Glutamate has been implicated in the pathophysiology of neuronal injury associated with cerebral hypoxia-ischemia. A model using chronic in utero microdialysis was developed to sample the extracellular space of the fetal brain. Using this model, we tested the hypothesis that glutamate efflux from the parasagittal parietal cortex of near-term fetuses would increase during maternal hemorrhage.
Twelve near-term fetal sheep were instrumented with vascular catheters, and a microdialysis probe(s) was implanted into the parasagittal parietal cortex. After a 3-day recovery period, the animals were subjected to maternal hemorrhage until either the fetal pH was < 7.00 or the fetus died. The extracellular glutamate concentration in the collected dialysate was determined by high pressure liquid chromatography (HPLC).
Maternal hemorrhage resulted in an 80-90% decrease in uterine blood flow, a decrease fetal po2, and a mixed metabolic and respiratory fetal acidosis. There were two groups of fetuses, survivors (n = 5) and nonsurvivors (n = 7). The nonsurvivor group showed a large increase (10-30-fold) in peak glutamate release (P = 0.0015). Survivors demonstrated a small (threefold) increase that was not statistically significant (P = 0.065), unless one animal with very low probe recovery was excluded (P = 0.0048).
Extracellular glutamate release from the fetal brain can occur during maternal hemorrhage with fetal acidemia. The pathophysiologic role (if any) of glutamate release in the survivors remains to be elucidated. Our results are consistent with the hypothesis that in utero release of glutamate occurs during periods of fetal asphyxia. This experimental preparation of chronic fetal brain microdialysis can be used to monitor the brain extracellular concentration of any dialyzable substance in response to stress, including maternal hemorrhage.
Key words: Anesthesia, obstetric; fetal asphyxia, Brain, fetus: excitatory amino acids; glutamate. Measurement techniques: microdialysis. Sheep.
THE release of the excitatory amino acid neurotransmitter glutamate can contribute to brain injury during cerebral ischemia. [1–4] It has been proposed that glutamate excitotoxicity may contribute to fetal hypoxic-ischemic brain injury. [5,6] Increased levels of glutamate have been detected in the cerebrospinal fluid (CSF) of asphyxiated human newborns.  Blockade of the N-methyl-D-aspartate (NMDA) subclass of glutamate receptors  as well as nonspecific glutamate receptor blockade  has been shown to protect neonatal brain from hypoxic-ischemic injury. These transmitter systems undergo extensive in utero development and change. [10,11] The normal function of these neurotransmitter systems is likely important in brain maturation and synaptogenesis. [12–14] An understanding of the normal development and response to asphyxia of these glutamatergic systems may help to explain the varied pathologic response of the fetal brain to hypoxic-ischemic insults.
One problem in studying the fetal brain's response to hypoxia is the difficulty in sampling from the fetal brain in utero. Most studies of biochemical and receptor changes of the immature brain to hypoxic-ischemic insult have been performed in ex utero animal pups (e.g., rats). [15–19] While useful, these studies are limited in that they do not account for the fact that the fetus exists in a relatively hypoxic environment in utero and is exposed to numerous maternal hormonal influences. Also, oxygen delivery to the fetus is profoundly influenced by changes in maternal cardiorespiratory status, uterine perfusion, and placental function. In addition, there are marked blood flow changes (e.g., increased pulmonary blood flow) that occur normally after birth.  It is difficult to accurately mimic the entire spectrum of blood flow changes occurring during in utero asphyxia in an ex utero model. The chronically catheterized fetal sheep model has been extremely useful in elucidating the fetal physiologic response to in utero insults including hypoxia-ischemia [21–23] and maternal hemorrhage. .
