Potential of therapeutic bile acids in the treatment of neonatal Hyperbilirubinemia

Neonatal hyperbilirubinemia or jaundice is associated with kernicterus, resulting in permanent neurological damage or even death. Conventional phototherapy does not prevent hyperbilirubinemia or eliminate the need for exchange transfusion. Here we investigated the potential of therapeutic bile acids ursodeoxycholic acid (UDCA) and obeticholic acid (OCA, 6-α-ethyl-CDCA), a farnesoid-X-receptor (FXR) agonist, as preventive treatment options for neonatal hyperbilirubinemia using the hUGT1*1 humanized mice and Ugt1a-deficient Gunn rats. Treatment of hUGT1*1 mice with UDCA or OCA at postnatal days 10–14 effectively decreased bilirubin in plasma (by 82% and 62%) and brain (by 72% and 69%), respectively. Mechanistically, our findings indicate that these effects are mediated through induction of protein levels of hUGT1A1 in the intestine, but not in liver. We further demonstrate that in Ugt1a-deficient Gunn rats, UDCA but not OCA significantly decreases plasma bilirubin, indicating that at least some of the hypobilirubinemic effects of UDCA are independent of UGT1A1. Finally, using the synthetic, non-bile acid, FXR-agonist GW4064, we show that some of these effects are mediated through direct or indirect activation of FXR. Together, our study shows that therapeutic bile acids UDCA and OCA effectively reduce both plasma and brain bilirubin, highlighting their potential in the treatment of neonatal hyperbilirubinemia.

www.nature.com/scientificreports/ that other anti-cholestatic medications could also be beneficial for unconjugated hyperbilirubinemia. In 2016, obeticholic acid (OCA) was FDA-approved as a treatment for primary biliary cholangitis, for patients in which UDCA does not provide sufficient relief. OCA is a selective agonist of the farnesoid-X-receptor (FXR, NR1H4), a biological sensor for BAs that, among other functions, controls BA homeostasis by modulation of their synthesis and enterohepatic transport 14,15 . In contrast to OCA, UDCA is not a ligand for FXR 16 .
In this study, we used the humanized hUGT1*1 mouse model for unconjugated hyperbilirubinemia. hUGT1*1 mice are transgenic for the entire human UGT1A locus, including the UGT1A1 promoter and the well-described phenobarbital response enhancing module (PBREM) 17 . Because UGT1A1 is expressed at an earlier age in mice as compared to humans, neonatal hyperbilirubinemia normally does not occur in mice 18,19 . Humanized UGT1*1 mice, however, display a similar transcriptional delay in expression of UGT1A1, resulting in severe unconjugated hyperbilirubinemia during the first three weeks of life. After three weeks, the bilirubin levels decrease upon UGT1A1 expression and the mice remain normobilirubinemic during adulthood 17 . Complete knockout of the Ugt1 locus in mice, when not rescued with expression of the human UGT1 locus, results in death within 4-11 days after birth 17,20 . The expression of the human UGT1 locus in hUGT1*1 mice rescues the lethality associated with severe neonatal hyperbilirubinemia as a result of the extrahepatic intestinal expression of the human UGT1A1 gene 17,21 . The physiological inducibility of hUGT1A1 expression in hUGT1*1 mice in both liver and intestine makes this an attractive model and superior to Ugt1 deficient models such as the Ugt1a knockout mice or Gunn rats, commonly used models for hyperbilirubinemia 22,23 . This inducibility of hUGT1A1 makes them similar to human neonates, and the only model to study the effects of drugs on human UGT1A1 expression and bilirubin conjugation. In this study, we assessed the effects of UDCA and OCA on bilirubin levels in plasma and brain in neonatal hUGT1*1 mice. In addition, Ugt1a deficient Gunn rats were used as a model to distinguish between UGT1A1-dependent and independent treatment effects.
