Main

Brain injury is a frequently encountered problem in neonatology and one for which no effective therapy exists (1–3). Asphyxia, IVH, and meningitis are significant causes of perinatal brain injury, and many such injuries result in lifelong disability or death (4). Several recent advances have delineated some of the cellular mechanisms involved in compensating for brain injury (5–7). These advances have created opportunities for designing rational new neuroprotective treatments. Among these advances is the discovery of the Epo system within the brain (7–9). Experiments in animals and cultured human cells indicate that during hypoxia, Epo production in the brain increases (10–12) and that Epo binds to specific Epo receptors on neurons (and other neural elements) where it exerts neuroprotective activity (7–9).

Although it is reasonably clear that Epo is a natural neuroprotectant from hypoxia, it is not clear whether administration of rEpo, either alone or as part of a “neuroprotective cocktail,” would diminish brain injury after perinatal asphyxia, IVH, or meningitis. Moreover, before this concept can be tested, two basic issues must be resolved. First, it is not known whether human neonates who have brain injury respond to such injury by elevating the Epo concentration in their brain (or in their CSF). Such an elevation would be consistent with the animal and in vitro evidence that Epo is a natural neuroprotectant in neonates. Second, it is not known whether, in human neonates, Epo crosses the blood-brain barrier. On the basis of its size (30.4 kD) and glycoprotein structure, passage would not be predicted. However, disruption of the barrier can occur in asphyxia, IVH, or meningitis, and passage may indeed occur under those conditions, thus providing a means by which rEpo given intravenously could reach the Epo receptor on central neurons.

Thus, the present study was performed to test the following two hypotheses:1) Epo concentrations in CSF are higher among neonates with brain injury (asphyxia, IVH, or meningitis) than in those without such injury;2) in neonates who have no evidence of brain injury, Epo does not cross the blood-brain barrier, but, in neonates with asphyxia, IVH, or meningitis, Epo does.

METHODS

CSF specimens.

One hundred twenty-two paired CSF and plasma samples were obtained from 114 neonates and children. All samples were obtained from the clinical laboratory of Shands Teaching Hospital after patients had undergone spinal taps that were clinically indicated. CSF samples were stored at −20°C until assayed. Cell counts, protein and glucose concentrations, and the presence of infection were documented for each spinal fluid sample.

Sixteen such paired samples were obtained within 2 d after asphyxia. Inclusion criteria during the perinatal period included the following:1) Apgar scores of ≤2 at 1 min and of ≤5 at 5 min in the context of a clinical setting consistent with asphyxia, such as placental abruption, CPR at birth, prolonged fetal bradycardia, or fetal acidosis;2) seizures within the first 12 h of life associated with fetal distress as noted above (and in the absence of other causes of seizures, such as trauma or intracranial hemorrhage); or 3) initial blood lactic acid concentration >8 mM/L associated with fetal distress as noted above and in the absence of other causes such as an inborn error of metabolism. Older infants were included in the asphyxiated group if they had a documented period of prolonged hypoxia with oxygen saturations <75% that was associated with clinical sequelae such as significant metabolic acidosis or the new onset of seizures. Thirty-one paired CSF and plasma samples were obtained from 26 neonates and children with culture-proven meningitis or presumed meningitis based on clinical presentation and elevated CSF cell counts. In four patients with meningitis, more than one spinal tap was evaluated. Eleven paired samples were obtained from neonates who had an IVH within 1 wk before the spinal tap. Twenty neurologically normal neonates were considered neonatal controls because they had a lumbar puncture to evaluate the possibility of meningitis, but such was excluded on the basis of normal CSF findings. The gestational ages of neonatal controls ranged from 24 wk to term. Twenty additional paired samples were obtained from children aged 1 mo to 12 y, and these were used as controls for children with meningitis. Eleven samples were obtained from eight individuals with miscellaneous neurologic disorders including seizures, apnea, CNS lymphoma, brain arteriovenous malformations, and progressive spinal muscular atrophy. Paired samples from an additional 12 infants treated with rEpo (Epogen, Amgen, Thousand Oaks, CA) were considered separately. These infants were given either 200 U·kg−1·d−1 i.v. as a continuous infusion in the hyperalimentation solution or 400 U/kg subcutaneously on Monday, Wednesday, and Friday. Investigators from this study did not influence whether infants received rEpo therapy. The present study was done with the approval of the University of Florida Institutional Review Board.

