Glial ER and GAP Junction Mediated Ca²⁺ Waves are Crucial to Maintain Normal Brain Excitability

Abstract

Astrocytes play key roles in regulating multiple aspects of neuronal function from invertebrates to humans and display Ca²⁺ fluctuations that are heterogeneously distributed throughout different cellular microdomains. Changes in Ca²⁺ dynamics represent a key mechanism for how astrocytes modulate neuronal activity. An unresolved issue is the origin and contribution of specific glial Ca²⁺ signaling components at distinct astrocytic domains to neuronal physiology and brain function. The Drosophila model system offers a simple nervous system that is highly amenable to cell-specific genetic manipulations to characterize the role of glial Ca²⁺ signaling. Here we identify a role for ER store-operated Ca²⁺ entry (SOCE) pathway in perineurial glia (PG), a glial population that contributes to the Drosophila blood-brain barrier. We show that PG cells display diverse Ca²⁺ activity that varies based on their locale within the brain. Ca²⁺ signaling in PG cells does not require extracellular Ca²⁺ and is blocked by inhibition of SOCE, Ryanodine receptors, or gap junctions. Disruption of these components triggers stimuli-induced seizure-like episodes. These findings indicate that Ca²⁺ release from internal stores and its propagation between neighboring glial cells via gap junctions are essential for maintaining normal nervous system function.

Graphical Abstract

Introduction

Glial cells regulate multiple aspects of brain function, including synapse formation, neuronal excitability, synaptic transmission and blood flow dynamics (Barres, 2008). Astrocytes, a prominent class of central nervous system (CNS) glia, modulate neuronal properties through the secretion of neuroactive agents (gliotransmission), neurotransmitter buffering and ion homeostasis, in addition to their role in synaptogenesis and blood-brain barrier function (Khakh & Deneen, 2019; Nagai et al., 2020). Astrocytes display Ca²⁺ fluctuations heterogeneously distributed throughout different cellular microdomains including cell bodies, processes and endfeet that directly contact synapses and blood vessels. Although the roles of glial Ca²⁺ dynamics are still being elucidated, they are hypothesized to allow astrocytes to respond to information from neighboring CNS cells and exert local modulatory control over various aspects of brain activity (Bindocci et al., 2017; Haustein et al., 2014; Jiang, Diaz-Castro, Looger, & Khakh, 2016; Khakh & Deneen, 2019; Otsu et al., 2015; Shigetomi et al., 2013; Stobart et al., 2018).

Multiple studies indicate mammalian astrocytes display complex and diverse Ca²⁺ signals. This glial Ca²⁺ activity can be spontaneous (Shigetomi, et al., 2013; Shigetomi, Tong, Kwan, Corey, & Khakh, 2011) or evoked by neuronal activity (Di Castro et al., 2011; Panatier et al., 2011), ranging from small microdomain events (Haustein, et al., 2014; Shigetomi, et al., 2013; Shigetomi, Kracun, Sofroniew, & Khakh, 2010) to global Ca²⁺ waves that encompass entire astrocytic cells (Haustein, et al., 2014). Furthermore, differences in Ca²⁺ activity of astrocytes from different brain regions provide evidence for functional heterogeneity (Clarke, Taha, Tyzack, & Patani, 2021). These diverse macroscopic Ca²⁺ events in different astrocytes or astrocytic processes may arise from diverse microscopic signaling cascades that are functionally segregated and molecularly distinct within the cell (Bindocci, et al., 2017; Shigetomi, Bowser, Sofroniew, & Khakh, 2008). Ca²⁺ sources in astrocytes include Ca²⁺ entry via plasma membrane Ca²⁺ channels (Dunn, Hill-Eubanks, Liedtke, & Nelson, 2013), release from endoplasmic reticulum (ER) intracellular Ca²⁺ stores (Haustein, et al., 2014; Straub, Bonev, Wilkerson, & Nelson, 2006) and release from mitochondria (Agarwal et al., 2017). However, signaling pathways mediating Ca²⁺ fluctuations in different cellular compartments are not well defined. An understanding of the subcellular distribution of signaling mechanisms is critical for dissecting how glial activity modulates brain development and function.

The Drosophila model offers a simple nervous system that is highly amenable to cell-specific genetic manipulations to address the role of different Ca²⁺ signaling pathways in glial function. Several Drosophila glial subtypes influence neuronal function via distinct mechanisms downstream of their Ca²⁺ dynamics. For example, astrocytic Ca²⁺ regulates neurotransmitter uptake (Y. V. Zhang, Ormerod, & Littleton, 2017) and secretion of neuromodulators (Ma, Stork, Bergles, & Freeman, 2016), while disruption of Ca²⁺ signaling in Drosophila cortex glia impairs K⁺ buffering capacity (Weiss, Melom, Ormerod, Zhang, & Littleton, 2019). Synchronized Ca^{2 +} waves in Drosophila subperineurial glia cells control nutrient-dependent reactivation of neural stem cells and subsequent brain growth (Holcroft et al., 2013; Speder & Brand, 2014).

The Drosophila CNS is separated from the surrounding hemolymph, the insect “blood”, by a barrier that is structurally and functionally similar to the mammalian Blood-Brain Barrier (BBB, Figure 1A, (DeSalvo et al., 2014; DeSalvo, Mayer, Mayer, & Bainton, 2011; Stork et al., 2008)). The Drosophila BBB covers the entire CNS with a flattened cell sheet consisting of two classes of glia, an outer layer of perineurial glia (PG) and an inner layer of subperineurial glia (SPG) (Yildirim, Petri, Kottmeier, & Klambt, 2019). Here, using a genetic screen for glial pathways that increase seizure susceptibility in Drosophila, we found that knockdown of dStim (the Drosophila homolog of the mammalian Stromal Interaction Molecule 1, Stim1, acting in the Store-Operated Calcium Entry pathway, SOCE) in PG cells leads to severe heat-shock (HS) induced seizures. By performing detailed Ca²⁺ imaging studies, we show that PG cells exhibit robust, complex and dynamic Ca²⁺ activity that significantly differs between PG cells that occupy different brain regions. In addition, we find these Ca²⁺ dynamics are independent of extracellular Ca²⁺, originate from internal ER stores and spread as waves through gap junctions. Similar to dStim knockdown in PG cells, knockdown of Orai, a plasma membrane Ca²⁺ channel that is gated by dStim leading to Ca²⁺ influx and restoration of the ER Ca²⁺ store, or of Ryanodine receptor (RyR) that releases Ca²⁺ from the ER, also disrupt PG Ca²⁺ dynamics and increase seizure susceptibility. Inhibiting the propagation of Ca²⁺ activity within the PG sheet through gap junctions recapitulates the behavioral phenotype of SOCE knockdown. Together, our data indicate that PG Ca²⁺ release from internal stores is essential to maintain normal nervous system function.

Figure 1: Glial Knockdown of dStim Increases Seizure Susceptibility.

(A) Schematic representation of the larval nervous system, showing five glial subtypes that occupy the CNS: perineurial glia (PG) and subperineurial glia (SPG) that form the blood-brain barrier (BBB), cortex glia (CG) and astrocyte-like glia (ALG) that directly contact neurons, and ensheathing glia (EG) that separate the cortex and neuropile. Neurons (N), septate junctions (SJ) and the neural lamina (NL) are also shown.

(B) Time course of heat-shock induced seizures (38.5°C, HS) for repo>dStim^RNAi is similar to that induced by the zyd mutation and enhanced in repo>dStim^RNAi on the zyd mutation background, suggesting the two manipulations disrupt independent pathways to enhance seizure susceptibility (p<0.05, Two-way ANOVA, N=4 groups of 20 flies/genotype, error bars are mean ± SEM, see Video 1).

(C) Representative voltage traces of spontaneous CPG activity recorded at larval 3^rd instar muscle 6 at 38°C in wildtype, zyd and repo>dStim^RNAi animals (n≥5 preparations/genotype), showing zyd and repo>dStim^RNAi animals lose normal rhythmic muscle activity at 38°C.

(D) qRT-PCR analysis of dStim and Orai expression levels. The housekeeping gene RPS3 (Ribosomal Protein S3) was used for normalization. Pan-glial dStim knockdown flies and pan-neuronal Orai knockdown flies have ^~35% of the control levels of dStim and Orai mRNAs, respectively. For complete data set, see Supplemental Table 1.

(E) Histogram summarizing the percent of flies exhibiting heat-shock induced seizures at 3 minutes (38.5°C) following conditional pan-glial knockdown of dStim. Rearing adult flies at the restrictive temperature (>30°C) with Gal80^ts allows expression of dStim^RNAi only at the adult stage. These manipulations partially reproduce the repo>dStim^RNAi seizure phenotype (p<0.0001, Student’s t-test, N=4 groups of 20 flies/genotype).

*=P<0.05, **=P<0.01, ****=P<0.0001.

Methods

Drosophila Genetics and Molecular Biology

Flies were cultured on a standard medium at 22°C unless otherwise noted. All Drosophila lines used in this study are listed in Table 1. The UAS/gal4 and LexAop/LexA systems were used to drive transgenes in glia using the indicated drivers. 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). UAS-myrGCaMP6s was constructed by replacing GCaMP5 in the previously described myrGCaMP5 transgenic construct (Melom & Littleton, 2013). To generate UAS- and lexAop-ER::GCaMP6f flies, OER:GCaMP6f (gift from Mikoshiba Hiroko, (Niwa et al., 2016)) was subcloned into either pBID-UASc or pBID-LexAop plasmids using standard methods (Epoch Life Science Inc.). Transgenic flies were obtained by standard germline injection (BestGene Inc). For all experiments described, both male and female larvae or adults were used. For survival assays, embryos were collected in groups of ^~50 and transferred to fresh vials (n=3). 3^rd instar larvae, pupae or adult flies were counted. Survival rate (SR) was calculated as:

$$SR=\frac{N^{live\hspace{0.17em}animals}}{N^{embryos}}$$

Table 1:

