A novel model to study mechanisms of cholestasis in human cholangiocytes reveals a role for the SIPR2 pathway

Background: Ductular reactivity is central to the pathophysiology of cholangiopathies. Mechanisms underlying the reactive phenotype activation by exogenous inflammatory mediators and bile acids are poorly understood. Methods: Using human extrahepatic cholangiocyte organoids (ECOs) we developed an injury model emulating the cholestatic microenvironment with exposure to inflammatory mediators and various pathogenic bile acids. Moreover, we explored roles for the bile acid activated Sphingosine-1-phosphate receptor 2 (S1PR2) and potential beneficial effects of therapeutic bile acids UDCA and norUDCA. Results: Synergistic exposure to bile acids (taurocholic acid, glycocholic acid, glycochenodeoxycholic acid) and TNF-α for 24 hours induced a reactive state as measured by ECO diameter, proliferation, lactate dehydrogenase activity and reactive phenotype markers. While NorUDCA and UDCA treatments given 8 hours after injury induction both suppressed reactive phenotype activation and most injury parameters, proliferation was improved by NorUDCA only. Extrahepatic cholangiocyte organoid stimulation with S1PR2 agonist sphingosine-1-phosphate reproduced the cholangiocyte reactive state and upregulated S1PR2 downstream mediators; these effects were suppressed by S1PR2 antagonist JET-013 (JET), downstream mediator extracellular signal-regulated kinase 1/2 inhibitor, and by norUDCA or UDCA treatments. JET also partially suppressed reactive phenotype after bile acid injury. Conclusions: We developed a novel model to study the reactive cholangiocyte state in response to pathological stimuli in cholestasis and demonstrated a contributory role of S1PR2 signaling in both injury and NorUDCA/UDCA treatments. This model is a valuable tool to further explore the pathophysiology of human cholangiopathies.


INTRODUCTION
Cholangiopathies are a group of liver diseases that share cholangiocytes as their primary target and have no highly effective or curative treatments.Therapies in use for cholangiopathies include several bile acid treatments.Chief among bile acid-based therapies in use for decades is ursodeoxycholic acid (UDCA), a hydrophilic bile acid present in limited amounts in human bile with both choleretic and anti-inflammatory properties. [1][4][5][6] Since norUDCA is resistant to side chain conjugation, it does not require conjugated bile acid transporters to enter the bile acid pool and appears to engage the cholehepatic shunt. [7]he pathophysiological basis of cholangiopathies is complex.[13][14] Subsequently, the reactive ductular phenotype further enhances cholangiocyte damage by bidirectional communication of mediators with newly recruited inflammatory cells.
Cholangiocytes are exposed to high concentrations of bile acids in normal physiology, principally on the apical surface facing the biliary compartment.In cholestatic states, the intrahepatic concentrations of bile acids rise, and the composition shifts to more hydrophobic and toxic bile acid species with physical exposure of basolateral as well as apical surfaces of cholangiocytes to bile acids, other biliary solutes, and various responsive mediators in the cholestatic liver from adjacent peribiliary cells and hepatocytes. [15][18] In addition to exposure to high concentrations of bile acids during cholestasis, cholangiocytes are also exposed to increased concentrations of proinflammatory cytokines that mediate reactive responses.Cell culture and animal studies indicate that cholangiocytes may be more susceptible to combined stimulation of bile acids and inflammatory mediators than to either of these alone. [19,20]While bile acids mediate their injurious effects through a variety of mechanisms, [21][22][23] recent studies have revealed that the transmembrane sphingosine-1-phosphate receptor 2 (S1PR2) may play a novel and central role in mediating bile acid and inflammation-mediated cholestatic pathology in vivo.S1PR2 responds to select conjugated bile acids to elicit specific patterns of signal transduction (including extracellular signal-regulated kinase [ERK] and serine/threonine kinase (AKT) pathways) in cholangiocytes. [24,25]In addition, some recent studies have reported upregulation of S1PR2 mRNA expression in liver samples from patients with biliary atresia. [26,27]ince the reactive cholangiocyte phenotype has been predominantly studied in animal models, the effects of bile acids on human cholangiocyte injury and activation of their reactive phenotype are poorly understood.In addition, the relevance of animal studies is limited because animals have bile acid compositions that are significantly different from humans [28] and animal studies have shown conflicting results. [29,30]he aim of this study was to develop an in vitro primary human cholangiocyte model that mimics key features of the cholestatic microenvironment.We used well-characterized extrahepatic bile duct-derived primary human cholangiocyte organoids that retain the morphologic, phenotypic, and functional characteristics of cholangiocytes over multiple passages. [31]These cells were exposed to elevated human bile acid concentrations together with increased proinflammatory cytokine levels in order to mimic key components of cholangiopathies to study cholangiocyte injury and activation of the reactive cholangiocyte phenotype.We also investigated the role of the S1PR2 pathway in cholangiocyte injury, reactive phenotype activation, and responses to bile acid-based therapies UDCA and norUDCA as a means to determine if these 2 therapies can elicit effective anti-cholestatic features in a controlled human cholangiocyte model.

