Liver and Kidney on Chips: Microphysiological Models to Understand Transporter Function

SY Chang1, EJ Weber2, KP Van Ness2, DL Eaton1 and EJ Kelly2

Because of complex cellular microenvironments of both the liver and kidneys, accurate modeling of transport function has remained a challenge, leaving a dire need for models that can faithfully recapitulate both the architecture and cell-cell interactions observed in vivo.The study of hepatic and renal transport function is a fundamental component of understanding the metabolic fate of drugs and xenobiotics; however, there are few in vitro systems conducive for these types of studies. For both the hepatic and renal systems, we provide an overview of the location and function of the most significant phase I/II/III (transporter) of enzymes, and then review current in vitro systems for the suitability of a transporter function study and provide details on microphysiological systems that lead the field in these investigations. Microphysiological modeling of the liver and kidneys using “organ-on-a-chip” technologies is rapidly advancing in transport function assessment and has emerged as a promising method to evaluate drug and xenobiotic metabolism. Future directions for the field are also discussed along with technical challenges encountered in complex multiple-organs-on-chips development.

HEPATIC MICROPHYSIOLOGICAL SYSTEMS reaching the systemic circulation—the hepatic first pass effect Basic aspects of xenobiotic metabolism in the liver (Figure 1). To model a correctly functioning hepatic in vitro The creation and maintenance of an in vitro microphysiological sys- microphysiological system, it is imperative that the biotransfor- tem that correctly mimics hepatic function to study drug and xeno- mation enzymes are expressed and active in order to accurately biotic metabolism must include functional biotransformation study the parent compound and metabolite transporter enzymes and transporters. In this section, we review the primary dynamics. hepatic phase I/II/III enzymes and transporters and survey hepatic Biotransformation enzymes are often called drug-metabolizing in vitro systems with an emphasis on how well they have been used enzymes. These enzymes are involved in xenobiotic disposition to study metabolism. Overall, there has been a trend for the devel- via the processes of absorption, distribution, metabolism, and opment of more complex systems that utilize multiple cell types, elimination and are classified as either phase I (generally oxida- physiologically relevant geometries, and microfluidics to capture the tion, reduction, or hydrolysis reactions), phase II (generally con- complex interactions within the liver. Together, these attributes cre- jugation or hydrolysis reactions), or phase III (transmembrane ate a favorable cellular microenvironment that allows for the expres- transporters), based on their metabolic function. sion and correct cellular orientation of the transporters. Generally, phase I enzymes are responsible for catalyzing the The liver is the main organ in which xenobiotics and endoge- hydrolysis, reduction, and oxidation of xenobiotics. Among the nous compounds are metabolized and excreted due to its physio- phase I enzymes, cytochromes P450 (CYPs; detailed below) are logical placement “downstream” from the gastrointestinal tract, the largest and most important superfamily: CYP3A4, CYP2D, high perfusion rate, large size, and high concentration of bio- and CYP2C subfamilies are responsible for 50%, 25%, and 20% transformation enzymes. of the biotransformation of all drugs, respectively.1 Phase II To predict a compound’s metabolic fate, including detoxifica- enzymes can transfer a functional group to xenobiotics—which tion or bioactivation to a toxic metabolite, it is very critical to includes acetylation, methylation, glutathione conjugation, sulfate understand both the biotransformation enzymes and transporters conjugation, and glucuronidation. expressed in the liver. Because the liver receives nearly all of the When xenobiotics are absorbed into hepatocytes from the por- blood perfusing the gastrointestinal tract, it is anatomically situat- tal vein and hepatic artery, transporters in the sinusoidal (basolat- ed to be able to potentially remove xenobiotics (drugs and other eral) membrane of hepatocytes assist xenobiotics to enter chemicals foreign to the body) absorbed from the gut before hepatocytes for later biotransformation.

1Department of Occupational and Environmental Health Sciences, University of Washington, Seattle, Washington, USA; 2Department of Pharmaceutics, University of Washington, Seattle, Washington, USA. Correspondence: EJ Kelly ([email protected]) Received 7 June 2016; accepted 15 July 2016; advance online publication 22 July 2016. doi:10.1002/cpt.436

464 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt hydrolase and cytosolic epoxide hydrolase); and paraoxonase (paraoxonase-1, paraoxonase-2, and paraoxonase-3).2

Reduction NAD(P)H- quinone oxidoreductases (NQO1 and NQO2), aldo-keto reductases, carbonyl reductase, CYP b5/NADH- cytochrome b5, nicotinamide adenine dinucleotide phosphate: P450 reductase, aldehyde oxidase.

Oxidation Examples of enzymes involved in xenobiotic oxidation include: aldehyde dehydrogenases 1 and 2, alcohol dehydrogenases-1, aldehyde oxidase, xanthine oxidase, monoamine oxidase, peroxidase- glutathione peroxidase, ﬂavin-dependent-monooxygenase 3, 4, and 5, and Cytochromes P450. CYPs are generally considered to be the most important group of oxidative enzymes involved in phase I biotransformation of xenobiotics. There are 56 different human CYP genes, but many are involved only in endogenous metabolic processes. Major human forms involved in xenobiotic biotransformation are: CYP1A1, IA2, 1B1, 2A6, 2A13, 2B6, 2C8, 2C9, 2C18, 2C19, 2D6, 2E1, 2F1, 2J2, 3A4, 3A5, 3A7, and 4F3.

Phase II enzymes: conjugation UDP-Glucuronosyltransferase. There are two major UDP- Glucuronosyltransferase (UGT) families involved in xenobiotic conjugation, UGT1 and UGT2. Within the UGT1 family are multiple gene variants that give rise to 10 (UGT1A1–1A10) dis- Figure 1 Liver kidney circulation and transporters. Graphic illustration of tinct gene products. The UGT2 family has two subfamilies, the relationship between subcellular localization of transporters in the liver UGT2A (3 members) and UGT2B (4 members). All UGT (top) and kidney (bottom) as well as the preferential flux directionality for enzymes use uridine diphosphoglucuronic acid as the cofactor. each individual transporter (arrows). Sulfotransferases (SULTs) are too multigene families of enzymes, with SULT1A, 1B, 1E, and 2A1 involved in xenobiotic The most abundant phase I/II enzymes and transporters conjugation. All SULT enzymes use 30 phosphoadenosine-50 expressed in the human liver are listed in Supplemental phosphosulfate as the cofactor. Glutathione S-transferases Table S1. One result of absorption, distribution, metabolism, (GSTs) are also involved in xenobiotic conjugation, with approxi- and elimination is to convert toxic xenobiotics into nontoxic, mately 15 different human genes, in five classes (GSTA1-5, water-soluble compounds that can easily be eliminated—a pro- GSTM1-5, GSTT1 and 2, GSTO1 GSTS1, and GSTZ1). In cess called detoxification or inactivation. However, in certain addition, GSTs use the tripeptide, glutathione, as a cofactor. cases, phase I and/or phase II enzymes can transform xenobiotics Other important conjugation reactions include N- into more toxic chemicals—a process called bioactivation. The acetyltransferase (NATs; NAT1 and NAT2), with the cofactor, concept of bioactivation has been adapted to the pharmacological acetyl coenzyme A, and several methyltransferases with different idea of the prodrug, in which the administered parent molecule substrate specificities (O-methyltransferase, N-methyltransferase, has little or no pharmacological activity but one or more metabo- and S-methyltransferase). lites act as the major contributor to the desired pharmacological response. For example, codeine is a morphine prodrug that Basolateral uptake transporters requires oxidative demethylation by CYP2D6 to form morphine These transporters mediate the hepatic uptake of xenobiotics and to achieve its pharmacological activity. endogenous substances, such as bile acid and cholesterol. These uptake transporters belong to a multigene family of solute carriers PHASE I ENZYMES OF PHARMACOLOGICAL AND (SLCs). In humans, the main uptake transporters are organic TOXICOLOGICAL SIGNIFICANCE anion-transporting peptides (OATPs; OATP1A2, 1B1, 1B3, Hydrolysis 2A1, and 2B1); the sodium-dependent taurocholate co- There are multiple gene products with a wide variety of potential transporting protein (Na1-taurocholate co-transporting poly- substrates, including: carboxylesterase; alkaline phosphatase; peptide [NTCP]); and the organic cation transporters (OCTs; dipeptidyl peptidase-4; epoxide hydrolases (microsomal epoxide OCT1 and OCT2.3 Many of these carrier proteins can transport

CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 100 NUMBER 5 | NOVEMBER 2016 465

substrates bidirectionally, depending on the substrate concentra- tauroursodeoxycholate, glycochenodeoxycholate, and glycocho- tion gradient. late. In addition, BSEP can transport pharmaceuticals, such as OATPs are integral membrane proteins with 12 transmem- pravastatin (reviewed by ref. 7). brane helices and are the main drug carrier proteins supporting BCRP transports a highly diverse range of hydrophobic sub- the sodium-independent hepatic uptake of organic anions.4 strates, including chemotherapeutic agents, such as mitoxantrone, Because of the prominent expression of OATPs on the basolat- methotrexate, topotecan, and irinotecan. BCRP can also trans- eral membrane of hepatocytes, OATPs are responsible for a criti- port hydrophilic-conjugated organic anions, particularly sulfated cal mechanism of chemical uptake into the liver, including a conjugates with high affinity (reviewed by ref. 13). variety of substrates containing steroidal or peptide structural Genetic defects or secondary consequences of hepatobiliary backbones and/or anionic or cationic chemicals. For example, obstruction or destruction can cause cholestasis, which are OATP1A2 is associated with the uptake of sulfobromophthalein, often involved in impaired function or a sustained inhibition BQ-123, [d-Pen2, d-Pen5]-enkephalin, fexofenadine, levofloxacin, of these apical efflux transporters. Inherited mutations in the ouabain, and methotrexate. OATP1B1 and OATP1B3 are the human MDR3 gene can cause progressive familial intrahepatic OATP1B isoforms expressed in human livers5 and can transport cholestasis type-3, a rare disease characterized by an early onset bilirubin and its glucuronide conjugates.6 OATP1B1 not only of cholestasis that leads to cirrhosis and liver failure before transports various statin drugs, but can also carry thyroxine, taur- adulthood.14 In addition, mutations in BSEP are responsible ocholate, and dehydroepiandrosterone sulfate (reviewed by ref. for patients with progressive familial intrahepatic cholestasis 7), whereas OATP2A1 supports the transport of prostaglandins, type-2 with high serum bile acid concentrations and low biliary 8 including PGE2 and PGF2a. bile acid but normal serum g-glutamyltranspeptidase activity Chemicals and drugs that belong to the small class I organic and cholesterol.15 cations, including tetramethylammonium, tetraethylammonium, tetrabutylammonium, thiamine, choline, dopamine, serotonin, Basolateral efflux transporters histamine, adrenalin, and noradrenalin, are transported in hepa- Removal of endogenous and xenobiotic chemicals from hepato- tocytes by OCT1 (reviewed by ref. 3). cytes to sinusoidal blood is mediated by transporters on the baso- NTCP predominantly transports bile salts and sulfated com- lateral side, including MRP3, MRP4, and organic solute and pounds in a sodium-dependent manner in addition to thyroid steroid transporter, Ost alpha-Ost beta (OSTa/b). hormones, estrone 3-sulfate, and certain statin drugs, such as MRP3 has a high affinity for glucuronide conjugates which rosuvastatin and pitavastatin.9–11 A recent study implicated are involved in detoxification and excretion of polar chemicals NTCP as the receptor for hepatitis B and D viruses, showing its that have undergone the process of glucuronidation, including clinical importance.12 morphine-3-glucuronide, bilirubin-glucuronide, etoposide- glucuronide, and acetaminophen-glucuronide. It is suggested Apical efflux transporters that MRP3 has a defense-related function and contributes to These transporters are located on the apical surface of hepato- the excretion of toxic anions, as expression is upregulated dur- cytes and pump out endogenous metabolites and xenobiotics via ing hepatic injury, such as cholestasis as it is associated with biliary excretion processes. These transporters belong to the bile acid homeostasis in spite of low affinity for bile acids ATP-binding cassette (ABC) transporter family containing ATP- (reviewed in ref. 7). binding domains that have ATPase activity to provide the energy MRP4 has a wide range of substrates, including antiviral agents necessary for active transport of substrates across the cell mem- (azidothymidin, adefovir, and ganciclovir), anticancer agents brane, most often against a 100 to 1,000-fold concentration gra- (methotrexate, 6-mercaptopurin, and camptothecins), and car- dient (reviewed by ref. 7). The most abundant and important diovascular agents (loop diuretics, thiazides, and angiotensin II transporters on the apical side of hepatocytes are the multiple receptor antagonists), as well as endogenous chemicals (steroid resistance proteins (MRPs; MRP2), multidrug resistance- hormones, prostaglandins, bile acids, and the cyclic nucleotides associated proteins (MDRs; MDR1 and MDR3), bile salt export cAMP and cGMP). MRP4 has higher affinity for sulfate conju- pump (BSEP), and breast cancer resistance protein (BCRP/ gates of bile acids and steroids. MRP4 is similar to MRP3, as ABCG2; reviewed by ref. 7). both are upregulated in cholestasis, suggesting a protective role in MRP2 is involved in the efflux of both hydrophobic preventing hepatotoxicity. Furthermore, MRP4-null mice devel- uncharged molecules and water-soluble anionic compounds oped cholestasis after bile duct ligation, implying that MRP4 is (reviewed by ref. 3). MDR1 (ABCB1, P-glycoprotein) can trans- important in bile acid homeostasis (review by ref. 7). port amphipathic organic cations and neutral compounds across OSTa/b proteins are present as heterodimers and/or hetero- the canaliculus membrane to the bile. MDR1 substrates include: multimers in the cell membrane. OSTa/b-mediated transport is glutathione, glucuronide, and sulfate conjugates; many macrolide bidirectional (uptake or efflux) and ATP-independent, depend- antibiotics, such as erythromycin, azithromycin, and clarithro- ing on the electrochemical gradient. Although it is expressed in mycin; and chemotherapeutics, such as tamoxifen, doxorubicin, high levels in the liver, OSTa/b is also expressed widely in the and paclitaxel (reviewed by ref. 3). MDR3 (ABCB4) can trans- small intestines, colon, kidneys, testes, ovaries, and adrenal gland, port phospholipids, whereas BSEP primarily transports conjugat- the latter of which is involved in steroid and bile acid homeosta- ed bile acids, including taurochenodeoxycholate, taurocholate, sis. The evidence from a study with Ost alpha null mice

466 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt LNCLPAMCLG THERAPEUTICS & PHARMACOLOGY CLINICAL Table 1 Current liver in vitro cell culture system Co-culture or inte- Validated functions Representative grated with other of phase I/II and Other liver-specific Methods used to image Platform Scaffold/format organ tissues transporters functions assess hepatotoxicity HepatoPac Micropatterned 24- Primary hepatocytes Metabolite measure- Albumin (ELISA) and MTT assay, ATP, and Hepregen32,33 and 96-well plates islands co-cultured ments: CYP1A: urea (colorimetric glutathione levels with 3T3 fibroblasts ethoxy-resorufin; endpoint assay) using Cell Titer-Glo CYP2A6: coumarin; secretion and GSH-Glo lumi- CYP2B6: bupropion; nescent kits CYP3A4: testoster- (Promega) one; UGTs and SULTs: 7- hydroxycoumarin OUE10NME OEBR2016 NOVEMBER | 5 NUMBER 100 VOLUME | mRNA expression : phase I CYPs, FMO, MAO, XO, EH; phase II- UGTs, SULTs, NATs, GSTs, methyltransferases transporters-MDR1, MRP3, OCT1, NTCP, BSEP canalicular flow by CMFDA fluorescent probe RegeneMed Liver 24-well Transwell Primary human or rat Phase I activity: Albumin, urea, fibrin- Levels of ATP and Transwell plate hepatocytes with CYP1A1, 2C9, and ogen, transferrin pro- GSH, LDH releasing RegenMed35 hepatic NPCs, includ- 3A4 by P450-Glo duction (ELISA); ing vascular and bile assays (Promega) glycogen synthesis duct endothelial Basolateral trans- (isotope labeling) cells, Kupffer cells, porter uptake mea- LPS-induced IL-1b, and hepatic stellate sured by isotope IL-6, IL-8, TNF-a,GM- cells labeled tracer- 3H- CSF: multiplex-based labeled estrone-3- ELISA sulphate 3D InSight Spheroids in 96-well Primary human hepa- CYPs activity/ Albumin secretion ATP, and GSH levels Microtissues plate tocytes with NPCs induction: CYP1A1, (ELISA) IL-6 and TNF- using CellTiter-Glo InSphero36 (Kupffer and endo- 2B6, 2C8, 2C9, a releasing (ELISA) and GSH-Glo lumi- thelial cells) 2C19, 2D6, CYP3A4 nescent kits (Prom- and 2E1 (IHC stain- ega) LDH releasing, ing) MDR1, BSEP mitochondrial (IHC staining) activity

3D print Liver 3D bioprinted liver Primary human hepa- CYP450 enzymes Albumin and trans- LDH releasing, GSH Organovo tissues in a Trans- tocytes with NPCs 3A4 (mRNA expres- ferrin production. and ATP level. Gene/ well plate sion levels and Immunologic: IL-6, protein expression probe metabolite GM-CSF,MCP-1, LPS proﬁling and histo- formation) stimulated TNF-a, IL- logical tissue 1b, IL-12p70, IL-10, assessment IL-2, IL-13, and IL-4

467 Table 1 Continued on next page 468 Table 1 Continued

Co-culture or inte- Validated functions Representative grated with other of phase I/II and Other liver-specific Methods used to image Platform Scaffold/format organ tissues transporters functions assess hepatotoxicity HUREL hepatic cul- mCCA biochip con- Primary human hepa- mRNA expression MTT assay, ATP, and tures, Hurel38 nected with pump. tocytes with NPCs and metabolite for- GSH levels using mCCA biochip can mation: CYP450 CellTiter-Glo and integrate multiple enzymes 1A2, 3A4, GSH-Glo lumines- organ tissues 2C19, SULT, UGT. cent kits (Promega), Metabolic clearance: LIVE/DEAD stain biliary efﬂux assay (Thermo Fisher) (LC-MS/MS) BSEP canaliculi (CMFDA) LiverChip System CN Multiwell plate plat- Primary hepatocytes Metabolite LIVE/DEAD stain Bio Innovations39–41 form with pneumatic co-cultures with formation: CYP450 (Thermo Fisher) pump NPCs and Kupffer enzymes 1A2, 2B6, cells 2C9, 2D6, 3A4

