BK K+ channel blockade inhibits radiation-induced migration/brain infiltration of glioblastoma cells

Infiltration of the brain by glioblastoma cells reportedly requires Ca2+ signals and BK K+ channels that program and drive glioblastoma cell migration, respectively. Ionizing radiation (IR) has been shown to induce expression of the chemokine SDF-1, to alter the Ca2+ signaling, and to stimulate cell migration of glioblastoma cells. Here, we quantified fractionated IR-induced migration/brain infiltration of human glioblastoma cells in vitro and in an orthotopic mouse model and analyzed the role of SDF-1/CXCR4 signaling and BK channels. To this end, the radiation-induced migratory phenotypes of human T98G and far-red fluorescent U-87MG-Katushka glioblastoma cells were characterized by mRNA and protein expression, fura-2 Ca2+ imaging, BK patch-clamp recording and transfilter migration assay. In addition, U-87MG-Katushka cells were grown to solid glioblastomas in the right hemispheres of immunocompromised mice, fractionated irradiated (6 MV photons) with 5 × 0 or 5 × 2 Gy, and SDF-1, CXCR4, and BK protein expression by the tumor as well as glioblastoma brain infiltration was analyzed in dependence on BK channel targeting by systemic paxilline application concomitant to IR. As a result, IR stimulated SDF-1 signaling and induced migration of glioblastoma cells in vitro and in vivo. Importantly, paxilline blocked IR-induced migration in vivo. Collectively, our data demonstrate that fractionated IR of glioblastoma stimulates and BK K+ channel targeting mitigates migration and brain infiltration of glioblastoma cells in vivo. This suggests that BK channel targeting might represent a novel approach to overcome radiation-induced spreading of malignant brain tumors during radiotherapy.


INTRODUCTION
Glioblastoma multiforme consists of cells with a highly migratory phenotype that may "travel" long distances throughout the brain [1]. Primary foci of glioblastoma usually show a characteristic diffuse and netlike brain infiltration which represents a major challenge for surgical tumor resection as well as for adequate coverage by radiotherapy [2]. This migratory phenotype in concert with a pronounced resistance to radiotherapy and chemotherapy probably contributes to frequent therapy failure and bad prognosis observed in the vast majority of patients with glioblastoma.
Sublethal IR has been demonstrated in vitro and/or in rodent tumor models to induce migration, metastasis, invasion and spreading of a variety of tumor entitites. In particular, a plethora of in vitro and in vivo studies suggest that IR induces migration of glioblastoma cells (for review see [3,4]). Three-dimensional-glioblastoma in vitro models, however, could not confirm this phenomenon [5] and whether or not IR induces migration of glioblastoma cells in vivo is still under debate.
If IR-induced migration, however, reaches relevant levels during fractionated radiotherapy of glioblastoma patients it might boost glioblastoma brain infiltration and -in the worst case -evasion of glioblastoma cells from the target volume of the radiotherapy. Along those lines, the chemokine SDF-1 (stromal cell-derived factor-1, CXCL12) via its receptor CXCR4 [6][7][8] stimulates migration of glioblastoma cells [9]. IR reportedly induces the expression of SDF-1 in different tumor entities including glioblastoma [10][11][12][13] as well as in normal brain tissue [7].
Collectively, these findings suggest that IR-induced migration may contribute to therapy resistance of glioblastoma. The present study, therefore, aimed to provide a quantitative analysis of IR-induced migration/ brain infiltration in an orthotopic xenograft model of human glioblastoma. Importantly, a previous in vitro study of our group disclosed IR-induced BK K + channel activation as a key event in IR-induced migration. Since BK channel blockade by paxilline, a toxin of the fungus Penicillium paxilli, suppresses IR-induced migration in vitro [14] the present study further tested whether glioma BK channel targeting with paxilline might be a powerful strategy to suppress IR-induced migration of glioblastoma cells in vivo. Our data strongly suggest that fractionated IR stimulates glioblastoma migration/ infiltration in vivo via auto-/paracrine SDF-1 signaling and subsequent BK channel activation.

