Previous reports have linked the spiking or two-phased character of calcium transients evoked by platelet-derived growth factor (PDGF) to the position of cells in the cell cycle without regard to cell–cell contact and communication. Because cell confluence can regulate growth factor receptor expression and dephosphorylation, we investigated the effect of cell culture confluence and cell cycle on calcium responses of PDGF-BB–stimulated A172 glioblastoma cells.
Digital imaging cytometry was used to correlate the peak and duration of calcium response with bromodeoxyuridine positivity and DNA content and with culture confluence on a cell-by-cell basis.
In serum-starved cultures, complete two-phase calcium signals and shorter, lower spikes occurred independent of cell cycle phase. However, the confluence of cell culture seemed essential for inducing a complete response because cells in sparse cultures exhibited mostly short spikes with lower peaks or no transients at all.
Platelet-derived growth factor receptor (PDGFR) is a receptor tyrosine kinase that plays an important role in the development of tumors of the central nervous system of glial origin, like multiform glioblastoma (1, 2). Ligand binding is followed by autophosphorylation of the receptor (3) and results in activation of various effector enzymes on its tyrosine residues, such as phospholipase-Cγ1 (4), p21ras GTPase activating protein (5), and phosphatidyl-inositol-3-kinase. Activation of phospholipase-Cγ1 leads to increased inositol-1,4,5-triphosphate and diacylglycerol levels, with the former mobilizing calcium from intracellular stores (6, 7). Another pathway leading to high intracellular Ca2+ levels upon PDGFR stimulation is the production of ceramide, sphingosine, and sphingosine-1-phosphate upon sphingomyelinase activation (1, 8). Sphingosine and sphingosine-1-phosphate release Ca2+ from the thapsigargin-sensitive intracellular pools independently of inositol-1,4,5-triphosphate (9, 10).
Ca2+ is a second messenger that is critically important during cell cycle progression. In many eukaryotic cells, growth factors and hormones that trigger the phosphoinositide pathway evoke a biphasic increase in intracellular free Ca2+ concentration: an initial transient release of Ca2+ from intracellular stores is followed by a sustained phase of Ca2+ influx. Blocking Ca2+ influx through the cell membrane or its mobilization from intracellular stores stops cells from entering G1 and S phase (11). Different patterns of Ca2+ signals may be responsible for different “messages” transmitted to the cell nucleus (12). Voltage-dependent channels may contribute to the influx phase of Ca2+ signals (13). In mouse fibroblasts PDGFR activation opens T-type voltage-gated Ca2+ channels (14). However, in PDGF-BB–stimulated glioblastoma cells blockers of L-type Ca2+ channels have no effect on the second, influx-based phase of the Ca2+ response (15); rather, depletion of intracellular Ca2+ stores seems to trigger Ca2+ channel opening (16). This is coherent with the currently accepted concept that growth factor and neurotransmitter-induced calcium influx is generally store dependent and is required for controlling various Ca2+-dependent processes such as synaptic secretion or cell division (17, 18).
The effects of growth factors have been shown to depend on the position of a cell in the cell cycle. In G1 phase, the absence of these factors causes reversible transition to G0 (19), and then the cells require growth factors to re-enter to the cell cycle (20, 21). It was previously shown that the expression level of PDGFR is also cell cycle dependent and regulates entering into G1 phase from G0 (22). However, PDGFR expression levels did not change significantly with the cell cycle in other studies (23).
Stimulation of PDGFR-expressing transformed oligodendrocytes and human embryonic kidney 293 cells with PDGF-BB homodimer can induce oscillatory and nonoscillatory Ca2+ responses (24), with the two types being uniformly distributed in asynchronously proliferating cell cultures and being exclusive for any given cell. In G0 arrest, the nonoscillatory Ca2+ signal was the most common; in G1 arrest, the oscillatory type occurred in a larger percentage of cells. Cells blocked at the G1/S border responded to PDGF-BB homodimer in a nonoscillatory manner, but upon entering S phase the percentage of oscillatory responses increased significantly. In G2 and M phases, cell cycle had no effect on the pattern of Ca2+ response (23). It has been suggested that cell cycle–dependent levels of sphingosine and sphingosine-1-P could be responsible for the oscillatory and nonoscillatory Ca2+ responses, respectively, of PDGF-stimulated cells (24, 25).
