Drosophila CaV2 channels harboring human migraine mutations cause synapse hyperexcitability that can be suppressed by inhibition of a Ca2+ store release pathway

Gain-of-function mutations in the human CaV2.1 gene CACNA1A cause familial hemiplegic migraine type 1 (FHM1). To characterize cellular problems potentially triggered by CaV2.1 gains of function, we engineered mutations encoding FHM1 amino-acid substitutions S218L (SL) and R192Q (RQ) into transgenes of Drosophila melanogaster CaV2/cacophony. We expressed the transgenes pan-neuronally. Phenotypes were mild for RQ-expressing animals. By contrast, single mutant SL- and complex allele RQ,SL-expressing animals showed overt phenotypes, including sharply decreased viability. By electrophysiology, SL- and RQ,SL-expressing neuromuscular junctions (NMJs) exhibited enhanced evoked discharges, supernumerary discharges, and an increase in the amplitudes and frequencies of spontaneous events. Some spontaneous events were gigantic (10–40 mV), multi-quantal events. Gigantic spontaneous events were eliminated by application of TTX–or by lowered or chelated Ca2+–suggesting that gigantic events were elicited by spontaneous nerve firing. A follow-up genetic approach revealed that some neuronal hyperexcitability phenotypes were reversed after knockdown or mutation of Drosophila homologs of phospholipase Cβ (PLCβ), IP3 receptor, or ryanodine receptor (RyR)–all factors known to mediate Ca2+ release from intracellular stores. Pharmacological inhibitors of intracellular Ca2+ store release produced similar effects. Interestingly, however, the decreased viability phenotype was not reversed by genetic impairment of intracellular Ca2+ release factors. On a cellular level, our data suggest inhibition of signaling that triggers intracellular Ca2+ release could counteract hyperexcitability induced by gains of CaV2.1 function.


Introduction
Episodic neurological disorders like migraine, epilepsy, and ataxia can result from underlying ion channel dysfunctions [1][2][3]. For many such disorders, little is known about how aberrant channel functions affect neuronal signaling paradigms. Cell-based and model organism-based examinations of disease-causing mutations could offer insights into disease-relevant biological processes. One Mendelian form of migraine-familial hemiplegic migraine type 1 (FHM1)results from gain-of-function missense mutations in human CACNA1A, which encodes the α1 subunit of Ca V 2.1 (P/Q)-type calcium channels [4]. Two FHM1-causing amino-acid substitutions alter highly conserved Ca V 2.1 α1 amino-acid residues, R192 and S218 [4,5]. The R192Q amino-acid substitution (RQ) causes "pure" FHM1, while the S218L substitution (SL) causes a severe combination of FHM1, seizures, and susceptibility to edema following head injury [4,5]. These two FHM1-causing amino-acid substitutions have been studied intensely [6], most notably in knock-in mouse models of FHM1 [7][8][9].
FHM1 knock-in mice display gain-of-function Ca V 2.1 phenotypes at neurons and synapses. Model synapses studied include the diaphragm neuromuscular junction (NMJ) [10,11], the calyx of Held [12][13][14], the trigeminal sensory neuron pathway [15][16][17], and cortical neurons [18,19]. At the mouse NMJ, both RQ and SL increase the frequency of spontaneous excitatory potentials [10,11]. These increases in quantal frequency are dependent on mutation dose and are more pronounced in SL versus RQ. SL also elicits broadening of evoked end-plate potentials at the mouse NMJ [11]. At the calyx of Held, both substitutions result in enhanced excitatory postsynaptic currents (EPSCs) [12][13][14], and it has been reported that SL causes an increase in the resting intracellular neuronal calcium, which could be responsible for some potentiation of synapse function [12].
It was recently reported that 2,5 0 -di(tertbutyl)-1,4,-benzohydroquinone (BHQ) reverses aspects of SL-induced gating dysfunction and short-term plasticity [20]. As part of that study, we found that BHQ also restores short-term synaptic plasticity to NMJs in fruit fly larvae expressing a transgene that encodes an S161L amino-acid substitution in Drosophila Ca V 2/ Cacophony-the functional equivalent of human Ca V 2.1 S218L [20]. Independent follow-up work in the mouse S218L model demonstrated that BHQ application also blunts cortical spreading depression susceptibility [21]. Given these collective results, a further examination of fruit fly synapses could be valuable for uncovering relevant molecular and electrophysiological consequences of Ca V 2.1 gains of function.
For the present study, we characterized the fruit fly as a way to model neuronal effects of FHM1-causing mutations. We neuronally expressed cacophony transgenes harboring the Drosophila melanogaster equivalents of RQ or SL-or both RQ and SL concurrently (denoted as "RQ,SL"). On the organismal level, neuronal expression of SL or RQ,SL transgenes drastically impaired overall health. On the synapse level, SL and RQ,SL transgenes markedly enhanced aspects of evoked and spontaneous neurotransmission, consistent with prior studies in mice. Through a combination of genetics, RNA interference, pharmacology, and electrophysiology, we uncovered evidence that impairment of a conserved intracellular signaling pathway that triggers store Ca 2+ release reverses hyperexcitability phenotypes in the context of gain-of-function Drosophila Ca V 2.

Transgenic Drosophila Ca V 2 "FHM1" channels cause coarse larval phenotypes and fly lethality
We utilized Drosophila melanogaster to study the impact that FHM1-inducing Ca V 2.1 aminoacid substitutions may exert on the level of individual synapses. Drosophila cacophony encodes the α1 subunit of fruit fly Ca V 2-type channels. We cloned two amino-acid substitutions that cause human FHM1 (S218L and R192Q) into the analogous codons of a functional Drosophila UAS-cacophony (cac)-eGFP transgene [22]. Single mutant transgenes were termed "SL" (UAScac-eGFP S161L ) [20] or "RQ" (UAS-cac-eGFP R135Q ) ( Fig 1A). We also generated a transgene containing both mutations in cis on the same cDNA clone, termed "RQ,SL" (UAS-cac-eGFP R135Q, S161L ). This is not a naturally occurring allele in humans with FHM1. We reasoned a priori that this complex allele could yield a genetically sensitized background for Ca V 2 gainof-function in Drosophila. "WT" signifies previously characterized wild-type UAS-cac-eGFP WT transgenes [22].
We expressed WT, RQ, SL, and RQ,SL UAS-cac-eGFP transgenes in post-mitotic Drosophila neurons using the elaV(C155)-Gal4 driver and the Gal4/UAS expression system [23,24]. We examined transgenic animals qualitatively for visible phenotypes. Neuronal expression of either SL or RQ,SL caused larvae to move in a jerky, uncoordinated manner. At the early third instar stage, SL-and RQ,SL-expressing animals developed protruding, anterior spiracles prematurely-well before the normal time point of wandering third instar stage and pupation ( Fig  1B).
Our initial observations indicated that SL-and RQ,SL-expressing animals were not present in expected Mendelian proportions. For each transgene (WT, RQ, SL, and RQ,SL), we set up test crosses (elaV(C155)-Gal4 females x Balancer Chromosome/UAS-cac-eGFP males) and counted the number of transgenic UAS-cac-GFP-expressing adult progeny and the number of sibling flies carrying a balancer chromosome. We also set up Gal4 and balancer chromosome control crosses lacking any UAS-cac-eGFP transgenes (Table 1). Compared to animals expressing the WT transgene, viability was dramatically diminished for animals expressing the SL and RQ,SL transgenes ( Fig 1C, Table 1). It was also diminished for SL-and RQ,SL-expressing animals compared to genetically matched control siblings carrying the elaV(C155)-Gal4 driver and a balancer chromosome (Table 1). WT-and RQ-expressing animals did not show significant defects in viability or statistical differences from the Gal4 control cross (Fig 1C, Table 1).
As expected, there was some depressed viability in animals carrying a balancer chromosome alone (Table 1).
Sex or dose of the SL and RQ,SL transgenes could influence viability. In Drosophila, Xlinked dosage compensation equalizes the expression of X-linked genes by doubling X-linked gene transcription in males [25][26][27]. The X-linked neuronal enhancer trap Gal4 line elaV (C155)-Gal4 should be expressed at higher levels in hemizygous elaV(C155)-Gal4/Y males than in heterozygous elaV(C155)-Gal4/+ females. Thus, effects of driving UAS transgenes could be stronger in males. Counting male vs. female progeny of SL-and RQ,SL-expressing flies revealed that while viability was starkly diminished for both sexes, it was also significantly showing diminished UAS-cac-eGFP mutant viability. "UAS-cac Viability Index" = # UAS-cac-eGFP transgenic adult progeny/# Balancer Chromosome siblings, normalized to 100% for WT female progeny counts (Table 1 for  lower in SL-and RQ,SL-expressing males than in SL-and RQ,SL-expressing females (Fig 1C,  Table 1).

