Monitoring bile acid transport in single living cells using a genetically encoded Förster resonance energy transfer sensor


  • Lieke M. van der Velden,

    1. Department of Metabolic Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Misha V. Golynskiy,

    1. Laboratory of Chemical Biology, Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ingrid T. G. W. Bijsmans,

    1. Department of Metabolic Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Saskia W. C. van Mil,

    1. Department of Metabolic Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Leo W. J. Klomp,

    1. Department of Metabolic Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Maarten Merkx,

    Corresponding author
    1. Laboratory of Chemical Biology, Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands
    • Laboratory of Chemical Biology, Department of Biomedical Engineering, Eindhoven University of Technology, Room HeO3.22, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
    Search for more papers by this author
    • fax: (31)-40-2451036

  • Stan F.J. van de Graaf

    Corresponding author
    1. Department of Metabolic Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    • Department of Metabolic Diseases, University Medical Center Utrecht, Room STR3.217, P.O. Box 85060, 3508 AB Utrecht, The Netherlands
    Search for more papers by this author
    • fax: (31)-88-7554295

  • Potential conflict of interest: Nothing to report.

  • Supported by Human Frontier of Science Program Young Investigator Grant RGY0068-2006, The Netherlands Organisation for Scientific Research (S. F. J. v. d. G. project 016.096.108), and the Netherlands Genomics Initiative (NGI-Horizon project 93511019). *These authors contributed equally to this work.


Bile acids are pivotal for the absorption of dietary lipids and vitamins and function as important signaling molecules in metabolism. Here, we describe a genetically encoded fluorescent bile acid sensor (BAS) that allows for spatiotemporal monitoring of bile acid transport in single living cells. Changes in concentration of multiple physiological and pathophysiological bile acid species were detected as robust changes in Förster resonance energy transfer (FRET) in a range of cell types. Specific subcellular targeting of the sensor demonstrated rapid influx of bile acids into the cytoplasm and nucleus, but no FRET changes were observed in the peroxisomes. Furthermore, expression of the liver fatty acid binding protein reduced the availability of bile acids in the nucleus. The sensor allows for single cell visualization of uptake and accumulation of conjugated bile acids, mediated by the Na+-taurocholate cotransporting protein (NTCP). In addition, cyprinol sulphate uptake, mediated by the putative zebrafish homologue of the apical sodium bile acid transporter, was visualized using a sensor based on the zebrafish farnesoid X receptor. The reversible nature of the sensor also enabled measurements of bile acid efflux in living cells, and expression of the organic solute transporter αβ (OSTαβ) resulted in influx and efflux of conjugated chenodeoxycholic acid. Finally, combined visualization of bile acid uptake and fluorescent labeling of several NTCP variants indicated that the sensor can also be used to study the functional effect of patient mutations in genes affecting bile acid homeostasis. Conclusion: A genetically encoded fluorescent BAS was developed that allows intracellular imaging of bile acid homeostasis in single living cells in real time. (HEPATOLOGY 2013)

Bile acids operate as signaling molecules with systemic endocrine functions that are very important for lipid, glucose, and energy homeostasis and are essential for intestinal absorption of dietary fat and fat-soluble vitamins.1 Bile acids are synthesized from cholesterol in the liver, a process that accounts for much of the cholesterol catabolism in the body. Bile acids play a major role in liver regeneration, cancer, and inflammation.2 Inherited or acquired defects in bile acid transport can result in cholestasis or the development of liver or gastrointestinal tumors and are a major cause of liver transplantation in children.3

Despite the clear biomedical importance of bile acids, many questions regarding basic regulation of bile acid synthesis and transport remain unanswered. Intracellular changes in bile acid levels are detected by the farnesoid X receptor (FXR or NR1H4), which binds naturally occurring bile acids and coordinates transcriptional control of bile acid synthesis, import, and export.4-6 However, it is unclear if and how bile acids enter the nucleus, where FXR is normally localized.7 Generally, little is known about subcellular transport of conjugated and unconjugated bile acids. Specific bile acid transporter proteins are known to mediate secretion of bile acids into bile, transepithelial transport of bile acids in the ileum, and uptake from the blood by hepatocytes.3 Current techniques to study bile acid transporters include the use of radiolabeled chemicals or luciferase reporter assays based on FXR-regulated transcriptional activation. These methods involve destruction of the sample (serum or cells) and provide a signal that is averaged over a large pool of cells. Therefore, they do not allow imaging of bile acid levels in single living cells and have limited time and spatial resolution. Studying bile acid efflux and/or intracellular transport processes is even more challenging and typically involves inside-out membrane preparations or isolated organelles and the use of fluorescent bile acid derivatives or radiolabeled bile acids, precluding direct translation of these results to the homeostasis of endogenous bile acids in living cells.

