Microglia control the spread of neurotropic virus infection via P2Y12 signalling and recruit monocytes through P2Y12-independent mechanisms

Neurotropic herpesviruses can establish lifelong infection in humans and contribute to severe diseases including encephalitis and neurodegeneration. However, the mechanisms through which the brain’s immune system recognizes and controls viral infections propagating across synaptically linked neuronal circuits have remained unclear. Using a well-established model of alphaherpesvirus infection that reaches the brain exclusively via retrograde transsynaptic spread from the periphery, and in vivo two-photon imaging combined with high resolution microscopy, we show that microglia are recruited to and isolate infected neurons within hours. Selective elimination of microglia results in a marked increase in the spread of infection and egress of viral particles into the brain parenchyma, which are associated with diverse neurological symptoms. Microglia recruitment and clearance of infected cells require cell-autonomous P2Y12 signalling in microglia, triggered by nucleotides released from affected neurons. In turn, we identify microglia as key contributors to monocyte recruitment into the inflamed brain, which process is largely independent of P2Y12. P2Y12-positive microglia are also recruited to infected neurons in the human brain during viral encephalitis and both microglial responses and leukocyte numbers correlate with the severity of infection. Thus, our data identify a key role for microglial P2Y12 in defence against neurotropic viruses, whilst P2Y12-independent actions of microglia may contribute to neuroinflammation by facilitating monocyte recruitment to the sites of infection. Electronic supplementary material The online version of this article (10.1007/s00401-018-1885-0) contains supplementary material, which is available to authorized users.


Other Supplementary Materials for this manuscript includes the following:
Supplementary Videos 1-11 Supplementary Figure 1. The impact of microglia depletion on the spread of virus infection in the brain. Bartha-DupGreen (BDG) was injected intraperitoneally or into the epididymal white adipose tissue (left side) to initiate retrograde transsynaptic spread of infection into the CNS, with or without the selective depletion of microglia with PLX5622. The propagation of infection in the brain was assessed 5 days after infection. Note the marked increase of PRV-immunopositive neurons (cyan) in the brain in the absence of microglia. This includes the spread of infection to higher cortical areas, including the piriform cortex, the insular cortex and the primary motor cortex after intraperitoneal injection. Due to the primarily sympathetic autonomic innervation of the white adipose tissue, the infection affects mostly the piriform cortex in microglia depleted animals, whereas the virus does not the reach the cerebral cortex in control mice. 3Vthird ventricle. Maximum Intensity Projection from confocal Z stack of 5 steps were made with 3,38 µm (cortex), 0,83 µm (PVN), 4,05 µm (brainstem) step size. Scale bars: 50 µm.

