Glial Ca2+signaling links endocytosis to K+ buffering around neuronal somas to regulate excitability

Glial-neuronal signaling at synapses is widely studied, but how glia interact with neuronal somas to regulate their activity is unclear. Drosophila cortex glia are restricted to brain regions devoid of synapses, providing an opportunity to characterize interactions with neuronal somas. Mutations in the cortex glial NCKXzydeco elevate basal Ca2+, predisposing animals to seizure-like behavior. To determine how cortex glial Ca2+ signaling controls neuronal excitability, we performed an in vivo modifier screen of the NCKXzydeco seizure phenotype. We show that elevation of glial Ca2+ causes hyperactivation of calcineurin-dependent endocytosis and accumulation of early endosomes. Knockdown of sandman, a K2P channel, recapitulates NCKXzydeco seizures. Indeed, sandman expression on cortex glial membranes is substantially reduced in NCKXzydeco mutants, indicating enhanced internalization of sandman predisposes animals to seizures. These data provide an unexpected link between glial Ca2+ signaling and the well-known role of glia in K+ buffering as a key mechanism for regulating neuronal excitability.


Introduction
Glial cells are well known to play structural and supportive roles for their more electrically excitable neuronal counterparts. However, growing evidence indicates glial Ca 2+ signaling influences neuronal physiology on a rapid time scale. A single astrocytic glia contacts multiple neuronal cell bodies, hundreds of neuronal processes, and tens of thousands of synapses (Halassa et al., 2007;Ventura and Harris, 1999). Cultured astrocytes oscillate intracellular Ca 2+ spontaneously (Takata and Hirase, 2008) and in response to neurotransmitters (Agulhon et al., 2008;Lee et al., 2010), including glutamate (Cornell-Bell et al., 1990). Glutamate released during normal synaptic transmission is sufficient to induce astrocytic Ca 2+ oscillations (Dani et al., 1992;Wang et al., 2006), which trigger Ca 2+ elevation in co-cultured neurons (Nedergaard, 1994;Parpura et al., 1994) that can elicit action potentials (Angulo et al., 2004;Fellin et al., 2006;Fellin et al., 2004;Pirttimaki et al., 2011). These astrocyte-neuron interactions suggest abnormally elevated glial Ca 2+ might produce neuronal hypersynchrony. Indeed, increased glial activity is associated with abnormal neuronal excitability (Wetherington et al., 2008), and pathologic elevation of glial Ca 2+ can play an important role in the generation of seizures (Gó mez- Gonzalo et al., 2010;Tian et al., 2005). However, the molecular pathway(s) by which glia-to-neuron communication alters neuronal excitability is poorly characterized. In addition, how glia interface with synaptic versus non-synaptic regions of neurons is unclear.
Several glia-neuronal cell body interactions have been reported for different glial subtypes (Allen and Barres, 2009;Baalman et al., 2015;Battefeld et al., 2016;Takasaki et al., 2010). A single mammalian astrocyte can be associated with multiple neuronal cell bodies and thousands of synapses (Halassa et al., 2007;Ventura and Harris, 1999). However, the complex structure of mammalian astrocytes and the diversity of their glia-neuron contacts makes it challenging to directly manipulate glial signaling only at contacts with neuronal cell bodies. Drosophila provides an ideal system to study glial-neuronal soma interactions as the Drosophila CNS contains two specialized astrocyte-like glial subtypes that interact specifically either with dendrites and synapses (astrocytes, Stork et al., 2014) or with neuronal cell bodies (cortex glia, Awasaki et al., 2008;Pereanu et al., 2005). Cortex glia encapsulate all neuronal cell bodies in the CNS with fine, latticelike processes (Awasaki et al., 2008;Coutinho-Budd et al., 2017) ( Figure 1A) and are thought to provide metabolic support and electrical isolation to their neuronal counterparts (Buchanan and Benzer, 1993;Volkenhoff et al., 2015).
Previous work in our lab identified zydeco (zyd), a cortex glial enriched Na + Ca 2+ K + (NCKX) exchanger involved in maintaining normal neural excitability (Melom and Littleton, 2013). Mutations in NCKX zydeco (hereafter referred to as zyd) predispose animals to temperature-sensitive seizure-like behavior ( Figure 1-video 1, see Materials and methods) and result in bang sensitivity (seizure-like behavior following a brief vortex). Basal intracellular Ca 2+ levels are elevated in zyd cortex glia, while near-membrane microdomain Ca 2+ oscillations observed in wildtype cortex glia are abolished . Whether the loss of Ca 2+ microdomain events in zyd is due to a disruption in the mechanism generating these events or secondary to a saturation effect from elevated basal Ca 2+ levels is unclear. Though the mechanism(s) by which cortex glia modulate neuronal activity in zyd mutants is unknown, disruption of glial Ca 2+ regulation dramatically enhances seizure susceptibility.
To determine how altered cortex glial Ca 2+ signaling in zyd mutants regulates neuronal excitability, we took advantage of the zyd mutation and performed an RNAi screen for modifiers of the seizure-like phenotype. Here we show that chronic elevation of glial Ca 2+ causes hyperactivation of calcineurin-dependent endocytosis, leading to an endo-exocytosis imbalance. In addition, knockdown of sandman, a K 2P channel, recapitulates the zyd phenotype and acts downstream of calcineurin in cortex glia, suggesting impaired sandman expression on cortex glial membranes is the cause of the zyd seizure phenotype. Indeed, cortex glial expression of GFP-tagged sandman shows that the protein is reduced on zyd cortex glial membranes. In addition, overexpressing a constitutively active K + channel in cortex glia can rescue zyd seizures. Together, these findings suggest glial Ca 2+ interfaces with calcineurin-dependent endocytosis to regulate plasma membrane protein levels and the K + buffering capacity of glia associated with neuronal somas. Disruption of these pathways leads to enhanced neuronal excitability and seizures, suggesting potential targets for future glialbased therapeutic modifiers of epilepsy.

