Glial Ca2+ Signaling Controls Endocytosis and K+ Buffering to Regulate Glial-Neuronal Communication at the Soma

Glial-neuronal signaling at synapses is widely appreciated, but how glia interact with neuronal cell bodies is less clear. Drosophila cortex glia are restricted to brain regions devoid of synapses, providing an opportunity to characterize interactions between glia and neuronal somas. Mutations in the cortex glial NCKX exchanger zydeco abolish microdomain Ca2+ oscillatory activity and elevate glial Ca2+, predisposing animals to seizures. To determine how cortex glial Ca2+ signaling controls neuronal excitability, an in vivo modifier screen for the NCKXzydeco seizure phenotype was performed. Our results indicate elevation of glial Ca2+ causes hyperactivation of calcineurin-dependent endocytosis and accumulation of early endosomes. Knockdown of sandman, a K2P channel, recapitulates NCKXzydeco seizures. Restoring glial K+ buffering by overexpressing a leak K+ channel rescues zydeco seizures. These findings indicate cortex glial Ca2+ couples to K+ buffering through calcineurin regulated endo-exocytotic balance and K2P channel expression to modulate 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. In the cortex, glia and neurons exist in equal abundance (Azevedo et al., 2009) and are intimately associated. 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 (Gomez-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.
All major CNS glial subtypes interact closely with neuronal cell bodies. Protoplasmic astrocytes form extensive contacts on neuronal cell bodies (Allen and Barres, 2009); satellite microglia form contacts between their cell bodies and those of neurons (Baalman et al., 2015); and specialized groups of perineuronal oligodendrocytes reside on neuronal cell bodies (Battefeld et al., 2016;Takasaki et al., 2010). While significant progress has been made in characterizing the contacts between glial cells and axons and synapses, less is understood of the signaling events between any type of glial process and neuronal cell bodies. In this regard, Drosophila provides an ideal system to study glial-neuronal soma interactions as the CNS is compartmentalized into two primary regions: the cell cortex (consisting of neuronal cell bodies and their proximal axons) and the synaptic neuropil (containing all CNS neurites and synapses). Two specialized astrocyte-like glial subtypes, astrocytes and cortex glia, are compartmentalized into these brain regions. Astrocytes extend processes that are confined to the neuropil and associate with synapses, axons, and dendrites . Drosophila astrocytes are similar to mammalian protoplasmic astrocytes in terms of their morphology, tiling behavior, expression profiles, and functions in synapse formation (Muthukumar et al., 2014;Stork et al., 2014), synaptic pruning , and regulation of synaptic function . Cortex glia, in contrast, are restricted to the cell cortex region that is devoid of synapses (Awasaki et al., 2008;Pereanu et al., 2005). Cortex glia extend fine processes into the cell cortex region and encapsulate virtually every neuronal cell body in the CNS (Awasaki et al., 2008) (Fig. S1A). Given these tight associations, cortex glia are thought to provide metabolic support and key nutrients to neurons (Volkenhoff et al., 2015). In addition, cortex glia have been shown to exhibit local microdomain Ca 2+ transients close to neuronal cell bodies that may regulate their function (Melom and Littleton, 2013). However, the role of cortex glia in modulating neuronal activity is unclear.
Previous work in our lab identified zydeco (zyd), a cortex glial enriched Na + Ca 2+ K + (NCKX) exchanger involved in maintaining normal neural excitability. Mutations in NCKX zydeco (hereafter referred to as zyd) predispose animals to temperature-sensitive seizures and result in bang sensitivity (seizures induced 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. These findings indicate disruption of glial Ca 2+ regulation triggers enhanced seizure susceptibility. To determine how cortex glial Ca 2+ signaling regulates neuronal excitability, we took advantage of the zyd mutation and performed an RNAi screen for modifiers of the seizure 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, sandman, a K2P channel, recapitulates zyd seizures when knocked down and acts downstream of calcineurin in cortex glia. Sandman was previously identified as the key K + channel that cycles to and from the plasma membrane in a group of central complex neurons to modulate their excitability and control sleep homeostasis in the Drosophila circadian pathway (Pimentel et al., 2016). Our findings suggest similar regulation of sandman in cortex glia could allow dynamic control of K + levels surrounding neuronal somas as a mechanism to gate neuronal excitability. Indeed, overexpression of 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 glial-based therapeutic modifiers of epilepsy. response is recorded. By progressively increasing stimulation intensity, the voltage threshold that triggers seizure activity in the GF circuit can be directly assayed, providing a readout of neuronal excitability. Most Drosophila bang-sensitive mutations display seizure induction at much lower voltages, consistent with their primary effect on membrane excitability thresholds within the neuron. For example, para bss1 , a gain-of-function bang sensitive mutation in the voltage-gated Na + channel, dramatically lowers seizure threshold (Fig. 1B) (Parker et al., 2011). In contrast, the voltage threshold for seizure induction in zyd was not significantly different from controls (Fig. 1B), indicating basic intrinsic neuronal properties are not altered in zyd animals at rest. 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. Panneuronal 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 neuronal activity (Fig. S1J) and rescued zyd TS-induced seizures (Fig. 1C). 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 (Tables S1, S2). 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 zyd (5), enhanced zyd seizures (37) or triggered seizures (3) in a wildtype background (Fig. 1D, Table S1). 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 identified multiple candidates controlling glial-toneuron communication, including regulators of vesicular trafficking, neurotransmitter receptors and ion channels, and Ca 2+ effector proteins that suppress zyd seizures when knocked down specifically in glia (Fig. 1D, Table S1).
