Main

Until recently, the action of Epo was considered to be restricted to erythropoiesis(1–3), but there is now evidence that certain nonerythroid cells express Epo-R which respond to rEpo in vitro and in vivo. These receptors have been demonstrated on murine cell lines of neural origin(4), in cultured murine embryonic neurons, in the developing brain and spinal cord of rodents(5–7), and in mid-trimester human fetal spinal cords(8). Regulated Epo synthesis has also been demonstrated in rodent cells of CNS origin(9, 10), and we have shown that significant concentrations of Epo are detectable in the spinal fluid of normal preterm and term human neonates, although its source is not clear(11). It is not known, however: 1) which cells within the human CNS express Epo, 2) whether this expression changes with development, 3) which cells within the CNS express Epo-R, 4) whether Epo-R expression within the CNS changes with development, and 5) whether these Epo-R are functional, and if so, what this function might be.

One mechanism by which Epo promotes erythropoiesis is by preventing the programmed cell death of erythrocyte precursors (CFU-E) via up-regulation of the bcl-2 and bcl-xL pathways(12). We proposed that Epo might have similar mechanisms of action in the CNS, by decreasing apoptosis of neurons during normal brain development, or by exerting a neuroprotective effect under adverse conditions such as hypoxia.

We hypothesized that neurons and glial cells express Epo and Epo-R, that expression of this cytokine and its receptor change with advancing gestational age, and that Epo binding to Epo-R in neurons provides protection from programmed cell death. A series of studies were designed to test these hypotheses.

METHODS

Human fetal specimens. Spinal cords, brainstems, and brains were collected from fetuses of 5-9 wk of gestation for primary cell culture. Tissues were transported in cold DMEM (GIBCO, Gaithersburg, MD). Additional spinal cords and brains (ranging from 7 to 24 wk of gestation) were identified, washed in sterile PBS, and immediately preserved in liquid nitrogen for later RNA extraction. Only fetuses that were normal by ultrasound examination and underwent elective pregnancy termination were studied. Gestational age was determined by fetal foot and long bone length(13, 14). The members of our study group had no contact with the mothers and made no attempt to influence decisions about pregnancy termination. The studies were approved by the University of Florida Institutional Review Board.

Mixed primary cell culture. Spinal cord, brain stem, and brain, ranging in gestational age from 5 to 9 wk, were placed in separate 100-mm Petri dishes containing 30 mL of isotonic salt solution (125 mM NaCl, 5 mM KCl, 5 mM Na2HPO4, 1.2 mM KH2PO4, 6 mM glucose, and 60 mM sucrose), containing 100 U of penicillin, 100 μg of streptomycin, and 0.25 μg of Fungizone, pH 7.2. Membranes and blood vessels were stripped from the tissue surfaces, and the tissues washed twice in isotonic salt solution, before chopping into pieces of approximately 2 mm3. Tissues were then suspended in 0.25% (wt/vol) trypsin in isotonic salt solution and placed in a 37°C shaking water bath for 15 min. Sedimented tissue pieces and dissociated cells were collected and mixed with an equal volume of DMEM containing 20% FBS (Sigma Chemical Co., St. Louis, MO). Dissociated cells were pooled and treated with 160 μg of DNAase I for 5 min at 37°C, then centrifuged at 800 × g for 10 min, and washed with 40 mL of DMEM containing 10% FBS. Cells were counted by hemocytometer and plated onto poly-L-lysine-coated dishes. Three days after plating cells, the medium was removed and replaced with DMEM containing 10% FBS and 1% cytosine ARA-C. After 2 d the medium was changed to DMEM containing 10% FBS(15). Cultures medium was changed every 3 d using fresh DMEM containing 10% FBS, until approximately 3-4 wk of age. For“glialenriched” cultures, the same procedure as above was used, except the cells were incubated with DMEM containing 10% FBS without 1% ARA-C.

NT2 and hNt cell culture. NT2 cells are a human committed neuronal precursor cell line derived from teratocarcinoma (Stratagene, La Jolla, CA), which can be induced by retinoic acid to differentiate in vitro into postmitotic CNS neurons (hNT cells)(16, 17). In the course of the differentiation process the NT2 cells lose neuroepithelial cell markers and gain markers specific for mature neurons. We carried out experiments on NT2 cells and hNT cells (Stratagene), and also on hNT cells which we induced in our laboratory following the prescribed protocol. In brief, cells are incubated with 10 μM retinoic acid in DMEM for 5 wk, then incubated with medium containing mitotic inhibitors (10 μM 5-fluoro-2′-deoxyuridine, 10 μM uridine, 1 μM ARA-C) for 10 d, at which time they are replated using a “neuron knock off” technique onto plates prepared by poly-L-lysine and Matrigel® coating.