Cac-GFP localizes normally and levels are comparable across transgenic constructs
We investigated why SL-and RQ,SL-expressing animals were showing overt phenotypes. We considered the possibility that excessive quantities of α1 protein generated via the GAL4/UAS expression system could reduce viability. Opposing this idea, neuronal overexpression of WT UAS-cac transgenes renders no reported structural, behavioral, or electrophysiological abnormalities [22,28]. Moreover, overexpressed Cac-GFP protein efficiently localizes to active zone structures at synapses like the larval neuromuscular junction (NMJ) [20,22,[29][30][31][32].
Using wandering third instar larvae and elaV(C155)-Gal4 driver, we first checked Cac-GFP localization of several transgenic lines: WT (published line, UAS-cac-eGFP 786c ) [22], RQ,SL (UAS-cac-eGFP RQ,SL(2M) ) (this study), SL (UAS-cac-eGFP SL (3-2M) ) [20], and RQ (UAS-cac-eGFP RQ(1M) ) (this study). We used an anti-GFP antibody to detect Cac-GFP and co-stained with a monoclonal antibody against the presynaptic ELKS/CAST active zone protein Bruchpilot (Brp) [33]. In all cases, Cac-GFP localized as expected in the larval central nervous system  (Fig 2A-2D, red channel). It also predominantly localized to presynaptic active zone sites at neuromuscular junction (NMJ), as expected ( Fig 2E-2H), consistent with the reports for the original WT constructs [22,29]. We checked Cac-GFP levels for the different transgenic constructs by Western Blot. We drove the transgenes neuronally using elaV(C155)-Gal4 and collected adult heads for analysis, blotting for Cac-GFP (239 kDa) with anti-GFP and anti-Actin as a loading control. Compared to elaV(C155)-Gal4 line controls, each UAS-cac-eGFP transgenic line showed an additional, faint band that migrated at a size consistent with Cac-GFP (Fig 2I). This band was expressed at comparable levels between the WT, RQ, SL, and RQ,SL lines ( Fig 2J), with no appreciable difference in control levels of actin between the lines (Fig 2K). Wild-type and mutant Cac-GFP successfully localized to NMJ active zones, as indicated by co-staining with anti-Brp (green) and anti-GFP (red). Scale bar 5 μm. (I) Western blots of fruit fly head lysates (10 heads/lane, single sex per Western), from flies expressing either elaV-Gal4 alone or the indicated UAS-cac-eGFP transgene driven by elaV-Gal4. Blots were probed with anti-GFP (top) and anti-Actin (bottom) antibodies. The band corresponding to Cac-GFP is indicated. Other bands are non-specific. (J) Compared to WT, there was no statistically significant change in Cac-GFP expression for any of the transgenic lines utilized in this study (band normalized to actin; p > 0.65, one-way ANOVA with Dunnett's multiple comparisons vs. WT; GAL4 alone control excluded from analysis). (K) Actin levels were also steady across all transgenic lines (p > 0.71, one-way ANOVA with Dunnett's multiple comparisons vs. WT). https://doi.org/10.1371/journal.pgen.1007577.g002

RQ,SL-expressing NMJs show small changes in bouton number and glutamate receptor coverage
Even in the absence of localization or expression-level differences, transgenic mutant Cac-GFP expression could affect synapse growth or development. Previously, we found no significant abnormalities in NMJ synaptic growth for SL-expressing flies [20]. We extended our analysis to the RQ,SL transgene line by co-staining third instar larval NMJs with antibodies against the Drosophila PSD-95 homolog, Discs Large (Dlg) and the GluRIIA glutamate receptor subunit (Fig 3A-3D). We observed a very small decrease in the number of Dlg-positive synaptic boutons at RQ,SL-expressing NMJs compared to control WT-expressing NMJs. This decrease was statistically significant only for segment A2, muscle 6/7 ( Fig 3E). We found no significant change in the number of glutamate receptor clusters per NMJ comparing WT-expressing synapses and RQ,SL-expressing synapses (Fig 3F).  At RQ,SL-expressing NMJs, the percentage of the synaptic area covered by the GluRIIA clusters-normalized to total Dlg area-was slightly but significantly increased ( Fig 3G). In principle, an expansion of the synaptic area capable of receiving neurotransmitter could underlie gains in synaptic transmission [34]. The magnitude of any such change based on this postsynaptic staining profile alone would likely be small but was uncertain based on these measures. We needed to conduct finer analyses by electrophysiology, both to document possible changes in synaptic function and also to test for potential presynaptic contributions when mutant cac transgenes were expressed.