Key resources

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Genetic reagent (D. melanogaster)	w1118
Genetic reagent (D. melanogaster)	zyd 1	(Melom & Littleton, 2013)		Zyd
Genetic reagent (D. melanogaster)	repo-Gal4	(Lee & Jones, 2005)
Genetic reagent (D. melanogaster)	GMR85G01-Gal4	(Kremer, et al., 2017)	RRID:BDSC_40436	Perineurial glia
Genetic reagent (D. melanogaster)	GMR54C07-Gal4	(Kremer, et al., 2017)	RRID:BDSC_50472	Subperineurial glia
Genetic reagent (D. melanogaster)	R54H02-Gal4	(Kremer, et al., 2017)	RRID:BDSC_45784	Cortex glia
Genetic reagent (D. melanogaster)	R77A03-Gal4	(Kremer, et al., 2017)	RRID:BDSC_39944	Cortex glia
Genetic reagent (D. melanogaster)	R86E01-Gal4	(Kremer, et al., 2017)	RRID:BDSC_45914	Astrocytes
Genetic reagent (D. melanogaster)	R56F03-Gal4	(Kremer, et al., 2017)	RRID:BDSC_39157	Neuropile ensheathing glia
Genetic reagent (D. melanogaster)	R75H03-Gal4	(Kremer, et al., 2017)	RRID:BDSC_39908	Tract ensheathing glia
Genetic reagent (D. melanogaster)	Moody-Gal4		Gift from Dr. Andrea Brand
Genetic reagent (D. melanogaster)	46F-Gal4		A gift from Dr. Vanessa Auld
Genetic reagent (D. melanogaster)	Gli-Gal4		A gift from Dr. Vanessa Auld
Genetic reagent (D. melanogaster)	Nrv2-Gal4		A gift from Dr. Vanessa Auld
Genetic reagent (D. melanogaster)	Alarm-Gal4		RRID:BDSC_67031 RRID:BDSC_67032
Genetic reagent (D. melanogaster)	NP2222-Gal4	(Hayashi, et al., 2002)	RRID:DGGR_112830
Genetic reagent (D. melanogaster)	Elav-Gal80	(Yang, et al., 2009)
Genetic reagent (D. melanogaster)	UAS-dStim-RNAi	Verified in (Petersen, Wolf, & Smyth, 2020)	RRID:BDSC_27263	dStimRNAi#1
Genetic reagent (D. melanogaster)	UAS-dStim-RNAi		RRID:BDSC_41759	dStimRNAi#2
Genetic reagent (D. melanogaster)	UAS-dStim-RNAi		RRID:BDSC_51685	dStimRNAi#3
Genetic reagent (D. melanogaster)	UAS-dStim-RNAi		RRID:BDSC_52911	dStimRNAi#4
Genetic reagent (D. melanogaster)	UAS-dStim-RNAi		RRID:FlyBase_FBst0478081	dStimRNAi#5
Genetic reagent (D. melanogaster)	UAS-Orai-RNAi		RRID:FlyBase_FBst0450445	OraiRNAi#1
Genetic reagent (D. melanogaster)	UAS-Orai-RNAi	Verified in (Petersen, et al., 2020)	RRID:BDSC_53333	OraiRNAi#2
Genetic reagent (D. melanogaster)	UAS-SERCA-RNAi		RRID:FlyBase_FBst0465735	SERCARNAi#1
Genetic reagent (D. melanogaster)	UAS-SERCA-RNAi		RRID:FlyBase_FBst0479267	SERCARNAi#2
Genetic reagent (D. melanogaster)	UAS-IP3R-RNAi	Verified in (Kohn et al., 2015)	RRID:FlyBase_FBst0451656	ItpRRNAi#1
Genetic reagent (D. melanogaster)	UAS-IP3R-RNAi	Verified in (Kohn, et al., 2015)	RRID:FlyBase_FBst0470335	ItpRRNAi#1
Genetic reagent (D. melanogaster)	UAS-IP3R-RNAi		RRID:BDSC_51686	ItpRRNAi#2
Genetic reagent (D. melanogaster)	UAS- IP3R -RNAi		RRID:BDSC_51795	ItpRRNAi#3
Genetic reagent (D. melanogaster)	UAS- RyR -RNAi		RRID:BDSC_65885	RyRRNAi#1
Genetic reagent (D. melanogaster)	UAS- RyR -RNAi		RRID:BDSC_31540	RyRRNAi#2
Genetic reagent (D. melanogaster)	UAS- RyR -RNAi		RRID:BDSC_29445	RyRRNAi#3
Genetic reagent (D. melanogaster)	UAS-TRPA	(Y. V. Zhang, et al., 2017)
Genetic reagent (D. melanogaster)	UAS-ChR2XXL	(Dawydow et al., 2014)
Genetic reagent (D. melanogaster)	UAS-Shits		RRID:BDSC_66600	Shits
Genetic reagent (D. melanogaster)	UAS-inx1-RNAi		RRID:BDSC_55601	Inx1RNAi
Genetic reagent (D. melanogaster)	UAS-inx2-RNAi		RRID:BDSC_42645	Inx2RNAi#1
Genetic reagent (D. melanogaster)	UAS-inx2-RNAi		RRID:BDSC_80409	Inx2RNAi#2
Genetic reagent (D. melanogaster)	UAS-inx1DN		Gift from Dr. Andrea Brand	Inx1DN
Genetic reagent (D. melanogaster)	UAS-inx2DN		Gift from Dr. Andrea Brand	Inx2DN
Genetic reagent (D. melanogaster)	Tub-Gal80ts		RRID:BDSC_7018 RRID:BDSC_7019
Genetic reagent (D. melanogaster)	UAS-IVS-mCD8::GFP		RRID:BDSC_32186	mCD8::GFP
Genetic reagent (D. melanogaster)	UAS-Esyt2::mCherry	(Kikuma, et al., 2017)	RRID:BDSC_77130	Esyt2::mCherry
Genetic reagent (D. melanogaster)	C155-gal4 Elav-gal4		RRID:BDSC_458 RRID:BDSC_8760
Chemical compound	Dextran	Sigma Aldrich	#30024	1 mM
Chemical compound	Thapsigargin	Sigma Aldrich	#T9033	10μM
Antibody	Mouse monoclonal anti-repo	DHSB	RRID:AB_528448	1:50
Antibody	Rat monoclonal antielav	DHSB	RRID:AB_528218	1:100
Antibody	Rabbit polyclonal antiGFP-488	Invitrogen	#A21311 RRID:AB_221477	1:500
Antibody	DyLight 649 conjugated anti-HRP (horseradish peroxidase)	Jackson ImmunoResearch	#123-605-021	1:2000
Antibody	Goat polyclonal antiMouse405	Life technologies	#A31553 RRID:AB_221604	1:3000
Antibody	Goat polyclonal anti-Rat555	Invitrogen	#A21434 RRID:AB_2535855	1:3000
Reagent	EZ-RNA II kit	Biological Industries, Israel	#20-410-100
Reagent	High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor	Thermo Scientific	AB-4374966
Reagent	Fast SYBR® Green Master Mix	Applied Biosystems	AB-4385612
Reagent	Vectashield	Vector Laboratories	RRID:AB_2336789

For conditional expression using Tub-gal80^ts, animals of the designated genotype were reared at 22°C with gal80 suppressing gal4-driven transgene expression (dStim^RNAi, Orai^RNAi and inx2^DN). Adult flies were then transferred to a 31°C incubator to inactivate gal80 and allow gal4 knockdown for the indicated period. For UAS/gal4 inhibition by low temperature, PG> Orai^RNAi animals were reared at 18°C to suppress gal4-driven transgene expression. Adult flies were transferred to a 25°C incubator upon eclosion to allow gal4 knockdown/overexpression for one day before testing for HS-induced seizures. For inhibiting transgene expression specifically in neurons, elav-gal80 (Yang et al., 2009) was used.

Behavioral analysis

For assaying temperature-sensitive seizures, 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 & Littleton, 2013). For screening purposes, only flies that showed normal wild-type-like behavior (i.e. walking up and down on vial walls) after >2min of heat-shock were counted as not seizing. For assaying seizures in larvae, 3^rd instar larvae were gently washed with PBS and transferred to 1% agarose plates or empty fly vials and heated to 38°C. Larval seizures were defined as continuous unpatterned contraction of the body wall muscles that prevented normal crawling behavior (Melom & Littleton, 2013). For determining seizure temperature threshold, groups of 10 animals were heat-shocked to the indicated temperature (ranging 30–39.0°C in 0.5°C increments). The threshold was defined as the temperature in which > 50% of the animals were seizing after 1 minute.

For assaying bang sensitivity, adult male flies in groups of ^~10–20 (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.

For larval activity monitoring, wandering 3^rd instar larval activity was assayed using a multi-beam system (MB5, TriKinetics) as previously described (Green et al., 2015). Briefly, individual animals were inserted into 5 mm × 80 mm glass pyrex tubes. The activity was recorded following a 5 minutes acclimation period. Throughout each experiment, animals 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 with n ≥ 3.

For adult speed assays, 5–10 days post-eclosion flies were used. Experiments were performed in a custom-built, fully automated apparatus (Bielopolski et al., 2019; Claridge-Chang et al., 2009; Parnas, Lin, Huetteroth, & Miesenbock, 2013; Rozenfeld, Lerner, & Parnas, 2019). Single flies were placed in clear polycarbonate chambers with a constant air flow (3 l/min) that was controlled with mass flow controllers (CMOSens PerformanceLine, Sensirion). The air flow was split between 20 chambers resulting in a flow rate of 0.15 l/minute per chamber. The 20 chambers were stacked in two columns each containing 10 chambers and were backlit by 940 nm LEDs (Vishay TSAL6400). Images were obtained by a MAKO CMOS camera (Allied Vision Technologies) equipped with a Computer M0814-MP2 lens. The apparatus was operated in a temperature-controlled incubator (Panasonic MIR 154) at 25°C. A virtual instrument written in LabVIEW 7.1 (National Instruments) extracted fly position data from video images. Data were analyzed in MATLAB 2018a (The MathWorks).

Immunocytochemistry and structural imaging

3^rd instar wandering larvae were reared at 25°C and dissected in hemolymph-like HL3.1 solution (70 mM NaCl, 5 mM KCl, 4 mM MgCl₂, 10 mM NaHCO3, 5 mM trehalose, 115 mM sucrose, and 5 mM HEPES, pH 7.2). Larvae were fixed for 45 min in HL3.1 buffer containing 4% paraformaldehyde and washed three times for 20 min with PBT (PBS containing 0.1% Triton X-100), followed by a 2 hours incubation in block solution (5% NGS in PBT). Samples were incubated overnight at 4°C and washed with two short washes and three extended 20 min washes in PBT, and then incubated with secondary antibodies at room temperature for 2 h or at 4°C overnight. Finally, larvae were rewashed and mounted in Vectashield for imaging. Antibodies used for this study include the following: anti-repo, anti-elav, anti-RFP, DyLight 649 conjugated anti-HRP (horseradish peroxidase), anti-GFP Alexa Fluor 488. Immunoreactive proteins were imaged on either a ZEISS LSM 800 microscope with Airyscan using a 63X oil immersion objective or a Leica TCS SP5 using a 20X oil immersion objective. Images were processed using ImageJ.

RNA purification, cDNA synthesis, and quantitative real-time PCR analysis.

