we collected CSF samples from individuals with and without MS4A6A mutations in the CABLE cohort (Supplementary Table S1) and measured the levels of AD-related biomarkers. We identified 31 SNPs of the MS4A6A gene in our cohort, 28 of which have not been previously studied in the context of AD. We found that the majority of MS4A6A SNPs showed significant associations with reduced CSF Aβ42 levels, after controlling for the effects of age and gender (Fig. 1b). While decreased levels of Aβ42 in CSF are often related to increased amyloid deposits in the brain [45, 46], the consistent effects on Aβ42 could also be indications of enhanced amyloid clearance. Importantly, SNPs in the MS4A6A gene seems to selectively affect AD-related pathologies, since we did not observe significant correlation between MS4A6A SNPs and progranulin or α-synuclein levels, and only one SNP (rs7232) correlated with increased soluble TREM2 levels as previously reported [11, 12]. These data strongly suggest that MS4A6A gene likely modulates AD pathogenesis via affecting amyloid pathologies, which we directly examined in vivo.

It is worth noting that all the SNPs reported above locate outside the exon regions of the gene, thus they likely influence disease risk by modulating the expression levels of MS4A6A. To directly test this, we engineered isogenic point mutation rs646924: T > C (A > G) in human microglia cell line (HMC3) using CRISPR-Cas9 (Fig. 1c). We found that rs646924 A allele (risk allele) significantly reduced expression of MS4A6A compared to the G allele (Fig. 1c). These results suggest that decreased MS4A6A transcription could lead to increased AD risk, likely via affecting amyloid-related processes.

To examine the effect of MS4A6A deficiency, we aimed to establish a MS4A6A knockout mouse model. However, the human MS4A6A gene is not present in mice. Previous studies have identified Ms4a6d in mice as a homologue gene for human MS4A6A [34, 47]. We examined the protein sequence and predicted structures of human MS4A6A and the MS4A gene family in mice, and found that Ms4a6d showed the highest similarity to human MS4A6A (Supplementary Fig. S2 and Supplementary Table S2). Therefore, we generated a Ms4a6d knockout mouse model (Supplementary Fig. S3). To directly measure the impact on amyloid-related processes, we cross-bred the Ms4a6d knock-out mice with AD-like transgenic line APP/PS1, and performed intracerebral microdialysis to assess Aβ clearance efficiency in 1-month-old mice before the formation of amyloid deposits. After administering γ-secretase inhibitor to block Aβ production, we observed a decline of Aβ_1-x levels in interstitial fluid (ISF), indicating the rate of Aβ clearance. We found that the rate of Aβ clearance was slower in Ms4a6d-deficient mice compared to control APP/PS1 mice (Fig. 1d). These results demonstrated that Ms4a6d deficiency impairs Aβ clearance in the brain.

We next examined the cellular mechanism underlying the modulation of Aβ depositions by MS4A6A gene. Previous studies indicated that the MS4A6A is selectively expressed in microglia in the brain [47]. We further examined the expression pattern of MS4A6A with human postmortem brain tissues through immunofluorescence staining. Consistent with previous reports, confocal images showed that human MS4A6A was specifically enriched within microglia but not in astrocyte (Fig. 1e). Thus, we hypothesized that MS4A6A may be an important regulator for microglia functions. We also tested the expression pattern of Ms4a6d with in situ hybridization on mice brain sections. The results showed that above 92% on average, and many times 100%, of the Ms4a6d was expressed in microglia. These data indicated that Ms4a6d expression is microglia specific (Fig. 1f), consistent with the results in a latest large-scale proteomic study across murine models [48]. In addition, as previous studies identified elevated levels of MS4A6A is associated with risks of AD onset in humans [19] and elevated expression of Ms4a6d in APP/PS1 mouse [15], we found that there is an increased expression level of Ms4a6d as mice age (Fig. 1g), which could represent a compensatory mechanism to promote amyloid clearance.

