Multiparameter digitized video microscopy of toxic and hypoxic injury in single cells.

There is no clear picture of the critical events that lead to the transition from reversible to irreversible injury. Many studies have suggested that a rise in cytosolic free Ca2+ initiates plasma membrane bleb formation and a sequence of events that lead ultimately to cell death. In recent studies, we have measured changes in cytosolic free Ca2+, mitochondrial membrane potential, cytosolic pH, and cell surface blebbing in relation to the onset of irreversible injury and cell death following anoxic and toxic injury to single hepatocytes by using multiparameter digitized video microscopy (MDVM). MDVM is an emerging new technology that permits single living cells to be labeled with multiple probes whose fluorescence is responsive to specific cellular parameters of interest. Fluorescence images specific for each probe are collected over time, digitized, and stored. Image analysis and processing then permits quantitation of the spatial distribution of the various parameters with the single living cells. Our results indicate the following: The formation of plasma membrane blebs accompanies all types of injury in hepatocytes. Cell death is a rapid event initiated by rupture of a plasma membrane bleb, and it is coincident with the onset of irreversible injury. An increase of cytosolic free Ca2+ is not the stimulus for bleb formation or the final common pathway leading to cell death. A decrease of mitochondrial membrane potential precedes the loss of cell viability. Cytosolic pH falls by more than 1 pH unit during chemical hypoxia. This acidosis protects against the onset of cell death.


Introduction
Until recently, research on hypoxic and toxic cell injury was, almost without exception, directed towards large populations of cells (e.g., whole tissue, cell suspensions, cell cultures). However, in such preparations, the onset of irreversible injury is not synchronous. Following hypoxia for example, minutes or even hours can even pass between death of the first cell in a population and the last cell. Thus, the events that lead to irreversible injury and cell death are obscured when whole cell populations are studied, especially if these events occur rapidly or suddenly. The determination of the temporal sequence of such rapidly occurring events may simply be impossible.
We are attempting to overcome these drawbacks by measuring multiple cellular functions in individual digital converter and frame memories from Imaging Technologies, a time-lapse video cassette recorder from RCA, computer-driven filter wheels for changing excitation and emission light, and software for image processing and analysis (1). The high sensitivity of the ISIT camera, together with the frame averaging capability of the computer, allows us to work at very low levels of excitation energy and fluorophore concentration, thus preventing photobleaching or photodamage that might lead to disruption of normal cell activity. The ISIT camera can also be used to obtain bright field images, allowing the acquisition of phase and fluorescent images from the same cells. In a typical experiment, we monitor and record phasecontrast images continuously at low illumination with brief interruptions to collect digitized fluorescent images. With the phase-contrast images, we can see changes in cell structure that are then correlated with quantitative changes in the fluorescent images.

Ratio Imaging
The use of MDVM not only provides quantitative information regarding the amount of fluorophore emission, but also the spatial and temporal characteristics of this emission. However, because MDVM provides a two-dimensional image of a three-dimensional object (i.e., the cell), misleading estimates of the fluorophore concentration in different parts of the cell will be obtained due to differences in pathlength or accessible volume (Fig. 1). To correct for this problem, images are acquired from cells at two different excitation wavelengths, one that is parameter sensitive and one that is not. By dividing the images obtained at both wavelengths, errors introduced by differences of pathlength, regional fluorophore concentration, dye leakage over time, photobleaching, and accessible volume are corrected. Such ratio imaging is employed with Fura-2 to determine cytosolic free Ca2' and with BCECF to determine cytosolic pH.

