Main

Hypothermia significantly prolongs the survival time of young animals subjected to HI injury(1) and reduces brain damage associated with cardiac arrest in man(2), as well as in the newborn dog(3). There is evidence to show that reductions of several degrees in brain or core temperature could have a therapeutic role in reducing ischemic brain injury. Mild intraischemic hypothermia markedly attenuates the release of glutamate into the brain's extracellular space and significantly diminishes the release of dopamine(4). Moderate brain hypothermia reduces infarct size in focal ischemia(4). For adult animal ischemia models, 31P NMR has shown that hypothermia improves cerebral energy metabolism and intracellular pH compared with normothermic and hyperthermic groups, but no direct correlation with brain damage has been demonstrated(5, 6). For a piglet complete ischemia model, brain temperature correlated with 31P metabolite levels as well as with the energy utilization rate(7). Recent studies in the neonatal rat (7 d old) have shown that moderate hypothermia during HI reduces core body and brain (core/brain) temperature and brain damage(8–10). However, hypothermia was not shown to be effective if instituted during a 3-h recovery period immediately after 3-h HI with this model(8).

We have previously reported that, at the control temperature of 37 °C, hemispheric 31P NMR measured and integrated during HI could be used to predict neuropathologic brain damage(11). Whereas 3 h of HI at 37 °C reliably produces brain damage(12, 13), a shorter duration of 2.5 h produces a full range of injury, with the level of 31P metabolite deficits corresponding to subsequent histology and area morphometry(11). If large perturbations in cellular energy metabolism are necessary for subsequent brain damage, then we hypothesize that a lowering of temperature during HI should reduce damage by decreasing enzyme reaction rates and metabolic demand, leading to a preservation of high energy phosphates. Accordingly, we chose to use the same reduced temperatures with the same neonatal model already shown to be neuroprotective(8). Our main objective was to determine the relation between cerebral energy metabolism and brain damage under conditions of modest (mild to moderate) hypothermia.

METHODS

Experimental model. The effect of modest hypothermia on intracellular brain metabolism was evaluated using an established PND 7 rat model of HI(14–17). PND 7 (11-15 g) Wistar rats (Charles River) of either sex were subjected to right common carotid artery ligation, via a midline neck incision(15, 16). During the 3-4-min surgical procedure, each rat was anesthetized with a mixture of halothane (4% induction, 1.5% maintenance), 30% oxygen, and the balance nitrous oxide. The rats were returned to their dams for 2.5 h of recovery followed by 3 h of hypoxia in 8% oxygen. This combined HI results in damage localized mainly to the posterolateral region of the cerebral hemisphere ipsilateral to the ligation. Cystic damage includes neuronal necrosis and reactive astrocytosis involving mainly the dorsolateral cortex at the posterior level, hippocampus, and dorsolateral striatum(11, 17–19). Developmentally, the brain of a PND 7 rat has been considered by some standards to be similar to that of a 32-34-wk preterm infant(17, 18, 20) and as such is a good model for the study of perinatal asphyxia.

Non-NMR studies. Parallel bench-top experiments were performed with ligated pups placed in 500-mL airtight jars (two per jar), which were 50% submerged (water depth of 3.8 cm) in a Haake circulating water bath, maintained at either 37 °C(12, 16, 21), 34 °C, or 31 °C. Core temperature was monitored via a rectal thermocouple probe (T-type, Omega Engineering Inc., Stanford CT) in randomly selected pups. Air temperature at the center of each jar (3-4 cm from bottom) was 32-33 °C for the 37 °C control. The pups were exposed to 30 min of air, followed by 3 h of hypoxia in an 8% oxygen-balance nitrogen gas mixture. Afterward, the jars were opened to room air, and the survivors were returned to their dams. Initially, control jar experiments were performed to determine the appropriate number of pups per jar. Serial rectal temperature was measured for each temperature during HI, and brain swelling was measured 42 h later.

