Fractalkine (CX3CL1) enhances hippocampal N-methyl-d-aspartate receptor (NMDAR) function via d-serine and adenosine receptor type A2 (A2AR) activity

Background N-Methyl-d-aspartate receptors (NMDARs) play fundamental roles in basic brain functions such as excitatory neurotransmission and learning and memory processes. Their function is largely regulated by factors released by glial cells, including the coagonist d-serine. We investigated whether the activation of microglial CX3CR1 induces the release of factors that modulate NMDAR functions. Methods We recorded the NMDAR component of the field excitatory postsynaptic potentials (NMDA-fEPSPs) elicited in the CA1 stratum radiatum of mouse hippocampal slices by Shaffer collateral stimulation and evaluated d-serine content in the extracellular medium of glial primary cultures by mass spectrometry analysis. Results We demonstrated that CX3CL1 increases NMDA-fEPSPs by a mechanism involving the activity of the adenosine receptor type A2 (A2AR) and the release of the NMDAR coagonist d-serine. Specifically (1) the selective A2AR blocker 7-(2-phenylethyl)-5-amino-2-(2-furyl)-pyrazolo-[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidine (SCH58261) and the genetic ablation of A2AR prevent CX3CL1 action while the A2AR agonist 5-(6-amino-2-(phenethylthio)-9H-purin-9-yl)-N-ethyl-3,4-dihydroxytetrahydrofuran-2-carboxamide (VT7) mimics CX3CL1 effect, and (2) the selective blocking of the NMDAR glycine (and d-serine) site by 5,7-dicholorokynurenic acid (DCKA), the enzymatic degradation of d-serine by d-amino acid oxidase (DAAO) and the saturation of the coagonist site by d-serine, all block the CX3CL1 effect. In addition, mass spectrometry analysis demonstrates that stimulation of microglia and astrocytes with CX3CL1 or VT7 increases d-serine release in the extracellular medium. Conclusions CX3CL1 transiently potentiates NMDAR function though mechanisms involving A2AR activity and the release of d-serine.

It is well established that NMDARs play an important role in a variety of physiological and pathological processes such as excitatory neurotransmission, synaptic plasticity and excitotoxicity [18,19]. NMDAR function is regulated by agents acting on a number of sites other than the Glu binding site [20], such as glycine [21] and D-serine that act with different affinities on the same binding site [22]. A recent report describes that D-serine preferentially acts on synaptic NMDARs, whereas glycine acts on extrasynaptic NMDARs [23]. In addition, D-serine is highly localized on glia in areas of the brain particularly enriched in NMDARs [24], thus suggesting that these cells are likely playing key roles in regulating NMDAR-dependent processes including synaptic transmission and synaptic plasticity. D-serine is degraded by D-amino acid oxidase (DAAO) and is mainly synthesized from L-serine by serine racemase (SR) present primarily on astrocytes. Recent studies also identified neurons and microglia as other important cellular sources of serine in the brain [25][26][27][28][29] even thought the role of neuronal and microglial D-serine remains to be determined.
Microglia have long been characterized by their innate immune activity in the nervous system, however recent evidence has shed new light on microglia function. These cells are now recognized as full time partners in neuronal function and the crosstalk between neurons and microglia has been described in several physiological and pathological conditions. Microglia secrete a number of factors, including Glu, nitric oxide, purines, trophic and anti-inflammatory substances [30], but also soluble factors affecting NMDAR and AMPAR such as glycine, cytokines, serine proteases and adenosine [31][32][33][34]. We have previously shown that stimulation of microglial cells with CX 3 CL1 induces the release of adenosine, which mediates the reduction of AMPAR function [13,[15][16][17], but scarce information is available on specific CX 3 CL1 effects on hippocampal NMDARs [35] that are highly involved in synaptic plasticity processes.
In the present work we demonstrate that CX 3 CL1 modulates NMDA-mediated synaptic transmission in the hippocampal CA1 region, through the activity of the A 2A R and the release of D-serine from glia.
For field recording experiments individual slices (350 μm thick) were then transferred to the interface slice-recording chamber (BSC1, Scientific System Design Inc., Mississauga, Ontario, Canada) with a total fluid dead space of approximately 3 ml. Slices were maintained at 30 to 32°C and constantly superfused at the rate of 2 ml/min. Solutions were applied to the slices by a peristaltic pump.
For patch-clamp experiments, individual slices (250 μm thick) were submerged in ACSF in the recording chamber at room temperature. The ACSF was perfused at a rate of approximately 1 ml/min.

