Main

Bacterial sepsis is one of the major causes of neonatal morbidity and mortality (1,2). The incidence ranges from 1 to 10 per 1000 live births (2). The mortality of neonatal sepsis is correlated to prematurity, and is particularly high in neonates with sepsis-associated neutropenia.

The increased susceptibility of the neonate for septic bacterial infections is the result of a variety of factors. They include a delayed maturation of the specific humoral and cellular immune response of neonatal B and T cells, defective activation of the complement system, and deficiencies of the myelopoetic system (3,4). The late prenatal expansion of the myelopoetic cell lineage leads to particular kinetics of proliferation of neutrophilic granulocytes. In the case of a bacterial infection an early exhaustion of the mobilizable granulocytic pool results in neutropenia in the peripheral blood (57). Studies of hematopoietic growth factors like G-CSF suggested a deficient expression and production of these mediators as the most likely cause of neonatal neutropenia (810). There is little information on IL-8 in neonatal sepsis, an important regulator of neutrophil activation and migration (11). Elevated IL-8 serum levels have been reported in adults with sepsis syndrome and in cord blood in cases of chorioamnionitis (12,13).

The interaction between the endothelial cell and the leukocyte is an essential event for the initiation of the inflammatory response. The leukocyte has to stop in the bloodstream at the site of infection, adhere to the endothelial cell, and migrate through the vessel wall. Important communicative elements between the leukocyte and the endothelial cell are the adhesion molecules on the endothelial side, such as the ICAM-1. The expression of ICAM-1 is induced by proinflammatory cytokines, such as TNF-α or IL-1β. In adults, the soluble form of ICAM-1 (sICAM-1) is elevated in a variety of disease states. In the neonate, elevation of sICAM-1 has been reported to be specific for septic bacterial infections (14). Buzby et al. (15), however, showed a decreased sICAM-1 mRNA expression in human endothelial cells derived from the umbilical vein when compared with that from the adult aorta.

The expression of the cytokines IL-6, TNF-α, and IL-1β in sepsis has been studied extensively in animal models and in clinical investigations in adults. They are thought to be the key mediators in the pathogenesis of the sepsis syndrome (1618). The typical clinical signs of elevated plasma levels of these mediators, such as fever or a prompt rise in acute phase proteins, are, however, generally absent in the septic neonate, especially in cases of early onset sepsis. This well known observation might suggest a different pathophysiologic course of cytokine secretion in the very young.

Early onset sepsis is defined as sepsis in the newborn beginning within the first 4 d of life (19). It is thought to develop as the result of an ascending infection around the time of delivery from maternal rectovaginal pathogens and subsequent chorioamnionitis, or, rarely, by the hematogenous route. To study the pathophysiologic role and the potential diagnostic impact of cytokines in early onset sepsis, we determined the time course of plasma levels and gene expression of six of these mediators. We investigated G-CSF, involved in neonatal hematopoiesis and neutrophil response to bacterial infections; TNF-α and IL-1β, as representative cytokines of the first line immune response; and IL-6, IL-8, and sICAM-1, as second line after the expression of TNF-α, and IL-1β and, thus suggested to be more consistently elevated.

METHODS

Study population. The study was performed at the University Children's Hospital and the University Hospital of Obstetrics and Gynecology in Freiburg, Germany, from June 1993 to July 1994. The hospital is a tertiary care center that serves an urban and rural population of roughly 400,000 inhabitants. The birth rate at the hospital in 1993 was 979 (171 premature newborns) and in 1994 1062 (172 premature newborns). The overall birth rate in the region is approximately 5000 live births/y. During the study period 543 newborns, 184 of which were premature, were admitted to the pediatric hospital.

Patients enrolled in the study were comprised of mature newborns admitted to the Children's Hospital within the first 72 h of life to rule out sepsis; premature infants admitted directly after birth for reasons of immaturity, and, therefore, estimated to be at increased risk for infection; and a healthy control group of mature infants delivered spontaneously. Cord blood samples of premature infants were collected without knowledge of whether the patient was infected. With the exception of the control group, cord blood samples of mature newborns were taken only if sepsis was suspected by clinical signs or maternal or peripartal risk factors. When possible, venous blood samples of the respective mothers were obtained at the time of delivery. For patients who were admitted within the first 72 h of life, to rule out sepsis, and who were neither premature nor had had a suspicion of sepsis at the time of delivery, cord blood and maternal samples were not available.

