Main

Much work has been done in the past decades to identify various risk factors for bilirubin encephalopathy or kernicterus. As with many other items, the role of infection in increasing the risk of bilirubin encephalopathy still remains controversial because the data available are largely from observational studies only. Reports from in vivo investigations on the risk of bilirubin encephalopathy among newborns with sepsis are not consistent. Convincing experimental evidence of their relationship is still lacking.

In animal models of bacterial meningitis endotoxin from the outer membrane of gram-negative bacteria has been linked to an increase of the permeability of the BBB; Temesväri et al.(1) observed a dose-dependent response in decreasing the time needed for opening the BBB after intracisternal injection of Escherichia coli O111 B4 LPS, which was the toxic ingredient of endotoxin, in newborn piglets. However, Hansen et al.(2) were unable to show a similar effect with endotoxin although an increase of the net accumulation of bilirubin in the brains of adult rats was observed.

It is now clear that apart from causing direct injury to certain cells, endotoxin also initiates in animals a cascade of inflammatory responses involving the release of various cytokines. Among them TNF-α plays a more important and direct role. TNF-α and other cytokines have also been found to be involved in a similar inflammatory response in humans(3).

Our study has demonstrated the combined effects of LPS and TNF-α in enhancing the bilirubin cytotoxicity in a mouse fibroblast cell culture which is the TNF-α target cell line, L-929 (ATCC, CCL-1).

METHODS

Reagents. Bilirubin (Sigma Chemical Co., St. Louis, MO, B4126) was dissolved in 0.1N NaOH and diluted with deionized distilled water to make up a 4280 µmol/L concentration. Stock solutions of different B/A molar ratios were obtained by fixing the albumin concentration (Sigma, A1653) at 4 g/dL (667 µmol/L) and varying the volume of bilirubin solution added. The pH was corrected to 7.4. A stock solution of MTT was prepared by dissolving MTT (Sigma, M2128) in PBS (pH = 7.2) at 1.5 mg/mL. AMD (Sigma, A9415) was dissolved in PBS to 20 µg/mL and stored at -20°C. Phenol-extracted LPS from E. coli 0111:B4 (Sigma, L2630) was stored as 1 mg/mL in PBS at 4°C. Recombinant human TNF-α (Genzyme Diagnostics, Cambridge, MA, TNF-H) was stored at -70°C. One unit per mL of recombinant human TNFα corresponds to 100 pg/mL. Horseradish peroxidase type I (Sigma, P8125) was prepared by dissolving 1 mg of this enzyme in 1 mL of distilled water. Analytic grade of ethyl hydrogen peroxide of 5% (wt/vol) was obtained from Polysciences Inc. (Warrington, PA).

Cell culture. In a pilot experiment, we have tested the cytotoxic effects of TNF-α and bilirubin on several human cell lines, including neuroblastoma cell (ATCC, HTB-10, SK-N-MC), glioblastoma cell (ATCC, CRL 1690, T98G), umbilical vein endothelial cell (ATCC, CRL 1730, HUV-EC), and liver cell (ATCC, CCL 13, Chang Liver, Hela markers). Bilirubin cytotoxicity could be seen in all the cell lines tested. TNF-α alone, however, did not demonstrate an effect even at concentration up to 1000 U/mL. Therefore, we elected to use mouse fibroblasts L-929, which is a well-known target cell line for TNF-α(4), in the study. The cells were grown in Eagle's minimal essential medium with penicillin (100 U/mL), streptomycin (100 µg/mL), glutamine (2.5 mmol/L), and 10% fetal calf serum and incubated at 37°C in 5% CO2. Cells were grown until they were subconfluent, then trypsinized, washed, and resuspended in fresh Eagle's minimal essential medium. A concentration of 3 × 105 cell/mL of the cultured cells was prepared. A total of 100 µL of the cell suspension was seeded into each well of a 96-well microplate with a multichannel pipette. The cells were allowed to attach onto the plate through 24-h incubation.

Testing the effect of TNF-α and LPS on L-929 cells. It is known that AMD, a DNA and RNA synthesis inhibitor, could greatly increase the sensitivity of TNF-α analysis(4–7). We have therefore also included AMD in some of our TNF-α studies. A total of 100 µL per well of solutions containing either LPS, TNF-α, TNF-α + AMD, or TNF-α + LPS in different concentrations was added to the wells as prepared above to replace the old medium and incubated for another 24 h. Each test was done in triplicate and each experiment was repeated on several days.

