Oligodendrocytes are the myelinating cells of the CNS and, like neurons, are highly sensitive to ischemic damage. However, the mechanisms underlying cytotoxicity in oligodendrocytes during hypoxic/ischemic episodes are not fully understood. TASK-1 is a K+ leak channel that mediates hypoxic depolarisation in neurons. The expression and function of TASK-1 in oligodendrocytes had not previously been addressed. In this study, we investigate the expression of TASK-1 in oligodendrocytes and its role in white matter ischemic damage. Expression of TASK-1 in oligodendrocytes was investigated in the mouse brain using immunostaining. TASK-1 channel function was identified by established pharmacological and electrophysiological strategies, using the whole-cell patch clamp technique in cell cultures of oligodendrocytes from the optic nerve, a typical white matter tract. The role of TASK-1 in hypoxia was examined in isolated intact optic nerves subjected to oxygen glucose deprivation (OGD). Oligodendrocytes are strongly immunopositive for TASK-1 throughout the brain. Patch-clamp identified functional TASK-1-like leak currents in oligodendrocytes using two recognised means of inhibiting TASK-1, decreasing extracellular pH to 6.4 and exposure to the TASK-1 selective inhibitor anandamide. Incubation of optic nerves with methanandamide, a non-hydrolysable form of anandamide, significantly protected oligodendrocytes against hypoxic disruption and death in OGD. Our data demonstrate for the first time that oligodendrocytes express functional TASK-1 channels and provide compelling evidence they contribute to oligodendrocyte damage in hypoxia. Since oligodendrocyte damage is a key factor in ischemic episodes, TASK-1 may provide a potential therapeutic target in stroke and white matter disease.

Animals

Mice aged postnatal day (P)7–15 were used throughout, and killed humanely, by cervical dislocation, in accordance with the UK Animals (Scientific Procedures) Act, 1986. Wild type mice of the C57BL/6 strain were used, or transgenic mouse strains in which the fluorescent reporters DsRed and eGFP are driven by the oligodendroglial genes PLP1 and Sox10, respectively (Azim and Butt, 2011).

Optic nerve explant cultures

Optic nerve explant cultures were prepared as previously described (Greenwood and Butt, 2003). In brief, optic nerves from P7–10 mice were isolated intact in ice cold artificial cerebrospinal fluid (aCSF) comprising of (in mM): NaCl 133; KCl 3; CaCl2 1.5; NaH2PO4 1.2; MgCl2 1; d-glucose 10; HEPES 8.55, at pH 7.4. Optic nerves were finely chopped with a scalpel to small segments less than 1 mm in length and explants were cultured onto poly-l-lysine/laminin (Invitrogen/Sigma) coated coverslips (1 optic nerve segment per coverslip) in modified Bottenstein and Sato (B&S) medium without thyroid hormones, plus 10 ng/ml recombinant human platelet derived growth factor (PDGF-AA; R&D Systems) and 0.1% gentamicin. After 3–4 days in vitro (DIV) explants were then incubated in B&S with 0.5 mM dibutyryl cyclic adenosine monophosphate (dbcAMP) medium up to 10 DIV to stimulate differentiation.

Immunolabelling

Brain tissue from P15 mice or optic nerve explant cultures were fixed with 4% paraformaldehyde in phosphate buffered saline (PBS, pH7.4); tissue was fixed for either 1 h at room temperature (RT) or overnight at 4 °C, and for 10 min for cultured cells on their coverslips, followed by washes in PBS. For sectioning, brain tissues were placed in cryoprotectant (30% w v− 1 sucrose in PBS) overnight at 4 °C, then embedded in Cryo-M-Bed (Bright Instruments Company Ltd), before rapidly freezing at − 70 °C, sectioning using a cryostat (10 μm; Leica CM3050 S), and transference of sections onto Polysine® coated slides (Thermo-Scientific). Sections and cells were treated the same in subsequent stages. A blocking stage was performed by incubation for 1 h at RT in 5% normal goat serum (NGS) and 0.2% triton-X-100 in PBS. After blocking, samples were incubated overnight at 4 °C with primary antibodies diluted in NGS-PBS: rabbit anti-TASK-1 raised against an intracellular (C-terminal) peptide corresponding to amino acid residues 252–269 of human TASK-1 (Alomone Laboratories), used at 1:200; rat anti-MBP 1:300 (Chemicon); and mouse anti-APC 1:300 (Calbiochem). After washes in PBS, tissues were incubated for 1 h at RT with the appropriate secondary antibodies conjugated with 488Alexafluor or 568Alexafluor (1:500, Molecular Probes). For TASK-1, control experiments were carried out in which tissue was preabsorbed with antigen peptide overnight prior to incubation in the primary antibody; other antibodies have been validated in many studies (Azim and Butt, 2011), and in these cases controls were carried out by omitting primary antibody. Following immunolabelling, coverslips/sections were mounted with Vectasheild® (VectorLabs) and images acquired using a LSM 5 Pascal Axioskop2 confocal microscope (Zeiss), maintaining variables constant between images.