Immunohistochemistry. Mixed primary cultures were fixed for 1 h in 4% paraformaldehyde, then rinsed twice, for 10 min, with PBS. Two additional 10-min rinses with 1% horse serum, PBS, and 0.4% Triton X (pH 7.2) followed. Cultures were then blocked with 1:20 horse serum to PBS containing 0.4% Triton X. The primary antibody was then incubated with the cells for 2 d at 4°C. Antibodies used included: monoclonal anti-MAP2 (Zymed Laboratories, San Francisco, CA) diluted 1:1500 (19 dishes stained)(18), monoclonal GFAP (Sigma Chemical Co., St. Louis, Mo) diluted 1:800 (17 dishes stained)(19), monoclonal von Willebrand factor (Novocastra Laboratories, Newcastle upon Tyne NE2 4AA. UK) diluted 1:500 (10 dishes stained)(20), and polyclonal antibody to the Epo-R (Santa Cruz Biotechnology; specificity of antibody confirmed by personal communication, Dr. Renee White), diluted 1:100 (six dishes stained). Biotinylated horse anti-mouse or anti-rabbit antibodies were used as a secondary antibodies, as appropriate to each respective antibody, diluted 1:100 for 60 min. Four 10-min rinses with 1% horse serum, PBS, and 0.4% Triton followed. Vectastain ABC reagent (Vector ABC immunokit) was applied, rinsed with PBS, and incubated in 0.05 M Tris for 5 min. Diaminobenzidine (Sigma Chemical Co.) was used as the chromogen. To control for the presence of endogenous peroxidase, sections were incubated in the absence of specific antibody. Human pulmonary artery endothelial cells (a gift from Dr. Gary Visner, University of Florida), were used as a positive control for von Willebrand antibody. Absence of the primary antibody was used as a negative control for each antibody, and in addition, for Epo-R staining, a specific blocking peptide was used (Santa Cruz). All negative controls showed absence of staining.

Epo assay. Epo concentrations in the conditioned media samples were assayed using the Quantikine™ IVD™ human Epo immunoassay ELISA (R&D Systems, Minneapolis, MN). Aliquots (100 μL) of conditioned medium which had been concentrated 10-fold (Centricon, Inc. Amicon, Beverly, MA) were assayed in duplicate (n = 6). Variability was less than 2% between duplicate samples. Sensitivity of this assay has been determined at 0.6 mU/mL. This assay has been tested for cross-reactivity with other members of its cytokine family, and the specificity of the assay is greater than 98%.

Preparation of total RNA. Total RNA was extracted from the washed tissues using either the guanidinium thiocyanate method(21), or using the RNeasy elution kit, a method of RNA extraction based on the selective binding of RNA to a silica membrane (Qiagen, Chatsworth, CA)(22). Total RNA was treated with RNase-free DNAse I (GIBCO) before further experimentation.

RT of RNA and amplification of cDNA. RT of RNA and amplification of cDNA was performed using a DNA thermal cycler (Perkin-Elmer Corp., Foster City, CA). Total RNA (2.0 μg) was combined with 2.0 μM oligo(dT) primers (GIBCO), heated to 70°C for 10 min, then placed on ice. The mixture was then combined with 250 μM dNTP (GIBCO), 0.01 M DTT, 50 mM Tris, pH 8.3, 75 mM KCl, and 3 mM MgCl2. The mixture was incubated at 42°C for 2 min; 2 μL of Superscript II (GIBCO) were added, and then the mixture was incubated for an additional 50 min. The reaction was inactivated by heating to 70°C for 15 min. Amplification cycles were carried out under the following conditions: 10 mM Tris, pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 2.0 mM dNTP, 0.2 μM upstream and downstream primers, 5% reverse transcriptase mix, 0.1 U/μL Ampli Taq 94°C for 1 min, 56°C for 1 min, 72°C for 2 min for 30 cycles, followed by 10-min elongation at 72°C. Primer pairs used to identify Epo, Epo-R, and β-actin are shown in Table 1(23, 24).