RQ-, SL-, and RQ,SL-expressing NMJs display hyperexcitable evoked synaptic discharges
Coarse phenotypes from neuronally expressed RQ,SL and SL transgenes (Fig 1)  . Expression of RQ numerically increased average NMJ EPSP amplitudes, but this increase was not statistically significant ( Fig 4A). Neither estimated quantal content (QC) (Fig 4B) nor calcium cooperativity of release for mutant lines were significantly different than WT across this range of 0.2-0.5 mM [Ca 2+ ] (Fig 4C) [20] (but see more detailed quantal analyses later).
We noted that the EPSP waveforms of RQ, SL, and RQ,SL animals were sometimes abnormal (Fig 4D and 4E). In addition to increases in EPSP amplitude (Fig 4D), we observed two distinct EPSP waveform phenotypes: 1) 'extra discharges' ("ED"), in which supernumerary spiking events occurred during the decay phase of the EPSP waveform (Fig 4E, left); and 2) 'shoulders,' in which there was an extended discharge during the decay phase of the EPSP ( Fig  4E, right), causing a discontinuity in the decay. These phenotypes were somewhat reminiscent of a broadening of the end-plate potential previously reported at the NMJs of SL knock-in mice [11]. The SL-expressing NMJs produced only the extra discharge type of abnormal waveform, whereas the RQ-expressing NMJs produced only the shoulder form (Fig 4E and 4F). Consistent with both mutations being present in the RQ,SL line, those NMJs exhibited both types of abnormal waveform ( Fig 4F).
We were also able to generate "RQ only" animals-functional null X-ray cac HC129 mutant [35] larvae rescued to viability by elaV(C155)-Gal4-driven neuronal expression of the RQ transgene. The cac HC129 allele works well for this type of genetic maneuver [22, 28], eliminating endogenous cac gene expression, while adding back transgenic cac. In the case of "RQ only", the waveform dysfunction closely matched that shown by the RQ-expressing NMJs (Fig 4F)i.e. a shoulder waveform phenotype was present. We were unable to generate "SL only" or "RQ,SL only" animals, possibly due to deleterious gains of function from the SL mutation.
We assessed the severity of the extra discharge phenotype by counting the number of extra discharge events per 30 evoked pulses (30 recording sweeps at 1 Hz per NMJ). Quantification confirmed that SL-and RQ,SL-expressing NMJs were highly dysfunctional, suggesting neuronal hyperexcitability (Fig 4G). A previous study in Drosophila demonstrated that higher levels of magnesium in the recording saline can mask hyperexcitability of neurons [36]. Therefore, we conducted additional WT and RQ,SL recordings in saline with lowered [MgCl 2 ] (6 mM vs. 10 mM for normal saline, see Materials and Methods). RQ,SL-expressing NMJs displayed extreme dysfunction in low MgCl 2 , both in terms of the percentage of NMJs that produced any supernumerary discharges (100%, Fig 4H) and the number of extra discharges counted Drosophila models of Ca V 2 migraine mutations per 30 presynaptic pulses (Fig 4I). By contrast, WT-expressing NMJs showed almost no such dysfunction (Fig 4H and 4I).
Finally, we conducted an additional series of recordings in normal saline, this time with the larval CNS left intact to check if the hyperexcitability might reflect an in vivo state for Drosophila larvae. With this experimental maneuver, it was possible to discern "native circuit" differences between WT-and RQ,SL-expressing animals. The "CNS intact" condition resulted in trains of spontaneous activity. Compared to WT, the RQ,SL-expressing NMJs displayed a high degree of spontaneous activity, marked by rapid, continuous large pulses (Fig 4J; see several measures and explanation in legend).
In conclusion, SL-and RQ,SL-expressing NMJs displayed evoked gain-of-function phenotypes consistent with prior mammalian FHM1 mutant analyses. By contrast, RQ-expressing NMJs only displayed a mild gain-of-function shoulder phenotype.

SL-and RQ,SL-expressing NMJs show enhanced spontaneous miniature EPSPs with respect to both amplitude and frequency
Mammalian models of FHM1 show dysfunctional spontaneous neurotransmission [10,11]. We extended our electrophysiological analyses at the Drosophila NMJ to quantal neurotransmission. We observed a striking phenotype: for SL-and RQ,SL-expressing NMJs, there was an enhancement in both amplitude and frequency of spontaneous miniature EPSPs (mEPSPs) (Fig 5A-5E, Table 2). By contrast, neither an increase in spontaneous mEPSP amplitude nor mEPSP frequency were observed for RQ-or WT-expressing NMJs compared to non-transgenic w 1118 controls (Fig 5B-5E, Table 2).
Since the mutations examined are in a voltage-gated calcium channel, it was important to document electrophysiological behavior at various calcium concentrations. At both 0.5 mM and 0.4 mM extracellular [Ca 2+ ], analyses of thousands of individual spontaneous events revealed that increases in spontaneous amplitudes were due to an overall increase in the size distribution of the events at SL-and RQ,SL-expressing NMJs (Fig 5D-5G). Additionally, at both 0.5 mM and 0.4 mM extracellular [Ca 2+ ], we noted that the spontaneous events at SLand RQ,SL-expressing NMJs included a minority of gigantic spontaneous events (10-40 mV) that were never seen in w 1118 or WT-expressing controls or in RQ-expressing NMJs (Fig 5A  and 5D-5G). Notably, these gigantic events were seen in the complete absence of presynaptic nerve stimulation in nerves that had already been severed from the central nervous system.
It was uncertain if enhanced spontaneous excitability was due to real-time expression of gains of function in Ca V 2 channel gating kinetics, long-term developmental alterations at the synapse-or if both factors could contribute. We considered altered Ca V 2 kinetics. It was previously demonstrated that the SL mutation causes complex biophysical alterations to Ca V 2.1 gating function, both by enhancing voltage-dependent activation [8,9,37] and by inhibiting calcium-dependent facilitation [38]. Follow-up work showed that the drug 2,5 0 -di(tertbutyl)-1,4,-benzohydroquinone (BHQ) opposes those effects, reversing SL-induced gains of function [20]. As part of the same study, we showed that BHQ restores a form of short-term synaptic plasticity at SL-expressing Drosophila NMJs [20]. We extended those prior analyses of BHQ effects on Ca V 2 gating, this time by examining the distribution of spontaneous events. We found that acute application of 5 μM BHQ was partially effective at reversing the increased size distribution of events for SL-and RQ,SL-expressing NMJs, without changing the distribution of WT events (Fig 5F and 5G; Table 2). Notably, 5 μM BHQ did not abolish gigantic events (Fig 5F and 5G). Interestingly, a higher concentration of 10 μM BHQ did abolish gigantic events for SL-and RQ,SL-expressing NMJs, but it also significantly decreased the size distribution of WT mEPSPs, which could indicate off-target postsynaptic effects ( Fig 5F). Our BHQ application data are consistent with the idea that that spontaneous neurotransmission gain-offunction phenotypes are driven in part through gating changes at Ca V 2 channels.
To test if long-term developmental alterations at the synapse could also play a role, we engineered stage-specific UAS-cac transgene expression. We utilized the temperature-sensitive Gal80 TS /TARGET system to temporally control expression of the RQ,SL transgene [39]. To conduct this experiment, we generated elaV(C155)-Gal4 >> UAS-cac-eGFP RQ,SL animals with a ubiquitous Gal80 TS transgene [39]. Gal80 TS protein halts GAL4-induced gene expression at permissive temperatures (25˚C) but not at restrictive temperatures (29˚C). For our experiment, animals raised at 25˚C throughout life had no discernible spontaneous neurotransmission hyperexcitability (Fig 5H). By contrast, animals started at 25˚C and shifted to 29˚C for the final 24 or 48 hours before third instar NMJ recording showed progressively more spontaneous hyperexcitability (Fig 5H and 5I). This experiment indicates that developmentally regulated expression of gain-of-function Ca V 2 channel subunits also underlies some of the spontaneous neurotransmission gain-of-function phenotypes.

