The w¹¹¹⁸ strain was used as a wild-type (wt) control, unless otherwise noted. Drosophila alleles used were raised at 25°C, unless otherwise noted in Figure Legends, and all flies were raised on standard molasses food. The following Drosophila stocks were used: GluRIIA^sp16, Elav^C155-Gal4 (BDSC BL458), and OK371-Gal4 (BDSC BL26160). The α2δ-3^k10814 and α2δ-3¹⁰⁶ alleles were generously provided by Dr. Yuh-Nung Jan (University of California, San Francisco; Howard Hughes Medical Institute) and Cac^sfGFP was a generous gift from Dr. Kate O’Connor-Giles (Brown University). Flies bearing UAS-stj-3HA (F001252) were obtained from FlyORF (University of Zurich).

The α2δ-3 coding sequence was cloned from UAS-stj-3HA (FlyORF F001252) transgenic flies and inserted into the pENTR-D-TOPO vector using the pENTR Directional TOPO Cloning Kit (Invitrogen K240020). Mutations were introduced in the MIDAS motif (DGSGS to AGAGA, UAS-α2δ-3^DSS-AAA) or RLR site (RLR to AAA, UAS-α2δ-3^RLR-AAA) of α2δ-3 using the Q5 site-directed mutagenesis kit (NEB E0554S). For generation of final constructs, pENTR vectors were recombined with destination vector pUASg-HA_attB (DGRC 1423) using LR clonase II enzyme (Invitrogen 11791020). Once final constructs were generated, they were sent to BestGene Inc. (Chino Hill, CA) for injection, using ZH-86Fb (III) as the targeted insertion strain. UAS-α2δ-3^DSS-AAA-3HA and UAS-α2δ-3^DSS-AAA-3HA transgenic lines were validated through PCR and sequencing, and these lines were subsequently used in experiments.

The protein sequences used for alignment were as follows: human CA2D1 (P54289), CA2D2 (Q9NY47), CA2D3 (Q8IZS8), mouse CA2D3 (Q9Z1L5), rat CA2D3 (Q8CFG5), and Drosophila CA2D3 (Q7K0H4, isoform B). The alignment was performed using CLUSTAL O (1.2.4). To generate predicted structures for Drosophila wild-type CA2D3 and CA2D3 with DSS-AAA or RLR-AAA mutations, we utilized ColabFold (Mirdita et al., 2022), which combines the fast homology search of MMseqs2 (Many-against-Many sequence searching) with AlphaFold2 (Jumper et al., 2021). For alignment in Figure 1E, we used the cryo-EM structure of human CA2D1 (pdb8FD7; Chen et al., 2023a) and wild-type Drosophila CA2D3 structure predicted by AlphaFold2. The alignment and presentation of wild-type and mutated Drosophila CA2D3 structures were performed using RCSB PDB in Figures 2A,B. The protein structures of wild-type and mutated Drosophila CA2D3, custom Python codes for AlphaFold protein structure presentations, and AlphaFold pLDDT statistics are available at: https://github.com/wanglab-georgetown/alphafold_a2d3.

Sharp-electrode recordings were made from muscle 6 at abdominal segments 2 and 3 of male and female 3^rd instar larvae using an Axoclamp 900A amplifier (Molecular Devices) as previously described (Frank et al., 2006). For α2δ-3 rescue experiments using Elav^C155-Gal4 (located on the X chromosome), only male larvae were included in recordings. HL3 saline solution containing the following concentrations (in mM) was used for all current clamp recordings: 70 NaCl, 5 KCl, 10 MgCl₂, 10 NaHCO₃, 115 Sucrose, 5 Trehalose, 5 HEPES, and 0.3 CaCl₂. EPSP and mEPSP traces were analyzed using StimFit (0.15.8) and MiniAnalysis (6.0.3 Synaptosoft). To rapidly induce PHP, larvae were incubated in 20 μM Philanthotoxin-433 (PhTX, Santa Cruz Biotechnology 276684-27-6 or AOBIOUS AOB0876) in an un-stretched, partially dissected preparation for 10 min, following the method described in Frank et al. (2006). For each NMJ, the average amplitudes of evoked EPSPs are based on the mean peak amplitudes in response to 20–30 individual stimuli. mEPSPs were recorded continuously for 60–90 s. Quantal content was estimated for each NMJ as the ratio of EPSP amplitude/mEPSP amplitude. The mean value across all NMJ for a given genotype is reported.

