|Home | About | Journals | Submit | Contact Us | Français|
Over the years, various liver-derived in vitro model systems have been developed to enable investigation of the potential adverse effects of chemicals and drugs. Liver tissue slices, isolated microsomes, perfused liver, immortalized cell lines, and primary hepatocytes have been used extensively. Immortalized cell lines and primary isolated liver cells are currently most widely used in vitro models for liver toxicity testing. Limited throughput, loss of viability, and decreases in liver-specific functionality and gene expression are common shortcomings of these models. Recent developments in the field of in vitro hepatotoxicity include three-dimensional tissue constructs and bioartificial livers, co-cultures of various cell types with hepatocytes, and differentiation of stem cells into hepatic lineage-like cells. In an attempt to provide a more physiological environment for cultured liver cells, some of the novel cell culture systems incorporate fluid flow, micro-circulation, and other forms of organotypic microenvironments. Co-cultures aim to preserve liver-specific morphology and functionality beyond those provided by cultures of pure parenchymal cells. Stem cells, both embryonic- and adult tissue-derived, may provide a limitless supply of hepatocytes from multiple individuals to improve reproducibility and enable testing of the individual-specific toxicity. This review describes various traditional and novel in vitro liver models and provides a perspective on the challenges and opportunities afforded by each individual test system.
The gold standard toxicological approach for evaluating chemical toxicity involves complex in vivo studies which are both time consuming and costly. Due to concerns about animal welfare, time and cost constraints, and the ever increasing number of chemicals that need testing, establishing workable in vitro culture systems has become a priority for the toxicology community. In addition, the predictive accuracy of rodent in vivo testing for human adverse health effects has become a matter of dispute in recent years, in part due to poor concordance of animal study results to disease phenotypes observed in heterogeneous human populations.1, 2 The use of in vitro model systems in toxicity testing has many advantages including the decrease in animal numbers, the reduced cost of animal maintenance and care, small quantity of a chemical needed for testing, shortening of the time needed, and increase in throughput for evaluating multiple chemicals and their metabolites.3, 4In vitro systems also allow to study chemical metabolism, evaluate the mechanisms of toxicity, measure enzyme kinetics, and examine dose-response relationships.4
The Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM) at the U.S. National Toxicology Program (NTP) works in conjunction with the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM) to establish and validate alternative methods to in vivo toxicological testing. The three main tenants for animal toxicity studies of these agencies are known as the three R's: refinement, reduction, and replacement. The ultimate goal is “the validation and regulatory acceptance of test methods that are more predictive of adverse human and ecological effects than currently available methods, supporting improved protection of human health and the environment.”5
The landmark report released by the National Research Council (NRC), “Toxicity Testing in the 21st Century: A Vision and a Strategy”,2 identified the challenges of modern toxicology and provided strategies for developing alternatives to in vivo research. The report supports the movement towards the use of in vitro systems instead of in vivo toxicological studies and, as requested by the U.S. Environmental Protection Agency (EPA), has developed a long-term objective of decreasing the use of in vivo studies for toxicity testing and proposed an initial strategy towards achieving that goal. In addition to the limitations mentioned above, the report acknowledges that in vivo study results cannot evaluate the much lower concentrations and mixtures of chemicals that humans are exposed to, lack information regarding modes and mechanisms of actions, and cannot account for human variability in responses and susceptibility.2 The report asserts that in vitro work can elucidate cellular-response networks and toxicity pathways, modes and mechanisms of action, allow for high-throughput studies, enhance dose-response relationships, evaluate many more concentrations than in vivo work, use concentrations relative to human exposure, provide information for generation of pharmaco-kinetic and -dynamic models, and lead to genome based investigations into perturbations of toxicity pathways.
There are three important factors which hinder the ability to use animal models to predict human adverse effects and the National Research Council report posits that the use of in vitro models should aid in overcoming this challenge. First, in vivo studies typically use high doses of compounds which are orders of magnitude greater than those humans are exposed to. Dose-response relationships are complex so extrapolation from these high doses to lower, human, exposure levels is difficult and results in many inaccuracies. Second, in vivo studies examine the response of a standard laboratory animal to a toxicant. This response may or may not occur in humans. Even more important is that the human population is very heterogeneous and a single strain of animal cannot accurately predict the variability in responses seen in the human population. Finally, in vivo exposures in toxicity testing are usually composed of a single compound. While this allows for close examination of the results of that single compound, humans are constantly exposed to mixtures of compounds daily and the effects of these co-exposures need to be examined. Conducting in vivo studies using mixtures of chemicals would be a monumental task due to the number of different chemicals and combinations of chemicals as well as the time it takes to complete each in vivo study.
To overcome these limitations to in vivo studies, the National Research Council developed four criteria important to designing a new toxicity-testing paradigm: 1) broaden the studies to include a greater number of chemicals and chemical mixtures as well as more end points and life-stages; 2) decrease the time it takes to obtain results for risk assessment by reducing costs and length of tests and increasing efficiency and flexibility of said tests; 3) reduce the number and suffering of animals involved in testing; and 4) more closely examine mechanistic and dosimetry information to provide risk assessment organizations with a broader range of information. The major components of the proposed strategy to establish this paradigm are chemical characterization, toxicity testing, and dose-response and extrapolation modeling. The National Research Council has recognized the fact that this new toxicity-testing paradigm will require years of research and assay validation and may never fully eliminate the need for in vivo work; however the benefits of reducing the number of animal studies are well worth the effort.