Total RNA from 60 adult heads was extracted using EZ-RNA II kit (Biological Industries, Israel) for each biological replicate. Reverse transcription of total RNA (1000 ng) into complementary DNA (cDNA) was performed using High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (Thermo Scientific, USA). qRT-PCR reactions were performed using Fast SYBR® Green Master Mix (Applied Biosystems, USA) in a StepOnePlus instrument (Applied Biosystems, USA). Primers (Table 2) were calibrated, and negative control was performed for each primer pair. Samples measured in technical triplicates and values normalized according to mRNA levels of an RPS3 or β-Tubulin housekeeping genes. The amplification cycles were 95°C for 30 seconds, 60°C for 15 seconds, and 72°C for 10 seconds. At the end of the assay, a melting curve was constructed to evaluate the specificity of the reaction. The fold change for each target was subsequently calculated by comparing to the normalized value of either Elav-gal4 parent or Repogal4. Quantification was assessed at the logarithmic phase of the PCR reaction using the 2^−ΔΔCT method, as described previously (Livak & Schmittgen, 2001). For the complete dataset, see Table S1.

Table 2:

Primers used for qRT-PCR analysis

Gene Name	Forward Primer	Reverse Primer
RPS3	ATGAATGCGAACCTTCCGATT	TGATCTCAGTGCGAGAGGGG
β-Tubulin	CCAAGGGTCATTACACAGAGG	ATCAGCAGGGTTCCCATACC
dStim	CCAGCTTGCATCGTCAGCTA	TCCTCCCGCAAAAAGTCATCG
Orai	TCTTCTGACCTCATCTGCGTA	GCGTTCGTATAGACACCACATT
RyR	AAGACAGCTCGTGTCATCCG	CTGTTTCTCCTCGTGCTCCAT

Electrophysiology

Intracellular recordings of wandering 3^rd instar male larvae were performed in HL3.1 saline (in mm: 70 NaCl, 5 KCl, 4 MgCl₂, 1.5 CaCl₂, 10 NaHCO₃, 5 Trehalose, 115 sucrose, 5 HEPES-NaOH, pH 7.2) using an Axoclamp 2B amplifier (Molecular Devices) at muscle fiber 6/7 of segments A3-A5. For recording CPG output, the CNS and motor neurons were left intact. The temperature was controlled with a Peltier heating device and continually monitored with a microprobe thermometer.

In vivo Ca²⁺ imaging

Cyto::GCaMP6s, myr::GCaMP6s and ER::GCaMP6f were expressed in PG cells using the drivers described above. PG-gal4 and UAS-myrGCaMP6s was used for most experiments, except for imaging in RNAi knockdowns (PG>dStim^RNAi and PG>Orai^RNAi) where PG-lexA and LexAop-myr::GCaMP6s was used. For live imaging of undissected 2^nd instar larvae, animals 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. Under these experimental conditions, Ca2+ activity was recorded from a ventral view of the VNC, through the larval cuticle, as the dorsal surface of the VNC is not accessible for imaging. For imaging of semi-dissected brains and VNCs, 3^rd instar larvae were dissected 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). A small incision was made above the brain, with the rest of the organs left largely intact. Under these experimental conditions, Ca²⁺ activity was recorded from a dorsal view of the brain and VNC. 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 20× water-immersion objective using Volocity Software. Single optical planes on the surface or a mid-section of the ventral nerve cord (VNC) or brain hemisphere were imaged. Due to frequent movements in RNAi knockdown animals, only brain-PG were used for comparative analysis.

Ca²⁺ imaging analysis

Both Region-Of-Interest (ROI)-dependent and event-dependent signal detection methods were used for analysis. Ca²⁺ oscillations were analyzed within the first 4 minutes of imaging at room temperature. Maximal myrGCaMP6s signals in PG cells were quantified in the central thoracic and abdominal segments of the VNC and the brain. For ROIs-dependent analysis, ROIs were manually assigned to avoid regions containing non-perineurial glial cells (i.e. midline glia and neurons, see Figure S2, S5A) based on morphology. Selected ROIs were circular with a 10 μm diameter, except for ROIs at the VNC where the diameter was 20 μm. For single-event detection imaging, data were processed using Astrocyte Quantitative Analysis (AQuA, run on MATLAB GUI, (Wang et al., 2019)). Default parameters were used with the following modifications: for all data sets the detection threshold (thrARScl), the temporal cut threshold (thrTWScl), and the Rising time uncertainty (cRise) were set to 3. In addition, as Ca²⁺ imaging at the VNC displayed higher noise levels, a stronger smoothing (smoXY) of 0.8 was used. The frequency (temporal density) of events was calculated by dividing the number of detected events that share a spatial footprint and a similar size (Network - Temporal density with similar size only) by the duration of the imaging session.

Blood-Brain-Barrier Permeability Assay

3^rd instar larval brains were dissected in HL3.1 and incubated with Alexa fluor 647-conjugated 10 Kd dextran (Sigma Aldrich #30024) for 5 minutes before image acquisition. Brains were then fixed in 4% PFA in PBS for 5 minutes, washed briefly in PBS, mounted in VectaShield H-1000 (Vector Laboratories) and imaged by confocal microscopy. Subperineurial knockdown of Su(H) was used as a positive control in each batch.

Statistical Analysis

All statistical analyses were performed using GraphPad Prism as described in the figure legends. No statistical methods were used to predetermine sample size. All n numbers represent biological replicates. Data were pooled from 2 to 3 independent experiments. Ca²⁺ imaging experiments were randomized and blinded. Students’ t-test, One-wat ANOVA and Two-way ANOVA were used, and P values are represented as *=P<0.05, **=P<0.01, ***=P<0.001, ****=P<0.0001. p<0.05 was considered significant. Data are expressed as mean ± SEM or the median.

Results

Knockdown of Glial dStim Increases Seizure Susceptibility

We recently observed that chronic Ca²⁺ increase in cortex glia predisposes animals to stimulation-induced seizures (Weiss, et al., 2019), while acute increases in intracellular Ca²⁺ in astrocyte-like glia drives neuronal silencing and paralysis (Y. V. Zhang, et al., 2017), demonstrating key roles for two Drosophila glial cell populations in modulating neuronal excitability. To identify additional glial signaling pathways that modulate neuronal excitability, we performed a genetic screen using the pan-glial driver repo-Gal4 to drive expression of RNA interference (RNAi) constructs targeting ^~850 genes encoding membrane receptors, secreted ligands, ion channels and transporters, vesicular trafficking proteins and known cellular Ca²⁺ homeostasis and Ca²⁺ signaling pathway components (Weiss, et al., 2019). This screen revealed that pan-glial knockdown of dStim led to severe HS-induced seizures-like episodes (Video 1, middle, hereafter referred to as seizures), similar to those we previously identified in NCKX^zyd (zyd) mutants that disrupt a Ca²⁺ ion exchanger in cortex glia (Figure 1B) (Melom & Littleton, 2013; Weiss, et al., 2019). Pan-glial knockdown of dStim (repo>dStim^RNAi) on the zyd mutation background led to ^~30% of flies showing room-temperature seizures and enhanced HS-induced seizures (Figure 1B), indicating the two manipulations disrupt independent pathways that enhance seizure susceptibility. Recordings of the motor central pattern generator (CPG) output at the larval neuromuscular junction (NMJ) demonstrated that 3^rd instar repo>dStim^RNAi larvae lose normal rhythmic firing at 38°C and instead display continuous neuronal firing, as observed in zyd mutants (Figure 1C). The seizure phenotype that resulted from dStim knockdown was similar when dStim was targeted using four additional partially overlapping dStim RNAi constructs (Figure S1A). All five dStim^RNAi lines also showed seizures when exposed to acute mechanical vortex (a phenotype referred to as bang-sensitivity), though to a lesser extent than following a HS (Figure S1B). To test the efficiency of the dStim RNAi construct, we measured the effect of dStim RNAi #1 on expression level using quantitative real-time polymerase chain reaction (qRT-PCR). dStim RNAi strongly reduced dStim levels (65 ± 16.8% decrease relative to repo-GAL4 control, Figure 1D, Table S1). The actual knockdown effect of dStim RNAi might be stronger, as the remaining expression partially originates from neuronal expression of dStim. For the remaining experiments, we used the dStim RNAi #1 (see Methods).

The seizure phenotype observed from pan-glial knockdown of dStim might result from a developmental role for glial dStim in the CNS. To test for a developmental effect of dStim knockdown, we conditionally expressed a single copy of dStim RNAi using Gal4/Gal80^ts (see Methods) only in adult flies. Adult flies reared at the permissive temperature for Gal80^ts for 3 days (>30°C, to allow dStim RNAi expression) exhibited significantly more seizures, with ^~60% of flies displaying seizure phenotypes (Figure 1E). dStim, together with the Ca²⁺ release-activated Ca²⁺ channel protein, Orai, are implicated in the SOCE pathway. Pan-glial knockdown of Orai using two non-overlapping RNAis (repo>Orai^RNAi) was embryonic lethal, while dStim knockdown was largely viable and showed seizure phenotypes as described above (Figure S1C). To test the efficiency of the Orai RNAi construct, we measured the effect of Orai RNAi #1 on expression levels using qRT-PCR. As pan-glial knockdown of Orai was embryonic lethal, elav-GAL4 was used to knockdown Orai in neurons. Elav>Orai RNAi animals were viable and displayed no seizures upon exposure to either HS or vortex stimuli. Orai RNAi strongly reduced Orai expression levels (65 ± 13.1% decrease relative to elav-GAL4 control, Figure 1D, Table S1). The actual knockdown effect of Orai RNAi might be stronger, as the remaining expression partially originates from glial cells. For the remaining experiments, we used the Orai RNAi #1 (see Methods). Collectively, these results suggest that the SOCE pathway is likely to be essential in glia, regulating neuronal excitability and susceptibility to seizures.

Knockdown of dStim in Perineurial Glial Cells Increases Seizure Susceptibility

To further characterize how dStim knockdown contributes to seizure susceptibility, we performed a secondary screen in which dStim was knocked down specifically in different glial subpopulations. For this screen, we used a series of previously described glial drivers (Hayashi et al., 2002; Kremer, Jung, Batelli, Rubin, & Gaul, 2017; Lee & Jones, 2005; Xie & Auld, 2011). Surprisingly, knockdown of dStim in the two glial subpopulations that are best positioned to influence neuronal activity, cortex glia (CG) and astrocyte-like glia (ALG), failed to recapitulate the phenotype of the pan-glial knockdown (Figure 2Aa’). We found that only knockdown of dStim using two different perineurial glia (PG) drivers (46F-Gal4 (Xie & Auld, 2011) and GMR85G01-Gal4 (Kremer, et al., 2017)), could recapitulate the HS induced seizure phenotype (Figure 2A, B, Video 1), although only ^~85% of PG>dStim^RNAi flies displayed seizures. The weaker phenotype caused with the PG drivers suggests these drivers may result in a less efficient knockdown of the transcript compared to the pan-glial driver. Co-expressing Dicer-2 with dStim^RNAi did not enhance the phenotype (Figure 2Aa’’), suggesting involvement of dStim function in other glial sub-types as well.

