Novel 3D Culture Systems for Studies of Human Liver Function and

Review pubs.acs.org/crt

Novel 3D Culture Systems for Studies of Human Liver Function and Assessments of the Hepatotoxicity of Drugs and Drug Candidates Volker M. Lauschke,*,† Delilah F. G. Hendriks,† Catherine C. Bell,† Tommy B. Andersson,†,‡ and Magnus Ingelman-Sundberg† †

Section of Pharmacogenetics, Department of Physiology and Pharmacology, Karolinska Institutet, SE-17177 Stockholm, Sweden Cardiovascular and Metabolic Diseases, Innovative Medicines and Early Development Biotech Unit, AstraZeneca, Pepparedsleden 1, Mölndal, 431 83, Sweden

‡

ABSTRACT: The liver is an organ with critical importance for drug treatment as the disposition and response to a given drug is often determined by its hepatic metabolism. Patient-specific factors can entail increased susceptibility to drug-induced liver injury, which constitutes a major risk for drug development programs causing attrition of promising drug candidates or costly withdrawals in postmarketing stages. Hitherto, mainly animal studies and 2D hepatocyte systems have been used for the examination of human drug metabolism and toxicity. Yet, these models are far from satisfactory due to extensive species differences and because hepatocytes in 2D cultures rapidly dedifferentiate resulting in the loss of their hepatic phenotype and functionality. With the increasing comprehension that 3D cell culture systems more accurately reflect in vivo physiology, in the recent decade more and more research has focused on the development and optimization of various 3D culture strategies in an attempt to preserve liver properties in vitro. In this contribution, we critically review these developments, which have resulted in an arsenal of different static and perfused 3D models. These systems include sandwich-cultured hepatocytes, spheroid culture platforms, and various microfluidic liver or multiorgan biochips. Importantly, in many of these models hepatocytes maintain their phenotype for prolonged times, which allows probing the potential of newly developed chemical entities to cause chronic hepatotoxicity. Moreover, some platforms permit the investigation of drug action in specific genetic backgrounds or diseased hepatocytes, thereby significantly expanding the repertoire of tools to detect drug-induced liver injuries. It is concluded that the development of 3D liver models has hitherto been fruitful and that systems are now at hand whose sensitivity and specificity in detecting hepatotoxicity are superior to those of classical 2D culture systems. For the future, we highlight the need to develop more integrated coculture model systems to emulate immunotoxicities that arise due to complex interactions between hepatocytes and immune cells.

■

CONTENTS

1. Introduction 2. Hepatocyte Sandwich Cultures 3. Spheroid Models 3.1. HepG2 Spheroid Systems 3.2. HepaRG Spheroid Systems 3.3. Upcyte Hepatocytes 3.4. Primary Human Hepatocyte Spheroid Systems 3.5. Stem-Cell Derived Hepatocyte Spheroid Systems 3.6. Spheroid Models Summary 4. Hollow Fiber 3D Bioreactors 5. Organ-on-a-Chip Platforms 5.1. Micropatterned Cocultures 5.2. Perfused Multiwell Plates 5.3. Microfluidic Liver Biochips 5.4. Microfluidic Multiorgan Chips 5.5. 3D Liver Bioprinting © XXXX American Chemical Society

6. Conclusions Author Information Funding Notes Biographies Acknowledgments Abbreviations References

A D E E E E

M M M N N N N N

G

1. INTRODUCTION The liver is an organ with major physiological functions. These include control of blood sugar and ammonia levels, synthesis of

H I I J J J K K L

Special Issue: Mass Spectrometry and Emerging Technologies for Biomarker Discovery in the Assessment of Human Health and Disease Received: May 3, 2016

A

DOI: 10.1021/acs.chemrestox.6b00150 Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology

