Wide field of view quantitative imaging of cellular ATP release

Abstract

Although several mechanical stressors promote ATP secretion from eukaryotic cells, few mechanosensitive pathways for ATP release have been precisely characterized and none have been clearly identified. To facilitate progress, we report here a wide field of view (∼20 × 20 mm sample area) imaging technique paired with a quantitative image analysis to accurately map the dynamics of ATP release from a cell population. The approach has been tested on A549 cells stretched at high initial strain rate (2–5 s⁻¹) or swelled by hypotonic shock. The amount of ATP secreted in response to a series of five graded stretch pulses (5–37% linear deformation, 1-s duration at 25°C) changed nonmonotonically with respect to strain amplitude and was inhomogeneous across the cell monolayer. In a typical experiment, extracellular ATP density averaged 250 fmol/mm², but the area of detectable signal covered only ∼40% of the cells. In some areas, ATP accumulation peaked around 900 fmol/mm², which corresponded to an estimated concentration of 4.5 µM. The total amount of ATP released from the combined stretch pulses reached 384 ± 224 pmol/million cells (n = 4). Compared with stretch, hypotonic shock (50%, 30°C) elicited a more homogeneous ATP secretion from the entire cell population but at a lower yield totaling 28 ± 12 pmol/million cells (n = 4). The quantitative extracellular ATP mapping of several thousand cells at once, with this wide field of view imaging system, will help identify ATP release pathways by providing unique insights on the dynamics and inhomogeneities of the cellular ATP secretion that are otherwise difficult to assess within the smaller field of view of a microscope.

INTRODUCTION

Extracellular adenosine 5′-triphosphate was recognized as a signaling molecule several decades ago, and many of its target purinergic receptors of the P2X_1-7 and P2Y_{1,2,4,6,11-14} families have since been sequenced and pharmacologically characterized (3). Yet several aspects regarding the modes of signaling of this nucleotide remain perplexing. In particular, while mechanical deformations of all kinds have been found to stimulate ATP release, no mechanosensitive pathways permitting ATP secretion have been unambiguously characterized as of today (9, 10, 18). There are currently two major acknowledged modes of cellular ATP release: lytic and nonlytic. Cell lysis resulting from plasma membrane rupture allowing the free passage of cytosolic ATP can be induced by physical or chemical stressors; the damages can be transient or lead to cell death (16). The nonlytic mode of ATP release is generally divided into conductive (ionic channels) and nonconductive pathways, with the latter further splitting up into regulated and nonregulated exocytosis (9).

Regardless of the stimulus under investigation, an accurate measurement technique for ATP is necessary to identify the transmembrane pathways used by this nucleotide. The most widely used technique to detect and quantify ATP is the luciferase-luciferin (LL) assay, a photon-producing enzymatic reaction whose light intensity serves as an indicator of ATP content in the solution. An alternative to the soluble luciferase is the pmeLUC cell expression system, featuring a plasma membrane-bound and extracellular-facing luciferase enzyme. It is a versatile tool suitable for in vitro as well as in vivo applications (12). However, in most cell-based experimental assays, the quantity of released ATP is tiny, and either soluble or membrane-bound luciferase approaches yield a very faint bioluminescent signal that is barely detectable with a conventional microscopy imaging system. Therefore, the standard procedure to quantify cellular ATP release generally involves collecting bulk extracellular solution for subsequent offline determination of ATP content using a luminometer or a plate reader. Although such approaches constitute a highly sensitive biochemical assay for ATP detection, it cannot provide information on the spatial kinetics of ATP release, a prerequisite to fully characterize and solve the identity of ATP release pathways.

To accurately map cellular ATP release, it is essential to combine a microscopy setup that has a strong light-gathering power with a very sensitive light detection apparatus. Currently available digital cameras with maximal light sensitivity, such as an electron-multiplying charge-coupled device (EMCCD), although well suited for low-light applications in fluorescence microscopy, are not yet adequate for such extremely low bioluminescent signal. A simple technique to circumvent this weak source of luminescence is to capture light for longer periods of time (8, 11). With this approach, however, the advantages of an imaging system mostly vanish as the spatial kinetics of ATP efflux (light emission) is lost. Image intensifier has been used with success for ATP imaging (6, 17), but it is an expensive and not widely available solution.

The brightness of an image acquired with a microscope increases with the objective’s numerical aperture (NA) but decreases with the lateral magnification (M) (Eq. 1). Since in general the NA of microscope objectives changes in proportion to their magnification, very little brightness can be gained by choosing an objective with the highest NA available. Besides, in most modern microscopes, the need to fit complex optical components within the main optical path results in longer lens-to-lens distance and thereby substantial loss of peripheral and off-axis image-forming rays.

$$\text{Image\hspace{0.17em}brightness\hspace{0.17em}}\propto {\left(\frac{\text{NA}}{\text{M}}\right)}^2$$ (1)

Thus, to boost the light intensity signal, we chose a macroscopic imaging approach similar to the one described by Feranchak et al. (5), but instead of a 1:1 object-to-image magnification, we chose a threefold reduction (M = 0.33) with an objective lens of NA = 0.15. Based on Eq. 1, the theoretical increase in brightness from those optical parameters alone compared with, for example, a ×5/0.25 objective is [(0.15/0.33)/(0.25/5)]² = 83-fold. The true increase in captured luminescence with our imaging system can surpass this ratio, however, due to its shorter optical path compared with nowadays microscope. While this higher light-gathering power came at the cost of reducing the image resolution, our imaging system with its wide field of view (FOV) that spans the whole of our experimental chambers (∼20 × 20 mm), coupled with our approach to map and quantify over time the distribution and rate of secreted ATP in our experimental chambers, provided new insights on the overall nature and dynamics of the cellular ATP secretion otherwise difficult to assess with standard microscopes and their small FOV.

