P2X4 receptor controls microglia activation and favors remyelination in autoimmune encephalitis

Following peripheral nerve injury, microglia in the spinal dorsal horn exhibit a reactive phenotype and upregulate expression of a variety of genes, including purinergic P2x4r (Tsuda et al, 2003; Beggs et al, 2012). Accordingly, we detected previously an increase in P2x4r mRNA expression in multiple sclerosis (MS) samples and at the peak of the immune attack in the acute EAE model (Vázquez‐Villoldo et al, 2014). We have further analyzed the time course of the P2x4r expression in EAE mice immunized with myelin oligodendrocyte glycoprotein (MOG). Levels of P2x4r expression were increased at the peak of the disease and remained elevated during the recovery phase (30 days; Fig 1A). Interestingly, there was a strong correlation between P2x4r expression and the neurological score, both at the peak and at recovery (Fig 1A; r² = 0.99 and 0.61, respectively). Previous data have demonstrated that interferon regulatory factor 8 (IRF8)–IRF5 transcriptional axis is a critical regulator for shifting microglia toward a P2X4R⁺‐reactive phenotype (Masuda et al, 2014). Accordingly, we observed that Irf8 and Irf5 transcription factors were upregulated at the peak and recovery phases of the disease and their expression correlated well with P2x4r expression (Fig 1B). P2x4r upregulation was also detected in FACS‐isolated microglia (Cd11b⁺CD45^high) in the spinal cord at the EAE recovery phase (Fig 1C).

We then tested the role of P2X4R in EAE pathogenesis in mice treated daily with P2X4R antagonist TNP‐ATP (10 mg/kg) from the onset of the disease at 10 days postimmunization (dpi). This time window is coincident with microglia activation, as previously reported (Ajami et al, 2011), and does not interfere with immune priming. Microglia die at early stages of EAE induction, and this population is replenished by infiltrating monocytes, promoting progression to paralysis (Ajami et al, 2011). Because P2X4R blockade was previously demonstrated to prevent LPS‐induced microglial cell death (Vázquez‐Villoldo et al, 2014), we reasoned that blockade of microglial cell death by TNP‐ATP would prevent replacement by monocyte and improve clinical signs of EAE. In contrast, blockade of P2X4R with TNP‐ATP exacerbated EAE disease (Fig 2A). Accordingly, the latency of the corticospinal tract was significantly increased in TNP‐ATP‐treated mice (Fig 2A), indicative of higher demyelination. We next stained spinal cord sections at the end of the experiment for ionized calcium‐binding adapter protein 1 (Iba1), which is a marker commonly used to identify microglia/macrophages in the CNS. TNP‐ATP‐treated mice showed a significant increase in Iba1⁺ cells in the white and gray matter of the spinal cord versus vehicle‐treated EAE mice (Fig 2B). The increase was observed even at the same neurological score (Fig EV1), probably indicating that the increase in microglia/macrophage number is not only the consequence of the higher EAE severity after P2X4R blockade.

We next confirmed the role played by P2X4R in EAE pathogenesis using P2X4^−/− mice. We first checked whether P2X4R deficiency could affect microglia and oligodendrocytes in normal conditions. We did not detect any change in the number or morphology of Iba1⁺ cells nor in the number of Olig2⁺ oligodendrocytes in the spinal cord of 2‐month‐old P2X4^−/− mice (Appendix Fig S1). Then, we compared neurological score in WT and P2X4^−/− MOG‐injected mice. In accordance with results obtained with TNP‐ATP, P2X4^−/− mice showed an exacerbated EAE, higher latency of the corticospinal tract, and an increase in the number of Iba1⁺ microglia/macrophage cells (Fig 2C). To further assess that the effect of TNP‐ATP was P2X4R‐dependent, we treated P2X4^−/− mice with TNP‐ATP from the onset of the disease. TNP‐ATP failed to alter the course of EAE disease in P2X4^−/− mice (Fig 2D). All these data confirmed the role played by P2X4R in EAE pathogenesis.

