Stem cell-derived polarized hepatocytes
Nature Communications volume 11, Article number: 1677 (2020) Cite this article
Human stem cell-derived hepatocyte-like cells (HLCs) offer an attractive platform to study liver biology. Despite their numerous advantages, HLCs lack critical in vivo characteristics, including cell polarity. Here, we report a stem cell differentiation protocol that uses transwell filters to generate columnar polarized HLCs with clearly defined basolateral and apical membranes separated by tight junctions. We show that polarized HLCs secrete cargo directionally: Albumin, urea, and lipoproteins are secreted basolaterally, whereas bile acids are secreted apically. Further, we show that enterically transmitted hepatitis E virus (HEV) progeny particles are secreted basolaterally as quasi-enveloped particles and apically as naked virions, recapitulating essential steps of the natural infectious cycle in vivo. We also provide proof-of-concept that polarized HLCs can be used for pharmacokinetic and drug-drug interaction studies. This novel system provides a powerful tool to study hepatocyte biology, disease mechanisms, genetic variation, and drug metabolism in a more physiologically relevant setting.
A major function of the liver is to filter blood from the digestive tract before it passes through the body. This task is performed by hepatocytes, which filter and process blood nutrients, metabolites, hormones, drugs, and other compounds for storage and excretion. Hepatocytes thus form a crucial cell layer engaged in two counter-current flow systems, which, on the one hand, involve uptake, processing, and secretion of sinusoidal blood components, and on the other hand, synthesis and secretion of bile 1 .
To mediate these functions, hepatocytes have a unique polarization with multiple basolateral membranes facing the sinusoids, and multiple apical membranes forming bile canaliculi (Fig. 1a , right panel). Within this peculiar structure, cell signaling, membrane trafficking, protein secretion, and bile transport are highly organized 2 . Although well described morphologically, little is known about the molecules that orchestrate polarization or their regulation, as a robust polarized system to study authentic hepatocyte function is lacking. Research tools are limited because few human hepatoma cell lines can be polarized 1 ; and those that can be polarized are typically de-differentiated with altered proliferative, metabolic, immune, and apoptotic responses.
Fig. 1: Stem cell-based differentiation on transwells to generate polarized hepatocyte-like cells (pol-HLCs).
a Schematic depicting the cellular organization of non-polarized, columnar polarized, and hepatocyte/multi-polarized cells. Blue boxes are tight junctions separating apical and basal membranes. b Stem cell differentiation protocol on transwells to generate pol-HLCs. c Representative immunofluorescence images of hESCs (day 0), definitive endoderm (day 5), hepatic progenitor (day 9 and 13), immature HLCs (day 16), and pol-HLCs (day 21). Cells were stained for immature hepatocyte markers AFP (green) and FoxA2 (magenta) or for mature hepatocyte markers ALB (green) and HNF4α (magenta). Scale bars = 500 μm/250 μm. d–f RT-qPCR analysis of the indicated genes along the pol-HLC differentiation protocol relative to D0 (n = biological replicates). g Pol-HLCs metabolize carboxyfluorescein diacetate (CFDA) and indocyanine green (ICG), and store glycogen as evidenced by periodic acid-schiff staining (PAS) staining. Scale bars = 250 μm h Paracellular permeability of polarized HLCs incubated O/N with 4 kDa FITC-dextran or 70 kDa RITC-dextran in the absence or presence of EDTA, plotted relative to diffusion across empty filters (n = biological replicates). i Rate of albumin (left panel) and urea (right panel) secretion into either the top or bottom compartment during pol-HLC differentiation (n = biological replicates). Statistical analysis was performed using a two-tailed unpaired t test with Bonferroni adjustment for multiple comparisons.
Full size image
Although primary human hepatocytes (PHH) offer a better alternative, they are not extensively used due to their limited availability, high donor-to-donor variability, and limited usefulness for genetic manipulation. In addition, PHH often de-differentiate upon plating and lose their hepatic morphology and functions 1 , 3 . To overcome these limitations, PPH can be expanded in vivo in liver injury mouse models and subsequently plated for in vitro studies 4 , 5 . Yet, potential co-purified mouse cells may, depending of the assay applied, complicate the interpretation of results. Sandwich or micropattern PHH cultures with supporting stromal cells 3 maintain hepatic polarization, but neither the biliary membrane nor the cargo excreted into the closed bile canaliculi are readily accessible (Fig. 1a , right panel). Furthermore, the extracellular matrix overlay hinders solute diffusion and complicates live cell imaging studies. Hepatocyte-based research would thus benefit from more reliable, physiologically relevant, and more experimentally tractable hepatocellular polarity systems.
