Abstract
Detection of apoptotic cell death in cells and tissues has become of paramount importance in many fields of modern biology, including studies of embryonic development, degenerative disease, and cancer biology. In addition to methods that employ biochemical analysis of large populations of cells, cytochemical methods have recently been extensively used both in individual cells and in tissues. Most of these methods exploit properties of dying cells that are more or less specific for the apoptotic process. However, considerable confusion exists over the interpretation of some of these methods and their usefulness in all settings. This review attempts to summarize the more recent advances in cytochemical detection of apoptosis and emphasizes some of the pitfalls that confuse the interpretation of results of these methods.
Keywords
T
Cell Death
Cells can die by either of two major mechanisms: necrosis or apoptosis. Necrosis is the death of cells through external damage, usually mediated via destruction of the plasma membrane or the biochemical supports of its integrity. Such a death is analogous to “ldquo;cell murder.” The necrotic cell exhibits a swollen morphology and the plasma membrane lyses, releasing cytoplasmic components into the surrounding tissue spaces. This release of necrotic debris attracts inflammatory cells, leading to the tissue destruction characteristic of inflammation. The death of single cells by this mechanism might be resolvable in some tissues, but a large number of cells dying by necrosis usually results in inflammation and subsequent repair and scarring, leading to compromise and permanent alteration of tissue architecture. Necrosis can occur in a matter of seconds (Collins et al. 1997).
The other major form of cell death is based on the concept of programmed cell death. Programmed cell death is a paradigm in which cells contain a genetically coded program of elements that lead to the death of cells. The biochemical and morphological events that effect this death usually lead to a unique and highly controlled series of events. The terminal events of this process are termed apoptosis (Kerr et al. 1972) [although it has been previously pointed out that some programmed cell death may not involve the mechanisms of apoptosis (reviewed by Sanders 1997)]. Apoptosis is a much slower series of events than necrosis, requiring from a few hours to several days, depending on the initiator. This form of cell death is more analogous to “cell suicide,” in which death is the consequence of molecular signals contained within individual cells. The importance of this suicide mechanism was first delineated in studies of embryonic development. Subsequent studies have focused attention on apoptosis, however, in that the apoptotic mechanism is widely employed in tissue homeostasis and forms a major component of many pathological processes, from the response of cancer cells to chemoradiotherapy to the death of cells in neurodegenerative diseases. Although the genetically programmed nature of this pathway was emphasized by its role in development, the same terminal mechanisms become a permanent part of the cell machinery and can be activated by extrinsic signals that require no new gene expression. Therefore, most cells possess a preordained mechanism of death that awaits an appropriate trigger. The mechanics of this death are essentially the same, whether induced by new genetic signals or induced through external initiators. The manifestations of apoptosis, both biochemical and morphological, are unique and are completely different from those of necrosis.
The terminal events of apoptosis involve the activation of a specific series of cytoplasmic proteases, termed caspases. The activation of these self-catalytic caspases in the cytoplasm is tightly regulated. The initiators of apoptosis that set off this cascade of events leading to caspase activation are multiple, and two major pathways of initiation of this terminal pathway have been identified. One pathway involves so-called death receptors at the cell surface, a pathway that can directly activate upstream caspases in the cytoplasm. Another pathway involves the participation of mitochondria through the induction of leakiness of the external mitochondrial membrane, leading to the release of cytochrome c into the cytosol. These two pathways also intersect, in that the death receptor pathway can be amplified through mitochondrial damage. Downstream from these initiator mechanisms are terminal caspases that lead to the morphological and biochemical consequences of apoptosis.
