Local Ca²⁺ signals couple activation of TRPV1 and ANO1 sensory ion channels

Single-cell fluorescence imaging can measure currents through CaCCs

Imaging of halide-sensitive enhanced yellow fluorescent protein (EYFP) mutants has been previously used to monitor Cl⁻ channel activity (32–34). The method is based on the permeability of most anion channels to iodide ions (I⁻), which, when added to the extracellular solution, quenches EYFP fluorescence in a manner dependent on anion channel opening. However, previous applications of this technique, including a single-cell approach developed by us (19), do not allow correlation of intracellular Ca²⁺ dynamics with I⁻ influx. Furthermore, the agonist-independent quenching complicates quantification and masks the response obtained by agonists. Here we developed a triple-wavelength imaging approach to simultaneously monitor halide-sensitive EYFP quenching and ratiometric fura-2 Ca²⁺ imaging (Fig. 1A). DRG neurons abundantly express GABA_A Cl⁻ channels (35–37); thus, we used GABA responses from cultured rat DRG neurons transfected with the halide-sensitive mutant EYFP H148Q/I152L (EYFP-QL) to optimize the extracellular I⁻ concentration and minimize agonist-independent quenching (Fig. 1B–E). Equimolar replacement of 5–30 mM NaCl with NaI revealed that 5 mM NaI offered the lowest rate of agonist-independent EYFP-QL quenching while providing sufficient dynamic range for the reliable detection of Cl⁻ channel opening.

We then performed simultaneous halide and Ca²⁺ imaging of DRG neurons transfected with EYFP-QL and loaded with fura-2AM. B₂R receptors were stimulated by bath-application of BK (250 nM) in the presence of 5 mM NaI while performing triple-wavelength excitation (340/380/488nm). Consistent with our previous study (19), BK induced robust transient [Ca²⁺]_i rises which were temporally correlated with the EYFP-QL fluorescence quenching (Fig. 2A, B). In contrast to the Ca²⁺ rises, which were transient in nature, the EYFP-QL quenching did not normally recover within the period of observation (reflecting that once I⁻ is taken up by a cell during CaCC activation, it is not easily extruded under our experimental conditions). The mean amplitude of the peak fura-2 Ca²⁺ signal (ΔR/R₀) and the mean drop in the normalized EYFP-QL fluorescence (ΔF/F₀) induced by BK are shown in Fig. 2C (left).

To confirm that this quenching was indeed mediated by ANO1 and not another Cl⁻ channel flux, transfected cells were preincubated and perfused with two ANO1 inhibitors, T16A-inhA01 and Ani9 (38). In the presence of T16A-inhA01 (50 μM), EYFP-QL quenching induced by BK was abolished despite the presence of Ca²⁺ transients (Fig. 2C). Ani9 (500 nM) also inhibited BK-induced quenching, although to a lesser extent. The BK-induced [Ca²⁺]_i rises were smaller in the presence of both ANO1 inhibitors, compared to control conditions (Fig. 2C). This observation is consistent with the hypothesis that ANO1 is not only a target of the Ca²⁺ signals, but may also facilitate ER Ca²⁺ release because of structural interactions between the PM and the ER (28). Hence, both T16A-inhA01 and Ani9 may interfere with this process. However, BK application still induced measurable rises in [Ca²⁺]_i in the presence of T16A-inhA01 and Ani9, whereas EYFP-QL quenching was either abolished (T16A-inhA01) or significantly reduced (Ani9). This indicates that our imaging approach allows reliable, all-optical measurement of ANO1-mediated CaCCs and concurrent Ca²⁺ dynamics in individual DRG neurons.

TRPV1-ANO1 coupling requires Ca²⁺ release from the ER

ANO1 activation is coupled to intracellular Ca²⁺ release from IP₃Rs in DRG neurons (19). To confirm this coupling using our imaging approach, BK was applied in the absence of Ca²⁺ in the extracellular bath solutions (Fig. 2D). The amplitude of EYFP-QL quenching was not affected by the removal of extracellular Ca²⁺ (Fig. 2E), despite a significantly lower induced [Ca²⁺] rise, compared to control conditions. This finding suggested that even though Ca²⁺ influx through PM Ca²⁺ channels contributed to the global Ca²⁺ rises induced by BK (8), it was the ER Ca²⁺ release that specifically activated ANO1/CaCC following B₂R activation. Indeed, as shown previously (19), Ca²⁺ influx through voltage-gated Ca²⁺ channels (VGCCs) is poorly-coupled to ANO1 activation in DRG neurons. To verify this finding using our imaging approach, we stimulated Ca²⁺ influx through VGCCs by depolarizing neurons with an extracellular solution containing 50 mM KCl (fig. S1A, B). The majority (15/23 or 65%) of DRG neurons did not respond to 50 mM K⁺ with any measurable EYFP-QL quenching, despite strong rises in [Ca²⁺]_i. The remaining 35% (8/23) neurons did display quenching (fig. S1A, B). However, significant EYFP-QL quenching induced by 50 mM KCl was observed in some DRG neurons in the absence of extracellular Ca²⁺ as well (fig. S1C, D). Because [Ca²⁺]_i rises were not recorded under these conditions, we conclude that depolarization-induced quenching observed in the minority of DRG neurons was likely not CaCC-related. In summary, our data (Fig. 2, A to E and fig. S1, A to D) established a reliable optical approach for measuring CaCC activation in DRG neurons and confirmed preferential coupling of ANO1/CaCC to IP₃R-mediated Ca²⁺ release from the ER rather than Ca²⁺ influx through PM Ca²⁺ channels in DRG neurons.

