Hemolysis is a primary ATP-release mechanism in human erythrocytes

Matching oxygen supply with demand requires specific mechanism(s) to increase blood flow in response to reduced tissue oxygen level. In addition to serving as major oxygen carriers, red blood cells (RBCs) have been proposed to function as oxygen sensors that respond to local tissue hypoxia with controlled ATP release.¹  Subsequent activation of vascular endothelial P2Y purinergic receptors stimulates the release of nitric oxide and other mediators of vasodilation, leading to vessel caliber alteration and enhanced blood flow.^2-4  In vitro studies have shown that shear stress and mechanical deformation are major stimuli of RBC ATP release, which is also induced by hypoxia.^5-8  Elevated ATP levels have been found in vivo in venous effluent from exercising forearm muscle^1,9  and further augmented by exercise performed in hypoxia.^2,10  Moreover, it has been demonstrated that only when the vessels were perfused with RBCs did the venous effluent ATP level increase and vessels dilate in response to low extraluminal O₂.²  All these results implicated RBCs in sensing low extraluminal O₂ and contributing to ATP-dependent local blood flow regulation. It has been hypothesized that shear stress- and hypoxia-induced ATP release involves activation of the cyclic adenosine monophosphate (cAMP) signaling pathway, opening up ATP-conducting channels.¹¹  Consistent with this hypothesis, ATP release has also been observed in response to other stimuli that elevate cAMP, such as agonists of prostacyclin,^12,13  or β-adrenergic receptors¹⁴  (for a review, see Ellsworth and Sprague⁴ ). Other stimulators of ATP release from RBCs include hypotonic shock¹⁵  and elevated temperature.¹⁶ 

Because mature mammalian RBCs are devoid of intracellular organelles, the only plausible pathways for such release appear to be limited to nonvesicular mechanisms, such as ATP-conducting channels.¹⁷  Pannexin-1,¹⁸  voltage-dependent anion channels, and cystic fibrosis transmembrane conductance regulator have been implicated in conductive ATP release.^13,19  However, hemolysis also is an important source of extracellular ATP, and intravascular hemolysis occurs in vivo as a consequence of hypoxia and mechanical trauma to RBCs.^20-22  Some degree of hemolysis is unavoidable during in vitro studies. Although it has been considered to be a potential factor contributing to stimulated ATP release in most previous investigations, its actual involvement has not yet been systematically investigated, eg, by paired measurement of free hemoglobin and ATP in each and every sample.

In the present study, we began by evaluating different experimental approaches to determine stimulated ATP release from RBCs and quantify hemolysis to determine their limitations and identify potential pitfalls. The most accurate methods were thus used to measure hemoglobin and ATP content in RBC supernatant samples. To directly assess the contribution of hemolysis to ATP release, free hemoglobin was always quantified in the same supernatant samples for which ATP content was reported.

Human blood was collected from hemochromatosis patients undergoing phlebotomy treatment. After obtaining informed verbal consent with institutional Ethic Committee approval in accordance with the Declaration of Helsinki, blood was drawn by venipuncture, anticoagulated with citrate phosphate dextrose solution, and stored at 4°C. Although most experiments were performed with fresh or 1- to 2-day-old blood samples, qualitatively similar results were obtained with cells that had been stored for up to 14 days. All samples were centrifuged (500g, 10 minutes, 4°C) prior to the experimental procedure. Blood plasma and buffy coat were removed by aspiration, and erythrocytes were washed 3 times with physiological salt solution (PSS) containing (in mM) 140.5 NaCl, 4.7 KCl, 2 CaCl₂, 1.2 MgSO₄, 21 Tris-base, 11.1 dextrose, and 0.5% bovine serum albumin (pH 7.4, osmolality 307 mOsm/kg). Unless stated otherwise, the cell count in all experiments was adjusted to 10⁶ cells/µL (∼10% hematocrit). Osmolality of the solutions was measured with a freezing point osmometer (Model 3300, Micro Osmometer; Advanced Instruments Inc., Norwood, MA).