Genetically encoded fluorescent sensors provide excellent tools to determine the concentrations and monitor transport of intracellular metabolites in single living cells in real time.8, 9 Here, we report the development of a genetically encoded Förster resonance energy transfer (FRET) bile acid sensor (BAS) based on the ligand-binding domain (LBD) of FXR. The robust change in emission ratio observed upon bile acid binding to the sensor allows monitoring bile acid levels in real time in single living cells. The sensor is sensitive to a range of physiological and pathophysiological bile acid species and can be readily targeted to different subcellular locations. The reversible nature of the sensor makes it particularly suitable to monitor the activity of bile acid import and export proteins, allowing unique insight into the dynamics of (intracellular) bile acid transport and signaling.


ACP, acyl carrier protein; ASBT, apical sodium bile acid transporter; BAS, bile acid sensor; CA, cholic acid; CDCA, chenodeoxycholic acid; DAPI, 4′,6-diamidino-2-phenylindole; dr, Danio rerio; GCDCA, glycochenodeoxycholic acid; FRET, Förster resonance energy transfer; FXR, farnesoid X receptor; L-FABP, liver fatty acid binding protein; LBD, ligand-binding domain; LCA, lithocholic acid; NCoA2, nuclear coactivator 2; NLS, nuclear localization signal; NTCP, Na+-taurocholate cotransporting polypeptide; OST, organic solute transporter; SNP, single-nucleotide polymorphism; TCDCA, taurine-conjugated CDCA; TLCA, taurolithocholic acid; UDCA, ursodeoxycholic acid.

Materials and Methods

DNA Constructs.

A synthetic DNA construct encoding BAS-0 with noninteracting fluorescent domains was ordered from Genscript (Piscataway, NJ). The cerulean and citrine fluorescent domains present in BAS-0 contained a V224L mutation and a C-terminal deletion of nine residues. The BAS-0 construct contained amino acid residues 258-486 of the human FXR LBD (GENE ID: 9971 NR1H4, amino acid numbering referring to FXRα2) and the nuclear interaction domain 2 peptide (NID2) from nuclear coactivator 2 (NCoA2) (amino acid residues 685-697; GENE ID: 10499 NCOA2). The nuclear localization signal (NLS) from Simian virus 40 (SV40) was included at the BAS C-terminus for nuclear targeting of the sensor. The synthetic BAS-0 construct was cloned into a pET28a vector (Novagen) using the NdeI and XhoI restriction sites. BAS-1 was obtained by introduction of a Q204F mutation in both fluorescent domains of BAS-0 using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) (see Supporting Table 1 for primer sequences). For mammalian cell culture experiments, the BAS-1 coding DNA was cloned into pECFP-C1 (Clontech, Leusden, The Netherlands) to generate NucleoBAS, replacing the ECFP native to that plasmid. CytoBAS was created by KpnI digestion of NucleoBAS, which removed the NLS, followed by self-ligation. More detailed information on the design and construction of the (additional) bile acid sensor constructs and transport proteins is available in the Supporting Information.

In Vitro Bile Acid Titrations.

The bile acids chenodeoxycholic acid (CDCA), taurine-conjugated CDCA (TCDCA), ursodeoxycholic acid (UDCA), lithocholic acid (LCA) and cholic acid (CA) were purchased from Sigma (St. Louis, MO). Cyprinol sulphate was kindly provided by Dr. Alan Hofmann (University of California, San Diego, CA). Sensor proteins were expressed in bacteria and purified as described10 with minor modifications (see Supporting Information).

Live Cell Confocal Experiments for the Intracellular Detection of Bile Acids.

Cells were plated on glass coverslips or Lab-Tek-I chambers (Nunc). Prior to transfection, medium was replaced with Dulbecco's modified Eagle's medium without phenol red supplemented with 1% charcoal–treated fetal bovine serum, 100 U/mL penicillin/streptomycin, 20 mM HEPES, Glutamax, and Puryvate (Gibco). Transfected cells were grown another 18 hours prior to microscopy experiments. Live cell fluorescent imaging was performed at 37°C using a Zeiss LSM 710 Meta confocal laser scanning microscope using the 63×/1.40N.A Plan-Apochromat objective (Carl Zeiss, Jena, Germany). Sensor proteins were excited at 405 nm. The cerulean (450-520 nm) and citrine (520-600 nm) channels were monitored simultaneously. Bile acids were added between cycles. Each experiment ended with the addition of 5 μM GW4064 (Sigma) which generated maximal FRET signal of the sensor. ImageJ software (National Institutes of Health, Bethesda, MD) was employed for data analysis.