31P NMR studies. During cerebral HI there is a depletion of adenine nucleotides due to impaired synthesis and enhanced degradation. The decrease in ATP ([ATP]) is preceded by a fall in [PCr] and a rise in [Pi] (via the creatine kinase reaction), thereby providing a direct estimate of phosphorylation potential. Accordingly, we used 31P NMR to monitor ATP and the Pi to PCr ratio. All 31P NMR studies were performed with a Bruker AM400 wide-bore spectrometer operating at 162.0 MHz. Each unanesthetized rat pup was positioned in an upright vertical orientation inside a specially constructed probe, which included a gas-tight temperature regulated chamber(12). The temperature at the thermocouple position just above the center of the heating pad was maintained at 36.3-37.0 °C for the normothermic control temperature, and 3 or 6 °C lower for the hypothermic values. For convenience, we refer to the “environmental temperatures” as 37, 34, and 31 °C. A lower ambient temperature recorded in proximity to the head was 31-33 °C for the controls. Twenty-three spectra were acquired in 10- or 20-min time blocks, beginning with a 30-min prehypoxic period using air, followed by 3.0 h of hypoxia in an 8% oxygen mixture and 2.5 h of recovery in air. ATP levels were quantified using the area of the β-ATP phosphate resonance, and pH was determined from the chemical shift of Pi. Values of Pi/PCr were corrected by a relaxation factor of 1.18 due to saturation during the 2-s repetition time. Detailed NMR setup and metabolite data analysis have been described previously, along with representative phosphorus spectra(12).

Brain damage measurements. After both NMR spectroscopy (n = 25) and the experimental jar study (n = 65), hemispheric WC was determined at 42 h of recovery, when brain swelling (edema) is advanced(22). Hemispheric WC was computed as a fraction of wet weight(12, 15, 16). Brain swelling was then determined by normalizing the increase in the hemispheric WC relative to the dry weight fraction of the contralateral WC according to the formula(22):

For a pilot control study of histology versus edema, two separate litters of animals (n = 13) were exposed to HI (31 or 37 °C). The animals were either saved for 42-h WC determination or were returned to their dams for 23 d, at which time they were decapitated, and their brains were stored in fixative before histologic sectioning(18). The severity and distribution of brain damage was determined in a blind fashion by microscopic examination and scoring of coronal sections stained with hematoxylin-eosin and immunoperoxidase stain for glial fibrillary acidic protein(11, 18). A score of one was given to each of 12 possible regions, and an additional score of one to those regions exhibiting cystic infarction.

TTC staining (n = 12) as well as multislice T2-weighted and diffusion-weighted MRI experiments (n = 12) were performed in additional studies for evaluation of the distribution of damage after 3.0 h of HI. For the TTC staining(9, 23) at 18-h recovery from HI, 2-mm coronal slices were taken from the posterior half of the brain, approximately at the level of the mid-infundibulum. Each section was immersed in a 2% TTC solution (0.9% saline) and heated at 37 °C for 8 min. The T2 images were acquired at ambient temperature at approximately 1 h and 42 h post HI, with an echo delay (TE) of 70 ms and repetition time of 2000 ms. Field of view was 24 mm; slice thickness and separation were 1 mm with an intraplane pixel resolution of 128 × 128. Diffusion-weighted imaging was performed using a spin-echo sequence (b value = 1000 s/mm2, δ = 10 ms, Δ = 31 ms)(24) and then was evaluated using standard procedures(24–26). A summary of the experiments performed on PND 7 rats that were subjected to 3 h of HI is given in Table 1.