Gigantic spontaneous events require extracellular calcium and sodium channel activity
Prior work proposed that mammalian neuronal dysfunction downstream of FHM1 mutations may be calcium-dependent [12]. We tested whether the observed effects on quantal size in our model could be calcium-dependent. First, we reduced the extracellular [Ca 2+ ] in the recording saline to 0.2 mM. Consistent with classic characterizations of Drosophila NMJ properties [40], low calcium did little to change the distribution of mEPSP size, the median mEPSP size, or the 25 th -75 th percentiles of mEPSP size-all of which remained normal for WT and elevated for SLand RQ,SL-expressing NMJs (Fig 6A and 6B, Table 2). However, lowering extracellular [Ca 2+ ] almost completely abrogated gigantic (10-40 mV) spontaneous events at SL-and RQ,SLexpressing NMJs-and it completely eliminated the very largest ones (Fig 6A and 6B). This suggested that these gigantic events somehow relied on a sufficient driving force of presynaptic calcium influx-and potentially on spontaneous presynaptic nerve firing.
We extended these analyses by altering the recording saline in three additional ways: 1) zero extracellular calcium; 2) adding the membrane-permeable calcium chelator, 1,2-Bis (2-aminophenoxy) ethane-N,N,N 0 ,N 0 -tetra acetic acid tetrakis (acetoxymethyl ester) (BAP-TA-AM, 10 μM); or 3) adding tetrodotoxin (TTX, 3 μM) to block voltage-gated sodium channels. We compared WT-expressing and RQ,SL-expressing NMJs (and SL-expressing NMJs in the case of zero calcium). All three manipulations produced a similar effect on mEPSP size for the gain-of-function mutants: an elimination of gigantic spontaneous events, but a persistence of overall elevated mEPSP size (Fig 6C-6F, Table 2). By contrast, these manipulations had little to no effect on the distribution of mEPSP amplitudes at WT-expressing NMJs (Fig 6C and 6D, Table 2).
Finally, we recorded spontaneous events in more in vivo-like condition, using an intact CNS, without severing the motor nerve. In order to do this, we revisited the intact CNS condition ( Fig 4J)-this time adding TTX to the recording saline (0.5 mM [Ca 2+ ]). This left the full network anatomy intact, while quieting spontaneous trains of activity. Under these conditions, the spontaneous event amplitude profile of RQ,SL-expressing NMJs was still larger than that of WT-expressing NMJs-and as expected, there were no gigantic events (Fig 6G and 6H). Interestingly, however, the difference between WT-expressing NMJs and RQ,SL-expressing NMJs was muted (Fig 6G and 6H; compare to Fig 5D and 5E). These data suggest that in living animals, network effects could potentially influence the spontaneous gain-of-function activity.

Large spontaneous events are due to multi-vesicular release
The presence of gigantic spontaneous mEPSPs that were sensitive to low calcium, calcium chelation, and TTX treatment suggested the possibility of spontaneous multi-vesicular release at SL-and RQ,SL-expressing NMJs. If this were true, traditional analysis of spontaneous mEPSPs would result in an overestimation of average quantal size ( Fig 5B) and underestimation of average QC (Fig 4B) for SL-and RQ,SL-expressing NMJs. Average mEPSP amplitudes ± SEM and mEPSP frequencies ± SEM for selected conditions. Also given: median mEPSP amplitudes and maximum mEPSP amplitudes achieved for spontaneous events analyzed (~100 per NMJ). w 1118 is a non-transgenic wild-type control. WT, RQ, SL, and RQ,SL are shorthand for the indicated UAScac-eGFP transgene being driven in male progeny presynaptically by the elaV(C155)-Gal4 driver. These data illustrate differential effects when lowering extracellular [Ca 2+ ], chelating Ca 2+ with BAPTA-AM, or inactivating Na V channels with TTX. Electrophysiological data were analyzed in two ways as average per NMJ and as cumulative distributions.  We utilized the method of failures to better resolve questions about quantal size and QC. At very low concentrations of extracellular calcium, synapses like the NMJ are essentially limited to a one-or-none evoked response in which stimulation of the presynaptic nerve either leads to the release of a single vesicle or fails to release any vesicles [41]. By conducting failure analyses, it is possible to measure the distribution of quantal events and also to estimate QC in a way that eliminates confounds of higher concentrations of calcium. First, we conducted failure analysis recordings at 0.14 mM [Ca 2+ ] e for WT-, RQ,SL-, and SL-expressing NMJs (Fig 7A-7C). For this condition, the evoked events for SL-and RQ,SL-expressing NMJs were far larger on average than those observed WT-expressing NMJs (Fig 7C-EPSP). This was due to a large proportion of events of > 2 mV for the SL-and RQ,SL-expressing conditions (compare Fig 7A  and 7B). Furthermore, even in this low level of extracellular Ca 2+ , many of the RQ,SL and SL events represented multi-vesicular release rather than the release of a single large vesicle. We calculated values of QC of > 2 for both mutant conditions at 0.14 mM [Ca 2+ ] e (QC = m = ln [(# trials)/(# failures)] [42]) (Fig 7C).
To test if lower calcium could generate a leftward shift in event size, we applied a more restrictive condition of 0.1 mM [Ca 2+ ] e to RQ,SL-expressing NMJs. At 0.1 mM [Ca 2+ ] e the proportion of failures was very high for RQ,SL-expressing NMJs, with events over 4 mV all but absent, and events greater than 1.5 mV also less prevalent (Fig 7D). The first peak in the distribution of events, which is reflective of single vesicle size [42], was centered near 0.7 mV ( Fig  7D), a value consistent with single-vesicle responses of normal size for the Drosophila NMJ [40]. Together, these data suggested that the observed large events at SL-and RQ,SL-expressing NMJs-regardless of whether spontaneous or failure analysis-evoked-were likely due to multi-vesicular release (see Fig 7E, spontaneous and failure analyses distributions side-byside).
If larger spontaneous events are multi-vesicular (or at the very least include a proportion of multi-vesicular events), this property should also be reflected in slowed spontaneous event rise time kinetics. We analyzed the rise time kinetics of several thousand spontaneous events for w 1118 , WT-, RQ-, SL-, and RQ,SL-expressing NMJs. Average rise times were slowed only for RQ,SL-expressing NMJs (Fig 7F). However, the rise times for larger events were markedly slower for all genotypes, not just RQ,SL (Fig 7G). For SL-and RQ,SL-expressing NMJs there was a much larger proportion of such events. Collectively, our data suggest that large events (> 2 mV) include several that are multi-vesicular.