Bile Acid Sensor Design and In Vitro Characterization.

The bile acid sensor is a fusion protein consisting of a donor (cerulean) and acceptor (citrine) fluorescent domain, the FXR LBD and a peptide derived from NCoA2, an FXR coactivator protein (Fig. 1A). This peptide contains an LXXLL motif that binds to the FXR LBD in the bile acid bound state, changing the conformation of the sensor and reducing the energy transfer (FRET) between cerulean and citrine. A novel engineering concept was employed that ensures maximal FRET efficiency in the ligand-free state by promoting the intramolecular association of the fluorescent domains through introduction of Q204F and V224L mutations in both domains (Fig. 1A,B).11, 12 The bile acid induced interaction between the FXR LBDs and the NCoA2 peptide disrupts the interaction between the fluorescent domains, resulting in a robust decrease in energy transfer between ligand-free and ligand-bound states. To ensure proper separation between the fluorescent domains in the ligand-bound state, the coactivator peptide was fused close to the core of the citrine domain, deleting the last nine amino acids of citrine's flexible tail (Fig. 1A,B and Supporting Fig. 1).

Figure 1.

BAS design and in vitro characterization. (A) Schematic representation of the BAS. In the absence of bile acids, the FRET signal is maximal due to the presence of mutations that promote intramolecular association between cerulean and citrine. Bile acid binding to FXR LBD recruits an FXR-cofactor peptide (LXXLL) to the FXR LBD, resulting in separation of the fluorescent domains and a decrease in FRET. (B) Domain architecture of the BAS. (C) Emission spectrum of purified BAS-1 in the absence (black line) and presence of 0.5 mM CDCA (red line). (D) Bile acid titration experiments showing the emission ratio of BAS-1 as a function of the concentration of CA (black circles) or CDCA (red triangles). Solid lines represent fits to equation 1 (Supporting Information). The gray symbols show the same titration experiments using a BAS variant (BAS-0) with noninteracting fluorescent domains (BAS-0). (E) Bile acid titration experiments showing the emission ratio of BAS-1 as a function of the concentration of TCDCA (black triangles), UDCA (red squares) or LCA (blue circles) (F) Bile acid titration experiments showing the emission ratio of BAS-1 as a function of the concentration of exogenously added CA in the absence and presence of 50% human serum. Error of the Kd is depicted as the SEM.

In vitro characterization of purified BAS-1 sensor protein showed a high FRET efficiency in the absence of bile acids (Fig. 1C, black line). Addition of bile acid to BAS-1 resulted in a two-fold decrease in the emission ratio, showing that the sensor switches to a ligand-bound, low FRET state in which the fluorescent domains are separated (Fig. 1C, red line). The affinities of BAS-1 for CDCA (Kd = 1.5 ± 0.8 μM) and CA (Kd = 39 ± 4 μM) (Fig. 1D) are similar to determined affinities for the FXR LBD.4, 5 The sensor is also responsive to several other bile acids, including TCDCA, UDCA, and LCA. The titration with LCA can be fit with a single Kd of 0.67 ± 0.06 μM, whereas a biphasic response was observed for TCDCA and UDCA (Fig. 1E). A comparable titration curve was obtained in the presence of 50% human serum, showing that the conformational switching mechanism is maintained in this complex background (Fig. 1F). A sensor construct lacking the Q204F mutations in cerulean and citrine (BAS-0) showed a low level of FRET in the absence of ligand and no significant response upon addition of bile acid (Fig 1D; gray symbols), which clearly establishes the importance of using self-interacting fluorescent domains to obtain a functional sensor. To confirm that the FRET change observed upon bile acid stimulation was actually triggered by the interaction between the FXR LBD and coactivator peptide, we replaced essential leucine residues of the coactivator peptide with alanine residues. The emission ratio of this BAS-1-aa sensor was even higher than that of BAS-1 in the absence of ligand. However, ligand-induced changes in emission ratio were only observed at extremely high concentrations of CDCA, rendering this sensor variant in effect a negative control under physiological conditions (Supporting Fig. 2).

Subcellular Targeting of the Bile Acid Sensor and In Situ Functionality.

