Mechanisms of delayed cell death following hypoxic-ischemic injury in the immature rat: evidence for apoptosis during selective neuronal loss

MOLECULAR BRAIN RESEARCH ELSEVIER

Molecular Brain Research 29 (1995) 1-14

Research

report

Mechanisms of delayed cell death following hypoxic-ischemic injury in the immature rat: evidence for apoptosis during selective neuronal loss Erica J. Beilharz

a, Chris E. Williams a,* , Mike Dragunow Peter D. Gluckman a

b, Ernest

S. Sirimanne

a,

aResearch Centre for Developmental Medicine and Biology, Auckland, New Zealand b Department of Pharmacology, University of Auckland, Pricate Bag, Auckland, New Zealand Accepted 13 September 1994

Abstract The mechanisms leading to delayed cell death following hypoxic-ischemic injury in the developing brain are unclear. We examined the possible roles of apoptosis and microglial activation in the 21-day-old rat brain following either mild (15 min) or injury. The temporal and spatial patterns of DNA degradation were assessed using severe (60 min) unilateral hypoxic-ischemic gel-electrophoresis and in-situ DNA end-labelling. Microglial activation, mitochondrial failure and cell death were examined using lectin histochemistry, 2,3,5,triphenyl-H-tetrazolium chloride (TTC) staining and acid fuchsin staining, respectively. Selective neuronal death produced by the 15 min injury was associated with the development of apoptotic morphology, DNA laddering and acidophilia from 3 days post-hypoxia. The 60 min injury accelerated this process with some cells showing signs of DNA degradation at 10 h post-hypoxia. However, in the cortex, which developed infarction after the 60 min injury, a different pattern of cell loss occurred. The DNA and mitochondria remained intact, and cells basophilic, until after 10 h post-hypoxia, then widespread necrosis developed by 24 hr. In contrast to regions of selective neuronal loss, DNA degradation was initially random (at 24 hr), with 180bp DNA ladders not detected until 3 days post-hypoxia. There was no morphological evidence of apoptosis. Microglial activation coincided with the onset of DNA degradation in regions of selective neuronal loss but not infarction, suggesting a possible role in selective neuronal death. The results suggest that cortical infarction, which was delayed for at least 10 h, was necrotic, and occurred independently of microglial activation and apoptosis. In contrast, selective neuronal death was apoptotic. Keywords:

Necrosis;

Apoptosis;

Microglial

activation;

Infarction;

1. Introduction Asphyxial injury to the neonatal brain is believed to be a major cause of permanent neurological damage. The mechanisms leading to neuronal death following hypoxic-ischemic (HI) injuries remain unclear [201. Increasing evidence indicates that neuronal death can occur in two distinct phases. During hypoxia-ischemia,

* Corresponding author. Fax: (64) 9-3737497. E-mail: chrisw@dev - phys.auckland.ac.nz. 0169-328X/95/$09.50

0 1995 Elsevier Science B.V. All rights reserved

SSDI 0169-328X(94)00217-7

Selective

neuronal

death;

DNA ladder;

Hypoxia-ischemia

acute energy failure leads to loss of ion homeostasis, intracellular sodium and calcium accumulation and osmotic swelling which, if severe, can lead to cell lysis. In addition cytotoxic actions of glutamate and free radicals may exacerbate the injury. A secondary phase of neuronal death can occur some hours later [38]. While many of the processes responsible for primary cell death may contribute to secondary cell death, they do not clearly explain the delay in onset of cell death. In order to explain this cell loss additional processes have also been suggested, such as apoptosis or cytotoxic activity of microglia and macrophages [17,22,43]. The patterns of cell loss after HI are likely to depend

E.J. Beilharz et al. /Molecular

Brain Research 29 (1995) I-14

Fig. 1. Histological outcome after HI injury. Thionin-stained sections showing the extent of damage present 5 days after the 15 min (A) am i 60 min (B) HI injuries. The ligated hemisphere is on the right. In A, selective neuronal loss occurs predominantly in the CA1 region of the hippocampus and cortical layers 3-5 (C), on the ligated side. In B, widespread damage is seen in the cortex (Cl, hippocampus (H), thalamus (T) and striatum (9, on the ligated side. No damage was detected on the non-ligated hemisphere.

E.J. Beilharz et al. /Molecular Brain Research 29 (1995) l-14

on the severity of the injury: selective neuronal loss developing after brief injuries, and infarction (tissue necrosis), where there is also loss of glia, resulting from more severe injuries. Different mechanisms of damage are likely to be associated with these distinct patterns of cell loss. Generally cells are thought to die either by apoptosis or necrosis. Apoptosis can be distinguished from necrosis morphologically. Necrosis is associated with swelling of the cytoplasm and organelles, with little change initially to the nucleus, whereas apoptosis is characterized by condensation of the cytoplasm and nucleus, which eventually breaks into small fragments [27]. Another characteristic of apoptosis is the activation of specific endonucleases which cleave the chromatin at internucleosomal points, leading to a DNA ladder [531. In addition, de novo synthesis of so-called ‘killer proteins’ may or may not be required, as evidenced by the ability of protein and RNA synthesis inhibitors to prevent apoptosis in some but not all cases [12]. The loss of excess neurons that occurs during development can occur via apoptosis [39]. Increasing evidence now shows that apoptosis is not confined to physiological cell death, but also occurs in pathological situations. Evidence for the involvement of apoptosis in HI injury comes from studies showing that protein and RNA synthesis inhibitors can reduce delayed, selective neuronal loss produced by brief HI injuries in rodents [22,43]. This evidence however remains controversial [9]. The activation of brain resident microglia is known to occur in the brain after injury. While their phagocytic role has been long recognized, it is now becoming evident that they may have additional roles. For example, they can produce neurotoxins, such as nitric oxide, hydrogen peroxide and glutamate, and therefore may contribute to delayed neuronal death [32]. In this study, we examined some of the possible mechanisms of damage that may lead to selective neuronal loss or infarction. We used a transient, unilateral HI injury [44] based on a modified Levene preparation, in the 21-day-old rat brain. In particular we have examined the temporal and spatial relationships between activation of microglia and cell loss, as determined by DNA degradation, mitochondrial failure and neuronal acidophilia following either mild (15 min) or severe (60 min) HI injuries. Two methods have been used to examine DNA degradation patterns. Firstly, extracted DNA was analyzed after electrophoresis, allowing discrimination between randomly degraded DNA (smear) and an oligonucleosomal ladder, which are thought to represent necrotic and apoptotic cell death respectively. Secondly, an in situ method of 3’ end-labelling of cut DNA strands allowed the distribution and morphology of individual cells with evidence of endonuclease-cut DNA to be determined.