Review

can retrospectively detect the toxicity of several hepatotoxins, their capability to accurately predict the potential of new chemical entities to induce DILI is largely limited, primarily because of the diversity and complexity of toxicity mechanisms and the long exposure times required. Before the emergence of cell culture systems, animal models served as the tool of choice for the screening of pharmacological compounds and are still required by legislation to evaluate the safety and efficacy of all newly developed compounds. 22 However, animal studies face important limitations as drug pharmacokinetics and dynamics differ considerably between animal models and humans; a study of 150 compounds revealed that only 63% and 43% of human toxicity was detected by nonrodent and rodent experiments, respectively.23 Thus, screening in animal models results in substantially high fractions of false-negative as well as falsepositive findings, which are both detrimental for drug development. Therefore, the need for methodologies that can reliably forecast drug responses to newly developed medications in humans has driven the development of human cell culture methodologies. Classical 2D monolayer culture systems have been and still are extensively used for studies in early stages of the drug development pipeline as they are easy to use and amenable to screening large numbers of compounds in a short amount of time. Recently, members of an Innovative Medicines Initiative (IMI, MIP-DILI)24 undertook an interlaboratory assessment of a number of human hepatic 2D cell models (primary human hepatocytes, HepaRG, HepG2, and Upcyte hepatocytes) in order to determine their utility in predicting the hepatotoxic potential of known DILI compounds.25 In this study, the authors found that primary human hepatocytes (PHH) were the most sensitive cell type, identifying eight out of nine hepatotoxic test compounds, followed by HepG2 cells (6/9), whereas HepaRG cells and Upcyte hepatocytes were relatively insensitive (3/9). Yet, mounting evidence indicates that biochemical cues and cell− cell communications necessary for maintaining the physiological phenotype of the cells are lost in 2D monolayer culture, also for PHH.26 Furthermore, the stiffness of the culture substrate, usually glass or plastic, has been shown to tremendously impact expression signatures and cell fates.27 Hepatocytes respond very profoundly to these nonphysiological microenvironments and culture conditions, which manifests in the rapid loss of expression of hepatic marker genes and acquisition of fetal-like phenotypes within hours in a process called dedifferentiation.28,29 As a consequence, longterm repeated dose studies, which better reflect in vivo exposure patterns, are not feasible in hepatocyte 2D monolayer cultures, hampering the translation of results to patients that undergo standard drug treatment regimens for prolonged periods. Ideally, in vitro systems for studies of drug metabolism should possess several features to be reckoned as reliable tools for pharmacological applications and toxicity assessments: (1) Cellular phenotypes at the molecular level should accurately correspond to the phenotypes observed in vivo, including the expression of P450 and phase II enzymes, transporters, and nuclear receptors. (2) Viability, functionality, and phenotypes of cells should be stable for multiple weeks in culture to allow chronic toxicity studies, as DILI often only manifests after repeated exposures in humans. (3) Systems should permit the coculture of hepatocytes with nonparenchymal cells (NPCs) to mimic the cellular repertoire of the intact organ, in order to capture also complex DILI events that only manifest due to an

various hormones, storage of vitamins and iron, as well as detoxification of a plethora of endogenous and exogenous substances. Furthermore, the liver is distinguished in its ability to regenerate after physical or chemical injury, which was already recognized by the ancient Greeks in the myth of Prometheus. Yet, when liver damage surpasses a certain threshold and becomes too severe, the liver loses the ability to perform its physiological functions, and orthotopic liver transplantation is the only therapeutic option and such endstage liver diseases represent the prime cause for liver transplantations.1 The most significant liver diseases are (i) viral hepatitis,2 (ii) alcoholic liver disease,3 (iii) nonalcoholic fatty liver disease (NAFLD), which encompasses a whole spectrum of disorders from steatosis to nonalcoholic steatohepatitis (NASH) to fibrosis,4 (iv) cirrhosis, the formation of fibrous tissue surrounding nodes of regenerating cells which leads to the collapse of liver structures,5 (v) primary liver cancer,6 (vi) primary biliary cirrhosis, an autoimmune disease leading to the progressive destruction of bile ducts,7 and (vii) primary sclerosing cholangitis, a disease in which chronic inflammation causes scarring and eventual obstruction of bile ducts.8 Besides liver diseases, adverse drug reactions (ADRs) are major causes of liver injury. In total, around 7% of hospital admissions are caused by ADRs,9 and up to 15% of inpatients experience drug-related adverse reactions.10 Furthermore, 0.3% of all hospitalized patients develop fatal drug-related complications, causing more than 100,000 deaths annually in the US.11 Moreover, ADRs and safety issues are prime causes for the termination of drug development projects as well as for the withdrawal of drugs from the market.12,13 They can arise from unanticipated interactions with coadministered medications, be the consequence of sensitization due to acute or chronic concomitant diseases, or be facilitated by genetic factors, such as variations in cytochrome P450 (P450) genes and drug transporters, which have been shown to be highly polymorphic.14,15 While the contribution of genetic variants has been estimated to account for 20 to 30% of the interindividual variability in drug response,16 recent research indicated that up to 40% of this variability is due to rare mutations, which are not assessed by current clinical genotyping platforms.15,17 Druginduced liver injury (DILI) is among the most common ADRs, rendering the liver a priority target for drug safety studies.18 Furthermore, liver toxicity has received particular attention as both European (European Medicines Agency, EMA) and American (US Food and Drug Administration, FDA) medical agencies have released regulatory guidelines to assess and interpret hepatotoxic signals in preclinical development.19 As many hepatotoxicities only manifest rarely and only in specific patients, such side effects are often not seen until the drug has been on the market, sometimes for years, further increasing the withdrawal costs for the pharmaceutical industry. As a consequence, due to the proration of the expenses for these failed projects, the costs for each new chemical entity reaching marketing stage have been estimated at US $2.6 billion per new drug.20 DILI reactions are often complex events that require delicate interactions between parenchymal hepatocytes and nonparenchymal cell types, such as Kupffer cells or stellate cells. These ADRs can manifest via a wide variety of mechanisms, including mitochondrial dysfunction leading to apoptosis and/or necrosis, inflammation, cholestasis, steatosis as well as immune-mediated reactions.21 While conventional 2D models B



