CD73-derived adenosine controls inflammation and neurodegeneration by modulating dopamine signalling

Abstract

Ectonucleotidase-mediated ATP catabolism provides a powerful mechanism to control the levels of extracellular adenosine. While increased adenosine A_2A receptor (A_2AR) signaling has been well-documented in both Parkinson’s disease models and patients, the source of this enhanced adenosine signalling remains unclear. Here, we show that the ecto-5′-nucleotidase (CD73)-mediated adenosine formation provides an important input to activate A_2AR, and upregulated CD73 and A_2AR in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced Parkinson’s disease models coordinatively contribute to the elevated adenosine signalling. Importantly, we demonstrate that CD73-derived adenosine-A_2AR signalling modulates microglial immunoresponses and morphological dynamics. CD73 inactivation significantly attenuated lipopolysaccharide-induced pro-inflammatory responses in microglia, but enhanced microglia process extension, movement and morphological transformation in the laser injury and acute MPTP-induced Parkinson’s disease models. Limiting CD73-derived adenosine substantially suppressed microglia-mediated neuroinflammation and improved the viability of dopaminergic neurons and motor behaviours in Parkinson’s disease models. Moreover, CD73 inactivation suppressed A_2AR induction and A_2AR-mediated pro-inflammatory responses, whereas replenishment of adenosine analogues restored these effects, suggesting that CD73 produces a self-regulating feed-forward adenosine formation to activate A_2AR and promote neuroinflammation. We further provide the first evidence that A_2A enhanced inflammation by antagonizing dopamine-mediated anti-inflammation, suggesting that the homeostatic balance between adenosine and dopamine signalling is key to microglia immunoresponses. Our study thus reveals a novel role for CD73-mediated nucleotide metabolism in regulating neuroinflammation and provides the proof-of-principle that targeting nucleotide metabolic pathways to limit adenosine production and neuroinflammation in Parkinson’s disease might be a promising therapeutic strategy.

CD73, adenosine A_2AR, microglia, neuroinflammation, Parkinson’s disease

Introduction

Microglia-mediated neuroinflammation is implicated in the progression and exacerbation of a wide range of neurodegenerative disorders (Frakes et al., 2014; Hong et al., 2016; Salter and Stevens, 2017; Joshi and Singh, 2018). Sustained microglia activation and prominent neuroinflammation are pathological hallmarks of multiple diseases including Alzheimer’s and Parkinson’s diseases (Kim and Joh, 2006; Wang et al., 2015; Li and Barres, 2018). Harnessing and targeting the neural-immune and neuroinflammation-related machinery to intervene disease progression has revealed beneficial effects in disease treatment (Adolfsson et al., 2012; Jucaite et al., 2015; Ohnishi et al., 2016); however, mechanisms underlying the endogenous regulation of neuroinflammation remain poorly understood.

Extracellular adenosine plays an important role in modulating microglia-mediated neuroinflammatory responses (Hasko et al., 2005). All four subtypes of adenosine receptors appear to regulate immunoresponses in microglia (van der Putten et al., 2009; Luongo et al., 2014; Merighi et al., 2015), with A_2A receptor (A_2AR)-mediated signalling being the main pathway involved (Rebola et al., 2011; Mills et al., 2012; Yao et al., 2012; Mohamed et al., 2016; Colella et al., 2018). A_2AR inactivation has been shown to enhance microglial inflammatory responses and exacerbate disease progression in experimental autoimmune encephalomyelitis (Mills et al., 2012; Yao et al., 2012). By contrast, other studies have shown that A_2AR antagonism prevents lipopolysaccharide (LPS)-induced microglia activation (Rebola et al., 2011) and reduces microglia-mediated pro-inflammatory profile in brain injury models (Mohamed et al., 2016; Colella et al., 2018). Thus, adenosine-A_2AR signalling regulates microglial function in a context-dependent manner (Dai et al., 2010; Ingwersen et al., 2016). In addition, although adenosine-A_2A signalling is recognized as an important modulator of microglia function (Farber and Kettenmann, 2006; Caetano et al., 2017), whether and how the production of adenosine is regulated to control A_2A activation and microglia-mediated neuroinflammation in neurodegenerative disorders remains to be clarified.

Parkinson’s disease is a common neurodegenerative disorder characterized by the progressive loss of dopaminergic neurons in substantia nigra pars compacta (SNc) and prominent neuroinflammation in the nigrostriatal pathway (Double, 2012; Koshimori et al., 2015; Kaur et al., 2017). Increased activity of A_2AR signalling in Parkinson’s disease (Calon et al., 2004; Morissette et al., 2006; Ramlackhansingh et al., 2011; Villar-Menendez et al., 2014; Hu et al., 2016), as well as the inverse relationship between the increased consumption of caffeine (an A_2AR antagonist) and the decreased onset of Parkinson’s disease has been well-documented (Xu et al., 2002; Boison, 2011; Wills et al., 2013; Madeira et al., 2017). A_2AR antagonism has been demonstrated to be beneficial in both Parkinson’s disease models (Chen et al., 2001; Ikeda et al., 2002; Yu et al., 2008; Wei et al., 2011b; Laurent et al., 2016) and preclinical trials (Hauser et al., 2011; Lopes et al., 2011; Hauser et al., 2014; Kondo et al., 2015) by improving motor behaviour deficits and attenuating dopaminergic neurodegeneration. A_2AR is known to coordinate motor control through a tight interaction with dopamine D₂ receptor (D₂R) on the striatopallidal neurons in the indirect pathway (Fuxe et al., 2005; Ferre et al., 2007; Azdad et al., 2009). Although studies have reported that glial cells are involved in the neuroprotective role of A_2AR antagonism in Parkinson’s disease (Yu et al., 2008; Gyoneva et al., 2014b; Boia et al., 2017), little is known about whether and how enhanced A_2AR signalling controls neuroinflammation to affect the disease progression and where the adenosine resource that leads to enhanced A_2AR signalling comes from in the disease.

Extracellular adenosine can arise from either the transmembrane transport of intracellular adenosine (Meghji et al., 1989; Parkinson et al., 2005; Wall and Dale, 2013) or the breakdown of extracellular ATP or ADP (James and Richardson, 1993; Zimmermann, 2006; Cunha, 2016). Increasing evidence suggests that enzymatic hydrolysis of ATP by two coupled ectonucleotidases, CD39 (encoded by the ectonucleoside triphosphate diphosphohydrolase 1, Entpd1) and CD73 (encoded by the ecto-5’-nucleotidase, Nt5e) serves as a main source for extracellular adenosine (Fredholm et al., 2007; Zimmermann et al., 2012; Matyash et al., 2017). We and others have shown that CD73 and A_2AR co-localize and physically interact on striatopallidal neurons (Augusto et al., 2013; Ena et al., 2013) and that CD73 provides extracellular adenosine to activate striatal A_2AR under physiological conditions (Augusto et al., 2013). Intriguingly, both CD73 (Schoen et al., 1992; Braun et al., 1997; Braun et al., 1998) and A_2AR are upregulated on microglia under disease states (Orr et al., 2009; Rebola et al., 2011; Gomes et al., 2013). Moreover, elevated adenine and dysfunction in nucleotide metabolism are associated with chronic inflammasome activation and low-grade inflammation in patients with cardiovascular diseases (Furman et al., 2017). These findings therefore prompted us to hypothesize that disease-sensitive CD73-produced adenosine may activate A_2AR on microglia to promote neuroinflammation and Parkinson disease progression.