Supplementary Figure 2. Pseudorabies virus infects neurons but does not induce productive infection in microglial cells in vivo. a) PRV-immunolabeling is completely absent in virus-free animals. Transmission electron microscopic (TEM) images
showing the presence of anti-PRV immunogold labeling in infected, and its complete absence in noninfected animals. Microglial processes (dark DAB-precipitate, yellow pseudocolor) are found in the vicinity of neurons (cytoplasm: cyan pseudocolor, nucleus: green psudocolor). The infected neuron contains a high level of anti-PRV immunogold particles (black grains), viral capsids and mature virions (insert). The non-infected animal is void of viral particles and PRV-immunogold particles, confirming the specificity of the antibody. Scale bar: 1 µm. b) TEM images show that microglial processes (dark DAB-precipitate, yellow pseudocolor) find and contact PRV-infected neurons even at the early phases of infection, when only low levels of anti-PRV immunogold labeling is present (cytoplasm: cyan pseudocolor, nucleus: green psudocolor). Arrowheads mark sites of microglial contacts, arrows point to a synapse on the dendrite of an infected neuron. Scale bar: 300 nm on left panel, 200 nm on right panel. c) TEM pictures (first panel) show that DAB-positive microglia (yellow pseudocolor) completely lack viral capsids and anti-PRV immunogold particles, in contrary to PRV-infected neurons (cyan pseudocolor). Confocal pictures (second panel) showing that microglia do not express the immediate early GFP marker, in contrast to neurons, where a strong GFP expression can be seen in both the cell bodies and neurites. Scale bars: 1 µm on TEM images, 50 µm, 10 µm on the confocal panels. Neuronal cultures were infected with a low titre of BDG to induce slow propagation of infection enabling the investigation of apoptotic and dead cells by using neuron-specific live annexin V staining and Propidium iodide (PI), respectively. Infected neurons show neither annexin V binding nor PI uptake until several hours following the appearance of the immediate early reporter protein, GFP. Neuron-specific live annexin V staining was first detected in neurons with decaying GFP expression. Note GFP high /annexin -(asteriks) and GFP low /annexin + (arrowheads) cells. PI entered cells only with intense annexin V labelling and little or no detectable GFP expression (arrow) indicating advanced infection. Immunostaining of the same cell culture after 32h postinfection revealed viral structural proteins (PRV) both in the GFP high and GFP low cells, indicating that both immediate-early reporter protein expression and viral protein expression precede the onset of infection-induced neuronal apoptosis. The images were captured with a Zeiss Axiovert epifluorescent microscope. Scale bar: 50 μm. b) Propidium iodide (PI, red) uptake was not altered in infected neurons 16 hours after infection. This particular PI uptake experiment was performed in parallel cultures to those used for HPLC analysis (Fig4.a,b). Scale bar: 50 μm. Figure 7. P2Y12 is expressed by Iba1-positive microglia in the brain. P2Y12 immunofluorescence precisely identifies microglia in the brain that are co-expressing the microglia/macrophage marker Iba1. Note that virtually all microglia are P2Y12-positive. P2Y12 immunostaining is absent in P2Y12 -/mice. P2Y12 immunofluorescence co-localises with Iba1 and GFP in Cx3Cr1 +/gfp microglia recruited to infected neurons expressing the immediate early marker, DSRed. Data are expressed as mean ± s.e.m. n.s.not significant, Scale bar: 100 µm (top), 50 µm (bottom). Figure 8. The amount of disintegrated and dead cells is markedly increased in the brain in microglia-depleted and P2Y12 -/mice. a) Cresyl violet staining was performed on brain sections derived from control, P2Y12 -/and microglia-depleted mice 6 days after BDG infection. The number of disintegrated cells were increased (b), while the number of intact cells were markedly decreased (c) in the hypothalamus (PVN) of P2Y12 -/and microglia-depleted mice compared to control animals. d) The number of morphologically intact cells (based on cresyl violet staining) show negative correlation with the number of intact, PRV-immunopositive cells (as presented in Fig.2a). e) The number of disintegrated, PRV-positive cells in Fig.2a correlates with morphologically damaged cells based on cresyl violet staining. b,c, one-way ANOWA, followed by Tukey's post hoc test, b, *P<0.05 Control vs P2Y12 -/-, **P<0.01 Control vs Depleted. c, one-way ANOWA, followed by tukey's post hoc test, **P<0.01 Control vs P2Y12 -/-; P2Y12 -/vs Depleted, ****P<0.0001 Control vs Depleted. Scale bar: 10 μm. Figure 9. Selective depletion of microglia does interfere with main blood cell populations. Cx3Cr1 +/gfp mice were fed a PLX5622 chow diet for 21 days to deplete microglia. On the 16th day of the diet, a group of mice were injected with Bartha-DupGreen (BDG). Blood samples were collected from control mice and from mice 5 days after virus infection with or without microglia depletion, and labelled with mixtures of specific antibodies against leukocyte markers followed by flow cytometric analysis. Total blood cell counts were calculated by using 15 µm polystyrene microbeads (Polybead Microspheres, 18328-5). The number of CD8 -(P5 gate) T lymphocytes, CD11b + Ly6c + SSC high Ly6G + granulocytes (P8 gate), CD11b + Ly6c high Cx3Cr1 + (Ly6G -) monocytes (P7 gate) and MHCII + CD11b -B cells (P18 gate) were no different between control, depleted, control infected and depleted infected animals. In association with the profoundly increased CNS infection in microglia depleted mice, numbers of CD3 + CD8 + T lymphocytes (P4 gate) and CD11b + Ly6c + SSC high Ly6G + granulocytes (P8 gate) showed significant difference between depleted and depleted virus infected animals. Data are expressed as mean ± s.e.m. one-way ANOVA, n=4. n.s. -not significant. Figure 10. Selective depletion of microglia does interfere with leukocyte numbers in the spleen. Cx3Cr1 +/gfp mice were fed a PLX5622 chow diet for 21 days to deplete microglia. On the 16 th day of the diet mice were injected with Bartha-DupGreen (BDG). Blood samples were collected from control mice and from mice 5 days after PRV infection, and labelled with mixtures of specific antibodies against leukocyte markers followed by flow cytometric analysis. Numbers of CD8 + T cells (P3 gate), CD8 -T cells (P4 gate) and MHCII + B cells (P9 gate) were not different between control, depleted, control infected and depleted infected animals. Spleen macrophages (Cx3Cr1-gfp + F4/80 + cells, P8 gate), granulocytes (P3* gate, CD11b + , Ly6c + , SSC high cells) and monocytes (P4# gate, CD11b low Ly6c vhigh cells) were not different between control, depleted, control infected and depleted infected mice. Data are expressed as mean ± s.e.m. One-way ANOVA, n=4. Figure 11. Virus infection and selective manipulation of microglia do not alter the numbers of T lymphocytes in the brain. Cx3Cr1 +/gfp mice were fed a PLX5622 chow diet for 21 days to deplete microglia. On the 16 th day of the diet mice were injected with Bartha-DupGreen (BDG), to assess the role of microglia in inflammatory cell recruitment into the brain. Brain cells were isolated and T lymphocytes labelled with CD8, CD3 antibodies followed by flow cytometric analysis. Numbers of CD8+ T cells (P9 gate) were not different between control, depleted, control infected and microglia depleted infected animals. Data are expressed as mean ± s.e.m. One-way ANOVA, n=4. Figure 12. Deficiency for P2Y12 does not influence infiltration of leukocytes into the brain upon virus infection. a, Flow cytometric dot plots showing infiltrating monocytes (P5) in the brain. b, Microglial numbers do not change in response to viral infection compared to control conditions. c, In response to viral infection monocytes infiltrate into the brain in P2Y12 -/-mice. Data expressed as mean±s.e.m. b,c, n.s. not significant c, ****p<0.0001 n=10 unpaired t-test.

