Na₂ATP, AZ106 (AZ10606120) and minocycline were from Tocris Bioscience (Bristol, UK). Thapsigargin, U73122, U73343, LY294 (LY294,002) and wortmannin were from Sigma–Aldrich (Castle Hill, NSW, Australia). Drugs were added to final solutions from frozen stock aliquots.

BV2 cells (American Type Culture Collection, ATCC; Manassas, VA, USA) were cultured (37°C, 5% CO₂) in DMEM, supplemented with 10% fetal bovine serum and penicillin (100U/ml)/streptomycin (100 mg/ml) (Invitrogen) and passaged every 2–3 days (60–80% confluence). For high-content functional imaging, ∼5,000 BV2 cells were plated into each well of a 384 multi-well plate (BD Falcon) and incubated for 24 h. Wells typically contained 7.5–10 × 10³ cells at the time of experiments. For fluo-4 two-photon recordings, BV2-seeded round cover-slips were directly used for staining and imaging.

BV-2 cells were incubated in medium containing various treatments for 3 h before whole cell extracts were prepared in 1x SDS loading buffer. Denatured samples and calibration curves (4, 8, and 16 μl of mixed whole cell preparations) were loaded onto 26 well, 4–15% Criterion TGX Stain Free gels (Bio-Rad, Hercules, CA, USA) and run for 4 min at 200 V. Protein was transferred to nitrocellulose membrane at 100 V for 30 min, incubated in Pierce Miser solution (Pierce, Rockford, IL, USA) and blocked. Primary antibodies Stim1S (1:500, mouse monoclonal, #610954, BD Biosciences, Lexington, Kentucky, USA) and Stim2 (1:1000, rabbit polyclonal S8572, Sigma–Aldrich, Sydney, NSW, Australia) were diluted in 1% BSA-PBS with 0.025% Tween and membranes incubated overnight at 4°C and 2 h at RT. After incubating with a secondary antibody (goat anti-mouse IgG, goat anti-rabbit IgG, HRP conjugated, 1:20,000), the membrane was coated with chemiluminescent substrate (West Femto, Thermo Scientific, Rockford, IL, USA). Each gel-lane was normalized to the total protein of that sample, determined from Stain Free gel, then to the three point calibration curve thereby providing a quantitative approach to western blotting (Murphy and Lamb, 2013). Data from most samples are the average of duplicate runs.

A Nikon A1 multiphoton microscope equipped with a Coherent^® Ti:Sa laser was used. BV2 cells were loaded with 5 μM fluo-4 AM (Invitrogen, Germany) in normal saline (in mM: NaCl 140, KCl 5, CaCl₂ 2, MgCl₂ 1, HEPES 10 and glucose 10, pH 7.4) for 20 min at 37°C and washed. Before recordings, cells were either flushed with Ca²⁺ free medium for 5 min or kept in normal saline, depending on the experiment. Two-photon excitation was done at 900 nm, and backscattered fluorescence recorded in de-scanned mode. Fluorescence intensity was recorded with a photomultiplier tube after passing through a 525/50 bandpass filter. Time lapse 512 pixels × 512 pixels recordings (XYT) were performed at 0.937 fps. Laser power was <3%. All settings were kept constant between recordings. Chemicals were added from stock solutions to the bath medium to obtain the final concentrations indicated. XYT image stacks were manually analyzed, assigning ROIs to individual BV2 cells and extracting temporal fluo-4 profiles. Individual traces were normalized to the average fluorescence within the first 10 s under initial resting conditions and converted to relative changes ΔF/F₀ for group analysis. Individual data traces are mostly presented as raw fluorescence intensities.