Review

Figure 1. 3D PHH systems constitute versatile tools for a multitude of applications. Overview of applications in which human hepatocytes in 3D culture systems have unique advantages over classical 2D monolayer cultures. (A) PHH cultured in various 3D platforms exhibit increased viability and functional stability compared to those of 2D monolayer cultures, emphasizing their utility for chronic toxicity applications.71 Furthermore, histological staining and the analysis of transcriptional signatures can give indications about specific toxicity mechanisms. (B) Hepatocytes in various 3D models maintain the expression of nuclear receptors, such as CAR, PPARα, and PXR and are therefore useful for determining the inductive potential of drug candidates. (C) Moreover, by virtue of their physiological phenotype, PHH 3D cultures accurately reflect the metabolic pattern of drug candidates as evidenced the metabolic routes of AZD6610, which is primarily glucuronidated in 3D HepaRG cultures, while in PHH and humans, it is primarily hydroxylated by P450 enzymes.120 Gluc indicates glucuronide. (D) The accessibility of hepatocytes in culture allows hypothesis-driven knock-downs of candidate genes in a physiological context, thus allowing the validation drug targets. (E) PHH 3D cultures can be tailored to reflect specific pathogenicities. By modulating culture conditions regarding, e.g., the fatty acid content in the media or by incubating with steatogenic compounds, hepatocytes in PHH 3D culture can be induced to accumulate free fatty acids. In addition, coculturing steatotic spheroids with Kupffer cells or challenging them with inflammatory cytokines provides a promising tool to emulate progression of steatosis to NASH. By incubating cells with drug candidates and monitoring the effect on bile acid retention and production, 3D hepatocyte cultures can be used to screen for the cholestatic potential of the compounds in question.37 Finally, coculturing hepatocytes with nonparenchymal stellate cells or infecting them with viruses can be used to generate fibrosis and hepatitis models. These tools are useful to (i) determine whether candidate compounds are prone to cause drug-induced steatosis, cholestasis, or fibrosis61,71 and (ii) to evaluate whether particular patient subgroups might be at risk for increased compound toxicity. Histological images were obtained with permission from ref 168. C


Chemical Research in Toxicology supporting cells may interfere with read-outs validated for drug metabolism and toxicity promising physiological functionalities proven most recent development promising for multiorgan toxicity testing

binding of drugs to scaffold

difficult to handle difficult to handle

high medium •• • • • •

high medium •• •• • • • •

•• (1−4 weeks) ••• (>4 weeks)

medium medium •• •• ••• • •

medium medium •• •• •• • •

••

••

••

•••

Hollow-fiber bioreactors Micro- patterned cocultures Perfused multiwell plates Microfluidic liver biochips Microfluidic multiorgan devices

••

••• (>5 weeks) ••• (>5 weeks) ••• (>5 weeks) •• (2 weeks) •• 3D spheroids

•

•

•

•

-

•

high

high

mostly scaffoldfree synthetic polymers collagen coated islands ECM-coated polymer wafer mainly scaffoldfree mainly scaffoldfree medium low ••• •• ••• •

• •• (2 weeks) •

•

comments scaffold

collagen collagen or matrigel low medium medium medium ••• • ••• ••

-

complexity D

-

highthroughput capability characterized and benchmarked

versatility

required cell numbers

Table 1

molecular phenotype

long- term stability

coculture with NPCs

bile canaliculi

perfusion/shear stress/ hemodynamics

2. HEPATOCYTE SANDWICH CULTURES Hepatocytes in vivo are polarized epithelial cells with distinct apical and basolateral domains that are segregated by tight junctions.26 The basal domains interact with extracellular matrix (ECM) components in the space of Disse, which contributes to the preservation of hepatocyte polarity and liver-specific functionality. Accordingly, hepatocyte-ECM interaction can be modeled in vitro by culturing hepatocytes between two layers of ECM, in what is referred to as a sandwichconfiguration, resulting in reduced cytoskeletal changes, decreased cellular flattening and accompanying stabilization of cell−cell contacts.32 In this configuration, hepatocytes retain characteristics and functionality, such as cellular polarization, formation of functional networks of bile canaliculi, and expression of relevant transporters, which are lost in the 2D monolayer culture. Consequently, this model is particularly suitable for studies of hepatobiliary transport and (druginduced) cholestatic liver injury.33−37 Sandwich-cultured PHH remain functionally stable for 2 weeks in culture before becoming necrotic as indicated by albumin secretion and LDH leakage, respectively.38 Proteomic analyses of PHH cultured for 2 weeks in sandwich configuration showed typical signs of hepatocyte dedifferentiation, accompanied by features of epithelial-to-mesenchymal transition.28 In agreement with this proteomic assessment, the activity of several key P450 enzymes, including CYP2C8, CYP2C19, and CYP2D6, gradually decreases during the first days in PHH sandwich culture, whereas CYP1A2 and CYP3A4 were less affected.39 Combined,

2D monolayer Sandwich culture

interaction of these hepatic cell types. Importantly, the NPCs have to remain functional and viable for extended time periods. (4) The culture platform should closely reflect in vivo morphology including the formation of bile canalicular structures. Furthermore, physiological parameters such as oxygen levels and perfusion should be emulated. (5) Systems should be validated and benchmarked using a panel of known hepatotoxins as training compounds. (6) Platforms should be compatible with high-throughput applications to facilitate the screening of drug candidates. Thus, assay setups should require low numbers of cells and be relatively simple. (7) The culture platforms should be amenable and allow genetic predispositions or concomitant diseases to be taken into account, which can modulate DILI risk. In the last decades, mounting evidence has indicated that culturing cells in three-dimensional structures could lead to drastic phenotypical improvements across a variety of cell types, including liver cells, by providing a permissive context that preserves cellular phenotypes on a molecular and histological level by mimicking the architecture and cell−cell contacts of the intact tissue in question.30,31 Because of these improvements, 3D systems constitute superior tools for a multitude of applications in drug development (Figure 1). Hepatic 3D models can be systematized using different taxonomy concepts: First, they can be classified by their experimental paradigm into sandwich cultures, spheroid models, hollow-fiber bioreactors, and organ-on-a-chip systems (Table 1). Second, they can be grouped based on the origin of cells employed into systems that are based on hepatic cell lines, stem cell-based approaches, and strategies using primary cells. In this review, we provide an overview of novel developments in the area of human hepatic 3D culture systems and comprehensively highlight their potentials and limitations with emphasis on their amenability to toxicity screening.

