Opposing functions of β-arrestin 1 and 2 in Parkinson’s disease via microglia inflammation and Nprl3

Although β-arrestins (ARRBs) regulate diverse physiological and pathophysiological processes, their functions and regulation in Parkinson’s disease (PD) remain poorly defined. In this study, we show that the expression of β-arrestin 1 (ARRB1) and β-arrestin 2 (ARRB2) is reciprocally regulated in PD mouse models, particularly in microglia. ARRB1 ablation ameliorates, whereas ARRB2 knockout aggravates, the pathological features of PD, including dopaminergic neuron loss, neuroinflammation and microglia activation in vivo, and microglia-mediated neuron damage in vitro. We also demonstrate that ARRB1 and ARRB2 produce adverse effects on inflammation and activation of the inflammatory STAT1 and NF-κB pathways in primary cultures of microglia and macrophages and that two ARRBs competitively interact with the activated form of p65, a component of the NF-κB pathway. We further find that ARRB1 and ARRB2 differentially regulate the expression of nitrogen permease regulator-like 3 (Nprl3), a functionally poorly characterized protein, as revealed by RNA sequencing, and that in the gain- and loss-of-function studies, Nprl3 mediates the functions of both ARRBs in microglia inflammatory responses. Collectively, these data demonstrate that two closely related ARRBs exert opposite functions in microglia-mediated inflammation and the pathogenesis of PD which are mediated at least in part through Nprl3 and provide novel insights into the understanding of the functional divergence of ARRBs in PD.


Introduction
Parkinson's disease (PD) is the second most common neurodegenerative disorder after Alzheimer's disease (AD), and affects~2-3% of the world's population over the age of 65 [1,2]. Although the etiology and pathogenic mechanism of PD are not fully understood, a variety of genetic factors and environmental exposures have been identified to contribute to the pathological progression of PD, and the possible mechanisms include the changes in dopamine metabolism, mitochondrial dysfunction, endoplasmic reticulum stress, impaired autophagy, and deregulated immunity [3]. Chronic neuroinflammation in the substantia nigra pars compacta (SNc), the progressive loss of dopaminergic (DA) neurons in SNc, and the presence of Lewy bodies in different nuclei of the nervous system are the neuropathological hallmarks of PD [4].
Microglia are the main immune cells of the brain and their activation-mediated inflammatory processes in PD have been investigated extensively [5,6]. It is now increasingly apparent that sustained microglia activation is a major contributor to neuroinflammation and responsible for exacerbated neurodegeneration in PD [7,8]. One of the characteristics of activated microglia is the enhanced production of pro-inflammatory cytokines, such as tumor necrosis factor α (TNFα), interleukin (IL) 1β, IL-6, and interferon-γ (IFN-γ), which are toxic to DA neurons [9][10][11][12].
Activated microglia, as well as concentration of proinflammatory factors, are increased in the SNc of PD patients [13][14][15]. Therefore, identification of the regulators involved in microglia activation may create new avenues for drug design in the treatment of PD.
There is considerable evidence supporting the diverse roles of ARRBs in the pathogenesis of central nervous system diseases, such as ischemia, AD, and multiple sclerosis [29][30][31][32][33]. Several recent studies have also implied a role for ARRB2 in PD. For example, ARRB2 in microglia mediates the functions of dynorphin/κ-opioid receptors in controlling endotoxin-elicited pro-inflammatory responses and protecting tyrosine hydroxylase (TH) positive neurons from inflammation-induced toxicity [34]. ARRB2 is involved in the development of L-DOPA-induced dyskinesia [35]. Studies from our laboratory have shown that ARRB2 negatively regulates the assembly and activation of NODlike receptor protein-3 (NLRP3) via direct interaction in astrocytes, which contributes to the anti-neuroinflammatory effect of dopamine D2 receptors in PD [36]. However, noting is known about the functions of ARRB1 in the pathological progression of PD.
The purposes of this study are to investigate the possible functions of both ARRB1 and ARRB2 in the pathogenesis of PD and to elucidate the underlying mechanisms. We demonstrate that by virtue of their abilities to differentially regulate nitrogen permease regulator-like 3 (Nprl3) and the NF-κB and signal transducers and activators of transcription 1 (STAT1) pathways in microglia, ARRB1, and ARRB2 opposingly affect DA neuron degeneration and microglia inflammation both in vitro and in vivo. These data not only reveal novel functions and regulatory mechanisms of ARRBs in PD but also suggest a potential therapeutic approach for the disease.