A control mechanism of cell proliferation in multicellular organisms is the inhibition of cell growth by cellular contact even in the presence of growth factors. PDGFR is a substrate of low-molecular-weight protein tyrosine phosphatases that control PDGFR-triggered pathways (26) and play a role in contact inhibition according to experiments showing that total protein tyrosine phosphatase activity increases in high-density cell cultures (27). In accordance with this, decreased tyrosine phosphorylation of β type PDGFR can be detected in confluent as opposed to sparse cell cultures (28).
A172 human glioblastoma cells express relatively large amounts of PDGF-BB and PDGFRβ that is constitutively activated by PDGF-BB in an autocrine loop (29). These cells, when confluent in culture, respond to PDGF-BB with a two-phase calcium transient consisting of intracellular release and store depletion-dependent calcium entry from the extracellular space (15, 16). However, even in confluent cultures, some cells show short spikes of intracellular calcium release or no change of calcium concentration after PDGF-BB stimulation. These cells appear to be more separated from their neighbors and are possibly in or entering M phase (15). In the present study, we examined whether cell cycle actually influences the type of calcium response exhibited by these cells, thereby causing G2/M cells to produce short calcium spikes or no peaks at all. Further, in light of the observations that PDGFR phosphorylation may be regulated by cell confluence, we investigated whether the lack of confluence and consequential cell–cell communication could be a cause of altered calcium response in cells without neighbors. Our results suggest that in A172 glioblastoma cells the complete two-phase calcium signal can occur independently of the G0/G1 or G2/M cell cycle stages of the cells. However, the confluence of cell culture seems to be a necessary condition for this complete response because cells in sparse cultures exhibit mostly short spikes with lower peaks or no transients at all.
MATERIALS AND METHODS
PDGF-BB homodimer was graciously provided by Dr. Glenn Pierce (AmGen Inc., Thousand Oaks, CA, USA) and used at a final concentration of 20 ng/ml, a dose known to cause two-phase calcium responses in confluent cell cultures (15, 16). Indo-1 free acid and Indo-1-AM were from Molecular Probes (Eugene, OR, now Invitrogen, Carlsbad, CA, USA), ionomycin from Calbiochem (San Diego, CA, USA), and all other reagents, unless otherwise indicated, were from Sigma-Aldrich (St. Louis, MO, USA).
A172 human glioblastoma cells (30) (American Type Culture Collection, Rockville, MD, USA; catalog number CRL-1620) were cultured in a humidified atmosphere with 5% CO2 at 37°C in Dulbecco's Minimal Essential Medium (Sigma) supplemented with 10% fetal calf serum (Sigma) and without antibiotics. Cells were propagated every 3 to 4 days. For experiments, cells were seeded at low density (20,000/cm2) onto coverglass chambers (Nunc, Naperville, IL, USA) or onto coverslip slices cut with a diamond pencil and used at a confluency of 70% to 80% (usually after 5 days) or less than 20% (after 2 days). Before stimulation experiments, the cultures were starved in serum-free Dulbecco's Minimal Essential Medium for 8 h.
Loading of Cells With Indo-1 Calcium Indicator
Starved cells were incubated with 2 μM Indo-1-AM added from a 1-mM stock solution of dimethylsulfoxide for 50 min at 37°C. Cells were then washed three times with warm HEPES buffer (HB; containing 20 mM HEPES, 123 mM NaCl, 5 mM KCl, 1.5 mM MgCl2, 1 mM CaCl2, 5 mM Na-pyruvate, pH 7.2) and incubated for another 30 min to allow for complete hydrolysis of the dye, which was confirmed by complete quenching of the dye fluorescence by Mn2+ in the presence of ionomycin. For measuring cells in suspension, Indo-1–loaded cells were trypsinized and washed twice with HB using centrifugation at 700g.