Effect of UDCA and OCA on hUGT1A1 expression in liver and intestine. To assess whether UDCA or OCA affected UGT1A1 expression in hUGT1*1 neonatal mice, we determined mRNA and protein levels in liver and intestine. Classically, the liver is regarded as the main organ responsible for bilirubin conjugation 24 . As expected, UGT1A1 expression was not detectable in the liver of neonatal hUGT1*1 mice 17 , and was also not induced by UDCA or OCA (data not shown). In contrast to the liver, UDCA treatment strongly increased levels of UGT1A1 mRNA in all intestinal sections (duodenum, jejunum, ileum and colon; Fig. 2a,c,e,g), whereas OCA treatment ( Fig. 2i,k,m,o) only caused a 1.7-fold induction in jejunum (p < 0.05) (Fig. 2k). To assess whether the increased intestinal UGT1A1 gene expression also resulted in increased protein expression, we performed Western blot analysis. In line with its mRNA levels, UGT1A1 protein levels were undetectable in liver whereas both UDCA (Figs. 2b,d,f,h; Fig. S2) and OCA (Figs. 2j,l,n,p; Fig. S2) treatment significantly increased hUGT1A1 protein levels in duodenum (both p < 0.01) and jejunum (both p < 0.01), but not in ileum. In colon, UGT1A1 protein expression was only induced after OCA treatment (p < 0.01) (Fig. 2p).
FXR activation by UDCA and OCA. OCA is a selective FXR agonist, and therefore its actions can be potentially explained by activation of FXR. UDCA, on the other hand, is not an FXR agonist, but since it modulates BA homeostasis, it is possible that it can still indirectly affect FXR activity. To determine whether OCA and UDCA (directly or indirectly) caused FXR activation, we determined mRNA levels of the established FXR targets short heterodimer partner (Shp, Nrob2), cholesterol 7α-hydroxylase (Cyp7a1) and the bile salt exporting protein (Bsep, Abcb11) in liver and of Shp and fibroblast growth factor 15 (Fgf15) in ileum (Fig. 3). Both UDCA and OCA treatment resulted in hepatic and ileal FXR activation, although Bsep expression was not significantly affected by UDCA (Fig. 3). www.nature.com/scientificreports/ Based on these findings, we hypothesized that the effects on UCB levels in plasma and brain could be mediated through FXR activation, either directly (OCA) or indirectly (UDCA). To further evaluate the potential involvement of FXR, we also tested the UCB lowering effects of the synthetic, non-BA, FXR-agonist GW4064, which is known to decrease BA production by suppression of CYP7A1. Treatment of hUGT1*1 pups with GW4064 (50 mg/kg/day) (n = 6) resulted in a similar gene expression pattern matching with FXR activation (Fig. S3). GW4064 also significantly reduced TPB by 13% (p < 0.05) and brain UCB by 31% (p < 0.01) (Fig. 4a,b), along with a twofold induction of UGT1A1 protein expression in duodenum (Fig. 4d). GW4064 also significantly increased UGT1A1 mRNA expression in ileum, but this was not accompanied by increased protein expression (Fig. 4g,h). In jejunum and colon, GW4064 did not induce UGT1A1 (Fig. 4e,f,i,j). Collectively, these observations support a role for FXR in the UCB lowering effects of tested anti-cholestatic drugs.
Contribution of UGT1A1 to the bilirubin-lowering effects of UDCA and OCA. Although hUGT1*1 mice are dependent on hUGT1A1 expression for survival, this does not mean that the UDCA-and OCA-induced effects in these mice are necessarily or exclusively UGT1A1-mediated. To assess the role of UGT1A1, we therefore also tested both compounds in the Ugt1a-deficient Gunn rat model. In line with our previous observations in adult Gunn rats 13 , UDCA caused a significant decrease in plasma bilirubin levels of neonatal Gunn rats (26%, p < 0.05) whereas no effect was observed on brain bilirubin levels. In contrast, both OCA and GW4064 did not affect plasma or brain bilirubin levels in neonatal Gunn rats (Fig. 5). Unlike OCA and GW4064, UDCA treatment did not activate FXR in liver and intestine, indicating that the hypobilirubinemic effects of UDCA in Gunn rats were independent of FXR activation (Fig. S4).