MATERIALS AND METHODS

Cells.

Human lung carcinoma A549 cells were grown on cell culture dishes (Cellstar; Greiner Bio-One, Frickenhausen, Germany) in HyClone High Glucose Dulbecco’s modified Eagle’s medium (GE Healthcare Life Sciences, HyClone Laboratories, Logan, UT) supplemented with 10% fetal bovine serum, 2 mM l-glutamine, 50 U/mL penicillin-G, and 50 μg/mL streptomycin sulfate (antibiotics were only included in media of cells grown for hypo experiments) at 37°C and 5% CO₂. For stretch experiments, trypsinized A549 cells were seeded at ∼700 cells/mm² (~25,000 cells/36 mm²) with culture media on a bottom-stretchable chamber (see below) and kept at 37°C and 5% CO₂ until the time of the experiment. For hypotonic shock, A549 cells after trypsinization were seeded with culture media on 15-mm-diameter glass coverslips and used after they reached 90% confluence (∼400–500 cells/mm²; 24–72 h after plating). All added media constituents were from Gibco/ThermoFisher Scientific (Burlington, ON, Canada).

Stretch experiments.

Stretch chambers were made with a Sylgard 184 silicone elastomer base and curing agent (Dow Corning, Midland, MI) as described previously (6, 7). The elastomer base and curing agent were poured into the stretch chamber’s mold and then baked for 1 h at 60°C. To facilitate the attachment of cells onto silicone, the 2-mm-wide stretchable groove was coated with 20% collagen type 1 from rat tail diluted in distilled H₂O (Sigma-Aldrich, St. Louis, MO). During all stretch experiments, cells were bathed in 200 µL of Gibco’s phenol red-free DMEM (Invitrogen Canada, Burlington, ON, Canada) containing the ATP detection reagents prepared as follows: one vial of ATP assay mix-lyophilized powder (Sigma-Aldrich) was reconstituted with 5 mL of sterile double-distilled water and isotonically adjusted with 5× physiological solution (LL). For ATP imaging, 100 µL of the isotonically adjusted LL was added to 100 µL of Gibco’s phenol red-free DMEM (LL-DMEM).

The stretch chamber mounted on the stage of our custom-designed ATP imaging system (see below) was clamped to the STREX NS-600W Cell Stretching System (STREX, Osaka, Japan). The stretching apparatus applied a horizontal stretch onto the silicone chamber for1 s. All stretch experiments consisted of five consecutive stretch pulses of increasing amplitude at 5-min intervals. Effective applied strains, along the main axis of load application, were measured on images acquired after the experiments with a repetition of the stretch sequence.

Hypotonic challenge.

A549 cells on glass coverslips were placed in a glass-bottomed (25-mm diameter coverslip no. 1) chamber (RC-40LP; Warner Instruments, Hamden, CT), bathed in LL-DMEM solution (500 µL), and warmed at 30°C with a DH-35 Culture Dish Heater (Warner) wired to a TC-324B Automatic Heater Controller (Warner) until hypotonic challenge. To keep the ATP detection reaction reagents at prehypotonic levels, the added hypotonic LL solution (hypoLL) contained the same level of LL reagent as LL-DMEM (for details see Stretch experiments). The luciferase reaction is well known for its sensitivity to ionic content, however (1). For instance, if the ionic concentration of an isotonic medium is lowered by 50%, it will result in approximate doubling of the luciferase reaction speed. As a result, before converting the pixel values into ATP (see ATP calibration and quantification), we divided by 2 the rate of light units (LU) per second. In addition, to keep divalent cation concentration constant during the course of the experiments, the hypoLL also contained (in mM) 1 MgCl₂ and 1 CaCl₂ (Fisher Scientific, Pittsburgh, PA). ATP release was stimulated by adding 500 µL of hypoLL solution to the chamber (∼50% hypo final).

ATP release imaging.

We chose an inverted configuration for the imaging system, with the EMCCD camera Evolve 512 (Photometrics, Tucson, AZ) placed underneath the experimental stage (see Fig. 1A). To position the stretch or hypotonic shock chamber, we used a standard manual mechanical XY translational stage. Focus was provided by moving vertically the camera positioned on a pinion-and-rack system. We selected a parallel path (sometimes referred to as infinity space) in our optical design, but instead of a distance between the two main lenses in excess of 200 mm, as typically used in infinity-corrected microscopes, we shortened it to 40 mm. This shorter distance means that more of the off-axis peripheral rays are captured, hence resulting in enhanced brightness. For the objective, we chose a standard bi-convex lens with f_obj = 75 mm (D_obj = 25 mm, NA = 0.15), and we fixed to the EMCCD camera a standard c-mount convex-type camera lens with f_cam = 25 mm (D_cam = 20 mm) that resulted in ×0.33 (f_cam/f_obj) magnification (3-fold reduction). It should be noted that this particular optical configuration with large-diameter lenses and magnification at or below ×1 is similar to macrophotography. Given the size of the camera chip per the manufacturer’s specifications of 6.7 × 6.7 mm (512 × 512, 13 μm pixel size), the resulting FOV was ∼20 × 20 mm. With such a wide FOV, large-diameter lenses are essential to minimize optical aberrations. Our imaging system is not entirely devoid of optical aberrations, however. In particular, coma aberrations from intense punctual light sources, e.g., from isolated cells suddenly ruptured by lysis, have been observed at the periphery of the field of view during hypotonic assay. Those image distortions are minor and not expected to diminish appreciably the exactness of the ATP quantification but will be addressed in the future.