RESULTS
Studies using human U-87MG glioblastoma cells to generate orthotopic mouse models report encapsulated and low brain infiltrative tumor growth [15]. Therefore, U-87MG glioblastoma seemed excellently suited for quantitative analysis of number and migration distances of individual glioblastoma cells. We used the U-87MG-Katushka clone stably transfected with the far-red fluorescent protein Katushka for histological glioblastoma cell tracking. The Katushka protein-expressing U-87MG cells were comparable to the wild type cells regarding growth kinetics and chemosensitivity against standard cytostatic drugs as shown in Supplementary Figure S1A-S1C. The BK inhibitor paxilline had no significant antiproliferative activity on U-87MG-Katushka cells upon long-term exposure at concentrations of up to 10 µM (Supplementary Figure S1D).
First, we studied in vitro both BK channel expression in U-87MG-Katushka cells and putative radiosensitizing effects of the BK channel inhibitor paxilline. Issuing the latter was plausible since pharmacological blockade of the BK-related Ca 2+ -activated IK channels reportedly radiosensitizes T98G and U-87MG glioblastoma cells [16]. Similar radiosensitizing action of paxilline would complicate the interpretation of any paxilline in vivo effect on tumor cell migration and brain infiltration.
As described for T98G and the parental U-87MG cells [14], the U-87MG-Katushka clone functionally expressed BK channels. This was evident from whole-cell patch-clamp recordings with K-gluconate in the pipette and NaCl in the bath. U-87MG-Katuska cells exhibited large outward currents in the range of several nanoamperes ( Figure 1A, left). These currents were outwardly rectifying and blocked by the BK channel inhibitor paxilline ( Figure 1A right and 1B) indicative of functional expression of BK channels. To test for a radiosensitizing action of BK channel targeting, the influence of paxilline on clonogenic survival of irradiated U-87MG-Katushka and T98G cells was determined by delayed plating colony formation assays. In contrast to IK channel targeting [16], BK channel blockade by paxilline did not radiosensitize either glioblastoma cell models ( Figure 1C and 1D).
Reportedly, IR stimulates in vivo the expression of the chemokine SDF-1 by the glioma invasion front [13]. Therefore, U-87MG-Katushka and T98G were tested in vitro for IR-induced BK channel activity, transfilter migration and the role of SDF-1 signaling herein in order to define radiation-induced signaling events upstream of BK channel activation. In on-cell patch-clamp recordings (KCl pipette-and NaCl bath solution) from U-87MG-Katushka cells (Figure 2A), large conductance ion channels (unitary conductance, g ≈ 200 pS, Figure 2B) became increasingly active with increasing positive voltage. In irradiated cells (2 Gy, 2-4.5 h after IR), channel activity was observed at highly significantly lower clamp-voltage than in non-irradiated cells (Figure 2 A, right, Figure 2C, closed triangles and Figure 2D, closed bar). In on-cell mode, a clamp-voltage between pipette and bath solution of 0 mV is recording the transmembrane currents at physiological membrane potential. Therefore, the current transitions at 0 mV observed in irradiated cells indicated the activity of the large conductance ion channel at physiological membrane potential. In unirradiated control cells, in contrast, channel activity was triggered only by clamp-voltages above +50 mV ( Figure 2D, open bar). Accordingly, mean macroscopic on-cell outward currents in irradiated cells exceeded significantly that of control cells by twofold ( Figure 2E, black symbols and Figure 2F, open bars). Notably, paxilline (5 µM; Figure 2E, grey symbols and Figure 2F, closed bars) blocked about half of the outward current in irradiated cells while having no effect in control cells. The paxilline-sensitive current fractions of irradiated cells showed typical outward rectification ( Figure 2G, closed triangles). In combination, voltage-dependence of open probability, high unitary conductance and paxillinesensitivity defined the large conductance channel as BK K + channel. Importantly, BK channels were active in irradiated cells at physiological membrane potential suggesting their functional significance for the irradiated glioblastoma cells. www.impactjournals.com/oncotarget BK channel activation in irradiated U-87MG-Katushka cells was paralleled by significantly faster chemotaxis when compared to unirradiated control cells as determined by FCS gradient-stimulated transfilter migration assays ( Figure 2H and 2I, open bars). The BK channel inhibitor paxilline (5 µM) did not affect the basal fraction of migrating cells, whereas the IR-induced migration was completely abolished ( Figure 2I, closed bars). Together, these data indicate radiation-induced migration in U-87MG-Katushka cells depending on IR-induced BK channel activity.