In preliminary experiments, we evaluated several commonly used methods of hemolysis determination, including Drabkin²³  and luminol assays,²⁴  and found that hemolysis readings based on hemoglobin absorbance measurements, such as the Harboe method, are most accurate, reproducible, and less susceptible to the presence of interfering substances, in agreement with a previous report.²⁵  In our hands, absorbance measurements allowed reliable detection of 20 to 30 hemolyzed cells/µL, corresponding to 0.002% to 0.003% hemolysis at our standard cell density of 10⁶ cells/µL (supplemental Methods; supplemental Figure 1, available on the Blood Web site).

ATP was quantified by luciferin-luciferase (LL) bioluminescence assay, with ATP Assay Mix and ATP Assay Mix Dilution Buffer supplied by Sigma-Aldrich Canada (Oakville, ON, Canada). Luminescence was measured by Turner TD-20/20 luminometer (Turner Designs, Sunnyvale, CA). Released ATP was expressed in fmol × (10⁶ cells)⁻¹. A calibration curve of LL luminescence (in arbitrary light units) vs standard ATP concentrations was produced before and after each experiment, and a separate calibration curve was plotted for each hypotonic and isotonic solution.

Following preliminary experiments, we chose to analyze ATP in RBC supernatants after centrifugation of RBC suspension samples (supplemental Methods; supplemental Figure 2A). With this approach, each and every supernatant sample underwent paired measurements of extracellular ATP and free hemoglobin, allowing direct monitoring of hemolysis contribution to ATP release.

Most ATP efflux experiments were performed in 6- or 12-well plate format. Each well contained ∼3 mL of RBC suspension in 6-well plates placed inside a 37°C incubator (Ambio High-Low Chamber; Laboratory-Line Instruments, Melrose Park, IL) on a 3D rotator platform (10° angle, 30 rpm; Laboratory-Line Instruments) for gentle mixing. A hypotonic stimulus was applied by diluting concentrated, washed erythrocyte suspensions (5 × 10⁶ cells/µL) to a final density of 10⁶ cells/µL by mixing appropriate volumes of concentrated cell suspension with saline solution of a reduced (eg, by 20%, 25%, 30%, and 35%) NaCl concentration. Cells were incubated in hypotonic solutions for 5 or 15 minutes with gentle mixing, centrifuged, and the supernatants analyzed for extracellular ATP and free hemoglobin (see above). All experiments were conducted at 37°C.

To study pharmacological stimulation of ATP release, experiments were performed in 6-well plate format, as described in the preceding section. Unless stated otherwise, before the experiments, forskolin (an adenylate cyclase activator) and papaverin (a phosphodiesterase inhbitor) stocks were prediluted with PSS to contain <10% dimethylsulfoxide (DMSO) and were added directly to RBC suspensions (final DMSO concentration <0.01%).

To apply shear stress, 2-mL samples of RBC suspensions (10⁶ cells/µL) were vigorously mixed in a 15-mL conical polypropylene tube by a VX100 vortex mixer (Labnet, Woodbridge, NJ) at ∼1000 rpm for 15 to 300 seconds. Supernatants were collected by centrifugation (1800g 10 minutes, 4°C) and kept on ice until ATP and free hemoglobin determinations. Computational fluid dynamics modeling with a similar vortex mixing system showed that at 1000 rpm agitation frequency, steady-state volume-averaged shear stress was ∼4 dyne/cm.^2,26 

RBC samples (in 6- or 12-well plate format) were placed in an airtight flow-through box with a continuously injected gas mixture containing 0% O₂ and 5% CO₂ balanced with N₂. After 30 minutes of equilibration with gentle mixing on the 3D rotator platform, hemoglobin oxygen saturation, measured by blood gas analyzer (ABL800 FLEX; Radiometer, London, ON, Canada), was within the 40% to 80% range compared with 97% to 99% in the controls.

To image ATP release from cells, a fresh droplet of RBCs was collected from healthy donors by finger prick and washed 2 to 3 times with PSS. Low-density RBC suspensions were deposited on poly-lysine–coated glass coverslips (0.1 or 0.01 mg/mL poly-lysine, molecular weight 2.7 kDa). To visualize cellular ATP release in real time, we deployed an imaging system that combined simultaneous high-sensitivity bioluminescence detection of ATP and infrared (IR) differential interference contrast (DIC) images of cells to identify ATP-releasing cells^27,28  (supplemental Methods).