Table 1 Summary of HI 7-day-old rat pup experiments performed*
Full size table

Statistics. Two-sample population comparisons (unpaired) were evaluated with a pooled variance t test. Data for the three temperatures were also evaluated by one-way ANOVA, followed by Bonferroni multiple t test comparisons. Differences were considered significant at p < 0.05. Time series data were evaluated using a two-way repeated measures ANOVA to assess between-group temperature and within-group time effects. This model assumed an autoregressive structure(27) for the within-animal errors and allowed for an interaction effect of time with temperature. Thus, by assuming that the serial data for each animal is dependent on its position in time, the correlation of the error is smaller for time points that are far apart, and a truer model is evaluated. The Mixed Procedure routine from SAS (SAS Institute Inc., Cary NC) was used for this analysis. Correlations were computed from both linear least squares and the nonparametric Spearman rank test. Both the FE test and the Wilcoxon-Mann-Whitney rank test were used to evaluate pups grouped by a damaged or normal category.

RESULTS

Determination of brain damage. In this study, tissue swelling (edema), measured at 42 h of recovery from HI was the chosen measure for evaluating irreversible brain injury in PND 7 rats, as justified by numerous studies with this animal model (cf. “Discussion”). To further support the use of edema as a reliable index of damage, two small litters (total n = 13) were exposed to HI at 31 or 37 °C, respectively, and were subsequently divided and evaluated for damage by either scored histology or by WC. The 31 °C pups all had either normal WC (n = 4) or histologic damage scores of 0 (n = 3). The 37 °C pups all had either significant edema (average WCR = 3.0% above normal, n = 3) or histologic scores of 17 out of a 24 maximum (n = 2), plus one death during early recovery. As with previous studies(11, 18), the high scores at 37 °C represent fairly uniform damage in each of 12 selected regions of the cortex, hippocampus, striatum, amygdaloid nucleus, thalamus, and globus pallidus with cystic infarctions in five of the regions. Although limited in sample size, the significance of these 37 versus 31 °C scores was high (p < 0.005 for swelling and p < 0.005 for histology). With consideration of the previous studies of damage, further sampling of histology was not considered to be necessary to show that edema is a reliable measurement.

1H MR T2-weighted images show some hyperintensity throughout the right hemisphere as early as 1 h post-HI (Fig. 1A). This effect intensified at around 42 h and is known to be associated with increased amounts of mobile water resulting from vasogenic edema(24–26). Also at 1 h of recovery, marked hyperintensity in the diffusion-weighted image corresponds to a drastic decline in the ADC seen throughout the entire ipsilateral hemisphere of 8/12 animals (Fig. 1B)(24). The remaining four pups indicated only a partial ADC decline in the cortex. The affected 1H MR imaging regions (Fig. 1,A and B) correspond spatially to the TTC staining results at 18-h recovery from numerous other animals (Fig. 1C). This suggests that damage from the “37 °C pups” is distributed throughout the right hemisphere in good agreement with the position of the 31P radiofrequency surface coil detector(12).

Figure 1
figure 1

Demonstration of the uniform spatial extent of brain damage, in the right hemisphere, from representative pups subjected to 3 h of HI in a 37 °C jar experiment. Coronal MR images at the posterolateral region were obtained at 1 h of recovery using T2 weighting (A), and diffusion weighting (B) for the same pup. The T2 image is sensitive to vasogenic edema (lighter region), whereas the hyperintensity in the diffusion image is indicative of cytotoxic edema due to restricted water (cf. text). At 18 h of recovery, the coronal TTC section at the same level (C) shows similar results with light gray (white) regions of irreversible ischemic damage (lacking electron transport chain enzymes) and dark gray (red) regions being normal.