Human osteosarcoma (U2OS) cells were initially chosen to test the performance of the BAS sensor in living cells. These adherent cells are readily transfectable, are suitable for confocal microscopy, are unable to synthesize bile acids, and do not endogenously express bile acid transport proteins. To monitor real-time bile acid concentration changes in different organelles, we targeted our sensor construct to the nucleus (NucleoBAS), cytosol (CytoBAS), and peroxisomes (PeroxiBAS) (see Supporting Fig. 1 for an overview of constructs). Sensor bearing a C-terminal NLS (NucleoBAS) was efficiently targeted to the nucleus as demonstrated by complete colocalization with 4′,6-diamidino-2-phenylindole (DAPI), whereas the sensor without this sequence (CytoBAS) mostly showed a cytosolic localization (Fig. 2). Stably transfected sensor constructs containing a C-terminal peroxisomal localization sequence (SKL; PeroxiBAS) colocalized with the peroxisomal marker protein Catalase (Fig. 2). Upon transient overexpression, peroxiBAS is also present in the cytosol (Supporting Fig. 3B), probably due to saturation of the peroxisomal targeting machinery.

Figure 2.

Subcellular targeting of the BAS. Confocal microscopy sections of U2OS cells transfected with NucleoBAS (top row), CytoBAS (middle row), or PeroxiBAS (bottom row). Nuclei were labeled with DAPI. An antibody directed against Catalase was used to identify peroxisomes.

We first measured the response of the BAS sensor to ligands that can readily diffuse through the plasma membrane, the unconjugated bile acid CDCA and GW4064, a potent synthetic FXR agonist with a ≈80 nM affinity.13 Addition of CDCA induced a concentration-dependent increase in FRET emission ratio for both NucleoBAS and CytoBAS (Fig. 3). Similar FRET changes were observed upon addition of CDCA or GW4064 when NucleoBAS or CytoBAS was expressed in a hepatocyte cell line (HepG2; Supporting Fig. 4), cholangiocyte cell line (H69), Madin-Darby canine kidney (MDCK) cells and human embryonic kidney (HEK293T) cells (data not shown). The change in FRET emission ratio upon addition of CDCA or GW4064 was found to be independent of sensor expression level and could also be detected using fluorescence-activated cell sorting (Supporting Fig. 5). Influx of CDCA or GW4064 had no effect on the FRET emission ratio of peroxisome-confined PeroxiBAS (Fig. 3). At present, it is unclear whether this results from inefficient transport over the peroxisomal membrane or differences between the cytosolic and peroxisomal milieu. However, the fraction of PeroxiBAS that is present in the cytosol does respond to addition of CDCA, which shows that the SKL peroxisomal targeting sequence by itself does not interfere with the functionality of the sensor construct (Supporting Fig. 3B).

Figure 3.

In situ functionality of the targeted BAS. (A,B) U2OS cells were transfected with NucleoBAS and CytoBAS. CDCA and GW4064 were added to the final concentrations as indicated. Data were normalized to the maximum emission ratio, which was determined by adding a saturating concentration of GW4064 (5 μM) at the end of each experiment. (C) U2OS cells stably expressing PeroxiBAS were transiently transfected with NucleoBAS. CDCA and GW4064 were added to the final concentrations as indicated. A single region of origin per cell was selected automatically by setting a threshold in image J and mean fluorescence was calculated. The uncorrected ratio between citrin and ceruluan channels is presented, showing similar starting ratio in peroxisome and nucleus. Upon addition of CDCA or GW4064, an increased ratio is only observed in the nucleus. One representative experiment for each construct is shown (three independent experiments; n = 4-6 cells measured per experiment). Error bars represent the SD.

Remarkably, ligand binding invariably induced an increase in FRET in situ, while a decrease in FRET was observed in vitro. This may be due to the presence of a ubiquitously expressed endogenous coactivator or corepressor protein that interacts with the sensor in a ligand-dependent manner. To determine whether the cytosolic environment affects the sensor Kd, we performed an in situ calibration by monitoring the relative response of NucleoBAS as a function of CDCA concentration. The Kd of 2.4 ± 0.2 μM that was determined in situ (Fig. 4, wt) is very similar to the Kd determined in vitro (Kd = 1.5 ± 0.8 μM), which shows that the reversed working mechanism in situ does not interfere with its bile acid affinity.

Figure 4.

Increased dynamic range of the BAS. U2OS cells were transiently transfected with sensor constructs that contained a variety of mutations in the FXR LBD. CDCA was added in final concentrations as indicated to perform in situ calibration of NucleoBAS and NucleoBAS mutants. Data were normalized to the maximum emission ratio, which was determined by adding a saturating concentration of GW4064 (5 μM) at the end of each experiment. Error of the Kd is depicted as the SEM (three independent experiments; n = 4-6 cells measured per experiment).

