The ER Ca²⁺ sensors STIM1 and STIM2 are essential regulators of SOCE, and play important roles in the maintenance of intracellular Ca²⁺ homeostasis (Chen et al., 2019a). In the present study, we used SiHa cells overexpressing STIM1 or STIM2, treated with thapsigargin (TG) for ER Ca²⁺ depletion, to investigate the ultra-fine structure of STIM proteins during SOCE activation. The morphology of STIM1 overexpressed in SiHa cells after TG treatment is identical to the endogenous expression of STIM1 (Supplementary Figure 1), but the former is conductive for characterizing the architecture of STIM1 (Chen et al., 2011, 2013a). These structures were observed by fluorescence imaging with different techniques, including conventional wide-field (WF) epifluorescence microscopy, confocal microscopy, and super-resolution imaging by direct STORM (dSTORM) (Figure 1A). The estimated localization precision was demonstrated in Supplementary Figure 2 and Supplementary Movie 1. The full width at half maximum (FWHM) of STIM1 estimated by dSTORM imaging was ∼250 nm, much smaller than the ∼650 nm measured by WF and confocal microscopy (Figure 1B). Therefore, dSTORM imaging provided a more accurate characterization of STIM1 morphology compared to conventional imaging methodologies. According to confocal imaging, TG induced the aggregation of STIM1 into “puncta-like” structures (Figure 1A). In striking contrast, dSTORM showed that STIM1 exhibited “powder-like” structures scattered through the cytosol at resting status, and aggregated into “trabecular-like” architectures toward the cell periphery after 10 min of TG treatment (Figure 1A). The sizes of the activated STIM proteins were analyzed using Integrated Morphometry Analysis (IMA) Measurement Parameters and MetaMorph software (Molecular Devices, San Jose, CA, United States) (Figure 1B and Supplementary Figure 3). Compared to STIM1, STIM2 showed no obvious aggregation after TG treatment, either when assessed by confocal images or by dSTORM (Figure 1C).

We also used dSTORM to observe the dynamic framework of STIM1 during SOCE. Figure 2A shows the time course of STIM1 activation induced by TG, indicating dynamic membrane trafficking. Figure 2B shows the quantitative analysis of STIM1 size during the activation and deactivation processes, which increased from 0.04 μm² to a maximum of 0.15 μm² after 10 min of TG treatment, and then decreased. We used dSTORM to map the spatiotemporal pattern of STIM1 spread in the cell following TG stimulation (Figure 2C and Supplementary Figure 3), and the most prominent alteration of STIM1 size was observed near the cell periphery, indicating the aggregation and trafficking of STIM1 to the plasma membrane during SOCE.

We have previously demonstrated the essential role of microtubules and end-binding protein 1 (EB1) for STIM1 trafficking (Chen et al., 2013a), and in this study we used dSTORM imaging to re-evaluate the functions of microtubules and EB1 during SOCE. Two-color mapping dSTORM imaging identified the morphological alterations and co-localization among STIM1, microtubules and EB1 in SOCE (Figure 3A). The molecular distance was calculated using FWHM according to the fluorescence intensity (Supplementary Figure 4). The distance between them was closest after TG treatment for 10 min (Figure 3B and Supplementary Table 1). The size of activated STIM1 could be variable while interacting with microtubules (Figures 3C,D), a finding which indicates that, once activated, STIM1 can bind to microtubules independent of aggregated state. This finding challenges the concept that only the STIM1 clusters interacts with microtubules (Liou et al., 2007; Soboloff et al., 2012; Chen et al., 2013a). We also used dSTORM imaging to clarify the framework of STIM2 during SOCE (Supplementary Figure 5). Although significant alterations were observed in the size and distance to microtubules of STIM1 after TG treatment, this phenomenon has not been observed in STIM2, an observation which implies that there was no significant aggregation or association between microtubules and STIM2 during SOCE. According to our previous study, the microtubule depolymerizer colcemid and the HDAC6 inhibitor tubastatin A both disrupted the microtubule network and subsequent SOCE (Chen et al., 2013a). In the present study, we used dSTORM imaging to assess this phenomenon. Except for colcemid and tubastatin A, we also applied another microtubule depolymerizer, nocodazole, to disrupt microtubules network, and which revealed that disruption of microtubule could inhibit the ER-plasma membrane trafficking of STIM1, but not aggregation, during SOCE (Figure 4 and Supplementary Figures 6, 7). The co-localization of STIM1 and microtubules was more disrupted by colcemid and tubastatin A than those between STIM1 and EB1 (Supplementary Figure 8). In summary, STIM1 therefore appears to interact with microtubules, regardless of undergoing aggregation, and the microtubule network is required for the ER-plasma membrane trafficking of STIM1.