One and half hours before experiments, culture medium was removed from 384-well plates and replaced by normal saline (control solution), supplemented with 2 μM pluronic acid and 2 μM fluo 4-AM (Molecular Probes). After staining, solution was replaced by control solution. For receptor stimulation, control solution contained 1, 10, 100, or 1,000 μM final ATP concentration ([ATP]). For drug screening, control and agonist (ATP) solutions were supplemented with minocycline (1, 10, and 100 μM), AZ106 (10 μM) U73122 (5 μM), U73343 (5 μM), LY294002 (5 μM) or wortmannin (100 nM), respectively. Three hundred and eighty-four-well plates were placed onto the motorized stage of an ImageXpress^® Micro XLS System (Molecular Devices, USA) and cells were imaged with a 10× objective. A mercury arc lamp, passing through a GFP dichroic mirror, was used to excite fluo-4 fluorescence and imaged with a CCD camera. The image resolution after 2 × 2 binning was 696 pixels × 520 pixels. The experimental protocol involved manual addition of 10 μl agonist or 30 μl agonist/drug solution at a rate of ∼0.5 ml/min, then imaging of each well 380 times at a rate of 2 Hz (300 ms exposure time).

Fluorescence images of cells were segmented and quantitated using a modified version of DetecTIFF^® software (Gilbert et al., 2009). Briefly, an averaged image was calculated from all images recorded within a single well, segmented using an iterative size- and intensity-based thresholding algorithm, and then the fluorescence signal of identified cells was calculated as the mean of all pixel values within the cell area. To evaluate drug effects on the resting intracellular Ca²⁺ concentration, the fluorescence intensity of identified cells was calculated from the first image of the image series as the mean of all pixel values within the cell area, using the averaged image as mask for quantitative analysis. The drug effect was calculated as the arithmetic mean of the mean fluorescence intensity of all cells within the averaged image. Cell morphology was quantified as a function of elongation (elongation factor) as:

An elongation factor of one represents a perfectly round cell, indicating a strong effect on cellular fitness or viability. A high value indicates normal morphology and unaffected viability. The drug effect was calculated as the arithmetic mean of the elongation factors of all cells. Individual [Ca²⁺]_i responses were reconstructed from image series of the same cells exposed to a defined ATP concentration or a combination of ATP and a drug. Each image typically contained 100–500 fluorescent cells for each tested well. For functional analysis of time-resolved [Ca²⁺]_i responses and subsequent classification and phenotyping of cell populations, a set of measures was taken from single cell-derived data. Responses were fitted with the LabView VI Curve Fitting Express VI using polynomial model type. The function finds the set of polynomial fit coefficients that best represents the input data. Each response was fitted twice with polynomial orders set to a value of 8 and 25. Responses fitted with polynomial order 8 were further evaluated using the LabView VI Peak Detector VI to identify a sustained calcium signal, indicating macro-pore formation, and activation of BV2 microglia cells. Responses fitted with polynomial order value 25 were analyzed for identification of ‘intermediate [Ca²⁺]_i oscillations’ occurring (i) between a transient [Ca²⁺]_i signal and sustained [Ca²⁺]_i increase, or (ii) sporadically during the entire span of the measurement using the residual of the fit. In addition, the fast-Fourier-transform (FFT) power spectrum of [Ca²⁺]_i responses was calculated for each cell using the LabView VI FFT Power Spectrum and PSD VI. Data obtained from curve fitting, peak and threshold detection and spectral analysis were used to calculate a total of 27 features (Table 1). Those data were used for classification and phenotyping of cell populations.

To identify distinct functional phenotypes, a three-stage procedure was applied. First, a training set comprising a total of 1,709 single cell-derived [Ca²⁺]_i responses was manually assigned to one of three classes using a custom-written LabView-based software. The classes were named and characterized as follows:

Next, a phenotype classification model was computed by a 10-fold, cross-validation of the training set using J48 decision tree algorithm available in WEKA 3.6 software (Hall et al., 2009). We accurately classified 1,683 or 98.5% of all responses correctly. A confusion matrix is shown in Table 2. Finally, all single cell-based ATP-induced [Ca²⁺]_i responses recorded in the presence of different ATP concentrations and ATP/drug combinations were analyzed based on the computed classification model.

Data were processed using MS Office 2010, SigmaPlot, Origin 7G, ImageJ and IrfanView. Biochemical protein data was analyzed using Prism GraphPad. Statistical analysis was done based on one-way and two-way ANOVA tests, checking for data normality and performing post hoc tests (Dunn or Bonferroni method). Asterisks in the graphs indicate significance levels; ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001.