The historical recognition that apoptotic cell death was a unique series of events was initially based on morphological evidence of changes in cell structures, especially the segmentation of nuclei. Other studies have shown that cells in tissue culture usually go through a unique series of surface morphological changes (e.g., Collins et al. 1997), as shown in Figure 1. Although the specific timing and the shape changes in cells vary from cell type to cell type, the overriding message that has evolved over the past decade is that apoptosis is a reflection of an extremely highly conserved mechanism that shows remarkable uniformity through a long evolution. This conclusion has allowed extrapolation from one species and one system to others, and from cell culture to intact tissue. The final mechanics of the apoptotic events are common but the initiating pathways into apoptosis are myriad and complex. In a sense, this makes the task of identifying apoptotic changes easier because they result in a common effector pathway, but the context in which they occur has to be interpreted with caution. Table 1 summarizes features of the apoptotic process that have been exploited for detection in individual cells, in mass cell cultures, and in tissues.
Surface morphological features of apoptotic cells in culture. These images demonstrate the hallmark sequential features (arrows) of apoptotic cells detected by phase-contrast microscopy, including blebs, echinoid spikes, and surface blisters. These images were generated using KB human carcinoma cells induced to go into apoptosis using ricin (Collins et al. 1997).
Categories of cellular changes that form the basis of apoptosis assays
Detection Methods in Individual Cells and in Tissues
Much of the knowledge about apoptosis mechanisms has derived from observations of isolated cells, such as tissue culture systems. However, the properties revealed in these assays are not always applicable to the study of intact tissues. For example, assays such as the binding of annexin V or the impermeability of propidium iodide depend on the ability to incubate intact impermeable cells with the reagent, a process not possible in fixed tissue sections. Assays based on the size of isolated cell fragments are not applicable to fixed tissue. The kinetic context related to the synchrony of apoptosis, the length of time necessary for the initiating event to lead to the changes that occur, and the removal of apoptotic cells in vivo through phagocytosis often make the assays of apoptosis that are useful in tissue culture less applicable to intact tissues. Furthermore, events that occur in individual cells can sometimes be implied from assays that are applicable only to the examination of large numbers of cells at one time. An analogy might be the ability to detect the low-level expression of products of specific transfected genes in mass culture through the use of sensitive enzyme assays (e.g., CAT assay) as opposed to the cytochemical identification of high levels of the specific protein product in a small number of individual cells (e.g., a GFP vector). These two approaches can sometimes yield conflicting results, often based on the heterogeneity or lack of synchrony of events in single cells vs the entire population. This has been a major point of confusion in the field of apoptosis.
Apoptosis Is Not Always Synchronous
A major and often unappreciated aspect of apoptotic cell death is that cells within a population may begin apoptosis at very different times after the addition of an initiator, and the length of the various stages of apoptotic morphological change can vary from cell to cell (Collins et al. 1997). For example, for apoptotic systems that depend on cell cycle alterations, the time from initial addition of an agent to the beginning of the first stages of apoptosis can be several days, and the actual completion of apoptosis to the point of cell lysis can require an additional day. The variability introduced can result in a highly asynchronous process. Some assays detect changes that occur early in the process, whereas other assays detect changes that occur only very late. Some changes are not stable and appear at one point in time, only to disappear as apoptosis progresses. Furthermore, the lysis that eventually occurs at the end of apoptosis is essentially the same membrane permeability event that occurs in necrosis, although the specific changes up to that point are very different. These problems point out the need for thoughtful analysis of the kinetics and synchrony within a system and the appropriate choice of assays that can detect those changes accurately.
Surface Morphological and Biochemical Changes
For tissue culture systems, a good point at which to start the analysis of apoptosis is the use of video time-lapse microscopy, a simple and clearly interpretable approach to defining the kinetics of apoptosis in cell culture (Fan et al. 1996; Pulkkinen et al. 1996; Collins et al. 1997). The usual sequence of morphological changes in adherent cultured cells are shown in Figure 1. These include the following: (a) a loss of adhesion to substratum, resulting in cell rounding; (b) a flurry of surface zeiotic blebbing, which may last for only a few hours (Laster and Mackenzie 1996); (c) shrinkage of the cell, with cessation of blebbing and other movements; (d) in some cells, the slow protrusion of elongated “echinoid spikes” from the cell surface (Collins et al. 1997); and (e) after a long delay of several hours, the “blistering” of the cell surface membrane with eventual lysis, similar to necrosis. Even these changes must be interpreted carefully, however, because some initiators (such as staurosporine) have profound effects on the motility mechanisms in cells, suppressing the normally dramatic series of surface morphological changes (Cartee et al. in press). The dynamic nature of the initial rounding and blebbing events makes them easy to spot in time-lapse recordings, and these events are a convenient point at which to designate the beginning of apoptosis. Unfortunately, evaluation of these events in intact tissue is not possible by this method.