To confirm previously published data on paxillinesensitive IR-induced migration [14], irradiated (0 or 2 Gy, 2-4.5 h after IR) T98G cells were on-cell patch-clamp recorded with KCl in the pipette and NaCl bath solution. Similar to the U-87MG-Katushka model, irradiated T98G glioblastoma cells showed voltage-dependent activity of large conductance (g ≈ 200 pS) channels ( Figure 3A-3C). Channels were active in irradiated T98G cells at physiological membrane potential (i.e., 0 mV clampvoltage, Figure 3B) and generated an outwardly rectifying macroscopic on-cell current ( Figure 3D, black closed   Figure 3E, 3rd bar) that was significantly larger than that of the unirradiated T98G control cells ( Figure 3D, black open circles, and Figure 3E, 1st bar). The BK inhibitor paxilline blocked the IR-induced outward current ( Figure 3D, grey closed triangles and Figure 3E, 4th bar) and the paxilline-sensitive current fraction showed typical outward rectification ( Figure 3F, closed triangles) indicative of IR-induced BK channel activation. Unirradiated T98G control cells, in contrast, did not exhibit significant paxilline-sensitive outward currents even at high clamp voltages ( Figure 3D, open circles, Figure 3E, 1st and 2nd bar, and Figure 3F, open circles).
IR-induced migration ( Figure 3G, black symbols and Figure 3H, 1st and 2nd bar) required BK activation, because shRNA-mediated knockdown of BK ( Figure 3G, insert) significantly decreased migration velocity of irradiated cells to the control values while not affecting basal migration of unirradiated control cells ( Figure 3H, grey symbols and Figure 3H, 3rd and 4th bar). These results suggest that the previously reported IR-induced and paxilline-sensitive migration of T98G cells [14] is mediated by IR-induced BK channel activation.
To estimate, whether IR-induced migration might be associated with an hyperinvasive phenotype and to identify radiation-triggered signaling events, abundances of selected mRNAs were compared between fractionated irradiated (5 × 2 Gy) and control (5 × 0 Gy) U-87MG-Katushka cells. Specifically, mRNAs encoding for BK, the matrix metalloproteinases MMP-2 and MMP-9, the chemokine SDF-1 (CXCL12) and the SDF-1 receptor CXCR4 were analyzed. As shown in Figure 4A, fractionated IR did not alter BK or CXCR4 mRNA but significantly increased the abundance of MMP-2, MMP-9, and SDF-1 mRNA collectively pointing towards an IRinduced hyperinvasive phenotype.
To estimate the functional significance of SDF-1 signaling in IR-induced BK activation and migration, the effect of conditioned medium harvested from control and irradiated U-87MG-Katushka cells (2 h after IR) on Ca 2+ signaling was determined in the presence and absence of the CXCR4 antagonist AMD3100 (1 µM). In fura-2 Ca 2+ imaging experiments, conditioned medium from irradiated cells evoked a significantly faster rise in cytosolic free [Ca 2+ ] i than conditioned medium from unirradiated control cells ( Figure 4F and 4G, open bars). Importantly, AMD3100, when washed-in together with the conditioned medium, completely abolished the IR effect on free [Ca 2+ ] i ( Figure 4G, closed bars). This suggests that IR induces enrichment of factors stimulating the CXCR4 signaling in the medium. Likewise, AMD3100 significantly blocked the BK channel activation by IR when applied before on-cell patch-clamp recording ( Figure 4H, 4I). Together, these data suggest the involvement of SDF-1/CXCR4 signaling in IR-induced BK channel activation of U-87MG-Katushka cells.
Similarly to U-87MG-Katushka cells, T98G cells expressed CXCR4 and SDF1 mRNA and protein (RT-PCR and immunoblot data not shown). IR (2 Gy, 2 h after IR) induced a significant increase in SDF-1 protein abundance in T98G cells as determined by immunofluorescence microscopy ( Figure 5A-5C). Conditioned medium from irradiated (2 Gy, 2 h after IR) T98G cells exhibited significantly higher SDF-1 concentrations than control cells as determined by ELISA ( Figure 5D). Accordingly, superfusion of conditioned medium from irradiated T98G cells (2 Gy, 2 h after IR) stimulated a significant increase in free [Ca 2+ ] i in T98G cells while medium from control cells did not ( Figure 5E and 5F, open bars). The CXCR4 antagonist AMD3100 (1 µM) prevented the Ca 2+ signals in T98G cells elicited by conditioned medium from irradiated cells ( Figure 5F, closed bars). Together, these data suggest IR-induced SDF-1 signaling also in T98G cells.