The technique of intracerebral microdialysis allows repeated sampling of the brain extracellular space in a minimally traumatic manner. Briefly, artificial cerebrospinal fluid (ACSF) is perfused via very fine catheters (diameter 0.5 mm or less) into and out of the area of the brain that is of interest. A portion of the tubing is replaced with a dialysis membrane that excludes molecules larger than a specified molecular weight (commonly 20,000 daltons). Substances, including most neurotransmitters, come to equilibrium with the ACSF and are collected in the effluent limb of the catheter. The collected dialysate can be analyzed for the presence of neurotransmitter substances such as glutamate.  Using this technique in an anesthetized and exteriorized fetal lamb, Hagberg et al.  demonstrated elevation of brain extracellular glutamate during periods of maternal aortic compression. With this in mind, we developed a chronic, unanesthetized ovine model to study the release of glutamate from the fetal brain in utero. To test the model, we studied the fetal response to asphyxia as produced by maternal hemorrhage. This is known to lead to decreased uterine blood flow, fetal acidemia and, if uncorrected, fetal death. .
Experimental Animal Preparation
All experiments were approved by the University of Iowa Animal Care Committee. Twelve mixed-breed pregnant ewes, obtained from a local supplier, were studied. Upon arrival, each animal received 2 ml procaine penicillin G (300,000 IU/ml) intramuscularly as preoperative infection prophylaxis. Animals were housed individually and allowed access to food and water ad libitum but were fasted for 24 h before surgery. Surgery was performed at 118–125 days of timed gestation (term = 147 days). The ewe was sedated with an intramuscular injection of ketamine (5 mg/kg). Intravenous access was obtained. Surgical anesthesia was induced with sodium thiopental (500 mg) and anesthesia maintained with 1–2% halothane in oxygen. The animals' lungs were mechanically ventilated throughout surgery. Maternal vascular catheters (polyethylene [PE]-240, Becton Dickinson, Parsippany, NJ) were inserted into the descending aorta (two) and inferior vena cava (one) via the left femoral artery, left mammary artery, and left mammary vein, respectively. Using the retroperitoneal route, the left uterine artery was isolated. An electromagnetic flow probe (Dienco, Los Angeles, CA) was placed around it for subsequent measurement of uterine blood flow (UBF). The uterus was exposed by a midline laparotomy. The lower half of the fetus was exteriorized through a small uterine incision. Fetal vascular catheters (PE-90) were inserted in both the right and left femoral artery and vein (total of four catheters). The tip of the catheter was positioned in the descending aorta or inferior vena cava. A fenestrated pressure monitoring catheter (MX566, Medex, Hilliard, OH) was secured to the fetal rump to serve as an amniotic pressure monitor. The fetus was returned to the uterus and the incision closed.
The fetal head was then exteriorized through a separate uterine incision. The head was secured in a custom-designed stereotaxic frame (David Kopf Instruments, Tujunga, CA). All microdialysis components, including the stereotaxic frame assembly, were cold gas-sterilized. Stereotaxic implantation of microdialysis probes was based on information from the atlas produced by Gluckman and Parsons  and modified as appropriate. Probes (CMA/12; 14 mm shaft length, membrane 2 mm in length and 0.5 mm in diameter, Bioanalytical Systems, West Lafayette, Indiana) were placed in the parasagittal parietal cortex through a hole drilled 5 mm left of the midline and 2 mm posterior to the coronal suture. The probe was placed through a guide cannula (CMA/12 guide cannula, Bioanalytical Systems). The final probe depth was approximately 12 mm below the dura. Immediately before implantation, the microdialysis probes were connected to the appropriate tubing (Bioanalytical Systems) and perfused with filter-sterilized and degassed artificial cerebrospinal fluid (ACSF) at 1 micro liter/min. The ends of the microdialysis probe tubing were heat-sealed. The millimolar (mM) composition of the ACSF was as follows: NaCl 134, KCl 2.0, MgSO41.3, KH2PO41.25, NaHCO317, CaCl22.5, and glucose 10. The microdialysis inlet and outlet tubing (Bioanalytical Systems) was protected from kinking by passing it through high-pressure tubing (Medex, Hilliard, OH) customized to accept the microdialysis probe. Dental pins (Coltene-Whaledent, New York, NY) were screwed into the skull in a pattern surrounding the microdialysis probe. The entire probe assembly was secured to the skull and to the pins with dental acrylic (L. D. Caulk, Milford, DE). The scalp was closed with sutures over the hardened cement, with the probe protruding through the incision. A plastic guard, formed from the cut-off top of a 125-ml screw-cap bottle with a hole drilled in the cap, was sutured over the entire scalp wound to further protect the microdialysis probe from twisting or breaking in utero. The placement and construction of the microdialysis assembly is detailed in Figure 1. The fetal head was returned to the uterus. The uterine incision was closed. Microdialysis probe tubing was exteriorized through a flank incision, the surgical wounds closed in layers, and the animal allowed to recover from anesthesia and surgery. The ewe received 80 mg gentamicin intravenously, and the fetus received 40 mg. The ewe received penicillin G (600,000 IU) intramuscularly every 48 h. Nalbuphine hydrochloride (5 mg) was injected intramuscularly as needed to control postoperative pain. All animals were allowed at least 3 days to recover from surgery before experimentation.