Full size image

In control temperature studies, serial core temperatures were measured during HI as a function of the number of pups per jar (n = 75). Yager et al.(8) have previously shown that core and brain temperatures are very similar during normoxia and HI for all three environmental temperatures. Figure 2 shows representative results with the water bath maintained at 37 °C (A) and at 31 °C (B). Results for two and four pups per jar (not shown) are intermediate to the displayed 1, 3, or 5 pups. Although not apparent at 37 °C, rectal temperature remains elevated with more pups per jar at 31 °C. Consistent with the expected relation of core/brain temperature with brain injury, more than 30% of pups suffered mild edema with five pups per jar. With one or two pups per jar, edema was rare at 31 °C (see below). From this result, and for practical reasons, two pups per jar was chosen for all other bench experiments. For this case, the typical variation of rectal temperature with time and among pups is shown in Figure 2,C and D, for both 37 and 31 °C. Core temperature measures about 36 °C (34.5-37 °C) during HI when the bath is at 37 °C and falls to below 31 °C for the 31 °C bath temperature. For the 34 °C bath, rectal temperature averaged 32-33.5 °C. To avoid interference during NMR acquisition, continuous rectal measurements were made only in a selective group. When there was only one restricted pup in the NMR chamber, core temperature closely followed the environmental temperature, after an initial period, similar to that of Figure 2,C and D.

Figure 2
figure 2

Core (rectal) temperature of representative rat pups during HI as a function of number of pups per jar. The water bath was maintained at 37 °C in (A) and at 31 °C in (B). Results for two and four pups per jar (not shown) are intermediate to the displayed one, three, and five pups. The variation of rectal temperature with two pups per jar is shown at 37 °C in (C) and at 31 °C in (D).

Full size image

Normal hemispheric WC was 87.77 ± 0.33% with values greater than 88.40% (± 2 SD) considered to be abnormal (edema). Because slight dehydration of the brain during recovery of damaged pups has been measured, the relative tissue swelling parameter has been found to be a more reliable indicator of damage than direct water content. Tissue swelling (and NMR metabolite parameters, described below) varied very significantly with the three temperatures for the NMR animals (Table 2). There was one death during late HI at 37 °C. Similar swelling results were obtained for the parallel jar experiments (n = 65) with edema in 21/21 pups at 37 °C and only 1/24 at 31 °C. There was also one death at 37 °C. For the jar experiments, percent tissue swelling was 0.01 ± 2.68, 20.34 ± 15.19, and 36.70 ± 6.40 at 31, 34, and 37 °C, respectively. For both protocols, one-way ANOVA showed high significance of tissue swelling with temperature (p < 0.0001).

Table 2 NMR metabolite and damage values versus temperature for HI rat pups*
Full size table

31P NMR studies. Figure 3A shows the 31P NMR time course of brain ATP expressed as a fraction of its initial baseline (normal) values for environmental temperatures of 31 °C (n = 8), 34 °C (n = 9), and 37 °C (n = 8). With regard to the statistical sample of NMR values, it should be noted that each animal was obtained from a separate litter and randomized for temperature during 6 mo of experiments. For absolute quantitation, the concentration of ATP in normal PND 7 rat brain is 2.80 ± 0.04 mM(15, 28). From Figure 3A, there was a dramatic preservation of ATP levels during HI with moderate hypothermia of 6 °C. The averaged curve for 31 °C briefly reaches a minimum above 70% relative to baseline, whereas under normothermia the curve decreases to a minimum of 40-45% for most of the last hour of HI. A significant preservation is also observed with hypothermia of only 3 °C, but these data are averaged from curves representing a mixture of moderately damaged and undamaged pups. In this model, pH has been shown to drop from its normal value of 7.16 ± 0.06, by only 0.1-0.3 U during the initial 1.5 h of HI(12, 13). During hypothermia, the decrease is less than 0.2 U. Pi/PCr (inversely proportional to the estimator of phosphorylation potential when pH changes are small) has been found to be the most useful NMR parameter for predicting edema and neuropathologic injury(11–13, 21, 29). This is due to its higher dynamic range than ATP, and being a ratio, is insensitive to animal movement away from the NMR surface coil. Averaged curves for Pi/PCr are shown in Figure 3B. This parameter reaches a maximum value, at about 2.5 h HI, of 4.6 for normothermia and only 2.0 for hypothermia of 6 °C.