Field potential recording
Slices were visualized with a Wild M3B microscope (Heerbrugg, Switzerland). At the beginning of each recording, a concentric bipolar stimulating electrode (SNE-100X 50 mm long Elektronik-Harvard Apparatus GmbH, Crisel Instruments, Rome Italy) was positioned in the stratum radiatum for stimulation of Schaffer collateral pathway projections to CA1. An ACSF-filled glass micropipette (0.5 to 1 MΩ) was positioned 200 to 600 μm from the stimulating electrode for recording the orthodromically-evoked NMDAR component of field excitatory postsynaptic potentials (NMDA-fEPSPs). Stimuli consisted of 100-μs-long constant current square pulses, applied at 0.05 Hz. The intensity of the stimulus was adjusted in each experiment to evoke approximately 50% of the maximal field potential amplitude without appreciable population spike contamination. Evoked responses were monitored online and stable baseline responses were recorded for at least 10 minutes. Only the slices that showed stable NMDA-fEPSP amplitudes were included in the experiments. To analyze the timecourse of the NMDA-fEPSPs slope, the recorded NMDA-fEPSPs were routinely averaged over 1 minute (N = 3). Experiments were performed in M-ACSF and averaged NMDA-fEPSPs (last 5 minutes of treatment or washout) were normalized to baseline values (5 minutes) prior to treatment. In the text, n refers to the number of slices analyzed/number of mice.
For the paired-pulse ratio (PPR) test, closely spaced consecutive stimuli (50 ms interval) were used, and PPR was calculated as the ratio between the NMDA-fEPSPs amplitude evoked by the second stimulus (A2) over the first (A1; A2/A1).
NMDA-fEPSPs were recorded and filtered (1 kHz) with an Axopatch 200 A amplifier (Axon Instruments, CA, USA) and digitized at 10 kHz with an A/D converter (Digidata 1322A, Axon Instruments). Data were stored on a computer using pClamp 9 software (Axon Instruments) and analyzed offline with the Clampfit 9 program (Axon Instruments). Membrane currents, recorded with a patch-clamp amplifier (Axopatch 200A; Molecular Devices, Foster City, CA, USA), were filtered at 2 kHz, digitized (10 kHz), and acquired with pClamp 9 software (Molecular Devices, Sunnyvale CA, USA). Excitatory postsynaptic currents (EPSCs) were evoked in CA1 pyramidal neurons by electrical stimulation with theta glass tubes pulled to a final tip diameter of 10 to 20 μm and filled with external solution. Stimulating electrodes were placed in the stratum radiatum to activate the Schaffer collateral pathway projecting to CA1. NMDA-EPSCs were evoked by stimulating at 5 to 50 V for 150 to 300 μs every 20 s. Recorded EPSCs were routinely averaged over 1 minute (N = 3) to analyze the timecourse of EPSC amplitude. Averaged NMDA-EPSCs (last 5 minutes of treatment or washout) were normalized to baseline values (5 minutes) prior to treatment. Stable EPSCs were monitored for at least 10 minutes before applying CX 3 CL1 and successively monitored during slice exposure to CX 3 CL1 (5 nM, a concentration close to the IC50 value for synaptic depression of AMPA mediated responses and to the concentration that inhibits NMDA-dependent LTP [15][16][17]. Data were analyzed offline with Clampfit 9 (Axon Instruments).

Glia and neuron primary cultures
Primary cortical glial cells were prepared from 0 to 2-day-old mice. Cerebral cortices were chopped and digested in 30 U/ml papain for 40 minutes at 37°C followed by gentle trituration. The dissociated cells were washed, suspended in Dulbecco's modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) (Invitrogen, Life Technologies, Monza, Italy) and 2 mM L-glutamine and plated at a density of 9 to 10 × 10 5 in 175 cm 2 cell culture flasks. At confluence (11 days in vitro), glial cells were plated (5.0 × 10 5 cells/cm 2 ) on dishes coated with poly-Llysine (100 μg/ml) in DMEM supplemented with 10% FBS, 100 U/ml penicillin, 0.1 mg/ml streptomycin. To detach and collect microglial cells, glial cells at confluence (11 days in vitro) were shaken for 2 h at 37°C. These procedures gave almost pure microglia (<2% astrocyte contamination) and astrocytes cell population (4% to 6% of microglia contamination), as verified by staining for glial fibrillary acidic protein (GFAP) and isolectin IB 4 .