current gold standard sensitive to variability in ECM batches

Review



Review

trovafloxacin, for which toxicity was exclusively detected in HepG2 spheroids after 6 days of repeated exposure.45 Furthermore, Fey et al. convincingly showed that culturing HepG2 cells as spheroids significantly improved the correlation of in vitro EC50 values (EC50 = median effective concentration = drug concentration causing 50% cell death after defined amount of time) to in vivo LD50 doses (LD50 = median lethal dose = the administered mass of drug per mass of laboratory animal that causes the death of half of the test individuals in the study after a defined period of time).50 Yet, as illustrated above, pronounced differences still exist, and the metabolic capacity especially of P450 enzymes remains poor compared to that of PHH. 3.2. HepaRG Spheroid Systems. The HepaRG cell line has been proposed as a surrogate for PHH because of similar transcriptomic profiles,51−53 but recent proteomic analyses revealed that expression of P450s, phase II enzymes, and transporters was strongly reduced in 2D monolayer cultures of HepaRG cells compared to that in PHH.54 2D cultured HepaRG cells also exhibit lower sensitivity than PHH to a panel of 16 known hepatotoxins (40% for PHH and 13% for HepaRG).44 Recent efforts have focused on improving the phenotype and functionality of these cells by employing 3D cultivation strategies. In spheroid culture, HepaRG cells express higher levels of liver-specific genes involved in drug metabolism, bile acid transport, and energetic pathways and secreted more albumin, glucose, and urea compared to those in 2D cultures.49,55 HepaRG cells form bile canalicular-like structures, accumulate F-actin at the canalicular membrane, and express relevant and functional bile acid transporters including MRP2 and P-gp in 2D as well as 3D culture.55−57 Determination of the drug-metabolizing capacity of HepaRG spheroids has been limited to CYP2E1, CYP3A4, and UGTs, which show stable activity during several weeks of culture.55,57 Furthermore, recent studies have proposed HepaRG spheroids as a useful model to perform P450 induction assays, due to their responsiveness to typical P450 inducers, including omeprazole and β-naphthoflavone (CYP1A2), phenobarbital (CYP2B6), and rifampicin (CYP2C9/CYP3A4).57−59 Evaluation of HepaRG spheroids as a model for drug toxicity screening has to our knowledge only been published using a few selected hepatotoxins. For instance, the toxicity of APAP was detected at lower concentrations in HepaRG spheroids compared to that in the 2D culture.55 HepaRG spheroids were also more sensitive to aflatoxin B1, whose toxicity is dependent on CYP3A4 bioactivation, in agreement with higher CYP3A4 activity in spheroid culture.60 In contrast, the toxicity of other hepatotoxins, such as troglitazone and chlorpromazine, was more pronounced in HepaRG 2D cultures, indicating that the sensitivity advantages of spheroid cultures might not be universal for all compounds.55,60 Recently, spheroid cocultures of HepaRG cells and hepatic stellate cells have been published as a model to study drug-induced fibrosis.61 In this study, repeated exposure to allyl alcohol and methotrexate induced typical fibrotic features, including elevated mRNA expression levels of markers of hepatic stellate cell activation (COL1A1 and LOXL2) and increased secretion of pro-collagen I peptides. In addition, the authors reported for the first time the fibrotic potential of APAP, which was subsequently confirmed in mice in vivo.61 3.3. Upcyte Hepatocytes. Upcyte hepatocytes are PHH that have been genetically modified to possess proliferative capacity by stably transducing them with the viral oncogenes E6

these studies indicate that hepatocyte dedifferentiation in sandwich culture is decelerated but not prevented. Xu et al. performed a screen in which they treated PHH on the third day of culture for 24 h with 300 hepatotoxic and nonhepatotoxic compounds.40 The authors obtained a true positive rate of 50− 60% when treating cells with 100x human cmax concentrations. This rather low sensitivity is likely caused by a combination of factors including the delayed onset of treatment and mass transfer barriers limiting the exchange of nutrients and compounds. Additionally, batch-to-batch variation of ECM substrates can pose a significant limitation of sandwich models.41 In conclusion, due to their polarity, hepatic sandwich cultures are widely used in studies of hepatobiliary transport, whereas their utility for noncholestatic, particularly long-term hepatotoxicity assessments is generally limited.