As well as EB1, the microtubule plus-end binding protein 3 (EB3) has been reported to play an important role in the microtubule-dependent translocation and aggregation of STIM1 (Honnappa et al., 2009; Chen et al., 2019a). We used SiHa cells transfected with EB1 and EB3 small interfering RNA (siRNA) to study the impact of EB1 and EB3 on SOCE. Decreased expression level of EB1 and EB3 in the presence of EB1- and EB3-specific siRNA was confirmed using western blotting (Figure 5A and Supplementary Figure 9). After the downregulation of either EB1 or EB3, the change of cytosolic Ca²⁺ concentration (Δ[Ca²⁺]_i) during SOCE was significantly decreased (Figures 5B,C). Two-color mapping dSTORM imaging demonstrated impairment of the microtubule network and accumulation of cytosol STIM1 (Figure 5D), which was a consequence of the defect in interactions between STIM1 and microtubules and in the ER-plasma membrane trafficking of STIM1 after the downregulation of EB1 (Figure 5E and Supplementary Figure 10). The interaction between STIM1 and the microtubule network were less strongly affected by siEB3 (Figures 5D,E). We examined the role of EB3 during SOCE after silencing EB1 (Supplementary Figure 11). Although the distance between STIM1 and EB3 was significantly decreased in the absence of EB1 (Supplementary Figure 11B), the impairment of STIM1 trafficking was still present. This finding implies that EB3 may partially compensate for the interactions between STIM1 and microtubules in the absence of EB1.

Stromal Interaction Molecule 1-dependent signaling plays an important role in cancer cell growth, migration, and angiogenesis (Chen et al., 2011). We aimed to uncover the impact of different STIM1 variants on SOCE and cancer cell behavior. According to COSMIC, a real-world database, there are six somatic mutations in Ca²⁺ binding domains and two in microtubule interacting domains (Figure 6A and Supplementary Table 2). The six STIM1 variants were generated into SiHa cells using site-directed mutagenesis. The migration activity was decreased in the variants of the Ca²⁺ binding domain, p.D76G and p.D84Y, but was increased in the variant of the microtubule interacting domain p.R643C, compared to wild-type STIM1 (Figures 6B,C). Pyk2 is uniquely located in focal adhesions and regulate cell migration. The Tyr402 autophosphorylation and focal adhesion targeting of Pyk2 are required for Pyk2-mediated cytoskeletal reorganization and the subsequent determination of focal adhesion turnover and cell migration (Mitra et al., 2006; Chen et al., 2011; Tsai et al., 2015). In consistent with the altered migration activity among different STIM1 mutations, the level of pTyr402-Pyk2 level upon epidermal growth factor (EGF) stimulation in SiHa cells is inhibited in p.D76G and p.D84Y variants, but is stimulated in the p.R643C variant, compared to wild-type STIM1 (Figures 6D–G). To sum up, these real-world STIM1 somatic variants from the COSMIC database affect the migration activity of cancer cells by regulating the turnover of focal adhesion.

After establishing the impact of different STIM1 variants on cancer cell migration, we further used dSTORM imaging to address STIM1 ER-plasma membrane trafficking in SiHa cells with p.D76G, p.D84Y, and p.R643C variants. There is distinct pattern of intracellular distribution and aggregation of STIM1 after TG treatment in SiHa cells harboring different variants (Figures 7A,B). The activation of STIM1 is inhibited in the p.D76G and p.D84Y variants but accelerated in the p.R643C variant. Similarly, the binding between STIM1 and microtubules is promoted in R643C variant (Supplementary Figure 12). The activity of SOCE is also consistent with the dynamic changes of STIM trafficking in wild-type and different STIM1 variants (Figures 7C,D). In summary, different STIM1 variants lead to dissimilar Ca²⁺ influxes during SOCE, which impact STIM1 dynamics and cancer cell migration (Figures 6, 7).

SiHa human cervical cancer cells were maintained in Dulbecco’s modified Eagle medium High Glucose supplemented with 10% FBS and 2 mmol/L L-glutamine. The cells were maintained at 37°C in an environment containing 5% CO₂. For the siRNA-mediated knockdown of EB1 or EB3, an siRNA pool of three different duplexes (Santa Cruz Biotechnology, Dallas, TX, United States) targeting EB1 or EB3 was used. Cells were transfected with 100 nM of either the targeting or the control siRNA (Santa Cruz Biotechnology) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, United States) for 48 or 72 h. The knockdown efficiency was determined by immunoblotting.

Thapsigargin was obtained from Cayman Chemical (Ann Arbor, MI, United States). Colcemid was purchased from Merck (Darmstadt, Germany). Tubastatin A was obtained from BioVision (Milpitas, CA, United States). Nocodazole was purchased from Abcam (Cambridge, United Kingdom). Fura-2/AM was obtained from Invitrogen. EGF Recombinant Human Protein was purchased from Thermo Fisher (Thermo Fisher Scientific, Waltham, MA, United States). The antibodies used in this study are shown in Supplementary Table 3.