Fluorescent Probes of Cellular Functions
The experimental approach we follow takes advantage of the availability of a number of different fluorescent compounds whose excitation and emission spectra are sensitive to various environmental parameters and which preferentially accumulate into different subcellular compartments in intact living cells. This enables us to study multiple cellular functions in individual living cells in response to external stimuli.
For the studies reviewed here, we used Fura-2 to measure free Ca2" (2), rhodamine 123 to monitor mitochondrial membrane potential (3,4), BCECF to determine intracellular pH (5), and propidium iodide to label the nuclei of nonviable cells (6). When excited at 340 or 350 nm, the fluorescence of Fura-2 shows a dependence on Ca2" concentration with a Kd of about 225 nM; when excited at 362 nm (isoasbestic point), there is no change in fluorescence intensity with a change in Ca2 . Corresponding pH-sensitive and pHinsensitive wavelengths for BCECF are 490 and 440 nm. Rhodamine 123 has been extensively employed as a fluorescent label of mitochondria and distributes electrophoretically in response to the mitochondrial membrane potential (4). Propidium iodide has seen widespread use in flow cytometry and labels nonviable cells exactly like trypan blue. Fluorescence specific to each probe is selected, employing the appropriate excitation and emission filters. In preparations loaded with several fluorophores, multiple parameters can be monitored quantitatively over time in single living cells. Introducing this multiparameter technique in 1985 (7), we showed that four or more variables of interest can be monitored in the same cell sequentially by selecting parameter-specific fluorophores with nonoverlapping emission and excitation spectra. cence is sensitive to free Ca2" but also to pathlength and accessible volume (as illustrated in 1). Ratio imaging can correct for these factors (as illustrated in 2). Fura-2 is excited at two different wavelengths, one that excites Ca2"-sensitive fluorescence and the other which excites Ca2+-insensitive fluorescence. The ratio of Ca2"-sensitive to Ca2"-insensitive fluorescence is proportional to Ca21 concentration, and errors due to pathlength, Fura-2 concentration, and accessible volume are factored out. As illustrated in 3, ratio imaging on a pixel-bypixel basis using our MDVM system produces a two-dimensional map of free Ca2" concentration. Background images are obtained at each wavelength and are subtracted from the experimental images at each wavelength before ratioing. Once ratioed, Ca2" concentration is determined by comparison to a standard curve stored in computer memory. Fura-2 and BCECF are loaded as their acetoxymethyl esters (5 jiM for 30 min in growth medium). Cytoplasmic esterases serve to cleave the acetoxymethyl ester bonds, releasing free Fura-2 or BCECF into the cytosol at concentrations in the range of 50 to 200 FM. Several technical issues concern the use of Fura-2 and BCECF. As summarized in Table 1 for Fura-2, these include: a) incomplete ester hydrolysis (8); b) sequestration into organelles (9)(10)(11)(12); c) shifts in excitation and emission spectra and Kd for Ca2+ dependent on viscosity, pH, polarity, and other factors (9,13); d) dye leakage from cells; and e) photochemical formation of fluorescent, Ca21-insensitive derivatives of Fura-2 (14).
All of these concerns are important, and Table 1 summarizes how each concern is addressed. However, the sequestration of Fura-2 into organelles is possibly most critical for correct qualitative interpretation of the data. For this reason, we have recently developed a definitive technique for the subcellular localization of Fura-2 and related dyes using MDVM (15,16). The principle of the approach is to co-load cells with Fura-2 (or other dye of interest) and additional fluorescent probes that are specific for various subcellular compartments: rhodamine 123 for mitochondria, rhodamine-dextran for lysosomes, and rhodamine microspheres (0.1 gm) for the pre-lysosomal endocytic compartment. Rhodamine-dextran is pulse-loaded a day earlier, and rhodamine 123 and rhodamine microspheres are co-loaded with the other dye of interest. In the multiply labeled cells, various compartments are sequentially opened by increasing concentrations of digitonin. A Ca21-free, succinate-containing buffer is employed to prevent mitochondrial depolarization after permeabilization of the plasma membrane. As illustrated for hepatocytes loaded with BCECF, 20 FM digitonin opens the cytosol but leaves mitochondria (Fig. 2), endosomes (Fig. 3), and lysosomes ( Fig. 4) intact. The subsequent additon of 100 jiM digitonin opens the lysosomal and endosomal compartments,  2), no additional loss of BCECF fluorescence occurred (E), but rhodamine-microsphere fluorescence disappearance was complete (F). This experiment demonstrates the absence of BCECF localization to diffuse prelysosomal endocytic compartment.
but mitochondria are preserved. More digitonin or another detergent (e.g., Triton X-100) releases the remaining mitochondrial fluorescence. From such experiments, we have quantitated the extent and amount of co-localization of Fura-2 in noncytosolic compartments from the digital images. In hepatocytes, we have had success in loading Fura-2 and BCECF with greater than 85% specificity for cytosol and less than 10% accumulation in the mitochondria or lysosomes. Our loading protocol also results in virtually complete hydrolysis of Fura-2-AM to Fura-2. However, the results from other cell types (e.g., cardiac myocytes, vascular smooth muscle cells, fibroblasts) can be quite different (9); thus it is essential to evaluate Fura-2 compartmentation in each new cell type studied.

