Main

Perinatal HI causes acute alterations in membrane function and failure of electron transport and oxidative phosphorylation, ultimately leading to rapid cell death in the vulnerable areas. The subsequent release of excitatory amino acid neuro transmitters and recurrent ischemic depolarizations trigger a metabolic cascade on a slower time scale, which leads to cell swelling and finally to dissolution of cell membranes in a zone surrounding the irreversibly damaged infarct core. This penumbral zone is characterized by a longer tolerance to ischemic stress and a slower rate of disintegration, because of a more moderate metabolic derangement and minor abnormalities in cerebral blood flow(1). There is a rapidly accumulating wealth of knowledge about neuroprotective strategies with regard to viability thresholds and probability of recovery of the ischemic penumbra(2,3).

Postasphyxic cytotoxic edema is one of the earliest events than can be observed by MRI(4). Of special importance is its size because it marks the brain areas at risk for HI damage(5). During the subsequent development of the HI injury, the relationships between the temporal evolution of the cerebral energy depression, the development of cytotoxic and vasogenic brain edema, and the size of the final infarction are of interest, because it contains important information for tailoring appropriate therapeutic intervention.

The aim of the present study was to investigate this relationship using repeated noninvasive MR spectroscopic and imaging measurements in combination with histopathologic examinations at specific times during asphyxia and subsequent recovery in 7-d-old rats.

METHODS

Animal model. The animal preparation has been described in detail previously(6). In brief, 34 7-d-old Sprague-Dawley rats weighting 16.8 ± 3.3 g (mean ± SD) were subjected to ligation of the RCCA under general anesthesia (1% halothane in a gas mixture of N2O:O2 of 70:30%). One hour after the operation, hypoxia was induced by lowering the inspired oxygen to 8% and maintained for 90 min while spontaneous respiration was monitored.

The animals were kept in a specially designed servo-regulated cage, which guaranteed a body temperature of 37°C during the entire experiment. The study was approved by the Ethical Committee of the University of Zurich.

Development of edema was observed with repeated MRI measurements, consisting of DWI, T2WI, and processed maps of the ADC, in 12 animals during hypoxia and at 0, 2, 12, 18, 24, 32, 38, and 52 h and 5 d of recovery. At the end of the observation period (5 d) these animals were killed for histologic examination. The energy metabolism was followed in another cohort of 10 animals using 31P-MRS at 0, 3, 6, 13, 20, 24, 28, 32, 38, 52, and 80 h of recovery after HI. Additional MRI at 32 h and 5 h allowed us to correlate the extent of the injury with the evolution of the brain energy metabolism. To correlate MRI findings with the corresponding morphologic alterations at the cellular level, two randomly selected animals (n = 12) were killed for histologic examination at specific times: at the end of hypoxia and at 2, 8, 15, 24, and 52 h of recovery. The control group consisted of 18 animals. Six of them did not have any intervention and underwent MRI and MRS examinations on postnatal days 7-11. The remaining animals were exposed to either 90 min of hypoxia (n = 6) or to ligation of the RCCA (n = 6). From each of the latter subgroups, two animals per time (n = 12) were killed for histologic examination after having completed the acquisition of MRI at 24 and 48 h and 5 d after the intervention.

31P-MRS was performed with a home-built three-turn surface coil designed to fit the affected hemisphere. The spectra were obtained using adiabatic 90° pulses with a one-pulse sequence. The homogeneity of the static magnetic field was optimized individually using the 1H signal. The spectra (band width 5 kHz; 512 averages; TR 8 s) were fitted fully automatically in the time domain with prior knowledge(7) using a combination of Lorentzian and Gaussian line shapes and 10-Hz line-broadening.

All measurements were performed on a 2-T whole-body MR system (Bruker-Medical, Faellanden, Switzerland) equipped with an actively shielded gradient insert with a 33-cm bore, maximal gradient strength of 30 mT/m, and 150-µs rise time. Operating frequencies were 85.15 MHz for protons and 34.47 MHz for phosphorus. During the MR procedure, the animals were kept in a thermoregulated animal holder under light general anesthesia (0.5% halothane in O2:N2O mixture). Respiration, pulse rate, and blood oxygenation were monitored using pulse oximetry. Body temperature was measured rectally.

