Validation of a Multiparametric, High-Content-Screening Assay for

Feb 2, 2017 - A multiparametric, live-cell, high-content-screening (HCS) cytotoxicity assay was first demonstrated in 2006 ( Arch. Toxicol. 2006 , 80 ...
1 downloads 0 Views 7MB Size
Article pubs.acs.org/crt

Validation of a Multiparametric, High-Content-Screening Assay for Predictive/Investigative Cytotoxicity: Evidence from Technology Transfer Studies and Literature Review Peter James O’Brien*,†,‡ and Anna Edvardsson†,‡ †

School of Veterinary Medicine, University College Dublin, Stillorgan Road, Belfield, Dublin 4, Ireland Advanced Diagnostic Laboratory, Park West Enterprise Centre, Lavery Avenue, Park West, Dublin 12, Ireland



ABSTRACT: A multiparametric, live-cell, high-content-screening (HCS) cytotoxicity assay was first demonstrated in 2006 (Arch. Toxicol. 2006, 80, 580−604) to be highly concordant with human hepatotoxicity, including idiosyncratic hepatotoxicities and other target organ toxicities in contrast to historical assays. The success of the assay was attributed to its simultaneous measurement of multiple appropriate “cytobiomarkers”: use of human cells with xenometabolic competence for toxicities mediated by metabolites, 72 h exposure to enable expression of slower-acting toxicants, exposure to a wide-range of concentrations from 30- to 100-fold the efficacious concentration, and normalizing the in vitro cytotoxic concentration to an estimate of the in vivo concentration of exposure. An overwhelming volume of evidence has accumulated over the last 10 years to support this approach as necessary in predictive toxicology. Equivalent assays have now been successfully applied in ∼50 studies across a wide variety of toxicants, toxicities, cell types, and disciplines. Review herein of the wider literature on cytotoxicity since the first assay was reported 100 years ago supports the selection of key cytobiomarkers along a final common pathway of cell injury, including cell proliferation, mitochondrial activity, apoptosis, lysosomal mass, oxidative stress, and cell membrane permeability. HCS studies without inclusion of such key cytobiomarkers or without testing to sufficiently high concentration have not been as successful. Furthermore, a subset of the original toxicants has been reanalyzed herein using the original HCS assay and has confirmed their high sensitivities and specificities across locations, HCS technologies, staff, laboratories, and time. A protocol is demonstrated for operational validation of the assay within laboratories to demonstrate proficiency and quality management.

1. INTRODUCTION AND LITERATURE REVIEW 1.1. Background. Cytotoxicity assays have long been used by drug discovery scientists in the pharmaceutical industry to screen for overt toxicity.1 However, historically, they have not be regarded as sensitive predictors of human toxicity and have been found to be far inferior to the predictivity of regulatory, preclinical, animal, and toxicity studies.1 Despite the high predictivity of regulatory animal studies for some toxicities, drug safety has been an important cause of marketplace attrition: over a 25-year period ending in 1999, there was an annual average of 0.7 approved drugs withdrawn and two drugs receiving black box warnings.2 From 1994 to 2006, the average safety attrition was even higher with 2.1 drugs per year and 38 drugs overall withdrawn from the market by the US Food and Drug Agency.3 Of drug candidates entering clinical development from 2006− 2010, nearly one-quarter failed due to safety issues.4 In 2000, Olson et al. reported a retrospective study of the concordance of human toxicity occurring with 150 drug candidates during the clinical phase of drug development with toxicity in preclinical regulatory animal studies.5 Concordance was 71% for both rodent and nonrodent species, 63% for nonrodents alone, and 43% for rodents alone. Toxicity was identified in studies of duration of 1 month or less for 94% of toxicities. Concordance of animal and human toxicity varied depending on the type of © 2017 American Chemical Society

toxicity: 91% if hematologic, 85% if gastrointestinal, 80% if cardiovascular, but only 37% if cutaneous (hypersensitivity), and 55% if hepatic. In addition to the predictivity being low for human hepatotoxicity and cutaneous toxicity, when they occurred, the probability was highest that the drug candidate would be terminated in its clinical development. In the 1980s, progress in understanding of cell biology and culture methods, and increasing interest in animal alternatives for toxicity assessment, led to a large growth in research and publications on cytotoxicity assessment.6 Previously, interest had focused on immune-mediated cytotoxicity and on anticancer therapy. Using the web browser PubMed and the search term “cytotoxic”, nine publications were retrievable by a worldwide web search for up to the year 1940, 24 up to 1955, 1,856 up to 1970, 17,249 up to 1985, 67,898 up to 2000, and 157,069 up to May 21, 2016. In 2015, there were 8934 publications. The growth of publications related to cytotoxicity is illustrated in Figure 1 with the most frequently cited papers identified. After approximately 1990, the implementation of cytotoxicity assays was made practical for screening purposes by the development and automation of multiwell microtiter-plate readers and the Received: October 26, 2016 Published: February 2, 2017 804

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

This validation was based on international consensus guidelines on how to verify performance for precision and accuracy.12 1.2. A Century of Evolution of in Vitro Cytotoxicity Assays and Identification of “Cytobiomarkers”. 1.2.1. Dye Uptake for Viability and Proliferation. Cytotoxicity assays were first developed a hundred years ago. In 1917, Alwin Pappenheimer demonstrated trypan blue was taken up into the nucleus of cells that had been injured by a wide variety of chemical and physical factors.13 The assay was modified to use eosin, propidium iodide, ethidium bromide, crystal violet, and other dyes, which are also excluded from healthy cells.14 Alternatively, or complementarily, toxicity could be identified by failure of uptake of supravital dyes such as neutral red into lysosomes.15 By dual staining with trypan blue and neutral red, both live and dead cells could be positively identified and counted microscopically. The first cytotoxicity assay developed for screening compounds in drug discovery16,17 was based on inhibition of proliferation of transformed cells cultured for 7 days using protein measurement as a biomarker of cell number.18 In 1956, Eagle and Foley first reported that antitumor agents were highly cytotoxic and used a protein assay to screen a library of 180 compounds of varying known antitumor activity.16 Their assay identified 80% of the antitumor agents but found 20% of the agents without antitumor activity to be cytotoxic. At the Upjohn Pharmaceutical Company, this assay was simplified and applied to similar group of compounds giving comparable results and then used to validate three other cytotoxicity biomarkers of cell proliferation: crystal violet exclusion, biomass as measured by carbon oxidation by potassium dichromate, and oxidation of 2,6-dichlorophenol indophenol by mitochondrial succinate dehydrogenase.19−21 Soon thereafter, more specific assays for antiproliferative effects of cancer therapeutic drugs on cell proliferation were developed that measured incorporation of tritiated thymidine or uridine into newly synthesized DNA or RNA of dividing cells.22 Radioisotopic labeling of DNA was replaced a couple decades later with 5-bromo-2-deoxyuridine, which was quantitated by enzyme immunoassay, greatly improving the convenience, cost, and safety of the assay.23 Quantitative 96-well microtiter-plate culture and cytotoxicity assays using spectrophotometric plate readers were introduced at the beginning of the 1980s, markedly facilitating cytotoxicity assaying by increasing rapidity, sensitivity, and precision.24 Crystal violet was first used as an indicator of loss of cell viability. Many of the microtiter-plate assays focused on measurement of cell mass, which increases with cell growth and proliferation but decreases with cytotoxicity, at least if exposure or its effect occurs over the duration of one or more cell cycles. This effect could be quantified by fluorescence, microtiter-based measurements of total DNA,25 total protein measured by sulfo-rhodamine B,26 or ATP content.27 The latter is determined by bioluminescent assays in which firefly luciferin-luciferase is used as an indicator. Most of these assays were largely eclipsed by the introduction of mitochondrial redox assays described below. These early cytotoxicity assays identified key cytobiomarkers of cell viability and cell proliferation that are the foundation for the current, highly effective HCS-based assays. Proliferative activity is still the single-most accurate predictor in in vitro cytotoxicity assays of human hepatotoxicity despite extraordinary developments in technology and understanding of cell biology.8 1.2.2. Leakage of Constituents from Injured Cells. Leakage of cell constituents, such as lactate dehydrogenase (LDH) or potassium ions, was demonstrated early to be indicative of cell

Figure 1. (left) Growth of cytotoxicity assays in scientific literature. (right) Most highly cited cytotoxicity assays.

development of a wide-range of relatively nontoxic, subcellular fluorescent dyes.7 High content analysis (HCA) is an automated technology that combines physiological incubation of cells in microtiter plates with automated liquid handling and multiprobes, epifluorescence or laser-based microscopy, fluorescence image acquisition, and quantitative morphometric analysis of individual cells and their organelles.7 It was introduced into a novel in vitro cytotoxicity assay using transformed human hepatocytes and looking at multiple cellular biomarkers involved in the pathogenesis of toxicity.8 The outcome was compared with that for seven conventional cytotoxicity assays. Testing of 243 drugs and chemicals of well-defined human toxicity demonstrated that the outcome in the HCA assay was concordant with human toxicity with 93% sensitivity and 97% specificity. Concordance with human toxicity was an order of magnitude higher for the HCA assay than any or all of the conventional assays. Additionally, concordance was similarly improved compared to that of regulatory preclinical animal studies.5,9 The relevance of high predictiveness of the HCA cytotoxicity assay8 was its use for more effective prediction of human toxicity and reduction of marketplace attrition of drugs due to adverse drug reactions.1 Use begins in drug discovery for optimization and prioritization of potential lead compounds alongside assays predictive of bioavailability and efficacy.1 By rapid and relatively low cost screening for toxicity potential, it may replace, reduce, and refine preclinical animal studies, substantially reduce financial costs, and accelerate drug development. However, for efficient use, the assay must meet many criteria including reliability, relevance, cost effectiveness, and sufficient throughput.10 If the assay is to be used for regulatory testing, it is also necessary for it to easily include routine quality controls for intraand interlocation variability.11 In this study, we first review the literature supporting the effectiveness of this HCA approach, including historical assays and concepts on which it is strongly rooted and numerous recent confirmatory studies. Additionally, we present new data from a study demonstrating successful technology transfer across HCA platforms, geographies, technologists, and laboratories and a strategy for operational validation and proficiency assessment of an HCA laboratory. The latter study is based on HCA analysis of a subset (67) of the 243 drugs presented in the original study.8 This data was generated on another HCA platform in three separate locations and assessed for six different markers of cytotoxicity (nuclear area (NA), nuclear intensity (NI), mitochondrial membrane potential (MMP), intracellular calcium (IC), and cell number (CN)). This study also includes the assessment of doxorubicin toxicity using an international consensus validation strategy for laboratory assays to demonstrate precision and accuracy. Five plates were assayed on separate days at three locations to show the robustness and reproducibility of the method when transferred between sites and platforms. 805