Statistical significance was assessed by nonparametric Mann-Whitney U test (STATISTICA 10 software, StatSoft, Tulsa, OK). Differences were considered to be significant at P < .05. All data were expressed as mean ± standard error of the mean (SEM).

Figure 1A depicts an experimental example where RBC suspension samples were exposed for 5 minutes to isotonic PSS or 25%, 30%, or 37% hypotonic PSS (ie, containing 140.5, 105.38, 98.35, or 88.52 mM NaCl, respectively). Extracellular ATP measured in supernatants significantly increased with rising hypotonicity. However, elevated hemolysis was also noted with RBCs exposed to hypotonic solutions, as revealed by hemoglobin absorbance measurement of the same supernatant samples (Figure 1B-C). By plotting released ATP against the number of lysed cells found in supernatant samples, we determined that released ATP (black circles) increases with the number of hemolyzed cells, and this dependence is well fitted by a linear relationship (correlation coefficient R = 0.9880, Figure 1D). Such linear dependence strongly suggests that hemolysis contributes, at least in part, to hypotonic shock-induced ATP release. To verify the extent of this contribution, Figure 1 also compares ATP released by hypotonic shock with ATP released by lysing the corresponding cell number of the same RBC batch with Triton X-100 (red triangles). The lysate data points closely overlap with hypotonic shock-induced ATP release, demonstrating that all ATP released by hypotonic shock originates from cell lysis. The slope of the linear relationships in Figure 1D corresponds to an intracellular ATP concentration of 1.42 mM, but it may vary between different RBC batches. This is illustrated in Figure 1E, where examples of similar hypotonic shock experiments performed with 3 different blood batches are presented. In all such experiments (n = 8), despite up to ∼10-fold variability of intracellular ATP concentration, we observed a tight, linear relationship between hemolysis level and released ATP (P < .0001). As might be expected for release mechanisms involving cell lysis, the pannexin channel inhibitor carbenoxolone (CBX) showed no effect on hypotonic shock-induced ATP release (Figure 1F).

The above experiments revealed that, besides hemolysis, there was no evidence of additional ATP release that could be attributed to other mechanisms triggered by hypotonic stress. To further confirm this observation, we performed luminescence ATP imaging, as described by Grygorczyk et al and Furuya et al.^27,28  RBCs were deposited on poly-lysine–coated glass coverslips and bathed in LL containing PSS (supplemental Methods). IR-DIC and luminescence images were simultaneously captured for the duration of the experiment. Figure 2A shows that all regions displaying ATP release perfectly matched sites where cells had lysed, demonstrating that ATP release occurred exclusively from lysed cells and was not detected from cells that remained intact. The time-course of luminescence responses in this experiment is depicted in Figure 2B, which discloses brief spikes of luminescence (average duration of 17.6 ± 6.7 seconds) due to ATP release from lysing cells. Although isolated lysis events could occasionally be seen under isotonic conditions, their frequency increased dramatically during 20% hypotonic shock. Interestingly, after ATP release, cell collapse/lysis was apparent after a delay of 47 ± 10.6 seconds (average ± SEM) (Figure 2C). The results confirm that cell lysis was the sole mechanism of ATP release in these experiments. It should be noted that spontaneous hemolysis of RBCs attached to the glass surface coated with high poly-lysine concentration may also contribute to RBC lysis, even in the absence of hypotonic stimulation.²⁹ 

When shear stress was applied to stimulate RBCs, ATP in cell supernatants progressively increased with the duration of stimulation (Figure 3 inset). It was accompanied, however, by elevated cell hemolysis showing a tight correlation between ATP and free hemoglobin in supernatant samples (Figure 3). Thus, the data are consistent with hemolysis as the primary ATP release mechanism in RBCs subjected to shear stress.

