Main

The moderate “physiologic” jaundice that develops after birth may be neuroprotective for the neonate (1), owing to the potent antioxidant properties of UCB (2, 3). By contrast, if the underlying immaturity of the hepatic transport processes or the postnatal increases in production and enterohepatic circulation of UCB are more severe (4), marked neonatal jaundice occurs, which may result in reversible neurotoxicity (bilirubin encephalopathy) (5, 6). This may progress to precipitation of UCB in focal areas of the CNS with permanent neurologic damage (kernicterus) (5).

The UCB that enters the CNS is derived from the free fraction of plasma UCB (BF) that is not bound to plasma proteins and lipoproteins (6). BF levels in plasma are normally very low, as a result of the tight binding of UCB to two sites on HSA (7). Recent data indicate that the affinity for UCB decreases markedly as HSA concentration increases (8, 9) and when chloride is added (9). Therefore, the accepted affinity constant of 6 × 107 L/mol (10), determined at an HSA concentration of 60 μM, overestimates by an order of magnitude the true affinity constant at physiologic albumin and chloride concentrations (9), with consequent marked underestimation of BF. In addition, most published studies of the neurotoxicity of UCB have been performed at total UCB levels vastly higher than those seen in jaundiced neonates with reversible bilirubin encephalopathy, and thus have questionable relevance to the clinical manifestations of neurotoxicity (11, 12).

In the present work, we have applied the affinity constants of UCB for HSA (KF), derived by serial ultrafiltration (9), to recalculate the BF levels present in published in vitro studies of UCB toxicity to neurons and astrocytes. Our aim was to test the hypothesis that in vitro neurotoxicity of UCB could be observed at BF values of 1.0 μM or less, in the range at which spontaneous precipitation of UCB would be unlikely to occur, even if BF was above the aqueous saturation limit of 70 nM. In 12 studies performed at relevantly low total UCB concentrations, toxicity usually occurred at BF levels near or modestly above the aqueous solubility of UCB.

METHODS

Selection of papers for meta-analysis.

We searched PubMed for papers under the following headings: bilirubin + (cells, cultured or cell lines) + (astrocytes or neurons). After eliminating duplicates and adding two related papers by Silberberg et al. (12a,b), we had 28 references. Of these, all but 12 were eliminated for the following reasons: paper published in Chinese 1; review article without original data, 1; did not perform in vitro incubations with UCB, 5; data duplicated results in a paper by the same group that was selected for meta-analysis, 1; only uptake and binding of UCB were studied, not toxicity, 1; incubations included whole human or bovine serum, or BSA, precluding estimation of BF values by applying our ultrafiltration-derived KF values for binding of UCB to delipidated HSA, 4 (also, in two of these, the source and purity of UCB were not given, one studied only UCB photodegradation products, and one performed studies only at BF values of ≥ 5 μM); studies were done only at BF values of 5 μM or greater, 2; studies were done at BF values below 5 μM, but examined only recovery of function after bilirubin washout and the control data duplicated results in another paper by the same group, 1.

One of the 12 papers selected one was done only at a single BF value of 0.5 μM in the absence of albumin (13), and in part duplicated data from another paper by the same group (14) that was performed over a range of BF values. In another selected paper (15), studies were done at BF values below 5 μM, but no toxicity was observed at BF values ≤ 1 μM. In three of the papers selected (13, 14, 16), the stock UCB solution was markedly supersaturated, so that precipitation and degradation likely occurred; thus, the true threshold for UCB toxicity may have been lower than the calculated BF values. Another selected paper studied BF values below 5 μM, but there were no studies done at BF values between 383 and 1761 nM, so that the true threshold could not be evaluated (17).

Calculations.

BF levels were calculated from a model that assumes independent binding of UCB to two sites on albumin, using equation 1 (10), where k1 and k2 are the binding constants for the first and second sites, respectively:

Applying the Solver function of Microsoft Excel 6.0 (Microsoft Corp, Redmond, WA, U.S.A.) to equation 1, BF values were calculated from the total UCB (BT) and total albumin [HSA] concentrations given in the selected papers, using affinity constants of solutions containing comparable concentrations of delipidated HSA and chloride. The value for k1 was set equal to the first site affinity constant (KF) of 14C-UCB for delipidated HSA, derived from serial ultrafiltration of 14C-UCB in solutions containing comparable HSA and chloride concentrations, after correction for the labeled degradation products of 14C-UCB that passed the filter (9). This is valid, because the ultrafiltration studies had intentionally been performed at UCB/HSA ratios of 0.25 or below (9), at which binding of UCB to the second, lower-affinity site is insignificant (10). k2 was calculated as k1/15 = k2 (10).

