Main

Studies of infants with severe birth asphyxia using 31P MRS have shown changes indicating reduced cerebral oxidative phosphorylation, which were frequently not apparent during the 1st d of life, but became most pronounced 2-4 d after birth(1). This impairment in cerebral energy metabolism has been called delayed or “secondary” energy failure because it was believed to have been initiated by an acute episode of hypoxia-ischemia causing “primary” energy failure during labor, which had been reversed by resuscitation at delivery. The extent of delayed energy failure was found to be directly related to the severity of subsequent neurodevelopmental impairment both at 1(1, 2) and 4(3) y of age.

The hypothesis that delayed cerebral energy failure could be initiated by an acute reversed episode of primary energy failure, designed to resemble that occurring during birth asphyxia, has been confirmed by 31P MRS studies in the newborn piglet(4) and rat pup(5, 6). In the piglet, cerebral [PCr]/[Pi],[NTP] (which is predominantly ATP)(7) and pHi fell during the acute insult, but returned to normal on resuscitation.[PCr]/[Pi] and [NTP] then fell again over the next 48 h, but pHi remained normal(4). These changes closely resembled those seen in the human infant(1).

When oxidative phosphorylation is impaired, increased glycolysis with the production of lactic acid is to be expected. 1H MRS allows the noninvasive observation of cerebral Lac and other metabolites, such as Naa, which is found particularly in neurones(8, 9). Studies of the brain in birth-asphyxiated newborn infants have shown changes in spectral peak-area ratios, indicating a rise in Lac(10–12), and a fall in Naa(10, 13), and these changes have been associated with an adverse outcome(10–13).

It has not yet been shown that the changes in 1H metabolite peak-area ratios seen in human infants can be provoked by acute reversed cerebral energy failure, and the time course of these changes has not yet been established. The raised cerebral Lac could be due to failure of eradication after the primary insult or later production associated with the progression of delayed energy failure.

To clarify these issues, continuous observations by interleaved 1H and 31P MRS were performed during and for 48 h after transient cerebral hypoxia-ischemia in the newborn piglet. We aimed to test two hypotheses:1) that changes in 1H metabolite peak-area ratios seen in birth-asphyxiated infants, indicating a rise in cerebral Lac and a fall in Naa, can be reproduced in the newborn piglet, and their time course determined; and 2) that changes in the Lac peak-area ratios are related to changes in the phosphorylation potential during primary and secondary cerebral energy failure as determined by 31P MRS.

METHODS

The study was carried out with Home Office approval and according to UK guidelines.

Subjects. Eighteen healthy Large White piglets weighing 1.62(0.23) kg (mean (SD)) were studied. They were born at term (gestation 115(2) d) and aged <24 h. Full details of the maintenance and monitoring of the piglets have been given previously(4). Briefly, after sedation with intramuscular midazolam(0.2 mg·kg-1), anesthesia was induced with 5% isoflurane, followed by ventilation with nitrous oxide, oxygen, and isoflurane (<1.5%) via an endotracheal tube, which was later replaced with a tracheostomy. Umbilical arterial and venous catheters were sited, and inflatable occluders positioned around both common carotid arteries. The anesthetized, ventilated piglet was then placed on a temperature-regulated water-filled mattress inside the bore of the magnet, and maintained there with monitoring of physiologic variables and full intensive care for the duration of the experiment.

MRS. 1H and 31P spectra were collected alternately using a 7-tesla Bruker Biospec spectrometer (21 cm bore;31 P frequency 121 MHzH frequency 300 MHz) and a 25-mm diameter double-tuned, inductively coupled surface coil placed on the intact scalp over the parietal lobes.

