Fluoxetine inhibited the activation of A1 reactive astrocyte in a mouse model of major depressive disorder through astrocytic 5-HT_2BR/β-arrestin2 pathway

C57BL/6 J WT mice (male, 2 month old and neonatal mice) were purchased from the Animal Core Faculty of Nanjing Medical University. β-arrestin1 knockout (β-arrestin1^−/−) mice were purchased from the Jackson Laboratory (Las Vegas, NV, USA) and β-arrestin2 knockout (β-arrestin2^−/−) mice were obtained from Gang Pei’s laboratory (Tongji University, Shanghai, China). β-arrestin1^−/− and β-arrestin2^−/−mice were bred and maintained in the Animal Resource Centre of the Faculty of Medicine, Nanjing Medical University. If not specifically clarified, mice were allowed access to food and water ad libitum and maintained at 22–24 ℃. Light on/off cycles were controlled in a 12/12 h shift. All animal experiments were carried out in compliance with the ethical regulations and approved by Institutional Animal Care and Use Committee of the Nanjing Medical University Experimental Animal Department.

The AAV PHP.eB viruses expressing mouse 5-HT_2BR small interfering RNAs (siRNA, 5-HT_2BR siRNA1/2, Additional file 1: Table S1) (AAV-mHT_2BR, 1.76E + 13 v.g./mL) or Control siRNA (AAV-Ctr, 1E + 13v.g./mL) (GeneChem Co, China) under the astrocyte-special gfaABC1D promoter were microinjected bilaterally into the hippocampal CA1 region (AP: − 2.0 mm; ML:1.8 mm; DV:− 1.5 mm) of C57BL/6 J mice (male, 2 month old) at a rate of 0.25 μL/min for 4 min (total 1 μL AAVs) in a stereotaxic apparatus. After 4 weeks, the mice were subjected to CMS to develop a MDD mouse model and administrated with fluoxetine (FLX) described as followed.

The CMS model was developed as described previously [30, 32, 33]. Briefly, each mouse was caged singly and adapted to the environment for 3 days. Mice randomly divided into Control + saline, Control + FLX, CMS + saline, CMS + FLX groups (n = 8 mice for each group); AAV-Ctrl + Control + saline (n = 9), AAV-Ctrl + Control + FLX (n = 9), AAV-Ctrl + CMS + saline (n = 11), AAV-Ctrl + CMS + FLX (n = 12), AAV-mHT_2BR + Control + saline (n = 9), AAV-mHT_2BR + Control + FLX (n = 9), AAV-mHT_2BR + CMS + saline (n = 12), AAV-mHT_2BR + CMS + FLX (n = 12) groups; WT + Control + saline (n = 10), WT + Control + FLX (n = 10), WT + CMS + saline (n = 10), WT + CMS + FLX (n = 11), β-arrestin2^−/− + Control + saline (n = 10), β-arrestin2^−/− + Control + FLX (n = 9), β-arrestin2^−/− + CMS + saline (n = 10), β-arrestin2^−/− + CMS + FLX (n = 12) groups. Mice were exposed to 2–3 times randomly scheduled, mild stressors, which were not repeated for 3 consecutive days per day for 10 weeks. The mild stressors included cold water (4 °C for 5 min), wet bedding (12 h), the reversal of the light–dark cycle (24 h), empty cage (12 h), restraint (6 h), food and water deprivation (12 h), clipping tails (3 min), strobe light (12 h) and cage tilting (45 °C, 12 h). Between 6 and 10 weeks, mice was injected with normal saline or fluoxetine (10 mg/kg, i.p., Sigma Aldrich, MO, USA) once-daily.

Neonatal mice (within 3 days after birth) were used for the isolation of microglia as described elsewhere [42]. Briefly, after stripping the meninges and vessels, the brain tissues were treated with 0.25% trypsin/EDTA (ThermoFisher, USA) and dissociated into single cells. The cells were plated into cell adherent reagent (C1010, Applygen, Beijing, China) pre-coated T75 flasks and cultured in DMEM/F-12 (Gibco, USA), 10% fetal bovine serum (FBS, Gibco) and 1% streptomycin–penicillin mixture (Gibco). The mix glial cells were cultured for 10–14 d and the medium was changed every 3 d. Then flasks were shaken gently at room temperature (RT) to collect the microglia. After centrifugation at 500 g for 10 min, the cells were plated in 24-well plates, incubated in serum-free base medium for 1 h, and treated with lipopolysaccharide (LPS, 100 ng/mL, Sigma Aldrich) for 12 h. The microglia-conditioned medium (MCM) were collected and centrifuged at 12,000 g for 10 min at 4 °C and the supernatant was collected to treat astrocytes. About 98% of these cells were positive for ionized calcium binding adapter molecule 1 (Iba1), a marker for microglial cell types.