Beyond regulators of the zyd phenotype, the screen also identified Stim (stromal interaction molecule), an ER Ca 2+ sensor for store operated Ca 2+ entry (SOCE), as a key Ca 2+ signaling component involved in TS seizure generation. Knockdown of Stim in wildtype adult animals caused seizures that were similar to those observed in zyd mutants (Fig. 1E), suggesting defects in glial microdomain Ca 2+ (zyd) or the SOCE pathway both increase neuronal seizure susceptibility. Recordings of motor central pattern generator (CPG) output at the larval neuromuscular junction (NMJ) showed that Stim RNAi larvae exhibit rapid, unpatterned firing at 38°C, similar to zyd mutants, whereas wildtype larvae retain motor neuron bursting necessary for normal crawling behavior (Fig. 1F). To determine if these distinct Ca 2+ entry pathways impinged on a similar mechanism for seizure induction, we tested genetic interaction between SOCE defects and zyd mutants. If both Ca 2+ entry mechanisms impinged on the same pathway, they should not show synergistic interactions when both are disrupted. However, knockdown of Stim in the zyd background enhanced zyd seizures (Fig. 1E), indicating additive effects and suggesting Ca 2+ is likely to interface with distinct glial-neuronal pathways depending upon the source and location of aberrant Ca 2+ signaling. Indeed, zyd-dependent seizures require calcineurin function, while SOCE pathway dysfunction does not (see below). Distinctions between microdomain and ER-dependent Ca 2+ signaling pathways have been previously observed in mammals as well (Bindocci et al., 2017;Savtchouk and Volterra, 2018).
Given TS seizures in zyd mutants can be fully rescued by reintroducing the wildtype zyd (NCKX) protein in cortex glia, we expected the genes identified in the RNAi pan-glial knockdown screen to function specifically within this population of cells. To directly examine cell-type specificity of the suppressor hits, the RNAi screen was repeated with drivers restricted to either astrocyte or cortex glial subtypes (Alrm-gal4 for astrocytes and NP2222-gal4 and GMR54H02-gal4 for cortex glia, Fig. 1D, Table S1). For the majority of suppressors, rescue with the glialsubtype specific drivers was weaker, likely due to lower RNAi expression levels compared to the stronger repo-gal4 driver line (Fig. 1D). In several cases (nompC, Nmdar2), however, near complete rescue of the zyd seizure phenotype could be achieved when the gene was knocked down using the astrocyte driver Alrm-gal4 (Fig.1D). This surprising observation indicates that although seizure induction in zyd mutants arises from abnormal Ca 2+ signaling within cortex glia, normal astrocyte function is required for either the propagation or maintenance of the subsequent excitability changes that drive the TS seizure phenotype. We recently discovered that Drosophila astrocytes employ Ca 2+ -mediated signaling pathways to dynamically regulate the surface levels of the GABA transporter, GAT, which negatively controls neuronal activity through modulation of synaptic GABA levels (Zhang et al., 2017). Together with the RNAi suppressor screen hits that function within astrocytes, these data indicate astrocytes regulate the seizure-inducing defects arising in cortex glia in the zyd mutant. For the remaining analysis, we focused on the characterization of the cortex glial Ca 2+ -dependent pathway that is 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. Calcineurin (CN) is a highly conserved Ca 2+ /Cam-dependent protein phosphatase implicated in a number of cellular processes in mammals (Rusnak and Mertz, 2000). In the RNAi screen for zyd interactors, panglial knockdown of the regulatory CN B subunit, CanB2, completely rescued both heat-shock and vortex induced seizures in zyd animals (Fig. 1D, 2A-C). In contrast to the continuous neuronal firing observed in zyd mutants, CPG recordings demonstrated that zyd;;repo>CnB2 RNAi#1 larvae exhibit normal rhythmic firing at 38°C similar to controls (Fig. 2B). Similar results were observed using three additional, non-overlapping CanB2 RNAi constructs ( Fig. 2A, 2F). To refine the glial subpopulation in which CanB2 activity is necessary to promote seizures in zyd mutants, we knocked CanB2 down using different glial drivers. CanB2 knockdown in astrocytes resulted in no rescue of the zyd phenotype (Fig. 1D). CanB2 knockdown with a cortex-glial specific driver (NP2222) greatly improved the zyd phenotype. Animals no longer displayed continuous seizures, but the rescue was less robust compared to pan-glial knockdown, likely due to lower expression of the RNAi (Fig. 2D, Fig. S2A). Indeed, rescue was greatly enhanced by cortex glial-specific knockdown of CanB2 using two copies of the CanB2 RNAi construct (Fig. 2D), 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 the CanB2 RNAi in neurons using C155-gal80 (a neuron specific gal4 repressor; Fig. 2D and S2A). To exclude a developmental effect of CanB2 knockdown within glia, we blocked gal4/UAS-driven CanB2 RNAi with gal80 ts and expressed a single copy of CanB2 RNAi only in adult zyd mutants. Adult flies reared at the permissive temperature for gal80 ts to allow CanB2 RNAi expression exhibited significantly less seizures after 3 days, with only ~20% of flies displaying zyd-like seizures by 5 days (Fig. 2E). 