1H spectra were obtained using a 1[horizontal bar over]1-2[horizontal bar over]2 binomial water suppression spin-echo sequence(14) optimized for Naa with echo time 270 ms, spectral width 4 kHz, 2048 quadrature data points, and 32 summed echoes sampled with a 16-bit analogto-digital convertor. EXORCYCLE(15) phase cycling was used. The TR was 5 s for baseline measurements, 1 s during and immediately after the acute hypoxic-ischemic insult (for rapidity of acquisition), and 5 s thereafter and in controls, making the duration of the collections 2.8, 0.7, and 2.8 min respectively. 1H spectra were analyzed first by baseline correcting and zero-filling each echo to 2048 points, followed by application of a convolution difference technique(16) to reduce spectrum baseline roll resulting from the unsuppressed wings of the water resonance. The preprocessed echoes were then Fourier transformed, and the spectra manually phased (zero and first order): a large first order phase adjustment was necessary to correct for the linear, chemical shift-dependent phase characteristic of the pulse sequence. A cubic spline baseline was fitted to the spectra in regions where either resonances were not expected or resonances were severely suppressed by the pulse sequence. The cubic spline baseline was then subtracted. Resonances were identified according to previously published assignments(17, 18). Lorentzian peaks were fitted to Cho, Cr, Naa, Lac, and other minor resonances using χ2 minimization with prior knowledge consisting of starting values for chemical shift and peak width(19). The Lac methyl resonance was fitted as a doublet, whereas Cho, Cr, and Naa were fitted as singlets.

For 31P, a pulse-acquire sequence (180 ° coil-centre flip angle) was used with a prior DANTE(20) sequence (500 10-μs pulses; 200-μs TR) for suppression of bone and phospholipid signals. The acquisition conditions were: TR 10 s, spectral width 14 kHz, 2048 quadrature data points, and 192 averaged FIDs for the baseline spectrum, 36 FIDs during and immediately after the acute insult, and 384 FIDs thereafter and in controls. (Collection times were thus 32, 6, and 64 min, respectively.) Full details of the 31P analysis methods have been given previously(4). NTP was quantified using the β-triplet as it contained no contributions from other metabolites, and [NTP] was expressed as a fraction of the [EPP], which consisted of the main phosphorylated metabolites involved in energy metabolism and was quantified as [Pi] +[PCr] + [(γ + α + β) NTP]. pHi was estimated from the chemical shift of Pi relative to PCr using the Henderson-Hasselbach relationship: pHi = 6.77 + log10 [(δPi - 3.29)/(5.68 - δPi)](21).

Acute cerebral hypoxia-ischemia. After baseline observations, 12 of the 18 piglets underwent a cerebral hypoxic-ischemic insult produced by reducing the inspired oxygen fraction to 0.12 and inflating the carotid artery occluders. This was continued until severe cerebral energy depletion was seen on the phosphorus spectra, which were plotted automatically during the acute insult. When PCr was either completely or almost totally depleted and NTP had fallen below about a third of its baseline level, resuscitation was commenced by releasing the carotid occlusion and increasing the inspired oxygen fraction transiently to 0.60 followed by reduction to normalize arterial Po2. The duration of the insult was 34-88 (median 47) min.

Observations were continued for 48 h after cerebral resuscitation. During this period, six piglets received between 5 and 40 (median 10) mL·kg-1 colloid (Gelofusine, B Braun Medical Ltd, Emmenbrücke, Switzerland) with the aim of maintaining blood volume and reducing tachycardia (a heart rate consistently above ≈230 beats·min-1). At the end of the 48 h the animals were killed by anesthetic overdose, except for one animal that died 42 h after cerebral resuscitation with evidence of catastrophic brain injury (absence of PCr and NTP for >24 h before death). During the 30 min before death this animal had been treated with an infusion of adrenaline; none of the other piglets received inotropic drugs, as MABP remained above 5.3 kPa.

Controls. Six piglets served as sham-operated controls, undergoing the same surgical procedure as described above, and maintained within the magnet for 48 h, but without being subjected to cerebral hypoxia-ischemia.

Data analysis. Data were examined for normality and equality of variance. Statistically significant differences from baseline values, and between the hypoxic-ischemic and control groups, were sought using parametric(paired and unpaired t tests) or nonparametric (Wilcoxon rank sum and Mann-Whitney tests) methods as appropriate. Where multiple tests were performed, only values of p < 0.01 were considered significant. Linear regression was used to investigate relationships between changes in1 H and 31P metabolites.