The medical records of all patients and their mothers were evaluated retrospectively by two senior staff members without knowledge of the results of the cytokine plasma levels of the prospectively collected blood samples. Data were recorded by a standardized protocol including maternal history, antibiotic treatment, treatment with dexamethasone, chorioamnionitis, rupture of membranes, type of delivery, color of amnion fluid, birth weight, gestational age, Apgar score, pH of cord blood, age at time of diagnosis of sepsis, respiratory status, fever, bacteriologic findings, CRP, CBC, and radiologic findings.

In the retrospective evaluation, sepsis was defined as the substantial clinical suspicion within the first 96 h of life based on the symptoms of poor peripheral circulation, respiratory distress, cyanosis, lethargy, irritability, apneic spells, tachypnea, fever, bradycardia, tachycardia, poor feeding, and 1) a blood culture was positive, or 2) at least three of the following criteria were met: (i) CRP > 2 mg/100 mL within 48 h after onset of clinically suspected sepsis, (ii) pneumonia as diagnosed on x-ray or by microscopic or cultural evidence in tracheal aspirate, (iii) gastric aspirate or urine with microscopic or cultural evidence of bacterial infection or positive latex agglutination test in urine for group B streptococci, (iv) proportion of immature (bands and less mature forms) to total neutrophils > 0.2 documented in any CBC within 48 h after clinically suspected sepsis, or (v) maternal fever or antibiotic treatment within 48 h of delivery or premature (>48 h) rupture of membranes before delivery or histopathology of amnionitis of the placenta or cord. According to this evaluation patients were classified into the following groups:

Group A. Neonates with (early onset) sepsis according to the criteria 1) or 2).

Group B. Neonates with initial clinical suspicion of sepsis who retrospectively did not meet the criteria 1) or 2). In this group an infection could not be confirmed.

Group C. The control group of mature, healthy neonates delivered spontaneously without maternal infection and risk factors or clinical or laboratory suspicion of sepsis.

The study design implied that, for a small number of premature infants, who at no time had a clinical suspicion of sepsis, cord blood samples were collected but not analyzed.

All patients with clinically suspected sepsis were treated initially by a standard antibiotic regimen of piperacillin and tobramycin. Before therapy, blood cultures (generally 1-2 mL of inoculum), urine samples, cutaneous swabs, and gastric aspirates were taken and cultured according to standard procedures. In addition, urine samples were analyzed for the presence of group B streptococcal antigen (Slidex méningite Strepto B, bioMérieux, Marcy-l'Etoile, France).

Parental informed consent was obtained for every patient before entry in the study. The study was approved by the local ethics committee.

Blood samples. Cord blood samples were obtained within 15 min of delivery. Samples were collected aseptically by squeezing or puncture of the cord. There was no significant difference in cytokine plasma levels between cord blood samples obtained by squeezing or puncture (data not shown). Blood samples other than cord blood were obtained by venipuncture or drawn from an indwelling arterial or venous line. For patients without available cord blood, the time of taking the initial blood cultures was the time of collecting the first blood sample for the study. All first blood samples were obtained at or within 24 h of delivery. Follow-up samples were obtained when blood taking was indicated for clinical reasons. There were no additional venipunctures for study purposes. Follow-up samples, as well as CBCs from group C were not available. Blood samples were grouped according to three time periods. First blood samples were obtained within 24 h after birth (group I), second samples between 24 and 48 h (group II), and third samples after more than 48 h after birth (group III). Blood samples of infants who were enrolled after 24 h of age were assigned to the respective time group.

Samples were collected in plastic tubes treated with EDTA. Plasma was separated from the blood cells within 30 min by centrifugation at 1000 × g for 10 min, aliquoted, and stored in plastic tubes at -30°C until further use. In the remaining whole blood, erythrocytes were hypotonically lysed by endotoxin-free distilled H2O, and leukocytes were washed several times with PBS. The remaining pellet was resuspended in a lysis buffer consisting of 4 M guanidine-isothiocyanate (Merck, Darmstadt, Germany), 25 mM sodium citrate (pH 7), 0.5% lauroylsarcosine (Sigma Chemical Co., Deisenhofen, Germany), and 100 mM 2-mercaptoethanol (Sigma Chemical Co.), and stored at -30°C until further use.

