I. INTRODUCTION

Top Previous Next References

A.  Defining Cell Death

1.  Cell death is an active process

This review is geared toward those working in the field of ischemic tissue damage and toward students and others who want to develop a critical overview of the area. The text carefully examines the evidence that certain chemical and ionic changes, protein and enzyme changes, and changes in key functions are involved in ischemic cell death. The long-term goal of this review is to make as cogent a model of ischemic cell death as is possible at present, which is done in the final section, and to highlight what is not known to excite rigorous, causally directed, experimental work. The latter is done throughout the text.

To study what causes cell death, cell death has to be recognized. However, there is no well-accepted definition of the point at which a cell dies (706, 1137). The most satisfactory definition of cell death, and the one that would be the most valuable to know therapeutically, is the point at which the cell becomes unable to recover its normal morphology and function even if all processes leading to dissoution are stopped pharmacologically (the point of no return). This would be analogous to cessation of breathing or cardiac function for death of the organism. At this stage, however, we do not know enough about the ischemic death process of a cell to be able to identify what the point is. Thus the only unequivocal definition of cell death is a morphological one, the elimination of the cell, which can arise either by cell disintegration or phagocytosis. These are both very clear end points, and they are often used in studying mechanisms of cell death. Before this there are certain metastable morphological states toward which dying cells tend that are almost certainly precursors of disintegration or phagocytosis (121, 814) and that are also commonly used as end points. In this review these are termed the end stages of the cell death process and are generally considered to represent dead cells, undergoing necrosis, autophagocytosis, or the end stages of apoptosis (Fig. 1, column 5). As defined in this way, cell death is an anatomical event that represents a state from which the cell cannot recover to even near normalcy. The state reflects major macromolecular change, and the study of cell death, at this stage, is the study of what leads to these drastic macromolecular changes.

View larger version (91K): [in this window] [in a new window]   Fig. 1. Pathways of ischemic cell death. This is an overview of processes involved in ischemic cell death that is consistent with organization of review. It illustrates major events that are hypothesized to contribute to cell death and also the extremely complex interactions between these events. This figure is not meant to be complete, but does include processes known or thought to be important at this stage. Column 5 lists 5 principal morphological forms taken by dying or dead cells after an ischemic insult. Determining how these end stages are reached is the ultimate goal of research on ischemic cell death. Column 4 lists 6 long-term functional or structural changes, all of which, except changes in membrane permeability, are known to occur as a result of ischemia. It is hypothesized that one or more of these account for one or more forms of ischemic cell death. As indicated by lack of 1:1 connections in figure, no direct causal sequence has been established between any of functional changes in this column and particular end stages shown in column 5. Potential relationships are discussed extensively in text (see sects. III and IV). Column 3 lists actions that are likely to cause long-term functional changes described in column 4. These are termed "perpetrators" because they are considered to be key damaging events in ischemic cell death. No direct effects of perpetrators on critical functional changes are insinuated in figure because none has been completely established. Many such effects are, however, reasonably well established and are discussed in text. Columns 2 and 1 show changes in many variables, initiated by original inhibition of electron transport, whose most important end result is considered to be activation of perpetrators, but which may also have more direct effects on cell damage, that are not shown. Major outputs of these changes, lumped together as initiators and activators, are changes shown in column 2, which activate perpetrators. Direct causal interactions between specific changes are shown for these earlier events because far more is known than about interactions at later stages of the process. However, all these interactions are not established with same degree of assurance, as discussed in text. These causal interactions are indicated by including changes within same toned horizontal band, as shown for columns 2 and 3. For example, both increased nitric oxide and free radicals lead to damaging actions of free radicals and peroxynitrite. They are also indicated by including causal changes within a box whose outline color is same as that of variable they are changing. For example, both increased calcium and gene activation contribute to formation of increased nitric oxide. This way of describing events indicates the very high level of interaction between different changes caused by ischemia, involving many positive-feedback loops. Interpreting the figure: as indicated above, colored boxes encompass different changes that cause change in variable represented by that color. For example, increased nitric oxide (peach color) results from increased calcium and gene activation. Colored horizontal arrowheads represent all events within box of that color. For example, red arrowhead (for ATP) represents loss of oxygen and increase in sodium, which are enclosed within red box associated with ATP (near bottom of figure). Using this notation allows the very large connectivity of system to be represented in a manageable way. For example, changes putatively contributing to increase in sodium, enclosed within green box at top of figure, are extremely numerous when "contents" of each arrowhead are considered. In addition, changes contributing to increase in cytosolic calcium, and hence calpain activity, are all those included in boxes and arrowheads in top gray band. Large number of positive-feedback cycles is readily seen. Depol, depolarization; pHi, intracellular pH; Nai, intracellular Na+; Cai, intracellular Ca2+; FFA, free fatty acids; PAF, platelet-activating factor; e Transport, electron transport.

Ultimately, it will be important to identify the point of cell death, defined above as the point of no return, because some agents might block the major morphological change but not block the irreversible damage to the cell. It seems reasonable that prolonged damage to the plasma membrane, mitochondria, cytoskeleton, or protein synthesis could each be a point of cell death defined as above, and section IV is devoted to evaluating their roles in ischemic cell death.

B.  Major Features of Ischemic Cell Death

1.  Three major modes of cell death

There appear to be at least three distinct modes of cell death that participate in ischemic cell death (149, 205, 706). Two of these, apoptosis and autophagocytotic cell death, clearly involve ordered physiological processes in that new structures are formed that are integral to the process, and the progress toward cell elimination is stereotyped. Furthermore, they are a part of normal cell function, during development and other times (205). There is some conflict as to whether autophagocytosis can be profound enough to cause cell death (149, 205, 835), but there are quite persuasive arguments by Clarke (205) that it is. These include the high percentage of neuronal area that is covered by autophagic vacuoles during some developmental cell death (205), as well as the ability of 3-methyladenine, a phosphatidylinositol kinase blocker (101) which prevents autophagocytosis, to completely prevent cell death in at least one case (149).

A third category of cell death, which has been termed necrosis or necrotic cell death, is fundamentally different from apoptosis and autophagocytotic cell death in that it only occurs after an exogenous insult; furthermore, no new structures are elaborated as part of the process. Thus it does not appear to be a programmed, or a normal, physiological process. Because the process is not yet well characterized, it is best defined now as cell death that is neither apoptotic nor autophagocytotic.