In this study, we present the first evidence that CD73-derived adenosine activates A_2AR and enhances pro-inflammatory responses in microglia. Moreover, we show that upregulated CD73 and A_2AR in Parkinson’s disease models promote neuroinflammation detrimental to dopaminergic neurons. Knockout of CD73 reduces the adenosine production and plays a neuroprotective role in Parkinson’s disease model by attenuating neuroinflammation. Our data suggest that targeting nucleotide metabolic pathways to limit adenosine production and neuroinflammation is a promising therapeutic strategy for Parkinson’s disease.

Materials and methods

Animals and reagents

Nt5e^+/+ mice and Cx3cr1^GFP/+ mice were obtained from the Jackson Laboratory. All mice were housed and genotyped according to the standard protocol. All procedures were performed in accordance with guidelines approved by the Animal Advisory Committee of Zhejiang University. For two-photon analysis, mice were bred to generate Cx3cr1^GFP/+:Cd73^−/− and Cx3cr1^GFP/+:Cd73^+/+ offspring.

The following primary antibodies were used: chicken anti-TH (Abcam, cat# ab76442) rabbit anti-Iba1 (Wako, cat# 019–19741), mouse anti-GFAP (Millipore, MAB3402), and mouse anti-A_2AR (Millipore, cat# 05–717). The customized rabbit anti-CD73 polyclonal antibody was made by immunizing rabbits with synthesized peptide corresponding to the KDELLKHDSGDQDIS peptide (514–528 aa) of the mouse CD73 protein. All the secondary antibodies, donkey anti-chicken Cy3, donkey anti-rabbit 488, and goat anti-mouse Cy5, were from Jackson ImmunoResearch or Invitrogen. High performance liquid chromatography (HPLC) grade adenosine, erythro-9 −(2-hydroxy-3-nonyl) adenine hydrochloride (EHNA hydrochloride), 2′,5′-dideoxyadenosine (ddAdo), and potassium phosphate monobasic (KH₂PO₄) were from Sigma-Aldrich. RIPA buffer and Enhanced BCA protein assay were from Beyotime. Protease inhibitor cocktail were from Sigma-Aldrich and phosphatase inhibitor were from Pierce. ELISA kits were purchased from eBioscience. LPS, 5′-(N-ethylcarboxamido)adenosine (NECA), adenosine deaminase, (+)-butaclamol hydrochloride (butaclamol), (−)-quinpirole hydrochloride (quipirole), and S-(−)-eticlopride hydrochloride (eticlopride) were from Sigma-Aldrich. CGS 21680 (CGS), and dopamine HCl (DA) were from Selleck, and preladenant was from MedChemExpress. Acetonitrile (HPLC grade), ProLong® Gold antifade reagent, TRIzol® Reagent and SuperScript™ II Reverse Transcriptase were from Thermo Fisher Scientific. Primers were synthesized from Thermo Fisher Scientific and SYBR® Premix Ex Taq was from TaKaRa. Other reagents were obtained from Sigma-Aldrich if not indicated.

Acute brain slice preparation, LPS stimulation and cytokine measurement

Wild-type and Cd73^−/− male mice (4 weeks of age) were anaesthetized with 1% sodium pentobarbital and decapitated. The brains were quickly dissected into ice-cold slicing solution containing 2.5 mM KCl, 25 mM NaHCO₃, 1.2 mM NaH₂PO₄, 20 mM glucose, 120 mM N-methyl-d-glucamine (NMDG), 7 mM MgCl₂, 2.4 mM (+)-sodium l-ascorbate, 1.3 mM sodium pyruvate, 1.0 mM CaCl₂ (pH 7.35, 300 mOsm) continuously bubbled with 95% O₂/5% CO₂. Coronal forebrain slices (300-μm thick) were prepared using a Leica 1200 S vibratome. After slicing, the striatal slices were transferred to oxygenated artificial CSF containing 126 mM NaCl, 2.5 mM KCl, 2.0 mM CaCl₂, 2.0 mM MgCl, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, and 10 mM glucose (pH 7.3, 300 mOsm) at 32°C for 1 h.

Brain slices were then transferred into 12-well plates (two slices per well) containing 2 ml oxygenated artificial CSF, continuously bubbled with 95% O₂/5% CO₂ in the presence or absence of 20 µg/ml LPS with different receptor agonist or CD73 inhibitor at 32°C for 4 h. The slices were snap-frozen in liquid nitrogen and artificial CSF removed and stored at −80°C until use. Slices were homogenized on ice in RIPA Lysis buffer with 1 mg/ml of protease inhibitor cocktail and 10 mg/ml of protease and phosphatase inhibitor. The protein concentrations were determined using the Enhanced BCA protein assay. The levels of IL-1β, IL-6 and TNF-α were measured using different ELISA kits from eBioscience: 88–7013–22 for IL-1β, 88–7064–22 for IL-6, 88–7324–86 for TNF-α, according to the manufacturer’s instructions. The absorbance at 450 nm was determined with an iMark Microplate Reader (Bio-Rad). The concentrations of cytokines in each treatment were calculated based on the readings and normalized to the protein contents for each sample.

Acute and subacute MPTP treatment

For the acute model, wild-type and Cd73^−/− male mice (8–10 weeks of age, 22–25 g, n = 16 for each genotype) were randomly divided into four groups. The experimental group (12 mice for each genotype) were intraperitoneally injected with two doses of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP, Sigma-Aldrich) at 20 mg/kg at a 2 h interval, and sacrificed at 1 day, 3 days, and 7 days later. The control mice (four mice for each genotype) were injected with an equal volume of normal saline and were sacrificed 7 days later.

For the subacute model, wild-type and Cd73^−/− male mice (8–10 weeks of age, 22–25 g, n = 24 for wild-type and n = 16 for Cd73^{− /−}) were randomly divided into two groups. The experimental group was intraperitoneally injected with 25 mg/kg/day of MPTP for five consecutive days, whereas the control mice were injected with equal volume of saline. Mice were sacrificed 7 days following the last injection, upon completion of open field and rotarod tests.

Behavioural analysis

All behavioural tests were carried out and analysed by two investigators unware of the genotype and treatment of animals.

For open field tests, mice were placed in the middle of a novel open field arena (50 cm length × 40 cm width × 30 cm height) under normal light conditions. The distance was recorded during 30 min using Open Field Video Tracking System.

