High-throughput Measurement of Plasma Membrane Resealing Efficiency in Mammalian Cells

Cell counting accuracy: HeLa cells are frequently used as a model mammalian cell line to explore membrane repair mechanisms. When assessing membrane repair at the cell population level, it is important to plate cells at the same concentration in all wells for proper data interpretation. It is also important to verify at the time of the assay that cell numbers are equivalent across wells. HeLa cells that constitutively express histone 2B fused to GFP (H2B-GFP) were introduced in this assay to automatically enumerate cells based upon detection of their fluorescent nuclei. To establish the accuracy in cell enumeration, twofold serial dilutions of HeLa H2B-GFP cells were plated in triplicate in a 96-well plate and cultured for 4 h. This amount of time is sufficient for cell attachment and provides limited cell division. Full wells were imaged under transmitted light (TL) and GFP fluorescence illuminations and cell counts were assessed based on GFP fluorescence using the plate reader analysis software (Figure 2A and 2B). The average cell counts ± standard deviation was plotted against cell seeding concentrations, and a line of best fit indicated a 1.08:1 ratio of cell count to cell seeding, demonstrating the accuracy of the counting (Figure 2C). By imaging all of the wells prior to a kinetic assay, it can be ensured that cell numbers are consistent among all wells. Also, by imaging wells after the kinetic assay, one can establish if exposure to the toxin caused cell detachment.

Expression of GFP does not interfere with propidium iodide (PI) intensity measurements (I_PI): to ensure that H2B-GFP nuclear association does not interfere with PI incorporation or PI fluorescence intensity measurement (I_PI), I_PI was compared in HeLa and HeLa H2B-GFP cells that were exposed, or not, to 1 nM LLO (Figure 3). In the absence of PI, there was a similarly low level of background fluorescence in HeLa and HeLa H2B-GFP, indicating that GFP does not bleed through PI fluorescence emission filters. In the presence of PI, but absence of LLO, there was a similar basal PI fluorescence emission in both HeLa and HeLa H2B-GFP that did not change over time. This confirmed that expression of GFP does not affect measurement of PI fluorescence and indicated that PI does not penetrate non-damaged cells over the time frame of the experiment. Addition of LLO resulted in an increase in PI fluorescence over time that was similar in both HeLa and HeLa H2B-GFP. This increase is due to cell wounding by LLO combined with PI association with nucleic acids. Together, these results establish that expression of histone-2B-GFP does not affect PI incorporation or measurement of its fluorescence.

PI fluorescence does not interfere with GFP-based cell counting: Reciprocally, it was important to verify that PI nuclear incorporation in wounded cells does not interfere with GFP-based cell counting. Representative fluorescence images of HeLa and HeLa H2B-GFP exposed to 1 nM LLO in the presence of PI showed that there was a marked accumulation of PI in wounded cells post kinetics, as expected (Figure 4A). Imaging also revealed that PI could bleed through GFP fluorescence detection (Figure 4A and Table 1). This fluorescence crossover was best appreciated on the post-kinetic images of HeLa cells that do not express GFP, but still displayed green fluorescent nuclei (Figure 4A). This crossover could also be evidenced by the measurement of GFP fluorescence intensity (I_GFP) in HeLa H2B-GFP cells, which significantly increased post-kinetic compared to pre-kinetic (Figure 4B). Importantly, PI fluorescence crossover does not affect cell counting because the segmentation process involved in the enumeration of nuclei is unaffected by an increase in GFP fluorescence (Figure 4C).

Measuring plasma membrane resealing efficiency: In this section, we present the basic methodology used to measure the efficiency of membrane resealing. To evidence the process of membrane resealing, HeLa H2B-GFP cells were exposed, or not, to 1 nM LLO in the presence (M1) or absence (M2) of extracellular Ca²⁺ (Figure 5). As expected, in the absence of LLO, I_PI remained constant in M1 and M2. Addition of LLO in Ca²⁺-containing medium resulted in a steady increase in PI fluorescence intensity (I_PI), whereas in the absence of extracellular Ca²⁺, there was a significantly steeper increase in PI fluorescence, reflecting the absence of membrane resealing. To assess the resealing efficiency, which is defined as the capacity of cells to repair in M1 relative to M2 (step 1.5.4.1), the area under the M1 and M2 curves (AUC) were determined and the efficiency of repair (E) was calculated to be 0.287.