There are at least two quite different forms of this cell death; by far the most common is ischemic/homogenizing cell change, where the cell first shrinks dramatically and becomes very electron dense. Less common is edematous cell change, where the cell swells greatly and organelles lose their form. There is no satisfactory term for these modes of cell death. It is widely termed necrosis or necrotic cell death, but necrosis is really the "decay" process after the point of cell death and occurs in all forms of cell death (1137). However, for the sake of continuity, necrotic cell death and necrosis will be used in this review. At this stage, it is not completely known why a particular mode of death occurs in a given cell as a result of a particular ischemic insult. However, it is apparent that one insult can spawn more than one mode of death in the same population of cells and also that in many cases one cell will manifest signs of more than one mode of cell death (205, 716). Thus neurons have the potential for exhibiting all modes of death in response to an ischemic insult. The pathway taken must be a function of the nature of the insult, the cell type, age, and very probably the state of the cell at the time of the insult (716). This issue is considered in section III.

In addition to its multimodal nature, ischemic cell death is also characterized by a long delay between the insult and manifestation of major cell damage. This delay varies greatly, depending on the nature of the insult and the brain region being affected. In some cases it is as long as several days or even weeks (276, 582), whereas in others it is a few hours or less (742). It is clear that, all other factors being equal, the greater the energy deprivation, the less time it takes for damage to develop (579). The very prolonged delay is quite an extraordinary phenomenon; it shows that very short-lived, albeit profound, derangements of metabolic patterns can initiate the development of damage and cell death that takes place, between a day and 3 wk later. Remarkably, normal homeostatic mechanisms are unable to prevent this process. Thus lethal ischemic insults are those that move cells far enough from their steady state to eliminate the possibility of reattaining that steady state; sublethal insults produce early damage, but this is able to be recouped by the cell. One very important and unresolved issue is what the basis is for this "destabilization" of the cell. It is discussed theoretically in section IX.

C.  Overview of Ischemic Cell Death

1.  Stages of ischemic cell death

Ischemic cell death is initiated by changes that result directly from inhibition of oxidative phosphorylation and create an early maelstrom of activity. These changes include decreased pH, decreased ATP, initiation of free radical production by the mitochondrial chain, increased cell Na⁺ and membrane depolarization as a result of the loss of ATP substrate for the Na⁺-K⁺ pump. These lead to secondary changes in ion and chemical concentrations which, in concert with the initiating changes, result in activation of damaging processes. These damaging processes are termed "perpetrators" in this review and are characterized by their abilities to produce long-term changes in macromolecules. Activation of one or more of the perpetrators is considered the key step in necrotic aspects of ischemic cell death. The perpetrators include proteases, phospholipases, and free radical actions. The macromolecular changes they cause, and the ensuing functional changes, are considered to be the proximal causes of cell death. At present, the way that ischemic apoptosis is activated is not well enough understood to include the same perpetrators in that process with any certainty, although caspases certainly are involved. Thus the process of cell death has at least three major stages: an early development of ionic and chemical changes, a resulting activation of perpetrators, and a subsequent change in critical functions and structures that denote or lead to cell death by one of three different routes.

The overall process is extremely complex due to the large number of interactions between the variables, illustrated by Figure 1. Although incomplete, Figure 1 provides a good working description of the probable flow of damage and the very complex interactions between different changes.

Two very important limitations in our understanding are not apparent from Figure 1.

The first is the degree of certainty that the different changes shown are actually involved in ischemic cell death. It is fairly certain that most, or all, of the activators and perpetrators in Figure 1 are actually involved in ischemic cell death, although, as will be seen, experimental evidence is often surprisingly weak. Much less is known about how these changes lead to cell death (Fig. 1, columns 4 and 5). As will be seen, the identities of the critical functional changes, how they actually might lead to cell death, and the end stages are little known for any of the three modes of cell death.

The temporal aspect of cell death is also missing from Figure 1. End stages or cell disintegration develop anywhere from 6 h to several days after the ischemic insult. The basis for these time delays is not yet known. Gene activation and subsequent synthesis of damaging proteins such as cytokines, inducible cyclooxygenase-2 (COX-2), inducible nitric oxide (NO) synthase (iNOS) may be important. However, it is also likely that there are long-term low-level activations of perpetrators when the insult is relatively mild so that major changes in cell function occur only after prolonged periods. The factors that determine the temporal aspects of ischemic death are poorly understood at the moment.

Figure 1 provides a framework for the organization of the review. After the different model systems used for studying ischemic damage are considered (see sect. II), the main body of the review begins from the right-hand side of Figure 1 and essentially walks backwards; for example, section III describes the end stages of ischemic cell death. Knowing the downstream events makes it easier to evaluate the role of a particular change. For example, understanding that mitochondrial dysfunction may be a mediator of cell death (see sect. IV) provides a focus for examining the targets of free radical action. Despite the connectivity of the different major parts, each section is designed to be usefully read independently of its order of appearance and to fully examine the possible contribution of a different variable to damage.

The achievable goal of the review is to critically evaluate the involvement of different variables in ischemic cell death. This entails describing the changes in the variable that occur during and after ischemia, determining how they are likely to occur, critically assessing pharmacological and gene manipulation studies done to show that the change contributes to cell death, and then trying to understand the mechanism by which the change is contributing to cell death.

In section IX, a still speculative attempt is made to synthesize the different sections and describe probable sets of events that lead to cell death in the different experimental models. It is hoped that this will provide useful directions for future studies.

	II. SYSTEMS USED TO STUDY ISCHEMIA

Top Previous Next References

Many experimental models are used to study ischemic damage. Mechanisms of cell damage are determined by testing effects of different manipulations on the extent of cell death in a model. To meaningfully compare results between models, the essential differences between the ischemic insults and the way damage is assessed in these models needs to be appreciated. The factors influencing development of damage in the different models need to be known, and problems in interpreting measurements of cell death have to be recognized.

The three main classes of in vivo rodent models are global ischemia, focal ischemia, and hypoxia/ischemia. In the latter, vessel occlusion is combined with breathing an hypoxic mixture; this modified Levine preparation is now almost exclusively used in young animals. Larger mammals have been used to study ischemia (534), as have nonmammalian vertebrates (453), but this review will focus on what has been learned from rodent models. In vitro models that include exposure of neuronal, neuronal/glial or organotypic slice cultures, or the freshly isolated hippocampal brain slice, to anoxia or to anoxia in the absence of glucose (in vitro ischemia) have provided important information and are discussed when they provide insight into events in vivo. Other single-cell or synaptosomal models have been used very effectively (307, 340) but will not be considered because the metabolic states of these preparations may be quite different from in vivo (307).

