Main

Bacterial infections are a major cause of morbidity and mortality among VLBW infants (1). Mechanistic explanations for the excessive infectious complications of these infants include birth before the complete development of their phagocytic immune capacity (2). The phagocytic defects of VLBW infants include quantitative and qualitative deficiencies of neutrophils and macrophages, limitations that are ameliorated experimentally by the administration of rGM-CSF (3–5). Indeed, considerable enthusiasm for rGM-CSF administration to VLBW infants was engendered by studies that indicated a rGM-CSF dose-dependent increase in neutrophils and monocytes in the blood and bone marrow with no apparent short-term adverse effects (6).

Any salutary effects of rGM-CSF administration to VLBW infants must be weighed against any offsetting risks of such treatment. Relevant to the latter issue, recent studies suggest that GM-CSF administration to VLBW infants could have actions outside of the hematopoietic system. GM-CSF acts by specific binding with its cognate receptor (GM-CSF-R). The presence of functional receptors and access to its ligand are, therefore, essential prerequisites for GM-CSF function. Animal studies indicate that GM-CSF-R are expressed on cells in the developing CNS (7, 8) and that GM-CSF can cross the blood-brain barrier (9). Further, it seems that these receptors can be functional, as GM-CSF has been shown to have trophic effects on astrocytes and certain types of neurons (8, 10, 11). The extent of nonhematopoietic functions of GM-CSF in the human fetus and neonate has not been defined. It is not known which cells in the fetus or preterm neonate (aside from those in the bone marrow) express GM-CSF-R, whether receptors at nonhematopoietic sites are functional, and which cells produce GM-CSF. We reasoned that obtaining this information is essential if rGM-CSF is to be tested as a potential treatment for VLBW infants. As a first step toward this goal, we designed a study to define the distribution of cells producing GM-CSF and GM-CSF-R mRNA and protein during human fetal development.

METHODS

Human Fetal Specimens

Human fetal specimens were obtained after elective pregnancy termination or surgical removal (tubal pregnancies). Fetuses were normal by ultrasound examination. Gestational age was determined by fetal foot and long bone length (12–14). Studies were approved by the University of Florida Institutional Review Board. Organs including brain, spinal cord, eyes, clavicle, heart, lung, liver, spleen, adrenal, kidney, stomach, small bowel, and placenta were collected from fetuses of 8 and 16 ± 2 wk postconception. Organs were identified using a dissecting microscope and placed either in Bouin's fixative for immunohistochemistry or snap-frozen for RNA extraction. Additional CNS tissues ranging from 8 wk to adult were obtained from stillbirths and autopsy specimens. Organs used for RNA studies were preserved using liquid nitrogen within 15 to 30 min of delivery.

Mixed primary cell culture.

Spinal cord, brain stem, and brain were cultured as described previously (15). Briefly, spinal cord, brain stem, and brain ranging in gestational age from 5 to 9 wk were placed in isotonic salt solution (125 mM NaCl, 5 mM KCl, 5 mM Na2HPO4, 1.2 mM KH2PO4, 6 mM glucose, 60 mM sucrose, containing 100 U penicillin, 100 μg streptomycin, and 0.25 μg Fungizone, pH 7.2). Membranes and blood vessels were stripped from the tissue surfaces, and the tissues were washed twice before they were chopped into pieces of approximately 2 mm3. Tissues were then suspended in 0.25% (wt/vol) trypsin and shaken at 37°C for 15 min. Dissociated cells were pooled and treated with DNAase I (GIBCO, Gaithersburg, MD) for 5 min at 37°C, centrifuged at 800 ×g for 10 min, and washed with Dulbecco modified Eagle medium containing 10% FBS. Cells were counted by hemocytometer and plated onto poly-L-lysine–coated dishes (Fisher). For neuron-enriched cultures, glial cell growth was inhibited by incubation with 1% cytosine β-D-arabinofuranoside for 2 d. For glial-enriched cultures, this step was omitted. These cultures were characterized by immunohistochemistry and light microscopy after 3 wk of culture. The neuron-enriched cultures contained 90–95% neurons with occasional glial cells and fibroblasts (as determined by MAP-2, glial fibrillary acidic protein, and vimentin immunostaining). No endothelial cells were present (defined by the absence of von Willebrand factor staining in the presence of positive controls). Glial cultures were approximately 95% pure.