Fig. 1.

Experimental setup for imaging ATP release from cultured cells. A: a Photometrics Evolve 512 electron-multiplying charge-coupled device (EMCCD) camera is mounted on a pinion-and-rack device for z-axis focus and installed under a translational (x-y-axis) mechanical stage. A pair of lenses with a 3:1 object-to-image ratio and a working distance of ∼70 mm are attached on the camera. Cells are seeded onto either a 2 × 18 mm collagen-coated groove of a flexible silicon chamber for stretch experiments or a 15-mm-diameter glass coverslip for hypotonic shock challenge. During the experiments, cells are covered with DMEM containing LL reagent. The stretch chamber is attached to a stretching apparatus fixed on the movable stage. The hypotonic chamber is fixed directly on the stage and kept warm with an incubator (not shown). An incandescent lamp installed above the stage and centered in the system’s optical axis is turned on only during pre- and postexperiment brightfield image acquisition. B: the coverslip placed in hypotonic chamber and the rectangular stretchable groove area can be visualized in brightfield acquisition mode.

To further amplify the signal, we set the on-chip camera binning at 2 × 2, achieving a final resolution of ∼78 μm/pixel. Control of the EMCCD and data acquisition were done via AxioVision software (AxioVs40 V 4.8.2.0; Carl Zeiss MicroImaging, Jena, Germany). Images were acquired at 0.5-Hz frequency and 1-s exposure time for stretch and for hypo at 0.2 Hz and 5-s exposure and gain settings at the lowest readout speed at 5 MHz. Stacks of 900 images were captured and stored for offline analyses with Metamorph (Meta Series Software 7.7.9; Molecular Devices, Downingtown, PA) or ImageJ open-source software (15).

To acquire brightfield images of the stretch or hypotonic chamber (Fig. 1B), we installed an external variable illumination source (incandescent bulb, 15 W, T6/collector lens f_coll = 25 mm, D_coll = 40 mm) in-line with a standard condenser (f_cond = 60 mm, D_cond = 40 mm, NA = 0.3). We also placed a diaphragm in the parallel optical path for additional light intensity control and fine-tuning for the brightfield-acquired images. Finally, to minimize light background, the entire imaging system was installed inside an opaque box and was accessible through a front door panel.

ATP calibration and quantification.

Extracellular luminescence, in contrast to intracellular fluorescence, is not enclosed within the boundary of cellular membranes; the spatial distribution of the light-emitting reaction is governed by time-dependent diffusion processes and is dispersed throughout the fluid covering the cells. Since the light intensity captured in the image is proportional not only to [ATP]_o but also to the depth (in the object space) of the in situ generated luminescence, an ATP calibration technique independent of the dimensions of the light-generating reaction is required. This is achieved by calibrating the rate of light generation with respect to the quantity of ATP_o molecules rather than its local concentration.

$$\text{Luciferin}+{\text{O}}_2+\text{ATP}\stackrel{\text{luciferase}}{\to }\text{oxyluciferin}+{\text{CO}}_2+\text{AMP}+{\text{PP}}_{\text{i}}+\text{γ}$$ (2)

$$\text{ATP}\left[\text{S}\right]\stackrel{\text{luciferase}\left(V_{\max },K_{\text{m}}\right)}{\to }\text{γ}\left[\text{P}\right]$$ (3)

Although ATP is only a cofactor in the LL reaction and light (γ) a by-product of it (see chemical reaction 2, PP_i: inorganic pyrophosphate), when using this reaction for ATP determination with [luciferin] ≫ [ATP], we can posit ATP as the sole reactant and light as the main product as described in the simplified chemical reaction 3 (the rate of production of oxyluciferin can be directly inferred from the rate of light generation). By using the Michaelis-Menten equation Δ[P]/Δt = (V_max × [S])/(K_m + [S]) and assuming K_m ∼200 µM (4) > [S] = [ATP], we obtain this simplified relation: (Δγ/Δt)/V = (V_max/K_m) × [ATP] (V: volume of reaction), where the rate of photon production (Δγ/Δt) per unit of volume is linearly proportional to the concentration of ATP. Multiplying both sides of the equation by V yields the volume-independent relationship: Δγ/Δt = (V_max/K_m) × ATP, which directly relates the rate of photon emission to the quantity of ATP molecules. We rearranged this equation and combined the enzymatic reaction constants and other parameters pertaining to the specifics of our imaging system into a single calibration factor (see Eq. 4). An example of parameters intrinsically included in this factor is the numerical aperture. The imaging system collects only a fraction of all the photons generated by the reaction. The capture ratio of those photons is determined by the NA, but it remains constant for the entire FOV and is unaffected by the signal intensity. In Eq. 4, the rate of photon generation (Δγ/Δt) is replaced by the pixel intensity value [identified as light unit (LU)] divided by the exposure time (Δt) of the image. The quantity of ATP_o for a region of interest (ROI) in the recorded image covering the whole chamber or a subarea within and containing n pixels can then be obtained from the sum of individual pixel values divided by exposure time and multiplied by the calibration factor:

$$\begin{gathered} {\text{ATP}}_{\text{ROI}}\left(\text{moles}\right)=\text{calibration\_factor}\left(\frac{\text{moles}\cdot \text{s}}{\text{LU}}\right) \\ \times \frac{1}{\Delta t\left(\text{s}\right)}{\displaystyle \sum }_{i=1}^{n\text{\hspace{0.17em}pixels}}{\text{pixel\_value}}_{\text{i}}\left(\text{LU}\right) \end{gathered}$$ (4)

