Gossypetin ameliorates 5xFAD spatial learning and memory through enhanced phagocytosis against Aβ

To evaluate the therapeutic potential of gossypetin in AD, 5xFAD mice were treated with gossypetin through intragastric administration (Fig. 1A). 5xFAD is an AD model mouse that expresses human APP and PSEN1 transgenes with familial AD-linked mutations. These mutant APP and PSEN1 drive rigorous accumulation of Aβ in the brain and lead to progression of AD pathology [13]. We used only female mice for our experiments because they were reported to show more profound AD pathology [14]. After the drug treatment, spatial learning and memory were evaluated through a Y-maze alteration test and a Morris Water Maze (MWM) test. Y-maze measures short term, spatial working memory [15]. In the Y-maze test, 5xFAD mice treated with gossypetin showed a remarkable improvement in alteration percentage compared to 5xFAD mice treated with vehicle (Fig. 1B). There was no difference in total arm entry between each group (Fig. 1C). We also conducted MWM test to evaluate hippocampus dependent spatial learning and memory. The mice were trained for 5 days prior to a probe test for MWM (Fig. 1D). In the probe test, we observed a significant increase in target quadrant occupancy in the 5xFAD mice treated with gossypetin (Fig. 1E). Although there was no statistical significance, gossypetin-treated 5xFAD mice showed an increased trend in the number of target crossings compared to vehicle treated 5xFAD mice (Fig. 1F). There was no difference in traveled distance during the probe test (Fig. 1G). Various flavonoid compounds show an inhibitory effect against acetylcholinesterase (AChE) [16]. To confirm whether gossypetin had induced the improvement in behavior through AChE inhibition, we performed in vitro AChE inhibition assay. Gossypetin showed a very weak AChE inhibitory effect even at a very high concentration (100 μM) when compared to donepezil (1 μM), which was used as a positive control (Fig. 1H). Taken together, these results suggest that gossypetin ameliorates the impaired spatial learning and memory found in 5xFAD mice.

Gossypetin was reported to have an inhibitory effect against Aβ aggregation in vitro [12]. Therefore, we conducted immunohistochemistry to determine whether the gossypetin improves the behaviors of 5xFAD mice by decreasing Aβ plaques in the brains of 5xFAD mice. We found that gossypetin significantly decreased the number and size of Aβ plaques in the dentate gyrus of the hippocampus and in the cortex of 5xFAD mice (Fig. 2A–E). To validate whether gossypetin was able to decrease other forms of Aβ, we measured the level of Aβ oligomer through dot blot assay. We observed slight decrease of Aβ oligomer in hippocampal lysate of 5xFAD treated with gossypetin (Fig. 2F, G). Then, we conducted Western blot to see whether the level of Aβ monomer was affected (Fig. 2H). Surprisingly, gossypetin significantly decreased Aβ monomer level in the hippocampus as well (Fig. 2I), while the level of amyloid precursor protein (APP) was unaffected (Fig. 2J). To further confirm whether gossypetin decreased the overall level of Aβ, we quantified the level of Aβ in the hippocampus through enzyme-linked immunosorbent assay. We were able to observe that the soluble and insoluble Aβ40 and Aβ42 were all significantly decreased in the hippocampus of gossypetin-treated 5xFAD mice (Fig. 2K–N). Therefore, gossypetin ameliorated behavioral deficits in 5xFAD mice by reducing not only the Aβ plaques but the overall Aβ level in the brain as well.

Since results suggested that gossypetin reduces both Aβ plaques and Aβ monomers, we wanted to elucidate how gossypetin was able to decrease Aβ in 5xFAD mice. We hypothesized that gossypetin would affect either the Aβ production or the clearance pathway. Therefore, we first investigated whether gossypetin modulates the Aβ production pathway. Many flavonoid compounds were previously reported to decrease the activity of β-secretase, which is the rate limiting enzyme of Aβ production [6, 17]. Thus, we measured the activity of β-secretase in hippocampal lysates. Although there was increased β-secretase activity in 5xFAD mice, there was no difference in β-secretase activity between vehicle- and gossypetin-treated 5xFAD hippocampal lysates (Fig. S1 A-G). Next, we measured the mRNA and protein levels of the β-secretase and γ-secretase subunits. We confirmed that gossypetin produced no difference in protein levels of β-secretase and γ-secretase (Fig. S1, H–L) and the mRNA level of BACE1, nicastrin, APH-1, or PEN-2 (Fig. S1, M–P). Based on these results, gossypetin does not seem to affect the production of Aβ.