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 responses (Fig. S2B) and adult locomotion (Fig. S2C). In contrast to rescue of the zyd phenotype, CanB2 knockdown could not rescue TS seizures induced by increasing glial Ca 2+ through overexpression of the Ca 2+ channel TRPA (Fig S2D) or RNAi knockdown of Stim (Fig. S2E), indicating glial-triggered seizures in these genetic backgrounds employ distinct Ca 2+ -regulated pathways to alter neuronal excitability. We conclude that CanB2 is required in cortex glia to promote zyd TS seizure activity.
CN is a heterodimer composed of a ∼60 kDa catalytic subunit (CanA) and a ∼19 kDa EFhand Ca 2+ -binding regulatory subunit (CanB). Both subunits are essential for CN phosphatase activity. 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. The Drosophila genome contains three genes encoding CanA (CanA1, Pp2B-14D and CanA-14F) and two genes encoding CanB (CanB and CanB2) (Takeo et al., 2006). We found that pan-glial knockdown of two CanA subunits, Pp2B-14D and CanA-14F, partially rescued zyd heat-shock and vortex induced seizures (Fig. 2C, F). 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. Rescue was enhanced by knockdown of both Pp2B-14D and CnA-14F (Fig. 2G), with more than ~90% of zyd;;repo>2xCanA RNAi flies lacking seizures, suggesting a redundant function of these two subunits in glial cells. To further verify the effect of CanA knockdown and the glial subpopulation in which CN activity is required, we overexpressed a dominant negative (DN) 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 seizureresistant, regardless of the driver used ( Fig 2H). Imaging of intracellular Ca 2+ in cortex glia with GMR54H02-gal4 driven myrGCaMP6s revealed that CN knockdown had no effect on either the elevated basal Ca 2+ levels or the lack of microdomain Ca 2+ events observed in zyd cortex glia ( Fig. 2I-J). 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 glialneuronal communication 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 system as a reporter for CN activity. CalexA (Ca 2+ -dependent nuclear import of LexA) (Masuyama et al., 2012) was originally designed for labeling active neurons in behaving animals. In this system, 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 in active neurons (for schematic representation, see Fig. S2F). 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 (Fig.  3A). CN activity and the resulting GFP expression was enhanced in zyd cortex glia (Fig. 3B). 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 as well (23 ± 3% enhancement, Fig 3D-E). RNAi knockdown of CanB2 greatly reduced CalexA GFP expression as expected (27 ± 1% reduction, Fig. 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 targeting of the glial calcineurin pathway rescues zyd seizures
Several seizure mutants in Drosophila can be suppressed by commonly used antiepileptic drugs (Kuebler and Tanouye, 2002;Song and Tanouye, 2008), indicating conservation of key mechanisms that regulate neuronal excitability. CN activity is strictly controlled by Ca 2+ levels, calmodulin, and CanB, and can be inhibited by the immunosuppressants cyclosporine A (CsA) and FK506. The CN inhibitor FK506 has been previously shown to reduce seizures in a rodent kindling model (Moia et al., 1994;Moriwaki et al., 1996), suggesting CN can modulate epilepsy in mammals. To assay if zyd TS seizures can be prevented with anti-CN drugs, adult flies were fed with media containing different anti-CN drugs (CsA, FK506 and CN585 (Erdmann et al., 2010)) and tested for HS induced seizures after 0, 3, 6, 12 and 24 hours of drug feeding (red arrowheads in Fig. 4A, Fig. 4B-C). Zyd flies fed with 1 mM CsA for 12 hours showed ~80% less seizures then control (Fig. 4B, C), while two other inhibitors, FK506 and CN585, had much weaker effects (Fig. 4B). Seizure rescue by CsA was dose-dependent, with less robust suppression when flies were fed with 0.3 mM CsA ( Fig. 4D-E). The CsA rescue was reversible, as zyd seizures reoccurred following 12 hours of CsA withdrawal (Fig. 4C). We conclude that pharmacologically targeting the glial CN pathway can improve the outcome of glial-derived neuronal seizures in the zyd mutant.