Immunohistochemistry (IHC) was conducted using standard protocols as previously described (Wang et al., 2014, 2020). Briefly, dissected third instar larvae were fixed in ice-cold ethanol for 5 min on ice, followed by rinses in PBST (PBS with 0.01% Triton X-100) for 6 times, 10 min each at room temperature. The preparations were then incubated in blocking solution (PBST with 5% normal goat serum) for 2 h at room temperature. Rabbit anti-GFP (1:200, Invitrogen G10362) and mouse anti-Bruchpilot (Brp) antibodies (1:200, DSHB nc82) were used for overnight incubation at 4°C with rotation. For confocal imaging, after the primary antibody incubation, the preparations were rinsed for 6 times, 10 min each in PBST. They were then incubated with Alexa 488-conjugated goat anti-rabbit (1:300, Invitrogen A-11008), Alexa Cy3-conjugated goat anti-mouse (1:300, Invitrogen A-10521), and Alexa 647-conjugated goat anti-HRP (1:100, Jackson Immuno Research Laboratories 123-605-021) antibodies. The preparations were mounted with Fisher Brand #1 coverslips using VECTASHIELD antifade mounting medium (H-1000-10). For STED imaging, after the primary antibody incubation, the preparations were incubated with Alexa 594-conjugated goat anti-mouse (1:300, Invitrogen R37121) and ATTO 647 N-conjugated goat anti-rabbit antibodies (1:500, Rockland 611-156-122S) in PBST for 2 h at room temperature. After 6 times, 10 min rinses in PBST, the preparations were mounted with Fisher Brand #1.5 coverslips using Prolong Gold mounting medium (Invitrogen P36930). In experiments to examine calcium channel abundance during acute PHP, a 10 min incubation with 20 μM Philanthotoxin-433 (PhTX, AOBIOUS AOB0876) was performed to induce PHP prior to fixation. For each IHC experiment, the consistency of immunolabeling is maintained by dissecting, fixing and staining preparations of all genotypes simultaneously. This ensures that all preparations were subjected to identical experimental conditions, such as antibody incubation, thereby minimizing variability.

Confocal images of the NMJ were acquired using a laser scanning confocal microscope (LSM 880, Carl Zeiss). Z-stacks of the NMJ on muscle 6/7 segments A2 and A3 were captured using a 63x objective (Plan-Apochromat 63x/1.40 Oil DIC M27), and maximum projections were used for analysis in Fiji (NIH). Fluorescence intensities in each channel of the confocal images spanned a dynamic range without reaching saturation. STED images of single 1b boutons (three most distal boutons per synapse) were obtained from the NMJ on muscle 6/7 segment A2 using a Nikon Eclipse Ti2-E confocal microscope equipped with an Abberior STEDYCON unit and a Nikon Plan Apo 100× NA1.45 Lambda Oil DIC N2 objective. The Alexa 594 and ATTO 647 N channels were scanned using 561 nm (excitation)/775 nm (STED) and 640 nm (excitation)/775 nm (STED) lasers, respectively. This setup provides a resolution of 40-60 nm on the X/Y plane. Each Z-stack consisted of 4 sections with a 0.5 μm step size to minimize bleaching. All images were captured in 16-bit format with a size of 8 μm × 8 μm. To maintain consistency of confocal and STED imaging, samples from each batch of IHC experiments were imaged using identical settings, and all within the same day. Maximum projections were used for analysis in Fiji (NIH) and presentation in Figures.