Figure 3
figure 3

Average 31P NMR values (±SEM) for PND 7 rat pups maintained at environmental temperatures of 31 °C (n = 8), 34 °C (n = 9), and 37 °C (n = 8). The vertical lines indicate the beginning of the 8% O2 administration and recovery in air. (A) Serial ATP expressed as a fraction of its initial prehypoxic (baseline) value. (B) Serial Pi/PCr measured from the same spectra as in (A). From two-way repeated measures ANOVA analysis, all three curves are statistically different from each other during the last 2 h of HI, when all of the data are modeled (cf. “Results”).

Full size image

Evaluation of these serial results (Fig. 3) from individual points does not properly account for the animal variability with time after the insult. To reflect the severity and duration of energy failure, the NMR data for each animal has been integrated during a time interval of HI, which maximizes the differences of the populations. Integration of the data also provides an enhanced signal-to-noise ratio. Table 2 summarizes this integrated assessment for the last 2 h of HI. With 6 °C of hypothermia, ATP is preserved by 73%, and Pi/PCr is reduced to 40% of its normothermic value (PCr/Pi is 150% greater). Both parameters of energy metabolism were also compared using a modeled two-way repeated measures ANOVA. In this type of analysis, it is important to model the autoregressive nature of the within animal variability. The analysis included the entire 6 h of data collection, but temperature effects were assessed only during the final 2 h of HI using linear contrasts. By including the complete data, a better estimate of variability is obtained. Pi/PCr was analyzed on a log scale. p values were multiplied by 6 to adjust for Bonferroni multiple comparisons. In this case, all three curves were significantly different for both ATP and Pi/PCr. For ATP, the two-tailed significance values were p < 0.0001 for 31 versus 37 °C, p = 0.0006 for 34 versus 37 °C, and p = 0.0012 for 31 versus 34 °C. For Pi/PCr, the values were p < 0.0001 for 31 versus 37 °C, p = 0.0006 for 34 versus 37 °C, and p = 0.038 for 31 versus 34 °C. When the serial PCr values were calculated without consideration of Pi, an excellent separation of curves, very similar in appearance (and statistical significance) to that for ATP, is apparent (Fig. 4). Note also that for PCr, as with ATP, the rate of recovery for the normothermic pups was much slower and less complete during the 2.5-h recovery period.

Figure 4
figure 4

Serial PCr values as a function of temperature, obtained as described in Fig. 3.

Full size image

Correlation with brain swelling. When individual integrated metabolite levels were compared with the corresponding tissue swelling (Fig. 5), there was a significant correlation of both ATP (r = -0.84, p < 0.0001) and Pi/PCr (r = 0.83, p < 0.0001). The Spearman rank correlation was r = -0.87 and 0.78, respectively. In addition, edema can be reliably predicted from a threshold level of these metabolite values (p < 0.0001, FE test and Wilcoxon-Mann-Whitney rank test for edema). For ATP, our chosen threshold for the normalized parameter was taken at 0.67 relative to baseline (Fig. 5A). One 34 °C point failed our criterion, and several values were borderline. For Pi/PCr (Fig. 5B), a threshold of 3.6, relative to an average baseline, separates all damaged from normal pups. Thus, the FE test will give the most significant results for separating damaged from normal pups based on a threshold model. The correlation tests and Wilcoxon-Mann-Whitney rank test, however, do not make any assumption of a metabolite threshold for their high significance. Also displayed in Figure 5 are our identically obtained 37 °C published results(12), which have now been converted from WCR to tissue swelling. These results are distributed similarly with the present data with 8/9 damaged pups having values clearly defined by the threshold. The remaining animal is borderline and indeterminate by our method. Not shown is the correlation of PCr by itself (r = -0.75, p < 0.0001). The distribution is similar to that of ATP, except that 3/12 normal pups gave low PCr results, which was not the case with Pi/PCr.