For rotarod tests, mice were first trained twice daily (10 min per session) for three consecutive days at the speed of 10 rpm before MPTP injection. Seven days after the last MPTP injection, mice were placed onto the rotarod and tested, with 3 min on–off the rotarod for five times daily to assess overall coordination and balance. The latency to fall was recorded at a speed of 45 rpm (accelerated from 0 rpm to 45 rpm over a 2-min period) using Rota-Rod test apparatus (UGO Basile).

Tissue and slice preparation

Mice were deeply anaesthetized with 1% sodium pentobarbital and perfused transcardially with saline. Brains were removed and cut into halves: one half was used for acute dissection of cerebral cortex, hippocampus and striatum, and snap-frozen in liquid nitrogen; the other half was fixed in 4% paraformaldehyde (PFA), transferred to 20% sucrose solution for 12 h, and 30% sucrose solution for 2 days, followed by coronal sectioning (30-μm thickness) using a cryostat (Leica Microsystems).

HPLC analysis of adenosine from mouse tissue

Snap-frozen brain tissues were minced in ice-cold mobile phase containing 98% (v/v) 40 mM KH₂PO₄ and 2% (v/v) acetonitrile in the presence of an adenosine deaminase inhibitors, 25 μM EHNA hydrochloride and an inhibitor of adenylyl cyclase, 2.5 μM 2′,5′-dideoxyadenosine (ddAdo), followed by sonication and centrifugation at 4°C, 13 700 rpm for 30 min. The supernatants were filtered using 0.2 μm Supor® membrane (PALL) and divided into two parts: 5 mg/ml adenosine was added to one part to indicate the peak location of adenosine and the other part was used to determine the adenosine content. The HPLC separation was carried out using an Ultimate AQ-C18, 5 μm, 4.6 × 250 mm column (Welch), with a 1 ml/min flow rate using the mobile phase described above, followed by UV detection at 260 nm.

Two-photon microscopy imaging and data processing

Anaesthetized Cx3cr1^GFP/+:Cd73^+/+ and Cx3cr1^GFP/+:Cd73^−/− mice (8–10 weeks) were perfused with ice-cold slicing artificial CSF containing (in mM): 110 choline chloride, 7 MgCl₂, 0.5 CaCl₂, 2.5 KCl, 25 NaHCO₃, 1.3 NaH₂PO₄, 20 glucose, saturated with 5% CO₂ and 95% O₂. Brains were acutely dissected and coronally sliced (300-μm thick) on a vibratome (MicromHM650V), then transferred to recording ACSF (32°C, 1 h) containing (in mM): 125 NaCl, 1.3 MgCl₂, 2 CaCl₂, 2.5 KCl, 25 NaHCO₃, 1.3 NaH₂PO₄, 10 glucose, saturated with 5% CO₂ and 95% O₂. Slices containing the striatum were used for live imaging on an Olympus two-photon microscope with a DeepSea Mai-Tai® Sapphire laser.

Microglia were imaged 50-μm deep into the slice surface to avoid activated cells in the mechanically damaged surfaces. Microglial process outgrowth and retraction under basal state were recorded for 10 min using a water-immersion objective (Olympus, 40 × , NA = 0.8). To trigger rapid microglia clustering, a laser bleach was applied to the slice as described by previous study (Davalos et al., 2005) and the responses of microglia toward the injury site were recorded with a z-stack step size of 1 μm (22 steps) at a resolution of 512 × 512 pixels for 30 min.

All z-stacks and time-lapse images were recreated to four dimensional representations using Imaris software (Bitplane AG, Switzerland). To track the speed of microglial processes towards the laser ablation site, objects greater than 1.5 μm in diameter were first picked up in the ‘spots’ module. Then microglial processes were detected and reconstructed into different spots. The ‘autoregressive motion’ algorithm was used to connect spots in different time frames into movement tracks. The maximal distance between spots was set up as 4 μm and the maximal gap size (connected frames between spots) was set up as 3. Finally, the track distance and duration of processes moving towards the injury sites were generated to calculate track speed.

To analyse the range and speed of microglial responses after laser injury, a classical method was adopted (Davalos et al., 2005) using the Imaris software. Initially, the laser ablation created a small injury sphere about 15 μm in diameter. Then, the neighbouring microglia extended processes towards the sphere and eventually occupied the space. Two areas were defined: area Y is the outer annular area, which is 70 μm in diameter (not including the injury sphere); and area X is the inner circular area of 35 μm in diameter with the injury sphere set as its centre. Microglial processes movement across these areas into the injury site was tracked across time. Spots 35–70 μm away from the injury centre were considered as located in area Y, accordingly spots that were less than 35 μm away from the injury centre were considered as located in area X. Thus, the increased number of spots in area X over time [R_X(t)] reflected responsive processes moving towards the injury sphere. To exclude the influence of cell body and nonresponsive processes, the number of spots within area X in the first image (taken immediately after laser ablation) was considered as background [R_X(0)]. To rule out the variability in the number of processes in different images, the number of spots within area Y in the first image was considered as cardinal number [R_Y(0)]. Therefore, the response of microglial processes at any time point is defined by the equation:

$$R\left(t\right)=\left[R_X\left(t\right)-R_X\left(0\right)\right]/\left[R_Y\left(0\right)\right]$$

(1)

All statistical analyses on microglial morphology and migration speed were carried out in a double-blinded manner, with the tester and analyser unaware of the genotype of the animals.

Cell culture

Primary mouse microglia were cultured as described previously (Wu et al., 2014), with minor modifications. Briefly, a mixed glial culture was obtained from the whole cerebrum of neonatal mice and maintained in minimum essential medium (MEM) containing 10% foetal bovine serum (FBS). The medium was completely replenished on the second day and half-changed on the fifth day. After 10–11 days of culture, mature microglia were identified as semi-adhesive cells with bright edges over the mixed glial culture. Mature microglia were collected by quick and gentle shaking, centrifuged, resuspended in MEM containing 3% FBS and transferred to 6-well cell culture plates (Corning Costar®). Microglia were allowed to attach to the plate for 2.5 h before further pharmaceutical treatment.

Immunofluorescence analysis

Brain slices were baked at 60°C for 1 h and rehydrated in Tris-buffered saline (TBS), followed by heat-induced antigen retrieval in 10 mM citrate buffer containing 0.1% Tween-20. Immunostaining was performed as previously described (Zhang et al., 2017). Slices were stained with anti-TH antibody (1:2000) and anti-Iba1 antibody (1:400) to label dopaminergic neurons and microglia, respectively. For CD73 immunostaining using the customized CD73 antibody, the antibody (1:50) was pre-incubated with Cd73^−/− brain slices to absorb the non-specific reactivity, followed by immunostaining using the regular procedure. Slides were mounted with ProLong® Gold antifade reagents. Images were acquired with an Olympus VS120 virtual slide microscope for TH-positive cell counting analysis, or with an Olympus FV1200 laser scanning confocal microscope for data presentation.