To distinguish purinergic receptor-mediated Ca²⁺ entry from intracellular ER Ca²⁺ release, BV2 cells were incubated in Ca²⁺-free media and internal stores emptied by 5 μM thapsigargin (TG), while two-photon excited fluo-4 fluorescence was recorded. Two-photon microscopy has the advantage of an excitation volume in the μl³ range, which substantially reduces dye bleaching and allows cellular long-term fluo-4 recordings. Figure 1A shows recordings from three different BV2 dishes. There was a small transient increase in fluo-4 fluorescence after TG application. Adding 2 mM external Ca²⁺ elicited a robust Ca²⁺ transient, while Ca²⁺ stores remained emptied. This is a hallmark of store-operated Ca²⁺ entry (SOCE). Applying ATP after an almost complete decline of the SOCE transient evoked a sustained increase in intracellular fluo-4 fluorescence in a concentration-dependent manner. This was mediated by purinergic receptor-operated Ca²⁺ entry (ROCE) to almost maximum Ca²⁺ permeability for ATP concentrations starting from 0.1 mM, as determined from subsequent Ca²⁺ ionophoresis (2 μM ionomycin). We confirmed this in several dishes in more than 20 cells for each condition (Figure 1C). Lower ATP concentrations (50 μM) evoked significantly smaller peak ΔF/F₀ levels than concentrations of 0.1 and 1 mM ATP, for which levels were similar to those in ionomycin-induced results. Not emptying internal stores and adding external Ca²⁺ again did not induce SOCE (Figure 1B), but did reproduce the behavior of purinergic ROCE with empty stores, albeit at a somewhat lower level (Figures 1B,C). One explanation could be that an ongoing blockade of SERCA with TG resulted in a small residual component of SOCE to the resulting Ca²⁺ entry (mainly ROCE). Figure 1B shows that also in the case of sole purinergic ROCE activation (without prior store depletion), the sustained intracellular Ca²⁺ increase was almost complete. This is because the subsequent unspecific breakdown of the sarcolemma by triton X-100 evoked only a brief minor increase before all of the dye dissipated from the cells, resulting in a loss of fluorescence. In other experiments, we addressed the question of whether purinergic receptor activation without external Ca²⁺, but with internal stores not emptied, would induce PLC-mediated (P2Y) signaling to empty internal stores and transiently increase fluo-4 fluorescence. Figure 1B shows a representative Ca²⁺ response demonstrating that this did not occur. Purinergic ROCE was only detected once external Ca²⁺ was re-introduced. Again, ATP-induced Ca²⁺ entry was almost complete compared with subsequent Ca²⁺ ionophoresis.

When we further differentiated early purinergic ROCE responses in BV2 cells in highly spatio-temporally resolved multiphoton recordings, two distinct patterns were observed: oscillatory Ca²⁺ fluctuations and the sustained slow Ca²⁺ increase. Both patterns were immediately initiated after ATP addition. Figure 2A shows a representative immediate oscillatory pattern followed by the sustained increase in intracellular Ca²⁺ due to purinergic ROCE (stores not emptied) over several minutes. Such oscillation patterns were rarely observed in response to 0.1 mM ATP in multiphoton experiments. They were sometimes observed with 50 μM ATP, still at a low frequency (∼15% of experiments). Figure 2B shows the concentration-response relationship for the three ATP concentrations tested in the scenario of recording purinergic ROCE with untouched internal Ca²⁺ stores, and in the presence of 2 mM Ca²⁺ (BV2 cells never exposed to Ca²⁺-free environment in experiments of Figure 2). ΔF/F₀ was significantly larger at 1 mM ATP than at lower concentrations. Due to rare detection in highly resolved two-photon experiments, we further studied Ca²⁺ oscillations in a setting of low-resolution, high-content screening (see below).