2. Materials

and methods

2.1. HI injury These experiments were approved by the Auckland Animal Ethics Committee. Briefly, 21-day-old Wistar rats weighing between 40 and 50 g had the right carotid artery ligated under 2% halothane anaesthesia. The rats were allowed to recover for 2 h and were then exposed to either 15 min (moderate injury) or 60 min (severe injury) inhalational hypoxia (8% oxygen) at 34°C and 70% humidity [29,44]. The rats were euthanised with pentobarbital and the brains were collected at the following timepoints: 5 h, 10 h, 1 day, 3 days and 5 days. One set of brains (n = 3/timepoint) was snap frozen on dry-ice and stored at -70°C until cut into 12 ym coronal sections at the level of the hippocampus. In situ DNA end-labelling, lectin histochemistry and acid-fuchsin staining were performed on parallel sections from each of these brains. Another set of brains used for DNA extraction was collected at the following timepoints: 10 h, 1 day, 3 days and 5 days after the 60 min injury, and 1 day, 3 days and 5 days after the 15 min injury (n = 3/timepoint). A third set of brains was collected at 10 h, 1 day, 3 days and 5 days after the 60 min injury for analysis of mitochondrial function (n = 3/timepoint). The brains from untreated 21-day-old Wistar rats were used as controls for each of the techniques used. 2.2. In situ DNA end-labelling In order to identify cells containing cut DNA (produced by either specific or random endonuclease activity), a method of end-labelling DNA in situ, using terminal deoxy-transferase (TdT), was used to examine coronal sections from brains collected at a number of time-points after either the moderate or severe injuries. The method for in situ labelling of DNA fragments was based on the TUNEL technique of Gavrieli et al. [15], and modified for use on frozen sections. The frozen sections were fixed in 4% paraformaldehyde, then washed twice in PBS. Endogenous peroxidase activity was blocked by 3% H,O,. After further PBS washes, the sections were preincubated in TdT buffer, followed by a 60 min incubation at 37°C in the following reaction mixture: 0.75 ~1 TdT (15 units/F0 (Gibco BRL, Gaithesburg, MD), 0.75 ~1 0.4 mM biotin-14-dATP (Gibco BRL), TdT buffer, in a total volume of 75 pi/section. These concentrations of enzyme and labelled nucleotide were found to give the most optimal ratio of positive-signal to background. After washing in PBS, the sections were incubated with extravidin-peroxidase (1: 200) (Sigma, St. Louis, MO) for 1 h at room temperature, and the reaction developed using diaminobenzidine (DAB) as the chromogen. The sections were then dehydrated, mounted in DPX and examined using light microscopy. As a negative control, sections were subjected to the above procedure but with either TdT or biotin-14-dATP omitted. As a positive control, sections were treated with 1 pg/ml DNAse 1 (Gibco BRL) for 15 min at 37°C prior to incubation with the TdT reaction mixture. As positive cells labelled by this method may contain either laddered or randomly digested DNA, it cannot be used to distinguish apoptotic from necrotic cells unless morphological criteria are also used. Cells with highly pyknotic and fragmented nuclei were identified as being apoptotic. 2.3. DNA extraction and analysis To complement the in situ results, gel electrophoresis [“‘P]dCTP-labelled DNA extracted from a separate set of brains used to analyze the pattern of DNA degradation. Tissue from hippocampus, the striatum and the cortex was dissected from ligated side of the brains subjected to the 60 min injury, at 10

of was the the h, 1

4

E.J. Beilharz et al. /Molecular

Brain Research 29 (1995) l-14

day, 3 days and 5 days post-injury (n = 3/timepoint). Hippocampal tissue was dissected from the ligated side of brains subjected to the 15 min injury at 1 day, 3 days and 5 days post-injury (n = 3/timepoint). The DNA was extracted using standard procedures. The tissue was ground into small pieces and incubated overnight in 0.5 mg/ml proteinase K and 1% SDS in 50 mM Tris, 100 mM NaCl and 100 mM EDTA. Phenol/chloroform extractions and ethanol precipitation were followed by RNAse treatment and proteinase K treatment, further phenol/chloroform extractions and ethanol precipitation. Five hundred nanograms of DNA was end-labelled with [“*P]dCTP (Amersham, Buckinghamshire, UK) using TdT. The samples were separated by electrophoresis on a 1.5% agarose gel (40,000 cpm/lane). After alkaline Southern transfer to nylon membrane (Genescreen plus, DuPont, NEN Products, Boston, MA) the filter was exposed to Kodak film for lo-24 h. Laser scanning densitometry (Personal Densitometer, Molecular Dynamics, Sky Valley, CA) was used to scan the autoradiographs and generate density profiles of DNA degradation. The approximate lengths of the DNA fragments were compared against a 123 bp ladder (Boehringer Mannheim, Mannheim, Germany). 2.4. Leclin histochemistry Microglia were detected using the Griffonia lectin-horseradish peroxidase conjugate (Sigma). bated overnight with B,-isolectin at 4°C. After DAB was used as the peroxidase substrate. examined using light microscopy.

simplicifolia B,-isoSections were incuwashing with PBS, The sections were

2.5. Acid-fuchsin / thionin staining Sections from the fresh frozen brains were stained with acidfuchsin and thionin using standard methods. Pink staining of acidophilic cells is an unequivocal marker of cell death [4]. 2.6. Mitochondrial failure Brains subjected to the 60 min injury were stained with 2,3,5, triphenyl, H-tetrazolium chloride (TTC) in order to detect the timing of mitochondrial failure and development of infarction. TTC stains mitochondrial succinate dehydrogenase enzymes, meaning that viable tissue is stained bright red, while irreversibly damaged tissue remains white [6]. Brains were collected at 5 h, 10 h, 1 day and 3 days after the 60 min injury (n = 3/timepoint) and immediately chilled in PBS for 20 min. Two millimetre slices were then cut at the level of the hippocampus and incubated in 2% ‘FTC at 37°C for 30 min. The TTC was then removed and the brain slices fixed in 10% formalin for photography.