Spectrofluorometric Measurement of Intracellular Free Calcium Concentration
Coverslip slices were fitted at a 30° angle to the excitation beam in 10 × 10-mm polystirol cuvettes and measured in a Perkin-Elmer MPF 44-B spectrofluorometer. Alternatively, cells in suspension at a concentration of 105/cm3 (in 2.5 cm3) were placed into the cuvettes. Excitation was at 360 ± 8 nm, emission was measured using a custom-built T-setup at 405 ± 8 nm through a monochromator and at 485 ± 20 nm through a bandpass filter. The ratio of emission intensities was saved with a time resolution of 0.5 Hz. PDGF-BB was added at 20 ng/ml final concentration and gently mixed into the cuvette by pipetting a few cycles.
Image Cytometric Measurement of Intracellular Free Calcium Concentration
Intracellular Indo-1 fluorescence was digitally imaged using an ACAS 570 stage-scanning laser fluorescence cytometer (Meridian Instruments Inc., Okemos, MI, USA) in a humidified thermostated atmosphere. Excitation using the 350- to 360-nm band of an Ar-ion laser was done through a 380-nm dichroic mirror and an Olympus DAPO 40×/1.3 numerical aperture oil-immersion objective. Emission was split with a 450-nm dichroic mirror and detected using two photomultiplier tubes behind 405 ± 20 and 485 ± 20 nm bandpass filters. The ratio of background subtracted emissions was used to calculate free calcium concentration.
Using two-dimensional scans with 0.5-μm pixel resolution, fluorescence images of both emission wavelengths in 240- × 220-μm areas were taken at a frequency of 0.1 Hz (Fig. 2a). The emission ratio was converted into absolute calcium concentration using a calibration curve and plotted as free Ca2+ concentration versus time after integrating inside each region of interest (ROI) corresponding to a cell of interest (Figs. 2a and 2b). PDGF-BB was added at 20 ng/ml final concentration in 0.25 ml volume to allow for quick mixing with the buffer already over the cells.
Calibration of Calcium Measurements
To be able to convert Indo-1 fluorescence ratios into absolute free calcium concentrations (31), 5 μM Indo-1 (pentapotassium salt) was added to buffers simulating the intracellular environment in terms of ion concentration, pH, and viscosity/polarity and containing various concentrations of free calcium. The buffer composition was 10 mM 2-(N-morpholino)ethane sulfonic acid, 115 mM KCl, 20 mM NaCl, 1 mM MgCl2, 1 mM EGTA (ethylene glycol bis [beta-aminoethylether]-N,N′-tetraacetic acid), and 40% glycerol, pH 7.4. Calcium concentration spanning a 10- to 2,000-nM range was set using 0.1 to 1.5 mM CaCl2. The solutions were imaged with instrument settings identical to cell imaging and averaged, background-subtracted emission ratios were matched to free calcium concentrations calculated using a Kd of 105 nM for Ca2+ and EGTA in the buffer used.
Bromodeoxyuridine Incorporation and Measurement of Cellular DNA Content
Before calcium measurement, cells were incubated with 10 μM 5-bromo-2′-deoxyuridine (BrdU) for 1 h at 37°C. After having measured the calcium response to PDGF, cells were washed twice in HB, once in phosphate buffered saline (PBS), and fixed in 70% ethanol (in PBS) at 4°C for 20 min, followed by a 1-h incubation in denaturing solution (2.5 M HCl/0.5% Triton X-100), two washes in PBS/0.1% Triton X-100, blocking with 1.5% nonfat dry milk in PBS/0.1% Triton X-100, and two washes again. Monoclonal mouse anti-BrdU at 20 μg/ml in PBS was then used overnight at 4°C, followed by washes (PBS/0.1% Triton X-100), incubation with 10 μg/ml FITC-anti-mouse goat immunoglobulin G for 2 h at 25°C, two washes, incubation with 1 μg/ml propidium iodide (PI) for 10 min, two washes, and mounting in Mowiol (0.1 M Tris-HCl, pH 8.5, 25 w/v% glycerol, and 10% Mowiol 4-88, Hoechst Pharmaceuticals, Frankfurt, Germany) (32). Samples were imaged with an ACAS 570 cytometer using 488-nm laser excitation, 490-nm dichroic mirror, and the standard FITC and PI filter sets in front of the two emission photomultiplier tubes.
