The chemokine CXCL16 modulates neurotransmitter release in hippocampal CA1 area

Chemokines have several physio-pathological roles in the brain. Among them, the modulation of synaptic contacts and neurotransmission recently emerged as crucial activities during brain development, in adulthood, upon neuroinflammation and neurodegenerative diseases. CXCL16 is a chemokine normally expressed in the brain, where it exerts neuroprotective activity against glutamate-induced damages through cross communication with astrocytes and the involvement of the adenosine receptor type 3 (A3R) and the chemokine CCL2. Here we demonstrated for the first time that CXCL16 exerts a modulatory activity on inhibitory and excitatory synaptic transmission in CA1 area. We found that CXCL16 increases the frequency of the miniature inhibitory synaptic currents (mIPSCs) and the paired-pulse ratio (PPR) of evoked IPSCs (eIPSCs), suggesting a presynaptic modulation of the probability of GABA release. In addition, CXCL16 increases the frequency of the miniature excitatory synaptic currents (mEPSCs) and reduces the PPR of evoked excitatory transmission, indicating that the chemokine also modulates and enhances the release of glutamate. These effects were not present in the A3RKO mice and in WT slices treated with minocycline, confirming the involvement of A3 receptors and introducing microglial cells as key mediators of the modulatory activity of CXCL16 on neurons.

the frequency of mIPSCs. This effect was long-lasting, with a slow wash, and is similar to what was observed for other chemokines affecting synaptic transmission [33][34][35][36] . The modulation of mIPSCs frequency was also evident as a leftward shift of the cumulative probability plot for the inter-event intervals (IEI) and as a significant increase in the mean mIPSC frequency (CTRL: 2.71 ± 0.17 Hz, CXCL16: 3.52 ± 0.36 Hz, N = 8, p = 0.008, Fig. 1C). In contrast, CXCL16 did not alter mIPSC amplitude (CTRL: 16.67 ± 1.19 pA, CXCL16: 16.18 ± 1.21 pA, N = 8, p = 0.29, Fig. 1D). These data indicate that CXCL16 enhances the spontaneous GABA release at inhibitory synapses impinging onto CA1 pyramidal cells. As a control for specificity, bath application of CXCL16 in CXCR6 KO slices, did not affect the frequency (CTRL: 4.55 ± 0. 31  CXCL16 regulates evoked GABA release. We then investigated whether CXCL16 regulates GABA responses evoked by extracellular stimulation with paired stimuli (inter stimulus interval = 50 ms) in the stratum radiatum of CA1 hippocampal region. Bath application of CXCL16 (10 nM) reduced the amplitude of the evoked IPSC (eIPSC) to 72.7 ± 8% of control within 20 min (Fig. 1E The results obtained on the evoked release of GABA are apparently different from those on miniature recordings, as CXCL16 application increased the spontaneous GABA release events. To better address this issue, we investigated the effect of CXCL16 application on spontaneous GABA currents (sIPSCs). Bath application of CXCL16 decreased the mean sIPSCs amplitude (CTRL: 35.85 ± 2.76 pA, CXCL16: 31.01 ± 1.69 pA, N = 7, p = 0.039) and shifted toward the left the cumulative distribution histogram ( Fig. 2A,B). Although the mean frequency of sIPSCs did not change upon CXCL16 application (CTRL: 5.43 ± 0.89 Hz, CXCL16: 5.37 ± 1.05 Hz, N = 7, p = 0.88, Fig. 2C), closer inspection of the traces revealed that the number of large amplitude sIPSCs is reduced, whereas the number of small-amplitude sIPSCs is increased. This effect was quantified in a pooled amplitude histograms showing that the number of events falling within 20 pA (the average of mIPCs amplitude) was increased in the presence of CXCL16 whereas the number of events larger than 80 pA was decreased (Fig. 2D). Overall, these results support the evidence that CXCL16 is able (i) to produce a presynaptic increase of the probability of spontaneous, action potential-independent GABA release, because the frequency of both the smallest sIPSCs and mIPSCs is increased and (ii) to reduce action-potential dependent GABA release, as demonstrated by the decrease in the amplitude of both eIPSCs and large sIPSCs.
In order to elucidate possible mechanisms underlying the effects of CXCL16 on evoked GABAergic activity, we investigated the role of metabotropic GABA-B receptor that has been reported to specifically reduce the release of evoked GABA onto CA1 pyramidal neurons [37][38][39] .