Another important protective microglia function is the clearance of amyloid via phagocytosis [27]. Thus, we assessed phagocytosis function in plaque-associated microglia by immunostaining the phagosome marker CD68 [50] and the amyloid fragments. We found that microglia lacking Ms4a6d showed reduced intracellular amyloid contents (Fig. 2f), and correspondingly the levels of phagosomes (Fig. 2g). These findings indicate that in the absence of Ms4a6d, both aspects of microglia protective functions are diminished, pointing to a synergized effect to exacerbate amyloid pathology.

Our previous studies indicated that microglia wrapping around plaques leads to compaction of amyloid materials. Following the lead that Ms4a6d deficiency caused reduced microglia wrapping and increased diffuse amyloid plaques, we then assessed the effect of Ms4a6d deficiency on the compactness of amyloid deposits by analyzing the Thioflavin S fluorescence intensity decay from the center of the deposits to the periphery regions (Fig. 3f). Compact plaque structures exhibit well-defined, distinct boundaries between the edge of the labeled amyloid deposits and the surrounding brain tissue, with fluorescence values changing sharply when crossing this boundary. In contrast, loose filamentous structures would exhibit a more gradual decline in fluorescence values along the radius as the labeled fibers extending outwards. In order to quantify this difference, we measured the fluorescence intensity decay from all directions for individual plaques, and calculated the slopes for the averaged trace (Fig. 3g). We found that amyloid deposits in Ms4a6d-deficient mice exhibited a more diffused morphology (Fig. 3h). Overall, consistent with the expected phenotypes from reduced microglial envelopment and phagocytosis of amyloid, Ms4a6d deficiency led to exacerbated amyloid deposits in mice.

Activation of NF-κB signaling is associated with subsequent activation of NLRP3 (Nod-like receptor pyrin domain containing 3), which is a critical regulator for inflammasome that controls the release of pro-inflammatory cytokines including IL-1β in AD [53, 54]. We then measured the expression levels of NLRP3 in brain tissues from control and Ms4a6d knockout mice using immunoblotting (Fig. 4h). APP/PS1 mice lacking Ms4a6d showed robust elevation of NLRP3 expression compared to control APP/PS1 mice. (Fig. 4i). Furthermore, we measured the levels of IL-1β using a commercially available ELISA kit and found concurrent increase in IL-1β levels in the brains of Ms4a6d knockout APP/PS1 mice (Fig. 4j).

Finally, we conducted behavioral assessments on Ms4a6d-knockout mice. Utilizing the novel object recognition test, we mapped the trajectories of the mice in an open field containing both familiar and novel objects. The heat maps provided a visualization of the changes in the mice's positions, with red indicating longer durations spent in specific areas. Our findings revealed that, compared to age-matched APP/PS1 mice, the Ms4a6d-knockout mice exhibited a decreased proportion of their total exploration time on the novel object (Fig. 5d). This lack of preference for the novel object suggests more severe cognitive impairment in the Ms4a6d-knockout mice.

The above results indicated that MS4A6A/Ms4a6d deficiency could disrupt microglia response to amyloid plaques thus exacerbating the risks of developing AD. It is thus possible that enhancing MS4A6A expression could provide a protective effect. To explore this possibility, we performed MS4A6A overexpression in the HMC3 cell line. RNA sequencing of MS4A6A overexpressed cells revealed 851 DEGs (FDR < 0.1, |log2FC|> 0.5) (Supplementary Fig. S5, Supplementary Material 2). MS4A6A overexpression exerted a broad range of gene expression regulation for microglia including inflammatory signaling pathways (e.g., NF-κB and response to LPS), glial activation, protein kinase activity regulation (regulation of phosphorylation), phagocytosis-related pathways (encapsulating), and migration-associated modules (Fig. 5e). Consistent with our results, we find the general effect of MS4A6A to be anti-inflammation and at the same time activating plaque-responsive genes such as TREM2 (Fig. 5f) [55]. This is consistent with prior reports of coordinated MS4A-TREM2 upregulation in protective microglial states [12], supporting the potential protective effect of MS4A6A upregulation.

Altogether, these data indicated that Ms4a6d deficiency disrupts microglia phagocytosis and envelopment of amyloid, and exacerbates neuroinflammation and neurotoxicity.