Experimental Models of Cell Injury
We have used four different protocols to produce toxic or anoxic injury in 24-hr cultured rat hepatocytes. True anoxia was established by perfusing an lysosomes. A hepatocyte was loaded with rhodamine-dextran 48 hr earlier by IP injection. After 1 day in culture, the cell was loaded additionally with BCECF. BCECF fluorescence from the cell was diffuse (A); whereas rhodamine-dextran fluorescence was punctate consistent with its localization to secondary lysosomes (B). After 20 jM digitonin, virtually all BCECF fluorescence was released (C), but rhodamine-dextran was unchanged (D). After 100 jM digitonin, no additional loss of BCECF fluorescence occurred (E), but rhodamine-dextran fluorescence was lost entirely (F). This experiment demonstrates the absence of BCECF localization to lysosomes. Since 20 FM digitonin released BCECF completely, but left lysosomes, endosomes, mitochondria, and presumably other membranous organelles intact, it can be concluded that BCECF was localized exclusively to the cytosol (including nucleoplasm).
environmental chamber mounted on the stage of the microscope with medium containing 1 mg protein/mL submitochondrial particles and 5 mM succinate (17). Submitochondrial particles consume oxygen quantitatively just as tissue does in ischemia. Because the submitochondrial particles were premixed with the medium, anoxia was established as soon as the medium containing succinate and the submitochondrial particles was perfused into the chamber. KCN (2.5 mM) and iodoacetate (0.5-10 mM) were employed to inhibit ATP production by oxidative phosphorylation and glycolysis (6,16,18). This treatment mimics the fall in ATP that accompanies anoxia and is termed chemical hypoxia. HgCl2 (50,M) was used to attack membrane transport processes (toxic injury), and cystamine (5 mM) was employed to cause oxidative stress by forming mixed disulfides with plasma membrane thiols (19,20).

Stages of Bleb Formation
Bleb formation occurred in all the experimental treatments (6,7,16,17,19,20). Three stages could be recognized. Stage 1 consisted of numerous small surface blebs (Fig. 5A). In chemical hypoxia, stage 1 blebbing began within 10 min of addition of KCN and IAA. Stage 2 began after about 20 min. During stage 2, blebs grew through a process of coalescence and bleb fusion (Fig. 5B) until one to three large terminal blebs were left (Fig. 5C). Bleb enlargement was a slow process requiring minutes. Bleb fusion was extremely rapid and occurred within consecutive video frames (33 msec; data not shown). Stage 3 of blebbing was initiated by rupture of one of the terminal belbs (Fig.  5D). Blebs were phase-lucent during all stages, indicating the absence of granular organelles such as mitocondria and lysosomes. The time required for progression of cells through stages 1 and 2 varied from cell to cell and depended on the type of injury. Thus, determination in single cells of bleb progression in relation to alterations in other cellular paarameters was essential for our studies.

Onset of Irreversible Injury
Alterations in the structure and function of the plasma membrane have received attention as a major focus in the evolution of hypoxic damage and in the transition from reversible to irreversible injury. Our results demonstrate direct plasma membrane injury in the form of bleb formation and eventual bleb rupture. Bleb rupture precipitates the onset of cell death, but the transition from reversible to irreversible injury may occur earlier. Therefore, we undertook a study to document the onset of irreversible injury with respect to plasma membrane blebbing, bleb rupture, and onset of cell death (17). Our strategy was to reoxygenate anoxic hepatocytes at various stages of blebbing and determine at what point the injury was no longer reversible. Cells were made anoxic by incubating them on the microscope stage with submitochondrial particles and succinate. Reoxygenation was achieved by the infusion of fresh, oxygenated medium that did not contain submitochondria particles or succinate. When cells in stages 1 or 2 of bleb development were reoxygenated, their blebs were resorbed (Fig. 6A-L). There was no apparent reoxygenation injury, and the hepatocytes continued to exclude propidium iodide ( Fig. 6D and  H). However, in stage 3 cells that had already taken up propidium iodide as a consequence of bleb lysis, reoxygenation did not reverse bleb formation or lead to any other change in cell structure (Fig. 6I-L). Thus in anoxic hepatocytes, the onset of cell death and the transition from reversible to irreversible injury appeared to occur at the same time.