To determine the value of the calibration factor, we first recorded the light intensity generated by known concentrations of ATP (5 nM to 50 µM, in 10-fold increments) within the stretch chamber for three volumes (18, 36, and 54 µL) of solution. The volumes were chosen to achieve heights of liquid of 1, 1.5, and 2 mm. Those height increments serve to validate that a large proportion of the luminescence is captured longitudinally along the optical axis of our imaging system, a property dependent on optical characteristics such as the depth of field. This validation requires that while keeping the ATP concentration unchanged, a step increase in solution height will augment the light signal in proportion. The ATP-Mg₂ dissolved in DMEM was pre-mixed 1:1 with LL-DMEM (2×) and then immediately transferred within the stretch chamber and images acquired as described above (see ATP release imaging) during 1 min. The slowly declining rate of LU per second (minus background level) for the whole chamber was then averaged over the first 10 s. We reported the results for the three series of volume (corresponding height) and [ATP] in Fig. 2A and the calculated quantity of ATP for the 15 combinations of volume and [ATP] in Fig. 2B.

Fig. 2.

Calibration factor determination. A: light signal in ΣLU/s (background corrected) against a series of [ATP] for 3 different heights of solution within the stretchable groove. Note that for a given ΣLU/s value, the [ATP] is undetermined without knowing the height of the reaction. Also, for a given [ATP], the light signal increases proportionally with the height of the reaction: with 1 mm signal as reference, the increases in light signal for all [ATP] are 1.5 ± 0.15 (1.5:1 mm) and 1.9 ± 0.22 (2:1 mm). B: light signal (ΣLU/s) against total ATP. In contrast to A, there is only one corresponding quantity of ATP for a given light signal value. C: calibration factor varies slightly with [ATP]. The average of the measured calibration factor was used to calibrate all ATP experiments (stretch and hypo). LU, light unit.

Note that while there are three possible rates of light generation (1 for each height) for a given [ATP] (Fig. 2A), only a unique value exists for a given quantity of ATP (Fig. 2B). Note also how the ATP-light signal relationship on Fig. 2B displays a strong linear proportionality over several orders of magnitude of ATP molecule content. Another important observation from those calibration measurements is the additive property of the relationship, whereby the rate of light generation from a 2-mm layer of a given concentration of ATP is approximatively twice the value of a 1-mm layer at the same concentration (light signal at 2 mm/1 mm = 1.9 ± 0.22). Thus, the 2-mm layer can otherwise be regarded as two stacked layers of 1 mm each. We believe this additive property to remain valid even with two hypothetically immiscible layers containing different ATP concentrations. We further reason that it can be extended to a gradient of ATP concentration, whereby a series of layers of infinitesimal height can be integrated over the entire depth to obtain the quantity of ATP molecules. In short, the rate of light production is a direct measure of the current quantity of ATP molecules independent of the spatial distribution of ATP. The calibration factor calculated from the average of individual ratio of ATP (moles) over light signal (LU/s) for each data point (n = 15) in Fig. 2B is 5 ± 2.6 × 10⁻¹⁸ moles·s⁻¹·LU⁻¹ (mean ± SD). In other words, 1 LU/s corresponds approximately to 5 amol or 3 × 10⁶ molecules of ATP. Although the calibration factor is not influenced by the height of the reaction (Fig. 2C), there is a small deviation from linearity with respect of [ATP] (Fig. 2C). As a result, using the average of the calibration factor ratio comes with the caveat that as the light signal gets bigger, ATP underestimation grows. Even though this small deviation from linearity is not believed to significantly impact the accuracy of the measurements, a complete 3D modeling of diffusion processes and enzymatic kinetics could be implemented to minimize the anticipated underestimation and further improve ATP quantification.

Based on our pseudo-first-order reaction assumption, we have calculated the half-life of the calibration signal decay at two submicromolar concentration for the three volumes of solution and obtained: t_1/2 = 6.6 ± 0.2 min, R² > 0.90 at 0.05 µM (n = 3) and t_1/2 = 6.8 ± 0.2 min, R² > 0.95 at 0.5 µM (n = 3). The magnitude of this parameter indicates that the detectable luminescence takes considerable time to decay. At higher ATP concentration (>5 µM), however, the measured half-life from our calibration assay shortens: t_1/2 = 5.6 ± 0.1 min, R² > 0.95 at 5 µM (n = 3) and t_1/2 = 4.2 ± 0.9 min, R² > 0.99 at 50 µM (n = 3). The reason for this deviation is well known and is a consequence of the typical “flash kinetics” of the LL reaction at elevated reactant concentration that had been explained mainly by the generation of the autoinhibitory product dehydroluciferyl-adenylate (L-AMP) (14). In the context of experimental stretching of cells this is not a concern, since L-AMP diffuses away from the main sites of reaction and has a negligible impact on the speed of the reaction. To demonstrate this, we simulated the ATP efflux originating from several cells as a large punctual source by injecting (PLI-100A Picoliter Microinjector; Warner) a small volume of ATP solution (<10 µL) at high concentrations (100 µM) in a RC-40LP glass-bottomed cell-free chamber filled with LL solution (∼1 mL) and measured the half-life from the signal coming from the initially small but bright source and then continuously expanding but fading luminescence to obtain t_1/2 = 7.6 ± 1.0 min, R² > 0.90 (n = 3). Thus, even though we used a high level of ATP concentration (100 µM), there was no diminution in half-life compared with our calibration assay at submicromolar values of 0.05 µM and 0.5 µM (unpaired t-test, two-tailed, P = 0.24 and P = 0.29, respectively).