Measurement of D-serine by liquid chromatography and mass spectrometry (MS) analysis
Mixed glia, microglia and neuronal cultures at 11 days old were pretreated in M-ACSF for 40 minutes and stimulated for either 20 minutes, 2 h and 20 minutes and 4 h with CX 3 CL1 10 nM or vehicle. After this time, media were collected, centrifuged at 4°C for 10 minutes at high speed, and the resulting supernatants were kept to −20°C until use. Cells remaining in the dish were analyzed for protein content with a BCA assay (Pierce, Euroclone Pero, Italy).
Aliquots of the culture medium was derivatized according to the method described by Berna and Ackermann [40]. Briefly, samples (25 μl) were incubated with 200 mM (pH 7.5) sodium bicarbonate solution and 10 μl of 1% Marfey's reagent (w/v in acetone) at 60°C for 1h.
Next, the samples were evaporated to dryness at 60°C under a stream of nitrogen, and the dried residues were reconstituted with 12.5% methanol in 15 mM ammonium acetate. The samples were centrifuged at 600 g for 5 minutes and the supernatants diluted with 200 μl of mobile phase. Injections of 20 μl were performed for liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. Marfey's adducts of D standard mixture were prepared by appropriate dilution of the 100 μg/ml stock solution of D-serine in highperformance liquid chromatography (HPLC)-grade water and following the same derivatization procedure.
The LC-MS/MS system consisted of a PerkinElmer 200 Series binary pump and autosampler (PerkinElmer, Norwalk, CT, USA) and an SCIEX API2000MS/MS triple quadrupole mass spectrometer including a Turbolon Spray W ionization source (Applied Biosystem-MDS SCIEX, Thornhill, Ontario, Canada).
Chromatographic separation was performed on a Synergi Polar RP column (150 × 2.0 mm, 4 μm), protected by a guard column with identical packing material (4 × 2.0 mm; Phenomenex, Torrance, CA, USA) The mobile phase consisted of a linear gradient (30% to 70% with respect to acetonitrile) formed by combination of 5 mM ammonium formate buffer in water (pH 4.0, eluent A) and acetonitrile (eluent B) at a flow rate of 0.2 ml/min. D-Serine derivatives were detected using multiple reaction monitoring (MRM) in negative ionization mode. Selected ion masses of the protonated precursors and fragmented ions (m/z) were 359.1 > 192. Chromatographic peaks were integrated using Analyst™ software (version 1.4.1, SCIEX). The detection limits (LOD) and quantification limits (LOQ) for D-serine were 1.71 ng/ml and 5.18 ng/ml respectively. The Marfey's derivated isobaric compounds L-Ser and D-Ser (2,4-dinitro-phenyl-5-Lalanine amide-(DNPA-)L-Ser and DNPA-D-Ser) were well resolved chromatographically as shown in Additional file 1: Figure S1. An example of the blank samples (control M-ACSF) is shown in Additional file 2: Figure S2, where no signal corresponding to DNPA-L-Ser or DNPA-D-Ser are present.

Statistical analysis
The values were reported as mean ± SEM. Unless otherwise specified, data values refer to number of slices/ number of animals analyzed. Where appropriate, t tests or one-way analysis of variance (ANOVA) were used. Since in all experiments data passed the normality and the equal variance tests, we performed parametric ANOVA followed by Tukey or Holm-Sidak multiplecomparison test (Sigma Plot, 11.0 software). If not specified, P values refer to ANOVA post hoc analyses. . Adenosine deaminase (ADA, Roche, Germany). Acetonitrile, methanol, both gradient grades were purchased from Merck (Darmstadt, Germany). Ultrapure water was prepared using a Millu-Q system (Millipore, MA, USA).
Drugs were dissolved in ACSF just before application.