Externalization of phosphatidylserine (PS) and phosphatidylethanolamine is a hallmark of the changes in the cell surface during apoptosis (Koopman et al. 1994; Martin et al. 1995; Emoto et al. 1997; Lecoeur et al. 1997; O'Brien et al. 1997; van Engeland et al. 1997). Annexin V binding can be used as a marker of PS externalization, using either microscopy or flow cytometry with fluorescently labeled annexin V (Zhang et al. 1997; van Engeland et al. 1998). These phospholipids are normally sequestered within the cell on the cytoplasmic face of the plasma membrane and other internal membranes. The exact reason for the appearance of PS on the external surface is not clear. In experiments to be presented elsewhere, we have examined this externalization and have found that annexin V binding appears to occur only on a minority of cells (less than one third in our system) during apoptosis. This occurs relatively early, just after the segmentation of nuclei (summarized in Figure 2). Other investigators have found similar labeling of subpopulations with annexin V (Boersma et al. 1996; Ferlini et al. 1997; Gatti et al. 1998). It is not clear why only a minority of cells show PS externalization during apoptosis, or whether they are in some way different from those that do not. However, it has been shown that inhibition of caspase activity also blocks the appearance of PS on the cell surface (Rimon et al. 1997). This suggests that the externalization of PS is caused by the same caspase activities that cause the other manifestations of apoptosis. Care must be exercised with the use of this assay, because the final lysis stage of apoptosis allows access to the PS that is normally inside all cells, creating a highly labeled cell fragment. Such labeling would also be evident in necrotic cells.
The permeability of the plasma membrane is also a central difference between necrosis and apoptosis. Large molecular weight DNA binding dyes, such as propidium iodide (PI), cannot enter intact cells because of their large size and, without permeabilization treatments, do not label apoptotic cells until the final lysis stage. On the other hand, smaller dyes, especially those that can attach to DNA, can label both apoptotic cells and normal cells. Using flow cytometry, one can distinguish apoptotic from necrotic cells as those that show internal DNA labeling with a small dye (such as DAPI, Hoechst 33342 or 33258, or calcein-AM), while not labeling with PI (Bussolati et al. 1995; Gatti et al. 1998). As cells generate apoptotic bodies that release some of their segmented nuclei, the strength of the internal DNA signal decreases, and this can be used as an assay of apoptotic progression. When the cell finally lyses and its membrane becomes permeable, then the larger-sized markers (such as PI) can label any DNA left within the cell. For microscopy, the altered morphology of nuclei selectively labeled with DNA-specific fluorescent dyes, such as DAPI or Hoechst dyes, can be a very easily interpreted assay for apoptosis.
Nuclear Morphology Changes and DNA Changes May Not Be the Same
An early observation concerning apoptosis was that cells that entered apoptosis from nonmitotic parts of the cell cycle showed dramatic and characteristic changes in nuclear shape and organization. It is perhaps correct to say that the characteristic change in nuclear morphology is the most accurate indicator of the involvement of apoptosis in the death of a cell. This is true even in the face of the ironic observation that nuclear segmentation is not actually required for other aspects of apoptosis. That is, cells that have been enucleated still undergo the other changes associated with apoptosis (Schulze-Osthoff et al. 1994). This points to the cytoplasmic location of the effector systems of the apoptotic machinery. However, in cells that contain nuclei, the change in nuclear morphology is still an early and relatively unequivocal hallmark of apoptosis. This nuclear shape change occurs at an early point in the series of apoptotic morphological events, usually soon after the beginning of surface blebbing (Collins et al. 1997).