In cells, the bile acid sensor containing the wild-type FXR LBD was responsive to a concentration range of 1-10 μM of CDCA (Fig. 4, wt). This range was expanded by mutagenizing the LBD of FXR in the NucleoBAS sensor construct. Mutations were chosen that were shown to result in reduced FXR-mediated transcriptional response to CDCA.14-16 The change in FRET ratio upon addition of 5 μM GW4064 was very similar between all sensor variants. The affinity of the NucleoBAS-M328T mutant (Kd = 2.3 ± 0.2 μM for CDCA) was not different from NucleoBAS wild-type (Fig. 4). The affinities of NucleoBAS-L287Y (Kd = 5.9 ± 0.4 μM), NucleoBAS-Y361F (Kd = 8.1 ± 0.3 μM) and NucleoBAS-N354K-I372V (Kd = 10.8 ± 0.8 μM) were significantly lower than that of wild-type NucleoBAS, increasing the dynamic range of bile acid concentrations that can be detected in situ to 30 μM (Fig. 4).

Subcellular Dynamics of Bile Acids.

It was shown that depletion of liver fatty acid binding protein (L-FABP), the most prominent cytoplasmic bile acid binding protein in hepatocytes, affects the activity of FXR.17, 18 Therefore, we determined whether the presence of L-FABP would affect the nuclear FXR–available concentration of bile acid. To this end, we expressed mCherry-tagged L-FABP in U2OS cells, stably expressing the NucleoBAS sensor. CDCA kinetics were compared between cells with and without L-FABP (Fig. 5). In cells without exogenous L-FABP, CDCA rapidly crossed the plasma membrane and subsequently entered the nucleus virtually unimpeded (Supporting Fig. 6). However, the presence of L-FABP significantly slowed down nuclear accumulation of CDCA (Fig. 5). Na+-taurocholate cotransporting polypeptide (NTCP)-mediated uptake of bile acids was unaffected by L-FABP expression (Fig. 5D).

Figure 5.

L-FABP slows down nuclear entry of bile acids. (A) L-FABP, tagged with the red fluorescent protein mCherry, was transiently expressed in U2OS cells stably expressing NucleoBAS. (B) Nuclear bile acid entry upon addition of 10 μM CDCA was monitored in time and fitted to a single exponential association curve. (C) Time constant (tau) of control cells and L-FABP expressing cells (n = 13-15 cells). Error bars indicate the SEM. (D) Tritium taurocholate uptake in U2OS cells cotransfected with NTCP and pcDNA3 (EV), NTCP, and L-FABP or L-FABP only. Error bars indicate the SD (n = 4).

Monitoring Import of Conjugated Bile Acids.

Most endogenous bile acids are conjugated to polar amino acids, which prevents their transport over the cellular membrane by passive diffusion. Instead, transport over the cellular membrane requires the activity of bile acid transport proteins (Fig. 6A). The BAS sensor provides a unique opportunity to study the action of these bile acid transporters in real time in single living cells without a need for labeling the bile acids with radionuclei or fluorescent tags. Coexpression of NucleoBAS and NTCP resulted in a rapid increase in sensor emission ratio upon addition of TCDCA (Fig. 6B, red line), whereas no change in FRET signals was observed for cells expressing only NucleoBAS (Fig. 6B, black line). Similar results were obtained in HepG2 cells stably transfected with SNAP-tagged NTCP or without exogenous NTCP (SNAP only) (Supporting Fig. 4). NucleoBAS was functional in cells without coexpression of NTCP, because increased FRET signal was observed upon addition of CDCA (Fig. 6B, black line). We noticed that the normalized emission ratio of the sensor at the beginning of an experiment is consistently higher for cells transfected with NTCP compared with cells lacking NTCP. Most likely, this is the result of intracellular bile acids imported by NTCP from the culture medium, resulting in an initial activation of the sensor. NTCP-mediated import of even low concentrations of TCDCA (0.3 μM) resulted in almost full activation of the sensor, because no further FRET increase was observed when 3 μM CDCA was added, and only a small signal increase was seen after addition of a saturating concentration of GW4064 (5 μM; Fig. 6B, red line).

Figure 6.