CellAsic42 Primary human Metabolite forma- LIVE/DEAD stain hepatocytes tion: CYP450 (ﬂuorescent probe; enzymes 1A1/2, Thermo Fisher) 2C9, 3A4, UGT, BSEP canaliculi (CMFDA)

SQL-SAL, University Nortis chip (Nortis) Primary human hepa- Metabolite forma- TNF-a releasing LDH leakage, fluo- of Pittsburgh43 with microfluidic flow tocytes with cell tion: CYP450 rescent protein ROS

OUE10NME OEBR21 | 2016 NOVEMBER | 5 NUMBER 100 VOLUME lines derived from enzymes 1A, 2C9, biosensors. Stellate NPCs 3A4, IHC staining: cell activation and MDR1, MRP2, BSEP, migration BRCP IdMOC 24 or 96-well plate Primary human hepa- MTT assay, ATP, and In Vitro ADMET Lab72 format integrated tocytes with 3T3 GSH levels using with multiple organs cells CellTiter-Glo and GSH-Glo lumines- cent kits (Promega) Microcompartment Hollow ﬁber Primary human liver Urea production, LDH and AST hollow ﬁber BAL37 cells (not pure albumin synthesis, releasing hepatocytes) glucose secretion, and lactate produc- www.wileyonlinelibrary/cpt tion, IHC: expressed cell marker of Kupffer cells

ATP, adenosine triphosphate; BAL, bioartificial liver; BSEP, bile salt export pump; CMFDA, carboxy-20,70-dichlorofluorescein diacetate; CYP, cytochrome; EC, endothelial cells; EH, epoxide hydrolase; ELISA, enzyme-linked immunosorbent assay; FMO, flavin-dependent-monooxygenase; GM-CSF, granulocyte-macrophage colony-stimulating factor; GSH, glutathione; GST, glutathione S-transferase; HC, hepatocyte cells; HSC, hepatic stellate cell; IHC, immunohistochemical; IL, interleukin; LC-MS/MS, liquid chromatography tandem mass spectrometry; LDH, lactate dehydrogenase; LPS, lipopolysaccharide; MAO, monoamine oxidase; MDR, multidrug resistant; MRP, multidrug resistance-associated protein; MTT, methylthiazol tetrazolium; NATs, N-acetyltransferases; NPC, nonparenchymal cell; NTCP, Na1-taurocholate co-transporting polypeptide; OCT, organic cation transporter; ROS, reactive oxygen species; SQL-SAL, sequentially layered, self-assembly liver; SULT, sulfotransferase; TNF, tumor necrosis factor; UGT, UDP-glucuronosyltransferase; XO, Xanthine oxidase (XO). demonstrated OSTa/b as a target for interrupting the enterohe- Human liver subcellular fractions, microsome/supersomes, patic circulation of bile acids. OSTa/b substrates include steroid cytosol fractions, and S9 fractions hormones and endogenous compounds, such as estrone sulfate These subcellular fractions contain CYPs, UGTs or NATs, and dehydroepiandrosterone sulfate, bile acids, and PGE2, as well SULTs, and GSTs, which are useful for xenobiotic biotransfor- as the cardiac glycoside digoxin (reviewed by ref. 16). mation research. These assays are traditionally used for in vitro- The study of hepatic transporter function has relied on (1) in based predictions of metabolic clearance and drug-drug interac- vitro/ex vivo hepatocytes in suspension or two-dimensional (2D) tions. However, due to the loss of structural integrity of the cell, plated monolayer cell formats for uptake and/or accumulation and the optimization of enzyme kinetic conditions by adding efflux assays using radioactive or fluorescent probe substrates, and cofactors, such as nicotinamide adenine dinucleotide phosphate (2) in vivo pharmacokinetic studies on mutant animals with defi- and 30 phosphoadenosine-50 phosphosulfate that are at concen- ciencies in specific transporter genes or transporter gene knock- trations irrelevant to physiological conditions, the results using out mice (reviewed by ref. 17). There are several in vitro methods these methods cannot be accurately used for transporter studies used in assessing human drug metabolism and active transport of and quantitative estimations of in vivo human biotransformation. the drug, including the use of immortalized cell lines with tran- sient or stable overexpression of transporters. In vitro transporter Cell lines cultured in 2D assays can help determine whether a compound is taken up at the The HepG2 cell line is the most frequently used and best charac- sinusoidal surface by hepatocytes and/or whether its metabolites terized immortalized human hepatoma cell line. However, com- can be eliminated at the canalicular membrane for biliary excre- pared with primary human hepatocytes, overall CYP activity tion. Uptake and inhibition assays often involve OATP1B1, remains low.22 Expression profile of transporters in HepG2 cells OATP2B1, and OATP1B3, and biliary efflux assays may include is not highly correlated to human liver tissue so the HepG2 cell MDR1, MRP2, and BCRP, which can be used to predict com- line is not a suitable model for human transport assays. In gener- pound disposition. The BSEP inhibition assay can be applied to al, 2D culture conditions cannot provide the optimal microenvi- screen whether the compound can lead to cholestasis or hyperbi- ronment for cells to establish polarization; thus, the use of 2D lirubinemia (reviewed by ref. 18). cell cultures has architectural limitations in transporter assays. In some instances, preclinical in vitro cell model systems poorly A new human liver cell line derived from a hepatocellular car- predict biotransformation and elimination in humans. First, cinoma (HepaRG) recently drew substantial attention in the field extrapolation from in vitro findings to the in vivo situation of pharmaceutics and toxicology. Differentiated HepaRG cells remains complex with poor in vitro-to-in vivo correlation. expressed high levels of phase I/II enzymes, including transporter Additionally expression levels of phase I/II enzymes and levels comparable to freshly isolated human hepatocytes.22 Hep- transporter in transformed cell lines, such as HepG2 (human aRG cells can maintain a proliferative state in undifferentiated hepatoma cells), are very low and variable (reviewed by ref. 19). culture medium for several weeks and can differentiate into hepa- OCT1 and OATP1B1 mRNA were abundantly expressed in tocytes and biliary epithelial cells by adding differentiation cul- human liver tissue, whereas these two transporters were expressed ture medium after reaching confluence.22 at low levels in HepG2 cells.20 Thus, transporter expression in HepG2 did not match the tissue expression pattern. Primary human hepatocytes suspension and cultured in 2D Furthermore, due to overlapping substrate specificities and lack Primary human hepatocyte cultures are a preferred in vitro system of selectivity with currently available inhibitors, it is challenging for predicting in vivo drug biotransformation and clearance as to fully understand the role of a given transporter in the disposi- they maintain critical metabolic features. After isolation by colla- tion of specific drugs. Even though in vivo pharmacokinetic stud- genase perfusion, primary human hepatocytes in suspension are ies can provide integrated analysis of a drug’s disposition, the viable for only a few hours thus they are immediately used as preclinical results from transgenic or mutant animal models rodent models for kinetic characterization of transporter func- sometimes fail to predict the clinical outcomes. Expression pro- tion. However, this rapid use and characterization is generally files of transporters in laboratory animals, such as rats and mice, not possible for human hepatocytes. Thus, studies with human are different from humans. For example, rodent mdr1a and hepatocytes rely on establishing primary cultures. Once plated in mdr1b genes are correlated to the MDR1 gene in humans despite a monolayer culture, human primary hepatocytes maintain suit- functional studies showing MDR1, mdr1a, and mdr1b expressing able viability for several days. However, they usually lose cell- cells demonstrating substrates for rodent mdr1a and mdr1b are specific functions, such as albumin production and CYP expres- unlikely to be substrates for human MDR1, showing species- sion, as both decline quickly over the first 24–48 hours of culture dependency in the spectrum of drug efflux activity.21 when the cells lose their differentiation status. Due to the scarcity We review in vitro/ex vivo human hepatic cell systems used in of available human liver tissue and successful cryopreservation biotransformation and transporter studies from traditional assays techniques, a good supply of human primary hepatocytes has to the most recently advanced three-dimension (3D) cultures, become commercially available. In culture, previously cryopre- including microsomes, cell lines cultured in 2D, primary hepato- served hepatocytes can recover and maintain phase I/II enzyme cyte suspensions, liver slices, sandwich cultures, 3D culture sys- activity after thawing for at least 7 days.23 Individual donor varia- tems, and microphysiological system (MPS) culture systems tion in metabolic enzyme activity due to genetic polymorphisms (Figure 2 and Table 1). and other factors can be compensated for by the mixing of

CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 100 NUMBER 5 | NOVEMBER 2016 469