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology toxicity and more sensitive28 than dye exclusion or cell growth assays. The latter, as measured by total protein or cell counts, were ineffective with acute toxicity because too little proliferation could have occurred during the period of exposure. Trypan blue exclusion was also found to be an insensitive indicator of loss of cell viability, changing much later than LDH release, which in turn was much less sensitive than release of potassium and influx of sodium. This difference was likely attributable to the much smaller size of an ion versus a large enzyme. Release of radioactive chromium 51 ion from prelabeled cells has been one of the most used assays for quantitation of cell-mediated cytotoxicity,29 although LDH release may be more convenient, more precise, and less expensive.30 Release of calmodulin, which is a smallmolecular-weight calcium-binding protein that is eukaryotespecific, was found to be effective for discrimination in cocultures of host cells versus infective bacterial injury, as both contain LDH.31,32 1.2.3. Increase or Loss of Mitochondrial Redox Activity. Although Grady21 first introduced a mitochondrial redox assay for cytotoxicity screening in 1960, it was not widely used for a couple of decades. However, since 1983, many cytotoxicity assays have focused on assessment of mitochondrial electron transport using redox dyes and spectrophotometry or fluorometry with high-throughput microtiter plate readers. Mossmann introduced the first of these microtiter assays33 in a publication that has been cited ∼38,000 times according to a Google search (May 22, 2016), which is more than any other cytotoxicity assay. It demonstrated that 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was reduced to a colored formazan precipitate, especially by NADH generated during cell metabolism. Therefore, a decreased signal could be used as a measure of cytotoxicity, and an increased signal was a measure of either proliferation or of cell activation. However, if these cell responses occurred together, occurrence of one could confound the interpretation of the other response because of their opposite effects on the signal. This fact argues for single cell determinations of activity. Other tetrazolium salts followed and became popular, such as 2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide (XTT),34 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2Htetrazolium (MST),35 and WST-1,36 which yield soluble formazans along with introduction of intermediary electron acceptors such as phenazine methosulfate to facilitate dye reduction.37 The sensitivity of this approach was increased 10-fold, halved the assay time, and reduced costs by introduction of resazurin dye (Alamar Blue) as the indicator.38 Alamar Blue reduction showed a linear relationship and good correlation with neutral red uptake, LDH release, total protein, and cell density.39 In a comparison of sensitivity of the different indicators for detection of cytotoxicity produced by 48 h exposure of neuronal cells to diquat at four different concentrations, decreased cell number and increased protein content per cell were the most sensitive indicators, decreased trypan blue exclusion and increased MTT reduction/cell were least sensitive, and LDH release was intermediate in sensitivity.40 The decrease in total MTT reduction proportional to the loss of cells but increase in MTT reduction per cell at higher concentrations was apparently attributable to adaptive mitochondrial biogenesis. Other viability assays were developed that monitored changes in mitochondrial or membrane activities.40 Rhodamine 123 was the first mitochondrial membrane potential dye to become available and was demonstrated to provide an earlier indication of loss of cell viability than that of the trypan blue exclusion assay.41

1.2.4. Increase in Intracellular Ionized Calcium with Cell Injury. Intracellular ionized calcium was shown to be increased with cytotoxicity using the spectrophotometric dye arsenazo III.42 Relatively nontoxic fluorescent dyes for quantitation of intracellular ionized calcium first became available in 198243 and were used to assess viability.44 As there is a 10,000-fold gradient in free calcium concentration across the plasma membrane, a prolonged rise in intracellular calcium concentration, as opposed to a transient and sharp rise such as may be involved in cell signaling, could indicate impaired cell health. Nonspecific membrane-perturbing agents such as halothane were shown to produce dose-related increases in cytosolic free calcium.45 Monitoring intracellular ionized-calcium concentration has not become very popular in cytotoxicity assays. This is probably because the fluo-4 dye that is most commonly used is not highly sensitive at resting intracellular calcium concentrations and is subject to more artifacts than classical ratiometric monitoring with dyes.45 For example, as shown below, doxorubicin has intrinsic fluorescence at the same wavelength at which fluo-4 is monitored. This obviously interferes with nonratiometric fluo-4 monitoring but, on the other hand, can define the intracellular free concentration of doxorubicin. 1.2.5. Apoptosis. Cells also may undergo a programmed response to marked cell injury in which there is systematic regulated cell involution.46 The cells do not rupture and release their cell contents as occurs with necrosis, and thus, apoptosis will not be recognized in cytotoxicity assays by monitoring release of intracellular constituents. Furthermore, there will be no inflammatory response, although the apoptotic cells are phagocytosed by macrophages. The approximate sequence of events that are readily measurable using cell imaging or flow cytometry is as follows:46,47 caspase activation, cell and nuclear shrinkage, frequently a reduction in mitochondrial membrane potential, rise in intracellular calcium as pH falls, activation of endogenous endonucleases with phosphatidyl serine (PS) translocation from the inner to the outer leaflet of the plasma membrane, and formation of apoptotic bodies and cytoplasmic blebs. Many cytotoxicity assays have been specifically developed for assessment of apoptosis, especially by flow cytometry. The first measured a mild and early increase in cell membrane permeability to a wide range of DNA-staining dyes including classics such as ethidium bromide,48 acridine orange,49 Hoechst 33342,50 propidium iodide,51 7-amino actinomycin D,52 anthracyclines,53 and many of the recently developed highly fluorescent and low-toxicity cyanine dyes such as TOTO,54 TO-PRO-3,55 and mitochondrial DNA-staining picogreen.56 The most cited method57 for detection of apoptosis (4565 citations, Google Scholar, Jun 13, 2016) is a 1995 dual fluorescence assay58 that detects PS by immunoassay in the outer surface membrane using the phospholipid-binding protein annexin V, which has high affinity for PS. Because PS exposure also occurs with cell necrosis, the assay was designed to discriminate between the two types of cell death using a membrane permeability dye. Apoptotic cells excluded propidium iodide,57 whereas necrotic cells were stained by it. Thus, only the apoptotic cells were stained by labeled annexin V and unstained by propidium iodide. Activation of the proteolytic cascade of apoptosis can be assayed directly as activation of the cell-death-specific cysteinyl aspartases, called caspases, in intact cells by immunocytochemical detection of cytoplasmic-to-nuclear translocation of caspase 3.59 However, indirect measures are more simple and practical. Cell shrinkage occurs early in apoptosis60 without loss of cell membrane integrity 806

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology and is easily detectable by flow cytometric measurement of forward-angle and side-angle light-scatter as cells flow in single file across a laser beam.61,62 The forward scatter diminishes in proportion to the cell shrinkage while the side scatter increases in proportion to the increased intracellular complexity during karyorrhexis. Endonuclease activation results in DNA degradation causing chromatin condensation and nuclear shrinkage or pyknosis,63 which is readily measured by automated image analysis of fluorescence-stained nuclei.8,10 This pyknosis has been one of the most concordant of in vitro cytobiomarkers with human hepatotoxicity.8 Kerr,60 who originally proposed the term apoptosis and described it as shrinkage necrosis, noted that the first morphological phase was the formation of apoptotic bodies. These are small, round, cytoplasmic fragments sometimes containing pyknotic nuclear fragments that in liver had been referred to as Councilman bodies. This nuclear pyknosis and karyorrhexis is also readily detected and measured by fluorescent image analysis in lymphocytes.64,65 Only one of three known pathophysiological mechanisms of apoptosis requires mitochondria, although there may be crosstalk between pathways.46,47 All pathways lead to sequential activation of caspases and deoxyribonucleases and the above involutional changes. The two mitochondrial-independent pathways are immune-mediated and activated at the cell surface. In the death-receptor pathway, tumor-necrosis factor or related cytokines bind death receptors to activate initiator caspase 8, which activates executioner caspases 3 and 7, which in turn activate the endonuclease caspase-activated deoxyribonuclease (CAD). In the second nonmitochondrial pathway, pores are formed in the surface membrane by perforin for entry of degradative serine-protease granzymes, both of which are released by cytotoxic T-lymphocytes or natural killer cells.66 1.2.6. Mitochondrial Outer Membrane Permeabilization. The mitochondrial pathway of apoptosis is initiated by mitochondrial outer membrane permeabilization (MOMP).67 This process is highly controlled, primarily through interactions between pro- and antiapoptotic proteins that are related to a transmembrane oncoprotein located in the outer mitochondrial membrane called B-cell lymphoma 2 (Bcl-2), or Bcl-2-like proteins, based on the tumor where it was first discovered.68 Activation of Bcl-2 promotes tumorigenesis by preventing apoptosis. The Bcl-2-related proteins Bcl-2-associated X (BAX) protein and Bcl-2 killer (BAK) are well-studied, pro-apoptotic proteins that oligomerize to form the pores that permeabilize the membrane to large proteins.69 So called BH3-only proteins are also related to Bcl-2 but have homology only for its third domain. They have regulatory interactions with Bcl-2, BAX, or BAK. For example, BAD and Bcl-2 interacting killer (BIK) bind to and inhibit Bcl-2. BH3-interacting domain death agonist (BID), when cleaved by caspase 8 of the death-receptor pathway, binds to and activates BAX and inhibits Bcl-2. BH3-only Bcl-2interacting mediator of cell death (BIM) proteins bind to and inhibit certain Bcl-2. Numerous pharmaceuticals have been recently developed based on regulation of the mitochondrial pathway of apoptosis and are in clinical trials or early use for treatment of a wide range of cancers.68,69 These include BH3 and second mitochondriaderived activator of caspase (SMAC) mimetics and executionercaspase activators that may activate intrinsic apoptosis of tumor cells. Side effects of cancer therapy may be ameliorated by executioner-caspase inhibitors or inhibitors of prostaglandin E2 synthesis such as cyclo-oxygenase type 2 inhibitors or activators