In several animals (Table 1), two microdialysis probes were implanted. This was to ensure that at least one probe would be patent at the time of experimentation given the fragile nature of the probes. In some animals (Table 1), the second probe was implanted into the cortex posteriorly using coordinates approximately 1.5 cm posterior and 2 mm medial to the anterior probe. The implantation depths were identical. Because of the possibility that the changes in glutamate might differ from anterior to posterior, it was decided in further experiments to symmetrically implant two probes in the anterior parasagittal cortex, one on either side of the midline. When two probes were used, the probes were encased in a single block of dental acrylic. In all other respects, the protocol was identical for animals with one or two probes.
Unbled Controls. On the third postoperative day, five animals were brought to the recording facility and the vascular catheters attached to appropriate transducers for recording of maternal and fetal blood pressure and maternal uterine blood flow. Blood pressure was corrected for intraamniotic pressure. All hemodynamic and blood flow data was averaged in 10-s epochs and stored onto computer disk for subsequent analysis. The inlet and outlet tubing for the microdialysis probe was unsealed and the inlet attached to an infusion pump (Harvard model 22, Natick, MA) and infused with ACSF at a rate of 1 micro liter/min. The slow infusion rate was necessitated by the long (900 mm) length of inlet tubing to minimize the back pressure on the dialysis membrane. The dialysis probe was perfused for 1 h before the beginning of the experiment (time = 0). Microdialysis samples were collected continuously at 28-min intervals (28 micro liter outflow tubing deadspace volume times 1 micro liter/min) and the maternal and fetal blood pressure, heart rate, and uterine blood flow were recorded simultaneously. After four basal collection periods, the ewe was infused (intravenously) with normal (0.9%) saline at a rate of 1.5 ml/kg/h. The saline infusion was continued for 11 28-min periods. This duration was chosen to correspond to the anticipated maximum length of the hemorrhage protocol. Maternal and fetal arterial blood samples were taken to determine pH, PO2, and PCO2at the beginning and end of the saline infusion and once during the infusion to confirm the health of the fetus. At the end of the saline infusion, the microdialysis infusion was stopped. The inlet and outlet tubing were resealed by melting the ends of the tubing. Vascular catheters were disconnected, and the animal was allowed to return to its pen before the next day's hemorrhage protocol.
Hemorrhage Protocol. On postoperative day 4, all animals, including the five unbled controls, were brought to the recording lab. Vascular catheters and microdialysis tubing were connected as described above. After a similar equilibration time and four baseline collections (28 min again), the hemorrhage protocol was begun. Maternal hemorrhage was performed through a maternal arterial catheter. Maternal blood (5 ml/kg) was removed over the first 10 min of the 28-min collection period. The blood was collected in a heparinized collection system for later return to the ewe. Fetal and maternal arterial blood gases were obtained as in the unbled controls, and additional samples were obtained near the end of each hemorrhage period. A further 5 ml/kg of maternal blood was removed during each subsequent 28-min interval until there was either fetal demise (i.e., asystole) or the fetal pH decreased to < 7.00, whichever came first. Because lower pH values are associated with very high fetal lethality, 7.00 was chosen as the target pH (personal observation). In situations in which the fetal pH decreased to < 7.00, the maternal blood was retransfused as quickly as possible and two additional microdialysis fractions were collected. Microdialysis samples were frozen for later amino acid determination. After termination of the protocol, the animals were euthanized by a barbiturate overdose. The microdialysis probe position was confirmed by infusing methylene blue under sufficient pressure to rupture the dialysis membrane. This confirmed probe-tip placement in the parasagitial cortex.