Discussion
In this study, we show that the therapeutic BAs UDCA and OCA effectively decrease plasma bilirubin in a humanized mouse model of neonatal hyperbilirubinemia. Both compounds also significantly reduced brain levels of bilirubin in these mice, highlighting their potential in the prevention of bilirubin-induced neurotoxicity. Mechanistically, we demonstrated that these effects are at least partially dependent on induction of intestinal UGT1A1 expression and on FXR activation. Neonatal hyperbilirubinemia is caused by a limited capacity to metabolize bilirubin mainly due to delayed expression of UGT1A1. Since rodents express UGT1A1 at an earlier age than humans, they normally do not exhibit neonatal hyperbilirubinemia. For mice, it is known that a complete deficiency in UGT1A1 expression is lethal within the first 4-7 postnatal days 17,20 , and that this condition can be rescued by bilirubin lowering treatments such as intraperitoneal albumin administration, Ugt1a1 replacement gene therapy or by transgenic introduction of the human UGT1A1 gene and its upstream regulatory region 20,25 . In hUGT1*1 mice, UGT1A1 expression in neonates is almost exclusively intestinal, whereas in adults, it is expressed both in intestine and www.nature.com/scientificreports/ liver 17 . In human neonates, the intestinal UGT1A1 expression and its potential contribution to neonatal bilirubin metabolism have never been studied, but there are several indications supporting the concept that intestinal bilirubin conjugation is also important in human physiology. Besides the liver, UGT1A1 is also highly expressed in the small intestine. Already in the '90 s, McDonnel et al. demonstrated UGT1A1 expression and bilirubin conjugation activity in the human intestinal tract 26 . In adult human tissues, the intestinal conjugation of the UGT1A1 substrate estradiol even exceeds the hepatic conjugation capacity 27 . In premature human newborns, UGT1A1 gene expression is repressed in the liver during the first postnatal days 18,19 . It is likely that intestinal UGT1A1 expression is similar between hUGT1A1 mice and human newborns with intestinal UGT1A1 participating in the metabolism of serum bilirubin. However, this will be hard to establish both experimentally and ethically, in particular because healthy neonatal intestinal tissue is hardly available. www.nature.com/scientificreports/ Both UDCA and OCA induced UGT1A1 mRNA and protein expression in duodenum and jejunum, which are also intestinal expression sites in human 26,28,29 . This induction could potentially be FXR-mediated, as we observed a significant bilirubin lowering effect by both OCA and GW4064, which are two structurally unrelated FXR ligands. Despite this, the bilirubin lowering effect by GW4064 and its induction of intestinal Ugt1a1 were less prominent as compared to OCA. These differences are most likely explained by their different chemical structures and pharmacokinetic profiles. While OCA is a bile acid analogue which exhibits bile acid like pharmacokinetics (enterohepatic cycling), GW4064 is a non-steroidal (non-bile acid) synthetic FXR ligand with a relatively poor oral availability and which does not undergo enterohepatic cycling. These characteristics might result in a lower induction of intestinal Ugt1a1 by GW4064.
Alternatively, the induction could be indirectly regulated, for example through activation of the pregnane-Xreceptor (PXR, NR1I2), another member of the nuclear receptor superfamily. PXR activation results in a robust transcriptional induction of the UGT1A1 gene while PXR itself can also be activated by BAs and transcriptionally regulated by FXR 30,31 . Besides regulation at the transcriptional level, it is also possible that hUGT1A1 induction is (partially) mediated at the posttranslational level. OCA, for example, did not increase mRNA levels of hUGT1A1 in duodenum, but did induce the protein expression by 2.5-fold, which can potentially be explained by increased protein stability. Finally, FXR-independent mechanisms could play a role in the induction of hUGT1A1 by UDCA and OCA. To investigate this, however, it would require an Fxr-deficient humanized UGT1*1 mouse model, which is currently not available.