Measurement of CBC and CRP. The CBC (Sysmex M2000, München, Germany) included a white blood cell count, a platelet count, and a Hb value. The white blood cell count was corrected for the presence of nucleated red blood cells, and a 100-cell differential count was obtained. An I/T ratio was calculated as the ratio of immature neutrophils (band forms, metamyelocytes, myelocytes) to total neutrophils. Reference ranges of neutrophils for mature and premature newborns with respect to gestational age as well as postnatal age were used as indicated by Manroe et al. (20) and Mouzinho et al. (21). CBCs were obtained at the time of the initial plasma sample and during the first 96 h of life. Plasma CRP concentrations were measured by the nephelometric method (Array protein system, Beckman Instruments, München, Germany).

Cytokines. Frozen aliquots of plasma were thawed on ice at the time of analysis. Plasma levels were measured by double sandwich enzyme immunoassay technique using commercial kits specific for human cytokines G-CSF, TNF-α, IL-1β, IL-6, IL-8 (Quantikine, R&D Systems Europe, Abingdon, United Kingdom), and sICAM-1 (DPC, Cellfree ICAM-1, T Cell Diagnostics, Cambridge, UK), respectively. Detection limits of the assays as indicated by the manufacturers were 4.4 pg/mL for TNF-α, 3.9 pg/mL for IL-1β, 0.35 pg/mL for IL-6, 10.9 pg/mL for G-CSF, 4.7 pg/mL for IL-8, and 0.3 ng/mL for sICAM-1. The plasma volumes necessary for the investigations were the following: 200 µL for TNF-α, 200 µL for IL-1β, 100 µL for IL-6, 100 µL for G-CSF, 100 µL for IL-8, and 25 µL for sICAM-1. Duplicate measurements were performed for each plasma sample. Samples were diluted before analysis if necessary. Dilution buffer was provided by the manufacturer. For analysis of the time course of cytokine levels, parameters were grouped according to the postnatal age of the patient (<24 h, 24-48 h, >48 h). If more than one plasma sample of a patient within a time period was available, only the first sample was evaluated. Due to limitations by the sample volumes, follow-up plasma levels of TNF-α, and IL-1β were not measured

PCR-assisted mRNA amplification. Total RNA was prepared using a phenol-chloroform extraction method (22). RNA precipitates were pelleted at 4°C, washed once with 75% ethanol in diethylpyrocarbonate-treated distilled H2O, and repelleted at 12 000 × g for 10 min. Pellets were resuspended in 40 µL of diethylpyrocarbonate-distilled H2O. For reverse transcriptase reaction, an aliquot of 8.5 µL of RNA was incubated with 2 µL of oligo(dT) (10-18 mer, 25 ng/µL) at 65°C for 10 min. After cooling on ice, the mixture was incubated with 4 µL of 5× reverse transcriptase buffer (250 mM Tris-HCl, pH 8.3, 375 mM KCl, 15 mM MgCl2), 2 µL of DTT (0.1 M), 2 µL of dNTP (10 mM), 0.5 µL of RNase Inhibitor (Boehringer, Mannheim, Germany), and 1 µL of Moloney murine leukemia virus reverse transcriptase (SuperscriptR, Gibco BRL, Life Technologies, Eggenstein, Germany) in a final volume of 20 µL for 60 min at 37°C, then heated to 95°C for 10 min and cooled on ice. The resulting cDNA was stored at -30°C until further use.