The broad-spectrum tetracycline antibiotic, minocycline, is a known microglia activation inhibitor (Wang W. et al., 2015), with inhibitory action on P2X4 and P2X7 receptors (Stebbing, unpublished observations). To further explore its mechanisms of action on purinergic receptor activation and Ca²⁺ homeostasis, we performed two-photon Ca²⁺ fluorescence experiments in the presence of 0.1 mM minocycline and varying ATP concentrations. Pre-incubating BV2 cells with minocycline completely abolished the high [ATP]-induced response (1 mM), but not the ROCE oscillatory responses to lower concentrations (0.1 mM) (Figure 3A). However, even at the lower concentration, ATP-evoked ROCE ΔF/F₀ levels were already significantly reduced compared to the Ca²⁺ responses seen at 0.1 mM ATP under control conditions (no inhibitors present) and also remained well below the levels observed after Ca²⁺ permeabilisation with ionomycin (Figures 3A,C). Minocycline alone did not induce any Ca²⁺ responses (Figure 3A). To address potential PLC pathway activation, we tested the PLC inhibitor U73122. The Ca²⁺ response in BV2 cells to either low or high ATP concentrations were also inhibited by incubation with 5 μM U73122 (U73343 not tested here) as compared to controls without inhibitor present, but were not significantly different under low- or high-ATP conditions (Figures 3B,C; all experiments with 2 mM external Ca²⁺). To test whether minocycline or U73122 affected SOCE, we conducted similar experiments to those outlined in Figure 1. As shown in a representative recording in several BV2 cells, neither minocycline nor U73122 blocked SOCE in BV2 cells (Figure 3D).

To verify the effect of microglia activation on recruitment of SOCE at the protein level, quantitative Western blot analyses were carried out on BV2 cell lysates after differential treatment with microglia activators (LPS, ATP) and M1 polarization (minocycline) inhibitors. Figure 4A shows a representative gel and corresponding blots probed for Stim1S and Stim2: both are present in control BV2 cells, and 1 mM ATP stimulation appears to increase the amount of Stim1S. Compared to controls, only 1 mM ATP significantly upregulated Stim 1S protein content, and only minor or no changes were observed with 0.1 mM ATP or LPS, respectively (Figure 4B). Also, pre-treatment with minocycline blunted the upregulation of Stim1S. None of the microglia activation treatments (LPS, ATP) produced any upregulation in protein content for Stim 2 (Figure 4C).

Although the above shown two-photon excitation approach provides a high-resolution metrology to study Ca²⁺ homeostasis in single BV2 cells, in particular less frequent event patterns, like Ca²⁺ oscillations seen during ATP BV2 activation, are difficult to statistically evaluate. To address this and to enable a high-content screening for Ca²⁺ responses, we established an assay for single cell-based, high-content, functional screening of ATP-induced [Ca²⁺]_i responses.

To evaluate the fraction and concentration-dependence of ATP-induced ‘intermediate [Ca²⁺]_i oscillations’, BV2 cells were prepared for screening experiments in 384-well plates. Figure 5A shows a selection of representative [Ca²⁺]_i responses of the three classes Oscillation & Peak (upper panel), Peak (middle panel), and Other (lower panel). Figure 5B shows the fraction statistics of BV2 cells exposed to 0 (n = 3 replica), 1 (n = 10), 10 (n = 10), 100 (n = 11), and 1,000 μM ATP (n = 12), assigned to one of the three classes. The average fraction of cells assigned to Other decreases with increasing ATP concentration, whereas the average fractions of cells assigned to Oscillation & Peak and Peak increase with increasing ATP concentration. These data clearly demonstrate that the established screening assay enables parallel recordings of ATP-induced [Ca²⁺]_i responses in many BV2 cells and functional phenotyping and quantification of the fraction of cells displaying ‘intermediate oscillations.’ This high-content assay is thus, more robust in quantifying low probability intermediate oscillations as compared to the high-resolution but low-content, two-photon experiments when recording from less cells.

To assess the applicability of the established screening assay for employing ATP-induced ‘intermediate [Ca²⁺]_i oscillations’ as a readout for microglia activation-targeted compound screening, BV2 cells were incubated with one of six selected drugs (minocycline, U73122, U73343, wortmannin, LY294, AZ106) for at least 30 min. [Ca²⁺]_i responses were recorded after application of the agonist ATP (1 mM). Figures 6–8 are representative images of BV2 cells shortly (∼10 s) after ATP application. They illustrate the effect of drug pre-incubation on Fluo-4 intensity and elongation of cells exposed to control solution. Respective drug concentrations are also shown in the bar graphs. Recorded time-resolved fluo-4 signals were automatically assigned to one Oscillation & Peak, Peak or Other classes and phenotype fractions were calculated from four independent experiments each. Averaged values (±SD) as shown in Figures 6–8 are listed in Table 3.