Apoptosis detection by in situ assays in the context of morphological changes. The appearance of positive signals during apoptosis for in situ end-labeling of DNA strand breaks is very dependent on technical details, including fixation conditions and pretreatment procedures. For some studies, positive cells appear early or even before other apoptotic features, whereas in other studies the positive signals occur only late in the process.
In addition to changes in nuclear morphology, loss of DNA integrity also characterizes apoptosis. It was assumed that these two events were in some way related, as they may be. However, there is no compelling reason necessarily to assume that they are. When DNA extracted from apoptotic cells was analyzed using gel electrophoresis, a characteristic internucleosomal “ladder” of DNA fragments was found. Larger DNA fragments have also been seen at earlier times in apoptotic cell cultures. These gel results have been used as a hallmark of apoptotic detection. However, such analysis requires the extraction of DNA from large numbers of cells. In addition, the apoptosis must be relatively synchronous for this analysis, a synchrony that is not always present. It has not been directly demonstrated that DNA fragments in this way in intact cells. That is, it is possible that the internucleosomal breaks in DNA occur after or during DNA extraction procedures because of the potential fragility of caspase-treated DNA in removing other components of chromatin structure. Furthermore, the detection of strand breaks may be so sensitive that only a small number of apoptotic cells may be needed in a population to produce a detectable signal. That is, the entry of the majority of cells into apoptosis might occur at a much later time point than the first detection of “ladders.” Even so, these observations led to the development of in situ assays for the presence of single-strand or double-strand breaks in DNA. The interpretation and application of these methods have been somewhat controversial. Regardless of the specific methods used, the change in nuclear morphology often does not coincide with the appearance of detectable strand breaks in every cell.
The Terminology of DNA and Nuclear Changes in Apoptosis
Although many cytochemical methods are based on the detection of strand breaks or on the appearance of DNA fragments in gel electrophoresis, there is no reason to assume that nuclear morphological changes and detectable DNA strand breaks occur at the same time. As a point of terminology, there has been considerable confusion generated by the term “nuclear fragmentation.” This term mixes the concepts of the generation of DNA fragments (DNA fragmentation) with the changes in nuclear shape. A better alternative, which I prefer, is the use of the term “nuclear segmentation.” This describes the change in nuclear shape in which the normally round or oval nucleus segments into smaller, compact, homogeneous, variably sized chromatin masses, the characteristic change seen in apoptosis. This term does not imply any generation of DNA fragments or strand breaks but refers only to the shape of the nucleus. Because there is reason to believe that the shape and the DNA breaks may not necessarily be related in each cell, it appears better to use this more precise term to describe shape changes.
When Do DNA Strand Breaks Appear in Situ?