Figure 2 Comparisons of the commonly used methodologies used to model hepatic metabolism and transport, including the pros and cons of each modality, with increasing complexity illustrated from left to right. hepatocytes from multiple donors to generate homogeneous periods. The use of ECM overlays allows for a favorable cellular enzyme activities. Hepatocytes represent the majority of the attachment environment and is thought to be one of the main hepatic cellular mass (about 80%), whereas other nonparenchy- reasons why cells polarize in this type of culture. Hepatocyte mal cells (NPCs), including vascular and biliary epithelial cells sandwich cultures with various medium constituents have been (i.e., cholangiocytes), Kupffer cells, and hepatic stellate cells, pro- shown to maintain albumin secretion, viability, and cuboidal- vide key physiological functions. For example, cholangiocytes not shape morphology with phase I/II/transporter expression similar only can contribute to bile secretion, but also can enable the to that of liver tissue.27 Biliary excretion can also be evaluated in absorption of ions, bile acids, amino acids, glucose, and other sandwich cultures with both basal and inducible biliary enzyme molecules, playing an important role in the modification of activities that allow assessments of hepatobiliary disposition. hepatic canalicular bile (reviewed by ref. 24). In vivo, stellate and Taken together, sandwich cultures can provide a robust means to Kupffer cells have been shown to play an important role in the evaluate hepatic compound uptake, metabolism, efflux and biliary hepatoxicity of some compounds, and, thus, the absence NPCs in excretion, while closely mimicking in vivo characteristics. primary culture systems is a potentially serious limitation for toxi- Compared with other 2D models, sandwich cultures have signifi- cology studies. cant advantages and can be considered as a bridge between 2D and 3D cultures technologies (reviewed by ref. 28). Precision-cut liver slices Precision-cut liver slices have several advantages for drug biotrans- Transwell culture for drug efflux formation and toxicology studies as they maintain the native liver Evaluation of the human hepatocyte uptake and efflux transport- structure with multiple cell types and zonation, and have good in ers MDR and MRP2 have been performed in transwell systems vitro-in vivo correlations of drug biotransformation features. with transfected cell lines, including porcine kidney epithelial Cultured liver slices can retain phase II enzyme activity, albumin cells and Mad-Darby canine kidney (MDCK) epithelial cells. production, and gluconeogenesis for up to 20–96 hours, while Transwell culture cells are grown on a permeable membrane filter regulating gene expression of the uptake transporters, NTCP, that allows for the physical separation of the apical and basolat- OATP, and efflux transporters, BSEP, MDR1, and MRP2.25 eral domains and has been used to study drug uptake and efflux Despite the preservation of the overall hepatic architecture, transporter activity.29 drug biotransformation and intrinsic clearance rates are lower than isolated hepatocytes as necrosis can occur after 48–72 hours Co-culture systems, 3D culture systems, and MPS culture whereas CYP activities are greatly reduce within 6–72 hours.26 systems Traditional in vitro methods, such as microsomes and suspension Sandwich culture cultures, usually have too short of a time window to perform essen- Sandwich cultures with primary human hepatocytes plated tial disposition assays, which can lead to imprecisions in the predic- between two layers of extracellular matrix (ECM; collagen or tion of human biotransformation/clearance and toxicity. In order Matrigel, derived from Engelbreth-Holm-Swarm sarcoma) were to improve the predictability of drug safety and efficacy in clinical developed to maintain liver-specific functions over longer culture development, and to have a clearer perspective of toxicity outcomes

470 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt LNCLPAMCLG THERAPEUTICS & PHARMACOLOGY CLINICAL Table 2 Current kidney in vitro cell culture system Co-culture or inte- Validated functions grated with other of phase I/II and Other kidney-specific Methods used to Representative image Platform Scaffold/format organ tissues transporters functions assess nephrotoxicity Bioartiﬁcial Renal Permeable synthetic MDCK, primary por- Phase I Fluid reabsorption, Cytokine release Tubule Assist hollow-ﬁber cine, primary human 25-(OH)D3-12-hydrox- bicarbonate trans- 58,59 Device membrane proximal tubule cells ylase: 25-OH-D3 port, glucose trans- Transporter function port, GSH transport, OAT1/3: and metabolism, Para- Ammoniagenesis amminohippurate Human kidney proxi- Porous membrane Primary human proxi- Transporter function FITC-albumin uptake, LDH release (cisplat-

OUE10NME OEBR2016 NOVEMBER | 5 NUMBER 100 VOLUME | mal tubule-on-a- integrated within a mal tubule cells MDR1: calcein-AM alkaline phospha- in toxicity), apoptosis chip61 PDMS multilayer tase activity, glucose via TUNEL assay system transport

Human proximal Coated HFM ciPTEC Transporter function tubule epithelial OCT2: 4-(4-(dimethy- cells cultured on lamino)styryl)-N- HFMs65 methylpyridinium iodide (ASP1)

hKDCs cultured on SIS derived from por- hKDC, HK-2 cells N/A FITC-BSA uptake SIS66 cine small intestine to serve as a natural scaffold Kidney derived Decellularized kidney HK-2 cells N/A qPCR data of kidney micro-scaffolds fragments with intact associated genes, (KMS)67 ECM cytokine release

Nortis microphysio- Collagen type I Primary human proxi- N/A Cell-associated logical device68,69 matrix coated with mal tubule cells staining of KIM-1 collagen type IV ECM and HO-1 upon protein CdCl2 exposure

Perfused kidney-on- Porous polyester MDCK cells N/A Live/dead stain, a-chip70 membrane cut out KIM-1 concentration, from Transwell and permeability plates assessment in response to gentamicin exposure

BSA, bovine serum albumin; CdCl2, cadmium chloride; ciPTEC, conditionally immortalizing proximal tubule epithelial cell; ECM, extracellular matrix; FITC, ﬂuorescein isothiocyanate; GSH, glutathione; HFM, hollow ﬁber membrane; HK, human kidney; hKDC, human kidney-derived cell; HO, heme oxygenase; KIM, kidney injury molecule; LDH, lactate dehydrogenase; MDCK, Manine-Darby canine kidney; MDR, multidrug resistant; N/A, not applica- ble; OAT, organic anion transporter; OCT, organic cation transporter; OH, hydroxy; PDMS, polydimethyl siloxane; qPCR, quantitative polymerase chain reaction; SIS, small intestinal submucosa; TUNEL, terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labeling. 471

while reducing the use of animals for toxicity studies, recent secretion, are preserved over 5 weeks. MDR1 and BSEP are research efforts have been focusing on development of advanced in expressed in this microtissue, evidence that these cultures exhibit vitro models based on the applications of co-culture systems, 3D cell polarization and bile canaliculi formation. Inflammation- cultures, and MPS cultures that utilize microfluidic flow, often are mediated toxicity and chronic toxicity assays with acetaminophen referred as “organ-on-chips” or “organoid culture.”30 and diclofenac have also been evaluated in this system. However, The liver is a complex organ, consisting of hepatocytes, NPCs, to date, there are no published data that demonstrate that this and various ECMs. NPCs play an important role in hepatic phys- system can be applied to transporter assays. iological functions as well as hepatotoxicity. Many studies have Organovo (San Diego, CA) uses a 3D bioprinting technique to demonstrated that co-cultures of hepatocytes with NPCs can sus- generate small-scale hepatic tissues using human primary cells in tain liver-specific function, morphology, and expression of liver- a platform called exVive3D Liver. This product can maintain sta- specific transcription factors via the activation of cell adhesion ble viability and albumin secretion for up to 4 weeks and has molecules and redistribution of cytoskeleton involved in cell-cell rifampicin-inducible CYP3A4 activity, as midazolam biotransfor- and cell-matrix interactions (reviewed in ref. 31). Several liver mation to 1-hydroxy-midazolam was increased by rifampicin pre- organoid cultures based on the application of co-culture have treatment. As Kupffer cells are also present, this system can been reported, and some have been commercialized. These respond to immune stimulation (LPS) with the release of proin- include advanced 3D culture systems based on cellular microenvi- flammatory cytokines. ronment dynamics between ECMs, microperfusion flow rates, The 3D liver bioreactors designed by the Charite Universit€ats- and co-cultures of various cell types. The MPS model represents medizin Berlin are derived from bioartificial livers used in the an interconnected set of cellular constructs designed to recapitu- clinic. This hollow-fiber and perfusion-based bioreactor provides late the structure and function of human organs. Here, we review a continuous mass exchange of culture media and controlled oxy- current advanced hepatic culture systems (Table 1). genation with a scaffold for cells to maintain a physiologically rel- HepatoPac is a co-culture system of human hepatocytes with evant environment.37 This 3D bioreactor is co-cultured with mouse fibroblasts (3T3-J2 fibroblast), commercially available hepatocytes and NPCs and can maintain albumin secretion and through Hepregen (Medford, MA). This system consists of CYP activity (CYP1A, CYP2C9, and CYP3A4) for up to 2–3 micropatterned hepatocyte islands surrounded and stabilized by weeks, as well as expression of canalicular transporters (MRP2, stromal cells in a 24-well plate format. This culture system can MDR1, and BCRP). Limitations of this system for use in phar- maintain liver-specific function for up to 6 weeks, including sta- macokinetic and drug toxicity testing include a lack of zonation ble albumin secretion, urea synthesis, phase I/II drug biotransfor- seen in liver tissues and low throughput, as only one condition mation, and formation of bile canaliculi with efflux can be evaluated per system. transporters32 (reviewed by ref. 33). Studies with the human The HlREL microdevice, an MPS platform provided by HepatoPac platform have demonstrated an in vitro-in vivo corre- Hurel (Beverly Hills, CA), is an integrated and microfluidic sys- lation with hepatic uptake of faldaprevir by multiple transporters tem that assembles multiple units of microfluidic microscale cell (including OATP1B1 and Na1-dependent transporters) and bio- culture analogs (lCCA) cultured from hepatic tissue or other transformation by CYP3A4.34 organ tissues in parallel. The Hurel plastic biochips are connected RegeneMed 3D Liver (San Diego, CA) is a liver tissue co- to a fluid reservoir and pump system interconnected with a com- culture system used for screening hepatic absorption, distribution, plex set of tubing that serves to recirculate the media. The hepatic metabolism, and elimination, using the transwell culture co-culture system with NPCs forms a 2D monolayer and main- approach.35 NPCs are seeded in a nylon screen sandwich mesh tains high viability for up to 9 days, with higher expression of insert with a 140-lm pore size and stabilized for a week, followed CYPs, SULT, and UGT, and in vivo-like hepatic clearance of by incubation with hepatocytes to form a 3D liver tissue. Liver- diclofenac, indomethacin, and coumarin, compared with tradi- specific functions, including production of albumin, fibrinogen, tional static culture conditions.38 The formation of the bile cana- transferrin, and urea, can be maintained up to 3 months, and the liculi network was visualized by using carboxy-20,70- induction of CYP1A1, 2C9, and 3A4 activity for up to 2 months. dichlorofluorescein diacetate, a fluorescent substrate of BSEP. Co-culture with Kupffer cells allows for study of the inflammatory However, there was no evidence to demonstrate that this system response as the release of proinflammatory cytokines can be can maintain other hepatic functions, such as albumin secretion, observed with lipopolysaccharide (LPS) exposures. Basolateral urea excretion, or transporter activity. transport-mediated drug uptake activity using 3H-labeled estrone- A multiple-well plate platform, LiverChip, by CN Bio Innova- 3-sulphate as the tracer has been demonstrated in this system. tions (Hertfordshire, UK) uses a flow system driven by a pneu- Although images of bile canaliculi-like structures were presented, matic pump at the bottom of the plate. This system was designed efflux transporter activity or bile excretion were not reported. to recapitulate the hepatic microenvironment in terms of fluid The 3D microtissue spheroid culture, 3D InSight, provided by flow, oxygen gradient, and shear stress (145 lMto50lMata InSphero (Schieren, Switzerland), is a hanging drop co-culture flow of 0.25 mL/min). Compared with 2D static conditions, system that uses gravity-forced cellular self-assembly of hepato- hepatocytes cultured in this system can maintain CYP activity cytes and NPCs into spheroids using a 96-well format.36 This (1A2, 2B6, 2C9, 2D6, and 3A4), albumin secretion, expression format is well suited for high throughput applications and stable of phase II-UGT enzymes and transporters (MDR2, MRP2, and viability and liver-specific function, such as persistent albumin BCRP) for up to 7 days.39 Hepatocytes co-cultured with liver

