Main

In animals subjected to experimental cerebral hypoxiaischemia, large numbers of apoptotic cells can be found in the brain 12-48 h after the end of the hypoxic-ischemic insult(1–7). These results have important implications for the development of cerebroprotective therapies, especially as strategies are becoming available that prevent cells committing to apoptotic death(8, 9). However, it remains to be confirmed that a similar activation of apoptosis occurs in human infants suffering intrauterine cerebral injury. These infants are at risk of acute hypoxia-ischemia, but may also suffer less well defined insults with other or additional causes such as infection or inflammation.

The present study therefore addressed the hypothesis that in human infants severe in utero insults induce a significant proportion of brain cells to undergo apoptosis. The infants studied had suffered an intrauterine insult of such severity that it led either to apparently sudden intrauterine death, or to hypoxic-ischemic encephalopathy and death after delivery. Tissue obtained at postmortem examination was examined histologically to determine the mode of cell death in the cingulate gyrus by quantifying the fractions of apoptotic and necrotic cells.

METHODS

Patients. Two groups of infants were studied by morphologic analysis of brain tissue and review of clinical and pathologic data.

Group 1. Birth asphyxia: Six term infants who had been successfully resuscitated after severe intrapartum disruption to gas exchange, but subsequently died after developing hypoxicischemic encephalopathy and delayed cerebral injury.

Group 2. Unexpected sudden intrauterine death: Six infants at or close to full term, who died suddenly and unexpectedly before delivery.

Morphologic studies. Serial coronal brain sections taken at the level of the head of the caudate nucleus were examined in all infants. For quantitative histologic analysis, 5-μm sections of the cingulate gyrus were stained with Cole's hematoxylin and eosin. Counting was done using a ×40 objective and an eyepiece graticule with 100 grid squares. Fifteen fields were evaluated in the superficial cortex (layers II and III) and deeper layers (IV, V, and VI) from the cingulate sulcus and the crest of the cingulate gyrus. A minimum of 4000 nuclei were counted, and the number of apoptotic and necrotic cells were expressed as the percentage of the total nuclei counted. Morphologic analysis was performed by one researcher unaware of the clinical history of the infants under investigation. The results were then checked independently by a second observer who was also unaware of clinical details.

Apoptotic cells were identified using conventional morphologic criteria: 1) intense, uniform nuclear basophilia; 2) chromatin condensation with nuclear shrinkage (pyknosis) or fragmentation of the nucleus into several rounded and uniformly dense basophilic masses (karyorrhexis); or 3) the formation of apoptotic bodies (which were defined as plasma membrane enclosing cytoplasm that was not eosinophilic; which frequently but not always contained dense basophilic chromatin; and which were regarded as representing a single cell if several appeared together in a group). Cells were counted as necrotic if they demonstrated intensely eosinophilic cytoplasm, breakdown of the nuclear and plasma membranes, and cell swelling(10).

Cells with uniform condensation of chromatin at all or part of the nuclear periphery were counted as apoptotic, whereas cells with small rounded nuclei and irregular chromatin clumping at the nuclear margin were excluded as these may be mistaken for oligodendroglia, especially if there is technical artifact. Cells showing nuclear condensation with eosinophilic cytoplasm were not included in cell counts.

The presence of DNA fragmentation typical of apoptosis within cells was sought using ISEL in five subjects, two in group 1 and three in group 2. This was carried out according to the protocol of Ansari et al.(11) with the following modifications. Tissue digestion was with 10 μg·mL-1 proteinase K (Sigma Chemical Co., Poole, UK); the end-labeling mix (500 μL/section) contained 10 mM each of dGTP, dTTP, and dATP and 4 mM biotin-14-dATP (GIBCO BRL, Paisley, Scotland) in 50 mM Tris-HCl buffer (pH 7.5) supplemented with 5 mM MgCl2, 10 mM 2-mercaptoethanol, 0.005% BSA (fraction V, Sigma Chemical Co.), and the Klenow fragment of DNA polymerase I (Pharmacia Biotech, Milton Keynes, UK) at 5 U·mL-1. To block endogenous peroxidase, sections were immersed in 0.3% H2O2/PBS for 15 min at room temperature. Incorporated biotinylated nucleotides were detected with horseradish peroxidase using a commercial kit (Vectastain, Vector Laboratories, Peterborough, UK), and sections were then counterstained with hematoxylin. For negative controls, the Klenow fragment was omitted from the end-labeling mixture.