TRPV1 activation in DRG neurons and in heterologous expression systems results in activation of ANO1; furthermore, ANO1 potentiates TRPV1-induced depolarization in vitro and the sensory response to heat in vivo (10). ANO1 is itself a heat-sensitive channel, with its thermo-sensitivity modulated by Ca²⁺ (9). Moreover, both ANO1 and TRPV1 channels are targeted by BK signals (8, 19, 39). Thus, elucidating the relationships between these two channels may reveal aspects of thermal sensitivity and inflammatory pain mechanisms. Takayama and colleagues suggested that TRPV1 and ANO1 interact physically and that Ca²⁺ influx through TRPV1 is the main mechanism of coupled ANO1 activation. However, TRPV1 is a nonselective cation channel and under physiological conditions conducts mostly Na⁺ currents with Ca²⁺ influx contributing only 7–10% of the overall current (30). On the other hand, TRPV1 activation in DRG neurons induces PLC activation and ER Ca²⁺ release (31). Given the coupling of ANO1 to ER-released Ca²⁺, we used our optical recording approach to test if this Ca²⁺ source contributed to TRPV1-ANO1 coupling as well. TRPV1 activation with 1 μM capsaicin induced significant EYFP-QL quenching (Fig. 3A, B), which correlated with intracellular Ca²⁺ rises (Fig. 3A). Both the capsaicin-induced EYFP-QL quenching and Ca²⁺ transients were comparable to those induced by BK (Fig. 3C).

Next, we tested if Ca²⁺ release from the ER contributed to TRPV1-induced ANO1 activation. As a first step, we measured capsaicin-induced Ca²⁺ transients in untransfected DRG neurons after ER store depletion produced by inhibition of the sarco/endoplasmic reticulum Ca²⁺-ATPase by either thapsigargin (TG; 1 μM) or cyclopiazonic acid (CPA;1 μM; Fig. 3D, fig. S2A). Consistent with a previous report (31), Ca²⁺ transients were significantly reduced in the presence of both drugs. However, after removal of Ca²⁺ from the extracellular bath solutions, capsaicin still produced Ca²⁺ signals (Fig. 3D, fig. S2A). These were significantly smaller as compared to control conditions, yet clearly identifiable. These initial experiments indicate that TRPV1 can induce sizeable Ca²⁺ release from the internal stores, which contributes approximately one half of the global Ca²⁺ signal measured by fura-2.

Next, we used triple-wavelength imaging to investigate the effect of store depletion on ANO1 activation, as measured with EYFP-QL quenching. Depletion of ER Ca²⁺ stores with TG reduced not only Ca²⁺ transients after capsaicin application but also EYFP-QL quenching (fig. S2B, C). We also applied capsaicin in Ca²⁺-free extracellular solution and using our triple wavelength imaging approach determined that both Ca²⁺ transients and significant EYFP-QL quenching were still present (fig. S3A, B). These results provide further support for our proposal that CaCC activation through TRPV1 requires the ER component. We also tested whether ER Ca²⁺ ‘leak’ could induce CaCC activation in our experimental setup. Acute application of TG to EYFP-QL-transfected DRG neurons produced Ca²⁺ transients and some EYFP-QL quenching (fig. S3C, D).

To further probe the relationship between TRPV1 activation, ER Ca²⁺ release and ANO1 activation, we expanded our imaging approach to enable simultaneous but separate monitoring of Ca²⁺ levels in the ER and cytosol together with EYFP-QL quenching. To allow ER Ca²⁺ monitoring, we used the ER-targeted, Ca²⁺ sensing protein red-CEPIA (40). CEPIA fluorescence is reduced as Ca²⁺ is released from the ER. We transfected DRG cultures with both EYFP-QL and red-CEPIA and also loaded cells with fura-2AM, thereby allowing us to concurrently monitor CaCC activity, ER Ca²⁺ levels and cytosolic Ca²⁺ dynamics, respectively. A quadruple-wavelength (340/380/488/560nm) imaging protocol was utilized to simultaneously image all fluorophores in the experiments. Application of capsaicin produced Ca²⁺ transients and evoked significant EYFP-QL quenching but also resulted in a reduction in red-CEPIA fluorescence which indicates Ca²⁺ moving out of the ER after TRPV1 activation (Fig. 3E, F). This experiment also directly demonstrated good temporal correlation between ER Ca²⁺depletion, cytosolic Ca²⁺ elevation, and CaCC activation.