Our work also indicates that UDCA acts independently of UGT1A1, as UDCA treatment decreased TPB in Ugt1a1-deficient Gunn rats. This finding is in line with our previous studies in adult Gunn rats, in which UDCA decreased TPB by 21% after 7 days of treatment 13 . The TPB reduction in Gunn rat pups (-26%) is lower than the reduction observed in hUGT1*1 pups (-86%) and this can potentially be explained by a cumulative effect of UGT1A1-dependent and -independent UDCA effects in the latter.
There are several possible explanations for the hUGT1A1-independent effects. Upon treatment, UDCA becomes the major component of the BA pool, and thereby affects both the hepatic and intestinal BA composition 13 . Changes in intestinal BA composition could cause alterations in both hepatic and intestinal bilirubin metabolism. Hepatic conjugation is not possible in Gunn rats, and in previous work by our group, UDCA did not increase the biliary bilirubin concentration 13 . This makes hepatic interference of UDCA in bilirubin metabolism unlikely. However, in the same study, it was shown that UDCA caused a significant increase in fecal bilirubin, which indicates that UDCA either diminishes the intestinal bilirubin reabsorption or stimulates direct intestinal bilirubin excretion. Possibly, these mechanisms also play a role in the neonatal hUGT1*1 mice, but this could not be tested in the current study since collection of sufficient bile or feces was impossible at this young age.
The fact that OCA and GW4064 did not reduce TPB in Gunn rats, in contrast to hUGT1*1 mice, indicates that hUGT1A1 expression is essential for the bilirubin-lowering effect. Another possible explanation for the absent effect of OCA could be the lower OCA dose used in rats (25/mg/kg/day) compared to mice (50 mg/kg/ day). This approach was chosen because the high OCA dose was poorly tolerated in rats (data not shown). This lower dose could be responsible for the absent OCA effect, but not for the lack of effect by GW4064, which was administered at the same dose to Gunn rats and hUGT1*1 mice.
Although adult Gunn rats have been widely used to study neonatal hyperbilirubinemia, neonatal hUGT1*1 mice appear to be a more relevant model, since like human neonates, they have inducible hUGT1A1 expression www.nature.com/scientificreports/ that mediates the postnatal bilirubin course. Furthermore, it is a strictly neonatal model, which more closely mimics the human neonatal physiologic situation with a yet underdeveloped intestine, that is solely exposed to milk feeding. In adult humans and wild-type rodents, bilirubin levels barely depend on intestinal bilirubin metabolism and reabsorption, since in the presence of a fully-developed intestinal microbiome, bilirubin is effectively converted to urobilinogen and other derivates 32,33 . These metabolites are either not reabsorbed by the intestine, or efficiently excreted via the urine or bile after intestinal reabsorption 34 . However, in an inadequately colonized intestine, as typical for early neonatal period or after oral antibiotic treatment 32 , intestinal reabsorption profoundly contributes to the high TBP levels 35,36 . Therefore, it will be interesting to target intestinal bilirubin metabolism as a treatment strategy, especially in neonatal hyperbilirubinemia. UDCA and OCA alter the intestinal milieu and could thereby affect intestinal bilirubin handling, which could be especially beneficial during the neonatal period. The UGT1A1 gene in neonatal hUGT1*1 mice is transcriptionally regulated by nuclear receptors PXR, the constitutive-androstane receptor (CAR, NR1I3), and peroxisome proliferator-activated receptor α (PPAR α) 30,37,38 . Induction of UGT1A1 was one of the underlying mechanisms of treatment with phenobarbital, a CAR agonist used as a standard treatment for neonatal hyperbilirubinemia before the introduction of phototherapy 39,40 and used in Crigler-Najjar syndrome type 2 41 . hUGT1 mice do not only contain the UGT1A1 gene, but also the upstream regulatory region, including the distal regulatory PBREM that is activated by PXR and CAR 37 . This is important, since the hUGT1A1 promoter regions are not conserved between species and transcriptional regulation of the gene differs between rodents and humans. By merely studying potential therapies in Ugt1a1-deficient models, such as Gunn rats, clinical effects could be missed, as we show for OCA, which did not decrease bilirubin in Gunn rats.