An alternative to PI: PI has been ubiquitously used as a marker for plasma membrane damage. However, there are other nucleic acid binding dyes that are also suitable for this assay. For example, a membrane impermeant carbocyanine nucleic acid binding dye (CNABD) exhibits an emission spectrum reaching into the far red and has a high specificity for double stranded DNA. PI on the other hand binds both DNA and RNA^40,41. Unlike PI, the excitation and emission spectra of CNABD does not overlap with those of GFP allowing for better spectral resolution between the two fluorochromes. Furthermore, the CNABD used in this protocol has an extinction coefficient nearly twice that of PI, which means that for their respective excitation wavelengths, this dye is more capable of absorbing energy than PI resulting in a stronger fluorescence emission. Quantitative fluorescence analysis of CNABD and GFP images showed that this dye exhibits a large fluorescence dynamic range, does not significantly emit at the wavelengths of GFP fluorescence, and does not affect cell count (Figure 6A-D). Indeed, HeLa H2B-GFP cells incubated in M1 or M2 media containing CNABD and damaged by 1 nM LLO exhibited a 4- and 5.5-fold increase in I_CNABD relative to the non-damaged controls, respectively (Figure 7A). For comparison, cells exposed to 1 nM LLO in the presence of PI exhibited a 2.5- and 3-fold increase in PI fluorescence intensity in M1 and M2, respectively (Figure 5). Like PI, CNABD exhibits increasing fluorescence intensity with increasing LLO concentration in M1 (Figure 7A), and the resulting repair efficiency was calculated as described in step 1.5.4.1 (Figure 7B). We report that the cell resealing efficiency decreases as LLO concentration increases. This phenomenon reflects the fact that cells decrease their capacity to reseal when excessive damages are caused.

Quality assessment of the membrane repair assay: a critical aspect of any assay is its robustness or capability to detect and resolve differences between the negative and positive controls. Signal variation between positive and negative controls must display sufficient dynamic range and reproducibility. In this membrane repair assay, the positive and negative controls are cells exposed to LLO in repair-permissive (M1) and repair-restrictive (M2) conditions, respectively. Two approaches were taken to assess the robustness of this assay for high-throughput analysis. First, the Z-factor, or screening window coefficient, determines whether a given set of conditions provides a large enough dynamic range, while accounting for signal variability. Z-factors within the ranges 0 < Z ≤ 0.5 and 0.5 < Z ≤ 1 correspond to an acceptable and very good assay, respectively^42,43. A limitation of using the Z-factor for quality assessment is that tested conditions typically exhibit more moderate values compared to the extreme positive and negative controls. Therefore, as a second approach, we calculated the strictly standardized mean difference (SSMD, β), which can identify differences between experimental conditions that would otherwise be categorized as a negative result based on a qualified Z-factor^44,45. SSMD values can be categorized into effect strength ranging from no effect (β = 0) to extremely strong (β ≥ 5). Using the data from Figure 7A and AUC for comparison of M1 versus M2 conditions, exposure to 0.25 and 0.5 nM LLO produced Z-factor values of 0.3100813 and 0.137313, and β = 6.0672 and 4.803308, respectively, indicating that a window of LLO concentrations is suitable for the assay. As the concentration of LLO is increased beyond 1 nM, the gap in the I_CNABD kinetic curves between M1 and M2 conditions closes resulting in drastically reduced Z-factor and SSMD values (Figure 7B). Such high concentrations of LLO correspond to conditions in which the repair potential is outweighed by damage and thus negates the use of the assay in identifying factors involved in membrane repair. This is further illustrated by the decrease in repair efficiency (E) as LLO concentration increases (Figure 7B). All data (Z-factor, SSMD, and E) were generated with 3 biological replicates and 3 technical replicates per experimental condition for assay validation. Together, these data show that this assay has the expected robustness for a high-throughput assay with LLO concentrations inferior to 1 nM for HeLa cells.