MRI. The proton signal for MRI was acquired with a specially designed home built volume coil. T2-weighted images were acquired with a multislice RARE technique(8) with TR = 4 s, RARE factor = 16, and interecho interval = 22 ms, resulting in an effective echo time of 252 ms. The matrix size was 256 × 128 (pixel dimensions 156 × 312 µm), and the FOV was 4 cm × 4 cm. DWI was performed with stimulated echoes using the following acquisition parameters: echo time = 18 ms, TR = 2 s, b = 1290 s/mm2, matrix size = 128 × 64, and FOV = 4 cm × 4 cm(9). The slice packages for the T2WI and DWI consisted of eight 1.5-mm-thick slices in the axial plane interleaved by a 0.3-mm gap, covering the entire brain. To calculate quantitative ADC maps, a single-slice, single-shot sequence consisting of diffusion-sensitized spin echo followed by a RARE loop for signal acquisition was used(10). Data for eight images with respective b values of 5, 45, 101, 180, 282, 406, 552, and 718 s/mm2 were acquired from a 2.5-mm-thick axial slice placed at the level of the pineal recess of the third ventricle(11). Phase and amplitude correction(12) was applied to the raw data before calculating the images. Finally, the ADC was calculated for each pixel using a linear fit to the natural logarithm of the pixel intensities of the different images as a function of the b value, and ADC maps devoid of T2 effects were reconstructed in a matrix size of 128 × 64 covering an FOV of 4 cm × 4 cm.

Histopathologic examination. Animals were transcardially perfused under deep anesthesia with 30 mL of 0.9% PBS followed by 30 mL of 4% formaldehyde solution. Brains were then removed and placed in the same fixative solution for a minimum of 48 h, dehydrated, and embedded in paraffin. Consecutive coronal sections of 5-µm thickness were cut at the levels of the anterior forceps of the corpus callosum, the caudate putamen, and the rostral and caudal portions of the hippocampus. Sections were stained with HE for assessment of neuronal damage. For further analysis of astroglial activation, immunocytochemistry was performed with a polyclonal rabbit anti-GFAP antibody (1:300) (DAKO, Zug, Switzerland) and visualized with biotinylated swine-anti rabbit serum (1:250) (DAKO), avidin-peroxidase (DAKO), and diaminobenzidine (Sigma, Buchs, Switzerland).

To complement the HE staining in detecting apoptotic or necrotic cells, in situ end labeling of fragmented DNA (TUNEL) assay was performed. Sections were washed in Tris buffered saline (TBS) and rinsed in reaction buffer for terminal transferase (Boehringer) containing 1 mM CoCl2, 0.2 M potassium cacodylate, 0.25 mg/mL BSA, and 25 mM Tris-HCl, pH 6.6, and then incubated under cover slips with 60 mL of the labeling mix [100 U/mL terminal transferase (Boehringer) and 8 mg/mL Dig-DNA labeling mix (Boehringer)] in reaction buffer for terminal transferase for 60 min at 37°C. After rinsing in TBS, sections were blocked with 10% FCS (Sigma) and then treated for 60 min with an alkaline phosphate-labeled anti-digoxigenin antibody Fab-fragment at a dilution of 1:300 in 10% FCS. After extensive washing in TBS, the color reaction was visualized by nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indoylphosphate (Boehringer). Reaction was stopped in TBS, and sections were counterstained with hematoxylin and mounted in Aquamount (BDH Laboratory Supplies, Poole, United Kingdom).

Data analysis. The evolution of the brain energy metabolism was analyzed by plotting the peak area ratio of PCr and Pi as a function of time. Fractional extent of brain edema was calculated from MR images on each slice and summed to yield the entire volume of the edema. The temporal evolution of ADC values, taken from the ADC maps as mean values from 16 neighboring pixels, was plotted for different brain regions. Tissue damage was evaluated macroscopically and microscopically from histologic sections to estimate the extent of edema and the condition of neuronal and glial cells. The contralateral brain regions in the histologic sections and in the MR images served as internal controls. According to Edwards et al.(13), cells were considered necrotic when cytoplasmic eosinophilia accompanied by nuclear chromatin dispersion because of nuclear membrane damage was observed. Apoptotic cells were identified by nuclear pyknosis or chromatin fragmentation on HE and TUNEL-stained sections. At times of advanced cellular disintegration, earlier apoptotic cell death was confirmed by the presence of apoptotic bodies.

Statistics. Data sets of PCr/Pi and ADC values taken at different times were not always normally distributed. Therefore, all data sets were analyzed using nonparametric tests for sake of comparability. A level of p ≤ 0.01 was considered to be statistically significant.