The severity of the hypoxic-ischemic insult in individual animals was assessed using 31P MRS, as described previously(4), by calculating the time integral of depletion of[NTP]/[EPP] from the onset of hypoxia-ischemia until 1 h after resuscitation. The greater this value, the greater was the magnitude and duration of NTP depletion, and the greater the severity of the primary insult.

RESULTS

Physiologic variables. Values for arterial pH, Po2, Pco2, and base excess, and MABP, heart rate, blood glucose, and rectal temperature are given in Table 1a for the animals subjected to hypoxia-ischemia, and in Table 1b for the control group. In four animals in the hypoxic-ischemic group, tympanic temperature was measured as an estimation of brain temperature(22). During the acute insult, the arterial Po2 fell to a mean of 3.3 kPa, and the arterial base excess fell to a mean of -3.6 mmol·L-1. There was a rise in heart rate and fall in MABP. After resuscitation and for the duration of the experiment, physiologic variables remained stable, although the arterial base deficit had fallen significantly from baseline by 48 h in the hypoxic-ischemic group. Blood Lac measurements were made in three piglets in the hypoxic-ischemic group. Values remained≤2 mmol·L-1 throughout the experiment.

Table 1 Physiological variables in the (a) hypoxic-ischemic and (b) control groups
Full size table

1H MRS. A typical sequence of spectra from an animal subjected to hypoxia-ischemia is illustrated in Figure 1. An acute rise in Lac relative to the other peaks was observed during the insult which fell toward baseline on resuscitation. A second rise in the Lac peak was then observed over the subsequent 48 h. By 48 h, the Naa peak appeared broader but reduced in height relative to the other major peaks, and resonances in the region of 2.3-2.7 ppm, due largely to glutamate/glutamine, were more prominent. Alanine was also detected.

Figure 1
figure 1

Sequence of spectra from an animal subjected to acute hypoxia-ischemia. The spectrum illustrated during the insult was obtained immediately before resuscitation, and subsequent times were from resuscitation. Resonance identifications: 1, glutamate/glutamine;2, glycine/myo-inositol; 3, taurine/scyllo-inositol; 4, Cho; 5, Cr;6, glutamate/glutamine; 7, Naa; 8, Lac;9, β-hydroxybutyric acid; 10, alanine.

Full size image

Table 2a gives the values for changes in Lac relative to Naa, Cho, and Cr for animals subjected to hypoxia-ischemia, with results from the control group in Table 2b. Lactate peak-area ratios rose acutely during the insult and fell sharply after reoxygenation and reperfusion, but for the hypoxic-ischemic group as a whole, they remained significantly elevated above both baseline and control values at all stages after resuscitation.

Table 2 Lactate peak-area ratios in the (a) hypoxic-ischemic group and (b) control groups
Full size table

Table 3 gives results for the peak-area ratios Naa/Cho, Naa/Cr, and Cr/Cho. During the acute insult, both Naa/Cho and Naa/Cr fell significantly in comparison with baseline values. Subsequently, for the hypoxic-ischemic group, Naa/Cho values did not differ significantly from the controls, although they were below baseline at 48 h. Naa/Cr was below baseline values at both 24 and 48 h, and at 48 h the results were also significantly below those of the control animals. Naa/Cho and Naa/Cr values obtained from the piglet subjected to hypoxia-ischemia that died early with severe cerebral energy failure were consistently more than 3 SDs below the mean of the control group during the 24 h before death. Cr/Cho for the hypoxic-ischemic group was significantly above baseline both during the acute insult and at 24 and 48 h after resuscitation.

Table 3 Naa, Cho, and Cr peak-area ratios in the (a) hypoxic-ischemic and (b) control groups
Full size table

Figure 2 shows data from an individual piglet subjected to a hypoxic-ischemic insult, together with means and 95% confidence intervals of control data. The etiology of the changes in the ratio of Lac/Naa over time, illustrated in Figure 2A, is further defined by plotting Lac and Naa separately over the same denominator (Cr).