Results
Mutations in a cortex glial NCKX generate stress-induced seizures without affecting brain structure or baseline neuronal function We previously identified and characterized a Drosophila temperature-sensitive (TS) mutant termed zydeco (zyd) that exhibits seizure-like behavior ( Figure 1-video 1, hereafter referred to as 'seizures', see Materials and methods for definition) when exposed to a variety of environmental stressors, including heat-shock and acute vortex. The zyd mutation disrupts a NCKX exchanger that extrudes cytosolic Ca 2+ . Restoring zyd function specifically in cortex glial completely reverses the zyd seizure phenotype (Melom and Littleton, 2013). Cortex glia exhibit spatial segregation reminiscent of mammalian astrocytes, with each glial cell ensheathing multiple neuronal somas (Awasaki et al., 2008;Melom and Littleton, 2013). However, little is known about their role in the mature nervous system. In vivo Ca 2+ imaging using a myristoylated Ca 2+ sensitive-GFP (myrG-CaMP5) revealed small, rapid cortex glial Ca 2+ oscillations in wildtype Drosophila larvae (Figure 1video 2, Figure 1-figure supplement 1A). In contrast, zyd mutants lack microdomain Ca 2+ transients and exhibit elevated baseline intracellular Ca 2+ (Figure 1-video 3), indicating altered glial Ca 2+ regulation underlies seizure susceptibility in zyd mutants (Melom and Littleton, 2013). Given cortex glia regulate the guidance of secondary axons and maintenance of cortical structural integrity (Coutinho-Budd et al., 2017;Dumstrei et al., 2003;Spindler et al., 2009), we first tested zyd larvae for morphological brain changes. Examination of brain structure and cortex glial morphology using fluorescent microscopy revealed no apparent changes in zyd mutants ( Figure 1A), which showed similar cortex volume occupied by glial processes ( Figure 1B) and cell body volume ( Figure 1C) compared to wildtype cortex-glia. Closer examination of cortex glial process contacts with neuronal cell bodies using electron microscopy did not reveal morphological changes between controls and zyd mutants ( Figure 1D). To determine if loss of ZYD affected glial or neuronal cell survival, we quantified the cell death marker DCP-1 (cleaved death caspase protein-1 [Akagawa et al., 2015]) in control and zyd 3 rd instar larvae and adults. No change in cleaved DCP1 levels were found, indicating basal cell death was unaffected ( Figure 1-figure supplement 1B). These data suggest mutations in NCKX zydeco disrupt cortex glia function rather than morphology or development.
We previously found that conditionally restoring zyd function only during adult stages can reverse the zyd seizure phenotype (Melom and Littleton, 2013), providing additional evidence that the phenotype does not arise secondary to developmental changes or defective assembly of brain circuits. In a complementary approach, we knocked down zyd chronically throughout development or conditionally in adult stages following brain development using a UAS-zyd RNAi hairpin expressed with cortex glial-specific drivers (NP2222-gal4 and GMR54H02-gal4). ZYD was previously shown to be specifically required in cortex glia for the generation of seizures (Melom and Littleton, 2013). Both chronic and inducible cortex glial knockdowns mimicked the zyd TS seizure phenotype ( Figure 1E). Seizure characteristics, including temperature threshold for seizure initiation and seizure kinetics, were similar between 3 rd instar larvae and adults ( Figure 1F-G), indicating a comparable requirement for zyd at both stages. Together, these results suggest that the zyd TS seizure phenotype is not due to morphological or developmental changes in brain anatomy, or changes in the ability of cortex glia to ensheath neuronal cell bodies.
NCKX transporters use the Na + and K + gradients to extrude Ca 2+ , suggesting the loss of ZYD might alter the ionic balance of these ions that could affect neuronal membrane properties. Hence, we assayed if zyd animals displayed altered behaviors in the absence of the temperature trigger needed to induce seizures. We used a gentle touch assay (Ma et al., 2016;Zhou et al., 2012) to investigate whether the zyd mutation changed larval startle-induced behaviors, as elevated Ca 2+ activity in astrocytes was reported to correlate with elevated arousal in mice (Ding et al., 2013; Figure 1 continued for each genotype, p=0.138). (C) Quantification of cortex glial cell body surface area shows no difference between wildtype and zyd (n > 120 cells/ N = 4 animals for each genotype, p=0.0892). (D) Electron microscopy images of cortex glial contacts (arrowheads) with neuronal somas (N). Cortex glial processes between neuronal cell bodies are as thin as 50 nm in both wildtype and zyd. Scale bar = 500 nm. (E) Time course of Heat-shock induced seizures (38.5˚C, HS) following chronic or conditional knockdown of zyd with two different cortex glial drivers (NP2222 and GMR54H02) is shown. Rearing adult flies at the restrictive temperature (>30˚C) for gal80 ts (a temperature-sensitive form of the gal4 inhibitor, gal80, see Materials and methods) removes gal80 inhibition of gal4 and allows expression of zyd RNAi only at the adult stage. These manipulations reproduce the zyd mutant seizure phenotype (N = 4 groups of 20 flies/genotype). (F-G) Behavioral analysis of HS-induced seizures at 38.5˚C shows that larval and adult seizures have similar temperature threshold (F) and kinetics (G) (N = 4 groups of 10-20 animals/condition/treatment). (H) Recordings of the giant fiber system muscle output. Seizure thresholds in wildtype, zyd and Para bss1 (positive control) are shown. The voltage required to induce seizures in zyd is not significantly different from wildtype (35.32 ± 3.65V and 31.33 ± 2.12V, p=0.3191, n ! 7 flies/genotype). (I) Behavioral analysis of the time course of HSinduced seizures indicates neuronal knockdown of cac (C155>cac RNAi ) rescues the zyd seizure phenotype. Inset shows results after 1 minute of HS (p=0.0004, N = 4 groups of 20 flies/genotype). Error bars are SEM, ***=P < 0.001, Student's t-test. DOI: https://doi.org/10.7554/eLife.44186.002 The following video and figure supplement are available for figure 1: Figure supplement 1. Mutations in a cortex glial NCKX generate stress-induced seizures without affecting brain structure or baseline neuronal function. DOI: https://doi.org/10.7554/eLife.44186.003 Figure 1-video 1. The response of wildtype flies to a 38.5˚C heat-shock is shown, following by the response of zyd flies to the same condition. DOI: https://doi.org/10.7554/eLife.44186.004 Figure 1 Paukert et al., 2014;Srinivasan et al., 2015) and in Drosophila (Ma et al., 2016). Crawling 3 rd instar larvae touched anteriorly execute one of two responses: pausing and/or continuing forward (type I response) or an escape response by crawling backwards (type II response). We found that wildtype, zyd and NP2222>zyd RNAi larvae exhibited similar frequencies of type I and type II responses ( Figure 1-figure supplement 1C). In addition, adult zyd flies exhibited normal locomotion ( Figure 1-figure supplement 1D) and larvae exhibited normal light avoidance responses at room temperature ( Figure 1-figure supplement 1E), indicating baseline neuronal properties required for these behaviors are unaffected. Zyd mutants also showed normal voltage thresholds in classical assays for giant fiber seizure induction, in contrast to animals harboring bang-sensitive mutations altering neuronal sodium channels ( Figure 1H). Together with our previous observation that cortex glial knockdown of calmodulin (cam [Melom and Littleton, 2013]) completely reverses zyd seizures, these data indicate zyd mutants are unlikely to display baseline changes in ionic balance that alter intrinsic neuronal properties without the elevated temperature or vortex-induced hyperactivity.
To determine if elevated neuronal activity is required for the stress-induced seizures in zyd mutants, we assayed seizure behavior in animals with reduced synaptic transmission and neuronal activity. Pan-neuronal knockdown of Cacophony (cac), the presynaptic voltage-gated Ca 2+ channel responsible for neurotransmitter release (Kawasaki et al., 2004;Rieckhof et al., 2003), significantly reduced locomotion (Figure 1-figure supplement 1F) and rescued zyd TS-induced seizures ( Figure 1I). These findings indicate elevated neuronal activity in zyd mutants during the heat shock is required for seizure induction following dysregulation of cortex glial Ca 2+ .
A genetic modifier screen of the zyd seizure phenotype reveals glia to neuron signaling mechanisms To elucidate pathways by which cortex glial Ca 2+ signaling controls somatic regulation of neuronal function and seizure susceptibility, we performed a targeted RNAi screen for modifiers of the zyd TS seizure phenotype in adult animals. We reasoned that removal of a gene product required for this signaling pathway would prevent zyd TS seizures when absent. We used the pan-glial driver repo-gal4 to express RNAi to knockdown 847 genes encoding membrane receptors, secreted ligands, ion channels and transporters, vesicular trafficking proteins and known cellular Ca 2+ homeostasis and Ca 2+ signaling pathway components (Supplementary files 1, 2). Given the broad role of Ca 2+ as a regulator of intracellular biology, we expected elevated Ca 2+ levels in zyd mutants to interface with several potential glial-neuronal signaling mechanisms. Indeed, the screen revealed multiple genetic interactions, identifying gene knockdowns that completely (28) or partially (21) rescued zyd seizures, caused lethality on their own (95) or synthetic lethality in the presence of the zyd mutation (5) or enhanced zyd seizures (37, Supplementary file 1).
Given TS seizures in zyd mutants can be fully rescued by reintroducing wildtype ZYD in cortex glia, we expected the genes identified in the RNAi pan-glial knockdown screen to function specifically within this population of cells. To assay cell-type specificity of the suppressor hits, we knocked down the top 33 rescue RNAis with cortex glial specific drivers (NP2222-gal4 and GMR54H02-gal4, Supplementary file 1). For the majority of suppressors, rescue with cortex glial specific drivers was weaker, either due to lower RNAi expression levels compared to the stronger repo-gal4 driver, or due to the requirement of the gene in other glial subtypes as well. To validate the rescue effects we observed, non-overlapping RNAis or mutant alleles for these genes were assayed (Supplementary file 2). For the current analysis, we focused only on the characterization of cortex glial Ca 2+ -dependent pathways that are mis-regulated in zyd, and how this mis-regulation promotes neuronal seizure susceptibility.