Our results demonstrate a large transient Ca²⁺ entry in BV2 cells after store-depletion. The amplitude of this SOCE was approximately ten times larger than the increase in Ca²⁺ fluorescence during store depletion with thapsigargin, which in turn is about two times larger than the relative increase after store depletion by CPA in primary microglia (Michaelis et al., 2015). Purinergic ROCE followed a clear ATP concentration-dependence with a marginal sustained response to 50 μM ATP, and a more complete sustained response to 1 mM ATP. The former is indicative of P2X4 activation, the latter of P2X7 activation (James and Butt, 2001). Not activating SOCE (without TG), also did not evoke Ca²⁺ entry when Ca²⁺ was added to a Ca²⁺-free solution. Interestingly, in this situation, purinergic ROCE responses were dampened compared with purinergic responses after SOCE (Figure 1C). This indicates an interaction between both processes. Alternatively, it may also indicate some residual contaminating contribution of SOCE to the global Ca²⁺ entry carried by ROCE after ATP application. However, such a contamination may be minor at best, as SOCE transients had already mostly declined before activation of ROCE (Figure 1A).

In primary microglia cells from Stim1/2 wt (^+/+) and knockout (^-/-) mice, where ablation of SOCE was also substantially reduced, a similar situation as in our BV2 cells was seen, with a sustained late ATP-induced ROCE response, most probably mediated by P2X, whereas P2Y receptor activation mediated an early transient Ca²⁺ spike, also present in Ca²⁺ free solution (Michaelis et al., 2015; their Figure 4). However, the latter represents a striking difference to the ATP-responses seen in our BV2 cells under Ca²⁺-free conditions, where ATP did not evoke any increase in intracellular Ca²⁺, indicative of blunted P2Y response in BV2 cells (Figure 1B). Although P2Y receptor expression has been recently confirmed (P2Y6 isoform; Zhu et al., 2015), we are not aware of any study that explicitly tested for Ca²⁺ liberation from the endoplasmic reticulum upon ATP-activated, P2Y-mediated PLC-pathways in BV2 cells. Although under our external 2 mM Ca²⁺ containing single cell experiments, where inhibition of PLC by U73122 did not have a significant effect on Ca²⁺ levels comparing low (0.1 mM) or high ATP (1 mM), it already significantly reduced those levels compared with controls (Figure 3C). While this may point to some involvement of P2Y-response to ATP, the results of missing Ca²⁺ responses in Ca²⁺-free external solution (Figure 1B) provides a stronger argument against a major involvement of those receptors in purinergic responses in BV2 cells. In that regard, BV2 microglia cells might be particularly suitable to study P2X-mediated purinergic Ca²⁺ responses and to elucidate putative cross-talks between SOCE and ROCE. Such a cross-talk is also compatible with P2X7 activation by 1 mM ATP (but not P2X4 or LPS activation) significantly upregulating Stim1S protein content, without changing Stim2 (Figure 4). Moreover, a fast translational control over Stim1 protein expression within ∼6 h treatment of BV2 cells with 1 mM ATP, sufficient to initiate protein upregulation, is apparent from our results. These findings suggest that upregulation of Stim1 protein abundance may be an additional regulatory mechanism of SOCE, which primarily involves the translocation of Stim1 along the ER membrane to juxtapose Orai channels to trigger SOCE. Treatment with LPS for such periods is obviously not enough to induce Stim overexpression which seems different to primary microglia treated with LPS for periods of 3–24 h, resulting in a ∼2.5-fold overexpression of Stim1 (Heo et al., 2015). An interesting similarity between our ROCE responses in BV2 cells and primary microglia cells (Michaelis et al., 2015) is reflected by their averaged Ca²⁺ trace profiles (from up to 30 cells) suggesting a more sustained ATP-mediated P2X Ca²⁺ response in Stim1^+/+ primary microglia, while this seemed more transient for Stim2^+/+ primary microglia (Michaelis et al., 2015; their Figures 2B,C, respectively) when stimulated with a fixed ATP concentration of 0.3 mM. Although not systematically investigated by them, one may speculate that the ROCE of sustained type associated with high ATP concentrations (from 0.1 mM) in our study may be related to interaction with Stim1 while the more transient and even oscillatory ATP-ROCE responses at lower ATP (<0.1 mM) may be regulated by Stim2. Thus, P2X4 may interact with the Stim2 machinery while P2X7 may do so with Stim1. Future studies using isoform-specific knockdown of Stim in BV2 are required to address this hypothesis. Nevertheless, using 0.1 mM ATP to stimulate ROCE, Heo et al. (2015) also observed sustained Ca²⁺ responses in their murine primary microglia model (Heo et al., 2015).