The energy metabolism (PCr/Pi) from animals of the experimental group was analyzed for significant difference compared with control animals using the Mann-Whitney U test. Sets of ADC values from different brain regions measured at the same times were analyzed for significant differences using the Kruskall-Wallis ANOVA test.

RESULTS

Alterations of energy metabolism after HI. In healthy neonatal rats, the mean values ± SD of cerebral energy metabolism, expressed as PCr/Pi, increased from 1.43 ± 0.21 on day P7 to 1.56 ± 0.25 on day P11 (Fig. 1). Throughout the observation period, experimental animals showed significantly lower values of PCr/Pi compared with healthy animals. Immediately after HI, cerebral energy metabolism was severely depressed (PCr/Pi = 0.11 ± 0.09)). With reoxygenation this value recovered slowly. The highest levels during recovery of 1.12 ± 0.27 were below baseline and were attained at around 13 h after HI. The subsequent decline of PCr/Pi reached trough levels of 0.48 ± 0.22 between 20 and 30 h after HI, which were significantly less than the values of recovery at 13 h. After a second incomplete recovery of PCr/Pi between 30 and 40 h, the brain energy metabolism deteriorated during the subsequent days, until the animals finally died.

Figure 1
figure 1

Biphasic evolution of brain energy metabolism after HI under normothermic recovery (37°C). Low values after HI (0 h) recover slowly and incompletely (13 h) with reoxygenation and decline again, reflecting extensive brain tissue necrosis.

Full size image

Brain edema. Figure 2 shows diffusion-weighted images (upper row) and T2-weighted images (lower row) describing the evolution of the cytotoxic and vasogenic brain edema, respectively, in one representative animal. In DWI, the hyperintense regions are interpreted as cytotoxic edema and hypointense regions as vasogenic edema(4). The comparison of the regional distribution of the diffusion coefficient, i.e. the ADC maps (middle row), with the signal change in the diffusion-weighted images showed that the T2 contamination of the diffusion-weighted images did not alter the borders of the edema. Thus, for the evaluation of the lesion size from the diffusion-weighted images, the T2 contamination was negligible. In T2 WI, the vasogenic edema appeared hyperintense. The size of the final infarction was best assessed by T2WI.

Figure 2
figure 2

Development of edema visualized with DWI, ADC maps, and T2WI. At the end of HI (0 h), the primary edema shows only a cytotoxic contribution (DWI, ADC). Within 2 h, this edema almost completely vanished, leaving only the HI core region visible as showing both cytotoxic and vasogenic edema. The images at 12 h represent an intermediate stage of the secondary, delayed lesion. The maximal extent is attained at 32 h after HI. At 5 d the infarction is delimited but not completely resorbed.

Full size image

At the end of the hypoxic period, DWI clearly showed an extensive cytotoxic edema covering the entire ipsilateral cortex, hippocampus, basal ganglia, and even parts of the contralateral cortex (Fig. 2, 0 h). The fractional volumes of the edema measured at different time points are listed in Table 1. No vasogenic contribution could be observed in the T2-weighted images. After 2 h of recovery, the cytotoxic edema resolved almost completely except for a small region in the parietal cortex, which also showed weak vasogenic edema. Beginning at 5 to 6 h of recovery, a regrowing cytotoxic and newly forming vasogenic edema occupying the same cerebral volume could be observed. The maximal extent of edema was reached between 20 and 30 h and lasted up to 38 h after HI, comprising the ipsilateral cortex, hippocampus, and basal ganglia. This secondary edema was on average 12% smaller compared with the primary edema during the HI insult and did not cross over to the contralateral hemisphere. After 2 d of recovery the restricted water diffusion characterizing the cytotoxic edema in DWI became gradually overshadowed by the rapidly evolving vasogenic edema. During the following period, the T2-weighted images clearly delineated the vasogenic edema and finally the extent of the infarcted brain tissue. The final volume of brain infarct occupied approximately 80% of the maximal extent of the lesion measured at 32 h after HI, resulting in 35 ± 12% necrotic brain volume.

Table 1 The extent of edema during recovery (% brain volume) and the secondary energy failure for the experimental and control groups
Full size table

In the ADC maps (middle row), low values of ADC appear dark, indicating restricted apparent diffusion, i.e. cytotoxic edema. High ADC values appear hyperintense and represent vasogenic edema.