Evaluation of IVH.

All infants with birth weight <1500 g received cranial ultrasound examinations on or before the seventh day of age as part of routine care in our NICU. Ultrasounds were repeated sequentially if abnormalities were found.

Blood samples.

Blood samples drawn within 24 h of the spinal tap were collected. Blood was spun (1000 rpm for 5 min), and plasma was removed and stored at −20°C until use.

Epo assay.

Epo concentrations in CSF and plasma were assayed using the Quantikine IVD human Epo immunoassay ELISA (R&D Systems, Minneapolis, MN). A standard curve was created in duplicate by using control solutions ranging from 0 to 200 mU/mL Epo, with a third standard curve created in CSF to confirm the reliability of the assay in human CSF (13). Variability was <2% and sensitivity 0.6 mU/mL. This assay has been tested for cross-reactivity with other cytokines, and the specificity of the assay is >98%. We also tested the cross-reactivity of this ELISA with orosomucoid (a plasma protein that, like Epo, is highly glycosylated) at concentrations of 0.05, 0.5, and 5 mg/mL and found no nonspecific interactions with the glycosylated side chains.

Statistics.

The t test with a Bonferroni correction for multiple comparisons was used to compare the groups with CNS injury with the control groups. A p value of <0.05 was considered significant. Linear regression between paired samples was used to determine whether CSF and plasma concentrations of Epo correlated with one another. An r value of >0.6 with a p value of <0.05 was considered significant. Values in the text are given as mean ± SEM.

RESULTS

A total of 114 patients were enrolled in this study, 79 of whom were neonates. The ages of patients ranged from 24 wk gestation to 16 y of age. CSF Epo concentrations ranged from undetectable (in one sample) to 2350 mU/mL, and plasma Epo concentrations ranged from 1 to 18 501 mU/mL (Fig. 1). Several samples from asphyxiated infants required serial dilutions of up to 1:100 to bring the Epo concentration of the sample within the measurement range of the ELISA. Overall, the pairs of CSF and plasma Epo concentrations correlated with one another (r= 0.97, p= 0.002).

Figure 1
figure 1

CSF and plasma Epo concentrations from all patient samples are shown. The x axis shows plasma Epo concentrations, whereas the y axis shows the CSF Epo concentrations. Note that both axes are on a log scale. Patients of corrected age of 40 wk or less are shown by closed circles. Those between 41 and 52 wk are shown by open triangles, and those older than 52 wk are shown by open circles. A line of linear regression is also shown (r= 0.97, p= 0.002).

Full size image

To address whether Epo crosses the blood-brain barrier in healthy premature infants, we evaluated paired CSF and plasma samples from 12 infants who had lumbar punctures while being treated with rEpo (Fig. 2, lower left panel). CSF Epo concentrations in this group ranged from undetectable (<0.6 mU/mL) to 12 mU/mL (5.7 ± 0.9). Plasma concentrations ranged from 11 to 306.5 mU/mL (96.7 ± 23.9). There was no correlation between the CSF and plasma Epo concentrations (r= 0.3, p= 0.34).

Figure 2
figure 2

Individual panels show CSF and plasma Epo concentrations for patients in the control group, asphyxia group, those with IVH, meningitis, rEpo treatment, and in the miscellaneous category. The x axis shows plasma Epo concentrations, whereas the y axis shows the CSF Epo concentrations. Note that the axes for patients with asphyxia are on a log scale. Linear regressions for each group are shown.