CX 3 CL1 increases hippocampal NMDA responses
To investigate whether CX 3 CL1 influences NMDAmediated synaptic transmission in hippocampal CA1 region, we recorded the NMDAR component of the fEPSPs (NMDA-fEPSPs) elicited in the CA1 stratum radiatum of mouse hippocampal slices by electrical stimulation of the Schaffer collateral (at 0.05 Hz). To isolate the NMDA component of the fEPSPs slices were treated with the AMPA receptor blocker NBQX (10 μM) and with low (0.2 mM) magnesium [41]. In control slices, synaptic activity was stable for at least 30 minutes (not shown). We challenged the slices with CX 3 CL1 at concentrations of 5 nM. In the majority of the experiments (25 out of 35 slices tested, 28 mice; Figure 1A) bath application of CX 3 CL1 for 20 minutes induced a short-term potentiation (STP) of synaptic transmission. The potentiation of the NMDA-fEPSPs slope developed within 5 minutes of CX 3 CL1 application was 1.22 ± 0.03 of baseline (from 1.92 to 1.01 in different experiments, measured here and thereafter after 20 minutes of treatment, P <0.05) and was followed by partial recovery (to 1.15 ± 0.01 of baseline, P <0.05, measured here and thereafter at 20 minutes of CX 3 CL1 withdrawal) with a few exceptions where full recovery was observed. To exclude the possibility that CX 3 CL1 treatment induces a change in presynaptic cell excitability, such as the recruitment of more presynaptic fibers by the same stimulus, changes in the amplitude of extracellular fiber volley (an index of presynaptic excitation) were monitored by recording evoked fEPSPs during CX 3 CL1 application. As shown in Additional file 3: Figure S3A, in a subset of slices where the potentiating effect of CX 3 CL1 was maximal (41.68 ± 0.05 of baseline, n = 5, P <0.05) the amplitude of the afferent volley did not change after treatment with CX 3 CL1 (P = 0.45, paired t test). Patch-clamp experiments were carried out to measure CX 3 CL1-mediated modulation of NMDA current at the single cell level. Whole-cell recordings of NMDA-EPSCs were performed in CA1 pyramidal neurons in acute slices while repeatedly stimulating Schaffer collateral axons. CX 3 CL1 treatment caused an increase in NMDA-EPSC peak amplitude (to 1.25 ± 0.03 P < 0.05 vs baseline) that recovered upon withdrawal (to 1.06 ± 0.04 P >0.05 vs baseline, Figure 1B, five slices/five mice). These results confirmed the CX 3 CL1-mediated potentiation of NMDA responses also at the single cell level.

Paired pulse facilitation (PPF) ratio is not altered by CX 3 CL1 application
To analyze if CX 3 CL1 potentiates NMDA-fEPSPs by altering presynaptic glutamate release, we performed studies of short-term plasticity, that is, paired-pulse ratio (PPR) measurements. The PPR, which is the ratio of the amplitude of the second response vs the first, depends on the probability of vesicular release at synapse and its variation is generally associated with changes in transmitter release probability [42].
To evaluate PPR values we stimulated Schaffer collateral pathway projections to CA1 at 50 ms intervals. As shown in Additional file 3: Figure S3B, PPR was 1.53 ± 0.017 and 1.54 ± 0.026 before and after 20 min of CX 3 CL1 treatment (6 slices/4 mice, P = 0.74, paired t test), respectively.
Presynaptic release of Glu can be affected by removal of Mg 2+ from the external solution, changing the competition between Mg 2+ and Ca 2+ at terminal endings and altering PPR mechanisms [43]. To investigate this possibility we repeated the PPR measurements in whole-cell experiments where the NMDA current were isolated by depolarizing the cell in a standard ACSF medium. As shown in Additional file 3: Figure S3C, the PPR measured in this condition was 1.43 ± 0.08 and 1.44 ± 0.09 before and after 20 minutes of CX 3 CL1 treatment, (5 slices/5 mice, P = 0.81, paired t test). These data indicated that under our experimental conditions, CX 3 CL1 treatment did not cause obvious changes in presynaptic release probability, confirming the postsynaptic nature of the potentiation of NMDA-fEPSPs, in line with our previous findings [15,44]. CX 3 CL1-mediated effects require the presence of CX 3 CR1 on microglial cells To demonstrate that the effects of CX 3 CL1 were indeed mediated by CX 3 CR1, experiments were performed on CX 3 CR1 GFP/GFP mice where the murine CX 3 CR1 gene was replaced with the cDNA encoding EGFP (Jung et al. [6]). Application of CX 3 CL1 for 20 minutes to CX 3 CR1 GFP/GFP hippocampal slices (Figure 2A, seven slices/three mice) failed to induce any obvious basal NMDA-fEPSPs slope change (P > 0.05). These findings clearly show that CX 3 CL1-induced effects on NMDA-fEPSPs were mediated through CX 3 CR1 activation.
As shown in Figure 2B, in the presence of minocycline (20 μM, 1 h preincubation), the application of CX 3 CL1 did not significantly affect the NMDA-fEPSPs (P > 0.5, eight slices/two mice), demonstrating a microglia-driven chemokine effect.