Human extrahepatic cholangiocyte organoid culture, injury, and treatments
Human extrahepatic cholangiocyte organoids (ECOs) were generated from cholangiocytes isolated from the normal extrahepatic bile ducts of deceased organ donors as described before, tissue samples were excised after obtaining informed consent from the donor's family in accordance with research ethics committee approval. [31]ECOs were propagated according to the described protocol [31] in serum-free William E medium under 5% CO 2 and ambient O 2 ; passaged every 5-6 days by dissociation in cell recovery solution (Corning) for 30 minutes at 4°C, then collected by centrifugation at 300 g for 3 minutes, washed, and replated to form organoids.Two ECO lines were used for the experiments, data shown are combined results.ECO lines were amplified and frozen between passages 6 and 8.For each experiment, frozen cells were thawed and passaged at least 3 times.All experiments were performed within 15 passages and after organoids were allowed to grow for 5 days.
The human liver has concentrations of taurocholic acid (TCA), glycocholic acid (GCA), and glycochenodeoxycholic acid (GCDCA) ranging from 1 to 4 mM. [18]e first optimized injurious bile acid concentrations by exposing ECOs to either vehicle (DMSO) or individual bile acids (TCA, GCA, GCDCA; Sigma-Aldrich) ranging from 0.5 to 5 mM for 24 hours in William E medium.Using the information from these pilot experiments, ECOs were exposed to vehicle (DMSO for bile acids, PBS for TNF-α) or the optimal target doses of TCA (1 mM) or GCA (1 mM) or GCDCA (0.5 mM) individually; with or without costimulation of 20 ng/mL TNF-α (Peprotech) for 24 hours.NorUDCA and UDCA's [2] effective dosages for the TNF-α+TCA injury model were determined by dose-finding studies.To study potential recovery from the TNF-α+TCA injury model, either vehicle (DMSO) or 250 µM norUDCA or 62.5 µM UDCA were given 8 hours after injury induction by replacement of 50% of the media with fresh media containing the respective treatments for 24 hours.Experiments were performed in duplicate and repeated from 6 to 12 times.

Cyst diameter measurement
ECOs were imaged at the beginning (T0) and end (T24) of treatments using a brightfield, inverted digital microscope (Thermo-Fisher) at ×40 magnification.Image analysis was performed in Image J2 software. [32]ameters (as a morphological indicator of injury) were measured for 20-40 ECOs/sample.For each ECO, diameter change was calculated as diameter at T24-T0.The average diameter change was calculated per sample and then per group.Fold change in diameter was calculated as (average diameter change in treatment)/(average diameter change in vehicle control).