Cortex glial knockdown of the sandman two-pore-domain K + channel 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 (Table S3). 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 (K2P), caused adult flies to undergo TSinduced seizures similar to zyd mutants (Fig. 5A). Vortex-induced seizures in repo>sand RNAi were less severe than those observed in zyd, with only ~50% of sand RNAi flies showing seizures (Fig.  S3A). TS-induced seizures in repo>sand RNAi adults were found to have the same kinetics and temperature threshold as seizures observed in zyd mutants (Fig. 5A, B), and CPG recordings showed that repo>sand RNAi larvae exhibit rapid, unpatterned firing at 38°C, similar to zyd mutants (Fig. 5C). Cortex-glial specific knockdown of sand recapitulated ~50% of the seizure effect when two copies of the RNAi were expressed (Fig. 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 glialknockdown of sand, seizure generation was suppressed (Fig. 5A). Similar to zyd mutants, sand RNAi animals did not show changes in general activity and locomotion at room temperature ( Fig. 5D). Finally, unlike the case with Stim removal in the SOCE pathway, RNAi knockdown of sand did not enhance the zyd phenotype (Fig. S3B), suggesting 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 Kir4.1 potassium channels (KCNJ10) have been implicated in maintaining K + homeostasis, with mutations in the loci causing epilepsy (Haj-Yasein et al., 2011). However, Kir channels are unlikely to be the only mechanism for glial K + clearance, as Kir4.1 channels account for less than half of the K + buffering capacity of mature hippocampal astrocytes (Ma et al., 2014). To determine if cortex glial Kir channels regulate seizure susceptibility in addition to sand, we used repo-Gal4 to knock down all three Drosophila Kir family members (Irk1, Irk2 and Irk3). Pan-glial knockdown of the Drosophila Kir family did not cause seizures (Fig. S3C), while knock down of either Irk1 or Irk2 slightly enhanced the zyd phenotype (Fig. S3D). Similarly, repo-gal4 knockdown of other well-known Drosophila K + channels beyond the Kir family also did not cause seizures (Fig. S4C), indicating sand is likely to play a preferential role in K + buffering in Drosophila cortex glia.
The mammalian sand homolog, TRESK, is directly activated by CN dephosphorylation (Czirjak et al., 2004;Enyedi and Czirjak, 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 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 (Fig. 5E), 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 (Fig. 5F, G), 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 (i.e. heat shock or acute vortex) causes enhanced seizure susceptibility in zyd mutants.

Enhanced endocytosis in zyd cortex glia
We next sought to examine how elevated CN activity in zyd mutants alters sand function. The mammalian sand homolog, TRESK, is constitutively phosphorylated on four serine residues (S264 by PKA, and S274, S276 and S279 by MARK1) (Enyedi and Czirjak, 2015). Two of these residues are conserved in Drosophila sand (S264 and S276, see Fig. S3E for protein alignment). Constitutive dephosphorylation and subsequent activation of sand by CN would be predicted to increase K + buffering following hyperactivation of the nervous system by stressors, and thus less seizure activity would be expectedopposite to what we have observed. If this regulatory mechanism is active in Drosophila cortex glia, we hypothesized that knockdown of either PKA or Par-1 (the Drosophila MARK1 homolog) would lead to enhanced activity of sand and improvement or rescue of zyd seizures. However, neither pan-glial or cortex glial knockdown of either kinase (PKA-C1, PKA-C2, PKA-C3 and Par-1), or overexpression of a PKA inhibitory peptide (PKI), altered the zyd phenotype (Table S4). 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.
