Main

SIDS is defined as the sudden death of an infant under 1 y of age that remains unexplained after performance of a complete autopsy, examination of the death scene, and review of the clinical history (1). It is hypothesized that infants who die of SIDS have a brain abnormality (possibly at the brainstem level) or a maturational delay related to neuroregulation of cardiorespiratory control leading to repeated episodes of hypoxemia (2). SIDS victims present histologic and biochemical findings consistent with prior repeated episodes of hypoxemia in several organs (3–8). However, examination of the brains of SIDS victims by light microscopy has revealed only subtle abnormalities (9) that, in themselves, are insufficient to explain the death.

Programmed cell death (apoptosis) can be triggered acutely in the brain by a hypoxic insult (10–12) and, typically, leaves no scarring. A cell that dies by apoptosis retains an intact membrane and is engulfed and digested by neighboring cells without any inflammatory reaction (unlike the process of necrosis) (13). Thus, an attractive hypothesis for the cause of death in SIDS is that repeated episodes of hypoxemia first lead neuronal apoptosis in select vulnerable brain regions, then to the disappearance of a significant number of cells, and eventually to impaired function. Apoptosis involves a distinctive multistage process of DNA fragmentation by endonucleolytic degradation (14,15). A relatively new technique to identify such fragmentation allows quantification of this process in various tissues, including the brain (16,17).

The aim of the present study was to determine whether apoptosis is found in the brainstem and hippocampus of SIDS victims. We chose to include the hippocampus because we had found evidence of recent hypoxemic-ischemic insult predominant in this brain region in a group of SIDS victims (18). The hippocampus is composed of subregions that are either sensitive or resistant to hypoxic-ischemic insult (19–21), thus permitting us to determine whether enhanced apoptosis (if present) is predominantly located in hypoxia-sensitive subregions or is more generalized. Moreover, in the neonatal period, the brainstem and hippocampus display a distinct pattern of reaction to hypoxic-ischemic insults (22–24); also, the hippocampus is the region in which the histologic findings in SIDS victims (8) resemble those found in animals exposed to hypoxic-ischemic insults (25–27).

METHODS

We studied 29 SIDS (19 males) and nine control cases (six males). Four additional cases of SIDS who survived temporarily after resuscitation were also included. The control group consisted of infants who had died in circumstances similar to those of the SIDS victims: death occurred suddenly in asymptomatic infants with no prior evidence of either neurologic disorder or any disease characterized by chronic hypoxia. The diagnoses were as follows: bronchopneumonia, three cases; pyelonephritis with early septicemia, two cases; septicemia, one case; endocardial fibroelastosis, one case. Although the remaining two control cases had no definitive diagnosis after a full autopsy, the infants died suddenly in the awake state during a crying spell (possible arrhythmia). Inasmuch as true control cases are rare, these were identified first and age-matched SIDS cases were then found (age range 35 to 127 d). All infants had had a thorough autopsy at a pediatric center, and an expert panel reviewed all cases to ensure conformity to the accepted definition of SIDS (11).

We used two techniques to identify apoptotic cells. First, we used the in situ end-labeling of DNA, a method now well established for paraffin-embedded tissue that enables the quantification of the apoptotic process in subregions of the brain (16,17). Second, we confirmed the nuclear changes typical of apoptosis by morphologic criteria. Those were the following: a well-defined nucleus; condensed chromatin within the nucleus or at the periphery of the nuclear membrane; and, at later stages, condensation of the cytoplasm and the appearance of apoptotic bodies (28). To this end, we examined slides stained for apoptosis and slides from adjacent sections stained with HPS.

We used a commercial kit (Apoptag, ONCOR, Gaithersburg, MD) for the in situ labeling. We performed the test on 5-µm paraffin sections. The apoptotic cells were identified by direct immunoperoxidase detection of digoxigenin-labeled genomic DNA in situ. The Apoptag kit targets (for labeling) the 3′-OH ends generated by DNA fragmentation. The paraffin sections were mounted on poly-L-lysine-coated slides. They were then deparaffinized in xylene, dehydrated in ethanol, and treated with proteinase K (10 µg/mL in PBS) for 10 min to digest protein before being treated with 50% methanol for inactivation of endogenous peroxidase (30 min in 2% hydrogen peroxide). All sections then underwent treatment to add digoxigenin-dUTP to the 3′-OH ends of the DNA fragments by terminal deoxynucleotidyl transferase (Tdt). Next, an antidigoxigenin antibody carrying a conjugated peroxidase was added. The peroxidase enzyme catalytically generates an intense signal from chromogenic substrates in nuclei that have high concentrations of cleaved DNA 3′-OH ends. Apoptotic nuclei will then appear dark brown, whereas normal nuclei will have blue background coloration (Mayer's hematoxylin). All sections were processed in triplicate, with each run including a positive (adrenal tissue) and a negative control (Tdt omitted).