of tumor-associated macrophages such as colony-stimulating factor 1 receptor antagonists. MOMP causes cytochrome c, in the intermembrane mitochondrial space, and an apoptotic protease-activating factor (APAF) to leak out into the cytosol and bind to and activate caspase 9 to form the so-called “apoptosome”, which activates effector caspases 3 and 7. Also released through the mitochondria pore is the protein SMAC/DIABLO, which inhibits an X-chromosome-linked inhibitor-of-apoptosis protein (IAP) that binds to and inhibits caspases 9, 3, and 7.68 1.2.7. Mitochondrial Permeability Transition Pore. Apoptosis typically involves perturbation of the inner mitochondrial membrane prior to MOMP.70 Mitochondrial function is dependent on an electrochemical transmembrane potential across the inner mitochondrial membrane that consists of an electrical potential generated by a voltage gradient and a pH gradient generated by the respiratory chain activity. A large mitochondrial permeability-transition pore (MPTP) in the inner membrane opens when mitochondrial-matrix calcium increases. The pore is formed by integral membrane proteins such as ATP synthase70,71 and allows nonspecific small molecules of up to 1500 Da to cross over. Widespread and sustained opening stops ATP synthesis, causing loss from the mitochondrial matrix of respiratory substrates and nucleotides and cell death. In the mitochondrial-dependent pathway of apoptosis, mitochondrial membrane potential decreases early at the time of the MPTP opening. The most popular and effective fluorescent dyes for monitoring MPTP have been the carbocyanine dyes DiOC672 and JC-173 and, more recently, the rhodamine dyes, including rhodamine123,41 but especially TMRM74 and TMRE.75 These are lipophilic cationic dyes that equilibrate across the inner mitochondrial membrane and accumulate in the mitochondrial matrix in classical Nernstian fashion in inverse proportion to the membrane potential. The latter two dyes produce minimal artifacts such as mitochondrial binding or inhibition of the electron transport chain. MPTP opening causes a change in the signal intensity or in the emission spectra peak of these dyes. Fluorescence intensity increases with hyperpolarization but decreases with depolarization. Use of mitochondrial membrane potential dyes such as TMRM in cytotoxicity assays may suffer from artifacts associated with drugs that may be substrates for the p-glycoprotein that extrudes compounds from cells. For example, this is the case for doxorubicin, which subsequently causes a marked increase in TMRM fluorescence.65 1.2.8. Lysosomotropic Drugs and Phospholipidosis and Vacuolization. Lysosomes have long been recognized as mediating cytotoxicity of certain drugs. Drug-induced phospholipidosis by cationic amphiphilic drugs (CADS) has been recognized since 194876 with now over 50 drugs and hundreds of compounds known to be phospholipidotic, especially antidepressants, antimalarial, antibacterial, antiarrhythmic, and cholesterol-lowering drugs.77,78 CADS have a hydrophobic domain and a hydrophilic amine domain. When they passively diffuse into lysosomes, the latter domain protonates and thereby traps the drug in the organelle, whereas the hydrophobic domain binds with phospholipids, especially in inner leaflet of the lysosomal membrane lipid bilayer, which is enriched in negatively charged phospholipids, especially phosphatidyl 4,5-bisphosphate (PIP2).79 CADS bind to and inhibit lysosomal PIP2, resulting in inhibition of lysosomal interaction with late endosomes of the endocytosis pathway.80 The CADS−PIP2 interaction also inhibits detubulation of cells that develop a surface membrane 807

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

candidate drugs for human toxicity potential with a live cellbased assay, it is necessary to have a profile of biomarkers that will identify activation or a “catch all” of the different pathways of response to cell injury, including cytotoxicity, cytostasis, apoptosis, autophagy, and necrosis.7,9,88 At the same time, it should identify the pathogenetic process by which cytotoxicity is occurring, such as by inhibiting cell or mitochondrial proliferation, inhibiting cellular metabolism or mitochondrial activity, generating oxidative stress, accumulating in lysosomes, and/or permeabilizing cell membranes (see review in ref 7). For sensitivity of detection and for some differentiation of pathogenesis, the profile should identify the prelethal onset of the effect and response rather than merely cell death. The most current, and likely the most effective, assay will utilize the findings of the last century of cytotoxicity assays that were briefly reviewed above. Specifically, summarizing these findings, it should identify effects on cell proliferation, surface membrane permeability, oxidative stress, apoptosis, lysosomal and mitochondrial activity, and intracellular homeostasis such as indicated by ionized calcium concentration, ATP, or NADH. 1.3. Assessment of Predictivity of Cytotoxicity Assays for Human Toxicity. 1.3.1. Acute Human Toxicity Prediction. In vitro cytotoxicity assays have been demonstrated to be highly predictive of acute human toxicity. A seven-year, international, multicenter study90 involving 29 laboratories and 61 different cytotoxicity assays of the kinds described above evaluated the relevance of cytotoxicity testing in numerous cell types to in vivo human toxicity. This was one of the most comprehensive assessments of predictivity of cytotoxicity tests for acute human toxicity. A wide range of 50 chemicals was studied, including wellknown poisons such as potassium cyanide, thallium sulfate, arsenic trioxide, mercuric chloride, and methanol. Common prescription drugs were also studied, such as chloramphenicol, diphenylhydantoin, isoniazid, propranolol, verapamil, as well as over-the-counter drugs such as acetaminophen, aspirin, caffeine, chloroquine, and orphenadrine. Substances of abuse were also included such as amphetamine, diazepam, ethanol, nicotine, and phenobarbital. Common household substances were also included, such as 2,4-dichlorophenoxyacetic acid, hexachlorophene, isopropanol, sodium chloride, and sodium fluoride. The in vitro cytotoxic concentrations were compared with the known, acutely lethal doses in humans and the known doses causing 50% lethality in rodents. Correlation (Pearson’s r) of in vitro cytotoxic concentrations with the lethal blood concentration in humans was high, 81% with mouse LD50 values, but significantly higher, 88%, for a battery of three 24 h human cell line tests with end points of total protein, ATP, morphology, and pH.91 Several other important observations and conclusions can be drawn from the MEIC study. Assays performed on human cell lines gave the best prediction (r). Correlation was approximately 10% lower for other mammalian cell lines, 15% lower for fish lines, and 25% lower for plant and bacteria. This, and the observation that good predictivity was largely independent of cell type, indicates that the mechanism of toxicity was inhibition of a common, fundamental, and vital cell process. The few chemicals that were not predictive in the in vitro assays all acted on specific receptors, such as digoxin, malathion, nicotine, and atropine. Specific comparisons of predictivity of human cell-lines to primary cells were made for hepatocytes and keratinocytes but did not reveal significant differences. 1.3.2. Extension to Chronic and Idiosyncratic Human Toxicity: The Predictive HCA Cytotoxicity Model. HCA is rapidly becoming the tool of choice in studying cell biology

invagination, such as the T-tubules in cardiomyocytes, open canicular system of platelets, and demarcation membrane system of megakaryocytes.79 The inhibition of phospholipid degradation is either direct by the drug interaction or by its inhibition of lysosomal phospholipase. Consequently, phospholipids accumulate in concentric lamellar bodies that are found in multiple tissues, including lung, liver, brain, kidney, cornea, and myocardium. These are visualized by electron microscopy. Lysosomal accumulation of CADS also causes the formation of prominent cytoplasmic vacuoles that may be visualized by transmission light microscopy in cells such as peripheral blood monocytes and pulmonary alveolar macrophages. Use of fluorescent lysosomal biomarkers demonstrates that CADS accumulate in and cause expansion of the lysosomal volume.81,96 Some CADS and other chemicals with a protonating amine may progressively accumulate in the lysosome but not interact with the membrane phospholipid.82 These may act by osmosis to increase lysosomal volume and cause the formation of clear vacuoles82 without phospholipid accumulation. Numerous organic amines such as nicotine, atropine, and procaine (amide) have been recognized since 1962 to cause this vacuolization.82,83 1.2.9. Cytotoxic and Cytostatic Autophagy. Autophagy is the regulated process by which cytoplasmic constituents are delivered to the lysosomes for degradation and recycling.84 Its roles include energy provision during starvation, removal of nonfunctional proteins and organelles, removal of intracellular pathogens or aggregating proteins, and tumor suppression. However, like apoptosis, autophagy may cause programmed cell death and has been referred to as Type II cell death, second to the Type 1 cell death of apoptosis.85 Cytotoxic autophagy occurs with some drugs against some cancers,86,87 for example, the alkylating drug Temozolomide in glioblastoma, cannabinoid in melanoma, vitamin D with radiation in mammary carcinoma, and certain protein kinase inhibitors in metastatic carcinoma. Certain drugs may have a cytostatic autophagic effect in which the autophagy pathway is activated to the point where cancer growth is inhibited but cell death is not produced.86 Autophagy can be detected by an increase in lysosomal mass similarly to that described above for lysosomotropic drugs.7 1.2.10. Oxidative Stress. Oxidative stress has long been implicated as an important mechanism of drug-induced toxicity (reviewed in ref 7) either by the drug or metabolite reacting with electron transport systems and increasing production of reactive oxygen or nitrogen species that damage macromolecules by depleting the intracellular antioxidant glutathione or by being activated to a reactive metabolite that forms adducts with cellular macromolecules. In a study of nine cellular parameters affected by the cytotoxicity of 27 drugs and chemicals,88 production of oxidative stress from a superoxide was found to be involved in the pathogenesis of almost half of the cytotoxicities, including representative examples of glitazones (rosiglitazone), statins (simvastatin), fibrates (fenofibrate), macrolide antibiotics (erythromycin), redox cyclers (menadione, diquat, pentachlorophenol), and inhibitors of mitochondrial oxidative phosphorylation (rotenone, FCCP). Some compounds, well-known to cause glutathione depletion as their mechanism of toxicity89 (acetaminophen, diethylmaleate, furosemide), did not produce detectable superoxide, and their oxidative stress would need to be detected with a dye for oxidized sulphydryls. Fluorescent dyes are available for other reactive oxygen and nitrogen species or for glutathione content to identify other mechanisms of oxidative stress. 1.2.11. Defining the Catch-All Cytotoxicity, Cytobiomarker Profile that is Predictive of Human Toxicity. To screen 808

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology because of its high-throughput capacity for monitoring a wide range of morphometric, functional, and biochemical properties of cells.88 It can be considered to be a simple combination and computerized automation of (1) epifluorescence microscopy within an environmental chamber to maintain viability while monitoring cells, (2) multifluorescence-intensity, microtiterplate reading to collect the data with high content, sensitivity, and speed, and (3) software for image segmentation and analysis to rapidly analyze, reduce, and present the data generated. With the appropriate choice of cell model and use of nontoxic fluorescent dyes, cytotoxicity is easily detected at multiple levels. Up to four different fluorescent signals can be readily measured from the same subcellular point across the cell along with a wide range of morphometric parameters and cell or organelle counts for several thousand cells in each of the 96 wells in less than an hour. An additional four dyes can be used in parallel in second adjacent wells to give twice as many biomarkers. Furthermore, following a live cell assay, cells can be fixed and stained with multiple immune-labeled antibody markers. In 2006, a novel cell-based model using HCA was reported8 that demonstrated an order-of-magnitude increased concordance of in vitro cytotoxicity and human toxicity of 250 marketed drugs compared to that found for conventional cytotoxicity assays. This work confirmed earlier preliminary reports with smaller data sets.92,93 The HCA assay focused on hepatotoxicity, a major cause of market attrition, in particular, low-incidence or idiosyncratic hepatotoxicity. Since then, there have been numerous confirmatory reports of the effectiveness of the HCA approach (see review in ref 88), and it has been widely introduced into safety programs in drug discovery.98−105 There are approximately 50 published studies on the use of HCA in cytotoxicity assessment, usually similar to the original study or substituting biomarkers or cell type. Numerous studies from pharmaceutical companies have confirmed the effectiveness of multiparameter HCA cytotoxicity assays in the drug-screening setting, including noteworthy, comprehensive evaluations out of Pfizer,8,92−96 Astra-Zeneca,97−100 Abbott Laboratories,101 Roche,102 Lundbeck,103,104 Astellas,105 and Novo Nordisk.100 Numerous studies from most of the many vendors of HCA technology,106,107 and from several contract-research organizations8,108,109 that provide HCA cytotoxicity testing service, also confirm the validity of the approach. Finally, numerous studies from academic research programs110−115 and research institutes116 have confirmed the HCA approach as effective for assessment of cytotoxicity. Effective use of the HCA cytotoxicity assay has also now been extended to a wider range of toxic substances, cell types, and diverse disciplines in which toxicity is relevant. Its effectiveness at identifying cytotoxicity has been extended from the classic smallmolecule drug to a wide range of toxic substances, including peptides such as melittin,117,118 nanoparticles,119−122 endotoxic lipopolysaccharides, 123 and drug delivery polymers and carriers124 (see Table 1). Its applicability has been extended to a wide range of cell types (see Tables 1 and 2), including hepatocytes,8,101 intestinal,118,124 renal,121 and respiratory121 epithelium; fibroblasts,119 osteoblasts,112 skeletal,88 and cardiac109,114 myocytes; and neurons,121,125,126 lymphocytes,64,65 macrophages,121,123 and monocytes.124 Its range of applications has been substantially extended from pharmaceutical8 to environmental toxicology,123,127,128 food safety,129,130 cancer diagnosis,131 and in vivo toxicity of anticancer drugs.65,132 In addition to studying the toxicity of pure substances, it has been used to evaluate the toxicity of mixtures128,130 of compounds that