Forskolin, a potent adenylate cyclase activator, was reported in several previous studies to stimulate ATP release from RBCs. In an attempt to replicate these experiments, we tested the effect of forskolin on ATP release and found that adding a small aliquot of concentrated forskolin stock solution in DMSO directly to erythrocyte suspensions markedly increased extracellular ATP (Figure 4A). It also illustrates that ATP responses were indistinguishable from those evoked by an equivalent volume of DMSO alone, indicating that even when final DMSO concentration (∼0.27%) was harmless, some cells, presumably those in close proximity to the pipette tip, might have been exposed to high solvent levels causing their lysis. Indeed, Figure 4B reports that DMSO-evoked elevation of extracellular ATP closely followed extracellular hemoglobin. When the same amount of DMSO was prediluted 10-fold in PSS and then added to the cells, no increase in hemolysis or ATP release was seen. A similar result was obtained with forskolin, when its DMSO stock solution was prediluted 10-fold in PSS before addition to RBC suspension: neither extracellular ATP nor hemoglobin showed any increment, the latter remaining <0.05%, as in control supernatant samples.

In subsequent experiments, only prediluted stock solutions containing DMSO <10% in PSS were tested to avoid hemolysis when adding dissolved agents directly to cell suspensions. With this approach, in our hands, neither forskolin alone nor forskolin in combination with papaverine and isoproterenol had any significant effect on ATP release (Figure 4C inset). Extracellular ATP levels in individual samples varied within the same range as in untreated control cells and strongly correlated with hemolysis (P < .0001, R = 0.975), which remained <0.05%, comparable with control values obtained in our other experiments (Figure 4C). Luminescence ATP imaging experiments confirmed that forskolin did not stimulate ATP release from intact cells (supplemental Movie 3); it occurred only from cells that lysed either because of mechanical perturbations or coapplied hypotonic shock (supplemental Movie 4). Thus, ATP luminometry and luminescence ATP imaging experiments demonstrated that in response to forskolin or other cAMP agonists, there was no additional mechanism of ATP release beside hemolysis.

Exposure of RBC suspensions to low oxygen conditions for 30 minutes led to significantly increased extracellular ATP and heightened free hemoglobin levels in RBC supernatants (Figure 5A). Hemolysis and extracellular ATP varied between individual experiments, but in each case, they were significantly higher in hypoxia-treated samples than in the controls. Moreover, a tight linear relationship was observed between the 2 variables, demonstrating that hemolysis was the main mechanism responsible for ATP release in these experiments (Figure 5B).

In this study, we attempted to identify cell-regulated ATP release pathway(s) in human erythrocytes. By paired measurements of ATP and free hemoglobin in each and every sample of RBC supernatants, we found that basal and stimulated ATP levels always correlated tightly with extracellular hemoglobin, a marker of cell lysis. Unexpectedly, this was seen with all stimuli tested, strongly indicating that, for each stimulus, the only source of extracellular ATP was cell lysis. In the case of hypotonic shock and cAMP/forskolin stimulation, we confirmed our observations more directly by simultaneous luminescence ATP imaging and IR-DIC imaging of substrate-attached RBCs. These experiments identified single ATP-releasing cells and revealed that only lysing cells contributed to ATP release. Thus, despite our efforts, we obtained no evidence of regulated, nonlytic ATP release in RBCs.

Basal hemolysis levels in our experiments varied between different blood donors and were elevated after prolonged blood storage. Typically between 0.01% and 0.05%, they were somewhat augmented in stimulated RBC samples subjected to shear stress, hypoxia, or hypotonic shock, but remained ≤0.1%. Such hemolysis values are similar to or below those reported by other investigators.^5,12,13  In several studies, hemolysis was stated to have negligibly contributed to stimulated ATP release, and samples with heightened hemolysis were simply excluded from the analysis. In other investigations, released ATP was corrected by subtracting average ATP attributable to hemolysis. The latter was often found to be not different or comparable in stimulated and control RBC samples. To the best of our knowledge, in most previous studies, hemolysis was not systematically measured or analyzed, ie, in all individual samples for which ATP content was reported. As a result, its contribution to ATP release might have been underestimated or, in some cases, entirely overlooked.

