Main

Hypoxic-ischemic injury in the perinatal period remains a significant cause of mortality and neurodevelopmental morbidity. Understanding the mechanisms of damage are important if rational neural rescue strategies are to be developed(1).

Studies of asphyxiated infants and animals using 31P magnetic resonance spectroscopy have demonstrated that transient hypoxia-ischemia is followed by a delayed impairment in energy metabolism 8-12 h after the initial injury(2–4). Similarly, in fetal sheep a transient ischemic insult is followed by a delayed increase in electrical CI which reflects loss of cellular ionic homeostasis(5). The mechanisms responsible for these derangements in energy metabolism are unknown but could be due to a critical decline in CBF and substrate delivery(6, 7). We therefore hypothesized that the delayed phase of injury in fetal sheep is associated with vasoconstriction and a reduction in blood and mitochondrial oxygenation.

The hypothesis was investigated in chronically instrumented, late gestation fetal sheep. We used a well described fetal sheep model, in which cell dysfunction is observed by CI and ECoG activity, and the overall severity of cerebral damage measured by histologic assessment(8). Cerebral vascular tone was assessed by observing changes in [tHb] with NIRS. Cerebral mean oxygen saturation was assessed by examining the relative changes in [Hb] and [Hbo2] in relation to tHb, and an estimate of cerebral mitochondrial oxygenation was made using NIRS to detect changes in[Cyto2].

METHODS

Surgical procedure. All studies were approved by the Animal Ethical Committee of the University of Auckland. Fourteen singleton fetal sheep of known gestational age (range 119-133 d) were operated on under 2% halothane/oxygen general anesthesia using sterile techniques. The general experimental approach has been described previously(5, 8). Briefly, the 90-min procedure involved externalization of the fetal head, neck, and forelimbs and placement of polyvinyl catheters into the axillary arteries, amniotic cavity, and brachial vein for measurement of MAP and amniotic pressure. Through burr holes, three pairs of shielded stainless steel electrodes were placed on the dura overlying the parietal cortex, two pairs of ECoG electrodes (5 mm plus 15 mm anterior and 10 mm lateral to the bregma), and a third pair of stimulating electrodes to measure CI (10 mm anterior and 15 mm lateral to the bregma). Cerebral blood supply was restricted to the carotid arteries by ligation of the vertebro-occipital anastomoses, and a double ballooned inflatable occluder was positioned on each carotid artery.

Near infrared light was transmitted from laser diodes at four wavelengths(774, 826, 844, and 910 nm) and carried to and from the head via fiber-optic bundles. Optical prisms, at the ends of the bundles, were surgically fixed onto each side of the parietal region of the skull with dental cement, at a distance of 3.5 cm apart.

The fetus was then returned to the uterus; the connecting lines externalized through a uterine and maternal lateral skin incision, and the muscle and skin layers were closed. After surgery the ewes were housed in metabolic cages at a constant temperature (16°C) and humidity (50%) and given free access to food and water. Antibiotics (gentamicin, 80 mg i.v., to the fetus and penicillin, 500 mg intramuscularly, to the ewe) were administered daily for 3 d after surgery.

Near infrared spectroscopy. The study used a commercial near infrared spectrophotometer (NIRO 500 Hamamatsu Phototonics KK, Hamamatsu City, Japan). The technique depends on the transmission of near infrared light through tissue and its characteristic absorbance by three chromophores: Hbo2, Hb, and Cyto2. Cytochrome aa3 is the terminal enzyme of the mitochondrial transport chain and is composed of 13 protein subunits, 2 heme units (A and A3), and two copper atoms(CuA and CuB). In the oxidized state, CuA has a characteristic absorption for near infrared light.

Changes in chromophore concentration can be calculated from the changes in light absorption using a modification of the Beer Lambert law which describes optical absorption in a highly scattering medium(9). Using a previously described technique, the “path length factor,” that takes account of the scattering of light in tissue, was calculated to be 4.55 in the fetal sheep head(10). The absorption coefficients of Hbo2, Hb, and Cyto2 are known from previousin vivo and in vitro studies, and therefore changes in the intracerebral concentration of the chromophores can be calculated from the changes in optical absorption(11–13). Changes in [tHb], calculated as a sum of the changes in [Hb] and[Hbo2], relates to changes in cerebral blood volume(14).

Experimental protocol. Fetuses whose arterial blood gas tensions and lactate concentration were normal 2 d after surgery (Po2> 2.27 kPa, pH > 7.32, lactate <1.2 mmol·L-1) were entered into the study. NIRS, CI, MAP, and filtered ECoG recordings were commenced 12 h before the ischemic insult, continued for a further 4 d, and stored on computer hard disks for later analysis. In addition, raw ECoG was recorded for 1 h before the insult and continued for a further 36 h.