Sholl analysis

All z-stacks of Iba1-labelled microglia images were reconstructed to 3D representations using Imaris software. The ‘Filament’ module was applied to reconstruct the microglial framework. The ‘Autodepth’ algorithm was further used to completely and precisely restore microglial processes. The cell framework was outlined by setting the cell body as a central point and process terminals as the limit points. The spacing of Sholl analysis was set at 1 μm to indicate the extent of process branching. The total length of processes was also indicated. Quantitative analysis was performed within individual unit area (211.96 μm × 211.96 μm).

TH-positive neuron count

Images of TH-positive neurons for cell counting were acquired using an Olympus VS120 virtual slide microscope. Using ‘cell counter’ plugin of ImageJ Fiji software, cells were marked and then counted manually for subsequent statistics.

Quantitative RT-PCR

Total RNA from the mouse striatal tissue or primary microglia cultured was extracted using TRIzol® reagent and reverse-transcribed using SuperScript™ II Reverse Transcriptase according to the manufacturer’s instructions. Real-time PCR was performed using the SYBR® Premix Ex Taq™ on a LightCycler® 480 Instrument II Real-Time PCR Detection System (Roche). Primer sequences are provided in Supplementary Table 1. The relative expression was measured using the 2^−ΔΔCt method. Briefly, the threshold cycle (Ct) values of target genes were determined automatically by LightCycler 480 II software. ΔCt = Ct_{(Target genes)} − Ct_Actin. ΔΔCt = ΔCt_{(Target genes)} − ΔCt_{(average ΔCt of control)}. Relative fold change was determined by 2^−ΔΔCt (Livak and Schmittgen, 2001) and normalized to the expression level of actin.

Statistical analysis

All data were presented as mean ± standard error of the mean (SEM). All statistical analyses were performed with GraphPad Prism (version 6.01). Unpaired Student’s t-tests (two-tailed) were used to compare differences between two groups. One-way or two-way ANOVA and Tukey’s or Bonferroni’s tests were applied to assess differences among multiple groups. For all experiments, values of P < 0.05 was considered to be statistically significant. Not significant (ns), *P < 0.05, **P < 0.01 and ***P < 0.001.

Data availability

The data that support the findings of this study are available on request from the corresponding author.

Results

CD73 deficiency attenuated LPS-induced pro-inflammatory responses in microglia

To determine whether CD73 contributes to the regional adenosine production in the brain, we measured the adenosine concentration in different brain regions including the cerebral cortex, hippocampus and striatum from wild-type and Cd73^−/− mice by HPLC analysis (Fig. 1A). In accordance with previous reports (Birbeck and Mathews, 2013; Pani et al., 2014), the striatum contains the most abundant levels of adenosine, and CD73 deletion significantly reduced the adenosine contents in the striatum (P < 0.001) but not in the hippocampus and cortex, suggesting that CD73 is a key contributor to the high adenosine content within the striatum. These data are consistent with the recently-described peak activity of CD73 in the striatum (Augusto et al., 2013; Ena et al., 2013; Kulesskaya et al., 2013). Our enzyme histochemistry analysis (Zylka et al., 2008) also revealed prominent CD73-mediated AMP hydrolysis activity in the striatum and relatively lower activity in the SN (Supplementary Fig. 1). Analysis of additional regulators in adenosine metabolism in the wild-type and CD73 knockout brains found no apparent changes (Supplementary Fig. 2), verifying the importance of CD73 in the striatal adenosine production.

Figure 1

CD73 deficiency attenuates LPS-induced pro-inflammatory responses in microglia. (A) HPLC analysis of adenosine level in three brain regions (cortex, hippocampus and striatum) in wild-type and Cd73^−/− mice. Unpaired Student’s t-test, n = 3–6 mice. (B) A diagram for the acute preparation of striatal slices, followed by protein extraction and ELISA analysis. (C) Quantification of IL-1β, IL-6 and TNF-α from acute striatal slices from wild-type and Cd73^−/− mice in control conditions, treated with LPS (20 μg/ml) for 4 h in the absence or presence of CD73 inhibitor (100 μM AOPCP) or non-specific adenosine receptors agonist (100 μM NECA), n = 3 mice. (D) Primary cultured microglia derived from wild-type and Cd73^−/− mice were treated with 100 ng/ml LPS for 12 h, followed by quantitative PCR analysis of the pro-inflammatory cytokines and markers as indicated. (E) Primary cultured microglia derived from wild-type and Cd73^−/− mice were treated with NECA (30 μM) or adenosine deaminase (0.5 units/ml) for 1 h, followed by 100 ng/ml LPS treatment for 12 h, n = 3 independent cultures (D and E). AOPCP = adenosine 5′-(α,β-methylene)diphosphate. Two-way ANOVA with Bonferroni’s post hoc tests (C–E).

Previous studies on striatal adenosine signalling have primarily focused on its function in synaptic modulation (Schiffmann et al., 2007), little is known about how adenosine affects the activities of glial cells, in particular, microglia-mediated neuroinflammation. To test whether reduced extracellular adenosine levels affect neuroinflammation, acute striatal slices were prepared from wild-type and CD73^−/− mice and stimulated with LPS, followed by ELISA analysis of pro-inflammatory cytokines (Fig. 1B). While LPS stimulation significantly increased the levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor α (TNF-α) in wild-type slices, their induction was significantly reduced in CD73^−/− slices (Fig. 1C). In addition, the CD73 inhibitor AOPCP substantially reduced LPS-induced pro-inflammatory effects in wild-type slices, whereas the direct activation of adenosine receptors by NECA (non-hydrolysable and non-specific adenosine receptor agonist) significantly boosted the LPS-induced IL-1β and IL-6 production in Cd73^−/− slices (Fig. 1C), suggesting that CD73-derived endogenous adenosine enhances the synthesis of these pro-inflammatory factors.

Microglia cells are the major contributors in LPS-induced neuroinflammation in the brain. To verify that CD73-derived adenosine modulates microglia-mediated inflammatory responses, we treated primary cultured microglia derived from wild-type and Cd73^−/− mice with LPS and measured the levels of multiple pro-inflammatory cytokines. Consistent with the data observed in slices, LPS robustly enhanced the synthesis of IL-1β, iNOS, IL-6, and TNF-α in wild-type microglia but this effect was substantially attenuated in Cd73^−/− microglia (Fig. 1D). Addition of NECA significantly boosted the synthesis of pro-inflammatory cytokines (IL-1β, iNOS, IL-6) in wild-type microglia and partially recovered the levels of these pro-inflammatory cytokines in CD73^−/− microglia (Fig. 1E). By contrast, adenosine hydrolysis in the presence of adenosine deaminase attenuated the production of these cytokines in wild-type microglia (Fig. 1E). Notably, NECA robustly reduced the levels of TNF-α in both primary microglia culture and brain slices (Fig. 1C and E), in accordance with previous observations (van der Putten et al., 2009; Merighi et al., 2015; Newell et al., 2015). TNF-α usually rises at early phases of immunoresponses, while other cytokines such as IL-1β, iNOS increase at later stages. The differential effects of adenosine on different cytokines may partly reflect the temporal context-dependent nature of adenosine in regulating immunoresponses (Wei et al., 2011a).