BAS as a tool to monitor cellular bile salt import. (A) Mechanisms for bile acid entry and efflux as elucidated by our FRET sensor. Hydrophobic unconjugated bile salts (depicted by hexagons) can enter and exit the cell by passive diffusion over the plasma membrane, whereas hydrophilic (e.g., CA) or conjugated (taurine- or glycine-) bile salts (depicted by green hexagons with black triangles) are actively taken up via NTCP or ASBT. Export of conjugated bile salts can occur via facilitated diffusion through the heteromeric protein OSTαβ. (B) U2OS cells expressing NucleoBAS and acyl carrier protein (ACP)-tagged NTCP (red line) or NucleoBAS alone (black line). TCDCA, CDCA, and GW4064 were added at concentrations indicated. (C) Changes in emission ratio upon addition of various extracellular concentrations of TCDCA normalized to the maximum emission ratio (using 5 μM GW4064) determined at the end of each experiment (left y axis). The intracellular TCDCA concentration (right y axis) was calculated based on FRET changes and the calibration curve obtained for TCDCA in vitro (Fig. 1E). (D) U2OS cells expressing the zebrafish FXR-based sensor drBAS and zebrafish ASBT (red line) or drBAS alone (black line). (E) Changes in emission ratio of drBAS or CytoBAS (gray bars) upon addition of 10 μM cyprinol sulphate or 10 μM TCDCA. One representative experiment is shown as trace (B,D) or combination of three independent experiments (n = 4-6 cells per experiment) (C,E). Error bars represent the SD.

To further investigate the possible formation of a gradient of conjugated bile acids over the plasma membrane, we next compared the relation between the extracellular and intracellular TCDCA concentration between cells expressing NTCP and cells transiently cotransfected with Organic Solute Transporters α and β (OSTα and OSTβ).19 OSTαβ is a facilitator of conjugated bile acid transport, which implies that OSTαβ can also function as a bile salt importer.19 Coexpression of OSTα and OSTβ in U2OS cells resulted in plasma membrane expression of the OSTαβ heteromer (Supporting Fig. 7A). In cells with OSTαβ the intracellular TCDCA concentration as measured by FRET is similar to the extracellular TCDCA concentration. In contrast, in cells expressing NTCP saturation of the intracellular FRET sensor was observed already at low extracellular TCDCA concentrations (0.3 or 1.0 μM) (Fig. 6C). This result indicates that NTCP mediates active transport resulting in an accumulation of TCDCA inside the cell, whereas OSTαβ constitutes a passive bile acid permeation pathway.

We subsequently employed our sensor approach to functionally characterize the putative zebrafish (Danio rerio) apical sodium bile acid transporter (drASBT). Zebrafish bile does not contain bile acids but instead contains the bile alcohol cyprinol sulphate. We therefore exchanged the human FXR LBD in the sensor for the zebrafish counterpart (drFXR) and used this to determine drASBT-mediated cyprinol sulphate uptake. Expression of drASBT resulted in a rapid influx of cyprinol sulphate, visualized by a decrease in emission ratio between cerulean and citrine, whereas cells without ASBT did not show any FRET change upon addition of this conjugated bile alcohol (Fig. 6D). Using the original cytoBAs sensor, we could establish that drASBT also mediated influx of other conjugated bile acids, such as TCDCA (Fig. 6E, gray bars). Subtle difference in the sensor likely determine the directionality of the sensor, as the drFXR-based sensor shows decreased FRET upon addition of bile alcohols or GW4064.

Influx Analysis of Multiple Bile Acid Species.

Previous research has identified a wide range of bile acids with distinct functional properties, but information on their subcellular distribution, (sub)cellular transport, and binding to FXR is very limited. Current tools to answer these questions, including fluorescent or radiolabeled bile acid derivatives are limited in availability or have properties distinct from unlabeled bile acids. The bile acid sensor allowed us to investigate the influx of a variety of unlabeled bile acid species and whether this influx was dependent on expression of NTCP. Addition of CDCA, deoxycholic acid, LCA, or UDCA resulted in an increase in emission ratio in cells that do not express exogenous bile acid transport proteins, which indicates that these molecules bind the sensor and cross cellular membranes by passive diffusion (Fig. 7A-D, black lines). CA, glycochenodeoxycholic acid (GCDCA), TCDCA, and taurolithocholic acid (TLCA) only caused an increase in FRET ratio in cells that coexpressed NTCP (Fig. 7E-H), indicating that these relatively hydrophilic and conjugated bile acids require active transport by a bile acid import protein to enter the cell and can activate FXR. Furthermore, we demonstrate that all bile acids have sufficient affinity for the LBD of FXR to activate the sensor and that they all, including conjugated bile acids, readily enter the nucleus.

Figure 7.

Cellular influx of physiological and pathophysiological bile acids. (A-H) U2OS cells expressing NucleoBAS and ACP-NTCP (black lines) or NucleoBAS alone (gray lines) for CDCA (A), DCA (B), LCA (C), UDCA (D), CA, (E), GCDCA (F), TCDCA (G), and TLCA (H). Bile acids were added in concentrations as indicated. One representative experiment is shown for each bile acid species tested. For each bile acid, at least two independent experiments were performed in the presence and absence of ACP-NTCP (n = 3-6 cells measured per experiment). Error bars represent the SD.