Transient ischemia was induced by 30 min of inflation of the bilateral carotid occluders with saline and confirmed by an isoelectric ECoG and a rise in CI. Sao2, Po2, Pco2, lactate, glucose, and Hb were measured before and immediately after the end of the occlusion, and at frequent intervals during the study period.

Four days after the ischemic insult, ewes were killed with an overdose of pentobarbital. The brain was perfused through the common carotid arteries with 4% phosphate-buffered paraformaldehyde.

Data collection. ECoG was recorded continuously, and intensity spectra were obtained by real time spectral analysis as previously described(15). The ECoG signal was amplified 10,000 times, low pass filtered at 30 Hz, and sampled at 256 Hz. An eighth order Butterworth low pass filter was used with the cut-off frequency set with the -3 dB point at 30 Hz. A four-electrode technique was used to extract spectral ECoG and the CI signal from the ECoG. Increases in CI are associated with a fall in extracellular space and reflect loss of ionic homeostasis across cell membranes associated with cytotoxic edema(5, 16).

NIRS recordings of changes in [tHb], [Hbo2], and [Cyto2] were displayed on line every 10 s during the study period and recorded on magnetic disks for later analysis. The distance between the optodes was determined during surgery and confirmed postmortem using measuring callipers.

MAP, recorded from the brachial artery, was adjusted electronically for changes in amniotic pressure and thus changes in the ewe's posture. CI, ECoG, and corrected MAP recordings were processed on line, averaged over 1- and 10-min intervals, and recorded using a database acquisition program (Labview for Windows Version 2.5.1, National Instruments, Austin, TX).

Histologic assessment. Brains were divided sagittally, fixed in 10% formalin, and embedded in paraffin. Histologic sections were stained with thionine acid fuchsin as described previously(8, 17). Injury caused by bilateral carotid artery occlusion is bilateral, and therefore scoring was performed, by an independent assessor (E.C.M), on coronal subserial sections (8 μm) from one hemisphere as previously described(17, 18). Briefly, dead cells were recognized by an acidophillic (red) cytoplasm and contracted nuclei, or just a small rim of red cytoplasm with a pyknotic nuclei, whereas all other cells were considered viable.

Data analysis and statistics. Changes in [tHb], [Hbo2], and [Cyto2] were calculated from alterations in optical attenuation by least-squares multilinear regression using an algorithm that utilized accurate component spectra, took account of wavelength dependent scattering, and calculated the changes in chromophore concentration together with the residual error of the data.

ECoG intensity data were log transformed and analyzed to determine the onset of epileptiform activity as indicated by the development of intense low frequency activity(19). Seizures were confirmed by analysis of the raw ECoG data by a clinically applied electroencephalographic interpretation program (Monitor 7.0 Stellate systems, Quebec, Canada) that determines seizures by analyzing spike wave and polyspike activity(20).

Descriptive measurements of maxima, minima, and 10% rise and fall times were made on data that had been median-filtered over 500 s. Significant changes during the study period were determined by analysis of variance with time as the repeated measure. Specific changes were then determined by Newman Keul's multiple comparison test. Paired t tests and linear regression analyses were applied when appropriate. The univariate relation between changes in NIRS variables and histologic data, CI, ECoG, MAP, Paco2, Pao2, arterial Hb, lactate, and glucose concentrations were determined.

Off-line signal analyses were performed by Viewdac Data Acquisition, Version 2.1 (Keithley Data Acquisition Division, Keithley Instruments, Inc., Taunton, MA), and statistics were calculated using the Sigmastat Statistical Analysis System, Version 1.02 (Jandel Scientific, Erkrath, Germay). All results are presented as mean ± SEM. All time points referred to relate to the end of the carotid artery occlusion as time zero.

RESULTS

Fourteen subjects were investigated. Two of these animals were studied as sham operated controls. Six experiments were rejected: three due to inadequate carotid artery occlusion, two due to premature labor, and one due to sepsis before the end of the study period. The six fetuses fulfilling the study entry criteria and the two control subjects were comparable in gestational age (126± 2 d), weight (3.3 ± 0.2 kg), and biparietal diameter (5.4± 0.1).

Changes observed in the NIRS and electrophysiologic variables are expressed conventionally as change from a baseline obtained from the 12-h preinsult recordings. For the NIRS variables and ECoG log-transformed data, the baseline was normalized with respect to the mean value of preinsult recordings and changes recorded in μmol·L-1 and dB, respectively. CI measurements were expressed as percentages of preinsult values.