3. Results 3.1. Histology The patterns of cell loss following the 15 and 60 min injuries used in this study have been described in detail [29,44]. The pattern of cell loss in the present study was confirmed by acid-fuchsin/ thionin staining and found to be consistent with these previous results. In brief, the 15 min injury led to selective neuronal loss, predominantly of pyramidal cells of the CA1/2 region and in cortical layers 3-5 on the ligated hemisphere only (Fig. 1A). In 4/9 brains striatal neurons were also

Fig. 2. Photomicrographs showing controls for the in situ DNA end-labelling technique. A: no labelling was seen in this normally positive damaged cortex when TdT was omitted from the in situ DNA end-labelling reaction. B: as a positive control this section, from a non-damaged brain, was preincubated with DNAse 1, leading to widespread labelling of nuclei.

affected. In contrast, the 60 min injury led to infarction in the frontoparietal cortex, and severe neuronal loss in the hippocampus, the striatum and the thalamus in all brains, on the ligated side only (Fig. 1B). No cell loss was detected on the non-ligated hemispheres after either the 15 or 60 min injuries. 3.2. DNA degradation Control studies for the in situ DNA end-labelling technique showed that DNAse pretreatment of sections led to positive staining of all cells, while omission of the enzyme TdT or biotin-14-dATP from the reaction mixture abolished any positive signal (Fig. 2). No positive cells were detected in undamaged (control) brains, or on the non-ligated side of the injured brains. The distribution of positive cells detected after the 15 or 60 min injuries (described below) was consistent between brains collected at the same timepoints, except where indicated. 3.2.1. 1.5 min hypoxia In situ DNA end-labelling showed that very few positive cells (i.e. those containing either randomly or specifically degraded DNA) were present at 5 h posthypoxia. At 1 day post-hypoxia, only a few cells in small

E.J. Beilharz et al. /Molecular

areas of the striatum were positive in 2/3 brains, whereas by 3 days, pyramidal neurons of the CAl-2 regions of the hippocampus and cortical neurons of layers 3-5 were positive, in addition to striatal neurons (Fig. 3). The distribution of positive cells was the same at 5 days post-hypoxia, although the intensity of the signal had increased. The distribution of positive cells corresponded to that of acidophilic neurons on sections stained with acid-fuchsin. The positive signal was restricted to nuclei, which in general appeared intact. However, morphological signs of late stages of apoptosis (i.e. pyknotic, fragmented nuclei) were found in scattered CA1 pyramidal neurons in brains collected at 5 days post-hypoxia (Fig. 3B) [27]. Electrophoresis of labelled DNA extracted from the hippocampus on the ligated side of the brain showed results consistent with the in situ DNA end-labelling

Brain Research 29 (1995) l-14

5

(Fig. 4A). At 1 day post-hypoxia, there was no sign of DNA degradation. At 3 days and 5 days however, there was clear evidence of a DNA ladder, with DNA fragments of multiples of 180 bp in length. Only a small proportion of the DNA extracted from the hippocampus was degraded, as shown by the large band of intact, high molecular weight DNA. This was presumably derived from the many non-damaged cells in other areas of the hippocampus. 3.2.2. 60 min hypoxia In situ DNA end-labelling showed two general patterns of DNA degradation (Fig. 5). At early timepoints, i.e. 5 h and 10 h post-hypoxia, a few isolated, strongly positive cells were identified in specific areas such as the dentate gyrus granule layer and striatum, and there were scattered cells in the white matter and hippocam-

A

Fig. 3. In situ DNA end-labelling of sections from brains subjected to the 15 min HI injury. A: CA1 region of a brain collected 5 days post-hypoxia. Note that only neurons of the pyramidal layer (p) contain positive nuclei. B: high magnification of the CA1 pyramidal neuron marked in A. Note the characteristic apoptotic appearance of the nucleus. C: positive cortical neurons. Bars = 100 pm (Al, 10 pm (B), 40 pm (Cl.

360bp 360bp-

180bp-

180bpArbitrary Units

Arbitrary Units

D

NObp18Obp -

Arbitrary

Units

Fig. 4. Electrophoretic analysis of DNA degradation patterns after 15 and 60 min HI injury. [“P]dCTP-labelled DNA extracted from the hippocampus after the 15 min injury (A), and from the hippocampus (B), striatum (0 and cortex (D) after the 60 min injury, was separated on 1.5% agarose gels and transferred to nylon membrane. Shown are representative autoradiographs and the corresponding density profiles produced by densitometric scanning of the autoradiographs. Note the presence of 180 bp unit ladders. In A, DNA was extracted at 1 day (lane l), 3 days (lane 2), and 5 days (lane 3) post-hypoxia. In B, C and D, DNA was extracted at 10 h (lane l), 1 day (lane 21, 3 days (lane 3) and 5 days (lane 4) post-hypoxia. * Intact, high molecular weight DNA.

Fig. 5. In situ DNA end-labelling of sections from brains subjected to the 60 min HI injury. A, B, and D show the ventral arm of the dentate gyrus granule layer (delineated by the open arrows) from brains collected at (A) 10 h, (B and C) 1 day, (D) 3 days post-hypoxia. Note that nuclei of cells in the dorsal layer become positive first, with even distribution throughout the granule layer not seen until 3 days. C shows the same brain as in B, at lower magnification, emphasizing the relatively strong labelling of the inner layers at 1 day post-insult. E, F, and G show positive nuclei of granule cells at 5 days post-hypoxia at high magnification. Note the apoptotic appearance of a number of cells (arrows). H: compacted chromatin in a striatal cell at 1 day post-hypoxia. I: the positively labelled nucleus of a cortical cell at 3 days post-hypoxia, which does not show apoptotic morphology. J: widespread labelling of positive, non-apoptotic cells in the cortex at 1 day post-hypoxia. Bars = 50 pm (A,B,D), 100 pm (C,J), 40 pm (EL 10 pm (F,G,H,I).