472 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt sinusoidal endothelial cells enriched NPC fractions can maintain the liver section, we review renal transporters in the proximal high viability, CYP activity, albumin secretion, and urea excretion tubule and in vitro renal systems that have been used to study for up to 13 days.40 Kupffer cells incorporated in the model can their function. Knowledge of the function of renal transporters release proinflammatory cytokines by LPS stimulation. The in the proximal tubule is important for understanding the crucial intrinsic clearance values of hydrocortisone and its metabolites role this organ plays in drug disposition. Demonstration that generated in this system correlated well with human data, dem- these transporters are functional in an in vitro renal system is crit- onstrating that this system has great potential for high through- ical for method validation and proper data interpretation. put use in hepatic metabolic research.41 The nephron is the functional unit of the kidney responsible CellAsic (Hayward, CA) has a microfluidic liver sinusoid mod- for maintaining both the homeostasis of electrolytes as well as el with a microporous endothelial-like barrier that mimics liver fluid volume.45 However, from a drug-development perspective sinusoids. This platform utilizes a 96-well plate format contain- the nephron is the primary site of renal handling involving any ing 32 small units that utilized a perfusion system with a flow of combination of filtration and phase III mediated-secretion/ 10–20 nL/min and 250 cells in each unit. Hepatocytes cul- reabsorption. Structurally, the nephron is a series of segmented tured in this system can maintain high viability for up to 7 days tubules that individually play a critical role in its overall function. and respond to drugs.42 Other liver-specific functions and the The proximal tubule, found in juxtaposition to the glomerulus, characterization of transporters need to be further analyzed. has been shown to be the primary site of transport-mediated 43 Vernetti et al. developed a human, 3D, microfluidic, four- reabsorption/secretion xenobiotics. Active vectorial transport of cell, sequentially layered, self-assembly liver model based on an xenobiotics are achieved given the polarized configuration of the MPS device platform by Nortis (Woodinville, WA). The current proximal tubule epithelial cells involving both transporters found sequentially layered, self-assembly liver model uses a co-culture of on the brush-border containing apical side (facing urine) and the primary human hepatocytes along with human endothelial basolateral side (facing systemic circulation; Figure 1). (EA.hy926), immune (U937), and stellate (LX-2) cells in physiological relevant ratios that are viable and functional for at least 28 Basolateral transporters days under continuous flow. This model can maintain canaliculi The SLC transporters are the major family of multispecific trans- structure, phase I/II activity, albumin production, and urea excre- porters that mediate the proximal tubule uptake of both xenobi- tion. In addition, by the integration of protein-based fluorescence otic and endogenous substrates that are circulating within the biosensors, the system can be used in reporting drug-induced blood. SLC transporters include the organic anion transporters mechanistic toxicity, such as apoptosis and reactive oxygen species (OATs; OAT1, OAT2, and OAT3), OCTs (OCT2 and for high throughput screening (reviewed by ref. 33). OCT3), and OATP4C1. The above-described 3D or liver-on-a-chip platforms generally provide suitable microenvironments for recapitulating most in Apical efflux transporters vivo hepatic functions compared with traditional 2D hepatocyte The SLC transporters are also found on the apical side of the cultures. However, these state-of-the-art in vitro/ex vivo technolo- proximal tubule and are responsible for the removal of exogenous gies still have limitations in accurately predicting human hepato- and endogenous substrates that have accumulated within the cell toxicity and first pass drug clearance. Further development of via diffusion, reabsorption, and/or uptake from the circulatory advanced 3D hepatic culture systems needs to consider the com- side. SLC transporters found on the apical side of the proximal plexity of the liver, including the co-culture ratio between hepato- tubule include the OAT4, multidrug and toxin extrusion cytes and NPCs, the source of hepatocytes, and zonation effects. (MATE) protein (1 and 2-K), urate anion exchanger 1, and Because the liver has a wide range of diverse functions, hepatic OCTN1/2. Additionally, another superfamily of multispecific cells show large heterogeneity and plasticity of functions. Oxygen transporters found on the apical side includes the ABC transport- gradients, hormones, and ECM all can regulate zonal variations, ers, which use ATP hydrolysis to drive molecules across cell mem- which are reflected in drug biotransformation capabilities. It branes. The ABC transporters on the apical side include MDR1 would be interesting if these zonation characteristics could be or P-glycoprotein, BCRP, and MRP2/4. established and sustained on the 3D liver-on-chip systems.44 Although these 3D models have great potential in pharmacoki- Proximal tubule transport model–historical perspective and netic prediction of first pass drug clearance, additional “proof of current status concept” studies are needed to validate in vitro-to-in vivo correla- In recent studies, there have been vast improvements of the tions using selected clinical drugs. In addition, most current mod- modeling capabilities of xenobiotic transport using models and els are individual liver organ systems, lacking the gastrointestinal cell types that accurately recapitulate human physiology. and/or renal transport/metabolism modules, which are needed to Traditionally, researchers have used established immortalized cell understand the whole profile of drug absorption, distribution, lines from animals, including canine (MDCK), opossum, and metabolism, and elimination in vivo. porcine cells (LLC-PK1), as a cell source to optimize models.46–48 Transwell studies, using cells grown in monolayers on a Renal nephron microenvironment–proximal tubule permeable scaffold, are utilized to investigate the facilitated trans- The status of renal in vitro microphysiological systems to study port from one compartment to the other resembling the apical- transporter function is not as well developed as the liver. As with basolateral relationship seen in vivo.49

CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 100 NUMBER 5 | NOVEMBER 2016 473