PLCβ loss genetically suppresses spontaneous excitability
For SL-and RQ,SL-expressing NMJs, we hypothesized that specific cellular cues could dictate the various electrophysiological phenotypes we documented: multi-vesicular quantal events, gigantic TTX-sensitive spontaneous events, and enhanced NMJ excitability. We inquired as to what the molecular nature of those cues might be. Our experiments indicated that intracellular calcium or intracellular calcium signaling processes might be important (Fig 6). Additionally, recent data from the mouse calyx of Held demonstrated that S218L knock-in synapses have enhanced resting intracellular calcium [12]. We hypothesized that altered intracellular calcium signaling or handling could impact myriad intracellular signals and investigated which signaling pathways might be relevant. This line of inquiry spurred a genetic approach examining regulators of intracellular calcium to test if inhibition of any of these factors may influence gain-of-function Ca V 2 phenotypes at the synapse (Fig 8A). We sought to identify suppressors capable of reversing gains of Ca V 2 function caused by the SL and RQ,SL transgenes.
We targeted Plc21C gene expression in neurons with a previously verified UAS-Plc21C (RNAi) construct, Plc21C GD11359 [45,54]. Compared to the NMJs of w 1118 and WT controls, those in which only Plc21C had been knocked down presynaptically exhibited no discernable baseline changes in mEPSP size (Fig 8B, Table 3)-or as previously documented, EPSP size, or QC [45]. By contrast, in RQ,SL-expressing NMJs such Plc21C knockdown alleviated aspects of NMJ hyperexcitability. Specifically, there was a leftward shift in the distribution of spontaneous events (Fig 8B, Table 3). Interestingly, there was not a significant reversal of the enhanced mEPSP frequency phenotype (Table 3).

IP 3 R and RyR point mutations strongly suppress hyperexcitability
We hypothesized that Plc21C could exert effects on spontaneous neurotransmission via one of several components of its canonical signaling pathway (e.g. PIP 2 , DAG, or IP 3 ). Notably, IP 3 acts through the IP 3 receptor (IP 3 R), an intracellular calcium channel located on the endoplasmic reticulum (ER). At other model synapses, release of Ca 2+ from the intracellular stores can promote the release of neurotransmitter-laden vesicles and contribute to the amplitudes of spontaneous events [55][56][57][58]. Moreover, IP 3 R has been proposed to play a role in spontaneous vesicle release through calcium-induced calcium release (CICR) [59], and increased ER Ca 2+ release was recently shown to potentiate synaptic transmission at the Drosophila NMJ [60].
We examined the Drosophila IP 3 R gene (itpr). Homozygous itpr loss-of-function mutations are lethal, so we tested a heterozygous loss-of-function condition. Since IP 3 R clusters consist of multiple units, we hypothesized that we might be able to partially disrupt them through a loss-of-function point mutation, itpr ug3 , a mutant possessing a missense mutation in the IP 3 R ligand-binding domain [61]. itpr ug3 /+ phenocopied Plc21C knockdown at RQ,SL-expressing NMJs: the mEPSP amplitude was partially reduced toward WT levels, and the number of giant, spontaneous events was diminished (Fig 8C and 8D, Table 3). Importantly, on its own Average mEPSP amplitudes ± SEM and mEPSP frequencies ± SEM for selected experimental conditions. Also given are the median mEPSP amplitude and the maximum mEPSP amplitudes achieved for all spontaneous events analyzed per genotype (~100 per NMJ). w 1118 is a non-transgenic wild-type control. WT and RQ,SL are shorthand for the indicated UAS-cac-eGFP transgene being driven in progeny presynaptically by the elaV(C155)-Gal4 driver. This Drosophila models of Ca V 2 migraine mutations itpr ug3 /+ did not significantly affect the baseline amplitude or distribution of mEPSPs (Fig 8C,  Table 3). Finally, as with the RNAi experiment, the increased mEPSP frequency phenotype was not suppressed (Table 3). We performed analogous experiments with a Drosophila ryanodine receptor gene (RyR) mutation. Tetrameric RyR channels have been reported to contribute to CICR downstream of IP 3 Rs [59]. Additionally, gigantic spontaneous miniature potentials at other model synapses are mediated by RyR and rapid expulsion of calcium from presynaptic stores [62][63][64][65][66]. We found that the heterozygous RyR point mutant RyR E4340K /+ [67] almost completely suppressed the increased average mEPSP amplitude in the RQ,SL-expressing background (Fig 8C and 8E, Table 3). Additionally, the gigantic spontaneous events were abrogated (Fig 8C and 8E, Table 3). Control recordings showed that RyR E4340K /+ did not affect the baseline amplitude or distribution of mEPSPs (Fig 8C and 8E). As with Plc21C and itpr, impairment of RyR function did not significantly suppress the enhanced mEPSP frequency phenotype of RQ,SL-expressing NMJs (Table 3).
Because the RyR E4340K /+ background provided such a strong suppression of spontaneous mEPSP hyperexcitability at RQ,SL-expressing NMJs, we checked if it could also suppress hyperexcitability in the context of evoked excitation. As shown before, when incubated in low extracellular magnesium, 100% of the RQ,SL-expressing NMJs showed a hyperexcitability dysfunction, with high expressivity of extra discharges (Figs 3H and 3I; 8F). In a heterozygous RyR E4340K /+ genetic background, this hyperexcitability phenotype was partially suppressed, in terms of both the penetrance of NMJs with extra evoked discharges and the expressivity of the extra discharge dysfunction at individual NMJs (Fig 8F). On its own, the RyR E4340K /+ condition shows almost no baseline hyperexcitability phenotype (Fig 8F).

Spontaneous mEPSP hyperexcitability can be suppressed pharmacologically
Our data for genetic manipulations affecting Plc21C, IP 3 R, and RyR show that it is possible to attenuate RQ,SL-induced gain-of-function mEPSP amplitude and excitability phenotypes by genetically impairing factors known to promote intracellular Ca 2+ release. We wondered if pharmacological manipulations could also be effective. We turned to two agents to test this idea: lithium (10mM LiCl in larval food) and Xestospongin C (5 μM in recording saline). Chronic exposure to lithium inhibits inositol monophosphate phosphatase, eventually resulting in a disruption of the recycling process that generates PIP 2 [68,69]. Xestospongin C has been previously characterized as a membrane-permeable inhibitor of IP 3 receptors [70,71]. Either chronically feeding larvae LiCl or applying Xestospongin C to the recording bath caused a significant leftward shift in the overall size distribution of spontaneous amplitudes (Fig 9A-9C), reminiscent of the effects observed for Plc21C, itpr, and RyR losses of function. The acute Xestospongin C application seemed to exert a stronger suppression effect in this regard, while the chronic LiCl application exerted a stronger suppression of the gigantic spontaneous events (Fig 9A-9C, Table 3). Notably, neither pharmacological manipulation diminished baseline spontaneous neurotransmission in WT-expressing control NMJs, nor did either manipulation significantly suppress the elevated mEPSP frequency phenotype for RQ,SL-expressing NMJs (Fig 9A-9C, Table 3).