The degree to which blood flow is reduced directly determines the changes in three variables that are almost undoubtedly the genesis of all future changes. Two of these, inhibition of electron transport and decreased ATP, appear to be directly related to the extent of flow reduction (see Table 1). The third variable, free radical production, is likely to be a more complex function of flow rate because it is activated both by electron transport inhibition and also by oxygen so that it will be maximal at blood flow values greater than zero. (The precise relationship between flow rate and free radical production has not been determined.) Other damaging changes such as decreased pH and leukocyte accumulation are more severe in the presence of residual blood flow. Thus the amount of damage need not be a direct function of the decrease in blood flow. Elevated blood glucose enhances damage in many cases despite maintaining ATP levels so that the fall in ATP is not a reliable indicator of the efficacy of an insult either. As a further complication, the nature of damage (e.g., extents of apoptosis vs. necrosis) may be a complex function of the amount of the ATP/electron transport decrease and free radical increase and thus may be a complex function of the flow rate during ischemia (194, 870). Thus, except for increased duration of an insult, which always enhances and accelerates damage, it is difficult to estimate, a priori, the potential impact of an insult based simply on the degree of ischemia and fall in ATP.

In vivo, distinct neuronal populations have very different vulnerabilities to ischemia, and these have been reviewed relatively recently (169). For example, 5-min global ischemia caused delayed death in almost all CA1 pyramidal cell neurons with no effects on other populations, whereas 20 min caused cell death in CA3 neurons but had almost no effect on dentate granule cells in the hippocampus (579, 583) or on interneurons in CA1 (322). The basis for differential vulnerability has not been determined, although, as will be seen, there are many differences between vulnerable and less vulnerable cells that might well account for it, as described in Table 5. In addition to these, differences in presynaptic events may be responsible for apparent selective vulnerabilities of postsynaptic neurons as will be seen in section VIB. The large range of selective vulnerabilities is quite an extraordinary phenomenon. It indicates large differences in metabolism between different cell types that presumably reflect their different physiological functions and lead, serendipitously, to differential vulnerabilities to ischemia. The phenomenon has been used to provide insight into mechanisms of damage and is discussed in that context throughout the review.

The Levine model, in which hypoxia is combined with unilateral carotid occlusion (644), adapted in the mid-1960s in landmark studies by Brown, Brierly, and co-workers (121, 129, 742), established the basis for systematic study of brain cell damage in rodents. Most in vivo models now rely on vessel occlusions that predominantly affect the forebrain, developed between the late 1970s and early 1980s (331, 366, 461). They are generally divided into two categories, global and focal ischemia, but a modified Levine preparation (hypoxia/ischemia) has some properties of both models.

The three most widely used global ischemia models are four-vessel occlusion (4-VO) and two-vessel occlusion (2-VO) combined with hypotension in the rat, and two-vessel occlusion in the gerbil. The importance of transgenic mice has spawned the use of the 2-VO in the mouse also, and the latter is now being systematically studied (588, 1074). Results to date are similar to those with rat and gerbil (588).

A) 4-VO IN RAT. The 4-VO in rat involves permanent coagulation of the vertebral arteries (which has no deleterious effects) and temporary ligation of the two common carotids. Ligation of temporal muscles reduces flow more reliably (926, 928). In Wistar rats, which show loss of righting reflex within 15 s, blood flow is reliably <3% of control values in hippocampus, striatum, and neocortex (931). The electroencephalogram (EEG) generally becomes isoelectric within 30-40 s (935), and spontaneous cortical activity is abolished within 1 min when skull temperature is maintained at 37°C (1233). No anesthetic is required during ischemia, and usually none is used.

B) 2-VO IN RAT. This involves ligation of the common carotid arteries only, along with a blood pressure reduction to ~50 mmHg, and has slightly more profound effects than 4-VO. Blood flows fall to ~1% in hippocampus, neocortex, and striatum (1055, 1056), and the EEG generally becomes isoelectric within 15-25 s. However, in the hands of one group, blood flow is only reduced to 15% of control levels (1089). This less-intense insult is more easily protected against, for example, by N-methyl-D-aspartate (NMDA) receptor blockade, and this has led to some confusion in the literature that will be dealt with in section VIC. Anesthetic is used during ischemia.

C) 2-VO IN GERBIL. This is induced by temporarily ligating the carotid arteries, with no reduction in blood pressure. Because there are no posterior communicating arteries in gerbils, this produces profound forebrain ischemia (583). Changes are similar to those in the rat models; blood flow in cortex is <1% and in hippocampus is ~4% of control values (548), whereas EEG failure occurs within 20 s (1088). Damage has been successfully dissociated from the occurrence of postischemic seizures that also characterize the insult in gerbils (493). Anesthetic is not used during ischemia.

D) COMPLETE VERSUS INCOMPLETE GLOBAL ISCHEMIA. Finally, there is complete global ischemia, generally achieved by neck-cuff (259, 297, 682), cardiac arrest (228, 294), or by ligating or compressing all arteries stemming from the heart (560, 915). Blood flow to the whole brain is zero or <1% in these models (297).

Vessel occlusion is often termed "incomplete ischemia" because of the residual blood flow, and there is controversy as to whether complete or incomplete ischemia is more damaging, with the thought being that the larger fall in pH or increase in free radicals during incomplete ischemia might enhance damage. This does not seem to be the case. Complete ischemia for 10 min in rats, even without maintenance of temperature, causes complete destruction of cells in the CA1 pyramidal cell layer after 7 days (226). This damage is certainly as severe as the incomplete global ischemia models (see sect. IID3).

Overall, gerbil vessel occlusion is the most studied model because the operation is simpler than the rat. However, there are many differences between the two species, both in terms of metabolic and genetic changes during and after ischemia, and the abilities of different agents to protect against damage. Thus the processes of cell death may be different; results from one species cannot necessarily be generalized.

Global insults are generally short, between 3 and 30 min, with 5-20 min being the usual exposure times. Longer insults become lethal, and these shorter insults are quite good models for reversible cardiac arrest pathology. Levels of ATP fall quite rapidly during this period (Table 1). By 2 min, levels are 15% or lower in gerbil, and also in rat during complete ischemia. Unfortunately, there are no good kinetic studies for the rat vessel occlusion models; by 10 min, ATP is 10% or less of normal values (Table 1), but the rate of decay, which may be important in determining extent of damage, is not known.

Levels of ATP in vulnerable cell regions recover quite well after global ischemia so that delayed damage develops in the face of ATP levels that are at least 70-80% of normal for most of the reperfusion period (162, 929). Tissue metabolic rate is generally depressed by 50% or more, for at least 6 h after the ischemia (291, 306, 606, 760). The basis for this is not known, and its consequences have not been determined, but they are likely to be important.

A) DEVELOPMENT OF DAMAGE. A cardinal feature of global ischemic insults is a substantial delay between the end of the short insult and cell death. This is generally between 12 h and several days. The delay depends on the cell population and the duration of the insult. This "delayed neuronal death" was first noted and documented in rat by Pulsinelli and Brierly (926, 927) and in gerbil by Kirino (580). It is most widely studied in the CA1 region of the hippocampus, but the striatum, parts of the CA4 region of hippocampus, and layers 2 and 5 in cerebral cortex also show the phenomenon in response to short insults.