In the present investigation, we tested several stimuli, including hypotonic shock, previously demonstrated to induce regulated RBC ATP release. Although not exactly a physiological stimulus, hypotonic shock is often considered as a surrogate of mechanical stimulation,^30,31  and ATP release can be extrapolated to responses evoked by mechanical perturbations. We noted heightened hemolysis at ∼25% hypotonicity, which increased with augmented hypoosmotic shock and strongly correlated with extracellular ATP, as expected for lytic ATP release mechanism. Apart from hemolysis, nonlytic mechanism(s) may contribute to ATP release from intact erythrocytes. This would be seen as a “surplus” of released ATP above the hemolysis “background.” However, ATP found in lysates of known cell density (eg, 1000 cells/µL) was almost exactly the same as that actually released by the same number of cells when they lysed during hypotonic shock. Furthermore, despite different intracellular ATP concentrations and the resulting variability of ATP release observed with various blood batches (Figure 1E), it always increased linearly with rising hemolysis level, demonstrating that there was no “surplus” ATP above the hemolysis “background” in all such experiments. Because hemoglobin is a major protein constituent of the RBC cytoplasm, its leakage eventually leads to cell collapse. Interestingly, after ATP release, cell collapse/lysis occurred with an average delay that was ∼2.7-fold longer than ATP release duration (compare Figure 2Bb, Cb). Slower hemoglobin release might be expected due to its ∼3.7-fold smaller diffusion coefficient in water vs ATP (6.9 × 10⁻⁷ and 2.6 × 10⁻⁶ cm²/s, respectively). The data support the view that ATP release occurs simultaneously with hemoglobin leakage during RBC lysis and involves the same release pathway.

An interesting finding in our study was the effect of DMSO on ATP release. Because of its amphiphilic properties and high polarity, DMSO is widely used as a solvent, easily permeating biological membranes and facilitating drug transport.^32,33  However, in high concentrations, DMSO imposes osmotic stress and is toxic to cells, affecting water flow through membranes and other cell functions via incompletely understood mechanisms.³⁴  When added to RBC suspensions, DMSO readily diffuses into cells, resulting in cell swelling or eventual rupture (supplemental Figure 3; supplemental Movie 5). We noticed a strong effect of DMSO on ATP release when concentrated stocks of test compounds, dissolved in DMSO, were added directly to RBC suspensions. Even if the final DMSO concentration in a given volume of RBC suspension is harmless, some cells may undergo lysis, presumably exposed to the highest solvent concentration, in proximity to the pipette tip.^33,35  To exclude a direct effect of DMSO on cell lysis, we tested prediluted stocks of compounds containing <10% DMSO. In our hands, forskolin and other cAMP agonists did not stimulate ATP release when precautions were taken to avoid DMSO-induced osmotic stress and cell lysis. Interestingly, in several previous studies, cAMP agonists (forskolin 10 µM, and 3-isobutyl-1-methylxanthine 100 µM) were included in “stop” solution designed to prevent ATP release and hydrolysis in human blood samples, consistent with lack of cAMP-stimulated ATP release in these investigations.^16,36 

Recently, Mairbäurl et al (2013)³⁷  addressed the role of hemolysis in shear stress- and hypoxia-induced ATP release in more detail in their in vitro study. They noted that, in shear stress experiments under normoxia, up to 50% of ATP release was due to hemolysis. Its contribution, however, decreased to <5% when RBCs were exposed to shear stress and hypoxia at the same time. Although precise correlations between hemolysis and ATP levels were not investigated, most ATP release observed in the hypoxia experiments was attributed to a cell-regulated mechanism. This was not confirmed in our study: we saw that shear stress or hypoxia alone could induce dose-dependent elevation of both extracellular ATP and free hemoglobin in cell supernatants, with strong, positive correlations between the 2 variables. Interestingly, hypoxia was shown to decrease deformability of RBCs and increase susceptibility to cell damage.^22,38  Thus, the increment of extracellular ATP induced by hypoxia and hypoxia together with shear stress could be attributed to augmented cell lysis.