Figure 5
figure 5

Correlation of normalized (A) ATP and (B) Pi/PCr, calculated from spectra obtained during 1-3 h of HI, with cerebral tissue swelling, measured at 42 h recovery (p < 0.0001). Each symbol represents an individual rat pup at the insult temperature. Values to the left of the vertical dashed line are considered normal, which in all cases represent animals below the horizontal dashed line for Pi/PCr. All damaged pups (8/8 at 37 °C, 4/9 at 34 °C, and 1/8 at 31 °C) have values above this Pi/PCr threshold. Published 37 °C edema values (calculated from a previous study) are also displayed (×).

Full size image

For the fixed environmental temperature value of 34 °C, core temperature during HI (integrated during last 2 h) correlated poorly with subsequent brain swelling and was not predictive of animals that were undamaged (p = 0.40 for linear least squares and p > 0.50 for FE test). At 31 or 37 °C the pups are either all normal or undamaged, respectively, so that a good correlation would not be expected. However, at the environmental temperature of 34 °C, the NMR animals show a better association (Fig. 5) of metabolites with subsequent brain swelling (for Pi/PCr: p = 0.0026 for linear least squares fit and p = 0.0079 for FE test; for ATP: p = 0.048 for FE test).

DISCUSSION

This study demonstrates that 31P NMR can be used to reliably predict whether PND 7 rats subjected to HI will develop edema and brain damage over a range of environmental temperatures. Regardless of the temperature, a large change in integrated high energy metabolism during HI is a prerequisite for edema or brain damage(11–13). With moderate hypothermia of 6 °C, there is an insufficient change in ATP or Pi/PCr, and also minimal HI-induced tissue swelling. From Figure 5, it appears that the integrated value of ATP (and PCr/Pi) must fall below a threshold level before any damage will develop. This reservoir of substrate may represent a form of neuroprotection for the neonate. NMR, however, is less reliable for separating severe damage and infarction from lesser injury, and the swelling tends to maximize over a range of metabolite values. With these considerations, if tissue swelling is plotted as a function of Pi/PCr or ATP, a sigmoid or reverse sigmoid relation can be observed. The relation is more evident in our recent study of the effect of fasting on energy metabolism during HI(21, 29). This type of all-or-none mechanism has been found with other cooperative processes in biologic regulation.

The serial rectal temperature results from the jars and NMR chamber were very similar to that found by Yager et al.(8). In particular, we also observed a loss of thermoregulation during HI with a 6 °C drop in environmental temperature (poikilothermic behavior). However, the interesting result of significantly higher core temperature, and also edema, with up to five pups per jar requires further discussion. The neonates appear to adjust to their environment during hypoxia at a slower rate (maintain homeothermic state) inside the more crowded and warmer interior of the jar. The relationship between the rate constant of a reaction, the activation energy, and temperature is given by the Arrhenius equation(30). The Arrhenius dependence predicts a reaction rate decrease of 33% for a 6 °C drop in core temperature. Nevertheless, at the fixed environmental temperature of 34 °C, the poor correlation of core temperature, but not NMR, with damage is surprising. This result suggests that serial measurements of high energy metabolites are a more direct indicator of subsequent damage than is core temperature.

Yager et al.(8) performed very similar jar experiments in a circulating water bath at the same temperatures. They observed damage using gross neuropathologic grading in 90% of pups at 37 °C, 30% (6/21) at 34 °C, and none at 31 °C. Saeed et al.(9), following the same paradigm and experimental protocol as Vannucci and colleagues(8, 12, 14–17, 22), observed damage using TTC or death in 100% (22/22) of pups at 37 °C, and none (0/18) at 30 °C. We found edema in 100% (21/21) at 37 °C, 75% (15/20) at 34 °C, and 4% (1/24) at 31 °C. As expected, swelling was more severe at higher temperature. Edema from the NMR experiments was similar (Fig. 5) with 100% (9/9) at 37 °C, 40% (4/10) at 34 °C, and 11% (1/9) at 31 °C. Despite the similarity of the time course of rectal temperature for the NMR chamber and our jar experiments, there was more severe damage in the latter at 34 and 37 °C. This might be explained by the higher surface temperature due to increased thermal contact in the glass jars, at both the ventral as well as rostral (breathing) regions. The difference in damage in the two models might also be related to the extra freedom of mobility inside the jar as well as by the additional animal in each jar.