Calcium responses of individual cells were plotted as a function of the average free calcium concentration within cell boundaries against time. For each cell, peak and duration of the calcium signal was recorded, as well as its position on the coverslip chamber. After labeling for BrdU and PI, the coverslip chamber was relocated to the same position and images of anti-BrdU indirect fluorescence and PI fluorescence were recorded for the very same cells. BrdU and PI signals were integrated within each cell boundary and correlated with the peak and duration of calcium response.
Because peak and duration values of calcium response did not follow a normal distribution, comparison of these values in cell subsets was done using the Mann-Whitney rank sum test or, when more than two categories were established, the Kruskal-Wallis one-way analysis of variance on ranks, followed by Dunn's pairwise multiple comparison procedure adjusted for ties. Correlation of the peak and duration of calcium response with each other and DNA or BrdU content was tested using Spearman's rank correlation test. Correlation of peak and duration with confluence and cell cycle data (BrdU positivity, phase) was checked using binary or multinomial logistic regression procedures. To test whether the different outcomes in any given classification category (DNA, BrdU, confluence) were randomly distributed over the outcomes of the other categories, chi-square test was used.
RESULTS AND DISCUSSION
Suspending Otherwise Adherent Cells Shortens Calcium Response to PDGF
To test the effect of cell–cell communication on large populations, a trivial approach is to use spectrofluorometry for measuring free cytoplasmic calcium concentration. Figure 1 shows typical traces of fluorescence ratios (F485/F405). Cell populations adhering to the coverslip (solid line) responded after a lag to PDGF-BB stimulus by a rise of ratio (corresponding to a rise of free Ca2+ concentration) (31), followed by a slow fall to a concentration maintained above resting levels for at least 700 s. EGTA added at the end of the measurement caused a drop in ratio to below resting levels, indicating that membrane Ca2+ channels are still open to maintain free calcium concentration above resting levels. Restoring extracellular Ca2+ caused the elevation of intracellular concentration again. In the case of cells from a matching culture, but trypsinized and suspended (dotted line), the lag phase appeared to be longer, and the rate of rise slower, which possibly refers to the lack of synchronization caused by the loss of cell–cell contacts. Also, after reaching a peak, the fall in concentration is faster than in the case of adherent cells, and resting level is reached by 300 s. Probing with EGTA and re-addition of calcium revealed that membrane calcium channels were closed at this last stage. Although these observations support the notion that cell–cell contact provided for by cell confluence is necessary to achieve a complete, long, two-phase calcium response in these cells, one needs to consider the possibility that it is the adherence of cells to substrate, rather than their contacts with each other, that is needed for the complete response (33). Furthermore, although spectrofluorometric measurements provide data on large cell populations, they do not permit the investigation of individual adherent cells and correlating their response to stimuli with other properties, such as their position in the cell cycle.
Digital Image Cytometry Allows for Correlating the Dynamics of Individual Calcium Responses with other, More Static Cellular Parameters
To correlate the type of calcium response with cell confluence and cell cycle, we spiked cells with BrdU before measuring intracellular calcium concentration in confluent and sparse cultures. Indo-1 fluorescence was then imaged (Fig. 2a) at 485-nm (detector 1 data) and 405-nm (detector 2 data) emission wavelengths to obtain image pairs at a rate of 0.1 Hz. Image sequences were postprocessed by converting ratio images to free calcium concentration maps, marking cell boundaries for individual cells, and averaging the calcium concentration inside the boundaries for each time point. As apparent in Figure 2b, each cell produced a characteristic response to PDGF stimulus, which could be characterized by its peak value and duration. Cells were then processed to label the incorporated BrdU with indirect immunofluorescence and nuclear DNA with PI. Coverslips were removed from the chamber and mounted on microscopic slides. The area imaged for calcium concentration was relocated in the cytometer and PI and FITC images were taken. In Figure 2c, these two fluorescence values are displayed on a pseudocolor scale in one common image. The ROIs for each cell of interest were retrieved and FITC and PI fluorescences above threshold were integrated within each ROI. On rare occasions, such as cell 5 in the example, cells were lost during the preparation process and had to be omitted from the analysis. Peak value and duration of calcium response for each cell were then analyzed as a function of DNA content (PI fluorescence), BrdU positivity, and cell culture confluence.