Mutations targeting intracellular calcium release signaling can exacerbate lethality
Since genetic mutations that target intracellular calcium release ameliorate hyperexcitability phenotypes, we reasoned that the same (or similar) mutations might ameliorate the lethality phenotypes associated with expressing the RQ,SL transgene. We conducted lethality test crosses and progeny counts in a similar manner as before (Table 1). This time, we crossed females bearing the UAS-cac-eGFP RQ,SL transgene to males carrying both the elaV(C155)-Gal4 driver and a collection of loss-of-function genetic manipulations on Drosophila melanogaster Chromosome II for the Plc21C, RyR, or Gq genes. In addition to Plc21C and RyR, we chose Gq because canonical PLCβ signaling is downstream of Gαq function. Our prior work showed that Plc21C and Gq play a role in the maintenance of homeostatic plasticity at the NMJ [45]. The hypothesis to test was that female progeny carrying the driver, the RQ,SL transgene, and the intracellular calcium release manipulation could have improved viability versus female progeny carrying only the driver and the RQ,SL transgene. Male progeny siblings would not carry the driver-and would therefore not express the RQ,SL transgene-and could be used to control for parameters affecting lethality, independent of the RQ,SL transgene.
As expected, female progeny carrying the driver, the RQ,SL transgene, and no balancer chromosome had reduced viability compared to their male sibling counterparts (Table 4; see "+"). However, introducing heterozygous loss-of-function manipulations affecting Plc21C, RyR, and Gq did not ameliorate this phenotype. Surprisingly, those manipulations almost always further reduced viability, often strongly ( Table 4). The effect was particularly strong for all Plc21C and Gq loss-of-function conditions examined (Table 4). For RyR, the effect was strong only for the RyR 16 deletion allele (Table 4). Heterozygous RyR point mutant manipulations did not further enhance lethality in a statistically significant way, but they did not ameliorate the lethality phenotype either (Table 4). No manipulation examined resulted in significantly higher male lethality, compared to control males (Table 4). These results highlight the fact that molecular manipulations can have a salubrious effect in one context (synapse excitability) and an exacerbating effect in another (viability). This is true in our Drosophila system but possibly in other systems as well.

Discussion
We generated fruit flies designed to mimic the effects of FHM1-inducing Ca V 2.1 channel mutants, R192Q and S218L. Flies expressing the SL and RQ,SL transgenes for Drosophila Ca V 2/Cacophony displayed overt phenotypes, including reduced viability (Fig 1). They also displayed synaptic phenotypes, including enhanced evoked excitability (Fig 4), stark increases in quantal size and frequency (Fig 5), giant, spontaneous, sodium channel-dependent events (Figs 5 and 6), and enhanced probability of release at very low calcium (Fig 7). All of these neurotransmission phenotypes occurred without major alterations in active zone localization or overall synaptic architecture (Figs 2 and 3). RQ-expressing NMJs had only a mild phenotype: EPSP discharges with extended, shoulder-like waveforms (Fig 4). Genetic knockdown of Drosophila PLCβ or genetic mutations affecting the receptors that gate intracellular calcium stores (IP 3 receptor and Ryanodine receptor) partially alleviated some of the electrophysiological phenotypes (Fig 8), as did pharmacological manipulations targeting the same processes ( Fig  9). These results suggest that intracellular Ca 2+ signaling through IP 3 receptors and Ryanodine receptors could influence physiological dysfunction in a gain-of-function Ca V 2 background ( Fig 9D). Additionally, given the ability of TTX to block gigantic spontaneous events-and given our ability to quiet that phenotype through genetic and pharmacological meansimpairment of the IP 3 Receptor/Ryanodine Receptor pathway may limit spontaneous neuronal firing by as-yet undetermined mechanisms (Fig 9D).

Similarities between fly mutations and FHM1-causing human mutations
Evoked neurotransmission. Our discovery that SL-and RQ,SL-expressing Drosophila NMJs displayed increased evoked excitation, especially at low [Ca 2+ ] e (Figs 4 and 7) [20], was consistent with findings from diaphragm NMJs in SL knock-in mice [11]. In that context, the endplate potential (EPP) amplitudes were significantly increased at low levels of calcium (0.2 mM) but did not differ from those at wild-type NMJs at physiological calcium (2 mM) [11]. Interestingly, at the SL knock-in calyx of Held, excitatory postsynaptic currents (EPSCs) were increased, but this effect was most pronounced at high levels of [Ca 2+ ] e [12]. The EPSP discharges caused by expression of the SL-containing transgenic constructs in flies (Fig 4) were reminiscent of the EPP broadening at SL knock-in NMJs [11]. Finally, the severity of the dysfunction in the Drosophila NMJ waveform in the context of decreased extracellular magnesium (6 mM) (Figs 4 and 8) was consistent with a marked increase in calcium current in response to long action potential waveforms in calyces of Held expressing the RQ or SL mutant protein [12,14].
Enhanced quantal frequency. The enhanced mEPSP frequency at SL-and RQ,SLexpressing Drosophila NMJs (Fig 5, Table 1) was reminiscent of observations in prior FHM1 studies. In the RQ and SL knock-in mice, the NMJs exhibited significant increases in the frequency of mEPPs [9][10][11]. In principle, this spontaneous activity could correlate with a buildup of intracellular calcium or a change in intracellular calcium dynamics. In support of this view, at the calyx of Held in SL knock-in mice the frequency of spontaneous mEPSCs was enhanced and resting [Ca 2+ ] i was elevated [12]. In that case, the increase in quantal frequency was partially reversed by adding the cell-permeable calcium chelator EGTA-AM [12]. Evidence from several model synapses suggests that Ca V 2 channels can play a prominent role in spontaneous release. In granule cells of the hippocampus, stochastic activity of Ca V 2.2 (N-type) channels potentiates spontaneous miniature events, and the application of either BAPTA-AM or EGTA-AM is sufficient to inhibit them [72]. Other studies have demonstrated that P/Q-, N-, and R-type calcium channels also promote spontaneous release [73]. Notably, the differences in the spontaneous miniature phenotype between mice harboring the SL and RQ knock-in substitutions, or fruit flies expressing mimicking substitutions, suggest that the differences in cellular outcomes occur downstream of the Ca v 2 channel. This highlights a need for genetic approaches to uncover pathways that might contribute to the divergent phenotypes, as well those that are shared.

Differences between fly mutations and FHM1-causing human mutations
Quantal amplitudes. FHM1 mutations have been shown to enhance spontaneous miniature quantal release frequency in other systems [8][9][10][11], but there has been no report of increases in spontaneous miniature quantal size due to these mutations. In theory, an increase in the amplitude of mEPSP events at the Drosophila NMJ could be explained by an alteration to the expression and localization of postsynaptic proteins. Yet immunostaining of postsynaptic markers showed only a slight increase in postsynaptic glutamate receptor clustering ( Fig  3G). Instead, a combination of quantal analyses (Figs 5-7) points to alterations to the nature of spontaneous, presynaptic vesicle release-namely, that a certain percentage of quantal events in SL-and RQ,SL-expressing NMJs are multi-vesicular.
Why do SL-and RQ,SL-expressing NMJs in Drosophila show spontaneous multi-vesicular release? The synaptic preparation examined is likely critical. Evidence from other systems has demonstrated that calcium channel activity can have a profound effect on quantal size. For example, work at the C. elegans NMJ has demonstrated that calcium from intracellular and extracellular sources combines to dictate quantal size and frequency [56]. Additionally, spontaneous miniature events with large amplitudes ("maximinis") have been documented at fast inhibitory synapses of the cerebellum [55,74]. Similar to the NMJ activity documented in our study, these maximinis rely on the ability of ryanodine-sensitive stores to support spontaneous calcium transients large enough to cause multi-vesicular release. It is possible that the architecture of a giant synapse like the Drosophila NMJ-which contains hundreds of active zones clustered into individual boutons and has a low level of spontaneous, multi-vesicular release [75]-makes it exquisitely sensitive to small changes in intracellular calcium from both extracellular and store sources.
Evoked waveforms. As is the case for the Drosophila NMJ EPSPs (Fig 4), the diaphragm NMJ of FHM1 knock-in mice displayed EPP broadening [11]. However, the extra discharges we found at the Drosophila NMJ do not seem to be documented for the mammalian NMJ. An instructive parallel may be drawn between our data and cultures of Drosophila giant neurons, in which manipulation of the voltage-gated potassium current generated altered waveforms, including extra and extended discharges [76,77]. It is possible that some aspects of the FHM1 phenotypes may be caused by the perturbation of other voltage-activated currents, and by synapse excitability more generally [78]. This possibility is consistent with the fact that mutations in the Na + /K + ATPase gene also cause a form of pure FHM [79]. Given the effectiveness of the Drosophila system for uncovering complex relationships amongst ion channel activities, in particular potassium currents, the fly may be a good model for studying the cellular bases of disorders such as FHM1 [80][81][82][83][84].