All imaging analyses were executed on unsaturated raw data. For confocal data analysis, binary masks were generated by applying a threshold to the Brp (cy3) channel, which were then transferred to the Cac (488) channel to quantify the mean and integrated intensity of Cac at individual active zones (inside Brp puncta) for each synapse. For STED data analysis, Z-stacks of boutons were processed using Fiji (NIH). After splitting the channels and creating maximum-intensity projections, a threshold was applied to each channel to isolate the active zones (Alexa 594 channel) or Cac (ATTO 647 N channel). The threshold for each channel remained consistent for all images collected in one experimental repeat. Particle analysis was performed to quantify intensity and area. The mean intensity (average fluorescence), integrated intensity (sum fluorescence), and area of each particle were then measured and reported for each bouton (STED).

The data are presented as Mean ± Standard Error of the Mean (SEM), with the precise sample sizes indicated in the Figure Legends. Statistical analysis was performed using Prism (9.5.1, GraphPad). First, we assessed the normality of residuals from ANOVA tests across all datasets using the D’Agostino-Pearson test. Then, for datasets with normal residuals, we used Brown-Forsythe and Welch ANOVA (non-equal variance) tests. For datasets with non-normal ANOVA residuals, the nonparametric Kruskal-Wallis test was applied. In case of comparing more than two conditions, we used the original False Discovery Rate (FDR) method of Benjamini and Hochberg to correct for multiple comparisons. The q-values obtained from the FDR correction are reported in the Figure Legends.

The α2δ proteins are located at the extracellular face of presynaptic release sites and interact directly with the α1 subunit of VGCCs (Figure 1A). In mammalian α2δ-1 and α2δ-2, the vWA domain and the arginine-arginine-arginine (RRR) gabapentin-binding site are essential for calcium channel trafficking (Canti et al., 2005; Field et al., 2006; Hoppa et al., 2012). To investigate the domains involved in regulating calcium channel trafficking in α2δ-3, we aligned the protein sequences and identified conserved regions (Figures 1B,C). The MIDAS motif within the vWA domain of α2δ-1 and α2δ-2 is crucial for proper calcium channel distribution and synaptic transmission (Hoppa et al., 2012). We observed that the MIDAS motif (DxSxS) in α2δ proteins is highly conserved across different species (Figure 1C). Furthermore, gabapentin and pregabalin bind directly to the RRR site in α2δ-1 and α2δ-2, leading to a decrease in surface expression of calcium channels (Field et al., 2006; Hendrich et al., 2008). In α2δ-3, the first and third arginine residues are conserved, but instead of an RRR site like α2δ-1 and α2δ-2, mammalian α2δ-3 has an arginine-asparagine-arginine (RNR) sequence, while Drosophila α2δ-3 has an arginine-leucine-arginine (RLR) sequence (Figure 1C). Based on previous functional studies and the cryo-EM structure of mammalian α2δ-1 (Chen et al., 2023b), we hypothesize that the RLR site in Drosophila α2δ-3 may be involved in the regulation of calcium channel trafficking.

The structure of Drosophila α2δ-3 protein has not been experimentally determined. Therefore, we used the AlphaFold algorithm (Jumper et al., 2021; Mirdita et al., 2022) to predict the structure of Drosophila α2δ-3. The predicted structure of Drosophila α2δ-3, particularly the protein domains containing the MIDAS motif and RLR site, showed high confidence according to the pLDDT score (Figure 1D; Tunyasuvunakool et al., 2021). To validate the prediction, we compared the predicted structure of Drosophila α2δ-3 with the cryo-EM structure of human α2δ-1 (Ca_v1.2/Ca_vβ/Ca_vα2δ-1; Chen et al., 2023a). The alignment between the two structures yielded a template modeling-score (TM-score; Xu and Zhang, 2010) of 0.67 and a root mean square deviation (RMSD; Kufareva and Abagyan, 2012) of 3.05 Å. These results indicate that Drosophila α2δ-3 and human α2δ-1 exhibit similar structural characteristics, thereby confirming the accuracy of the AlphaFold prediction (Figure 1E).