The image noise distribution of the EMCCD camera at baseline (dark current; no illumination) was Gaussian (Kolmogorov-Smirnov test: P = 0.134 H₀: distribution is normal), and its standard deviation (σ_noise) tended toward 5 LU/pixel (e.g., for 1-s exposure this yields a rate of 5 LU/s/pixel equivalent to 25 amol ATP/pixel). At pixel level per 1-s exposure image, we suppose that this reading noise limits ATP detection to 3σ_noise = 75 amol or 0.075 fmol (per pixel). Note that the signal noise from the LL reaction may exceed this value.

The individual LU values per pixel for the 1-s acquisition time for stretch (or 5 s for hypo) were converted into ATP, with the calibration factor and linear fitting on ImageJ. The total amount of detectable extracellular ATP (ATP_o) within the chamber or a smaller specific ROI can be calculated and monitored over the entire course of the experiment. The quantity of secreted ATP for each stretch impulse is calculated by subtracting the prestretch ATP_o level from the stretch-generated peak value. The rate of ATP release, expressed in fmol/s or pmol/s, is the first derivative of ATP_o within any ROI over time. Color-coded images reporting ATP_o density in fmol/mm² (1 pixel image = 6 × 10⁻³ mm²) with respect to the surface substrate or rate of ATP release density in (fmol/s)/mm² were generated with ImageJ.

RESULTS

Cell stretch and amount of ATP secretion.

To test the sensitivity and resolution of our wide FOV quantitative ATP imaging approach, we first uniaxially stretched silicone substrate-adherent A549 cells. Application of a stretch pulse at high initial speed (2–5 s⁻¹) is known to result in rapid ATP secretion (7) and, as such, is well suited to assess the temporal resolution of our imaging system. All five graded stepped pulses, applied every five minutes, generated a detectable and quantifiable signal.

The ATP_o density images pre- and poststretch of a typical experiment are presented in Fig. 3A. Local increases in extracellular ATP can be detected as early as 5% strain, but ATP_o level rose mostly near the edge of the chamber, perhaps as a result of localized stress concentration. Some isolated areas reached an ATP density of ∼180 fmol/mm² (Fig. 3A, top right), although their contribution to the overall ATP secretion level was minimal (Fig. 3B). Whereas the detectable signal from the following pulse (9%) was more widespread, it remained concentrated mostly in the middle region of the stretched area, with a range covering (∼14 mm² or 40% of the stretchable groove) and a maximal ATP_o density of ∼250 fmol/mm². The 5-min interval between stretch pulses is too short for the complete breakdown of ATP_o by ectoATPases and LL reaction below the theoretical detection level of 12.5 fmol ATP/mm² (0.075 fmol/pixel ÷ 6 × 10⁻³ mm²/pixel). In consequence, ATP signal remnants from the previous stretch were still present before application of the 21% extension load, but the addition of newly released ATP can be measured by substracting pre- from poststretch level (Fig. 3C). The 21% deformation created a markedly inhomogeneous distribution of extracellular ATP, with two distinct regions where ATP_o density triples to >750 fmol/mm² (Fig. 3A). While ATP_o density continued to rise with the subsequent 27 and 37% strain, it was due mostly to additional ATP release in the vicinity of already existing high accumulation ATP, not as a result of higher ATP efflux. The stretch-induced ATP response was not monotonic. The 21% strain produced the highest response with ∼7 pmol of secreted ATP for the whole chamber, a magnitude more than twice the previous stretches, and declined thereafter for the subsequent loads. This threshold level at 21% strain had been identified for all four independent chambers tested and is comparable with that occurring during lung inflation. It has been estimated that at total lung capacity inflation, the distension of alveolar space measured from fixed rat lungs reaches a maximum of ∼37% surface strain (19), which corresponds to 17% linear strain, assuming an equibiaxial field of deformation.

Fig. 3.

Quantitative imaging of extracellular ATP for an entire stretchable area (sample no. 1). A: overlay color-coded images of ATP density or average [ATP]_o (reaction height, h_r, = 200 µm) over transmitted light (static) image of the stretch chamber (gray) with A549 cells (∼25,000). Images have been cropped to show only the stretchable groove. ATP release is shown in response to a sequence of 5 uniaxial 1-s duration stretch pulses of increasing amplitude (from 5 to 37% strain) given at 5-min intervals (25°C). Extracellular ATP density recorded seconds before (left) and after stretch (right). Note that with the stimulation protocol used in this experiment, the ATP-dependent luminescence in those images does not fully return to the background level between the consecutive stretches. B: the quantity of released ATP (in pmol) in function of time during the experiment illustrated in A. The timing of stretch stimuli is indicated by arrows. C: bar graph indicating ATP released (pmol) at each given stretch. The quantity of ATP secreted by cells was calculated from B by subtracting prestretch from after stretch peak value.

For autocrine/paracrine purinergic studies with ATP as an effector, it is essential to estimate the concentration of extracellular ATP. We can derive an average value for [ATP]_o in the vicinity of the cell monolayer from the map of ATP_o density level and the estimated extent of the LL reaction. The upper limit of the LL reaction (emitting a detectable signal) is not known, but assuming the reaction initially grows into a hemispherical shape, it should correspond to the lateral diffusion radius of small isolated ATP sources on the acquired images. The height of the reaction was varying and was not uniformly distributed throughout the chamber, but we found that a 200-µm limit was a good approximation for the experiment shown in Fig. 3A. We reported the average [ATP]_o on the alternate calibration bar for this estimated reaction height (h_r). Since the concentration is derived from ATP_o accumulation in the chamber, the largest area with the highest [ATP]_o at 4.5 µM for this experiment was seen after the last stretch (37% strain) took place. The [ATP]_o in vicinity of the cell surface ought to be higher, however, due to the existence of a gradient of ATP_o concentration near the cell surface and upward.