Five microliters of cDNA were amplified in the presence of 10 pM final concentration of the respective primer pairs, 200 µM dNTP, 1 U of Taq polymerase (Pharmacia, Freiburg, Germany), and 5 µL of 10× PCR buffer (500 mM KCl, 15 mM MgCl2, 100 mM Tris-HCl, pH 9.0) in a final volume of 50 µL. The reaction mixture was overlaid with the same volume of light mineral oil, and PCR was performed in a DNA thermal cycler (OmniGene, Hybaid, Teddington, England) for 30 cycles in all experiments: 60-s denaturation at 94°C, 60-s annealing at 60°C, and 60-s extension at 72°C. The reaction product was visualized by subjecting to electrophoresis 20 µL of the reaction mix at 80 V in 2% agarose in 0.5× Tris-borate-EDTA buffer. The expected sizes of the PCR products of the cellular cDNA were 237 bp for β-actin, 268 bp for β2-microglobulin, 263 bp for IL-1β, 427 bp for TNF-α, 260 bp for IL-6, 247 bp for IL-8, and 290 bp for G-CSF. Specificity of the amplified bands was validated by their predicted size. PCR-assisted mRNA amplification was repeated at least twice for each sample. Cellular samples were considered to be evaluable, if mRNA of β-actin was detectable.

Semiquantitative PCR. To compare separate samples quantitatively, PCR was performed on target cDNA serially diluted 4-fold using primers specific for β-actin, β2-microglobulin, IL-1β, TNF-α, IL-6, IL-8, and G-CSF in the presence of a constant amount of competitor control DNA consisting of 5′ and 3′ primer sequences in tandem array cloned on the plasmids pQA-1 and pQB-3, kindly provided by David Shire, Sanofi Elf Biorecherche (23). The size of the standard amplicon of the plasmid pQA-1 was 370 bp and of the plasmid pQB-3, 410 bp. During co-amplification, target and control DNA compete for the primers, and the amount of PCR product is proportional to the amount of input cDNA. Samples were determined to contain similar amounts of amplifiable cDNA when the dilution at which equally dense bands were visualized during electrophoresis was the same for both samples tested. Competitive PCR on the respective cytokines was evaluated in comparison with parallely performed PCR on β2-microglobulin (pQA-1), which has been reported to be abundantly expressed in lymphocytes and monocytes (24).

Statistical analysis. The differences between the median values of cytokine plasma levels of the three groups were analyzed by the Kruskal-Wallis test (one-way ANOVA). p values indicated for the differences between the individual groups are the results of the post hoc analysis to show which groups were significantly different (Wilcoxon-Mann-Whitney test). If only two groups were compared, the Wilcoxon-Mann-Whitney test was applied. All values are expressed as medians, quartiles, and ranges. For analysis of the differences between the clinical characteristics and the mRNA gene expression, the χ2 test was used. Analyses were performed using the SPSS software package (SPSS, Germany). Differences were considered significant at p < 0.05. The sensitivity was calculated as the percentage of patients defined to have sepsis who had a cytokine plasma level above the given cutoff. The specificity was calculated as the percentage of patients not meeting the criteria of sepsis who had plasma levels below the cutoff value. The predictive value of a positive test (PV+) gives the percentage of patients with plasma levels above the given cutoff who met the criteria of sepsis, the predictive value of a negative test (PV-) gives the percentage of patients with plasma levels below the cutoff who did not meet the criteria of sepsis.

RESULTS

Study population. A total of 167 neonates were enrolled in the study. Thirty-one infants had to be excluded because they developed late onset sepsis. Thus, 136 neonates were analyzed. Thirty-five newborns were classified belonging to the septic group A, 6 of them according to the criterion 1), 29 to the criterion 2). Sixty-six newborns were classed to the nonseptic group B, and 35 to the control group C. When calculating the incidence of sepsis by referring the septic newborns to all newborns born at the department of obstetrics, a high rate of about 3% is found. This is due to the fact that the hospital is a tertiary care center, where pregnancies at risk are referred to from peripheral hospitals in the region. When referring the sepsis incidence to the overall birth rate in the region during the study period (about 5000 live births), the rate is about 0.7%, which is in accordance with data from the literature (2).

Cord blood samples were available for 16 patients of group A, 43 patients of group B, and 35 patients of group C. First blood samples were available in 34 patients of group A, 61 of group B, and 35 of group C. There were 23 second and 31 third samples in group A; 15 second and 31 third samples in group B. Blood samples from 47 mothers could be obtained. In 47 cases paired samples of peripheral blood from the mother and cord blood from the newborn were available, 15 times in group A, 32 times group B, and none in group C patients.