Chemicals and Solutions
All chemicals were at least reagent grade quality and were obtained from a variety of commercial suppliers. High pressure liquid chromatography (HPLC) grade sodium acetate and methanol were used for the HPLC mobile phase.
Blood gas analysis was performed using a blood analyzer (Instrumentation Laboratory, System 1302, Lexington, MA), and hemoglobin and hematocrit determinations were made with a hemoximeter (model OSM3, Radiometer America, Carrollton, TX). The glutamate concentration of each dialysis sample was measured using HPLC by a previously published technique. [28,29] The technique involved precolumn derivitization of glutamate with ophthaladehyde/beta-mercaptoethanol to form a thio-substituted isoindole derivative, which was separated using isocratic reversephase HPLC. Quantitation was by fluorescence detection and comparison of peak areas with external standards. The lower limit of sensitivity for glutamate was 9 pmol/ml of dialysate.
Where data are presented as percent of baseline, baseline was calculated as the mean value for the four intervals before saline infusion or hemorrhage. It was necessary to report percent of baseline values rather than actual values to use each animal as its own control. In the case of glutamate concentrations, this was necessary because variations in in vivo microdialysis probe recovery were not known. Glutamate efflux before and after hemorrhage was compared by calculating the ratios of peaks before and after hemorrhage. The means of the logarithms of the ratios were compared to zero by one-sample, one-sided t tests.
Microdialysis Success Rates
The success rate for intact probes (i.e., membrane unruptured) on the day of the experiment exceeded 80%. For sheep with two implanted probes, one or more probes were successfully perfused allowing continuation of the experiment. Although the absolute concentration of glutamate varied between probe pairs, the peak ratios were similar. Before dual implantation it was twice impossible to continue the experiment due to probe failure. Probe perfusion was never equivocal; that is, the amount of dialysate collected always equaled the amount perfused or was zero. All probes that were patent during unbled control experiment were later patent during hemorrhage experiments.
Hemodynamic and Biochemical Data in Unbled Controls
During maternal saline infusion (Table 2), there were no physiologically relevant changes in maternal or fetal hemodynamic and biochemical data over the study period. The time course was chosen to encompass the anticipated time course for the subsequent hemorrhage experiment to be performed on the same animals.
(Table 1) describes the fate for the individual fetal sheep. In six animals, fetal pH < 7.00 during the hemorrhage. Fetus 2 died a few minutes into the retransfusion phase and was included in the nonsurvivor group for analysis. Seven fetuses died during hemorrhage or immediately during the retransfusion period. These were subsequently designated the nonsurvivor group. The number of hemorrhage periods required to reach the predetermined endpoint varied considerably (four to ten periods). For a typical 70-kg pregnant sheep, this corresponded to 1,400–3,500 ml of shed blood. No apparent relationship was found between the number of hemorrhage periods and fetal survivor or nonsurvivor status (Table 1). Also, no relationship was found between prehemorrhage hemodynamic or biochemical data and fetal survivor or nonsurvivor status (not shown).
Hemodynamic and Biochemical Data: Hemorrhage
(Table 3) summarizes the hemodynamic and biochemical data for both fetus and ewe during the hemorrhage protocol. In those fetuses that did not die first, fetal pH decreased to 7.00 or less by the end of the hemorrhage. As expected, hemorrhage was associated with a marked decrease in uterine blood flow. As previously reported, [24,30,31] this degree of uterine blood flow reduction was associated with a decrease in fetal pO2and an increase in fetal PCO2.* To avoid biasing the data with extreme postmortem changes, for those fetuses that died during the experiment, blood gas values obtained for the interval immediately before death were used to calculate the end hemorrhage value. The end recovery data reflects only those fetuses that survived the retransfusion of hemorrhage blood and therefore include only five animals.