CD73-derived adenosine signalling modulates microglial responses in an acute MPTP model

Microglia-induced neuroinflammation has been implicated in the progression of Parkinson’s disease. Since CD73 is distributed in the nigrostriatal pathway, a prominently affected area in the disease (Jackson-Lewis and Przedborski, 2007; Tristao et al., 2014), we tested whether CD73-derived adenosine affects microglia responses in an acute Parkinson’s disease model with reduced doses of MPTP administration (20 mg/kg, two doses) (Fig. 2A). This model allows us to track the dynamic changes of microglia in a time course with no significant dopaminergic neurons loss during the time window of investigation (Jakowec and Petzinger, 2004). In wild-type mice, acute MPTP treatment induced mild morphological changes of microglia at Day 1, but robust activation at Day 3, accompanied by prominent morphological changes of microglia in both caudate putamen and SNc, followed by a recovery at Day 7 (Fig. 2B and Supplementary Fig. 3A). By contrast, microglia in CD73 deficient mice exhibited a distinct response. Prominent microglial activation within striatum and SNc was readily observed at Day 1, while at Day 3, the morphological changes of microglia gradually recovered (Fig. 2B). By Day 7, microglia nearly converted back to the resting state as shown by Sholl and area under the curve (AUC) analyses of 3D constructed microglia morphology (Fig. 2C–E). Interestingly, AUC analysis also revealed significant differences in microglial morphology in the caudate putamen between wild-type and Cd73^−/− mice at resting state (Supplementary Fig. 3B and C), suggesting that endogenous adenosine may also control basal morphology of microglia. Together, these data suggest that CD73 deficiency modulates morphological changes during the activation process of microglia in the acute MPTP model of Parkinson’s disease.

Figure 2

CD73-derived adenosine modulates microglial responses in substantia nigra in an acute MPTP-induced Parkinson’s disease model. (A) A diagram for the acute Parkinson’s disease model with MPTP adminsitration. (B) Representative brain slices from the SNc region of wild-type and Cd73^−/− mice treated with saline or MPTP (20 mg/kg, twice every 2 h) were stained with anti-TH and anti-Iba1 antibodies to label dopaminergic neurons and microglia, respectively. Scale bar = 30 μm. (C) 3D reconstruction of microglia from wild-type and Cd73^−/− mice treated with saline or MPTP at Day 3 using Imaris v.7.2.3 software. Scale bar = 5 μm. (D) Quantification of microglial processes by Sholl and the total areas under curve (AUC) analyses (right). (E) Quantification of cell density per unit area, process terminals density per cell and total process length per cell. Two-way ANOVA with Bonferroni’s post hoc tests, n = 3–4 mice (total 278 cells).

CD73 regulates microglial motility and CD73 deficiency leads to enhanced microglia motility

The altered responses of microglia in the acute MPTP-model upon CD73 depletion may arise from enhanced motility of microglial processes. To test whether CD73-derived adenosine modulates microglia morphology and motility, we monitored the basal dynamics of microglial processes by two-photon microscopy live imaging of the striatal slices from Cx3cr1^GFP/+:Cd73^−/− or Cx3cr1^GFP/+:wild-type mice (Fig. 3A). Quantitative analysis revealed that CD73 deletion dramatically reduced the retraction rate of microglial processes (P < 0.05), but significantly increased the rate of microglial processes extension (P < 0.05), suggesting that CD73 deficiency promoted microglial process extension but suppressed process retraction (Fig. 3B). These data are in accordance with the differences observed in the above AUC analysis (Supplementary Fig. 3B) and previous findings showing the role of adenosine-A_2AR in mediating microglia process retraction (Orr et al., 2009; Gyoneva et al., 2014a). To assess whether CD73 deficiency affects injury-triggered directed movement of microglia, a laser injury was induced in the striatum (Fig. 3D) and the speed of migratory microglial processes towards the injury site was measured based on a 3D reconstruction of the migration trajectory (Fig. 3G). Quantification of either the mean speed (Fig. 3C) or the speed over a 30-min period (Fig. 3E and H) showed a faster migration of microglial processes in slices from Cd73^−/− mice. Meanwhile, whereas wild-type microglia maintained a constant response of movement towards the injury site, Cd73^−/− microglia showed an enhanced response 5 min after the injury and maintained at a faster speed until their processes reach the injury site (Fig. 3F and I). Together, these data suggest that CD73 deletion substantially promotes microglial motility in the striatum.

Figure 3

CD73 regulates striatal microglial motility and CD73 deficiency leads to enhance microglia motility. (A) Two-photon imaging of microglial dynamics during a 10-min time period under the basal state in striatal slices from Cx3cr1^GFP/+:wild-type and Cx3cr1^GFP/+:CD73^−/− mice. Scale bar = 30 μm. (B) Quantification analysis of microglia dynamics demonstrated increased rate of process extension but reduced rate of retraction in Cd73^−/− microglia. (D) Two-photon imaging of laser ablation (red circle)-induced chemotaxis of microglial processes towards the injury site in Cx3cr1^GFP/+:wild-type and Cx3cr1^GFP/+:Cd73^−/− striatal slices. Scale bar = 30 μm. (G) Reconstruction of microglial processes and moving tracks by polygonal lines using Imaris v7.2.3 software. Scale bar = 10 μm. (C) Quantification analysis of the mean track speed in Cd73^−/− microglia in comparison to wild-type. (E and H) Quantification of the track speed of microglial process toward the laser-induced injury site over a 30 min time period (H) and AUC analysis (E). (F and I) Quantification of microglial processes migrated into the laser-induced injury site over a 30 min time period (H) and AUC analyses (F). Unpaired Student’s t-test, n = 5 mice.

Restricting CD73-derived adenosine suppresses neuroinflammation and ameliorates neurotoxicity in the MPTP model

Since microglia in Cd73^−/− mice exhibited a reduced pro-inflammatory response, enhanced motility, and robust morphological changes, we next tested whether CD73 depletion attenuated neuroinflammation and neurotoxicity in a subacute MPTP model, which closely recapitulates pathogenesis of Parkinson’s disease with progressive loss of dopaminergic neurons and prominent neuroinflammation (Fig. 4A). As shown in Fig. 4B, 5 days of MPTP treatment induced robust microglial activation and clustering within the SNc in wild-type mice. The clustered microglia displayed a typical amoeboid morphology with short processes and large soma (Fig. 4B, top). By contrast, microglial accumulation in the SNc was substantially suppressed and microglia cells displayed a ramified morphology with long processes in MPTP-treated Cd73^−/− mice (Fig. 4C and D), consistent with the observations in the acute MPTP model (Fig. 2B–E). Measurement of pro-inflammatory factors in the striatum revealed that iNOS, IL-6, TNF-α, and Hif1α were significantly increased after MPTP treatment; however, their induction was substantially reduced in Cd73^−/− mice, suggesting a suppressed neuroinflammatory response in the absence of CD73 (Fig. 4E).