To validate our imaging results, we made whole-cell patch clamp recordings from cultured DRG neurons using physiological bath and internal pipette solutions. Application of capsaicin elicited robust inward currents (Fig. 4A). The current amplitude and kinetics of the response varied substantially; therefore, we quantified the area under the curve (A.U.C.) as a readout of TRPV1 activation. To test the contribution of ANO1 to this current, DRG neurons were preincubated with Ani9 (500 nM; Fig. 4A). In the presence of Ani9, the capsaicin-induced A.U.C. was significantly reduced, which was consistent with a previous study (10). To test if Ca²⁺ release from the ER was necessary for TRPV1-coupled ANO1 activation in DRG neurons, we pretreated neurons with TG (1 μM), which significantly reduced the capsaicin-induced A.U.C. (Fig. 4B, C). In the presence of TG, the majority of responses to capsaicin were lower compared to those evoked in control conditions (Fig. 4B, black trace; Fig. 4C). However, in 5 TG-treated neurons, capsaicin-induced responses were within the range of the average capsaicin response under control conditions. (Fig 4B, red trace; Fig. 4C). These results suggested three conclusions. First, ANO1-mediated Cl⁻ current contributed substantially to the macroscopic currents induced in DRG neurons by capsaicin. Second, TRPV1-induced release of Ca²⁺ from the ER contributed substantially to TRPV1-coupled ANO1 activation in the majority of DRG neurons. Third, in some neurons, TRPV1-mediated Ca²⁺ influx was sufficient to induce strong ANO1 activation. The latter observation also argues against the possibility that TG directly interferes with TRPV1 function.

TRPV1-mediated Ca²⁺ release from the ER requires crosstalk between plasmalemmal TRPV1 and PLC

One important question that needed to be addressed was the manner in which capsaicin activated ER Ca²⁺ release. The two most likely scenarios suggested in the literature are activation of PLCs that cleave phosphatidylinositol 4,5-bisphosphate (PIP₂) to produce IP₃, which in turn, induces Ca²⁺ release through IP₃Rs in the ER (41), and direct activation of ER-localized TRPV1 channels by capsaicin (42). To differentiate between the above mechanisms, we used two approaches. First, we performed fluo-4- Ca²⁺ imaging in DRG neurons treated with capsaicin after preincubation with (and in the presence of) ruthenium red (RR; 10 μM), a cell-impermeable TRPV1 blocker (Fig. 5A, B). Without access to the cytosol, RR should not inhibit putative ER-localized TRPV1 channels; therefore, capsaicin responses in the presence of this TRPV1 blocker could be attributed to this subset of channels (or any other intracellular TRPV1 channel pool). There was generally no response to capsaicin in the presence of RR (Fig. 5A, B). Upon wash out of RR and with capsaicin still present, Ca²⁺ responses were routinely recorded, suggesting that capsaicin did not induce noticeable Ca²⁺ transients through putative ER-resident TRPV1 channels in the majority of DRG neurons under our experimental conditions. Of 38 neurons that responded to capsaicin upon RR washout, only 4 cells produced discernable Ca²⁺ transients in the presence of RR and capsaicin. To specifically probe ER-Ca²⁺ release by TRPV1 activation, we performed imaging in DRG neurons expressing green-CEPIA to enable direct monitoring of ER Ca²⁺ levels. Capsaicin application produced significant ER-depletion only after RR washout (Fig. 5C, D).

Second, we utilized a cell-impermeable TRPV1 ligand to specifically activate PM-resident TRPV1 channels while avoiding activation of intracellular TRPV1s. Double-knot toxin (DkTx) is a vanillotoxin from the Chinese bird spider (Ornithoctonus huwena) which activates TRPV1 by binding to its extracellular outer pore turret (43–45). Here we used a single-knot derivative of DkTx, K2, which activates TRPV1 with improved efficacy (46). Both capsaicin and K2 (30 μM) evoked comparable Ca²⁺ transients in DRG neurons (Fig. 5E, F). In the absence of Ca²⁺, K2 still induced cytosolic Ca²⁺ transients (Fig. 5F and fig. S4). Finally, DRG cultures were transfected with red-CEPIA and K2 was applied to test whether Ca²⁺ release from the ER could be detected. K2 induced significant reduction in CEPIA fluorescence in 8/12 neurons tested (Fig. 5G, H). The data obtained with both a cell-impermeable TRPV1 agonist and a cell-impermeable antagonist suggested that activation of PM-localized TRPV1 channels was indeed capable of inducing Ca²⁺ release from the ER. Contribution of putative ER-localized TRPV1 channels to the capsaicin-induced cytosolic Ca²⁺ signal in DRG neurons appeared to be negligible.