E.J. Beilharz et al. /Molecular

Brain Research 29 (1995) l-14

’ . ,

1 ,

??

8

Fig. 6. Isolectin B,-positive microglia in the 15 min injury. B: high magnification brains collected at 1 day (0 and 3 days the infarcted cortex 3 days after the 60

E.J. Beilharz et al. /Molecular

Brain Research 29 (1995) 1-14

the injured brains. A: strongly labelled microgha in the pyramidal layer (p) of the CA1 region 5 days after of microglia in the damaged cortex 5 days after the 15 min insult. C,D: microglia in the dentate gyrus in (D) after the 60 min insult. Note the very weak staining at 1 day, compared with 3 days. E: microglia in min injury. Bars = 100 Frn (A), 40 pm (B,E), 50 pm (CD).

pal fimbria on the ligated hemisphere. There were no positive cells in the cortex at these timepoints. However, at 1 day, there was additional widespread, weaker labelling of nuclei throughout all areas affected by the injury, including the cortex, striatum, thalamus and the CA region of the hippocampus. The intensity was increased at 3 days but by 5 days post-hypoxia both the

number of stained cells and intensity of staining was dramatically reduced, although some dentate gyrus granule cells remained strongly labelled. As with the 15 min injury, labelling was restricted to nuclei. Neurons of the dentate granule layer showed an interesting pattern of DNA fragmentation (Fig. 5A-D). At early timepoints (up to 1 day post-hypoxia), most of

Fig. 7. Mitochondrial failure in damaged brains, detected by lack of TTC staining. TTC staining of 2-mm-thick fresh sections of (A) control brain, and (B) a brain collected at 1 day after the 60 min injury. The ligated hemisphere is on the right. The large area of non-staining tissue in B indicates regions of mitochondrial failure. C, cortex; H, hippocampus; T, thalamus.

A

c

10

E.J. Beilharz et al. /Molecular Brain Research 29 (19951 l-14

the positive cells were located in the inner-most layers of the granule layer, i.e. the ventral layers of the dorsal arm and the dorsal layers of the ventral arm. However by 3 days, positive cells were scattered throughout the entire structure. The signal from these cells was consistently stronger than in all other areas (except for some cells in the striatum). From 3 days onwards, groups of positive nuclei in the granule layer were pyknotic and fragmented (Fig. 5E-G), showing an apoptotic morphology similar to that of cells in the CA1 region after the 15 min injury (see above). While cells in all other areas showed only weak staining at 5 days, the fragmented nuclei found in the dentate granule layer at this timepoint were still strongly stained. A similar apoptotic morphology was also detected in a few striatal cells at 1 day post-hypoxia (Fig. 5H), while no morphological signs of apoptosis were detected in the cortex (Fig. 51,J). Electrophoresis of DNA extracted from the hippocampus, striatum and cortex following the severe injury agreed with the in situ DNA end-labelling results (Fig. 4B-D). In both the striatum and hippocampus there was a small amount of fragmented DNA, in the form of a DNA ladder (particularly evident in the striatum), in addition to a large amount of intact DNA, at 10 h post-hypoxia. At 1 day post-hypoxia both regions showed extensive DNA degradation, which appeared to be in the form of a smear (randomly degraded) in addition to a ladder. At 3 days post-hypoxia there was no intact DNA left, as seen by the lack of high molecular weight DNA, and by 5 days, there was very little DNA greater than 800 bp in length remaining, indicating that the degradation process had continued. In contrast, DNA extracted from the damaged cortex remained intact until 1 day post-hypoxia, at which time random degradation was detected. Laddered DNA was not clearly evident until 3 days posthypoxia. 3.3. Microglia 3.3.1. 15 min hypoxia Sections adjacent to those stained for in situ DNA end-labelling were stained with isolectin B,, to identify activated microglia. Activated microglia, identified by increased isolectin B, staining (compared with basal levels), showed characteristic short processes and large bodies [34]. At 1 day post-hypoxia, weak microglial staining could be detected in the striatum and hippocampus of the ligated hemisphere, while at 3 days and 5 days post-hypoxia, strongly stained microglia were present in all areas corresponding to neuronal damage i.e. CA1/2 pyramidal layer, cortical layers 3-5 and striatum (Fig. 6A,B). A faint signal was also present in the stratum oriens and stratum radiatum of the CA1/2 region.

3.3.2. 60 min hypoxia At early time-points (up to 1 day post-hypoxia), activated microglia could be detected in some areas of the ligated hemisphere (although the staining was relatively faint compared to later timepoints), including the striatum and hippocampus (Fig. 60, but there was no staining in the cortex. Activated microglia were not detected in the cortex until 3 days post-hypoxia. At 3 days post-hypoxia, very strongly stained activated microglia were distributed throughout all areas containing damaged cells (Fig. 6D,E). 3.4. Mitochondrial failure Undamaged (control) brains and brains collected 10 h after the 60 min HI injury stained strongly with TTC, indicating viable succinate dehydrogenase activity (Fig. 7A). In contrast, at 24 h after the 60 min injury tissue in the regions of cell loss remained unstained (Fig. 7B), suggesting that mitochondrial failure occurs between 10 h and 24 h post-hypoxia.