Primary astrocytes were isolated from WT, β-arrestin1^−/− and β-arrestin2^−/− neonatal mice (within 3 days after birth) by digested with 0.25% trypsin–EDTA for 5 min as described [43]. The cells were cultured in DMEM (Gibco) medium containing 10% FBS and 1% penicillin–streptomycin mixture. The medium was changed every 3 d. After 7 d, the cells were split onto plates and incubated in serum-free base medium for 1 h. About 98% of these cells were positive for glial fibrillary acidic protein (GFAP), a marker for astrocyte cell types. Cells were pre-treated with PBS or fluoxetine (10 μM) for 2 h as described previously [26]. Then, astrocytes were treated with following stimulators: A1 cocktail containing tumor necrosis factor (TNF)-α (30 ng/mL, 315-01A, Peprotech, USA), interleukin (IL)-1α (3 ng/mL, 211-11A, Peprotech) and C1q (400 ng/mL, 204,876, Millipore, USA), MCMs as mentioned above or IL-6 (100 ng/ml, CA92590, Chemicon, USA) for 24 h. 5-HT_2BR inhibitor RS-127445 (10 μM, s2698, Selleck) and Gq protein inhibitor YM-254890 (100 nM, QS 3666A, AdipoGen Life Sciences, USA,) were used to pretreat astrocytes for 30 min before fluoxetine administration according to the manufacturer’s instructions and previous report [44].

Fetuses (E15–16) from pregnant mice were used for the preparation of neurons as described previously [45]. Cortices and hippocampus were collected and dissociated with 0.125% trypsin/EDTA. Cells were plated on poly-L-lysine (PLL, 0.05 mg/ml, Sigma, St Louis, MO, USA) -coated 96-well plates at a density of 5 × 10⁵ cells/ml in DMEM/F-12 supplemented with 10% FBS and incubated at 37 °C in 5% CO2 for 6 h. Then, the media were removed and switched to neurobasal media (Gibco) with 2% B27 (Gibco) supplemented with 0.5 mM glutamine. Culture media were changed every 3 days. After 7 days, primary neurons were lysed. About 98% of these cells were positive for microtubule-associated protein 2 (MAP2), a marker for neurons.

siRNA targeting 5-HT_2BR (5-HT_2BR siRNA1) and a negative Control (NC) siRNA (GenePharma, Shanghai, China) (Additional file 1: Table S1) were transfected into astrocytes using LipofectamineTM RNAi MAX (Invitrogen, Life Technologies, CA, USA) in Opti-MEM reduced serum medium (Gibco) according to the manufacturer’s instructions for 8 h. The mixture was removed and the cells were cultured with normal medium for an additional 40 h (Additional file 2).

For immunohistochemistry (IHC) and immunofluorescence (IF), mice were anesthetized by an intraperitoneal injection of pentobarbital sodium (60 mg/kg, Sigma). After transcardial perfusion with cold PBS and 4% paraformaldehyde (PFA), the brains were harvested, post-fixed with 4% PFA overnight at 4 °C and dehydrated in 20 and 30% sucrose for 1 week. Then brain tissues were snap-frozen and cut by a Leica CM1860 cryostat (20 or 30 μm thick). The brain slices were stored at − 80 °C until use. For real-time reverse transcription polymerase chain reaction (RT-qPCR), western blotting and ELISA, mice were anesthetized and perfused with cold PBS, then the brains were harvested and hippocampus and cortex were collected in Eppendorf tubes and stored at − 80 °C until use. Blood samples were collected from eyeball and allowed to clot for 30 min at RT and overnight at 4 °C. To obtain the serum the blood was centrifuged at 800 g for 30 min at 4 °C and the supernatants were centrifuged at 5000 g for 10 min at 4 °C. The serum were stored at − 80 °C until use.

For immunohistochemical analysis [46], frozen 30 μm-thick hippocampal slices were deparaffinized with 3% H₂O₂ for 30 min. Nonspecific antibody binding was blocked by 5% bovine serum albumin (BSA, BIOFROXX, Guangzhou, China) in PBS supplemented with 0.3% Triton X-100 for 60 min. Subsequently, the sections were incubated with GFAP (1:1000, Abcam, MA, USA, ab7260) at 4 °C overnight. After washing with PBS, the slides were incubated with goat anti-rabbit antibody (1:1000, Invitrogen) for 60 min at RT. The GFAP positive cells in hippocampal were detected by the DAB staining system (Boster, China), observed under a stereomicroscope (Olympus, Tokyo, Japan) for imaging, and branch number of GFAP-positive cell were calculated by Image J software (National Institutes of Health, Bethesda, Maryland, USA).