The alternative possibility is that TRPV1 activation in DRG neurons was coupled to PLC activation and subsequent Ca²⁺ mobilization (31). To test this possibility, we transfected DRG neurons with an optical PIP₂ reporter, GFP-tagged plextrin homology domain of PLCδ (PLCδ-PH-GFP) and monitored its translocation from the membrane to cytosol following the capsaicin application as a measure of PIP₂ hydrolysis by PLC (47). Application of capsaicin (1 μM) induced translocation of PLCδ-PH-GFP from the PM to the cytosol (fig. S5A, B). Translocation was recorded even in the in the absence of extracellular Ca²⁺, albeit to a smaller extent (fig. S5B). Both of these findings are consistent with the previous report (31) and with our K2 experiments.

ANO1, TRPV1 and IP₃R cluster at ER-PM junctions

Our results above suggested that there were functional interactions between ANO1, TRPV1 and IP₃Rs that allowed TRPV1 to activate ANO1 through IP₃R-mediated Ca²⁺ release, supplementing the relatively small Ca²⁺ influx through TRPV1 itself (30). For this mechanism to occur, there must be close proximity between these channel proteins. Previous work suggests that at least some ANO1 channels in DRG neurons reside at ER-PM junctions (1, 19), which are sites for intracellular information transfer where two membrane types come into very close juxtaposition (~15 nM) (48). At these sites ANO1 channels may physically interact with the IP₃Rs (19, 49). To test if all three proteins were found in close association, we first utilized a proximity ligation assay (PLA) which detects two proteins that are closer than 30–40 nm from each other by manifesting characteristic 0.5–1 μm fluorescent puncta (50, 51). All antibodies used for PLA have been previously validated extensively by us in sensory neurons (19, 52).

Specific PLA signals indicating close proximity were detected for TRPV1-ANO1 (Fig. 6A), ANO1-IP₃R1 (Fig. 6B) and TRPV1-IP₃R1 (Fig. 6C) protein pairs. A particularly high concentration of puncta per neuron was observed for the TRPV1-ANO1 pair (Fig. 6D). Positive PLA controls using two primary ANO1 antibodies raised in rabbit and goat and corresponding secondary PLA probes displayed robust puncta (fig. S6A). When only the rabbit primary antibody was used with both secondary probes, no signal was detected (fig. S6B). No puncta were observed when PLA was performed between IP₃R1 and CD71 (also known as the transferrin receptor) (fig. S7A), a non-raft plasmalemmal protein (19, 53) that is not expected to co-localize with the members of the ANO1 junctional complex (19). These controls confirm PLA specificity in our experimental settings. Together, these data established that ANO1, TRPV1 and IP₃R1 in DRG neurons often are found in close proximity. Of the three pairs tested, TRPV1-ANO1 complexes were more easily detectable, perhaps reflecting the PM residence of these proteins, whereas ANO1-IP₃R1 and TRPV1-IP₃R1 interactions exist across the ER-PM junction.

STORM super-resolution microscopy allows direct visualization of individual ANO1-TRPV1-IP₃R complexes

To further probe the existence of multi-channel complexes suggested by the above data with greater precision, we employed multi-wavelength super-resolution microscopy that can detect individual protein molecules in the membrane of fixed primary cells. Stochastic optical reconstruction microscopy (STORM) (54, 55) allows multiple fluorophore excitation and detection using visible light, enabling localization of individual proteins, or complexes within a lateral resolution of less than 50 nm (52). Our previous work (52) describes in detail the extensive controls performed for such multi-wavelength STORM and the analysis paradigms that we have developed. The traditional STORM approach incorporates different “activator” fluorescent dyes (in this work we used Alexa Fluor 405, Cy2, and Cy3, excited at either 405 nm, 488 nm or 561 nm, respectively) that excite a common “reporter” fluorescent dye (for example, Alexa Fluor 647), that are both conjugated to secondary antibodies to identify which activator dye is the source of the excitation signal through tight temporal correlation. Each protein of interest is first labeled with appropriate primary antibodies that we have validated previously (19, 52). Because each secondary antibody is specific to the species of a corresponding primary antibody (mouse, rabbit, goat), we can simultaneously visualize individual complexes containing up to three different proteins. Control experiments by our group have shown a single tetrameric ion channel (such as voltage-gated K⁺ channel or a TRP channel) to be represented by a cluster of labeled “centroids” (single-molecule localizations) with a radius of ~30–50 nm, in which a single type of protein is represented by a cluster of densely proximal, like-colored centroids (52). Complexes containing two or three types of physically associated proteins are represented by clusters of centroids of two or three colors, with a slightly greater radial “footprint” of the antibody-dye label (52). Each standalone cluster of centroids of a single color represents a protein or a complex (such as a multimeric ion channel), recognized by a single type of a primary antibody. Clusters of centroids are objectively identified with a DBSCAN algorithm based on the original STORM localization data, which we previously described. Differential bleaching kinetics of Cy2 and Cy3 and Alexa Fluor 405 (fig. S8) was corrected.