Supplementary figure 13. Changes in inflammatory cytokines and chemokines in response to virus infection and selective elimination of microglia.
Left, in vitro data: Cytokine and chemokine levels were measured in the conditioned medium of microglial cell cultures by cytometric bead array after LPS treatment or exposure to PRV. Samples were collected 24 hours after infection. Right, in vivo data: Cytokine and chemokine levels were measured from homogenates of hypothalamic brain tissues of control and microglia-depleted mice infected with PRV in a retrograde transneuronal manner, 5 days prior to sample collection. All measurements were performed by cytometric bead array. One-way ANOVA, followed by Tukey's post-hoc test (left) or unpaired t test (right). n=5-6. Data are expressed as mean ± s.e.m.

Supplementary figure 14.
Microglia express P2Y12 in the human brain. a) Colocalization of microglial P2Y12 and Iba1 on perfusion-fixed, free-floating brain sections (frontal cortex). b) Proportion of P2Y12-positive microglia among Iba1+ cells. c) Immunohistochemistry showing Tmem119-positive and P2Y12-positive microglia in paraffinembedded brain sections (frontal cortex). d) Quantification of Tmem119-positive and P2Y12positive microglia in the brain. All samples are derived from patients with no known neurological disease.

Supplementary figure 15. Advanced HSV-1 infection is associated with marked neuroinflammatory responses in the human brain.
Recruitment of numerous CD45positive cells and CD68-positive brain macrophages is seen at sites of advanced virus infection characterized by large number of HSV-1-positive cells. Scale bar: 50μm. Figure 16. Numbers of Cx3Cr1 +/gfp microglia are reduced in areas of advanced PRV infection (heavy virus load) compared to that seen in uninfected (control) mice. Unpaired t test, n=5-6. Data are expressed as mean ± s.e.m.

Supplementary Video 8. Trajectories of P2RX7 -/microglial cells migrating for a 20 h period in untreated (left) or BDG virus infected (right) astroglial cultures.
For better visibility 20 randomly selected individual trajectories were centered to start from the origin on each side and superimposed with color codes changing from red to blue as time elapses. Note the more localized migration pattern of microglial cells in virus-infected culture on the right.

Supplementary Video 9. Trajectories of P2RY12 -/microglial cells migrating for a 20 h period in untreated (left) or BDG virus infected (right) astroglial cultures.
For better visibility 20 randomly selected individual trajectories were centered to start from the origin on each side and superimposed with color codes changing from red to blue as time elapses.

Mice
Mice were bred and genotyped at the SPF unit of the Animal Care Unit of the Institute of Experimental Medicine (IEM HAS, Budapest, Hungary) as described earlier [12,17]. Mice had free access to food and water and were housed under light-, humidity-and temperaturecontrolled conditions.
Selective elimination of microglia from the brain Mice were fed PLX5622 for 3 weeks, to eliminate microglia from the brain. We could not observe any sign of physiological illness (alterations in food intake, weight, physical appearance) or behavioural changes (social interactions, exploration) during the diet period, in accordance with other studies [10,17].

Neurotropic herpesvirus infection
Mice were randomly assigned to experimental groups and were injected either intraperitoneally or directly into the epididymal white adipose tissue with a genetically modified PRV-Bartha derivative, PRV-Bartha-Dup-Green (BDG) [3] to induce retrograde transsynaptic infection in the brain. To enable a defined, exclusively retrograde transsynaptic spread of infection in vivo, the neuroinvasiveness of BDG was modified by insertion of a GFP gene expression cassette to the putative antisense promoter (ASP) located at the repeated invert region of the virus. This construct also allows time-mapping the spread of infection in individual cells [3,4,7]. The activity of the strong immediate-early promoter/enhancer of the human cytomegalovirus is independent from the viral protein, enabling very early stages of infection to be detected by GFP expression. PRV genes encoding structural proteins are driven by late promoters, resulting in the appearance of the virus proteins in the late stage of the infection. In a set of studies, mice were infected with BDG on 16 th day of PLX5622 diet to assess the effect of microglia depletion on central propagation of virus infection. For in vivo two-photon imaging, Cx3Cr1 GFP/+ mice were infected with PRV-Bartha-DupDsRed (BDR) [2] enabling the co-detection of infected neurons with microglia. The virus was grown in porcine kidney (PK)-15 cells to a titer of 6x10 8 plaque-forming units (PFU)/ml. After virus injection (10 μl intraperitoneally or into the epididymal white adipose tissue), mice were let to survive for 6 or 7 days and were regularly monitored for neurobehavioral symptoms.

Tissue processing and immunostaining
Mice were transcardially perfused with saline followed by an ice-cold 4% PFA solution.