Table 1 PCR primers
Full size table

Confirmation of Epo and Epo-R PCR products. To confirm that the 197-bp RT-PCR product was indeed Epo-R, the PCR product derived from mRNA from primary neuronal culture was sequenced. The Taq DyeDeoxy Terminator protocol developed by Applied Biosystems (Perkin-Elmer Corp.) was used, and labeled extension products were analyzed on an Applied Biosystems model 373 DNA sequencer. The 234-bp product for Epo was likewise confirmed by direct sequencing.

Semiquantitative PCR. Semiquantitative PCR was performed relating synthesis of Epo and Epo-R cDNA to the internal standardsβ-actin and 18 S RNA, and to an external standard cyclophilin Competicon™ (Ambion, Inc, Austin, TX). The PCR reaction was carried out in the presence of 2.5 μCi/mL of 3000 Ci/mmol α-[32P]dCTP. Initial experiments to determine the optimal number of cycles required to achieve amplification in the linear range demonstrated that 30 cycles would be required (data not shown). After completion of the PCR, 30 μL of the reaction were separated by electrophoresis through a 3% agarose gel in 0.5× Tris-borate-EDTA buffer, and size was standardized with 100-bp DNA markers (GIBCO). Gels were photographed. The amplified products of appropriate size were then cut out of the gel, transferred to scintillation vials containing 5 mL of Cytoscint, and 32P activity was counted. The counts for Epo and Epo-R were related to the counts for β-actin, 18 S RNA, or cyclophilin, which were co-amplified.

Neuroprotection assessment. NT2 or hNT cells were grown in OptiMEM medium (GIBCO) containing 10% FBS, 2 mM L-glutamine, 100 U of penicillin, and 100 μg of streptomycin. In separate experiments, rEpo was added to quadruplicate 100-mm dishes exposed to hypoxia alone (1% oxygen, 5% carbon dioxide, 94% nitrogen) for 24, 36, or 48 h followed by 24 h of normoxic conditions. rEpo doses included 0, 0.1, 0.5, 1, 5, or 10 U/mL. A neutralizing polyclonal rabbit anti-human Epo antibody (Genzyme) was added at 1:1000 dilution to cell culture dishes containing 5 U/mL rEpo. The release of nuclear matrix protein into the media by dying cells was sought by ELISA (Calbiochem, Cambridge, MA)(25). The protein composition of the nuclear matrix determines the basic morphologic pattern of the nucleus, and as the nucleus becomes pyknotic during apoptosis, these proteins are released from the nucleus in soluble form. The amount of protein released is a function of the number of dead cells. In separate set of triplicate experiments done comparing cells cultured in the presence or absence of 5 U/mL Epo, cell viability was assessed by counting the number of live and dead cells after 36 h of hypoxia, and also after the 24 h of recovery. Cell viability was determined by trypan blue staining. The presence of apoptotic cells death was ascertained by a DNA fragmentation kit (Calbiochem, Cambridge, MA). In this assay, cells were first permeabilized with proteinase K for 5 min. Endogenous peroxide activity was quenched with hydrogen peroxide. Klenow was then bound to exposed 3′-OH ends of cleaved DNA fragments, and biotin-labeled nucleotides were incorporated in a template-dependent manner. These nucleotides were detected using streptavidin-horse radish peroxide conjugate. Diaminobenzidine was used as a chromogen, with methyl green as a counterstain. Incubation of cells with Dnase I was used to generate the positive controls, and water added in place of Klenow enzyme provided the negative controls. At the conclusion of triplicate experiments, the medium was removed and saved, the cell layer trypsinized, and the cells in the medium were combined with the trypsinized cells, an aliquot counted to determine cell density, and an estimated 3.25 × 105 cells were cytospun onto a slide and stained. Ten fields were counted per slide.

RESULTS

The first set of experiments were performed to determine whether the Epo-R mRNA detected in the fetal neuronal tissue was the result of blood contamination or meningeal endothelial cells. To exclude these possibilities, primary cultures of “neuron-enriched” and“glial-enriched” cells were grown for investigation. These cultures were characterized by immunohistochemistry and light microscopy after three weeks of culture. Figure 1A shows representative cultures exposed to 1% ARA-C to inhibit the growth of glial cells. These neuronenriched cultures are shown stained with the monoclonal MAP2 antibody, which reacts with a specific 280-kD cytoskeletal protein, the somatodendritic MAP of neurons. This is expressed to a lesser extent in astrocytes. Although these cultures were greatly enriched for neurons, some glial cells remained.Figure 1B illustrates a typical glial-enriched culture stained with GFAP, which stains intermediate filaments of characteristic 10-mm diameter that are specific to astrocytes.