Functional Analysis of NTCP Mutations.

Several ethnicity-dependent single-nucleotide polymorphisms (SNPs) have been identified in the SLC10A1 gene, which encodes NTCP.20 Here, we used a covalent fluorescent labeling strategy to verify NTCP plasma membrane expression at the single cell level and determine bile acid uptake activity of the same cell using the NucleoBAS sensor. We included NTCP p.S267N, a variant known to have low bile acid transport activity and a profound gain of function for rosuvastatin uptake, NTCP p.E257N and NTCP p.E257Q, carrying mutations at the E257 position, which is critical for bile acid uptake.20, 21 Wild-type and mutant NTCP proteins were coexpressed with NucleoBAS-N354K-I372V, a sensor variant that has a larger bile acid detection range than wild-type NucleoBAS (Fig. 4). Fluorescent labeling of cell surface resident NTCP revealed plasma membrane expression for all NTCP variants (Supporting Fig. 7C). Addition of TCDCA increased the FRET ratio for cells transfected with wild-type NTCP protein, but not for the mutant NTCP proteins, indicating that TCDCA was imported in cells expressing wild-type NTCP but not in cells expressing mutant NTCP (Supporting Fig. 7B). These results are in agreement with uptake experiments using tritium-labeled taurocholate.21 The starting FRET ratios of cells expressing wild-type NTCP was higher compared with cells expressing NTCP mutants, again indicating that wild-type NTCP protein imported bile acids from the medium but that the mutant NTCP variants could not.

Monitoring Bile Salt Export.

Although several methods exist to characterize the activity of bile acid import proteins in intact cells, studying bile acid export is much more challenging and is currently mostly done on inside-out vesicles isolated from virally infected insect cells or from the apical plasma membrane of hepatocytes.22-24 The direct and reversible read-out of our fluorescent sensor should allow the dynamics of both import and export to be measured in intact cells. First, we investigated whether binding of bile acids was indeed reversible when measured in situ. Alternating perfusion of cells with medium lacking bile acids and with medium containing 0, 1, or 3 μM CDCA resulted effectively and repeatedly in a high sensor emission ratio in the presence of bile acids, and a fast decrease in emission ratio upon washout of CDCA (Fig. 8A). This shows that the bile acid sensor can detect intracellular bile acid dynamics in a reversible manner. Next, we studied whether we could exploit this reversibility of the sensor to measure transporter protein-mediated export of conjugated bile acids in living cells. To this end, we performed perfusion of cells coexpressing NucleoBAS and OSTαβ with a high extracellular concentration of TCDCA (10 μM), which increased the FRET ratio, indicating influx of the conjugated bile acid (Fig. 8B). Importantly, a decrease in FRET ratio was subsequently observed when the media was changed to medium without bile acid (wash out), reflecting efflux of TCDCA (Fig. 7B). In the next experiment NTCP was coexpressed with OSTαβ. In this case, a low concentration of 0.3 μM TCDCA could be used compared with expression of OSTαβ alone, because NTCP is a high-affinity bile acid uptake protein (see also Fig. 6). Perfusion of TCDCA resulted in an increase in FRET ratio in cells expressing NTCP (Fig. 8C, gray line) and cells coexpressing OSTαβ and NTCP (Fig. 8C, black line). More importantly, changing to medium without TCDCA resulted in a decrease in FRET ratio for cells that coexpressed the bile salt exporter OSTαβ, but not for cells that only expressed NTCP. These results accurately identify NTCP as bile acid importer and OSTαβ as mediator of bile acid exchange and illustrate the unique ability of the bile acid sensor to study both import and export of (conjugated) bile acids in living cells in real time.

Figure 8.

Dynamic bile acid sensing allows monitoring of bile salt import and export. (A) U2OS cells transfected with NucleoBAS were used in live cell confocal microscopy experiments in combination with a perfusion system. Bile acids were added or washed out as indicated. (B,C) Sensor emission ratio upon addition and washout of bile acids was determined in (B) U2OS cells expressing NucleoBAS and pCDNA3.1-OSTαβ or (C) U2OS cells expressing NucleoBAS and ACP-NTCP (gray line) or NucleoBAS, pCDNA3.1-OSTαβ and ACP-NTCP (black line) by confocal microscopy in combination with a perfusion system. One representative experiment is shown (NTCP, n = 5; OSTαβ and NTCP, n = 2; n = 4-6 cells measured per experiment). Error bars represent the SD.