Round glass coverslips 20 mm in diameter (No. 1.5, Paul Marienfeld GmbH, Thuringia, Germany) were cleaned using sonication for 15 min with 1 M aqueous potassium hydroxide, then dried. Before use, the slides were rinsed with 100% ethanol, dried, and coated with poly-D-lysine (SI-P0899, Sigma-Aldrich, St. Louis, MO, United States) for seeding cells. Fixed cells at different time points of SOCE were performed by dSTORM super-resolution microscopy.

Super-resolution microscopy in epifluorescence mode was performed using an IX81 inverted microscope (Olympus, Tokyo, Japan) equipped with the MetaMorph Microscopy Automation and Image Analysis Software (Molecular Devices, San Jose, CA, United States) with a super-resolution module using a 60× oil immersion objective (NA = 1.49). For dSTORM imaging, the exposure time was set as 50 ms, and the electron-multiplying gain as 100× (Leduc et al., 2018; Fan et al., 2020). Alexa Fluor 647 (Jackson ImmunoResearch Laboratories, West Grove, PA, United States) or CF 568 (Biotium, Fremont, CA, United States) fluorescence was bleached using the full laser power until individual fluorophores began to blink (Supplementary Figure 2 and Supplementary Movie 1). We imaged with exciting 561-nm laser following with 647-nm laser, which can avoid the bleaching of Alexa Flour 647 by 561-nm laser illumination. The 647-nm laser was used to excite fluorescence from Alexa Fluor 647 molecules. Before acquiring dSTORM images, we used relatively weak 647-nm light (∼0.05 W/cm²) to illuminate the sample and recorded the conventional fluorescence image before a substantial fraction of the dye molecules were switched off. We then increased the 647-nm light intensity (to ∼2 kW/cm²) to rapidly switch the dyes off for STORM imaging. The 561-nm laser followed the same steps. 15,000 frames were recorded to generate the final super-resolution image of the molecular ultrastructure. All imaging buffers were supplemented with the oxygen scavenging system, consisting of 10% (w/v) glucose (Sigma-Aldrich), 100 mM cysteamine (MEA) (M9768, Sigma-Aldrich), 313 mM glucose oxidase (G2133-250KU, Sigma-Aldrich), and 0.16 mM catalase from bovine liver (C40, Sigma-Aldrich). The oxygen scavenging system is critical for the reliable photoswitching of the fluorophores. For dSTORM imaging, 15,000 256 × 256 pixel frames were captured at with an electron multiplying charge coupled devices camera (Andor Technology, Belfast, United Kingdom). Image collection and fitting was performed using WaveTracer (MetaMorph, Molecular Devices, San Jose, CA, United States), which sets a photon-detection threshold to identify genuine blinking events, and uses a wavelet segmentation algorithm to fit a centroid to each blinking event within each 10 nm × 10 nm pixel. The point locations of individually blinking fluorophores in each of the 15,000 frames are collated to form a super-resolution image. TetraSpeck microspheres 0.1 μm on diameter (Thermo Fisher Scientific) were used to verify that there was no significant drift during image acquisition.

The super-resolution image was randomly selected from a single cell at a view area of 40 × 40 μm² including cytosol and cell periphery in each experiment. To analyze STIM1 quantitatively, we used a DIC image to identify cell membrane and used MetaMorph software to define the distance from cell membrane in dSTORM imaging. With an appropriate and fixed threshold, STIM1 boundary and size were analyzed by IMA Measurement Parameters (Supplementary Figure 3). To measure the distance between two different molecules, we used line fitting from the raw localization data. We lined the region of interest (ROI) between two different molecules, each peak with a high enough signal-to-noise ratio was fitted to a Gaussian function by non-linear least-square fitting, positions as well as photon counts were extracted from the fitted peaks for rendering and quantification purposes (Olivier et al., 2013; Franke et al., 2017). The full width at half-maximum (FWHM) was calculated from the resulting histogram, and we calculated the molecular distance by minus the peak-to-peak distance (Supplementary Figure 4).