For these reasons, human embryonic or induced pluripotent stem cell (hESC/iPSC)-derived hepatocyte-like cells (HLCs 6 , 7 ) could offer an attractive option to fill this need for a polarized hepatocyte culture model. Here we report a novel stem cell-based differentiation protocol that generates columnar polarized HLCs. These polarized HLCs secrete cargo directionally and allow non-invasive sampling of both compartments. Albumin, urea, and lipoproteins are secreted basolaterally, with bile acids secreted apically. Enterically transmitted hepatitis E virus (HEV) progeny particles are secreted basolaterally as quasi-enveloped particles and apically as naked virions, recapitulating the natural history of infection in vivo. We also provide proof-of-concept that polarized HLCs offer an attractive platform to test and model drug–drug interaction studies.
Generation of HLCs exhibiting columnar polarization
As described above, maintenance of cell polarity is essential for retaining PHH functions, yet, HLCs are conventionally differentiated in culture dishes under two-dimensional (2D) culture conditions, a process that is inefficient, variable, and yields less-polarized cells. The resulting HLCs are immature and resemble fetal rather than adult hepatocytes 6 . When HLCs are differentiated in spheroids 8 , 9 or cultured in micropatterned co-cultures 10 they better recapitulate hepatocyte functions. This suggests that three-dimensional (3D) architecture assists hepatocyte maturation.
Most epithelial cells, such as lung or intestinal cells, are columnar polarized (Fig. 1a ). Some evidence exists that hepatocyte differentiation passes through a columnar intermediate in vivo 1 . Therefore, we adapted an existing stem cell-based HLC differentiation protocol 11 to generate columnar polarized HLCs on transwell filters (Fig. 1b ). In the past, transwell filters have been extensively used for columnar polarization (Fig. 1a ) of a range of cancer epithelial cells. As the transwell filter is permeable, the configuration permits uptake and secretion of molecules by both, the basolateral and apical sides of the cell, allowing metabolic activities to occur in a more physiological fashion.
We first differentiated hESCs to definitive endoderm (DE) in culture dishes. By day 5, cells expressed lower levels of the pluripotency marker Nanog compared with hESC cells (Fig. 1d ) and high levels of DE markers forkhead box protein A2 (FoxA2) (Fig. 1c ) and C-X-C Motif Chemokine Receptor 4 (CXCR4) (Fig. 1d) . To induce hepatic specification, we then seeded the DE cells on matrigel-coated transwell filters in serum-free, hepatocyte growth factor (HGF)-containing medium. Expression of the biliary markers keratin-19 (KRT19) and prominin-1 (PROM1) increased by day 9 (Fig. 1e ) and plateaued by day 13, at which point the cells expressed nuclear hormone receptor HNF4α, an accepted marker of human hepatic progenitor (HepProg) cells 12 (Fig. 1c ). We further matured the HepProg cells by exposing them to basolateral medium without growth factors in the top compartment, and to complete medium supplemented with HGF and dexamethasone in the bottom compartment. This process yielded immature hepatocytes (ImHep), which by day 16 expressed high levels of the fetal liver marker alpha-fetoprotein (AFP) (Fig. 1c and f). ImHeps were further matured by exposing them to basic hepatocyte culture medium (HCM) in the top compartment and complete HCM medium supplemented with oncostatin-M in the bottom compartment. By day 21, this process yielded HLCs that expressed high levels of the adult hepatocyte marker albumin (ALB) (Fig. 1c and f). By counting ALB-positive cells, we determined that ~80% of the final cell population consisted of HLCs (Fig. 1c ). Notably, HLCs underwent hepatic multipolar polarization and formed apparent bile canaliculi when grown on transwell filters with fully supplemented medium in both compartments and overlaid with matrigel throughout differentiation.