Most of the papers in the meta-analysis added the UCB ± HSA to cells incubated in protein-free DMEM, which has a total chloride concentration of 118 mM; two papers (20,21) used a chloride concentration of only 1 mM. Because the KF values from the ultrafiltration studies had been obtained only in the presence of 50 mM chloride or no chloride, KF values obtained at 50 mM or 0 mM chloride were applied, respectively. Based on published measurements of the affinity of chloride ions for delipidated HSA (18), these approximations may result in overestimation of KF by, at most, 14 to 26% (see “Discussion”).

All but three of the papers used unpurified UCB. Only the studies from the Lisbon group (19–21) used UCB that was purified by alkaline extraction of impurities and recrystallization from chloroform (22). All studies that included HSA, including the reference ultrafiltration studies of UCB-HSA binding, used delipidated HSA (Sigma Chemical Co, St. Louis, MO, U.S.A.). In all studies, the UCB had been dissolved in 0.1-1.0 N NaOH, then added to the buffered solution of HSA (if used), and then neutralized with HCl.

RESULTS

We report only comparisons of reported toxic effects of UCB with BF levels calculated from the total UCB, HSA, and chloride concentrations provided in the published papers that were selected for the meta-analysis. We performed no direct measurements of the BF levels in these media and no studies of the effects of UCB on cultured cells.

Studies with Unpurified UCB

Astroglial cells (Table 1). Cultured cerebral glial cells from rat embryos showed a significant decrease in mitochondrial function (MTT activity) when exposed for 2 h to BF levels of 500 nM or higher (Fig. 1) (16). Trypan blue release by the same cells increased significantly at BF levels of 1560 nM, but not at 119 nM; intermediate BF levels were not tested. In contrast, cultured cerebral astrocytes from neonatal rats were damaged by ≥ 24-h exposure to BF levels as low as 71 nM, exhibiting significant dose-related decreases in viability (increased LDH release) and mitochondrial function (MTT test) (Fig. 2) (23). A similar 24- to 48-h exposure of these cells to UCB likewise significantly increased LDH release, but the threshold BF level was 721 nM (Fig. 3), with no effect at 194 nM (24).

Table 1 Toxic effects of UCB on cultured rodent glial cells and astrocytes, with threshold BF levels
Full size table
Figure 1
figure 1

Effects on cultured embryonic rat cerebral glial cells of exposure for 2 h to 10 to 10,000 nM of unpurified UCB in the absence of albumin. Impairment of mitochondrial function (MTT test) was similar in cells cultured for 4 or 8 d. *p < 0.05 vs control. Derived from Figure 4 of Amit and Brenner (16).

Full size image
Figure 2
figure 2

Effects on cultured neonatal rat astrocytes of exposure to varied concentrations of unpurified unbound UCB (BF) for 24 h. A significant decrease in both mitochondrial function (MTT activity) and membrane integrity (increased LDH release) occurred at BF levels at or above 71 nM, just above the aqueous solubility limit for UCB (70 nM, vertical dashed line). *p < 0.05, **p < 0.01. Derived from data of Chuniaud et al. (23).

Full size image
Figure 3
figure 3

Effects on cultured neonatal rat astrocytes of exposure for 24, 36, or 48 h to concentrations of unpurified unbound UCB (BF) ranging from 194 to 26,818 nM. *p < 0.05 at threshold BF of 721 nM, as well as all higher concentrations. The effects above BF of 6000 (not shown) are similar to those at 5400 nM. Derived from data of Rhine et al. (24).

Full size image

Another study with cultured brain astrocytes from newborn rats, performed without albumin, tested BF values as low as 1 μM, but found no effect on the uptake of T3 at BF of 10 μM (15); at BF ≥ 25 μM, UCB caused dose-dependent inhibition of T3 uptake, with a Ki of 31 μM. Biliverdin, bilirubin ditaurate and bilirubin glucuronides were progressively more effective inhibitors than UCB, supporting the concept that the inhibition was competitive and not a result of cytotoxicity.