Figure 2
figure 2

(A) Lac/Naa, Lac/Cr, and Naa/Cr, and(B) [PCr]/[Pi], [NTP]/[EPP], and pHi, from an individual piglet subjected to hypoxia-ischemia (•). Blood Lac measurements were 1.3, 1.6, 0.9, 0.9, and 1.1 mmol·L-1 at baseline, during the insult, and at 2, 24, and 48 h after resuscitation. The mean control values are shown (○) with 95% confidence intervals.

Full size image

Relation between 1H and31P metabolites. The degree of rise of Lac peak-area ratios varied substantially in different piglets. This is largely explained by varying degrees of severity of the acute hypoxic-ischemic insult, as assessed by the time integral of depletion of [NTP]/[EPP](4). The time integral of primary [NTP]/[EPP] depletion was positively related to the maximum Lac/Naa observed during delayed cerebral energy failure 24-48 h after resuscitation (linear regression, r = 0.66, p = 0.02), and also to the maximum Lac/Cr (r = 0.62,p = 0.03). Thus the more severe the primary insult, the greater the relative rise in the Lac signal during secondary energy failure.

The time course in a single piglet of changes in [PCr]/[Pi],[NTP]/[EPP] and cerebral pHi, as well as in the 1H metabolites, is illustrated in Figure 2. The timing of the changes observed by 31P MRS mirrored those seen with 1H MRS, except that, although a severe intracellular acidosis developed during the acute insult, pHi subsequently remained normal despite the secondary rise in the Lac peak-area ratios. Values for [PCr]/[Pi] and pHi from both groups of animals are given in Table 4. There was no significant difference in [PCr]/[Pi] from baseline or control values until 24 h after resuscitation.

Table 4 [PCr]/[Pi] and pHiin the (a) hypoxic-ischemic and (b) control groups
Full size table

Figure 3 shows the relation during delayed energy failure between the maximum Lac/Naa value and (A) the minimum[PCr]/[Pi], and (B) the minimum [NTP]/[EPP]. The maximum Lac/Naa rose steeply when minimum [PCr]/[Pi] was low; however, maximum Lac/Naa rose steadily as minimum [NTP]/[EPP] fell (r = -0.94,p < 0.0001). Maximum Lac/Cr and Lac/Cho were also closely related to minimum [NTP]/[EPP] (r = -0.90, p = 0.0001 for both).

Figure 3
figure 3

Relation between the maximum Lac/Naa value and(A) the minimum [PCr]/[Pi] and (B) the minimum[NTP]/[EPP] value observed between 24 and 48 h after resuscitation in the hypoxic-ischemic (•) and the control (○) piglets. The regression line(r = -0.94, p < 0.0001) for the hypoxic-ischemic piglets is shown in B.

Full size image

DISCUSSION

This study confirmed our hypothesis that changes in 1H metabolite peak-area ratios previously reported in birth-asphyxiated human infants(10–12) could be provoked in the piglet by an episode of transient cerebral hypoxia-ischemia. It is highly likely that the changes observed in the human infant were initiated in the same way, by a severe transient episode (or episodes) of perinatal cerebral hypoxia-ischemia. Using 1H MRS to study infants with evidence of perinatal hypoxia-ischemia aged 2-160 (median 42) h, we demonstrated a significantly elevated Lac peak relative to Naa, Cho, and Cr in comparison with normal infants(12). The data given in Figures 2 and 3, and Tables 2 and 3, show that closely similar changes took place in the piglets with increases in Lac relative to Cho, Cr, and Naa, above both control and baseline values.