Table 1. Demonstrated Applications for the HCA Cytotoxicity Assay cell types used in

substance toxicity predictions

hepatocytes (HepG2, primary, HepRG) neurons myocytes cardiac myocytes enterocytes renal lymphocytes monocytes

drugs in discovery, development, and on the market chemicals peptides nanoparticles, microparticles organic dust components polymers poisons/toxins environmental substances (e.g., concentrates, condensates) food contaminants

macrophages Zebrafish drug efficacy prediction anticancer drugs resolution of cell injury protection from oxidative stress

elucidation of cellular pathogenesis of substance toxicity specific classes of compounds (e.g., statins, glitazones, NRTI) cardiotoxicants anticancer drugs idiosyncratic hepatotoxicants renotoxicants neurotoxicants genotoxicity

industries and disciplines used in

in vivo translational blood biomarkers of toxicity and pathology

pharma: safety assessment pharma: drug discovery pharma: drug delivery

lysosomal toxicity: phospholipidosis, vacuolization neoplasia: leukemia/lymphoma mitochondrial toxicity: metabolic inhibitors, oxidants NRTI toxicity (mitochondrial DNA depletion) carcinogenesis: micronuclei in culture blood lymphocytes efficacy and toxicity of anthracyclines oxidative stress

environmental toxicity food safety oncology academia: cell biology research

may be more relevant for environmental toxicology and food safety. In the 2006 HCA cytotoxicity assay,8 12-point concentration− response curves for drug-induced cytotoxic changes were determined by three-day exposure of HepG2 hepatoma hepatocytes to progressively increasing (doubling) concentrations of drug. Up to seven drugs were tested per 96-well microtiter plate along with controls. Cytotoxic changes monitored were mitochondrial membrane potential using TMRM, intracellular ionized calcium using fluo-4, surface membrane permeability to small molecules using toto-3, and cell proliferation and nuclear area using Hoechst-33258 nuclear staining. Concentration−response curves were horizontal for nontoxic compounds but had an inflection upward or downward, beyond a background signal-noise level, for cytotoxic compounds. The concentration at which deflection occurred was defined as the cytotoxicity concentration. Interpretation of the significance of this was dependent upon normalizing to the serum concentration (Cmax) at which the drug was efficacious to generate a therapeutic index (cytotoxic concentration/Cmax). Drugs for which this value exceeded 100 were considered safe with 90% confidence. In a head-to-head comparison of the HCA cytotoxicity assay with conventional cytotoxicity assays, such as were run for the 809

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology Table 2. Validation Studies for Multiparametric HCA Cytotoxicity Assaya year

HCA platform

1

1994

2

1999

3 4 5 6 7 8 9 10

2001 2003 2004 2004 2006 2006 2006 2006

11

2007

MMM, Carnegie Mellon MMM, ParkDavis MMM, Pfizer KSR, Cellomics MMM, Harvard KSR, Cellomics KSR, Cellomics KSR, Cellomics KSR, Cellomics ArrayScan, Cellomics KSR, Cellomics

12

2008

13 14 15

2008 2008 2008

16

2008

ArrayScan, Cellomics In Cell, GE KSR, Cellomics MMM, AstraZeneca KSR, Cellomics

17

2009

not stated

18

2010

19

2010

20

2010

Image Xpress, MD ArrayScan, Cellomics In Cell, GE

21 22 23

2010 2012 2012

24

2012

25 26

2012 2012

27

2013

28

2014

29

2014

30 31

tests done

dyes

live/ fixed

first author

ref

3

L, F

mouse fibroblast

proposal

Ca (ratiometric), Ab

Taylor

140

tacrine, melittin

3

L

rat H

demo

Ca, MP, P

Plymale

117

3 34 100 14 2 243 2 8

thiazolidinediones drugs, chemicals drugs drugs, chemicals drugs drugs, chemicals quantum dots phospholipidosis

4 4 3 4 4 4 4 3

L L F L L L L L

rat H HepG2 Hum H HeLa HepG2 Hum H HepG2 Hum H HepG2 Hum H Hum fibroblast MΦ

demo validation validation demo statin toxicity validation demo validation

Ca, MP, NA, P Ca, CN, MP, NA,P D, 2 of 10 Ab Ca, CN, MP, NA, P Ca, CN, MP, NA, P Ca, CN, MP, NA, P CN, P, Ab DNA, L, L lipid

Haskins O’Brien Perlman Xu Diaz O’Brien Zhang Morelli

138 92 141 93 94 8 119 97

24

4

L

HeptG2 Hum H

demo

3

L

HepG2 Hum H

validation

Ca, CN, MP, NA, P, MM, ROS, MDNA CN, MP, NA, P

Davila

51

idiosyncratic hepatotoxicants drugs, chemicals

Abraham

101

2 16 12

quantum dots drugs, chemicals drugs

4 4 3

L L L, F

HepG2, neuron HepG2 Hum H PHH

Ca, CN, MP, NA, P Ca, CN, MP, NA, P MP, NA, Ab

Jan O’Brien Ainscow

120 10 98

344

drugs, chemicals

4

L

PHH

demo demo Ximelagatran toxicity hepatotoxicity

Xu

antiobesity, cancer drugs nanoparticles

4

L

not stated

demo

CN, MP, NA, ROS, GSH CN, NA, MP, L, pH

Garippa

102

4

L

demo

Ca, CN, MP, P

George

122

drugs, chemicals

2

F

Hum respiratory, MΦ Hum neuron

validation

CN, Ab

Mundy

125

polymer

4

F

Rawlinson

124

4 4 4

L L L

pDMAEMA toxicity demo validation validation

Ca, CN, NA, NI, MP, P

lymphotoxicity steatosis hepatotoxicity

Caco-2 enterocyte, monocyte lymphotoxicity HepG2 Hum H HepG2 Hum H

Ca, CN, NA, NI, MP, P CN, MP, ROS, P, lipid Ca, CN, MP, P, ROS

O’Brien Donato Tolosa

132 110 111

4

L

HepG2 Hum H

peptide toxicity

Ca, CNB, NA, MP, P

Walsh

118

4 2

L L

Caco-2 enterocyte HepRG Hum H

Patulin toxicity validation

Ca, CNB, NA, MP, P CN, ROS, NA, NI

Mohan Mennecozzi

129 116

2

37 3 16 1

substance tested for cytotoxicity

cell type

objective

parameters tested

96

95

In Cell, GE ScanR, Olympus ScanR, Olympus; In Cell, GE In Cell, GE

4 22 78

1 93 100

drugs, chemicals

4

L

HepG2 Hum H

validation

CN, NA, MP, L, P

Persson

103

cholestasis

3

L

PHH

validation

Persson

104

3

nanoparticles

4

L

validation

Anguissola

121

3 36

dust particles neurotoxicity

4 4

L L, F

validation validation

CN, MP, NA, P, L CN, MM, 2 Ab

Ramery Wilson

123 126

32

2014

KSR, Cellomics

27

4

L

validation

88

2014

In Cell, GE

144

4

L, F

HepG2, PHH

validation

Garside

99

34

2014

240

hepatotoxicity

4

L

HPSCD H

validation

Sirenko

107

35

2014

Image Express, MD Operatta, PE

329

4

L, F

Hum osteosarcoma

validation

Martin

112

36

2014

In Cell, GE

4

chemical library screening lymphotoxicity

Ca, CN, MP, NA, P, MM, ROS, MDNA, L CN, MP, NA, ROS, L, lipid, Ab CN, NI, C, MP, MM, L, lipid CN, NA, MP, Ab

O’Brien

33

idiosyncratic hepatotoxicants hepatotoxicity

A, B, E, HepG2, N, MΦ, R Hum macrophage SH-SY5Y, PC12, HPSCD Hum N HepG2 Hum H

CN, 2 fluorescent bile analogues CN, MP, NA, NI, P, L

2014 2014

In Cell, GE ArrayScan, Cellomics ArrayScan, Thermo Fisher ArrayScan, Thermo Fisher ArrayScan, Thermo Fisher In Cell, GE In Cell, GE

antimicrobial peptide toxicity mycotoxin drugs, chemicals

4

L

validation

Ca, CN, NA, NI, MP, P

Domingos

37

2015

53

phospholipidosis

2

L

validation

CN, L

Bauch

108

38

2015

L, F

validation

127

3

3

L

bovine R

mixture toxicity

CN, NA, MM, MP, 3 of 6 Ab CN, NA, NI, MM, MP

Shah I

2015

environmental toxicants mycotoxins

4

39

Clarke

130

40

2015

1

cardiotoxicity

3

HPSCD CM

Talbert

109

2015

mitochondrial hepatotoxicity

4

Ponatinib toxicity statin toxicity

CN, ROS, lipid

41

ArrayScan, Thermo Fisher ArrayScan, Thermo Fisher CellInsight, Thermo Fisher ArrayScan, Thermo Fisher ScanR, Olympus