Having successfully generated HLCs on transwell filters, we next examined several characteristic hepatic functions. The cells stained positive with indocyanine green, a tricarbocyanine dye that is taken up by hepatocytes (Fig. 1g ). The HLCs also demonstrated the ability to synthesize glycogen, as tested by periodic acid-Schiff staining. We further demonstrated that HLCs could metabolize non-fluorogenic carboxyfluorescein diacetate into fluorogenic carboxyfluorescein (Fig. 1g ). Next, we showed that the cells formed a tight monolayer by measuring transepithelial electrical resistance (~400 Ω/cm2) and assessing diffusion of small fluorescent dextrans (4 kDa and 70 kDa), which could not be detected in the opposite transwell compartment in the presence of polarized HLCs (Fig. 1h ). From day 16 of the differentiation protocol, HLCs secreted characteristic hepatocyte cargos such as ALB and urea (Fig. 1i ). This coincided with positive ALB staining (Fig. 1c ).
This protocol was generalizable to multiple stem cell lines, including hESC lines RUES2 13 and HUES8-iCas9 14 (Supplementary Fig. 4 ) as well as the iPSC line iPS.C3A 15 . Importantly, the majority of ALB and urea, normally secreted by native hepatocytes from their basolateral membrane into the bloodstream, was secreted into the bottom compartment (Fig. 1i ). These results indicate that the cells were differentiated into metabolically functional polarized HLCs (pol-HLCs), with what appeared to be simple, columnar epithelial polarization (Fig. 1a ). Of note, pol-HLCs continued to secrete ALB and urea for another 22 days and 17 days, respectively (Supplementary Fig. 1 ).
Properties of transwell-polarized HLCs
We then compared the cellular structure of pol-HLCs with conventionally differentiated HLCs on Matrigel-coated culture dishes. As shown in Fig. 2 , conventional HLCs exhibited some level of cellular polarity with what appeared to be apical villi (Fig. 2a ) and polarized distribution of some of the proteins analyzed (Fig. 2d ). Yet, their degree of polarization is markedly less compared with pol-HLCs differentiated on transwells, as demonstrated in the following experiments. For simplicity, we refer to conventionally plated HLCs that have not been through our columnar polarization protocol as “nonpol-HLCs”.
Fig. 2: Structural polarization and organization of polarized HLCs.
a Transmission electron microscopy, cross-sectional view of nonpol- and pol-HLCs. No, nucleolus; Nu, nucleus; Mi, mitochondria; ER, endoplasmic reticulum; Lys, lysosome; TJ, tight junctions. Scale bar in insets = 2 μm. b xy images of pol-HLCs stained for the tight-junction marker ZO-1 (green), breast cancer resistance protein (BCRP, green), multi-drug resistance protein 2 (MRP2, green), or scavenger receptor-B1 (SR-BI, green), and DAPI (blue). Scale bars = 200 μm c Transferrin-conjugate binding to pol-HLC. Pol-HLCs were incubated with 25 µg/mL Transferrin-594 (green) for 10 min at 37 °C prior to washing and staining with anti-ZO-1 (magenta). Bottom panels: xz images from cross sections indicated by the dashed line in the corresponding xy-images above. Scale bars = 15 μm. d Cross-sectional views (xz) of nonpol- and pol-HLCs stained for indicated marker. Yellow arrows = basolateral membrane of pol-HLCs. * = autofluorescence of transwell pore. Images are representative of three independent differentiations.