IF staining was carried out as described previously [42]. Frozen 20 μm-thick brain sections were blocked in 5% BSA with 0.3% Triton X-100, and incubated with specific the primary antibodies, including mouse anti-GFAP (1:1000, Millipore, MAB360), rabbit anti-C3 (1:100, Abcam, ab11887), rabbit anti-Serping1 (1:500, Proteintech, IL, USA 12259–1-AP) or rabbit anti-5-HT_2BR (1:500, SAB Signalway, Nanjing, China, 32964) at 4 °C overnight. The slides were incubated with Alexa 594- or Alexa 555-conjugated secondary antibodies (1:500, Invitrogen) for 1 h and stained with DAPI (Invitrogen), and images were observed by confocal microscopy (Olympus) and analyzed by Image J software. The area of positive signal of GFAP, double positive signal for C3, or Serping1, and GFAP was measured using the interest grayscale threshold analysis with constant settings for minimum and maximum intensities for each staining marker [47‐49]. Three brain sections per mouse were used for quantification and the average of three sections was used to represent a data for each mouse.

Total RNA from mouse brain tissues or cultured astrocytes were extracted using Trizol reagent (Invitrogen), and reverse transcribed using HiScript Q RT SuperMix for qPCR Kit (Vazyme Biotech, Nanjing, China) according to the manufacturer’s instructions. The cDNA obtained was mixed with AceQ qPCR SYBR Green Master Mix (High ROX Premixed) (Vazyme Biotech) and gene-specific primers (Additional file 1: Table S1) for real-time PCR in a StepOnePlus instrument (Applied Biosystems, CA, USA). The cycle time values were standardized to GAPDH of the same sample. A relative quantity was calculated using the −ΔΔCT (showed in heatmap) and 2^−ΔΔCT (showed in bar chart) methods [17, 50].

Mouse tissues or primary cultured cells were lysed in RIPA buffer (50 mmol/L Tris.HCl (pH 7.5), 150 mmol/L NaCl, 10% Nonidet P-40 (NP-40), 0.5% Sodium Deoxycholate, 0.1% SDS, and 1 mmol/L PMSF (add before used)) supplemented with protease inhibitor cocktail (Roche, USA). The lysates were centrifuged at 16,000 g for 15 min at 4 °C, and the supernatants were used for immunoblot analysis after determining the protein concentrations by the BCA method (BCA Protein Quantitation Assay, Keygen Biotech, China). The extracts were resolved by SDS/PAGE on an 8–12% gradient separating gel and blotted on PVDF membranes (PALL Gelman, Ann Arbor, MI, USA) using Mini-PROTEAN® Tetra Handcast Systems and Mini Trans-Blot® Cell (Bio-Rad Laboratories, Hercules, CA, USA). Electrophoresis buffer (14.4 g/L Glycine, 3.0 g/L Tris.HCl and 1.0 /L SDS) and transfer buffer(2.93 g/L Glycine, 5.82 g /L Tris.HCl and 200 ml/L Methanol) were used. The membranes were blocked for 1 h at room temperature with 5% fat free milk in Tris-buffered saline Tween-20 (8.0 g/L NaCl, 2.4 g/L Tris base and 1 mL/L Tween-20) and then incubated with the following primary antibodies at 4 °C overnight: rabbit anti-C3 (1:500, ThermoFisher, PA5-21,349), rabbit anti-5-HT_2BR (1:1000, SAB Signalway, 32,964), rabbit anti-β-arrestin1 (1:1000, Santa Cruz, CA, USA, SC-53780), rabbit anti-β-arrestin2 (1:1000, Cell Signaling Technology, USA, 3857) and mouse anti-β-actin (1:5000, YI FEI XUE Biotechnology, China, YFPA1596). After the membranes were washed three times and incubated with the corresponding HRP-conjugated secondary antibody (1:5000, YI FEI XUE Biotechnology) for 1 h at RT, the bands were detected by enhanced chemiluminescence western blot detection reagents (Pierce) and analyzed with the ImageQuant™ LAS 4000 imaging system (GE Healthcare, USA). The band intensity was determined by densitometry with the aid of the Bio-Rad Gel Doc XR documentation system.

