Introduction

Hyperbilirubinemia is a common condition in preterm neonates that results from the accumulation of bilirubin in the blood. Bilirubin is the normal byproduct in the catabolism of hemoglobin. However, rapid turnover of red blood cells and the limited ability to conjugate bilirubin cause neonates to frequently develop hyperbilirubinemia.1,2 Preterm infants are at higher risk for neurotoxicity at lower concentrations of bilirubin than full-term infants.3,4 This increased susceptibility is due to the migration and proliferation of the principal neurons of brain regions, such as the cerebellum, during late gestation and early postnatal development.5 Acute bilirubin neurotoxicity in preterm infants usually occurs without obvious clinical symptoms. However, it may result in lifelong cerebellar dysfunction leading to impaired gait, coordination, and balance among other symptoms in later childhood and adults.6

While there may be many mechanisms of neurotoxicity due to hyperbilirubinemia, we have focused on the disruption of detergent resistant membranes (called lipid rafts (LR)) and have shown that bilirubin impairs LR function in vitro.7,8 LR are dynamic membrane assemblies consisting of cholesterol, sphingolipids, GM1 gangliosides, and proteins that play an important role in facilitating cell signaling9,10 and trafficking.11,12 We use L1 cell adhesion molecule (L1CAM) as a reporter protein for lipid raft function. L1CAM is critical to the development of the central nervous system and functions through LR-dependent signaling.13 The cytoplasmic domain of L1CAM contains multiple tyrosine residues whose phosphorylation status regulates signaling and trafficking allowing for neurite outgrowth. The disruption of lipid rafts with the cholesterol chelator, methyl-β-cyclodextrin, completely inhibits L1CAM-mediated neurite outgrowth.14

Choline is a precursor to both phosphatidylcholine and sphingomyelin, essential components of the plasma membrane and LR, respectively. Choline deficiency is common amongst pregnant women15 and preterm infants.16 We have found that choline pretreatment ameliorates bilirubin-induced LR-dependent dysfunction in vitro.8

The Gunn rat is a model of hyperbilirubinemia. The homozygous Gunn rat (jj) lacks the ability to conjugate bilirubin which results in the accumulation of unconjugated bilirubin bound to albumin. Administration of sulfadimethoxine (SDMX) results in acute elevation of free bilirubin (Bf), which can cross the blood brain barrier and perturb developing regions of the brain.17 We have previously shown that preterm hyperbilirubinemia can be modeled using this animal by administering SDMX on P5 (ref). The Gunn rat dams were made choline deficient to model the common condition of choline deficiency,15 making this animal model even more relevant to preterm infants.16 We have previously shown in this animal model that cerebellar behaviors are impaired only in jj rat pups more than 3 weeks after SDMX administration and that this effect is reduced by choline supplementation. The study reported here is the first to show that in vivo, 1) that bilirubin induces a decrease in L1CAM tyrosine phosphorylation, L1CAM Y1176 dephosphorylation, and causes a redistribution of L1CAM in LR, and 2) choline supplementation ameliorates these effects.

Methods

Antibodies and materials

Goat polyclonal anti-neuronal cell adhesion molecule L1CAM antibody to the cytoplasmic domain of L1CAM (anti-L1CD) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA) (catalog #sc-374046). Mouse monoclonal antibody to the dephosphorylated Y1176 of L1CAM (74-5H7) (catalog #MMS-172R) was obtained from Covance (Berkeley, CA). Mouse monoclonal antibody to phosphotyrosine (PY100) (catalog #9411) was obtained from Cell Signaling Technology (Danvers, MA). Horseradish peroxidase (HRP) conjugated cholera toxin B subunits (CTxB) was obtained from Sigma-Aldrich (St. Louis, MO). Mouse monoclonal anti-transferrin receptor antibody was obtained from Invitrogen (Carlsbad, CA). HRP -conjugated goat anti-mouse IgG (H + L) and HRP-conjugated donkey anti-goat IgG (H + L) secondary antibodies were obtained from Jackson Immuno-Research Laboratories (West Grove, PA).

The choline-restricted diet was obtained from Dyets (Bethlehem, PA) (catalog #518753). Choline chloride was obtained from Sigma (St. Louis, MO).