Proof of principle: Once the robustness of the assay was established, we performed additional experiments as a proof of principle that this assay has the sensitivity and resolution to identify a defect in the repair process. Also, we considered that high-throughput assays are used as a screening process to identify "hits" within large screens, which may involve less than 3 biological replicates. Thus, it is pertinent that the experimental design provides the detection power to identify "hits" within a single assay. Therefore, under a high-throughput screen, the assay layout can be adjusted to accommodate 4 technical replicates to increase statistical power in a single experiment. Cells were plated in quadruplicate and were pre-treated 1 h prior to the assay with desipramine, a pharmacological inhibitor of the lysosomal protein acid sphingomyelinase (ASM) which plays a role in plasma membrane repair^15,17,46. Importantly, treatment with desipramine did not affect cell counts throughout the assay, allowing for appropriate comparison between desipramine-treated and non-treated cells (Figure 8A). Inhibition of ASM in desipramine-treated cells resulted in a defect in membrane resealing efficiency upon exposure to LLO, as illustrated by the decrease in E and R_Eff (Figure 8B and 8C). Using a mixed effects model, a comparison of desipramine-treated to non-treated cells exposed to 0.25 and 0.5 nM LLO in M1 showed p-values of 0.0010 and 1 x 10^-10, respectively. Together, data indicate that 0.25 and 0.5 nM LLO are appropriate concentrations to identify defects in repair in a high-throughput experimental setting, with possible statistical analyses of a single experiment once the technical replicates are increased to four. Note that the statistical approach of the mixed effects model between quadruplicate will not account for potential variations across multiple biological replicates. Any significant findings using one biological replicate should be verified in additional experimental settings.

Figure 1: Experimental design. The flow diagram depicts a representative plate design configured to test the effect of seven test conditions in comparison to control non-treated cells. Additional controls should be included if appropriate, as for example drug vehicles. Cells are plated (plate 1) 24 h prior to the experiment. On the day of the experiment, cells in plate 1 are washed with M1 or M2 medium pre-warmed at 37 °C, and the plate is imaged (TL, GFP and PI fluorescence) pre-kinetic. During the 15 min of imaging, reagents are added on ice to plate 2. After imaging, plate 1 is immediately placed on ice for 5 min, and 100 μL/well are transferred from plate 2 to plate 1. Plate 1 is placed in the plate reader to run the kinetic assay at 37 °C for 30 min, followed by imaging (TL, GFP, and PI fluorescence). Data are then analyzed to count cells and assess repair efficiency in all experimental conditions. In large data sets, analysis can be automated. Also, the number of technical replicates can be increased to 4 in high-throughput screens. Please click here to view a larger version of this figure.

Figure 2: Cell counting accuracy. HeLa cells expressing GFP-tagged Histone 2B were seeded in triplicates at the indicated concentrations. (A) Cells were imaged at 37 °C under transmitted light (TL) and GFP fluorescence (12 images/well) and a cell detection algorithm was used to delineate individual nuclei (in purple). Scale bar = 1 mm. (B) Higher magnification of TL, GFP, and cell detection (purple) images (zoomed 2X). Scale bar = 100 μm. (C) Image analysis software was used to count the number of cells and the cell counts were plotted against the initial cell seeding concentration. Please click here to view a larger version of this figure.

Figure 3: Propidium iodide fluorescence measurement is not affected by Histone 2B-GFP expression. Histone 2B-GFP expressing and non-expressing HeLa cells were exposed, or not, to 1 nM LLO in the presence (solid lines) or absence (dashed lines) of 30 μM PI in Ca²⁺-containing medium (M1). The kinetic assay measured PI fluorescence intensities by spectrofluorometry every 5 min for 30 min at 37 °C. Data are the average PI fluorescence intensity (I_PI) expressed in relative fluorescence units (RFU) ± SEM (n = 3 independent experiments, each performed in triplicates). Please click here to view a larger version of this figure.