Overall, gerbil hippocampus is slightly more sensitive than rat. With body temperature at 36°C, but brain temperature falling normally, 5 min of ischemia caused massive cell death in the hilus of the dentate gyrus (CA4) and the striatum within 24 h and complete cell loss in the CA1 region of the gerbil hippocampus within 3-4 days (223, 580). Ischemia for 3.5 min caused 75% cell death in CA1 after 7 days (847), whereas 2 min caused no measurable damage in that region (543). It did cause extensive damage in the somatostatin-containing cells in the hilar region (735).

When rat brain was maintained at 36-37°C, 6-10 min of 4-VO caused near-complete CA1 pyramidal cell death in 3 days (647, 681), and 10 min of 2-VO caused almost complete cell death in 7 days (210, 519, 819, 921). Ischemia for 5 min caused 50% loss (519) compared with complete cell loss in gerbil, and 2-min ischemia caused a (very minor) neuronal loss in CA1 and CA4 (1055). As with the gerbil, hilar and striatal neurons die much more rapidly than CA1 pyramidal cells (896, 997), although, interestingly, a greater insult duration was necessary to initiate the cell death in striatal neurons than in hippocampal neurons (997). This shows a dissociation between factors that initiate the damage and those involved in its development.

Damage from successive insults can be cumulative. When spaced 1 h apart, three insults that normally gave no damage individually caused major damage (543).

B) MORPHOLOGICAL CHANGES DURING DELAYED NEURONAL DEATH. There are four major sets of changes during development of the delayed death in gerbil. Most notable is a massive proliferation of stacks of endoplasmic reticulum (ER) that peaks after ~2 days; ER fragments accumulate, with lysosomes, around the nucleus (579, 581). After ~3 days, there is a regression of the proliferated ER and the formation of a large number of autophagosomes with a very large increase in the lysosomal proteases, cathepsins B, H, and L (834). The autophagic vacuoles may arise from the expanded ER as the two have been associated during developmental cell death (205). Thus a long-term program for autophagocytosis may have been activated by 5 min of ischemia.

There is also a gradual increase in the numbers of small dense bodies ~1 µm or less in diameter in both somata and dendrites, which largely localize under the plasmalemma. Throughout the 3 days, the dendrites show progressive swelling, microvacuolization, and disappearance of microtubules, beginning immediately after ischemia (1236). After early swelling, mitochondria remain normal until ischemic cell change (ICC) or edematous cell change (ECC) become apparent.

The delayed death in the rat appears very different, although it develops at about the same rate. The extent of ER proliferation is far less (251, 581, 897) and, perhaps relatedly, no autophagasomes have been explicitly identified. The small dense bodies are somewhat more prevalent than in the gerbil and may be undegraded proteins that participate in damage (251, 582), although there is no explicit evidence.

As the duration of the insult increases, and so presumably the extent of the biochemical changes, so does the rate of cell death increase (579, 897). The nature of the process may ultimately change; ER proliferation, for example, is absent in gerbils after a 30-min insult that causes cell death in 12 h (897).

At a certain point, these rather subtle changes are transformed into quite abrupt major changes in cell structure associated with the end stages of cell death. These are discussed in section III. The mechanisms of this dramatic change are not known; presumably accumulated changes become great enough to cause what is essentially a "phase change," as discussed by Gallyas et al. (350).

Almost all focal ischemic models, whether larger mammal such as cat (1077) or nonhuman primate (1092) or rodent, primarily involve occlusion of one middle cerebral artery (366). In some cases, the carotid artery is also ligated. The insult is thus differentiated from global ischemia in two very important ways. First, even at the core of the lesion, the blood flow is almost always higher than during global ischemia so that longer insults are required to get damage. Secondly, there is a significant gradation of ischemia from the core of the lesion to its outermost boundary, and hence there are different metabolic conditions within the affected site. Because of its duration, and heterogeneity, the insult is much more complex than global ischemia, but it is an invaluable model for stroke and is thus widely studied.

In permanent focal ischemia, the arterial blockage is maintained throughout an experiment, usually for 1 to several days, whereas in temporary focal models, vessels are blocked for up to 3 h, followed by prolonged reperfusion. Despite the big difference in insults, maximal lesion sizes are comparable in the two cases, and the progression of damage, in terms of the numbers of damaged neurons and the extent of that damage over 6-72 h, is remarkably similar (1283). Nevertheless, there are clear differences illustrated by effects of different protectants so that the mechanisms of cell death may well be different. As with global ischemia, mouse models have now been introduced (434, 588) but will not be discussed explicitly.

There are variations within each of the two basic classes of insult: temporary and permanent ischemia. Several different rat strains are used (one is a spontaneously hypertensive strain that has a less extensive collateral circulation during ischemia and is more easily damaged; Refs. 137, 222, 391). Different strains show minor quantitative differences in lesion locations and sizes but qualitative differences in responses to different protectants.

Also, different sites and techniques of artery occlusion are used, to try and obtain reproducibility of flow reduction and lesion size (462, 860). There are two principal occlusion sites. In proximal occlusion, the middle cerebral artery (MCA) is occluded close to its branching from the internal carotid, before the origin of the lenticulostriate arteries (860, 1104, 1105). Appropriate care can produce lesion sizes with coefficients of variation that are <20% (860). A newer and now widely used approach to proximal MCA occlusion is the insertion of a nylon suture into the carotid artery past the point at which the MCA branches so that the latter is occluded at its origin (79, 598, 655, 798). Used optimally, the coefficient of variation in lesion size is ~8% and of blood flow is 35% (79). A similarly small coefficient of variation in lesion size, along with a large lesion of ~250 mm³ at 24 h, was obtained with the silicone-coated suture in Fischer 344 rats (44). There are some potentially important artifacts with this widely used method. Flow following temporary ischemia is somewhat compromised by the partial occlusion of the carotid arteries with the filament (79). Also, even a temporary occlusion leads to a long-term increase in core temperature at ~1.5°C due to decreased circulation and damage to the hypothalamus, at least in Wistar rats (1241, 1297). Finally, there is quite extensive damage to small arteries in the ischemic field, with loss of endothelium and muscle damage (832). This occurs after normal insults (e.g., 6-h ischemia or 2-h ischemia and several hours of reperfusion). It has been suggested that this damage may affect subsequent neuronal cell death, for example, by exacerbating the leukocyte response in the period after ischemia (435). This is not at all proven but needs to be considered.