The major methods developed for detection of DNA strand breaks involve the detection of 3′-OH ends of single-stranded DNA (in situ end-labeling; ISEL). Addition of labeled nucleotides to these ends, either using E. coli polymerase (or its Klenow fragment) by in situ nick-translation (ISNT) (Figure 3) or using terminal transferase (TUNEL) (Modak and Bollum 1970, 1972; Gavrieli et al. 1992; Hall et al. 1994; Tatton et al. 1998), allows the cytochemical demonstration of free DNA strand ends (Ansari et al. 1993; Fujita et al. 1997). Another method for detecting these single strand ends is the use of monoclonal antibody reactive with single-stranded DNA (Naruse et al. 1994; Frankfurt et al. 1996a,b). Such single strand ends are not specific for apoptosis and can also be seen in necrotic cells. Double strand breaks can be detected by other methods, such as the use of hairpin oligos (Didenko et al. 1998). A clear conclusion from the use of such methods, however, is that the preservation methods used to prepare cells can have dramatic effects on detection of strand breaks (Negoescu et al. 1996; Hikim et al. 1997; Labat-Moleur et al. 1998; Tateyama et al. 1998). False-positivity can be generated through nonapoptotic mechanisms (Kockx et al. 1998; Mizoguchi et al. 1998). On the other hand, inaccessibility of sites can prevent the detection of strand breaks (Nakamura et al. 1997). This has led to the use of protease treatment to unmask such inaccessible sites. This is somewhat ironic because the process to be detected is the consequence of endogenous protease activity. It has been suggested that exogenous protease may render sites in apoptotic cells accessible specifically because of the increased fragility of apoptotic DNA that has been affected by activated caspases in the cell. However, the impact of this exogenous protease treatment on the specificity of labeling is not clear. In some cases, methods have been sufficiently sensitive to detect DNA repair mechanisms that are independent of apoptosis. Depending on permeabilization and fixation protocols, some methods detect so-called preapoptotic nuclei in which strand breaks are detected but no nuclear morphological changes have occurred. In other cases, the positive detection of strand breaks does not correlate with individual cells that show nuclear segmentation. In other cases, detection of DNA strand breaks occurs only very late in the process, corresponding to the lytic stage at the end of apoptosis, far removed from the changes in nuclear morphology (Collins et al. 1997). Therefore, although these methods are useful, they are neither theoretically nor practically specific, as yet, for apoptosis. Many authors have pointed out the need for simultaneous interpretation of these in situ methods along with apoptotic nuclear shape changes. The major advantage of these methods, however, is that they are applicable to intact tissue sections directly.
Apoptotic Cells May Not “Live” Long In Vivo
Dying cells in vivo are removed from tissues by phagocytic cells, a process that emphasizes the usefulness of the apoptotic mechanism for normal tissue remodeling. However, the removal of cells can occur with different efficiencies in different tissues. Massive apoptosis can overwhelm the phagocytic potential of a tissue, and virtually all apoptotic cells will remain for extended periods. In the case of limited apoptosis, however, phagocytic or other adjacent cells may remove the apoptotic remnants as rapidly as they are generated. Unlike cell culture, in which the cell remnants remain undisturbed, in vivo apoptosis may yield only a hint of an occasional cell remnant in a phagocytically active tissue. This problem makes in vivo measurements and interpretation of apoptosis by in situ cytochemical methods extremely difficult.
Detection of DNA strand breaks in apoptotic cells using ISNT. KB cells induced to enter apoptosis by ricin treatment were fixed, permeabilized with detergent, and labeled using in situ nick-translation (ISNT) with rhodamine fluorescence detection. Nuclear segments (
Other Biochemical Changes
Recent studies have revealed many other steps in the apoptotic machinery inside cells. Among these are the activation of caspases and the cleavage of specific caspase substrates. Incorporation of specific caspase substrates into living cells and the detection of cleavage products (e.g., using fluorescence resonance energy transfer) have been presented as new assays of apoptosis (Xu et al. 1998). Whether or not this will provide better insight into the process at the cytochemical level is not yet clear. My own experience with one of these substrates for caspase 3 (Phiphilux; OncoImmunin, College Park, MD) has been that the signal generated in cells appeared to occur only late in the apoptotic process (results not presented). The other aspect of these assays is that they would probably not be useful for intact tissues.
Methods for detection of activity of endogenous caspase substrates (such as PARP) have also been presented recently (Davis et al. 1998). On the other hand, a polyclonal affinity-purified antibody (CM-1) has been recently described that recognizes a cleaved subunit of caspase 3 (Srinivasan et al. 1998). With this antibody, apoptotic neurons could be selectively detected in cultures and in paraffin sections of mouse embryos. The point in the apoptotic process at which this antibody reaction can be detected was not precisely determined, and some cells appeared to undergo apoptosis without becoming reactive for this antibody. However, this approach may be a promising one for identification of apoptotic cells in intact tissues.