Mean basal (spontaneous) glutamate efflux (plus/minus SE) in dialysate samples was 330 plus/minus 95 ng/ml. There were no changes in glutamate efflux in microdialysate samples over the course of the experiments in the unbled controls (Figure 2). There were two patterns of glutamate efflux during maternal hemorrhage, that of survivors and that of nonsurvivors. In the survivor group, the glutamate efflux appeared to be pulsatile. Figure 3depicts a typical experiment for a surviving fetus. In that group, there was a peak in glutamate release sometime during the hemorrhage period, typically followed by a return to basal levels. The timing of the peak was not predictable. With the small group size, it was not possible to delineate a fetal pH, fetal pOsub 2, or UBF threshold for glutamate efflux. An approximately threefold increase in peak glutamate release (P = 0.065) was seen in four of five surviving fetuses (Figure 4). One fetus (3) failed to show any rise in peak glutamate levels during maternal hemorrhage. In this animal, actual glutamate concentrations were at or just above background in all samples. This may reflect a low relative recovery of glutamate, perhaps due to excessive debris around the probe. However, because of the small numbers of animals in the survivor group, we believed that excluding this fetus would not be warranted. Had we excluded this animal, there would have been a significant increase in peak glutamate efflux during hemorrhage in the survivor group (P = 0.0048).
A typical experiment from the nonsurvivor group is shown in Figure 5. Animals in the nonsurviving group were less likely to show significant peaks of glutamate release during the hemorrhage period, except close to or at the time of fetal death, when there was a dramatic increase in glutamate efflux (P = 0.0015;Figure 4).
Fetal Microdialysis Model
This is the first report of the chronic in utero implantation of a microdialysis probe into the fetal brain. In control experiments, we have demonstrated detectable, stable levels of basal glutamate efflux over a 7-h experiment 3 days after probe implantation. The acute effects of surgery and probe implantation are minimized by allowing 3 days to intervene as a recovery period. In rats, it has been demonstrated that reliable results can be obtained up to 7 days after probe implantation.  There is some concern that chronic probe implantation for this length of time will lead to fibrin and glial buildup around the probe.  This potentially would limit the transfer of substances (including glutamate) to the dialysis perfusate. It has been suggested that the appearance of glutamate in dialysate after ischemia may be delayed in chronically versus acutely implanted microdialysis probes.** It is possible that a failure to detect appreciable elevations in glutamate may be due to diminished recovery, secondary to the long probe implantation. A limitation of microdialysis in general, and chronic implantation particularly, is the possibility that high or toxic concentrations of glutamate are present in the synaptic cleft but are not detected in the dialysate due to diffusion barriers and/or active reuptake by glia or neurons. Although we do not wish to ignore this possibility, it is reassuring that detectable levels of glutamate were present under basal conditions. Also, there was an obvious relationship between peak glutamate levels and survival versus nonsurvival, thus preserving the expected relationship between high glutamate levels and hypoxia-ischemia. Finally, the negative effect of prolonged implantation on substrate recovery is minimized using the low flow rate of 1 micro liter/min. At slow flow rates, the ACSF transit time allows near-equilibrium of solute between the brain's extracellular compartment and the dialysate. It has been demonstrated that, at such low flow rates, recovery is near-maximal. .
The parasagittal parietal cortex was chosen as the site of probe implantation. This has been shown to be an area of neuronal injury in term humans  and in fetal lambs.  The hippocampus and basal ganglia are other areas demonstrating neuronal injury.  They have not been studied because available microdialysis probes are not long enough to reach these areas. Experiments on these brain areas are currently underway using custom designed probes (CMA, Sweden).
Additional limitations with the technique of in utero microdialysis include the high cost of this complex preparation. Unlike many other microdialysis applications, our probes were not reusable because of the extensive cement surrounding each probe. It is likely that this technique will be limited to research questions that are not answerable in smaller animals. Also, the necessity of relatively long (28 min) collection intervals can limit the interpretation of glutamate efflux data by making it difficult to temporally relate changes in glutamate efflux with shorter-term changes in hemodynamic and/or metabolic status. Because of the sensitivity of the glutamate analysis, it will be possible in future experiments to use smaller aliquots for assays, thereby decreasing the fraction collection time and increasing the time resolution of the method.