Testing the effect of TNF-α and bilirubin on L-929 cell. Studies were designed to mimic two coexisting clinical situations of hyperbilirubinemia and sepsis with alternating sequential conditioning of hyperbilirubinemia and sepsis. In one set of experiments bilirubin was incubated with the cells for 24 h before adding TNF-α for another 24-h incubation. In another set, cells were precultured for 24 h with TNF-α before bilirubin was added for further incubation for 24 h. In either case, cell cultures with bilirubin alone or with TNF-α alone were also processed for comparison. Actinomycin D of 100 ng/mL was added to enhance the effect of TNF-α.

Testing the effect of TNF-α, LPS, and bilirubin on L-929 cells. Cells were incubated with three different B/A molar ratios of 0, 1.4, or 1.8, respectively, for 24 h, then with the bilirubin solution remaining, another solution of 100 µL/well containing one of the following was introduced: LPS, TNF-α + LPS, TNF-α + AMD, TNF-α + LPS + AMD. The cells were further incubated for another 24 h before measurement of the cell viability. The concentrations of these reagents were TNF-α 5 U/mL, LPS 10 µg/mL, and AMD 20 ng/mL.

Modified MTT method to measure cell viability. We have developed a modified MTT method(8) based on Mosmann's method(9) and Denizot and Lang's modification(10). After exposure to bilirubin, LPS, and TNF-α, the reaction medium was removed and 90 µL of fresh medium and 10 µL of MTT were added to the cell culture and incubated for 4 h to allow the viable cells to cleave the MTT into the purplish blue formazan; 100 µl/well of 0.04N HCl-isopropanol was then added to dissolve the MTT crystals. The supernatant containing the dissolved formazan was transferred into a new plate and the optical density was measured with an ELISA reader (SLT Labinstruments Ges. m.b.H., Salzburg, Austria) at 570 nm and a reference wavelength of 630 nm. The intensity of the optical density is proportional to the number of viable cells(9).

Bilirubin-albumin titration studies with horseradish peroxidase oxidation method. We have established in our laboratory an automated peroxidase oxidation method for measuring the concentration of free bilirubin(11). A microprocessor controlled Cobas Fara analyser (Diagnostica Division, F. Hoffmann-La Roche & Co. Ltd., CH-4133, Basel, Switzerland) was used and the enzyme kinetic mode for the peroxidase oxidation method was used. TNF-α, LPS, and AMD were added to 2-mL aliquots of albumin solution. Dosages used were exactly the same as in the tissue culture system for detecting the effect of these factors on bilirubin cytotoxicity. They were TNF-α 5 U/mL, LPS 10 µg/mL, and AMD 20 ng/mL. The concentration of albumin in each titration study was 4 g/dL. Bilirubin stock solution was added to the aliquots of samples in increasing amounts. Albumin concentration, total and free bilirubin concentrations were determined. The free bilirubin concentration was calculated from the delta absorbance. Bilirubin-albumin titration curve was compared with those added with TNF-α, AMP, LPS; positive control to demonstrate bilirubin displacement was also tested with sodium salycilate. Albumin content was assayed by the bromocresol green method of Doumas et al.(12).

Statistical methods. ANOVA was used to test the mean difference between groups, among different B/A molar ratios and on different days when experiments were carried out. In case of unbalanced data GLM was applied. p < 0.05 was considered significant.

RESULTS

Effect of LPS and TNF-α on L-929 cells. LPS had some degree of cytotoxicity on L-929 cells. ANOVA analysis showed a tendency of dose-response relationship (p < 0.02). The difference due to dosage, however, was small; the mean viabilities for LPS concentrations of 0, 10, 50, and 100 µg/mL were 100%, 84%, 81%, and 81.5%, respectively.

Table 1 shows that recombinant human TNF-α ablated the mitochondrial dehydrogenase activity on L-929 cells in a dose-responsive manner, as assessed by the conversion of MTT to a formazan derivative (ANOVA, p < 0.02). With increasing TNF-α concentrations from 0 to 640 U/mL, the cell viability decreased progressively from 100% to 56% (ANOVA, p < 0.0001). Our experiments have confirmed that AMD can significantly enhance the cytotoxicity of TNF-α on L-929 cells (ANOVA, p < 0.0001).

Table 1 The mean v%* of L-929 cells after TNFα treatment in the presence of actinomycin D
Full size table

Two sets of TNF-α concentrations of 1.25 U/mL and 2.5 U/mL were used to test the effect of LPS on TNF-α cytotoxicity. At LPS concentration of less than 10 µg/mL, and in the absence of AMD, there was no difference in the cell viability at the two different TNF-α concentrations (ANOVA, p > 0.05). With added AMD, an increased cytotoxicity by TNF-α could be found (ANOVA, p < 0.003). When LPS was increased to above 10 µg/mL, e.g. 50 µg/mL, a significant increase of cytotoxicity with the same TNF-α concentrations was observed even without AMD (ANOVA, p < 0.0001).