Bilirubin neurotoxicity is caused by brain bilirubin deposition (kernicterus) and therefore, strongly depends on the maturity and permeability of the blood-brain barrier 36 . In homozygous and heterozygous Gunn rats, we observed that brain bilirubin levels are significantly higher in neonates as compared to (young) adults and that the ratio between brain and plasma bilirubin is higher in neonates than in adults (data not shown). This phenomenon has also been described in piglets, in which the ratio between brain bilirubin and free bilirubin in plasma decreases fivefold between P2 and P14 42 . Therefore, we used neonatal animals to study the effects of UDCA and OCA on brain bilirubin deposition and the consequent neurotoxicity risk.
Interestingly, we show that UDCA only decreased plasma, but not brain bilirubin levels in neonatal Gunn rats, whereas it decreased both plasma and brain bilirubin levels in neonatal hUGT1*1 mice. This phenomenon could have several explanations. Firstly, a more profound decrease in plasma bilirubin levels might be needed to induce a decrease in brain bilirubin in the Gunn rat. However, in the hUGT1*1 mice, the relative decrease in brain bilirubin seemed to follow closely the corresponding decrease in plasma. This possibly results from the fact that UGT1A1 is expressed in brain of hUGT1*1 mice. Although bilirubin conjugation in the brain has never been demonstrated, UGT1A1 was previously shown to mediate estradiol conjugation in the brain 43 . Therefore, brain UGT1A1 could potentially contribute to local bilirubin biotransformation and export. Secondly, bilirubin is exported from the brain by species-specific transporters such as Mdr1a P-glycoprotein (Abcb1a), which are differentially regulated between mice and rats 44 . Potentially, UDCA may upregulate a mouse transport protein that is not present or not regulated similarly in Gunn rats.
In conclusion, we show that both UDCA and OCA strongly decrease bilirubin in plasma and brain of neonatal hUGT1*1 mice. Mechanistically, our work indicates that intestinal, rather than hepatic hUGT1A1, plays an important role in these effects. Furthermore, it suggests that OCA acts via hUGT1A1-dependent mechanisms, whereas UDCA also has partial hUGT1A1-independent effects. Both UDCA and OCA are FDA-approved drugs, and UDCA is even already approved for neonatal use, making them readily available for clinical studies. Both compounds could potentially offer a prevention strategy for neonatal hyperbilirubinemia.

Methods
Animals. The generation of humanized UGT1*1 mice (> 99% C57BL/6J background) has been described previously 17   . UDCA was given at 250 mg/kg/day once daily via oral gavage. Both OCA and GW4064 were given at a dose of 50 mg/kg/ day, divided over 2 doses daily. Also, hUGT1*1 mice were treated with the clinically recommended UDCA dose for neonates (30 mg/kg/day in 2 doses per day). All mice were treated daily from P10-14. Mice were sacrificed within 6 h after the last dose on P14 by decapitation under isoflurane anesthesia. Blood was collected in EDTAcoated capillaries from the trunk and stored on ice in the dark upon centrifugation and plasma was stored at -80° C under argon until analysis. Liver and spleen were snap-frozen upon harvesting and subsequently pulverized. The intestine was flushed with ice-cold PBS before snap freezing. Brains were snap frozen and stored in ambercolored tubes at -80 °C. Since OCA is associated with hepatotoxicity, we tested its effect on plasma activities of ALT and AST. OCA treatment induced moderately increased activities of plasma ALT but had no significant effect on plasma AST, suggesting that hepatotoxicity under these conditions was limited (Fig. S5). www.nature.com/scientificreports/ Gunn rat pups were treated with UDCA (250 mg/kg/day) divided over 2 daily doses from P7-14 or with OCA (25 mg/kg/day) or GW4064 (50 mg/kg/day) from P10-14 in one daily dose. Gunn rat pups were terminated on P14 within 6 h after the last dose by cardiac puncture under isoflurane anesthesia and tissues collected and stored as described. All experiments with hUGT1*1 mice were performed at UCSD and approved by the local Animal Ethics Committee of University of California San Diego. All experiments with Gunn rats were performed at the UMCG with the approval of the local Ethics Committee for Animal Experiments of the University of Groningen. All experiments were performed in accordance with relevant guidelines and regulations (including laboratory and biosafety regulations).