It is not clear why some animals were able to survive the hemorrhage insult better than others. We could find no relationship between initial fetal or maternal blood gases or hemodynamic data and subsequent group outcome. That is, the nonsurvivors did not start out less “healthy” than the survivors. The UBF decrease was comparable in both groups. Some animals died before reaching a pH < 7.00. This may be because of the time resolution of our method. We were limited to 28-min collection intervals because of the dead space volume of the microdialysis tubing and the low ACSF flow rates we required. It is possible that, had shorter collection periods been used or more frequent fetal blood gas samples taken, we would have detected a fetal pH < 7.0 sooner. If detected, it may have been possible to intervene in time to prevent fetal death.
In four of five fetuses in the survivor group, there was pulsatile release of glutamate during the hemorrhage period. In one fetus, no release was seen, but in that animal, basal levels of glutamate were barely detectable, suggesting that there may have been a problem with recovery in that probe. The increase fell just short of statistical significance, (P = 0.065; excluding the animal with questionable recovery allowed a high level of statistical significance, P = 0.0048). Nevertheless, we argue that the threefold increase likely underestimated the concentration of glutamate at the synaptic cleft. Therefore, it is possible that these concentrations may represent excitotoxic levels of glutamate. Studies are being initiated to assess histologic outcome associated with this magnitude of glutamate release in surviving fetuses.
The hypoxic-ischemic threshold for glutamate has not been determined in fetal lambs. In an acute insult such as severe maternal hemorrhage, it is likely that the release threshold is very near that which will lead to fetal death. The massive release seen in the nonsurvivor group is likely a result of terminal anoxic depolarization. In subthreshold situations, such as may have been responsible for the pulsatile glutamate release in the survivor group, it is likely that reuptake mechanisms are sufficiently robust to remain functional in this degree of hypoxemia and acidosis. This reuptake would limit the amount of glutamate available to be dialyzed. There is evidence that, in rat astrocyte cultures, hypoxia alone is insufficient to inhibit glutamate reuptake. Glucose depletion is required in addition to hypoxia, suggesting that ischemia rather than hypoxia is necessary to inhibit astrocyte reuptake of glutamate.  This is relevant, because survivable in utero asphyxia is likely hypoxic rather than ischemic in nature. .
Fetal brain microdialysis in utero is a powerful new tool for assessing the response of the fetal brain to a wide variety of external influences. We have shown that the probes are well tolerated by the fetus and that there is a reasonable probe patency rate 3–4 days after implantation. Using this technique, we have demonstrated in utero release of glutamate in the fetal cerebral cortex during hypoxia-ischemia. Maternal hemorrhage leads to dramatic glutamate efflux from the fetal brain in nonsurviving fetuses (P = 0.0015). Surviving fetuses have smaller increases in glutamate release from the fetal brain (P = 0.065). This may be due to failure of a sufficient number of neurons to achieve threshold for release. It is also possible that impaired probe recovery, active reuptake of glutamate by neurons or glia, or some combination of these factors limited the magnitude of glutamate efflux. Nevertheless, the several-fold increases in glutamate efflux seen in the surviving fetuses may represent high or toxic amounts of glutamate at the synaptic cleft. This animal model can be used to explore the mechanisms of in utero excitotoxicity and its possible relation to fetal neurologic injury.
The authors thank James F. Brien, Ph.D., and Michael M. Todd, for their advice and encouragement in this research.
*The fetal and maternal hemodynamic and biochemical changes associated with maternal hemorrhage are well known and anticipated. We did not statistically test these changes, to preserve statistical power to analyze the glutamate efflux data.
**Grabb MC, Sciotti VM, Van Wylen DGL: Changes in dialysate amino acids and purines during cerebral ischemia: Chronically versus acutely implanted microdialysis probes (abstract). Society for Neuroscience 19:1661, 1993.