Full size image
We then tested pol-HLC for modeling an antiviral combination drug regimen that has well characterized pharmacokinetics (Fig. 5c ). For this we chose STRIBILD, a fixed dose anti-HIV combination therapy. This regimen includes emtricitabine (FTC), tenofovir disoproxil fumarate (TDF), elvitegravir (EVT), and cobicistat (COBI). To model STRIBILD disposition, we incubated pol-HLCs from their basolateral membrane with FTC, TDF, and EVT combined at a ratio mimicking the approved antiviral regimen with or without COBI. Unlike the other three compounds, which are direct virus replication inhibitors, COBI blocks the drug-metabolizing CYP3A4 enzymes, as well as drug transporters. We then measured the absorption and secretion from either basolateral or apical membrane over a 2 h period using LC-MS. As shown in Fig. 5c , the cells absorbed each drug at a different rate, as evidenced by the steady decrease of free compound in the bottom transwell compartment. EVT and TDF were rapidly absorbed, whereas FTC, which is known to have a long plasma half-life 44 , was slowly absorbed. Concomitant with the decrease in the bottom transwell compartment, the amount increased in the top compartment, suggestive of partial biliary release from the apical membrane. We also detected a slow release of tenofovir (TF) in both compartments, demonstrating that pol-HLCs metabolized TDF converting it to TF. Co-administering COBI only affected EVT absorption, which unlike FTC and TDF, is metabolized by the CYP3A4 enzyme family. COBI decreased the rate of EVT absorption by approximately threefold and reduced apical EVT release by ~50%. This is in agreement with EVT’s increased plasma half-life (from 3–9 h) when boosted by COBI, allowing it to be dosed once daily 45 , 46 . These observations show that COBI helps sustain basolateral levels of EVT in blood by reducing its uptake and metabolism in hepatocytes. In contrast, COBI did not affect FTC or TDF levels, which suggested that the observed decrease of EVT is not owing to diffusion between the two transwell compartments but rather is limited by the action of enzymes and transporters. Therefore, the pol-HLC system recapitulates observations made in humans and demonstrates the potential application to pre-clinical drug development 46 .
Cell polarity is based on the asymmetric organization of cellular components, and is a pre-requisite for fundamental biological processes. It enables a polarized cell to ensure directional cargo transport and release while maintaining a barrier within the epithelium. Although most epithelial cells typically establish a columnar apical-basal polarity, hepatocytes distinguish themselves by their multipolar organization, which allows them to exert their particular activities in the 3D environment of the liver. Current in vitro models for hepatocyte polarity studies are suboptimal. Only few hepatoma cell clones can be columnar polarized, but their transformed nature makes them undesirable for many applications. Transferring already differentiated, non-proliferative PHHs onto the transwell membrane does not give rise to PHH with columnar polarization but rather induces restoration of their previously formed multipolar structure, which restricts apical and basal cargo sampling (data not shown). Similarly, plating human fetal hepatoblasts (usually from week 15 to 21 of gestation) from different donors onto the transwell membrane also fails to form a tight monolayer and columnar polarized hepatocytes (data not shown). Here, we report a stem cell-based differentiation protocol to generate polarized HLCs on transwell filters. The use of transwell filters yielded a columnar polarization that was likely achieved and supported by the one-sided exposure to ECM and appropriate nutrient gradients.
We first examined the structural polarization of pol-HLCs by analyzing the localization of hepatocyte membrane proteins. Immunofluorescence staining showed that pol-HLCs have clearly defined basolateral and apical membranes separated by tight junctions. We also showed that pol-HLCs possess functional polarity. Hepatic cargo destined to reach the bloodstream such as albumin, urea, and lipoproteins were secreted basolaterally. In contrast, bile acids, destined for secretion into the biliary system, were secreted apically from pol-HLCs. We further found that pol-HLCs can recapitulate the directionality of HEV infection in vivo: HEV progeny particles are secreted basolaterally as quasi-enveloped particles and apically as naked virions. Knocking down CYP8B1, a key enzyme in bile acid metabolism, revealed that co-secreted bile acids strip the envelope from apically released virions rendering them highly infectious. Finally, by showing that pol-HLCs faithfully replicated hepatocyte uptake and biliary/blood excretion of the once-daily anti-HIV regimen STRIBILD, we provided proof-of-concept that polarized HLCs can be used for pharmacokinetic and drug–drug interaction studies.
We are aware that despite recapitulating some hepatocyte functions better than nonpol-HLCs, pol-HLCs still retain an immature phenotype as evidenced by the transcriptome comparison with PHHs (Supplementary Fig. 2 ). Efforts are ongoing by multiple labs to optimize protocols and identify compounds and conditions that can enhance maturation of HLCs. Success and combination with the protocol described here may eventually yield highly functional polarized HLCs that better recapitulate fully mature hepatocyte functions. However, the system described here has several immediate advantages. Unlike PHHs, pol-HLCs are derived from a renewable, reproducible, and cost-effective source. Furthermore, the resulting apical-basal polarity and release of cargo in separate compartments allows easy sampling and analysis over time. This, together with the ability to genetically manipulate cellular proteins of interest such as membrane transporters, will help advance our general understanding of the polarized hepatic trafficking machinery and how components are selectively targeted to the apical versus the basolateral compartment.