NT2 and hNT cell culture.

NT2 cells are a human committed neuronal precursor cell line derived from teratocarcinoma (Stratagene, La Jolla, CA) that can be induced by retinoic acid to differentiate in vitro into postmitotic CNS neurons (hNT cells) (16, 17). These cells were cultured as previously described (15).

Identification of Specific mRNA

Preparation of total RNA.

Total RNA was extracted from the washed tissues and from cultured cells with the RNeasy elution kit (Qiagen, Santa Clarita, CA). Purity and concentration of extracted RNA was determined by measuring UV absorbance at 260 and 280 nm. Total RNA was treated with RNAse-free DNAse I (GIBCO) before further experimentation.

Reverse transcription of RNA and amplification of cDNA.

Reverse transcription (RT) of RNA and amplification of cDNA were performed using a DNA thermal cycler (Stratagene, La Jolla, CA). Total RNA (2.0 μg) was combined with 2.0 μM oligo(dT) primers, heated to 70°C for 10 min, and then placed on ice. The mixture was combined with 250 μM dNTP, 0.01 M DTT, 50 mM Tris pH 8.3, 75 mM KCl, and 3 mM MgCl2. After a 42°C incubation for 2 min, 2 μL of Superscript II reverse transcriptase (GIBCO) was added, and the mixture was incubated for 50 min. The reaction was terminated by heating to 70°C for 15 min. Amplification of the cDNA was carried out as follows: 10 mM Tris pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 2.0 mM dNTP, 0.2 μM upstream and downstream primers, 0.1 U/μL Ampli Taq DNA polymerase (GIBCO) 94°C for 1 min, 51°C (GM-CSF-R) or 56°C (GM-CSF) for 1 min, 72°C for 1.5 min for 30 cycles followed by 10-min elongation at 72°C. Primer pairs used for GM-CSF amplification were synthesized by GIBCO: 5′-ACA-CTG-CTG-AGA-TGA-ATG-AAA-CAG-TAG-3′ and 5′-TGG-ACT-GGC-TCC-CAG-CAG-TCA-AAG-GGG-ATG-3′, with an expected fragment length 286 bp. For GM-CSF-R amplification, 5′-TGT-CAC-CTG-GAT-AAC-CTG-TC-3′ and 5′-GCA-GCT-CTG-ATC-TTC-ACA-CT-3′ were used, with an expected fragment length of 365 bp. Every sample was tested for RNA integrity by using β-actin primers. β-Actin controls were amplified using 5′-TGA-CGG-GGT-CAC-CCA-CAC-TGT-GCC-CAT-CTA-3′ and 5′-CTA-GAA-GCA-TTT-GCG-GTG-GAC-GAT-GGA-GGG-3′, which yielded a 661-bp product. Each sample also was tested for adequacy of DNAse I digestion by running a negative control PCR reaction with DNAse-treated RNA that had not been reverse transcribed. Isolated human monocytes were used as a positive control. U937 monocyte/macrophage (ATCC, Bethesda, MD) was used as a positive control.

Specific GM-CSF and GM-CSF-R identification.

Specificity of the RT-PCR products was confirmed by direct sequencing of an amplified PCR product with the Taq DyeDeoxy Terminator protocol developed by Applied Biosystems (Perkin-Elmer Corp., Foster City, CA). The labeled extension products were analyzed on an Applied Biosystems model 373 DNA sequencer.

Immunohistochemistry

Tissues were fixed in Bouin's solution for 6 h, removed to 70% alcohol for overnight incubation and then paraffin-embedded, or fixed in 10% buffered formalin overnight and paraffin-embedded. Six-micrometer sections were deparaffinized in xylene and rehydrated through a graded series of alcohols. A 30-min incubation in 10 mg/L saponin at room temperature was used for antigen retrieval. Tissues were stained using the Ventana NEXUS automated immunohistochemical staining system (Ventana, Tucson, AZ) (18). Primary antibody reactions were performed for 38 min at 37°C. Sections were lightly counterstained with hematoxylin and bluing reagent for 2 min each.

GM-CSF immunoreactivity.