FHM and non-FHM Migraine: Treatments
Our data suggest that a fly model could uncover molecules that could be targeted to mitigate effects of gain-of-function calcium channel activity associated with migraine. A novel and intriguing finding of our study is factors controlling intracellular calcium store release can be targeted to mitigate FHM1-like hyperexcitability (Figs 8 and 9). Indeed, the RyR channel blocker dantrolene has established uses in the clinic [85,86]. Moreover, significant evidence indicates that blockade of RyR by dantrolene could have neuroprotective applications [87]. In the context of FHM1, store operated calcium release would be a novel pathway to consider. Furthermore, lithium (Fig 9) has been employed in treating migraine, but only in limited cases.
One caveat to our findings is that impairing intracellular calcium release signaling pathways did not reverse all phenotypes associated with SL-and RQ,SL-expressing NMJs. For instance, in the case of increased mEPSP frequency, there is no significant reversal (Table 3). Another caveat is our fly lethality data (Table 4), which suggest that the excitability of single nerve terminals or circuits is not be the only factor to consider. In the case of our fly model, global impairment of factors controlling calcium store release dampened hyperexcitability but enhanced lethality (Table 4). These results point to the fact that gain-of-function Ca V 2 substitutions may cause multiple, separable, pleiotropic effects. It is possible that the neuronal hyperexcitability phenotypes are somehow protective for fruit fly viability or health-or are a reflection of a protective process that gets blunted when PLCβ and RyR are diminished. Similar considerations could be important in the context of any human migraine treatment.
There is no single, gold-standard pharmaceutical treatment for forms of hemiplegic migraine [88]. Several treatments have been employed in clinical settings [89,90], each with serious drawbacks. Some agents employed to treat hemiplegic migraine include calcium channel blockers like flunarizine [91,92] and verapamil (Ca V 1-blocking and potentially Ca V 2-blocking at higher doses) [93,94]. Blocking of voltage-gated calcium channels would seem to be an intuitive way to counter gain-of-function CACNA1A mutations; yet there would be obvious side effects of interfering with Ca V 2.1 function globally. Other agents reported to be effective in treating hemiplegic migraine are lamotrigine (targeting Na V and Ca V 2 channels), sodium valproate (several targets including Ca V 3 channels, resulting in increased inhibitory signaling), and acetazolamide (a pH modulator via carbonic anhydrase inhibition) (see for detailed review [89,95]). Finally, in cases where hemiplegic migraine attacks are frequent, prophylactic use of triptans has been employed [88,89]. Triptans are a standard treatment for generalized migraine attacks, but since they are vasoconstrictors, there has been some thought that they may not be appropriate for hemiplegic migraine.
Why might a new model be useful, specifically for FHM? FHM is unlike other chronic migraine conditions due its underpinning in central and cortical hyperexcitability and susceptibility to cortical spreading depression [96]. In recent years, calcitonin gene-related peptide (CGRP)-based and peripheral approaches have been the focus of generalized migraine treatment. From recent work in mice, there is evidence that CGRP induces migraine-reminiscent photophobia both peripherally and centrally [97]. Yet it is uncertain whether CGRP-based therapies would be effective for FHM. For one consideration, CGRP injections do not induce migraine in individuals with FHM in the same manner that it does for other chronic migraineurs sensitive to CGRP levels [98,99]. Recent clinical trials support the use of anti-CGRP receptor antibodies for migraine prophylaxis [100][101][102], and the Food and Drug Administration (FDA) of the United States has recently approved the anti-CGRP receptor antibody drug Erenumab as a therapeutic [103]. Yet the supporting studies did not use individuals with a history of hemiplegic migraine and the antibodies likely act by peripheral mechanisms because they cannot readily cross the blood/brain barrier [104][105][106]. Finally, triptan-based treatments act via reduction in CGRP release and act peripherally where they reverse the effects of CGRP on vasculature [107][108][109]. Given these facts, a new model by which to screen for pharmacogenetic targets of FHM-causing mutations-such as use of coarse phenotypes of electrophysiological phenotypes in flies-may be valuable.

Limitations and future directions
One strength of Drosophila is the power of genetic manipulation. The genetic toolkit afforded to Drosophila neuroscience makes the NMJ a useful model synapse. One caution regarding the model we generated for this study is that it utilizes over-expression of wild-type or mutant UAS-cacophony transgenes. The wild-type version of this transgene recapitulates wild-type cac function without generating hyper-excitability phenotypes [22,28], and we also controlled for potential overexpression phenotypes (Fig 2). Nevertheless, downstream analyses can be obscured by the need to separate overexpression (hypermorphic) gain-of-function versus mutant (neomorphic) gain-of-function analyses. Other methods, such as CRISPR-based knock-in mutations or expression of a genomic cac construct (as employed in [110]) could yield expression levels more similar to endogenous cac. Although mutations in the endogenous cac locus would be advantageous, we do not expect that these particular limitations detract from our core findings.
We have shown that genetic or pharmacological impairment of an intracellular calcium release signaling pathway suppresses some gain-of-function Ca V 2 electrophysiological phenotypes. Yet the precise mechanism and sequence of events underlying hyperexcitability suppression we observe are unclear. Potentiation of the baseline activities of the IP 3 R and RyR channels by mutant Ca V 2 channels is one possibility [111][112][113][114]. In principle, potentiated RyR or IP 3 R activity could feedback to and further potentiate Ca V 2 channels. Another possibility is that these gain of function mutations result in chronically increased of intracellular [Ca 2+ ] (as in [12]), which could then be reversed indirectly by targeting store pathways. Yet another possibility is that impairment of Ca 2+ store-release mechanisms somehow dampens Ca V 2 gating functions-effectively reversing gating gains of function that result from FHM1-causing mutations. Many future directions are possible, utilizing reagents that exist for Drosophila work. A mechanistic refinement could be aided by visual data-for instance by combining inhibition of Ca 2+ store release along with visual analyses of action potential waveforms via voltage imaging [115] and measurements of Ca V 2-mediated influx of Ca 2+ via genetically-encoded indicators tethered to active zone sites [116]-and separately by examining Ca 2+ dynamics at the stores themselves by using an ER-localizing, genetically encoded calcium sensor developed for Drosophila [117].
The implication of PLCβ activity and intracellular calcium in hyperexcitability is novel within the context of FHM1 mutations, but in hindsight, it also fits with results of prior studies. One recent RNA profiling analysis of the cerebellum of SL knock-in mice revealed an overrepresentation of several signaling components, including PLCβ [118]. Moreover, PLCβ and the release of calcium from intracellular stores have been implicated in signaling by CGRP [119][120][121], whose levels are correlated with generalized migraine [122][123][124]. Beyond work in the Drosophila model, further investigation will be needed to establish whether there is actually a causative link between the action of intracellular calcium stores either in inducing migraine or in precipitating neurological events that precede some forms of migraine, like aura and cortical spreading depression.