Figure 4

Restricting CD73-derived adenosine suppresses microglial-mediated neuroinflammation in a subacute MPTP model. (A) A paradigm for the subacute MPTP Parkinson’s disease model with MPTP administration. (B) Substantia nigral slices from wild-type and Cd73^−/− mice treated with saline or MPTP (25 mg/kg for 5 days) were stained with anti-TH and anti-Iba1 antibodies. Images in the right panels are enlarged images from the substantia nigra area in the left panels. Scale bar = 300 μm (left), 30 μm (right). (C) The cell density (left, n = 6 mice) and total process length per single cell (right, n = 4 mice) were quantified using Imaris v7.2.3 software. (D) Quantification of microglial processes by Sholl and AUC analyses, n = 4 mice. (E) Striatum was dissected from MPTP- or saline-treated wild-type or Cd73^−/− mice, followed by quantitative PCR analysis of the pro-inflammatory cytokines, n = 3 mice. Two-way ANOVA with Bonferroni’s post hoc tests.

To examine whether reduced neuroinflammation in the absence of CD73 played a neuroprotective role in the Parkinson’s disease model, we assessed the loss of dopamine neurons by immunofluorescence staining of TH in the SNc (Fig. 5A) and caudate putamen (Fig. 5B). Compared to the dramatic loss of dopamine neurons within the SNc in wild-type mice treated with MPTP, dopamine neuronal loss was significantly reduced in Cd73^−/− MPTP-treated mice (Fig. 5C and D). Similarly, the MPTP-induced loss of terminals of dopamine neurons in the caudate putamen was also markedly reduced in Cd73^−/− mice (Fig. 5D). Further, we found that CD73 deficiency significantly attenuated MPTP-induced deterioration of the spontaneous motor activity and coordinated movement of MPTP-treated mice in the open field and rotarod tests (Fig. 5E). Taken together, these data demonstrate that deficiency of CD73 partially prevents MPTP-induced DA neuronal loss and behavioural impairment.

Figure 5

Restricting CD73-derived adenosine ameliorates MPTP-induced neurotoxicity in a subacute MPTP model. (A) Dopaminergic neurons were stained with anti-TH antibody in SNc. (B) The number of TH-positive neurons on serial sections from the SNc (one side) was counted (left) and AUC were analysed (right), n = 5 mice. (C) Quantification of TH-positive neurons in the SNc (both sides), n = 5 mice. (D) The terminals of dopaminergic neurons were stained with anti-TH antibody in caudate putamen. Images in the lower panels are enlarged from the white boxes in the upper panels. Scale bar = 300 μm. (E) Quantification of the intensity of TH-positive terminals in caudate putamen, n = 4–6 mice. (F) Motor behavioural analysis of wild-type and Cd73^−/− mice in the open field test (OFT) and rotarod test, n = 8 mice; box plots indicate the median, 25th, and 75th percentiles; whiskers show the minimum and maximum. Two-way ANOVA with Bonferroni’s post hoc tests.

Adenosine-dependent A_2AR elevation promotes inflammatory responses

We have previously shown that CD73-produced adenosine is closely connected to the A_2AR activation (Augusto et al., 2013). In the subacute MPTP model, we found that the levels of both CD73 (Fig. 6A–C) and A_2AR (Fig. 6D) were increased, suggesting an increased adenosine production and A_2AR signalling in this disease model. Notably, MPTP only induced A_2AR elevation in wild-type but not Cd73^−/− mice (Fig. 6D). Previous studies have suggested that A_2AR level is significantly elevated in microglia in response to immune insults (Orr et al., 2009; Rebola et al., 2011; Gomes et al., 2013). To determine whether CD73-derived adenosine promotes the up-regulation of A_2AR in microglia, we examined the expression of A_2AR in both wild-type and Cd73^−/− microglia. Notably, LPS-induced elevation of A_2AR in microglia was significantly attenuated in the absence of CD73; whereas the addition of NECA significantly restored the levels of Adora2a (A_2AR) mRNA (Fig. 6E). Moreover, NLRP3 and HIF-1α, two effectors downstream of A_2AR (Ouyang et al., 2013), were also substantially reduced in the absence of CD73 (Fig. 6F). Furthermore, activation of adenosine A_2AR by CGS (CGS 21680, specific A_2AR agonist) selectively enhanced the synthesis of IL-1β in Cd73^−/− microglia (Fig. 6G), whereas blockade of A_2AR by preladenant (a specific A_2AR antagonist) reduced IL-1β production (Fig. 6G), suggesting that adenosine enhances inflammatory response via A_2AR-mediated signalling.

Figure 6

CD73-derived adenosine-dependent A_2AR elevation in microglia and MPTP models. (A) Quantitative PCR analysis of Cd73 mRNA levels in saline- or subacute MPTP-treated wild-type mice. Unpaired Student’s t-test, n = 3 mice. (B) Striatal slices from the saline- or subacute MPTP-treated wild-type mice were immunostained with anti-CD73 antibody. Scale bar = 300 μm. (C) Quantification of CD73 intensity in the striatum. Unpaired Student’s t-test, n = 5 mice. (D) Quantitative PCR analysis of A_2AR mRNA levels in saline- or subacute MPTP-treated wild-type and Cd73^−/− mice, n = 3 mice. (E–G) Primary cultured microglia derived from wild-type and Cd73^−/− mice were treated with different reagents or control for 1 h, followed by 100 ng/ml LPS treatment for 12 h, and quantitative PCR analysis. (E) Quantitative PCR analysis of A_2AR mRNA in microglia under control condition, or treated with control, LPS, or LPS+ NECA (30 μM). (F) Quantitative PCR analysis of the NLRP3-inflammasome related molecules and anti-inflammatory molecule arginase 1. (G) LPS-primed microglia were treated with CGS (specific A_2AR agonist, 20 μM) or preladenant (specific A_2AR antagonist, 10 μM). CGS = CGS 21680 HCl; Prel = preladenant. Two-way ANOVA with Bonferroni’s post hoc tests (D–G), n = 3 independent cultures (E–G).