Effect of TNF-α on bilirubin cytotoxicity. Figure 1 illustrates an obvious additive effect of TNF-α on bilirubin cytotoxicity. In the test range, both TNF-α and bilirubin could separately cause cell death in a dose-related manner (Table 2, ANOVA, p < 0.0001). These two reagents in combination produced more pronounced cytotoxicity. Changing sequence of treatment with TNF-α or bilirubin to the cell culture did not alter the results.

Figure 1
figure 1

Additive effect of TNF-α on bilirubin cytotoxicity in mouse fibroblasts. Cells were first treated with bilirubin for 24 h before addition of TNF-α in actinomycin D (100 ng/mL) for another 24 h. Significant differences between the cell viabilities at different TNF-α concentrations and different B/A molar ratios were obtained (Table 2, ANOVA, p < 0.0001, p < 0.0001).

Full size image
Table 2 ANOVA results of the effect of TNFα and bilirubin on the L-929 cells in the presence of actinomycin D 100 ng/ml (r2 = 0.84)
Full size table

Effect of LPS and TNF-α on bilirubin cytotoxicity. LPS (10 µg/mL) alone had no additional effect on bilirubin cytotoxicity (GLM, p > 0.05). When LPS was combined with TNF-α an additive cytotoxic effect on bilirubin cytotoxicity could be demonstrated (Fig. 2). Further addition of AMD did not alter the result (GML, p > 0.05).

Figure 2
figure 2

Additive effect of TNF-α and LPS on bilirubin cytotoxicity in mouse fibroblasts. Cells were incubated with bilirubin first for 24 h then additional solutions containing TNF-α, LPS, and AMD were added for another 24-h incubation. Note that LPS alone caused an apparent but insignificant effect on bilirubin cytotoxicity (GLM, p > 0.05). Combinations of TNF-α + LPS + BILIRUBIN, TNF-α + AMD + BILIRUBIN, and TNF-α + AMD + LPS + BILIRUBIN all caused more cell death than incubating cells with bilirubin alone (GLM, p < 0.05).

Full size image

Effects of LPS, TNF-α and AMD on the bilirubin titration curves. With the same concentrations used in above none of these agents generated more unbound bilirubin from the bilirubin-albumin titration studies (ANOVA, p > 0.05).

DISCUSSION

Infection has been implicated to be a risk factor of bilirubin encephalopathy although definitive evidence is lacking. Pearlman et al.(13) reported autopsy findings of kernicterus in four near-term infants, all had antemortem culture proven sepsis. The infants were more than 2200 g in weight and were more than 36 wk gestational maturity; apparently prematurity or low birth weight could not be incriminated. The authors suggested that persistent bacterial sepsis might be a critical predisposing factor in the development of the kernicterus even though the serum bilirubin concentrations were only 8.6-15.6 mg/dL. The occurrence of kernicterus in a large number of Chinese term infants many of whom were also inflicted by bacterial infection, especially omphalitis(14–17) adds further indirect evidence of the causal association of infection with kernicterus.

TNF-α, an important infection induced cytokine, is a well-characterized mediator of cytotoxicity although the mechanism of its action is ill defined. Because the discovery of its cytotoxicity in 1975(18), TNF-α assay has been measured by its activities on mouse fibroblast cell line L-929, which is the sensitive target cell(19). The sensitivity of this assay can be further enhanced by addition of AMD as confirmed by our study(4–7,20).

Recombinant human TNF-α administered to rats, dogs, and primates mimics the observed response to endotoxin or overwhelming gram-negative infection(21–23). Pretreatment of primates with monoclonal antibodies to TNF-α confers a distinctive survival advantage during experimental E. coli bacteremia(24). Demonstrable presence of serum TNF-α is associated with bacterial infections(25). In patients of meningococcal meningitis and in childhood sepsis(26), high levels of TNF-α have been correlated with increased fatality. Wääge et al.(25) studied the relationship between TNF-α in the serum and outcome in patients with meningococcal meningitis and septicaemia. They detected TNF-α from 10 of 11 patients who died but in only 8 of 68 survivors. All five patients with serum TNF-α levels over 440 U/mL died. The appearance of TNF-α in experimental models of meningitis has been predictive of neurologic damage(27). TNF-α level in the cerebrospinal fluid was also found to be correlated with the degree of BBB impairment(28). Apparently TNF-α possesses significant tissue damaging effect in vivo.