One of the current major benefits of in vitro systems is their utility for screening chemicals for prioritization purposes. The EPA ToxCast™ project was established mainly for the purpose of prioritization of chemicals for further toxicological study by using high-throughput in vitro assays to establish “toxicity signatures” of chemicals, identify toxicity pathways, and predict potential toxicity.6 To do this, ToxCast™ uses a large number of cell-free or cell-based in vitro high-throughput assays. The different assays measure perturbations in pathways that could lead to cellular toxicity and the results of ToxCast™ assays will be used to link particular genes to toxicity endpoints. Researchers hope to use data produced by the high-throughput assays of ToxCast™ to supplement data from genomics, proteomics, and metabolomic studies to further elucidate toxicity pathways.7 Unfortunately, the high-throughput screening results do not correlate well with in vivo toxicity and cannot account for biological processes related to toxicity such as exposure, biotransformation, toxicokinetics, and individual genetic diversity.8 Thus, other in vitro methods must be examined to supplement these high-throughput assays.
The liver is the major source of metabolism and drug biotransformation, thus liver cells are a logical choice for toxicological and pharmacological testing. Indeed, in vivo liver toxicity testing comprises a large portion of in vitro toxicology. High throughput in vitro liver models would be a great resource and enable the shift to a greater use of the alternative toxicity testing methods. In this review we discuss the benefits and pitfalls of traditional liver-derived in vitro systems and examine novel in vitro liver models and their usefulness for toxicity testing.
Many different in vitro liver models have been employed over the years with the hopes that the effects after chemical treatment will be predictive of in vivo responses. Liver slices, cell lines, and primary hepatocytes have consistently remained the leading models in in vitro liver toxicity testing. Advantages and disadvantages of these systems vary greatly.
Liver tissue slices can be a beneficial model as they retain liver structure, contain all the cell types found in vivo, have good in vitro/in vivo correlation of xenobiotic metabolism, and maintain zone-specific cytochrome activity (allowing for cellular and zonal toxicity) and mechanisms of toxicity.9 Phase II enzymes, albumin production, and gluconeogenesis were shown to decrease slightly but remain fairly stable for up to 20-96 hours when slices were cultured.10-13 Toxicogenomics studies found that rat liver slices correlated more closely to in vivo rat livers over 24 hours when compared to two cell lines and primary hepatocytes in both conventional and sandwich culture14 and human liver slices showed low levels of CYP down-regulation over the same time period.15
Studies using tissue slices typically range in culture length from 30 minutes to 5 days10, 16-18 and culture conditions such as oxygen tension, media and supplements, and culture system (i.e. shaken flasks, multiwall plates, stirred wells, etc.) have been modified to increase cell viability and reduce degenerative changes in the tissue over the culture period.9 Unfortunately, even with these modifications, necrosis still occurs after 48-72 hours in culture,10, 18, 19 and metabolic enzyme levels are greatly reduced after 6-72 hours.17, 20, 21 In addition, rates of drug metabolism and intrinsic clearance were found to be lower in liver slices than in isolated hepatocytes,22, 23 Km values are usually higher in slices than in isolated hepatocytes,22 and investigators surmise that a gradient in chemical exposure exists within the tissue slice resulting in not all hepatocytes participating in the metabolism of compounds.9
Most of the available liver-derived immortalized cell lines do not possess phenotypic characteristics of the liver tissue.24-26 Common immortalized liver-derived cell lines in use are Fa2N-4, HepG2, Hep3B, PLc/PRFs Huh7, HBG, and HepaRG.27, 28 The HepG2 line was generated in the 1970s and expresses many liver-specific genes;29 however, the expression profile of genes involved in phase I and phase II metabolism has been shown to vary between passages and as a result data can be difficult to interpret across laboratories and passages.28 The more recently developed human hepatoma cell line, HepaRG, retains the expression of many liver-specific functions as well as many cytochrome P450s, nuclear receptors, membrane transporters, and phase II enzymes.30 HepaRG cells have a stable karyotype, can differentiate into either hepatocyte or biliary lineages, have a high proliferative capacity, and have shown to produce data that is both reproducible and consistent among experiments.28, 31, 32 Still, the expression of liver-specific functions in HepaRG cells is still much lower on average than that of primary hepatocytes31 and they represent a phenotype from a single donor, thereby reducing their predictive value for the human population.
Most hepatocyte isolation protocols use collagenase digestion to disrupt the bonds between cells and allow for single-celled suspensions. Suspensions of hepatocytes are an easy to use method for moderately high-throughput toxicity studies. Griffin and Houston 33 demonstrated that suspensions of hepatocytes provided a more accurate estimate of internal clearance rate when compared to conventional monolayer cultures. Suspensions also retain high levels of functionality allowing for a more accurate correlation to in vivo toxicity than cultured cells.34-37 However, it is well known that most isolation protocols result in damage to cell surface receptors and antigens, cell junctions, cell membranes, and cytosolic contents.9, 38 Collagenase digestion also induces oxidative stress in the hepatocytes leading to a loss of cytochrome enzyme activity with greatest loss observed 4-8 hours after isolation.39-41 The loss of contact with extra cellular matrix and other cell types results in loss of cell polarity, integrity, and differentiation42, 43 which can only be re-established after restoring normal intercellular contacts. Finally, as Hewitt et al. 35 and Burke et al. 44 describe, often hepatotoxicity is a process that occurs over several hours (such as with acetaminophen toxicity) and thus, hepatocytes in suspension cannot maintain viability for the time necessary for the development of toxicity.
Cultures of primary hepatocytes have been the gold standard for in vitro testing as they can maintain functional activities for 24-72 hours, can be used for enzyme induction and inhibition studies, allow for medium-throughput screening of compounds, and are ideal for examining interspecies and inter-individual differences in metabolism.34, 45 Primary hepatocytes are often employed with the expectation that chemicals will affect or be affected by an isolated cell in the same manner that would occur in the whole organ and will in this way be a predictive model.4 Unfortunately, primary hepatocyte cultures are limited as in vitro systems as well.