Image analysis

Image analysis was carried out using Volocity 6.1 software (Perkin Elmer). Image acquisition of the different laser lines was performed using multichannel sequential scanning, narrow bandwidths, and minimal laser power and gain to prevent cross-talk between the channels. A pinhole of 1 airy unit or less was used, with an average of 4 scans per image. The number of z-sections and resolution were optimised using the Zeiss acquisition software, and in general z-sections were < 0.75 μm thickness, and approximately 30 z-sections were taken for each cell. For confocal photomicrographs, two-dimensional flattened images of the z-stacks are presented. For colocalisation analyses, the more accurate technique of Barlow and colleagues was used (Barlow et al., 2010), in which the degree of separation between pixels from the red and green channels was determined in single z-sections to provide measurements of signal overlap. First, images were thresholded to separate the positive signal (positive immunolabelling) from background. The threshold was determined by measuring the background intensity value for each channel in negative control sections and setting the threshold as the mean background intensity plus three standard deviations (averaged from a minimum of 6 images). Then, the thresholded Pearson’s correlation coefficient (PCC) and Mandersons’ overlap coefficients (M1 and M2) were determined as previously described (Barlow et al., 2010), using Volocity 6.1 software. The thresholded PCC determines the statistical strength of the linear relationship between fluorescent intensities from the red and green channels, and the M1 and M2 overlap coefficients provide accurate measurements of the true degree of overlap of red and green. A colocalisation channel was generated from the thresholded PCC to illustrate in three-dimensions the voxels in which the two channels overlap with the same intensity.

Recombinant expression

Optic nerve explant cultures were transiently transfected with mTASK-1pTagFP635, in which mouse TASK-1 cDNA is fused to a N-terminal far-red fluorescent protein TagFP635. This was generated by subcloning mTASK-1 from mTASK-1pEXO into the pTagFP635 mammalian expression vector (Clontech). The construct was verified by sequencing and cells in antibiotic free medium were transfected at 8 DIV using Lipofectamine™ 2000 (Invitrogen), following the manufacturer’s guidelines. Cells were fixed 48 h post transfection with 4% paraformaldehyde in PBS, pH7.4 for 10 min with three subsequent washes in PBS prior to mounting with Vectasheild® (VectorLabs) and examined using an LSM 710 Axiovert confocal microscope (Zeiss).

Western blotting

Whole brain and optic nerves (pooled from 3 animals) were removed from P15 mice and tissue was homogenised in buffer containing (in mM): NaCl 12.5; Tris HCl 2, pH 8.0; phenylmethylsulphonylfluoride (PMSF) 0.2, in distilled water with 1x complete mini protease inhibitor cocktail (Roche). Samples were centrifuged at 4 °C, at high speed (14,000 rpm/20,800 rcf) for 4 min and pellets resuspended in PBS. Quantification of protein concentration was carried out using the bicinchoninic acid assay (Sigma) with a standard bovine serum albumin (BSA) concentration curve and UV spectrophotometer absorbance readings at 550 nm. Samples were mixed with Laemmli sample buffer, heated at 95 °C for 10 min with β-mercaptoethanol and 10 μg of protein per lane was loaded for 8% acrylamide sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Proteins were then electrophoretically transferred to a polyvinyllidene difluoride membrane (Amersham) which was then incubated in blocking solution (0.5–5% w.v− 1 dried milk in Tris buffered saline (TBS; 150 mM NaCl, 10 mM Tris pH 7.4) with 0.05% v.v− 1 Tween 20). Incubation in rabbit anti-TASK-1 antibody at 1:1000 (Alomone Laboratories) was carried out overnight at 4 °C, and following washes, the secondary antibody horseradish peroxidase-conjugated swine anti-rabbit (Dako) was added at 1:10,000 for 1 h at RT; controls were preincubated with the competitive peptide from which the TASK-1 antibody was raised. Extensive washing of the membranes in TBS with 0.05% v.v− 1 Tween 20 was performed after each incubation and immunocomplexes were detected using an enhanced chemiluminescence method (Amersham).