We also examined the HPS slides for features of neuronal necrosis, features comprising an increased cytoplasmic density (eosinophilic cytoplasm), indistinct nuclear-cytoplasmic boundaries, hyperchromatic pyknotic nuclei, and retracted cells (29–31). The presence of such damaged neurons attests to the occurrence of a recent neuronal insult; their regional distribution within the hippocampus helps to determine the time of such an insult (32–34).

One section containing the hippocampus (Fig. 1) and usually two to three sections of the right brainstem were available to us in paraffin blocks. The brainstem blocks had been taken from two places: one usually at the obex and the other at the open medulla near the level of the cranial nuclei X and XII. The nuclei studied at each level are shown in a diagrammatic representation of the brainstem in Figure 2.

Figure 1
figure 1

Overview of the hippocampus showing the location of the analyzed regions. The surface of the counted area represents about one half of the CA4 and subiculum (SUB) region. Except for the boundary regions (the transition is not always clearly evident), the entire area of the CA3, CA2, and CA1 subregions were included in the counts. The average number of neurons counted per region were as follows: CA4 = 245, CA3 = 191, CA2 = 96, CA1 = 79, and SUB = 277.

Full size image
Figure 2
figure 2

Computer-generated diagram of the brainstem sections. Left panel, level of the obex; right panel, level of the open medulla. Abbreviations and average number of neurons counted per nuclei: Acc Oliv, accessory olive; Cun, cuneate (163); Grac, gracilis (93); Hypo, hypoglossal (29); Inf Oliv, inferior olive (63, obex; 104, open medulla); Inf Vest, inferior vestibular (98); Lat Cun, lateral cuneate (89); Lat Ret, lateral reticular formation (115, obex; 50, open medulla); NA, nucleus ambigus; NTS, nucleus of the tractus solitarius (107, obex; 89, open medulla); Sp Trig, spinal trigeminal (158, obex; 80, open medulla); Sp trig T, spinal trigeminal tract; Vag, vagus (36, obex; 42, open medulla).

Full size image

For neuronal counts, each slide stained for DNA end-labeling was reviewed by two investigators, both ignorant of the diagnosis (SIDS or control brains). We studied all hippocampal subregions and the whole hemibrainstem on each slide. We examined the cells under a ×40 objective to ensure the presence of adequate staining, to permit differentiation between neuronal and nonneuronal cells, and to ascertain the morphologic characteristics of apoptosis and/or necrosis.

For the hippocampus, all neurons were counted in an area of each subregion at magnification ×10. We counted normal neurons (negative staining), positively stained neurons having the features of apoptosis, and positively stained neurons that could not be identified as apoptotic (damaged neurons, possibly by necrosis). The number of apoptotic neurons was expressed as a percentage of the total number of neurons. The cases were considered negative (physiologic apoptosis) if only rare neurons were positive (<1% positive cells). Normal and necrotic neurons on the HPS sections were counted and the number of necrotic neurons expressed as a percentage of the total number of neurons.

For the brainstem, we counted apoptotic neurons in the same way as for the hippocampus. For small, well-defined nuclei (nucleus tractus solitarius, nucleus of the spinal trigeminal tract, for instance), the whole nucleus was counted. For large nuclei (gracile and cuneate, for example), a region was chosen within each nucleus and at least 100 cells were counted.

We used the Mann-Whitney test for comparison between SIDS and control groups and for comparison between subgroups of SIDS victims. The Fisher exact test was used to compare proportions between different groups. p < 0.05 was considered significant.

The present study received approval from our Institutional Review Board. Permission to obtain information from all cases of sudden death was obtained from the Office of the Chief Coroner of the province and the Bureau d'Accès à l'information du Québec.

RESULTS

Neuronal apoptosis was found in 23 of the 29 SIDS victims (79%). Comparing SIDS cases that exhibited apoptosis with those that did not, we discerned no effects of age, sex, postmortem interval, or formalin-fixation duration. All cases were consistently identified as positive or negative by both observers. For positive cases, we elected to group the data in the following three categories: a) mild apoptosis, 1 to 9%; b) moderate apoptosis, 10 to 19%; and c) marked apoptosis, >20% of positive neurons. This was done because, when the data were examined, there was an apparent grouping into these three categories. Although some variability occurred in the results of percent positive neurons between observers, all cases ended up in the same categories when we grouped them as above.

No control cases were positive in the hippocampus (0/9 versus 15/27, p < 0.005, control versus SIDS), whereas three control cases were positive in the brainstem (3/7 versus 22/23, p = 0.006).