The concentrations of IL-6 in serum, cortex and hippocampus of mice were measured using commercially available ELISA kits (DY406, R&D, USA) according to the manufacturer’s instructions. Serum were diluted to 5 times using Reagent Diluent (1% BSA in PBS, pH 7.2–7.4, 0.2 µm filtered) and the extracts of cortex and hippocampus were diluted to 5 mg/mL using RIPA buffer. Standards were diluted using Reagent Diluent and the final concentrations are 1000, 500, 250, 125, 62.5, 31.25 and 15.625 pg/mL. Coat a 96-well microplate with 100 μL per well of the diluted Capture Antibody and seal the plate overnight at RT. Aspirate each well and wash with Wash Buffer (0.05% Tween® 20 in PBS, pH 7.2–7.4) for three times. After the last wash, remove any remaining Wash Buffer by aspirating or by inverting the plate and blotting it against clean paper towels. Block plates using Reagent Diluent and incubate for 1 h at RT. Repeat the aspiration/wash as above and the plates are ready for sample addition. Add sample or standards (100 μL per well), cover with an adhesive strip and incubate 2 h at RT. Repeat the aspiration/wash as above, and add Detection Antibody (100 μL per well), cover with an adhesive strip and incubate 2 h at RT. Repeat the aspiration/wash as above, and then add working dilution of Streptavidin–HRP (100 μL per well), cover the plate and incubate for 20 min at RT. Avoid placing the plate in direct light and repeat the aspiration/wash as above, and then add Substrate Solution (100 μL per well) and incubate for 20 min at RT. Add Stop Solution (10% H₂SO₄, 50 μL per well) and gently tap the plate to ensure thorough mixing. Determine the optical density of each well immediately, using a microplate reader (Thermo) set to 450 nm, with background correction read at 590 nm. A linear standard curve was created according the optical density (O.D.) values and standard concentration. The concentration of the samples was then calculated using the equation of the standard curve and taking into account the dilution of each sample.

Cell viability was determined by measuring the LDH release in the supernatants of astrocytes using a LDH assay kit (Jiancheng Bioengineering technology, Nanjing, China) according to the manufacturer’s instructions.

In this study, C57BL/6 J mice were subjected to the CMS for 6 weeks to establish a mouse model of MDD and followed by chronic fluoxetine treatment for 4 weeks (Additional file 1: Fig. S1A). After CMS stimulation for 6 weeks, the sucrose preference percentage markedly decreased (5–10 weeks: P < 0.001), which could be significantly improved by fluoxetine treatment for 3 and 4 weeks (Fig. 1A) (9–10 weeks: P < 0.001). Then, their behavioral changes were analyzed by TST, FST, OFT, NSFT and SIT. The immobility time of mice were significantly increased in TST and FST after CMS stimulation (Fig. 1B, C) (TST: P < 0.01 and FST: P < 0.05). In OFT experiments, the bouts and duration in center were reduced in mice that were exposed to CMS (Additional file 1: Fig. S1B–D) (bouts: P < 0.01 and duration: P < 0.05). Moreover, the latency time of mice to the first sniff in SIT and to the first feed > 3 s in NSFT were increased after CMS stimulation (Additional file 1: Fig. S1E, F) (P < 0.001). All these depressive-like and anxiety-like behaviors induced by CMS were clearly mitigated by chronic fluoxetine administration for 4 weeks (Fig. 1B, C, Additional file 1: Fig. S1B–F) (FST, duration and NSFT: P < 0.05; TST and bouts: P < 0.01; SIT: P < 0.001). Besides, IHC and IF results showed that fluoxetine treatment also reversed the reduction in the number and area of GFAP positive astrocytes in hippocampus of mice induced by CMS stimulation (Fig. 1D–G) (CON + saline vs CMS + saline P < 0.001; CMS + saline vs CMS + FLX P < 0.05 in IHC; P < 0.001 in IF). These data indicated that fluoxetine administration relieved the impairment of astrocyte in CMS mice. Next, we determined the effects of fluoxetine on the types of astrocyte by assessing mRNA levels of A1-specific and A2-specific markers. The transcripts of A1-specific markers (Serping1, Ligp1, Psmb8, Amigo2 and C3) were increased in hippocampus and cortex of CMS mice (hippocampus: P < 0.01 in C3; P < 0.001 in Serping1, Ligp1, Psmb8 and Amigo2; cortex: P < 0.001 in Serping1, Ligp1, Psmb8, Amigo2 and C3), which was reversed by fluoxetine treatment (Fig. 1H, Additional file 1: Fig. S2A) (hippocampus: P < 0.05 in C3; P < 0.001 in Serping1, Ligp1, Psmb8 and Amigo2; cortex: P < 0.001 in Serping1, Ligp1, Psmb8, Amigo2 and C3). However, CMS stimulation and fluoxetine administration failed to affect the levels of A2-specific transcripts (Clcf1, Ptx3, S100a10, Sphk1, Cd109, Ptgs2, Emp1, Slc10a6, Tm4sf1, B3gnt5 and Cd14) in hippocampus and cortex of mice (Fig. 1H, Additional file 1: Fig. S2A). As C3 is one of the characteristics of A1 astrocyte and the levels of C3 reflex the activation of A1 astrocyte[14], we also found that chronic fluoxetine treatment decreased the protein levels of C3 in hippocampus and cortex of CMS mice (Fig. 1I, J, Additional file 1: Fig. S2B, C) (CON + saline vs CMS + saline P < 0.01 in cortex; P < 0.001 in hippocampus; CMS + saline vs CMS + FLX P < 0.001). The enhancement of the percentage of A1 astrocytes, which are C3 and GFAP double-positive (C3⁺GFAP⁺) cells and Serping1 and GFAP double-positive (Serping1⁺GFAP⁺) cells in hippocampus of CMS mice were reduced by chronic fluoxetine treatment (Fig. 1K–N) (P < 0.001). These findings suggest that fluoxetine decreases A1 astrocyte reactivity in the CMS mouse model, which might contribute to the antidepressant effects of this drug.