Patch clamp

Patch-clamp recordings were performed on oligodendrocytes from optic nerve explant cultures at 10–12 DIV, as described previously (Bay and Butt, 2012). TASK-1 currents were examined using an established protocol adapted from Skatchkov and colleagues (Skatchkov et al., 2006). Patch electrodes (2–8 MΩ) were backfilled with intracellular solution (ICS) containing (in mM): potassium acetate 90; KCl 20; HEPES 40; MgCl2 3; EGTA 3; CaCl2 1, at pH 7.4; plus 3 mM Na2ATP, to block KATP, and 100 μM Spermine (Acros), to block the outward component of Kir currents. The extracellular solution (ECS) was aCSF, plus BaCl2 (100 μM), to block Kircurrents, tetraethyl ammonium chloride (TEA, 3 mM), to block KD and BK channels, and 4-aminopyridine (4-AP, 6 mM), to block transient KAchannels. Blockade of these glial K+ currents reveals the background K2Pcurrents and, at the concentrations used, the K+ channel blockers would not affect TASK-1 currents (Han et al., 2002; Lesage and Lazdunski, 2000; Reyes et al., 1998; Skatchkov et al., 2006). The residual glial cell currents were subject to voltage step protocols, which were then repeated in the presence of the specific TASK-1 channel inhibitor anandamide (10 μM) (Maingret et al., 2001; Plant et al., 2002), altered pHo to 6.4 and 8.4, and increasing [K+]o to 30 mM (substituting NaCl). Cells were held at − 50 mV in voltage clamp before applying the voltage step protocols and currents recorded using a CV 203Bcl Headstage (Axon Instruments), and an Axo-patch 200B amplifier (Axon Instruments), with frequencies > 1 kHz filtered, and digitisation achieved (sampling at 5 kHz) through a DigiData 1440A interface (Axon Instruments). The pCLAMP10 software package (Axon Instruments) was used for data acquisition and analysis, and significance between peak currents determined by t-test, using Prism3.0 (Graphpad).

Oxygen-glucose deprivation

Optic nerves from P11–13 transgenic PLP-DsRed or Sox10-eGFP mice were placed in aCSF for 30 min at 37 °C with 5% CO2. Controls were incubated for a further 1 h in normal aCSF containing glucose with 5% CO2/95% O2. Oxygen-glucose deprivation (OGD) was achieved using the method of Fern and colleagues (Salter and Fern, 2005), by incubating nerves for 1 h at 37 °C in glucose-free aCSF (osmolarity was maintained by replacing glucose with sucrose), and switching the chamber atmosphere to 5% CO2/95% N2. Pharmacological modulators were added to the aCSF solution during OGD: the specific TASK-1 channel inhibitor methanandamide (10 μM, Sigma), a non-hydrolysable analogue of anandamide; or lidocaine (1 μM, Sigma), a non-specific TASK channel inhibitor. Cell death was measured using propidium iodide (PI, 1 μg/μl; Sigma). At the end of experiments, nerves were fixed for 1 h in 4% PFA, whole-mounted with vectashield and examined using a LSM 5 Pascal Axioskop2 confocal microscope (Zeiss). Images were captured at a depth of 15 μm beneath the pial surface, and 15 × 1 μm z-sections were acquired within a constant 20 × 20 μm field of view (FOV). In nerves from PLP-DsRed mice, the cell bodies and primary processes were analysed by cell counts and by measuring pixel intensity in a 20 × 20 × 15 μm region (Salter and Fern, 2005), using Volocity 6.1 software, and maintaining all variables constant between images. Sox10-GFP mice were used to measure cell death, taken as the number of PI +/Sox10 + cells per FOV. Data were expressed as mean ± SEM, where each ‘n’ value represents a separate nerve, and significance was determined by ANOVA and Newman–Keuls multiple comparison post-hoc analysis, using Prism3.0 (Graphpad).