Cell proliferation
ECOs were cultured in glass bottom chamber slides (80807, IBIDI).Ki67, a well-known nuclear marker for cells undergoing mitosis [33] was used to detect proliferating cells.An immunostaining assay was conducted as described [31] using an anti-Ki67 antibody (NB500-170, NovusBio) with Alexa 488 (Invitrogen) secondary antibody.DAPI (Invitrogen) was used to label nuclei.Images were acquired with a spinning disc confocal microscope (Leica) at ×200 magnification and analyzed in Velocity 7 software (Quorum Technologies Inc.).The percentage of Ki67-positive cells in each image field was calculated as ([no. of Ki67-positive nuclei]/[no. of DAPI-positive nuclei] × 100).Six to eight random fields/ samples were counted for analysis.

Organoid permeability assay
Organoids grown on glass bottom chamber coverslips (ibidi) were used for live imaging with Leica SP8 lightning confocal microscope (Leica, Wetzlar, Germany) inside humidified 37°C chambers with 5% CO 2 .The imaging protocol was established based on the previously published method. [34]Briefly, before image acquisition, 1 mM lucifer yellow dye (L0144, Sigma Aldrich) was added to the culture media.Images were acquired in 10 random fields with organoids for 20 minutes at 5-minute intervals after dye addition, at ×200 magnification.Image analysis was performed using Leica LAS X core software.Average background baseline fluorescence was measured from images acquired 1 minute after dye addition.For quantification, the background-subtracted mean fluorescence intensities at 20 minutes were measured inside and outside the organoids within a set region of interest.% Dye intensity inside organoids was calculated by dividing fluorescence intensity inside with total intensity inside + outside organoids.Ten random organoid fields were imaged/sampled in 6 experiments.

Gene expression
Total RNA was extracted in TRIzol (Invitrogen).PuroSPIN total RNA purification kit (NK051-50, Luna Nanotech) was used for RNA purification and genomic DNA removal.Superscript IV first-strand synthesis system (Invitrogen) was used for reverse transcription.Gene expression was analyzed by quantitative real-time PCR using quantitative real-time PCR kit (800-435-UL, Wisent).Primers are listed in Supplemental Table S1, http://links.lww.com/HC9/A813.Gene expression is calculated using the 2 −ΔΔCT method [35] by normalizing against the vehicle control group as 1-fold and relative to the housekeeping gene glyceraldehyde-3phosphate dehydrogenase.

Cytotoxicity
Lactate dehydrogenase (LDH) release is a hallmark of cell death.An LDH-Cytox Assay Kit (426401, BioLegend) was used to access cytotoxicity by measuring LDH activity in cell culture media following 24-hour treatment.The assay was performed according to the manufacturer's instructions; total cell lysate and cell culture media from vehicle control groups were used as high and low controls, respectively.

ERK1/2 phosphorylation
The kinase activity of ERK proteins is regulated by dual phosphorylation at Threonine 202 and Tyrosine 204 in ERK1, and Threonine 185 and Tyrosine 187 in ERK2.ERK1 and 2 phosphorylation at these sites and total endogenous ERK1/2 levels were measured independently by ELISA Kit (ab176660, ABCAM). [36]Cell lysate was collected 24 hours after treatments and assayed according to the manufacturer's instructions.The percentage of phosphorylated ERK1/2 was expressed relative to total ERK1/2.

Statistical analysis
Experiments were repeated at least 4 times and the data were reported as mean ± SD.One-way ANOVA was used to analyze the variation between groups followed by multiple comparisons test by either Dunnet, Benjamini, Fisher's least significant difference, or student t test, where appropriate, using GraphPad Prism (Graph-Pad, San Diego, CA).A value of p < 0.05 was considered statistically significant.