Figure 2: Perineurial Knockdown of Store-Operated Ca²⁺ Entry Pathway (SOCE) Components Impairs Locomotor Activity and Increases Seizure Susceptibility.

(A) Histograms summarizing the percent of flies exhibiting HS-induced seizures at 3 minutes (38.5°C). (a’) An array of glial specific Gal4 drivers were used to knock down dStim (see methods). Only knockdown of dStim using perineurial glia (PG) drivers (46F-Gal4 and GMR85G01-Gal4) recapitulated the pan-glial HS-induced seizure phenotype (p<0.0001, Two-way ANOVA), while knockdown with a third PG driver (NP6293) failed to recapitulate the phenotype. (a’’) Inhibiting Gal4 expression of the RNAi in neurons with Gal80 (C155-Gal80) does not alter the seizures observed with GMR85G01 knockdown, indicating the seizure phenotype does not arise from neuronal knockdown of dStim. Co-expressing Dicer2 with dStim RNAi (to enhance the knockdown) using the GMR85G01 driver does not enhance the seizure phenotype (p>0.05, Two-way ANOVA, N=4 groups of >10 flies/genotype, error bars are mean ± SEM).

(B) Time course of heat-shock induced seizures (38.5°C, HS) for repo>dStim^RNAi and PG>dStim^RNAi are shown (p>0.05, Two-way ANOVA, N=4 groups of 20 flies/genotype, error bars are mean ± SEM, Video 1).

(C) Expression pattern of the perineurial glial driver GMR85G01. GMR85G01 expression of a membrane tethered GFP (mCD8::GFP) reveals high expression in PG cells that enclose the entire larval CNS (brain and VNC); 98 μm projection, scale bar= 100 μm (green: mCD8::GFP, cell membranes; red: anti-elav, neuronal nuclei). For complete expression analysis see Figure S2. GMR85G01 hereafter referred to as PG driver.

(D-E) CPG activity in PG>dStim^RNAi showing PG>dStim^RNAi lose normal rhythmic muscle activity under heat-shock (HS) conditions (38°C). (D) Representative voltage traces of spontaneous CPG activity recorded at larval 3^rd instar muscle 6 at 38°C in control and PG>dStim^RNAi animals (n≥5 preparations/genotype). (E) Quantification of precent muscle potential bursting for CPG recordings of PG>dStim^RNAi animals at room temperature and 38°C HS (marked with pink shadings) (p<0.01, Student’s t-test, n ≥ 5 preparations/genotype).

(F, G) Activity level of adults (F) and 3^rd instar larvae (G) expressing dStim RNAi using the PG driver show significant reduction in total locomotor activity (p<0.0001 for adult flies, p<0.05, p<0.01 for larvae, Student’s t-test, N=8 larvae/genotype, N>100 adult flies/genotype, median is presented).

(H, I) Perineurial conditional knockdown of Orai using Gal4/Gal80^ts. Rearing adult flies at the restrictive temperature for Gal80^ts (>30°C, marked with pink shadings) allows expression of Orai^RNAi only in adults (N=4 groups of 20 flies/condition, error bars are mean ± SEM). PG> Orai^RNAi animals that were reared at 18°C to suppress Gal4-driven transgene expression (marked with blue shading), displayed higher survival rate.

(G) Over the course of several days at 31°C, the majority of PG>Orai^RNAi/Gal80^ts flies died (^~80% mortality after 7 days, p<0.0001, Two-way ANOVA). (H) A significant increase in seizures (p<0.0001, Two-way ANOVA) was seen after seven days of rearing flies at the restrictive temperature for Gal80^ts (31°C). Approximately 30% of surviving adults showed seizures and the rest displayed severe locomotor defects. PG> Orai^RNAi animals that were reared at 18°C to suppress gal4-driven transgene expression (marked with blue shading), displayed increased seizure susceptibility.

*=P<0.05, **=P<0.01, ****=P<0.0001.

The GMR85G01-Gal4 driver line was previously shown to drive uniform expression in all PG cells of the adult CNS covering the whole brain and ventral nerve chord (VNC) (Kremer, et al., 2017). Similarly, we found the GMR85G01-Gal4 driver line drives uniform expression in all PG cells of the larval CNS (Figure 2C, S2A), covering the whole brain and VNC (Figure S2B). GMR85G01-Gal4 also drives expression in PG cells of the larval peripheral nervous system (PNS), partially enwrapping peripheral nerves (Figure S2C), as previously described for PG cells (Stork, et al., 2008). While the expression of the GMR85G01 driver in the larval brain is restricted to the PG sheet (Figure S2A), similar to the observations in the adult CNS (brain and VNC), we found that at the larval VNC it also drives weak expression in glial cells other than PG cells (Figure S2D) and a small subset of neurons (Figure S2E). Nevertheless, our driver screen suggests the seizure phenotype caused by dStim knockdown does not arise from dStim suppression in cell types other than PG, as knock down with other cell-type-specific drivers did not lead to seizures (Figure 2Aa’), and neuronal suppression of dStim knockdown (using elav-gal80, Figure 2Aa’’) failed to recapitulate the pan-glial knockdown phenotype. For the remaining experiments, we used the GMR85G01-Gal4 driver, hereafter referred to as PG-Gal4.

Recordings of CPG output at the larval NMJ demonstrated that 3^rd instar PG>dStim^RNAi larvae lose normal rhythmic firing at 38°C and instead display continuous neuronal firing (Figure 2D, E). Adult PG>dStim^RNAi exhibited a significant reduction in locomotor activity levels at room temperature (Figure 2F). Thus, basal locomotor activity is also impaired, suggesting a homeostatic effect rather than an acute effect that only occurs during HS. Taken together, these results indicate that the role of glial SOCE in neuronal excitability is primarily required in PG cells.

To further characterize the effects of disrupting SOCE, Orai was specifically knocked down in PG cells. Orai knockdown with the PG driver (PG>Orai^RNAi) was adult lethal with most animals surviving until late pupal stages, thus preventing the characterization of adult animals. 3^rd instar PG>Orai^RNAi larvae showed a significant defect in locomotor activity (Figure 2G) and HS-induced seizure-like activity when placed at 38°C (Video 2, left). PG>dStim^RNAi larvae showed similar activity impairment as PG>Orai^RNAi larvae (Figure 2G). To exclude the possibility that a developmental effect of Orai knockdown in PG cells leads to lethality, Orai^RNAi was conditionally expressed with Gal4/Gal80^ts (see Methods) only in adult flies. PG>Orai^RNAi/Gal80^ts animals reared at 25°C survived to adulthood and showed no HS-induced seizures 1-day post eclosion (Figures 2H, 2I). However, over the course of several days at 31°C, the majority (^~80%) of PG>Orai^RNAi/Gal80^ts flies showed progressive loss of motor control and death (Figure 2H), with ^~35% of the surviving flies displaying seizures after 7 days (Figure 2I). These results indicate Orai function in PG cells is crucial for normal brain function. Consistent with the more severe phenotype of Orai knockdown compared to dStim knockdown with repo-Gal4 or PG-Gal4, the temperature threshold for seizures in PG>Orai^RNAi larvae was significantly lower compared to PG>dStim^RNAi (Figure S3A). The stronger effect of Orai knockdown might be due to stronger suppression of the SOCE pathway, or due to Orai functioning in a dStim-independent manner (Deb, Pathak, & Hasan, 2016). To reduce the expression of Orai^RNAi we reared PG>Orai^RNAi animals at 18°C, which is below the optimal activation temperature of the UAS/Gal4 system (Duffy, 2002). Under these conditions, ^~50% of PG>Orai^RNAi animals survived to adulthood and showed no HS-induced seizures following eclosion. However, flies that were moved to 25°C after eclosion showed HS-induced seizures one day later (^~80%, Figures 2H, 2I, Video 2). Over the course of several days at 25°C, the majority (^~80%) of PG>Orai^RNAi/Gal80^ts flies rapidly deteriorated and died. To further support the role of ER-originated Ca²⁺ signaling in PG cells, we knocked down other central components of ER-related Ca²⁺ signaling. While knockdown of the inositol 1,4,5-trisphosphate receptor (IP₃R) was viable (Figure S3B) and showed no apparent behavioral phenotype (Figure S3C), knockdown of SERCA (Sarco/Endoplasmic reticulum Ca²⁺-ATPase) was adult lethal (Figure S3B). Collectively, these results suggest a critical requirement for ER-related Ca²⁺ signaling in PG cells, independent of IP₃R, that is necessary to maintain normal brain function.

Knockdown of dStim in Perineurial Glial Cells Does not Affect Blood-Brain-Barrier Integrity

PG cells are thought to influence the development, integrity and function of the blood-brain-barrier (BBB) formed by the SPG layer, as suggested for astrocytes in the mammalian CNS (Abbott, Ronnback, & Hansson, 2006). Perineurial glia also contribute to the deposition of the neural lamella, thus participating in regulating brain shape and stiffness. Alterations in the neural lamella can disrupt brain shape and migration of PG cells (Yildirim, et al., 2019). Hence, alteration of PG Ca²⁺ signaling in dStim and Orai knockdowns could lead to seizures secondary to a role for PG ER-related Ca²⁺ signaling in controlling brain development. However, this seems unlikely given conditional knockdown of dStim or Orai in adult flies recapitulates the seizure phenotype (Figures 1E and 2I). Nevertheless, to test for an effect of glial SOCE on brain development and PG migration, we co-expressed dStim^RNAi or Orai^RNAi together with mCD8::GFP specifically in PG cells, as done previously (Speder & Brand, 2014). We found no apparent changes between control, PG>dStim^RNAi and PG>Orai^RNAi animals in brain wrapping by PG or in brain size of 3^rd instar larvae (Figure S3D). To test whether PG SOCE knockdown compromises the function of the BBB, we incubated PG>dStim^RNAi brains with Alexa647-conjugated 10kD dextran and monitored brain penetration of the dye. In both parental control and PG>dStim^RNAi brains, fluorescent dextran remained at the periphery of the brain (Figure S3E, arrowheads), indicating dStim knockdown does not grossly alter the permeability of the BBB, while in SPG>Su(H)^RNAi (positive control), significant uptake of the dye was observed (Figure S3E, right, arrowheads), indicating dysfunction of the BBB in these animals. Taken together, these results suggest that PG knockdown of SOCE does not affect the gross integrity of the Drosophila blood-brain barrier.