Neuronal cells (Table 2). In the absence of HSA, MTT activity was impaired after 24 h of exposure of embryonic rat hippocampal neurons to 250 nM UCB (p < 0.05; figure not shown) (25) and of 14-d embryonic rat forebrain neurons to UCB concentrations as low as 400 nM (Fig. 4) (14) or 500 nM (data not shown) (13). The last two studies also demonstrated that exposure to 500 nM UCB for 24 to 96 h caused a large decrease in [3H]thymidine incorporation, accompanied by increases in subsequent [3H]thymidine release and in apoptosis. [3H]leucine incorporation into cell protein was affected also after 6 h, with a triphasic response (13). A line of mouse neuroblastoma cells exposed to UCB/HSA systems for 22 h showed significant, dose-related decreases in MTT activity and [3H]thymidine incorporation only at BF levels of 775 nM and above (Fig. 5) (26). In a related paper from the same group (17), a line of rat neuroblastoma cells exhibited significant, multifunctional UCB toxicity after only 2-4 h of exposure to UCB/HSA systems, but only at BF levels in excess of 1700 nM; the true thresholds may have been much lower, because BF levels between 400 and 1700 nM were not tested. The functions impaired included 42K+ uptake, [3H]thymidine incorporation into DNA, MTT activity, and incorporation of [35S]methionine into cellular protein.

Table 2 Toxic effects of UCB on cultured rodent neuronal cells, with threshold BF levels
Full size table
Figure 4
figure 4

Effects on cultured forebrain neurons from 14-d-old rat embryos of exposure for 72-96 h to varied concentrations of unpurified unbound UCB (BF). A significant decrease in mitochondrial function (MTT activity) was observed at BF levels at or above 400 nM, and an increase in apoptosis at a BF level of 500 nM, 5.7 and 7.1 times, respectively, the aqueous solubility limit for UCB (70 nM, vertical dashed line). **p < 0.01. Derived from data of Grojean et al. (14).

Full size image
Figure 5
figure 5

Effects on cultured mouse neuroblastoma cells (NBR10A line) of exposure for 22 h to varied concentrations of unpurified unbound UCB (BF). A significant decrease in both mitochondrial function (MTT activity) and [3H]thymidine uptake occurred at BF levels at or above 780 nM, 11 times the aqueous solubility limit for UCB (70 nM, vertical dashed line). *p < 0.05, **p < 0.01. Derived from data of Schiff et al. (26).

Full size image

Studies with Purified UCB

Cortical astrocytes from neonatal rats showed a dose-related decrease in uptake of [3H]glutamate after only 15 min of exposure to purified UCB at BF levels above 300 nM (Table 1 and Fig. 6) (19). Apoptosis was observed also, but was studied only at BF levels of 6063 nM or higher (20). Neonatal and embryonic cortical neurons from rats exhibited dose-related apoptosis when exposed for 4 h to BF levels of 85 nM or higher (Table 2 and Fig. 7) (20, 21). At and above this threshold BF level, the apoptosis was accompanied by release of cytochrome c from mitochondria, as well as activation of caspase-3, and cleavage of ADP-ribose polymerase (21). By contrast, other mitochondrial changes (translocation of Bax and collapse of membrane potential) were not observed until BF levels reached 50 μM.

Figure 6
figure 6

Effects on cultured cortical astrocytes from 2-d-old neonatal rats of exposure for 15 min to varied concentrations of purified unbound UCB (BF). A significant decrease in [3H]glutamate uptake was observed at BF levels at or above 309 nM, 4.4 times the aqueous solubility limit for UCB (70 nM, vertical dashed line). *p < 0.05, **p < 0.01. Derived from data of Silva et al. (19, 20).

Full size image
Figure 7
figure 7

Effects on cultured cortical neurons and astrocytes from neonatal and embryonic rats of exposure to varied concentrations of purified unbound UCB (BF) for 4 h. A significant increase in apoptosis of neurons occurred at BF levels at or above 85 nM, slightly above the aqueous solubility limit for UCB (70 nM, vertical dashed line). Astrocytes were affected at 6063 nM, but lower BF levels were not tested. *p < 0.05, **p < 0.01, ***p < 0.001. Derived from data of Silva et al. (20) and Rodrigues et al. (21).