When the slices were pre-incubated for 30 min with the GABA-B receptors antagonist CGP 55845 (1 μ M) and then continuously superfused, CXCL16 was not able to affect the peak amplitude and the PPR of eIPSCs (fold increases relative to baseline: peak amplitude: 0.98 ± 0.07, p = 0.73; PPR: 1.01 ± 0.04, p = 0.41, N = 6, Fig. 3). By contrast, in parallel experiments were CGP55845 was not applied, CXCL16 reduced the peak amplitude to 0.75 ± 0.06 (fold increase relative to CTRL, N = 6, p = 0.02) and increased the eIPSC PPR to 1.33 ± 0.14 (N = 6, p = 0.03). Together these results suggest that the inhibitory effect of CXCL16 on evoked GABAergic transmission is mediated by the GABA-B receptors activity.
Excitatory pre-synaptic transmission is potentiated by CXCL16 application. The results presented above showed that CXCL16 affects inhibitory synaptic transmission by modulating GABA release. To further investigate its action on synaptic transmission, we tested the effects of CXCL16 on glutamatergic one.
CXCL16 increased the frequency of miniature excitatory post-synaptic currents (mEPSCs), as illustrated in the time course from a representative experiment showed in Fig. 4A,B. Specifically, during bath application of CXCL16 (10 nM, 20 min) the cumulative distribution curve for IEIs was shifted towards the left and, correspondingly, the average frequency of mEPSC was increased (CTRL: 0.62 ± 0.13 Hz, CXCL16: 0.85 ± 0.13 Hz, N = 8, p = 0.05, Fig. 4C and inset). In contrast the mean amplitudes of mEPSCs were similar before and after CXCL16 treatment (15.03 ± 0.58 pA and 14.26 ± 1.14 pA, respectively, N = 8, p = 0.2, Fig. 4D inset) and no changes in the cumulative probability plot for amplitudes were observed (Fig. 4D). These results indicate that CXCL16 is able to increase the spontaneous glutamate release onto CA1 pyramidal neurons.
In addition, we elicited excitatory post-synaptic currents by Schaffer collateral stimulation (eEPSCs) with paired stimuli (ISI = 50 ms). We found that CXCL16 significantly increased the peak amplitude of eEPSC CXCL16 fails to affect neurotransmission in adenosine receptor A3R Knock-Out mice. We previously reported that CXCL16 exerts neuroprotective effects through the activation of the A3R on astrocytes 31 . Here we examined the effect of the chemokine on synaptic transmission in slices obtained from A3R KO mice.

Minocycline treatment prevents CXCL16-dependent modulation of synaptic transmission.
Microglia is known to modulate synaptic transmission [40][41][42] , and microglia also express CXCR6 31 . We asked whether the changes in synaptic transmission observed upon CXCL16 administration might also required microglia neuron interplay. To test this hypothesis, slices were pre-treated for 1 h with the antibiotic minocycline (500 nM) to "inhibit" microglia activation 43 , and then continuously superfused during the recording. In this experimental condition, CXCL16-mediated modulation of GABAergic transmission was completely

Discussion
CXCL16 is up-regulated in the brain upon different pathological states. Nevertheless, the observation that CXCL16 and its unique receptor CXCR6 are constitutively expressed in the CNS by neurons and glia suggests that this pair might have some important functions in brain homeostasis. In the present paper we describe for the first time that CXCL16 has neuromodulatory action in hippocampal CA1 region, tuning both the GABAergic and the glutamatergic transmissions, and that this modulation requires glia-neuron interplay.
At inhibitory synapses, CXCL16 enhances the action-potential independent release of GABA, increasing the frequency of mIPSCs. By contrast, the evoked release is reduced and the PPR ratio was increased. Interestingly, in the frequency distribution histogram of sIPSCs recordings, we found that, although CXCL16 did not affect the mean frequency, the number of small-sIPSCs (up to 20 pA, that resemble mIPSCs) was increased, whereas the largest were reduced, supporting the idea that CXCL16 increases the spontaneous GABA release while reduces the action-potential dependent ones. The opposite modulation of spontaneous and evoked release of neurotransmitter by different mediators has been previously reported [44][45][46][47][48][49][50] . Several mechanisms could contribute to the opposite regulation of spontaneous and evoked release. For instance, large differences in the amount of intraterminal Ca2 + levels have been shown to be associated with a block in evoked release or an increase in mEPSC frequency 51,52 . Moreover, differences in presynaptic fusion machineries giving rise to spontaneous and evoked release or segregation of spontaneous and evoked neurotransmission can account for the differential effects of CXCL16 on spontaneous and evoked GABA transmission 53,54 .
Alternatively, the enhancement of GABA release, associated with the increase in the frequency of mIPSCs, could activate presynaptic GABA-B receptors, reducing the probability of evoked release, as reported for hippocampal inhibitory neurons in CA3 55 . In line with this last hypothesis, we demonstrated that the pharmacological inhibition of GABA-B receptors prevents the effect of CXCL16 on evoked GABA release, suggesting their involvement in the mechanism underlying the opposite modulation by CXCL16 on spontaneous and evoked GABAergic activity.