The production of A1 astrocytes were induced by three cytokines together including IL-1α, TNF-α and C1q which were released by activated microglia [14]. To investigate the functions of fluoxetine in the transition of A1 astrocyte in vitro, primary astrocytes (Additional file 1: Fig. S3A) were cultured and treated with A1 cocktail (TNF-α, IL-1α and C1q) or activated (LPS) microglial-conditioned media (LPS-MCM) for 24 h. Compared to control group, fluoxetine treatment reduced the enhancements in the levels of A1-specific mRNA (H2–T23, Serping1, H2–D1, Ggta1, Ligp1, Gbp2, Fbln5, Ugt1a, Fkbp5, Psmb8, Srgn, Amigo2 and C3) induced by A1 cocktail and activated MCM in astrocytes (Fig. 2A, B) (CON + A1 vs CON + PBS P < 0.05 in Fkbp5, P < 0.001 in other markers; CON + A1 vs FLX + A1 P < 0.05 in Fkbp5, P < 0.01 in H2–T23, Amigo2 and C3, P < 0.001 in other markers; CON + activated (LPS) MCM vs CON + non-activated MCM P < 0.05 in Ugt1a, P < 0.01 in Ggta1 and Fkbp5; P < 0.001 in other markers; CON + activated (LPS) MCM vs FLX + activated (LPS) MCM P > 0.05 in Ugt1a, P < 0.01 in Serping1, Ligp1, Gbp2, Fbln5, P < 0.001 in other markers). While, A1 cocktail or LPS-MCM stimulation as well as fluoxetine administration had no effects on the expression of A2-specific transcripts in primary astrocytes (Fig. 2A, B). In addition, the protein levels of C3 were significantly enhanced in astrocytes after A1 cocktail or LPS-MCM stimulation, which were reversed by fluoxetine treatment (Fig. 2C–F) (CON + A1 vs CON + PBS P < 0.01; CON + A1 vs FLX + A1 P < 0.05; CON + activated (LPS) MCM vs CON + non-activated MCM, and CON + activated (LPS) MCM vs FLX + activated (LPS) MCM P < 0.001). Furthermore, LDH release in supernatant was measured and the results showed that fluoxetine decreased the death of astrocyte induced by A1 cocktail and LPS-MCM stimulation (Additional file 1: Fig. S3B, C) (CON + A1 vs CON + PBS, CON + A1 vs FLX + A1, and CON + activated (LPS) MCM vs CON + non-activated MCM P < 0.001; CON + activated (LPS) MCM vs FLX + activated (LPS) MCM P < 0.05). Meta-analysis of adult, adolescent and children studies found a positive association between depressive symptoms and IL-6, a pleiotropic cytokine with roles in immunity [51, 52]. We showed that the levels of IL-6 were significantly increased in serum, cortex and hippocampus of mice after CMS stimulation, which were attenuated by fluoxetine treatment (Additional file 1: Fig. S2D) (CON + saline vs CMS + saline P < 0.05 in hippocampus, P < 0.001 in serum and cortex; CMS + saline vs CMS + FLX P < 0.05 in hippocampus and cortex, P < 0.001 in serum). We also found that although the mRNA levels of A2-special markers did not change, A1-specific markers (Serping1, Ligp1, Psmb8, Srgn, Amigo2 and C3) were increased in astrocytes after IL-6 stimulation (Fig. 2G) (P < 0.05 in C3; P < 0.01 in Psmb8 and Amigo2; P < 0.001 in Serping1, Ligp1 and Srgn). Moreover, the expression of C3 protein and release of LDH were also enhanced in astrocytes after IL-6 stimulation (Fig. 2H–I, Fig. S3D) (P < 0.001). Fluoxetine treatment significantly blocked the activation of A1 astrocyte and release of LDH induced by IL-6 (Fig. 2G–I, Additional file 1: Fig. S3D) (P < 0.05 in C3, Srgn and Amigo2; P < 0.01 in LDH; P < 0.001 in Serping1, Ligp1 and Psmb8 and C3 protein). These data above indicate that fluoxetine inhibits the activation of A1 astrocyte under three different stimulation in vitro.