Hippocampus. Neuronal apoptosis in the hippocampus was found in 15 (55%) of the 27 SIDS victims for which a good section was available. The presence of apoptosis in the hippocampus correlated very well with the presence of necrosis (p < 0.001). In the SIDS cases, 13 of the 15 cases positive for apoptosis were also positive for necrosis. Two control cases displayed necrotic neurons but no apoptosis.

A clear regional distribution of apoptosis and necrosis was noted within the hippocampus (Table 1). The CA4 and subiculum subregions were most often involved in the apoptotic process and, in two cases, these subregions alone displayed positive findings. CA2 was least often involved (four cases) and, when involved, all other regions of the hippocampus were involved as well. This finding may suggest a somewhat more severe insult in these cases. As for necrosis, the subiculum, CA4, and CA1 were involved in the 13 positive SIDS cases, whereas other subregions were negative. In all cases, the number of neurons that had undergone apoptosis much exceeded the number that had undergone necrosis. Representative examples of apoptosis and necrosis are shown in Figures 3 and 4.

Table 1 Neuronal apoptosis in the hippocampus
Full size table
Figure 3
figure 3

Hippocampus: (a) Portion of CA4 showing the distribution of apoptosis; apoptotic cells are stained in brown by the in situ end-labeling method; (b) higher power magnification of the CA4 subregion showing apoptotic neurons adjacent to normal neurons; and (c) apoptotic bodies (arrow).

Full size image
Figure 4
figure 4

Hippocampus: (a) Normal neuron in CA1. (b) Necrotic neuron in CA1; compared with the normal neuron, the necrotic neuron is dense and retracted, its cytoplasm is eosinophilic, and the boundary between the nucleus and the cytoplasm is indistinct. (c) Cells with the morphologic features of apoptosis (arrow).

Full size image

Brainstem. Neuronal apoptosis was present in 22 of the 23 SIDS cases (96%) for which a good section was available and in three of the control cases (all having died of bronchopneumonia or septicemia). Also, the brainstem showed a regional distribution of neuronal apoptosis, with the greatest number of positive nuclei in the dorsal (sensory) region. A representative example of apoptosis in the spinal trigeminal nucleus is shown in Figure 5. The data for each of the positive nuclei are presented in Table 2. Although three cases were positive for apoptosis in the control group, the distribution and degree of apoptosis differed from that of the SIDS victims (Table 2).

Figure 5
figure 5

Brainstem: (a) Overview of the spinal trigeminal nucleus with normal (open arrows) and apoptotic (black arrows) neurons. (b) Higher power magnification of the spinal trigeminal nucleus with a normal and an apoptotic neuron (brown staining).

Full size image
Table 2 Neuronal apoptosis in brainstem nuclei
Full size table

Neuronal apoptosis in temporary survivors. We studied the brainstem and hippocampus of four infants who were successfully resuscitated and survived temporarily for 6, 13, 24, and 96 h (organ donor), respectively. Neuronal apoptosis was present in both the brainstem and the hippocampus of two of these four infants (survival of 13 and 24 h) and in the hippocampus only in one infant (survival of 6 h). The infant who survived 96 h showed no apoptosis but rather a marked diffuse loss of neurons throughout both the brainstem and the hippocampus. The highest degree of apoptosis (>50% of neurons in each subregion of the hippocampus and in nuclei of the brainstem) was noted in the infants who survived 13 and 24 h. The pattern distribution of apoptosis in both the brainstem and hippocampus of the temporary survivors was similar to that of the other SIDS victims.

DISCUSSION

The degree of apoptosis found in the present study is very significant. Apoptosis in 2 to 3% of neurons can signify a cell regression rate of 25%/d (35). Therefore, the extent of apoptosis described in this study (often >20%) implies a high rate of cell death, with potential significant dysfunction in the affected brain regions of these infants.

The presence of both apoptotic and necrotic neurons in the hippocampus of our SIDS victims provides evidence of a serious insult at some time, probably within 24 to 48 h before death. The exact time of the insult is impossible to determine accurately. It could not have occurred less than 4 to 6 h before death because this is the earliest period during which necrosis-induced changes are observable by light microscopy (32,33). Yet, the interval between the insult and the nuclear changes characteristic of apoptosis could not have been much more than 24 h, because the initiation of apoptosis-induced changes must be underway by that point (16,28,34). In addition, the predominant involvement of specific hippocampal subregions also provides information on the time of the triggering event for both types of neuronal injury. After a hypoxemic-ischemic insult, the necrotic process in CA4 occurs within hours of the insult, but the involvement of CA1 is typically delayed, for up to 3 d in adults and less in immature subjects (19–21,36). Our finding in the infants who were resuscitated and survived temporarily seems to confirm the time of 24 to 48 h. Indeed, there was almost no apoptosis in the survivor of 6 h, a marked degree in the survivors of 13 and 24 h, and nothing but extensive cell loss in the survivor of 4 d.