Cortex glial calcineurin activity is required for seizures in zyd mutants
We previously observed that knockdown of glial calmodulin (cam) eliminates the zyd seizure phenotype (Melom and Littleton, 2013), suggesting a Ca 2+ /cam-dependent signaling pathway regulates glial to neuronal communication. Cam is an essential Ca 2+ -binding protein that regulates multiple Ca 2+ -dependent cellular processes and is abundantly expressed in Drosophila glia (Altenhein et al., 2006), although its role in glial biology is unknown. In the RNAi screen for zyd interactors, pan-glial knockdown of the regulatory calcineurin (CN) B subunit, CanB2, completely rescued both heat-shock showed that in contrast to the continuous neuronal firing observed in zyd mutants, recordings from zyd;;repo >CanB2 RNAi#1 larvae exhibit normal rhythmic firing at 38˚C similar to wildtype controls ( Figure 2C,D). The rescue effect of CanB2 knockdown on zyd seizure phenotype was similar when CanB2 was targeted using three additional, partially-overlapping CanB2 RNAi constructs (Figure 2A,G). In addition, the zyd seizure phenotype was partially rescued when combined with a heterozygous CanB2 knockout (zyd;CanB2 KO /+) allele Figure 2A To refine the glial subpopulation in which CanB2 activity is necessary to promote seizures in zyd mutants, we knocked down CanB2 using glial subtype specific drivers. CanB2 knockdown in astrocytes resulted in no rescue of the zyd phenotype, while CanB2 knockdown with a cortex-glial specific driver (NP2222-gal4) greatly improved the zyd phenotype ( Figure   glial knockdown of CanB2 and CanA-14F rescues zyd vortex-induced seizures (N = 3 groups of 20 flies/genotype). (C) Representative voltage traces of spontaneous CPG activity at larval 3 rd instar muscle 6 at 38˚C in wildtype, zyd and zyd;;repo >CanB2 RNAi#1 animals (n ! 5 preparations/genotype). (D) Quantification of average bursting duration for CPG recordings of the indicated genotypes at 38˚C (n ! 5 preparations/genotype). (E) Detailed analysis (see Materials and methods) of HS induced behaviors of zyd/CanB2 RNAi flies. Cortex glial knockdown of CanB2 leads to seizure rescue in~30% of zyd; NP2222>CanB2 RNAi flies, with the remaining~70% displaying partial rescue. Cortex glial CanB2 knockdown with two copies of the RNAi (zyd; NP2222>CanB2 2xRNAi ) recapitulates the full rescue seen with pan-glial knockdown. Inhibiting gal4 expression of the RNAi in neurons with gal80 (C155-gal80) does not alter the rescue observed with cortex glial knockdown, and astrocyte specific (alrm-gal4) CanB2 knockdown does not rescue zyd seizures (N = 3 groups of >15 flies/genotype, see Figure 2-figure supplement 1D for complete dataset). (F) Cortex glial conditional knockdown of CanB2 using gal4/gal80 ts . Rearing adult flies at the restrictive temperature (>30˚C) for gal80 ts allows expression of CanB2 RNAi only at the adult stage. A significant reduction in seizures (p<0.0001) was seen after four days of rearing flies at the restrictive temperature for gal80 ts (31˚C), with only~25% of adults showing seizures. The reduction in seizures was enhanced when adults were incubated at 31˚C for longer periods (N = 3 groups of >10 flies/ genotype). Inset shows analysis after 4 days of incubation at 31˚C (p=0.0001). (G) Pan-glial knockdown of the Drosophila calcineurin (CN) family (CanA1, CanA-14D/Pp2B-14D, CanA-14F, CanB and CanB2) indicate CanB2 knockdown completely rescues zyd seizures, CanA-14D and CanA-14F knockdowns partially reduce seizures (N = 4 groups of >10 flies/genotype). (H) Pan-glial knockdown of Pp2B-14D and CanA-14F partially rescues the zyd HS seizures phenotype (~25% rescue for Pp2B-14D, p=0.0032; and~60% rescue for CanA14F, p<0.0001). Knocking down the two genes simultaneously rescues zyd seizures, with only~10% of flies showing seizures (~90% rescue, p<0.0001, N = 3 groups of >10 flies/genotype). (I) Overexpressing a dominant-negative form on Pp2B-14D (CanA H217Q ) rescues~50% of zyd seizures regardless of the driver used (repo: p<0.0001; NP2222: p=0.0006; GMR54H02-p=0.0004. N = 3 groups of >10 flies/genotype). (J) Larval Ca 2+ imaging in cortex glia expressing myrGCaMP6s indicates the elevated basal Ca 2+ fluorescence at 25˚C observed in zyd mutants relative to wildtype cortex glia (p=0.0003) is not altered following CanB2 knockdown (zyd;;repo>CanB2 RNAi , p=0.6096. n ! 5 animals/genotype). (K) Microdomain Ca 2+ oscillations observed in wildtype cortex glia expressing myrGCaMP6s are abolished in zyd cortex glia and are not restored following either CanB2 or CanA14F knockdown (n ! 5 animals/genotype). Error bars are SEM, **=P < 0.01, ***=P < 0.001, ****=P < 0.0001, Student's t-test. DOI: https://doi.org/10.7554/eLife.44186.007 The following video and figure supplement are available for figure 2: construct ( Figure 2E, Figure 2-video 3), with~90% of zyd;NP2222>CanB2 2xRNAi adult animals lacking seizures. The effect of CanB2 knockdown was specific to cortex glia, as it was insensitive to blockade of expression of CanB2 RNAi in neurons using C155-gal80 (a neuron specific gal4 repressor; Figure 2E and Figure 2-figure supplement 1D). To exclude a developmental effect of CanB2 RNAi knockdown within glia, we conditionally expressed a single copy of CanB2 RNAi (with gal4/gal80 ts , see Materials and methods) only in adult zyd mutant flies. Adult flies reared at the permissive temperature for gal80 ts (>30˚C) to allow CanB2 RNAi expression exhibited significantly fewer seizures after 3 days, with only~20% of flies displaying zyd-like seizures by 5 days ( Figure 2F). Zyd seizure rescue by CanB2 knockdown did not result from simple alterations in motility, as repo>CanB2 RNAi and zyd;;repo>CanB2 RNAi animal exhibited normal larval light avoidance response ( . We conclude that CanB2 is required in cortex glia to promote zyd TS seizure activity. Calcineurin (CN) is a highly conserved Ca 2+ /cam-dependent protein phosphatase implicated in a number of cellular processes (Rusnak and Mertz, 2000). CN is a heterodimer composed of a~60 kDa catalytic subunit (CanA) and a~19 kDa EF-hand Ca 2+ -binding regulatory subunit (CanB). Both subunits are essential for CN phosphatase activity. The Drosophila CN gene family contains three genes encoding CanA (CanA1, Pp2B-14D and CanA-14F) and two genes encoding CanB (CanB and CanB2) (Takeo et al., 2006). Previous studies in Drosophila demonstrated several CN subunits (CanA-14F, CanB, and CanB2) are broadly expressed in the adult Drosophila brain (Tomita et al., 2011) and that neuronal CN is essential for regulating sleep Tomita et al., 2011). CN function within glia has not been characterized. We found that pan-glial knockdown of two CanA subunits, Pp2B-14D and CanA-14F, partially rescued zyd heat-shock and vortex-induced seizures ( Figure 2B,G). The rescue was more robust for vortex-induced seizures than those induced by heat-shock, suggesting heat-shock is likely to be a more severe hyperexcitability trigger ( Figure 2B). Rescue was enhanced by knockdown of both Pp2B-14D and CanA-14F ( Figure 2H), with more than~90% of zyd;;repo>Pp2B-14D RNAi ,CanA-14F RNAi flies lacking seizures, suggesting a redundant function of these two subunits in glial cells. Similar to CanB2, knockdown of both Pp2b-14D and CanA-14F on a wildtype background was viable ( . We next overexpressed a dominant negative form of CanA (Pp2B-14D H217Q ) using either pan-glial (repo-gal4) or two different cortex-glial specific drivers (NP2222-gal4 and GMR54H02-gal4). Overexpressing Pp2B-14D H217Q resulted in~50% of zyd/Pp2B-14D H217Q flies becoming seizure-resistant regardless of the driver used ( Figure 2I). These results indicate CN activity is required in cortex glia to promote seizures in zyd mutants. Imaging intracellular Ca 2+ in cortex glia with GMR54H02-gal4 driving myrGCaMP6s revealed that CN knockdown had no effect on wildtype cortex glial Ca 2+ oscillatory behavior or the elevated basal Ca 2+ levels and the lack of microdomain Ca 2+ events in zyd cortex glia ( Figure 2J-K, Figure 2-video 4 and 5). These observations indicate CN function is required downstream of elevated intracellular Ca 2+ , rather than to regulate Ca 2+ influx or efflux in cortex glial cells. Together, these results demonstrate that a CN-dependent signaling mechanism in cortex glia is required for a glia to neuronal pathway that drives seizure generation in zyd mutants.