Statistical analysis. Data groups were tested for normality using the Kolmogorov-Smirnov test and compared using ANOVA or Kruskall-Wallis ANOVA as appropriate. Relations between continuous variables were computed by least-squares linear regression after equal variance was confirmed. The results of statistical tests were only recorded as significant if p < 0.05.

RESULTS

Clinical details. Group 1. Birth asphyxia: Four boys and two girls, of birth weight 2.84-4.37 kg (median 3.21 kg), were born at gestational age of 39-42 wk (median 41.5) after pregnancies that were uneventful until the onset of labor. Five infants showed cardiotochographic abnormalities. Two underwent spontaneous vaginal deliveries, whereas two were delivered by forceps and two by emergency cesarian section. Apgar scores at 5 min ranged from 1 to 5 (median 3), and the base deficit of the first arterial blood sample taken after birth from 20.0 to 25.8 mmol (22.1 mmol). All infants required mechanical ventilation, and all developed hypoxic-ischemic encephalopathy grade 3 according to the scoring system of Sarnat and Sarnat(12). EEG recordings were abnormal in all infants, and all developed seizures. Five of the six were studied by 31P magnetic resonance spectroscopy at 12-96 (median 90) h: the ratio of PCr to Pi concentration ([PCr]/[Pi]) was below normal range values in all infants, ranging from 0 to 0.78 (median 0.40). The infants died 64-159 (median 150) h after birth.

The general pathologic examination was reported to show evidence of hypoxia-ischemia in multiple organs in five infants. Placental pathology was available in two cases, and showed perivillous fibrin deposition in one and chorioamnionitis, intervillitis, and funisitis in the other; this infant was found to be infected by group B streptococcus. A third placenta was found to have a complete abruption at cesarian section but was not examined microscopically. The thymus was reported as showing a stress reaction in five infants, and thymus weight ranged from 4.0 to 11.2 g (median 8.6 g). General neuropathologic examination showed extensive hypoxic-ischemic injury in multiple brain areas in all six infants.

Group 2. Unexpected stillbirth: Four boys and two girls, of birth weight 2.45-3.89 kg (3.00 kg), were stillborn at 36-40-wk (median 39) gestation. Three pregnancies were completely uneventful. Two mothers (one suffering diabetes and the other pregnancy-induced hypertension) had reported minor reduction in fetal movements at some stage of pregnancy although no obstetric intervention was initiated. The remaining infant, whose mother's blood contained anti-LA antibodies, was mildly growth-retarded (birth weight 2.45 kg) with slight reduction in amniotic fluid volume. Fetal growth and amniotic fluid volume was normal in other infants. In three cases, absent fetal heart was noted on admission to the labor ward. Labor was spontaneous in four; one labor was induced because of maternal hypertension and the other because of intrauterine death. Two infants were delivered by forceps and none by cesarian section.

A general pathologic examination was reported to reveal evidence of generalized hypoxic-ischemic injury in only one infant. Placental histopathology showed focal calcification in one infant but was normal in the remainder. A stress reaction was seen in the thymus in one infant; thymus weight ranged from 7.5 to 13.8 g (median 11.9 g) and was significantly greater than in group 1 (p < 0.05). General examination of the brain was reported to show a subdural hemorrhage in one infant, considered to be caused postmortem by instrumental delivery, and some gliosis of periventricular white matter without widespread damage in one other. No infant had evidence of congenital infection, either on microscopic analysis or from culture or serologic tests.