Study design

Gunn rats were obtained from the Rat Resource & Research Center and a breeding colony was established at Case Western Reserve University. All breeding and experimental procedures were approved by the Case Western Reserve University Institutional Animal Care and Use Committee. Litters were generated by pairing jj males and Nj females. Females were checked daily for the presence of a mucus plug. The day a plug was observed was designated as gestational (G)0. Dams were placed on a choline-restricted diet on G5. On postnatal (P)1, pups were genotyped to determine Nj and jj status. Each pup was assigned to receive either saline or choline by block randomization within each genotype. Pups were weighed daily. Pups were injected subcutaneously with a single injection of either choline chloride (100 mg/kg bodyweight, 5 µl/g) or an equivalent volume of saline daily from P1 to P5. On P5, pups were further randomly assigned to be injected intraperitoneally with either SDMX (200 mg/kg) or an equivalent amount of saline. The SDMX solution was prepared such that 10 µL per gram of bodyweight of the pup or an equivalent volume of saline was administered to the pup as a single injection 1 h after SQ saline/choline. Due to limitations caused by the capacity of the ultracentrifuge, there were three treatment groups per genotype. In Nj pups, the groups were as follows: saline from P1 to P5 and saline on P5 (Nj SS, n = 3), saline from P1 to P5 and SDMX on P5 (Nj SSD, n = 3), and choline from P1 to P5 and SDMX on P5 (Nj CSD, n = 3). In jj pups the groups were as follows: saline from P1 to P5 and SDMX on P5 (jj SSD, n = 3), choline from P1 to P5 and saline on P5 (jj CS, n = 3), and choline from P1 to P5 and SDMX on P5 (jj CSD, n = 3). Pups were sacrificed on P6, 24 h after saline/SDMX administration.

Tissue preparation

The animals were rapidly decapitated into ice cold Tris-buffered saline (TBS) and the cerebella were dissected from the whole brain in petri dishes placed on ice. Each cerebella was homogenized in ice cold TBS containing 0.5% Triton X100, 10% glycerol, 10 mM Na-vanadate, 2 µM aprotinin, 0.1 mM phenylmethylsulfonyl fluoride, 1 µM leupeptin, 1 µg/mL pepstatin A, 10 µg/mL turkey trypsin inhibitor, 100 pM cypermethrin, phosphatase inhibitor cocktail I (Sigma, St. Louis, MO), and phosphate inhibitor cocktail II (Sigma, St. Louis, MO). The tissue extract from each animal were centrifuged at 13,000 × g for 10 min at 4 °C and individual supernatants were collected. Protein concentration was determined by Bradford Assay (Sigma-Aldrich, St. Louis, MO).

Immunoprecipitation

Supernatants containing equal amount of protein were pre-cleared with 1 µg rabbit serum for 1 h, followed by a 1 h incubation at room temperature with protein A/G agarose. The samples were centrifuged at 1000 × g for 4 min and the supernatant was removed. Two µg anti-L1CD antibody and 30 µg protein A/G agarose beads were added to the supernatants, and the samples were incubated overnight at 4 °C. The beads were washed 3 times with phosphate buffered saline, and boiled in 30 µL 1 × sample buffer for 5 min. The sample buffer containing the immunoprecipitate was collected by centrifugation of 1000 × g for 4 min and analyzed by immunoblotting.

Western blots

Samples were separated with SDS-PAGE (12% gel) run at 110 V for 1.5 h and transferred at 20 V for 45 min to a polyvinylidene difluoride membrane. This membrane was blocked in TBS containing 5% skim milk powder and 0.1% Tween-20 at room temperature for 30 min. For tyrosine phosphorylation of L1CAM, membranes were incubated with antibodies to phosphotyrosine (PY100), washed and probed with HRP-goat anti-mouse IgG, stripped and reblotted with anti-L1CD. For Y1176 dephosphorylation of L1CAM, membranes were incubated with monoclonal antibody to the Y1176 tyrosine of L1CAM (74-5H7) and probed with HRP-rabbit anti-mouse IgG. Membranes were then stripped for 5 min at room temperature with WesternSure ECL stripping buffer obtained from Li-Cor (Lincoln, Nebraska) and reprobed with anti-L1CD with a secondary antibody of HRP-Rabbit anti-goat IgG. Primary antibodies were diluted into block buffer at a concentration of 1:1000 and incubated for 2 h at room temperature. Secondary antibodies were diluted into block buffer at a concentration of 1:5000 and incubated for 1 h at room temperature. The relative intensity of the bands was quantified using transmittance densitometry with Image J (NIH). For each lane of the immunoblots, the phosphotyrosine band density of the L1CAM immunoprecipitate and the Y1176 band density of the L1CAM western blot were normalized to the total L1CAM band density to calculate the proportion of the phosphorylated or dephosphorylated L1CAM to total L1CAM.