Evidence of Cerebral Injury

Biophysical measurement. Figure 1 shows CI and ECoG in the study group and demonstrates the delayed changes seen in previous studies using this preparation(5). There are two periods when CI increased above baseline: 1) during ischemia and2) a delayed phase commencing 17.5 ± 2.3 h postischemia, peaking at 42.3 ± 2.4 h and lasting 66.0 ± 4.9 h. ECoG was depressed during and after ischemia, and increased in amplitude at 13.6± 3.0 h due to seizures, confirmed by inspection of raw ECoG, that persisted for 25.4 ± 3.2 h. Values for CI and ECoG during and after ischemia related to changes in [tHb] are shown in Table 1. Worse histologic outcome was related to a more depressed ECoG amplitude at 96 h postinsult (r = 0.93; p < 0.01) and a lesser recovery of CI to baseline after the insult (r = 0.82; p < 0.05).

Figure 1
figure 1

Changes in (a) cortical impedance and(b) electrocortical activity after transient cerebral ischemia. In(a) the changes in cortical impedance (ΔCI) over 96 h during and after transient cerebral ischemia (n = 6) are shown. Symbols represent the mean of averaged data for each fetus as percentage (%) change from preinsult baseline (100%). There was an increase in CI during the insult, a period of recovery, and a delayed increase commencing several hours later. In (b) the changes in electrocortical activity (ΔECoG) in decibels (dB) from preinsult zero baseline are shown. ECoG is depressed during and after cerebral ischemia, recovers toward baseline, and then is finally depressed to an extent that relates to severity of histologic outcome(p < 0.01).

Full size image
Table 1 Changes in [tHb] and other variables during and following cerebral ischemia
Full size table

Histologic analysis. Table 2 gives the histologic results. Damage was seen in a similar distribution in all subjects, but the severity of damage was variable. Laminar necrosis was observed in the parasagittal region of the cortex with a lesser degree of injury in the hippocampus, lateral cortex, striatum, and dentate gyrus.

Table 2 Neuronal loss 4 d after transient cerebral ischemia
Full size table

Cerebral Hemodynamics and Oxygenation

Total cerebral Hb concentration. Figure 2a shows changes in [tHb] in the six fetuses. Inspection of these data revealed a characteristic pattern in all subjects, with two increases of [tHb] above baseline. Changes in [tHb] could be defined post hoc into five phases, here given with a mean time of onset and the duration of each phase± SEM: 0, preischemic baseline (-12; 11.5 h); 1) fall in [tHb] during carotid artery occlusion (-0.5; 0.5 h); 2) first postischemic increase in [tHb] (0.5 ± 0.1; 2.3 ± 0.4 h);3) [tHb] returned to baseline (2.7 ± 0.4; 10.0 ± 2.4 h); 4) second postischemic increase in [tHb] (12.8 ± 2.0; 43.1 ± 5.2 h); 5) return toward baseline at end of study period (55.9 ± 5.2; 7.0 ± 4.6 h).

Figure 2
figure 2

Changes in NIRS variables after transient cerebral ischemia averaged into variable time bias. Changes in (a) total cerebral Hb (Δ[tHb]), (b) oxyhemoglobin(Δ[Hbo2]), and (c) oxidized cytochrome oxidase(Δ[Cyto2]) from preischemic normalized baseline (n = 6). In each graph, different symbols represent data from each fetus. A fall in[tHb] during ischemia is followed by a first postischemic increase in [tHb], then a return to baseline, and a second postischemic increase in [tHb] commencing several hours later. [tHb] returns toward baseline at the end of the study period. In (b) Δ[Hbo2] demonstrates similar changes although a persistent increase in [Hbo2], when [tHb] returns toward baseline, demonstrates an increase in the mean cerebral oxygen saturation during this time. In (c) a fall is demonstrated in[Cyto2] during the insult, a period of relative stability and then a progressive fall until 78-80 h postinsult.