Cell morphology in the cingulate gyrus. Cells undergoing apoptosis and necrosis are shown in Figure 1. Apoptotic and necrotic cells were seen in both groups of infants. The number of cells showing nuclear condensation with eosinophilic cytoplasm was always less than 10% of the total number and usually much less. Nonquantitative neuropathologic analysis by an experienced neuropathologist (M.V.S.) showed that apoptosis was more common in cortical layers I, II, and III, whereas necrosis tended to predominate in the deeper layers. However, apoptotic and necrotic cells were distributed through all areas, and often sited contiguous to each other. Preliminary analysis using morphologic criteria alone suggested that apoptosis seemed to occur in neurons, glia, and endothelial cells, but phagocytes containing apoptotic bodies appeared to be very infrequent.

Figure 1
figure 1

Photomicrograph of the cingulate gyrus from an infant dying after birth asphyxia, showing characteristic apoptotic (arrows) and necrotic (arrowheads) cells. Stained with hematoxylin and eosin; magnification, ×600.

Full size image

An example of ISEL staining is given in Figure 2. ISEL detected fragmented DNA in most pyknotic and some karyorrhectic cells. However, a few necrotic cells were also stained by this technique. In control experiments, omitting the Klenow fragment of DNA polymerase resulted in a complete absence of ISEL (data not shown).

Figure 2
figure 2

Photomicrograph of the cingulate from an infant suffering apparently sudden unexpected intrauterine death. Tissue is stained by in situ labeling of fragmented DNA, and positive staining is seen in an apoptotic neuron in the center of the field. Magnification, ×240.

Full size image

Quantitation of the fractions of apoptotic and necrotic cells after hematoxylin and eosin staining showed that a mean 8.3% (95% CI for the population, 3.7-12%) of all cells counted in asphyxiated infants and 6.7% (CI 0.2-13.6%) in stillbirths were apoptotic. This was not significantly different. Of all cells counted, 20.8% (CI 11-30.6%) were necrotic in the asphyxiated group, but significantly fewer, 3% (CI 0.4-5.6%), in the stillborn infants. All data groups were normally distributed.

The observed variation in cell counts was related to differences in total cell death, as is displayed in Figure 3. The fraction of apoptotic cells increased significantly in proportion to total cell death in both subject groups. From the gradient of the relation: (fraction of all cells that were apoptotic/sum of the fraction of all cells that were apoptotic and the fraction that was necrotic) it was calculated that 78% (CI 41-100%) of dead cells were apoptotic in stillbirths. Data from asphyxiated infants was most precisely fitted by a second order regression model, but a simple first order model showed reasonable fit and demonstrated that 26% (CI 1-51%) of dead cells were apoptotic.

Figure 3
figure 3

Figure showing the percentage of all cells that were apoptotic (Apoptosis%) related to the percentage of cells showing either apoptotic or necrotic death (Total cell death%). Infants suffering birth asphyxia and delayed cerebral injury are shown in (a) and those with unexpected and apparently sudden intrauterine death are represented in (b).

Full size image

DISCUSSION

Apoptosis. Apoptosis is the physiologic mechanism that allows a multicellular organism to remove redundant or hazardous cells(13). It plays an important role in brain development, as up to 50% of cells in the developing brain die by apoptosis(14), and there is evidence to suggest that immature cells are more prone to apoptosis(2, 15). Apoptotic cells are rapidly phagocytosed by neighboring cells before lysis; this prevents spillage of the cytoplasmic contents into the surrounding tissue, thus avoiding an inflammatory reaction, so that, despite a high rate of apoptotic death, fewer than 1% of cells are visibly apoptotic in brain tissue at any time during development(13, 16, 17).

All cells seem to retain the ability to undergo apoptosis, and activation of the apoptotic cascade can be a normal response to a variety of forms of cellular injury, including DNA damage, viral infection, inflammation, and hypoxia-ischemia(18). The apoptotic program seems to be executed by “killer” proteins generally present constitutively in cytoplasm, but normally repressed by mechanisms such as survival signals from other cells(13). Once activated, the apoptotic cascade requires time and energy to complete(19, 20), although the time taken for individual cells to die is variable(4). After hypoxia-ischemia, apoptotic cells are usually first detected soon after the insult and become most prominent after 24 h(5, 6). Thereafter increased apoptosis can be detected for days or weeks(4, 21).