Traditional primary hepatocyte cell culture involves the method of plating cells on a rigid substratum in which they proceed to form a monolayer across the bottom of the culture plate well. Under these conditions, primary hepatocytes undergo changes in cell morphology, structure, polarity, gene expression, and liver-specific functions,3, 46, 47 a process referred to as de-differentiation. Hepatocyte de-differentiation is an important limitation to this in vitro system. As the isolated cells have lost their normal microenvironment structure, cell to cell interactions, and cell membrane structures, response to chemical exposures can be different than those occurring in vivo. Finally, one of the most problematic disadvantages of the use of primary hepatocytes for toxicological research is the drastic decrease in cell functionality over time. Liver specific functions such as albumin production and cytochrome P450 expression decline quickly over the first 24-48 hours of culture as the cells begin to lose their differentiated status.3, 34, 46, 48 Extending culture longevity, both in terms of liver specific functions as well as basic cellular functionality, has been a great obstacle the field has not yet been able to overcome.
In an effort to maintain liver-specific functionality over longer culture periods, a simple sandwich configuration was developed. In this system, hepatocytes are placed between two layers of matrix (traditionally collagen or Matrigel®.47 Maintaining hepatocytes in a sandwich culture configuration increases and maintains albumin secretion, restores polygonal morphology similar to that of in vivo cells, prevents loss of cell viability, increases basal and induced enzyme activities, and mimics in vivo biliary excretion rates.47-51 Some of these functions are retained for relatively long periods of 24 days to 7 weeks in rat hepatocytes. Investigators attribute these positive effects to cell polarity which was induced by the ECM.47 Research has shown that sandwich culture prevents the formation of stress fibers in addition to increasing transferrin, fibrinogen, and bile salt secretion, and stabilizing urea secretion in rat hepatocytes over 42 days.52
Despite the positive effects attributed to the sandwich culture technique, expression of genes responsible for many liver-specific functions decreases over time (90 hours). Even though phase I enzyme gene expression decreases over time and phase II enzymes remain expressed at relatively high levels, data indicate that P450 gene expression in sandwich cultured mouse hepatocytes is still more stable than in monolayers of primary hepatocytes.53 It has been suggested that the sandwich culture model is most useful for the mechanistic studies of hepatobiliary toxicity.54
Approaches to surmount the challenges of the in vitro liver test systems have been proposed, including adjusting components of the culture medium, altering the extracellular matrix, changing the cell culture format (such as monolayer, spheroid, or 3D cultures), adding flow to the culture system, and culturing hepatocytes with other cell types, all of which are summarized in multiple reviews.3, 34, 54-57 Although a number of investigators have found certain approaches to benefit hepatocyte culture in their own laboratories, difficulties in comparison of experimental results occur and continue to hinder the identification and adoption of optimal culture conditions. In addition, choice of species, strain, and sex of animal vary greatly across experiments. Conditions of cell isolation are also variable as are the methods for cell purification, conditions of cell culture, and endpoint assay methods. Yields and viabilities of final cell preparations can differ depending on types of collagenase and perfusion methods used. Species differences in optimal culture conditions abound as well. These factors must be taken into consideration when employing primary hepatocyte cultures for toxicity testing and developing the proper procedures in one's own laboratory.
Traditional 2D hepatocyte cultures are not particularly well suited for high-throughput screening applications. Still, the availability of the robotics-equipped facilities and a relatively low per sample cost of data generation in these systems has prompted exploration of novel culture formats. For example, a well-established role for inflammation in toxicity – both developmental and postnatal – has been demonstrated in many organ systems and disease states, from endometriosis58 to idiosyncratic drug toxicity.59
An in vitro drug/inflammatory cytokine/inflammatory mediator co-treatment approach was used to reproduce clinical drug hepatotoxicity signatures, particularly for idiosyncratic drugs, in cultured primary human and rat hepatocytes, as well as HepG2 cells.60 Using this approach, drug-cytokine hepatotoxicity synergies were shown to be detectable for multiple idiosyncratic hepatotoxicants, but not for non-toxic. The authors ascertained that TNF, IL-1 alpha, and LPS were especially informative within the context of multi-cytokine mixes. Further studies using this method (i.e., combinations of hepatotoxic drugs and cytokines) combined with multiplex phosphoprotein signaling and cytotoxicity measurements in cultured hepatocytes from multiple human donors produced multi-dimensional data amenable to modeling the signal-response data.61 The authors showed that drug/cytokine liver toxicity is integratively controlled by four key signaling pathways: Akt, p70 S6 kinase, MEK-ERK, and p38-HSP27. Furthermore, this modeling predicted, and experimental studies confirmed, that the MEK-ERK and p38-HSP27 pathways contribute pro-death signaling influences in drug-cytokine synergy. These studies may be especially useful for illuminating consensus mechanisms of toxicity in human cells as they are amenable to high throughput, multiplexed intracellular signaling measurements, and reveal signaling networks that are difficult to ascertain from either in vivo or traditional in vitro culture systems.
Primary hepatocytes and hepatocyte-like cells have limited utility under standard culture conditions for certain types of toxicity testing. A continuing challenge in the culture of both primary hepatocytes and hepatocyte-like cells is the long-term maintenance of their functionality.3, 34, 46, 48, 62, 63 In addition, lack of proper absorption, distribution, metabolism, and excretion (ADME) properties as a result of cellular disconnect from the circulatory and other organ systems also make assessment of chemical exposure results difficult with primary hepatocyte cultures.4, 34 At this time, novel culture strategies are being devised to counteract the de-differentiation process, establish a more heterotypical microenvironment, and create more predictive in vitro liver toxicity models. In this section, four of those strategies will be reviewed in more detail: three-dimensional culture systems, bioartificial livers, co-cultures, and stem cell models.
Three-dimensional culture of hepatocytes is a rapidly expanding field of study as engineers, biomaterial scientists and biologists attempt to recreate the complex cellular microenvironment of the liver in hopes of extending primary hepatocyte culture longevity and functionality. Three-dimensional cultures range in complexity from monolayer sandwich culture and spheroids to more advanced systems involving porous materials, packed-bed reactors, hollow fibers, and perfusion flow.