In the septic group A, subgroups according to the inclusion criteria 1) and 2) were compared with respect to clinical features, hematologic parameters, and plasma levels of G-CSF, TNF-α, IL-1β, IL-6, IL-8, sICAM-1, and CRP in the initial blood sample and in the cord blood. No statistical differences were found for the median value of any parameter. Therefore, both subgroups were combined in group A for further evaluation. When comparing the clinical characteristics of the septic neonates in group A and the nonseptic neonates in group B, there was a predominance of premature newborns in the nonseptic group B (Table 1), because prematurity was one of the inclusion criteria. In addition, maternal fever (20% versus 5%), greenish color of amnion fluid (46% versus 17%), and histologically proven chorioamnionitis (31% versus 15%) were seen more frequently in group A patients (p < 0.05) (Table 1). Likewise, duration of antibiotic treatment (median 7 versus 2 d) was longer in group A than in group B patients (p < 0.05). Between groups B and C, there was a significant difference for gestational age and birth weight (Table 1), as expected by the definition of the groups. These differences were eliminated by evaluating group B combined with group C in comparison with group A. There were no differences between group A and B in the frequency of premature rupture of membranes, maternal treatment with antibiotics or dexamethasone, maternal edemaproteinuria-hypertension gestosis, premature rupture of membranes, mode of delivery, gender, differences in Apgar values (at 5 min), duration of hospitalization, days in the intensive care unit and days of mechanical ventilation, as well as outcome (data not shown). When comparing clinical characteristics of premature and mature newborns within group A, there was a predominance of preterm infants only in the frequency of premature rupture of membranes and the presence of chorioamnionitis (data not shown).

Table 1 Clinical characteristics of the neonates in the septic group A, the nonseptic group B, and the control group C

Bacteriologic findings. In the septic group A, the organisms detected in blood, urine, and gastric aspirates were group B streptococci (16 times; in blood 6 times and in the urine samples 10 times) and Escherichia coli (one time in the urine samples). Twice, Gram-positive cocci were seen microscopically in the gastric aspirate, but were not cultured. In the nonseptic group B, no pathogens were cultured in blood, urine, or gastric aspirates. For patients in group A, cytokine plasma levels, CRP values, and the incidence of neutropenia did not differ between newborns with proven group B streptococci infection and those with isolation of other or no pathogens (data not shown).

Neutropenia. Neutropenia was noted in 9 of 34 patients in group A, and in 7 of 49 patients in group B (26% versus 14%, p = 0.17). Five of 26 patients in group A and 4 of 45 patients in group B displayed neutropenia in the first available blood count (19% versus 9%, p = 0.27). Comparing cytokine plasma levels from samples with and without neutropenia, there was no difference in any parameter analyzed including G-CSF (data not shown). Similarly, there was no difference in cytokine plasma levels of the first sample of patients who were or became neutropenic during the following 96 h compared with those who did not become neutropenic (data not shown).

Cytokine plasma levels. Newborns of the septic group A had significantly higher plasma levels of G-CSF, TNF-α, IL-1β, IL-6, and IL-8 in the first evaluable plasma sample than those of the nonseptic group B or the control group C (Table 2). Plasma levels of sICAM-1, in contrast, did not differ between newborns of group A and those of group C, but were higher in infants of group A than of group B (Table 2). There was no difference in the plasma levels of G-CSF, TNF-α, IL-1β, IL-6, and IL-8 between groups B and C. However, plasma levels of sICAM-1 were significantly higher in patients of group C than in group B (p < 0.0001). This result might be explained by differences in maturity and mode of delivery between the groups. sICAM-1 levels were significantly higher in mature than in premature newborns when analyzing the combined group B and C (p < 0.0001). Likewise, in infants spontaneously delivered, sICAM-1 levels were higher than those in infants born by cesarean section when analyzed for the combined group B and C (p = 0.004) and for group B separately (p = 0.02). Within the septic group A, IL-1β was more elevated in premature newborns (median 142 versus 20.5 pg/mL; p = 0.04*). There was no difference for any parameter when analyzed separately for septic patients born spontaneously or be cesarean section, and for patients born to mothers treated with antibiotics before birth or not. Within the nonseptic group B, there was neither a statistical difference between premature and mature newborns, newborns born spontaneously or by cesarean section, not between newborns born from mothers treated with antibiotics or not. Gestational age, maternal gestosis, or maternal treatment with dexamethasone did not correlate with plasma levels when analyzed in groups A, B, and C separately, in group B and C or in the whole study population. Neonatal sepsis could be predicted with a high sensitivity and specificity by either parameter: G-CSF, TNF-α, IL-1β, IL-6, or IL-8, from blood samples taken within 24 h after birth (first blood samples) (data not shown).