In this study, we present a genetically encoded FRET sensor that allows real-time monitoring of bile acid concentrations in single living cells. In comparison with existing methods for studying bile acid flux, this fluorescent sensor protein has several key advantages. The robust change in cerulean-to-citrine fluorescence ratio that is observed upon addition, or removal, of bile acids allows for detection of changes in concentration of multiple physiological and pathophysiological bile acid species. Unlike assays employing radioactive bile acids, the use of the genetically encoded fluorescent sensor allows imaging of bile acid uptake at a single cell level with subcellular resolution and monitoring of bile acid levels in complex media such as human serum. Monitoring the kinetics of bile acid transport through ratiometric live cell imaging is relatively straightforward using these new tools. For example, functional characterization of zebrafish ASBT showed that this protein is able to mediate influx of cyprinol sulphate, the main fish bile alcohol, functionally confirming its postulated physiological role in fish. Furthermore, we showed that the functional consequences of coding SNPs or disease-associated amino acid substitutions in NTCP could be assessed by simultaneously studying the effect of these mutations on their cellular localization and bile acid import activity. The bile acid sensor also provides a unique system for direct, real-time measurements of bile acid export in single living cells. To date, bile acid efflux capacity of transport proteins required either the generation of polarized monolayers by stable transfection or viral infection to determine transepithelial flux, or inside-out membrane vesicles, which were generally obtained from insect cells.10, 23 In contrast, the bile acid sensor described here allowed evaluation of the bile acid efflux activity of OSTαβ using a simple expression system, as well as sequential determination of the activity of import and export proteins in living cells. Therefore, this genetically encoded fluorescent sensor should be well suited to investigate bile acid efflux using high-throughput screening approaches, for example, to test novel drugs for potential bile acid–related hepatotoxicity25 or to identify novel genes implicated with familial intrahepatic cholestasis.26

An important distinction between FXR-based transcription activation assays and the FXR LBD based fluorescent sensor reported here, is that the latter directly monitors bile acid–induced conformational changes in the LBD, whereas the transcription-based assays report on the net result of all steps involved in ligand-induced transcriptional activation. Previous studies showed complete loss of CDCA-induced transactivation upon mutagenesis of several FXR residues, including Y361, M328T, and L291,14, 15 whereas only modest effects on the apparent CDCA affinity were observed in this study using a FRET sensor. The deleterious effects of these mutations on transcriptional activation can therefore not solely be attributed to impaired ligand binding, but are likely due to other steps in the transcriptional activation pathway. In fact, the BAS sensor carrying these mutations proved beneficial as they allowed for a larger range of bile acid concentrations to be monitored (Fig. 4). Consistent with earlier work by Cui et al.14 we showed that mutagenesis of two key residues (N354 and I372) by the residues present in the mouse FXR yielded a sensor with a four-fold decrease in CDCA affinity. A further expansion of both the affinity and specificity can be envisioned by exchange of the human FXR LBD by that of other species, as demonstrated by the exchange of the human FXR LBD with the drFXR LBD.

We demonstrated rapid influx and efflux of conjugated and unconjugated, hydrophobic and hydrophilic bile acids into cytosol and nucleus. These and previous results using fluorescent bile acid mimetics27 suggest that bile acids enter the nucleus without apparent delay or obvious transporter requirement. Furthermore, using the sensor, we could demonstrate the role of L-FABP in nuclear entry of bile acids. L-FABP expression impedes CDCA accumulation in the nucleus. This is in line with data from mice lacking L-FABP (L-FABP−/−) displaying altered bile acid metabolism, likely caused by activation of FXR.17, 18 L-FABP expression did not affect NTCP-mediated cellular uptake of bile acids (Fig. 5). This suggests a mechanism in which L-FABP buffers bile acids to prevent excessive (nuclear) FXR activation while allowing cellular transport.

In conclusion, a genetically encoded FRET-based bile acid sensor was developed that provides the unique opportunity to study bile acid dynamics in single living cells with high spatial and temporal resolution. Because this probe can be delivered via simple plasmid transfection protocols, it represents an easily accessible and convenient tool to study cellular bile acid import and export in physiological and pathophysiological conditions in a direct manner in living cells and tissues. The availability of this sensor protein should allow much more detailed studies of bile acid homeostasis in a variety of organelles and a wide range of cell types and facilitate approaches to screen for novel molecular players important for bile acid homeostasis or bile acid–induced signaling.


We thank Paul Dawson for providing the OSTαβ expression construct; Alan Hofmann for providing cyprinol sulphate; and Rina Wichers, Laurens Lindenburg, and Ger Arkenstein for technical assistance.