For normal tissue a mean ADC value of 1.06 ± 0.08 × 10-3 mm2/s was obtained at P7. This value is higher than that reported in the literature for adult rats(12) and reflects the unmyelinated stage of the immature brain with much higher extracellular space volume(13,14). During the observation period, the ADC declined slightly to 0.97 ± 0.07 × 10-3 mm2/s at P11 owing to maturational changes. The evolution of the regional distribution of the ADC values after HI is shown in Figure 3. Immediately after HI, the values of ADC were lowest in the ipsilateral parietal cortex and normalized gradually to the periphery. Corresponding to the severity of the lesion, three zones with distinct patterns of temporal evolution of the ADC were found (Fig. 4). Zone 1 (Fig. 4) comprised the parietal cortex and corresponded to the core of the HI lesion. The ADC recovered transiently at 2 h after HI (Fig. 3) and decreased again beginning at approximately 5 h after HI. Trough values of 0.43 ± 0.12 × 10-3 mm2/s, corresponding to 41% of baseline, were attained at 12 h and lasted until 24 h. Thereafter the ADC increased to finally attain values of free water at 5 d (2.9 ± 0.06 × 10-3 mm2/s at 37°C). This tissue was severely affected by HI and became necrotic within 2 h after hypoxia. At 5 d of recovery it was found in an advanced stage of resorption.

Figure 3
figure 3

Evolution of the ADC for different brain regions: the mildly affected contralateral cortex, and moderately affected basal ganglia, and the severely affected core region, together with the adjacent penumbra in the ipsilateral cortex. The values during the secondary decline are related to local lesion severity as long as the vasogenic component of edema is weak (approximately 10-18 h). The subsequent period is characterized by elevation of the ADC, corresponding to extensive cell lysis and resorption. The error bars are not drawn for visibility. Until 24 h, the SD are ± 0.09 × 10-3 mm2/s and later increase to ±0.15 × 10-3 mm2/s.

Full size image
Figure 4
figure 4

Schematic distribution of brain regions with different severity of hypoxia-ischemia. Zone 1 represents the irreversibly damaged tissue (core of lesion). Zone 2 comprises the penumbra, i.e. reversibly damaged tissue that undergoes delayed necrosis. Zone 3, formed by thin margins of the lesions together with parts of the contralateral cortex, recovers within hours after HI and does not undergo delayed necrosis within 5 d.

Full size image

In zone 2 (Fig. 4), which lay around the core lesion and included large parts of the hippocampus and the basal ganglia, the ADC evolved as described for zone 1, but base values after 12 h were significantly higher (0.71 ± 0.13 × 10-3 mm2/s, corresponding to 67% of baseline). Thereafter the ADC increased and stabilized between values of normal tissue and values of free water. Microscopically, the corresponding tissue was found to undergo delayed damage and to become completely necrotic. It was partially resorbed at 5 d after HI.

Comparing zones 1 and 2, we found the ADC values of differ significantly from each other and from baseline during the first 24 h after HI (Fig. 3). This was attributed to the different severity of damage in the two brain regions. During the subsequent elevation of the ADC, the differences between animals became larger, but not significant. The ADC reflected the tissue disorder caused by massive cellular disintegration and resorption, occurring at slightly different times in different animals. The interindividual differences vanished with progressive replacement of the infarcted brain tissue by fluid.

At the end of HI in zone 3, comprising the ipsilateral border of the HI lesion together with a ribbon of the contralateral cortex, the ADC value was slightly, but significantly, reduced compared with baseline (>0.87 × 10-3 mm2/s, 83% of baseline) and significantly higher than the ADC of zones 1 and 2 (Fig. 3). Within 2 h after HI, the ADC of zone 3 recovered and stabilized at normal levels such that no significant difference could be observed when compared with control animals later during recovery. Histopathologic examinations showed that the corresponding tissue recovered completely and did not undergo secondary damage during the 5-d observation period.

Histopathologic alterations. Immediately after the HI insult neuronal necrosis could be observed in the core, which consisted of an area of the ipsilateral cingulate and parietal cortex (Fig. 5, 0 h. HE). In addition, swollen neurons were found throughout the ipsilateral cingulate and parietal cortex, the hippocampus, and the thalamus, and even in the contralateral parietal cortex. Neuronal swelling became gradually less with increasing distances from the core lesion. The extent of this primary cytotoxic edema exactly matched with the areas of restricted diffusion in the MRI (Fig. 2, DWI, 0 h). No apoptosis, swelling of only a few glial cells, and no glial reaction were observed at these early times.