During EAE, T cells are primed in the mouse peripheral immune system before the onset of the clinical signs (Stromnes & Goverman, 2006), suggesting that treatment with TNP‐ATP starting after the onset of the disease would have no impact on T‐cell infiltration. In order to corroborate this hypothesis, we analyzed immune cell infiltration (CD4⁺ T cells, CD8⁺ T cells, γδ T cells, neutrophils, and macrophages) in the brain (Fig EV1) and spinal cord (Fig 2E) at the peak of disease in vehicle‐ and TNP‐ATP‐treated mice by flow cytometry in another set of mice. We found similar proportions for all of CNS‐infiltrating CD45⁺ leukocytes in TNP‐ATP‐ and vehicle‐treated mice at the peak of EAE except for CD8⁺ T cells in the spinal cord (Fig 2E) and brain (Fig EV1) which were significantly increased in TNP‐ATP‐treated mice. We further assessed the role of P2X4R on T‐cell function directly by stimulating isolated splenocytes in vitro with anti‐CD3/CD28 in the presence or absence of TNP‐ATP for 3 days and then measured proliferation of CD4⁺ and CD8⁺ T cells. Furthermore, we determined T‐cell cytokine production via intracellular cytokine staining of PMA/ionomycin‐stimulated cells. Importantly, in vitro assays showed that blockade of P2X4R on lymphocytes did not alter T‐cell proliferation or cytokine production (Fig 2F).

To further exclude the involvement of P2X4R in adaptive immune system, we performed an additional EAE experiment and treated mice with TNP‐ATP during the priming phase (0–17 dpi). TNP‐ATP treatment did not affect disease development (Fig 3A). At the peak, we quantified by flow cytometry the immune response in periphery (spleen and lymph nodes) and in the spinal cord. Treatment with TNP‐ATP during the priming phase did not change the number of CD4⁺ T cells, CD8⁺ T cells, and γδ T cells in spleen, lymph nodes, or the spinal cord (Fig 3B). To further assess the CD4⁺ T‐cell response, we measured mRNA for Foxp3 and Ror, transcription factors that specify Tregs and Th17, respectively, and Ifng, signature cytokine for Th1 cells. We did not detect any change in transcript expression after TNP‐ATP treatment (Fig 3C). Finally, to check whether P2X4R blockage could alter infiltration of T cells during the immune priming phase, we measured BBB disruption by PET imaging using a radioligand that enabled tracking of matrix metalloproteinase (MMP) activity as a marker of early lesions and ongoing leukocyte infiltration (Gerwien et al, 2016). An increase in MMP activity was detected in the lumbar spinal cord at the peak of the disease; however, treatment with TNP‐ATP during the immune priming did not change the MMP‐PET signal in EAE (Fig 3D). Altogether, these data suggest that blockade of P2X4R did not interfere with the efficacy of immunization and the T‐cellular immune response against MOG.

Since myelin clearance is necessary for remyelination and recovery (Li et al, 2005; Kotter et al, 2006; Neumann et al, 2009) and phagocytosis and remyelination are modulated by microglia/macrophage polarization (Miron et al, 2013), we hypothesized that P2X4R could be involved in this process. We first checked the status of microglia/macrophage in TNP‐ATP‐ and vehicle‐treated EAE mice. We performed gene expression profiling from the lumbar spinal cord of vehicle‐ and TNP‐ATP‐treated EAE mice at the peak and recovery phases (Fig 4A and B). Expression of pro‐inflammatory and anti‐inflammatory genes involved in microglia/macrophage activation was analyzed using a 96.96 Dynamic Array^™ integrated fluidic circuit (Fluidigm). In accordance with previous data showing that macrophages and microglia showed an intermediate activation status in MS (Vogel et al, 2013), most pro‐inflammatory and anti‐inflammatory genes were significantly increased at the EAE peak (Fig EV2) and recovery phase (Fig 4B). Blockade of P2X4R with TNP‐ATP did not significantly alter anti‐inflammatory gene expression, but it significantly increased pro‐inflammatory gene expression at the recovery phase (Fig 4B), but not at EAE peak (Fig EV2). A higher increase in pro‐inflammatory gene expression was also detected on microglia FACS‐sorted (Cd11b⁺CD45^high; see gating in Fig 1C) from EAE P2X4^−/− mice versus EAE WT (Fig 4C). Accordingly, we found an increase in iNOS expression in microglia/macrophage after EAE in P2X4^−/− mice and in TNP‐ATP‐treated mice (Fig 4D and E). These data suggest that P2X4R could be influencing microglia/macrophage activation.