The time course of changes of cerebral Lac peak-area ratios in asphyxiated human infants has not yet been established, but it is likely that it will closely resemble that found in the present study. A clear biphasic response has been demonstrated. During the acute cerebral hypoxic-ischemic insult, there was a steep rise in Lac relative to the other metabolites which was associated with a profound fall in cerebral pHi(Figs. 2 and 3; Tables 2 and 4), presumably largely due to the production of lactic acid in cerebral tissue. Immediately after reoxygenation and reperfusion, the Lac ratios fell toward baseline values, and pHi returned to normal. This suggests the removal of lactic acid from the brain either by local metabolism(23), or by transport(24). After an interval of some 12-24 h, the Lac ratios rose again, despite the maintenance of normal values for arterial Po2, MABP, and blood glucose. Values for blood Lac were ≤2 mmol·L-1 throughout the experiment in the three piglets in which the observations were made. We did not measure absolute concentrations of cerebral Lac, but preliminary results from quantitative studies of birth-asphyxiated infants indicate cerebral Lac concentrations greater than those in blood(25). Thus, the elevated Lac seen during delayed energy failure is principally due to renewed production of Lac in the brain tissue. Astrocytes appear to produce far greater amounts of Lac than neurones, but Lac moves freely across the glial membrane(26). The protons associated with this delayed production of Lac must have been buffered or removed from the brain as, in contrast to the elevation of Lac during the acute insult, cerebral pHi remained normal. A similar dissociation of cerebral Lac and acidosis has been observed in other animal models of brain damage(27), and may be due to the operation of proton extrusion mechanisms other than Lac transport, such as Na+/H+ exchange(24). These results are also consistent with observations in asphyxiated human infants, where a tendency for pHi to rise rather than fall during delayed energy failure was noted(1). The combination of elevated cerebral Lac and intracellular alkalosis has also been found during chronic cerebral infarction in adult stroke patients(28).

During both primary and secondary energy failure, the changes in the Lac peak-area ratios closely mirrored those seen with 31P MRS (Fig. 3). As [PCr]/[Pi] and [NTP]/[EPP] fell, so the Lac ratios rose. In primary energy failure, the increased Lac ratios are attributable to reduced oxygen supply resulting in decreased oxidative phosphorylation, causing enhanced glycolysis via the Pasteur effect. Reduced oxygen supply, however, is unlikely to account for the increased Lac during secondary energy failure. Cerebral oxygenation during this period is elevated, rather than reduced, as demonstrated by near infrared spectroscopy in birth-asphyxiated human infants(29, 30) and in this piglet model (our unpublished observations). Explanations for the delayed rise in cerebral Lac include inhibition of mitochondrial respiration or irreversible mitochondrial damage. Mitochondrial respiration may be inhibited by the increased synthesis of nitric oxide(31, 32), and mitochondrial damage may be caused by free radicals and calcium overload(33). Increased Lac ratios correlate closely with falling [NTP]/[EPP] (Fig. 4), and a possible explanation for this is that a fall in ATP concentration is required before the glycolytic enzymes phosphofructokinase and pyruvate kinase are significantly activated. Another contributory factor to the delayed rise in Lac could be the removal of dead or dying cells by phagocytosis. The activation of human monocytes and macrophages enhances glycolysis leading to increased Lac production(34).

The mechanisms accounting for the progression from primary to secondary energy failure, with its associated neuronal loss(35), are still unclear. The release into the synaptic clefts of excitatory neurotransmitters, particularly glutamate, in response to hypoxia-ischemia is likely to be important. Glutamate can be detected by 1H MRS, especially in the regions between 2.3 and 2.6 ppm and at ≈3.75 ppm(17, 18). In the present studies, there were relatively minor changes between 2.3 and 2.6 ppm during the acute insult (Fig. 1), but during secondary energy failure, there was a marked increase in the amplitudes of signals in this region, relative to other major peaks. However, overlapping multiplet resonances from glutamate and other amino acids such as glutamine, Naa, and γ-aminobutyrate cause difficulties in the assignment of resonances and accurate measurement of peak areas in this region of the spectrum. Also, in severe secondary energy failure, as Naa falls, a shoulder appears on the left-hand side of the“Naa” peak (final spectrum, Fig. 1) which is probably related to further glutamate, glutamine, and γ-aminobutyrate resonances at 2.1-2.3 ppm(18). We have therefore not attempted detailed analysis of changes in glutamate, pending further studies of the assignment of resonances in this region of the spectrum using the linear combination of model spectra method(36).