Figure 5
figure 5

Histopathologic findings in the area of cortical lesion. The upper row shows the extent of damage in representative whole-mount sections in a time course (HE, original magnification ×5). Second row, High-power magnification of the area labeled in upper row. The parietal cortex shows hypereosinophilic neurons at an early time point (0 h), followed by extensive neuronal swelling (8 h) and neuronal loss (24 h, 5 d) (HE, original magnification ×250). No glial activation is detectable after 8 h (third now, GFAP, original magnification ×250). Prominent reactive gliosis is first detectable after 24 h. At 5 d after HI, gliosis is detectable only a thin ribbon, separates normal tissue from the infarct. The TUNEL technique (fourth row, original magnification ×250) together with the HE staining demonstrates that after 8 h mainly neurons undergo apoptosis and necrosis. In contrast, at later stages, extensive and advanced damage of all cell types in detected.

Full size image

At 2 h of recovery after HI (Fig. 5, 2 h) these histologic alterations disappeared to a large extent. Histologic changes persisted and were much smaller in size only in animals with abnormal MRI (Fig. 2, 2 h). The lesion consisted of necrotic neurons and evidence of interstitial edema restricted to a small part of the cingulate and upper parietal cortex exactly matching the hyperintensities seen simultaneously in both DWI and T2WI. In the immediately adjacent regions, swollen and occasionally apoptotic neurons were found. No astrocytic reaction, as assessed by GFAP immunostaining, was visible at that time.

Between 8 and 15 h after HI, progressive extension of cortical lesions expanding to basal ganglia with extensive neuronal necrosis and apoptosis could be observed. During the same period, increasing numbers of newly activated, swollen, GFAP-positive glial cells contributed additionally to the cytotoxic edema. Furthermore, an interstitial edema became progressively evident.

At 24 h after HI, the final extent of the lesion was clearly delineated and the maximum of neuronal necrotic and apoptotic cell death was detected in the injured area. Disintegrating neurons and increased number of macrophages, together with marked interstitial edema, were apparent throughout the lesion and correlated spatially with the MRI observations. In addition extensive glial activation was evident. Generally, the temporal sequence of cell response, e.g. neuronal death and glial reaction, were related to the severity of HI and to the territory of the RCCA.

At 50 h after HI, the histologic picture was that of an early stage of resorption. Interstitial edema, apoptotic and necrotic neurons, macrophages, and onset of glial disintegration were detectable throughout the lesion. The intensity of the cellular response of neuronal and glial cells is summarized in Table 2.

Table 2 Cellular response in the penumbra during recovery after hypoxia-ischemia
Full size table

During the following days, complete lysis of the injured brain tissue with partial or complete cystic infarction developed. HE staining and TUNEL assay revealed extensive staining, arising from the death of all involved cell types. Although the duration of the HI insult was the same for all animals, at 5 d after the HI insult a range of severity of reactions was noted. This is reflected by the individual extents of the brain lesions. Mild lesions resulted in complete tissue loss in the parietal and cingulate cortex and parts of the hippocampus. With increasing severity the infarct progressively covered the hippocampus together with parts of the basal ganglia and the thalamus (data not shown). Very severe lesions additionally showed selective neuronal necrosis in the contralateral cingulate and parietal cortex.

DISCUSSION

The present work was designed to study the temporal evolution of HI brain injury in 7-d-old rats during a period of 5 d.

Previously we reported that there was a biphasic evolution of brain edema after HI brain injury, with emphasis on delayed events during recovery(9). In the present study, more attention was given to the early edema because it indicated the entire region at risk. This included the territory of the RCCA, where the most severe injury was found in the parietal and cingulate cortex, and crossed over to the contralateral hemisphere (Fig. 2 DWI, 0 h). The latter edema can be interpreted as a response of the most vulnerable cerebral tissue to the combination of hypoxia and general hypoperfusion. The primary edema had a cytotoxic origin and was composed. Of three distinct zones(1), probably related to the cellular response to the gradual impairment of regional cerebral blood flow(14,15): a core of infarct zone (Fig. 4, zone 1), a surrounding zone at risk, adjacent to the core lesion and corresponding to the penumbra (zone 2), and a thin border zone (zone 3), which fully recovered because the HI threshold was not yet surpassed(6).

At the cellular level, extensive neuronal swelling but relatively little glial swelling without glial activation suggests that the main contribution to the primary edema arose from neurons. Indeed, neurons appear to be more prone to swelling with HI than glial cells because they have more pathways for ion movement and do not contain glycogen reserves(16). The regional difference in severity of the HI injury determined by histopathology and ADC values showed a direct relationship to the early phases of cellular swelling during ischemia, as postulated by Vorisek and Sykova(17). During phase I after cardiac arrest in young rats, these authors found no significant changes of the extracellular space, followed by neuronal swelling during phase II and glial and further neuronal swelling during phase III. Degradation of membrane structures, corresponding to phase IV, was not observed in our experiments immediately after HI.