To investigate the functional role of α2δ-3 in regulating calcium channel trafficking during PHP, we generated two mutant forms of α2δ-3 by introducing specific mutations in either the vWA domain or the RLR site. For the vWA domain mutation, we replaced three conserved metal coordinating residues within the MIDAS motif (DGSGS) with alanine residues (AGAGA) in the UAS-α2δ-3^wt transgene, resulting in the UAS-α2δ-3^DSS-AAA mutant (Figures 1B,C, 2A). Additionally, we generated a transgenic line carrying mutations in the RLR site, where the RLR sequence was mutated to AAA, yielding the UAS-α2δ-3^RLR-AAA mutant (Figures 1B,C, 2B).

Using AlphaFold, we predicted the structural impact of these mutations in the MIDAS motif and RLR site. The predicted structures of both the wild-type and mutated forms of α2δ-3 showed high confidence for the MIDAS motif and RLR site (Figure 2C). In the DSS-AAA mutant, the negatively charged or polar residues of the MIDAS motif were replaced by alanine residues, which have a shorter carbon backbone and smaller hydrophobic side chains. Similarly, in the RLR-AAA mutant, the positively charged residues of the RLR site were substituted with alanine residues. These mutations may alter the local conformation of the vWA domain and RLR site, thereby disrupting potential interactions with binding partners in the binding pockets (Figures 2A,B). While the exact impact of these mutations requires experimental validation, the overall Alphafold-predicted structure of the α2δ-3 protein was not significantly affected by these mutations, as indicated by the RMSD scores of 0.84Å and 1.45Å when comparing the wild-type α2δ-3 to the DSS-AAA and RLR-AAA mutants, respectively.

To assess the effect of mutations in the MIDAS motif and RLR site of α2δ-3 on PHP, we induced PHP by applying the glutamate receptor antagonist Philanthotoxin-433 (PhTX, 20 μM; Frank et al., 2006) and performed electrophysiological recordings at the neuromuscular junction (NMJ). In all genotypes, including the wild-type, we observed a reduction of approximately 50% in the average amplitude of miniature excitatory postsynaptic potentials (mEPSPs) upon PhTX application (Figures 2D–F). In the wild-type condition, the pharmacological inhibition of postsynaptic glutamate receptors resulted in a significant increase in presynaptic neurotransmitter release, indicated by an increase in quantal content (QC), which compensated for the perturbation and restored the excitatory postsynaptic potential (EPSP) amplitude back to the normal baseline level (Figures 2D,E,G,H). However, in the α2δ-3 homozygous mutant (α2δ-3^k10814/α2δ-3¹⁰⁶), when PHP was induced with PhTX, we observed no change in quantal content, and the EPSP amplitude was dramatically reduced, indicating a complete loss of PHP (Figures 2D,E,G,H). These findings are consistent with our previous observations that α2δ-3 is necessary for the induction of PHP (Wang et al., 2016).

In order to investigate the functional effects of mutations in the MIDAS motif and RLR site of α2δ-3 on PHP, we conducted tissue-specific rescue experiments. Specifically, we selectively overexpressed wild-type α2δ-3 (UAS-α2δ-3^wt) or α2δ-3 with mutations in the MIDAS motif (UAS-α2δ-3^DSS-AAA) or RLR site (UAS-α2δ-3^RLR-AAA) in neurons using the Elav^C155-Gal4 driver in the background of the α2δ-3 homozygous mutant (genotypes as indicated in the Figure Legends; Figures 2D–H). Consistent with previous findings, neuronal overexpression of wild-type α2δ-3 (α2δ-3^wt rescue, Elav^C155-Gal4;α2δ-3^k10814/α2δ-3¹⁰⁶;UAS-α2δ-3^wt) restored the compensatory increase in quantal content in the α2δ-3 mutant background (Figures 2E,H). Although the EPSP amplitude was not fully restored to the baseline level, there was a significant increase in quantal content when we induced PHP using PhTX in the α2δ-3^wt rescue condition (Figures 2D,E,G,H). These results suggest that wild-type α2δ-3 expression in neurons rescues the loss of PHP.