Besides its role on GABAergic transmission, CXCL16 affects also excitatory transmission enhancing both spontaneous and evoked glutamate release. Specifically, we observed that CXCL16 increased the frequency of mEPSCs and the amplitude of eEPSCs but reduced the PPR, strongly indicating that CXCL16 enhances glutamate release.
Presently, we are not able to distinguish whether CXCL16 acts directly on (i) glutamatergic and GABAergic terminals, (ii) via intermediate inhibitory interneuron(s), or (iii) by the action of glial cells, given the diffuse distribution of CXCR6 31 . However, the neuroprotective activity of CXCL16, which requires complex interplay between cell types 31 , suggest that glia-neuron interaction could be a general mechanisms of CXCL16 action. Supporting this complex interaction, we showed that inhibiting microglia function with minocycline, the effects of CXCL16 on GABAergic and glutamatergic transmission were abolished. It can be argued that minocycline can have additional effects on neurons 43,[56][57][58][59] . However at the concentration we used (500 nM), minocycline did not alter basal synaptic transmission, because we did not observe differences in mIPSC and mEPSC amplitude and frequency. In addition we demonstrated that A3Rs are required for the neuromodulatory activity of CXCL16, since upon pharmacological or genetic inactivation of A3Rs, the chemokine is ineffective in modulating GABAergic and glutamatergic transmission. The relevance of A3Rs in mediating CXCL16 functions has been previously reported for the neuroprotective activity in ischemia 32 . Although the A3Rs are expressed in the brain, their function in the CNS is poorly understood 60   generally thought that activation of A3Rs in the CNS might occur only under pathological conditions (such as hypoxia, ischemia or epilepsy) which increase extracellular levels of adenosine in the brain 65,67-69 . However, it has been recently shown that A3Rs can be tonically active in mouse hippocampal slices, tuning glutamatergic activity in combination with A1Rs and A2ARs 70 but the consequences of A3Rs activation on basal synaptic transmission are poorly characterized. In the present paper, we propose A3R as key mediator of the presynaptic regulation of both GABA and glutamate release by CXCL16. Considering the widespread distribution of A3Rs, at present it is not clear whether neuronal or glial A3Rs are involved. Since it has been shown that A3R agonists potentiate excitatory post-synaptic currents both in cortical and hippocampal neurons, via post-synaptic mechanisms 7,64,70,71 , according to our data we can probably exclude the involvement of neuronal A3Rs in CXCL16-dependent modulation of glutamatergic transmission. In addition, as far as we know, there is only one paper that shows a link between A3Rs activation and the regulation of GABA release 72 , in an experimental model of pain, but the location of A3Rs has not been determined. We have previously demonstrated that, acting in concert with astrocytic A3R, CXCL16 mediates the release of CCL2 from astrocytes 31 . In the last years several evidences led the emerging concept that CCL2 may function as a neuromodulator in the CNS 73,74 , and in particular it has been shown that in the CA1 region of rat hippocampus CCL2 enhances neuronal excitability and synaptic transmission via presynaptic mechanisms 75 with effects similar to those we found for CXCL16. According to all these findings we could speculate that the release of CCL2 from astrocytes could contribute to the modulation of glutamatergic and GABAergic neurotransmission by CXCL16 with a mechanism described in Fig. 8.
The increased excitation/inhibition ratio induced by CXCL16 is in apparent contrast with its neuroprotective activity 31 . However the two effects are induced by different CXCL16 concentrations and stimulation times, likely reflecting distinct mechanisms, not necessary correlated.
Overall, we demonstrated for the first time that CXCL16 is able to modulate neurotransmitter release on inhibitory and excitatory synapses in the hippocampus, with mechanisms requiring microglia and A3Rs activation. These data further reinforce the current idea that chemokines, key players in the interplay between neurons and glia cells, play unexplored roles in physiological conditions as regulators of synaptic transmission, representing endogenous modulators of brain homeostasis.

Materials and Methods
Animals. Procedures using laboratory animals were in accordance with the Italian and European guidelines and were approved by the Italian Ministry of Health in accordance with the guidelines on the ethical use of animals from the European Community Council Directive of 22 September 2010 (2010/63/EU). All efforts were made to minimize the number of animals used and their suffering.
Hippocampal slices were routinely obtained from 1-month-old C57BL/6 J. When specified, hippocampal slices were obtained also from: i) A3R knockout mice (A3RKO) 76 ; ii) cxcr6 gfp knock-in mice 77 in which the coding region of the receptor has been substituted with the coding region of the Green Fluorescent Protein (GFP). In the paper we refer to these mice as CXCR6KO mice: in particular heterozygous CXCR6KO were mated to obtain both homozygous CXCR6KO, and their CXCR6+ /+ littermates as controls. Both A3RKO and cxcr6gfp knock-in mice were backcrossed at least 10 times on a C57BL/6J background.