To test, whether SDF-1 can mimic IR-induced BK channel activation and migration in U-87MG-Katushka cells, free [Ca 2+ ] i was recorded during wash-in of SDF-1 ( Figure 6A, 6B). Acute application of SDF-1 (50 nM) stimulated a significant long-lasting increase in free [Ca 2+ ] i . In addition, acute application of SDF-1 induced significant paxilline-sensitive outward currents in on-cell patch-clamp recordings ( Figure 6C-6F). Remarkably, the SDF-1stimulated current fraction was outwardly rectifying for BK channels.  Figure 6E) and closely resembled the radiation-induced currents (compare Figure 6E with Figure 2G, closed triangles) in voltage-dependence and absolute values. Finally, SDF-1 (50 nM) significantly increased transfilter migration of U-87MG-Katushka cells ( Figure 6G and 6H, open bars). The BK channel inhibitor paxilline (5 µM) blocked the SDF-1 induced augmentation of transfilter migration without significantly inhibiting basal migration ( Figure 6H, closed bars). In summary, SDF-1 very similarly to radiation stimulates migration that depends on BK channel activation.
Analogous to U-87MG-Katushka, acute application of SDF-1 (50 nM) induced in T98G cells a long-lasting increase in free [Ca 2+ ] i , ( Figure 7A, 7B) and an activation of macroscopic outward current in on-cell patch-clamp recordings ( Figure 7C, 7D). Single channel analysis ( Figure 7E) revealed BK-like large conductance channels (p ≈ 170 pS; Figure 7F) that activated increasingly with increasing voltage ( Figure 7G). In summary, the in vitro data on the U-87MG-Katushka clone demonstrated IR-induced BK K + channel-dependent migration similar to that observed in the present study and/or reported previously for T98G and the parental U-87MG cells [14]. Moreover, these in vitro experiments strongly suggest that IR-induced SDF-1 signaling is triggering upstream of BK at least part of the IR-induced migration.
For generation of orthotopic glioblastoma, U-87MG-Katushka ( Figure 8A) cells were stereotactically inoculated into the right striatum of NSG mice ( Figure 8B). Inoculation resulted in the formation of solid and most widely encapsulated glioblastoma ( Figure 8C) that grew exponentially during the first 3 weeks after tumor challenge ( Figure 8D). As illustrated in Figure 8E, 8F, the glioblastoma-bearing right hemisphere of isoflurananaesthesized mice was fractionated irradiated (6 MV photons) on days 7-11 with daily fractions of 2 Gy by the use of mouse holders mounted in the radiation beam of a linear accelerator. The mouse torso and head outside the radiation field were shielded by multileaf collimator and/ or an 8 cm thick on-body lead block. The film dosimetry revealed steep dose decline outside the radiation field restricting the area bounded by the 50% isodose line to about 0.8 cm × 0.5 cm = 0.4 cm 2 (with about 1/3 of this field outside the animal ( Figure 8F, 8G). Fractionated IR (5 × 2 Gy) was well tolerated by the mice as deduced from the only minute IR-associated decline of body weight ( Figure 8H, black symbols). The mice receiving paxilline developed mild and transient ataxia and showed a body weight loss of some 10-15% as compared to the respective control groups ( Figure 8H, pink symbols). After end of paxilline treatment, mice recovered completely. On day 21, mice of the 4 treatment groups were sacrificed and brains excised for histological analysis. Importantly, U-87MG-Katushka cells continued to express BK channels when grown in mouse brain as demonstrated by immunohistochemistry ( Figure 8I-8K).
In addition to SDF-1, unirradiated and fractionated irradiated orthotopic U-87MG-Katushka cells expressed CXCR4 chemokine receptor protein ( Figure 9D, 9E). CXCR4 protein abundance was similar in fractionated irradiated ( Figure 9F) and in unirradiated tumors and localized in the plasma membrane (green fluorescence) and the cytoplasm (merged yellow fluorescence due to colocalization with the far-red Katushka protein, Figure 9G).