Calcineurin activity is enhanced in zyd cortex glia
To characterize CN activity in wildtype and zyd cortex glia we used the CalexA (Ca 2+ -dependent nuclear import of LexA) system (Masuyama et al., 2012) as a reporter for CN activity. In this assay, sustained neural activity induces CN activation and dephosphorylation of a chimeric transcription factor LexA-VP16-NFAT (termed CalexA) which is then transported into the nucleus. The imported dephosphorylated CalexA drives GFP reporter expression (Figure 3-figure supplement 1). The CalexA components were brought into control and zyd mutant backgrounds to directly assay CN activity. A substantial basal activation of CN was observed in control 3 rd instar larval cortex glia at room temperature using fluorescent imaging ( Figure 3A). CN activity and the resulting GFP expression was enhanced in zyd cortex glia ( Figure 3B) and greatly reduced in zyd/CanB2 RNAi cortex glia ( Figure 3C-C'). Western blot analysis of CalexA-induced GFP expression in adult head extracts revealed enhanced cortex glial CN activity in adult zyd mutants compared to controls (23 ± 3% enhancement, Figure 3D-E). RNAi knockdown of CanB2 reduced CalexA GFP expression as expected (27 ± 1%, Figure 3C-E). These results demonstrate CN activity is enhanced downstream of the elevated Ca 2+ levels in zyd mutant cortex glia, and that CN activity can be efficiently reduced by RNAi knockdown of CanB2.

Pharmacological inhibition of calcineurin rescues zyd seizures
Several seizure mutants in Drosophila can be suppressed by commonly used anti-epileptic drugs (Kuebler and Tanouye, 2002;Song and Tanouye, 2008), indicating conservation of key mechanisms that regulate neuronal excitability. The catalytic activity of CanA is strictly controlled by Ca 2+ levels, calmodulin, and CanB, and can be inhibited by the immunosuppressants cyclosporine A (CsA) and FK506. To assay if zyd TS seizures can be prevented with an anti-CN drug, adult flies were fed with media containing CsA and tested for HS induced seizures after 0, 3, 6, 12 and 24 hr of drug feeding (red arrowheads in Figure 4A, Figure  dependent, with less robust suppression when flies were fed with 0.3 mM CsA ( Figure 4B-D). The CsA rescue was reversible, as seizures reoccurred following 12 hr of CsA withdrawal ( Figure 4B). Feeding flies with a second anti-CN drug, FK506, resulted in a partial rescue of the phenotype ( Figure 4D). Although we cannot exclude off-target effects of these compounds in Drosophila, these data suggest pharmacologically targeting the CN pathway can improve the outcome of glial-derived neuronal seizures in the zyd mutant.
Cortex glial knockdown of the two-pore-domain K + channel, sandman, mimics zyd seizures To explore how CN hyperactivation promotes seizures, we conducted a screen of known and putative CN targets using RNAi knockdown with repo-gal4. We concentrated our screen on putative CN target genes that are involved in signal transduction (Supplementary file 3). This screen revealed that pan-glial knockdown of sandman (sand), the Drosophila homolog of TRESK (KCNK18) and a member of the two-pore-domain K + channel family (K 2P ), caused adult flies to undergo TS-induced seizures similar to zyd mutants ( Figure 5A, Figure 5-video 1). Vortex-induced seizures in repo>sand RNAi were less severe than those observed in zyd, with only~50% of sand RNAi flies showing seizures ( Figure 5-figure supplement 1A). TS-induced seizures in repo>sand RNAi adults were found to have the same kinetics and temperature threshold as seizures observed in zyd mutants ( Figure 5A, B), and CPG recordings showed that repo>sand RNAi larvae exhibit rapid, unpatterned firing at 38˚C, similar to zyd ( Figure 5C,D). Cortex-glial specific knockdown of sand recapitulated~50% of the seizure effect when two copies of the RNAi were expressed ( Figure 5A,B). The less robust effect observed with the cortex-glial driver could be due to less effective RNAi knockdown or secondary to a role for sand in other glial subtypes. To determine if sand functions in other glia subtypes to mimic the zyd seizure pathway, we expressed sand RNAi using the pan-glial driver repo-gal4 and inhibited expression specifically in cortex glia with GMR54H02>gal80. In the absence of cortex glial-knockdown of sand, seizure generation was suppressed ( Figure 5A). Similar to zyd mutants, sand RNAi animals did not show changes in general activity and locomotion at room temperature ( Figure 5E). To assess whether the seizure phenotypes in zyd and sand RNAi originate from the same pathway, we tested flies in which the zyd mutation is combined with sand RNAi . We tested different genetic combinations for seizure temperature threshold ( Figure 5B), light avoidance ( Figure 5E) and seizures kinetics ( Figure 5-figure supplement 1B). Pan-glial or cortex glial knock down of sand did not enhance the hemizygote zyd phenotype, judged by seizure temperature threshold ( Figure 5B) and kinetics ( Figure 5-figure supplement 1B). Heterozygous zyd flies expressing two copies of the sand RNAi in cortex glia (zyd/+;NP2222>sand 2xRNAi ) showed the same seizure frequency (~60% of flies seizing after two minutes of heat-shock), with similar seizures characteristics as those observed when only sand 2xRNAi was expressed ( Figure 5B, Figure 5-figure supplement 1B). Together, these results suggest seizures due to loss of sand and zyd impinge on a similar pathway. Mammalian astrocytes modulate neuronal network activity through regulation of K + buffering (Bellot-Saez et al., 2017), in addition to their role in uptake of neurotransmitters such as GABA and glutamate (Murphy-Royal et al., 2017). Human K ir4.1 potassium channels (KCNJ10) have been implicated in maintaining K + homeostasis, with mutations in the loci causing epilepsy (Haj-Yasein et al., 2011). In addition, gain and loss of astrocytic K ir4.1 influence the burst firing rate of neurons through astrocyte-to-neuronal cell body contacts (Cui et al., 2018). However, K ir channels are unlikely to be the only mechanism for glial K + clearance, as K ir4.1 channels account for less than half of the K + buffering capacity of mature hippocampal astrocytes (Ma et al., 2014). To determine if cortex glial K ir channels regulate seizure susceptibility in addition to sand, we used repo-gal4 to knock down all three Drosophila K ir family members (Irk1, Irk2 and Irk3). Pan-glial knockdown of the Drosophila K ir family did not cause seizures ( Figure 5-figure supplement 1C), while knock down of either Irk1 or Irk2 slightly enhanced the zyd phenotype ( Figure 5-figure supplement 1D). Similarly, repo-gal4 knockdown of other well-known Drosophila K + channels beyond the K ir family also did not cause seizures ( Figure 5-figure supplement 1C), indicating sand is likely to play a preferential role in K + buffering in Drosophila glia.
The mammalian sand homolog, TRESK, is directly activated by CN dephosphorylation (Czirják et al., 2004;Enyedi and Czirják, 2015), while Drosophila sand was shown to be modulated in sleep neurons by activity-induced internalization from the plasma membrane (Pimentel et al., 2016). Regardless of the mechanism by which CN may regulate the protein, we hypothesized that Knockdown of sandman (sand) in different glial subtypes: pan-glial (repo), cortex glial (NP2222) and in all glia other than cortex glia (repo>sand RNAi / GMR54H02>gal80 in which gal80 is constitutively inhibiting gal4 activity and sand RNAi expression only in cortex glia). Inset shows analysis after 2 minutes of HS (p=0.0006 for NP2222>sand 2xRNAi , p<0.0001 for repo>sand RNAi /GMR54H02>gal80, N = 4 groups of >10 flies/genotype). (B) Temperature threshold of repo>sand RNAi (p=0.5185) and NP2222>sand 2xRNAi (p=0.2302) seizures in comparison to zyd (N = 3 groups of 10/temperature/genotype). (C) Representative voltage traces of spontaneous CPG activity at larval 3 rd instar muscle 6 at 38˚C in wildtype, zyd and repo>sand RNAi (n ! 5 preparations/genotype). (D) Quantification of average bursting duration for CPG recordings of the indicated genotypes at 38˚C. (E) Light avoidance assay reveals no defect in this behavior at 25˚C (N = 3 groups of 20 flies/genotype). (F) Behavioral analysis of HS-induced seizures. Seizures in repo>sand RNAi animals were not suppressed with either CanB2 RNAi#1 or by feeding flies with 1 mM CsA (N = 3 groups of 20 flies/genotype/treatment). (G-H) Ca 2+ imaging in larval cortex glial cells using myrGCaMP6s. (G) The average rate of microdomain Ca 2+ events was reduced in repo>sand RNAi cortex glia relative to wildtype (20.36 ± 5.5 and 69.83 ± 5.3, p<0.0001). Knockdown of sand on the zyd background did not restore zyd Ca 2+ microdomain events (n ! 5 animals/genotype). (H) Average myrGCaMP6s fluorescence in cortex glia at 25˚C. Elevated basal fluorescence of GCaMP6s in zyd relative to wildtype cortex glia (p=0.0003) is not altered following sand knockdown (zyd;;repo>sand RNAi , N = 4 animals/genotype). Error bars are SEM, ***=P < 0.001, ****=P < 0.0001, Student's t-test. DOI: https://doi.org/10.7554/eLife.44186.017 The following video and figure supplement are available for figure 5: sand is epistatic to CN in controlling zyd-mediated seizures. Indeed, inhibition of CN by RNAi or CsA did not alter sand RNAi -induced seizures ( Figure 5F), placing sand downstream of CN activity. Furthermore, knockdown of sand in the zyd background does not alter the elevated basal Ca 2+ or the lack of microdomain Ca 2+ events in zyd mutants ( Figure 5G,H), and cortex glial Ca 2+ oscillatory behavior in repo >sand RNAi is similar to wildtype ( Figure 5G,H and Figure 5-video 2), suggesting sand is downstream to the abnormal Ca 2+ signaling in zyd. Overall, these findings suggest elevated Ca 2+ in zyd mutants leads to hyperactivation of CN and subsequent reduction in sand function. These results suggest that impairment in glial buffering of the rising extracellular K + during elevated neuronal activity and stress conditions (heat shock or acute vortex) causes enhanced seizure susceptibility in zyd mutants.