Dynamic Ca²⁺ Transients Occur in Perineurial Glia in vivo

Multiple components of the vertebrate BBB, including astrocytes, show fluctuations in intracellular Ca²⁺. Therefore, we explored whether PG cells that contribute to the Drosophila BBB also exhibit fluctuations in Ca²⁺. To examine in vivo Ca²⁺ dynamics in PG cells, we expressed a myristoylated variant of GCaMP6s (myr::GCaMP6s) that tethers it to the plasma membrane to monitor Ca²⁺ dynamics in fine processes (Melom & Littleton, 2013; Weiss, et al., 2019) specifically in PG cells. We first performed imaging experiments in live, undissected 2^nd instar larvae as previously described (Melom & Littleton, 2013). PG expression of myr::GCaMP6s revealed Ca²⁺ transients in PG cells at the ventral surface of the ventral nerve cord (VNC, Figure 3A–E and Video 3A) and in peripheral nerves (Figure 3F–H and Video 3B).

Figure 3: Dynamic Ca²⁺ Transients Occur in Perineurial Glia In Vivo.

Ca²⁺ imaging of PG>myr::GCaMP6s in live, non-dissected 2^nd instar wildtype Drosophila larvae.

(A) Top, schematic representation of the Drosophila larval brain shows the relative field of view at the ventral surface of the VNC (light blue). The dorsal-ventral (D-V) axis is shown. Bottom, time-lapse image series of perineurial glial Ca²⁺ at the ventral surface of the VNC (CNS-PG). Arrowheads mark peaks of Ca²⁺ transients. Scale bar, 20 μm.

(B) 10 second activity projection of PG cells at the ventral side of the VNC, showing Ca²⁺ elevations are diverse in size and reoccur in the same regions. Events were detected using single-event detection (AQuA, see methods). The square marks the field of view in panel A. Arrowhead marks a Ca²⁺ wave that is presented in panels D, E. Scale bar, 20 μm.

(C) Representative traces of mean fluorescence (% ΔF/F) in active regions. Ca²⁺ elevation in the upper trace is shown in panels D, E.

(D-E) Slow Ca²⁺ waves occur in VNC-PG cells in vivo. (D) Time-lapse image series of a single slow Ca²⁺ elevation (^~15 sec duration) that spreads as a wave across large distances. (E) Heatmap summarizing the rise time and spread of the wave shown in panel D. Scale bars, 20 μm.

(F-G) Fast Ca²⁺ elevations occur in PNS-PG in vivo. (F) Top, schematic representation of the Drosophila larval brain shows the relative field of view at a peripheral nerve (light blue). Bottom, time-lapse image series of perineurial glial Ca²⁺ signals in an abdominal segment peripheral nerve (PNS-PG). Event-based detection revealed that Ca²⁺ elevation in PNS-PG is synchronized across wide areas. Scale bar, 20 μm.

(G) Representative trace of the mean fluorescence (% ΔF/F) of an active region of PNS-PG.

(H) Heatmap summarizing the rise time of a single fast event in PNS-PG. Scale bars, 20 μm.

(I-K) Histograms comparing Ca²⁺ transient characteristics in PG cells of the VNC (CNS-PG) or enwrapping peripheral nerves (PNS-PG). (I) Transient amplitudes are significantly larger in PNS-PG (% ΔF/F, p<0.0001, Student’s t-test, n=99 transients/23 ROIs/4 VNCs, n=65 transients/5 nerves/5 animals). (J) Transient durations are significantly shorter for PNS-PG (seconds, p<0.0001, Student’s t-test, n=61 transients/23 ROIs/4 VNCs, n=61 transients/5 nerves/5 animals). (K) Transient frequencies are significantly lower in PNS-PG (transients/minute, p<0.01, Student’s t-test, n=23 ROIs/4 VNCs, n=5 nerves/5 animals).

**=P < 0.01, ****=P < 0.0001.

PG cells that enwrap the VNC (CNS-PG) show Ca²⁺ signals that range from small, localized elevations to events that cover large areas (Video 3A). These events recur frequently in the same regions (Figure 3B, C and K), occasionally spreading as waves across large distances (Figure 3D, E), between neighboring segments, and even across the midline (Figure S4A), suggesting adjacent PG cells can laterally transfer information through Ca²⁺ waves. The duration of CNS-PG Ca²⁺ transients was 12.96 ± 0.59 seconds and exhibited a mean ΔF/F of 21.08 ± 0.9% (Figures 3I, J). In contrast, PG cells that enwrap peripheral nerves (PNS-PG) show fast elevations in Ca²⁺ along the whole imaging field (Figure 3F–H and Video 3B). These transients recur frequently (^~4 events/min, Figure 3K), with a duration of 2.32 ± 0.142 seconds and a mean ΔF/F of 49.11 ± 2.61% (Figure 3I, J). Astrocyte Quantitative Analysis (AQuA, (Wang, et al., 2019)) software for single-event detection recapitulated these observations (for example see Figure S4B, for complete analysis see Figure S4C–E) and revealed that PNS-PG also display small, localized sporadic events dispersed between the large Ca²⁺ events (Figure S4F–H). Together, these data indicate PG cells display dynamic Ca²⁺ signaling in vivo, with PG cells in the CNS and PNS displaying distinct patterns of Ca²⁺ activity.

Characterization of PG Ca²⁺ Activity Reveals Unique Signatures in Cells that Occupy Different CNS Territories

To examine whether influx of external Ca²⁺ underlies the observed Ca²⁺ transients, and to minimize muscle contractions, we performed imaging in an external solution containing no added Ca²⁺ (i.e. nominal [Ca²⁺]_out). Under these conditions, analysis of assigned ROIs (see methods and Figure S5A) revealed that the dorsal perineurium enwrapping the VNC (VNC-PG) exhibited robust population-wide Ca²⁺ elevations (Video 4, Figures 4A), suggesting PG Ca²⁺ signaling relies on intracellular Ca²⁺ stores rather than extracellular influx. Ca²⁺ elevations at the VNC appear to be highly correlated between neighboring ROIs (with an average Pearson correlation of r=0.67 ± 0.01, Figure S5B–D), with only a small dependency on the distance between ROIs (Figure S5E). These larger Ca²⁺ elevations spread as waves across the entire VNC (Figure 4B) and displayed a mean ΔF/F of 53.63 ± 1.48% (Figure 4O) and duration of 20.20 ± 0.95 seconds (Figure 4P). Single event detection (AQuA) recapitulated these population-wide Ca²⁺ elevations (Figure S5E) and revealed that these events spread as waves across the entire VNC (Figure 4B, Video 4). Furthermore, this analysis revealed that PG cells at the VNC also display more localized Ca²⁺ elevations that represent the majority of VNC-PG activity (over 90% of detected events, Figure 4D and Video 4). Analysis of these two event-populations revealed the duration of wide-spread waves was significantly longer (3.44 ± 0.14 seconds longer, Figure 4E, Figure S5G), while the amplitude was not significantly different (Figure 4F). Analysis of the temporal density of Ca²⁺ elevations (see Methods) revealed no significant difference between small events and larger, wide-spread waves (^~7 events per minute, Figure 4G, S5H).

Figure 4: Characterization of PG Ca²⁺ Reveals Distinct Signatures in Cells that Occupy Different CNS Territories.

Imaging of PG>myr::GCaMP6s in dissected 3^rd instar wildtype Drosophila larvae.

(A-J) Ca²⁺ activity of PG cells at the dorsal surface of the VNC.

(A) Top, schematic representation of the Drosophila larval brain shows the relative field of view at the dorsal surface of the VNC (light blue). Bottom, representative traces of the mean fluorescence (% ΔF/F) of manually assigned ROIs (see methods and Figure S5A).

(B) Top, heatmaps summarizing the rise time and spread of three representative slow waves that occurred in the same region of the VNC. This analysis revealed that waves are highly variable, and do not have a preferable direction of spread. Scale bar, 100 μm. Bottom, a whole trace of mean fluorescence (% ΔF/F) summarizing the detected activity. The waves shown in the upper panel are marked with pink shadings.

(C) 10 second activity projection showing small Ca²⁺ elevations that occur between slow waves. These events reoccur in the same regions. Events were detected using single-event detection (AQuA). Scale bar, 100 μm.

(D) Histogram summarizing the frequency (%) of event size (μm²). Small events (<500 μm²) represent the majority of Ca²⁺ activity at the VNC (>90%). The frequency distribution of event size follows a Gaussian distribution, with an R²=0.87.

(E-G) Histograms comparing different characteristics of small area events (<500 μm²) versus wide-spread Ca²⁺ waves (>500 μm²). Data was derived from event-based detection (AQuA, see methods). N=5 animals/ 4 minutes imaging session each. Event durations (J) are significantly shorter for small area events (3.44 ± 0.14 seconds longer, p<0.0001, Student’s t-test), while transient amplitudes (E, % ΔF/F) and temporal density of events (G, see methods) are not significantly changed between the two event-populations (Student’s t-test).

(H-J) Ca²⁺ activity of single PG cells co-expressing myr::GCaMP6s and nuclear-localized mCherry (magenta: mCherry.nls, PG nuclei; green: myr:GCAMP6s, PG membrane).

(H) Time-lapse image series of PG Ca²⁺ showing a Ca²⁺ wave spreading through multiple adjacent PG cells (also see Video 5). Scale bar, 20 μm.

(I-J) Analysis of single cell Ca²⁺ activity. ROIs were assigned to single cells by mCherry.nls as shown in Figure S5I.

(I) Representative traces of mean myr::GCaMP6s fluorescence (% ΔF/F) of single cells reveal dynamic Ca²⁺ activity.

(J) Ca²⁺ activity of single VNC-PG is mostly asynchronized between single cells across the VNC (also see Figure S5J, K) independently of distances between cells. An exponential fit (R²=0.19) shows a decay of tau=50.8. (n=228 cells/ N=3 animals).

(K-M) Ca²⁺ activity of PG cells at the brain (brain-PG).

(K) Top, schematic representation of the Drosophila larval brain shows the relative field of view at a mid-section through a brain hemisphere (light blue). Bottom, time-lapse image series of Ca²⁺ imaging in PG that enwrap a brain hemisphere (brain-PG). Mid-section is shown. Ca²⁺ activity at brain-PG can be localized to small areas or spread as waves across large distances (also see Figure S6F, G and Video 6).

(L) Representative traces of mean fluorescence (% ΔF/F) show highly variable dynamic Ca²⁺ activity (Video 6).

(M) Heatmap summarizing the rise time and spread of a single wide-spread Ca²⁺ wave at the brain. This analysis revealed that waves are highly variable, and do not have a preferable direction of spread (see also Figure S6F). Scale bar, 20 μm.

(N) Wide-spread Ca²⁺ waves (>500 μm²) represent only a small fraction of brain-PG Ca²⁺ activity (6.8% of the total detected activity). Overall, the frequency distribution of event size follows a Gaussian distribution, with an R²=0.83.