The microenvironment of the hepatocyte in vivo is very important to the maintenance of normal function, including its response to endogenous and exogenous substrates, and can be complex to mimic. Limitations of traditional two-dimensional culture, namely reduction of gene activity and expression of genes involved in drug metabolism, mainly occur within the first 24-48 hours and these cultures have been shown to have a low sensitivity of drug hepatotoxicity detection.54 Reestablishment of cellular polarity has been suggested by many investigators as being crucial to maintaining gene activity and expression as well as proper function of hepatocytes while in culture.52 Hepatocytes possess not only a single apical and basal surface, but have multiple apical surfaces (bile canalicular surfaces) and two basolateral surfaces.52 For this reason, re-establishing cell polarity in vitro is more challenging than with other types of epithelial cells. Different extracellular matrices (ECMs) have been extensively evaluated in 2D and 3D hepatocyte culture in an attempt to re-establish and subsequently maintain hepatocyte cell polarity. Many configurations of matrices have been evaluated to determine optimal conditions for restoration of hepatocyte polarity. The use of 3D liver cultures can overcome deficits of the 2D culture system by providing models that reestablish cellular polarity and create more complex local environments.
Three-dimensional culture systems have evolved greatly within the last decade in an effort to extend culture period and retain hepatocyte function. Initially, hepatocyte spheroids were constructed, with the assumption that cellular aggregates better mimic liver tissue characteristics. Working with self-forming hepatocyte spheroids was initially a complex undertaking involving a prolonged culture period for spheroid formation prior to the hepatocytes being available for any toxicological or pharmacological experimentation. In 1996, Wu and colleagues64 optimized a protocol for quick production of rat hepatocyte spheroids through use of spinner flasks. They observed rat hepatocyte spheroid physiology and found that oxygen supply was critical in the proper formation of spheroids and that these cells maintained a more differentiated state as compared to the monolayer culture. Cyp1a2 and Cyp1a1 expression dropped rapidly in a traditional 2D monolayer culture of mouse hepatocytes, while in 3D spheroids the levels remained high over 5 days,65, 66 which indicates that spheroid culture is preferred over monolayers for studies evaluating levels of gene expression. Spheroid culture of mouse hepatocytes can also maintain expression of Cyp2b9 and Cyp2b10 for several days at levels equal to in vivo.67
Du et al. 68 were able to develop a unique spheroid-based monolayer culture configuration by conjugating cell adhesion peptides Arg-Gly-Asp (RGD) with galactose ligand and polyethylene terephthalate. They determined that the rat hepatocyte 3D spheroid configuration was ideal; by using this new culture system cells would anchor well to the substrata and have a limited ability to spread while still keeping the functional characteristics of spheroids. Liver-specific functions of cells on the new substratum, such as albumin secretion and urea synthesis, were similar to that of traditional spheroids. These cells also exhibited greater sensitivity to acetaminophen-induced hepatotoxicity than those cultured on traditional 2D collagen. However, due to oxygen and nutrient diffusion difficulties, spheroid culture is limited in its ability to be used for bioartificial liver models or long-term culture; thus, additional strategies for 3D models are being explored.
Gelled materials have been devised to combine the complexity of a Matrigel®-type extracellular matrix with spheroid formation. The idea of culturing hepatocytes encapsulated within a gel was originally developed for use in a bioartificial liver where semi-permeable hollow fiber cartridges were used as a scaffold for hepatocyte attachment.54 Miccheli et al.69 entrapped rat hepatocytes in alginate beads and placed them in a bioreactor system with a continuous flow of medium. In this system, energy metabolism, viability, and redox state was stable within 3 hours and the proper shape, microvilli, tight junctions, and bile canaliculi were reformed within 8 hours. The authors confirmed previous observations that adequate supply of oxygen and substrates is critical for proper cellular reorganization in these systems. They observed increased glycerophosphorylethanolamine levels in the static cultures indicating problems with membrane phospholipid metabolism regulation contrasting low levels in the bioreactor system. Gels of agarose have been employed to sustain viability of mouse hepatocyte cultures for 21 days. Within these gels, hepatocytes typically form aggregates and excrete significantly higher amounts of albumin than their 2D counterparts.70 Shen et al. 71 demonstrated that gel-entrapped hepatocytes expressed higher levels of liver-specific functions and phase I metabolism, maintained higher intracellular ATP levels and mitochondrial membrane potential, and accumulated less lipids within the cytoplasm than those cultured in traditional 2D monolayer. In addition, measurements of phase I metabolic enzymes were similar to levels found in vivo. Furthermore, gel entrapped cultures were able to accurately reflect hepatotoxicity in over twenty reference compounds.72 A general limitation of the gel culture systems is impairment of oxygen and nutrient transfer to cells. In addition, these systems are labor-intensive, low throughput and require large amounts of reagents.
An extension of this line of research was the development of models that use scaffolding to provide structural support. A semi-permeable hollow fiber-based system serves as an example. Some of the fiber systems incorporated gel entrapment while others did not. Rat hepatocyte cultures with increased functionality have been obtained through use of scaffolding composed of polyvinyl formal resin and nanofibers.73, 74 The scaffolding increased cell adhesion, even at higher seeding densities, exhibited lower LDH release, and maintained biochemical functions such as albumin secretion, urea synthesis, and glycogen synthesis. Stabilization of albumin and urea levels after an initial decline was noted in a functional 20-day rat culture.73 Shen and colleagues75 examined a simple micro-hollow fiber reactor system for rat hepatocytes and determined that functionality (in terms of urea synthesis and albumin secretion) was increased in this system when compared to simple monolayer culture. Such a system also improves the metabolic capability of cells, as indicated by greater sensitivity to chemical insult than seen in cells in monolayer culture. Still, these models are difficult to produce in large quantities and also lack flow to facilitate oxygen and nutrient delivery to the cultured cells.