To quantify the emigration activity of untreated and irradiated tumors, the number of evaded glioblastoma cells was counted and the migration distances determined. The margin of untreated tumors was usually clearly delimited with tangentially oriented glioblastoma cells at the tumor surface ( Figure 10A). Irradiated tumors ( Figure 10B), in contrast, exhibited zones where glioblastoma cells invaded in the adjacent brain parenchyma giving the tumor margin a fringed appearance. Figure 10C depicts numbers and emigration distances of glioblastoma cells emigrated from unirradiated (5 × 0 Gy, open circles) and fractionated irradiated (5 × 2 Gy, closed triangles) glioblastoma. IR regimes were applied in the absence ( Figure 10C) or presence ( Figure 10D) of systemic paxilline administration (8 mg/kg B.W. i.p., 6 h prior to and 6 h after each IR fraction). IR significantly stimulated emigration from the tumor ( Figure 10E, left) and paxilline administration prevented this IR-induced migration but did not affect basal emigration ( Figure 10E, right). Importantly, neither IR nor paxilline changed the glioblastoma volume ( Figure 10F).
In summary, our data on IR-induced migration of glioblastoma cells acquired in vitro by the use of the 2D-cultures and those obtained in vivo in an orthotopic glioblastoma mouse model strikingly coincided. The IR-induced induction of migration possibly occurs via IR-induced stabilization of HIF-1α. The subsequent up-regulation of the HIF-1α target gene SDF-1 was observed in vitro and in vivo, suggesting that SDF-1 can stimulate Ca 2+ transients that lead to BK K + channel (E) Mean (± SE) conductance of the clamped membrane patch as calculated from (D) for the outward currents for control and irradiated cells in the absence (open bars) or presence (closed bars) of paxilline. (F) Mean (± SE) paxilline-sensitive current fractions of control (open circles) and irradiated (closed triangles) cells (data from D). (G) Knockdown of BK channels in T98G cells abrogates radiationinduced migration. Mean (± SE, n = 4) impedance as measure of transfilter migration of empty vector control (ev, black) or BK-specific shRNA-expressing T98G cells (grey) irradiated with 5 × 0 Gy (open circles) or 5 × 2 Gy (closed triangles). The experiment started 24 h after the last IR fraction. The insert (left) shows the housekeeper-normalized BK-encoding mRNA abundance (left) as well as BK and β-actin protein abundances of T98G cells stably transduced with empty vector control (ev) or BK-specific shRNA. (H) Mean (± SE, n = 8) normalized migration velocity as calculated from the data in (G) for the first 0.25 h of transfilter migration in control (0 Gy, open bars) and irradiated (2 Gy, closed bars) cells expressing non or BK-specific shRNA. *, ** and *** indicate p ≤ 0.05, p ≤ 0.01, and p ≤ 0.001, respectively, ANOVA (D, G, H) or Welch corrected t-test (G). * and n. s. in Figure F indicate significantly (p ≤ 0.05) and not significantly different from 0, respectively, two-tailed one-sample t-test.   activation. In vitro and in vivo, BK channel targeting prevented IR-induced migration indicating that BK is functionally involved in the stress response of irradiated glioblastoma cells.

DISCUSSION
Whether or not IR-induced glioblastoma cell migration occurs in vivo is highly controversially debated and of relevance for the radiotherapy. Our study was conceptualized to get a quantitative answer to this question on cellular resolution. The new findings of our study are that in an orthotopic glioblastoma mouse model only 5 fractions of irradiation with the clinical relevant dose of 2 Gy were sufficient to increase the number of cells infiltring the brain parenchyma by factor of two. Thereby, IR-induced BK channel activity seemed to be a key event since BK blockage by paxilline abrogated IR-induced brain infiltration. Finally, our in vivo and in vitro experiments strongly suggest that IR-induced auto-/paracrine SDF-1/CXCR4 signaling contribute to BK channel activation.
Our in vivo glioblastoma model, U-87MG-Katushka, develops largely encapsulated gliomas in mouse brain. One might say that U-87MG cells, therefore, do not represent the majority of glioblastoma that grow highly infiltative, which certainly limits the generalization of our findings. On the other hand, the encapsulated growth of U-87MG cells only enabled us to quantitatively analyze number and migration distances of cells that emigrated out of the primary tumor lesion and infiltrated the brain. Moreover, U-87MG cells have the advantages to be tumorigenic, to quickly generate tumor mass with highly reproducible tumor volumes, to exhibit IR-induced migration in vitro and to express all components required for IR-induced migration as defined so far.    Another limitation of our in vivo study is that the applied glioblastoma treatment protocol (5 × 2 Gy) only partially reflects the trimodal therapy (surgery, 30 × 2 Gy radiotherapy and temozolomide chemotherapy) of glioblastoma patients [17]. Previously reported in vivo data on IR-induced migration of glioblastoma were acquired with cells or brains pre-irradiated prior to transplantation [18,19], whole brain irradiation with a single dose of 8 Gy [10], partial brain irradiation with large irradiation fields (1 cm 2 ) and single doses of 8 and 15 Gy [13], or stereotactical glioblastoma irradiation with a single dose of 50 Gy [20]. Although, each of these studies has the above mentioned limitations, these pieces of evidence combined strongly suggest that IR-induced migration is a general phenomenon and may occur during fractionated radiation therapy of glioblastoma patients.