Differential expression of ARRBs in PD mouse models
As an initial approach to define the functions of ARRBs in PD, we measured their expression in the midbrain of lipopolysaccharides (LPS)-and 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP)-induced PD mouse models [37]. ARRB1 and its mRNA were increased by~150%, whereas ARRB2 and its mRNA were decreased by more than 50%, in LPS-induced PD model (Fig. 1a-c). Such opposite regulation of ARRB1 and ARRB2 expression was also observed in MPTP-induced PD model (Fig. 1a-c). More interestingly, immunostaining showed that both ARRBs were highly expressed in ionized calcium binding adapter molecule 1 (Iba1) + microglia (Fig. 1d), but barely in Glial fibrillary acidic protein (GFAP) + astrocytes (Supplementary Fig. S1a) and neuronal nuclei (NeuN) + neurons ( Supplementary Fig. S1b) in the SNc of LPS-and MPTPinduced PD mice. Consistent with their expression in PD models in vivo, ARRB1 expression was markedly augmented, whereas ARRB2 expression was attenuated in primary cultures of microglia after LPS plus IFN-γ stimulation (Fig. 1e, f). These data suggest that the expression of ARRB1 and ARRB2, particularly in microglia, is differentially regulated in both inflammation-and toxin-induced mouse models of PD.

Effects of ARRB knockout on DA neuron loss, microglia activation and neuroinflammation in PD models in vivo
We then used Arrb1 −/− and Arrb2 −/− knockout mice to generate PD mouse models by LPS or MPTP challenge and studied DA neuron loss and microglia activation in the SNc. Ablation of ARRB1 or ARRB2 was confirmed by immunoblotting and one isoform knockout did not affect the expression of the other isoform ( Supplementary Fig. S2). As expected, LPS challenge caused remarkable DA neuron death and microglia activation as measured by staining with antibodies against TH and Iba-1, respectively, in wild-type (WT), Arrb1 −/− and Arrb2 −/− mice. However, LPS-induced neuron loss and microglia activation were significantly alleviated in Arrb1 −/− − mice, but exacerbated in Arrb2 −/− mice, as compared with those in WT mice ( Fig. 2a-h).
We next determined the effects of ARRB knockout on neuroinflammation by measuring the expression of inflammatory markers in the mouse midbrain. All proinflammatory markers tested, including Il6, Il1b, Tnf, and Nos2 genes and inducible nitric-oxide synthase (iNOS, encoded by Nos2), were significantly decreased, whereas anti-inflammatory markers, including Arg1, Ym-1 and Mrc1 genes and CD206 (encoded by Mrc1), were increased in Arrb1 −/− mice after LPS challenge, as compared with those in WT mice. In marked contrast, the pro-inflammatory markers were enhanced and the anti-inflammatory markers were reduced in Arrb2 −/− mice as compared with those in WT mice ( Fig. 2i and Supplementary Fig. S3a-d). Similar to the results observed in LPS-induced PD models, knockout of ARRB1 and ARRB2 produced opposite effects on DA neuron loss, microglia activation, and neuroinflammation in MPTP-induced PD mouse models in vivo (Supplementary Fig. S3e-q).
To define the role of microglial ARRBs in the PD mouse model, AAVs carrying the microglia-specific promoter F4/ 80 ( Supplementary Fig. S4a) were used to deliver siRNA to knock down ARRB1 or ARRB2 in microglia as confirmed by immunostaining ( Supplementary Fig. S4b). Similar to the results observed in Arrb1 −/− and Arrb2 −/− mice, MPTPinduced loss of DA neurons, activation of microglia and neuroinflammation all were significantly alleviated by AAV-mediated knockdown of microglial ARRB1, but exacerbated by knockdown of microglial ARRB2 in vivo ( Fig. 2j-r and Supplementary Fig. S3r-u). These results indicate that microglial ARRB1 and ARRB2 play opposite roles in PD mouse models.