The broad-spectrum antibiotic minocycline inhibits microglia activation during brain inflammatory processes (Wang N. et al., 2015; Henry et al., 2008). It also blocks LPS-stimulated pro-inflammatory cytokine secretion in BV2 cells and reduces microglial toll-like-2 receptor expression (Henry et al., 2008). One potential mechanism of minocycline preventing pro-inflammatory microglial M1 polarization is through inhibited NF-κB upregulation (Kobayashi et al., 2013). Because purines released from inflamed or damaged parenchyma sustain M1 polarization via purinergic activation, its link between microglia inflammasome activation and Ca²⁺ signaling appears obvious. However, to date, this action of minocycline remained undetermined. Therefore, our results provide an important element to understand the actions of minocycline in microglia. We suggest minocycline prevents M1 polarization by blocking P2X4- and P2X7-mediated ROCE, which inhibits Ca²⁺-induced activation of pro-inflammatory pathways and the inflammasome (Stebbing, unpublished observations). From our study, minocycline appears to have higher blocking efficiencies for purinergic Ca²⁺ influx via P2X7 over P2X4 receptors (Figure 3C).

In Ca²⁺-containing external solutions in which SOCE is not activated, purinergic ROCE was ATP concentration-dependent and maximal for high concentrations (≥1 mM), activating P2X7 receptors (Figure 2B). Selective Ca²⁺ permeabilization of the membrane confirmed no additional increase in fluo-4 fluorescence. In particular, for lower ATP concentrations, early Ca²⁺ oscillations occasionally occurred before the delayed sustained rise in Ca²⁺ fluorescence. These oscillations were rarely observed in our high-resolution two-photon imaging at high ATP concentrations, which suggests an association with P2X4 (but not P2X7) activation. Ca²⁺ oscillations have been reported to occur spontaneously in BV2 cells as a result of cell-cell communication mediated by extracellular messengers, i.e., ATP (Wu et al., 2013). ATP-evoked Ca²⁺ oscillations are not well characterized in microglia. In a recent study in microglia cells from a murine Alzheimer’s disease model, ATP induced a more transient response, whereas spontaneous signals co-occurred with Ca²⁺ oscillations in adjacent neurons and astrocytes (Brawek et al., 2014). As differential Ca²⁺ responses in microglia may be of paramount importance for differentially regulating purinergic response patterns, we sought to establish a classification system of ROCE-responses in a high-content setting. This has the advantage of more reliably following rather rare events that may be missed in single cell multiphoton recordings, because the latter focus on high spatial resolution in individual cells rather than on observations of large numbers of cells. During our multiphoton recordings, it already became apparent that there were at least two distinct Ca²⁺ responses (classes): Oscillations & Peak and Peak. We expanded the classes to include a third, Other, which summarized all other events. The assay has been applied here in a proof-of-concept study including ATP concentration-response experiments and functional imaging in cells exposed to a selection of drugs selectively perturbing ATP-dependent Ca²⁺ signaling. Our results revealed that the fraction of cells assigned to one of the three functional classes changed with increasing ATP concentrations, demonstrating that the experimental and analytic approaches for studying distinct ATP-dependent Ca²⁺ signaling in BV2 cells were applicable. The effects of minocycline, U73122, U73343, wortmannin, LY294 and AZ106 were subsequently tested on the Ca²⁺ responses.