Previous study implied that fluoxetine has relatively high affinity with 5-HT_2BR in astrocytes [28]. In this study, we found higher concentrations of 5-HT_2BR proteins in primary cultured astrocytes than in primary cultured microglia and neurons (Fig. 3A, Additional file 1: Fig. S3A). Immunofluorescent imaging also showed that 5-HT_2BR was mainly colocalized with GFAP positive astrocytes in the hippocampus of mice (Fig. 3B). Therefore, we used 5-HT_2BR inhibitor RS-127445 to investigate whether 5-HT_2BR participate in the effects of fluoxetine on the A1 astrocyte reactivity in vitro. As shown in Fig. 3C, D, RS-127445 had no effect on the mRNA levels of A1-specific markers (Serping1, Ligp1, Psmb8, Srgn, Amigo2 and C3) in astrocytes after A1 cocktail or IL-6 stimulation, while blocked the inhibitory effects of fluoxetine on the transcripts of A1-specific markers (CON + A1 vs CON + PBS P < 0.001; CON + A1 + RS-127445 vs CON + PBS + RS-127445 P < 0.01 in Amigo2; P < 0.001 in other markers; CON + A1 vs FLX + A1 P < 0.01 in Amigo2; P < 0.001 in other markers; FLX + A1 vs FLX + A1 + RS-127445 P < 0.05 in Amigo2; P < 0.001 in other markers; CON + IL-6 vs CON + PBS P < 0.01 in Srgn and Amigo2; P < 0.001 in other markers; CON + IL-6 + RS-127445 vs CON + PBS + RS-127445 P < 0.01 in Srgn; P < 0.001 in other markers; CON + IL-6 vs FLX + IL-6 P < 0.05 in Amigo2; P < 0.01 in Srgn; P < 0.001 in other markers; FLX + IL-6 vs FLX + IL-6 + RS-127445 P < 0.01 in Amigo2; P < 0.001 in other markers). Next, the expression of 5-HT_2BR was reduced by siRNA (Additional file 1: Fig. S4) and the functions of fluoxetine in the A1 astrocyte reactivity were determined using RT-qPCR. Fluoxetine treatment decreased the activation of A1 astrocyte induced by IL-6 (P < 0.001), which was reversed by 5-HT_2BR knockdown in astrocytes (Fig. 3E) (P < 0.001). These results suggest that 5-HT_2BR contributes to the inhibitory roles of fluoxetine in the A1 astrocyte reactivity in vitro.