In distal MCA occlusion, flow to the basal ganglia is not blocked so that damage is only cortical. The occlusion can be made surgically, in which case the lesion is readily reversible (137) or, less invasively, by creating a (permanent) thrombus by laser irradiation when the artery is perfused with Rose Bengal (714). Distal MCA occlusion has to be combined with occlusion of the ipsilateral carotid artery to get adequate flow reduction (122, 189).

F.  Focal Ischemia: Core and Penumbra

1.  Blood flows

In both proximal and distal MCA occlusions, there are core ischemic regions, served primarily by the occluded MCA, where blood flow is reduced to <15%, and there are penumbral regions where flow is <40%. There are also extrapenumbral cortical regions in which flow is >40%. These approximate flow-based definitions of core (284, 816, 977, 1104) and of penumbra (48, 369, 462) as well as the extrapenumbral (or peri-infarct) region are generally agreed upon, although there is still much discussion. Using flow is a simple and physiologically meaningful way to differentiate the regions, which suffer fundamentally different insults as indicated by efficacies of protectants, and major morphological changes. Given a long enough temporary (or permanent) insult, both core and penumbra become infarcted (1299), whereas the extra penumbral zone only shows death of isolated neurons. The density of cell death is not great enough to cause infarct.

In proximal MCA occlusion, the core is in the lateral portion of the caudate putamen and overlying parietal/somatosensory cortex (106, 751, 798, 1104). Most other regions of neocortex and entorhinal cortex, as well as the medial caudate-putamen, constitute the penumbra (79, 106, 751, 798). In the distal MCA occlusion/carotid artery occlusion, the striatum is spared, and core and penumbra span the parietal cortex, with penumbra occupying the more inferior regions. There is significant variation in the extent of the different regions among Sprague-Dawley rats, spontaneously hypertensive rats, and Wistar rats (48, 122, 137, 714).

With development of on-line measurements of flow using positive emission tomography (PET) scans, and the use of more detailed time course studies using iodoantipyrene, there is emergent evidence that blood flow varies throughout the penumbral region and also decays with time, although the extent of decay varies widely in different models (79, 438, 798, 1297). The implications of this decay are currently being considered. Clearly, it intensifies the insult in that region over time, and so may be important in the development of penumbral damage (325). It is considered in greater depth in the above-referenced papers dedicated to flow measurements and analysis.

The core undergoes rapid anoxic depolarization within 1-3 min with a concomitant rise in extracellular K⁺ to ~70 mM, which may show some short intermittent returns to baseline (359, 818). There is a large decrease in extracellular Ca²⁺ within 1-2 min at flow levels that are present in the core (428, 429), signifying entry into the tissue. During the reperfusion period, after 2 h temporary focal ischemia, extracellular K⁺ returns to control levels for 6 h before rising slightly (to 5 mM) for the remainder of a 24-h recirculation period. Despite this return to near normalcy, the insult produces severe damage with a large infarct including all of the core (359).

Events in the penumbra are less drastic, although they still lead to infarct. There is EEG silence (1077) and very much reduced evoked transmission (977), but there is no permanent anoxic depolarization with concomitant ion changes during the ischemia. Presumably ATP levels, which are maintained at 50-70% of normal, do not fall enough to allow the anoxic depolarization. However, there are sporadic transient depolarizations (369, 759) that are here termed "intraischemic depolarizations" (462, 818). They arise differently from anoxic depolarizations. Anoxic depolarizations are independent of NMDA-type glutamate receptors (394, 611, 635, 934, 1231) and only very slightly (1231) or not at all (635) dependent on non-NMDA glutamate receptors. In contrast, the frequency of intraischemic depolarizations is markedly dependent on both classes of ionotropic glutamate receptors (47, 188, 483, 759). The depolarizations may actually originate from glutamate release in the infarct core. Nitric oxide synthase knockout mice show reduced frequencies along with less release of glutamate in the core of the lesion (1031). The depolarizations occur very abruptly, last between 3 and 5 min, and have the characteristic form of spreading depressions. They do not occur during reperfusion (818). As discussed in section IIH, they may well enhance damage.

Levels of ATP fall to ~25% of basal values in the core and remain there between 5 min and 4 h of ischemia (328, 329, 1084, 1201) (Table 1). Later measurements have not been reported. After damaging durations of temporary ischemia, core ATP levels return to about two-thirds normal, for at least 4 h (329), although damage is as great as in permanent ischemia. Oxygen levels return to normal (801), showing that compromised flow is not the basis for the damage that occurs during reoxygenation.

Levels of ATP in the penumbra average ~50-70% of normal during ischemia (Table 1). Glucose utilization is actually elevated in the penumbral region (48, 816, 1252) at early times but falls to ~50% normal, equal to the core, by 3.5 h (1252), suggesting significant compromise of cell function at this point. Unfortunately, there are no reports of oxygen utilization during the insult, which would allow a more complete assessment of the rate of energy metabolism. Although penumbral blood flow declines during the insult, it remains far higher than in the core; infarct develops despite this flow difference.

Levels of ATP during the reperfusion period after temporary ischemia do not seem to recover much from their ischemic values (489, 590), although further measurements are needed to confirm this conclusion. Glucose utilization in some penumbral tissue is reduced to well below normal during early reperfusion, and this is considered a reliable indicator of the tissue that will later become part of the infarction (79, 1299). Presumably, the low glucose metabolism indicates severe cell damage of some kind at this early stage.

Thus, as with global ischemia, the core of the focal lesion undergoes massive ion and metabolite changes that recover quite strongly after a temporary insult. Despite this recovery, it is almost impossible to prevent damage by any intervention during the reperfusion period. It has also proven extremely difficult to prevent core damage by any single intervention, even before the onset of ischemia, when ischemia lasts for 2-3 h. The penumbra can be rescued by many interventions. These differences suggest different mechanisms of damage in the two regions. As with global ischemia, postischemic energy metabolism appears to be compromised in both core and penumbra.

In contrast to ion and metabolite changes, a recent study demonstrated very little free radical change in the core as compared with penumbra (1060). Free radicals, as measured by hydroxylation of salicylate in microdialysates, increased early during ischemia in the penumbra, remained elevated throughout 3 h, and then became further elevated with the onset of reperfusion. In marked contrast, there was no increase in the core during 3 h of ischemia. There was an increase in the core after ~4 h of reperfusion. Thus the exposure of the core to free radicals during the main period of damage development must be considered minimal. The opposite is true of the penumbra.

Unlike global ischemia, which leads to neuronal cell death in isolated regions, focal ischemia produces a contiguous mass of damaged brain tissue termed the infarct. Rather than measuring damage by counting dead cells, as in global ischemia, damage is generally expressed as the volume of the infarct, for which there are standard measurement techniques (e.g., Ref. 860). This is a rough measurement in that the appearance of infarct is predicated on a large fraction of the cells in the region being damaged; otherwise, individual cell damage is noted.