Incorporating fluid flow into three-dimensional culture systems was an important step in tissue engineering. Packed-bead, hollow fiber, or other type of “reactors” integrate micro-circulation of medium to recreate the fluid flow within the liver. Addition of flow to 3D cultures is important for combating the issues of poor oxygen and nutrient diffusion through spheroids and aggregates of cells and extracellular matrices. Systems incorporating shear flow have shown to promote round cell aggregates that are more similar to in vivo morphology and increase and maintain liver-specific functions.76 Incorporating flow into the culture system has been shown to increase and maintain a higher level of urea synthesis77 and to increase albumin, lactate, and glucose secretion.78 Physiologic oxygen gradients can be established in these microfluidic systems which can assist in elucidating the physiological mechanisms involved in hepatotoxicity.79 Furthermore, the organotypic physiological geometries and flow properties in these systems allow for the study of drug-drug and cell-cell interactions.80 Although these methods of three-dimensional hepatocyte culture show improvement over traditional two-dimensional cultures they are still insufficient for long-term culture or use in clinical therapy (such as bioartificial livers) due to the continued loss of metabolic gene expression after cell isolation.
There continues to be a need for more complex 3D systems. There are a number of efforts to engineer more advanced 3-D culture systems by applying the existing knowledge regarding important cellular environmental dynamics between extracellular matrix, micro-circulation, and cell type and density. The overall goal of many of these efforts is to form a fully functional liver culture model that can be used for toxicological and pharmacological research or that can be modified into a bio-artificial liver for clinical use. A few systems will be briefly discussed here.
Chao et al. 80 published an evaluation of a novel microfluidic device, the HμREL® biochip (an earlier model of this system was previously described by Sin et al. 81). Four biochips are enclosed in a polycarbonate housing connected, by tubing lines, to a fluid reservoir and peristaltic pump. Each biochip has one or more separate compartments in which different cells can be housed. The compartments are microfluidically connected, in a linear path, to allow for interaction between the cell types. A new design allows for separate microfluidic experiments to be run in parallel on a larger set of housing plates. Evaluation of the system using primary human hepatocytes indicated the system preserved cell viability and metabolic competency at least as high as, and sometimes higher than, the traditional static culture conditions.80, 82 The utility of this model for toxicity testing was explored through prediction of in vivo clearance rates. It was demonstrated that data from this model system was more correlative with in vivo data than that derived from the static hepatocyte cultures and this correlation further improved when co-cultures were used.82 In addition, in vivo-like absorption, distribution, metabolism, bioaccumulation, and toxicity of naphthalene was demonstrated when lung, adipose, and liver cells were fluidically connected.83 The microscale design allows for microscopic imaging, oxygen sensing, physiologically relevant ratios of chamber sizes and liquid residence times in each compartment, physiological hydrodynamic shear stress, physiological liquid to cell ratios, and requires less media and cells. The model is severely limited, however, by the fact that sample removal is difficult without disturbing the system dynamics. Furthermore, the recirculation of the medium involves a complex set of tubing lines and reservoirs, and cells on chips form monolayers and not physiological tissue constructs.
A hollow-fiber based bioreactor was described by Schmelzer et al.84 for the regeneration and culture of human hepatocytes. Cells are seeded into the extracapillary space and are surrounded by three independent capillary membrane systems. The capillary systems are composed of porous polyethersulphone and hydrophobic multilaminate hollow fiber membranes which allow for gas exchange. The capillary layers are interwoven around the extracapillary space and two of the capillary systems are perfused in a counter-current flow with either culture medium or plasma while the third allows for decentralized oxygenation and supply of nutrients. The bioreactor functions through use of a perfusion device that uses pressure-regulated pumps to control the medium flow. A gas mixing unit is also used to provide the system with, and control the rates of, air, O2, and CO2. Utilizing this system, human hepatocyte functionality was stable for up to four weeks in this bioreactor system. Although this bioreactor system allowed for a microenvironment more similar to in vivo conditions, the system was large and required high numbers of cells and large amounts of reagents. A microscale prototype version of this bioreactor was then developed.85 In this smaller model, the bioreactor is comprised of four cell chambers each of which contains four compartments. The first compartment houses the cells, two contain culture medium, and the last provides the oxygen supply. All compartments are connected to provide the cells with a physiologically-based environment. This micro-bioreactor has thus far been used for human fetal liver cell culture and has sustained cell viability and differentiation for ten days. The prototype allows for small numbers of cells and limited reagent use, microscopic evaluation of the cells, and monitoring of oxygen concentrations. In addition, the counter-directional perfusion method is unique in providing a more physiologically similar flow. Subsequently, a similar system was used for studies of pharmacokinetics and drug toxicity. Albumin synthesis and CYP activity was maintained for 2-3 weeks in co-cultures of parenchymal and non-parenchymal liver cells.86 Limitations of this system include a lack of physiologic gradients normally seen in liver tissue, the complication of numerous tubing lines, and limited throughput as only a small number of different culture conditions can be evaluated on a single system.