Does IR-induced migration/infiltration contribute to the apparent high radioresistance of glioblastoma? After clinical radio(chemo)therapy, most (70-90%) volume of the recurrent glioblastoma reportedly lays within the IR target volume [2,21,22]. At a first glance, this suggests that the overall contribution of target volumeemigrated tumor cells on tumor recurrence -if existent -is low. On the other hand, one might argue that recurrent glioblastoma preferentially and much faster re-expand into irradiated and necrotic brain volumes than infiltrating intact brain parenchyma. Along those lines, detailed imaging analysis has suggested that significant volume of the recurrent glioblastoma lays in the outermost zone of the IR target volume, i.e., outside the initial gross tumor volume or biological target volume as defined by MRI or PET [2]. This might fit to the idea of re-settling the IR-target volume by glioblastoma cells that formerly emigrated from the gross tumor volume.
Glioblastoma migration is programmed by Ca 2+ signaling involving CaMKII (for review see [3,4]). Besides glioblastoma, IR-stimulated Ca 2+ signaling has been described in leukemia [23][24][25] suggesting IR-induced Ca 2+ signaling as a general phenomenon. Like IR, SDF-1 occupation of the G protein coupled chemokine receptor CXCR4 induces Ca 2+ signaling and migration/invasion in glioblastoma [9,26] and pancreatic cancer [27]. In particular, SDF-1 has been shown to induce Ca 2+ release from the Ca 2+ stores via activation of phospholipase C, and formation of inositol 1, 4, 5-trisphosphate [28]. Notably, the Ca 2+ -activated BK channels have been demonstrated to be directly linked to inositol 1, 4, 5-triphosphate receptors via lipid rafts [29]. The present study identified BK channel targeting as effective in vivo strategy to prevent IR-induced migration. BK channels reportedly fullfil a dual function in glioblastoma migration. They contribute, both, to cell volume changes that motorize migration (for review see [3]) and to Ca 2+ signaling that trigger migration [14].
BK channels are expressed in neurons of the central nervous system, e.g., in hippocampus, where they can be found in pre-and postsynaptic membranes [30]. The BK channel blocker paxilline, applied in the present study, is a neurotoxin produced by the endophytic fungus Penicillium paxilli and causes "ryegrass staggers" in sheep which is characterized by ataxia and uncontrollable tremors [31]. Paxilline is a very specific BK channel inhibitor which works in the nanomolar range [32] .The mechanism of BK blockage is largely unknown. Paxilline has been proposed to be an allosteric inhibitor which stabilizes the closed conformation of the channel [33]. Systemic paxilline administration in our in vivo experiments (8 mg/kg BW, twice per day) provoked ataxia due to the blockade of BK channels in the cerebellum [31] indicating that paxilline crosses the blood-brain barrier and reached effectiv concentrations in the brain.
Paxilline at the applied dose induced besides reversible ataxia no severe side effects and was well tolerated by the mice. This might suggest that BK channel targeting might be applied in glioblastoma patients. As a matter of fact, drugs with BK channel modulating side effects are already in clinical use. Classical neuroleptics such as haloperidol or chlorpromazine inhibit BK channels with an IC 50 in the low micromolar range [34]. Reportedly, haloperidol may accumulate in the human brain up to micromolar [35] and chlorpromazine up to several ten micromolar concentrations [36] suggesting that the therapeutic concentrations of the classical neuroleptics affect BK channel activity.
In conclusion, fractionated radiation stimulates migration of glioblastoma cells in vivo. This phenomenon might lead to enhanced tumor spreading during fractionated radiotherapy and might contribute to therapy failure. Radiation-induced BK K + channel activation triggers and BK channel blockage suppresses IR-induced migration suggesting BK channel targeting as useful-tool to overcome IR induced migration during radiotherapy.