While a complete 3D modeling will be required to fully characterize the vertical distribution of ATP in the medium, we attempted nonetheless at estimating the [ATP]_o near the cell surface by hypothesizing a linear distribution of ATP_o. Our rationale is based on the ATP breakdown activity of endogenous cell-attached ecto-ATPases coupled with the consumption from LL reaction being proportional to the concentration of ATP. We expect those processes to lessen the concentration gradient and thereby “linearize” the vertical distribution of ATP. We assumed the [ATP]_o upward and away from the monolayer surface to be low: [ATP]_away ∼1 nM. The formula for the [ATP]_o near the monolayer at the location of highest [ATP]_o can be expressed as: [ATP]_average = ([ATP]_near − [ATP]_away)/2 = 4.5 µM. Solving for the unknown gives [ATP]_near = 9 µM.

Cell stretch and rate of ATP secretion.

While the amount of secreted ATP and ATP_o density map reveal the extent of the bulk release of ATP per chamber and per normalized surface as well as the nature of the inhomogeneous secretion pattern, the true dynamics of ATP secretion are best revealed with the time derivative of the ATP accumulation response. The highly transient nature of ATP secretion by stretch is depicted in Fig. 4 and can be seen for all five incremental loads. As opposed to the accumulated ATP, the rate of release already went back to prestretch level before the subsequent stretch as depicted on the color-coded images (Fig. 4A). But similarly to the amount of ATP secreted, a threshold was also crossed at 21% strain (Fig. 4B) with a peak rate of 1.2 pmol/s. Whether a small lingering ATP efflux persisted poststretch and/or that ATP-induced ATP release took place between load pulses could not be resolved. This ATP efflux, if it existed, ought to have been small since it failed to override the feeble but continuous degradation of ATP from LL reaction and ecto-ATPases, as shown on Fig. 4B,with the return to pre-stretch baseline level in less than ∼1 min for all applied deformations.

Fig. 4.

Quantitative imaging of ATP efflux for whole stretchable groove (sample no. 1). A: overlay images of time derivative of total ATP signal (described in Fig. 3) over brightfield image of the A549 cell-adherent stretchable groove. Left and right columns show density efflux of ATP recorded seconds before and after stretch, respectively. Note that, as opposed to the prestretch images in Fig. 3A, the rate of ATP-dependent luminescence in those images returns to near background levels before the following stretch. B: ATP efflux in function of time during the experiments illustrated in A. The timing of stretch stimuli is indicated by arrows.

In addition, half-life analysis of the decay signal past the culminating ATP accumulation suggested that very little, if any, additional ATP had been released. The average half-life from all interpulse decay signals from the four stretch chambers we analyzed in this report was t_1/2 = 9.3 ± 2.2 min, R² > 0.90 (n = 20). This average half-life value is slightly higher than the ones from our cell-free experiments (t_1/2 ~7 min) described in materials and methods. If additional ATP had been released after the stretch pulse, it would have supplemented the level of ATP and accordingly increased the half-life to a value higher than the one we measured in the absence of cells. The amplitude of this increase, however, is extremely small. For instance, in Fig. 3B, the half-life from the decaying signal following the 21% deformation load was t_1/2 = 9.8 min. In the same graph at ∼16 min, corresponding to ∼4 min past the 21% strain peak of cumulated ATP and just before the next incoming stretch, there was a total of ∼7.4 pmol in the chamber. Had the half-life been 7 min instead of the measured 9.8 min, the remaining ATP would have been 6.7 pmol (−0.7 pmol) and the difference in rate of ATP release during this period would have amounted to only +0.003 pmol/s, a tiny value when compared with reported efflux at Fig. 4B. Thus, the luminescent signal between stretch pulses is almost entirely remnants from the previous stretch and a consequence of the slow degradation of ATP but unlikely a result of substantial ATP secretion.

Regional analysis of ATP secretion.

The ability to monitor extracellular ATP fluctuation, not only for the whole chamber but also for any region of interest (ROI), is another interesting feature of our integrative ATP quantification technique. The total production rate of photons in any region in the chamber, represented by the sum of LU in this region on the image, corresponds to the quantity of ATP_o present in this location from the surface of cells and above. To illustrate this, we tracked the variation in ATP_o content for five different ROIs of 0.4 mm² from another stretch assay (Fig. 5). The quantity of ATP detected in the extracellular medium in response to stretching this chamber and the ATP efflux (Figs. 5, B–C) was less than that described in Fig. 3 and had a different distribution (Fig. 5A) but presented the same strain threshold. The cumulated ATP_o for all five regions and their corresponding efflux is shown in Fig. 5, D and E. As expected, the values are much lower than the whole chamber (see Fig. 5B), with <10 fmol per pulse stretches for most ROIs, with the exception of ROI no. 3 reaching ∼40 fmol at 21% strain pulse. While the ATP efflux kinetics of the individual regions ROI no. 3–5 located near or within the main response area (see Fig. 5A) resembled, albeit at lower magnitude, the ones observed for the whole chamber, the isolated regions (ROI nos. 1 and 2) presented a contrasting response with near-maximal ATP response at lower strain values (5 and 9%).

Fig. 5.