Sivaraman et al. 45 used a 3D bioreactor system to evaluate the functionality of rat hepatocyte spheroids. They had hypothesized that a system that included heterotypic cell interactions, fluid flow stresses, and microarchitecture similar to those in vivo would provide the appropriate microenvironment for cells to function as they do in vivo. This system housed spheroids of rat hepatocytes attached within the many channels of a silicon scaffold. Perfusion of the medium was directed both over the top of the channels as well as through the tissue in the channels. A high-resistance filter controlled the perfusion through the channels so flow was uniform through all channels of the scaffold. Flow was initially in a downward direction to trap the cells in the channels and later, after the cells had adhered to the channel walls, the flow was reversed. Spheroids seeded in the bioreactor system had an increased and sustained functionality when compared to spheroids in both static culture and single cells seeded into the bioreactor system. Both urea synthesis and albumin excretion were increased and more stable with the spheroid bioreactors. This bioreactor system can be scaled up easily to incorporate more channels (and thus, more cells) and up to 12 different reactors controlled by the same unit while still maintaining cell viability and albumin production for 7 days.87 The throughput and ease of handling of this model system is greater as compared to other 3D systems. Toxicity testing-relevant uses of this system include studies showing that clearance rates of compounds with known liver metabolism are comparable to those obtained in vivo, and promising results with known inflammatory-mediated idiosyncratic toxicants.88 The scaffolds in these perfused bioreactors promote tissue morphogenesis and allow for microscopic examination via light or two-photon microscopes; however this cell culture system still lacks high-throughput capabilities, does not accommodate large numbers of cells, and needs to be examined with hepatocytes from other species.
Although much progress has been made over the traditional two-dimensional static culture system, there still is a need for a model in which all levels of liver-specific function can be maintained for long periods of time. Many of the available models have yet to establish whether the systems can incorporate and sustain hepatocytes from different species (e.g., rodents and humans). Furthermore, studies of applications of these methods in toxicity testing have thus far been limited and further research is needed to establish whether they may provide increased benefit over traditional culture systems.
Bioartificial livers (BALs) are a promising avenue for ex vivo therapies for patients with liver disease or failure. If properly designed and manufactured, the devices could be used for toxicological studies as well. The most common designs for BALs are hollow-fiber systems and flat-plate systems, using aggregates or single layers of hepatocytes respectively.89 In these systems plasma can either be in direct contact to, or separated from, the hepatocytes.
Three primary criteria must be fulfilled in the design of a fully functional bioartificial liver: the device must be able to (1) use and maintain a large number of differentiated hepatocytes, (2) decrease transport limitations, and (3) prevent the need for large amounts of plasma/medium.90
A number of bioartificial liver models are under development and many use selective membranes to prevent direct contact between the blood and hepatocytes.89 The ability to evaluate hepatotoxicity in the presence of human blood allows for a more predictive model by providing non-hepatic soluble factor influences. Zhang et al. 91 developed a multilayer flat-plate BAL that caused a reduced immune response in the recipient. These researchers co-cultured fresh porcine hepatocytes with bone marrow mesenchymal stem cells and used the BAL device to treat beagles in acute liver failure. Low levels of antibodies and complement indicate that this BAL has a high level of immunosafety. Davidson and colleagues92 determined that zonal distribution of oxygen is one of the most important factors in BAL design. Their results showed that flow rate of plasma/medium was key to creating metabolic zones in such devices. A liver organoid system was developed by Lu and associates93 which utilized hollow fibers to assemble collagen-hepatocyte mixtures into three-dimensional constructs. In this system, glucose-related metabolic activity and hormonal responses were maintained at high levels over the culture period. Another device used microcarriers and alginate to allow for a large number of cells, minimal migration, and sufficient oxygen diffusion.94
The spatial heterogeneity that a bioartificial liver provides would be a great advantage for toxicological studies. For example, urea cycle enzymes are expressed in high levels in the periportal and centrilobular regions, however not in the perivenous hepatocytes.90 Hepatotoxins such as acetaminophen can, over a period of repeat exposure, modulate expression of cytochrome P450 enzymes across the acinus, leading to new expression patterns.89 Bioartificial liver devices can create a more similar environment to that of in vivo liver than 2D and 3D culture and can allow for maintenance of functional heterogeneity.
Although advances in BALs are promising, Bikhchandani and colleagues95 indicate that the clinical relevance of BALs is still questionable and no statistically significant benefit in regards to patient survival has been shown by any single device. However, the lack of survival benefits for patients is not a disadvantage to their use in toxicity studies. Yet there are other attributes of BALs that are detrimental to toxicity studies, including their need for large numbers of highly functional cells (not widely available), lack of proper microenvironment for hepatocytes (including structure and interactions with other important non-parenchymal cell types), short life-span of hepatocytes in culture, and lack of proper complex membranes (to allow for oxygen, nutrient, and small molecule exchange).96
It has been suggested that any bio-artificial device composed of primary cells will require non-parenchymal cells in order to maintain tissue-specific functions.97 Bioartificial liver devices are being developed to not only assist in the treatment of liver failure but also for use in drug metabolism and toxicological studies.98 Whether or not a bioartificial liver or other type of three-dimensional model would best predict in vivo responses to chemical exposure remains to be seen; however the incorporation of key micro-environmental requirements, such as the presence of non-parenchymal cells, is critical to the development and application of these systems.