Figure 1
figure 1

(A) A typical mixed cell culture identified as neuron-enriched. These cells have been stained with the monoclonal anti-MAP2 antibody. Magnification, ×200. (B) A typical glial-enriched culture. This has been stained with the monoclonal anti-GFAP, which stains intermediate filaments specific to astrocytes. Magnification, ×200. These mixed cell cultures contain other nondifferentiated neurons and microglia.

Full size image

RT-PCR was performed on total RNA extracted from cultures that were neuron-enriched, or glial-enriched to ascertain whether these cells expressed mRNA for Epo-R and Epo. Human neuronal precursors (NT2 cells), and their mature, differentiated counterparts (hNT cells) were also investigated, as these provided a pure growth of neuronal cells without the glial supporting cells, or possible contamination with endothelial cells. All cells tested expressed both Epo (Fig. 2A) and Epo-R (Fig. 2B) mRNA. Fig. 2C shows theβ-actin control PCR reactions. Negative controls, done using either water or RNA, which had not been reverse transcribed, were negative.

Figure 2
figure 2

RT-PCR products using primers specific to human Epo(A), Epo-R (B), and β-actin (C) are displayed for mixed neuronal culture (lanes 2 and 3), for mixed glial-enriched cultures (lanes 4 and 5), and for NT2 and hNT cells (lanes 6 and 7). Lanes 1 and 8 contain 2 μg of a DNA ladder marking 100-bp increments.

Full size image

A rabbit polyclonal antibody specific to amino acids 489-508 of human Epo-R was used to identify cells derived from the fetal CNS that express Epo-R on their surfaces. Neurons (Fig. 3A) and astrocytes (Fig. 3B) stained positively for Epo-R.Figure 3C shows a negative control for Epo-R, incubated in the absence of primary antibody. To test endothelial cell contamination as a source of Epo-R(26, 27), cells were stained with von Willebrand factor (Fig. 3D). None of the neuron or glialenriched cultures stained positive for von Willebrand factor, although human endothelial cells used as positive controls were positive (data not shown). All cells stained in the absence of primary antibody were negative.

Figure 3
figure 3

Polyclonal anti-Epo-R antibody 1:500 was used to identify cells expressing Epo-R. Demonstration of neuronal staining(A) and lighter staining of glial cells (B). Original magnification, ×160. (C) A neuron-enriched culture to which no primary antibody was added (negative control). Original magnification,×400. (D) Lack of cell staining after incubation of a glial-enriched culture with von Willebrand antibody. Original magnification,×200. Bar = 9.2 μm for C, 20 μm for A, B, and D.

Full size image

To identify which cell types show immunoreactivity to Epo-R in vivo, brain sections taken from a brain at 20 wk of gestation and from an adult brain were stained. Figure 4 demonstrates that in vivo, certain types of neurons (Cajal Retzius neurons(28) shown in Fig. 4A), astrocytes (Fig. 4B), and choroid plexus (Fig. 4C) all stained positively for Epo-R. Figure 4D shows a slice of brain adjacent to that shown in Fig. 4C, after incubation with the Epo-R-blocking peptide (negative control).Figure 4,E and F, shows human fetal liver at 22 wk of gestation, as a positive and negative control (with the Epo-R-blocking peptide), respectively. In adult brain, Epo-R staining was identified predominantly in astrocytes, not in neurons (data not shown).

Figure 4
figure 4

These photomicrographs show specific cell types within the brain that stain positively with anti-Epo-R antibody. (A) An example of Cajal Retzius neurons (arrows) in the subpial granular layer of the developing 20-wk brain. The underlying cortical plate is identified as CORT. Original magnification, ×400. (B) Specific staining of astrocytes within the deep white matter of an adult brain. Original magnification, ×400. An arrow identifies one astrocyte.(C) The positive staining of choroid plexus (CP) from an adult brain. Original magnification, ×200. (D) A negative control using the specific blocking peptide for the Epo-R antibody. Original magnification, ×160. (E) A positive control, human fetal liver from 22 wk of gestation (original magnification, ×200). (F) A negative control, human fetal liver from 22 wk of gestation, pretreated with blocking peptide. Original magnification, ×200.