3. SPHEROID MODELS Spheroids, which are cellular 3D aggregates generated by stirring in bioreactors or gravitational aggregation in hangingdrop cultures or on ultralow attachment (ULA) surfaces, constitute models that allow high-throughput screenings in 96well format and require minimal technical investments. They can be generated using, e.g., hepatic cell lines, primary cells, or stem cell-derived hepatocyte-like cells. Overall, cells in spheroid culture exhibit several beneficial features in comparison to their 2D cultured counterparts, which are elaborated below. 3.1. HepG2 Spheroid Systems. Human liver cell lines such as HepG2 are routinely used for the assessment of drug hepatotoxicity due to their availability, robustness, and lack of interdonor variability. The phenotype and functionality of these cells in conventional 2D culture, however, is only in poor agreement with hepatic phenotypes seen in vivo as evidenced by significant differences between HepG2 cells and PHH at the functional, transcriptomic, and proteomic levels.28,42−44 HepG2 cells cultured as spheroids stop proliferating after an extended period in culture, which coincides with the establishment of a more differentiated phenotype.45 HepG2 spheroids show defined cell−cell contacts, cell polarity, and form functional bile canaliculi-like structures, which improves liver functionality, as judged by the increased secretion of albumin, apoB, and urea as well as by elevated mRNA levels of nuclear receptors (AhR, NR1I3 encoding CAR and NR1I2 encoding PXR) and various phase I, II, and III enzymes.45−49 In addition, enzymatic activities of CYP2C9, CYP2D6, and CYP3A4 are increased in HepG2 spheroids compared to 2D cultures, whereas CYP2E1-mediated 6′hydroxychlorzoxazone formation showed an inverse profile as it was absent in HepG2 spheroids but could be detected in 2D monolayer culture.45 The rate of SULT-mediated acetaminophen (APAP) sulfation was similar in 2D and spheroid culture, whereas UGT-mediated diclofenac glucuronidation could exclusively be detected in the HepG2 spheroids, indicating an overall improvement of hepatic functionality in 3D culture.45 Evaluation of the sensitivity of HepG2 spheroids regarding drug toxicity using a set of hepatotoxic (isoniazid, valproic acid, trovafloxacin, nimesulide, troglitazone, APAP, bosentan, and diclofenac) and nonhepatotoxic compounds (aspirin, dexamethasone, fluoxetine and dextromethorphan) showed similar toxicity profiles in HepG2 spheroids as in 2D HepG2 cultures after 24 h exposure, whereas repeated exposure for 6 days considerably sensitized the HepG2 spheroids to the hepatotoxic compounds.45 This sensitization effect was most prominent for E



Review

Figure 2. Primary human hepatocyte spheroids cultured in ultralow attachment plates closely resemble the in vivo liver. (A) Heatmap visualizing whole proteome analysis of primary human liver samples (n = 5) after 24 h and 7 days in a 2D monolayer culture compared to spheroid cultures after 7 days. Note that in vivo liver samples (black) and spheroids (green) cluster closely together, while the proteomes of samples cultured in 2D (24 h = blue; 7 days = red) are distinctly different. (B) Principle component analysis separates proteomes from liver and PHH spheroids from 2D monolayer-cultured samples. (C) Interindividual differences observed in vivo are preserved in 3D culture, with each of the 3D samples clustering with the respective liver piece from the same donor. (D) Expression levels of metabolic enzymes (CYP2C8, CYP2C9, CYP2D6, CYP3A4), drug (SLCO1B1) and bile transporters (ABCB11), critical hepatic transcription factors (HNF4A), and secretory products (ALB) are significantly elevated between 14- and 1834-fold in the 3D spheroid system after 7 days compared to those in 2D cultures at the same time point. (E) Functional activities of CYP1A2, CYP2C9, CYP2D6, and CYP3A4 remained overall constant throughout 5 weeks of culture (n = 20 spheroids from 3 donors per time point). Error bars in D and E indicate SEM. (F) Immunofluorescence imaging of PHH spheroids after 8 days and 5 weeks in culture shows the persistent presence of key hepatic proteins. Note also the formation and stability of bile canaliculi as evidenced by MRP2 staining. Data are from ref 71 and S. Vorrink (unpublished data). F