Immuno-electron microscopy
For combined immunogold-immunoperoxidase stainings, animals were perfused with 4% paraformaldehyde (PFA) and 0.1% glutaraldehyde in 0.1 M phosphate buffer (PB) pH 7.4 for 40 minutes and 50 μm thick sections were cut. Sections were washed in PB, treated with 0.5% sodium-borohydride for 5 minutes, and further washed in PB and TBS. This was followed by incubation in 1% human serum albumin (HSA, #A1653-5G, Sigma-Aldrich), diluted in TBS. Then the sections were incubated for 48 hours in the following solutions of primary antibodies diluted in TBS: either guinea-pig anti-Iba1 (1:500, #234 004, Synaptic Systems) or chicken anti-eGFP (in case of the Cx3CR1 GFP/+ mice, 1:1000, #A10262, ThermoFisher Scientific) was mixed with rabbit anti-PRV (1:500, from Lynn Enquist). To eliminate non-specific binding, the anti-PRV antibody solution was incubated with control brain sections for one day before use. After repeated washes in TBS, sections were treated with blocking solution (Gel-BS) containing 0.5% cold water fish skin gelatin (#900.033, Aurion, The Netherlands) and 0.5% HSA in TBS for 1 h. This was followed by 24 h incubation in the following solutions of secondary antibodies diluted in Gel-BS: 1. Sections were washed in TBS, and tris-buffer (TB) pH 7.6, and the immunoperoxidase reaction was developed using 3,3-diaminobenzidine as chromogen (DAB, #32750-25G-F, Sigma-Aldrich). After repeated washes in TBS and enhancement conditioning solution (ECS, #500.055, Aurion), gold particles were intensified using the silver enhancement solution (SE-EM, #500.044, Aurion) for 60 minutes at room temperature. After subsequent washes, sections were treated with 0.5 % osmiumtetroxide in PB for 20 minutes. Then sections were dehydrated in ascending ethanol series and acetonitrile, and embedded in epoxy resin (Durcupan, #44610-1EA, Sigma-Aldrich). During dehydration sections were treated with 1% uranylacetate in 70% ethanol for 20 minutes. After polymerization, 70 nm thick sections were cut on a Leica EM UC6 ultramicrotome (Nussloch, Germany), and picked up on formvarcoated single-slot copper grids. The sections were examined using a Hitachi H-7100 electron microscope (Tokyo, Japan) and a side-mounted Veleta CCD camera (Olympus Soft Imaging Solutions).

Correlated confocal laser-scanning microscopy, electron microscopy and electron tomography
Before the immunofluorescent staining, the 50 µm thick brain sections from BDG-injected Cx3CR1 GFP/+ mice were washed in PB, treated with 0.5% sodiumborohydride for 5 minutes, and further washed in PB and TBS. This was followed by blocking for 1 hour in 1% human serum albumin (HSA). After this, sections were incubated in mixtures of primary antibodies: chicken anti-eGFP (1:4000, #A10262, ThermoFisher Scientific), rat-anti-mouse CD68 (1:250, #MCA1957GA, AbD Serotec) and rabbit anti-PRV (1:500, gift of Lynn Enquist). To eliminate non-specific binding, the anti-PRV antibody solution was incubated with control brain sections for one day before use. After repeated washes in TBS, sections were treated with blocking solution containing 0.5% cold water fish skin gelatin and 0.5% HSA in TBS for 1 h. After incubation, sections were washed in TBS, and incubated overnight at 4 °C in the mixture of biotinylated goat-anti-chicken (1:500, #103-065-155, Jackson ImmunoResearch), Alexa 594 conjugated donkey-anti-rat (1:500, #A21209, Invitrogen) and Alexa 647 conjugated donkey-anti-rabbit (1:500, #711-605-152, Jackson ImmunoResearch) diluted in GelBS. Secondary antibody incubation was followed by washes in TBS and PB. Sections were mounted in PB, coverslipped, sealed with nail-polish. Immunofluorescence was analysed using a Nikon Eclipse Ti-E inverted microscope (Nikon Instruments Europe B.V., Amsterdam, The Netherlands), with a CFI Plan Apochromat VC 60X water immersion objective (NA: 1.2) and an A1R laser confocal system. We used 488, 561 and 642 nm lasers, and scanning was done in line serial mode. Image stacks were obtained with NIS-Elements AR software with a pixel size of 50x50 nm in X-Y, and 150 nm Z-steps. Stacks were deconvolved using Huygens Professional software (www.svi.nl), and 3D-reconstruction was performed using the IMOD software package [16]. After imaging, sections were washed 0.1 M PB, TBS and incubated in ABC (1:300, Vectastain Elite ABC HRP Kit, #PK-6100. Vector Laboratories) diluted in TBS. Sections were washed in TBS, and tris-buffer (TB) pH 7.6, and the immunoperoxidase reaction was developed using 3,3-diaminobenzidine as chromogen. After subsequent washes, sections were treated with 0.5 % osmiumtetroxide in PB for 20 minutes. Then sections were dehydrated in ascending ethanol series and acetonitrile, and embedded in epoxy resin. During dehydration sections were treated with 1% uranylacetate in 70% ethanol for 20 minutes. After polymerization, 70 or 150 nm thick sections were cut on an ultramicrotome, and picked up on formvar-coated single-slot copper grids. The sections were examined using a Hitachi H-7100 electron microscope and a side-mounted Veleta CCD camera. For the electron tomographic investigation, we used the 150 nm thick sections. Grids were put on drops of 10% HSA in TBS for 10 minutes, dipped in distilled water (DW), put on drops of 10 nm gold conjugated Protein-A (#AC-10-05-05, Cytodiagnostics) in DW (1:3), and washed in DW. Finally, we deposited 5-5 nm thick layers of carbon on both sides of the grids. Electron tomography was performed using a Tecnai T12 BioTwin electron microscope equipped with a computer-controlled precision stage and an Eagle™ 2k CCD 4 megapixel TEM CCD camera. Acquisition was controlled via the Xplore3D software (FEI).
Regions of interest were pre-illuminated for 4-6 minutes to prevent further shrinkage. Tilt series were collected at 2 degree increment steps between -65 and +65 degrees at 120 kV acceleration voltage and 23000x magnification with -1.6 --2 µm objective lens defocus. Reconstruction was performed using the IMOD software package. Isotropic voxel size was 0.49 nm in the reconstructed volumes. Segmentation has been performed on the virtual sections using the 3Dmod software.