Because gossypetin did not have any effect on the Aβ production pathway, we thought that gossypetin might stimulate the clearance of Aβ. Glial cells, especially the microglia and the astrocytes, were reported to play a central role in modulating the Aβ clearance and degradation [5]. Therefore, we performed single-cell RNA sequencing (scRNA-seq) from both cortex and hippocampus of 5xFAD mice to see whether gossypetin modulates transcriptome of glial cells. We used the brain of one wild type (WT) mouse and two 5xFAD mice, each from the vehicle- or gossypetin-treated group for analysis. We were able to obtain 17,629 cells from the WT mice and 31,363 cells from the 5xFAD mice (Fig. S2). Graph-based unsupervised clustering after correcting the batch effects in scRNA-seq data identified Aif1-, Cd68-, Fcgr1-, Itgam-, and Trem2-positive clusters as microglia (Fig. S2, A and B; and Fig. S3 A). We observed that the majority of identified cells were microglia while neurons, which is the major cell type in brain tissues, were found to be depleted. This depletion is likely due to the cell dissociation process. Enzymatic dissociation at warm temperature causes relatively selective deaths of neurons and astrocytes compared to mechanical dissociation at low temperatures, establishing a cell population enriched with microglia [18–20]. We decided to focus on microglia and further subclustered the microglial population into eight subtypes in both the cortex and hippocampus microglia according to their differentially expressed genes (DEGs) and inferred cell cycle (Fig. 3A, B; and Fig. S3, A-C). Microglia from the WT mice were mostly homeostatic, while disease-associated microglia (DAM) and other microglial subpopulations were found in the 5xFAD mice (Fig. S3, B and D) [21]. We were unable to find any changes in the number or the proportion of microglial subpopulations driven by treatment of gossypetin.

We next performed gene ontology biological processes (GOBP) enrichment analysis for each microglial subtype. The treatment of gossypetin upregulated largest number of GOBP in hippocampal homeostatic microglia of 5xFAD mice. Notably, terms related to phagocytosis and antigen presentation via MHC II were significantly upregulated in gossypetin-treated 5xFAD mice. An increased trend of antigen presentation via MHC II was found in most of the hippocampal microglial subtypes, while a similar trend was found only in the homeostatic microglia and stage 1 DAM in the cortex (Fig. 3C; Fig. S4 A; and Additional file 1). We further calculated the normalized enrichment score (NES) by gene set enrichment analysis (GSEA) for hippocampal homeostatic microglia. Gossypetin-treated 5xFAD showed a decreased NES score in gliosis related gene sets (Fig. 3D; and Fig. S4 B). As we have previously mentioned, the upregulated genes in gossypetin-treated hippocampal homeostatic microglia showed high enrichment of gene sets related to phagocytosis and antigen presentation via MHC II (Fig. 3D). We also examined the individual DEGs of 5xFAD microglia affected by gossypetin treatment. Among those DEGs, H2-DMa, Lpl, and Clec7a were upregulated, while Ftl1, Apoe, and Spp1 were downregulated in homeostatic and stage 1 and 2 DAM subtypes (Fig. 3E; Fig. S5 A; and Additional files 2 and 3). This result was very interesting because although Apoe, Spp1, Lpl, and Clec7a are all well-known markers for DAM, Apoe and Spp1 showed completely opposite response to gossypetin treatment when compared to Lpl and Clec7a with gossypetin treatment. We calculated the DAM signature score and found a significant decrease in DAM signature from the stage 1 DAM population in the hippocampus and cortex (Fig. S5 B). Previous studies report that Apoe and Spp1 promote inflammation while Lpl and Clec7a sustain phagocytic activity of microglia [22–25]. Since DAM was already reported to have high gene expression associated with phagocytosis, it was very interesting to see gossypetin-mediated increase in phagocytic genes even in DAM subtypes [21]. Taken together, our scRNA-seq results suggest that gossypetin reduced overall transcriptomic signature of DAM in various microglial subpopulations. However, gossypetin upregulated genes associated with phagocytosis, which presumably led to the reduction of Aβ by enhancing clearance through microglia and ameliorated microgliosis in 5xFAD mice.