Table 2 Plasma levels of cytokines, sICAM-1, CRP and neutrophil counts in the first blood samples

Similar results were obtained when analyzing cord blood samples only. Neonatal sepsis could be predicted also from cord blood plasma levels with a high sensitivity and specificity by either parameter; G-CSF, TNF-α, IL-1β, IL-6, or IL-8 (Table 3). In group A, sICAM-12 was significantly elevated in cord blood when compared with group B, but not when compared with group C. Again, the difference between sICAM-1 levels of groups B and C, as well as between premature and mature newborns within the combined group B and C was highly significant. Within the septic group A, however, there was no difference between premature and mature infants in cord blood levels of either sICAM-1, G-CSF, TNF-α, IL-6, or IL-8.

Table 3 Sensitivity and specificity of cord blood cytokine plasma levels for the presence of neonatal sepsis

Maternal cytokine plasma levels. To evaluate whether neonatal sepsis can be predicted from maternal cytokine plasma levels, blood samples of the respective mothers from the time of delivery were compared. No significant differences were found in plasma levels of G-CSF, IL-1β, IL-6, IL-8, or sICAM-1 in mothers of septic and nonseptic infants (data not shown), whereas TNF-α levels (median 42.4 versus 4.4 pg/mL, p = 0.004) and CRP levels (median 1.7 versus 0.6 mg/100 mL, p = 0.04) were higher in mothers of septic infants. There was no difference in cytokine levels between mothers of mature and premature newborns, whereas mothers of mature newborns had significantly higher values for CRP than mothers of premature infants (median 4.6 versus 0.6 mg/100 mL, p = 0.02). Similarly, there was no difference in cytokine plasma levels between mothers delivering spontaneously or by cesarean section (data not shown).

Maternal versus cord blood levels. To confirm the endogenous cytokine production by the neonate, cytokine plasma levels in cord blood and the corresponding maternal blood samples were compared. Cord blood levels of G-CSF, IL-1β, IL-6, and IL-8 were significantly higher in septic infants than in their respective mothers (Fig. 1), CRP, however, was more elevated in the mothers than in their infants. There was no significant difference in TNF-α and sICAM-1 values between mothers and infants. Comparing cord blood and maternal values in the combined nonseptic group (group B + C), there was no difference in either G-CSF, TNF-α, IL-1β, IL-6, or IL-8 values (data not shown). sICAM-1 (p = 0.0003) and CRP (p < 0.0001) were, however, significantly more elevated in maternal plasma than in their newborns, similar to the group A samples.

Figure 1
figure 1

Cytokine plasma levels in cord blood (A) and maternal blood (AM) of patients of the septic group A presented as box plots (25). The median of the batch is marked by the center horizontal line of the central box. The lower and upper hinges (H) comprise the edges of the central box representing the interquartile range. The H spread is the absolute value of the differences between the values of the two hinges. The whiskers () show the range of values which fall within 1.5 H spreads of the hinges. Values outside the inner fences (±1.5 H spread) are plotted with asterisks (*). Values outside the outer fences (±3 H spread) are plotted with empty circles (). Differences were considered significant at p < 0.05; CRP (p = 0.008), sICAM-1 (p = 0.23), G-CSF (p = 0.0001), IL-6 (p = 0.001), IL-8 (p = 0.0001), and IL-1β (p = 0.004), and TNF-α (p = 0.59).

Time course of plasma levels. For septic patients in group A, there was a highly significant decrease in the plasma levels of G-CSF, IL-6, and IL-8 from the initial value to those obtained more than 48 h later (p < 0.0001, p < 0.0001, and p = 0.004, respectively) (Fig. 2). The fastest decrease in cytokine levels was noted for IL-6 with a significant drop within the first 48 h (p < 0.01), whereas the CRP increased during the same time period (p < 0.0001). In contrast to all other cytokines analyzed, the plasma levels of sICAM-1 increased when comparing time groups I and III (p = 0.0008).