Hum/Can lymphocytes HepG2 Hum H, rat CM HepG2 Hum H

Ca, CN, MM, P, MP, ROS

Tolosa

113

5

18

967

20

L

HepG2 Hum H

810

65

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology Table 2. continued tests done

substance tested for cytotoxicity

Image Xpress, MD ScanR, Olympus

10

ArrayScan, Thermo Fisher ScanR, Olympus CellInsight, Thermo Fisher

year

HCA platform

42

2015

43

2016

44

2016

45 46

2016 2016

dyes

live/ fixed

4

L, F

4

L, F

28

hepatic and cardiac toxicity phospholipidosis, steatosis hepatotoxicity

4

L

28 7

steatosis pollutant mixtures

4 4

L L

8

cell type

objective

HPSCD Hum H/ CM HPSCD Hum H

validation

HepG2/HepRG Hum H HepRG Hum H HepG2

validation

validation

validation validation

first author

ref

Ca, CN, MM, ROS, lipid, Ab CN, MM, MP, ROS, L

parameters tested

Grimm

114

Pradip

100

CN, MM, MP, ROS, GSH CN, MP, ROS, P, lipid CN, NA, NI, MP, MMP, ROS

Saito

105

Tolosa Wilson

115 128

a A = astrocytoma; Ab = fluorescent-tagged antibody; B = blood brain barrier cells; Ca = ionized calcium; Can = canine; CM = cardiomyocytes; D = DNA; E = endothelial; GE = General Electric Healthcare; GSH = glutathione; H = hepatocyte; HPSCD = human pluripotent stem cell-derived; Hum = human; KSR = KineticScan Reader; L= lysosome; MDNA = mitochondrial DNA; MC = Molecular Devices; MM = mitochondrial mass; MMM = multimode light microscope; MP = mitochondrial membrane potential; MΦ = mouse macrophage; N = neuronal; NA = nuclear area; NI = nuclear intensity; P = permeability; PE = PerkinElmer; PHH = primary human hepatocytes; R = renal; ROS = reactive oxygen species.

cytobiomarkers to monitor have been defined by the previous century of work on cytotoxicity assays as reviewed above and include cell proliferation, mitochondrial activity, cell membrane permeability, lysosomal volume, apoptosis, and intracellular calcium. All of these were included as cytobiochemical or cytomorphological biomarkers in the original HCA assay for predictive cytotoxicity (see refs 8 and 88). To assess whether other cytobiomarkers were more effective, several more were investigated. Oxidative stress biomarkers were examined88,96,133 because this a common mechanism of drug toxicity.7 Mitochondrial mass was studied,88,96 as mitochondrial biogenesis is a common response to cell injury.7,40 Lysosomal mass was evaluated,88,96,121,123 because of the involvement of lysosomes in drug-induced phospholipidosis7 and vacuolization,7,88 in apoptosis, autophagy, and nanoparticle121 and microparticle123 toxicity. Nuclear staining intensity64,65 was included as it is altered by competition for binding by permeability dyes or drugs such as doxorubicin that intercalate into DNA. All cytobiomarkers were used in one series of studies88,96 to give up to 10 simultaneous cytobiomarkers using eight different fluorescent dyes and duplicate wells with four different dyes in each duplicate well,88 although little advantage was obtained for most cytotoxicities upon extending beyond the original four dyes. The inclusion of 10 cytobiomarkers did, however, give more information on the toxicity mechanisms.88 Schoonen and colleagues134−136 used manual methods for each of multiple assays and demonstrated, without the use of the automated HCA platform, that multiple assays were more effective than single assays at predicting toxicity. Similar results were obtained by Tomida et al.137 However, such manual and single-signal assays are obviously substantially more labor intensive, time consuming, and resource requiring, as well as less precise than those of an automated multiplex assay. Multiplex cytotoxicity assays using three or four simultaneous signals from single cells had first been developed using intracellular fluorescent probes with nonoverlapping spectra and microscopic imaging117,138 or flow cytometry.139 The latter technology, however, provides much less morphometric and subcellular and usually less kinetic information. An automated, robotic, fluorescence microscopy multiplex assay approach using a platform with a controlled environmental chamber was apparently first proposed for cytotoxicity assessment in 1994 by Taylor,140 and this technology was demonstrated in a tricolour cytotoxicity assay in four-chambered glass slides by Plymale et al. in 1999117 and in a quadprobe assay in eight-chambered coverglasses by Haskins et al. in 2001.139 A few years later, it was

above MEIC study, concordance with human toxicity was assessed.8 All assays had almost 95% specificity, giving few false positives. However, false negatives typically exceeded 80% for the conventional assays but were less than 10% for the HCA approach.8 Frequently, cytotoxicities were found for drugs without known human hepatotoxicity potential. These could be attributed to the cytotoxic effect that they had in other human target organs, such as cardiotoxicity, nephrotoxicity, or neurotoxicity. Thus, the use of hepatocytes identified not only human hepatotoxicity potential but toxicity potential for other organs. There are several novel features of the HCA model that distinguish it from conventional cytotoxicity assays and, collectively, likely explain the basis for its enhanced predictivity. 1.3.3.1. Cellular and Subcellular Fluorescence Imaging. First, single-cell microscopic imaging allows subcellular spatial resolution to the basic level at which pathology and signal generation occurs with cytotoxicity. This includes at the nucleus where apoptosis or cell division can be visualized and measured, at the cytoplasm for oxidative stress, at the plasma membrane for cell integrity, and in organelles such as mitochondria for energy production and lysosomes for autophagy and accumulation of waste products or foreign material. This sensitivity, specificity, resolution, and directness of measurement increases precision and accuracy and reduces artifacts such as could occur from extracellular fluorescence, dead cells, or cells of another type. Such cells can be automatically excluded from the analysis based on specific definable and measurable features. HCA imaging allows the full diversity and asynchronicity of individual cellular changes to be seen, which are summated in conventional assays based on population measures. As previously described, measurement at the entire well level can produce artifacts if, for example, cell proliferation was decreased but cells underwent an adaptive increase in a signal canceling out a decrease in another cell due to its loss of viability. This has been described, for example, in diquat toxicity.40 1.3.3.2. Multiple-Biomarker Catch-All Profiles. After direct subcellular imaging, the next most important feature behind HCA’s effectiveness in cytotoxicity testing is multiparametric monitoring. Up until the introduction of HCA, no more than two fluorescence signals could be monitored simultaneously. With HCA, not only can the intensities of four separate fluorescent signals of intracellular activities be kinetically monitored at the same time, but the subcellular structures from which they come can be morphometrically analyzed. This yields such information as number, shape, and size of cells, nuclei, and organelles such as mitochondria and lysosomes. The effective parameters or 811

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology applied in a novel and optimized cell-based model using the first commercialized, fully automated platform with 96-well microtiter plates.92−94 Full validation using 250 human toxicants with well-described toxicity profiles followed in 2006.8 The number of parameters that can be assessed in cytotoxicity live-cell assays can be extended by use of different multiplex assays in parallel wells and also by subsequent fixation of cells and application of immunostains. A single-cell assay for profiling drug-induced DNA profiling was developed by Perlman et al.141 in which they exposed cells for 1 day to 13 doses of 100 different drugs, then fixed the cells and stained them for DNA and ten immunostains (two at a time) for different structural or regulatory cellular proteins involved in calcium signaling, microtubules, microfilaments, mRNA splicing, regulatory protein kinase, cytokinesis, cell receptor signaling to nuclei, transcription, and tumor suppression. A combination of live cell activities and fixed cell immunostains has been used effectively in several HCA studies of cytotoxicity (see Table 2). However, this approach may be limited for predictive cytotoxicity by not using live cells and the cytobiomarkers of toxicity described above that have been identified over the last century. 1.3.3.3. Living Cells with Xenobiotic Metabolic Competence. Of importance to HCA cytotoxicity assays is their capability for live cell measurements of structural, functional, and biochemical parameters in an environmental chamber in which gaseous composition, temperature, and humidity are finely and precisely regulated to optimal conditions for cell viability. This minimizes or eliminates environmentally induced changes that would inevitably occur and obscure or confound chemicalinduced changes. In fact, unless local well humidity is well controlled, the so-called edge effect8 may occur in which evaporation occurs at the edge, and especially at corner wells, and may have effects confounding interpretation of drug effects. There are a number of limitations to the use of the HCA cytotoxicity assay that relate to choice of cell type used for assessment. Just as described above for conventional cytotoxicity assays if the toxicity was pharmacologically mediated or mediated by a cell-specific receptor, then it typically would not be expressed in cells without those receptors. However, most, although not all, of the toxicities of drugs in the marketplace have a more basic mechanism of toxicity that is common to multiple cell types and independent of pharmacology. Second, many hepatoxicities are idiosyncratic and arise from toxic metabolites, especially from the cytochrome p450 system. Thus, if the chosen cell line does not have appropriate xenobiotic metabolic capacity, it will generate false negatives. Fortunately, during the 3 days of exposure of HepG2 hepatoma hepatocytes to drugs, metabolic competence develops and toxicity expresses itself for most drugs.8,108,136 During day 1 of incubation, only 25% of drugs are toxic, but by the end of day 3, 75% were toxic.8 Side-by-side comparisons of HCA assay sensitivity with 24 versus 72 h of drug exposure found the former to be more sensitive for nuclear counts, nuclear area, and membrane integrity. Westerink et al. demonstrated that metabolic competence developed during this time.136 Primary human hepatocytes and numerous other human hepatocyte lines than HepG2 have been proposed and studied for use in cytotoxicity studies (see Table 2). Some have been transformed to develop their metabolic competence, and some have been derived from human stem cells. However, none have been shown to be significantly superior to HepG2 with respect to predictivity, although with the exception of primary human hepatocytes that apparently have lower predictivity,95 the others may have similar predictivity. On the other hand, they are often

less well-studied and less practical, being less globally available or more expensive. 1.3.3.4. Duration and Kinetics of Monitoring. Although not specific to HCA cytotoxicity assays, it was demonstrated that most toxicities require more than 1 day to be expressed.8 At least 3 days of exposure to preacclimatized cells is needed, and unless specific mitochondrial biomarkers are included in the assay, there are rare classes of compounds such as the nucleoside reverse transcriptase inhibitors and the oxazolidinone antibiotics that inhibit mitochondrial DNA or protein synthesis, respectively, that will need up to a week to clearly express their toxicity. Thus, before the HCA cytotoxicity assay is performed, cells must be not only acclimatized for a day to the microtiter plates and incubator conditions but must also be incubated with the test compounds for 3 full days or more. In the original HCA studies of drug-induced cytotoxicity,8 cells were monitored each half hour for 10 measurements. However, it was rapidly identified that, whereas most toxic changes occurred over such a period, it was only after exposure at a critical threshold concentration and duration of time.8 1.3.3.5. Automation with Rapid Throughput. In addition to the high content of information that is collected, the assay is performed at a relatively high throughput compared to that of the conventional cytotoxicity assays. The advantage of this is in the increased user-friendliness and decreased instrument operator-time required for assaying, which can be of the order of weeks versus hours. Also important, interassay variability is reduced because far more samples may be run at once. This high throughput also facilitates inclusion of increased points for more precise definition of concentration−response curves as well as the inclusion of additional positive and negative controls. Finally, automation gives higher precision because it minimizes variance due to human handling and human error. 1.3.3.6. Full Range Dose−Response Curves. For the nontoxic and cytotoxic concentrations for each cytobiomarker to precisely defined, there must be numerous doses studied over a wide range of concentrations. Typically, doses are serially doubled from a low value and are extended to at least one hundred times the blood concentration that is effective in vivo. 1.3.3.7. Determination of Therapeutic Index. As pointed out by Paracelsus some 500 years ago, most drugs are toxic if they are bioavailable and the dose is sufficiently high.88 Thus, interpretation of the significance of cytotoxicity requires evaluation of the dose at which this occurs compared to the dose at which the drug is used therapeutically. This can be estimated by dividing the cytotoxic concentration in vitro by the efficacious concentration as represented by the maximal serum concentration (Cmax) at the therapeutic dose. This therapeutic index is less than 100 for toxic substances.8