Based on the results from scRNA-seq, we investigated whether the changes in GOBP terms by gossypetin could be reproduced in 5xFAD mice. First, we investigated whether gliosis was ameliorated in 5xFAD mice through immunohistochemistry since gliosis is a major hallmark of AD. Since gliosis occurs in both microglia and astrocyte in 5xFAD, we stained microglial marker Iba-1 and astrocyte marker GFAP to measure gliosis. We observed a dramatic decrease of gliosis in both the microglia and the astrocytes in the hippocampus and cortex of 5xFAD mice treated with gossypetin (Fig. S). 6

Then, we evaluated whether phagocytic activity of microglia was increased as well, like the result from scRNA-seq suggested. We co-stained hippocampus and cortex with Iba-1 and CD68, which are microglial and lysosomal markers respectively. We observed increased phagosome formation in microglia at the hippocampus and cortex of 5xFAD mice treated with gossypetin (Fig. 4A–D). This indicates that gossypetin increased the phagocytic activity of microglia in 5xFAD mice. To further confirm whether gossypetin was able to enhance the phagocytic activity of microglia, we cultured primary microglia and BV2 mouse microglial cell line. We treated these cells with recombinant 488-tagged fluorescent Aβ (488-Aβ) that emits fluorescence only when it is accumulated inside a cell. Primary microglia and BV2 pretreated with gossypetin showed increased uptake of 488-Aβ as they showed enhanced fluorescence in the cell (Fig. 4E, F; Fig. S7 A and B). To confirm that the increased uptake of 488-Aβ in primary microglia was caused by phagocytosis, we applied cytochalasin D (Cyto D), which is a phagocytosis inhibitor. When Cyto D inhibited phagocytosis, only minimal fluorescence was detected in primary microglia, indicating that gossypetin could increase the phagocytosis of microglia (Fig. 4E, F). Since we have confirmed that gossypetin increases uptake of Aβ through phagocytosis, we wanted to know whether phagocytic dynamics and overall phagocytic capacity of microglia were affected by gossypetin treatment. We conducted live cell imaging to observe the phagocytic activity of BV2 cell during 488-Aβ uptake. We found that gossypetin treatment induced microglia to uptake 488-Aβ more quickly. Also, overall capacity for Aβ uptake was increased in BV2 cells when treated with gossypetin (Fig. S7 C; Additional files 4, and 5). Therefore, we confirmed that gossypetin mediates Aβ clearance by increasing the phagocytic dynamics and capacity of microglia.

Because our scRNA-seq results also showed increased antigen presentation via MHC II in microglia due to gossypetin treatment, we wanted to determine whether we could observe such trends in primary microglia and brain tissues. In primary microglia culture, we observed a further increase in the number of MHC II⁺ microglia in gossypetin-treated primary microglia after Aβ treatment (Fig. 5A, B). Also, we found increased number of MHC II⁺ microglia in the cortex and hippocampus of gossypetin-treated 5xFAD mice (Fig. 5C–F). This confirmed the scRNA-seq data that gossypetin increases MHC II expression in microglia.

5xFAD (Tg6799, Jackson Laboratory) male mice were mated with B6/SJL (#100012, Jackson Laboratory) female mice and the genotypes of the offspring were determined by PCR analysis of tail cut samples. The non-transgenic female littermates were used as wild-type control animals and only female 5xFAD mice were used in this study.

Eight-week-old wild-type (WT) and 5xFAD (Tg) mice were administered with 10 mg/kg of Gossypetin (Boc Science) in 150 μl every day for 13 weeks through intragastric administration. Mice of control groups (WT-Vehicle, 5xFAD-Vehicle) were administered the same volume of vehicle (10mM Tris pH7.5, 200mM NaCl with 1% DMSO or 1% carboxymethyl cellulose with 1% DMSO).