The full-length cDNA of EGFP-STIM1 was kindly provided by Liangyi Chen of the Institute of Molecular Medicine, Peking University, China (Xu et al., 2006). pEX-SP-YFP-STIM1(23-685) (Addgene; #18857) was purchased from Addgene (Watertown, MA, United States). The mutant STIM1 was generated using site-directed mutagenesis kits (Agilent Technologies, Santa Clara, CA, United States). Primers for the site-directed mutagenesis of STIM1 variants were used to generate the cEF hand and S/TxIP mutant libraries (Supplementary Table 4). Reactions were amplified using the following protocol: 95°C denaturation for 2 min, followed by 20 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 30 s, and extension at 68°C for 2.5 min, followed by a 10 min extension at 68°C. Three STIM1 sequencing primers were used for DNA sequencing to confirm the mutations: 5′-atggatgtatgcgtccgtcttgc-3′, 5′-atcacctcaaggacttcatgctggt-3′, and 5′-acagcattgcgggagc-3′. SiHa cervical cancer cells were transfected with plasmids using Polyplus-transfection reagents (Polyplus-Transfection SA, New York, NY, United States) and selected using G418 (Sigma-Aldrich). The stable pools of cells overexpressing STIM1 or STIM1 somatic variants were isolated using the MoFlo XDP Cell Sorter (Beckman Coulter, Brea, CA, United States).

Stromal interaction molecule 1 variant analysis was carried out using the COSMIC database (Sondka et al., 2018). A lollipop diagram was created in trackViewer (Bioconductor software).

For immunoblotting, cells were harvested with ice-cold modified radioimmunoprecipitation assay (RIPA) buffer containing a protease inhibitor cocktail (Roche Diagnostics, Basel, Switzerland), 100 mM potassium chloride, 80 mM sodium fluoride, 10 mM ethylene glycol tetraacetic acid, 50 mM h-glycerophosphate, 10 mM p-nitrophenyl phosphate, 1 mM vanadate, 0.5% sodium deoxycholate, and 1% NP40. Protein concentrations were determined with using a Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, CA, United States). Equal amounts of protein lysates were separated by SDS-PAGE and then transferred to nitrocellulose blotting membranes (Pall Corporation, Port Washington, NY, United States). Immunoblots were blocked, incubated with the primary antibody, washed, and incubated with the corresponding horseradish peroxidase-conjugated secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, United States), and visualized using western blotting luminol reagent (Santa Cruz Biotechnology). Western blot bands were imaged and quantified using the iBright^TM FL1000 Imaging System (Thermo Fisher Scientific).

Cells attached on glass-bottom dishes were loaded with 2 μM Fura-2/AM in serum-free culture medium at 37°C for 30 min. Cells were then washed three times with phosphate buffered saline (PBS) before [Ca²⁺]_i measurement. ER Ca²⁺ release was induced with 2 μM TG in the absence of extracellular Ca²⁺, followed by the activation of SOCE with the addition of 2 mM Ca²⁺. The Fura-2 was excited alternatively between 340 nm (I340) and 380 nm (I380) using a Polychrome IV monochromator (Till Photonics GmbH, Gräfelfing, Germany) and images were detected using an inverted microscope (IX71, Olympus, Tokyo, Japan) equipped with a xenon illumination system and an IMAGO CCD camera (Till Photonics GmbH, Gräfelfing, Germany). The fluorescence intensity of excitation at 510 nm was monitored, to calculate [Ca²⁺]_i using HCImage Analysis (Hamamatsu Photonics, Shizuoka, Japan). The imaging parameters were acquired every 2 s per frame. We have used this protocol to measured SOCE in the series of studies on cervical cancer cells.

For the migration assays, the ibidi Culture-Insert 2 Well (ibidi GmbH, Planegg, Germany), which has two cell culture reservoirs separated by a 500 μm wall, was used. Cells were plated at 80,000 cells per well and allowed to attach overnight. On the following day, the culture inserts were removed, and light microscopy (IX71, Olympus, Tokyo, Japan) images were acquired. Cells were maintained under standard culture conditions while migrating toward the cell-free gap area, and images were acquired 48 h later. Images were analyzed using ImageJ software. The migration area was represented as the area remaining uncovered by cells.

SiHa cervical cancer cells were fixed for 10 min at room temperature with 4% paraformaldehyde, or with methanol at −20°C for 10 min. Following fixation, the cells were washed two to three times in PBS, and then blocked with 3% BSA for 30 min at room temperature. Samples were incubated with primary antibodies in PBS at 4°C overnight and rinsed at least three times in PBS with Tween 20 (PBST), incubated with Alexa Fluor 647 secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, United States) for 1 h at room temperature, and rinsed at least three times in PBST. To detect the nucleus, cells were stained with Hoechst 33258 (Invitrogen) for 1 h at room temperature. The cells were then washed and mounted, and the fluorophores were excited by a laser at 405 nm or 640 nm, and detected using a scanning confocal microscope (FV-3000; Olympus, Tokyo, Japan). The fluorescence intensity of phosphor-Pyk2 (Tyr402) immunostaining was measured by the analysis program included in the FV-3000.

Quantitative data are shown as mean ± SEM unless otherwise specified. The analyzed number (n) of samples is listed for each experiment.