Full size image

Table 1 shows the changes in [tHb] and other variables during these phases. Linear regression analysis relating [tHb] to other variables showed that: (a) the duration of the first rise in [tHb] was longer in fetuses with more severe histologic outcome (p < 0.05)(Fig. 3a). (b) The time of onset of the second increase in[tHb] was earlier in subjects with more severe histologic outcome (p< 0.01) (Fig. 3b). (c) The duration of the second rise in [tHb] was shorter in subjects with more severe histologic injury(p < 0.01) (Fig. 3b). (d) In all animals the onset of the delayed increase in [tHb] preceded the onset of the second increase in CI (p < 0.05). (e) MAP increased during ischemia and during phase 4, but linear regression analysis showed no significant relation between [tHb] and MAP. (f) Arterial Pao2, arterial oxygen Sao2, Paco2, and Hb concentration were constant throughout. Arterial pH and glucose and lactate concentrations altered during ischemia, but not during the later phases of the experiment. No other significant relations were found(Tables 1 and 3).

Figure 3
figure 3

The relation between Δ[tHb] and histologic outcome after transient cerebral ischemia. In (a) is shown the linear relation between the duration of the first increase in [tHb](r = 0.82; p < 0.05) and overall neuronal loss score, and in (b) is shown the relation between time of onset (•)(r = 0.93; p < 0.01) and duration (○) (r= 0.97; p < 0.01) of the second postischemic increase in [tHb] and overall neuronal loss score. A worse histologic outcome is associated with a prolonged first increase in [tHb] and a second postischemic increase in[tHb] that commences early and is short-lived.

Full size image
Table 3 Changes in [Hbo2], [Cyto2], and arterial blood gases during and after cerebral ischemia
Full size table

Cerebral oxygenation

Figure 2 shows the changes in concentration of(b) [Hbo2] and (c) [Cyto2] in the six fetuses. Changes in [Hbo2] and [Cyto2] during the five phases in[tHb] are given in Table 1. During phase 5, [Hbo2] remained increased, although [tHb] returned toward baseline, demonstrating an increase in the mean cerebral oxygen saturation.

The optical signal attributed to [Cyto2] declined during the insult and immediately after. It then showed no significant change until 28-30 h postinsult when it fell progressively to a minimum of -5.0 ± 2.8μmol·L-1 at 78-80 h. The maximal fall was variable among the subjects and a greater maximum fall in [Cyto2] at 78-80 h was related to worse histologic outcome (r = 0.82; p < 0.05)(Fig. 4a) and greater depression of final ECoG amplitude(r = 0.83; p < 0.05) (Fig. 4b).

Figure 4
figure 4

The relation between maximum fall in [Cyto2] and outcome after transient cerebral ischemia. Shown is the relation between(a) neuronal loss (r = 0.82; p < 0.05) and(b) final ECoG (r = 0.87; p < 0.05) and maximum fall in [Cyto2] at 78-80 h postischemia, indicating that a greater fall in the optical signal for [Cyto2] is associated with a worse histologic and electrophysiologic outcome.

Full size image

Control Subjects

In two control animals the carotid arteries were not occluded, but the subjects were observed for the full 96-h study period. There was no significant acute or delayed change in CI, ECoG, NIRS variables, MAP, or blood gases during the study period.

DISCUSSION

Near infrared spectroscopy. The accuracy of NIRS for detecting alterations in [tHb], [Hbo2], and [Cyto2] has been described previously(21). It requires precise absorption spectra, the validity of the Beer Lambert relationship, and a constant optical path length throughout the experimental period.

Accurate component spectra are available, and multicomponent analysis has confirmed that the brain contains no other mobile chromophore(12, 13, 22). The validity of the Beer Lambert law has been confirmed by theoretical modeling and experimental measurement of light in scattering media, and is further supported by the close match between spectra measured in vivo and in vitro(21). Surgical fixation of the optodes to the fetal skull, which at this gestation has fused sutures, eliminated any alteration in optical path length due to movement, although changes in the scattering properties of the brain after ischemia have not been investigated and need to be considered when interpreting of these results.

When the arterial [Hb] and cerebral:peripheral hematocrit ratio remains constant, changes in [tHb] measured by NIRS are a precise measure of changes in cerebral vascular volume(23). This reflects changes in cerebrovascular tone, which unlike measurements of CBF, can be interpreted without considering the arterial blood pressure. As NIRS does not determine whether vasodilatation occurs in arteries, capillaries, or veins, it offers an estimate of average vascular tone.

[tHb] is the sum of [Hbo2] and [Hb] and observation of these component parts allows an estimation of the overall state of oxygenation in the cerebral blood. This can be related to the signal derived from[Cyto2], which under suitable conditions reflects the oxidation-reduction balance in mitochondria(24). However, cytochrome aa3 is probably present only in very small concentrations in the fetal sheep brain and is more susceptible to error due to subtle changes in tissue optics than Hb measurements, so that[Cyto2] results should be interpreted with caution.