A rabbit polyclonal anti-human GM-CSF antibody (Genzyme, Cambridge, MA) was used at 1:30 dilution. Absence of primary antibody and incubation with preimmune serum were used as negative controls.

GM-CSF-R immunoreactivity.

A polyclonal rabbit antibody raised against amino acids 379–396 of the carboxyl terminus of the precursor form of human GM-CSF-R (Santa Cruz, Santa Cruz, CA) was used at 1:50 dilution. Incubation with preimmune serum in the absence of antibody or preincubation with a 10-fold excess of blocking peptide was used as negative control.

Macrophage immunoreactivity.

A mouse MAb HAM-56 (Ventana) was used to identify tissue macrophages. This antibody is purchased prediluted and optimized for use on Ventana immunostainers. Liver was used as a positive control, and incubation with preimmune serum was used as a negative control.

Spinal Fluid Specimens

Thirty-nine paired samples of CSF and serum were obtained from 34 patients ranging in age from preterm neonates at 26 wk gestation to 16 y. 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 all but five samples. Five of the 34 patients had a second spinal tap. Control patients were those who underwent lumbar puncture for suspected meningitis, which was excluded by the CSF cell count and culture analysis.

ELISA

GM-CSF concentrations were measured in 39 paired samples of plasma and CSF with the Quantikine human GM-CSF immunoassay ELISA (R&D Systems, Minneapolis, MN). A standard curve was constructed in duplicate with control solutions ranging from 0 to 500 pg/mL GM-CSF, with a third standard curve constructed to confirm the reliability of the assay in human CSF. Aliquots of 100 μL of CSF or plasma were assayed. Sensitivity of the assay has been determined at 2.8 pg/mL. This assay has been tested for cross-reactivity with 76 human cytokines and with 22 nonhuman cytokines at a concentration of 50 ng/mL, and no significant cross-reactivity was noted.

RESULTS

GM-CSF.

A summary of all samples tested for GM-CSF and GM-CSF-R mRNA and protein is shown in Table 1. At 8 wk postconception, significant GM-CSF mRNA expression in somatic tissues was limited to spleen. Immunohistochemistry revealed diffuse reactivity in the spleen at 8 wk gestation. All other somatic organs tested showed no organ-specific immunoreactivity. Macrophages were strongly reactive in several organs including lung and liver.

Table 1 GM-CSF/GM-CSF-R evaluation in fetal tissues
Full size table

At 16 wk, GM-CSF mRNA expression in somatic tissues was detected in lung, spleen, adrenal, and placenta. Strong cell-specific immunoreactivity for GM-CSF was shown in macrophages of the lung, with no other organ-specific staining (Fig. 1A). Moderate cellular immunoreactivity was noted in the adrenal cortex (but not in the adrenal medulla) (Fig. 1B) and diffuse weak reactivity in the spleen, with stronger reactivity of specific hematopoietic elements (Fig. 1C). The developing kidney showed reactivity in the collecting tubules (Fig. 1D). In the placenta, syncytiotrophoblasts were strongly immunoreactive (Fig. 1E). None of the other somatic organs tested showed any reactivity to anti–GM-CSF antibody.

Figure 1
figure 1

GM-CSF immunostaining of human fetal organs. Panel A shows immunoreactivity of scattered macrophages (brown) in 17-wk fetal lung (original magnification ×200). Panel B shows diffuse immunoreactivity of an 8-wk adrenal cortex with no immunoreactivity in the adjacent adrenal medulla (×200). Panel C shows diffuse reactivity in the fetal spleen at 16 wk gestation (×100). Panel D shows the nonspecific binding of GM-CSF in the collecting system of the 17-wk kidney (×200). Placenta is shown in panel E (×100). Panel F shows immunoreactivity of astrocytes from a 32-wk brain (×400), whereas panel G shows neurons from the same patient (×400). Panel H shows a negative control.

Full size image

GM-CSF mRNA expression was detected in total RNA extracted from fetal brain, brain stem, and spinal cords taken from fetuses 8–11 wk postconception. Cultures of differentiated neurons (hNT cells), undifferentiated neuroepithelial cells (NT2 cells), and primary cultures of neurons and glia derived from fetuses of 5- to 9-wk postconceptual age also expressed GM-CSF mRNA. Immunohistochemical localization revealed weak reactivity of both neurons and astrocytes in a 20-wk fetal brain and strong reactivity of these cells in the brain of a 32-wk infant who lived 8 h before succumbing to group B streptococcal sepsis (Fig. 1, F and G ).