The Y-maze apparatus has three V-shaped arms (40 × 3 × 12 cm) at 120° angles from each other. To test spontaneous alternation, a mouse was placed at the end of an arm and allowed to freely explore the Y-maze during 8 min under the dim light conditions (50 lux). Alternation was defined as consecutive entry in three different arms. The total number and the sequence of individual arm visits were recorded, and alternation ratio was calculated [45, 46].

The water maze tank (120 cm diameter) was filled with water and white non-toxic tempera paint was diluted to make the water opaque. Four different visual cues were placed each cardinal points. The water temperature was set at 22 ± 1 °C before the everyday trial. Mice were habituated to the behavior test room for at least 15 min before starting the test. Three training trials (60 s each, 1 h intervals) were performed per day for five consecutive days. For each training trial, each mouse was allowed to remain on the platform for 15s after finding the platform and was removed from the maze and returned to its home cage. If the mouse did not find the platform within 60s, it was guided to the platform and returned to its home cage after 15s. A probe trial was performed on the sixth day. The trajectories of swim were recorded and analyzed with a video tracking system (SMART v2.5 and SMART v3.0, Panlab, RRID:SCR_002852) [45, 46].

Mice were treated with CO₂ for them to lose consciousness. Then, mice were transcardially perfused with ice-cold phosphate buffer saline (PBS, pH 7.4). After decapitation, skull was removed to collect the brain. Brains were cut into two hemispheres. One hemisphere was fixed in 4% paraformaldehyde (PFA) for overnight at 4 °C and placed into 30% sucrose in PBS at 4 °C until it settles down. Then it was frozen in O.C.T compound (Sakura, 4583) and kept at – 80 °C for further histological analysis. For the other brain hemisphere, it was further dissected to collect hippocampus and cortex. Collected tissues were quickly frozen in liquid nitrogen and kept at – 80 °C for further biochemical analysis. Frozen hippocampus and cortex were mechanically crushed to make them into powder. Powdered hippocampus and cortex were aliquoted for them to be used in protein and RNA extraction [45].

To divide soluble and insoluble protein extraction in hippocampus and cortex, two steps of serial extraction protocol was used. To prepare the soluble extract, hippocampal and cortical samples were weighed and homogenized at 100 mg/ml in cold TBS (50 mM Tris pH 7.5, 150 mM NaCl) with protease inhibitor cocktail (Thermo, A32953). The TBS homogenates were centrifuged at 15,000 rpm for 1 h at 4 °C. The supernatants were stored at − 80 °C until further use. The pellets were reserved for the extraction of insoluble proteins. To prepare the insoluble extract, the previously obtained pellets were dissolved in 5M guanidine buffer (5M guanidine HCl, 150mM NaCl, pH 7.5) and vortexed until all pellets got dissolved completely. After centrifugation at 15,000 rpm for 1 hrat 4 °C, the supernatants were stored at − 80 °C until further use [45].

Soluble fraction of brain lysates was used for western blot. Fifteen to thirty micrograms of brain lysates were separated by SDS-PAGE on SDS gels (5% stacking, 10~15% resolving) and transferred to nitrocellulose membranes (110V constant for 90min, wet transfer). Some of the protein samples were separated on Bolt^TM 4–12% Bis-Tris Plus gels with MES running buffer (200 V constant, for 25 min) and transferred to nitrocellulose membranes using Bio-Rad wet transfer system (100 V constant for 1 h). The membranes were blocked with 5% skim milk in TBST (200mM Tris, 3M NaCl, 1% TWEEN-20, pH 7.6). The membranes were labeled with primary antibodies (anti-GAPDH, 1:1000, Bethyl #A300-641A, RRID:AB_513619; anti-PEN2, 1:500 ,Cell Signaling #8598, RRID:AB_11127393; anti-APH-1, 1:100, Thermo Fisher #PA1-2010, RRID:AB_2227105; anti-Nicastrin, 1:1000, Cell Signaling #5665, RRID:AB_10694688; anti-BACE1, 1:1000, Cell Signaling #5606, RRID:AB_1903900; anti-β-amyloid, 1:100, BioLegend #803002, RRID:AB_2564654), for overnight at 4°C and then labeled with the HRP-conjugated secondary antibodies (anti-Rabbit IgG HRP-conjugate, 1:5,000, Promega #W4011, RRID:AB_430833; anti-mouse IgG Secondary antibody HRP, 1:5000, Thermo Fisher #31430, RRID:AB_228307). Chemiluminescence images were captured by ImageQuant LAS 4000 (Fuji, RRID:SCR_014246).