Intriguingly, our findings from the neuronal overexpression of α2δ-3^DSS-AAA (α2δ-3^DSS-AAA rescue, Elav^C155-Gal4;α2δ-3^k10814/α2δ-3¹⁰⁶;UAS-α2δ-3^DSS-AAA) and α2δ-3^RLR-AAA (α2δ-3^RLR-AAA rescue, Elav^C155-Gal4;α2δ-3^k10814/α2δ-3¹⁰⁶;UAS-α2δ-3^RLR-AAA) in the α2δ-3 homozygous mutant background revealed distinct effects on PHP. Specifically, the overexpression of α2δ-3^DSS-AAA failed to rescue the compensatory increase in neurotransmitter release observed during PHP (Figures 2D–H). However, in contrast, the overexpression of α2δ-3^RLR-AAA fully restored PHP (Figures 2D–H). The EPSP amplitude remained significantly reduced in the α2δ-3^DSS-AAA rescue condition in the presence of PhTX (Figure 2G). On the other hand, overexpression of α2δ-3^RLR-AAA precisely restored the EPSP amplitude back to the baseline level after PhTX application (Figure 2G). These results provided compelling evidence that the MIDAS motif, but not the RLR site, in α2δ-3 is crucial for the rapid induction of PHP. The MIDAS motif appears to play a critical role in mediating the compensatory increase in neurotransmitter release during PHP, while the RLR site is not be directly involved in this process.

We examined the EPSP amplitude and quantal content in the absence of PhTX in the α2δ-3 homozygous mutant and various rescue conditions. We found that in the α2δ-3 homozygous mutant (α2δ-3^k10814/α2δ-3¹⁰⁶), there was a significant reduction in both EPSP amplitude and quantal content, suggesting that α2δ-3 is crucial for normal neurotransmitter release at baseline (Figures 2D,G,H). As expected, the neuronal-specific overexpression of α2δ-3^wt restored the EPSP amplitude and quantal content in the α2δ-3 mutant, bringing them back to wild-type levels (Figures 2D,G,H). However, in contrast, the neuronal overexpression of either α2δ-3^DSS-AAA or α2δ-3^RLR-AAA did not restore the EPSP amplitude or quantal content in the α2δ-3 mutant at baseline (Figures 2D,G,H). We observed a slight yet statistically significant increase in mEPSP amplitude in the α2δ-3^RLR-AAA rescue as compared to the wild-type control (Figure 2F). This observation suggests potential postsynaptic changes associated with the α2δ-3^RLR-AAA rescue. However, we noted that there was still a dramatic decrease in quantal content in the α2δ-3^RLR-AAA rescue condition compared to the wild-type control (Figure 2H). These results highlight that both the MIDAS motif and RLR site in α2δ-3 are necessary for normal presynaptic neurotransmitter release under basal conditions in the absence of PhTX.

In summary, our findings indicate that the MIDAS motif within the vWA domain and RLR site in α2δ-3 play distinct roles in controlling basal synaptic transmission and PHP. The MIDAS motif is essential for both processes, while the RLR site is specifically required for baseline neurotransmitter release and is not involved in PHP. Despite the absence of direct binding between mammalian α2δ-3 and gabapentin (Marais et al., 2001), our study provides evidence that the RLR site in Drosophila α2δ-3 is critical for normal synaptic transmission. We hypothesize that the RLR site interacts with endogenous binding partners to regulate the trafficking of VGCCs. These findings shed light on the functional significance of specific protein domains in α2δ-3 and their involvement in the regulation of synaptic plasticity.