Full size image

Epo concentrations measured in the supernatant of neuronenriched and glial-enriched cells ranged from below the limit of detection (0.6 mU/mL) to 2 mU/mL. Unconditioned medium which had been similarly concentrated did not contain measurable Epo concentrations, ruling out cross-reactivity with the endogenous Epo in the FBS. Semiquantitative PCR of triplicate NT2 cultures showed an increase in Epo mRNA after exposure to hypoxia (mean Epo/actin value in normoxic conditions 0.35 ± 0.05; mean ± SEM, whereas values obtained for cells maintained in hypoxia overnight were 0.74 ± 0.06,p = 0.006 by a t test).

To identify changes in Epo and Epo-R mRNA expression during fetal development, spinal cords from fetuses aged 7-16 wk of gestation and brains from fetuses aged 8-24 wk of gestation were studied. RT-PCR showed that Epo and Epo-R were expressed in all samples (Fig. 5,A and B).β-Actin controls for each RNA sample are shown in Fig. 5C.

Figure 5
figure 5

RT-PCR products using primers specific to human Epo(A), Epo-R (B), and β-actin (C) are displayed for spinal cords and brains ranging in gestational age from 7 to 24 wk. Lane 1 contains 2 μg of DNA ladder marking 100-bp increments. Lanes 2 through 10 contain cDNA products from spinal cords of 7, 8, 9, 10, 11, 12, 13, 15, and 16 wk of gestation, respectively. Lanes 11 through 14 contain cDNA products from brains of 17, 21, 23, and 24 wk of gestation.

Full size image

Semiquantitative PCR using internal and external controls revealed no significant change in spinal cord expression of Epo-R between 7 and 16 wk of gestation (Fig. 6A); however, statistical analysis using a Spearman's rank correlation showed an increase (p = 0.007,rs = 0.656) in expression of brain Epo-R between 8 and 24 wk of gestation. Similarly, no significant changes in spinal cord Epo mRNA expression were noted (Fig. 6B); however, like Epo-R, increased Epo expression was noted in brains between 8 and 24 wk (p= 0.008, rs = 0.838 by Spearman's rank correlation test).

Figure 6
figure 6

(A) Demonstration of relative Epo-R expression (y axis) as a function of gestational age in weeks(x axis), whereas (B) shows similar results for Epo expression. Spinal cord samples are shown by open circles, whereas brain samples are denoted by filled diamonds. Data from three semiquantitative PCR runs were combined. Results are shown ± SEM.

Full size image

To assess whether Epo might function by affecting the process of apoptosis, we cultured NT2 cells and hNT cells (in separate experiments) in 1% oxygen for 24, 36, or 48 h followed by a 24-h recovery period, conditions which we have found induce apoptotic cell death in a proportion of the cells (data not shown). Increasing concentrations of rEpo were added to quadruplicate cultures. After 24 h of hypoxia, a minority of cells underwent apoptosis. After 36 and 48 h, the number of affected cells increased. In a dose-dependent manner, rEpo decreased the effect of hypoxia on both NT2 cells and hNT cells exposed to hypoxia for 24, 36, or 48 h. After 24 h of hypoxia, 10 U/mL rEpo reduced NMP release by mature neurons by 30%. After 36 h of hypoxia, followed by 24 h of normoxia, NMP release by NT2 cells incubated with 0, 0.5, 5, or 10 U/mL rEpo was 102.00 ± 7.76, 81.44 ± 19.72, 66.22 ± 4.90, and 57.59 ± 7.49, respectively, and this effect was reversed by adding a 1:1000 dilution of Epo-neutralizing antibody to cultures containing 5 U/mL rEpo. NMP release by cells grown to 70% density in normoxic conditions are compared with these data and are shown in Figure 7

Figure 7
figure 7

Nuclear matrix protein is plotted in units/mL on the y axis, whereas culture conditions are denoted on the x axis. Condition 1 is cells grown in normoxic conditions as a baseline control. Conditions 2 through 5 are: hypoxia alone (36 h of hypoxia followed by 24 h of normoxia), hypoxia plus 0.5 U rEpo/mL, hypoxia plus 5.0 U rEpo/mL, and hypoxia plus 10.0 U rEpo/mL. Condition 6 is hypoxia plus 5.0 U/mL rEpo plus 1:1,000 dilution of neutralizing anti-Epo antibody. Results are shown as mean ± SEM. *p < 0.01.