The gold standard of infarct delineation is a pallid region when tissue is stained with hematoxylin and eosin or pure Nissl stains (860). The pallor results from vacuolization of the neuropil, including swelling of glia, presynaptic elements and dendritic elements, and the loss of Nissl staining in the neuropil (355). So identified, it signifies cell damage but not necessarily cell death. This is exemplified by a delayed reduction in infarct size in animals in which iNOS is blocked (799).

In permanent ischemia, the infarct is first seen after 3-12 h (76, 137, 276, 355, 359, 814, 860). The wide time range appears to reflect strain differences, with the early infarct delineation seen in Sprague-Dawley rats (860) and the later delineation seen in Wistar rats (355), possibly because the lesion is less uniform in the latter (351, 355). The infarct begins in the core but reaches close to its maximal size, which includes core and penumbra, 6-24 h after the onset of ischemia. The infarct continues to grow after 1 day, although at much slower rate, so that between 24 and 72 h it grows by ~30% in Wistar rats (355). Thus damage continues to increase in penumbral regions.

Different durations of temporary ischemia lead to graded involvement of different regions in the infarct as measured after 1-2 days reperfusion (752). Ten to twenty minutes produces scattered dead neurons in the core (653, 752), whereas 1 h leads to infarct in the core (354, 526, 752). Infarct in the penumbra develops fully when the temporary occlusion is 2-3 h (354, 526, 752). In this case, the size of the infarct is the same as it is in permanent ischemia (e.g., Refs. 752, 1283), although it develops slightly more slowly (1283). After ~7 days, there is almost complete loss of cellular elements in the infarcted region; this is termed pan-necrosis. As little as 30 min of temporary ischemia also leads to an infarct in the core region, but this only first appears after 3 days and takes 2-3 wk to mature to its full size (276).

In summary, development of the infarct is quite rapid, but it matures over several days, indicating a significant delayed component of damage in tissue where the insult intensity is relatively small. As indicated above, expansion of the infarct indicates expansion of the region in which the large preponderance of cells are damaged. The relationship of infarct to cell death is considered in section IIG.

Diffusion weighted magnetic resonance imaging (DWMRI) delineates the core of the (future) infarct as early as 1 h after the onset of ischemia, and at later times, DWMRI quite faithfully defines the developing infarct (151, 362). It is very likely that the altered signal intensity represents regions of tissue in which there is intracellular edema (nicely discussed in Ref. 362).

Staining frozen sections with 2,3,5-tetraphenyltetrazolium chloride (TTC) (359), which is colored red by electron transport in active mitochondria, defines a region in which mitochondrial function is severely compromised. Damage is not necessarily irreversible. The region is slightly smaller than the infarct size defined by hematoxylin and eosin (H and E) at early times but overlaps the H and E staining at 24 h (955).

Finally, the loss of mitogen-activated protein (MAP) 2 immunostaining, which is almost certainly a reflection of increased cytosolic Ca²⁺ and activation of calpain (Y. Zhang and P. Lipton, unpublished data) quite faithfully follows the development of the infarct as defined by H and E staining (233). Thus several markers of cell dysfunction parallel the expansion of the infarct as measured by loss of Nissl stain, indicating that the cellular elements that are poorly stained are metabolically compromised.

It is critical to relate the infarct to cell death. Individual cell damage within the infarct develops slowly, but more rapidly than delayed cell death in global ischemia, suggesting that the prolonged focal insults have more impact than the shorter, more intense global insults. The best descriptions of this cell damage have come from Garcia and co-workers (351, 355, 1283). Cell death, defined as eosinophilic shrunken neurons (advanced ischemic cell change) or neurons showing autophagocytotic morphology (178), and also neurons showing apoptotic morphology (178, 652) (see Fig. 1), first becomes significant at ~6-12 h in the core of the lesion, with a major increase between these two time points. At 12 and 24 h, ~80% of the neurons have suffered cell death. A similar although less extensively described time course is seen in spontaneously hypertensive rats after 80 min of ligation (202). Although development of cell death in the penumbra appears to lag development in the core, it is hard to be sure, from the data provided. A less-detailed time course study found that 2 days after 2-h focal ischemia, 55% of the cells in the core and penumbral infarcts were dead, whereas 30% were damaged but did not show signs of death; 15% appeared healthy. These may represent different cell types, but that was not discussed (656).

At least for the first 72 h, the rate and extent of developments of cell death caused by 2-h temporary ischemia are the same as they are for permanent ischemia (1283). Furthermore, the number of apoptotic cells in the penumbra is the same in the two models throughout this 72-h period (785).

For the early infarct (before pan-necrosis), on which most measurements are made, the implicit assumption is thus that the uniform pallor and other indicators within the infarct largely result from neurons so severely damaged that they will die or have already died. Overall, this is justified in that a large percentage of neurons within the infarct die after 24-72 h, at least in Wistar rats. Thus, if a treatment reduces the radius of the early infarct, it is very likely to be slowing down cell death in penumbral neurons. On the other hand, in at least one case, infarct size was reduced after 48 h when delayed protective treatments were applied (1278), indicating that the majority of cells at the borders of the infarct are not yet dead after 48 h (1278). Overall, it is most satisfactory to measure infarct size after 72 h, but this is often dificult to do.

In general, there may not be a precise 1:1 correspondence between infarct size and the amount of cell death. Infarct requires a certain percentage of the cells to die (as witnessed by the peri-infarct area that has cell death but no infarct), indicating that the infarct in a certain region may be abolished but that significant cell death may well remain. It would be much more useful for determining cell death mechanisms if cell counting as well as infarct size were used to assess cell death whenever possible. This being said, the infarct size provides a reliable semiquantitative measure of the extent of neuronal death and is far easier to measure than individual cell death.

The final stage of infarct development in focal ischemia is pan-necrosis, in which the neuronal death is accompanied by glial and vascular cell death and loss of cellular elements. This does not occur in most global models that lead to widespread neuronal death, although it does occur in extremely insulting conditions, such as low pH (522) or a prolonged very intense insult (259). It occurs about 1 wk after most focal insults. The basis for pan-necrosis is not clear, but for some reason, both the endothelial and glial cells become major targets, suffering early changes and eventual death (914). In hyperglycemic global ischemia, which leads to infarct, glial pH falls to 4.3 (607), a level that can kill these cells (608). No measurements were made of endothelial cell pH. It would be very useful to measure focal ischemic effects on glial (or endothelial cell) pH to see whether drastic lowerings might explain development of the infarcts in those conditions.