For the on-cell experiments (Figures 2-4, 6-7) the pipette solution contained (in mM) 0 or 0.005 paxilline in 0.1% DMSO, 130 KCl, 32 HEPES, 5 d-glucose, 1 MgCl 2 , 1 CaCl 2 , titrated with KOH to pH 7.4. We used a high K + -containing pipette solution in order to have a direct quantitative measure for IK channel activity by analyzing the inward currents at highly negative voltages (since BK currents are negligible at these negative voltages). IK channel activity is also induced by radiation [16]. In some experiments, the chemokine stromal cell-derived factor-1 (SDF-1, CXCL12, 50 nM, Immuno Tools, Friesoythe, Germany) and paxilline (5 µM) was added to the bath solution. Whole-cell and macroscopic on-cell currents were analyzed by averaging the currents between 100 and 700 ms of each square pulse. Applied voltages refer to the cytoplasmic face of the membrane with respect to the extracellular space. In the current tracings (wholecell and macroscopic on-cell currents), the individual current sweeps recorded at the different clamp-voltages are superimposed. Outward currents, defined as flow of positive charge (here: K + ) from the cytoplasmic to the extracellular membrane face, are positive currents and depicted as upward deflections of the original current tracings.

Colony formation assay
To test for clonogenic survival, U-87MG-Katushka and T98G cells were preincubated (0.5 h), irradiated (0, 2, 4 or 6 Gy) and post-incubated (24 h) in RPMI-1640/10% FCS medium additionally containing paxilline (0 or 5 µM in 0.1% DMSO). 24 h after IR, cells were detached, 300 and 600 cells were re-seeded in inhibitor-free medium on 3 cm wells and grown for further 2-3 weeks. The plating efficiency was defined by dividing the number of colonies by the number of plated cells. Plating efficiencies of control and paxilline-treated cells were 0.23 ± 0.001 and 0.22 ± 0.001 for T98G (n = 36) and 0.10 ± 0.003 and 0.06 ± 0.006 (n = 12) for U-87MG-Katushka cells, respectively. Survival fractions as calculated by dividing the plating efficiency of the irradiated cells by those of the unirradiated controls were fitted by the use of the linear quadratic equation.

Transfilter migration
The lower and upper chamber of a CIM-Plate 16 (Roche, Mannheim, Germany) were filled with 160 µl (lower chamber) and 100 µl (upper chamber) of RPMI-1640 medium containing 5% and 1% FCS, respectively, equilibrated at 37°C and 5% CO 2 for 30-60 min. The upper and lower chamber additionally contained SDF-1 (0 or 50 nM) and paxilline (0 or 5 µM in 0.1% DMSO). After CO 2 equilibration and resetting the impedance to zero, 100 µl of cell suspension containing 40.000 of unirradiated (SDF-1 experiments) or irradiated cells (0 or 2 Gy, 1-2 h after IR) in RPMI-1640/1% FCS were added to the upper chamber. After sedimentation and adherence of the cells (2-3 h after IR), migration was analyzed in real-time by measuring the impedance increase between electrodes which cover the lower surface of the filter membrane and the reference electrode in the lower chamber. Upon trans-filter migration, cells adhere to the filter electrode surface and increase the impedance. To compare between individual experiments the impedances were normalized to the 0.5 h values of the respective controls.

Orthotopic mouse model of human glioblastoma
Animal experiments were carried out according to the German animal protection law and approved by the local authorities. Fluorescent U-87MG-Katushka cells ( Figure 8A) were inoculated stereotactically into the right striatum of 12 week old immunocompromised male and female NOD/SCID/IL2Rγ null (NSG) mice. The skullcap was trepanated 2.6 mm laterally and 0.5 mm caudally of the bregma (as indicated in the drawing of Figure 8B) by the use of a dental driller and 30.000 U-87MG-Katushka cells (in 10 µl of FCS-free EMEM medium) were injected in 3 mm depth from the dura surface into the right striatum. Starting at day 7, Isofluran-anaesthesized mice were immobilized under a 6 MV linear accelerator (LINAC SL25 Philips) and the right hemispheres were irradiated with daily fractions of 0 or 2 Gy 6 MV photons using mouse holders and shieldings as described in Figure 8E, 8F. On the days of radiation, paxilline (0 or 8 mg/kg BW i.p. in 70 µl 90% DMSO) was administered 6 h prior to and 6 h after each radiation fraction to some of the mice. In particular, 6 mice of the 5 × 0 Gy and 5 × 2 Gy control groups received vehicle alone while 14 mice were not i.p. injected. The data between the vehicle-receiving and non-receiving control mice did not differ in the 5 × 0 Gy or 5 × 2 Gy group and were pooled in each group. For dosimetry, Gafchromic 3 films (Ashland Inc., Covington, KY) placed in a mouse phantom (in 5 mm depth form the phantom surface) were exposed. By comparison with unshielded calibration films, dose distribution was defined by the film blackening inside and outside the target volume. For the dosimetry film shown in Figure 8G, background blackening (as defined by unexposed films) was subtracted.