First, we used STORM to probe for the same combinations of protein complexes tested in the PLA experiments (TRPV1-ANO1, ANO1-IP₃R1 and TRPV1-IP₃R1; Fig. 7A, B, C). Clusters of centroids representing individual ANO1, TRPV1 and IP₃R1 channels were detected in each corresponding experiment and, as expected, a fraction of these centroid clusters were co-mingled within a combined radius of <120 nm, indicating close association. Even when in tight proximity (<100 nm), the footprint of each channel protein, as labeled by the STORM-visualized antibody/dyes could be individually resolved from the associated, coupled protein (Fig. 7A, B, C). Of the total detected clusters of centroids for TRPV1 and ANO1, 74.1 ± 6.5% were identified as co-localized channels based on our analysis paradigm (52) (Fig. 7D). The observed overlap of the two channels (as represented by their respective like-colored centroid clusters) in 2D space in the STORM image was usually not complete, suggesting discrete channels that were physically coupled in the PM within very close proximity. Non-colocalized ANO1 (2.3 ± 0.5%) and TRPV1 channels (23.5 ± 6.4%) were also detected which were a lower fraction compared to the co-localized population (table S1).

In our experiments, STORM was performed on a microscope platform similar to that of total internal reflection microscopy (TIRF), in which illumination of cells is typically restricted to <400 nm from the surface of the glass, with the bulk of the light internally reflected (56, 57). In STORM, the angle of the laser light can be altered so that the illumination is not fully internally reflected, but adjusted so it is sufficiently oblique to penetrate up to ~1 μm, allowing illumination of the cytosol or organelles (such as the ER membrane) within 1 μm of the plasma membrane. Thus, we used STORM to examine the presence of complexes containing individual TRPV1 or ANO1 channels in the PM, and closely proximal to ER-resident IP₃Rs. In double-labeling of ANO1 and IP₃R1 under STORM, we detected co-localized ANO1-IP₃R1 complexes, represented by clusters of centroids of two colors, accounting for 54.2 ± 4.3% of all detected centroid clusters observed in DRGs (Fig. 7E). Clusters of centroids of a single color, representing individual ANO1 channels or IP₃R1s comprised 23.8 ± 4.0% and 22.0 ± 8.1% of all clusters, respectively. We also probed for complexes containing TRPV1 and IP₃R1, and found co-localization of the two proteins, accounting for 51.8 ± 4.0% of detected clusters (Fig. 7F). Individual TRPV1 and IP₃R1 clusters amounted to 34.0 ± 4.0% and 14.1 ± 0.8% of total clusters, respectively (table S1). As with the PLA data, the fraction of total TRPV1-ANO1 complexes per cell was more abundant as compared to complexes containing ANO1-IP₃R1 or TRPV1-IP₃R1. Although this cross-comparison of protein pairs (both by STORM and PLA) suggests the co-localization of ANO1, TRPV1 and IP₃R1 that we can visualize, it does not provide quantitative evidence for the relative abundance of individual pairs, because the primary antibody affinities to their respective epitopes are not uniform and cannot be meaningfully normalized.

Because the PLA approach is limited to the detection of protein pairs only, we performed triple-labeled STORM imaging to probe for complexes containing all three proteins (ANO1, 405/647; TRPV1, 488/647; IP₃R1, 561/647; Fig. 8A, B, C). The fraction of ANO1, TRPV1, and IP₃R1 detected as co-localized together was 40.2 ± 0.2%, which accounted for the majority of the co-localizations compared to the pairs of only TRPV1-ANO1 (15.7 ± 1.6%), ANO1-IP₃R1 (4.2 ± 0.04%), and TRPV1-IP₃R1 (6.4 ± 1.6%) without a third protein observed (table S1). The ANO1-IP₃R1-only and TRPV1-IP₃R1-only populations were small and under the noise threshold of interactions detectable under STORM that we previously determined. This “noise threshold” was experimentally determined as ~10% of total clusters (based on spurious interactions of KCNQ2 and KCNQ4 channels) and in which the binned cluster radii are consistently less than 1% of the total fraction of clusters (52). Clustering between IP₃R1 and CD71 amounted to 10.6 ± 1.6% of total clusters (fig. S7B), confirming the stringency of our threshold criteria and agreeing with the PLA results (fig. S7A).

In triple-labeled cells, TRPV1-ANO1 co-localizations remained prevalent, consistent with the large number of physical interactions of these two PM proteins in the double-label STORM experiments. However, the mean radius and centroid density of the ANO1-TRPV1-IP₃R1 clusters were significantly larger than these of the TRPV1-ANO1 (no IP₃R1) clusters, presumably due to the addition of the IP₃R1s (tables S2 and S3). This is important when comparing these data to the double-labeled STORM experiments, in which the cluster populations of co-localized pairs of proteins were similar in size and centroid density (tables S2 and S3). Based on the proportion of complexes, we can surmise that ANO1 and TRPV1 are more likely to cluster in the presence of co-localized IP₃R1, as evident in the significantly greater number of triple-color clusters including IP₃R1s.