Soluble hippocampal lysates (4 μg) were diluted in 100μl of TBS. The diluents were loaded on nitrocellulose membrane using the Bio-Dot® microfiltration apparatus (BIO-RAD) and filtered under vacuum. The membrane was washed with TBST and followed by blocking with 10% skim milk. Primary antibody, anti-oligomer A11 (1:2000, Thermo Fisher #AHB0052, RRID:AB_2536236), and HRP-conjugated secondary antibody (anti-Rabbit IgG HRP-conjugate, 1:5,000, Promega #W4011, RRID:AB_430833) were used for detecting soluble oligomers. Chemiluminescence image was captured by ImageQuant LAS 4000 (Fuji, RRID:SCR_014246).

Aβ40 and Aβ42 sandwich ELISA kits (Invitrogen, KHB3481, KHB3442) were used to analyze the soluble and insoluble Aβ levels. The soluble extracts were diluted in the ratio of 1:50 and 1:200 for Aβ40 and Aβ42, respectively, and the insoluble extracts were diluted in the ratio of 1:5,000 and 1:20,000 for Aβ40 and Aβ42, respectively. The extracts were diluted with ELISA diluent buffer. The assay was performed as the manufacturer’s protocol. Synthetic Aβ40 and Aβ42 peptides in the assay kit were used for creating standard curve.

Acetylcholinesterase inhibition assay was performed with acetylcholinesterase inhibitor screening kit (colorimetric, Biovision, Milpitas, CA, USA) according to manufacturer’s protocol. One hundred micrometers of gossypetin dissolved in DMSO was used for acetylcholinesterase inhibition assay.

Beta-secretase Activity Assay Kit (Fluorometric) (ab65357) was used to analyze activity of beta secretases in mouse hippocampal lysate. Assay was performed as manufacturer’s protocol.

Coronal sections (10μm) of mouse brains were cut using a cryostat (Leica CM1850) and stored at – 20 °C until use. Upon use, brain sections were washed three times with PBS and blocked with blocking buffer (5% Fetal Bovine Serum (FBS) + 3% Bovine Serum Albumin (BSA) + 0.3% Triton X-100 in PBS) for 1 h at room temperature (RT). Brain sections were then incubated with primary antibodies (anti-β-amyloid, 1:100, BioLegend #803002, RRID:AB_2564654; anti-Iba-1, 1:200, Wako #019-19741, RRID:AB_839504; anti-GFAP, 1:200, Abcam #ab7260, RRID:AB_305808; anti-CD68, 1:200, Abcam #ab53444, RRID:AB_869007; anti-MHC class 2, 1:100, Thermo Fisher #14-5321-82, RRID:AB_467561) in blocking buffer overnight at 4 °C. Sections were washed three times with PBS and treated with secondary antibody (Anti-mouse IgG Alexa Fluor 405, 1:100, Thermo Fisher #A-31553, RRID:AB_221604; Anti-mouse IgG Alexa Fluor 488, 1:500, Thermo Fisher #A-11001, RRID:AB_2534069; Anti-Rabbit IgG Alexa Fluor 488, 1:200, Thermo Fisher #A-11008, RRID:AB_143165; Anti-Rabbit IgG Alexa Fluor 594, 1:200, Thermo Fisher #A-11012, RRID:AB_2534079; Anti-Rat IgG Alexa Fluor 594, 1:500, Thermo Fisher #A-11007, RRID:AB_10561522; Anti-Rat IgG Alexa Fluor 488, 1:200, Thermo Fisher #A-11006, RRID:AB_2534074) antibody in blocking buffer for 1h protected from light. Brain sections were then washed three times again, and Hoechst 33342 (10μg/ml) diluted in PBS (1:2500) was stained for 10 min protected from light. Lastly, sections were washed again with PBS three times and mounted with fluorescence mounting medium (Dako, S3023). Images were captured with ZEISS Axioplan 2 fluorescent microscope (RRID:SCR_020918) and Olympus confocal microscope (FV3000, RRID:SCR_017015). The GFAP or Iba-1 positive cell regions, number of Aβ (6E10) plaques and area, and phagosomes in prefrontal cortex and dentate gyrus of hippocampus were analyzed by Image J (Fiji) software (RRID:SCR_002285). Stained images were opened with Image J. Dentate gyrus and cortical regions were specifically selected in each image with “freehand selection” tool. After selecting the area for analysis, images were made into 8-bit. Then, fluorescence positive area was visualized through “threshold” tool and quantified with “Analyze Particle” tool [45].