Lipid raft isolation

Prior to sacrifice of the animals, all equipment and buffers were cooled to 4 °C. One mL of supernatant was mixed with an equal amount of 80% sucrose solution and then overlaid with 4 ml each of 32% and 5% sucrose solutions. The gradient was then centrifuged at 180,000 × g for 24 h at 4 °C. At 4 °C, sequential 1 mL fractions were drawn off the top of the gradient. Fractions containing LR were identified by dot blots for the presence of both GM1 ganglioside and transferrin receptor. All GM1 ganglioside-containing fractions were combined into a LR containing pool and all remaining fractions were combined in a non-lipid raft pool (N). Proteins from an equivolume of each LR and N pool were precipitated using chloroform/methanol and resuspended in Laemmli sample buffer, then immunoblotted as described above.

Statistical analysis

Repeated measures ANOVA was performed to assess differences in subject weights over time.

To determine changes in the proportion of L1CAM with tyrosine phosphorylation or dephosphorylation, the pixels in either the phosphorylated or the dephosphorylated L1CAM band were divided by the pixels in the L1CAM band of each lane representing each animal. An initial analysis was conducted for the Nj treatment groups (Nj SS, Nj SSD, Nj CSD). Each blot was normalized to the control condition (PY100x/L1CAMx)/(PY100Nj SS/L1CAMNj SS) or (L1CAM-Y1176x/L1CAMx)/(L1CAM-Y1176Nj SS/L1CAMNj SS). The results were expressed as mean ± SEM using Microsoft Excel software (Redmond, WA).

To determine differences in the percentage of L1CAM in LR, pixels in the LR band were divided by the sum of the pixels in both the LR and N bands and multiplied by 100, then normalized to the control (Nj SS) group. The results were expressed as mean ± SEM using Microsoft Excel software (Redmond, WA).

A one-way ANOVA was conducted to determine whether the heterozygous (Nj) groups were significantly different across treatment conditions and no significant differences were found. These animals were combined into a single control group (Nj combined) to increase statistical power.

The main analysis employed a one-way ANOVA on the homozygous (jj) animal treatment groups with the combined heterozygous (Nj) group serving as a reference control for subsequent comparisons. Tukey’s Honestly Significant Difference (HSD) test was used for post-hoc pairwise comparisons to determine specific treatment effects. Statistical analysis was performed using JASP software (Amsterdam, The Netherlands).

Results

Choline supplementation has no impact on weight gain of pups

No differences were observed in the initial and final weights of both Nj (n = 11) and jj (n = 11) pups. There was no significant difference between Nj pups supplemented with choline or saline nor jj pups supplemented with choline or saline, indicating that choline supplementation and genotype had no effect on the growth of animals from the start of supplementation until sacrifice (Fig. 1).

Fig. 1: Choline supplementation has no effect on weight gain of rat pups.
figure 1

Growth curves of pups from postnatal days 1 to 5 prior to SDMX administration. N = 6 for Nj saline and jj choline, N = 5 for Nj choline and jj saline. Repeated measures ANOVA, Tukey HSD post hoc test showed no significant different between groups.

Full size image

Choline reduces bilirubin inhibition of L1CAM tyrosine phosphorylation

A one-way ANOVA revealed a significant effect of treatment on L1CAM tyrosine phosphorylation, F(3,14) = 10.87, p < 0.001, η2 = 0.70. Post-hoc pairwise comparisons using Tukey’s HSD indicated that treatment with SDMX in the saline supplemented jj Gunn rat group (jj SSD) showed a significant decrease in the proportion of L1CAM tyrosine phosphorylation in comparison to all other groups (p < 0.05) (Fig. 2). Choline supplementation in the SDMX treated jj Gunn rat group (jj CSD) showed a significant increase in the fraction of L1CAM tyrosine phosphorylation in comparison to jj SSD (p = 0.017). There were no significant differences between all the other groups.

Fig. 2: Choline reduces bilirubin inhibition of L1 tyrosine phosphorylation.
figure 2

a Representative immunoblot of phosphorylated L1CAM (PY-L1) and total L1CAM (L1). The experimental groups are as follows: Nj combined (Nj saline/saline (SS), Nj saline/SDMX (SSD), Nj choline/SDMX (Nj CSD)), jj saline/SDMX (SSD), jj choline/saline (CS) and jj choline/SDMX (CSD). b Bar graph of mean ± SEM of fraction of relative densitometric units of PY-L1 divided by L1CAM (L1), normalized to Nj combined. Nj combined, N = 9, for each of the three jj groups N = 3, One-way ANOVA, Tukey HSD post hoc test, *p < 0.05.