Cell culture and transfection

DRGs were extracted from 7 day old Wistar rats and neurons were dissociated as previously described (8, 19). Cultures were maintained on 10 mm glass coverslips pre-coated with poly-D-lysine and laminin and cells were left to grow for 48 hours in DMEM supplemented with GlutaMAX I (Invitrogen), 10% fetal bovine serum, penicillin (50 U/ml), and streptomycin (50 μg/ml) in a humidified incubator (5% CO2, 37°C). For transfection with the EYFP-QL or CEPIA constructs, the Lonza Nucleofector I was used and transfection was done before plating out the cells as described (70).

Live imaging

DRG cultures were loaded with 2 μM fura-2AM (Life Technologies) in the presence of 0.01% pluronic F-127 (Sigma) at 37°C for 45 minutes. Imaging was done using a Nikon TE-2000 microscope (inverted) with a 40× objective and the imaging setup consisted of a Polychrome V monochromator, IMAGO CCD camera and TILLVision 4.5.56 (TillPhotonics) or Live Acquisition 2.2.0 (FEI) software packages. For standard fura-2 imaging (no EYFP-QL transfection), excitation at 340 and 380nm was performed. For fluo-4 imaging, untransfected cells were loaded with fluo-4AM using the same method as for fura-2AM loading. An imaging system comprising of a Nikon TE-2000 E microscope with a DQC-FS camera and a Livescan™ Swept-field confocal system, was utilized for fluo-4 imaging. Analysis was done using Nikon Elements software.

To simultaneously image Ca²⁺ and I⁻, DRG neurons (~20 μm diameter) successfully transfected with EYFP-QL and loaded with fura-2 were selected for imaging; glial cells (identified by bipolar processes) were excluded from the analyses. The simultaneous Ca²⁺/I⁻ imaging was performed using the UV-2A filter cube (Nikon) and excitation was performed at 340/380 (fura-2) and 488nm (EYFP-QL) wavelengths. Exposure time (typically within 50–200ms for fura-2 and 400–500ms for EYFP-QL) was set for each wavelength on a cell to cell basis, depending on the basal fluorescence intensity. For quadruple wavelength imaging, DRG (transfected with EYFP-QL) loaded with fura-2 were imaged using 340, 380, 488 and 560nm wavelength light using a custom-made dichroic/filter combination DC/59022bs-XR-360-UF1 and DC/59022m; Chroma).

Standard extracellular bath solution consisted of (mM): NaCl (160); KCl (2.5); MgCl₂ (1); CaCl₂ (2); HEPES (10) and Glucose (10); pH adjusted to 7.4 with NaOH (All from Sigma). I⁻ containing solutions were produced by equimolar substitution of the NaCl with NaI (30 mM, 10 mM or 5mM). Compounds were added to the NaI containing solution whilst imaging: GABA, 100 μM (Tocris); capsaicin, 1 μM (Sigma); bradykinin, 250 nM (Merck), ruthenium red, 10 μM (Sigma). K2 was a generous gift from Dr. Avi Priel (The Institute for Drug Research, School of Pharmacy, Faculty of Medicine, The Hebrew University of Jerusalem). The depolarizing 50 mM K⁺ solution was produced by equimolar substitution of NaCl with KCl. Ca²⁺-free solutions were produced with increased NaCl (165 mM) to compensate for Cl⁻ lost with the removal of CaCl₂; 1 mM EGTA was also added. Pre-incubation with thapsigargin (1 μM; LKT Laboratories) or cyclopiazonic acid (1 μM; Tocris) was done during the fura-2 loading process where required and the compound was also kept in solutions throughout imaging. Gravity bath perfusion was used for all experiments apart from those involving K2, for which an Automate Scientific SmartSquirt Microperfusion system was utilized. A fine 100μm tip was used with this system and pressure was maintained at 3psi for all experiments using this perfusion method. The tip was positioned over DRG neurons selected for testing under 10× or 40× magnification before being activated.

Imaging data analysis

Analysis of primary data and statistics were performed in Microsoft Excel, Origin Pro and GraphPad Prism 8. Fura-2 ratios were normalized to R value at t=0 (R₀) and expressed as a difference from R/R₀ (ΔR/R₀). EYFP-QL fluorescence intensity was normalized to fluorescence at t=0 (F₀) and expressed as a difference from F₀ (ΔF/F₀), for each cell. Data from all cells were temporarily aligned to the point of R/R₀ increase by 3 times the standard deviation of the baseline. Agonist-dependent EYFP-QL fluorescence quenching or red-CEPIA fluorescence intensity change were quantified from this point until the end of agonist application. Where possible, linear agonist-independent run-down of fluorescence signals was corrected before final analysis.