2. MATERIALS AND METHODS 2.1. Geographic Location of HCA Centers. For this study, three instruments at three different locations were used: University College Dublin (site 1), Advanced Diagnostics Laboratory Ltd. (site 2), both located in Dublin, Ireland, and GE Healthcare (site 3) situated in Cardiff, UK. 2.2. Materials: Reagents, Drugs, and Technology. Human hepatocellular carcinoma (HepG2) cells were acquired from GE Healthcare (Cardiff, UK) at passage 99. Fetal bovine serum (FBS; 10108-165), L-glutamine (25030-024), and trypan blue (15250-061) were obtained from Invitrogen (Bio Sciences, Dublin, Ireland). Dulbecco’s modified Eagle’s medium (DMEM; M4526), penicillin− streptomycin mixture (PenStrep, P4333), nonessential amino acids mixture (NEAA, M7145), and poly-D-lysine hydrobromide (PDL, 812

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology molecular weight 30,000−70,000; P7280) were obtained from SigmaAldrich (Wicklow, Ireland). Nunc flat-bottomed 96-well plates with lids (TKT-180-070U) were obtained from Thermo Fisher Scientific (Ireland) and 96-well MatriPlates (0.17 mm bottom thickness; 28-9323-99) were acquired from GE Healthcare. Fluorescent dyes were obtained from Invitrogen, including Hoechst 33342 (H1399), Fluo-4 (F14201), TMRM (T668), and TOTO-3 (T3604). DMSO (D2650), FCCP with ≥99% purity (21857), ionomycin with ≥98% purity (I3909), and Triton X-100 (234729) were acquired from SigmaAldrich. Fluvastatin, pravastatin, and isoproterenol were obtained from Merck Chemicals (UK), and all other drugs were acquired from SigmaAldrich. All drugs and chemicals screened for cytotoxicity potential were of the highest purity possible. 2.3. Methods of Analyses. 2.3.1. Preparation of Cell Plates for the HCA Sublethal Cytotoxicity Assay. HepG2 cells were maintained as previously described8 and seeded in 100 μL of media/well in 96-well plates that had been coated with poly-D-lysine as described previously.8 Cell counts were determined at sites 1 and 3 before dilution to the final seeding concentration, and cell dilution was based on the provided stock concentration at site 2. Seeding densities were adjusted so that there would be sufficient cells for analysis but not overgrowth of cells: for the 5-day validation, 4500 cells/well were seeded in Nunc plates, and for the cytotoxicity assessment, 9000 cells/well were seeded in MatriPlates. 2.3.2. Drug Preparation and Treatment. A subset of drugs (67) were selected from the Pfizer8 study. These drugs were classified into four groups based on their human hepatotoxicity: severely hepatotoxic, moderately hepatotoxic, nontoxic, or toxic to other organs. Positive and negative control compounds were also included. Compound effects were tested over 12 concentrations up to 100 μM or 30 times the maximum total serum concentration (Cmax) of the drug reported to be circulating at the therapeutic dose. Drugs were prepared as concentrated stock solutions and dissolved in water or DMSO. If DMSO was used as solvent, the final concentration was 0.5% (volume/volume) in all wells used to test the drug. A “drug plate” was prepared with 1.5-fold concentrated drug. Each row contained one drug, which was 3-fold serially diluted (in contrast to the Pfizer study in which a 2-fold serial dilution was used), starting with the highest concentration in column 12 and the lowest in column 2. The wells in column 1 contained no drug. Fifty microliters of media was removed from each well in the cell plate before adding the drugs in 100 μL of volume. The cell plate was incubated for 72 h before being assayed. For the 5-day validation, duplicate wells were treated with 5 concentrations of doxorubicin (at 10-fold increasing concentrations from 0.002 to 20 μM) for a period of 24 h. 2.3.3. Incorporation of Fluorescent Probes for the HCA Sublethal Cytotoxicity Assay. A cocktail of dyes (0.8 μM Hoechst 33342, 1 μM Fluo-4, 20 nM TMRM, and 1 μM TOTO-3) was made up in media containing 10% FBS, and 150 μL was added to each well of the plate following the protocol of O’Brien.10 Plates were protected from light and incubated for 1 h at 37 °C before high-content analysis. 2.3.4. Internal Quality Control. Three chemicals, FCCP, ionomycin, and Triton X-100, were used as positive controls as previously described8 and were added in duplicate (5-day validation) or triplicate (cytotoxicity assessment). Negative controls were wells containing drug solvent but without any drug. 2.3.5. HCA Analyzer Specifications and Settings. Plates were analyzed by epifluorescence microscopy using an automated microplatereading analyzer (IN Cell Analyzer 1000, GE Healthcare), which is a different instrument than the Kinetic-Scan HCS Reader (Cellomics, Pittsburgh, PA) used in the Pfizer study. The instrument enables fully automated monitoring of fluorescent intensity at four wavelengths and image acquisition. The imaging system of the IN Cell Analyzer 1000 was controlled via a personal computer (Dell XW6200 Workstation) with dual core processors and 240 GB hard drive with 4 GB of random access memory (RAM). Fluorescence was monitored in live cells 1 h after addition of the four dyes (at site 3, images were acquired 1, 2, and 3 h after dye addition, but only data from the 1 h acquisition were used in this study). The 20× objective (NA 0.45) was used to collect images for all four fluorescence channels with an appropriate dichroic (Quad C86224). Dyes were excited, and their fluorescence was

monitored at excitation and emission wavelengths of (1) 360 and 460 nm for Hoechst 33342 on channel 1, (2) 475 and 535 nm for Fluo-4 on channel 2, (3) 570 and 620 nm for TMRM on channel 3, and (4) 620 and 700 nm for TOTO-3 on channel 4. Exposures were optimized and set as follows: Hoechst 33342 for 200 ms, Fluo-4 for 1000 ms, TMRM for 1000 ms, and TOTO-3 for 500 ms. Four fields-of-view per well were monitored, which took approximately 40−45 min. The IN Cell Analyzer 1000 plate heater was set to 37 °C during image acquisition, but no extra CO2 was fed to the plate. Microplate lids were left on the assay plate to prevent evaporation. Cell counts were made after acquisition with the 20× objective. Ten fields per well were imaged using the 10× objective with the same settings for Hoechst 33342 as above. 2.3.6. HCA Data Capture and Image Analysis. For each plate, the following data were collected: nuclear area and nuclear intensity were determined by the area and fluorescence intensity of Hoechst 33342 in channel 1; intracellular calcium was measured as the intensity of Fluo-4 in an intracellular collar region centered at the nucleus in channel 2; mitochondrial membrane potential was defined as total TMRM fluorescence intensity in inclusion granules of the collar region of the cell for channel 3; and TOTO-3 fluorescence intensity in the nuclear region was measured for channel 4 to assess plasma membrane permeability. Cell number was defined as the number of objects identified based on the Hoechst 33342 fluorescence in the images acquired with the 10× objective. 2.3.7. Assessment of Assay Imprecision, Accuracy, and Predictivity. The imprecision was determined to assess random variation and built-in artifacts in the assay. It can occur for all six measured parameters and at various levels. Hence, the imprecision of each parameter was determined for (1) cell-to-cell within one field, (2) field-to-field within one well, (3) well-to-well within one plate, (4) plate-to-plate within one site, and (5) between sites. All calculations are based on five plates assayed on separate days at each site with two wells per plate and four fields per well (three fields per well at site 3). The wells were either untreated or treated (2 μM doxorubicin) and had been incubated for 24 h (sites 1 and 2) or 72 h (site 3). The mean, standard deviation (SD), and percent coefficient of variation (%CV) from each well (n = the number of cells in each well) was calculated from the “cell measures” data, and the %CV from each well over 5 days was then averaged to obtain cell-to-cell variation. In a similar manner, the “summary by field” data was used to calculate the average %CV from both wells in a plate for the field-to-field variation (n = 3, 4, or 10 fields per well). For well-to-well variation, “summary by well” data was used to calculate the average %CV on each plate (n = 2 wells per plate). The well-to-well data was also used to determine the plate-to-plate variation (n = 5 plates) by calculating the average %CV for each parameter on all plates. Site-to-site variance was estimated by taking the cell-to-cell averaged %CV from each site and calculating the mean, SD, and %CV for each parameter. This was also calculated for the field-to-field, well-to-well, and plate-to-plate data, and an average %CV was calculated for each parameter based on the four levels of imprecision. Statistical analysis was carried out using the program GraphPad Prism v5.02 (2008). Student’s t test was used to compare two groups, and multiple groups were compared with 1-way ANOVA followed by Bonferroni’s multiple comparison test. P < 0.05 was selected as the criteria for statistical significance. No individual cells, fields, or wells were excluded, but up to one plate from each site was excluded if the value was more than ±3 SD from the mean of the others, resulting in a lower n (n = 4) for these parameters. Sensitivity and specificity were calculated using the formula previously described.8 Parameters were normally measured with a 20× objective in many cells per field in each of four fields. However, cell numbers were measured as a single number per field. Thus, with only a single number for each of the four fields, there was high imprecision. Accordingly, both the number of fields and the number of cells assessed were increased to determine cell number with less imprecision. This was accomplished by using the 10× objective and counting cells in 10 fields. For this approach to be validated, imprecision was compared for both approaches. 2.3.8. Definition of Positive and Negative Test Responses. To be comparable with the results in the Pfizer study, the cytotoxicity assessment results in this study are also reported in terms of (1) the degree of positivity, (2) the ratio (TI) of the lowest cytotoxic concentration to the 813