SAND is not differentially phosphorylated in zyd cortex glia
We next sought to examine how elevated CN activity in zyd mutants alters sand function. We generated a transgenic Drosophila strain to express GFP-tagged sand specifically in cortex glia (UASsand:eGFP, see Materials and methods). First, we used these lines to verify sand knockdown by sand RNAi , and found that SAND:GFP expression is reduced by~45% when both sand and sand RNAi are expressed using a pan glial driver (repo-gal4), while the expression of a control RNAi did not change SAND:GFP expression ( Figure 6A,B).
The mammalian sand homolog, TRESK, is constitutively phosphorylated on four serine residues (S264 by PKA, and S274, S276 and S279 by MARK1 [Enyedi and Czirják, 2015]). Two of these residues are conserved in Drosophila sand (S264 and S276, see Figure 6-figure supplement 1A for protein alignment). Constitutive dephosphorylation and subsequent activation of sand by CN is predicted to increase K + buffering following hyperactivation of the nervous system by stressors, and thus fewer seizures would be expected -opposite to what we have observed. To directly asses sand phosphorylation status in wildtype and zyd cortex glia, we used Mn 2+ phosphate binding tag (Phostag) gel electrophoresis (Kinoshita et al., 2006) to separate phosphorylated species of cortex glial SAND:GFP (expressed by the GMR54H02 driver). We found that SAND:GFP was not differentially phosphorylated in wildtype versus zyd mutants ( Figure 6C). Multiple bands were detected when samples were pre-treated with alkaline phosphatase ( Figure 6C, arrowheads), indicating SAND:GFP is indeed phosphorylated on several phosphorylation sites in cortex glia. This analysis also revealed that SAND:GFP expression level was reduced in zyd relative to wildtype cortex glia (see below). Together with the prediction that regulation of sand by dephosphorylation should lead to seizure suppression, these results argue against enhanced sand dephosphorylation as the primary cause of zyd seizures.