Immunofluorescence microscopy and immunohistochemistry of brain sections
Three weeks (in pilot experiments one or two weeks) after tumor challenge, mice were sacrificed and brains were fixed (2% paraformaldehyd in phosphate buffered solution, PBS for 24 h), cryo-protected (30% sucrose in PBS for 24 h), frozen at -80°C in Richard-Allan Scientific ™ Neg-50 ™ Frozen Section Medium (Thermo Scientific, Germany), and cryosectioned (20 µm). For glioblastoma cell migration, cryosections were directly coverslipped in Vectashield Antifade Mounting Medium with DAPI and Katushka and DAPI fluorescence was evaluated by conventional fluorescence microscopy. For each tumor, all emigrated cells were summarized to generate one data point. The total number of emigrated cells per tumor was then compared between all four groups (5 × 0 Gy-and 5 × 2 Gy-irradiated glioblastomas in absence and presence of systemical application of BK channel inhibitor paxilline).
For SDF-1 protein immunostaining, sections were post-fixed 15 min (4% paraformaldehyde in PBS) and processed identically to the protocol described above for the cultured cells. After mounting, SDF-1-specific FITC was analyzed by confocal fluorescence microscopy. Antibody specificity was confirmed by the isotype which didn't produce any considerable fluorescence staining (data not shown). For CXCR-4 staining, the brain sections were fixed and permeabilized with 100% icecold methanol for 10 min instead of PFA fixation. As primary antibody, an anti-CXCR4 antibody (rabbit polyclonal antibody, Abcam #ab2074) was used in a 1:100 dilution.
For BK channel staining ( Figure 8I-8K), sections were washed three times for 5 min with PBS, fixed and permeabilized for 10 min with 100% icecold methanol, again washed with PBS for 5 min and blocked for one hour in blocking solution (PBS containing 1% BSA, 0,2% Glycin, 0,2% Lysin, 5% goat serum and 0,3% Triton X-100). Sections were incubated overnight (4°C) with rabbit anti BKa (674-1115) antibody [30], 1:500 in blocking solution and after washing three times for 5 min with PBS incubated for 1 h with the biotinylated secondary antibody, 1:200 (anti rabbit IgG, Vector Laboratories) in blocking solution. The staining was visualized with the alkaline phosphatase method and the sections were covered with Aquatex (Merck-Millipore). For positive and negative control hippocampus sections from a NSG mouse ( Figure 8I) and a BK -/mouse [30] ( Figure 8K) were used, respectively.

BK knockdown
BK channels were down-regulated in T98G cells by transduction with a pool of five BK-specific MISSION ® shRNA Lentiviral Transduction Particles and as a control with MISSION ® pLKO.1-puro Empty Vector Control Transduction Particles (SHCLNV-NM_002247 and SHC001V, Sigma-Aldrich). Cells were transduced according to the provided experimental protocol positively transduced clones were selected by the use of 2 µg/ ml puromycin in the culture medium. Down-regulation of BK was controlled by quantitative RT-PCR and immunoblotting ( Figure 3G, insert).

SDF-1 ELISA
T98G cells (250.000) were seeded in 75 cm 2 cell culture flasks in RPMI-1640 medium containing 10% FCS and grown over night. Cells were washed with PBS and serum depleted for 24 h. Thereafter, cells were washed with PBS and overlaid with NaCl solution (see patch-clamp section) containing protease inhibitors (Roche, cOmplete Mini, EDTA-free, #04693159001). After 30 min incubation, cells were irradiated with 0 or 2 Gy. After further 2 h the medium was harvested and the SDF-1 concentration determined using an ELISA assay kit (R&D Systems, Human CXCL12/SDF-1 DuoSet ELISA, #DY350).