Full size image

Choline reduces bilirubin inhibition of L1CAM Y1176 dephosphorylation

L1CAM Y1176 dephosphorylation differed significantly among groups, F(3,14) = 16.25, p < 0.001, η2 = 0.78. Post-hoc pairwise comparisons using Tukey’s HSD indicated that in jj Gunn rats treated with saline/SDMX (jj SSD) there was a significant decrease in the proportion of L1CAM Y1176 dephosphorylation in comparison to all other groups (p < 0.05) (Fig. 3). jj Gunn rats treated with choline/SDMX (jj CSD) showed a significant increase in L1CAM Y1176 dephosphorylation in comparison to jj SSD (p = 0.012). There were no other differences between all the other treatment groups.

Fig. 3: Choline reduces bilirubin inhibition of L1CAM Y1176 dephosphorylation.
figure 3

a Representative immunoblot of dephosphorylated L1CAM (Y1176) and total L1CAM (L1). The experimental groups are as follows: Nj combined (Nj saline/saline (SS), Nj saline/SDMX (SSD), Nj choline/SDMX (Nj CSD)), jj saline/SDMX (SSD), jj choline/saline (CS) and jj choline/SDMX (CSD). b Bar graph of mean ± SEM of the fraction of relative densitometric units of Y1176 divided by total L1CAM, normalized to Nj combined. Nj combined N = 9, for each of the three jj groups N = 3, One-way ANOVA, Tukey HSD post hoc test, *p < 0.05.

Full size image

Choline reduces bilirubin-induced redistribution of L1CAM in lipid rafts

Analysis of L1CAM distribution in LR by one-way ANOVA revealed significant group differences, F(3,14) = 10.01, p < 0.001, η2 = 0.68. In jj Gunn rats treated with saline/SDMX (jj SSD), the percentage of L1CAM in lipid rafts doubled to 13.8% of total L1CAM, compared to 6.14% in the Nj combined group (p < 0.001) (Fig. 4). The prior supplementation of choline in SDMX treated jj (jj CSD) Gunn rats significantly decreased the percentage of L1CAM in lipid rafts down to 6.7% when compared to jj SSD (p = 0.009). The jj saline/SDMX group (jj SSD) was significantly different to all other groups (p < 0.05). There were no significant differences between all the other treatment groups.

Fig. 4: Choline reduces bilirubin-induced L1CAM redistribution in lipid rafts.
figure 4

a Representative immunoblot of L1 distribution in lipid raft (LR) and non-LR fractions (N). The experimental groups are as follows: Nj saline/saline (SS), Nj saline/SDMX (SSD), Nj choline/SDMX (Nj CSD), jj saline/SDMX (SSD), jj choline/saline (CS) and jj choline/SDMX (CSD). b Bar graph of mean ± SEM of percent L1 in lipid rafts (LR / (LR + N)) x 100. N = 3, One-way ANOVA, Tukey HSD post hoc test, *p < 0.05.

Full size image

Discussion

Our prior work supports that one of the mechanisms by which free bilirubin exerts its neurotoxic effects is by localizing to LR and inhibiting downstream L1CAM signaling. This inhibition has been shown to be reduced with choline pretreatment in vitro.8 The results of this current study are twofold: 1) that bilirubin affects L1CAM function and signaling in vivo; and 2) that choline supplementation of jj Gunn rat pups whose dams are on a choline restricted diet can mitigate effects the effects of bilirubin.

Management of hyperbilirubinemia consists of lowering total serum bilirubin (TSB) concentrations using intensive phototherapy, and in extreme cases, double volume exchange transfusion. However, there are several drawbacks to this approach. The reduction of bilirubin neurotoxicity is better predicted if the treatment is based on Bf and not TSB, as a reduction in TSB does not correlate with improved outcomes.18 Additionally, phototherapy increases mortality risk in extremely low birth weight infants.19 Exchange transfusion is known to be a high risk procedure.20 Therefore, there is a need for safer and more effective therapies such as the use of choline.