Enhanced endocytosis in zyd cortex glia leads to reduction of plasma membrane SAND
A second mechanism to link elevated Ca 2+ and CN hyperactivation to sand regulation is suggested by a previous study demonstrating sand expression on the plasma membrane of neurons involved in sleep homeostasis in Drosophila is regulated by activity-dependent internalization (Pimentel et al., 2016). Cam and CN activate several endocytic Ca 2+ sensors and effectors that control Ca 2+ -dependent endocytosis (Xie et al., 2017). If hyperactivity of CN leads to enhanced internalization of sand and subsequent seizure susceptibility due to decreased K + buffering capacity, we hypothesized that sand would be differentially distributed in zyd cortex glia. Indeed, examination of SAND:GFP distribution within cortex glial cells showed a significant reduction in SAND:GFP fluorescence on cortex glial membranes (53 ± 6.85% reduction, Figure 6D,E). Consistent with this observation, SAND:GFP . SAND:GFP expression in sand RNAi knockdown is reduced by~45% (p=0.0011), while the expression of a control RNAi does not significantly change SAND:GFP expression (N ! 2 experiments, three head extracts per sample). GFP signals in each experiment were normalized to control. (C) Mn 2+ phosphate binding tag (Phos-tag) gel electrophoresis analysis shows that SAND:GFP is not differentially phosphorylated in zyd relative to wildtype cortex glia, indicated by a single SAND:GFP band. Multiple bands were detected when samples were pre-treated with alkaline phosphatase, indicating SAND:GFP is phosphorylated on multiple phosphorylation sites in cortex glia (N = 3, five heads/sample). (D) Immunofluorescence of 3 rd instar larval ventral nerve cords (VNCs). Cortex glial levels of SAND:GFP are reduced in zyd relative to wildtype (magenta: anti-repo, glial nuclei; green: anti-GFP, SAND:GFP; scale bars: 10 mm in upper panels, 5 mm in lower panels). (E) Average sand:GFP fluorescence in cortex glia of wildtype and zyd. SAND:GFP fluorescence is reduced by 53 ± 6.85% in zyd cortex glia. Figure 6 continued on next page expression level detected by western blot analysis was reduced by~34 ± 15% in zyd cortex glia ( Figure 6F-G). The reduction in expression was specific to sand, and not due to a general disruption in membrane homeostasis, as no difference was detected in membrane tethered GFP (mCD8:GFP) expressed in wildtype and zyd cortex glia ( Figure 1A, Figure 6-figure supplement 1B-C). These results indicate that sand undergoes enhanced internalization and increased degradation in zyd relative to wildtype cortex glial cells.
If hyperactivity of CN leads to enhanced internalization of sand and subsequent seizures, interrupting cortex glial endocytosis should suppress zyd seizures. To test this model, we used cortex glial-specific RNAi to knock down genes involved in endocytosis and early endosomal processing and trafficking. Cortex glial knockdown of several essential endocytosis genes, including dynamin-1 and clathrin heavy and light chains, caused embryonic lethality (Figure 2-figure supplement 1A). In contrast, cortex glial knockdowns of Rab5, a Rab GTPase regulator of early endosome (EE) dynamics (Dunst et al., 2015;Langemeyer et al., 2018), and Endophilin A (EndoA), a BAR-domain protein involved in early stages of endocytosis (Verstreken et al., 2002), were found to be viable (Figure 2-figure supplement 1A) and to completely suppress zyd TS seizures ( Figure 7A, Figure 7-video 1). A second, non-overlapping hairpin and a dominant negative (DN) construct for Rab5 (Rab5 DN ) resulted in early larval lethality, likely due to more efficient Rab5 activity suppression. Cortex glial knockdown of Rab5 was previously shown to cause a morphological defect in which cortex glia fail to infiltrate the cortex and enwrap neuronal cell bodies (Coutinho-Budd et al., 2017). The loss of neuronal wrapping by cortex glia in Rab5 RNAi might influence the glia-to-neuron signaling pathway that is activated in zyd animals to increase their seizure susceptibility. To test whether the rescue effect of Rab5 knockdown is due to an impairment in the structure of glial-neuronal contacts, we conditionally expressed Rab5 RNAi and Rab5 DN both in 3 rd instar larvae and in adult cortex glia. Adult flies incubated for 16 hr at the restrictive temperature for gal80 ts (31˚C) to allow Rab5 RNAi expression, showed a partial rescue in zyd seizures, with only~25% of flies showing HS-induced seizures ( Figure 7B). Adult flies and 3 rd instar larvae conditionally expressing Rab5 DN showed a partial rescue of zyd seizures with~40% of animals displaying wildtype behavior ( Figure 7B). These results suggest that Rab5, a master regulator of endocytosis and EE biogenesis, plays a key role for the pathway that is activated in zyd cortex glia to promote seizures. To assay if excess endocytosis secondary to CN hyperactivity disrupts membrane trafficking in cortex glia, we imaged endosomal compartments by over-expressing GFP-tagged Rab5 in cortex glial cells (with GMR54H02-gal4) in control and zyd animals. We found that large (>0.1 mm 2 ) Rab5-positive early endosomes accumulated in zyd cortex glia compared to controls ( Figure 7C,D). Feeding zyd larvae the CN inhibitor, CsA (1 mM), restored the number of Rab5 compartments to control levels ( Figure 7D). These results indicate CN hyperactivation secondary to elevated Ca 2+ levels in zyd mutants increases endocytosis and the formation and accumulation of early endosomes in cortex glia.

Chronic inhibition of dynamin-mediated endocytosis rescues zyd seizures
Our previous analysis of zyd indicated that basal intracellular Ca 2+ is elevated in cortex glia, with Ca 2+ levels increasing more when zyd animals are heat-shocked (Melom and Littleton, 2013). The temperature-induced elevation in Ca 2+ could further enhance CN activity and endocytosis beyond that observed at rest. These data raise the question of whether the basal enhancement of endocytosis or the additional heat shock-induced Ca 2+ increase is the primary cause for seizure susceptibility in zyd mutants. To directly assay the role of endocytosis in seizure susceptibility in zyd flies, we :GFP puncta in zyd cortex glia relative to wildtype cortex glia. Rab5::GFP was expressed using a cortex glial-specific driver (GMR54H02-gal4; scale bar = 5 mm. n ! 5 animals/genotype). (D) Analysis of the number of large (>0.1 mm 2 ) Rab5::GFP puncta in wildtype and zyd cortex glia. The number of Rab5::GFP puncta in zyd cortex glia was increased relative to wildtype (average of 4.64 ± 0.58 and 2.85 ± 0.37 puncta/ROI respectively, p=0.0088). The number of large Rab5::GFP puncta in zyd treated with 1 mM CsA for 24 hr was decreased relative to zyd (average of 3.19 ± 0.37 puncta/ROI, p=0.0378; n ! 25 ROIs/3 animals/genotype/treatment). (E-F) Conditional inhibition of endocytosis by cortex glial overexpression of shi ts . (E) Schematic representation of the experimental design. Adult zyd; NP2222 >shi ts male flies (>1 day old) were incubated at the shi ts restrictive temperature (31˚C) for 3 or 6 hr and then tested for HS-induced seizures (red arrowheads, N = 3 groups of >15 flies/time point). (F) Behavioral analysis of HS induced seizures. Left: A significant reduction in seizures is observed in flies that were incubated at 31˚C for 3 hr (p=0.0283) or 6 hr (p=0.013). (N = 3 groups of >15 flies/time point). Right: zyd;NP2222 >Shi ts flies seizures re-occur after removal from the Shi ts restrictive temperature (25˚C, N = 4 groups of 10-15 animals/time point; 12 hr: p<0.0001). Error bars are SEM, *=P < 0.05, **=P < 0.01, ***=P < 0.001, ****=P < 0.0001, Student's t-test. DOI: https://doi.org/10.7554/eLife.44186.023 Figure 7 continued on next page conditionally manipulated endocytosis by overexpressing a TS dominant-negative form of Dynamin-1 (Shi ts ) in zyd cortex glia. This mutant version of Dynamin has normal activity at room temperature and a dominant-negative function upon exposure to the non-permissive temperature (>29˚C, Figure 7E). Acute inhibition of endocytosis by inactivation of Shi ts in cortex glia did not suppress zyd seizures, suggesting further enhancement of CN activity and endocytosis specifically during the heat shock is not likely to be the cause of the rapid-onset seizures observed in zyd mutants. Given the chronic enhancement in CN activity and endocytosis in zyd mutants demonstrated by enhanced CalexA signaling (Figure 3) and early endosome accumulation ( Figure 7C-D), we hypothesized that inhibiting endocytosis prior to exposing animals to a heat shock might improve their phenotype by altering the plasma membrane protein content over longer timescales. We incubated zyd; NP2222>Shi ts flies at a non-permissive temperature for Shi ts (31˚C) for either 3 or 6 hr, and then tested for heat shock-induced seizures at 38.5˚C ( Figure 7E-F). Zyd mutants alone do not undergo seizures at 31˚C, nor does pre-incubation at 31˚C alter the subsequent seizure phenotype observed at 38.5˚C. In contrast, inhibition of endocytosis for 6 hr at 31˚C in zyd mutants co-expressing Shi ts suppressed the subsequent seizures observed during a 38.5˚C heat shock in~80% of animals ( Figure 6F). A shorter 3 hr inhibition caused a less significant improvement in seizures. The seizure suppression observed after 6 hr of Dynamin inhibition was reversible, as adults tested 12 or 24 hr after return to room temperature regained the zyd seizure phenotype ( Figure 7F, right). We conclude that chronic hyperactivation of CN and endocytosis caused by elevated basal Ca 2+ in zyd cortex glia is the primary cause for zyd seizures.