Electrophysiology

Whole cell voltage clamp recordings were performed as previously described (8). Patch electrodes were pulled from borosilicate glass and fire polished to a final resistance of 2–4 MΩ. An EPC-10 patch clamp amplifier in combination with Patchmaster version 2×35 software (HEKA) was used. Standard extracellular bath solution consisted of (mM): NaCl (160); KCl (2.5); MgCl₂ (1); CaCl₂ (2); HEPES (10) and glucose (10); pH adjusted to 7.4 with NaOH (all from Sigma). The internal pipette solution consisted of (mM): CsCl (150); MgCl₂ (5); K₂ATP (1); NaGTP (0.1); EGTA (1) and HEPES (10). pH adjusted to 7.4 with CsOH (all from Sigma). After obtaining a gigaseal, whole cell configuration was attained and gap-free recordings performed at a holding voltage of −60mV. Capsaicin (1 μM) was applied to DRG neurons with or without 30 min pre-incubation with Ani9 (500 nM; Tocris) or thapsigargin (1 μM). Fitmaster v2×73.5 (HEKA) has been used for data analysis.

PLCδ-PH-GFP translocation

Cultured DRG neurons were transfected with the PLCδ-PH-GFP construct using the same method as for EYFP-QL and CEPIA transfection (see above). The Nikon confocal imaging and analysis system used for fluo-4AM live Ca²⁺ imaging was also used for PLCδ-PH-GFP translocation assays. GFP fluorescence intensity in the cytosolic regions of the cell was used as a measure of the probe translocation.

Proximity Ligation Assay (PLA)

PLA was performed using PLA kits (Sigma) as previously described (19). Antibodies used were ANO1 (1:200; Santa Cruz or 1:500; Abcam), TRPV1 (1:500; Neuromics) and IP₃R1 (1:500; Calbiochem), CD71 (1:500, Santa Cruz). For anti-guinea pig antibodies (such as TRPV1), PLA ‘minus’ probes were manually conjugated onto an anti-guinea pig IgG antibody using a PLA conjugation kit. DRG cultures were prepared and plated on microscope slides (coated with poly-d-lysine and laminin) and permeabilized using acetone:methanol (1:1) on ice for 20 minutes and washed with PBS 3 times. Hydrophobic barriers were made on the slides using an ImmEdge hydrophobic barrier pen (Vector laboratories) to delimit reactions to ~1cm² and blocking was done using the PLA blocking reagent for 30 minutes in a humidified incubator (37°C). Primary antibodies were then applied and left at 4°C overnight. The following day, PLA probes were applied to the samples; signals were amplified and detected according to the manufacturer’s instructions; and slides were sealed with DAPI-containing mounting medium (Sigma). Samples were imaged using the Zeiss LSM700 confocal microscope. Green fluorescent puncta (0.5–1μm) were counted per cell using Zen imaging software (Zeiss).

Statistical analysis

Data are presented as mean ± S.E.M unless stated. Data were tested for normality and either parametric (Student’s t-test, ANOVA) or non-parametric (Wilcoxon signed-rank test; Mann Whitney test, Kruskal-Wallis ANOVA) tests were carried out. Difference between means was considered significant at p<0.05. Approach to analysis of STORM data is described below.

Stochastic Optical Reconstruction Microscopy (STORM)

All images were acquired on a Nikon N-STORM super-resolution system (Nikon Instruments Inc.), consisting of a Nikon Eclipse Ti inverted microscope and an astigmatic 3D lens placed in front of the EMCCD camera to allow the Z coordinates to be most accurately determined. The laser set-up of our STORM system allows for the multiple, largely non-overlapping activator dyes, Alexa Fluor 405 carboxylic acid (Invitrogen, #A30000), Cy2 bisreactive dye pack (GE Healthcare, #PA22000), and Cy3 mono-reactive dye pack (GE Healthcare, #PA23001), to be conjugated to affinity-purified secondary antibodies from Jackson Immunoresearch, along with the reporter dye, Alexa Fluor 647 carboxylic acid (Invitrogen, #A20006) in order to acquire images from three different protein labels in the same experiment. Using the same primary antibodies used in the PLA experiments, we determined the extent of co-localization of TRPV1, ANO1, and IP₃R1, as individual multi-protein complexes with either two or three photo-switchable dye activators, Alexa Fluor 405, Cy2, and Cy3, coupled to three distinct secondary antibodies (raised in different species) labeling the proteins in DRGs. The activator and reporter fluorophores were conjugated in-house to an appropriate unlabeled secondary antibody (71). STORM imaging was performed in a freshly prepared imaging buffer that contained (in mM): 50 Tris (pH 8.0), 10 NaCl and 10% (w/v) glucose, with an oxygen-scavenging GLOX solution (0.5mg/ml glucose oxidase (Sigma-Aldrich), 40 μg/ml catalase (Sigma-Aldrich), and 10 mM cysteamine MEA (Sigma-Aldrich). MEA was prepared fresh as a 1 M stock solution in water, stored at 4°C and used within 1 month of preparation. Acquisitions were made from between 2–4 different experiments for labeling and imaging. Images were rendered as a 2D Gaussian fit of each localization. The diameter of each point is representative of the localization precision (larger diameter, less precise), as is intensity (more intense, more precise). Signal-noise thresholds were handled as peak height above the local background in the N-STORM software (Elements). A detected peak was set as the central pixel in a 5×5 pixel area, and the average intensity of the 4 corner pixels was subtracted from intensity of the central pixel. Using a 100× objective and 16×16μm pixel area of the iXon3 camera, this corresponds to a 0.8×0.8μm physical neighborhood.