During prolonged hypoxia/ischemia, which also leads to infarct, there was an early activation of pinocytotic transport across the endothelial layer, which was quite specific to conditions in which infarct was developing (895, 898). This might be important, but it is not known whether it occurs in focal ischemia. Direct exposure of brain to free radical-generating systems leads to infarction-like changes (174), suggesting that prolonged production of free radicals may be leading to pan-necrosis, but there is no explicit evidence. Thus there are reasonable lines of investigation of the pan-necrotic mechanism. It does not seem to relate to neuronal death because it occurs so late, but it might.

The bases of the intraischemic depolarizations or spreading depression-like depolarizations (818) discussed previously are not completely known; they may well be the spread of increased extracellular K⁺ and glutamate from the core of the lesion (463, 818). Their frequencies are halved by either NMDA or DL--amino-3-hydroxy-5-methylisoxazole-propionic acid (AMPA) antagonists, indicating that glutamate-mediated depolarization contributes markedly. The evidence linking the development of damage in the penumbra to these depolarizations is extensive but is largely correlative (47, 151, 462, 463, 818). For example, factors that reduce the frequency of the intraischemic depolarizations, such as glutamate antagonists and hypothermia (462, 818), as well as blockade of iNOS (1031), reduce the size of the infarct that develops in the penumbral region. The most persuasive direct evidence that they are damaging is that artifactually increasing the numbers of depolarizations during ischemia, by about twofold, using either KCl (151) or electrical stimulation (47), increased the size of infarct, and peri-infarct damage, by 30-100%, respectively. In one case, this was measured 24 h after distal MCA occlusion and in the other 2 h after the onset of proximal MCA occlusion, using DWMRI (255). Thus the artifactual depolarizations do appear to be damaging in ischemic conditions, and this suggests that the endogenous depolarizations are also damaging. Unfortunately, it is very difficult to specifically target the latter with a drug, to more directly establish their importance.

The basis for the putative damaging effect of the depolarizations is not known; similar magnitude and duration depolarizations are not damaging in nonischemic tissues (818), although in those cases there is an increased blood flow that may help to maintain energy levels. One possibility is that they lower ATP levels because blood flow is reduced and that this is damaging. There is evidence, from an abstract published several years ago (1101), that the depolarizations are responsible for the reduction in ATP in the penumbra. However, ATP only falls to ~70%, and it is not clear at all that this would be damaging during the 5-min depolarizations. It would be very useful to measure the effects of the artifactually increased frequencies of intraischemic depolarizations on ATP levels to see if there were a correlation between the latter and the increased damage.

One of the most intriguing lines of study to develop over the past several years has been three lines of evidence that white blood cells contribute to focal ischemic damage after temporary insults (41, 204, 248, 431, 592). These are 1) the appearance of neutrophils in ischemic tissue at an appropriate time; 2) the protection afforded against damage by agents that should prevent the accumulation or activation of neutrophils and 3) known damaging actions of these cells when they accumulate in tissue (1200). Effects of various manipulations suggest that the neutrophils are important in temporary, but not permanent, focal ischemia (203, 775, 1285). The evidence is generally strong and is bolstered by the probable involvement of neutrophils in ischemic injury in other tissues (399, 933). However, there are important conflicting reports and alternate interpretations of data that need to be considered. Furthermore, it remains to be determined whether damage results from effects on blood flow or from direct toxic effects on the vasculature or the neurons.

There is an accumulation of neutrophils in infarcted tissue 24 h after permanent or temporary ischemia (63, 120, 542, 1244). At this time, neutrophils are present in the vasculature and have, in most cases, infiltrated the parenchyma (63, 353). A very notable exception to the latter is human stroke where neutrophils accumulate early in severe cases (9) but, in the one relevant study, were almost exclusively confined to the vessels even more than a day after the stroke (665).

If it helps cause neuronal damage, neutrophil accumulation should be present well before 24 h after ischemia, and this does seem to be the case. Polymorphonuclear leukocytes accumulate in blood vessels within 4 h of the onset of permanent (353, 592), or temporary (563) ischemia (the latter measurements made in baboons). Perhaps the most striking evidence correlating neutrophil accumulation with damage emerges from studies using PET scans on humans. Accumulation begins within 6 h of the stroke, and there is a striking positive correlation between the amount and duration of leukocyte accumulation and both the size of the infarct and the severity of neurological outcome (9).

An important issue is whether the neutrophils act after infiltration through the blood vessels. Significant infiltration into the parenchyma only begins 12 h after the onset of permanent occlusion, by which time there is major cell necrosis (353, 1283), but it is more rapid following temporary ischemia, the insult for which there is evidence that neutrophils are damaging. There was significant accumulation 6 h after 1- to 2-h occlusion in two different studies (202, 542, 1283), which increased by 50% at 12 h (1283). This time course is consistent with parenchymal neutrophils causing cell damage. However, in one of the studies, the location of the early accumulation was not described (202) and in the other the accumulation did not reach significance in cortex at 6 or 12 h (1283); cortex is the region in which damage is prevented by antileukocyte treatment. Thus there is no well-documented evidence that accumulation in the parenchymal tissue is likely to be causing damage.

In contrast to the large number of results showing timely accumulation of neutrophils in at least the vasculature, Bartus and co-workers (435) found that there was no significant neutrophil accumulation until 21 h after 2-h distal MCA occlusion occlusion despite the occurrence of infarct by 8 h. There is other evidence that neutrophil accumulation does not always accompany cell damage. Two very careful sets of studies on spontaneously hypertensive rats, using essentially identical methodologies, showed either a large neutrophil accumulation 6-12 h after 80 min of MCA occlusion in the penumbral tissue (63, 202) or no neutrophil accumulation except in severely damaged tissue after 60 min of MCA occlusion (642). In the latter, there was a major activation of microglia/macrophages ~6 h after the occlusion. Thus as yet unidentified factors may determine the nature of the inflammatory response to an insult.

It has been suggested that artifactual mechanical effects of the insult may actually cause the neutrophil accumulation (for example, filament occlusion or laser irradiation) (435), and this has not been ruled out by careful studies. However, neutrophils do accumulate in the vasculature after human stroke (9, 665), where there is clearly no mechanical artifact. At present, there is no simple explanation for the conflicting data on neutrophil accumulation in rodents. Certainly they suggest that these white blood cells may not always be important in damage.

Neutrophil accumulation in vessels requires interactions between several adhesion molecules. These include intracellular adhesion molecule (ICAM)-1 and E- and P-selectins on the endothelial cells, fibronectin and laminin in the extracellular matrix, and integrins and L-selectins on the white blood cells. There are excellent short reviews of the process (204, 964, 1216) and, indeed, antibodies and peptides that bind to these molecules profoundly inhibit the accumulation of neutrophils after transient or permanent ischemia (197, 1243, 1245, 1285, 1287).