GM-CSF was detected in 10 of the 39 CSF samples tested (Table 2). Eight of these patients had culture-proven or presumed meningitis based on clinical presentation and CSF white blood cell counts. Three of the patients with detectable CSF GM-CSF did not have detectable concentrations in the paired plasma sample. Two of these patients had meningitis without an associated bacteremia. The third child (patient 9) did not have a proven infection. This child was born at 23 wk of gestation to a mother with chorioamnionitis. The mother was treated with antibiotics before delivery. The spinal tap was performed on d 4 of life because of a worsening clinical condition despite antibiotic therapy. Nineteen of the 39 plasma samples tested had detectable GM-CSF. Of these, nine had proven sepsis and an additional five had presumed sepsis.

Table 2 Clinical characteristics of patients who had CSF and plasma evaluated
Full size table

GM-CSF-R.

GM-CSF-R mRNA was detected in all tissues at 8 and 16 wk and in all cultured cells tested. Immunohistochemistry of fetal organs at 8 wk postconception revealed only cell-specific staining for macrophages resident in each organ. At 14 wk gestation, osteoclasts and developing granulocytes in clavicle were also immunoreactive (Fig. 2A). At 16 wk gestation, weak immunoreactivity of cells of the adrenal cortex was noted, although the adrenal medulla was not immunoreactive (Fig. 2B). In all other somatic tissues tested, only scattered resident macrophages were immunoreactive. Representative photomicrographs of 16–17-wk postconception spleen, lung, bowel, and kidney are shown in Figure 2, C - F . GM-CSF-R immunoreactivity in the placenta was also restricted to macrophages. In contrast with somatic organs, immunoreactivity in the nervous system was prevalent. We observed weak immunoreactivity of giant motor neurons in the spinal cord, moderate reactivity of dorsal root ganglion cells, and moderate reactivity of Clarke's column neurons in the thoracic spinal cord (Fig. 3A). Neurons of the basis pontis and basal pontine fibers were also immunoreactive beginning at 29 wk postconception (Fig. 3B). The third nerve nuclei were also immunoreactive with anti–GM-CSF-R antibody (Fig. 3C), as were Purkinje cells in the 32-wk postconception cerebellum (Fig. 3D). Cells in the developing neural retina were also reactive (Fig. 3E). In term postnatal brains and in adults, neurons and astrocytes continued to be immunoreactive.

Figure 2
figure 2

GM-CSF-R immunostaining of human fetal somatic organs. Panel A shows immunoreactivity (brown diaminobenzidine (DAB) reaction) of 14-wk postconception bone marrow (×100). Panel B shows diffuse immunoreactivity of the adrenal cortex with strong immunoreactivity of scattered cells and negative staining of the adrenal medulla (×200). Panels C–F show 16–17-wk spleen, lung, small bowel, and kidney, respectively (original magnification ×200).

Full size image
Figure 3
figure 3

GM-CSF-R immunostaining of neural tissues. Panel A shows GM-CSF-R immunoreactivity of neurons (arrow) in Clarke's column of an adult thoracic spinal cord (original magnification ×200). Panel B shows immunoreactivity of neurons (arrow) of the basis pontis and of basal pontine fibers (×200). Panel C shows immunoreactivity of the third nerve nuclei (arrow). Panel D shows immunoreactive Purkinje cells (arrow) in the cerebellum (×400). Panel E shows the neural retina (arrow) from a fetal eye 9 wk postconception (×200). Panel F shows a negative control in which a 10-fold excess of blocking peptide was used (×200).

Full size image

DISCUSSION

Erythropoietin and granulocyte-colony stimulating factor are hematopoietic factors that belong to the same cytokine family as GM-CSF, and they have been shown to have functional receptors in somatic and CNS tissues unrelated to hematopoiesis (15, 19–21). The present study was undertaken to determine whether GM-CSF and its receptor may also be present in nonhematopoietic organs of the developing human fetus. This issue is particularly relevant to the developing neonate, because treatment of granulocytopenia with hematopoietic factors is being increasingly considered as a therapeutic modality for septic or granulocytopenic newborns (6).