(O-P) Comparisons of Ca²⁺ transient characteristics in VNC-PG and brain-PG. Data was derived from event-based detection (AQuA, see methods). N=5 animals/ 4 minutes imaging session each/ CNS region. Brain-PG transients show significantly higher amplitudes (p<0.0001, Student’s t-test) (O) and shorter event durations (p<0.0001, Student’s t-test) (P). ****=P<0.0001.

The widespread Ca²⁺ activity likely represents the activity of multiple PG cells. To characterize the activity of single PG cells, we co-expressed myr::GCaMP6s and nuclear mCherry (mCherry.nls) in PG cells and assigned ROIs to single cells based on nuclear labeling (Video 5, Figures 4H and S5I). Single cells at the VNC showed Ca²⁺ activity patterns in which cells alternate between active and silent periods (Figure 4I). Single-cell activities were mostly asynchronized (with an average Pearson correlation of r=0.1926 ± 0.003, Figures 4J and S5J, K). The activity of neighboring cells can be more synchronized (Figure S5L), consistent with waves that travel through adjacent cells. The mean duration of single-cell VNC-PG transients was 20.42 ± 0.56 seconds, similar to what was measured with hemi-segment ROIs (Figure 4O).

In contrast to the slow wide-spread waves observed in VNC-PG, the PG sheet on the surface of the brain (Brain-PG) showed fast, asynchronous activity (Video 6A, Figure S6A–C). These transients recurred frequently in the same regions (^~6 events/min, Figure S6B). Single event detection revealed that this localized activity occasionally spreads between neighboring cells (Figure S6D, E), however wide-spread waves as observed at the VNC were not detected. Imaging Ca²⁺ from a mid-section through a brain hemisphere revealed that brain-PG exhibit Ca²⁺ waves that spread through large areas of the PG sheet (Figures 4K–M, Figure S6F and Video 6B, note that sporadic, small events are also visible). These waves recurred frequently in the same regions (Figure S6G) and represent a small fraction of the total brain-PG activity (^~7% of detected events, Figure 4N). Interestingly, the maximum amplitude of these wave events does not necessarily overlap with the initiation site of the event (Figure S6H, I), suggesting these events spread through a propagation mechanism rather than through passive diffusion. Relative to VNC-PG, Ca²⁺ waves observed in brain-PG exhibited larger amplitudes, with a mean ΔF/F of 66.38 ± 2.91% (Figure 4O) and shorter duration with a mean duration of 9.71 ± 0.26 seconds (Figures 4P). PNS-PG cells did not show Ca²⁺ oscillations under these experimental conditions, suggesting that PNS-PG signaling relies more on extracellular Ca²⁺. Together, these data indicate that PG cells show complex and diverse Ca²⁺ activity patterns based on their location, indicating functional diversity within the PG cell population.

Ca²⁺ Transients in Perineurial Glia Originate from Internal ER Ca²⁺ Stores

The occurrence of PG Ca²⁺ transients in a low-Ca²⁺ external solution suggests that PG Ca²⁺ signaling is likely to rely on internal ER Ca²⁺ stores. To test this hypothesis, we repeated the Ca²⁺ imaging experiments in an external solution containing 10 μM Thapsigargin (Tg, see Methods) to pharmacologically inhibit the restoration of ER Ca²⁺ stores. Under these conditions, PG Ca²⁺ signaling in both the VNC and the brain is almost completely abolished (Figures 5A and Video 6C). Genetic knockdown of the SOCE pathway is predicted to disrupt restoration of ER Ca²⁺ stores and reduce amplitudes of ER-originated Ca²⁺ transients. To test this hypothesis, PG Ca²⁺ transients were imaged in PG knockdowns of dStim and Orai. Indeed, Ca²⁺ activity of both VNC-PG (Figure 5B) and brain-PG (Figures 5C–H) were significantly reduced (Video 7). Single-event detection revealed that the frequency of Ca²⁺ events (Figure 5C), including large-area events (>500 μm², Figure 5D), was significantly lower in SOCE knockdown animals. Furthermore, event area, amplitude, duration and temporal density of the remaining brain-PG activity were significantly reduced in SOCE knockdown animals relative to control (Figure 5E–H). These data indicate that both pharmacological and genetic depletion of ER Ca²⁺ stores significantly reduce PG Ca²⁺ signaling, supporting the model that PG Ca²⁺ activity relies solely on Ca²⁺ signals that originate from internal ER Ca²⁺ stores.

Figure 5: Pharmacologically and Genetically Inhibiting ER Ca²⁺ Signaling Abolish Perineurial Ca²⁺ Activity.

(A) Pharmacologically inhibiting ER Ca²⁺ signaling in dissected 3^rd instar wildtype Drosophila larvae. Top, schematic representation of the Drosophila larval brain shows the relative field of view at a section through a brain hemisphere (light blue). Bottom, representative traces of mean myr::GCaMP6s fluorescence (% ΔF/F) of randomly assigned ROIs in brain-PG show dynamic Ca²⁺ activity under control conditions (1% DMSO) that is completely abolished when samples are incubated in 10 μM Thapsigargin (Tg) for 2 minutes prior to imaging. Rapid decrease in basal myr:GCaMP6s fluorescence prevented automated detection of small events.

(B- H) Imaging of PG-lexA>LexApo-myr::GCaMP6s following genetic inhibition of SOCE.

(B) Top, schematic representation of the Drosophila larval brain shows the relative field of view at the dorsal surface of the VNC (light blue). Bottom, representative traces of mean fluorescence (% ΔF/F) in ROIs assigned to VNC hemi-segments show significant reduction in VNC-PG Ca²⁺ activity in SOCE knockdowns (PG>dStim^RNAi and PG>Orai^RNAi, also see Video 7).

(C) Top, schematic representation of the Drosophila larval brain shows the relative field of view at a mid-section through a brain hemisphere (light blue). Bottom, 10 seconds activity projection showing the total temporal density of Ca²⁺ elevations in control, PG>dStim^RNAi and PG>Orai^RNAi animals. Events were detected using single-event detection (AQuA, see methods). Scale bar, 100 μm.

(D-H) Comparisons of Ca²⁺ transient characteristics in brain-PG of control, PG>dStim^RNAi and PG>Orai^RNAi animals. Data was derived from event-based detection (AQuA, see methods). N=5 animals/ 4 minutes imaging session each/genotype.

(D) Brain-PG activity in SOCE knockdown (dStim and Orai RNAis) show significantly less large area events (>400 μm², p<0.01, One-way ANOVA). Control Brain-PG transients show significantly higher temporal density (E, p<0.0001, One-way ANOVA), larger areas and amplitudes (F, G, p<0.0001,One-way ANOVA) and longer event durations (H, p<0.0001, One-way ANOVA).

** P<0.01, ****=P<0.0001.

Propagation and Spread of PG Ca²⁺ Waves Through Gap Junctions are Crucial for the Prevention of Seizures

To examine the role of PG Ca²⁺ elevations and PG Ca²⁺ waves in the generation of seizures, we artificially elevated Ca²⁺ levels in PG cells by over expressing the heat-sensitive Ca²⁺ channel TRPA and the light-activated Channel Rhodopsin (ChR^XXL). Activation of these channels in PG cells with temperature shifts or light did not alter the PG>dStim^RNAi seizure phenotype or cause a behavioral phenotype by themselves (Figure S7A), suggesting the phenotype arising from SOCE knockdown is due to an impairment in an ER related signaling pathway rather than from an alteration in basal [Ca²⁺]_i.

Ca²⁺ waves in mammalian astrocytes spread via direct communication between adjoining cells through gap junction channels or by release of gliotransmitters that activate neighboring cells via membrane receptors. These two mechanisms are thought to work in parallel to coordinate Ca²⁺ activity between neighboring cells (Scemes & Giaume, 2006). To examine the mechanism that mediates Ca²⁺ wave spread within the PG cellular sheet, we first manipulated PG secretion by overexpressing the temperature-sensitive (ts) allele of the Drosophila Dynamin homolog, Shibire (Shi^ts). Conditionally inhibiting endocytosis (and subsequent exocytosis) in PG cells, either acutely (by subjecting PG>Shi^ts flies directly to 38°C) or constitutively (by pre-incubating PG>Shi^ts flies at the restrictive temperature, 30°C, before a 38°C HS) had no effect on seizure susceptibility of either wildtype or PG>dStim^RNAi animals (Figure S7B). These results suggest gap junction communication may be the dominant mode of Ca²⁺ wave spread in Drosophila PG cells rather than through the secretion of exogenous factors.

Within the Drosophila BBB, SPG cells exhibit Ca²⁺ waves that spread through neighboring SPG cells via gap junctions (Holcroft, et al., 2013; Speder & Brand, 2014). Astrocytes and pericytes at the vertebrate BBB also exhibit Ca²⁺ waves that spread through gap junctions (Burdyga & Borysova, 2018). To test whether the spread of Ca²⁺ waves between neighboring PG cells is mediated by gap junctions, we over-expressed a dominant-negative (DN) form of one of the Drosophila gap junction homologs, Inx2 (Inx2^DN), in PG cells. PG overexpression of Inx2^DN significantly inhibited Ca²⁺ activity in PG cells in the VNC and the brain (Figure 6A and Video 8). Single-event detection revealed that the total Ca²⁺ activity is significantly reduced (Figure 6B, C), although the reduction in the occurrence of large-area events (>500 μm²) in Inx2^DN animals was variable and not significant (Figure 6D). Furthermore, event amplitude and duration of the remaining brain-PG activity were significantly reduced in Inx2^DN animals relative to control (Figure 6E–F). Similar to other genetic manipulations described above, RNAi-mediated knockdown of either Inx1 or Inx2, or overexpression of Inx2^DN, was adult lethal. Conditionally expressing Inx2^DN with Gal4/Gal80^ts (see Methods) only in adult flies significantly increased their seizure susceptibility, with ^~60% of the flies showing seizures after 15 hours at the restrictive temperature for Gal80^ts (>30°C, Figure 6G, Video 8). Together, these results suggest a key role for gap junctions in the propagation of Ca²⁺ waves through adjacent PG cells.

Figure 6: Ca²⁺ Waves Spread is Necessary for Preventing Neuronal Hyperexcitability.

Imaging of PG>myr::GCaMP6s in dissected 3^rd instar Drosophila larvae expressing Inx2^DN.

(A) Top, schematic representation of the Drosophila larval brain shows the relative field of view at a mid-section through a brain hemisphere (light blue). Bottom, representative traces of mean myr::GCaMP6s fluorescence (% ΔF/F) show that PG Ca²⁺ activity is significantly reduced in PG>Inx2^DN animals compared to control..

(B) The total temporal density of Ca²⁺ events is significantly reduced in PG>Inx2^DN animals relative to control (P<0.0001, Student’s t-test).

(C) 10 second activity projection showing the total temporal density of Ca²⁺ elevations in control, and PG>Inx2^DN animals. Scale bar, 100 μm.