Figure 4: Cell enumeration is unaffected by PI fluorescence. (A) Representative pre- and post-kinetic images (TL, PI, and GFP) of cells exposed, or not, to 1 nM LLO in M1. Scale bar = 100 μm. (B) Quantitative fluorescence microscopy analysis (I_GFP± SEM) revealed the increased GFP fluorescence measurement due to PI nuclear incorporation upon cell wounding by LLO (post-kinetic). (C) GFP-based cell enumeration was unaffected by the increase in GFP intensity. Cell count per well was expressed as average ± SEM. (In B and C: black bars = pre-kinetic data; red bars = post-kinetic data; n = 3 independent experiments, each performed in triplicates; a two-tailed Student's t-test was used to analyze quantitative fluorescence intensity and cell counts from acquired images, **p < 0.01). Please click here to view a larger version of this figure.

Figure 5: Measuring plasma membrane resealing efficiency. Histone 2B-GFP expressing HeLa cells were exposed, or not, to 1 nM LLO in Ca²⁺-containing (M1) or Ca²⁺-free (M2) medium containing 30 μM PI. Kinetic data represent PI fluorescence intensity (I_PI) in relative fluorescence units (RFU) ± SEM, measured for 30 min at 37 °C. n = 3 independent experiments, each performed in triplicates. The resealing efficiency was measured as indicated in protocol step 1.5.4. Please click here to view a larger version of this figure.

Figure 6: Carbocyanine nucleic acid binding dye as an alternative dye to assess membrane resealing. HeLa H2B-GFP cells were exposed, or not, to 0.5 nM LLO for 30 min at 37 °C in the presence of 1 μM CNABD in Ca²⁺-containing (M1) or Ca²⁺-free (M2) medium. (A) Images of HeLa H2B-GFP cells were acquired pre- and post-kinetic in M1 containing the dye. Scale bar = 100 μm. (B and C) Integrated CNABD and GFP fluorescence intensities were measured using the imaging cytometer and expressed in relative fluorescence units (RFU) ± SEM. (D) GFP fluorescence images were processed to enumerate HeLa H2B-GFP cells (black bars = pre-kinetic data, red bars = post-kinetic data, n = 3 independent experiments, each performed in triplicates). A two-tailed Student's t-test was used to analyze quantitative fluorescence intensity and cell counts from acquired images, **p < 0.01, ***p < 0.001) Please click here to view a larger version of this figure.

Figure 7: Effect of LLO concentration on resealing efficiency, Z-factor, and SSMD. (A) HeLa H2B-GFP cells incubated in M1 or M2 containing 1 μM CNABD were exposed to increasing concentrations of LLO and subjected to the kinetic assay for 30 min at 37 °C. Data are expressed as CNABD intensity (I_CNABD) in relative fluorescence units (RFU) ± SEM. (B) The Z-factor and strictly standardized mean difference (SSMD) were calculated as a quality assessment for the robustness of the membrane repair assay, using the area under the curve (AUC) as a metric for the kinetic curves^42,43,44,45. The resealing efficiencies were calculated as described in the protocol and results sections (n = 3 independent experiments, each performed in triplicates). Please click here to view a larger version of this figure.

Figure 8: Cell exposure to desipramine causes defects in resealing. HeLa H2B-GFP cells (plated in quadruplicate) were pre-treated with 30 μM desipramine (or not) for 1 h at 37 °C and then exposed to 0.25 or 0.5 nM LLO in the presence of 1 μM CNABD in Ca²⁺-containing (M1) or Ca²⁺-free (M2) medium. Cells were imaged (pre- and post-kinetic) and subjected to the kinetic assay at 37 °C for 30 min. (A) Cells were enumerated; data are expressed as average cell counts ± SEM. (B) CNABD fluorescence intensity (I_CNABD) is expressed in relative fluorescence units (RFU) ± SEM. (C) Resealing efficiencies were calculated in the presence and absence of desipramine. A mixed effects model was used on log-transformed intensity values assuming a random intercept for each technical replicate. To capture both a shift and change in shape of the kinetic curves, the main effect of treatment condition and the interaction effect between treatment condition and time were jointly tested for statistical significance. A two-tailed Student's t-test was used to analyze cell counts from acquired images. The p-value was calculated using the mixed effects model. (4 technical replicates, one experiment). Please click here to view a larger version of this figure.

Table 1: GFP, PI, and CNABD excitation and emission peaks and the green and red channel excitation and emission bandpasses for the imaging cytometer.