4. Discussion Using various techniques we have examined events occurring during the secondary phase of cell death following HI injuries to the 21 day old rat brain. In addition to determining the temporal and spatial relationships between DNA degradation, microglial activation and mitochondrial failure, we have provided evidence for the involvement of apoptosis in selective neuronal loss in the developing brain. Following the 15 min injury, DNA degradation could not be detected until 3 days post-hypoxia, with the exception of some striatal neurons, which were positive at 1 day. This delayed neuronal death is consistent with the timing of neuronal death reported after brief HI injuries in both the adult rat and gerbil [28,38]. Following the 60 min injury, striatal neurons also degenerated rapidly, in agreement with other adult studies [8,46]. Together with granule cells of the dentate gyrus, some striatal neurons showed DNA degradation at 5 h posthypoxia. In contrast, DNA degradation was not seen in the cortex until 24 h post-hypoxia. This corresponded to the timing of mitochondrial failure, as determined by TIC staining. Both the in situ end-labelling and gel electrophoresis results showed that cell death was not complete by 1 day post-hypoxia, with additional cells becoming positive for DNA degradation between 1 day and 3 days. Therefore, it appears that even after the 60 min injury, the majority of cells destined to die may have been viable until at least 10 h post-hypoxia. Our results have provided evidence in two forms for selective neuronal death due to apoptosis. Firstly, using the in situ DNA end-labelling technique we were able

E.J. Beilharz et al. /Molecular

to identify positive cells with an apoptotic morphology. Secondly, DNA ladders, made up of multiples of 180 bp units, were detected after electrophoresis of DNA extracted from damaged regions of the brains. Such laddering is often associated with apoptosis, and has been shown to be due to the internucleosomal digestion of the chromatin by specific calcium-dependent endonucleases [53]. This contrasts with the smear produced by predominantly random degradation in necrotic cells. Similar techniques have been used to show DNA laddering occurring in CA1 and striatal neurons within 24 h of global ischemia in the adult rat [25]. Caution must be used when interpreting DNA laddering results, as it has been shown that necrotic cells also produce DNA ladders in some circumstances [7,14]. However, the initial presence of DNA laddering together with the morphological changes of dying cells suggests strongly that much of the selective neuronal loss that occurred after an HI injury was apoptotic. The appearance of laddered DNA occurred relatively early in some regions of the brain after the 60 min injury, with striatal and dentate granule cells containing laddered DNA as early as 10 h post-hypoxia. These cells also showed evidence of an apoptotic morphology. Therefore, it appears that the apoptotic process was accelerated by the more severe injury. What induces apoptosis in this situation is not known, although loss of trophic support is one possibility. The dependence of dentate gyrus granule cells on hormonal support has previously been shown by the specific induction of apoptosis in these cells following adrenalectomy [4.5]. Interestingly, we have previously shown that IGF-1 treatment after HI injury in the adult rat reduces neuronal loss, with the areas showing the greatest neuroprotection being the dentate gyrus and the striatum [23]. While the mode of action of IGF-1 has not been established, it is known to prevent apoptosis in various cell types [35,41] and may act in the same way in the injured brain. Another possible contributor to the death of these neurons, i.e. activated microglia, will be discussed later. A number of previous reports have addressed the question of whether HI injury involves programmed cell death. Some studies have shown that selective neuronal loss, produced by brief injuries (similar to the 15 min injury used in this study) can be reduced by administration of protein and RNA synthesis inhibitors [22,43], although these results are controversial 191. While these results could reflect the prevention of synthesis of ‘killer’ proteins responsible for programmed cell death, other explanations are possible. For instance, the production of neurotoxins by nonneuronal cells could equally be affected. In addition, the potentially neuroprotective hypothermic effects of cycloheximide cannot be discounted. The endonuclease inhibitor aurintricarboxylic acid (ATA) has been shown

Brain Research 29 (1995) 1-14

11

to reduce both ischemic and NMDA induced toxicity in vivo and in vitro [40,42,54], although again the mechanism of action is unclear, as ATA has also been shown to be an NMDA receptor antagonist [54]. Our results support the hypothesis that selective neuronal death is caused by apoptosis. Whether this is the classical form of protein and RNA synthesis-dependent programmed cell death remains to be determined, because some types of apoptosis may be produced by constitutively expressed proteins whose activation may occur posttranslationally (e.g. phosphorylation). However, in a previous study we have shown, using the same 1.5 min HI injury, that the mRNA and protein for the immediate-early gene c-jun is induced by 24 h post-hypoxia [lo] in the same regions (e.g. CAl/CA2). This result shows that the apoptosis occurring in the 15 min preparation may be transcriptionally-regulated, and therefore programmed, by prior expression of c-j,,, which may function as a suicide transcription factor in neurons 1111. Pathophysiological studies suggest that excitatory amino acids and calcium accumulation may contribute to selective neuronal loss. For instance, Andine et al. [2,3] have shown that several hours after an ischemic neurons is injury, Ca* + uptake into CA1 pyramidal increased and extracellular levels of excitatory amino acids rise. Excitatory amino acids such as glutamate are known to be neurotoxic, and glutamate can increase calcium influxes and induce apoptosis in neurons under certain conditions [30]. Increased intracellular Ca*+ is a feature of apoptosis (as well as necrosis) and may lead to the activation of endonucleases responsible for the DNA laddering seen in this study. Furthermore treatment with the NMDA receptor antagonist MK801 6 h after an ischemic injury can reduce selective neuronal loss [47]. There is also evidence for a gradual reduction in mitochondrial efficiency in CA1 pyramidal neurons following brief HI injuries [ll, with mRNA and protein levels for the mitochondrial enzyme cytochrome oxidase c decreasing significantly within hours of the injury. This may leave the neurons increasingly susceptible to glutamate toxicity. Analysis of the severely injured brains indicated that the onset of cortical infarction occurred later than 10 h post-hypoxia. No DNA degradation was seen in the cortex until 1 day post-hypoxia, at which time there was widespread mitochondrial disruption, necrosis and acidophilia. Random DNA degradation, seen at I day, preceded specific laddering, which was not detected until 3 days post-hypoxia, and there were no morphological signs of apoptosis seen at any stage, with the positive nuclei appearing intact. While the presence of some DNA laddering was clearly evident, it is unlikely that many of the cells died via apoptosis. For instance, it has been demonstrated that as little as 2% of apoptotic cells amongst necrotic cells can be detected as a