Artificially increasing glial K + uptake rescues zyd-dependent seizures
Genetic analysis of zyd indicate the primary cause of seizure susceptibility is chronic enhancement in Ca 2+ -dependent CN activity and subsequent increases in endocytosis in cortex glia. We hypothesize that this enhancement in endocytosis leads to increased internalization of sand, which in turn disrupts K + uptake and buffering by cortex glial cells during periods of intense neuronal activity ( Figure 8A). If the cause of zyd seizures is a reduction in K + buffering, we predicted that increasing it by over-expression of sand will improve potassium homeostasis and reduce seizures. Indeed, panglial overexpression of sand:GFP partially suppressed zyd HS-induced seizures ( Figure 8C). It is likely that a full rescue was not observed because the over-expressed channel is still internalized and degraded in zyd cortex glia beyond that found in controls ( Figure 6D-G). To test this model further, we assayed if artificially increasing cortex glial K + uptake in zyd mutants by overexpressing another K + leak channel could suppress the seizure phenotype. Constitutive cortex glial overexpression of the open K + channel EKO (White et al., 2001) rescued vortex-induced seizures in~75% of zyd mutants ( Figure 8B). During a heat shock, cortex glial overexpression of EKO led to a dramatic change in the behavior of~60% of zyd animals, showing partial recovery from the seizure phenotype to bottom dwelling and hypoactivity ( Figure 8C). CPG recordings revealed that zyd;NP2222>EKO larvae regain rhythmic muscle activity at 38.5˚C ( Figure 8D-E), indicating cortex glial K + buffering is critical for neuronal excitability during states of intense excitation following heat shock or acute vortex. Together, these results indicate that enhanced CN-induced endocytosis lead to internalization and degradation of sand, impairment of cortex glial K + buffering and increased seizure susceptibility in zyd mutants.

Discussion
Significant progress has been made in understanding glial-neuronal communication at synaptic and axonal contacts, but whether glia regulate neuronal function via signaling at somatic regions remains largely unknown. Drosophila cortex glia provide an ideal system to explore how glia regulate neuronal function at the soma, as they ensheath multiple neuronal cell bodies, but do not contact synapses (Awasaki et al., 2008). In this study, we took advantage of the zydeco (zyd) mutation in a Na + /

Figure 7 continued
The following video is available for figure 7: Figure 7-video 1. The response of zyd/NP2222>Rab5 RNAi flies to a 38.5˚C heat-shock is shown. DOI: https://doi.org/10.7554/eLife.44186.024 Figure 8. Enhancing glial K + buffering by overexpressing a leak K + channel rescues zyd seizures. (A) A model for zyd function in seizure susceptibility is depicted. In wildtype cortex glia (left), oscillatory Ca 2+ signaling maintains normal cortex glia-to-neuron communication and a balanced extracellular ionic environment. In zyd cortex glia (right), the basal elevation of Ca 2+ leads to hyperactivation of CN and enhanced endocytosis with accumulation of early endosomes. This disrupts the endo-exocytosis balance of the K 2P leak channel sandman (and potentially other cortex glial membrane proteins) Figure 8 continued on next page Ca 2+ /K + exchanger to explore how cortex glia regulate neuronal excitability. We found that elevation of basal Ca 2+ levels in cortex glia leads to hyperactivation of Ca 2+ -CN dependent endocytosis. We showed that seizures in zyd mutants can be fully suppressed by either conditional inhibition of endocytosis or by pharmacologically reducing CN activity. Two well-characterized mechanisms by which glia regulate neuronal excitability and seizure susceptibility are neurotransmitter uptake via surface transporters and spatial K + buffering. Cortex glia do not contact synapses, making it unlikely they are exposed to neurotransmitters. Instead, cortex glial-knockdown of the two-pore K + channel (K 2P ) sandman, the Drosophila homolog of TRESK/KCNK18, recapitulates zyd TS seizures, while sand:GFP expression level in zyd is reduced in zyd relative to wildtype cortex glia. These findings indicate impairment in K + buffering during hyperactivity in zyd mutants underlies the increased seizure susceptibility, providing an unexpected link between glial Ca 2+ signaling and K + buffering. We propose that Drosophila cortex glia regulate the expression levels of sand (and potentially other K + channels and/or plasma membrane proteins) on the cell membrane in a Ca 2+ -regulated fashion in wildtype animals. When Ca 2+ is constitutively elevated in zyd mutants, this regulation is thrown out of balance. Together, these findings indicate that elevated Ca 2+ levels lead to hyperactivation of CN and elevated endocytosis, sand internalization, and impairment in K + buffering by cortex glia in zyd mutant animals ( Figure 8A).
Different glial subtypes exhibit dynamic fluctuations in intracellular Ca 2+ in vitro (Fatatis and Russell, 1992;Nett et al., 2002) and in vivo (Nimmerjahn et al., 2009;Porter and McCarthy, 1996). These early discoveries led to the model that astrocytes can listen to and regulate neuronal and brain activity. Accumulated findings indicate that glial Ca 2+ signaling influences neuronal physiology on a rapid time scale. However, the signaling pathways underlying these astrocytic Ca 2+ transients and their relevance to brain activity are poorly defined and controversial (Fiacco and McCarthy, 2018;Savtchouk and Volterra, 2018). The Drosophila zyd mutation was identified in an unbiased genetic screen for behavioral mutants that triggered TS-dependent seizures, thus establishing the biological importance of the pathway before the gene mutation and cellular origin of the defect was known. The elevation in cortex glial Ca 2+ levels found in zyd mutants provides a mechanism to explore how this pathway influences neuronal excitability. We recently found that Ca 2+ elevation in astrocyte-like glia results in the rapid internalization of the astrocytic plasma membrane GABA transporter GAT and subsequent silencing of neuronal activity through elevation in synaptic GABA levels (Zhang et al., 2017). As such, Ca 2+ -regulated endo/exocytic trafficking of neurotransmitter transporters and K + channels to and from the plasma membrane may represent a broadly used mechanism for linking glial Ca 2+ activity to the control of neuronal excitability at synapses and cell bodies, respectively.
Effective removal of K + from the extracellular space is vital for maintaining brain homeostasis and limits network hyperexcitability during normal brain function, as disruptions in K + clearance have been linked to several pathological conditions (David et al., 2009;Leis et al., 2005;Somjen, 2002). In addition to ion homeostasis, astrocytic K + buffering has been suggested as a mechanism for promoting hyperexcitability and engaging network activity (Bellot-Saez et al., 2017;Wang et al., 2012). Two mechanisms for astrocytic K + clearance have been identified, including net K + uptake (mediated by the Na + /K + ATPase pump) and K + spatial buffering (via passive K + influx) (Bellot-Saez et al., 2017). While many studies indicate K ir4.1 , a weakly inward rectifying K + channel exclusively expressed in glial cells, is an important channel mediating astrocytic K + buffering, it is unlikely to be the only mechanism for glial K + clearance, as K ir4.1 channels account for less than half of the K +