A second mechanism to link CN hyperactivation to sand regulation is suggested by a previous study demonstrating sand expression on the plasma membrane of neurons involved in sleep homeostasis 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, 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. In contrast, cortex glial knockdown of Rab5, a Rab GTPase regulator of early endosome (EE) dynamics (Langemeyer et al., 2018), and Endophilin A (EndoA), a BAR-domain protein involved in early stages of endocytosis (Verstreken et al., 2002), completely suppressed zyd TS seizures (Fig. 6A). A second, non-overlapping hairpin and a dominant negative (DN) construct for Rab5 (Rab5 DN ) resulted in early larval lethality (Table  S5), likely due to more efficient Rab5 activity suppression. However, conditionally expressing Rab5 DN in adult cortex glia partially rescued zyd seizures (Fig. 6B). Rab5 is engaged in the initial step of vesicle endocytosis and recycling (Dunst et al., 2015), suggesting a key role for this process in regulating glial-to-neuronal soma signaling. To determine if the observed rescue was specifically associated with early endocytosis defects, we assayed seizure suppression in zyd mutants following knockdown of the entire family of Drosophila Rab GTPases, most of which are expressed in Drosophila cortex glia (Coutinho-Budd et al., 2017). Beyond Rab5, none of the remaining Rab proteins altered the zyd seizure phenotype or caused a behavioral phenotype on their own following RNAi-mediated knockdown (Table S5).
To assay if excess endocytosis secondary to CN hyperactivity disrupts membrane trafficking, we imaged endosomal compartments by expressing GFP-tagged Rab proteins in cortex glial cells of control and zyd animals. We found that large (>0.1 μm 2 ) Rab5-positive early endosomes accumulated in zyd cortex glia compared to controls (Fig. 6C). Feeding zyd larvae the CN inhibitor, CsA (1 mM), restored the number of Rab5 compartments to control levels (Fig.  6D). These results indicate CN hyperactivation secondary to elevated Ca 2+ levels in zyd mutants increases endocytosis and the formation of early endosomes in cortex glia.

Chronic inhibition of dynamin-mediated endocytosis rescues zyd seizures
Our previous analysis of the zyd mutant indicated that basal intracellular Ca 2+ is elevated in cortex glia, with Ca 2+ levels increasing even more when zyd animals are heat-shocked (Melom and Littleton, 2013). This additional elevation in Ca 2+ could potentially 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 test these two models, we 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, Fig. 6F). 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 (Fig. 3) and early endosome accumulation ( Fig. 6C-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 hours, and then tested for heat shock-induced seizures at 38.5˚C (Fig. 6E-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 hours 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 (Fig. 6F). A shorter 3 hour inhibition did not cause a significant improvement in seizures. The seizure suppression observed after 6 hours of Dynamin inhibition was reversible, as adults tested 12 or 24 hours after return to room temperature regained the zyd seizure phenotype (Fig.  6G). 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.

Increasing glial K + uptake rescues zyd-dependent seizures
Genetic analysis of the zyd mutant 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 this enhancement in endocytosis leads to increased internalization of plasma membrane proteins such as sand, which in turn disrupt K + uptake and buffering by cortex glial cells during periods of intense activation of the nervous system (Fig. 7A). 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. Indeed, constitutive cortex glial overexpression of the open K + channel EKO (White et al., 2001) rescued vortex-induced seizures in ~75% of zyd mutants (Fig 7B). During a heat shock, cortex glial overexpression of EKO led to a dramatic change in the behavior of ~60% of zyd animals, changing the seizure phenotype to hypoactivity or complete paralysis (Fig. 7C). CPG recordings revealed that zyd;;repo>EKO larvae lose all bursting and normal CPG activity at 38˚C (Fig. 7D, middle), while zyd;NP2222>EKO regain normal wildtype-like rhythmic activity (Fig. 7D, right). These results indicate cortex glial K + buffering is critical for neuronal excitability during states of intense excitation as observed following heat shock or acute vortex. During these periods of intense neuronal activity in zyd mutants, defective cortex glial K + buffering surrounding neuronal somas leads to seizures. Enhancing K + buffering can either reverse the seizure phenotype or push the scales toward neuronal hypo-excitability and paralysis. Future studies will be required to determine if sand function is dynamically modulated by normal microdomain Ca 2+ oscillatory activity in wildtype cortex glia in response to changes in neuronal activity, which would provide a robust glial-based homeostatic mechanism to maintain neuronal spiking rates in acceptable ranges.

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 unstudied. A single mammalian astrocyte can be associated with multiple neuronal cell bodies and thousands of synapses, making it challenging to direct manipulations that alter glial signaling only at neuronal somas. Drosophila cortex glia ensheath multiple neuronal soma but do not contact synapses (Awasaki et al., 2008). These tight associations with neuronal cell bodies make cortex glial an ideal system to explore how glia regulate neuronal function at the soma. In this study, we took advantage of the zydeco TS seizure mutation in a NCKX Ca 2+ exchanger to explore pathways by which cortex glial cells regulate neuronal excitability. We found that elevation of basal Ca 2+ levels in cortex glia leads to hyperactivation of Ca 2+ -CN dependent endocytosis. 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 (K2P) sandman, the Drosophila homolog of TRESK/KCNK18, recapitulates zyd TS seizures. These findings suggest 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. Consistent with this model (Fig. 7A), enhancing cortex glial K + uptake by overexpressing a constitutively open K + channel (EKO (White et al., 2001)) reduces neuronal activity in zyd mutants, rescues vortex-induced seizures, and reverses the TS behavioral phenotype from seizures to paralysis.