In the body, cells function by way of complex interactions and signals from other cells. For this reason, it is safe to assume that cultures of hepatocytes alone may not properly represent in vivo functionality especially when examining drug related hepatotoxicity. Hepatocyte co-cultures are usually comprised of hepatocytes with one other cell type, often other liver cells,99-102 non-liver epi- or endo-thelial cells,103, 104 fibroblasts or cell lines.105-107 Co-cultures have also been prepared using hepatocytes and more than one additional cell type, such as hepatocyte-non-parenchymal cell (NPC) cultures. In these cultures the hepatocytes are cultured with other liver cells: Kupffer cells (resident macrophages), sinusoidal endothelial cells, and stellate cells (stromal cells). Co-culturing hepatocytes with other cell types have been shown to be one of the most successful techniques for maintaining hepatocyte function under in vitro conditions.28, 108-110
Griffith and colleagues111 demonstrated that hepatocytes seeded with bovine aortal endothelial cells could form three dimensional structures in a polymer scaffolding system with flow. Within these structures hepatocytes bound to the substrata and the endothelial cells flattened and “covered” the hepatocytes. These studies were later expanded to demonstrate that hepatocyte co-cultures with liver-derived endothelial cells in a microfabricated perfusion reactor resulted in formation of endothelial network structures and greater retention of hepatocellular function that those without co-cultured cells.112,87, 88 Of note, co-cultures including liver-derived immune cells, along with other liver non-parenchymal cells, were responsive to inflammatory cues, rendering cultures susceptible to idiosyncratic toxicity of certain drugs that display toxicity when their metabolism synergizes with inflammation.88 Other investigators have found increases in albumin secretion and urea synthesis in rat hepatocytes when cultured with mouse embryonic fibroblasts, 3T3-J2, or NIG-3T3 cells. Other benefits included retention of hepatic polygonal morphology, visible bile canaliculi, and distinct nuclei.113 Mouse hepatocytes demonstrated better maintenance of albumin secretion and urea synthesis over a 7 day culture period when allowed to form 3D aggregates with mouse liver fibroblasts. The importance of the physical interactions between the two cell types was evident. Increased liver-specific functionality developed when the cells were in contact in the same culture rather than on different substrata.114
Results from a study culturing micropatterned hepatocytes (produced by seeding on a stencil and then removing the stencil) on a 3T3-J2 fibroblast feeder layer indicate that this type of co-culture can lead to increased and maintained albumin and urea production as well as retained hepatocyte morphology for up to 42 days.115 This collagen micropatterned substrate system is now used by Hepregen Corporation in their HepatoPac™ Bioengineered Microliver Platform. Hepatocytes selectively adhere to ECM-coated domains while stromal cells attach to remaining areas surrounding the parenchymal cells.115 When compared to human liver microsomes, S-9 fractions, and primary hepatocyte suspensions, the Hepregen system allowed for longer incubations with 27 known liver-metabolized compounds and was able to generate a greater proportion of the major human metabolites normally found in vivo.116 A similar system is produced by Transparent, Inc. in which mouse fibroblasts are used as feeder cells. In this system, the feeder cells migrate from underneath and surround the 3D hepatocyte structure and form a Disse-like space between the two cell types. Results from this system indicate that hepatocyte CYP basal expression and induction and transporter activity can be maintained, increased, or measured in levels higher than traditional monolayer culture for 7, 14, or even 54 days.117 The importance of the heterotypic interface between the hepatocytes and fibroblasts further exemplifies the importance of cell-cell contacts in hepatocyte culture.118
Not only can cell-cell contacts with fibroblasts, cell lines, and epi- and endothelial cells affect hepatocytes in culture, but the soluble factors excreted by other cell types have been shown to be extremely important as well. Cultures of rat hepatocytes in medium conditioned by cardiac endothelial cells functioned similarly to hepatocytes cultured under a layer of endothelial cells and gel.119 Hepatocytes were shown to modulate endothelial activation states as well.120 Morin et al. 121 observed similar increases in albumin secretion in culture systems in which the endothelial cells and hepatocytes were not in contact as cultures in which the cells were in direct contact. These non-contact co-cultures have had varied success and limited reproducibility between laboratories, however, thus cell-cell contacts continue to be one of the most important features of co-cultures.57, 122
The addition of liver non-parenchymal cells to cultures of hepatocytes is required to reconstruct an organotypic environment that can approximate in vivo conditions. Many publications report on the benefit of non-parenchymal cells on hepatocyte function. Ries et al. 99 described a hepatocyte-NPC co-culture model in which hepatocytes were seeded on top of the NPC layer (containing Kupffer, stellate, and sinusoidal endothelial cells). In their cultures, sinusoidal endothelial cells survived and maintained their fenestrae for 6 days. Other cell types remained viable for up to 14 days. They were able to demonstrate that not only did the presence of the NPCs maintain hepatocyte differentiation, but the hepatocytes contributed to survival of the NP cells. Human hepatocytes co-cultured with non-parenchymal cells, in a system with flow, were shown to have greater drug metabolism capabilities and better maintenance of metabolic capacity than cultures composed solely of hepatocytes.82 Other cultures of hepatocytes with non-parenchymal cells found that stimulation of hepatocyte DNA replication after treatment with known hepatotoxicant WY-14,643 was enhanced compared to cultures of hepatocytes alone, indicating that the co-cultures result in more in vivo-like responses to treatment.123 Tukov et al. 100 examined the effects of Kupffer cells cultured in contact with rat hepatocytes. They found that by adding Kupffer cells to the cultures they could mimic in vivo drug-induced inflammatory responses thus providing a model which could be useful in predicting such interactions prior to clinical trials. Co-cultures of rat primary hepatocytes and stellate cells on a poly (DL-lactic acid) surface quickly formed spheroidal aggregates and maintained liver-specific functions for over 2 months. Cytochrome P450 enzyme induction was maintained for almost the entire culture period.124 Stellate cells have also been shown to promote cell proliferation when in contact with hepatocytes and maintain hepatocyte differentiation when cultured without direct contact.102
Finally, RegeneMed Inc. has demonstrated the beneficial effects of seeding primary hepatocytes on nylon screen mesh constructs containing established cultures of non-parenchymal cells. These co-cultures formed 3D-like tissue constructs and maintained albumin synthesis and cytochrome induction for up to 48 days in culture.125 More recently, the company has established that cells within the 3D co-culture system can respond to chemical exposures (such as Wy-14643, Phenobarbital, TCDD, and IL-6) and induce gene expression in multiple pathways in a manner similar to that found in in vivo tissue.126
Current procedures used to isolate primary hepatocytes tend to be labor intensive and results vary greatly between laboratories and protocols. Use of cryopreserved hepatocytes is becoming more popular; however cell quality upon thawing is not always consistent. A more robust and reproducible source of hepatocytes would greatly benefit the field of toxicity testing and assist in standardizing research. Theoretically, stem cells would represent a renewable source of cells and would potentially provide large numbers of functionally equivalent cells that could be stored for later use. A toxicity model which incorporates a relatively unlimited supply of human hepatocytes with defined phenotypes would allow for better predictivity in responses to drug treatment and could account for the genetic diversity within the human population as well. Two sources of stem cells will be discussed here: embryonic and induced pluripotent cells.