Three-day-old ICR mice were used for primary microglial culture. Mice were anesthetized in cold ice and decapitated. After removal of skull and meninges, cortex and hippocampus were extracted and dipped in HBSS in 15 ml conical tube. Tissue was mechanically dissociated through pipetting. Then, trypsin was added to tissue in 1:1 ratio (4 ml each) with HBSS and incubated in 37 °C water bath for 15 min. During incubation, conical tube was inverted every 3 min. Then, 2 ml of FBS was added and centrifugated at 700 rpm for 15 min. Supernatants were removed, and tissue was dissociated and cultured with primary culture medium (FBS 5 ml, Horse serum 5 ml, penicillin-streptomycin 500 μl, glutaMAX 500 μl, 25 ng m-CSF, DMEM F12 in 50 ml) in cell culture flask (SPL life science) with 16 ml of media in each flask. Media was changed 2~3 days after the incubation. Then, half of the media was changed every 2 days until harvest. Whole media was changed a day before harvest and primary microglia was harvested by shaking culture flasks in 260 rpm for 3h. Then, media was collected and centrifuged at 700 rpm for 15 min. Then, cell pellets were harvested and cultured on glass chip on 12~24 well plate and used for analysis [46].

Primary microglia and BV2 cells grown on glass chips in 12 or 24-well plate were used for phagocytosis assay. For live cell imaging, BV2 cells were seeded on confocal dish. Upon seeding microglial cells on glass chips, gossypetin (25 μM) or equal volume of DMSO was treated to each well and incubated for 24 h. Before the treatment of 488-Aβ, some wells were treated with cytochalasin D (20 μM) for 30 min before 488-Aβ treatment. After all the media were removed, cells were treated with media containing 24 h preincubated 488-Aβ (250 nM) for 1 h. For wells that were pre-treated with cytochalasin D, media with 488-Aβ were added with cytochalasin D. Cells were washed once with warm PBS and were fixed with 4% PFA for 40min at RT. After washing three times, cells were treated with 0.5% NP-40 in PBS for 30 min and washed once. Cells were incubated with blocking solution (5% FBS, 2.5% BSA in PBS) for 2 h at RT. Then, primary antibody (anti-Iba-1, 1:200, Wako #019-19741, RRID:AB_839504; anti-MHC class 2, 1:100, Thermo Fisher #14-5321-82, RRID:AB_467561) was treated and incubated at 4 °C for overnight or 4h at room temperature. Cells were washed three times and treated with secondary antibody at RT for 2h. After washing three times, cells were stained with Hoechst 33342 (10 μg/ml) diluted in PBS (1:2500) for 10 min and washed twice additionally. Cells were finally mounted with fluorescence mounting medium (Dako, S3023). For live cell imaging, after the media was exchanged with media containing preincubated 488-Aβ, live cell images were taken every 2 min for 1 h. Images were captured using Olympus confocal microscope (FV3000, RRID:SCR_017015) and analyzed by Image J (Fiji) software (RRID:SCR_002285). After selecting Iba-1 positive area with “create selection” tool, selected area was applied to 488-Aβ positive area to measure 488-Aβ that had been phagocytosed by microglia. After measuring the 488-Aβ positive area, it was divided by number of microglial cells. Number of microglial cells were measured by counting the nucleus stained by Hoechst. “Multi-point” tool was used to count the number of nuclei.