Drosophila genetics and molecular biology
Flies were cultured on standard medium at 22˚C unless otherwise noted. zydeco (zyd 1 , here designated as zyd) mutants were generated by ethane methyl sulfonate (EMS) mutagenesis and identified in a screen for temperature-sensitive (TS) behavioral phenotypes (Guan et al., 2005). The UAS/gal4 and LexAop/LexA systems were used to drive transgenes in glia, including repo-gal4 (Lee and Jones, 2005), a pan-glial driver; NP2222-gal4, a cortex-glial specific driver; GMR54H02-gal4, expressed in a smaller set of cortex glial cells; and alrm-gal4, an astrocyte-like glial cell specific driver. The UAS-dsRNAi flies used in the study were obtained from the VDRC (Vienna, Austria) or the TRiP collection (Bloomington Drosophila Stock Center, Indiana University, Bloomington, IN, USA). All screened stocks are listed in supplementary material (Supplementary file 2). UAS-myrG-CaMP6s was constructed by replacing GCaMP5 in the previously described myrGCaMP5 transgenic construct (Melom and Littleton, 2013). UAS-sand:GFP was constructed by fusing the ORF of sand (Drosophila melanogaster sandman (sand), mRNA, NM_136505) to eGFP and inserted into pBID-UASc between restriction sites EcoRI and XbaI (Epoch Life Science, Inc). Transgenic flies were obtained by standard germline injection (BestGene Inc). For all experiments described, only male larva and adults were used, unless otherwise noted. In RNAi experiments, the animals also had the UAS-dicer2 transgenic element on the X chromosome to enhance RNAi efficiency. For survival assays, embryos were collected in groups of~50 and transferred to fresh vials (n = 3). 3 rd instar larvae and/or pupae were counted. Survival rate (SR) was calculated as: For conditional expression using Tub-gal80 ts ( Figure 1E, 2F, 7B), animals of the designated genotype were reared at 22˚C with gal80 suppressing gal4-driven transgene expression (zyd RNAi , CanB2 RNAi and Rab5 DN and respectively). 3 rd instar larvae or adult flies were then transferred to a 31˚C incubator to inactivate gal80 and allow gal4 expression or knockdown for the indicated period. For inhibiting transgene expression in cortex glia ( Figure 5A) GMR54H02-lexA (a kind gift from Gerald Rubin collection) was used to express gal80 from LexAop-gal80. For inhibiting transgene expression specifically in neurons ( Figure 2E, Figure 2-figure supplement 1D), C155-gal80 was used.

Behavioral analysis
All experiments were performed using groups of~10-20 males.

Temperature-sensitive seizures/zyd modifier screen
Adult males aged 1-2 days were transferred in groups of~10-20 flies (n ! 3, total # of flies tested in all assays was always >40) into preheated vials in a water bath held at the indicated temperature with a precision of 0.1˚C. Seizures were defined as the condition in which the animal lies incapacitated on its back or side with legs and wings contracting vigorously (Melom and Littleton, 2013; Figure 1-video 1). For screening purposes, only flies that showed normal wildtype-like behavior (i.e. walking up and down on vial walls, Figure 1-video 1, Figure 2-video 4 and 5) during heatshock were counted as successful rescue. To analyze behavior in a more detailed manner, we characterized four behavioral phenotypes: wall climbing (flies are climbing on vial walls), bottom dwelling (flies are on the bottom of the vial, standing/walking without seizures), partial seizures (flies are on the bottom of the vial, seizing most of the time) and complete seizures (flies are constantly lying on their side or back with legs twitching). For assaying seizures in larvae, 3 rd instar males were gently washed with PBS and transferred to 1% agarose plates heated to 38˚C using a temperature-controlled stage (Melom and Littleton, 2013). Larval seizures were defined as continuous, unpatterned contraction of the body wall muscles that prevented normal crawling behavior (Melom and Littleton, 2013). For determining seizure temperature threshold, groups of 10 animals were heat-shocked to the indicated temperature (either 37.5, 38, 38.5 or 39˚C). Threshold was defined as the temperature in which > 50% of the animal were seizing after 1 minute.

Bang sensitivity
Adult male flies in groups of~10-20 males (n = 3) were assayed 1-2 days post-eclosion. Flies were transferred into empty vials and allowed to rest for 1-2 hr. Vials were vortexed at maximum speed for 10 seconds, and the number of flies that were upright and mobile was counted at 10 s intervals.

Light avoidance
These assays were performed using protocols described previously following minor modifications. Briefly, pools of~20 3 rd instar larvae (108-120 hr after egg laying) were allowed to move freely for 5 minutes on Petri dishes with settings for the phototaxis assay (Petri dish lids were divided into quadrants, and two of these were blackened to create dark environment). The number of larvae in light versus dark quadrants was then scored (n = 4). Response indices (RI) were calculated as:

RI ¼ N Dark total
Activity monitoring using the MB5 system Adult flies activity was assayed using the multi-beam system (MB5, TriKinetics) as previously described (Green et al., 2015;McParland et al., 2015). Briefly, individual males aged 1-3 days were inserted into 5 mm Â80 mm glass pyrex tubes. Activity was recorded following a 20-30 minutes acclimation period. Throughout each experiment, flies were housed in a temperature-and light-controlled incubator (25˚C,~40-60% humidity). Post-acquisition activity analysis was performed using Excel to calculate activity level across 1 minute time bins (each experimental run contained eight control animals and eight experimental animals, n ! 3).

Gentle touch assay
3 rd instar male larvae (108-120 hr after egg laying) were touched on the thoracic segments with a hair during forward locomotion. No response, a stop, head retraction and turn were grouped into type I responses, and initiation of at least one single full body retraction or multiple full body retractions were categorized as type II reversal responses. Results were grouped to 20 males per assay (n = 3).

Electrophysiology
Intracellular recordings of wandering 3 rd instar male larvae were performed in HL3.1 saline (in mm: 70 NaCl, 5 KCl, 4 MgCl 2 , 0.2 CaCl 2 , 10 NaHCO 3 , 5 Trehalose, 115 sucrose, 5 HEPES-NaOH, pH 7.2) containing 1.5 mm Ca 2+ using an Axoclamp 2B amplifier (Molecular Devices) at muscle fiber 6/7 of segments A3-A5. For recording the output of the central pattern generator, the CNS and motor neurons were left intact. Temperature was controlled with a Peltier heating device and continually monitored with a microprobe thermometer. Giant fiber recordings were performed and seizure thresholds were defined as previously described (Howlett and Tanouye, 2013;Pavlidis and Tanouye, 1995). Shortly, seizure-like activity was defined as uncontrolled, high-frequency (>100 Hz) motoneuron activity evoked by HFS stimulation and recorded in the DLM (Kuebler and Tanouye, 2000). Seizure threshold was the lowest HFS voltage that evoked seizure-like activity.
In vivo Ca 2+ imaging UAS-myrGCaMP6s was expressed in glia using the drivers described above. 2 nd instar male larvae were washed with PBS and placed on a glass slide with a small amount of Halocarbon oil #700 (Lab-Scientific). Larvae were turned ventral side up and gently pressed with a coverslip and a small iron ring to inhibit movement. Images were acquired with a PerkinElmer Ultraview Vox spinning disk confocal microscope and a high-speed EM CCD camera at 8-12 Hz with a 40 Â 1.3 NA oil-immersion objective using Volocity Software. Single optical planes within the ventral cortex of the ventral nerve cord (VNC) were imaged in the dense cortical glial region immediately below the surface glial sheath. Average myrGCaMP6s signal in cortex glia was quantified in the central abdominal neuromeres of the VNC within a manually selected ROI excluding the midline glia. Ca 2+ oscillations were counted within the first minute of imaging at room temperature, and then normalized to the ROI area.

Drug feeding
Cyclosporin-A (CsA, Sigma Aldrich) or FK506 (InvivoGen) were dissolved in DMSO to a final concentration of 20 mM. The feeding solution included 5% yeast and 5% sucrose in water. Adult males less than 1 day old were starved for 6 hours and then transferred to a vial containing a strip of Wattman paper soaked in feeding solution containing the designated concentration of CsA/FK506 or DMSO as control. Flies were behaviorally tested following 6, 12 or 24 hours of drug feeding.