Astrocytes, as well as other 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). Despite decades of studies on astrocytic Ca 2+ activity, signaling pathways underlying these transients and their in vivo relevance to brain activity are poorly defined and controversial (Fiacco and McCarthy, 2018;Savtchouk and Volterra, 2018). Prior studies examining astrocytic Ca 2+ signaling used several approaches to artificially increase intracellular Ca 2+ (transgenic receptor expression, caged Ca 2+ photolysis and pharmacological or optogenetic stimulation) (Savtchouk and Volterra, 2018). However, it is unclear if these methods mimic physiological astrocytic responses and how these manipulations alter in vivo behavior (Agulhon et al., 2010;Fiacco et al., 2007;Wang et al., 2012a;Wang et al., 2012b). Likewise, the distinction between signaling events mediated by local microdomain Ca 2+ oscillatory activity that require extracellular Ca 2+ entry versus the more well-studied IP3-dependent release of Ca 2+ from ER stores has only been recently appreciated (Bindocci et al., 2017;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. While we focused on CN-dependent endocytosis and K + buffering by cortex glial cells in zyd animals, our study uncovered evidence for additional pathways by which glia modulate neuronal excitability as well. Knockdown of Stim in glia, a key regulator of SOCE, also caused TS seizures but did not require CN activity, suggesting glia possess more than one Ca 2+ -sensitive pathway that regulates neuronal excitability. In addition, we found evidence that astrocyte-like glia modulate the expression of seizures that arise from elevated Ca 2+ signaling within cortex glia, as several RNAi suppressor hits rescue zyd seizures when knocked down using an astrocyte-specific driver. 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.
CN is the only Ca 2+ /Cam-dependent phosphatase encoded in the genome and is highly expressed throughout the brain (Furman and Norris, 2014;Goto et al., 1986a, b;Kuno et al., 1992;Polli et al., 1991). CN interacts with numerous neuronal substrates to modulate diverse functions, including receptor and ion channel trafficking, ion channel function and gene regulation (Baumgartel and Mansuy, 2012). Neuronal CN also controls synapse loss, dendritic atrophy, synaptic dysfunction, and neuronal vulnerability (Abdul et al., 2010;Reese and Taglialatela, 2011). In contrast, astrocytic CN expression has been shown to increase during aging, injury and disease. Glial cells rely on CN signaling pathways to regulate phenotypic switching/cellular activation, immune-like responses and cytokine production (Furman and Norris, 2014), with the pathway being intimately involved in neuroinflammation (Furman and Norris, 2014;Kataoka et al., 2009;Nagamoto-Combs and Combs, 2010;Rojanathammanee et al., 2013;Shiratori et al., 2010). The role of CN in astrocytes during normal brain states is unclear . The identification of calmodulin (Melom and Littleton, 2013) and CN as suppressors of glial-originated seizures in zyd mutants indicates a Cam/CN-dependent pathway is hyperactivated and impairs normal cortex glial-to-neuronal soma signaling. Although the mechanism by which CN activity is upregulated in Drosophila cortex glia is different from that observed in mammals, the ability of CN to alter neuronal activity appears similar to mechanisms employed in glial-dependent neuroinflammatory states. During injury or disease states in mammals, activated astrocytes exhibit Ca 2+ dysregulation with higher intracellular Ca 2+ levels, more frequent Ca 2+ oscillations and an elevated expression of Ca 2+ channels and Ca 2+ -regulated proteins (Kuchibhotla et al., 2009;Lin et al., 2007;Sama and Norris, 2013). While its clear how these changes would hyperactivate CN signaling, which has extreme consequences for neuronal function, it is unclear what role basal CN activity has in glial signaling. CN regulates the expression of several key Ca 2+ mediators in multiple cell types Genazzani et al., 1999;Graef et al., 1999;Groth et al., 2007), including plasma membrane Ca 2+ channels, intracellular Ca 2+ release channels, and Ca 2+dependent enzymes (Baumgartel and Mansuy, 2012). However, no obvious phenotypes were observed when we decreased CN activity in CanB2 RNAi cortex glia in control animals (Fig. 3C), suggesting the primary function of CN is only engaged following states of Ca 2+ elevation in cortex glia during periods of strong neuronal activity.