CX 3 CL1-mediated NMDAR potentiation is mediated by ARs
We have previously shown that CX 3 CL1 treatment induces the release of adenosine from microglia [12] and that CX 3 CL1-mediated modulation of synaptic transmission and plasticity involves AR activity [17,48]. In addition, it is known that AR activity in the hippocampus modulates NMDA receptor functioning [49]. For all these reasons we investigated whether CX 3 CL1 effects on NMDARs require AR activity. We first treated hippocampal slices with the broad-spectrum AR antagonist CGS15943. The applications of CGS15943 (1 μM) abolished CX 3 CL1-mediated STP of NMDA-fEPSPs ( Figure 3A, P > 0.05 vs baseline, nine slices/four mice) indicating that the activity of ARs is indeed involved in CX 3 CL1 effects.
To disclose the specific role of distinct AR subtypes, we performed additional experiments with selective A 1 R, A 2A R and A 3 R antagonists ( Figure 3B-D).
Conversely, when slices were treated with the selective A 2A R antagonist SCH58261 (10 nM), CX 3 CL1 was unable to affect NMDA-fEPSPs slope ( Figure 3D, P > 0.05 vs baseline P < 0.01 vs CX 3 CL1 alone, six slices/three mice), indicating that A 2A R activity is associated with CX 3 CL1 effects on NMDAR-dependent basal transmission.
The role of ARs in CX 3 CL1 effects was further investigated by means of mice selectively lacking AR subtypes. As shown in Figure 4A application of CX 3 CL1 in A 1 R −/− mice increased NMDA-fEPSPs slope (to 1.14 ± 0.01, P < 0.05 vs baseline) that partially recovered after washout (to 1.04 ±0.02, P > 0.05 vs baseline, ten slices/ Microglia activation is necessary for CX 3 CL1 action. Slices pretreated for 1 h with minocycline (last 10 minutes in the graph) and then continuously superperfused. Cotreatment with CX 3 CL1 did not increase NMDA-fEPSPs (n = 8/2). three mice). Similar results were obtained in A 3 R −/− mice, where CX 3 CL1 elicited a significant increase of NMDA-fEPSPs slope (to 1.11 ± 0.01, P < 0.05 vs baseline) that recovered upon withdrawal (to 1.02 ± 0.01, P > 0.05 to baseline, Figure 4B, ten slices/two mice). Interestingly, in accordance with pharmacological data, in A 2A R deficient mice (A 2A R −/− ) CX 3 CL1 was unable to affect NMDA-fEPSPs. (P > 0.05, vs baseline, 11 slices/3 mice, Figure 4C), thus pointing to an essential role of A 2A R activity in the CX 3 CL1 effect. To further support this observation, we tested whether a selective A 2A R agonist, VT7 (50 nM), was able to mimic CX 3 CL1 effect. The results in Figure 5A demonstrate that in the presence of adenosine deaminase (ADA, 1 U/ml), which removes ambient adenosine [50], VT7 induced an increase of NMDA-fEPSPs slope (to 1.18 ± 0.01, P < 0.05 vs baseline, nine slices/two mice) that did not recover after washout (to 1.15 ± 0.01, P < 0.05 vs baseline).
Considered together, our findings indicate that the influence of CX 3 CL1 on NMDARs is mediated by the A 2A R activity.