(D-G) Comparisons of Ca²⁺ transient characteristics in brain-PG of control and PG>Inx2^DN animals. Data was derived from event-based detection (see methods). N=5 animals/ 4 minutes imaging session each/genotype. Brain-PG activity in Inx2^DN animals show non-significant decrease in wide-spread event occurrence (D, >400 μm²). In contrast, Ca2+ events in control Brain-PG display significantly higher amplitudes (E, p<0.001, Student’s t-test), and longer durations (F, p<0.0001, Student’s t-test).

(G) Histogram summarizing the percent of flies exhibiting heat-shock induced seizures at 2 minutes (38.5°C) following conditional PG>Inx2^DN expression only in adult flies. Flies in which Inx2^DN expression was induced for 15 hours show enhanced seizures relative to controls (p<0.001, Student’s t-test, N=4 groups of 20 flies/genotype).

*=P<0.05, ** P<0.01, ***=P<0.001, ****=P<0.0001.

Ca²⁺ Signals in Perineurial Cells Occur at ER-Plasma Membrane Contacts

Astrocytes exhibit highly complex and dynamic fluctuations in Ca²⁺ that vary between different cellular compartments (Khakh & Deneen, 2019). To further explore the spread of Ca²⁺ activity in PG cells, we used a soluble variant of GCaMP6s (cytoGCaMP6s). This variant of GCaMP6s did not fill the entire volume of PG cells, and the recorded events displayed small amplitudes and a low signal-to-noise ratio (Figure S7C–D). To capture Ca²⁺ activity that occurs proximate to the ER in PG cells, we generated a transgenic Drosophila line expressing GCaMP6f tethered to the external surface of the ER ((Niwa, et al., 2016), ER::GCaMP6f, see Methods). We hypothesized that imaging ER-derived Ca²⁺ signals in proximity to its source might reveal faster events using the GCaMP6f variant that has faster kinetics. Surprisingly, the recorded events using the myr::GCaMP6s and ER::GCaMP6f sensors were similar (see below), indicating the slow kinetics observed using myr::GCaMP6s was not an artifact.

Imaging Ca²⁺ activity of brain-PG using ER::GCaMP6f revealed robust activity (Figures 7A, S7E–H and Video 9), with single events showing a similar waveform to that observed with myr::GCaMP6s (Figure 7B). Although GCaMP6f signals are reported to display smaller amplitudes than signals recorded with GCaMP6s (Chen et al., 2013), events recorded using ER::GCaMP6f had significantly larger amplitudes than those observed with membrane-bound myr::GCaMP6s (mean ΔF/F of 53.02 ± 2.9%, Figure 7C). This observation suggests PG Ca²⁺ events are being recorded close to their source. These Ca²⁺ transients recurred frequently in the same regions (^~3–4 events/min, Figure 7A, D), similar to what was observed using myr::GCaMP6s, providing additional support that all events observed in PG cells originate from internal ER Ca²⁺ stores. Single event detection revealed that ER::GCaMP6f detected events occasionally spread through several adjacent cells (Figure S7G, H), similar to what we observe for myr::GCaMP6s events (Figure S6D, E). The similar waveforms of the transients measured with ER::GCaMP6f and myr::GCaMP6s together with similar frequencies indicate that the majority of events in PG cells originate from ER Ca²⁺ stores.

Figure 7: ER Originated Ca²⁺ Waves in PG Cells Display Striking Complexity.

Imaging of PG>ER::GCaMP6f in dissected 3^rd instar wildtype Drosophila larvae.

(A) Top, schematic representation of the Drosophila larval brain shows the relative field of view at a section through a brain hemisphere (light blue). Bottom, Representative traces of mean ER::GCaMP6s fluorescence (% ΔF/F) show dynamic Ca²⁺ activity.

(B-D) myr:GcaMP6s and ER::GCaMP6f Ca²⁺ transients in brain-PG reveal similar kinetics.

(B) Superimposition of the means of myr:GcaMP6s and ER::GCaMP6f Ca²⁺ transients in brain-PG. (n>100 events/ N=3 animals/ sensor).

(C-D) Comparison of isolated Ca²⁺ transient amplitudes (% ΔF/F) recorded with myr:GcaMP6s and ER::GCaMP6f shows significant amplitude increase (C) and similar frequency (D) in ER::GCaMP6f recorded events relative to myr:GcaMP6s (p<0.0001, Student’s t-test, N=3 animals/ 4 minutes imaging session).

(E) Structural imaging of the cellular localization of ER-plasma membrane (ER:PM) contacts in PG cells. The PG driver was used to co-express a plasma membrane tethered GFP (mCD8::GFP) and a mCherry tagged Esyt2 (ESyt2::mCherry, as a marker of ER:PM contacts). Z-stack projection (3 μm) reveals substantial accumulation of ESyt2::mCherry around cell-cell contacts (visible as two parallel membranes marked with GFP, arrowheads). Scale bar, 5 μm.

****=P<0.0001.

The observation that activities recorded using myr::GCaMP and ER::GCaMP are similar, suggests PG Ca²⁺ activity localizes to ER:plasma membrane (PM) contacts. Together with the low signals recorded with cyto::GCaMP, these data suggest ER accumulates close to the PM to form ER:PM contacts. Indeed, examination of the cellular localization of ER::GCaMP6f in VNC-PG revealed that ER fills the entire cellular volume and accumulates around the nucleus and in the periphery of the cell where PG cell-to-cell contacts are formed (Figure S7I), consistent with a role for ER signaling in cell-to-cell communication. To further confirm this observation, we co-expressed the ER:PM-contact marker Esyt2 (Giordano et al., 2013) tagged with mCherry (Esyt2::mCherry, (Kikuma, Li, Kim, Sutter, & Dickman, 2017)) and a membrane-tethered GFP (mCD8::GFP) specifically in PG cells, and found extensive accumulation of Esyt::mCherry at cell-to-cell contact sites (Figure 7E). Together, these data indicate PG cells exhibit complex and diverse ER-originated Ca²⁺ activity that propagates through adjacent cells to form robust Ca²⁺ waves that spread over long distances in the Drosophila nervous system.

RyR Mediated Perineurial Ca²⁺ Wave Propagation is Crucial for the Prevention of Seizures

Ca²⁺ release from the ER through IP₃R and/or ryanodine receptors (RyR), is thought to be the major intracellular Ca²⁺ mobilization pathway (Bazargani & Attwell, 2016). However, our results suggest ER-related Ca²⁺ signaling in PG cells is independent of IP₃R. Activation of RyR dependent Ca²⁺-induced Ca²⁺-release (CICR) in astrocytes has been observed in response to various stimulations (Hua et al., 2004; Rodriguez-Prados, Rojo-Ruiz, Garcia-Sancho, & Alonso, 2020). As such, RyR-mediated CICR may be the primary mechanism that generates PG Ca²⁺ waves and acts to amplify Ca²⁺ signaling.

To investigate the role of RyR in PG Ca²⁺ waves, we first knocked down RyR using the PG driver. Similar to other genetic manipulations described above, RNAi-mediated knockdown of RyR was adult lethal (either partial or complete lethality, Figure S7J). To test the efficiency of the RyR RNAi, we measured the effect of RyR RNAi #1 on expression levels using qRT-PCR. RyR RNAi strongly reduced RyR expression levels (36.47 ± 0.9% decrease relative to Repo-GAL4 control, Figure S7K, Table S1). Recordings of CPG output at the larval NMJ demonstrated that 3^rd instar PG>RyR^RNAi larvae lose normal rhythmic firing at 38°C (Figure 8B, C), as observed in PG knockdown of dStim (Figure 2D, E). In RyR knockdowns that caused partial lethality, ^~40% of PG>RyR^RNAi flies that survived to adulthood showed HS-induced seizures (Figure 8C). Lastly, Ca²⁺ imaging (Figure 8D) revealed that PG>RyR^RNAi animals show a significant reduction in the temporal density, size, amplitude and duration of Ca²⁺ signals (Figure 8E–J), and significant inhibition of Ca²⁺ wave spread in PG cells (i.e. decrease in the abundance of events larger than 500μm², Figure 8G). Together, these data demonstrate that PG Ca²⁺ waves are mediated by RyR-dependent CICR and spread through neighboring cells via gap junctions. Blocking the spread of these Ca²⁺ waves recapitulate the behavioral phenotypes of SOCE knockdown, suggesting propagation of PG Ca²⁺ waves is crucial for maintaining normal brain excitability and preventing seizure activity.

Figure 8: RyR Mediated Perineurial Ca²⁺ Wave Propagation is Crucial for the Prevention of Seizures.

(A, B) CPG activity in PG>RyR^RNAi.

(A) Representative voltage traces of spontaneous CPG activity recorded at larval 3^rd instar muscle 6 at 38°C in control and PG>RyR^RNAi animals (n≥5 preparations/genotype).

(B) Quantification of percent time bursting for CPG recordings of PG>RyR^RNAi animals at room temperature and after a 38°C heat-shock (HS, marked with pink shading) (p<0.0001, One-way ANOVA, n≥5 preparations/genotype).

(C) Histogram summarizing the percent of flies exhibiting heat-shock induced seizures at 2 minutes (38.5°C) following PG knockdown of RyR with 3 non-overlapping RNAi hairpins (L= adult lethal for RNAi#1, p<0.0001 for RNAi#2, p<0.01 for RNAi#3, One-way ANOVA, N=4 groups of >10 flies/ genotype).

(D-J) Imaging of PG>myr::GCaMP6s in dissected 3^rd instar PG>RyR^RNAi#1 Drosophila larvae.

(D) Top, schematic representation of the Drosophila larval brain shows the relative field of view at the dorsal surface of the VNC (light blue). Bottom, representative traces of mean (% ΔF/F) show that PG Ca²⁺ activity is decreased in PG>RyR^RNAi#1 animals compared to controls.

(E) 10 seconds activity projection showing the total temporal density of Ca²⁺ elevations in control and PG>RyR^RNAi animals. Scale bar, 100 μm.

(F-J) Comparisons of Ca²⁺ transient characteristics in brain-PG of control and PG>RyR^RNAi animals. Data was derived from event-based detection (AQuA, see methods). N=5 animals/ 4 minutes imaging session each/genotype. Brain-PG activity in RyR^RNAi animals show significant decreases in the total temporal density of Ca²⁺ events (F, P<0.0001, Student’s t-test) and wide-spread event occurrence (G, >400 μm², P<0.01, Student’s t-test). The remaining Ca²⁺ activity in RyR^RNAi animals show significantly smaller area (H, P<0.001, Student’s t-test), smaller amplitudes (I, p<0.01, Student’s t-test), and shorter event durations (J, p<0.001, Student’s t-test).

*=P<0.05, **=P<0.01, ***=P<0.001, ****=P<0.0001.