We found no evidence of mRNA expression or immunoreactivity for GM-CSF in somatic organs with the exception of spleen, which showed both. The kidney showed no evidence of mRNA expression for GM-CSF but did show some immunoreactivity in the collecting system, suggesting nonspecific binding of the protein. Specific binding is unlikely because the GM-CSF-R was not localized to the same region.

In contrast with GM-CSF, GM-CSF-R mRNA was identified by RT-PCR in all organs tested. Identification of cell-specific reactivity by immunohistochemistry showed that receptors for GM-CSF were present primarily on resident tissue macrophages, not on organ-specific cell types. The only exception to this in somatic organs was in the fetal spleen. We identified GM-CSF and GM-CSF-R mRNA and protein expression in spleen at both 8 and 16 wk. As we did not investigate splenic expression of GM-CSF and its receptor in the postnatal infant, the relevance of our finding to the newborn is not known.

The brain has been previously identified as a site of GM-CSF production. In humans, GM-CSF mRNA expression has been documented in neuroblastomas and in stimulated fetal and adult microglia and astrocytes, indicating the ability of neurons to express GM-CSF (22, 23). Murine and simian studies have shown that GM-CSF mRNA is expressed by lipopolysaccharide-stimulated cultures of astrocytes, microglia, oligodendrocytes, and neuronal cell lines and that this GM-CSF enhances the inflammatory response within the CNS (8, 24–26). GM-CSF expression has not been identified previously in unstimulated murine or human cells of the CNS, either during embryonic development or in adults (23, 24, 27). In contrast, we found both GM-CSF-R and GM-CSF mRNA and protein in normal unstimulated human CNS tissues beginning as early as 8 wk postconception and persisting throughout the period tested (32 wk postconception and in adult spinal cord). These differences in results may be due to differences in methods used. Lee et al. (23) used Northern analysis to identify mRNA expression, whereas we used RT-PCR, a more sensitive technique. Primary cultures of human neurons and astrocytes as well as neuroepithelial cells (NT2) and differentiated neuronal cell lines (hNT) also expressed mRNA for GM-CSF and its receptor.

One suggested function for GM-CSF in the postnatal brain has been immunosurveillance, as GM-CSF production by astrocytes promotes the expansion, differentiation, and function of the microglial population (8, 28, 29). This proposed role of GM-CSF in the CNS is consistent with increased GM-CSF concentrations in the spinal fluid of a high percentage of patients with meningitis, suggesting that GM-CSF production in brain is up-regulated in the presence of infection (30). We cannot, however, rule out the transfer of GM-CSF across the blood-brain barrier, as it has previously been shown that transfer of this cytokine across the blood-brain barrier occurs in a saturable manner (9). The fact that GM-CSF concentrations were higher in the spinal fluid than in the plasma of some of the patients makes this a less likely possibility. It should also be noted that although GM-CSF was most frequently found in patients with meningitis, it was also present in some patients without culture-proven infection. This may be a reflection of 1) the culturing techniques in which often small (and possibly inadequate) volumes are taken for culture in young infants, 2) false-negative cultures due to pretreatment with antibiotics, or 3) an indication that infection is not the only stimulus that induces GM-CSF in the CSF.

In the fetal brain, one role for GM-CSF may be neurodevelopmental, as GM-CSF has a trophic effect on central cholinergic neurons (11). Its effects on the microglial population may also influence the remodeling that occurs during brain development. It was interesting that many of the sites of neural immunoreactivity for GM-CSF-R observed in preterm infants (Purkinje cells, basal pontine neurons, and neurons in Clarke's nuclei) were related to movement and control of movement. This was not exclusively true, however, as proprioceptive cells in the basal root ganglion were also immunoreactive. The role of GM-CSF in the brain needs further investigation. It is possible that it may have different functions during neurodevelopment than in later life when it may have a more immunosurveillant function.

In summary, GM-CSF-R are present on macrophages throughout the developing human fetus, and also in the developing spleen, and in the developing and mature human CNS. We did not find GM-CSF-R on organ-specific cells of somatic organs such as is seen with erythropoietin and granulocyte colony-stimulating factor. We speculate that GM-CSF has important neurodevelopmental and immunosurveillant roles in the brain.