Furthermore, we investigated the role of the vWA domain and RLR site in α2δ-3 in the long-term maintenance of PHP. PHP can be induced by genetic deletion of the postsynaptic glutamate receptor subunit, GluRIIA (Petersen et al., 1997). Since this is a genetic mutation during the lifetime of the animal, it has been used to assess the long-term maintenance of PHP. We generated double mutant animals with homozygous mutations in both GluRIIA and α2δ-3 genes. The GluRIIA mutation alone resulted in a significant reduction in average mEPSP amplitude, accompanied by a compensatory increase in presynaptic release, indicating the induction of PHP (Figures 3A–C,E). As a result, EPSP amplitude was restored toward baseline levels in the GluRIIA mutant (Figures 3A,D). In contrast, in the GluRIIA,α2δ-3 double mutant, we observed a dramatic decrease in mEPSP amplitude (q < 0.001), similar to the α2δ-3 single mutant, with no significant change in quantal content (q = 0.30), indicating a complete block of PHP (Figures 3A–C,E).

Interestingly, when UAS-α2δ-3^wt or UAS-α2δ-3^RLR-AAA is expressed presynaptically in neurons in the GluRIIA,α2δ-3 double mutant background, the quantal content is dramatically increased. This finding suggests that both the wild-type α2δ-3 and the RLR site-mutated α2δ-3 are capable of rescuing the long-term maintenance of PHP (Figures 3A,B,E). However, when we overexpressed UAS-α2δ-3^DSS-AAA in neurons in the GluRIIA,α2δ-3 double mutant background, there was no change in quantal content compared to the baseline value observed in the α2δ-3^DSS-AAA;α2δ-3 mutant alone (Figures 3A,B,E). These observations indicate that the MIDAS motif-mutated α2δ-3 is unable to rescue PHP. Therefore, we concluded that the MIDAS motif, but not the RLR site, in α2δ-3 is required for chronic PHP. In summary, the MIDAS motif within the vWA domain is necessary for acute PHP, chronic PHP, and synaptic transmission at baseline, whereas the RLR site is only required for synaptic transmission at baseline.

To further investigate the roles of the MIDAS motif and RLR site in regulating calcium channel localization at presynaptic zones during PHP, we employed confocal and STED super-resolution imaging techniques (Figure 4A). Initially, we aimed to determine whether the MIDAS motif and RLR site in α2δ-3 directly influence presynaptic calcium channel abundance at baseline, utilizing confocal imaging method. In Drosophila, the cacophony (cac) gene encodes the α1 subunit responsible for forming the pore of Ca_v2.1 calcium channels. To assess the synaptic expression of Cac, we used a CRISPR knock-in allele, wherein the endogenous Cac is GFP-tagged (Cac^sfGFP; Gratz et al., 2019).

To assess Cac abundance, we performed immunolabeling and confocal imaging experiments. Cac^sfGFP, along with the active zone component Bruchpilot (Brp; Kittel et al., 2006), was co-labeled, and the fluorescence intensity of Cac^sfGFP within Brp puncta was quantified. Our results revealed that synaptic Cac^sfGFP levels were reduced by approximately 50% in the α2δ-3 mutant compared to the wild-type (Figures 4B,C,G,H). Notably, motoneuron-specific overexpression of wild-type α2δ-3 (UAS-α2δ-3^wt) resulted in a substantial increase in Cac^sfGFP intensity and partially restored calcium channel abundance in the α2δ-3 mutant, approaching wild-type levels (Cac^sfGFP;α2δ-3,OK371-Gal4;UAS-α2δ-3^wt; Figures 4B,D,G,H). In contrast, when the MIDAS motif-mutated form of α2δ-3 (Cac^sfGFP;α2δ-3^k10814,OK371-Gal4/α2δ-3¹⁰⁶;UAS-α2δ-3^DSS-AAA) or the RLR site-mutated form of α2δ-3 (Cac^sfGFP;α2δ-3^k10814,OK371-Gal4/α2δ-3¹⁰⁶;UAS-α2δ-3^RLR-AAA) was expressed in motoneurons in the α2δ-3 mutant background, calcium channel abundance remained significantly reduced compared to the wild-type (Figures 4B,C,E−H). These findings indicate that the mutated forms of α2δ-3 expressed in neurons lack the ability to restore normal calcium channel distribution. Hence, the MIDAS motif and RLR site in α2δ-3 are both essential for maintaining proper calcium channel abundance at baseline.