To further analyze the influence of P2X4R on microglia polarization in vitro, cells were primed with colony‐stimulating factors to differentiate into pro‐inflammatory and anti‐inflammatory microglia according to a previous protocol (Fig 5A; see details in Materials and Methods) and analyzed by immunocytochemistry using pro‐inflammatory (iNOS) and anti‐inflammatory (mannose receptor; MRC1) markers. Blockade of P2X4R with TNP‐ATP induced a significant increase in iNOS⁺ cells and a significant reduction in MRC1⁺ cells (Fig 5B). Accordingly, qPCR analysis revealed an increase in pro‐inflammatory genes and a decrease in anti‐inflammatory genes after TNP‐ATP treatment during polarization (Fig 5C). Similar results were obtained in P2X4^−/− microglia (Fig EV3). Altogether, these data suggest that P2X4Rs modulate microglial polarization.

A switch from a pro‐inflammatory to an anti‐inflammatory phenotype occurs in microglia and peripherally derived macrophages during remyelination in MS, and this change is essential for efficient remyelination (Miron et al, 2013). These data led us to the hypothesis that P2X4R blockade in microglia could be indirectly affecting oligodendrocyte differentiation and remyelination. We first characterized the expression and function of P2X4R in oligodendrocytes and microglia. Double‐immunocytochemistry analysis showed that P2X4R expression is virtually absent from Olig2⁺ oligodendrocytes and highly enriched in isolectin B4⁺ microglial cells in microglia–oligodendrocyte progenitor cell (OPC) mixed cultures (Fig 6A; Appendix Fig S2). On the contrary, P2X7 receptors are expressed in both cell populations. The function of P2X4R in microglia and oligodendrocytes was further analyzed by electrophysiology in acute slices from CXCR3‐GFP and PLP‐DsRed mice, respectively. When ATP was applied to whole‐cell‐clamped microglia at −70 mV, it elicited a massive inward current (619 ± 100 pA; n = 8) that was significantly reduced by A438079 (10 μM), antagonist of P2X7R (172 ± 69 pA; n = 8; Fig 6B). However, the remaining ATP‐evoked current was virtually abolished in the presence of TNP‐ATP (Fig 6B). These data suggest that P2X7R and P2X4R are the main contributors to ATP‐evoked currents in microglia. In contrast, ATP‐evoked currents in oligodendrocytes (80 ± 21 pA) were totally abolished in the presence of A438079 (Fig 6B), indicating the lack of functional P2X4R on these cells. To further exclude any direct role of P2X4R on oligodendrocyte differentiation, we stimulated culture oligodendrocytes with ATPγS (10 μM, 3 days) at low concentrations to avoid P2X7R activation. ATPγS did not induce any change in oligodendrocyte differentiation (Fig EV4). Altogether, these results suggested that oligodendrocytes lack functional P2X4Rs.

Next, we analyzed whether microglia P2X4R could play a role on oligodendrocyte differentiation. To this end, we conditioned oligodendrocyte culture medium (SATO; see Materials and Methods) with control, pro‐inflammatory, and anti‐inflammatory microglia for 24 h and then incubated oligodendrocyte progenitors with microglia‐conditioned media for 3 days (see cartoon in Fig 6C). All microglia‐conditioned media induced an increase in the number of mature MBP⁺ oligodendrocytes as compared to polarization factors alone (Fig 6C). However, anti‐inflammatory microglia induced a higher increase in oligodendrocyte differentiation, an effect that was blocked in the presence of TNP‐ATP (Fig 6C). Since BDNF enhances oligodendrocyte differentiation and myelination (Wong et al, 2013) and P2X4R stimulation in microglia has been linked to BDNF release (Tsuda et al, 2003; Coull et al, 2005; see also Fig EV4), we analyzed BDNF production in differentially polarized microglia. Western blot analysis revealed a significant increase in BDNF production in anti‐inflammatory microglia, which was significantly reduced in the presence of TNP‐ATP (Fig 6D). Accordingly, Bdnf mRNA expression during the recovery phase of EAE was significantly reduced in the presence of TNP‐ATP, a fact that correlated with the downregulation of Mbp expression (Fig 6E). Bdnf mRNA was also reduced in FACS‐isolated microglia from control and EAE P2X4^−/− mice (Fig EV4).