We further elucidated the function of astrocytic 5-HT_2BR in the effects of fluoxetine on A1 astrocyte reactivity in vivo. AAVs carrying the astrocyte-specific promoter gfaABC1D (Additional file 1: Fig. S5A) were used to knock down 5-HT_2BR in astrocyte of hippocampus as confirmed by western blot and immunostaining (Additional file 1: Fig. S5B, C). After AAVs injection in the hippocampus for 4 weeks, the mice were subjected to CMS for 6 weeks and subsequently administrated with fluoxetine for 4 weeks, and then behavioral tests were measured (Additional file 1: Fig. S5D). For depressive-like behaviors, the improved functions of fluoxetine in SPT,TST and FST of CMS mice were significantly reduced by AAV-mediated knockdown of astrocytic 5-HT_2BR (Fig. 4A, B) (AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CON + Saline, and AAV-mHT_2BR + CMS + Saline vs AAV-mHT_2BR + CON + Saline P < 0.01 in FST; P < 0.001 in SPT and TST; AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CMS + FLX P < 0.05 in FST; P < 0.01 in TST and P < 0.001 in SPT; AAV-mHT_2BR + CMS + FLX vs AAV-Ctrl + CMS + FLX P < 0.05 in TST and FST; P < 0.01 in SPT). For anxiety-like behaviors (OFT, SIT and NSFT), the attenuated roles of fluoxetine in CMS mice were also blocked by 5-HT_2BR knockdown in astrocytes (Additional file 1: Fig. S6A–E) (AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CON + Saline, and AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CMS + FLX P < 0.05 in NSFT; P < 0.001 in Bouts, Duration and SIT; AAV-mHT_2BR + CMS + Saline vs AAV-mHT_2BR + CON + Saline P < 0.001; AAV-mHT_2BR + CMS + FLX vs AAV-Ctrl + CMS + FLX P < 0.05 in Bouts; P < 0.01 in Duration and NSFT; P < 0.001 in SIT). Next, we determine A1 astrocyte reactivity in mice using immunostaining, western blot and RT-qPCR. As shown in Fig. 4C–F, the percentage of C3⁺GFAP⁺ A1 astrocytes and the protein levels of C3 in hippocampus of CMS mice were markedly decreased by fluoxetine treatment (AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CON + Saline P < 0.01 in western blot; P < 0.001 in IF; AAV-mHT_2BR + CMS + Saline vs AAV-mHT_2BR + CON + Saline, and AAV-Ctrl + CMS + Saline vs AAV-Ctrl + CMS + FLX P < 0.05 in western blot; P < 0.001 in IF), which were restored by astrocytic 5-HT_2BR knockdown (P < 0.05 in western blot; P < 0.001 in IF). In addition, the inhibitory effects of fluoxetine on the levels of A1-specific transcripts (Serping1, Ligp1, Psmb8, Amigo2 and C3) in the hippocampus and cortex of CMS mice were blocked by AAV-mediated astrocytic 5-HT_2BR knockdown (Fig. 4G, Additional file 1: Fig. S6F) (P < 0.001). These data indicate that astrocytic 5-HT_2BR is required for the effects of fluoxetine on behavioral impairments and A1 astrocyte reactivity in CMS mice.

Activation of 5-HT_2BR stimulates classical Gq protein pathways and β-arrestins-dependent signaling, leading to diverse functions [53, 54]. We explored the possible roles of these two signaling pathways in the effects of fluoxetine on A1 astrocyte reactivity. First, to measure the role of the classical G protein pathway, YM254890, an Gq protein inhibitor was used. As shown in Fig. 5, the enhancement in the mRNA levels of A1-special markers induced by either A1 cocktail or IL-6 stimulation were decreased by fluoxetine treatment, and YM254890 failed to affect the above effects of fluoxetine (CON + A1 vs CON + PBS, CON + A1 + YM254890 vs CON + PBS + YM254890, CON + A1 vs FLX + A1, and CON + A1 + YM254890 vs FLX + A1 + YM254890 P < 0.001; FLX + A1 + YM254890 vs FLX + A1: P > 0.05; CON + IL-6 vs CON + PBS P < 0.001; CON + IL-6 + YM254890 vs CON + PBS + YM254890 P < 0.05 in Amigo2; P < 0.01 in Psmb8; P < 0.001 in other markers; CON + IL-6 vs FLX + IL-6, and CON + IL-6 + YM254890 vs FLX + IL-6 + YM254890 P < 0.01 in Psmb8; P < 0.001 in other markers; FLX + IL-6 + YM254890 vs FLX + IL-6: P > 0.05). The data indicate that the classical G protein pathway isn’t the downstream signaling of astrocytic 5-HT_2BR regulated the inhibitory functions of fluoxetine in A1 astrocyte activation.