Patch clamp technique. Whole-cell patch clamp recordings were performed on CA1 pyramidal neurons at room temperature by using a Multiclamp 700B amplifier (Molecular Devices, USA). The ACSF was perfused at a rate of approximately 2 ml/min by using a gravity-driven perfusion system. Cell capacitance was constantly monitored over the time and experiments were access resistance changed more than 20% were discarded. Glass electrodes (3-4 MΩ) were pulled with a vertical puller (PC-10, Narishige). Pipette were filled with 140 mM Cs Methanesulfonate, 10 mM Hepes, 0.5 mM EGTA, and 2 mM Mg-ATP, Na3-GTP 0.3 mM, MgCl2 2 mM (295-300 mOsm, pH 7.2). Signals were acquired (sampling 10 kHz, low-pass filtered 2 kHz) with DigiData-1440A using pCLAMP-v10 software (Molecular Devices, USA).
GABAergic outward membrane currents were recorded with the neuron clamped at 0 mV. At this voltage, Cl− -mediated inhibitory events are outward currents (estimated ECl = − 80 mV) whereas excitatory currents are inward but of small amplitude as they would occur close to their reversal potential. Although it was possible to isolate sIPSCs pharmacologically, by using 10 μ M CNQX plus 50 μ M AP-5 to block both the AMPA and NMDA receptor components of spontaneous excitatory postsynaptic currents (sEPSCs), this antagonist mixture sometimes attenuated or occasionally completely blocked sIPSCs 78 . This presumably reflected impediment of excitatory synaptic drive to the inhibitory interneurons that were responsible for sIPSC generation. In view of this variable effect of CNQX/AP-5 on sIPSCs, we elected to use a holding potential of 0 mV rather than pharmacological methods to separate sIPSCs from sEPSCs. The validity of this approach is supported by the observation that 100μ M picrotoxin completely eliminated all spontaneous outward current activity recorded at 0 mV (not shown).
By using the same conditions, excitatory post-synaptic currents (EPSCs) were recorded clamping the cell at − 80 mV. In a subset of experiments, the glutamatergic nature of the mEPSC recordings was confirmed at the end of the experiment by total blockade of mEPSCs by DNQX (20 μ M; data not shown).
Miniature recordings. Miniature EPSCs/IPSCs were recorded during an initial 10 min baseline period, followed by application of TTX (0.5μ M, Tocris Bioscience, Bristol, United Kingdom) for 15 min. After stabilization of TTX effect, CXCL16 was applied for 20 min. Only data from the last 5 min of each recording epoch was analyzed to ensure that drugs had fully equilibrated. Analysis was performed off-line using MiniAnalysis software (Mini Analysis, Synaptosoft Fort Lee, NJ, USA) with the threshold for detection set at 5 pA. Whole-cell currents were recorded at 20 kHz and filtered with a Bessel filter at 4 kHz.

Paired-pulse experiments.
A concentric bipolar stimulating electrode (SNE-100 × 50 mm long Elektronik-Harvard Apparatus GmbH, Crisel Instruments, Rome, Italy) was positioned in the stratum radiatum to evoke IPCSs from CA1 pyramidal neurons or in CA3 region to evoke eEPSCs. Pairs of stimuli (ISI 50 ms) were applied every 20 sec. Stimulus intensity was 0.27 ± 0.03 mA (range, 0.2-0.4 mA), delivered through a A320R Isostim Stimulator/Isolator (WPI). PPR was calculated as the ratio between the amplitude evoked by the second Figure 8. Schematic diagram summarizing the proposed mechanism of CXCL16 action. Exogenous CXCL16 might act on both microglia and astrocytic CXCR6 (1). Activation of microglial CXCR6 induces the release of adenosine (2). On astrocytes, the simultaneous activation of CXCR6, by exogenous CXCL16, and A3R, by adenosine, induces the release of CCL2 (3) that, acting presynaptically potentiates the spontaneous release (mEPSC) and evoked release (eEPSC) of glutamate on excitatory synapses (4). A similar mechanism might be involved in CXCL16 enhancement of spontaneous GABA release (mIPSC) at GABAergic synapses (4′ ). Increased GABA release activates presynaptic GABA-B receptors (5) reducing the probability of evoked release (eIPSC) (6). The inhibition of microglia activation with minocycline prevents the effects of CXCL16 on both glutamatergic and GABAergic transmission (7).
Statistical analysis. The values were reported as mean ± SEM. Unless otherwise specified, data values refer to number of cells analyzed. Student's paired t tests was used for inter-group comparison. For cumulative probability plot comparisons we used Kolmogorov-Smirnov test (Mini Analysis, Synaptosoft Fort Lee, NJ, USA). Levels of significance were set as *p < 0.05; **p < 0.01; ***p < 0.001.