These data suggested that P2X4R blockade in microglia inhibits oligodendrocyte differentiation by shifting microglia toward a pro‐inflammatory phenotype, an effect that could impede remyelination. To assess that possibility, we further checked the impact of P2X4R on remyelination on the lysolecithin‐induced demyelination model using ex vivo organotypic cerebellar slice culture (Fig 6F), a model independent of adaptive immune system. Indeed, slices treated with TNP‐ATP (10 μM) during the remyelination phase (7 days) showed a significant decrease in their remyelination capacity, as revealed by analyzing MBP expression by Western blot (Fig 6F).

Phagocytosis of myelin debris by microglia is essential for an efficient regenerative response (Kotter et al, 2006; Ruckh et al, 2012). We observed a higher accumulation of disrupted or fragmented myelin after EAE induction in TNP‐ATP‐treated mice (Fig 7A) as well as in P2X4^−/− mice (Fig EV5). Fragmented myelin yielded higher immunoreactivity likely due to additional exposed epitopes of the MBP antibody. Myelin debris accumulation could be the result of higher demyelination or a defect on myelin phagocytosis. Indeed, we observed higher accumulation of myelin debris not surrounded by microglia phagocytic processes in TNP‐ATP‐treated mice (Fig 7A) at the recovery phase. Because of that, we challenged the hypothesis that these features could be the consequence of a failure on microglia/macrophage phagocytosis. For that, myelin was isolated from adult rat whole brain using sucrose gradient (Norton & Poduslo, 1973), labeled with the dye Alexa 488, and added to microglia cultures. In order to efficiently clear up myelin, microglia should internalize myelin and deliver it to lysosomes to degrade it. We monitored by confocal microscopy myelin endocytosis and myelin degradation time course. At initial stages, myelin endocytosis (1 h) was significantly increased in anti‐inflammatory microglia (Fig 7B), in accordance with recent data (Healy et al, 2016). Blockade of P2X4R with TNP‐ATP significantly reduced the anti‐inflammatory‐dependent increase in myelin endocytosis (Fig 7B). We then checked myelin degradation in the different polarized microglia populations. After 1‐h incubation with Alexa 488‐myelin, microglia were kept in label‐free growth medium for up to 6 days in the presence of polarizing factors. We could not eliminate polarization factors because microglia activation quickly reverses without them (unpublished observations). We found that microglia retained Alexa 488‐labeled myelin for up to 6 days and only about 30% of internalized myelin was degraded in that time in control microglia (Fig 7C). Anti‐inflammatory microglia showed a higher capacity to degrade myelin at 3 and 6 days, whereas pro‐inflammatory microglia degraded myelin less efficiently (Fig 7C). Moreover, P2X4R blockade by TNP‐ATP significantly attenuated the increase on myelin degradation observed in anti‐inflammatory microglia and exacerbated the deficits on myelin degradation characteristic of pro‐inflammatory microglia (Fig 7D). These results suggest that P2X4R blockade could contribute to the failure on myelin phagocytosis necessary prior to regeneration.

We then explored the therapeutic potential of P2X4R potentiation. Ivermectin (IVM) is an FDA‐approved semi‐synthetic macrocyclic lactone used in veterinary and clinical medicine as an anti‐parasitic agent. IVM allosterically modulates both ion conduction and channel gating of P2X4Rs (Priel & Silberberg, 2004). We first analyzed the role of P2X4R potentiation in EAE pathogenesis in mice treated daily with IVM (1 mg/kg) after the onset of the disease (14 dpi). IVM induced a rapid and significant amelioration of the motor deficits of EAE (Fig 8A) that lead to an improvement in corticospinal tract function in the recovery phase, as revealed by the decrease in the latency of corticospinal tract (Fig 8A). These data indicate that IVM treatment effectively promoted myelin recovery. Accordingly, IVM significantly enhanced remyelination in the lysolecithin‐induced demyelination model in organotypic cerebellar slices (Fig 8B).