Full size image

To establish whether Epo was affecting early cell death (during the period of hypoxia) versus late cell death (during the 24 h of recovery), in separate experiments, we counted trypan blue-stained cells in the medium, and in the trypsinized cell layer, after the 36 h of hypoxia, and again after the 24-h recovery period. We found that, in both the control and rEpo-treated cells, about one-third of the cell death occurred during the hypoxia exposure(24-36.1% for control cells versus 12.5-31% in rEpo-treated,p = 0.31), thus rEpo did not change the timing of cell death during the hypoxia. In separate experiments, we counted viable cells (rEpo-treated versus control) at the end of the 24-h normoxia recovery period. Cell counts from these experiments showed the rEpo-treated cells had approximately double the number of viable cells, as did the controls (3.13× 105 ± 0.28 × 105 versus 1.5× 105 ± 0.41 × 105, respectively, p< 0.05), despite equal numbers of cells having been plated at the beginning of the experiment.

To detect DNA fragmentation, which typifies apoptosis, we labeled cytospun control and Epo-exposed cells with Klenow and counted 10 fields per slide. Under conditions of hypoxia followed by normoxia in the absence of added rEpo, 21.5% cells (161/750 cells counted) had no evidence of DNA fragmentation, whereas when 5 U/mL rEpo was added to the cultures, 50.5% of cells (361/715 cells counted) had no DNA fragmentation.

DISCUSSION

The action of Epo on erythroid progenitor cells has been the subject of intense investigation(1, 29–32). In contrast, the action of Epo on nonhematopoietic cells is only beginning. It is known that Epo-R exist on a wide variety of nonerythroid cell types including endothelial cells, cardiomyocytes, mesangial cells, smooth muscle cells, placental tissues, and cells of neuronal origin(4, 6, 10, 26, 33–36). Wald et al.(35) recently reported that addition of rEpo to cardiomyocytes results in a mitogenic and chemotactic response (like endothelial cells), and that this response is most likely mediated by Na+-K+-ATPase activity. Carlini et al.(37) reported that incubation of endothelial cells with rEpo results in an increase in endothelin release.

Although the physiologic role of Epo in the CNS is still unclear, there is increasing evidence that the Epo-Rs expressed in neural tissues are functional. Early data showed that addition of rEpo to neural cells resulted in an increase in monoamine concentrations, and an increase in intracellular calcium(4, 10, 38). Circumstantial evidence that Epo and Epo-R might serve a role in neurogenesis is provided by studies demonstrating the presence of this ligand-receptor pair in the embryonic and fetal murine brain, with evidence of developmental regulation(5, 6). Further support for developmental regulation of Epo within the CNS was provided in our previous study in which Epo concentrations were measured in spinal fluids obtained from patients ranging in age from 24 wk of gestation to adulthood. Significant concentrations (up to 32 mU/mL) were found in normal premature and term infants untreated with rEpo, but these decreased by 5 mo of age to less than 2 mU/mL(11). Epo has been shown to increase in vivo survival of lesioned rat cholinergic septal neurons produced by fimbria-fornix transections, and to augment choline acetyltransferase activity in cultured neurons(38, 39). There is also data to suggest that Epo might function as a neurotropic factor, influencing neuronal development, differentiation, maintenance, and regeneration(38, 40). Most recently, Morishita et al.(7) has shown that, at concentrations present in the cerebrospinal fluid of normal infants, rEpo prevented glutamate-induced neuronal death if cells were preincubated for a minimum of 8 h. This suggests that at physiologically relevant concentrations Epo might be an important mechanism for neural preservation in adverse circumstances such as hypoxia.