Notably, we observed a mild but not significant reduction in the presynaptic active zone area (as indicated by Brp immunolabeling) in the α2δ-3 homozygous mutant compared to the wild-type (Figures 4B,C,I). This finding suggests that the observed impairment in calcium channel abundance in the α2δ-3 mutant is not a result of severe deficits in active zone organization. Furthermore, neuronal expression of UAS-α2δ-3^wt in the α2δ-3 mutant led to a mild reduction in active zone area, although the decrease was not statistically significant (Figure 4I). Therefore, the rescue of calcium channel abundance observed upon overexpression of UAS-α2δ-3^wt is unlikely to be due to the expansion of presynaptic active zones. Collectively, these findings suggest that the disruption of the MIDAS motif or RLR site in α2δ-3 significantly reduces presynaptic calcium channel abundance, reflecting the deficits in synaptic transmission observed under these conditions using electrophysiological methods.

Recent studies utilizing super-resolution imaging have indicated that the abundance and structural organization of presynaptic scaffolding proteins and calcium channels undergo modulation during the rapid induction and long-term maintenance of PHP (Bohme et al., 2019; Gratz et al., 2019; Ghelani et al., 2023). These findings highlight the importance of presynaptic protein trafficking and stabilization in homeostatic plasticity. To further investigate the changes in presynaptic calcium channel abundance during acute PHP, we employed the STED super-resolution imaging technique. With a resolution of 40–60 nm in the X/Y dimensions, STED imaging allowed us to examine the localization of calcium channels within presynaptic active zones (approximately 200 nm in diameter). Based on our electrophysiological data suggesting the requirement of the MIDAS motif in acute PHP, we focused on understanding the role of the MIDAS motif in controlling calcium channel distribution during PHP.

To investigate this, we performed immunolabeling of Cac^sfGFP and Brp in five different genotypes: wild-type, the α2δ-3 mutant, motoneuron overexpression of UAS-α2δ-3^wt in the α2δ-3 mutant (Cac^sfGFP; α2δ-3^k10814,OK371-Gal4/α2δ-3¹⁰⁶;UAS-α2δ-3^wt), overexpression of UAS-α2δ-3^DSS-AAA in the α2δ-3 mutant (Cac^sfGFP; α2δ-3^k10814,OK371-Gal4/α2δ-3¹⁰⁶;UAS-α2δ-3^DSS-AAA), and overexpression of UAS-α2δ-3^RLR-AAA in the α2δ-3 mutant (Cac^sfGFP; α2δ-3^k10814,OK371-Gal4/α2δ-3¹⁰⁶;UAS-α2δ-3^RLR-AAA). Immunolabeling was performed in the presence and absence of PhTX for each genotype. Upon inducing PHP by applying 20 μM PhTX for 10 min, we observed a significant increase in the mean intensity, integrated intensity, and area of Cac^sfGFP in the wild-type, indicating an upregulation of calcium channel abundance at presynaptic active zones during acute PHP (Figures 5A,F–H). However, this increase in calcium channel abundance during acute PHP was completely absent in the α2δ-3 homozygous mutant (Figures 5B,F–H). This deficit in the upregulation of presynaptic calcium channels correlates with the impairment of the compensatory increase in neurotransmitter release observed in the α2δ-3 mutant (Figures 2D,E,H).