Oligodendrocytes express TASK-1 throughout the brain

Immunohistochemical and western blot analysis of TASK-1 expression was performed in the brain and optic nerves of wild-type mice and transgenic mice in which the myelin gene proteolipid protein (PLP) drives expression of DsRed (Fig. 1). TASK-1 immunostaining had not previously been performed on oligodendrocytes, and so we first validated the TASK-1 immunostaining in the cerebellum, where Purkinje neuron cell bodies and primary dendrites are strongly immunopositive for TASK-1 (Fig. 1A, asterisks) (Leonoudakis et al., 1998; Rusznak et al., 2004). Negative controls for immunostaining consisted of preabsorbing the primary antibody with antigen peptide, and these sections displayed no immunoreactivity (Fig. 1Ai). Secondly, we performed western blot analysis of total protein cell lysates to establish the specificity and expression of TASK-1 channels in the brain and optic nerve, with predicted molecular weights of around 65 kDa, and elimination of the positive bands with the competitive peptide (Fig. 1D). Notably, the presence of TASK-1 channels was not restricted to neurons, and we found prominent TASK-1 immunostaining in oligodendrocytes, which we identified by their expression of the PLP-DsRed promoter (Figs. 1B, C, stars). We also observed TASK-1 immunostaining of astrocytes and Bergmann glia, but not to the same level as oligodendrocytes (not illustrated). Oligodendroglial immunostaining for TASK-1 is most clearly demonstrated in the optic nerve, where oligodendrocyte somata are arranged in discrete interfascicular rows, and there was clear co-localisation between TASK-1 and APC (CC1), a definitive label for differentiated oligodendrocyte somata (Fig. 2A, asterisks). Oligodendrocytes were the cells showing the highest immunoreactivity for TASK-1 in all white matter regions analysed, as illustrated for the corpus callosum (Fig. 1B), cerebellum (Fig. 1C), and optic nerve (Fig. 2A), and to a lesser extent also in the cerebral cortex and hippocampus (not illustrated). No immunoreactivity was observed in oligodendrocytes in negative controls, following preabsorption with the antigen peptide (Fig. 1Bi). Further verification of oligodendroglial expression of TASK-1 was provided in optic nerve explant cultures, using the PLP-DsRed reporter (Fig. 2Bi-iii), and by expression of transfected fluorescently tagged TASK-1 cDNA, which demonstrated oligodendrocytes have the endogenous mechanisms for plasmalemmal expression of TASK-1 (Fig. 2Biv).

Fig. 1

TASK-1 expression in oligodendrocytes of the mouse brain. (A) Immunostaining for TASK-1 in the cerebellum identifies Purkinje neuron somata (asterisks) and their dendritic trees, which serves as a positive control, and TASK-1 immunostaining was absent …

Fig. 2

Localisation of TASK-1 to oligodendrocyte somata. Immunostaining for TASK-1 with APC to identify oligodendroglial somata in the optic nerve (A), shows TASK-1 expression preferentially on interfascicular rows of oligodendrocyte cell bodies (asterisks), …