Although established immortalized cell lines have remained as was demonstrated using the smaller RAD unit with albumin- the gold standard for in vitro transport studies, there are draw- induced oncotic pressure in comparison to fluid reabsorption at backs associated with their use, particularly the poor expression baseline and ouabain-inhibited state. To show the transport capa- and/or absence of human-specific transporters.50 To combat this bility of cells cultured within the RAD, a series of transport problem, investigators have used molecular techniques to trans- studies were demonstrated using bicarbonate, glucose, and para- fect the established animal cell lines with specific human trans- aminohippurate. Bicarbonate transport was assessed using the porters.51 Additionally, investigators have immortalized primary larger RAD unit and significant inhibition was observed in the renal cortex cells from human kidneys (HK-2) and have shown presence of specific carbonic anhydrase inhibitor, acetazolamide. that not only do they retain proximal tubule cell phenotypes, but Glucose reabsorption was assessed using the larger RAD unit and they also possess the functional aspect of transport and sensitivity significant inhibition of reabsorption was achieved in a dose- to toxin.52 However, as demands increase to resemble the in vivo dependent manner using the SGLT inhibitor, phlorizin. Para- phenotype to its highest extent, there are shortcomings with the aminohippurate secretion was assessed using the larger RAD unit immortalized human cell lines that research groups have and significant inhibition of secretion was achieved using the addressed.53 As models continue to become more advanced, there inhibitor, probenecid. To assess glutathione transport and metab- has been a widespread shift from immortalized cells to primary olism, the smaller RAD unit was used and significant inhibition renal cells to more accurately represent the native in vivo micro- of glutathione removal was achieved using the specific inhibitor environment.54 Brown et al.55 has shown, using primary human of g-glutamyl transpeptidase, acivicin. Last, the functional aspect proximal tubule cells grown on permeable filter supports, differ- of the RAD was demonstrated by both the production of ammo- entiated cells expressing a wide array of transporters as well as nia in response to a decrease in perfusion pH as well as the meta- functional transport activity of OAT1/3, MRP2/4, OCT2, bolic capacity to convert 25-OH vitamin D3 to its active form, 58 MDR1, and MRP2. 1,25-(OH)2D3. Future experiments illustrated the functional consistency of the RAD, via the integration of human proximal Engineered transport modeling of the proximal tubule tubule cells, in combination with a hemofiltration cartridge that The utility of bioengineered models to replicate function of the resembles an efficient metabolic replacement of kidney function kidney proximal tubule is summarized in Table 2. These have in both animals as well as a patient population with acute renal ranged from a macroscale designed for clinical use in patients failure.59,60 The RAD system has not only served as a translation- with kidney failure to microphysiological scales designed to study al therapy targeted against renal failure but a foundational model biochemical and toxicological processes. when approaching recapitulation of the functional aspect of the proximal tubule microenvironment. Bioartificial renal tubule assist device Documented as one of the earliest systems to effectively model Human kidney proximal tubule-on-a-chip the microenvironment of the kidney, researchers developed the Successful mimicry of the renal microenvironment is a complex bioartificial tubule renal assist device (RAD) with the perspective undertaking given the 3D architecture of the proximal tubule of improving renal replacement therapies and outcomes of and the fluidic environment to which it is exposed. A number of patients suffering from endstage renal disease.56 The RAD is research groups have developed multilayer systems using polydi- described as a perfusion bioreactor system in which cells are methyl siloxane scaffolds in combination with a porous mem- grown in a monolayer on a permeable synthetic hollow-fiber brane to which cells can culture in 3D. Jang and Suh61 developed membrane (270 lm inner diameter, 35 lm wall thickness, and their multilayer microfluidic device using two compartments 2.5 cm length). Earlier developments used MDCK cells as the (one static chamber and one flow chamber) separated by a porous cell source for system validations in which functional confluence membrane to which rat renal tubule cells were cultured and and fluid transport was demonstrated using radiolabeled inulin. polarized on. The multilayer microfluidic device provided a fluid- MacKay et al.57 observed >98.9% recovery of perfused radiola- ic microenvironment (fluid shear-stress of 1 dyne/cm2) similar to beled inulin in monolayer systems as well as a significant increase what has been observed in the nephron, which has been proposed in fluid reabsorption in response to albumin-induced oncotic to being a key component in tubule cell cytoskeletal reorganiza- pressure when compared to both a baseline and ouabain- tion and remodeling of junctional complexes.61 inhibited state. Further improvements of the RAD demonstrated Leading the organ-on-chips focus at the Wyss Institute, the its ability to reabsorb fluid, transport various solutes, glutathione Ingber Laboratory has improved upon the multilayer microfluidic transport and metabolism, ammoniagenesis, and vitamin D acti- device by microfabricating a polydimethyl siloxane microfluidic vation, using two different RADs varying in length, wall thick- device containing both an interstitial fluidic compartment and ness, and diameter (smaller unit: 200 lm inner diameter, 40 lm luminal flow channel (fluid shear-stress 0.2 dyne/cm2) separated wall thickness, and 17 cm length; larger unit: 250 lm inner by a porous membrane coated with extracellular matrix protein, diameter, 70 lm wall thickness, and 12 cm length). Using prima- collagen type IV. Using primary human kidney proximal tubule ry porcine cells from harvested proximal tubule segments, Humes epithelial cells cultured within the device, the fluidic proximal et al.58 first used radiolabeled inulin leakage (<5–10%) as a selec- tubule microenvironment was demonstrated by the presence of tion criterion for confluent monolayers to be used in other func- tight junction protein, ZO-1, as well as basolateral distribution of tional experiments. Like the previous study, fluid reabsorption Na/K-ATPase and cytoplasmic expression of aquaporin 1.

474 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt Furthermore, acetylated tubulin staining was used to visualize pri- submucosa to show the correct phenotype of the proximal tubule mary cilia, a fluid shear stress mechanosensor important for regu- using immunohistochemical staining as well as detecting base- lation of tubular morphology. Functionally, the microfluidic ment membrane proteins and microvilli. Finesilver et al.67 pre- device was used to evaluate albumin uptake, cellular alkaline pared kidney micro-scaffolds and cultured HK-2 cells within the phosphatase activity, and transepithelial glucose transport. Jang microstructures in comparison with traditionally cultured HK-2 and Suh61 observed, in response to static cultures, a significant cells. Cells grown within the kidney micro-scaffolds for 25 days uptake of fluorescein isothiocyanate (FITC)-albumin, a signifi- showed a significant (>twofold) increase in expression of genes cant level of alkaline phosphatase activity, and significant levels of AQP-1, ATP1B1, LRP-2, CCL-2, SLC23A1, and SLC5A2 glucose transported by fluidic cultures of proximal tubule epithe- (except for 21–24 days). Additionally, the research group showed lial cells. Additionally, the microfluidic device was used as a mod- less cytokine release in response to stimulation for HK-2 cells el of nephrotoxicity using the prototypical nephrotoxin, cisplatin grown in the kidney micro-scaffolds compared with cultures (100 lM), in the absence and presence of cimetidine, which has grown traditionally.67 Adler et al.68 and Kelly et al.69 are develop- been previously shown to suppress cisplatin-induced injury. ing a microfluidic system that uses primary renal epithelial cells Using lactate dehydrogenase (LDH) release and apoptosis, via cultured in 3D on a collagen scaffold coated with ECM proteins. annexin V staining and a terminal deoxynucleotidyl transferase- In response to cadmium chloride exposure, the microphysiologi- mediated deoxyuridine triphosphate nick-end labeling assay, sig- cal system showed sensitivity and expressed biomarkers (Heme nificant cisplatin-induced cellular injury was observed in static oxygenase-1/kidney injury molecule-1) of injury showing the cultures compared with fluidic cultures as well as a significant potential of the model to serve as a suitable predictor of toxicity decrease in injury when cimetidine was co-administered with cis- ex vivo. In recent light, additional groups are interested in the platin. Finally, to assess the transport capabilities of the microflui- toxicity modeling aspect, as evidenced by Kim et al.70 exposing dic device, cellular accumulation of calcein-AM was evaluated in fabricated microfluidic devices with MDCK cells cultured on the presence and absence of the P-glycoprotein ATP-binding cas- porous membranes to gentamicin over time. In response to genta- sette membrane transporter inhibitor, verapamil (100 lM). micin exposure, the researchers observed a significant difference Static cultures showed higher accumulation of calcein-AM of injury between culture types (static/shear) of HK-2 cells as (presence/absence of verapamil) in comparison with fluidic well as dosing regimens (D1 5 bolus mimicking regimen; D2 5 cultures lending evidence to the concept of the fluidic microen- continuous infusion regimen). vironment enhancing transporter expression and improving the phenotype of the proximal tubule cells.62,63 FUTURE DIRECTIONS Using unique conditionally immortalizing proximal tubule epi- Although great strides have been made in developing isolated thelial cells (ciPTECs) previously characterized by their laborato- hepatic and renal microphysiological systems, the integration of ry group, Jansen et al.64 functionally tested ciPTECs grown on the two organ systems to model complex metabolic processes is hollow fiber membranes (HFMs) using microfluidics. Upon mat- the current focus of several groups. Our group has developed uration of ciPTECs within the HFM, collagen IV ECM was an integrated liver-kidney MPS system for identifying poten- visualized using immunocytochemistry lending evidence to renal tially nephrotoxic liver-metabolized chemicals by connecting a lineage. Additionally, using FITC-inulin, transepithelial barrier liver-on-a-chip with primary rat or human hepatocytes, to a function was significantly greater (and caused less FITC-inulin kidney-on-a-chip with human proximal tubule epithelial cells 69 leakage) for cells containing HFM in comparison to cells with in MPS devices developed by Nortis (Woodinville, WA). To empty HFM. Morphologically, the mature cultured ciPTECs dis- test the hypothesis that first pass hepatic clearance of a neph- played well-developed organelles as well as cell-surface microvilli, rotoxic chemical might have significant importance in deter- a typical marker of the proximal tubule epithelial cell. To assess mining ultimate kidney toxicity, we utilized aristolochic acid I, the polarity of the ciPTECs cultured in the HFM, confocal a well-known nephrotoxin and carcinogen that undergoes microscopy was implored using immunocytochemistry to observe extensive hepatic metabolism to form toxic metabolites. Our the high expression of tight junction protein, ZO-1, and correct results provide mechanistic insights into the important role of localization of basolateral transport protein, OCT2. hepatic biotransformation for the kidney-specific toxicity of Furthermore, OCT2 transport activity was measured using real- AA-I toxicity. This integrated in vitro/ex vivo MPS model pro- time fluorescent 4-(4-(dimethylamino)styryl)-N-methylpyridi- vides a novel approach for investigating the mechanisms that nium iodide (ASP1) uptake in the absence and presence of underlay pharmacokinetically and toxicologically important OCT2 inhibitors, cimetidine (100 lM) and uremic toxin mix, organ-organ interactions. showing significant inhibition of OCT2-mediated uptake.65 Other integrated MPS models include the HlREL microdevice platform that assembles multiple culture units of hepatic tis- Additional models of the human renal microenvironment sue with other organ tissues in parallel. Used in combination There are a number of groups currently developing outstanding with a mathematical modeling approach (pharmacokinetic/ models of the renal proximal tubule model that faithfully recapit- pharmacodynamic modeling), this novel platform provides ulate the microenvironment, but have yet to assess the transport improved predictability for drug biotransformation, clearance, capabilities. Hoppensack et al.66 used highly proliferative human and toxicity.71 The integrated discrete multiple organ co-culture kidney-derived cells cultured in monolayers on small intestinal by In Vitro ADMET Laboratories (Columbia, MD) uses the

CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 100 NUMBER 5 | NOVEMBER 2016 475

“wells-in-a-well” concept, interacting multiple cell types via the those of the Agency. The Environmental Protection Agency does not overlying medium. This model can mimic multiple organs in a endorse any products or commercial services mentioned in this publication. human body interacting via the systemic circulation.72 The ultimate goal of accurate assessment of human drug toxicity will rely on the development of in vitro platforms with multiple organs, VC 2016 American Society for Clinical Pharmacology and Therapeutics including the immune system, with each organ represented by multiple cell types and communication among organs achieved 1. Nebert, D.W. & Russell, D.W. Clinical importance of the cytochromes P450. Lancet 360, 1155–1162 (2002). using human plasma or equivalent. It would be interesting to see 2. Anzenbacher, P. & Anzenbacherova, E. Cytochromes P450 and more results of selected clinical drugs treated in these integrated metabolism of xenobiotics. Cell. Mol. Life Sci. 58, 737–747 models for validating in vitro-to-in vivo correlation. (2001). 3. Faber, K.N., Muller,€ M. & Jansen, P.L. Drug transport proteins in the There are still a number of technical challenges in multiple- liver. Adv. Drug Deliv. Rev. 55, 107–124 (2003). organs-on-chips development that have been important research 4. Kalliokoski, A. & Niemi, M. Impact of OATP transporters on areas for this field (reviewed by ref. 73). (1) Microfluidic volume pharmacokinetics. Br. J. Pharmacol. 158, 693–705 (2009). 5. Kullak-Ublick, G.A. et al. Organic anion-transporting polypeptide problems are the physiological relevant scaling of fluid volumes B (OATP-B) and its functional comparison with three other and flow rates that are associated with individual organ micro- OATPs of human liver. Gastroenterology 120, 525–533 physiological systems are difficult to acquire, as delicate and reli- (2001). 6. Niemi, M., Pasanen, M.K. & Neuvonen, P.J. Organic anion able engineering systems are required for the connection of transporting polypeptide 1B1: a genetically polymorphic transporter reservoirs, pumps, and tubing to deliver accurate flow rates to the of major importance for hepatic drug uptake. Pharmacol. Rev. 63, cultured cells. In addition, the determination of the physiological- 157–181 (2011). ly relevant flow rate for hepatic cells cultured in microphysiologi- 7. Klaassen, C.D. & Aleksunes, L.M. Xenobiotic, bile acid, and cholesterol transporters: function and regulation. Pharmacol. Rev. cal systems is difficult to determine given that the liver is a 62, 1–96 (2010). complex organ architecturally with varying fluid channel areas 8. Kanai, N., Lu, R., Satriano, J.A., Bao, Y., Wolkoff, A.W. & Schuster, and perfusion rates that can expose hepatic cells to a wide range V.L. Identification and characterization of a prostaglandin transporter. Science 268, 866–869 (1995). of shear forces. One approach to this problem would be to incor- 9. Visser, W.E., Wong, W.S., van Mullem, A.A., Friesema, E.C., Geyer, J. porate vascular systems within hepatic or renal chips to allow for & Visser, T.J. Study of the transport of thyroid hormone by cell-controlled flow rates between the vascular system and organ- transporters of the SLC10 family. Mol. Cell. Endocrinol. 315, 138– 145 (2010). specific cells. (2) Universal cell culture media are individual cell 10. Ho, R.H. et al. Drug and bile acid transporters in rosuvastatin hepatic types, especially primary cells, which require customized media uptake: function, expression, and pharmacogenetics. for optimal cell culture performance. A universal culture media Gastroenterology 130, 1793–1806 (2006). 11. Fujino, H., Saito, T., Ogawa, S. & Kojima, J. Transporter-mediated that can sustain multiple cell types from different organs will influx and efflux mechanisms of pitavastatin, a new inhibitor of need to be developed to successfully co-culture cells from multi- HMG-CoA reductase. J. Pharm. Pharmacol. 57, 1305–1311 ple organs with an optimal balance of nutrients, osmolality, pH, (2005). 12. Yan, H. et al. Sodium taurocholate cotransporting polypeptide is a and supplements. functional receptor for human hepatitis B and D virus. Elife 1, To address these challenges, the Defense Advanced Research e00049 (2012). Projects Administration, the National Center for Advancing 13. Ni, Z., Bikadi, Z., Rosenberg, M.F. & Mao, Q. Structure and function of the human breast cancer resistance protein (BCRP/ABCG2). Curr. Translational Sciences of the National Institutes of Health, the Drug Metab. 11, 603–617 (2010). Food and Drug Administration, and the Environmental Pro- 14. Falguie`res, T., A€ıt-Slimane, T., Housset, C. & Maurice, M. ABCB4: tection Agency have funded several groups for MPS develop- insights from pathobiology into therapy. Clin. Res. Hepatol. Gastroenterol. 38, 557–563 (2014). ment and organotypic culture models for predictive toxicity et al 74,75 15. Kubitz, R. . Autoimmune BSEP disease: disease recurrence and pharmacokinetic study. The European Commission is after liver transplantation for progressive familial intrahepatic funding a Body-on-a-Chip project to many collaborated groups cholestasis. Clin.Rev.AllergyImmunol.48, 273–284 (2015). between multiple European academic and industrial partners 16. Ballator, N., Li, N., Fang, F., Boyer, J.L., Christian, W.V. & Hammond, to achieve the goal of developing a comprehensive in vitro C.L. OST alpha-OST beta: a key membrane transporter of bile acids model that allows identification of multiorgan toxicity and and conjugated steroids. Front. Biosci. (Landmark Ed) 14, 2829– pharmacogenetics. 2844 (2009). 17. Zhang, Y., Bachmeier, C. & Miller, D.W. In vitro and in vivo models for assessing drug efflux transporter activity. Adv. Drug Deliv. Rev. 55, Additional Supporting Information may be found in the online version of 31–51 (2003). this article. 18. Sahi, J. Use of in vitro transporter assays to understand hepatic and renal disposition of new drug candidates. Expert Opin. Drug Metab. ACKNOWLEDGMENTS Toxicol. 1, 409–427 (2005). This work was supported by the National Center for Advancing 19. Brandon, E.F., Raap, C.D., Meijerman, I., Beijnen, J.H. & Schellens, J.H. An update on in vitro test methods in human hepatic drug Translational Sciences (NCATS) of the National Institutes of Health biotransformation research: pros and cons. Toxicol. Appl. Pharmacol. under award UH3 TR000504, NIEHS grant 5P30ES070033 and in part 189, 233–246 (2003). under Assistance Agreement No. 83573801 awarded by the U.S. 20. Hilgendorf, C., Ahlin, G., Seithel, A., Artursson, P., Ungell, A.L. & Environmental Protection Agency. It has not been formally reviewed by Karlsson, J. Expression of thirty-six drug transporter genes in human the Environmental Protection Agency. The views expressed in this intestine, liver, kidney, and organotypic cell lines. Drug Metab. document are solely those of the authors and do not necessarily reflect Dispos. 35, 1333–1340 (2007).