We proceeded to investigate whether the trafficking of calcium channels is necessary for the observed increase in calcium channel abundance during acute PHP. Specifically, expressing UAS-α2δ-3^wt in motoneurons in the α2δ-3 mutant background led to a significant increase in the mean and integrated intensity and area of Cac^sfGFP upon PhTX application (Figures 5C,F–H). These findings clearly indicate the critical role of α2δ-3 in facilitating the compensatory increase in calcium channel abundance to the presynaptic membrane during acute PHP. In contrast, overexpression of UAS-α2δ-3^DSS-AAA specifically in motoneurons of the α2δ-3 mutant did not result in any changes in calcium channel abundance upon induction of acute PHP (Figures 5D,F–H). This observation suggests that the MIDAS motif in α2δ-3 is essential for the regulation of calcium channel localization to active zones during PHP. Interestingly, when we overexpressed UAS-α2δ-3^RLR-AAA specifically in motoneurons of the α2δ-3 mutant, we observed a significant increase in the mean and integrated intensity and area of Cac^sfGFP following PhTX application, suggesting a rescue of calcium channel localization during PHP in the α2δ-3^RLR-AAA condition (Figures 5E–H). This result indicates that the RLR site in α2δ-3 is not required for the upregulation of calcium channels during the rapid induction of PHP, consistent with our observation that the RLR site is not necessary for the increase of neurotransmitter release in PHP (Figures 2D,E,H). In summary, by using STED super-resolution imaging approach, we have provided evidence that α2δ-3 plays a crucial role in dynamically regulating the abundance of presynaptic calcium channels during acute PHP. Furthermore, we have found that the MIDAS motif within the vWA domain, but not the RLR site, is required for the regulation of calcium channel localization in the context of acute PHP. These observations align with the essential role of the MIDAS motif in regulating presynaptic neurotransmitter release in PHP (Figures 2D,E,H).

Consistent with confocal imaging data, STED super-resolution imaging experiments revealed a significant reduction in the intensities of Cac^sfGFP in the α2δ-3 mutant (Figures 5B,F–H). Furthermore, these metrics were restored to wild-type levels upon overexpression of UAS-α2δ-3^wt in motoneurons of the α2δ-3 mutant (Figures 5C,F–H). However, neither overexpressing UAS-α2δ-3^DSS-AAA or UAS-α2δ-3^RLR-AAA in motoneurons rescued the intensities of Cac^sfGFP in the α2δ-3 mutant (Figures 5D–H). These results further validate the role of α2δ-3 in regulating the abundance of presynaptic calcium channels under baseline conditions.

Finally, we aimed to determine if α2δ-3 is necessary for the compensatory increase observed in presynaptic calcium channel abundance during chronic PHP. To accomplish this, we conducted STED imaging at the NMJ and performed immunolabeling of Cac^sfGFP and Brp in various genetic backgrounds, including wild-type, GluRIIA, α2δ-3, and GluRIIA,α2δ-3 double mutant. Our findings revealed a significant increase in presynaptic calcium channel abundance in the GluRIIA mutant compared to the wild-type, as evidenced by the elevated Cac^sfGFP intensities (Figures 6A–C). However, the observed increase in Cac^sfGFP intensities was completely abolished in the GluRIIA,α2δ-3 double mutant compared to the α2δ-3 mutant alone, indicating the essential role of α2δ-3 in the compensatory increase of presynaptic calcium channels (Figures 6A–C). In line with previous findings, we also observed an elevation in the mean intensity of Brp in the GluRIIA mutant compared to the wild-type. However, this upregulation of Brp intensity was diminished in the GluRIIA,α2δ-3 double mutant when compared to the α2δ-3 mutant alone (Figure 6D). Collectively, these findings strongly suggest that α2δ-3 is crucial for both the upregulation of neurotransmitter release during chronic PHP and the regulation of presynaptic calcium channels in homeostatic plasticity.