Then, we analyzed the effect of IVM on microglial polarization and phagocytosis in vitro. IVM (3 μM) induced a significant increase in polarization to an anti‐inflammatory phenotype (Fig 8C). Accordingly, qPCR analysis revealed a decrease in pro‐inflammatory genes and an increase in anti‐inflammatory genes after IVM treatment during polarization (Fig 8D). Finally, we analyzed the effect of IVM on myelin phagocytosis. We did not detect a potentiation of myelin endocytosis (1 h) in anti‐inflammatory microglia, perhaps because myelin endocytosis capacity is already saturated. However, myelin endocytosis was significantly increased in pro‐inflammatory microglia and even in control microglia in the presence of IVM (Fig 8E). Regarding myelin degradation, long‐term treatment with ivermectin potentiated myelin degradation in control (Fig EV5), pro‐inflammatory, and anti‐inflammatory microglia (Fig 8F). The effect of IVM in myelin phagocytosis was absent in P2X4^−/− microglia (Fig EV5). Previous studies have described P2X4R to be located intracellularly in late endosomes and lysosome membranes where they form functional ATP‐activated cation channels regulated by luminal ATP in a pH‐dependent manner (Huang et al, 2014). We observed that IVM induced lysosome fusion, an effect dependent on P2X4R (Fig EV5). Because fusion of lysosomes with late endosomes generates acidic endolysosomes with high cathepsin activity for phagocytic degradation, we then measured intralysosomal pH in microglia. Anti‐inflammatory microglia showed an acidic shift of lysosomes, which correlates with its higher phagocytic capacity. Moreover, IVM treatment (3 μM, 16 h) induced a significant acidification of lysosomes (Fig EV5). Altogether, data suggested that P2X4R potentiation by IVM could favor microglia switch necessary to efficient remyelination.

The therapeutic potential of ivermectin in MS could be compromised by its possible influence in neuropathic pain. Thus, a recent study reported the anti‐allodynic effect of a novel P2X4R antagonist in mice with traumatic nerve damage, although the antagonist did not affect acute nociceptive pain and motor function (Matsumura et al, 2016). We therefore measured the effects of IVM in mechanical allodynia at different stages of EAE before the appearance of severe motor deficits. Withdrawal thresholds were significantly diminished in both vehicle‐ and IVM‐treated mice at EAE onset and before EAE peak relative to baseline responses, but there were no significant differences between the two cohorts at any stage (Fig 8G).

All experiments were performed according to the procedures approved by the Ethics Committee of the University of the Basque Country (UPV/EHU). Animals were handled in accordance with the European Communities Council Directive. Mice were kept under conventional housing conditions (22 ± 2°C, 55 ± 10% humidity, and 12‐h day/night cycle) at the University of the Basque Country animal facilities. All possible efforts were made to minimize animal suffering and the number of animals used. Generation of P2X4^−/− mice was described previously (Sim et al, 2006).

EAE was induced in 8‐ to 10‐week‐old female P2X4^−/− (backcrossed for > 15 generations onto C57Bl/6 background; Sim et al, 2006) and their wild‐type littermate C57Bl/6 mice by subcutaneous immunization with 300 μl of myelin oligodendrocyte glycoprotein 35–55 (MOG; 200 μg; Sigma) in incomplete Freund's adjuvant supplemented with 8 mg/ml Mycobacterium tuberculosis H37Ra. Pertussis toxin (500 ng; Sigma) was injected on the day of immunization and again 2 days later. Motor function was recorded daily and scored from 0 to 8 following standard scales (Matute et al, 2007). Mice were daily treated with TNP‐ATP (10 mg/kg, i.p.; Tocris), ivermectin (1 mg/kg, i.p.; Sigma), or vehicle from EAE onset to the end of the experiment except for the experiments designed to check the effect of TNP‐ATP on immune priming, where TNP‐ATP was administered daily from day 0 to EAE peak. All mice were randomized before the immunization and before the appearance of EAE symptoms. Conduction velocity of the corticospinal tract was assessed at the end of the experiment in anesthetized mice with tribromoethanol (240 mg/kg, i.p.; Sigma) using stimulatory and recording electrodes placed in the primary motor cortex and in the vertebral canal at the L2 level, respectively (Matute et al, 2007). Neurological score and latency recording were undertaken by readers blinded to the study. Disease phases were assigned according to days after onset as follows: peak, 6–10 days after onset; and recovery, score stabilized and 18–30 days after onset.