D-serine is involved in CX 3 CL1-dependent modulation of NMDARs
Recent evidence indicates that D-serine serves as indispensable cofactor at the glycine binding site of NMDARs representing a new and important glia-derived neuromodulatory factor [51]. In fact, in glial cells, D-serine is synthesized and metabolized endogenously by the enzymes serine racemase (SR) [52] and D-amino acid oxidase (DAAO) [53], respectively.
To investigate whether D-serine is involved in CX 3 CL1-mediated modulation of NMDARs, we first treated the slices with the selective NMDAR glycine-site antagonist 5,7-dichlorokynurenic acid (DCKA) at a concentration (750 nM) known to reduce NMDA responses by around 25% [54] and able to prevent D-serine mediated increase of NMDA-fEPSPs slope (n = 7, data not shown). At this concentration, DCKA abolished CX 3 CL1 effects on NMDA-fEPSPs (P > 0.05 vs baseline, eight slices/four mice, Figure 6A).
To disclose a possible D-serine versus glycine coagonist requirement, we treated the slices with DAAO (0.1 U/ml). Excess catalase (300 U/ml) was added in the solution in order to avoid the possibility of hydrogen peroxide  production from the enzymatic degradation of D-serine by DAAO. Under this condition, DAAO completely prevented CX 3 CL1 action (P > 0.05 vs baseline, ten slices/ four mice, Figure 6B), indicating that D-serine is necessary. To further confirm the involvement of D-serine in the modulation of NMDARs by CX 3 CL1, we applied exogenous D-serine (10 μM, six slices/four mice, Figure 6C) to hippocampal slices. As shown in Figure 6C, D-serine produced an increase in NMDA-fEPSPs slope reaching a plateaux effect within 20 minutes (up to 64.4 ± 0.01 after 40 minutes treatment, P < 0.05 vs baseline). Subsequent coapplication of CX 3 CL1 failed to further increase NMDA-fEPSPs slope (P > 0.05 vs D-serine alone), indicating that saturation of the glycine site by D-serine interferes with chemokine action.
Since we have demonstrated that the CX 3 CL1 effect is prevented by hampering both D-serine and A 2A R functions, we investigated whether the two pathways share some common mechanisms of action. For this reason we applied the A 2A R agonist VT7 (in the presence of ADA, 1 U/ml) while partially blocking the NMDAR D-serine site by DCKA. In this condition, VT7 failed to increase NMDA-fEPSPs slope (P > 0.05 vs baseline, Figure 5B, 11 slices/ 3 mice). In addition, slices pretreatment with DAAO (0.1 U/ml, in the presence of catalase, 300 U/ml, and ADA, 1 U/ml) inhibits VT7 action (>0.05 vs baseline, Figure 5C, seven slices/three mice). Taken together this evidence suggests that NMDAR modulation mediated by A 2A R requires D-serine.

CX 3 CL1 and VT7 stimulate the release of D-serine from glia cells
To further investigate the possible source of D-serine in CX 3 CL1-induced NMDAR modulation, we measured D-serine levels in glia-conditioned medium. The control medium (M-ACSF) did not contain D-serine, as it is shown in Additional file 2: Figure S2.
Primary cortical glial cells and isolated microglial cell populations were stimulated with CX 3 CL1 (10 nM) for different times (20 minutes, 2 h and 20 minutes or 4 h). After stimulation, mass spectrometry was performed on media samples. Treatment of both primary purified microglia and mixed glia with CX 3 CL1 resulted in a large increase in D-serine in the medium (Figure 7). In particular, in microglial cells CX 3 CL1 induced a significant increase in D-serine release at all the incubation times considered (1.53 ± 0.08 after 20 minutes, 1.26 ± 0.07 after 2 h and 20 minutes and 2.63 ± 0.80 after 4 h, P < 0.05 in all cases vs untreated, n = 8, Figure 7A). Similarly, CX 3 CL1 treatment augmented D-serine level in the medium of mixed glia (1.76 ± 0.19 after 20 minutes, 1.66 ± 0.11 after 2 h and 20 minutes and 1.52 ± 0.24 after 4 h, P < 0.05 in all cases vs untreated, n = 8, Figure 7B).
We then tested whether also the A 2A R agonist VT7 was able to modulate the release of D-serine from glia. For this purpose we stimulated the cells with VT7 (50 nM) in the continuous presence of adenosine deaminase (ADA, 1 U/ml). As for CX 3 CL1, the A 2A R agonist was able to induce an increment in the basal level of D-serine. Specifically, VT7 treatment produced an increase in D-serine release in both primary purified microglia (up to: 1.28 ± 0.13 after 20 minutes, 1.38 ± 0.10 after 2 h and 20 minutes and 1.60 ± 0.30 after 4 h, P < 0.05 in all cases vs untreated, n = 9, Figure 7C) and mixed glia medium (up to: 1.53 ± 0.10 after 20 minutes, 1.37 ± 0.06 after 2 h and 20 minutes and 1.47 ± 0.12 after 4 h, P < 0.05 in all cases vs untreated, n = 12, Figure 7D), respectively. Interestingly, the control basal level of D-serine in purified microglia medium was unchanged at different time points, whereas in mixed glia it accumulates upon time, and it had similar values (at 20 minutes, both in M-ACSF or M-ACSF plus ADA, P >0.05 for all cases). In addition, the level of basal D-serine released by mixed glia or microglia was in the order of 10 fmol/cell, accordingly to previously reported values [25].
We further tested the ability of VT7 to induce the release of D-serine form neurons in the presence of ADA. During treatment, no variation in the basal level of D-serine was observed, at all time points considered (P > 0.05 vs untreated, n = 12). Notably, the basal level of D-serine released by neurons was much lower compared to glia (at 20 minutes: 37.24 ± 7.02 ng/mg of protein, n = 12).