Changes in the Naa, Cho, and Cr peak-area ratios occurred during the acute hypoxic-ischemic insult. Peak-area ratios are affected by metabolite T2 values as well as concentrations. The fall in Naa/Cr and rise in Cr/Cho were probably due to an increase in the T2 of Cr, which has been found in this piglet model during acute hypoxia-ischemia(37). The fall in Naa/Cho is less easy to explain, as Naa and Cho T2 values did not appear to alter during the insult(37). After resuscitation, these ratios reverted to normal, with no significant differences from baseline or control values. Later changes in Naa peak-area ratios resembled those previously found in human infants. For example, other authors have reported reductions in Naa/Cho in birth-asphyxiated infants at mean ages of 7(10) and 14(13) days, but in our own studies of birth asphyxiated infants examined earlier (median age 1.8 d), and for whom control data were available, there was no reduction in Naa/Cho, but there was a trend toward a reduction in Naa/Cr(12). Hanrahan et al.(11) studying infants within the first 18 h after birth-asphyxia, did not find a reduction in Naa/Cr. The data from the present study show a significant reduction in Naa/Cr by 24 h after the insult and a major reduction by 48 h (Table 3). Naa/Cho was not significantly reduced at 24 h and only modestly so, against baseline but not control values, at 48 h. The Naa T2, in particular, has been shown to increase with the progression of secondary energy failure in this model(37), enlarging the Naa signal and potentially masking a fall in [Naa]. Therefore, the falls in Naa/Cr and Naa/Cho found here may not accurately reflect the extent of the reduction in [Naa], which may also be the case in birth-asphyxiated infants studied acutely using peak-area ratios. Naa is found predominantly in neurones, but also, of importance in the neonatal brain, in immature oligodendrocytes and oligodendrocyte type 2 astrocyte precursors(8). Falls in Naa peak-area ratios may therefore reflect loss of both neurones and differentiating glial cells after acute hypoxia-ischemia. Naa/Cr appears to be a more sensitive early indicator than Naa/Cho of this loss. The increase in Cr/Cho over the 48 h after resuscitation (although T2 changes may contribute) is likely to be explained by a relatively greater fall in [Cho] after hypoxia-ischemia, as a rise in [Cr] is unlikely. Preliminary results of quantitation of cerebral metabolites in babies support significant falls in[Cho] as well as [Naa] after perinatal hypoxia-ischemia(25).

The results of this investigation have further implications for the interpretation of data from birth-asphyxiated human infants. As shown previously(4), and in the present study,[PCr]/[Pi] returned to normal after the acute cerebral insult and fell again only 12-24 h later, resembling the changes seen in infants. The extent of the fall in [PCr]/[Pi] in infants is strongly related to prognosis(1–3), hence the ability to predict this fall could facilitate controlled studies of cerebroprotective strategies designed to reduce the chances of neurodevelopmental impairment or death. The fact that the Lac ratios remained significantly raised after resuscitation at a time when [PCr]/[Pi] was still normal suggests that they may be a useful marker predicting the development of secondary energy failure. Some indication that this is also the case in human infants has been found by 1H MRS studies of birth-asphyxiated neonates on the 1 st d of life(11, 12), a time when 31P MRS is characteristically normal(1). Further studies are required in asphyxiated infants focusing on the first few hours after resuscitation, when intervention is likely to be most effective. The close relation of indices for assessing the severity of delayed energy failure measured by 1H and 31P MRS (Fig. 4) is also of potential clinical importance, where either study time or technical factors may prevent both studies from being carried out.

We conclude that: 1) changes in 1H metabolite peak-area ratios closely resembling those seen in birth-asphyxiated human infants have been reproduced in the newborn piglet after an acute reversed hypoxic-ischemic cerebral insult, and the time course of these changes has been determined;2) increases in the Lac peak-area ratios were related to falls in[PCr]/[Pi] and [NTP]/[EPP] during both primary and secondary energy failure. However, the Lac peak-area ratios remained significantly elevated in the interval between primary and secondary energy failure, when[PCr]/[Pi] and [NTP]/[EPP] were normal.