Discussion

Ca²⁺ signaling is considered to be essential for astrocyte function (Khakh & Deneen, 2019; Khakh & McCarthy, 2015), with release from intracellular Ca²⁺ stores as one of the major underlying mechanisms. Genetic deletion of the inositol 1,4,5-trisphosphate receptor (IP₃R) type-2 (IP₃R2), the major IP₃R expressed in astrocytes, was reported to eliminate ER Ca²⁺ release in astrocytes. However, recent studies have shown the presence of astrocytic Ca²⁺ transients in IP₃R2 knockout mice (Agarwal, et al., 2017; Kanemaru et al., 2014; Rungta et al., 2016; Srinivasan et al., 2015). These IP₃R2-independent Ca²⁺ signals are thought to be mediated by plasma membrane Ca²⁺ influx (Rungta, et al., 2016; Shigetomi, et al., 2011; Srinivasan, et al., 2015) or Ca²⁺ release from mitochondria (Agarwal, et al., 2017). In contrast to the vast interest in astrocytic IP₃R related signaling, the role of ryanodine receptors (RyRs) in astrocyte Ca²⁺ signaling is largely unstudied and controversial (Rodriguez-Prados, et al., 2020). The expression and function of RyR3 in astrocytes has been reported (Chai et al., 2017; Matyash, Matyash, Nolte, Sorrentino, & Kettenmann, 2002), although the physiologically relevant mechanism for astrocytic RyR activation remains elusive. In this study, we found that genetic manipulations in the ER SOCE pathway in Drosophila PG cells leads to severe HS-induced seizures. We performed a detailed characterization of PG Ca²⁺ signaling that revealed PG cells exhibit robust Ca²⁺ activity. These Ca²⁺ transients are independent of extracellular Ca²⁺ and originate from internal stores as they are sensitive to Tg, knockdown of RyR, or knockdown of components of the SOCE pathway. We further show that Ca²⁺ waves propagating through the glial network via gap junctions are crucial for the generation of single PG cell Ca²⁺ transients. Ca²⁺ signals in PG cells are essential for controlling neuronal excitability, as knockdown of SOCE components or manipulation of gap junction function impairs basal motor activity and increases seizure susceptibility. These data indicate PG Ca²⁺ signaling involves store-dependent Ca²⁺ signaling and is essential for maintaining normal nervous system function.

A key question moving forward is how PG Ca²⁺ waves mechanistically regulate brain function. Maintenance of neuronal excitability requires a fine-tuned extracellular ion balance and a steady supply of nutrients and metabolites. This homeostasis is achieved by evolutionary conserved specialized structures that form the blood-brain barrier (BBB). The primary function of the BBB is to maintain homeostasis by regulating influx and efflux transport, a role that requires tight cell-to-cell interactions. The mammalian BBB consists of endothelial cells, astrocytes, pericytes, neurons, and microglia which shape the homeostatic function of the barrier (Alvarez, Katayama, & Prat, 2013). Several components of the BBB, including endothelial cells, astrocytes and pericytes exhibit fluctuations in intracellular Ca²⁺, suggesting a generalized role for Ca²⁺ activity in BBB function (Fujii, Maekawa, & Morita, 2017; Nagashima et al., 1997; Paemeleire, de Hemptinne, & Leybaert, 1999; Scemes & Giaume, 2006). In Drosophila, the BBB is formed by two glial layers: the PG and SPG cells (Figure 1A). The main barrier function is attributed to SPG cells that form pleated septate junctions and prevent paracellular diffusion, similar to tight junctions in the mammalian endothelial BBB. PG establish the first diffusion barrier, perform structural roles (i.e. secretion of the neural lamella and providing rigidity to the CNS) and provide SPG with metabolic support, although their exact contribution to BBB function is not fully understood (Limmer, Weiler, Volkenhoff, Babatz, & Klambt, 2014; Schirmeier & Klambt, 2015; Stork, et al., 2008; Yildirim, et al., 2019). We did not find defects in the gross morphology of PG when PG Ca²⁺ waves were disrupted. Similarly, the BBB diffusion barrier as assayed with dextran dye penetration was unaffected. As such, PG Ca²⁺ waves are likely to regulate either small molecule diffusion, SPG cell function or secretion of unknown factors from PG cells to shape neuronal excitability.

While interference with SOCE and the spread of Ca²⁺ waves across the PG sheet alters behavior and increases seizure susceptibility, the mechanism(s) downstream of intracellular Ca²⁺ changes that alter neuronal excitability is unknown. Based on our current observations, the different signatures in PG Ca²⁺ signaling across different brain regions suggest that although all PG cells utilize ER-store-dependent Ca²⁺ signaling, distinct PG cell populations may employ intracellular Ca²⁺ signaling in unique ways. The Drosophila hemolymph-brain barrier (PG and SPG) expresses numerous transporters and receptors that selectively move nutrients, metabolites and other compounds in and out of the brain (DeSalvo, et al., 2014). Hence, one possible mechanism by which PG Ca²⁺ could alter neuronal activity is through regulation of transport across the BBB. First, PG Ca²⁺ activity could modulate exocytotic/endocytotic cycling of membrane proteins within the PG layer itself, similar to its role in regulating surface levels of K⁺ channels in cortex glia (Weiss, et al., 2019) and GABA transporters in astrocyte-like glia (Y. V. Zhang, et al., 2017). Second, PG Ca²⁺ may regulate transport in SPG cells via a cell non-autonomous mechanism. A recent study found that efflux transporters in Drosophila SPG cells were regulated by a circadian clock in PG cells (S. L. Zhang, Yue, Arnold, Artiushin, & Sehgal, 2018). In this study, changes in PG Mg²⁺ balance were shown to regulate the activity levels of Pgp transporters (belong to the ATP-binding cassette (ABC) transporter family) in SPG cells (S. L. Zhang, et al., 2018). Though the gross anatomy of the BBB and basic diffusion of large molecules across the barrier were not affected, regulation of more subtle aspects of BBB transport of small molecules such as ions and metabolites could be altered. PG cells also play structural roles by secreting proteins composing the neural lamella and providing rigidity to the CNS (Yildirim, et al., 2019). As such, impairments in the fine structure of the BBB might indirectly affect transport. At the Drosophila neuromuscular junction (NMJ), Ca²⁺ release from the ER is involved in microtubule stabilization (Wong et al., 2014), and disruption in this process impairs synaptic growth, synaptic vesicle release probability and decrease synaptic transmission. Thus, loss of PG SOCE and subsequent depletion of ER Ca²⁺ stores might directly influence the secretion of proteins that compose the neural lamina, or induce destabilization of PG microtubules and alter rigidity that is crucial for brain homeostasis and function (Yildirim, et al., 2019). Further studies will be required to define how PG Ca²⁺ waves ultimately control neuronal excitability and whether different PG populations use distinct intracellular Ca²⁺ signaling pathways that are dependent on their unique Ca²⁺ wave dynamics.

Single-cell transcriptomic analyses of mammalian glial subtypes, including astrocytes, have advanced our understanding of astrocyte diversity. Astrocytes from different brain regions, as well as within the same region, have distinct transcriptomic profiles that allow classification into novel subpopulations with unique spatial distribution and signaling pathways (Khakh & Deneen, 2019; Yu, Nagai, & Khakh, 2020). To date, no molecular differences have been described between PG cells derived from the CNS or PNS, and despite morphological differences, PG cells are thought to share similar functional properties (Yildirim, et al., 2019). Transcriptomic analysis of Drosophila surface glia (PG and SPG together) demonstrated these cells collectively show molecular signatures similar to vertebrate brain-vascular endothelial cells that form the BBB (DeSalvo, et al., 2014). However, this analysis lacked single-cell resolution required for a molecular distinction of PG cells from different brain areas. Single-cell transcriptomic analysis of the Drosophila brain displayed relatively low coverage of PG cells (^~70 cells, (Davie et al., 2018)), preventing any distinction between possible subpopulations. Our data suggest PG cells derived from different brain regions can be distinguished based on Ca²⁺ wave dynamics, and further transcriptomic analyses might yield insights into the diversity of PG cells and glial cells in general.

Although astrocytes were traditionally considered to serve only supportive functions in the brain, the discovery of astrocytic intracellular Ca²⁺ signals has changed our view of how these cells contribute to brain function. Accumulating data indicate astrocytes can respond to neuronal activity and regulate neuronal function via intracellular astrocytic Ca²⁺ signaling. However, the functional consequences of glial Ca²⁺ signaling on neuronal physiology and brain function are not fully understood. One of the controversies in the field of glial biology is the functional distinction between ER-mediated somatic Ca²⁺ oscillations and near-membrane microdomain Ca²⁺ oscillations in glial processes. A central mechanism in intracellular Ca² signaling is the SOCE pathway which re-fills ER Ca²⁺ stores upon depletion triggered by a signaling cascade. In this pathway, the gating of the plasma membrane Ca²⁺ channel, Orai, is controlled by the ER-localized Ca²⁺ sensor, Stim, leading to Ca²⁺ influx and restoration of the ER Ca²⁺ store. While Orai and Stim expression have been detected in mammalian astroglia, the role of SOCE in glial biology is yet to be fully characterized. Our functional analysis indicates SOCE is a critical Ca²⁺ signaling pathway in PG cells that can act independently of IP₃ receptors.

Accumulating evidence indicate glia are likely to play a central role in a host of neurological disorders, with multiple disease-associated genes enriched in glial subtypes (Kelley, Nakao-Inoue, Molofsky, & Oldham, 2018). Studies investigating the mechanisms underlying epileptic seizures have primarily focused on neuronal origins, though accumulating evidence highlights an important role of non-neuronal cells in both the generation and spread of epileptic seizures in the brain. In this study, we found that alterations in the SOCE pathway in the Drosophila BBB lead to seizure-like episodes without affecting basic barrier function. This suggests that SOCE within the BBB regulates more subtle processes such as the regulation of active transport of small molecules or ions. In vertebrates, a tight connection between seizures and BBB dysfunction has also been found, with some studies showing that prolonged seizures or brain injury can lead to changes in BBB properties and subsequent BBB dysfunction, and other studies suggesting a causative role for BBB dysfunction in epileptogenesis (Gorter, Aronica, & van Vliet, 2019; van Vliet, Aronica, & Gorter, 2015). Future characterization of how glial Ca²⁺ signaling within the Drosophila BBB actively shapes neuronal excitability should shed light on the broader role of BBB function in the generation of seizures and suggest potential new treatment targets for epilepsy.

Main Points:

• Perineurial glia show diverse Ca²⁺ dynamics that vary based on their locale within the brain.

• Ca²⁺ waves are mediated by store-operated Ca²⁺ entry (SOCE), RyR and gap junctions.

• Disruption of Ca²⁺ waves triggers stimuli-induced seizures.

Data Availability Statement:

The data that support the findings of this study are available from the corresponding author upon request.