Bilirubin analysis. TPB analysis in hUGT1*1 mice was performed using a UNISTAT Bilirubinometer (Reichert Technologies, Depew, NY). TPB analysis in Gunn rats was performed using the Bilirubin Total Gen 3 kit (Roche Diagnostics, Rotkreuz, Switzerland) on a Roche/Hitachi Cobas 501 Analyzer (Hitachi, Tokyo, Japan). Analysis of bilirubin in the brain tissue was performed as described previously 45 . Briefly, approximately 100 mg of homogenized brain was mixed with 50 μL of 5 μM internal standard mesobilirubin (Frontier Scientific, Logan, UT) and homogenized with glass rod. Bilirubin from this mixture was extracted into methanol/chloroform/ hexane (10/5/1, v/v/v). Afterwards, the resulting phase was mixed with n-hexane and carbonate buffer (pH 10), vortex-mixed and centrifuged. Fifty μL of the resulting polar droplet was loaded onto reverse C-8 column (Luna 3 µm, 150 × 4.6 mm, Phenomenex, Torrance, CA) and the amount of bilirubin was determined on HPLC (Agilent 1200 DAD, Agilent, Santa Clara, CA). Concentrations of bilirubin in the brain tissue were expressed in µmol/L of tissue homogenate.
Gene expression analysis. Total RNA was isolated from the liver, small intestine (duodenum, jejunum, ileum) and colon of mice using Trizol (Life Technologies, USA) and reverse transcribed into cDNA using iSCRIPT (Bio-Rad). For quantitative PCR (qPCR), cDNA was amplified using Hi-ROX SensiMix™ SYBR green (Bioline, London, UK) or Taqman fast mix (Applied Biosystems, Foster City, CA) using the Quant Studio Real-Time qPCR (Applied Biosystems) with gene-specific primers (Table S1). Each sample was quantified using a standard curve of the pooled samples and normalized for cyclophilin.
Protein analysis. For immunoblot analysis, tissue lysates were obtained using RIPA lysis buffer (Thermo-Fischer Scientifc, Waltham, MA). After determination of protein concentration, 30 μg protein was loaded on a pre-cast NuPAGE BisTris gel (Thermofisher, Waltham, MA). The resolved protein was transferred onto a 45 µm PVDF membrane (Merck Millipore, Burlington, MA). After blocking, the membrane was incubation with hUGT1A1 primary antibody (ab170858, Abcam, Cambridge, UK) and GAPDH (sc-32233, Santa Cruz Biotechnology, Dallas, TX) and secondary antibody (anti-mouse IgG HRP-linked 7076S and anti-rabbit IgG HRPlinked 7074, Cell Signaling, Danvers, MA) detected by the Clarity Western ECL Substrate (Biorad, Hercules, CA) and visualized using the Bio-Rad ChemiDoc imaging system. Statistical analysis. GraphPad Prism 5.00 software package (GraphPad Software, San Diego, CA) was used to perform statistical analysis. Since the that the data was not normally distributed, significance was determined using nonparametric Mann Whitney U-test. All values are given as means ± standard error unless stated otherwise. Significance is indicated as *P < 0.05, **P < 0.01, ***P < 0.001.

Data availability
The data generated in the current study are available from the corresponding author upon request.