Full size image

DISCUSSION

Applying the corrected affinity constants (KF) (9), BF exceeds maximum aqueous UCB solubility (70 nM) at BT well below those at which the first binding site on HSA becomes saturated (Fig. 8A). At the normal adult HSA concentration of 600 μM, this occurs when BT exceeds 80-85 μM (4.7-5.0 mg/dL; Fig. 8B). Except in Crigler-Najjar syndrome, BT levels are rarely this high in adults with unconjugated hyperbilirubinemia. At the 25% lower mean albumin concentrations in newborn plasma (5), supersaturation would occur at BT above 82 μM (>4.8 mg/dL; Fig. 8B), values commonly observed in uncomplicated neonatal hyperbilirubinemia. Thus, except possibly when HSA levels are low in jaundiced patients with cirrhosis, only neonates are exposed to plasma BF levels above aqueous solubility.

Figure 8
figure 8

UCB/HSA ratios (left) and total UCB concentrations (right) at which calculated BF values equal the aqueous solubility of unbound UCB at pH 7.4 over a range of albumin (HSA) concentrations. Areas above each curve are supersaturated with unbound UCB, whereas areas below the line are unsaturated. At mean normal plasma albumin concentrations of 450 μM for neonates and 600 μM for adults, supersaturation occurs at UCB/HSA ratios as low as 0.2 and at total UCB concentrations above 80 μM. BF values were calculated by applying corrected affinity constants of UCB for HSA in the presence of chloride (9) at the albumin concentrations indicated. The calculations (equation 1) assumed independent binding of UCB to two sites on albumin (10).

Full size image

At BF levels above saturation, self-aggregation of UCB diacid must occur, progressing through three stages (7, 27–29). Oligomers of UCB diacid appear just above saturation; although they are too small to precipitate, they can dissociate reversibly and serve as a reservoir to replenish UCB monomers removed by cells. At higher UCB concentrations, larger colloidal aggregates form, stabilized by UCB mono- and dianions adsorbed on their surfaces (7). These microsuspensions may precipitate with prolonged standing (ripening) or neutralization of the charges by a decrease in pH (28). At yet higher UCB concentrations, coarser aggregates appear that precipitate spontaneously. Limited available data suggests that metastable aggregates are present at BF as low as 1-2 μM at pH 7.0 to 7.4, but probably not at 500 nM (28, 29).

Our recalculated BF levels for published in vitro studies (Tables 1 and 2) reveal that neurotoxic effects of even purified UCB can be observed at BF levels ranging from slightly above to 11 times aqueous solubility (71-7 70 nM), although higher thresholds were obtained in some of the studies. Only a few of the original papers attempted to calculate or measure free bilirubin concentrations, and, when doing so, they used methods that have been shown to be flawed, as discussed elsewhere (8, 9). The variation of more than 10-fold among studies in our recalculated toxic thresholds for unbound bilirubin concentrations (Tables 1 and 2) is not unexpected, as the 12 papers used different cell systems from different species (26) of differing maturity, different cell functions, and different durations of culture (16) and exposure to UCB. Thus, the variable thresholds may simply reflect different susceptibilities of different cell systems to different types of injury.

Minor components in the DMEM, including those released by the cells themselves, might have influenced binding also. Although the different batches of delipidated HSA may have differed somewhat in their affinities for UCB, we have found that the binding affinity for UCB among four different batches of delipidated albumin from the same manufacturer (Sigma Chemical Co) varied by less than 4% (Ostrow JD, unpublished data). Thus, only a small error in BF is introduced by our assumption that the binding affinity of the HSA used in our ultrafiltration study (9) is representative of the batches used in the studies in the meta-analysis.

Figures 2,5, and 6 reveal a trend toward decreased viability or function at BF levels below those at which a statistically significant impairment was attained. In all four cases, if those trends are assumed to be real effects, the resultant lower thresholds are still all slightly below (Fig. 2 and 6) to modestly above (Fig. 5) the solubility limit for unbound UCB. Some figures show gradual declines in function with increasing BF levels, whereas in others, the threshold appears to be abrupt. Such differences, however, may be more apparent than real, depending on whether enough data points were obtained both above and below the true threshold.