Another example of antibodies specific for caspase-generated cleavage products was a study describing an antibody specific for a caspase-generated fragment of actin, termed “fractin” (Yang et al. 1998). In this study, individual apoptotic neurons could be identified on the basis of their fractin content. Studies in vitro suggested a surprisingly rapid time course for the appearance of this fragment when colchicine was used as an apoptosis-inducing agent, perhaps more rapid than one would expect in most cells. Whether this reflects a cell type-specific sensitivity or changes that could occur in the context of morphological events independent of apoptosis will require further study. The authors suggested that fractin might reflect events involved in surface blebbing at the beginning of apoptosis. Blebbing itself, however, is also seen in other settings, so an important clarification will be whether this fragment appears selectively only in blebbing induced by apoptosis or whether it might also appear in other contexts independent of cell death. Even so, this type of approach may provide important insights into the design of new assays useful for tissue sections.
Transglutaminase activation is another biochemical change that has been proposed as an indicator of apoptosis, but whether or not this would be applicable to all systems is not yet clear (Kashima et al. 1997). Certain surface and intracellular antigens have been identified that are expressed as part of the apoptotic process, and these may in the future present new avenues for cytochemical assays.
Mitochondria have been implicated in the apoptotic pathway due to many death inducers, even those that act primarily through surface death receptors but are then amplified through mitochondrial damage. The primary event proposed as important for apoptotic induction is the leakiness of the external membrane of the mitochondrion, resulting in leakage of cytochrome c into the cytosol. This is different from the overall loss of membrane potential of the entire mitochondrion (the permeability transition) as measured by vital dyes such as rhodamine 123 or Mitotracker, but both events may occur in dying cells (Lemasters et al. 1998). Therefore, cytochemical detection of cytochrome c localization is a potential assay for apoptotic induction, although the amount necessary for apoptotic death (how many mitochondria must be leaky for death) is not yet clear. Furthermore, there is some evidence that this leakiness may be reversible, especially in cells in which caspases are inhibited. Antibodies have been identified that detect conformational changes in cytochrome c that appear during apoptosis, but the timing of this change, and whether this might be exploited as a general histochemical assay of apoptosis, is not yet completely clear (Jemmerson et al. 1999; Varkey et al. 1999). Measurements of the metabolic activity of mitochondria have also been used to measure cell death, but this would not necessarily be specific for apoptosis. Some internal antigens in mitochondria (Zhang et al. 1996; Koester et al. 1997) are inacessible in intact mitochondria that are well preserved by fixation. Therefore, the accessibility in apoptotic cells for these antigens in the cytosol (selectively detected using permeabilization with saponins) has also been presented as a potential assay for apoptosis, especially using flow cytometry (Koester et al. 1997). Although many of these methods deal with important mechanisms during the apoptotic process, they still are not easily applicable to detection of apoptotic cells in intact tissue sections.
Conclusions
This brief review emphasizes some of the recently employed cytochemical methods for the detection of apoptosis in cells. Unfortunately, there is no single assay that can be used blindly with perfect specificity and sensitivity. Ironically, the most specific assay is perhaps the oldest, the detection of nuclear shape changes in the early stages of apoptosis. In combination with other methods, this morphological interpretation usually allows a relatively accurate interpretation of apoptosis. For tissue sections, many investigators recommend labeling of DNA strand breaks (ISNT, TUNEL, anti-SS DNA) together with analysis of nuclear morphology. Although detection of DNA strand breaks is a cornerstone method for use in tissue sections, it requires care in interpretation, especially in the details of cell fixation, permeabilization, and processing. For cultured cells, a direct and easily interpretable assay for apoptosis is the observation of surface morphological features with time-lapse microscopy. For flow cytometry using DNA binding dyes, detection of apoptotic bodies as a pre-G1 peak is a simple and rapid assay for large numbers of cells. Labeling of surface-exposed phosphatidylserine is a useful although imperfect method for both microscopy and flow cytometry. Perhaps the only comforting conclusion from these observations is that the mechanism of apoptotic death appears to be extremely highly conserved in eukaryotes. As a consequence, the development of new techniques will have to deal with the many facets of only a single, albeit complex, biological mechanism.