Effects of ARRB depletion on microglia-induced DA neuron damage
To define if the effects of ARRBs on DA neuron loss were indeed caused by their actions on microglia activation as observed in the PD mouse models in vivo, we measured Fig. 1 Expression of ARRB1 and ARRB2 in PD mouse models. a ARRB1 and ARRB2 expression in the SNc of LPS-and MPTPinduced PD mouse models. b Quantitative data shown in a. c mRNA of ARRB1 and ARRB2 in the SNc of PD mice measured by RT-PCR. d Expression of ARRB1 and ARRB2 in microglia (Iba-1) in the SNc of PD mouse models. Similar results were obtained in three separate experiments. e Expression of ARRB1 and ARRB2 in primary microglia after LPS plus IFN-γ stimulation. f Quantitative data shown in e. Quantitative data are mean ± s.e. (n = 3). **P < 0.01 and ***P < 0.001. Scale bars, 50 µm. the effects of conditioned medium (CM) collected from microglia with or without LPS + IFN-γ treatment on DA neuron apoptosis, death and survival in vitro. The CM from microglia treated with LPS + IFN-γ strongly lowered the expression of anti-apoptotic Bcl-2, but elevated the expression of pro-apoptotic Bax in the neurons (Fig. 3a-d) and reduced the viability of the neurons (Fig. 3e, f). The neurons exhibited apoptotic features, including chromatin condensation and nuclear fragmentation ( Fig. 3g-j). The CM from microglia treated with LPS + IFN-γ also decreased the expression of TH (Fig. 3a-d) and shrunk the length of DA neuron axons ( Fig. 3k-n). All of these deleterious effects on the DA neurons were clearly mitigated by the CM from ARRB1 knockout microglia, but intensified by the CM from ARRB2 knockout microglia (Fig. 3). These results indicate that ARRB1 knockout can rescue, whereas ARRB2 depletion further amplify, the DA neuron damage induced by microglia inflammatory responses. j-q Immunohistochemistry (j, l, n, and p) and stereological counts (k, m, o, and q) of TH + DA neuron (j-m) and Iba-1 + microglia (n-q) in the SNc of MPTP-induced PD models after AAV injection (n = 5). Scale bars, 200 µm (upper panels) or 40 µm (lower panels) in j, l, n and p. r mRNA levels of pro-and anti-inflammatory markers in the midbrain of PD mice after AAV injection (n = 3). Quantitative data are mean ± s.e. *P < 0.05, **P < 0.01, and ***P < 0.001.