CD73-derived adenosine promotes microglia inflammatory responses by inhibiting dopamine signalling

In exploring mechanisms underlying the modulation by clues for adenosine in inflammation, we tested whether the antagonistic interplay between adenosine and dopamine receptors affects inflammatory responses in microglia by treating primary cultured microglia with different dopamine and adenosine receptor inhibitors. First, we examined the expression of Adora2a (A_2AR) and Drd2 (D₂R) mRNA in microglia. In primary cultured microglia, the expression of both D₂R and A_2AR was low, with A_2AR expression being an order of magnitude lower than D₂R expression. In LPS-primed microglia, the expression of A_2AR s dramatically increased by approximately 50 times, whereas the levels of D₂R was slightly reduced (Fig. 7A). Blocking dopamine receptors by butaclamol (a non-selective dopamine receptors antagonist) substantially boosted IL-1β, iNOS, and IL-6 synthesis in LPS-primed wild-type microglia; however, this induction was robustly inhibited in the presence of adenosine deaminase or of selective A_2AR antagonist (preladenant) (Fig. 7B). Intriguingly, addition of dopamine (pan dopamine receptor agonist) or quinpirole (a selective dopamine D₂R agonist) only mildly reduced the synthesis of pro-inflammatory cytokines (Fig. 7C and D). Conversely, blocking dopamine receptors with butaclamol or D₂R receptors with eticlopride (a selective dopamine D₂R antagonist) markedly promoted the levels of pro-inflammatory cytokines IL-1β, iNOS and IL-6 (except TNFα) (Fig. 7C and D), suggesting that dopamine signaling is not limiting in suppressing inflammation in microglia. Notably, the pan-dopamine (butaclamol) or the D₂R antagonist (eticlopride)-induced pro-inflammatory cytokine induction was substantially attenuated in Cd73^−/− microglia (Fig. 7C and D), suggesting that dopamine and adenosine signalling may antagonize each other to modulate inflammatory responses in microglia.

Figure 7

CD73-derived adenosine promotes microglia inflammatory responses by inhibiting dopamine signalling. (A) Primary cultured microglia derived from wild-type mice were treated with 100 ng/ml LPS for 12 h, followed by quantitative PCR analysis of A_2AR and D₂R. (B) Quantitative PCR analysis of IL-1β, iNOS and IL-6 in LPS-primed microglia treated with butaclamol (non-specific dopamine receptors antagonist, 20 μM) in the presence of adenosine hydryolysis enzyme adenosine deaminase or A_2AR antagonist preladenant. One-way ANOVA with Tukey’s post hoc tests. (C) LPS-primed wild-type and Cd73^−/− microglia were treated with DA (non-specific dopamine receptors agonist, 200 μM) or butaclamol (20 μM) and analysed for different inflammatory cytokines. (D) LPS-primed wild-type and Cd73^−/− microglia were treated with quinpirole (specific dopamine D₂R agonist, 25 μM) or eticlopride (specific dopamine D₂R antagonist, 25 μM). Buta = (+)-butaclamol hydrochloride; DA = dopamine; Quin = (−)-quinpirole hydrochloride; Etic: S-(−)-eticlopride hydrochloride. Two-way ANOVA with Bonferroni’s post hoc tests (A, C and D), n = 3 independent cultures.

Discussion

Increased adenosine A_2AR signalling has been reported in Parkinson’s disease models and patients (Calon et al., 2004; Morissette et al., 2006; Ramlackhansingh et al., 2011; Villar-Menendez et al., 2014; Hu et al., 2016); however, the source of this enhanced adenosine A_2AR signalling has not been defined. The present study now shows that the ectonucleotidase CD73 provides a self-regulating feed-forward adenosine formation to activate striatal A_2AR. Furthermore, this study adds a new dimension to the mechanism of A_2AR-mediated control of Parkinson’s disease: while the development of A_2AR antagonists to manage Parkinson’s disease was prompted by the hypothesis that Parkinson’s disease involved an excessive A_2AR function over-riding D₂R function in controlling the activity of the indirect pathway (Svenningsson et al., 1999; Schiffmann et al., 2007), we now demonstrate that the over-function of A_2AR in Parkinson’s disease might also involve the control of microglia-mediated neuroinflammation. The present study further provides the first evidence to document an interaction between A_2AR and D₂R in the control of neuroinflammation. In fact, we demonstrate that limiting CD73-derived adenosine substantially attenuated microglia-mediated neuroinflammation in MPTP-induced Parkinson’s disease models, improving the viability of dopaminergic neurons and motor behaviours.

It is well established that the antagonism of A_2AR is beneficial in both Parkinson’s disease models (Chen et al., 2001; Ikeda et al., 2002; Yu et al., 2008; Wei et al., 2011b; ,Laurent et al., 2016) and clinical trials (Hauser et al., 2011; Lopes et al., 2011; Hauser et al., 2014; Kondo et al., 2015), in agreement with the inverse correlation between the incidence of Parkinson’s disease and the consumption of caffeine that antagonizes adenosine receptors (Xu et al., 2002; Boison, 2011; Wills et al., 2013; Madeira et al., 2017). We have recently demonstrated a physical and functional tight association between A_2AR and CD73 in the striatum (Augusto et al., 2013) and this is now extended to a model of brain disease involving a nigrostriatal dysfunction: we show that CD73 ablation phenocopies A_2AR antagonism to afford a robust neuroprotection in an MPTP model of Parkinson’s disease. This provides evidence that the known increased release of ATP as a danger signal in the brain (Rodrigues et al., 2015), namely in the injured nigrostriatal pathway (Melani et al., 2012), provides the source of adenosine responsible for the over-function of A_2AR that is involved in the pathophysiology of Parkinson’s disease. It should be noted that adenosine signalling can undergo biphasic responses in a temporal manner in diseases such as ischaemia and multiple sclerosis (Chen et al., 2007; Ingwersen et al., 2016). In the current study, we have mainly focused on the effects of CD73 inactivation in acute responses of microglia in the disease models. Assessing the long-term effects of the CD73-derived adenosine in chronic Parkinson’s disease models would be helpful to further validate the contribution of adenosine signalling in the disease.

Further, our study provides a novel mechanistic insight on the role of CD73-mediated formation of extracellular adenosine and over-activation of A_2AR in the emergence of nigrostriatal dysfunction in Parkinson’s disease. The development of A_2AR antagonists as anti-parkinsonian drugs was based on the known interaction of A_2AR and D₂R in the control of information processing in the indirect pathway of striatal circuits (Svenningsson et al., 1999; Schiffmann et al., 2007). However, different studies have pointed out that glial cells, in particular microglia, also play a critical role in Parkinson’s disease (Kim and Joh, 2006; Joers et al., 2017; Li and Barres, 2018; Joe et al., 2018; Yun et al., 2018). In agreement with the previous indications that the neuroprotection afforded by A_2AR antagonists might involve glia-mediated mechanisms (Yu et al., 2008; Gyoneva et al., 2014b; Boia et al., 2017), we now show that A_2AR antagonism plays a neuroprotective role, at least partially, by alleviating microglial A_2AR promoted neuroinflammation. In fact, we show that both CD73 and A_2AR are elevated in a subacute MPTP-induced Parkinson’s disease model and this upregulated CD73-adenosine-A_2AR signalling boosts pro-inflammasome-dependent cytokine production in microglia, which is deleterious to dopaminergic neurons. Blocking adenosine production suppressed neuroinflammation, improved motor behaviour and ameliorated the loss of dopaminergic neurons in a MPTP-induced Parkinson’s disease model. Together with the previously demonstrated increased release of ATP from ‘activated’ microglia (Imura et al., 2013; George et al., 2015), the localization and disease-prone upregulation of CD73 in microglia (Schoen et al., 1992; Braun et al., 1997, 1998) and the parallel upregulation of A_2AR upon injury suggest a potential role of microglia ATP-derived, CD73-mediated A_2AR over-activation in the development of neurodegenerative disorders (Madeira et al., 2014; Cunha, 2016) (Fig. 8).