Minocycline was reported to inhibit microglia activation by blocking NF-κB nuclear translocation (Giuliani et al., 2005). Our data from both the two-photon experiments and high-content studies indicate it directly inhibits early P2X-induced purinergic ROCE in BV2 cells. The observation of reduced and increased fractions of the Oscillation & Peak and Peak classes, respectively, at a concentration of 100 μM minocycline in large numbers of BV2 cells suggests the drug inhibits P2X4 receptors (Stebbing, unpublished observations). These results also indicate that oscillatory and sustained Ca²⁺ signals cannot be unambiguously attributed to either P2X4 or P2X7 receptors (Figure 6). Such assignment to either channel requires future work involving specific P2X-receptor knockout models or transient siRNA knock-down in expression systems.

U73122 affects ATP-induced Ca²⁺ signaling by inhibiting the transient [Ca²⁺]_i increase which involves PLC to release Ca²⁺ from intracellular stores (Takenouchi et al., 2005). Furthermore, U73122 inhibits P2X7-mediated sustained [Ca²⁺]_i increase, i.e., those [Ca²⁺]_i responses that are mainly represented by the Peak class and partly by the Oscillation & Peak class. The pronounced reduction in the Oscillation & Peak and Peak classes (Figure 7C) confirms this assumption. As shown in Figure 7C, the fraction of cells assigned to the Peak class is significantly reduced compared with controls. Interestingly, inhibiting PLC-pathways with U73122 and blocking P2X7 with minocycline did not affect SOCE in BV2 cells (Figure 3D). The former is consistent with our finding that probably, P2Y-activated, PLC-mediated pathways are not, or at least to a negligible degree, present in BV2 cells (Figure 1B), in contrast to primary microglia cells (Michaelis et al., 2015). Although not explicitly tested in primary microglia cells yet, one could extrapolate to strong blocking effect of U73122 on the P2Y-mediated transient Ca²⁺ release from internal stores measured in Ca²⁺-free solution (for example Figure 2A, Michaelis et al., 2015).

Wortmannin is a natural inhibitor of phosphatidylinositol 3-kinase and enhances microglia-mediated brain inflammation by promoting lipopolysaccharide (LPS)-induced NO release (Pyo et al., 2003). Besides a slightly increased fluorescence signal of cells treated with the drug (Figure 8), the data indicate no effect on ATP-induced Ca²⁺ signaling compared to controls. However, the presence of wortmannin resulted in reduced overall Ca²⁺ signals as compared to data with LY294 and AZ106 which is in line with a previous report on the suppressing effect of wortmannin on P2X4 in BV2 cells (Bernier et al., 2008).

LY294 is a morpholine-containing chemical compound and potent phosphoinositide 3-kinase inhibitor (Maira et al., 2009). The data in Figure 8 suggest perturbation of ATP-induced Ca²⁺ signaling in cells displaying a sustained [Ca²⁺]_i increase only, but not in cells showing ‘intermediate Ca²⁺ oscillations’ plus sustained [Ca²⁺]_i increase.

AZ106 is a selective P2X7 antagonist and inhibits ATP-induced Ca²⁺ signaling (Bhaskaracharya et al., 2014). This is reflected by a strong decrease of cells with Oscillation & Peak and Peak classes, and an increase in the number of cells assigned to the Other class.

A model of the mode of actions of purinergic receptors (P2X, P2Y), and their involvement in downstream pro-inflammatory signaling and inhibitors used in this study is given in Figure 9.

In summary, our study shows the existence and crucial contribution of SOCE to ROCE in immortalized murine BV2 microglia. Interestingly, ATP appears to more selectively activate P2X than P2Y receptor responses in BV2 cells. Our ‘low-resolution’ high-content platform provides a more robust representation of purinergic Ca²⁺ signaling to overcome constraints of cell-to-cell variability in single cell high-resolution experiments. It will be useful for identifying compounds modulating microglial activation in the context of neuro-inflammation. To further develop our high-content screening platform to understand mechanisms of different purinergic phenotypes, future studies will include expression systems with defined receptor expression profiles, BV2 cells and primary cells, for comparison.

Although there are numerous examples of cell signaling studies carried out on BV2 cells alone, it is important to mention that some data obtained with cell lines may be prone to artifacts due to the immortalization procedure and may differ from data observed in primary cells. Thus, data reported in this paper are confined to BV2 cell and transferability has to be considered preliminary until cross-validation in primary microglia cells.