Yager et al.(8) also found an absence of a protective effect of 3.0 h of moderate hypothermia (31 °C) when initiated after 3.0 h of HI at 37 °C. From this, it would appear to be too late to rescue the pups once there is damage, although in one study, with the same paradigm, early rescue at 2.25 h with the drug allopurinol has been demonstrated(16). This supports our results which suggest that a sequence of events leading to early primary damage, is directly affected by the severity of high-energy metabolic changes during the insult. These events include a reduction of ATP-dependent processes, including a loss of the Na+, K+, Cl-, and Ca2+ gradients across the plasma membrane(14). The intracellular accumulation of Na+ and Cl- draws water into the cell from the extracellular space causing cytotoxic edema with cellular swelling and a narrowing of the extracellular space (decreased ADC). The opening of voltage-sensitive Ca2+ channels contributes to cell death by activating enzymes that disrupt the cellular structural integrity. A sufficient disruption in the blood-brain barrier during HI results in additional osmotic edema and increases with the length of insult(22). Our early diffusion-weighted images (Fig. 1B) appear to be quite sensitive to this apparent irreversible damage after long-term (3 h) HI(24). Rumpel et al.(25) using the same model, but after a shorter and harsher insult, did not observe a clear association of diffusion-weighted imaging with irreversible damage. Upon further recovery from HI, substantial breakdown of the blood-brain barrier leads to further swelling and brain damage, which can be measured by WC at 42 h or by T2-weighted imaging. Another indicator of damage during early resuscitation is the slow recovery of high energy phosphate reserves, necessary for endergonic reactions to resume(14). For example, in Figures 3A and 4, ATP and PCr clearly fail to recover after 2.5 h of air at the control temperature, but they do recover for modest hypothermia. Other possible contributions to this pathophysiologic cascade are from free radicals, produced during HI and recovery, which may induce membrane breakdown by attacking fatty acyl groups. Also, excitatory amino acids and neurotransmitters are released, leading to a further disruption in homeostasis(14). Hypothermia appears to modulate their initial release during ischemia(31) or else hypothermia may prolong the time until terminal depolarization(32). This offers alternative or additional mechanisms which merit further investigation with the present HI model. However, it has been shown that glutamate does not become neurotoxic until after intracellular energy levels are reduced(33). The correlation of swelling with integrated metabolite levels (Fig. 5), over variable duration and temperature of insult, provides strong evidence for the role of energy metabolism.

Using a histochemical luminescence technique for recording ATP levels in frozen brain sections, the distribution of ATP has been measured in a very similar PND 7 model of HI under normothermia(34). At 0 h post HI, patterns of severe ATP depletion were found throughout the cortex, striatum, hippocampus, and thalamus of the ipsilateral hemisphere. In particular, the depletion in the cerebral cortex resembles the patterns of histologic injury and often remains depleted at 2 h post HI(34). A few animals suffered less reduction of ATP with the decrease restricted to the cortex and striatum. These patterns are very similar to histologic results that we have obtained for mild and severe damage(11, 18) and helps to explain why surface coil NMR is adequate to detect even mild damage. Radiofrequency field simulations have shown that about half of the NMR signal is due to cortex metabolites, with remaining contributions as deep as the thalamus(11). 31P NMR is advantageous in that it allows one to quantify the in vivo serial changes of several metabolites during both HI and initial recovery and then make comparisons with brain damage measurements, all from the same animal.