Discussion
The main finding of the present study is that NMDAR function, in CA1 hippocampal region, is potentiated by CX 3 CL1 application. The activation of the glial A 2A R and the release of the NMDAR coagonist D-serine are likely major causative factors for the observed NMDA potentiation. Interestingly this effect is postsynaptic in nature, depends on microglia and requires the presence of CX 3 CR1. Specifically, the parameter related to changes in the probability of vesicular release at the synapse, such as PPR, was not affected by CX 3 CL1, indicating that the chemokine did not potentiate NMDA-fEPSPs by altering presynaptic glutamate release (see also [15,44]). In addition, the use of genetically modified mice lacking CX 3 CR1 or inhibiting microglia function by minocycline treatment prevented CX 3 CL1-mediated potentiation of NMDAR function.

A 2A R activity is involved in CX 3 CL1 effect
It is established that adenosine acts as neuromodulator of NMDAR function; in particular A 2A Rs exert a facilitatory role on NMDARs [55,56]. Nevertheless, the consequences of A 2A R activation on synaptic transmission in physiological conditions of adenosine release remained elusive. Since the ambient level of adenosine is sufficient to activate inhibitory A 1 R but not facilitatory A 2A R, possibly due to differences in affinity [57], it is believed that specific stimulation such as high-frequency bursts is needed to increase the level of adenosine to values able to activate A 2A R [58]. Indeed, we have previously demonstrated that stimulation of microglial cells by CX 3 CL1 induces an increase in extracellular level of adenosine [12].
Here, we reported that CX 3 CL1-mediated activation of A 2A R influences the NMDA-mediated component of excitatory synaptic transmission. In particular, the specific A 2A R blocker SCH58261, as well as the genetic ablation of A 2A R, prevented the short-term potentiation of NMDA fEPSPs induced by CX 3 CL1. In addition, the specific A 2A R agonist VT7 was able to mimic CX 3 CL1 effect, producing a similar potentiation of NMDA-fEPSPs.