Members of the CYP450 family and other phase I, II, and III drug-metabolizing enzymes have poor expression and induction levels in hepatoma cells, which is why they are unsuitable for drug metabolism and disposition studies. As mentioned before, HLCs have been proposed as an attractive alternative to overcome these limitations. In addition, as HLCs can be generated from patient-specific iPSCs 47 , studying the impact of genetic polymorphism and inter-individual variations on drug exposure and toxicity can facilitate the development of personalized therapies. The ability of pol-HLCs to excrete drugs from either the basolateral or apical membrane will improve these types of studies.
In conclusion, this novel stem cell differentiation protocol provides a powerful cell culture system to study proper hepatocyte function, which will lead not only to a better understanding of normal liver physiology, but also holds promise for informing therapeutic options and drug development.
Reagents and antibodies
The following antibodies were used for immunofluorescence staining or western blot analyses: anti-FoxA2 (used at 1:400, Cell Signaling), anti-HNF4α (used at 1:500, Cell Signaling), anti-AFP (used at 1:1000, Sigma-Aldrich), anti-ALB (used at 1:1000, Cedarlane, Burlington, Canada), anti-ZO-1 (used at 1:1000, Thermo Fisher), anti-E cadherin (used at 1:500, Cell Signaling), anti-SR-BI (used at 1:100, Novus Biologicals), anti-BCRP (used at 1:100, Millipore), anti-MRP2 EAG5 48 (used at 1:200, a kind gift from Anne Nies, IKP Stuttgart 48 ), anti-Apo-CIII (used at 1:500, Abcam), anti-CYP8B1 (used at 1:100, Abcam), anti-ORF2 (used at 1:400, a kind gift from Suzanne U. Emerson, NIH) and anti-HAV capsid (used at 1:1000, a kind gift from Stanley M. Lemon, UNC School of Medicine). Alexa Fluor 488 and 549 anti-mouse (used at 1:1000) and Alexa Fluor 488 and 549 anti-rabbit (used at 1:1000) antibodies were purchased from ThermoFisher. Alexa 594-conjugated transferrin was purchased from ThermoFisher. Tenofovir, Tenofovir disoproxil fumarate, and Emtricitabine were obtained through the AIDS Reagent Program, Division of AIDS, NIAID, NIH. Elvitegravir and Cobicistat were purchased from SelleckChem. Sofosbuvir was purchased from Acme Bioscience. BX795, oleic acid and lomitapide were purchased from Sigma-Aldrich.
Polarized HLC stem cell differentiation
In all, 2×105 cells/cm2 of ESC or iPSC were differentiated to DE by harvesting them with gentle cell dissociation reagent (Stemcell Technologies) and plating onto Matrigel (Corning, Catalog number 354230)-coated culture dishes (Corning) in mTeSR1 medium (Stemcell Technologies). The next day, DE differentiation was initiated by using the STEMdiff Definitive Endoderm Kit (Stemcell Technologies). To induce hepatic differentiation, DE cells were harvested using Accutase (Innovative Cell Technologies), re-seeded in Matrigel-coated transwells (Transwell, Corning, Catalog number 3460) as described in Results and cultured in the presence of medium A (basolateral medium (BM): CTS KnockOut DMEM/F12, 10% KnockOut Serum Replacement, 0.5% GlutaMAX supplement, and 0.5% non-essential amino acids all from ThermoFisher Scientific, supplemented with 100 ng/ml HGF), 1% DMSO) for 8 days followed by incubation in medium B (BM, 100 ng/ml HGF, 1% DMSO, 40 ng/ml dexamethasone) for three 3 days. Cells were further matured in HCM (Lonza, omitting the EGF) supplemented with 20 ng/ml oncostatin-M for 5–7 days. For nonpol-HLCs, DE cells were seeded on Matrigel-coated culture plates and step-wise matured in the media described above. HGF was purchased from Peprotech, dexamethasone from Sigma, and oncostatin-M from R & D Systems.
Isolation and culture of PHHs
PHHs were isolated from patient liver tissue after partial hepatectomy. The protocol was authorized following written informed consent of the patients and approved by the ethics commission of Hannover Medical School (EthikKommission der MHH, Nr. 252-2008). Isolated PHHs were seeded on collagen-plated plates and used for studies 48 h post plating.