TASK-1 is highly localised to oligodendrocyte somata

The results indicated that TASK1 is preferentially localised to oligodendrocyte somata compared to the myelinated fascicles, which appeared as largely immunonegative gaps between interfascicular rows of immunopositive oligodendrocyte somata (Fig. 2Ai). Colocalisation of fluorescence labelling is most frequently presented as overlays of red and green channels, with areas of yellow indicating colocalisation (Figs. 2Ai, Bi). However, this simplistic approach can be misleading. We therefore used the more accurate technique of Barlow and colleagues (Barlow et al., 2010), to perform quantification of the degree of colocalisation between TASK-1 immunolabelling and PLP-DsRed on high resolution confocal z-sections (Fig. 2C). The PLP-DsRed transgenic mouse line enables clear identification of oligodendrocyte units, comprising cell body, primary processes and cytoplasmic elements of internodal myelin sheaths within dense white matter myelinated tracts. Images were thresholded, using the background intensity measured in negative control sections (averaged from a minimum of 6 images), and the thresholded Pearson’s correlation coefficients (PCCs) were determined in single z-sections, to provide accurate and unbiased measurements of the strength of the linear relationship between fluorescent intensities from the red and green channels for each individual pixel (Fig. 2Cii). The thresholded PCC was used to generate colocalisation channels, where only individual voxels in which the two channels overlap with the same intensity appear yellow, and these illustrate in three-dimensions the degree of true colocalisation of TASK-1 on oligodendrocytes and myelin (Fig. 2Ci). The thresholded PCC allows for both positive and negative values to be calculated without bias, where + 1 is a perfect overlap of pixels from the two channels and − 1 indicates no relationship (displacement of at least 10 pixels between the two channels), and low positive or negative values indicate some overlap and displacement between pixels (Barlow et al., 2010). The mean positive unbiased PCC value demonstrated a statistically significant colocalisation of TASK-1 on oligodendrocytes, with a PCC of 0.15 ± 0.02 on the somata and 0.02 ± 0.02 on myelin sheaths (Fig. 2Cii), indicating < 3 pixels displacement between TASK-1 and PLP-DsRed on somata, and < 4 pixels on myelin. Notably, the DsRed reporter is cytoplasmic and at the borders of the cell somata there was perfect overlap between the red and green channels, indicating plasmalemmal expression of TASK-1 (appears yellow in Fig. 2Ci, whereas cytoplasm appears red). The M1 and M2 overlap coefficients provide accurate measurements of the overall degree of overlap of red and green channels throughout the FOV. Hence, the M1 overlap coefficient shows 55.0 ± 3.2% of PLP-DsRed voxels were also TASK-1 positive, indicating very strong expression of TASK-1 in oligodendrocytes, bearing in mind the DsRed reporter is cytoplasmic, whereas TASK-1 is plasmalemmal. The M2 overlap coefficient shows that 50% of TASK-1 immunopositive voxels were not localised to PLP-DsRed, indicating non-oligodendroglial expression of TASK-1, as demonstrated for neurons and astrocytes (Fig. 1). Overall, colocalisation analysis confirms the qualitative observation that TASK-1 is preferentially, but not exclusively, localised to oligodendrocyte cell bodies, and to a lesser extent on myelin.

TASK-1-like pH sensitive ‘leak’ currents in oligodendrocytes

Patch clamp examination of oligodendroglial TASK-1 channels was performed in optic nerve explant cultures using a protocol based on previous studies on Müller glia, in which classical glial K+ channels are blocked and the residual leak current was analysed for TASK-1 channel characteristics (Skatchkov et al., 2006). TASK-1 is relatively insensitive to Ba2 +, Cs+, TEA, and 4-AP (Han et al., 2002; Lesage and Lazdunski, 2000), but is specifically and directly blocked by the endocannabinoid anandamide (Maingret et al., 2001; Plant et al., 2002). A further notable property of TASK-1 channels is they are inhibited by extracellular acidosis with a pK of ~ 7.4, squarely in the physiological range (Lesage and Lazdunski, 2000; Plant et al., 2002). We examined oligodendrocyte whole cell currents for these properties, therefore, using a cocktail of both intracellular (ICS) and extracellular (ECS) K+channel blockers: BaCl2 (100 μM) to inhibit Kir; TEA (3 mM) to block KDand BK; 4-AP (6 mM) to block KA; Na2ATP (3 mM, intracellular) to inhibit KATP; and spermine (250 μM, intracellular) to block outward Kir channel currents (Skatchkov et al., 2006). Whole cell currents were first recorded in the presence of ICS blockers only (Fig. 3Ai), which blocked inward K+currents, then in addition to ECS K+ channel blockers, which markedly reduced outward K+ currents to reveal residual K2P currents (Fig. 3Aii), and finally in the K+ channel blocker cocktail plus 10 μM anandamide, which significantly reduced further the residual leak K+ current (Fig. 3Aiii). Plots are illustrated of mean (± SEM) current–voltage (I–V) relations (Fig. 3B), and of peak currents at + 50 mV (Fig. 3C). There was a significant anandamide-sensitive residual current (as observed in Müller glia (Skatchkov et al., 2006), and the peak current at + 50 mV was significantly inhibited by anandamide by 33.38 ± 18.64% (p < 0.05, t-test, n = 7). In addition, residual currents (in the presence of ICS and ECS K+ channel blockers) were significantly pH-sensitive (Fig. 4Ai–iii), with a shift in I–V relations that was more marked with a rise in extracellular pH from 7.4 to pH 8.4, than with a reduction to pH 6.4 (Fig. 4B), suggesting K2P in addition to TASK-1. Statistical analysis of peak currents at + 100 mV confirmed a significant pH-dependent effect (Fig. 4C).