12

E.J. Beilharz et al. /Molecular Brain Research 29 (199.5) l-14

ladder [7]. Another possibility is that high intracellular calcium levels lead to the activation of endogenous endonucleases which cut at internucleosomal sites, and therefore mimic the ladder associated with apoptosis, as suggested by Tominaga et al. [50], who found DNA laddering in the striatum and cortex 1X h after global ischemia in the adult rat. Therefore, it is possible that the different patterns of cell death described above reflect both apoptosis and necrosis occurring in parallel in different regions. We have previously demonstrated that severe HI injury c-Fos / Fos-related protein synthesis in corsuppresses tical neurons [24] supporting the hypothesis that DNA fragmentation in the infarcted cortex is a consequence of necrosis rather than programmed apoptosis. However, the possibility also exists that the severe model produces unprogrammed apoptosis [ 121. Based on isolectin B, histochemistry, we observed that microglial activation following the 15 min injury coincided with DNA degradation, i.e. both were first detected in the striatum at 1 day, and in the CA1 and cortex at 3 days post-hypoxia. Microglial activation also coincided with cell loss in those areas of the severely damaged brains showing apoptotic morphology i.e. the striatum and dentate gyrus, although the microglial staining at early timepoints was relatively weak. In contrast, the widespread cell loss in the cortex at 1 day post-hypoxia occurred before any signs of microglial activation. This suggests that the role of microglia may differ depending on the type of cell death. It is possible that activated microglia participate in cell death in the selective neuronal loss model, as has previously been suggested [32]. A number of studies have shown that activated microglia appear before neuronal death [16,18,33,34]. Microglia can release a number of cytotoxins including glutamate, hydrogen peroxide, nitric oxide, tumour necrosis factor LY(TNFcu) and a novel class of compounds as yet not fully characterized [ 17,19,37,51]. Neurotoxic effects of microglia in culture have been shown to be mediated at least in part by nitric oxide and hydrogen peroxide [13,49]. Inhibition of microglial activity by chloroquine and colchicine leads to a reduction of ischemic neuronal loss in the rabbit spinal cord [18]. Also, administration of macrophage inhibitory factor (MIF) reduces axotomy-induced ganglion cell degradation [48]. Our results are consistent with a possible role for activated microglia in areas where apoptotic cells are seen. However, a more detailed study, using further timepoints and additional markers for microglial activation would be needed to determine the exact temporal relationship between microglial activation and neuronal death. It is known that external factors can trigger apoptosis. Activated microglia may be such a trigger. Macrophages have previously been shown to be necessary for tissue remodelling, which involves programmed cell death, in

the eye in vivo [31]. It has been shown that cells undergoing apoptosis, but not necrosis, express the Le(y) antigen [261. It is possible that sick neurons produce antigens that target themselves for killing by microglia. Alternatively, dying neurons may release substances that initiate the cytotoxic activities of microglia. In contrast to a possible role for microglia in selective neuronal death, the onset of cortical infarction following the 60 min injury is clearly independent of microglial activation, which was a late event. Instead the microglia probably play a role in the clearance of debris and regulation of gliosis and wound repair [17]. Our results indicated that widespread cortical cell necrosis occurred suddenly, after a delay of at least 10 h, and the late activation of endonuleases suggests that DNA degradation is a result rather than an initiator of cell death. It also suggests that intervention may be possible for several hours after the injury. A number of factors may potentially contribute to the development of the infarct. In vivo, studies of HI injury in the late gestation fetal sheep brain have shown that the development of cortical infarction is associated with a progressive loss of membrane function and the development of epileptiform activity from around 8 h post-hypoxia [52]. The 60 min infant rat shows a similar timecourse of seizures (E.S.Sirimanne, unpublished observations) and is at an equivalent stage of neurological maturation to the late gestation fetal sheep. Suppression of the epileptiform activity with the NMDA receptor antagonist MK801 postpones the loss of membrane function in the fetal sheep [47]. This suggests that NMDA receptor-mediated activity accelerates the development of cortical infarction. It is likely that energy depletion contributes to the susceptibility of the neurons to NMDA-mediated damage [36]. TTC staining of succinate dehydrogenase suggested that mitochondria were functional until about 24 post-hypoxia. However, it is possible that decreasing levels of other mitochondrial enzymes, as shown during selective neuronal loss [l], also occurs, reducing the viability of neurons. The development of an infarct also involves the death of glial cells, which are much less susceptible to oxygen deprivation than neurons [21]. Damage to glial cells might be indirect and perhaps due to the loss of trophic support. For instance, oligodendrocytes have been shown to be dependent on axon-derived growth factors for survival, and can be rescued by IGF-1 or CNTF after axotomy [5]. This hypothesis is supported by results showing that exogenous IGF-1, when administered 2 h post-hypoxia, reduces the rate of infarction in a dose dependent manner following HI injury in the adult rat [23]. In summary, the pathogenesis of selective neuronal loss and infarction following HI injury in the 21 day old rat brain are distinct. In areas of selective neuronal

E.J. Beilharz et al. /Molecular

loss, apoptosis appeared to be a major form of cell death. Activated microglia may have participated in this process. However, in the severely damaged cortex, the lack of early DNA laddering and morphological changes indicated that delayed onset necrosis, rather than apoptosis, was the primary mechanism of delayed cell death. These results may have important therapeutic implications with respect to the mechanisms of injury following hypoxic-ischemic damage to the developing brain, for instance during the perinatal period.

Acknowledgements The work was supported by the Health Council of New Zealand and the Neurological tion of New Zealand.