Many laboratories have, with some success, developed protocols to isolate embryonic stem cells (ESCs) and induce them to form hepatocyte-like cells. Hamazaki et al. 127 evaluated the capability of murine embryonic stem cells to differentiate into mature adult hepatocytes in vitro. The researchers found that although ESCs could not differentiate past the initial stages of mid-endodermal differentiation (expression of TTR, AFP, AAT, ALB) with simply basal media alone; the addition of growth factors allowed the cells to differentiate into further stages of hepatic lineage (expression of mature hepatic genes). Other investigators have produced metabolically active hepatocyte-like cells from human ESCs using supplemented media128-130 and recently researchers used HepG2 pre-conditioned medium to induce similar formation.131 The researchers treated the hepatocyte-like cells with ethanol and observed similar effects (down-regulation of genes and proteins as well as cell mortality) that are typical of this treatment in vivo.
Unfortunately, all differentiation protocols result in highly variable functionality within the cell population and cells begin to lose hepatic characteristics after a few days much like standard culture conditions.62 Zamule et al. 132 employed WEM with normal hepatocyte supplements (HEPES, glutamine, dexamethasone, etc.) and type I collagen extracellular matrix to culture human ESCs for 10-20 days. Over this period of time, and under these conditions, the cells became committed to hepatic lineage differentiation. Although the differentiated cells expressed many liver specific functions at high levels, the levels were still not comparable to primary human hepatocytes indicating that the cells were not fully mature. Other researchers have found many other substances that induce differentiation of human ESCs along the hepatic lineage: FGF,128, 133-136 BMP,129, 134, 135 HGF,128, 136 dexamethasone,135, 137, 138 insulin,134, 138, 139 and transferrin.134 Mouse and rat models have shown to support the proliferation and function of transplanted adult hepatocytes in cases where the liver has been severely injured.140-142 Unfortunately, hepatocytes transplanted into normal functioning liver tissue do not respond in the same proliferative manner.143
Researchers are developing methods to overcome these disadvantages of functional variability, low expression levels, and loss of functionality over time, however embryonic stem cells are a controversial source of cells for scientific research and this is not likely to change in the near future. A different source of stem cells could eliminate this ethical controversy.
Induced pluripotent stem cells (iPSCs) transform into hepatocyte-like cells by following the same differentiation processes as stem cells; however they originate from somatic cell types which must first be “reprogrammed”. IPSCs retain great capacity for cell division and differentiation.144 Since these cells do not come from an embryo, the use of these cells is not as ethically controversial.
In 2010, a method that produced hepatocyte-like cells from murine fibroblasts was published.63 This study was important as it produced cells that had mRNA expression similar to that of hepatocyte-like cells derived from embryonic stem cells indicating iPSCs could be as good a model as ESCs. Liu et al. 145 noted that generating human iPSCs from hepatocytes was a faster process than generation from other somatic cell types, however functionality was not determined beyond minimal evaluation (albumin, CYP3a4, and CYP1a2). Si-Tayeb and colleagues63 postulated that if enzyme expression was to remain at high levels proper culture methods would have to be established (similar to 2D culture). These scientists also used a method that did not use any undefined factors such as serum or primary feeder cells which increased reproducibility and allowed for a large proportion of cells to exhibit primary hepatocyte characteristics.
Regrettably, expression levels of xenobiotic metabolism genes in iPSCs are still not equal to those found in whole liver or freshly isolated primary hepatocytes, enzyme levels decrease quickly over the culture period (again, similar to 2D hepatocyte culture), and other minor obstacles continue to impede the progress towards using these cells for in vitro toxicology (variability in lines, incomplete programming within cell populations, uncharacteristic response to prototype hepatotoxicants, etc.).146 Unexplained differences in protein-coding and miRNA gene expression between ES and iPS cells were reported.147 It was also found that the origin of the iPSCs could influence their differentiation, tumorigenic properties, and gene expression and epigenetic features.148, 149 Finally, Liu and colleagues145 were not able to determine if the final hepatocyte-like cells derived from their iPSCs were mature hepatocytes containing proper gene expression profiles at levels comparable to whole liver.
Fully functional hepatocytes derived from induced pluripotent stem cells would be a valuable tool for drug development or evaluation of the contribution of genetic variation to variable responses. However, due to current disadvantages and technical limitations, neither induced pluripotent nor embryonic stem cells are yet a widely accepted option for toxicological and pharmacological studies.
Much progress has been made towards improving traditional in vitro models that sustain liver-specific functions and can accurately predict in vivo responses to toxicants. These models, however, still lack the ability to provide long-term culture without cell necrosis or de-differentiation. Although each traditional in vitro system has advantages and disadvantages, the use of three-dimensional systems with dynamic flow for primary cell and/or stem cell cultures appears to be the most promising for toxicological studies. One can imagine the implications of a physiologically accurate 3D system or BAL and the availability of a large number of highly functional stem cell populations. In addition, the ability to provide large biomasses of human-disease relevant cells could significantly reduce the need for animal use in toxicological studies. The use of co-cultures could further enhance the in vivo–like characteristics of a culture device and would provide more predictive results. As the needs for toxicity testing of ever greater number of chemical substances are not likely to diminish in the next decade, the development of the models for reliable, quick, and economical toxicity testing is of major importance. The need for quantities of hepatocytes capable of in vivo-like functions over long-term culture is urgent and with these new developments it appears that this goal may soon be within reach.
Supported, in part, by grants from the National Institutes of Health: P42 ES005948, and R01 ES015241.