Figure 2
figure 2

Time course of cytokine plasma levels in patients of the septic group A presented as box plots. For analysis of the time course of cytokine levels, parameters were grouped according to the postnatal age of the patient: group I (<24 h), group II (24-48 h), group III (>48 h). If more than one plasma sample of a patient within a time period was available, only the first sample was evaluated. The median of the batch is marked by the center horizontal line of the central box. The lower and upper hinges (H) comprise the edges of the central box representing the interquartile range. The H spread is the absolute value of the differences between the values of the two hinges. The whiskers () show the range of values which fall within 1.5 H spreads of the hinges. Values outside the inner fences (±1.5 H spread) are plotted with asterisks (*). Values outside the outer fences (±3 H spread) are plotted with empty circles (). Differences were considered significant at p < 0.05. Indicated within the figure are the differences between the three time groups as analyzed by the Kruskal-Wallis test (one-way ANOVA). The differences between the individual groups (Wilcoxon-Mann-Whitney test) are the following. CRP: I vs II (p < 0.001), I vs III (p = 0.36), II vs III (p < 0.0001). sICAM-1: I vs II (p = 0.27), I vs III (p = 0.008), II vs III (p = 0.08). G-CSF: I vs II (p = 0.06), I vs III (p < 0.0001), II vs III (p = 0.005). IL-6: I vs II (p = 0.008), I vs III (p < 0.0001), II vs III (p < 0.0001). IL-8: I vs II (p = 0.14), I vs III (p = 0.004), II vs III (p = 0.04).

Cytokine mRNA levels in total nucleated blood cells. mRNA expression of G-CSF, TNF-α, IL-1β, IL-6, and IL-8 was measured in total nucleated blood cells of the same blood samples in which cytokine plasma levels had been measured. IL-1β and IL-8 mRNA expression was seen in 82 and 45% of patients in group A, 71 and 42% in group B, and 63 and 75% in group C, respectively. mRNA expression of IL-6 was detectable in 27% of group A patients, 8% of group B patients, and 13% of group C patients. TNF-α mRNA was detectable in only 27% of group A patients and none of groups B and C. With the exception of TNF-α (p <0.05), the difference in the frequency of detectable gene expression between groups A, B, or C was not significant. G-CSF mRNA was not detectable in any sample, ever after PCR with 40 cycles was performed. As determined by semiquantitative PCR, in a few patients of group A (n = 3), an about 60-fold stronger gene expression of TNF-α, IL-1β, and IL-6, and IL-8 than in the nonseptic and control patients was observed, whereas β2-microglobulin expression was equally high (data not shown). There was no difference in median cytokine levels between patients with detectable or not detectable gene expression. There was also no difference in gene expression between newborns born from mothers treated with antibiotics before birth and those who were not, or between premature and mature newborns (data not shown).

DISCUSSION

Previous studies in adults and children have demonstrated the important role of TNF-α, IL-1β, and IL-6, and IL-8, in the pathogenesis of the sepsis syndrome. In the last few years, additional studies have been performed in neonates. Although a defective cytokine production of neonatal cells has been observed in vitro, in vivo studies did not uniformly confirm these findings. Some investigators described a reduced cellular capacity of IL-6 production in neonates (2628), but others observed high plasma levels of IL-6 in septic neonates (2934). Likewise, a reduced capacity of neonatal mononuclear cells to produce G-CSF has been described in vitro (8,35), whereas elevated G-CSF plasma levels were seen in normal full-term neonates (36) and neonates with bacterial infections (37,38). Contradictory results have also been reported for TNF-α, IL-1β, and IL-8 (13,3941). In this study, we demonstrate highly elevated plasma levels and, to a lesser extent, gene expression of a panel of cytokines including G-CSF, TNF-α, IL-1β, IL-6, and IL-8 in newborns with early onset sepsis. Plasma levels of the cytokines allowed the diagnosis of sepsis with a high sensitivity and specificity. The study was focused on neonatal early onset sepsis, because the pathogenesis of early onset sepsis is thought to be uniformly due to an infection of the newborn around the time of delivery, whereas sepsis later in life is more variable in its origin. The significant drop in plasma levels of G-CSF, IL-6, and IL-8 in septic neonates after birth support the hypothesis that the infection and the stimulation of the immune system occurs before birth. At birth, a high gene expression of the respective cytokines was detectable in only a few patients. Recently, a similar observation was described in chorioamnionitis with undetectable mRNA levels of IL-6 in mononuclear cells of cord blood, whereas plasma levels of IL-6 were elevated (28). The authors suggest that decidual or maternal cells are the origin of IL-6 production. The results of this study, demonstrating a highly significant difference between maternal and neonatal plasma levels of proinflammatory cytokines in septic infants, indicate that placental transfer might contribute only to a small extent to the newborn's plasma level. One might speculate that the maximum of cytokine gene expression occurs before birth at the beginning of the infection, and thus has decreased to levels only rarely detectable at the time of birth. Moreover, besides mononuclear blood cells, umbilical endothelial cells can produce high amounts of IL-6 after stimulation by IL-1β (42), and are likely to contribute considerably to the neonatal cytokine production.