Hemispheric WC measurement may be less sensitive than some methods when there are low levels of damage. However, it can be rapidly applied by us and has been extensively used in the past by our collaborators to determine HI injury. It has been shown that the magnitude and duration of edema formation during the recovery period is dependent upon the severity of tissue injury. In particular, the extent of infarction increases proportionately with severity of cerebral edema(22). Studies with the same paradigm have shown that WC measurements produce the same distributions of abnormal brains as does gross neuropathology, histology, and brain morphometry(16). In the present study, we have shown that 1H MR T2-weighted and diffusion-weighted images obtained at only 1-2 h of recovery from HI correspond spatially to the damaged areas routinely visible in subsequent TTC stains. Saeed et al.(9) have shown that TTC is a reliable method for the assessment of brain injury in this neonatal model. Because the damage is usually widespread, methods to obtain further metabolite spatial localization using spectroscopy would not be expected to provide much improvement upon the quantitation of damage. Also the signal/noise of the 31P spectra would be greatly reduced.

Measurement of high energy metabolite perturbations during HI has also been shown to be a reliable indicator as to whether neonatal pups will sustain edema(12, 13, 21, 29) or histologic and morphometric (atrophy) damage(11, 21, 29) during recovery. These additional studies include preinjection with drugs (allopurinol and deoxycoformycin), which reduce degradation of ATP, and administration of ketone body substrate (β-hydroxybutyrate) exogenously, or endogenously via fasting. In fact, by relating the measures of brain damage with our measure of metabolite levels, regions of edema have been shown to correspond to subsequent histologic damage and even more closely with brain atrophy(21). The measurement of apparent irreversible injury shortly after the beginning of recovery is consistent with the observation that HI damage in the immature rodent brain evolves more rapidly than its adult counterpart(17, 19–20). Thus, it was shown by Towfighi et al.(19) that very similar regions of cell injury are visible by histology at 5 min to 1 h as are present at 3 wk of recovery from HI. In particular, the vast majority of hippocampal neurons are irreversibly damaged by 2 h of recovery from HI, and only a small number of additional neurons became damaged within the next 22 h(19). These results are consistent with a substrate limited metabolism during neonatal HI(20). Most adult models are oxygen limited and are not reliable for predicting damage from measurement of high energy metabolites. For example, Busto et al.(35) using an adult four-vessel occlusion model of ischemia, measured severe depletion of brain ATP and phosphocreatine at brain temperatures of 30, 33, and 36 °C. Because the two lower temperatures conferred a marked protective effect without preserving energy metabolism, this hypothermia mechanism cannot be compared with our neonatal HI model. A recent study in young rats (PND 21) subjected to only 15 min of HI injury (also 8% oxygen) showed that prolonged hypothermia (22 °C environmental temperature for 72 h) during recovery was neuroprotective(36). Thoresen et al.(37) also showed that some PND 7 animals (litter placed in a 4-liter chamber) were rescued from more severe damage by providing 3 h of moderate hypothermia after 2 h of HI. Secondary energy impairment may be an important contributor in these investigations. In contrast, our PND 7 model of prolonged HI seems to produce primary damage at 37 °C that is developed mostly shortly after insult and thus may be considered to be irreversible(8, 9, 11–17, 21, 24, 28, 29).

Although extrapolation of results from the PND 7 rat to the preterm human infant may not be ideal, the present findings suggest that modest hypothermia could improve neurologic outcome from HI brain injury at birth. The lower rat core/brain temperature slows several key physiologic processes and sufficiently reduces energy consumption requirements. The clear correlation of the change in integrated high energy metabolites with subsequent brain damage provides evidence for the central role of ATP in the mechanism of HI injury. Measurement of these metabolite levels in the birth-asphyxiated patient has now been shown to correlate with outcome(38). These results support our findings of the prerequisite depression in cerebral high energy metabolism for HI damage. Future studies should be aimed at modifying our PND 7 model by using a less prolonged insult and extended hypothermia, while following spectroscopy and imaging parameters, and final histology.