The link between elevated glial Ca 2+ signaling and defects in K + buffering was an unexpected observation in the zyd mutant. 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 specific network activity (Bellot-Saez et al., 2017;Wang et al., 2012a). Two mechanisms for astrocytic K + clearance have been previously 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). However, the complete repertoire of K + channels involved in spatial K + buffering remains elusive (Ma et al., 2014). We found that cortex glial knockdown of the Drosophila KCNK18/TRESK K2P homolog sand triggered stress-induced seizures, indicating sand is involved in K + homeostasis in the brain. Several studies have linked other members of the K2P family, mainly TREK-1 and TWIK-1, to distinct aspects of astrocytic function. TREK-1 regulates fast glutamate release by astrocytes (Woo et al., 2012), TWIK-1 and TREK-1 mediate passive K + conductance in astrocytes (Hwang et al., 2014), and TWIK-1 is recruited to the astrocytic membrane by mGluR3 activation (Wang et al., 2016).
We considered two potential mechanisms for CN regulation of sand, including direct dephosphorylation and altered endocytic trafficking. At rest, the mammalian sand homolog TRESK is constitutively phosphorylated. Following generation of a Ca 2+ signal, TRESK is dephosphorylated and activated by CN (Enyedi and Czirjak, 2015). Constitutive dephosphorylation and activation of sand by CN could potentially result in higher K + efflux from cortex glia and neuronal depolarization, leading to higher seizure susceptibility. However, in this scenario, K + buffering upon hyperactivation of the nervous system should be more efficient, and thus less seizures are expected. Cortex glia knockdown of the two kinases that phosphorylate TRESK, PKA and Par-1, did not cause seizures. As such, we found no evidence that sand activity is regulated by CN dephosphorylation of the protein directly in cortex glia.
Beyond dephosphorylation, sand expression on the plasma membrane of specific sleep homeostat neurons in Drosophila is regulated by activity-induced internalization (Pimentel et al., 2016). Given astrocytes can shape neuronal circuit activity by actively altering their K + uptake capabilities (Wang et al., 2012a), we propose that Drosophila cortex glia regulate the expression levels of sand (and potentially other K + buffering proteins) on the cell membrane in a Ca 2+regulated fashion in normal animals. When Ca 2+ is constitutively elevated in zyd mutants, this regulation is thrown out of balance. Indeed, inhibition of endocytosis in cortex glia can rescue zyd seizures, suggesting that a membrane protein responsible for the neuronal hyperexcitability phenotype is being abnormally internalized in zyd cortex glia. Bypassing sand function by improving cortex glial K + buffering/uptake capacity through overexpression of a leak K + channel (EKO (White et al., 2001)) can reverse the zyd phenotype from seizures (caused by neuronal hyperactivity) to paralysis (caused by neuronal hypoactivity). Together, these findings indicate 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 (Fig. 7A).
Accumulating evidence indicate glia play contributive or even causative roles in several neurological disorders, neurodevelopmental syndromes and neurodegenerative diseases including epilepsy, Fragile X syndrome and SMA respectively. Increased glial activity is associated with abnormal neuronal excitability (Wetherington et al., 2008), and pathologic elevation of glial Ca 2+ has been suggested to play an important role in the generation of seizures (Gomez-Gonzalo et al., 2010;Tian et al., 2005). Approximately 50 million people worldwide have epilepsy, making it one of the most common neurological diseases globally (World Health Organization, 2018, http://www.who.int/en/). The traditional view assumes that epileptogenesis occurs exclusively in neurons. However, an astrocytic basis for epilepsy was proposed almost two decades ago (Gomez-Gonzalo et al., 2010;Tashiro et al., 2002;Tian et al., 2005). In a nonpathological state, glia display Ca 2+ oscillations spontaneously (Takata and Hirase, 2008) and in response to physiological neuronal activity (Wang et al., 2006). One widely-studied output of Ca 2+ oscillations is gliotransmission, the glial release of certain transmitters (Agulhon et al., 2008;Lee et al., 2010). Indeed, Ca 2+ -dependent glutamate release from astrocytes causes synchronous currents in neighboring neurons (Angulo et al., 2004;Fellin et al., 2006;Fellin et al., 2004), and is capable of eliciting action potentials (Pirttimaki et al., 2011), suggesting abnormally elevated glial Ca 2+ may produce neuronal hypersynchrony through enhanced gliotransmission. In addition, in vivo work demonstrated that several anti-epileptic drugs reduce glial Ca 2+ oscillations (Tian et al., 2005). Although increased glial activity has been associated with abnormal neuronal excitability, the role of glia in the development and maintenance of seizures, and the exact pathway(s) by which abnormal glial Ca 2+ alter glia-to-neuron communication and neuronal excitability are poorly characterized (Wetherington et al., 2008). A second proposed mechanism by which astrocytes regulate neuronal excitability and seizure susceptibility involves the uptake and redistribution of K + ions (Bellot-Saez et al., 2017;Wang et al., 2012a) and neurotransmitters (Rose et al., 2017) from the extracellular space. In this study, we explored the pathways that are activated within glial cells in response to abnormally elevated glial Ca 2+ that triggers seizures. We found that both mechanisms are at play in Drosophila cortex glial cells, with elevated intracellular Ca 2+ leading to impaired K + buffering. We identified several other suppressors of seizure induction in zyd mutants that are involved in GPCR signaling and vesicle trafficking, suggesting additional glial pathways may impact neuronal activity as well.