Functions of ARRBs in inflammatory response of primary microglia and macrophages
The gain-and loss-of-function approaches were used to further study the roles of ARRBs in microglia-mediated inflammation in response to LPS plus IFN-γ stimulation. In the gain-of-function studies, ARRB1 overexpression ( Supplementary Fig. S5a, b) significantly promoted, whereas ARRB2 overexpression ( Supplementary Fig. S5c, d) reduced, the expression of pro-inflammatory marker genes (TNF-α, IL-6, IL-1β, and iNOS) in microglia ( Fig. 4a, b). In the loss-of-function studies, ARRB1 knockout significantly inhibited (Fig. 4c), whereas ARRB2 knockout raised, the expression of the proinflammatory markers (Fig. 4d).
As microglia and macrophages have similar properties in mediating inflammation [38][39][40], bone marrow-derived macrophages (BMDMs) were used to confirm the functions of ARRBs in microglia-mediated inflammation. Similar to the results observed in microglia, ablation of ARRB1 markedly lowered the expression of proinflammatory marker genes and iNOS in BMDMs after LPS plus IFN-γ stimulation (Fig. 5a, c, d), whereas knockout of ARRB2 enhanced the expression of proinflammatory markers (Fig. 5b, e, f). Furthermore, ARRB1 knockout decreased, whereas ARRB2 knockout increased, the release of pro-inflammatory cytokines (IL-6, IL-1β, and TNF-α) as measured by ELISA (Fig. 5g, h). Immunofluorescent imaging showed that ARRB1 depletion attenuated the expression of CD16, a pro-inflammatory marker, in BMDMs upon stimulation with LPS plus IFN-γ ( Fig. 5i, j). In contrast, ARRB2 depletion enhanced CD16 expression (Fig. 5k, l) in BMDMs. These data demonstrate that ARRB1 and ARRB2 expression levels may directly control the inflammatory responses in a contrary manner in microglia and macrophages
ARRBs have been shown to interact with three molecules, IκBα, IKKα, and ΙΚΚβ [46][47][48] in the NF-κB pathway. To determine if ARRBs could bind other molecules in this pathway, we measured their interaction with p65. ARRB1 and ARRB2 were found to robustly interact with p65 in co-immunoprecipitation (co-IP) assays. More interestingly, both interactions fully depended on inflammatory stimulation (Fig. 6q, r). Furthermore, knockout of one ARRB isoform clearly potentiated the interaction of the other isoform with p65 ( Fig. 6s, t). These results suggest that two ARRBs physically associate with p65 likely in a competitive fashion.

Nprl3 as a novel effector of ARRBs in microglia
To elucidate the molecular mechanisms underlying the function of ARRBs, we measured the effects of their phosphorylation and ubiquitination, which are known to control their functions in GPCR trafficking and signaling [49][50][51][52], on microglial neuroinflammation, by measuring the production of inflammatory factors. Expression of ARRB1 phosphorylation mutants (S412A and S412D), ARRB2 phosphorylation mutants (S361A/T383A and S361D/T383D), ARRB1-Ub, and ARRB2-Ub produced similar effects on the expression of pro-inflammatory marker genes, as compared with their wild type counterparts ( Supplementary Fig. S6). These data suggest that phosphorylation and ubiquitination unlikely play a major role in ARRB-mediated inflammatory responses. To identify the effectors acting downstream of ARRBs, RNA sequencing (RNA-seq) was performed to compare genome-wide transcriptional profiles in microglia from WT or Arrb2 −/− mice in response to inflammatory stimulation ( Supplementary Fig. S7a). This strategy identified 130 genes upregulated and 56 genes downregulated in Arrb2 −/− mice, as compared with WT mice (Supplementary Fig. S7b). Analysis of the enriched biological processes showed that upregulated genes were related to positive regulation of immune responses, whereas downregulated genes associated with negative regulation (Supplementary Fig. S7b). Analysis of the enriched KEGG pathways also showed that upregulated genes were linked to inflammatory and immunological responses ( Supplementary Fig. S7c). Consistent with our results above (Fig. 4d), the RNA-seq data showed that the expression of pro-inflammatory genes, including Il1b, Tnf, Il6, and Nos2, was increased in Arrb2 −/− mice as compared with WT mice (Fig. 7a).
Il12rb1 and Lpar1 are receptors for IL-12 and lysophosphatidic acid (LPA), respectively, and both are well known to regulate inflammatory responses [53][54][55][56]. The functions of Nprl3, however, are poorly studied and thus, it was selected to be studied in microglia activation. As our data demonstrated that Nprl3 was downregulated in Arrb1 −/− mice and upregulated in Arrb2 −/− mice, we determined  Fig. S8a, b) enhanced the expression of pro-inflammatory marker genes (Il6, Il1b, Tnf, and Nos2), as well as the activation of p65 and STAT1 in ARRB1-knockout microglia, as compared with cells transfected with control vectors (Fig. 8a, c-f). siRNAmediated Nprl3 knockdown ( Supplementary Fig. S8c, d) inhibited the expression of pro-inflammatory marker genes and the activation of p65 and STAT1 in ARRB2-knockout microglia (Fig. 8b, g-j). These results suggest that Nprl3 is a novel effector, acting downstream of both ARRBs and mediating their functions in microglia inflammatory responses and activation of the NF-κB and STAT1 pathways.