Full size image

The distribution of CSF and plasma Epo concentrations in asphyxiated infants varied widely, as can be seen in the upper right panel of Figure 2. CSF Epo concentrations in this group ranged from 6.7 to 2350 mU/mL (225.9 ± 155.2). Plasma concentrations ranged from 6.7 to 18 501 mU/mL (1806.8 ± 1254.1). There was a positive correlation between CSF and plasma Epo concentrations (r= 0.996, p< 0.0001). For the patients with perinatal asphyxia, the closer to birth the samples were obtained, the higher the Epo concentrations were. Overall, patients with asphyxia had higher CSF (p< 0.05) and plasma (p< 0.05) Epo concentrations than controls (Fig. 3).

Figure 3
figure 3

Mean ± SEM values for CSF and plasma Epo concentrations in each group of patients are compared. Control patients are shown by black bars, asphyxiated patients by striped bars, IVH patients by checked bars, patients with meningitis by open bars, and Epo-treated individuals by bars with crisscross stripes. ★p< 0.05, ★★p< 0.01 compared with the control group.

Full size image

Eleven patients were identified with IVH (three infants had grade I, three had grade II, four had grade III, and one had grade IV hemorrhage) and received a spinal tap within the week after the diagnosis of the hemorrhage. Half of these were >31 wk of gestation and presented with neurologic abnormalities such as apnea or seizures. The distribution of CSF and plasma Epo concentrations is shown as a function of postnatal age in the middle left panel of Figure 2. CSF Epo concentrations ranged from 4.3 to 63.1 mU/mL. In seven of these eleven paired samples, CSF Epo concentrations were higher than the corresponding plasma value, although some of these pairs fell within the normal range for CSF and plasma Epo. The five patients with grade III and IV hemorrhages did not have higher CSF Epo than the six with grade I and II hemorrhages. A trend was noted between CSF Epo concentrations and age, with those CSF samples obtained closer to birth (and presumably to the time of the hemorrhage) tending to show higher CSF Epo concentrations than those obtained later.

Mean CSF Epo concentrations in neonates with IVH were higher than those in neonatal controls (20.6 ± 6.9 versus 5.8 ± 0.7 mU/mL respectively, p< 0.05). Plasma concentrations from these infants ranged from 4.7 to 164.2 mU/mL, and, as a group, these were not different from plasma values for neonatal controls (29.8 ± 14.3 versus 5.6 ± 0.8 mU/mL, respectively, p= 0.52), as shown in Figures 2 and 3. There was no correlation between CSF and plasma Epo concentrations in these infants (r= −1.61, p= 0.64).

Twenty-six patients had meningitis, and 31 paired samples were analyzed from these because four individuals had one or more follow-up taps. Thirteen patients in this group were neonates, and 13 were older children with ages ranging from 6 wk to 12 y. As a group, patients with meningitis did not have CSF or plasma Epo concentrations different from controls (Fig. 3). No differences were noted between the neonatal and older groups; thus, they were evaluated together. When analyzed by linear regression, the CSF and plasma Epo concentrations did not correlate (r= 0.04, p< 0.83).

An additional group of patients with miscellaneous neurologic injuries was studied to determine whether a change in CSF Epo might be a nonspecific CNS response. This group included three infants with severe hypotonia (two of whom had courses compatible with spinal muscular atrophy), one child with CNS lymphoma, two infants with significant apnea, one child with unexplained seizures, and two neonates with large CNS arteriovenous malformations. The CSF and plasma Epo concentrations in this group were no different from controls (4.8 ± 4.2 and 8.0 ± 8.1, respectively).