Bleb Rupture
Scanning electron microscopy was used to characterize changes in membrane structural integrity after bleb lysis or rupture. A field of cells was viewed in the video microscope during chemical hypoxia. Blebs formed (Fig. 7a) and after the first bleb had ruptured in one of the cells (Fig. 7b), propidium iodide uptake by the cell was documented (Fig. 7c), and the coverslip was flooded with fixative and prepared for scanning electron microscopy (Fig. 7d). In cells not taking up propidium iodide, the cell surface was smooth and continuous with large, intact blebs (cell on the left in Fig. 7d). In the cell which had lost viability, a large discontinuity of the membrane surface was observed; it was rimmed by vesiculated fragments of membrane (cell on the right in Fig. 7d). Thus, bleb rupture led literally to a large hole in the plasma membrane. Chemical Hypoxia Cytosolic Free Ca2" during Chemical Hypoxia It has been suggested that a rise in cytosolic free Ca2" initiates a series of harmful processes that culminate in cell death (21)(22)(23)(24). Using MDVM and ratio imaging of Fura-2 fluorescence, we monitored cytosolic free Ca2' following chemical hypoxia with KCN plus iodoacetate in relation to mitochondrial membrane potential, bleb formation, and the onset of cell death (Fig. 8). Unexpectedly, chemical hypoxia was not associated with any change in cytosolic free Ca2" up to the point of cell death, even though these cells underwent blebbing.
To document that Fura-2 in these cells was responsive to change in cytosolic free Ca2", hepatocytes were treated with vasopressin, phenylephrine, and an epidermal growth factor-all of which increased free Ca', as measured by Fura-2 ratio imaging (6). Ionomycin, a calcium ionophore, also increased free Ca2", whereas addition of EGTA, which chelates extracellular free calcium, caused a slight decrease. Thus, Fura-2 responded to stimuli that both increased and decreased cytosolic free Ca2". Although substantial increases in free Ca2" were caused by these hormones, cells die not bleb and remained viable for long periods of time. In addition, we determined in multiple labeling experiments nearly all the internalized Fura-2 was nonmitochondrial and nonlysosomal (15). Moreover, the fluorescence spectrum of Fura-2 released from cells by detergent was nearly identical to that of the pure Fura-2 free acid in the same buffer and was fully responsive to Ca2 , indicative of the absence of unhydrolyzed or partially hydrolyzed Fura-2-AM.

Mitochondrial Membrane Potential during Chemical Hypoxia
Anoxic injury is initiated by an inhibition of ATP synthesis leading to a lack of energy supply for cellular reactions. Mitochondria serve as the primary site of ATP synthesis, and synthesis of ATP is associated with the generation of a potential difference across the mitochondrial inner membrane. Rhodamine 123 is a fluorescent dye that accumulates electrophoretically in mitochondria in response to the potential difference across the inner membrane (4). Following chemical hypoxia, the mitochondrial membrane potential decreased steadily, concomitant with bleb formation until the onset of cell death (Fig. 8) (6). After loss of cell viability, a large final drop in potential occurred and rhodamine 123 fluorescence could no longer be measured above background. Before bleb rupture, the potential decreased about 30 mV or 20% of the initial potential of 160 mV (25). A similarly persisting potential has been reported in hepatocyte suspensions during anoxia (26).

Cytosolic pH and Cell Injury
Acidosis is a salient feature of ischemic tissue injury. Accordingly, we measured intracellular pH (pHi) in 24-hr cultured rat hepatocytes by ratio imaging of BCECF fluorescence using MDVM (16). In normoxic cells, pHi closely followed changes of pH,.
Following chemical hypoxia at pH, of 7.4, pHi decreased by more than a unit within 10 min after the addition of the metabolic inhibitors (Fig. 9). There was little spatial heterogeneity of pHi in either normoxic or hypoxic cells (Fig. 10). After 20 to 30 min, pHi began to rise, and cell death, as indicated by propidium iodide nuclear staining, ensued after another 10 min (Figs. 9 and 10). At acidic pH, (Fig. 9) or when Na+/H+ exchange was inhibited with choline ( Fig. 11), pHi decreased to similar values during chemical hypoxia, but the duration of intracellular acidosis was prolonged an additional 20 to 30 min. Cell survival was prolonged to a similar degree.
Monensin prevented an acidic pHi from forming and accelerated the onset of cell death (Fig. 11). ATP depletion measured in cell suspensions at neutral and acidic pH, was identical and could not explain the prolonged cell survival (18).
These results indicate that intracellular acidosis develops rapidly in hepatocytes during ATP depletion, but pHi rises shortly before cell death. Inhibiting Na+/H+ exchange or placing cells in an acidic pH, prolongs intracellular acidosis and delays the onset of irreversible injury. Thus, intracellular acidosis protects against cell death from ATP depletion-a phenomenon that may represent a protective adaptation against hypoxic and ischemic stress.
Toxic and Oxidative Injury Cytosolic Free Ca2" during Toxic ijury with HgC12 and Cystamine In contrast to chemical hypoxia, exposure of cells to HgCl2 resulted in a large increase in free Ca2+, and the cells died more rapidly (Fig. 12) (19,20). Bleb formation was also rapid and actually preceded the rise in calcium (Fig. 12, arrow, top panel). A spatial gradient of calcium also developed in these cells with blebs having higher concentrations of free Ca21 (exceeding 1 ,uM) than the rest of the cell (Fig. 13). The source of this increased Ca2' was extracellular, since in low Ca21 medium, cytosolic free Ca2' did not change after HgCl2. The rate of cell killing, however, was unchanged.
Bleb formation also occurred in cells exposed to FIGURE 10. cystamine with bleb rupture occurring after 40 to 70 min (20). Similar to chemical hypoxia, cytosolic free Ca2" did not increase prior to blebbing or the onset of cell death (Fig. 12) nor did any spatial heterogeneity of Ca2' concentration develop (Fig. 13). These studies demonstrate that a rise in cytosolic free Ca2' is not a prerequisite for bleb formation or the final common pathway to cell death.