Discussion
In this study, we have demonstrated that ARRB1 and ARRB2, two closely related ARRBs, display functional antagonism in the pathogenesis of PD (Fig. 8k) which is mediated through their distinct actions on microglia inflammatory responses. We first found that the expression of ARRB1 and ARRB2 was adversely regulated in the SNc and microglia of PD mouse models. We then used Arrb1 −/− and Arrb2 −/− mice and microglia-specific depletion of Fig. 6 Effects of knockout of ARRB1 or ARRB2 on activation of NF-κB, STAT1 and ERK1/2 pathways. Activation of IKKβ (a, b, e and f), p65 (c, d, g and h), STAT1 (i-l) and ERK1/2 (m-p) in ARRB1 or ARRB2 knockout BMDMs with or without LPS plus IFN-γ stimulation for 2 h. q-r Interaction of ARRB1 or ARRB2 with p65. BMDMs were stimulated with LPS plus IFN-γ for 1 h and the cell lysates were immunoprecipitated with antibodies against ARRB1 (q) or ARRB2 (r). s Effect of ARRB1 knockout on ARRB2 interaction with p65. t Effect of ARRB2 knockout on ARRB1 interaction with p65. Blots shown in m-p are representatives of three independent experiments. Quantitative data are mean ± s.e. (n = 3). *P < 0.05, **P < 0.01, and ***P < 0.001.
ARRBs to demonstrate that ARRB1 ablation significantly ameliorated, whereas knockout of ARRB2 exaggerated, DA neuron degeneration, microglia activation, and neuroinflammation in two PD mouse models in vivo. The opposing functions of ARRB1 and ARRB2 were also observed in microglia-mediated DA neuron damage, inflammation, and the activation of inflammatory signaling pathways in the gain-and loss-of-function studies using primary cell cultures in vitro. These data demonstrate that the expression of individual ARRBs as well as their expression ratio, specifically in microglia, is a crucial tipping point to control microglia inflammatory responses which in turn affect DA neuron degeneration and eventually the development of PD (Fig. 8k).
We have also identified that Nprl3 is a novel effector, acting downstream of ARRBs and mediating their opposite effects on microglia inflammation. Nprl3 is a component of the gap activity towards rags 1 complex and regulates mTOR complex 1 signaling; it is associated with the pathogenesis of epilepsy and cancer [57][58][59]. However, its physiological functions remain largely unknown. In the current study, Nprl3 was identified as an effector of ARRBs by RNA-seq analysis of genome-wide transcriptional profiles, which was further confirmed by its ability to control the functions of ARRBs in microglia-mediated inflammatory responses, including the expression of proinflammatory factors and activation of the NF-κB and STAT1 pathways. It is worth noting that ARRB1 and ARRB2 competitively interact with p65 and the interaction is completely dependent on inflammatory stimulation, suggestive of p65 activation-dependent interaction. Previous studies have revealed that the effects of ARRBs on the activation of the NF-κB pathway depend on stimuli and cell types studied [28,60] and the underlying mechanisms may involve direct interaction with IκBα, IKKα, and ΙΚΚβ [46][47][48]. Our data not only suggest a novel mechanism by which ARRBs activate the NF-κB pathway, but also imply distinct functions of ARRBs in PD being attributable to their different abilities to interact with p65.