In addition, bilirubin is a potent antioxidant. In the term and preterm newborn, the antioxidant protective enzymes are not yet expressed, making bilirubin especially needed.21 Phototherapy may lead to increased oxidative damage due to the reduction in bilirubin. Supplementation with choline may reduce the neurotoxicity of bilirubin without reducing its antioxidant effects.

Choline is an essential nutrient that has been shown to ameliorate the behavioral and biochemical effects of ethanol and bilirubin in vitro and in vivo.8,17,22,23 Choline is taken up by neurons via the choline transporter and metabolized by three different pathways. Cholinergic neurons synthesis acetyl choline, a neurotransmitter. All cells, including neurons, synthesize phosphatidylcholine via the Kennedy pathway. Phosphatidylcholine is metabolized to sphingomyelin, a critical component of lipid rafts. Choline is also synthesized to betaine by choline dehydrogenase. Betaine is metabolized via methionine to S-adenosylmethionine, a 1-methyl donor for proteins, DNA and RNA modifications. Thus, choline may influence signaling through at least 3 different pathways.

Choline pretreatment in primary culture of cerebellar granule neurons mitigated the effects of bilirubin on L1CAM-mediated neurite outgrowth and reduced the inhibition of L1CAM tyrosine phosphorylation, L1CAM Y1176 dephosphorylation, and redistribution of L1CAM in lipid rafts.8 In the present study, we show that hyperbilirubinemic Gunn rat pups supplemented with choline have decreased bilirubin-induced biochemical deficits in L1CAM tyrosine dephosphorylation/phosphorylation and LR perturbations in the cerebellum. It is interesting that choline supplementation in choline restricted pups alone did not influence the L1CAM properties measured here. It may be that bilirubin alters the flux through these pathways in choline deficient animals and that choline supplementation restores them. More experiments are needed to determine the underlying mechanism.

A limitation of this study is that the effects of choline supplementation on L1CAM function were only evaluated in Gunn rat pups whose dams were on choline-restricted diets rather than dams that were on normal diets. This was done to replicate the common condition of choline deficiency that encompasses 90% of adults in the US.15 Preterm infants in neonatal intensive care also receive less choline than is present physiologically in the intrauterine environment.16 However, it cannot currently be concluded that choline supplementation would not benefit subjects that have adequate maternal choline intake.

Secondly, this study looked at 6 of the 8 total possible treatment groups due to constraints on the number of samples that can be ultra-centrifuged concurrently. Although the experimental design did not include all possible group combinations, specifically choline supplemented Nj treated with saline (Nj CS) and saline supplemented jj treated with saline (jj SS), the study design remained adequate to test our primary hypothesis that choline can ameliorate biochemical deficits in the cerebellum caused by bilirubin in hyperbilirubinemic rat pups. The omission of the Nj CS group does not compromise our conclusions, as this group would examine whether choline can further improve function in untreated control animals, which was not our primary research focus. Our preliminary analysis demonstrated no significant differences among any of the Nj experimental conditions, indicating that choline supplementation does not affect L1CAM phosphorylation in non-jaundiced animals regardless of the included treatment conditions. Similarly, the absence of the jj SS group does not limit our ability to evaluate choline’s therapeutic efficacy, as this group would function as an additional control that we believe is adequately covered by the Nj combined control group and the jj CS group. We have previously shown in vitro that choline administration in bilirubin-treated primary cerebellar granule neurons does not improve outcomes beyond untreated control. This conclusion is further confirmed in the current study, where the jj CS group showed no significant improvement in comparison to the Nj combined control. While a complete factorial design would provide additional mechanistic insights, the targeted approach employed directly addresses the key translational questions regarding choline’s potential as a neuroprotective treatment for acute bilirubin encephalopathy.

Future work will focus on elucidating the mechanism(s) by which choline exerts its neuroprotective benefits. Currently, the evidence for bilirubin’s impact on lipid rafts is purely correlative and requires further study to determine the specific mechanisms through which bilirubin may be perturbing lipid rafts. Further mechanistic studies will be done using fluorescence recovery after photobleaching to investigate the physical properties of the L1CAM kinetics around lipid rafts after treatment with bilirubin. Future experiments will also explore the impact of hyperbilirubinemia in other parts of the brain such as the auditory nuclei, hippocampus, and globus pallidus and the potential efficacy of choline supplementation in these regions.

This study identifies choline supplementation as a potential neuroprotection for neurodevelopmental defects in the cerebellum due to hyperbilirubinemia and provides additional evidence in vivo that disruption of lipid rafts may be a mechanism of bilirubin neurotoxicity.