Our first goal in the present study was to ascertain whether cells within the human fetal CNS express Epo. We observed that primary mixed cell cultures derived from first trimester spinal cords, enriched for either neurons or glial cells, synthesize Epo mRNA and protein. These cultures contained a mixture of astrocytes, microglia, neurons, undifferentiated cells, and possibly endothelial cells, thus it could not be conclusively stated which cell types were responsible for the Epo production. Although it is known that endothelial cells in the brain express Epo-R(27), it is unlikely that these cells contributed significantly to the Epo-R expression in our cultures because immunohistochemical staining with von Willebrand factor did not reveal any endothelial cells. Next we observed expression of both Epo and Epo-R in cultures of pleuripotent neuronal precursors (NT2 cells) and in terminally differentiated human neurons (hNT cells), neither of which contained any endothelial cell contamination. Our primary cell cultures were derived from fetuses of 5-9 wk of gestation, and we did not exhaustively rule out the presence of endothelial cells in these cultures. Thus, we could not definitively state, based on those studies, that expression of Epo and Epo-R was from neurons or glial cells, nor that this reflected in vivo expression in later gestation. We therefore performed additional studies, using an immunohistochemical approach to localize Epo-R to specific cell types within the brain. In addition to the expected endothelial cell staining(27), we observed Epo-R staining of early migrating neurons (Cajal Retzius neurons in the marginal zone of a 20-wk brain)(28), of astrocytes, and of the choroid plexus. In adult brain, Epo-R staining was located predominantly in astrocytes, not in neurons. Thus at different stages of development Epo-R are present in astrocytes and neurons in human brain. In mature mouse brain, the main Epo binding sites have been localized to white matter, including the capsula interna, corpus callosum, hippocampus, and to a lessor degree, the brainstem, mesencephalon, and lateral posterior thalamic nuclei(41). The localization of Epo and Epo-R within human brain and changes with development have not been fully delineated, and are beyond the scope of the present study.

Our next goal was to determine whether expression of Epo mRNA changed during fetal development. In murine studies, the synthesis of Epo and Epo-R in brain were limited to fetal life, and decreased with fetal maturity(5, 6). We found, by semiquantitative PCR, that expression of Epo and Epo-R mRNA did not change between 7 and 16 wk of gestation in the spinal cord, but both increased between 8 and 24 wk gestation in brain. One limitation of that study is, due to the difficulty of obtaining brain specimens in early gestation fetuses, we have no data points on either Epo or Epo-R expression by the fetal brain between 10 and 16 wk of gestation. Another caveat is that expression of this cytokine-receptor pair is likely to be specific to the precise anatomic location from which the RNA was derived within the brain, and this was not controlled for in our study. In addition, although statistically significant, it is unclear whether a doubling of Epo or Epo-R expression, detected by PCR, is physiologically significant.

The last goal of this study was to begin to elucidate the function of Epo-R within the developing human CNS. Murine and rodent neuronal tissues respond to hypoxic stimuli by increasing expression of Epo, in vitro and in vivo(9, 41). We exposed mixed primary cell cultures to hypoxia and found an increase in Epo mRNA expression. We previously described that the concentrations of Epo in the cerebrospinal fluid of preterm infants (including those treated with rEpo) ranged from undetectable to 51 mU/mL. The Epo concentrations we measured in conditioned media from mixed cell cultures were generally much lower than this, ranging from below the limit of detection to 2 mU/mL. It is possible that cells most responsible for contributing to cerebrospinal fluid Epo in babies 24 wk to term were not well represented in these cultures which were derived from CNS tissues of fetuses 5-9 wk of gestation, or, that the culture conditions are not representative of in vivo conditions. We tested whether or not Epo might have a neuroprotective effect in hypoxic conditions and found that it does. The rEpo concentrations used in these experiments were supraphysiologic, however, and these experiments must be repeated with concentrations more representative of those found in cerebrospinal fluid, perhaps with preincubation, as was published by Morishita et al.(7). Morishita found that when neurons were preincubated with rEpo, the effects of glutamate toxicity were counteracted. This suggests that the neuroprotection requires protein synthesis. Up-regulation of bcl-2 is an important mechanism by which neurons can be rescued from apoptotic death due to many stimuli(42–44). We speculate that Epo may rescue neurons from apoptosis via an increase in bcl-2 and bcl-xL, a mechanism by which Epo decreases programmed cell death in CFU-E(12).

It is clear that Epo and its receptor have a more widespread physiologic role than was once suspected. We conclude that both Epo and Epo-R are present within the CNS of developing humans; specifically that both are present in glial cells and neurons early in development. Further studies are needed to define the change in the pattern of expression of this receptor and its ligand as development proceeds. The CNS Epo-R appears to be functional, and may serve a role in neuroprotection by inhibiting apoptotic cell death under hypoxic conditions. The physiologic importance of these issues in human fetal development require further studies.