Figure 8

Model depicting the role of CD73-produced adenosine formation and A_2AR signalling in microglia-mediated neuroinflammation and neuronal degeneration. Insults such as LPS or MPTP activate microglia and induce pro-inflammatory responses and A_2AR expression in microglia. CD73-derived adenosine further enhances microglia-mediated inflammation by activating A_2AR, inducing A_2AR expression and promoting the synthesis of pro-inflammatory cytokines that are detrimental to neurons. In the absence of CD73, the reduced production of adenosine decreases the activity of A_2AR, suppresses A_2AR induction and attenuates the synthesis of pro-inflammatory cytokines, thereby playing a neuroprotective role in disease models.

Amongst the different functions fulfilled by microglia, we found that depletion of CD73-derived adenosine signalling promoted striatal microglial process extension, but suppressed process retraction under basal state, in agreement with previous studies showing that adenosine A_2AR signalling mediates microglial processes retraction (Orr et al., 2009; Gyoneva et al., 2014b; Matyash et al., 2017). This suggests that microglia in a low-adenosine environment may have a quicker response to microenvironmental changes. This concept is supported by the enhanced microglial process extension towards the laser-induced injury site in CD73 deficient mice. Moreover, microglia in CD73 deficient mice exhibited a response faster than wild-type mice after acute MPTP treatment. It is likely that the inability to hydrolyse AMP into adenosine contributes to local AMP and ATP accumulation, which serves as a strong chemoattractant for microglial movement and promotes a quicker microglial response (Di Virgilio et al., 2009; Koizumi et al., 2013; Eyo et al., 2015). Alternatively, adenosine depletion may prevent A_2AR-mediated process retraction that also enhances microglia chemotactic responses (Santiago et al., 2014).

The present study also unravelled a hitherto unknown interaction between adenosine and dopamine signalling in the control of microglia and nigrostriatal neuroinflammation in an animal model of Parkinson’s disease. This extends to the realm of neuroinflammation the known balanced homeostasis between dopamine and adenosine signalling in the striatum that coordinates locomotion and motivational behaviour (Kim and Palmiter, 2008). In fact, in addition to their neuromodulatory roles, both signalling systems have been independently implicated in regulating immunoresponses (Sitkovsky et al., 2004; Torres-Rosas et al., 2014). Under resting state, microglia express moderate levels of D₂R (Huck et al., 2015), but barely detectable levels of A_2AR (Rebola et al., 2011). In response to LPS or MPTP stimulation, the density of A_2AR is dramatically increased (Rebola et al., 2011; Ahmad et al., 2013; Pedata et al., 2014; Merighi et al., 2015), and this dramatic upregulation of A_2AR likely switches the homeostatic balance towards the A_2AR-mediated pro-inflammatory responses. In parallel, the gradual loss of dopaminergic neurons in Parkinson’s disease substantially reduces the dopamine levels. Altogether, these opposite changes prompt a predominance of adenosine signalling bolstering inflammatory responses, as marked by the prominent microglia activation in the affected areas. Our data from primary microglia cultures showed that A_2AR antagonism or absence of CD73 ameliorate dopamine-associated inflammation, indicating that the maintenance of the homeostatic balance between the A_2AR and dopamine receptors, in particular D₂R, is critical for microglia-mediated immunoresponses. These data further provide proof-of-principle for the combinatorial application of both dopamine replenishment and A_2AR antagonism, which restores the balance and likely contributes to the alleviation of neuroinflammation and better preservation of neuronal circuits. On the other hand, the development of potent CD73 inhibitors, and the combined use of CD73 and A_2AR inhibitors may provide more pronounced effects to restore microglia function and alleviate brain diseases.

The present observation of a parallel upregulation of CD73 and of A_2AR in Parkinson’s disease model entails a novel key concept in the understanding of the adaptive changes of the adenosine modulation system in brain diseases. In fact, our present findings show that there seems to be a simultaneous need of a parallel upsurge of A_2AR and of ATP-derived extracellular adenosine to trigger the maladaptive neurochemical and behavioural changes characteristic of Parkinson’s disease. Our findings further indicate that blocking CD73-mediated formation of extracellular adenosine not only inhibited downstream A_2AR signalling but also prevented A_2AR elevation in microglia in response to LPS-stimulation or MPTP toxicity, suggesting an adenosine-dependent positive regulatory loop to upregulate A_2AR and maximize the downstream pro-inflammatory effects. This posits CD73 at the core of coordination of the alterations of adenosine signalling in brain diseases, which prompts the proposal that targeting CD73 to reduce adenosine availability provides a promising strategy to antagonize the A_2AR-promoted neuroinflammation and to enhance neuroprotection in the treatment of Parkinson’s disease.

Therefore, the novel findings that CD73-produced adenosine modulates microglial motility and neuroinflammation in Parkinson’s disease models, reveals the importance of nucleotide metabolism in the regulation of immune responses under different physiological and pathological conditions. Other than targeting the downstream adenosine receptor-mediated signalling, targeting the upstream nucleotide metabolic pathway that controls adenosine production to modulate neural-immune interactions and neuroinflammation-related machinery, represents a promising therapeutic strategy for intervening disease progression of Parkinson’s disease.

Abbreviations

Abbreviations
 
• A_2AR

  A_2A receptor

 
• D₂R

  dopamine D₂ receptor

 
• LPS

  lipopolysaccharide

 
• NECA

  5′-(N-ethylcarboxamido)adenosine

 
• SNc

  substantia nigra pars compacta

Acknowledgements

We are thankful to Dr. Rodrigo A. Cunha at the University of Coimbra for excellent suggestions and comments on our study and manuscript. We thank Eli York and Anne Liu at University of British Columbia for critical comments on the manuscript. We thank Drs. Roseline Godbout and Xiaoping Tong for helpful suggestions on the manuscript. We are grateful to Research Assistant Yanwei Li at the Core Facilities of Zhejiang University School of Medicine for assistance of the HPLC analysis, Dr. Xiaodong Wang, Dr. Sanhua Fang and Qiaoling Ding at the Core Facilities of Zhejiang University Institute of Neuroscience for technical assistance.

Funding

This work was supported by the National Key Research and Development Program of China (2016YFA0501000; 2017YFA0104200), the National Natural Science Foundation of China (31471308; 31490590; 31671057; 81521062), the Fundamental Research Funds for the Central Universities (2018FZA7004) and the Chinese Ministry of Education Project 111 Program (B13026).

Competing interests

The authors report no competing interests.