Immunohistochemistry and immunofluorescence on post-mortem human brain samples
To investigate microglia recruitment in response to neurotropic virus infection in the human brain, formalin-fixed paraffin-embedded (FFPE) post-mortem brain tissues from five patients Representative pictures were taken on 20x magnification using a Nikon Ni-E C2+ microscope. To assess the proportion of P2Y12-positive microglia, P2Y12 and Iba1 double immunofluorescence has been performed on 50 µm thick freefloating brain sections. Images were captured using a Nikon Eclipse Ti-E inverted microscope (Nikon Instruments Europe B.V., Amsterdam, The Netherlands), with a CFI Plan Apochromat VC 60X water immersion objective (NA: 1.2) and an A1R laser confocal system.

Two-photon imaging
To assess microglia recruitment to infected neurons in the mouse brain in real-time, Cx3CR1 GFP/+ mice were i.p. injected with 10μl of the BDR virus. 7 days after virus injection, cranial window surgery was performed on anaesthetized (using 3% isoflurane for induction, 1.5-2% during surgery) mice and a circular craniotomy (3 mm diameter) was made above motor cortex (Bregma -0,82); on the right hemisphere. A custom made aluminium head plate was fixed to the skull using Cyanoacrylate glue. During skull opening the place of craniotomy was washed continuously with cold Ringer solution. The craniotomy was covered with a circular cover glass and dura mater remained intact under the glass to ensure that microglial activation is not induced [17]. The cover glass was fixed with Paladur mixed Cyanoacrylate.
Measurements were performed on a Femto2D-DualScanhead microscope (Femtonics Ltd., Hungary) coupled with a Chameleon Ultra II laser (Coherent, Santa Clara, USA) [5,13]. The wavelength of the laser was set to 980nm to measure DsRed and GFP signal simultaneously. Excitation was delivered to the sample, and the fluorescent signal was collected using a CFI75 LWD 16XW/0.8 lens (Nikon, 16x, NA 0.8) and then separated using a dichroic mirror (700dcxru) before the two channel detector unit, which was sitting on the objective arm (travelling detector system) as described in detail earlier [5,13]. The dichroic mirror and emission filters (490-550 nm for the green and 570-640 nm for the red channel) was purchased from Chroma Technology Corp. (Vermont, USA). Data acquisition was performed by MES softver (Femtonics Ltd.) A Z-stack from of 32 images (800x800 pixel, field-of-view=210x210 µ, Z stack contained 12 image planes with 4.6 µm step size (range= 80-135 µm from pial surface) was made at every 3 minutes. Two-photon image sequences were exported from MES and analysed using ImageJ.