Adjustment of STORM localizations for differential rates of dye bleaching

The STORM acquisitions were carried out for a minimum of 6,000 activator-reporter cycles and until at least one of three activator laser wavelengths no longer registered localizations. We observed that the Alexa Fluor 405-Alexa Fluor 647 dye pair persisted stably for more cycles than the Cy2-Alexa Fluor 647 or Cy3-Alexa Fluor 647 dye pairs, independent of the secondary antibody that was conjugated (fig. S8). Therefore, we concluded that the Cy2 and Cy3 dyes bleach out more rapidly than Alexa Fluor 405 after many repeated activation cycles. Such differential rates of photobleaching would cause us to over-estimate the number of complexes containing a protein labeled by Alexa Fluor 405, and possibly under-estimate those without. Hence, we corrected for this artifact to the best of our ability. To more adequately compare the STORM localizations in the triple-labeled conditions, we first determined a linear approximation of the differential exponential decay of localizations for each corresponding activator dye. The cumulative probability distributions demonstrated that Alexa Fluor 405 had a significantly greater number of localizations than equally labeled Cy2 or Cy3 at approximately 3,000 laser activation cycles (36,000 total frames) and beyond (fig. S8). The factor of decay was used to time-dependently correct for this by probabilistically excluding Alexa Fluor 405-specific localizations from the cluster detection analysis to compensate for this, preventing oversampling above that of Cy2 or Cy3 molecules. After this adjustment, the numbers of clusters of each combination of multi-channel complexes did not significantly change due to the details of our sampling paradigm, except for those of Alexa Fluor 405-only centroid clusters. We estimate that the applied compensation paradigm corrected for the above mentioned artifact by >90%.

Cluster size and localization proximity analysis

Unfiltered STORM localization data were exported as molecular list text files from Nikon-Elements and were analyzed with in-house software incorporating a density-based spatial clustering of applications with noise (DBSCAN) (72). A dense region or cluster was defined as localizations within a directly-reachable radius proximity (epsilon) from a criterion minimum number of other core localizations (MinPts). Density-reachable points were localizations that were within the epsilon radius of a single core point and thus considered part of the cluster. Localizations considered to be noise were points that were not within the epsilon distance of any core points of a cluster. We derived the appropriate epsilon parameter using the nearest-neighbor plot from single-dye labeled controls; cluster detection was determined for epsilon between 20 and 80nm, which were the nearest-neighbor localization distances representing 95% of area under the curve. DBSCAN parameters were verified by measuring goodness of fit to Gaussian distribution with cluster population data from single-dye labeled controls, and set for distance of directly reachable points at 50nm (epsilon) and 7 minimum points (MinPts). These parameters were found to be the most stringent possible, that also reliably fit the control data. For cluster detection, each localization was assessed based on its corrected X and corrected Y 2D spatial coordinates, and the associated activator dye was tracked throughout analysis. Detected clusters were tabulated by the composition of resident activator dyes contributing to the total neighborhood of localizations for that cluster. Clusters were categorized based on activator dye composition as Alexa Fluor 405 only, Cy2 only, Cy3 only, Alexa Fluor 405 + Cy2, Alexa Fluor 405 + Cy3, or Alexa Fluor 405 + Cy2 + Cy3, according to the dye conjugated to each antibody label. Cluster radius size (nm) data were placed in probability distribution histograms with bin size of 5nm. Based on the observation that a majority of cluster distributions displayed positive skew, distributions were fit by the generalized extreme value distribution function: f(x) = e^{−(x−1+ê−x)}. Cluster size values were reported as mean ± SEM. The β continuous scale parameter of each cluster distribution was derived from the mean value of the extreme distribution (Mean = μ + γβ, γ: 0.5772, Euler-Mascheroni constant) and was determined as a metric for “tailedness” of cluster size distribution. In addition, the percentage of clusters belonging to a labeling category out of total clusters was compared. Cumulative distribution functions of cluster categories were compared with the Kolmogorov-Smirnov test. Population statistics for each cluster category were derived from 6–12 cells per staining and imaging condition.