Quantitative imaging of regional ATP release by stretch (sample no. 2). A: ATP density and average [ATP]_o (reaction height, h_r, = 200 µm) overlaid on transmitted light image of the stretch chamber are shown in response to a sequence of 5 uniaxial 1-s duration stretch pulses of increasing amplitude (from 5 to 37% strain) given at 5-min intervals at 25°C. Extracellular ATP density images for each pulse recorded immediately before or after stretch are shown. Note that similarly to the experiment presented in Fig. 3A, the major ATP secretion activity is localized in the middle portion of the stretchable zone. B: quantity of released ATP (in pmol) by A549 cells in function of time for the whole chamber. Stretch stimuli occurrences are indicated by arrows. C: ATP efflux in function of time for whole chamber derived from B. D: ATP accumulation response from the 5 regions outlined in A. For clarity, the positions of the regions of interest (ROI) are shown on a single image, but analysis was performed over the complete experimental time course. E: rate of ATP release for the 5 regions indicated in A. Data were smoothed with a running average (n = 5) before the derivative calculation to reduce the signal noise from the luciferase-luciferin reaction and better highlight the peak values. Gray area is ±SD.

Mean normalized ATP secretion.

Figure 6A summarizes the stretch-induced ATP accumulation in extracellular medium over time for four independent experiments. The nearly continuous accumulation of ATP in the extracellular medium masks the very dynamic response of A549 cells under mechanical stretch. The ATP secretion activity occurred immediately after the onset of stretch and dissipated in <5 s. The average amount of secreted ATP after each pulse and the peak efflux during the burst of ATP release are reported on Fig. 6, B and C. Maximal responses were observed at 21% deformation, with ATP release reaching 184 ± 81 pmol/million cells (or 184 ± 81 amol/cell) and ATP efflux climbing to 26 ± 16 (pmol/s)/million cells (or 26 ± 16 (amol/s)/cell).

Fig. 6.

Summary of accumulated ATP release per chamber/pulse and rate per pulse. ATP secretion and max efflux per pulse of increasing strain are not monotonic. A: average total ATP secreted per chamber and normalized per million cells in response to pulse stretch of increasing strain (n = 4 chambers). Gray area is ±SD. B: ATP secreted per pulse. C: maximal ATP efflux per pulse of increasing strain. Data come from the same set of experiments analyzed in A. Note the similar pattern of response for total ATP secreted (in B) and ATP efflux per pulse. Mean ± SD (strain): 1.7 ± 0.2 (5%), 6.7 ± 8.2 (9%), 25.6 ± 16.2 (21%), 5.0 ± 1.9 (27%), and 5.4 ± 3.6 (37%) pmol·s⁻¹·million of cells⁻¹ (n = 4).

Given that an adherent A549 cell has a volume of ∼5–10 pL (2), and assuming the cytosolic fraction occupies the whole cell, with an estimated cytosolic ATP concentration of 1–5 mM, the total cytosolic ATP content would reach 5–50 fmol/cell. Thus, in response to a 21% strain, ∼0.4–4% of cytosolic ATP content has been expelled extracellularly. Our imaging system clearly showed that not all cells were responsive, however, and that higher values for a single stretch can be anticipated in high release areas. For example, in the experiment shown in Fig. 3A at 21% deformation, the level of ATP attained 280 pmol/million cells (280 amol/cell) for the whole chamber, but the density map revealed a peak region reaching ∼900 fmol/mm² from a prestretch value of ~200 fmol/mm². With a relative ATP density gain of ∼700 fmol/mm² and the seeding cell density of 700 cells/mm² (see materials and methods), we computed ∼1 fmol/cell or 2–20% of total cytosolic ATP content in this high-release zone from a single stretch. The rate of ATP release also showed considerable spatial heterogeneity. On Fig. 4A, for instance, the same region presented the highest ATP efflux, reaching ∼270 (fmol/s)/mm² and normalizing this rate with cell density yields ∼386 (amol/s)/cell.

Hypotonic shock and ATP secretion.

A sudden decrease in medium tonicity is another well-studied stimulus known to provoke ATP secretion (2). We tested A549 cells, which were seeded on glass coverslips and mounted on the stand of our real-time imaging system, for swelling-induced extracellular ATP elevation. To prevent cells from being acutely exposed to the remotely controlled addition of hypotonic solution, a small piece of the coverslip underneath the tip of the fluid dispenser was removed before testing, and hypotonic solution was added slowly (∼3–5 s) to minimize mechanical perturbation while achieving a steady mixing.

Compared with stretch experiments, the secretion dynamics progressed at a slower pace and was more homogeneous. In the experiment shown in Fig. 7A, the distribution of secreted ATP was evenly spread over the cell population during the initial phase of the response (t < 15 min; Fig. 7B) within 10 min after the addition of hypotonic fluid and ∼3 min after the rise of detectable ATP above background level (Fig. 7B). Afterward, a hot spot of ATP release gradually expanded near the edge of the coverslip. The ATP_o density, ranging beween 20 and 30 fmol/mm² for most of the coverslip during the rising phase of the response (Fig. 7A), climbed in excess of 100 fmol/mm² in this hot spot region but remained far from the density level observed during stretch loads (see Fig. 3A and 5A). This heterogeneous luminescence intensity, markedly distinct at t = 37.5 min (Fig. 7A), is likely a consequence of spatial differences in ATP efflux arising from regional variation in cell density. This may not be the sole explanation, however, since the heterogeneity of ATP accumulation was augmented even further after ATP secretion had ceased (t > 37.5 min). This nonhomogeneous spatial distribution in bioluminescence appears related in part to the chemistry of the ongoing enzymatic reaction with the spatial segregation and establishment of gradients of either reagents or ATP throughout the volume of the reaction rather than exclusively linked to cellular ATP efflux, although this cannot be answered firmly unless a more detailed three-dimensional diffusion analysis and enzymatic kinetics of the reaction is undertaken.

Fig. 7.