Our data presented in this study have demonstrated that, to the best of our knowledge, ARRB1 and ARRB2 are the only and first pair of closely related isoforms which produce completely opposing functions in both in vivo and in vitro models of PD. Based on these data, we can conclude that the normal function of ARRB1 in microglia is stimulatory, whereas the normal function of ARRB2 is inhibitory, with   Activation of p65 (c, d, g and h) and STAT1 (e, f, i and j) in microglia after Nprl3 overexpression (c-f) or knockdown (g-j) as above. k A schematic diagram showing the opposite roles of ARRB1 and ARRB2 in microglia-mediated inflammation and DA neuron degeneration via regulating Nprl3 and the inflammatory STAT1 and NF-κB pathways (see text for details). +, stimulatory; −, inhibitory. Quantitative data are mean ± s.e. (n = 3). *P < 0.05, **P < 0.01, and ***P < 0.001. respective to the expression of Nprl3, the activation of the STAT1 and NFκB pathways, inflammatory responses, and pathologic progression of PD (Fig. 8k). As such, these data imply a potential therapeutic intervention, using genetic and/or pharmacological approaches to inhibit ARRB1 function and enhance ARRB2 function simultaneously, It is interesting to note that distinct expression and function of ARRB1 and ARRB2 have been observed in other animal and cellular settings [60,61]. For example, in an experimental autoimmune encephalomyelitis (EAE) mouse model, ARRB1 knockout alleviates disease phenotypes [32], whereas ARRB2 deficiency exacerbates EAE symptoms [33]. In a cellular model, ARRB1 downregulation increases, whereas ARRB2 depletion reduces, angiotensin II receptor-mediated activation of extracellular signal-regulated kinases 1 and 2 [62,63], which are contradictory to our results in BMDMs in response to inflammatory stimulation. Similar to the results observed in this study, both ARRB1 and ARRB2 interact with NLRP3 inflammasome, but they produce functionally contrary effects on the inflammasome activation [64,65]. These data, together with our current study, demonstrate the extreme complexity of ARRB-mediated functions under different physiological and pathological conditions.
In summary, this study reveals for the first time important, but opposite, roles played by ARRB1 and ARRB2 in the microglia-mediated inflammation and pathogenesis of PD. Our results provide novel insights into the understanding of the functional divergence of ARRBs in PD and may aid in the development of drugs for the treatment of PD.