D-serine is essential for CX 3 CL1-mediated potentiation of NMDA-fEPSPs
It is well established that D-serine act as an essential coagonist of NMDAR. The production of D-serine in the mammalian brain results from the activity of SR, an enzyme that converts L-serine into D-serine [59]. Astrocytes synthesize and release D-serine [25], for instance following activation of AMPARs by glutamate released from nerve terminals [24,60]. In contrast, scarce information is available on microglia derived D-serine. It has been reported that microglia express serine racemase [27][28][29] and release D-serine after activation with amyloid β-peptide [25].
We demonstrated that D-serine is involved in CX 3 CL1mediated potentiation of NMDA-fEPSPs because its action is prevented by: (i) the selective NMDAR antagonist of the glycine (and D-serine)-binding site DCKA, (ii) the enzymatic degradation of D-serine by DAAO, and (iii) the saturation of the coagonist site by D-serine.
In addition, mass spectrometry analysis revealed that treatment of primary mixed glial cells and purified microglial population with CX 3 CL1 induces a significant increase of D-serine release in the extracellular medium. The glial population used for these studies was derived from cerebral cortices and we are aware that microglial phenotypes differ across brain regions and ages [61,62], so that results from these preparation cannot be directly transferred to hippocampal glial cells; nevertheless these data strongly support a direct role of CX 3 CL1 in controlling D-serine release from glia, that need further investigation, and clearly demonstrate a basal release of D-serine from microglial cells that can be further increased by CX 3 CL1 treatment.
Overall, the evidence suggests that both A 2A R activity and D-serine release are needed for CX 3 CL1-mediated potentiation of NMDA-fEPSPs. Are these factors interacting in some way? Our results point to a possible crosstalk between A 2A R activity and D-serine release. First of all, at Schaffer collateral-CA1 synapses the A 2A R agonist failed to increase NMDA-fEPSPs when the NMDA glycine site was partially blocked by DCKA, indicating an extracellular modulation (via D-serine binding) of NMDAR mediated by the activation of A 2A R. It is unlikely that the partial block of the NMDAR extracellular site could mask a possible A 2A Rmediated intracellular modulation of the receptor [63], the latter actually occurring at hippocampal mossy-CA3 synapse via calcium amplification signals [56]. In addition, enzymatic degradation of D-serine by DAAO prevented the A 2A R agonist mediated potentiation of NMDA-fEPSPs, supporting the D-serine requirement for VT7 effect.
Secondly, by means of mass spectrometry analysis, we have shown that primary glial cells and purified microglia are capable to release D-serine in the medium after stimulation with VT7. These observations suggest a direct role of A 2A R in controlling the release of D-serine from glia that in turn can modulate NMDAR functions. To the best of our knowledge, this is the first report on adenosine-mediated release of D-serine from glia. A 2A R are indeed expressed on microglia and astrocytes [49,64,65], usually coupled to G s proteins, enhancing PKA or PKC activity [56,58,66]. Interestingly, PKC activity seems relevant in regulating SR activity and D-serine availability [67][68][69]. Since it has been reported that neuronal cells can release D-serine, with controversial functions on synaptic NMDAR [70,71], we investigated whether VT7 stimulated D-serine release from neurons. The basal release of D-serine from neurons appeared markedly reduced compared to glia and the activation of neuronal A 2A R was unable to further increase the extracellular D-serine levels, although we cannot exclude an indirect modulation of neuronal D-serine via other mediators released by CX 3 CL1-activated microglia.
We have previously demonstrated that CX 3 CL1 inhibits AMPA-mediated transmission and LTP expression [15,17] while here we reported that CX 3 CL1 potentiates NMDAR.
We speculate that the persistent CX 3 CL1-mediated potentiation of NMDA function and the consequent calcium increase could interfere with the complex mechanisms of LTP induction that depends not only by the precise amplitude but also by the duration of postsynaptic Ca 2+ increase and by the recent history of synaptic activity. Indeed, it has been demonstrated that prior activation of NMDAR either by synaptic or pharmacological stimulation could effectively inhibit the subsequent LTP induction [72,73].
Since in our experimental setting we recorded synaptic NMDAR blocking AMPAR component, it would be informative to investigate CX 3 CL1-mediated NMDAR modulation in a more physiological condition, where neuronal and glial AMPAR are activable.
Interestingly, the effect of CX 3 CL1 seems to be specific for the synaptic NMDARs that are exclusively gated by the coagonist D-serine [23]. It has been reported that activation of synaptic NMDARs (NR2A-containing NMDARs) preferentially triggers cell survival pathways, whereas stimulation of the whole-cell population of NMDARs and/ or selective activation of extrasynaptic NMDARs (NR2Bcontaining NMDARs) signals toxic stimuli ( [74,75], but see also [23]). Therefore, CX 3 CL1-mediated potentiation of synaptic NMDAR function can possibly contribute to the neuroprotective action of CX 3 CL1 described in different in vitro and in vivo conditions [11,12,14].
Future studies are warranted to discern the role and the precise mechanism of CX 3 CL1-mediated potentiation of NMDA-fEPSPs in both physiological and pathological condition, with particular attention to the sequence of mediators and cell types involved, namely neurons, astrocytes or microglia. Nevertheless, considering overall data, we can speculate that on microglial cells the activation of CX 3 CR1 triggers the release of adenosine that in turn, via A 2A R activity, increases the release of D-serine from glia leading to a potentiation of NMDA function (Figure 8).

Conclusions
In summary, we have demonstrated a direct association between the effects of CX 3 CL1 on NMDAR activation and increased levels of D-serine and A 2A R activity within the CA1 region of the hippocampus. The CX 3 CL1-CX 3 CR1 axis therefore represents a pathway for direct communication between neural cells and microglia.  Figure 8 Suggested mechanism of CX 3 CL1 action. Activation of microglial CX 3 CR1 by exogenous application of CX 3 CL1 induces the release of adenosine (1) that in turn activates adenosine receptor type A2 (A 2A R) located on microglia cells and, possibly, on astrocytes stimulating the release of D-serine (2). The binding of the cofactor D-serine to the synaptically activated N-methyl-D-aspartate receptors (NMDARs) (3) potentiates the receptor function. Note that α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) are blocked and, for clarity, only factors demonstrated to play a role in the described mechanism are represented.