Bile acids and TNF-α exposure induce cholangiocyte injury and the reactive phenotype
In order to determine the response of ECOs to bile acids, we first performed dose-finding cytotoxicity studies for TCA, GCA, and GCDCA.As shown in Supplemental Figure S1, http://links.lww.com/HC9/A813, the concentrations of 1 mM TCA, 1 mM GCA, and 0.5 mM GCDCA were chosen since these were determined to be optimal for injury induction while minimizing cell death (< 5% increase in cell death and <10% increase in LDH activity compared to vehicle control; p < 0.05).In pilot studies, we exposed ECOs up to 72 hours with lower 100 and 200 µM doses of TCA without detecting significant changes in injury parameters (data not shown).
To evaluate the potential impact of cell death in our model, we measured caspase-3 expression and activity.Immunostaining assay showed between 1% and 3% cleaved caspase-3-positive cells after GCDCA stimulation, and even lower expression after TCA and GCA stimulation (Supplemental Figure S2A, B, http://links.lww.com/HC9/A813).There were also no significant changes in caspase-3 activity (Supplemental Figure S2C, http://links.lww.com/HC9/A813).Taken together, these data suggest a minimal contribution of apoptotic cell death in our injury model.
An organoid permeability assay was used to measure membrane-impermeable fluorescent dye lucifer yellow internalization after injury induction, with measurements taken 20 minutes after dye addition.Organoids exposed to vehicle or TNF-α had undetectable amounts of dye inside while TCA, GCA, and GCDCA exposure increased the % dye intensity inside the organoids by 14.5 ± 3.55, 12.17 permeability.TNF-α costimulation further increased dye internalization compared to bile acid alone in most groups (Figure 1D).These data suggest some synergistic effects of combined injury in increasing organoid permeability.

Protective effects of norUDCA and UDCA treatment in the primary human cholangiocyte injury model
We first explored optimal effective and potentially toxic doses of both norUDCA (250-1000 µM) and UDCA (62.5-250 µM) in the standard TNF-α+TCA injury model.As seen in Supplemental Figure S3, http://links.lww.com/HC9/A813, 250 µM norUDCA and 62.5 µM UDCA elicited effective suppression of IL-8 while leading to low levels of LDH activity.Therefore, these doses were used as treatments of ECOs for subsequent experiments.