Quality of Calcium Transients is Independent of G0/G1 and G2/M Phases
Figure 3a shows the plot of DNA content (PI fluorescence) versus BrdU immunofluorescence signal for all cells investigated. The threshold for BrdU positivity is at 230 (dotted line), clearly separating G1/G0 cells (BrdU signal < 200) on the left of the line from the others that exhibit BrdU signal above 250. This latter population is divided into three categories; early S with PI fluorescence centered around 260, late S with PI fluorescence centered around 520, and mid-S phase for DNA content in between. The frequency distribution histogram in Figure 3b confirms the peak values of DNA content characteristic of G1 and G2 to be 260 and 525, respectively. These values were used to normalize DNA content shown in Figures 3c and 3d. The one cell with high DNA content and no BrdU incorporation is in G2/M that got to that point before BrdU pulsing. For analysis purposes, cells with 4n DNA content, whether BrdU positive or not, were considered G2/M because immediately after measuring calcium response their PI fluorescence is maximal, whereas BrdU pulsing happened 1.5 to 2 h before that measurement. Cells that are BrdU positive but still below 1.1 times the diploid DNA content are considered early S, whereas those above this limit up to 1.8 times the diploid are considered S phase. The distribution of cells into these categories is confirmed by the significant correlation of BrdU positivity and DNA content (chi-square test, P < 0.001).
It appears from Figures 3c and 3d that cells in G1/G0, early S and G2/M can produce short or long calcium responses and higher and smaller calcium peaks. The only exception to this uniform distribution was the set of six cells in S phase, which uniformly produced short calcium spikes with no second (influx) phase. Statistical analysis (for details, see Table 1) corroborated that BrdU-positive and -negative cells did not differ significantly in calcium peak or duration, and coherently with this, no correlation was found between BrdU positivity and calcium response parameters. When considering the BrdU signal a continuous nonparametric variable alongside peak and duration, Spearman's test also failed to reveal significant correlation (R = −0.07, P = 0.56 and R = −0.16, P = 0.21 for peak and duration, respectively).
Table 1. Statistical Analysis of Peak and Duration of Calcium Response as a Function of Cell Cycle and Confluence*
Median peak (μM)
Median duration (s)
Median values of calcium peak and response duration are given with number of events for each category of BrdU incorporation (positive or negative), DNA content (diploid, greater than diploid), position in cell cycle (G1/G0, early S, S, G2/M), and cell confluence (confluent, sparse). The P values of statistical analysis are indicated below the medians. Differences in peak or duration between two categories were tested with the Mann Whitney rank sum test. Differences between the four categories of cell cycle were tested using Kruskal-Wallis one-way analysis of variance on ranks. Correlation of peak or duration with binary outcomes was tested using binary logistic regression; correlation with cell cycle phases was tested using multinomial logistic regression.
Position in cycle
Neither duration nor peak of the calcium response correlated with the DNA content being diploid or greater, and their values were not significantly different in the diploid and greater-than-diploid categories (Table 1). Spearman's test on DNA content as a continuous variable against peak or against duration showed no correlation at all (R = 0.03, P = 0.81 and R = 0.007, P = 0.96, respectively).
When considering the cells' position in the cell cycle as deduced from DNA content and BrdU positivity, Kruskal-Wallis one-way analysis of variance on ranks hinted at some differences between certain groups in terms of peak and even more so in terms of duration of response (Table 1). Therefore pairwise comparisons (Dunn's method) with correction for ties were applied to all four categories (G1/G0, early S, S, G2/M). In this series of tests, peak and duration of response were significantly different between S and G2/M cells (P = 0.05 and 0.01) but not between any other categories in the cycle, indicating that cells in S phase are more likely to produce short and not too high calcium spikes.