Review

and E7, inhibitors of p53 and Rb, respectively.62,63 The firstgeneration Upcyte hepatocytes were proposed as a surrogate to PHH since large amounts of hepatocytes from the same donor could be generated for screening purposes. However, in 2D culture, these cells have low phase I activity,62 and their proteome is very similar to that of the proteome of HepG2 cells.54 The Upcyte technology was recently modified to develop more differentiated hepatocytes, which possess higher expression of drug-metabolizing enzymes, responsiveness to P450 inducers, and increased sensitivity to hepatotoxins compared to those of the first generation.64,65 However, assessments of phenotypic stability of these second generation Upcyte hepatocytes during long-term culture as well as a comparison between other cell systems in comprehensive drug toxicity tests is still lacking. Currently, to our knowledge the use of 3D culture systems to improve the phenotype of these cells has not yet been explored. 3.4. Primary Human Hepatocyte Spheroid Systems. Mature PHH are currently considered the gold standard cell type for in vitro testing of hepatic drug toxicity.66 Yet, although the interest in PHH spheroid cultures has grown substantially over the past several years and many novel platforms have been introduced to the market, relatively few publications describing the performance of PHH spheroids in toxicity assessments have been published, at least in part due to the perceived difficulties in handling. Tostões et al. described the generation of PHH spheroids in a bioreactor system, in which functional bile canaliculi form and hepatocytes maintain phase I and phase II gene expression as well as inducible enzyme activity as evidenced by their increased P450-dependent deethylation capacity of 7-ethoxycoumarin following induction by rifampicin and β-naphtoflavone.67 While promising regarding the phenotype of these cultures, the bioreactor setup impedes effective dosing and is not suitable for screening purposes. To overcome these limitations, spheroids can be generated in 96-well plates using hanging-drop systems, resulting in spheroids of defined sizes that show stable ATP and albumin production for multiple weeks in culture.68 Importantly, the format is compatible with repeated dose toxicity experiments, and cells in this assay were shown to be relatively susceptible to the toxicity of APAP (EC50 = 754 μM, cmax = 139 μM) and diclofenac (EC50 = 178 μM, cmax = 7 μM) under chronic (14 days), repeated dose treatment. Interestingly, APAP was metabolized to its cytotoxic metabolite N-acetyl-p-benzoquinoneimine (NAPQI), as evidenced by NAPQI-protein adducts detected by mass spectrometry, highlighting the biological and toxicological relevance of the system.69 This system has proven useful for the detection of toxicity from nanomaterials contained in consumer goods in which upon repeated exposures, cytotoxicity, cytokine secretion, lipid peroxidation, and genotoxicity were increased, demonstrating the array of end points that can be measured from a small number of cells.70 Yet, nondisclosed media compositions impair result interpretation and question their physiological relevance, hampering the wider adoption of this system to biological applications. Furthermore, whether this system provides a specific and sensitive tool for the prediction of drug hepatotoxicity remains to be demonstrated in systematic studies using larger panels of compounds. As an alternative to the hanging-drop system, culture conditions to generate spheroids from PHH in 96-well ultralow attachment (ULA) plates were recently established (Figure 2).71 By using unbiased proteomics approaches, the phenotypes

of PHH spheroids were compared with liver tissues and 2D monolayer cultures from the same donors. While massive rearrangements of the molecular signatures were observed in 2D monolayer cultures already after 24 h, including significant reduction in P450 and transporter expression, perturbation of mitochondrial functionality, oxidative phosphorylation, and TCA cycle, spheroid cultures maintained phenotypes that were very close to the in vivo liver. Importantly, PHH spheroid cultures maintained viability and hepatic functionality at physiological levels for at least 5 weeks in culture as indicated by persistent high-level albumin secretion and stable metabolic activities of CYP1A2, CYP3A4, CYP2D6, and CYP2C9, whereas CYP2C8 activity was reduced after long-term culture. These properties render the system suited to support chronic toxicity tests. Upon long-term repeated dosing (28 days), the toxicity of amiodarone, bosentan, diclofenac, tolcapone, and fialuridine was detected at clinically relevant concentrations, recapturing the delayed onset of DILI in vivo.72 For fialuridine, the difference in sensitivity between acute and chronic exposure was most pronounced, dropping from no overt toxicity at 100 μM after 48 h to an EC50 of 0.1 μM after 28 days of exposure. Importantly, fialuridine hepatotoxicity was not detected in any preclinical study, including primary hepatocyte 2D cultures and rat, mouse, dog, and cynomolgus monkey in vivo studies.73 Yet, in clinical trials, severe hepatotoxicity was observed after multiple weeks, and 7 of 15 patients developed progressive lactic acidosis and worsening jaundice, 5 of whom died.74 Thus, prolonged viability and functionality of this spheroid system enables the detection of drug-induced human hepatotoxicity, not identified in standard animal models, showcasing its utility to flag compounds with pronounced hepatotoxic potential. In addition to applications for drug toxicity studies, the versatility of the ULA assay as a platform for disease model development was demonstrated. Cholestatic disease can be emulated when PHH spheroids are treated with chlorpromazine, as evidenced by a pronounced intracellular accumulation of bile acids. Moreover, cyclosporine A treatment resulted in the accumulation of neutral lipids, an effect that was inhibited by cotreatment with α-tocopherol as in vivo,75 indicating that this system can provide a suitable platform to screen the cholestatic and steatogenic potential of drugs as well as to identify antisteatotic drug candidates. Also, the culture conditions permit long-term cocultures of hepatocytes with various nonparenchymal cells, such as Kupffer, stellate, and biliary cells, opening possibilities to capture the development of complex diseases, such as fibrosis, that require an intricate interplay of multiple cell types.71 While offering tremendous improvements over 2D culture systems, the PHH spheroids described to date still do not fully recapitulate the gene expression pattern seen in vivo. Recent efforts to optimize spheroid functionality included the coculture of PHH with different stem cell populations. When cultured together with adipose-derived stem cells, PHH spheroids formed more rapidly and became more compact than hepatocyte monocultures.76 Rebelo et al. used a two-step strategy in which PHH first aggregate as monocultures.77 Subsequently, these spheroidal aggregates are incubated with bone marrow mesenchymal stem cells, which then surround the inner core of PHH. Cocultured hepatocytes retained the expression of HNF4α and CYP2C9 and showed elevated levels of albumin and urea secretion. It will be interesting to see which molecular interactions confer the enhanced hepatic phenotypes G