Although, for reasons noted above, the thresholds varied with different studies and the responses were not uniform, our findings clearly establish that that marked supersaturation and precipitation of UCB are not necessary to produce toxicity to CNS cells. This renders untenable the long-accepted concept that only coarse UCB aggregates, which may include coprecipitated albumin, are involved in early UCB toxicity (7). Even allowing for potential moderate inaccuracies in our calculated values of unbound bilirubin (9), our findings strongly suggest that toxicity develops only near or above the aqueous saturation limit of 70 nM, a range in which only UCB monomers, soluble oligomers, and metastable small colloids are likely to be present. By contrast, BF levels well below 70 nM [aqueous saturation (30)] appear to protect CNS cells against oxidative damage (25, 31), and this protection is lost because of the countervailing toxic effects of UCB at higher BF levels (25).

At these relatively low BF levels, both astrocytes and neurons were susceptible to impairment of mitochondrial functions (MTT activity and apoptosis), whereas diminished incorporation of [3H]thymidine was reported only for neurons. These toxic effects can account for the structural features of apoptosis that appear in the cerebellum and cochlear nucleus of jaundiced Gunn rat pups (32–34) and in the basal ganglia of kernicteric human infants (5). These early changes appear well before peak UCB levels are attained, but ultimately progress to atrophy of these CNS regions.

The modestly supersaturated BF levels also affected astrocyte membranes, as shown by increased LDH release (23) (Fig. 2) and impaired [3H]glutamate uptake (Fig. 6) (19). The extremely brief period of exposure to UCB probably explains why Warr et al. (35) did not detect changes in glutamate transporters, N-methyl-D-aspartate receptors, or electrical currents in retinal glial cells from salamanders after treatment with 10 μM UCB for only 10-50 s.

Effects of purified UCB on the membrane structure of neurons (21) and mitochondria (20) have been observed, however, only at highly supersaturated BF levels in the micromolar range, at which UCB precipitation is expected. This fits with historic concepts that precipitation of UCB in cell membranes alters membrane fluidity and the activity of integral membrane proteins (7), but it is problematic as to whether this is relevant to the modestly elevated BF levels associated with the reversible stages of bilirubin encephalopathy. Monomers of UCB diacid cannot penetrate deeply into membranes (36), but bind near the surface of the outer leaflet of the membrane (37–39). The resultant modest perturbation of membrane structure might be a factor in the early cellular toxicity of clinically relevant concentrations of UCB (39).

Ahlfors (8) applied a peroxidase-diazo method to reassess BF for historic data on plasma BT and HSA concentrations in jaundiced neonates and concluded that kernicterus was likely only when BF levels exceeded 60 nM (40), in apparent agreement with our results for in vitro systems. There are, however, important differences between plasma or serum and in vitro systems that limit comparisons between his study and ours. Plasma contains additional proteins that bind UCB, such as apo D (41), so that BF levels in plasma are lower than those in solutions containing the same concentration of purified albumin (42). On the other hand, FFA and other substances not present in defined solutions containing delipidated albumin may inhibit the binding of UCB to albumin (7). Finally, in vivo, neurons and astrocytes are not exposed directly to plasma, but are separated by the blood-brain and blood-cerebrospinal fluid barriers that may limit penetration of unbound UCB into the CNS (6). Thus, the media to which the CNS cells are exposed in vitro are the equivalent of the cerebrospinal fluid and extracellular fluid in the brain, where, in jaundiced Gunn rats, total UCB concentrations may be only one fifth those in plasma (43) and albumin concentrations are much lower than plasma. Overall, therefore, the threshold BF levels for UCB neurotoxicity are likely to be higher in plasma in vivo than in defined albumin solutions in vitro.

CONCLUSIONS

Because of the above-noted differences between in vitro and in vivo systems, as well as interspecies differences, it remains to be determined whether, to fully prevent bilirubin encephalopathy, treatment of neonatal hyperbilirubinemia should be instituted at plasma UCB levels lower than those that are currently recommended (5, 44). Nonetheless, our findings favor a role for small, soluble UCB aggregates, present at moderately supersaturated BF levels, in the often-reversible damage to mitochondria, and possibly plasma membranes of CNS cells that characterize the early stages of bilirubin encephalopathy.