Cell Culture Confluence is a Major Determinant of the Characteristics of PDGF-Evoked Calcium Response
In Figure 3c it is obvious that almost all triangle symbols representing confluent cells are in the 500- to 700-s range for duration of response, whereas circles representing cells from sparse cultures are below the 400-s line with one exception. Although not so apparent, a similar tendency can be observed in Figure 3d, where confluent cells tend to produce higher calcium peaks. The difference in peak and duration between confluent and sparse cells is statistically highly significant (P < 0.001), and a good correlation (P < 0.001) can be observed when fitting a logistic regression to confluence state versus peak or duration (Table 1). Because confluent cells are found in equal proportions in BrdU-positive and -negative groups (chi-square test, P = 0.54) and among cells with 2n or 4n DNA content (P = 0.68), the possibility of variations in cell cycle causing these observations can be ruled out. There is only one group of cells, those in S phase, that are mostly (five of six) from sparse cultures, and thus in their case one cannot decide with certainty whether the characteristic spiking response is related more to being in S phase or to lacking cell–cell contacts. However, in all other phases of the cycle, confluence seems to be the sole determinant of the outcome of growth factor stimulus in terms of calcium signalization. It was shown earlier that in neuronal cells pressure induced calcium influx is more pronounced upon reaching cell culture confluence but this was atrributed to G0/G1 blockade paralleled with differentiation (34). However, in our case there were no morphological signs of cell differentiation upon reaching confluence.
Serum-starved adherent A172 glioblastoma cell populations assayed in a spectrofluorometer responded to PDGF stimulus with a rise of intracellular free calcium followed by a level maintained above the resting concentration, whereas the same cells in suspension terminated their response after the initial peak. In single cell measurements, these cells responded with either a two-phase calcium transient, a single calcium peak quickly returning to resting levels, or remained inert. By pulsing cells with BrdU before PDGF stimulus and labeling BrdU and DNA after calcium measurements, we were able to correlate the peak and duration of the calcium response with cell cycle parameters and with cell confluence on a single cell level. Cells in G0/G1 and G2/M have shown all types of responses, which is partially coherent with findings on other PDGF-stimulated cells that showed diverse calcium transients in G2/M, but produced mostly two-phased responses on the border of G1/S (23). As for cells in S phase, our experience is that cells responded with a spike or nothing at all, with the former being parallel to findings on oligodendrocytes and HEK 293 cells (23). However, the parameter that best determines the quality of calcium transient was found to be the confluence of cells. The vast majority of cells in confluent cultures tended to give complete two-phase responses whether in G0/G1 (including very early S) or G2/M (including late S). Conversely, cells in sparse cultures usually failed to produce a two-phase calcium signal and produced a short spike only or did not respond at all. As most S phase cells belonged to this category, it cannot be unanimously ascertained whether in S phase the spikes or absence of response is attributed to cell cycle stage or the lack of confluence. Because cell confluence, by virtue of cell–cell contacts, is assumed to be an important regulator of proliferation, one might wonder what the significance of such a change in calcium signaling would be as the cell culture progresses from sparse to confluent. It has very recently been shown that contact inhibition and reduction in serum concentration select for the same phenotype of cell that increases saturation density and generates transformed foci (35). In view of this finding, one is tempted to speculate that these transformed cells “learn” to give themselves extra growth stimuli upon reaching confluence and being subjected to contact inhibition. In the case investigated here, one such stimulatory pathway would depend on the PDGF–PDGFR autocrine loop (29) exhibited by A172 glioblastoma cells. This pathway, by an unknown regulatory mechanisms, becomes more avid in confluent cultures and gives rise to maintained intracellular free calcium levels necessary for extended stimulation of protein kinase C and calmodulin-dependent kinases.
We thank Dr. Georg Kreysch and Merck KGaA for their generous donation of an ACAS 570 laser scanning cytometer. We also thank Elza Friedländer for valuable discussions.