RNA was extracted from hippocampus and cortex with TRI-solution (Bio Science Technology, TS200-001). Seven hundred microliters of TRI solution was added to each tissue and homogenized with pestle. Then, each sample was vortexed, and 140 μl of chloroform was added and vortexed. The mixture was incubated at RT for 10 min and centrifuged at 15,000 rpm in 4 °C for 10min. Three hundred thirty microliters of supernatants were transferred to new tube and mixed with equal volume of isopropanol and vortexed and incubated for 10 min in ice. Samples were centrifuged at 15,000 rpm in 4 °C for 10 min. Then, RNA pellets were washed with 100 μl of 75% ethanol. Then, samples were centrifuged in 15,000 rpm at 4 °C for 2 min. Supernatants were carefully discarded, and RNA pellets are dissolved in 20~60 μl of DEPC water. Concentration of RNA was calculated with NanoDrop 2000 spectrophotometer (Thermo Scientific, RRID:SCR_018042) and processed into cDNA through RT-PCR using ImProm-II reverse transcription system (Promega). qRT-PCR was performed using FastStart Universal SYBR Green Master (Roche, 04913914001) using following primers: BACE1 forward: AGAGGCAGCTTTGTGGAGAT; BACE1 reverse: CTGGTAGTAGCGATGCAGGA; NCSTN forward: GACTACATTGGCAGCTCACG; NCSTN reverse: AGACATGGGATCTGTGTGCA; aph1 forward: TGACAGACCGATCAGATGCA; aph1 reverse: AAGCCCTCATCTGCCTTCTT; PSENEN forward: GAGAAGTTGAACCTGTGCCG; PSENEN reverse: ATCACCCAGAAGAGGAAGCC; GAPDH forward: AAATGGTGAAGGTCGGTGTG; GAPDH reverse: TGAAGGGGTCGTTGATGG. Relative expression was quantified using the ΔC_t method.

Hippocampus and cortex were dissected from mouse brain as mentioned above. Tissues were chopped on ice and washed with cold HABG medium (Hibernate A medium supplemented with B27 and glutamine) [47]. Tissues were incubated in papain dissociation buffer (Hibernate without calcium medium with 20U/ml papain, 1 mM L-cysteine, 0.5mM EDTA) at 30 °C for 30 min with mild shaking at 300 rpm. Tissues were washed with HABG medium and triturated using fire-polished Pasteur pipets. The cell suspension was filtered by 70 μm strainer. Debris were removed using Debris Removal Solution (130-109-398, Miltenyi Biotec), and red blood cells were lysed using lysis buffer (555899, BD Biosciences). To minimize the artificial transcription changes during tissue dissociation, actinomycin D was added during trituration (3 μM) and incubation with papain (45 μM) [48]. After staining with propidium iodide, live cells were sorted using FACS Aria III (BD biosciences, RRID:SCR_018091).

Libraries for Single cell RNA seq were generated using Chromium single cell 3′ library & Gel Bead Kit v3 (PN-1000092, 10X Genomics), Chromium Single Cell B Chip kit (PN-1000074, 10X Genomics). Briefly, cells were mixed with reverse transcription (RT) reaction reagent and loaded onto the B chip aiming for 6000 captured cells per a channel. After RT reaction, gel bead-in-emulsions were transferred to tubes and performed RT reaction using thermal cycler (C1000 Touch, Bio-Rad, RRID:SCR_019688). cDNA was purified using Dynabead MyOne SILANE (Thermofisher) and further amplified. Subsequent steps for library construction were performed by the manufacturer’s instruction provided. Quality of the amplified cDNA and final libraries were monitored by Bioanalyzer (Agilent, RRID:SCR_019715). Libraries were sequenced with a 2 × 100 bp paired-end protocol on a Novaseq S4 platform from Illumina to generate minimum 20,000 read pairs per cell.