Apoptosis was defined using light microscopy with conventional morphologic criteria; in studies of newborn piglets we have confirmed that cells with these appearances are apoptotic by electron microscopy(1). As in our previous studies, ISEL detected DNA fragmentation characteristic of apoptosis in almost all pyknotic cells. However, consistent with these earlier results, ISEL was sometimes negative in karyorrhectic cells, perhaps because karyorrhexis represents the final stages of nuclear dissolution, and extreme disturbances to the quaternary structure of DNA could prevent DNA polymerase enzymes from binding, resulting in reduced labeling. ISEL also detected a small number of necrotic cells in the human tissue, which is consistent with earlier observations which showed that DNA degradation during necrosis can also be detected by ISEL(1). Consequently ISEL cannot be used for quantitative studies. However, the ISEL results were particularly useful in demonstrating that pyknotic cells were undergoing DNA fragmentation, and that the nuclear appearances was not due to technical or staining artefact.

Neuropathology. We examined the cingulate gyrus because we have previously found that a significant amount of apoptosis occurs in this region in the newborn piglet(1) and because it is a region of brain interrogated by our measurements of [PCr]/[Pi]. The results confirm that apoptosis is seen in the cingulate gyrus of infants suffering intrauterine insults. In animals apoptosis is found in most brain regions after hypoxicischemic injury(2); our preliminary results suggest that this is also true in newborn infants (our unpublished data), but further investigation is required to confirm this.

Although apoptosis occurred throughout the cingulate gyrus, it seemed on nonquantitative analysis to be more prominent in the superficial layers of the cortex, which is consistent with data suggesting that less mature cells are more prone to undergo apoptosis(2, 15). It seemed to occur in a variety of cell types, including neurons, glia, and endothelium, although morphologic classification of injured cells must be treated with great caution, and confident statements on the types of cells undergoing apoptosis must await further studies.

Pathologic investigation of human infants presented difficulties not found in experimental studies. First, there could be no human control subjects. However, a large number of studies in a wide variety of mammalian species have consistently shown that the fraction of apoptotic cells detectable in the developing brain is less than 1%(13, 14, 16). The amount of apoptosis found in the study was much higher than this, and comparable to that seen in newborn piglets subjected to hypoxia-ischemia(1). This, together with the increasing fraction of apoptosis that occurred with increasing severity of damage (as determined by the total number of dead cells), makes it most unlikely that the observed numbers of apoptotic cells represent basal values and argues strongly that apoptosis is occurring in response to injury. Second, variability of cell counts may have been increased by differences in the time interval between death and autopsy. However, as apoptosis is an energy-requiring process the postmortem interval is very unlikely to have increased the amount of apoptosis seen in either group of infants(19, 20); studies in animals have confirmed that the postmortem interval does not significantly influence the results of studies into DNA fragmentation(22).

Comparison of subject groups. The two subject groups represented different sequelae to intrauterine insults. Infants in group 1 suffered severe intrapartum asphyxia followed by resuscitation. They subsequently developed hypoxic-ischemic encephalopathy, and pathologic examination showed widespread hypoxic-ischemic damage with a marked stress reaction in the thymus. Delayed cerebral injury was demonstrated objectively by delayed impairment of cerebral energy metabolism in the five infants studied by magnetic resonance spectroscopy(23, 24). This secondary phase of injury is a period of massive cerebral destruction, and the magnitude of the delayed disruption of cerebral energy metabolism predicts the severity of later neurodevelopmental impairment(25). Many apoptotic and necrotic cells were found in the cingulate gyri of these infants.