DISCUSSION
The cholangiocyte reactive phenotype is a central pathophysiologic feature in many cholestatic diseases. [14]In cholestasis, cholangiocytes are exposed to a potentially toxic mixture of high concentrations of hydrophobic bile acids and proinflammatory mediators.How these stimuli induce the cholangiocyte-reactive phenotype and the mechanisms involved are not clearly understood and have not been explored in detail in human cells.Our goal was to create a robust primary human cholangiocyte injury model that emulates key aspects of cholestatic conditions in vitro to study the mechanisms underlying cholangiocyte injury and reactive phenotype activation.
The main findings of this study are (1) hydrophobic primary human bile acids elicit a reactive phenotype activation at high concentrations, many of these injury parameters can be aggravated by proinflammatory cytokine cotreatment; (2) both norUDCA and UDCA treatments are effective in suppressing cholangiocyte injury and reactive phenotype activation in the in vitro cholestatic condition; (3) the S1PR2 pathway may partially contribute to cholangiocyte-reactive phenotype activation after bile acid stimulation, and ERK1/2 and COX-2 signaling could be important downstream effectors; (4) both norUDCA and UDCA appear to protect cholangiocytes from injury and reactive phenotype activation in part due to suppression of the S1PR2 pathway.
Bile acid composition analyses in cholestatic patient serum [16] and bile [18] show markedly increased total bile acid levels.In biliary disorders, total ductular bile acid concentrations can range between 5 and 10 mM, which can be attributed predominantly to millimolar concentrations of primary conjugated bile acids (TCA 2-4 mM, GCA 1.5-2.5 mM, GCDCA 1-2.5 mM, and TCDCA 1-2 mM). [18,39]In our model, we tested these pathophysiological millimolar concentrations of three of the most abundant conjugated primary bile acids TCA, GCA, and GCDCA.Data from our injury models reveal that these high concentrations of bile acids can induce cholangiocyte injury and reactive phenotype activation, evident from their loss of diameter and proliferation, increase in membrane permeability and senescence, and upregulation of select reactive phenotype marker mRNAs.The bile acid concentrations needed to induce injury may correlate with their hydrophobicity and the implicit molecular characteristics of each bile acid. [40]e observed that the more hydrophobic bile acid GCDCA induced injury at lower concentrations than the less hydrophobic bile acids TCA or GCA.As expected, combinations of bile acids and the inflammatory mediator TNF-α as a seminal component of cholestatic conditions further increased injury and reactive phenotype expression.The TNF-α+TCA injury model was selected in this study as TCA is one of the most accumulating bile acids in the cholestatic liver. [41]DCA treatment in animal models of cholestasis appears to elicit its beneficial effects mainly by restoration of biliary bicarbonate secretion, improvement of choleresis, [42][43][44] and protection from cellular senescence and autophagy. [45]NorUDCA (a side chain shortened synthetic version of UDCA) has shown promising results in a recent phase 2 clinical trial of primary sclerosing cholangitis. [2]In some disease COX2, cyclooxygenase-2; ERK1/2, extracellular signal-regulated kinase 1/2; JTE-013, sphingosine 1-phosphate receptor 2/4 antagonist; S1P, sphingosine-1-phosphate; S1PR2, sphingosine-1-phosphate receptor 2; TCA, taurocholic acid; UDCA, ursodeoxycholic acid; Veh, vehicle.models of biliary obstruction and primary sclerosing cholangitis, norUDCA treatment showed enhanced protective effects compared to UDCA, [3][4][5] which may be attributed to norUDCA's ability to undergo cholehepatic shunting and enhance biliary bicarbonate secretion, [7] immunomodulation, [46] and promotion of cellular proliferation and tight junction integrity. [21,47]owever, very little is known about the molecular mechanistic role of either norUDCA or UDCA on cholangiocyte injury and reactive phenotype activation under cholestatic conditions.In our study, norUDCA was better tolerated at high dosages than UDCA; at high doses, UDCA had detrimental cytotoxic effects.In our TNF-α+TCA injury model, both norUDCA and UDCA treatments ameliorated the loss of cyst diameter associated with injury, suppressed the upregulation of cholangiocyte-reactive phenotype markers, and reduced LDH activity associated with cytotoxic effects of injury.NorUDCA also showed a superior ability to suppress cytotoxicity than UDCA in our model.These findings show that norUDCA and UDCA treatments can ameliorate human cholangiocyte injury and reactive phenotype activation under cholestatic conditions.Interestingly, only norUDCA was able to improve cell proliferation after injury.Cholangiocyte injury and activation of the reactive phenotype is a complex process that may be regulated via multiple signaling pathways.A recent study of obstructive cholestasis in a mouse model detected S1PR2 upregulation in cholangiocytes after bile duct ligation, and following TCA stimulation of isolated mouse cholangiocytes. [25]S1PR2 is also upregulated in livers of patients with cholestasis. [26,27]enetic knockout or inhibition of the S1PR2 receptor with antagonist JTE-013 reduced injury and fibrosis associated with bile duct ligation.This was associated with increased ERK1/2 phosphorylation and upregulation of COX-2, a potent modulator of