Raw fastq files were processed using the Cell Ranger pipeline (v3.1.0, RRID:SCR_017344). The cDNA sequences were mapped to the mouse genome (GRCm38) using the STAR (v2.5.1b, RRID:SCR_004463) aligner with the GRCm38.97 GTF file [49]. A gene-by-cell count matrix was generated with default parameters. To remove the putative empty droplets during the single-cell capturing, we used EmptyDrops function of DropletUtils (v1.8.0) R package with FDR < 0.01 [50]. Low-quality cells with less than 3.0 total log10-scaled UMI count and with more than 25% of UMIs assigned to mitochondrial genes were excluded, where the thresholds were determined by visually inspecting outliers in the PCA plot on the quality control metrics using the calculateQCMetrics function of the scater (v1.16.1, RRID:SCR_015954) R package [51]. To remove cell-specific biases, cells were clustered using the quickCluster function of the scran (v1.16.0, RRID:SCR_016944) R package with default parameters and cell-specific size factors were computed using the computeSumFactors function of the same package [52]. The aggregated gene-by-cell count matrix across samples was normalized by dividing the raw UMI counts by cell-specific size factors. The normalized counts were then log2-transformed by adding a pseudo-count of 1. We defined highly variable genes (HVGs) with respect to biological variability using the decomposeVar and the getTopHVGs function of the scran package (RRID:SCR_016944), with the parameter of fdr.threshold < 0.05.

To remove undesired technical variables among the samples, we first split the aggregated scRNA-seq data by the sample ID. Then, each sample were normalized again using NormalizeData function of Seurat (v3.2.0, RRID:SCR_016341) R package, and 2000 HVGs for it were selected using FindVariableGenes function of the same package. The anchors for integrating the samples were identified on first 15 PCs of 2000 HVGs using FindIntegrationAnchors function of the same package; then, the samples were combined with these anchors using IntegrateData function. The k-nearest neighbor graph was computed with FindNeighbors function on first 15 PCs and used for computing clusters using FindClusters function with resolution = 0.1. The 15 PCs were used for the calculation of UMAP using RunUMAP function of Seurat R package. For the 13 identified clusters, cell types were annotated manually based on known marker genes.

Microglial populations were separated from the scRNA-seq data; then, we divided it by sampling region (cortex and hippocampus) of the data. After the removal of outlier cells, the single-cell transcriptome for each region were batch-corrected then re-grouped into N and M clusters respectively, on the 15 PCs of 2000 HVGs using the same method as above. The microglia subtypes were annotated based on their major markers, DEGs and cell cycle. Three cell types among them were further re-named as homeostatic microglia, stage 1 DAM, and stage 2 DAM as they were previously identified.

We identified differentially expressed genes (DEGs) between the 5xFAD and WT, and gossypetin and vehicle condition, respectively for each microglia type using the limma (v3.44.3, RRID:SCR_010943) R package with P < 0.05 and absolute value of log₂FC > 0.25 [53]. To specify enriched biological processes in DEGs between each condition for each cell type, significantly upregulated or downregulated GO biological process (GOBP) terms (P < 0.01) were selected using the topGO (v2.40.0, RRID:SCR_014798) R package with the org.Mm.eg.db (v3.11.4) annotation data package. Further GSEA were done for the calculation of normalized enrichment score using fgsea (v1.14.0) R package. DAM signature scores were calculated using AddModuleScore function of Seurat R package; then, the difference between 5xFAD/Vehicle and 5xFAD/Gossypetin condition within stage 1 DAM was tested with Mann-Whitney U test.

All statistical analyses were performed using GraphPad Prism version 9.0 (RRID:SCR_002798). The significance of differences was assessed by the unpaired t-test, one- or two-way ANOVA, followed by the Tukey’s multiple comparison tests. A p-value of p < 0.05 was considered to represent a significance. All data are presented as mean ± SD or SEM.

Approval of the study protocol was obtained from the Pohang University of Science and Technology Institutional Animal Care and Use Committee (POSTECH IACUC) (Approval ID: POSTECH-2021-0058). All animal experiments were carried out according to the provisions of the Animal Welfare Act, PHS Animal Welfare Policy, and the principles of the NIH Guide for the Care and Use of Laboratory Animals. All mice were maintained under conventional conditions at POSTECH animal facility under institutional guidelines.

Not applicable

All other data generated or analyzed during this study are included in this published article and its supplementary information files.

The authors declare competing financial interests; D.L. is employed by Novmetapharma.

The scRNA-seq datasets generated and analyzed during the current study are available in the National Center for Biotechnology Information under accession number PRJNA766517.