Review

cell derived HLCs express very low levels of many phase I and phase II metabolic enzymes, resulting in metabolic activities one to multiple orders of magnitude lower than that in adult human hepatocytes (96 and 97 and references therein). Furthermore, they persistently express fetal genes, such as AFP, in agreement with an overall fetal-like cellular phenotype, significantly limiting their usefulness for studies of drug metabolism and liver biology.98 To alleviate this discrepancy between stem-cell derived and primary hepatocytes, several strategies have been pursued, including 3D culture methods,99,100 microfluidic systems,101 and bioreactors.102 Notably, it was shown that direct cell−cell contacts99,103 and cAMP signaling104 in 3D culture improve HLC differentiation states and hepatic phenotypes, resulting in elevated levels of albumin and alpha-1 antitrypsin secretion and a reduction in AFP levels.100 While the mechanisms by which cell−cell contacts facilitate hepatic differentiation are only poorly understood, cAMP signaling seems to promote expression of hepatic genes by upregulation of the transcriptional coactivator proliferator-activated receptor γ coactivator 1α (PPARGC1A).104,105 It has been found that cultivation of HLCs in 3D culture increased the expression of a selection of phase I and phase II enzymes and gave rise to the formation of bile canaliculi as evidenced by MRP2 staining.99,100,106 In addition, the inductive capacity of rifampicin and phenobarbital on CYP3A4100 as well as of omeprazole on CYP1A2107 was significantly increased in HLC 3D culture. Moreover, stem-cell derived HLC spheroids metabolized omeprazole and rifampicin to 5-hydroxy omeprazole, omeprazole sulfone, and 25-des acetyl rifampicin, respectively.107 APAP was metabolized to NAPQI, as evidenced by high levels of glutathione-conjugates, not found in HepG2 cells.107 Clearance of APAP occurs primarily via sulfation and glucuronidation by SULT- and UGT-family enzymes.108 While APAP-sulfate is generated in HLC aggregates, APAP is not metabolized to its glucuronide, probably due to significantly reduced levels of the responsible UGTs (UGT1A1, UGT1A6, UGT1A9, and UGT2B15).107,109 Takayama et al. performed a comparative screen of 24 hepatotoxins in 2D and 3D cultured stem cell-derived HLCs and found that the sensitivity to 20 of those compounds increased in 3D culture.110 This effect was most pronounced for clozapine, isoniazid, and flutamide, for which toxicity was exclusively detectable in 3D culture. In summary, stem cell-derived HLCs generated by current protocols only partly resemble adult hepatocytes with regard to phenotype, global gene expression signature, and metabolic activity. Approaches to yield more in vivo like hepatic cells include transdifferentiation experiments in which human fibroblasts are directly converted into HLCs by overexpressing a battery of six factors important for hepatic specification.111 The resulting HLCs expressed considerable levels of metabolic enzymes and showed high activity of CYP3A4-, CYP2B6-, and CYP1A2-dependent metabolism, whereas the hydroxylation of diclofenac and phenytoin by CYP2C9 and CYP2C19, respectively, remained low.111 Furthermore, innovative strategies that involve transplantation of hiPSC-derived HLC 3D aggregates into mice indicated a further improvement of HLC differentiation states seen exclusively upon exposure to the in vivo environment;112,113 yet, also these elegant approaches did not achieve terminal maturation of transplanted HLCs. Taken together, these experiments suggest that the gap between anticipated and achieved HLC differentiation states is due to a combination of (i) stem-cell intrinsic factors, such as

and whether the elevated expression levels of hepatic genes translate into increased sensitivity when challenged with chemical insults. PHH spheroids exhibit physiological phenotypes for prolonged culture times while requiring relatively low cell numbers rendering them a promising tool for application in drug toxicity assessments. In addition, given the appropriate culture conditions, PHH spheroids provide a highly versatile model system that can be used for a multitude of applications, such as drug target validations and induction studies. Furthermore, they can be utilized to emulate various hepatic diseases in vitro, thus enabling the possibility to investigate patient-specific factors. 3.5. Stem-Cell Derived Hepatocyte Spheroid Systems. The shortage of PHH, the fact that cells often originate from diseased livers, and their inability to be expanded in culture have fueled the interest in alternative systems to use for the predictive assessment of hepatotoxicity.78 Human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs) have been discussed as promising sources of cells for evaluating drug toxicity. Both hESCs and hiPSCs are pluripotent cells that are able to divide virtually unlimitedly in vitro.79 hESCs are highly pluripotent and versatile cells yet difficult to acquire due to ethical constraints and the limited availability of donor embryos.80 In contrast, hiPSCs can be generated from any somatic tissue including dermal fibroblasts,81 thus allowing one to retrospectively acquire cellular material from individuals of interest that, e.g., displayed an idiosyncratic drug response. Moreover, stem cell-derived hepatocyte-like cells (HLCs) have been suggested as useful model systems for viral and inherited metabolic disorders.82,83 Unlike hESCs, hiPSCs preserve an epigenetic memory of their somatic origin that predisposes them to differentiate into cell types close to the tissue from which they were generated.84,85 Importantly though, these epigenetic signatures are abrogated upon prolonged passaging of cells in culture and consequently, differentiation potentials of hiPSCs equalized after long-term culture.86 Besides their transient epigenetic memory, hiPSCs are distinguished from hESCs by unique gene expression signatures regardless of their origin or method by which they were generated due to differential binding of the reprogramming factors.87 Moreover, individual hiPSC clones can differ substantially regarding their level of pluripotency88 and, furthermore, exhibit high cell-to-cell variability at the transcriptomic level.89 Partly as a consequence of this heterogeneity, some hiPSC clones can show a propensity to spontaneously differentiate into various hemangioblastic and neural derivates.90,91 In a recent benchmarking study, 58 iPSC lines from different laboratories were functionally and genomically compared.92 Importantly, the authors detected adverse features, such as contaminations and karyotype abnormalities in 23 (40%) of these, emphasizing the difficulties in cross-study comparability and highlighting the importance of selecting a well-established and thoroughly characterized stem cell line to obtain physiologically relevant results. Hence, in recent years, much research has been focused on the development and improvement of differentiation protocols by which hESCs and hiPSCs can be induced to acquire hepatic characteristics.93−95 As a result of these efforts, stem cellderived HLCs that detectably express hepatic genes can now be generated with acceptable purity. However, current differentiation protocols still fall short of generating cells with adequately mature hepatic characteristics. In particular, stemH