The results presented here contradict much previously published experimental data on ATP-release mechanism(s) in RBCs. Owing to the absence of intracellular organelles, mature mammalian RBCs were considered as the ultimate example of a cell system where ATP release occurs exclusively via conductive pathways. Several channels were implicated in ATP release, including cystic fibrosis transmembrane conductance regulator, voltage-dependent anion channels, and pannexins based on observed stimulation of ATP release, eg, by cAMP agonists, or its inhibition by various compounds, such as DIDS, flufenamic acid, glibenclamide, CBX, or probenicid. All these compounds have a wide spectrum of effects, blocking other channels, transporters, or processes besides putative ATP channels.^39-42  Some inhibitors may directly interfere with the bioluminescence reaction used to measure ATP. For example, the stretch-activated ion channel blocker gadolinium (at 10 µM) and the anion channel blocker 5-nitro-2-[(3-phenylpropyl)amino]benzoic acid (at >250 µM) were found to inhibit LL bioluminescence.^31,43  Certain inhibitors, including CBX and glibenclamide, were found to deplete cellular ATP by interfering with cell metabolism in other cell types.^39,40  The effects of test compounds, or their solvents, on cell membrane stiffness/fragility, resulting in altered susceptibility to hemolysis, may also contribute to apparent ATP release modulation under specific experimental conditions.⁴⁴  Thus, in addition to tight hemolysis control, several other control experiments are needed to support the validity of RBC ATP release inhibition data.

The major finding of our study is that ATP released from RBCs by various stimuli correlated tightly with free hemoglobin, clearly demonstrating an important yet often underestimated role of hemolysis in RBC ATP release. Luminescence imaging of ATP release and IR imaging of RBCs provided direct evidence that ATP is released exclusively from lysing cells. Our data show that extreme caution must be taken when studying ATP release from RBCs. In in vitro experiments, from the moment of blood collection through storage, centrifugation, mixing, and pipetting to final assessment, erythrocytes are exposed to various physical perturbations that unavoidably cause rupture and hemolysis of some cells. Furthermore, cell lysis can be induced by several stimuli intended to trigger regulated ATP release, as we have demonstrated for hypoxia, mechanical stimulation, hypotonic shock, or the addition of test compounds dissolved in concentrated DMSO.

Importantly, hemolysis also occurs in vivo, and such ATP release mechanisms may have physiological and pathophysiological roles in local purinergic signaling and blood flow regulation. Intravascular hemolysis is particularly profound during intense exercise and mainly affects senescent RBCs, which are mechanically ruptured when they pass through capillaries in contracting muscles.^20,45  Furthermore, exercise training under hypoxia was shown to cause RBC senescence, reduce deformability under shear flow, and thus increase susceptibility to membrane rupture. Altered RBC fragility is observed in a number of disease conditions, including hereditary spherocytosis, chronic liver diseases, and thalassemia. Our study therefore suggests that mechanisms controlling RBC deformability and membrane fragility will also contribute to modulation of ATP release via hemolysis. For example, the Ca²⁺-activated K⁺ channel (Gardos channel), by controlling RBC dehydration, participates in the regulation of erythrocyte deformability and its activity is suppressed in senescent cells.²²  Moreover, elevated intracellular calcium is known to increase RBCs’ osmotic fragility, which may in part involve specific Ca²⁺-induced lytic vulnerability of membrane,⁴⁶  eg, via activation of scramblase and calpain and subsequent lysis of cytoskeletal proteins.^22,47 

In summary, the results presented here call into question current views on the mechanism of stimulated ATP release from RBCs and show that hemolysis is the primary release mechanism in these cells. In vivo, such an ATP release mechanism, involving lysis of senescent cells, might have physiologically relevant roles, eg, matching oxygen supply with demand during exercise and hypoxia, where elevated intravascular hemolysis is observed.

This study was supported by the Canadian Institutes of Health Research (MOP64364 to R.G.). The authors thank Dr R. Beaulieu, Hematology Department, Centre Hospitalier de l'Université de Montréal – Hôtel-Dieu, Montreal, for help in obtaining blood samples, and Dr J. W. Hanrahan, McGill University, for helpful discussions and comments on the manuscript.

Contribution: R.G., principal investigator in this study, contributed to the study design, its conduct, data analysis, and drafting the manuscript; J.S. performed all luminometry experiments and data analysis and drafted the graphs and “Results”; K.F. and R.G. contributed to luminescence imaging experiments and their analysis; and S.N.O. contributed to the study plan and research direction. All authors contributed to data review and provided comments on the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ryszard Grygorczyk, CRCHUM, Tour Viger, 900, rue St-Denis, Montreal, QC, Canada H2X 0A9; e-mail: ryszard.grygorczyk@umontreal.ca.