Research Founda-

References [l] Abe, K., Kawagoe, J., Lee, T.H., Aoki, M. and Kogure, K., Disturbance of a mitochondrial DNA expression in gerbil hippocampus after transient forebrain ischemia, Mol. Brain Res., 19 (1993) 69-7.5. I. and Hagberg, H., Enhanced calcium [2] Andine, P., Jacobson, uptake by CA1 pyramidal cell dendrites in the postischemic phase despite subnormal evoked field potentials: excitatory amino acid receptor dependency and relationship to neuronal damage, J. Cereb. Blood Flow Metab., 12 (1992) 773-783. [3] Andine, P., Orwar, O., Jacobson, I., Sandberg, M. and Hagberg, H., Changes in extracellular amino acids and spontaneous neuronal activity during ischemia and extended reflow in the CA1 of the rat hippocampus, J. Neurochem., 57 (1991) 222-229. [4] Auer, R.N., Kalimo, H., Olsson, Y. and Siesjo, B.K., The temporal evolution of hypoglycemic brain damage. II. Light- and electron-microscopic findings in the hippocarnpal gyrus and subiculum of the rat, Acta Neuropathol., 67 (1985) 25-36. M.D., Schmid, R., Sendtner, M. and ]51 Barres, B.A., Jacobson, Raff, M.C., Does oligodendrocyte survival depend on axons?, Curr. Biol., 3 (1993) 489-495. J.B., Pitts, L.H., Germano, S.M., Nishimura, M.C., ]61 Bederson, Davis, R.L. and Bartkowski, H.M., Evaluation of 2,3,5triphenyltetrazolium chloride as a stain for detection and quantification of experimental cerebral infarction in rats, Stroke, 17 (1986) 1304-1308. ]71 Collins, R.J., Harmon, B.V., Gobe, G.C. and Kerr, J.F., Internucleosomal DNA cleavage should not be the sole criterion for identifying apoptosis, Int. J. Radiat. Biol., 61 (1992) 451-453. W., Harrison, A. and Nadler, J., Selec[81 Crain, B.J., Westerkam, tive neuronal death after transient forebrain ischemia in the mongolian Gerbil: a silver impregnation study, Neuroscience, 27 (1988) 387-402. J., Bergstedt, K., Linden, T., Kalimo, H. and Wie]91 Deshpande, loch, T., Ultrastructural changes in the hippocampal CAl-region following transient cerebral ischemia - evidence against programmed cell death, Exp. Brain Res., 88 (1992) 91-105. M., Beilharz, E.J., Sirimanne, E., Lawlor, P., ]I01 Dragunow, Williams, C., Bravo, R. and Gluckman, P.D., Immediate-early gene protein expression in neurons undergoing delayed death but not necrosis following hypoxic-ischemic injury to the young rat brain, Mol. Brain Res., 25 (1994) 19-33. M., Young, D., Hughes, P., Macgibbon, G., Lawlor, [III Dragunow,

Brain Research 29 (1995) I-14

13

P., Singleton, K., Sirimanne, E., Beilharz, E.J. and Gluckman, P.D., Is c-Jun involved in nerve cell death following status epilepticus and hypoxic-ischaemic brain injury, Mol. Brain Res., 18 (1993) 347-352. a product of programmed and unpro]121 Eastman, A., Apoptosis: grammed cell death, Toxicol. Appl. Pharrnacol., 121 (1993) 160164. ]131 Fabian, R.H. and Rea, H.C., Neuronal toxicity by macrophages in mixed brain cell culture is augmented by antineuronal IgG and dependent upon nitric oxide synthesis, J. Neuroimmumol., 44 (1993) 95-102. of exten]141 Fukuda, K., Kojiro, M. and Chiu, J.F., Demonstration sive chromatin cleavage in transplanted Morris hepatoma 7777 tissue: apoptosis or necrosis?, Am. J. Pathol., 142 (1993) 935-946. of ]151 Gavrieli, Y., Sherman, Y. and Bensasson, S.A., Identification programmed cell death in situ via specific labeling of nuclear DNA fragmentation, J. Cell Biol., 119 (1992) 4933501. J., Bonnekoh, P., Miyazawa, T., Hossmann, K.A. [I61Gehrmann, and Kreutzberg, G.W., Immunocytochemical study of an early microglial activation in ischemia, J. Cereb. Blood Flow Metab., 12 (1992) 257-269. ]171 Giulian, D., Chen, J., Ingeman, J.E., George, J.K. and Noponen, M., The role of mononuclear phagocytes in wound healing after traumatic injury to adult mammalian brain, .I. Neurosci, 9 (1989) 4416-4429. C., Inhibition of mononuclear 1181Giulian, D. and Robertson, phagocytes reduces ischemic injury in the spinal cord, Ann Neural., 27 (1990) 33-42. ]I91 Giulian, D., Vaca, K. and Corpuz, M., Brain glia release factors with opposing actions upon neuronal survival, J. Neurosci., 13 (1993) 29-37. P.D. and Williams, C.E., When and why do brain ]201 Gluckman, cells die?, Dev. Med. Child Neural., 34 (1992) 1010-1014. 1211 Goldberg, M.P. and Choi, D.W., Combined oxygen and glucose deprivation in cortical cell culture - calcium-dependent and calcium-independent mechanisms of ncuronal injury, J. Neurosci., 13 (1993) 351-3524. [22] Goto, K., Ishige, A., Sekiguchi, K., Iizuka, S., Sugimoto, A., Yuzurihara, M., Aburada, M., Hosoya, E. and Kogure, K., Effects of cycloheximide on delayed neuronal death in rat hippocampus, Brain Res., 534 (1990) 299-302. [23] Guan, J., Williams, C., Gunning, M., Mallard, E.C. and Gluckman, P.D., The effects of IGF-1 treatment after hypoxic-ischemic brain injury in adult rats, J. Cereb. Blood Flow Metab.. 13 (1993) 609-616. [24] Gunn, A.J., Dragunow, M., Faull, R.L. and Gluckman, P.D., Effects of hypoxia-ischemia and seizures on neuronal and gliallike c-fos protein levels in the infant rat, Brain Res., 531 (1990) 105-116. F., [25] Heron, A., Pollard, H., Dessi, F., Moreau, J., Lasbennes, Ben-Ari, Y. and Charriaut-Marlangue, C., Regional variability in DNA fragmentation after global ischemia evidenced by combined histological and gel electrophoresis observations in the rat brain, J. Neurochem., 61 (1993) 1973-1976. [26] Hiraishi, K., Suzuki, K., Hakomori, S. and Adachi, M., Lety) antigen expression is correlated with apoptosis (programmed cell death), Glycobiology., 3 (1993) 381-390. [27] Kerr, J.F.R., Searle, J., Karmin, B.V. and Bishop, C.J., Apoptosis. In C.S. PoKen (Ed.), Perspecticw on Mammlian Cell Death, Oxford University Press, New York, 1987, pp. 93-128. (281 Kirino, T., Delayed neuronal death in the gerbil hippocampus following ischemia, Brain Res., 239 (1982) 57-69. [29] Klempt. N.D., Sirimanne, E., Gunn, A.J., Klempt, M., Singh, K., Williams, C.E. and Gluchnan, P.D., Hypoxia-ischemia induces transforming growth factor p, mRNA in the infant rat brain, Mol. Brain Res., 13 (1992) 93-101.