Mechanical allodynia was assessed by an e‐VF Electronic von Frey aesthesiometer (Ugo Basile SRL) at different stages of EAE: before immunization, EAE preonset (5 dpi), onset (10–11 dpi), and before peak (15 dpi). Mice, placed upon an elevated wire mesh surrounded by a Perspex box, were exposed to increasing mechanical pressure to the plantar hind paw through a metal filament. Withdrawal threshold was measured automatically from the initiation of mechanical stimulus to withdrawal of the paw three times in both left and right hind paws separated by at least 10 min between each stimulus. Mean results for each animal were calculated.

Cells were fixed in 4% PFA in PBS for 20 min and processed for conventional immunocytochemistry. For tissue, adult mice were deeply anesthetized with chloral hydrate (500 mg/kg, i.p.) and transcardially perfused with 0.1 M sodium phosphate buffer, pH 7.4, followed by 4% PFA in the same buffer. Antibodies used are described in Appendix.

Images were acquired using a Leica TCS STED SP8 confocal microscope or a Zeiss AxioVision microscope with the same settings for all samples within one experimental group. Olig2⁺ cells in corpus callosum and in longitudinal sections of spinal cord and Iba1⁺ cells in longitudinal sections of spinal cord were counted blindly using a 40× objective in an AxioVision microscopy (Zeiss). At least four different fields from three slices per animal were counted from each mouse. To quantify microglia polarization, immunoreactivity of iNOS and MRC1 was calculated with the ImageJ software (NIH) and normalized to the number of cells (eight fields per coverslip from at least four different experiments performed in triplicate). To analyze the effect on OPC differentiation, MBP⁺ cells were counted and the results were expressed in percentage versus total cells (15 fields per coverslip were analyzed by two observers from n = 3 different experiments performed in triplicate). To quantify microglia and oligodendrocyte expression of P2X receptors, regions of interest (ROIs) were generated with ImageJ software (NIH) in IB4⁺ and Olig2⁺ cells (12–15 cells per culture from three independent experiments).

Total protein was extracted from microglia by scraping the cells in SDS/sample buffer. Tissue from cerebellar organotypic slices was directly heated at 100°C in sample buffer (7 min). Samples were loaded and size‐separated by electrophoresis using Criterion TGX Precast 12% gels and transferred to Trans‐Blot Turbo Midi PVDF Transfer Packs (Bio‐Rad, Hercules, USA). Membranes were blocked in 5% skimmed milk and 5% serum in Tris‐buffered saline/0.05% Tween‐20 (TBS‐T) and proteins detected by specific primary antibodies to BDNF (#sc‐547, 1:200; Santa Cruz), to MBP (#SMI‐99P, 1:2,000, Covance), to GAPDH (#MAB374, 1:2,000; Millipore), and to β‐actin (#A2066, 1:1,000; Sigma), followed by secondary peroxidase‐coupled goat anti‐rabbit antibodies (#A6154, 1:2,000; Sigma) or sheep anti‐mouse antibodies (#A6782, 1:2,000; Sigma). After washing, blots were developed using an enhanced chemiluminescence detection kit according to the manufacturer's instructions (SuperSignal West Dura or Femto, Pierce). Images were acquired with a ChemiDoc MP system (Bio‐Rad) and quantified using ImageJ software. Values of BDNF and MBP were normalized to corresponding β‐actin and GAPDH signal, respectively.