the cellular stress response downstream of S1PR2. [25,38]o test the relevance of the S1PR2 pathway in human cholangiocyte-reactive phenotype activation, damage, and treatment, we first explored whether the S1PR2 pathway agonist S1P and antagonist JTE can modulate these responses in our standard TNF-α+TCA injury model.When the S1PR2 pathway was specifically activated by S1P, the cholangiocyte-reactive phenotype activation profile was recapitulated together with increased LDH activity; these effects were blocked by the S1PR2 antagonist JTE.This suggests that S1PR2 pathway activation can lead to human cholangiocyte reaction and injury.When the antagonist JTE was applied to the injury model, some of the reactive phenotype markers were suppressed, suggesting partial contribution of the S1PR2 pathway in cholangiocytereactive phenotype activation under cholestatic injury conditions.As expected, JTE did not block any of the markers associated with TNF-α exposure alone, as S1PR2 is a receptor for conjugated primary bile acids. [48]However, both norUDCA and UDCA treatments ameliorated the reactive phenotype and LDH activity associated with S1PR2 pathway activation by S1P.This, together with norUDCA and UDCA's ability to ameliorate injury and reactive phenotype after TNF-α +TCA exposure, indicates that the protective effects of these treatments likely encompass multiple contributing pathways, and the S1PR2 pathway may be an important component.Next, we explored the effects of injury and treatments on some of the known downstream mediators of the S1PR2 pathway-ERK1/2 and COX-2. [25,38]Direct activation of S1PR2 by its agonist S1P increased S1PR2 expression, ERK1/2 phosphorylation, and COX-2 expression; these effects were suppressed by antagonist JTE.These observations demonstrate the involvement of the ERK1/2-COX-2 signaling axis downstream of S1PR2 activation in primary human cholangiocytes.Cholestatic injury conditions also upregulated these mediators, but JTE did not show significant suppressive effects, further suggesting the involvement of multiple interconnected pathways that may exert positive feedback to S1PR2.However, when norUDCA or UDCA treatments were administered either after specific S1PR2 activation by S1P or after TNF-α+TCA injury, the S1PR2 downstream mediators were  1) Conjugated bile acids can bind and activate the transmembrane receptor S1PR2, leading to signal transduction via ERK1/2 phosphorylation.This promotes upregulation of downstream cell signaling and transcription regulators such as COX-2, which is a known regulator of inflammatory and fibrotic responses in cholangiopathies.The paracrine mediators released from injured cholangiocytes may initiate positive feedback responses on the S1PR2 pathway and can also contribute to cellular injury and cytotoxicity.(2) Signaling through other transmembrane receptors, (3) bile acid uptake via transporters, and (4) passive diffusion of hydrophobic bile acids through the cell membrane may also contribute to cholangiocyte injury and reactive phenotype activation.NorUDCA and UDCA treatments may protect cholangiocytes from injury by suppressing the S1PR2 pathway and also through modulation of the other injury mechanisms involved.Abbreviations: COX2, cyclooxygenase-2; ERK1/2, extracellular signal-regulated kinase 1/2; S1PR2, sphingosine-1phosphate receptor 2; UDCA, ursodeoxycholic acid.
suppressed.This suggests that norUDCA and UDCA may directly modulate the S1PR2 pathway and may also indirectly interact with other adaptive feedback mechanisms involved in cellular injury under cholestatic conditions.Using U0126, an inhibitor for ERK1/2 phosphorylation, we showed that ERK1/2 is indeed a downstream regulator of S1PR2 signaling in human cholangiocytes and that it can directly contribute to cholangiocyte-reactive phenotype activation.Taken together, our observations not only highlight the S1PR2 pathway as an important mediator of cholangiocyte injury and a potential target for treatment but also underline the importance of further mechanistic studies needed to discover other contributing pathways involved.Furthermore, this indicates that both norUDCA and UDCA can modulate S1PR2 signaling in cholangiocytes to suppress reactive phenotype activation and injury.These putative mechanisms of cholangiocyte injury in response to bile acids and the effect of norUDCA and UDCA treatments are depicted in the cartoon diagram in Figure 6.
In summary, by employing a primary human cholangiocyte model that incorporates the toxic microenvironmental conditions of the cholestatic liver in vitro, we were able to explore specific aspects of their molecular and cellular responses.These studies demonstrate that pathological concentrations of bile acids can induce cholangiocyte injury and the reactive phenotype, which can be aggravated by the presence of a proinflammatory mediator.Both norUDCA and UDCA treatments can reduce human cholangiocyte injury and reactive phenotype activation under test cholestatic conditions.On a molecular level, these studies highlight the S1PR2 pathway as one of the drivers of human cholangiocyte injury and reactive phenotype activation and that the S1PR2 pathway may be an important component of norUDCA's and UDCA's efficacy.We anticipate that this model can be used as a platform to test other cholestatic injury mechanisms and therapeutic interventions for human cholangiocytes.

6
Putative mechanisms of cholangiocyte injury and treatment.(