Quantitative imaging of ATP efflux induced by hypotonic shock. A549 cell swelling results in initially more homogeneous ATP secretion but slower efflux than stretch. A: color-coded images of ATP density or average [ATP]_o (reaction height, h_r, = 200 µm) overlaid on brightfield-acquired (static) image of the hypotonic chamber (gray) are shown in response to an acute diminution of extracellular fluid tonicity (50%) at 30°C. HypoLL solution was added at a location marked with an X at t = 5 min. Note that the addition of the hypotonic solution to the medium containing protein-rich LL solution created 2 air bubbles located at the periphery of the chamber wall. They did not interfere with the reaction but reflected some of the luminescence, and were especially visible at t = 37.5 min. B: cumulated ATP in the chamber over time has been calculated similarly as stretch experiments and is presented for the whole experiment until luminescence level went back to initial level. The detectable ATP efflux lasted for several minutes and ceased ∼33 min after the onset of hypotonic shock. C: total ATP secretion from stretch load (sum of all 5 consecutive stretch pulses; n = 4) and hypotonic challenge per million of A549 cells (n = 4). Secretion levels were different (*P < 0.05) by Mann-Whitney test (R, version 3.4.1 Open-source software http://www.R-project.org/).

Contrary to stretch, the hypotonically induced secretion lasted several minutes after the initiation of cell swelling (Fig. 7B). This sluggish response has been observed on three other independently tested coverslips. In a previous report from our laboratory, using an offline analysis and a perfusion chamber, the measured swelling-induced ATP efflux peaked around 1.5 min after the medium tonicity was lowered and reached a mean rate of 5 ± 1.1 pmol·s⁻¹·million cells⁻¹ (n = 9) or 5 ± 1.1 amol·s⁻¹·cell⁻¹ (2), but here, the efflux halted at 0.03 ± 0.01 (pmol/s)/million cells or 30 ± 10 (zmol/s)/cell and at a much later time: ∼15 min on average after medium dilution (n = 4). While this lower ATP secretion is mostly a consequence of diminished release, a prolonged but feeble ATP efflux tends toward underestimation. Indeed, when the ATP efflux is strong, it largely exceeds the ATP breakdown from the LL-reaction and ectoATPases, but a weak ATP efflux struggles to accumulate ATP_o in the chamber against that constant degradation. Several reasons, in addition to the ATP breakdown activity of the ecto-ATPases and LL reaction, could explain those differences. The first and foremost is the lack of perfusion resulting in a slower change in bulk medium tonicity, a critical factor determining the rate of cell swelling (13). Other factors include the lower experimental temperature, 30°C instead of 37°C in (2), and the absence of tonic fluid shear mixing the cell boundary layer in the present configuration. In addition to efflux, the total hypo-induced ATP release (n = 4) was similarly smaller at 28 ± 12 pmol/million cells (Fig. 7C) when compared with the former study from our lab (2) with 1,002 ± 154 pmol/million A549 cells (n = 9). Notice that the total amount of ATP secreted from the sum of the five stretch pulses reached 384 ± 224 pmol/million A549 cells (Fig. 7C).

DISCUSSION

The real-time ATP imaging technique described herein produces images with superior brightness when compared with a standard microscope, thanks to its threefold object-to-image reduction. It also expands considerably the field of view to encompass the whole of typical experimental chambers used for ATP release investigation and permits to exclusively visualize the heterogeneous nature of mechanosensitive ATP release from a cell monolayer. It is also a compact free-standing imaging system easy to enclose in a light-tight box for benchtop assays and a less expensive solution than other microscopy-based techniques. Other advantages of macroscopic imaging include a very long working distance (∼70 mm with the lenses described in this study) to accommodate a variety of experimental chambers and the option to change the magnification level with additional lenses.

A large proportion of earlier ATP release studies had reported dimensionless LU in lieu of quantity of ATP, making it difficult to compare the cellular responses between independent studies and further impeding the progress in search of the nature and identities of ATP release pathways. The quantification of ATP from recorded images as detailed in the present report will help eliminate this difficulty and facilitate interstudies comparisons. In addition, our ability to identify areas of higher rate of ATP secretion and accumulation despite the limited resolution of our imaging system enabled us to estimate peak ATP secretion at the cellular level, an important characteristic to help identify the mechanisms of ATP release. Furthermore, by estimating the extent of ATP diffusion, we could produce a fair approximation of the local concentration of ATP in the vicinity of the cell, an essential measure for purinergic signaling studies. Finally, we are also looking forward to improving the ATP mapping accuracy by estimating the ATP distribution in the extracellular fluid using a 3D model incorporating enzymatic kinetics and diffusion processes.

In conclusion, our macroscopic wide FOV quantitative ATP imaging approach, despite its lower imaging resolution when compared with that of a microscope, has nonetheless a very high level of ATP_o mapping resolution and quantification. Further, by revealing the big picture, this novel approach will enrich our understanding of the inhomogeneities and subtle dynamics of mechanically triggered ATP secretion from a cell monolayer.

GRANTS

This study was supported by Canadian Institutes of Health Research (MOP64364) and Natural Sciences and Engineering Research Council of Canada (R. Grygorczyk).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.J.T., O.P., R.G., and F.B. conceived and designed research; J.J.T. and O.P. performed experiments; J.J.T., O.P., and F.B. analyzed data; J.J.T., O.P., R.G., and F.B. interpreted results of experiments; J.J.T. prepared figures; J.J.T., R.G., and F.B. drafted manuscript; J.J.T., O.P., R.G., and F.B. edited and revised manuscript; J.J.T., O.P., R.G., and F.B. approved final version of manuscript.