Neurotropic herpesvirus infection of primary neuronal and glial cultures
The neuronal or astroglia cultures were infected with either PRV-Bartha-Dup-Green (BDG) virus or PRV-Bartha-DupLac (BDL) at a final titer of 2.5x10 5 PFU/ml, as described earlier [11]. The multiplicity of infection (MOI) was ~0,17 PFU/cell. The cells were incubated with the virus containing medium for 1h, at room temperature. In order to remove infective virus particles from the culture medium we washed the cultures at least three times after the 1h viral exposure. For quantification of ATP derivatives released by or retained in control vs. infected cells, embryonic (E16) primary cortical neuronal cultures were used. The cells were infected at DIV9. 16 hours post-infection GFP expression was initiated in most cells, while the cell membrane integrity was not compromised and the overall viability was not reduced based on MTT cell viability tests. The cell culture media and the cell fractions were collected 15 min after media change, as it is described below. For detemination of cytokine production as well as quantitative RT-PCR the following cultures and conditions were used.

Establishment of neuron/microglia and astroglia/microglia co-cultures
In cultures used for time-lapse recordings microglial cells isolated from mixed glial cultures were seeded on top of astrocytic or neuronal cell cultures in 10 000 cell/cm 2 density, immediately after the infection and the subsequent washing steps. If not CX3CR1 GFP/+ microglia were used, the cells were subjected to staining with the orange or red vital CellTracker dyes CMTMR/CMTPX (#C2927, #C34552, ThermoFisher Scientific) prior coculture establishment (30min, 37 o C).

Time-lapse microscopy
Time-lapse recordings were performed on a computer-controlled Leica DM IRB inverted fluorescent microscope equipped with a Marzhauser SCAN-IM powered stage and a 10x N-PLAN objective with 0.25 numerical aperture and 5.8 mm working distance. The microscope was coupled to an Olympus DP70 color CCD camera and a Zeiss Colibri LED epifluorescent illumination system. Cell cultures were kept at 37°C in humidified 5% CO 2 atmosphere in tissue culture grade Petri dishes in a stage-top incubator mounted on the powered stage of the microscope. Stage positioning, focusing, illumination and image collection were controlled by a custom-made experiment manager software on a PC. Phase contrast and epifluorescent images were collected consecutively every 10 minutes from several microscopic fields for durations up to 48 hours. Images were edited using NIH ImageJ software.

Cell motility data analysis
Cell tracking: Images were analyzed individually with the help of custom-made cell-tracking programs (G-track and Wintrack) enabling manual marking of individual cells and recording their position parameters into data files. At the magnification applied, the precision of this tracking procedure is estimated to be 10 μm, comparable to the average cell diameter (10-50 μm). In the following, the position of the i th cell at time t is denoted by x i (t) .
Trajectories: Cell positions, x i (t) , from several microscopic fields were recorded over 24 hours and each cell's trajectory was plotted. Subsequently, the starting positions of trajectories were aligned to the origin (x=0, y=0) and consecutive relative cell positions were plotted and superposed yielding groups of centered trajectories enabling the comparison of cell migration directionality. Trajectories in the monolayer cell cultures studied indicate a persistent random walk behavior with different velocity and directional persistence.
Displacement: The motion of individual cells is often evaluated in terms of average cell displacement, d, over a time period t as: where X i (t) denoting the center of cell i at time t, i  is an average over all possible cells, and t 0 is an arbitrary reference frame of the image sequence analyzed. The empirical d(t) curves indicate a persistent random walk behavior in monolayer cell cultures studied.
Average velocity: Average cell velocity was calculated directly from cell displacements in consecutive steps of cell tracking. The velocity, v i (t), of a given cell i at time t was calculated as where Dt is the difference of two consecutive steps of cell tracking and thus the time resolution data acquisition. To characterize the motility of an ensemble of cells, timedependent average velocity v(t) was calculated as where the summation goes over each N(t) cell being in the cell population. Average velocity, v, was calculated by averaging v(t) over all time steps of cell tracking as where K is the number of time steps of tracking.