Rat and mouse myelin was isolated as previously described (Norton & Poduslo, 1973). Briefly, brain was mechanically homogenized in 0.32 M sucrose and subjected to repeated sucrose gradient centrifugation and osmotic shocks to separate myelin from other cellular components. Myelin was incubated with Alexa 488‐NHS dye (A2000 Life Technologies) for 1 h 45 min at RT in PBS (pH 8). Dyed myelin was dialyzed for removing dye excess, resuspended in PBS (pH 7.4), vortexed for 60 s for fragmentation in homogeneous size aggregates, and added to microglia culture medium (1:100 dilution).

To evaluate myelin endocytosis, microglia were incubated with Alexa 488‐NHS‐labeled myelin for 1 h at 37°C, rinsed, and fixed. To evaluate myelin degradation, cells were subsequently chased at 3 and 6 days for rat microglia and at 3 days for mouse microglia. Cells were fixed with 4% PFA and stained using antibodies to Iba1 (#019‐19741, 1:500; Wako) and Hoechst 33258. Myelin was quantified on Iba1⁺ cells using ImageJ on individual microglial cells outlined with the Iba1 immunostaining as the defining parameter for the ROIs (at least 50 cells were analyzed in each experiment). Identical acquisition parameters were used for image capture of individual experiments.

The measurement of lysosomal/endosomal pH by confocal microscopy is based on the use of the ratio of the pH‐sensitive fluorescein fluorescence to pH‐insensitive rhodamine fluorescence as previously described (Majumdar et al, 2007). The same dye was used to measure lysosome area. Briefly, cells were incubated for 16 h with 5 mg/ml dextran conjugated to both fluorescein and rhodamine (70,000 mol. wt.; ThermoFisher), washed thoroughly, and then further incubated for 2 h to chase dextran into lysosomes. Cells were then examined by confocal imaging at 37°C. Lysosome fluorescence and lysosome area were quantified in defined ROIs corresponding to individual lysosomes using Fiji software. For pH measurements, the ratio of fluorescein to rhodamine fluorescence was determined. For all experimental sets, cross‐talk of the fluorophores was negligible. Calibration curves were generated after fixing and equilibrating the fluorescein–rhodamine–dextran‐loaded cells to a range of buffer pH values. We quantified about 40 ROIs per field (63× photographs) from at least 10 fields per condition in two independent experiments performed in duplicate.

Cortical slices (300 μm thick) were prepared from the brain of P15–P20 PLP‐DsRed and CXCR3‐GFP mice to record oligodendrocytes and microglia, respectively. Slices were obtained in ice‐cold solution containing (in mM): 215 sucrose, 2.5 KCl, 26 NaHCO₃, 1.6 NaH₂PO₄, 1 CaCl₂, 4 MgCl₂, 4 MgSO₄, 20 glucose, and 1.3 ascorbic acid bubbled with 95% O₂/5% CO₂, pH 7.4. Slices were then stored for at least 1 h at 32°C in artificial cerebrospinal fluid (aCSF) that contained (in mM) 124 NaCl, 2.5 KCl, 25 NaHCO₃, 1.2 NaH₂PO₄, 1.25 CaCl₂, 2.6 MgCl₂, and 10 glucose. Standard whole‐cell recordings of microglia and oligodendrocytes (Vh = −70 mV) were performed at 37°C on a Leica microscope (Leica DM LFSA) using the MultiClamp 700B amplifier (Axon). Recordings were low‐pass‐filtered at 2 kHz, digitized at 5 kHz, and stored as data files on a computer using the pClamp 8.2 program (Axon Instruments, CA) for later analysis. The extracellular bath solution contained (in mM) 124 NaCl, 26 NaHCO₃, 1 NaH₂PO₄, 2.5 KCl, 2 MgCl₂, 2.5 CaCl₂, and 10 glucose, bubbled with 95% O2/5% CO₂, pH 7.4. ATP was applied through a 5242 microinjector (Eppendorf) in divalent cation‐free extracellular solutions to maximize purinergic currents. Patch clamp pipettes (3–5 MΩ) were filled with a solution containing (in mM) 135 KCl, 4 NaCl, 0.7 CaCl₂, 4 MgATP, 10 HEPES, 10 BAPTA, and 0.5 Na₂‐GTP, pH 7.3.