The granulocyte is a critical component of the host defense against acute bacterial infections. Factors such as G-CSF that enhance production, mobilization, or bactericidal function of the neutrophil are of special interest in neonatal sepsis. Although highly elevated G-CSF plasma levels have been described in septic adults (43), neonates with bacterial infections were found to have moderately elevated (37,38) or normal plasma levels (44). The reference values for G-CSF in healthy neonates (median 112 pg/mL) given by Ishiguro et al. (45) correspond to those of our control group C (median 86.2 pg/mL) and nonseptic group B (median 88.8 pg/mL). In this study, however, septic neonates presented with excessively elevated endogenous G-CSF levels in cord blood; levels were higher than peak levels induced by G-CSF treatment for chemotherapy-induced neutropenia (data not shown). Although placental transfer of maternal G-CSF has been demonstrated in fetal rats (46), our data argue against a transfer of significant amounts of cytokines in man. If neutropenia was seen in septic neonates, it generally developed during the first hours of life, when G-CSF levels decreased. Therefore, G-CSF expression seems to be induced by the infectious process itself rather than by neutropenia. Moreover, high endogenous G-CSF levels were not able to prevent neutropenia in the septic neonate. In view of these findings it appears questionable whether septic newborns will benefit from postnatal G-CSF treatment in terms of the expansion of the neutrophil proliferative pool. Gillan et al. (37), however, demonstrated neutrophilia and functional activation of neutrophils after administration of G-CSF in neonates with presumed sepsis. In the animal, G-CSF administration to pregnant rats had a prophylactic effect for the newborn postnatally infected with group B streptococci (48). The underlying mechanism is most likely the expansion of myelopoesis with increasing numbers of peripheral neutrophils, but other mechanisms of host defense enhancement may also be involved. Görgen et al. (49) showed that G-CSF was capable of reducing TNF-α expression in macrophages in vitro and suggested a negative feedback mechanism.

Adhesion molecules are essential intercellular communicative elements in inflammatory processes. In neonates, sICAM-1 has been reported to be a good marker for sepsis (14). We could not confirm these results, although an increase of sICAM-1 levels similar to that of CRP was observed in the septic newborns during the first days of life. The sensitivity of sICAM-1 for early onset sepsis was, however, low. Our findings indicate either a slower kinetics of ICAM-1 expression compared with other cytokines, or a delayed shedding of the soluble molecule. The lower plasma levels of sICAM-1 in premature compared with mature infants may indicate that sICAM-1 production depends on the maturation of the immune system. It might also contribute to the increased susceptibility of the premature newborn for infection.

In conclusion, our data indicate that, in contrast to CRP and sICAM-1, cord plasma levels, but not the presence of mRNA expression of G-CSF, TNF-α, IL-1β, IL-6, and IL-8 can predict neonatal early onset sepsis with a high sensitivity and specificity. Because the predictive value does not depend on maturity, it will be also applicable to the premature newborn at high risk for infection. Cytokine plasma levels seem to be the result of the endogenous neonatal production which, in case of early onset sepsis, is stimulated by an infection that occurs before birth around the time of delivery. However, cell types other than blood cells are likely to be involved in cytokine secretion. Group B Streptococcus, the most important pathogen in neonatal sepsis, proves to be a competent stimulus of the neonatal immune system in terms of induction of cytokine expression.