14

E.J. Beilharz et al. /Molecular

[301 Kure, S., Tominaga,

T., Yoshimoto, T., Tada, K. and Narisawa, K., Glutamate triggers internucleosomal DNA cleavage in neuronal cells, Biochem. Biophys. Res. Commun., 179 (1991) 39-45. are required for cell [311 Lang, R.A. and Bishop, J.M., Macrophages death and tissue remodeling in the developing mouse eye, Cell,

74 (1993) 453-462. [321 Lees, G.J., The possible

contribution of microglia and macrophages to delayed neuronal dealh after ischemia, J. Neurol. Sci., 114 (1993) 119-122. [33] Morioka, T., Kalehua, A.N. and Streit, W.J., The microglial reaction in the dorsal hippocampus following transient forebrain ischemia, J. Cereb. Blood Flow Metab., 1 I (1991) 966-973. of [341 Morioka, T., Kalehua, A.N. and Streit, W.J., Characterization microglial reaction after middle cerebral artery occlusion in rat brain, J. Camp. Neural., 327 (1993) 123-132. of human erythroid I351 Muta, K. and Krantz. S.B., Apoptosis colony-forming cells is decreased by stem cell factor and insulin-like growth factor-l as well as erythropoietin, J. Cell Physiol., 156 (1993) 254-271. R.C., [3hl Novelli, A., Reilly, J.A., Lysko, P.G. and Henneberry, Glutamate becomes neurotoxic via the N-methyl-o-aspartate receptor when intracellular energy levels are reduced, Brain Rex, 451 (1988) 205-212. [371 Piani, D., Frei, K., Do, K.Q., Cuenod, M. and Fontana, A.. Murine brain macrophages induce NMDA receptor mediated neurotoxicity in vitro by secreting glutamate, Neuroscl. Lett., 133 (1991) 159-162. [381 Pulsinel li, W., Brierley, J.B. and Plum, F.. Temporal profile of neuronal damage in a model of transient forebrain ischemia, Alan. Neural., 11 (1982) 491-498. [391 Raff. M.C., Social controls on cell survival and cell death, Nature. 356 (1992)397F400. J.M., Marcy, V.R., Zhao, Y.H., Vaught, J.L., [401 Robertslewis, Siman, R. and Lewis, M.E., Aurintricarboxylic acid protects hippocampal neurons from NMDA-induced and ischemia-induced toxcicity invivo. J. Neurochem.. 61 (1993) 378-381. G., Collins, M.K.L., Garcia, I. and [411 Rodriguez-Tarduchy. LopezRivas, A., Insulin like-growth factor-l inhibits apoptosis in IL-3-dependent hemopoietic cells, J. Immunol.. 149 (1992) 535540. acid pro[421 Samples. S.D. and Dubinsky. J.M., Aurintricarboxylic tects hippocampal neurons from glutamate excitotoxicity invitro, J. Neurochem., 61 (1993) 382-385.

Brain Re.Teurch 29 (1995) l-14

[431 Shigeno,

T., Yamasaki, Y., Kato, CT., Kusaka, K., Mima, T., Takakura, K., Graham, D.I. and Furukawa, S., Reduction of delayed neuronal death by inhibition of protein synthesis, Neurosci. Lett., 120 (1990) 117-119. E.S., Gunn, A.J., Guan, J., Williams, C.E. and [441 Sirimanne, Gluckman, P.D., Two models for determining the mechanisms of damage and repair after hypoxic-ischemic injury in the developing rat brain, J. Neurosci. Methods, 55 (1994) 7-14. [451 Sloviter, R.S., Dean, E. and Neubort, S., Electron microscopic analysis of adrenalectomy-induced hippocampal granule cell degeneration in the rat: apoptosis in the adult central nervous system, J. Camp Neural., 330 (I 993) 337-351. [461 Smith, M.L., Kalimo, H., Warner, D.S. and Siesjo, B.K., Morphological lesions in the brain preceding the development of postischemic seizures, Acta Neuroputhol., 76 (1988) 253-264. [471 Tan, W.K.. Williams, C.E., Gunn, A.J., Mallard, C.E. and Gluckman, P.D., Suppression of postischemic epileptiform activity with MK-801 improves neural outcome in fetal sheep, Ann Neural., 32 (1992) 677-682. [4X1 Thanos, S., Mey, J. and Wild, M., Treatment of the adult retina with microglia-suppressing factors retards axotomy-induced neuronal degradation and enhances axonal regeneration in viva and in vitro, Neuroscience, 13 (1993) 455-466. [491 Thery, C., Chamak, B. and Mallat, M., Cytotoxic effect of brain macrophages on developing neurons, Eur. J. Neurosci., 3 (1991) 11.5-1164. [501 Tominaga, T., Kure, S., Narisawa, K. and Yoshimoto, T., Endonuclease activation following focal ischemic injury in the rat brain, Brain Rex, 608 (1993) 21-26. [51] Urban, J.L., Shepard, H.M., Rothstein, J.L., Sugarman, B.J. and Schreiber, H., Tumor necrosis factor: a potent effector molecule for tumor cell killing by activated macrophages, Proc. Natl. Acad. Sci. USA, X3 (1986) 5233-5237. [52] Williams, C.E., Gunn, A.J. and Gluckman, P.D., Time course of intracellular edema and epileptiform activity following prenatal cerebral ischemia in sheep, Stroke, 22 (1991) 516-521. [53] Wyllie. A.H., Glucocorticoid-induced thymocyte apoptosis is associated with endogenous endonuclease activation, Nature, 284 (1980) 555-556. [S4] Zeevalk, G.D., Schoepp, D. and Nicldas, W.J., Aurintricarboxylic acid prevents NMDA-mediated excitotoxicity - evidence for its action as an NMDA receptor antagonist, J. Neurochrm., 61 (1993) 386-389.