Review

Figure 3. Micropatterned hepatocyte coculture systems remain functionally stable in culture. (A) Schematics and photomicrographs depicting the assembly of the MPCC platform. (B) Expression levels of a selection of hepatic genes normalized to freshly isolated hepatocytes. Note that expression levels of MPCC cultured hepatocytes (gray) are substantially higher than those of hepatocytes in a conventional 2D monolayer (black) after 1 week in culture. (C) Activities of phase I (CYP2A6, CYP2B6, and CYP3A4) and phase II enzymes remain relatively constant over 3 weeks (shades of gray), while the drug metabolizing capacities in 2D hepatocyte monocultures (red) drastically decline. Reprinted with permission from ref 121. Copyright 2007 Macmillan Publishers Ltd.

an interesting step toward the development of personalized prediction models.

residual epigenetic memory or anomalous transcription factor binding patterns, that counteract complete differentiation and (ii) a poor understanding of the underlying frameworks that steer cells from fetal-like to more differentiated phenotypes in vitro. It is therefore of importance to mechanistically understand the molecular events that control differentiation and dedifferentiation of hepatocytes, as they might provide critical cues for the improvement of stem cell differentiation protocols. 3.6. Spheroid Models Summary. In conclusion, 3D spheroid culture of liver cells from hepatic cell lines, primary sources, or stem cells results in an overall improvement of cellular phenotypes and functionality. Yet, many studies lack a systematic evaluation of the sensitivity of these systems toward chronic drug toxicity. In addition, the competency of most 3D models has only been described in relation to the corresponding 2D cultures, whereas a direct comparison to the liver in vivo is lacking. Yet, this benchmark is important in order to establish these systems as relevant in vitro models to study liver function and toxicity. An important step toward phenotypically competent liver systems was the development of 3D PHH spheroids, in which long-term pharmacological drug effects can be studied. Furthermore, these systems allow the investigation of compound toxicities in cells from patients with different genetic backgrounds and/or disease states providing

4. HOLLOW FIBER 3D BIOREACTORS While spheroid cultures are amenable systems that overall well maintain hepatic phenotypes, they fail to provide perfusion. Yet, hemodynamics and shear stress have been shown to have profound impacts on hepatic function.114,115 Thus, in an attempt to generate models that more closely recapitulate the complex morphology of the liver in vivo and, specifically, to compensate for the decline in hepatic functionality that is observed in the liver during severe liver injury, a multicompartment hollow fiber membrane bioreactor technology for 3D perfusion was developed as a large-scale system (800 mL) for extra corporal liver support.116 The hollow fiber bioreactor technique is composed of three independent interwoven capillary systems for arteriovenous medium perfusion, oxygen supply, and carbon dioxide removal. When the bioreactor is inoculated with parenchymal and nonparenchymal liver cells, these cells self-organize in culture and assemble into tissue-like structures, including neo-biliary channels and neo-sinusoidal endothelialized structures. The bioreactor was miniaturized to a volume of 2 mL corresponding to 1.2 × 108 cells, to be useful for applications in preclinical drug testing.117 P450 activities were preserved for up to 23 days in culture supporting longI



Review

content screening.124 These results suggest that the sensitivity to identify liver toxic compounds is higher in repeated dosing studies in the MPCC platform compared to short-term studies in 2D culture. Yet, it cannot be ruled out that differences in sensitivity and specificity between platforms and dosing regimens are due to differences in the composition of compound panels between studies. The MPCC hepatocyte system has been applied in a crossspecies study comparing the metabolic profiles of wellcharacterized drugs.125 The qualitative pattern of the metabolite profiles investigated in the cocultures reflected the profiles in vivo, indicating the value of the model to identify human specific metabolites before proceeding with new chemical entities to cost-intensive clinical phases. The stability of the cocultures was demonstrated by studying the intrinsic clearance of drugs with known low metabolic turnover and hepatic clearance in vivo. When considering only low clearance compounds (

Novel 3D Culture Systems for Studies of Human Liver Function and

Recommend Documents