Current estimates suggest ~70% of children and adults with epilepsy can be successfully treated with current anti-epileptic drugs (World Health Organization, 2018, http://www.who.int/en/). The observation that several anti-epileptic drugs reduce glial Ca 2+ oscillations in vivo (Tian et al., 2005), together with the fact that ~30% of epilepsy patients are non-responders, suggest that pharmacologically targeting glial pathways might be a promising avenue for future drug development in the field. Several neuronal seizure mutants in Drosophila have already been demonstrated to respond to common human anti-epileptic drugs, indicating key mechanisms that regulate neuronal excitability are conserved from Drosophila to humans. Indeed, zyd-induced seizures can be rescued when animals are fed a CN inhibitor (Fig. 4), indicating pharmacological targeting of the glial CN pathway can improve the outcome of a glialderived seizure mutant. Prior studies have also shown improvement following treatment with the CN inhibitor FK506 in a rodent kindling model (Moia et al., 1994;Moriwaki et al., 1996). These data suggest CN activity regulates epileptogenesis in both Drosophila and mammalian models. Further characterization of how glia detect, respond, and actively shape neuronal excitability is critical to our understanding of neuronal communication and future development of new treatments for neurological diseases like epilepsy.

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. The UAS/Gal4 and LexAop/LexA systems were used to drive transgenes in glia, including repo-gal4, 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 (Table S2). UAS-myrGCaMP6s was constructed by replacing GCaMP5 in the previously described myrGCaMP5 transgenic construct. Transgenic flies were obtained by standard germline injection (BestGene Inc.). For all experiments described, only male larva and adults were used. 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 (Figs. 1A, 2E and 6B), animals of the designated genotype were reared at 22°C to adulthood with gal80 suppressing gal4-driven transgene expression (zyd RNAi , CanB2 RNAi and Rab5 DN , respectively). 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 (Fig. 5A) GMR54H02-lexA was used to express gal80 from LexAop-gal80. For inhibiting transgene expression specifically in neurons (Fig. 2D Transgenic lines generated for this study:

UAS-myrGCaMP6s
LexAop-myrGCaMP6s For a complete list of all RNAi stocks used in this study, see Table S2.

Behavioral analysis:
All experiments were performed using groups of ~10-20 males.
1. 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. Paralysis was defined as the condition in which the animal fell to the bottom of the vial and exhibited no movement. For screening purposes, only flies that showed normal wildtype-like behavior (i.e. walking up and down on vials walls) during heat-shock were counted as successful rescue. To analyze behavior in a more detailed manner we characterized four behavioral phenotypes: wall climbing (flies are climbing on vials 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. Larval seizures were defined as continuous, unpatterned contraction of the body wall muscles that prevented normal crawling behavior. For determining seizure temperature threshold, groups of 10 animal 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 h. Vials were vortexed at maximum speed for 10 seconds, and the number of flies that were upright and mobile was counted at 10 seconds intervals.

Light avoidance:
These assays were performed using protocols described previously following minor modifications. Briefly, pools of ~20 3 rd instar larvae (108-120 hours 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: = 4. Activity monitoring using the MB5 system: Adult flies activity was assayed using the multi-beam system (MB5, TriKinetics) as previously described. Briefly, individual males aged 1-3 days were inserted into 5 mm×80 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 8 control animals and 8 experimental animals, n≥3).
5. Gentle touch assay 3 rd instar male larvae (108-120 hours 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 MgCl2, 0.2 CaCl2, 10 NaHCO3, 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 as previously described.

In vivo Ca 2+ imaging
UAS-myrGCaMP6s was expressed in glia with 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 (LabScientific). 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 40X 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 (Sigma Aldrich), FK506 (InvivoGen) and CN585 (Millipore) were dissolved in DMSO to a final concentration of 20 mM. Fly 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 CN inhibitor, or DMSO as control. Flies were behaviorally tested following 6, 12 or 24 hours of drug feeding.
This disrupts the endo-exocytosis balance of cortex glial membrane proteins such as K2P leak channel sandman that regulate neuronal excitability. Knockdown of sandman mimics the zyd phenotype and the protein acts downstream of CN, indicating impaired extracellular K + homeostasis causes neuronal hyperexcitability and hyperactivity in zyd mutants. B-D.