Mitochondrial Membrane Potential
A decrease of rhodamine 123 fluorescence also occurred early in HgCl2 toxicity and actually preceded any change in free Ca2' (Fig. 14). The half-maximal decrease of fluorescence occurred after about 6 min in both low and high Ca2' mediums. Redistribution of rhodamine 123 fluorescence from mitochondrial to cytosolic compartments preceded the decrease of total cellular rhodamine 123 fluorescence (Fig. 15), indicating an even earlier collapse of the mitochondrial membrane potential. The very early mitochondrial response suggests that HgCl2 may target mitochondria for its toxic effect.

Conclusion
Our data indicate that the transition from the reversible stage of injury occurred simultaneously with bleb rupture and the onset of cell death, as assessed by propidium iodide uptake and loss of trapped cytoplasmic probes. A rise in cytosolic free Ca2' was not the stimulus for bleb formation and was not a requirement for the progression to irreversible Time (minutes) FIGURE 12. Cytosolic free Ca2" with 50 jM HgCl2 or 5 mM cystamine. Cytosolic free Ca21 was measured in single hepatocytes using Fura-2 ratio imaging and MDVM. Cells were also examined by phase microscopy to determine the onset of blebbing, as indicated by the arrows. The dotted line in each panel represents the time period over which individual cells lost viability in each experiment. Asterisks (*) signify the time at which all cells had lost viability. FIGURE 13. Pseudocolor map of cytosolic free Ca2' after toxic injury to cultured hepatocytes. Top, a hepatocyte was exposed to 50 jM HgC92.
Bottom, a hepatocyte was exposed to 5 mM cystamine. Note the increase in free Ca2' after HgCl2 which was especially prominent in blebs (red and yellow colors in third panel from left, top row). However, initiation of bleb formation occurred prior to the increase of free Ca21 (second panel from left, top row). After cystamine, bleb formation and the onset of cell death occurred without an increase of free Ca2". The final panel in each row was taken after the cells had lost viability. These images record the weak autofluorescence of the nucleus. Cytoplasmic fluorescence above background was absent. fluorescence that preceded the onset of blebbing and increase of cytosolic free Ca2+. The increase of total cellular rhodamine 123 fluorescence is attributed to early mitochondrial depolarization. Rhodamine 123 redistributed from the mitochondrial to the cytosolic compartment with unquenching of fluorescence before exiting the cell (Fig. 11).
FIGURE 15. Intracellular distribution of rhodamine 123 in HgCl2-treated hepatocytes. Before addition of HgCl2, rhodamine 123 was localized to mitochondria (A). After about 2 min of exposure to 50 jiM HgCl2, rhodamine 123 was released from mitochondria into the cytosol (B). A and B show reverse images. In A, mitochondria are bright yellow-green spheres. In B, mitochondria are pale, nearly nonfluorescent spheres showing some swelling, whereas the surrounding cytoplasm is brightly fluorescent.
injury and cell death. Rather, cytosolic pH and mitochondrial membrane potential were more important factors in the progression to irreversible injury. The question remains as to how these ATP-depleted cells maintain large gradients of Ca2" and hydrogen ions.
The simplest explanation is that anoxia induces a large decrease of ion permeability of the plasma membrane, but the mechanisms underlying such permeability changes remain to be elucidated.
We gratefully acknowledge the expert technical assistance in electron microscopy of Mr. Yukio Tanaka.
This work was supported in part by grants HL35490, DK30874, and AG07218 from the National Institutes of Health, and a grantin-aid from the American Heart Association, North Carolina Affiliate. G.J.G. is a recipient of an Individual National Research Service Award from the National Institutes of Health and is a Mayo Foundation Scholar.  I I I I I I I I I I I I I I I I I