Quantification of nucleotides and adenosine
The adenine nucleotides (ATP, ADP, AMP) and adenosine (Ado) were determined in extracts from cells and culture media by using HPLC method. The medium (400 µl) was separated into a cold Eppendorf tube which contained 50 µl of homogenization solution. The cell layer was frozen with liquid nitrogen and extracted in 100 µl volume of ice-cold homogenization solution. The homogenization solution was 0.1 M perchloric acid that contained theophylline (as an internal standard) at 10 µM concentration. The cell extract was centrifuged at 3510 g for 10 min at 0-4 o C and the pellet was saved for protein measurement. Perchloric anion from the supernatant was precipitated by 1 M potassium hydroxide, the precipitate was then removed by centrifugation. The culture media after the acid extraction was centrifuged as described above. The extracted purines were kept at -20 o C until analysis. The adenine nucleotides and adenosine in extracts from cells and culture media were determined by online column switching separation using Discovery HS C18 50 x 2-mm and 150 x2-mm columns. The flow rate of the mobile phases ["A" 10 mM potassium phosphate, 0.25 mM EDTA "B" with 0.45 mM octane sulphonyl acid sodium salt, 8 % acetonitrile (v/v), 2 % methanol (v/v), pH 5.2] was 350 or 450 µl/min, respectively in a step gradient application. The enrichment and stripping flow rate of buffer [10 mM potassium phosphate, pH 5.2] was during 4 min and the total runtime was 55 min. The HPLC system used was a Shimadzu LC-20 AD Analytical & Measuring Instruments System, with an Agilent 1100 Series Variable Wavelength Detector set at 253nm. Concentrations were calculated by a two-point calibration curve using internal standard method. The data are expressed as pmol per mg protein or pmol per mL.

Immunohistochemical staining for NTPDase1
After the fixative was washed out, the coronal brain sections were incubated in blocking solution (5% normal goat serum and 1 mg/ml BSA) for 1 h at 22 ˚C. Incubation in the solution of the polyclonal NTPDase1 antibody (1:500, #rN1-6L, I4), was performed overnight at 4 o C. Following three, 10 min washes in PBS, the sections were incubated with biotinylated secondary antibody for 2 h. The staining was performed with Vectastain ABC Elite kit using DAB as the chromogen. After washing thoroughly with distilled water, sections were postfixed in 1% OsO4, dehydrated in ethanol, stained with 1% uranyl acetate in 50% ethanol for 30 min, and embedded in Taab 812. Negative control experiments were performed using the same protocol but substituting pre-immune serum for the primary antibody. Ultrathin sections were cut using a Leica UCT ultramicrotome (Leica Microsystems, Milton Keynes, UK) and examined using a Hitachi 7100 transmission electron microscope (Hitachi; Tokyo, Japan).

Enzyme histochemistry for detection of ecto-ATPase activity
A cerium precipitation method was used for electron microscopic investigation of ecto-ATPase activity [14]. Briefly, thoroughly washed sections were incubated in a medium containing 1 mM ATP as substrate, 3 mM CeCl3 (precipitating agent for the liberated phosphate), 1 mM levamisole (inhibitor of alkaline phosphatases, Amersham, Poole, UK), 1 mM ouabain (Na+, K+-ATPase inhibitor; Merck, Darmstadt, Germany), 50 mM αβ-methylene ADP (5'-nucleotidase inhibitor) and 5 mM KCl in 70 mM Tris-maleate buffer (pH 7.4) for 30 min at room temperature. Incubation was followed by three rinses in Tris-maleate buffer and washing with distilled water. The tissue blocks were then postfixed, dehydrated and treated and embedded into Taab 812 resin for ultrathin sectioning and microscopic examination as described above. Control reactions were performed without adding the ATP substrate.
Flow cytometric analysis of brain, spleen and blood samples