Animals and PD models
Arrb1 −/− mice were purchased from the Jackson Laboratory (Las Vegas, NV) and Arrb2 −/− mice kindly provided by Dr. Gang Pei (Tongji University, Shanghai, China). C57BL/6J WT mice were from the Animal Core Faculty of Nanjing Medical University. All mice were allowed access to food and water ad libitum and maintained at 22-24°C with a 12 h light/dark cycle. 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 LPS-and MPTP-induced PD models were developed as described previously [36,66]. Briefly, mice (male, 3-4 months old) were randomly divided into groups (n = 11 for each genotype) and microinjected bilaterally with LPS (0.5 µg in 1 µl of saline, Sigma) at the SNc (AP: −3.0 mm; ML: ± 1.3 mm; DV: −4.2 mm) at a rate of 0.2 μl/min in a stereotaxic apparatus or administered with MPTP (20 mg/kg, i.p., Sigma) four times at 2-h intervals. After 7 days, the mice were sacrificed, and the brain tissues extracted and processed for immunoblotting, PT-PCR and immunohistochemistry.

Isolation and treatment of primary cells
Primary microglia were isolated from the brain tissues of neonatal mice (within 3 days after birth) by treatment with 0.25% trypsin/EDTA as described [67]. The cells were plated into PLL-coated T75 flasks and cultured in DMEM/ F-12 medium containing 1% penicillin/streptomycin and 10% FBS. The medium was changed every 3 days. After 10-14 days, the cells were split onto plates, incubated in serum-free base medium for 1 h, and treated with LPS (100 ng/ml) plus IFN-γ (20 ng/ml) at 37°C for 24 h. The CM was collected and centrifuged at 12,000 g for 10 min at 4°C and the supernatant was used to treat DA neurons.
BMDMs were isolated from the femur and tibia cavities of mice (male, 3 months old) and cultured in DMEM supplemented with 10% FBS, 1% streptomycin/penicillin, 1 mM sodium pyruvate, and 10 ng/ml GM-CSF as described [68]. The medium was changed every 3 days and the cells were used for further experiments after 7 days.

Cell transfection
In knockdown experiments, primary cells were transfected with individual siRNAs targeting Nprl3 (Supplementary  Table S1) using Lipofectamine RNAi MAX. In overexpression, phosphorylation and ubiquitination experiments, the cells were transfected with ARRB and Nprl3 plasmids using Lipofectamine 3000.

ELISA
The concentrations of cytokines, including TNF-α, IL-1β, and IL-6, in the supernatants of BMDMs were measured using ELISA kits, according to the manufacturer's instructions.

Quantitative RT-PCR
Total RNA extracted from mouse tissues or cells with TRIzol reagent was reverse transcribed with the TaKaRa Master Mix. The cDNAs obtained were mixed with Fas-tStart Universal SYBR Green Master and gene-specific primers (Supplementary Table S2) for RT-PCR in a Ste-pOnePlus instrument (Applied Biosystems, USA). GAPDH served as an internal control.

Immunohistochemistry
Frozen 30-μm-thick midbrain sections or primary DA neurons were deparaffinized by treatment with 3% H 2 O 2 for 30 min, blocked by 5% BSA plus 0.3% Triton X-100 for 60 min, and then incubated with anti-TH (1:1000) or anti-Iba-1(1:1000) antibodies at 4°C overnight as described [66]. After washing, the slides were incubated with secondary antibodies (1:1000) for 60 min. After staining with the DAB system and imaging in a stereomicroscope (Olympus, Japan), the axon lengths were measured and the positive cells in the SNc were stereologically counted by Microbrightfield Stereo-Investigator software (Microbrightfield, USA).

Cell viability
The viability of primary neurons was measured using CCK-8, according to the manufacturer's instructions.

Co-IP
BMDMs were lysed in RIPA buffer containing protease inhibitors as described [68]. The cell extracts were precleared with protein-A/G beads and then immunoprecipitated with 2 μg of antibodies against ARRB1 or ARRB2 at 4°C overnight followed by incubation with protein A/G beads at 4°C for 2.5 h. After washing for three times with RIPA buffer, the bound proteins were eluted and detected by immunoblotting.

RNA-seq
Total RNA was extracted from primary microglia with TRIzol reagent. RNA-seq libraries were prepared and sequenced on an Illumina HiSeq 4000 system by Novogene. Differential expression analyses were performed by DESeq.

Statistical analysis
Statistical tests were carried out using GraphPad Prism 7.0 software. Unpaired Student's t test was used for comparison between two groups. One-way analysis of variance (ANOVA) or two-way repeated-measures ANOVA were used to assess differences among multiple groups. Data are presented as the means ± s.e. Differences were considered statistically significant at P < 0.05.
Author contributions In this study, YF, GH, QJ, ML, and JD conceived and designed the experiments. YF, QJ, SL, HZ, RX, NS, XD, JL, MC, and MS carried out experiments. YF, QJ, GW, GH, and ML analyzed and interpreted the data and wrote the paper. GH critically reviewed and edited the work. All authors approved the final version of the manuscript.

Compliance with ethical standards
Ethics statement 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.

Conflict of interest
The authors declare no competing interests.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.