476 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt 21. Yamazaki, M. et al. In vitro substrate identification studies for p- 43. Vernetti, L.A. et al. A human liver microphysiology platform for glycoprotein-mediated transport: species difference and investigating physiology, drug safety, and disease models. Exp. Biol. predictability of in vivo results. J. Pharmacol. Exp. Ther. 296, 723– Med. (Maywood) 241, 101–114 (2016). 735 (2001). 44. Allen, J.W. & Bhatia, S.N. Formation of steady-state oxygen gradients 22. Hart, S.N., Li, Y., Nakamoto, K., Subileau, E.A., Steen, D. & Zhong, in vitro: application to liver zonation. Biotechnol. Bioeng. 82, 253– X.B. A comparison of whole genome gene expression profiles of 262 (2003). HepaRG cells and HepG2 cells to primary human hepatocytes and 45. Zhuo, J.L. & Li, X.C. Proximal nephron. Compr. Physiol. 3, 1079–1123 human liver tissues. Drug Metab. Dispos. 38, 988–994 (2010). (2013). 23. Richert, L. et al. Gene expression in human hepatocytes in 46. Liang, M., Ramsey, C.R. & Knox, F.G. The paracellular permeability of suspension after isolation is similar to the liver of origin, is not opossum kidney cells, a proximal tubule cell line. Kidney Int. 56, affected by hepatocyte cold storage and cryopreservation, but is 2304–2308 (1999). strongly changed after hepatocyte plating. Drug Metab. Dispos. 34, 47. Malstrom,€ K., Stange, G. & Murer, H. Identification of proximal 870–879 (2006). tubular transport functions in the established kidney cell line, OK. 24. Tabibian, J.H., Masyuk, A.I., Masyuk, T.V., O’Hara, S.P. & LaRusso, Biochim. Biophys. Acta. 902, 269–277 (1987). N.F. Physiology of cholangiocytes. Compr. Physiol. 3, 541–565 48. Gstraunthaler, G., Pfaller, W. & Kotanko, P. Biochemical (2013). characterization of renal epithelial cell cultures (LLC-PK1 and MDCK). 25. Elferink, M.G. et al. Gene expression analysis of precision-cut human Am. J. Physiol. 248(4 Pt 2), F536–F544 (1985). liver slices indicates stable expression of ADME-Tox related genes. 49. Fouda, A.K., Fauth, C. & Roch-Ramel, F. Transport of organic cations Toxicol. Appl. Pharmacol. 253, 57–69 (2011). by kidney epithelial cell line LLC-PK1. J. Pharmacol. Exp. Ther. 252, 26. Martin, H. et al. Morphological and biochemical integrity of human 286–292 (1990). liver slices in long-term culture: effects of oxygen tension. Cell Biol. 50. Kuteykin-Teplyakov, K., Luna-Tortos, C., Ambroziak, K. & Loscher,€ W. Toxicol. 18, 73–85 (2002). Differences in the expression of endogenous efflux transporters in 27. Schaefer, O. et al. Absolute quantification and differential expression MDR1-transfected versus wildtype cell lines affect P-glycoprotein of drug transporters, cytochrome P450 enzymes, and UDP- mediated drug transport. Br. J. Pharmacol. 160, 1453–1463 (2010). glucuronosyltransferases in cultured primary human hepatocytes. 51. Gartzke, D. & Fricker, G. Establishment of optimized MDCK cell lines Drug Metab. Dispos. 40, 93–103 (2012). for reliable efflux transport studies. J. Pharm. Sci. 103, 1298–1304 28. Swift, B., Pfeifer, N.D. & Brouwer, K.L. Sandwich-cultured (2014). hepatocytes: an in vitro model to evaluate hepatobiliary transporter- 52. Ryan, M.J., Johnson, G., Kirk, J., Fuerstenberg, S.M., Zager, R.A. & based drug interactions and hepatotoxicity. Drug Metab. Rev. 42, Torok-Storb, B. HK-2: an immortalized proximal tubule epithelial cell 446–471 (2010). line from normal adult human kidney. Kidney Int. 45, 48–57 (1994). 29. Keppler, D. Uptake and efflux transporters for conjugates in human 53. Jenkinson, S.E., Chung, G.W., van Loon, E., Bakar, N.S., Dalzell, A.M. hepatocytes. Methods Enzymol. 400, 531–542 (2005). & Brown, C.D. The limitations of renal epithelial cell line HK-2 as a 30. Esch, E.W., Bahinski, A. & Huh, D. Organs-on-chips at the frontiers of model of drug transporter expression and function in the proximal drug discovery. Nat. Rev. Drug Discov. 14, 248–260 (2015). tubule. Pflugers Arch. 464, 601–611 (2012). 31. LeCluyse, E.L., Witek, R.P., Andersen, M.E. & Powers, M.J. 54. Detrisac, C.J., Sens, M.A., Garvin, A.J., Spicer, S.S. & Sens, D.A. Organotypic liver culture models: meeting current challenges in Tissue culture of human kidney epithelial cells of proximal tubule toxicity testing. Crit. Rev. Toxicol. 42, 501–548 (2012). origin. Kidney Int. 25, 383–390 (1984). 32. Khetani, S.R. et al. Use of micropatterned cocultures to detect 55. Brown, C.D. et al. Characterisation of human tubular cell monolayers compounds that cause drug-induced liver injury in humans. Toxicol. as a model of proximal tubular xenobiotic handling. Toxicol. Appl. Sci. 132, 107–117 (2013). Pharmacol. 233, 428–438 (2008). 33. Bale, S.S. et al. In vitro platforms for evaluating liver toxicity. Exp. 56. Humes, H.D. The bioartificial renal tubule: prospects to improve Biol. Med. (Maywood) 239, 1180–1191 (2014). 18 34. Ramsden, D., Tweedie, D.J., Chan, T.S., Taub, M.E. & Li, Y. Bridging supportive care in acute renal failure. Ren. Fail. , 405–408 (1996). in vitro and in vivo metabolism and transport of faldaprevir in human 57. MacKay, S.M., Funke, A.J., Buffington, D.A. & Humes, H.D. Tissue using a novel cocultured human hepatocyte system, HepatoPac. Drug engineering of a bioartificial renal tubule. ASAIO J. 44, 179–183 Metab. Dispos. 42, 394–406 (2014). (1998). 35. Kostadinova, R. et al. A long-term three dimensional liver co- 58. Humes, H.D., MacKay, S.M., Funke, A.J. & Buffington, D.A. Tissue culture system for improved prediction of clinically relevant drug- engineering of a bioartificial renal tubule assist device: in vitro induced hepatotoxicity. Toxicol. Appl. Pharmacol. 268, 1–16 transport and metabolic characteristics. Kidney Int. 55, 2502–2514 (2013). (1999). 36. Messner, S., Agarkova, I., Moritz, W. & Kelm, J.M. Multi-cell type 59. Humes, H.D. et al. Initial clinical results of the bioartificial kidney human liver microtissues for hepatotoxicity testing. Arch. Toxicol. 87, containing human cells in ICU patients with acute renal failure. 209–213 (2013). Kidney Int. 66, 1578–1588 (2004). 37. Zeilinger, K. et al. Scaling down of a clinical three-dimensional 60. Saito, A. et al. Evaluation of bioartificial renal tubule device prepared perfusion multicompartment hollow fiber liver bioreactor developed with lifespan-extended human renal proximal tubular epithelial cells. for extracorporeal liver support to an analytical scale device useful for Nephrol. Dial. Transplant. 27, 3091–3099 (2012). hepatic pharmacological in vitro studies. Tissue Eng. Part C Methods 61. Jang, K.J. & Suh, K.Y. A multi-layer microfluidic device for efficient 17, 549–556 (2011). culture and analysis of renal tubular cells. Lab Chip 10, 36–42 38. Novik, E., Maguire, T.J., Chao, P., Cheng, K.C. & Yarmush, M.L. A (2010). microfluidic hepatic coculture platform for cell-based drug 62. Huh, D., Torisawa, Y.S., Hamilton, G.A., Kim, H.J. & Ingber, D.E. metabolism studies. Biochem. Pharmacol. 79, 1036–1044 (2010). Microengineered physiological biomimicry: organs-on-chips. Lab Chip 39. Vivares, A. et al. Morphological behaviour and metabolic capacity of 12, 2156–2164 (2012). cryopreserved human primary hepatocytes cultivated in a perfused 63. Jang, K.J. et al. Human kidney proximal tubule-on-a-chip for drug multiwell device. Xenobiotica 45, 29–44 (2015). transport and nephrotoxicity assessment. Integr. Biol. (Camb). 5, 40. Domansky, K., Inman, W., Serdy, J., Dash, A., Lim, M.H. & Griffith, 1119–1129 (2013). L.G. Perfused multiwell plate for 3D liver tissue engineering. Lab Chip 64. Jansen, J. et al. A morphological and functional comparison of 10, 51–58 (2010). proximal tubule cell lines established from human urine and kidney 41. Sarkar, U. et al. Metabolite profiling and pharmacokinetic tissue. Exp. Cell. Res. 323, 87–99 (2014). evaluation of hydrocortisone in a perfused three-dimensional 65. Jansen, J. et al. Human proximal tubule epithelial cells cultured on human liver bioreactor. Drug Metab. Dispos. 43, 1091–1099 hollow fibers: living membranes that actively transport organic (2015). cations. Sci. Rep. 5, 16702 (2015). 42. Lee, P.J., Hung, P.J. & Lee, L.P. An artificial liver sinusoid with a 66. Hoppensack, A. et al. A human in vitro model that mimics the microfluidic endothelial-like barrier for primary hepatocyte culture. renal proximal tubule. Tissue Eng. Part C Methods 20, 599–609 Biotechnol. Bioeng. 97, 1340–1346 (2007). (2014).

CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 100 NUMBER 5 | NOVEMBER 2016 477

67. Finesilver, G., Bailly, J., Kahana, M. & Mitrani, E. Kidney derived 72. Li, A.P. Evaluation of adverse drug properties with cryopreserved micro-scaffolds enable HK-2 cells to develop more in-vivo like human hepatocytes and the integrated discrete multiple organ co- properties. Exp. Cell. Res. 322, 71–80 (2014). culture (IdMOC(TM)) system. Toxicol. Res. 31, 137–149 (2015). 68. Adler, M. et al. A quantitative approach to screen for nephrotoxic 73. Wikswo, J.P. The relevance and potential roles of microphysiological compounds in vitro. J. Am. Soc. Nephrol. 27, 1015–1028 (2016). systems in biology and medicine. Exp. Biol. Med. (Maywood) 239, 69. Kelly, E.J. et al. Innovations in preclinical biology: ex vivo engineering 1061–1072 (2014). of a human kidney tissue microperfusion system. Stem Cell Res. Ther. 74. Andersen, M.E. et al. Developing microphysiological systems for use 4 (suppl. 1), S17 (2013). as regulatory tools--challenges and opportunities. ALTEX 31, 364– 70. Kim, S. et al. Pharmacokinetic proﬁle that reduces nephrotoxicity of 367 (2014). gentamicin in a perfused kidney-on-a-chip. Biofabrication 8, 015021 (2016). 75. Sutherland, M.L., Fabre, K.M. & Tagle, D.A. The National Institutes of 71. Sung, J.H., Kam, C. & Shuler, M.L. A microﬂuidic device for a Health Microphysiological Systems Program focuses on a critical pharmacokinetic-pharmacodynamic (PK-PD) model on a chip. Lab challenge in the drug discovery pipeline. Stem Cell Res. Ther. 4 Chip 10, 446–455 (2010). (suppl. 1), I1 (2013).

478 VOLUME 100 NUMBER 5 | NOVEMBER 2016 | www.wileyonlinelibrary/cpt