It is impossible for us to assess the presence of previous apoptosis or enhanced apoptosis as a continuous slow process secondary to repeated insults in our group of SIDS victims. Once triggered, the apoptosis process is fast and cells can disappear in 24 h. Our technique identifies only recent apoptosis. Previous apoptosis can be evaluated solely by identifying significant neuronal loss in the regions of interest. In fact, neuronal loss evaluation in various brain regions requires volumetric counting of serial sections. One cannot evaluate the number of neurons properly in a single section because neuronal populations (especially in brainstem nuclei) are not always uniformly distributed throughout the brain and brainstem nuclei.

How does the present finding of neuronal apoptosis in the hippocampus and brainstem fit into the current hypotheses regarding the cause of SIDS? It is now clear that repeated, at times severe, hypoxemic and/or asphyxial episodes occur in infants who are at increased risk for SIDS (37). In accordance with the most widely accepted hypothesis on the pathology underlying SIDS, the repeated events are secondary to a brainstem dysfunction possibly resulting from an adverse intrauterine milieu or to events occurring postnatally (9,38–40). The brainstem is more sensitive to hypoxemic-ischemic injury in fetal and early postnatal life than later on (22,23), and the immature, in contrast with the adult, brain appears to respond to injury with apoptosis rather than with necrosis (24,41–43). Future SIDS victims may well have experienced repeated triggering for apoptosis and, eventually, significant dysfunction because of neuronal loss in the brainstem. For instance, neuronal loss has been reported in the arcuate nucleus in brainstems of SIDS victims (44), a region thought to be involved in control of respiration. Perhaps, then, damage occurred at critical stages of the development of cardiorespiratory control. Such damage will predispose to cardiorespiratory pattern abnormalities that, in turn, could cause the hypoxic-ischemic insult leading to apoptosis in other brain and brainstem areas, as found in the present study. We did find significant neuronal apoptosis in the nucleus of the tractus solitarius, a region involved in autonomic and respiratory control. However, it was impossible for us to assess precisely the degree of apoptosis in the arcuate nucleus. The arcuate nucleus is a thin band of cells located at the periphery of the brainstem, and peripheral regions are often positive with the method we used even though the nuclei of the neurons lack the features of apoptosis (possible nonspecific staining because of drying).

Our findings of predominant apoptosis in the vestibular and trigeminal nucleus in the brainstem could have important implications. The vestibular system mediates postural reflexes and information concerning the orientation of the head; the trigeminal system mediates cutaneous and proprioceptive sensations from skin, muscles, and joints in the face and mouth. Perhaps, then, a dysfunction at the level of the vestibular and trigeminal nuclei (because of repeated events that triggered apoptosis in those regions) is linked to the fact that prone sleeping is a major risk factor for SIDS (45–47). The prone sleeping position (especially when the face is turned into the mattress) has been shown to lead to hypoxemic and hypercapnic episodes (48,49). It is generally thought that normal infants should be able to turn or lift their heads to avoid asphyxia. However, we speculate that if infants have repeated damage to the system that controls sensation of the face and orientation of the head, suffocation may occur.

Another possibility is that SIDS victims, compared with normal infants, are vulnerable to apoptosis because of a relative deficiency of protective factors. The bcl family of protooncogenes, for instance, encodes, proteins that protect cells from apoptosis. Their expression changes during development in the newborn rat (50), and their expression also changes during fetal life in humans (51). We do not know the time course or mode of expression of bcl-xL (expressed in human brain) in postnatal life in humans, nor whether the exposure to repeated hypoxia could modify it. Interestingly, whereas hypoxemia-ischemia leads to substantial apoptosis in the hippocampus of neonatal mice (wild-type), the overexpression of bcl-xL (transgenic mice) is neuroprotective (52). The absence of apoptosis in the hippocampus of our control infants and in those who have died of cardiac disease (53) favors the hypothesis that SIDS victims are more vulnerable to apoptosis than are other infants. However, several brain regions should be examined before such a conclusion is definitively drawn; a study of the expression of promotors and/or suppressors of apoptosis should then be undertaken.

In summary, we report, in 79% of a group of SIDS victims, significant neuronal apoptosis in the brainstem and in subregions of the hippocampus known to be vulnerable to hypoxemic-ischemic insults. The event that triggered the apoptosis seems to have occurred several hours before death. This finding, coupled with the reported repeated hypoxemic episodes in SIDS victims before their death, suggests that those infants may have suffered repeated apoptosis resulting in significant neuronal damage and, thus, functional loss in key brain regions. Our findings may well constitute an important first step in understanding the pathogenesis of SIDS and the cause of death in such victims.