Ischemia activates this adhesion system. On endothelial cells, ICAM is increased 2 h after 2-h MCA occlusion (542, 1286) and also after human stroke (665). E-selectin is upregulated some time between 4 and 24 h (427, 1287). On the neutrophils, the integrin complex CD11/CD18 may also be upregulated, although this is less certain (63).

There is reasonable but very incomplete evidence that one or both cytokines, interleukin (IL)-1 and tumor necrosis factor- (TNF-), which do upregulate ICAM (965), are mediators of this response. It seems likely that there is a cascade involving free radical generation, increased TNF- and IL-1, and the activation of the nuclear transcription factor NFB. (1186). The NFB upregulates more proximally than the cytokines and is probably a critical intermediary step. Other factors are also probably necessary (56). There may be important positive feedback between the cytokines and NFB and free radicals as they all are able to activate each other (1186), and it is not clear which of these is the initiating step.

Tumor necrosis factor- is increased within 30 min of the onset of focal ischemia in the circulation and in brain tissue (154, 636), as is IL-1 (1295). The only study directly assessing the importance of the cytokines showed that treatment with the IL-1 receptor antagonist reduced accumulation of neutrophils during permanent MCA occlusion (352). Less directly, but still significant, blockade of TNF- action with a recombinant receptor greatly enhanced the number of perfused microvessels at the end of 4-h permanent MCA occlusion (412).

The postulated interaction is supported by studies in cultured endothelial cells (1035). Hypoxia induced the formation of IL-1, the subsequent expression of ICAM and E-selectin, and increased adherence of leukocytes. Antagonizing IL-1 action attenuated leukocyte adherence by ~60% (unfortunately adhesion molecule changes were not explicitly measured). A similar effect was observed in microvascular cells from human brain (1068). Further studies on these cells show that activation of the transcription factor NFB is very probably critical to the expression of ICAM. Hypoxia and reoxygenation caused activation of NFB and, later, ICAM (464). Both these responses were blunted by the proteasome inhibitor n-tosyl-Phe-chloromethylketone (464), which blocks the activation of NFB. The ICAM upregulation was also blocked by the free radical scavenger pyrrolidine dithiocarbamate, and cerebral endothelial cells in culture produce abundant free radicals under these conditions (1226).

Although evidence for protective effects of blocking cytokine action is now abundant (see sect. VIIIE), this cannot be taken to show that leukocyte accumulation is damaging; blockade of cytokines reduces tissue damage even in permanent ischemia (352, 946), against which adhesion antagonists are ineffective (203, 775, 1285).

A very large number of studies show that conditions that reduce or completely block leukocyte accumulation greatly reduce infarct size. Infarct volume 24 h after 45- to 60-min ischemia was reduced by about two-thirds when circulating neutrophils in mice (215) or in rats (729) were completely depleted with an antineutrophil antibody. Inhibiting the activity of the endothelial adhesion molecule ICAM, either by creating null mice (215, 1062), or using monoclonal antibodies in rats (197, 730, 1285), strongly reduced neutrophil accumulation and reduced infarction size by 50% (rats) to 75% (mice). Slightly smaller reductions, ~33%, were produced in rats by antibodies to the neutrophil integrins CD10/CD18 (197, 730) and by peptides that interfere with fibronectin binding (1243, 1244) or laminin binding (1245). A peptide that binds to selectin (775) decreased infarct volume by >50% in spontaneously hypertensive rats. Many of these studies involved 45-min to 1-h exposures to ischemia, but in some, ischemia was 2 or 3 h so the protective effects are not restricted to short insults. These effects are quite dramatic; there is essentially no further development of infarct once reperfusion begins.

Importantly, leukocytes do not appear to be involved in damage during permanent focal ischemia, despite their accumulation in the tissue vasculature (120, 542). Several interventions against this insult, which are protective against temporary ischemic damage, were ineffective in preventing damage (203, 775, 1285).

Although these studies taken together are impressive indictments of neutrophils in ischemic damage, there is at least one exception, which has not been explained away. In the study where no neutrophil accumulation was measured, neutropenia had no protective effect (435). Because of this, it is important to consider alternate explanations for the effects of the antileukocyte accumulation studies.

Neutrophil depletion may be protective as a result of lowering blood levels of damaging cytokines, TNF- and IL-1, that can be produced by these cells (280). Measurements of the effects of neutropenia, or other protective treatments, on cytokine levels during ischemia would help interpret results. The adhesion molecules and basal laminar proteins may also allow critical damage to vasculature or neurons in ways other than via leukocyte accumulation. Indeed, there is a marked depletion of lamin and fibronectin that is thought to contribute to microvascular damage (416). Perhaps the laminin and fibronectin peptides that attenuate damage (1243, 1245) are acting by maintaining vascular integrity. Although these explanations are not particularly likely, it is important to ensure that they, or others, are not correct. This has not been done.

Leukocytes might induce damage by causing local vascular occlusion, or they might initiate toxic reactions, including free radical production by NADH oxidase, production of hypochlorous acid, or protease activation (504, 728, 1200). The latter two require prior free radical production.

A) CHEMICAL TOXICITY. Efforts to determine whether free radical production is important have not been conclusive. The one study in which effect of neutrophils on free radicals was actually measured, carried out by Kogure and co-workers (728), showed that neutropenia completely abolished the generation of free radicals in cortex after 60-min MCA occlusion. This very dramatic effect was seen by measuring formation of ascorbyl free radicals (728), a reliable semiquantitative method (143). Unfortunately, this is the only such report, and the only study in which this method has been used to measure free radical formation. Further experimental support for the conclusion is very necessary as it seems, a priori, unlikely that neurons and glia themselves do not produce free radicals during and after ischemia. If it is correct, it ascribes a very important role to neutrophils in damage, as free radical attenuation produces major reductions in infarct size. However, at this stage, it cannot be considered likely.

Neutrophils produce free radicals via activation of NADPH oxidase. Walder et al. (1185) studied transgenic mice lacking NADPH oxidase. Infarct size was reduced by 50%, suggesting that leukocyte NADPH oxidase mediated the damaging effect of the neutrophils. However, Walder et al. (1185) then found that selective elimination of either leukocyte or parenchymal NADPH oxidase did not reduce damage; both had to be eliminated. Thus elimination of neutrophil NADPH oxidase does not account for the protective effects of eliminating neutrophil accumulation; if it did, then eliminating the enzyme selectively in neutrophils would have prevented damage.

An hypothesis that is consistent with both sets of studies is that neutrophils are necessary for the generation of toxic levels of free radicals but themselves are not the major source of the free radicals. Thus large-scale free radical production would be dependent on leukocytes, as noted by Kogure and co-workers (728), but could be initiated by NADH oxidase in parenchymal tissue or in the leukocytes. The way in which leukocytes mig