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

Figure 2. Multifluorescence images of drug-induced changes in cytobiomarkers of toxicity. Nuclei are stained blue. Mitochondria membrane potential is shown in red. Intracellular calcium is represented by green, and membrane permeability is shown in magenta. Images were obtained with a 20× objective. (left column) (A) Healthy untreated cells with the nuclei stained in blue and mitochondria membrane potential in red. (B) HepG2 cells treated with 100 μM chloroquine for 72 h. Increased intracellular calcium levels are shown in green, and increased membrane permeability is shown in magenta. Chloroquine reduces the nuclear area, mitochondrial membrane potential, and cell number and increases nuclear intensity, intracellular calcium, and plasma membrane permeability. (C) No cytotoxic changes are observed after 72 h treatment with 100 μM of the nontoxic drug isoproterenol. (middle column) Representative images following 72 h of doxorubicin treatment at (A) 0, (B) 0.14, and (C) 11 μM concentrations. (B) The hormetic effects of increased nuclear area and mitochondrial membrane potential at low doxorubicin concentrations are visualized. (C) Decreased nuclear intensity is observed at the higher concentration of doxorubicin. Also visible is the stepwise reduction in cell numbers and increased intracellular calcium compared with untreated cells (A). (right column) Untreated cells from the same plate as the drug-treated cells are shown for each pair on the left. Treatment of HepG2 cells for 72 h with (A) 0.05% detergent Triton X-100, which ruptures cell membranes (mitochondrial and calcium stains are not shown), (B) 100 μM herbicide diquat, (C) 200 μM of the insulin sensitizer drug troglitazone, (D) 8,000 μM of the nonsteroidal anti-inflammatory drug ibuprofen, and (E) 3.8 μM of the antiparasitic and antirheumatic drug quinacrine. Blue staining is of nucleic acids with Hoechst 33342. Green staining is of ionized calcium by Fluo-4. Cytoplasmic red staining is of mitochondrial membrane potential by TMRM. Orange-to-red nuclear staining is of nucleic acids after membrane permeabilization due to cytotoxicity by Toto-3. Cytotoxic concentrations/therapeutic indices were 0.05%/(not relevant), 67 μM/15 for troglitazone, 10 μM/(not relevant) for diquat, 500 μM/2 for ibuprofen, and 1 μM/1 for quinacrine. total maximal drug concentration (Cmax) in serum that is associated with efficacy (where this information is available), and (3) the concentrations at which clear effects are first observed with each parameter. The degree of positive/negative results was based on a number of criteria. A strong positive result was assigned when at least two parameters were affected, and this occurred at lower concentrations than the highest concentration tested. A positive response was assigned when only one or two parameters were affected, and one of these was at the highest concentration tested. A negative result was assigned when one or no parameters were affected, and a high TI negative result was assigned when the TI was higher (>100) than the suggested safety margin.8 A >100 μM negative response was assigned if cytotoxicity was observed at >100 μM, and 30× Cmax was assigned if no cytotoxic response was observed and if the drug was not tested high enough to elicit a clear cytotoxic response.

in Figures 2 and 3, demonstrating the range of morphological and functional effects by different chemicals and drugs at different magnifications and on different cell lines. The pattern of change across the dose−response curve may indicate mechanisms of toxicity. In Figure 2, healthy untreated control HepG2 cells (Figure 2A) exhibit normal NA, NI, and MMP. Chloroquine treatment for 72 h (Figure 2B, left column) induces a reduction of NA, MMP, and CN and increases NI, IC, and PMP. No cytotoxicity is observed in cells treated with the nontoxic drug isoproterenol (Figure 2C, left column). In the right column, there are pairs of photomicrographs demonstrating the effects of treatment with five cytotoxic chemicals and drugs. Untreated cells are on the left for each pair, and on the right, cells treated with toxic doses are shown for: (i) the detergent Triton X-100 (0.05%), which ruptures cell membranes, (ii) the herbicide diquat (100 μM), (iii) the insulin sensitizer drug troglitazone (200 μM), which was withdrawn from the market due to the occurrence of lethal hepatotoxicity,

3. RESULTS 3.1. Representative Fluorescence Images of Cytotoxic Drugs. Representative images of cytotoxic changes are illustrated 814

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

Figure 3. (left column) Multifluorescence images (40×) of drug-induced cytotoxicity in human hepatocytes. Untreated cells from the same plate as the drug-treated cells are shown in the column on the left. Treatment of HepG2 cells for 72 h with (A) 100 μM statin fluvastatin, (B) 20 μM anthracycline doxorubicin, and (C) 100 μM of calcium-antagonist antihypertensive mibefradil. Blue staining is of nucleic acids with Hoechst 33342. Green staining is of ionized calcium by Fluo-4. Cytoplasmic red staining is of mitochondrial membrane potential by TMRM. Orange-to-red nuclear staining is of nucleic acids after membrane permeabilization due to cytotoxicity by Toto-3. Images were collected with a 40× objective. Cytotoxic concentrations/therapeutic indices were 0.5 μM/1.5 for fluvastatin, 0.1 μM/0.3 for doxorubicin, and 0.14 μM/0.4 for mibefradil. (right column) Multifluorescence images (10×) of drug-induced cytotoxicity in human lymphocytes. Untreated cells from the same plate as the drug-treated cells are shown in the column on the left. (A) Jurkat T-lymphocytes were treated for 72 h with 0.2 μM anthracycline mitozantrone. Images were collected with a 10× objective. A small number of cells in (A) are magnified in (B).

the hepatocytes treated with doxorubicin of the same class of drugs. In (B), there are magnifications of both the untreated and treated cells, clearly demonstrating the qualitative and dyestaining changes associated with toxicity. 3.2. HCA Assay Imprecision. Table 3 describes the site-tosite, plate-to plate, well-to-well, field-to field, and cell-to-cell variance of six cytotoxicity markers assayed over 5 days at different locations. Data for both untreated cells and cells treated with 2 μM doxorubicin for 24 h are tabulated. Imprecision for the measured parameters was consistent between sites with the wellto-well and cell-to-cell data of IC being the only parameter with statistically higher variance at site 2. Figure 4 contains representative graphs of imprecision depending on source and parameter. For the field-to-field data, CN is the parameter with the highest variance (Figure 4A and B), which is significantly higher than all other parameters in both untreated and treated cells. In untreated cells, IC was the least variable parameter. Its variance increased 5-fold in treated cells but remained the second most precise parameter. A 2.5-fold increase in variance changes PMP from being the second least variable parameter in untreated cells to being the second most variable in treated cells. No significant changes were observed between untreated and treated cells for the NI, MMP, and CN parameters. The greatest precision was typically observed between wells within one plate as illustrated in Figure 4C. The variances for field-to-field and plate-to-plate were similar for the different parameters measured, except for CN, where the field-to-field imprecision was up to 4× higher. The site-to-site imprecision was the second largest source of variance. The largest source of imprecision was the cell-to-cell measurements within one well. The variance was around 2−10× higher depending on which

(iv) the commonly used nonsteroidal anti-inflammatory drug ibuprofen (8 mM), and (v) the antiparasitic and antirheumatic drug quinacrine (3.8 μM). The detergent causes nuclear fluorescence to change from blue to red as the membrane permeability is increased and TOTO-3 inserts into and displaces Hoechst from the DNA. There is no change in cell number (as indicated by the nuclear stain). Diquat is shown to also cause loss of mitochondrial activity, a marked increase in intracellular calcium, and a marked decrease in cell number. Troglitazone and quinacrine show the same cytoplasmic changes as for diquat, but not the nuclear fluorescence change nor an obvious decrease in cell number, and ibuprofen shows all four color changes as well as the decrease in cell number. In Figure 3, the left column shows pairs of photomicrographs demonstrating cytotoxicity of three common drugs in HepG2 cells. Untreated controls are on the left for each pair, and cytotoxic effects are shown on the right for: (i) the cholesterollowering statin fluvastatin (100 μM), (ii) the antineoplastic anthracycline drug doxorubicin (20 μM), and (iii) the antihypertensive calcium-antagonist drug mibefradil (100 μM). For each toxic treatment, the number of cells is decreased with those remaining having markedly decreased mitochondrial activity (red stain), increased intracellular calcium (green stain), and loss of Hoechst (blue) staining with replacement by orange staining of the cell permeability dye. In Figure 3, the right column demonstrates that the appearance of cytotoxicity in lymphocytes is similar to that of cytotoxicity in hepatocytes shown in the previous figures. In (A), untreated Jurkat T lymphocytes are shown on the left with mitoxantrone-treated cells on the right. This antineoplastic anthracycline drug causes similar effects in the lymphocytes as in 815

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology Table 3. Five Day Imprecision Assessment of Six Cytotoxicity Markers within and across Sites parameter NA

NI

MMP

IC

PMP

CN

CN†

NA NI MMP IC PMP CN CN

b

location site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 3 site 1 site 2 site 1 site 2 site 1 site 2 site 1 site 2 site 1 site 2 site 1 site 2 site 1 site 2

site-to-site %CV ± SD

25.1 ± 22.5

25.4 ± 18.2

17.2 ± 11.0

45.1 ± 37.2

19.5 ± 3.6

17.6 ± 13.1

35.9 ± 24.7

28.9 ± 36.3 37.1 ± 44.7 11.9 ± 14.3 24.5 ± 16.6 13.3 ± 10.3 39.3 ± 36.2 24.3 ± 13.1

plate-to-plate %CV

well-to-well %CV ± SD

Imprecision of Untreated Cells 2.3 2.6 ± 2.0 3.0 2.9 ± 3.7 6.2 4.3 ± 4.3 8.4 4.3 ± 3.5 10.7 2.4 ± 2.2 3.4 4.0 ± 2.9 12.0 5.3 ± 4.8 12.1 5.7 ± 4.4 14.1 9.0 ± 8.0 2.7 1.0 ± 1.4 2.4 4.7 ± 0.6c 7.0 1.3 ± 0.2 7.1 3.8 ± 7.1 6.2 5.4 ± 4.0 9.4 3.7 ± 2.3 5.3 12.3 ± 7.7 10.5 14.3 ± 7.6 8.6 15.8 ± 15.2 2.1 7.1 ± 4.1 9.5 7.6 ± 7.8 10.0 13.0 ± 6.3 Imprecision of Treated Cells 1.7 4.4 ± 3.3 6.5 4.5 ± 2.4 2.9 2.7 ± 2.4 17.6 2.7 ± 1.5 16.5 7.1 ± 5.6 10.2 7.5 ± 4.5 8.3 2.3 ± 2.5 4.8 4.0 ± 3.5 11.2 5.4 ± 5.2 11.4 7.7 ± 4.0 14.8 25.5 ± 6.3 4.0 18.8 ± 13.6 13.2 5.7 ± 4.6 7.4 7.1 ± 6.4

field-to-field %CV ± SD

cell-to-cell %CV ± SD

8.5 ± 4.3 6.9 ± 1.5 7.0 ± 2.1 8.6 ± 4.9 7.3 ± 2.0 9.0 ± 2.7 12.5 ± 4.9 9.3 ± 3.3 14.4 ± 2.1 2.2 ± 0.7 3.0 ± 0.6 2.2 ± 1.2 7.1 ± 2.8 5.2 ± 1.6 6.3 ± 3.8 46.5 ± 14.4 44.0 ± 17.1 39.8 ± 13.6 28.0 ± 5.1 34.5 ± 7.4 28.8 ± 5.9