Infants in group 2 did not have definitive evidence of hypoxia-ischemia, and the etiology of injury was unsure. They seem to have undergone a different process from the asphyxiated infants, apparently sudden unexpected intrauterine death without the development of hypoxic-ischemic encephalopathy. The thymus showed a stress reaction in only one subject, and the thymus weights were significantly higher than in group 1. Widespread hypoxic-ischemic damage was seen in only one infant. Although the nature of the intrauterine episodes is unknown, it is likely that hypoxia-ischemia and/or infection contributed; certainly at a cellular level the cerebral insult must have included complete loss of substrate delivery at some stage. As a fraction of the total number of counted cells, the amount of apoptosis in stillborn infants was similar to that of the asphyxiated group, although fewer necrotic cells meant that apoptosis comprised a higher proportion of total cell death. However, the small numbers of infants studied, the similarity of the absolute quantity of apoptotic cells, and other issues discussed below make it premature to conclude that apoptosis is more important in the brain pathology associated with apparently sudden intrauterine death than in hypoxic-ischemic encephalopathy. Nevertheless, as the apoptotic program requires time to complete, these data imply that in stillborn infants the cerebral insult began some time before the apparently sudden death.

Apoptosis or necrosis in response to hypoxia-ischemia. Apoptosis was observed in both subject groups, and within the groups higher fractions of apoptotic cells were detected in brains with evidence of more severe injury. These results confirm the importance of apoptosis as a response to cerebral insults in the perinatal period. They suggest that intrauterine injury can trigger apoptosis with or without the generalized tissue destruction seen after hypoxia-ischemia with delayed cerebral injury.

Apoptosis comprised a relatively smaller proportion of dead cells in the infants who developed delayed cerebral injury, because necrosis was more common. There are several mutually compatible explanations for this.

First, delayed energy failure may cause necrosis rather than apoptosis, perhaps through increased free radical or extracellular glutamate production(26, 27). However, this is unlikely to be the sole explanation, as apoptosis not only occurs during delayed injury in many experimental models(1, 4, 5), but can also be specifically reduced by manipulations during the delayed phase(8).

Second, some cells which appeared necrotic may have been triggered to apoptosis but not undergone phagocytosis, because the normal mechanisms whereby apoptotic cells are cleared could be overwhelmed. As a final stage in their destruction membrane integrity would be lost and the cells undergone secondary necrosis(28). The small number of ISEL positive necrotic cells and cells of intermediate morphology may represent a stage in this process. The proportion of apoptotic cells may thus have been underestimated in the more severely damaged brains.

Third, subtle cellular insults might cause apoptosis, whereas necrosis results from more severe injury. Low concentrations of neurotoxins can trigger apoptosis in neural cells in culture, whereas high levels induce rapid necrosis; cells that lose mitochondrial function in these conditions undergo necrotic death, whereas those that retain the ability to produce ATP proceed to apoptosis(29, 30). In evolutionary terms, responding to an insult by apoptosis may have survival value for a multicellular organism, as unlike necrosis, it should not result in the release of damaging cellular contents into the extracellular matrix(13).

Neuroprotection by inhibition of apoptosis. Studies in rodents have shown that overexpression of the antiapoptotic gene bcl-2(31), attenuation of proapoptotic genes such as p53(32), or the administration of survival factors such as IGF-I(9) ameliorates neural damage after hypoxia-ischemia. Clinically applicable antiapoptotic strategies that are beneficial in animal models, such as moderate postinsult brain cooling(8), may be worth evaluating for their therapeutic potential.

Conclusion. These results demonstrate that a large number of brain cells undergo apoptosis in infants who die after intrauterine insults. Apoptosis follows both intrauterine hypoxia-ischemia and also less well defined injuries that lead to unexpected intrauterine death. Whether a particular cell undergoes necrosis or apoptosis may depend upon many factors, the severity of the cellular insult, the events associated with delayed energy failure, cellular energy reserves, the ability of the mitochondria to recover function upon resuscitation, the availability of trophic support from neighboring cells, the effectiveness of phagocytotic mechanisms, and on the cell having sufficient time and energy for the apoptotic program to complete. Nevertheless, apoptosis appears to be a consistent feature of cerebral damage in the perinatal period, and activation of the apoptotic cascade may be a final common pathway to cell death after a variety of intrauterine insults.