β-arrestins have two isoforms, β-arrestin1 and β-arrestin2, which regulate diverse physiological and pathophysiological processes [55]. We explored the roles of β-arrestin1 and β-arrestin2 in the inhibitory effects of fluoxetine on the activation of A1 reactive astrocyte by knocking out them separately. There were no changes in the mRNA expressions of A1 markers between WT and β-arrestin1-deficient (Additional file 1: Fig. S7A) astrocytes after fluoxetine treatment with or without A1 cocktail and IL-6 stimulation (Fig. 6) (WT CON + A1 vs WT CON + PBS, β-arrestin1^−/− CON + A1 vs β-arrestin1^−/− CON + PBS, and β-arrestin1^−/− FLX + A1 vs β-arrestin1^−/− CON + A1 P < 0.001; WT FLX + A1 vs WT CON + A1 P < 0.05 in Amigo2; P < 0.001 in other markers; β-arrestin1^−/− FLX + A1 vs WT FLX + A1 P > 0.05; WT CON + IL-6 vs WT CON + PBS, WT FLX + IL-6 vs WT CON + IL-6, and β-arrestin1^−/− FLX + IL-6 vs β-arrestin1^−/− CON + IL-6 P < 0.001; β-arrestin1^−/− CON + IL-6 vs β-arrestin1^−/− CON + PBS P < 0.05 in C3; P < 0.001 in other markers; β-arrestin1^−/− FLX + IL-6 vs WT FLX + IL-6 P > 0.05). Whereas, ablation of β-arrestin2 (Additional file 1: Fig. S7B) significantly abolished the ameliorative actions of fluoxetine on the transcript levels of A1 markers in astrocytes after A1 cocktail, IL-6 and activated MCM stimulation (Fig. 6, Additional file 1: Fig. S7C) (β-arrestin2^−/− CON + A1 vs β-arrestin2^−/− CON + PBS, β-arrestin2^−/− FLX + A1 vs WT FLX + A1, and β-arrestin2^−/− CON + IL-6 vs β-arrestin2^−/− CON + PBS P < 0.001; β-arrestin2^−/− FLX + IL-6 vs WT FLX + IL-6 P < 0.05 in Psmb8; P < 0.001 in other markers; WT CON + activated (LPS) MCM vs WT CON + non-activated MCM, and β-arrestin2^−/− CON + activated (LPS) MCM vs β-arrestin2^−/− CON + non-activated MCM P < 0.001; WT FLX + activated (LPS) MCM vs WT CON + activated (LPS) MCM P < 0.01 in C3; P < 0.001 in other markers; β-arrestin2^−/− FLX + activated (LPS) MCM vs WT FLX + activated (LPS) MCM P < 0.01 in Amigo2 and C3; P < 0.001 in other markers). These results suggest that β-arrestin2, rather than β-arrestin1 is involved in the inhibitory effects of fluoxetine on activation of A1 astrocyte in vitro.

Next, we further elucidated the role of β-arrestin2 in the effects of fluoxetine on activation of A1 astrocyte in MDD in vivo. Chronic fluoxetine treatment improved depressive-like behaviors, including decreased sucrose preference of SPT and increased immobility time of TST and FST in CMS mice, which were reversed by β-arrestin2 knockout (Fig. 7A, B) (WT CMS + Saline vs WT CON + Saline, β-arrestin2^−/− CMS + saline vs β-arrestin2^−/− CON + Saline, and WT CMS + FLX vs WT CMS + Saline P < 0.001; β-arrestin2^−/− CMS + FLX vs WT CMS + FLX P < 0.05 in TST and FST; P < 0.01 in SPT). In addition, β-arrestin2 deficient blocked the improved effects of fluoxetine on anxiety-like behaviors, including increased bouts and duration in center of OFT (Additional file 1: Fig. S8A–C) (WT CMS + Saline vs WT CON + Saline, and β-arrestin2^−/− CMS + saline vs β-arrestin2^−/− CON + Saline P < 0.001; WT CMS + FLX vs WT CMS + Saline, and β-arrestin2^−/− CMS + FLX vs WT CMS + FLX P < 0.05 in duration; P < 0.001 in bouts) and decreased latency time of mice to the first sniff in SIT and to the first feed > 3 s in NSFT after CMS stimulation (Additional file 1: Fig. S8D, E) (WT CMS + Saline vs WT CON + Saline, β-arrestin2^−/− CMS + saline vs β-arrestin2^−/− CON + Saline, and WT CMS + FLX vs WT CMS + Saline P < 0.001; β-arrestin2^−/− CMS + FLX vs WT CMS + FLX P < 0.01 in SIT; P < 0.001 in NSFT). We then assessed the activation of A1 astrocyte induced by CMS stimulation in WT and β-arrestin2 knockout mice with or without fluoxetine treatment. Immunostaining results showed that the proportion of A1 astrocytes in the hippocampus of CMS mice were reduced by chronic fluoxetine administration, which were abolished by β-arrestin2 knockout (Fig. 7C, D) (P < 0.001). Moreover, ablation of β-arrestin2 reversed the inhibitory roles of fluoxetine in the levels of C3 protein in hippocampus of mice after CMS stimulation (Fig. 7E, F) (P < 0.001). Furthermore, the ameliorative effects of fluoxetine on the transcript levels of A1 markers in hippocampus and cortex of CMS mice were blocked by β-arrestin2 deficient (Fig. 7G, Additional file 1: Fig. S8F) (P < 0.001). These data suggest that β-arrestin2 participates in the functions of fluoxetine in behavioral impairments and A1 astrocyte reactivity in CMS mouse model. Collectively, β-arrestin2 pathway is a downstream effector of 5-HT_2BR mediated the inhibitory effects of fluoxetine on A1 astrocyte reactivity in MDD in vitro and in vivo.