Cerebral vascular changes after ischemia. [tHb] increased during two discrete periods after cerebral ischemia. This was not contingent upon alterations in arterial Hb concentration, MAP, arterial blood gases, or electrophysiologic changes, and was too large and rapid to be ascribed solely to alterations in the cerebral:peripheral hematocrit ratio. The changes must therefore represent periods of cerebral vasodilation.

The first of these periods occurred immediately after the end of ischemia, and may be related to the increase in CBF noted by other groups investigating hypoxic-ischemic injury(6, 7, 25, 26).

The second period of vasodilatation occurred several hours later. Clinical studies of newborn infants suffering from birth asphyxia have found increased CBF and cerebral blood volume some 12 h or more after birth(23, 27). However, this has not previously been described under experimental conditions, and in some well characterized experimental models, there is a delayed reduction in CBF after perinatal hypoxia-ischemia(7, 28). It is not clear whether this discrepancy reflects different species, the effect of different experimental conditions, or the result of different techniques for measuring CBF. However, our results demonstrate that, after ischemia, delayed cellular dysfunction is accompanied by vasodilation rather than vasoconstriction, and this appeared to mirror the effects of birth asphyxia in newborn infants(23).

The excellent time resolution of NIRS allowed precise definition of the time course of these cerebrovascular changes. It demonstrated that the first increase in [tHb] was short-lived, that the second increase began some hours before CI began to increase, and that this was not directly related to the occurrence of seizures. This suggests that vascular changes occur before cellular function is sufficiently deranged to impair ion transport.

The duration and pattern of cerebrovascular changes were related to the severity of cerebral damage: more severe injury was associated with a larger first increase in [tHb], an earlier onset of the second increase, and a shorter duration of the second period of increased vascular volume. These results cannot yet be adequately interpreted, but suggest several hypotheses which can be investigated in this fetal sheep model. Among the questions that we intend to investigate further are: Are the two periods of vasodilatation determined by different chemical mediators? Does the first hyperemia reflect the production of a vasodilatory agent which may damage cells? What mediates the second period of vasodilation, and does this reflect an endogenous protective mechanism? We have recently shown that nitric oxide contributes to the extent of the delayed vasodilation(29).

Cerebral oxygenation after ischemia. Observation of changes in[Hbo2] and [Hb] demonstrated that [tHb] declined after reaching its second peak, and most of the fall was accounted for by a decrease in cerebral[Hb]. Thus relatively more Hb remaining in the brain was oxygenated, reflecting an overall increase in mean cerebral oxygen saturation. However, NIRS estimation of [Cyto2] suggested that, from about 20 h after reperfusion, mitochondrial oxygenation declined progressively.

These results may demonstrate decreased cerebral oxygen uptake due to disordered mitochondrial oxygen consumption. Changes in high energy phosphates consistent with mitochondrial damage have previously been demonstrated after hypoxia-ischemia in other species(2–4, 30–32). Disturbances in mitochondrial DNA expression(33) and disruption of cytochrome oxidase activity(34) occur during early reperfusion and are aggravated over time, suggesting that such a disturbance could cause progressive failure of energy production in cells that eventually die.

However, some complexities need to be considered. No direct measurement of oxygen consumption was made in this study, and, although unlikely, it is possible that increased cerebral oxygenation could be due to increased cerebral perfusion despite the coincident decline in cerebral blood volume. Further studies of cross-brain oxygen content and CBF will answer this question.

The measurements of [Cyto2] may also be misleading. The concentration of [Cyto2] in the fetal brain is probably very low; in rat pups it is only 1.2 nmol·g of brain-1(35). This is supported by the relatively small fall detected during total cerebral ischemia, when complete reduction of cytochromes could be expected. However, the larger late fall in [Cyto2] was closely related to two independent measures of the severity of brain injury: the number of dead cells and the final depression of the ECoG; it is therefore very unlikely to be an error.

It is possible that the change in [Cyto2] late in the experimental period reflected either the removal of cellular materials from the optical field as cells died, or subtle changes in the optical properties of tissue after the cerebral injury. This is an exciting possibility which may offer novel approaches to the investigation and diagnosis of delayed cerebral injury, and requires further investigation.

Conclusion. Contrary to our initial hypotheses, this study found that the delayed changes in cellular function were not associated with cerebral vasoconstriction and a fall in mean cerebral oxygen saturation. On the contrary, a complex biphasic pattern of vasodilatation, with a late increase in the oxygenation of cerebral blood was revealed. [Cyto2] apparently declined after ischemia, but further information on the optical characteristics of cerebral tissue is required before this can be interpreted confidently as evidence of mitochondrial damage.