Fig. 3

Oligodendrocytes express anandamide sensitive TASK-1-like currents. Whole-cell patch clamp analysis in oligodendrocytes from optic nerve explant cultures. (A) Anandamide-sensitive residual TASK-1-like currents. 10 mV depolarising voltage steps …

Fig. 4

pH sensitivity of oligodendroglial K2Pcurrents. Whole-cell patch clamp analysis in oligodendrocytes from optic nerve explant cultures. (A) Residual K2P currents are pH-sensitive. 10 mV steps from − 100 mV to + 100 mV …

TASK-1 channel inhibition protects oligodendrocytes from ischemia

Inhibition of TASK-1 channels has been shown to protect neurons from ischemia-induced depolarisation and cell death (Plant et al., 2002). The high expression of TASK-1 in oligodendrocytes may therefore be pivotal to their susceptibility to hypoxic injury (Back et al., 2007; Dewar et al., 2003). We examined the effects of TASK-1 channel inhibition in the isolated intact optic nerve from mice aged P11–13 exposed to OGD for 1 h (Fig. 5). Transgenic PLP-DsRed mice were used to examine changes in oligodendrocytes and their myelinating processes (Figs. 5A–C). Quantification of oligodendroglial processes (Fig. 5D) and number of live cells (somata) (Fig. 5E) were performed on single z-sections taken with identical settings and depths within the nerve from 2 separate FOV in each nerve (n = 4 nerves for each experimental group); relative changes in oligodendroglial processes were measured as the total number of fluorescent pixels within ROI devoid of cell somata. Compared to acutely dissected nerves, incubation in normal aCSF for 1 h at 37 °C in 5%CO2/95%O2 had no significant effect (p > 0.05, ANOVA and Newman–Keuls post-hoc tests) on oligodendrocyte processes (Fig. 5D), the number of live oligodendrocytes (Fig. 5E), or the number of propidium iodide (PI) labelled dead oligodendrocytes (Fig. 5F), demonstrating that nerves remained viable during incubation in normoxic and glucose conditions. In contrast, compared to controls incubated in O2and glucose (Fig. 5A), OGD resulted in a marked loss of oligodendrocytes and their processes (Fig. 5B), which was statistically significant (p < 0.05, ANOVA and Newman–Keuls post-hoc tests) for both oligodendrocyte processes (Fig. 5D) and somata (Fig. 5E). Inhibition of TASK-1 with 10 μM methanandamide, a non-hydrolysable form of anandamide, preserved oligodendrocytes and their processes (Fig. 5C), which were statistically greater than in OGD (p < 0.05, ANOVA and Newman–Keuls post-hoc tests) and not statistically different from controls (p > 0.05, ANOVA and Newman–Keuls post-hoc tests) (Figs. 5D, E). The loss of oligodendroglial somata in OGD is indicative of cell death, and so we examined this using PI labelling in optic nerves from Sox10-eGFP transgenic mice; PI fluoresces red and so could not be used in PLP-DsRed mice. OGD resulted in a significant increase in oligodendrocyte cell death compared to controls, and this was blocked in the presence of the TASK-1 channel inhibitor methanandamide and the non-specific TASK channel inhibitor 1 μM lidocaine (p < 0.05, ANOVA and Newman–Keuls post-hoc tests) (Fig. 5F); oligodendrocyte cell death in normoxic and glucose conditions was unaffected by methanandamide or lidocaine (p > 0.05, ANOVA and Newman–Keuls post-hoc test). The results indicate that inhibition of TASK-1 channels is protective against ischemia-induced oligodendrocyte disruption and death.

Fig. 5

TASK-1 inhibition reduces ischemia-induced oligodendrocyte cell damage and death. Optic nerves from P11–13 transgenic PLP-DsRed or Sox10-EGFP mice were exposed to 1 h OGD in the absence and presence of TASK-1 inhibitors and compared to …

This work was funded by the Medical Research Council. We thank Dr Anthony Lewis for help with the transfection, Csilla Brasko with the cell culture and immunostaining, and P. Enyedi and G. Czirjak for the mTASK-1pEXO construct.