Gain-of-function cacophony constructs
To generate UAS-cac-eGFP SL transgenes, we used PCR to alter the serine 161 codon to leucine in the pUAST-based UAS-cac-eGFP DNA construct [20,22]. This substitution corresponds to S218L in mammalian CACNA1A. To generate UAS-cac-eGFP RQ transgenes, we used PCR to change the arginine 135 codon to glutamine. This substitution corresponds to R192Q in mammalian CACNA1A. For the UAS-cac-eGFP RQ,SL transgene, both mutations were incorporated into the same UAS-cac-eGFP construct using PCR to link the overlapping RQ and SL fragments. Transgenic lines were generated by injection of UAS-cac-eGFP constructs into a w 1118 background (The Best Gene, Chino Hills, CA) and mapped and backcrossed.
Electrophysiological data were collected using Axopatch 200B or Axoclamp 900A amplifiers (Molecular Devices, Sunnyvale, CA). Sharp electrode (> 10 MO) recordings were taken from muscle 6 of abdominal segments 2 and 3, as described previously [30,31,134]. Prior to muscle V m measurements, the Axoclamp 900A was bridge balanced. For the Axopatch 200B, the amplifier was placed in bridge mode (using I-CLAMP FAST for sharp electrode recordings). Before recording from each muscle, electrode resistance was measured and properly compensated by applying a step input and adjusting series resistance. Muscles with a V m more hyperpolarized than -60 mV and an input resistance of greater than 5 MO were deemed suitable for recording [30]. Data were digitized using a Digidata 1440A data acquisition system (Molecular Devices) and recorded using the pCLAMP 10 acquisition software (Molecular Devices). Spontaneous activity was recorded, followed by evoked activity. For presynaptic nerve stimulation, a Master-8 pulse stimulator (A.M.P. Instruments, Jerusalem, Israel) and an ISO-Flex isolation unit (A.M.P. Instruments) were utilized to deliver suprathreshold stimuli (1 ms unless otherwise indicated) to the appropriate segmental nerve. For each NMJ, the average amplitude of spontaneous miniature excitatory postsynaptic potential EPSPs (mEPSPs) was quantified by measuring approximately 100-200 individual spontaneous release events per NMJ. The average per-NMJ mEPSP amplitudes were then averaged for each genotype. Evoked EPSP amplitude was calculated for each NMJ as the average of 30 events (1 Hz). Quantal content (QC) was determined in two different ways. At very low extracellular [Ca 2+ ], QC was calculated by the method of failures, as m = ln[(# trials)/(# failures)], as described elsewhere [42]. At higher extracellular [Ca 2+ ], QC was calculated by dividing EPSP/mEPSP, as described in the text. For analyses conducted across different calcium concentrations, QC was corrected for non-linear summation [135]. For histograms displaying mEPSP amplitude frequencies, the same number of spontaneous events was analyzed for each NMJ (per genotype or experimental condition). This ensured that no individual NMJs were overrepresented or underrepresented in the aggregate analyses.
For analysis of fluorescence intensity and area, experimental and control larval preparations were stained in the same container, mounted on the same slide, imaged using identical acquisition settings, and analyzed using the same procedure and thresholds. Bouton and glutamate receptor cluster numbers were quantified semi-automatically using the 'Spots' function in Imaris x64 v7.6.0 (Bitplane, Zurich Switzerland). Any errors in automated counting were corrected by hand to arrive at the final value. GluRIIA and Dlg levels were assessed using ImageJ 1.48s/Java 1.6.0_24 (64-bit) with Fiji plugins. Z-stack images were compressed using the maximum projection function; ROIs were hand drawn to exclude non-synaptic structures; a minimum threshold was set for each channel to eliminate background fluorescence; and the Measure function was used to assess fluorescence intensity and area.

Western blotting
10 adult fly heads/sample were prepared in sample buffer using standard methods. SDS-PAGE was performed using the Novex NuPAGE SDS-PAGE system with 4%-12% Bis-Tris gels run at 125 V for 10 minutes and 150 V for 2.5 hours. Transfer to PVDF membrane (Bio-Rad, Hercules, CA) was performed using a Trans-Blot-SDSemi-Dry Transfer Cell (Bio-Rad, Hercules, CA). Blocking was performed in 5% BSA for GFP blots or 5% milk for actin blots in 1X PBS with 0.1% Tween 20. Primary antibodies were obtained from the DSHB, mouse anti-actin (JLA20) 1:1000, or from Torrey Pines Biolabs, rabbit anti-GFP 1:2000. Horseradish peroxidase-conjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) was used at 1:5000 for actin blots. Horseradish peroxidaseconjugated goat anti-rabbit secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) was used at 1:5000 for GFP blots. All antibodies were diluted in blocking buffer. Blots were developed with Super-Signal West Pico Chemiluminescent Substrate (Thermo Scientific, Waltham, MA) and imaged with Amersham Hyperfilm ECL film (GE Healthcare Limited, Buckinghamshire, UK). Band intensity was quantified using ImageJ.

Statistical analyses and data plots
Most electrophysiological comparisons were made across multiple data sets. As appropriate, statistical significance was either assessed by one-way ANOVA with Tukey's post-hoc analysis for multiple comparisons (assumes Gaussian distribution), or a non-parametric Kruskal-Wallis ANOVA with Dunn's post-hoc analysis for multiple comparisons (does not assume Gaussian distribution). Other statistical tests utilized included Fisher's exact tests for viability counts and for counts of gigantic mEPSP events; Log-rank tests for survivability curves; linear regression analyses for calcium cooperativity; and Student's T-Tests for direct comparisons between one control group and one experimental group. p values of Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001, and ÃÃÃÃ p < 0.0001 were considered significant. The values reported or plotted on regular bar graphs are mean ± SEM. The values reported and plotted on box-and-whisker graphs are: box (25 th -75 th percentiles), whiskers (1 st -99 th percentiles), line (median), + symbol (average), and individual raw data points plotted outside the 1 st and 99 th percentiles. Statistical analyses were performed in GraphPad Prism (GraphPad Software).