On reoxygenation, the primary edema regressed; however, PCr/Pi was still <30% of baseline. Obviously, the recovery initiated by reoxygenation was sufficient to restore the failure in cell membrane potential and to regulate cell volume. The remaining depression of the cellular energy reserve likely may have reflected the ongoing hypoperfusion and impairment of higher cellular function. The persistent edema visible in the ipsilateral cortex between 1 and 3 h after HI (Fig. 2) represented irreversibly damaged tissue, i.e. the histologic core zone(18,19) of the lesion. This delineated also the true HI core lesion, occupying only approximately 5% of the brain volume, inasmuch as considerable lesion enlargement by extensive cell death occurred later (Table 1).

The secondary edema originated in the cortex adjacent to the core approximately 3 to 4 h after HI, at a time when the animals were once again well oxygenated. Its progression is thought to be associated with a secondary depolarization of the cell membranes caused by a permanent hypoperfusion(5,15), delayed release of neurotransmitters(20–22), and increased concentration of lactic acid(20,23,24) in the extracellular space. From histopathology we interpreted the persisting cytotoxic edema as arising from activated glial cells, whereas the necrotic and apoptotic neurons were gradually enhancing the vasogenic edema. These findings are in concert with a partially recovered energy metabolism at approximately 13 h (Fig. 1), reflecting viable activated glial cells, and emphasize a different time scale of pathology for neurons and glial cells (Table 2).

The subsequent decline of the energy metabolism beyond 13 h (Fig. 1) was similar to that observed in newborns after severe perinatal asphyxia(25–27) and in newborn piglets and infant rats(28–30) after HI. Consequently, it is reasonable to attribute the breakdown of glial energy metabolism predominantly to the observed glial deterioration, caused by energy-consuming processes such as proliferation of processes during glial activation(31), ongoing apoptosis(32–34), the attempt to restore ion homeostasis(9), and repeated spreading depressions(35–37) in an environment with disturbed cell supply. Trough values of PCr/Pi coincided with the maximal extension of the secondary edema. In particular, given the advanced neuronal damage, both the SEF and the maximal size of edema appear too late to be used as indicators for specific neuronal rescue therapy.

It should be noted that the further development of the energy metabolism is characterized by a slight elevation of the fraction PCr/Pi at approximately 30 h. This can be explained at least in part by the elimination of the Pi from the injured volume, increasing the ratio(29). Therefore, this rise does not reflect a true recovery. In contrast, the final drop always heralded agonal deterioration.

Remarkably, although the HI injury comprised initially the full range of severity, i.e. from necrosis in the core to mild cytotoxic edema in the periphery, the exacerbation during the secondary injury did not maintain this graded correspondence, but led to uniform cell death in the penumbra. Thus, tissue with ADC <85-90% of baseline at 12 h finally became necrotic, whereas tissue with ADC >90% of baseline recovered. The latter condition was fulfilled only in a thin border of the secondary lesion. The final size of the injury was therefore similar to that arising from an often-used animal preparation for neonatal HI brain injury, which consists of 3-4 h of rest after vessel ligation followed by 3 h of hypoxia in 8% inspired O2(38–40) and produces severe injuries with large core regions surrounded by a thin penumbra immediately after HI. Thus, the extent of the edema at 1 h after HI was similar to that of the late vasogenic edema(41,42), and the majority of hippocampal neurons were irreversibly damaged already at 2 h after HI(43). In comparison, the animal preparation used in this study produced moderate HI injuries, which were situated at the threshold of core formation. Correspondingly, the core regions were small and the penumbra were large and contained the full range of reversible damage levels, thereby realizing a simulation of moderate rather than severe perinatal asphyxia. The extents of the entire lesion and of the core can be assessed early from DWI at 0 and 2 h, respectively, allowing an individual assessment of the effectiveness of therapeutic interventions.

Conclusions. We conclude that the biphasic evolution of brain edema and energy metabolism reflects early neuronal and late glial damage in response to moderate HI injury. The primary hypoxic-ischemic and the delayed metabolic phases of brain injury are well separated by a temporal window, during which the core lesion can clearly be delineated by DWI. The SEF observed during recovery reflects predominantly glial activation and subsequent glial death. Therefore, it appears well beyond the time during which neurons can be rescued.