In our study, TNF-α has been found to have an additive effect on bilirubin cytotoxicity. The mechanism of this synergism remains to be determined. TNF-α has not been shown to displace bilirubin from albumin binding, which would enhance the cytotoxicity. AMD inhibits DNA and RNA synthesis(29,30). As shown in Figure 2, the addition of AMD to the mixture of TNF-α + LPS + bilirubin did not significantly alter the result suggesting that such effect did not require mRNA transcription.

It is well known that TNF-α exhibits its cytotoxic effect through binding to the cell receptor(31). TNF-α receptors have been found ubiquitously distributed on various tissues(31). Because the enhanced cytotoxicity of TNF-α was observed in both pre- and postexposure to bilirubin, the effect is most unlikely to be due to other mediators. The observation of synergism of cytotoxic effect of TNF-α, LPS, and bilirubin even on L-929 cells, which are not particularly sensitive to LPS and bilirubin, suggests that a similar effect could occur in other TNF-α target tissues also.

The direct effect of endotoxin has been studied in vitro in various systems. Different results have been obtained depending on the cells and the assays used. Harlan et al.(32,33) reported that endotoxin in concentration up to 10 µg/mL did not induce detectable cytotoxicity in human endothelial cells derived from umbilical vein, pulmonary artery, or pulmonary vein. In contrast, significant cytotoxicity was observed in bovine aortic endothelial cells exposed to endotoxin as low as 0.01 µg/mL. Thomas et al.(34) demonstrated that endotoxin was not toxic to either endothelial cells or neutrophils, as measured by the release of lactate dehydrogenase. However, preincubation of endothelial cells with endotoxin at concentration of 1.0 to 10 µg/mL would enhance their ability to bind to neutrophils. The effect of endotoxin on L-929 cells has been studied and concentrations up to 100 ng/mL from E. coli was without effect(35). A single intravenous injection into healthy human subjects of E. coli endotoxin at 3-4 ng/kg has been shown to induce fever, cytokinemia, and several hematological and endocrinological changes characteristic of an infection. These subjects demonstrated elevated levels of TNF-α and IL-1β(36), which reached peak plasma levels at 90 and 180 min, respectively. Similar findings have also been observed in animals(37). However, in tissue culture system it is difficult to elicit the toxic effect of endotoxin, which might require the potentiation of cytokines. Our studies have shown significantly different cytotoxic effects with different concentrations of endotoxin (p < 0.02), the mean percentage of cell viability for LPS of 10, 50, and 100 µg/mL were 84%, 81%, and 81.5%, respectively, which, however, represents only marginal dose-response effects.

Pfister et al.(35) reported an enhanced cytotoxic effect between TNF-α and LPS on L-929 cells. In their work such enhancement was undetectable without AMD and LPS alone was without any cytotoxic effect(35). Interestingly we have found a synergism of LPS and TNF-α in inducing cytotoxicity on L-929 cells both in the presence and absence of AMD, although such effect was undetectable when the LPS level was less than 10 µg/mL. These results suggest a dose-related response for such synergism.

The effect of an infection on bilirubin encephalopathy probably involves many other mechanisms such as the permeability of the BBB. Previous observations have shown that polymyxin B, an endotoxin-neutralizing drug, improves the outcome of the experimentally induced bacterial meningitis(38). There is also evidence that TNF-α can cause injury to bovine aortic endothelial cells(39).

Our experimental findings have provided additional evidence that there is an increased risk of bilirubin encephalopathy when the severely jaundiced neonate is also suffering from sepsis. This risk is demonstrated to be due to the direct additive effect of TNF-α, endotoxin, and bilirubin on the cells themselves. To our knowledge this is the first time such evidence is demonstrated at the cellular level. This study also suggests that the prevention of kernicterus in newborn infants should not rely only on an assessment for the potential availability of free bilirubin or the total bilirubin, it should also include other aspects of newborn care, such as the eradication of infection when present.

It would be interesting to find out whether the sudden occurrence of kernicterus in term infants among the early discharged newborns in the western communities(40,41) is related to infection and inflammatory cytokines. Heavy bacterial colonization might have occurred in the home environment; these would generate an increase of bacterial toxin and inflammatory cytokines in some of these jaundiced infants, thus enhancing the bilirubin brain toxicity in addition to inducing hyperbilirubinemia.

In summary, our study has indicated that jaundiced newborns who have developed sepsis would be at increased risk of developing bilirubin encephalopathy. We believe this is the first time that evidence of enhanced bilirubin cytotoxicity has been shown by the effects of infection.