For flow cytometric analyses, cells were stained in FACS buffer containing 0.1% BSA and 1 mM EDTA with fluorochrome‐conjugated monoclonal antibodies. Details about antibodies used are described in Appendix. Stained cells were washed and resuspended in 300 μl FACS buffer. Cells were measured using a BD LSRFortessa, and data were analyzed with FlowJo software (Tree Star).

For microglia sorting, samples were sorted using CD11b (#101205, 1:100; Biolegend), CD45 (#103134, 1:100; Biolegend), Ly6C (#128012, 1:100; Biolegend), Ly6G (#127622, 1:100; Biolegend), and CCR2 (#FAB5538, 1:100; R&D Systems) to distinguish between resident microglia (CD11b⁺/CD45^low) and invading macrophages (CD11b⁺/CD45^high; Szulzewsky et al, 2015). Cells were sorted using a FACSAria IIIu (BD Bioscience).

Total RNA from control and EAE lumbar spinal cords and from microglia cultures was isolated using TRIzol (Invitrogen) according to the manufacturer's instructions. RNA from sorted microglia was isolated using RNeasy Plus Micro Kit (Qiagen). Subsequently, cDNA synthesis was conducted using SuperScript III retrotranscriptase (200 U/μl; Invitrogen) or AffinityScript Multiple Temperature cDNA Synthesis Kit (Agilent Technologies Inc.) and random hexamers as primers (Promega). cDNA from EAE experiments was processed via Fluidigm real‐time PCR analysis and GenEx software, and the results were depicted as relative gene expression according to the ΔΔC_q method ( $2^{-\mathrm{\Delta }\mathrm{\Delta }{C}_{\mathrm{t}}}$) and expressed in base 2 logarithmic scale. For in vitro experiments, real‐time quantitative PCRs were performed with SYBR Green using a Bio‐Rad CFX96 Real‐Time PCR Detection System as described previously (Domercq et al, 2016).

[¹⁸F]BR‐351, the matrix metalloproteinase inhibitor used for PET imaging (¹⁸F‐MMPi), was prepared from its tosylate precursor according to a previously reported procedure (Wagner et al, 2011) using a TRACERlab FX_FN synthesis module (GE Healthcare). Radiochemical yields (non‐decay‐corrected) were in the range 12–16%, and radiochemical purity was always > 95% at the time of injection. Naïve control mice and EAE mice ± TNP‐ATP (treatment from day 0 to day 17 postimmunization) at the peak of the EAE were subjected to μPET imaging using the radiotracer [¹⁸F]BR‐351 for assessment of MMP activity. Thirty‐minute static μPET scans were acquired for 100 min after intravenous injection of 10 MBq. CT acquisitions were performed to provide anatomical information and the attenuation map for reconstruction using 2DOSEM. PET images were co‐registered to the anatomical data of the CT of the same mouse, and volumes of interest (VOIs) were manually drawn in the inner part of the brain and spinal cord. The PET signal was expressed as percentage of injected dose per gram of tissue (%ID/g).

Data are presented as mean ± s.e.m. with sample size and number of repeats indicated in the figure legends. Comparisons between two groups were analyzed using paired Student's two‐tailed t‐test for data coming from in vitro experiments and unpaired Student's two‐tailed t‐test for data coming from in vivo experiments, except in MOG‐EAE experiments where statistical significance in neurological score was determined by Mann–Whitney U‐test. Comparisons among multiple groups were analyzed by one‐way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison tests for post hoc analysis. Statistical significance was considered at P < 0.05.

Organotypic cultures, primary cultures, electrophysiology protocols, and EAE experiments in wild‐type or P2X4^−/− mice were approved by the Comité de Ética y Bienestar Animal (Animals Ethics and Welfare Committee) of the University of the Basque Country. All the experiments were conducted in accordance with the Directives of the European Union on animal ethics and welfare.