RT-PCR and real-time quantitative RT-PCR
Total RNA was isolated from cell lysates using the RNeasy Mini Kit (Qiagen) followed by reverse transcription using Superscript III Reverse Transcriptase (Thermo Fisher Scientific). Gene expression was quantified using the LightCycler SYBR Green I Master mix (Roche Life Science, Indianapolis, IN) on a LightCycler 480 Instrument I (Roche Life Science) with primers as listed in Supplemental Table 1 . Relative expression data were calculated into fold changes based on cycle thresholds.
Fluorescent dextran assay
Polarized HLCs were incubated with 4 kDa fluorescein isothiocyanate–dextran (Sigma-Aldrich) or 70 kDa rhodamine B isothiocyanate–dextran (Sigma-Aldrich) diluted in either apical or basolateral HCM medium to a final concentration of 1 mg/ml overnight (O/N) at 37 °C. To dissociate cell–cell junctions, 2.5 mm EDTA (Gibco) was added. Fluorescence was measured using a FLUOstar Omega plate reader (BMG Labtech) (FITC-dextran: Exc: 485 nm and Em: 544 nm and rhodamine B-dextran: Exc: 520 nm and Em: 590 nm).
Transmission electron microscopy
Cells were fixed with 2% paraformaldehyde (PFA) and 2.5% glutaraldehyde in 0.075 m sodium cacodylate buffer pH 7.4. Subsequently, cells were washed in the buffer, post-fixed with 1% osmium tetra-oxide for 1 hr, underwent en bloc staining with 1% uranyl acetate for 30 min, dehydrated by a graded series of ethanol, infiltrated with a resin (Eponate12, Electron Microscope Sciences) and embedded with the resin. After polymerization at 60 °C for 48 h, ultra-thin sections were cut, underwent post-staining with 2% uranyl acetate and 1% lead citrate and were examined under a JEOL JEM 1400Plus transmission electron microscope equipped with SerialEM 49 in montage mode and the digital imaging system (Gatan Digital Micrograph 1000) (a gift from the Helmsley Charitable Trust). EM data were processed by IMOD 42 .
Immunofluorescence staining and analysis
Cells were fixed in 4% PFA in phosphate-buffered saline (PBS) at room temperature (RT) for 30 min and blocked with 3% bovine serum albumin in PBS (3% B-PBS). Cells were incubated with primary antibodies in 3% B-PBS at 4 °C overnight. Secondary antibodies conjugated to Alexa Fluor 594 (Thermofisher) or Alexa Fluor 488 (Thermofisher) in 3% B-PBS were added and incubated at RT for 1 hr, followed by several washes with PBS. The staining with transferrin-594 conjugate (Thermofisher) was performed following the manufacturer’s protocol. For HLC cultures grown on transwell filters, the filter was removed from the hanging insert and submerged, cell side up, in PBS using a slice anchor (Warner Instruments). Cultures were imaged on an Upright BX61WI microscope (Olympus) with a UMPlan FL 60 × 1.0 NA water dipping objective (Olympus) or UMPlan FL ×10, 0.3 NA water dipping objective and an Orca Flash 4.0 digital CMOS camera (Hamamatsu) using MetaMorph image acquisition software (Molecular Devices). Deconvolution was performed using the standard adaptive point spread settings in Autoquant (Media Cybernetics). Image analysis was conducted using Fiji.
Approximately 100 ng of total RNA isolated from biological duplicate samples using the RNeasy mini Kit (Qiagen) was used as input. Sequencing libraries were constructed using the TruSeq Stranded mRNA LT Sample Prep Kit (Illumina) and sequenced on an Illumina HiSeq with a read length of 51 nt (single end, reverse stranded). Output sequencing read quality was analyzed using the Seqtk (version 1.2) and fastx_toolkit (version 0.0.14) software tools and used without further processing. The reads were mapped to the Ensembl human genome assembly GRCh37 (also known as hg19) using Tophat2 (version 2.0.12; Bowtie version 2.2.7) with first-strand library type, no novel junctions, and otherwise default options. Mapped reads were counted using featureCounts (from subread version 1.4.6) with default options. Statistical analysis was performed using a count-based workflow 50 with the edgeR Bioconductor package (version 3.12.1). In brief, gene counts were normalized to counts per million reads and genes with