DISCUSSION

Epo is a major determinant of tissue oxygenation, performing this function by regulating the production of red blood cells. Circulating Epo concentrations increase after tissue hypoxia, reflecting an increase in production primarily from kidney and, to a lesser extent, from liver (10, 14). Increased Epo concentrations in cord blood and amniotic fluid have been identified as markers for fetal hypoxia (15–17), suggesting that this feedback mechanism, mediated in part by hypoxia-inducible factor 1 (18), is intact in the fetus. Epo production in rat and primate brain is also increased after hypoxia (10–12), suggesting that CNS Epo synthesis may be regulated by similar mechanisms. Epo production by astrocytes is, however, also affected by hormones such as IGF, glucocorticoids, and thyroid hormone (19–21). We and others have previously reported the presence of Epo within the CSF of premature and term infants (13, 22), but little is known about the metabolism of Epo within the human brain or how different mechanisms of brain injury affect Epo CSF concentrations.

In the present study, we observed that neonates treated with rEpo did not have higher CSF Epo concentrations than untreated controls despite significantly elevated plasma Epo concentrations. This supports the hypothesis that rEpo does not cross the intact blood-brain barrier in the otherwise healthy premature infant. This is reassuring, because the function of Epo within the CNS is still not well understood. There are reports of both beneficial and deleterious CNS effects associated with rEpo therapy. Reported beneficial effects in adults and children with renal failure include improved cognitive function and overall well being (23–27), whereas experiments in animal models and cell culture show that Epo can have neuroprotective (7–9) and neurotrophic effects (28). Deleterious effects reported in patients with renal failure include seizures and hypertensive leukoencephalopathy (29–31). Adverse effects of rEpo administration have not been shown in preterm infants, although no long-term neurodevelopmental studies have been reported. As it is unlikely that rEpo crosses the intact blood-brain barrier, the CNS findings documented after use of systemic rEpo may reflect secondary effects of rEpo mediated by factors such as improved hematocrit or increased endothelin-1.

In contrast with our findings in rEpo-treated patients, we found that patients who had sustained a hypoxic injury had higher CSF and plasma Epo concentrations, and the values were highly correlated. It is possible that the hypoxic insult resulted in increased permeability of the blood-brain barrier, with “spillover” of Epo into the CSF. Alternatively, it is possible that hypoxia resulted in an increase in both systemic and CNS Epo production. Our data cannot differentiate between these two possibilities.

Infants who had sustained an intracranial hemorrhage showed a different pattern of Epo concentrations in CSF and plasma than infants with asphyxia. As a group, CSF Epo concentrations were increased compared with age-matched controls, but plasma Epo concentrations were not. Unlike asphyxiated patients, there was no correlation between CSF and plasma Epo concentrations. More than half the patients with IVH had CSF Epo concentrations higher than circulating Epo concentrations, supporting the concept that the Epo measured in the CSF originated in the brain. This increased CSF Epo may reflect regional hypoxia in areas of the brain affected by the hemorrhage. Not all patients we examined with IVH had elevated CSF Epo concentrations. This may have been because we missed the peak Epo response or because there was not an increase in CNS Epo production in these individuals. More patients need to be studied to answer this question. Similarly, further studies are needed to determine whether an increase in CSF Epo correlates with decreased periventricular leukomalacia or improved neurologic outcome.

Patients with meningitis did not have an increase in either CSF or plasma Epo, suggesting that inflammation does not up-regulate somatic or CNS Epo production. Patients with other neurologic diseases such as progressive hypotonia also did not have elevated CSF Epo concentrations. This supports the hypothesis that Epo elevations in both CSF and plasma are a specific response to hypoxia rather than a nonspecific marker for CNS injury.

In conclusion, we have shown that in the patient groups we studied, hypoxia, either local or systemic, seemed to be the major stimulus for increased net Epo in the CSF. Other pathologic processes such as inflammation or progressive neurodegeneration did not increase CSF Epo concentrations. Further studies are needed to determine the significance of Epo within the CNS. We speculate, on the basis of animal studies showing the neuroprotective effects of rEpo in vitro and in vivo, that the higher concentration of CSF Epo associated with hypoxic injury is one mechanism by which the brain attempts to protect itself from hypoxia.