42.0 ± 0.8 39.2 ± 1.6 38.6 ± 2.2 40.6 ± 6.1 38.2 ± 3.5 31.6 ± 6.0 61.1 ± 2.8 53.0 ± 2.4 56.1 ± 2.3 8.5 ± 1.2 10.2 ± 4.5c 8.8 ± 0.9 37.4 ± 7.6 36.1 ± 4.9 26.9 ± 14.4 NA NA NA NA NA NA

16.2 ± 8.1 12.3 ± 3.2 7.9 ± 1.7 10.1 ± 3.2 16.3 ± 6.2 15.2 ± 1.8 12.6 ± 7.6 13.2 ± 6.0 13.8 ± 7.4 17.8 ± 4.1 34.2 ± 10.1 42.7 ± 7.1 26.1 ± 5.2 33.7 ± 6.4

58.9 ± 2.1 49.6 ± 1.6 30.4 ± 3.8 46.1 ± 4.4 65.7 ± 4.3 61.2 ± 2.3 45.0 ± 27.0 59.5 ± 17.4 68.0 ± 5.7 77.2 ± 6.8 NA NA NA NA

a Cell-to-cell, field-to field, well-to-well, plate-to-plate, and site-to-site imprecision of the six parameters (NA, nuclear area; NI, nuclear intensity; MMP, mitochondrial membrane potential; IC, intracellular calcium; PMP, plasma membrane permeability; and CN, cell number) measured in the HCA cytotoxicity assay. Presented imprecision is the averaged %CV (±SD) from five plates assayed on separate days at different locations. Calculations for all assay parameters are based on two wells per plate and four fields per well (three fields per well at site 3) acquired with a 20× objective. Imprecision is shown for both untreated and treated (2 μM doxorubicin) cells. “Cell measures” data (individual cell data) was used to calculate the mean, SD, and %CV from each well (n = the number of cells in each well). The %CV from each well over five days was then averaged to represent cell-to-cell variation. “Summary by field” data (mean data per field) was used to calculate the mean, SD, and %CV from two wells on each plate (n = 3, 4, or 10 fields per well). The average %CV over five days was calculated to represent field-to-field variation. “Summary by well” data (mean data per well) was used to calculate the mean, SD, and %CV on each plate (n = 2 wells per plate). The average %CV across five days was calculated to represent well-to-well variation. The well-to-well data was also used to determine the plate-to-plate variation (n = 5 plates). The well-towell mean for each parameter was used to calculate the mean, SD, and %CV. The average %CV represents plate-to-plate variation. Site-to-site variance was estimated by calculating the mean, SD, and %CV using the %CV values from the respective site for each of the other four sources of imprecision. The average %CV of these four values represents the site-to-site imprecision. The well-to-well, field-to-field, and cell-to-cell data were statistically analyzed for each parameter of the treated (1-way ANOVA) and untreated (t test) cells to assess eventual differences between sites. For plate-to-plate analysis data, all parameters were grouped for each site and analyzed with paired 1-way ANOVA or t tests as above. bTen fields per well data (10× objective) were also calculated for CN. cP < 0.05 is considered statistically significant

but %CV was significantly decreased by 30% when comparing field-to-field data. 3.3. Reproducibility of Dose−Response Curves. Figure 5 shows 6- and 12-point dose−response curves of doxorubicin. In the 6-point curves, HepG2 cells were incubated with 10-fold dilutions of doxorubicin (0−20 μM) for 24 h. The experiment was repeated five times at each site. In the 12-point dose− response curves, which were performed once at each site, cells

source and parameter with which it was compared. The exception to this trend was IC in untreated cells where the cell-to-cell variance was unusually low and where the site-to-site imprecision was the worst. The results presented in Figure 4D show the effect of field number per well on CN for three sources of imprecision. When the number of fields was increased from 4 to 10, there was no improved precision for plate-to-plate or well-to-well measures, 816

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

points in the graphs at which parameter values either increase or decrease for the various parameters are obvious. A rapid scan of these curves therefore rapidly identifies the concentration of drug at which cytotoxicity occurs and identifies which parameters are affected. A rapid scan indicates that cell number, a surrogate for cell proliferation over the 72 h of exposure, is ubiquitously affected by cytotoxicity. Sometimes this is the only parameter affected, such as for stavudine and pravastatin; outliers are rapidly identified, such as doxorubicin, which produces three remarkable artifacts. It is the only drug that (a) displaces the DNA dye due to its anticancer mechanism of action, (b) dramatically increases the apparent Ca concentration due to the drug’s intrinsic fluorescence at the same wavelength as the calcium probe, and (c) dramatically increases the apparent mitochondrial membrane potential due to the drugs inhibition of the p-glycoprotein that extrudes the mitochondrial dye. Rapid scanning of the dose−response curves in Figure 6 also identifies, for example, drugs that cause an early rise in intracellular calcium, such as chloroquine, quinacrine, troglitazone, and FCCP. Biphasic effects, possibly attributed to hormesis, are rapidly recognized as well, such as for chloroquine and quinacrine. 3.5. Comparative Cross-Site Cytotoxicity Assessment. From the Pfizer study,8 67 compounds from different classes were selected for a comparative study. Doxorubicin was tested twice at separate locations and is therefore counted twice. Two new drugs (4-chloro-m-cresol and troglitazone) were also included in the present study with 70 assessments in total. Comparative results from these two studies are shown in Table 4. In both studies, the toxic drugs pravastatin and astemizole were assessed as false negatives due to high TI ratios, whereas the nontoxic drug picotamide was found to be toxic. No cytotoxicity was noted for four of the 59 compounds where a positive result was expected. Cytotoxicity was observed for fenfluramine in the present study but at higher concentration compared with that in the previous study, resulting in a TI ratio above 100 μM. No explanation was found for why no cytotoxicity was detected in the present study for ketorolac, a severely hepatotoxic drug. In contrast, a positive result was observed for ryanodine for which a false negative result was noted in the Pfizer study.8 A positive result was noted for the negative control folate in the present study. However, this was at a concentration higher than the 100 μM set as the upper limit for the negative controls, and it was therefore assigned a negative result. Determination of positive/ negative results for valproate and hydralazine are entirely based on CN data where normal cell numbers were observed. No cells were found with the 20× objective at concentrations higher than 0.9 and 0.07 μM of the respective drugs. However, by studying data from different time points after dye addition, cells treated with valproate could be detected at all concentrations 3 h after dye addition (data not shown). The sensitivities of the assay for the severely and moderately hepatotoxic drugs, the drugs toxic to other organs, and the positive controls were 95, 95, 83, and 100%, respectively (Table 4). The sensitivities for the same compounds in the Pfizer study8 are 100, 96, 92, and 75% for the respective groups. The specificity for the nontoxic drugs was 86% in both studies and 100 and 75% for the negative controls in this and the Pfizer studies,8 respectively. The overall assay sensitivity and specificity for the comparable 67 compounds were 93 and 89% in the present study and 97 and 89% in the Pfizer study, respectively.8 For all 243 drugs tested in the previous study, the overall sensitivity was 93% and specificity was 98%.

Figure 4. Imprecision in the HCA assay due to the cytobiomarker tested and HCA laboratory. Representative data (field-to-field) showing variance in a number of parameters in (A) untreated HepG2 cells and in (B) cells treated with 2 μM doxorubicin for 24 h. Statistically significant differences (p < 0.05) in imprecision due to treatment are indicated by an asterisk. (C) Imprecision at each of the five levels for the MMP data. (D) Cell number variance for data collected at two magnifications. With the 20× objective, only four fields were assessed per well for parameters. However, cell number is determined per field rather than per cell, as for the other parameters. To decrease imprecision, cell numbers per field and numbers of fields assessed were increased for cell counting by using a 10× objective magnification and counting 10 fields. The two strategies were compared for imprecision. Data presented are the averaged (mean ± SEM) results over five days from three (untreated cells) or two (treated cells) different sites. Statistically significant differences (p < 0.05) in imprecision for the different strategies for counting cells are indicated by an asterisk.

were incubated with 3-fold dilutions of doxorubicin (0−100 μM) for 72 h. For comparison, 12-point curves generated in the Pfizer study8 are also included, which assessed 2-fold dilutions of doxorubicin. The 6-point dose−response curves were reproducible on different days and sites. The lowest concentration of drug in which a cytotoxic effect was noted was identical for four of the six parameters studied and was one point earlier for NA and CN cytotoxicity. The 12-point dose−response curves at site 1 and 3 were reproducible with cytotoxicity detected at identical or the most similar concentrations. In the 6-point curves, only a few cells could be observed at 20 μM after 24 h of doxorubicin treatment. Similar results are seen in the 12-point curves after 72 h where no cells are found above 33 μM. Clear hormetic responses for NA and MMP can be seen in the 12-point curves with 3-fold dilutions assessed at sites 1 and 3. 3.4. Representative Dose−Response Curves for Cytotoxic Drugs. Representative dose−response curves of drug- or chemical-induced cytotoxic changes are illustrated in Figure 6, demonstrating the range of patterns of change, including the variation in sequence of parameter changes. Figure 6 shows the 12-point dose−response curve for ibuprofen-induced cytotoxicity. Values for six parameters are plotted vertically against the concentration cells were exposed to for 72 h. The cell count decreased over the last three concentrations, whereas the other five parameters changed only at the last concentration. The 12-point dose−response curves are shown in Figure 6 for 21 drugs with varying degrees and types of toxicity. Inflection 817

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

Figure 5. Dose−response curves of doxorubicin showing reproducibility of the six cytobiomarkers of toxicity. Variation in doxorubicin-induced changes in six different HCA parameters are shown: nuclear area (NA) as a marker of apoptosis; nuclear staining intensity (NI), mitochondrial membrane potential (MMP), intracellular ionized calcium (IC), plasma membrane permeability (PMP), and cell number as a marker of cell proliferation. HepG2 cells were incubated with 10-fold drug dilutions (0−20 μM) and incubated for 24 h before being assayed (6-point graphs, left). The experiment was repeated over five days at each site, and the data presented are the averaged daily mean ± SEM 12-point drug treatment with 3-fold dilutions (0−100 μM) and 72 h incubation performed once at respective site as shown on the right (12-point graphs). Curves from the original Pfizer study (O’Brien et al., 2006) are also included (far right), which involved 2-fold dilutions of doxorubicin and 72 h incubation (no data is available for concentrations higher than 4 μM (Pfizer study) or 33 μM (sites 1 and 3) because no cells could be detected at these concentrations). Data presented for sites 1 to 3 are averaged field data ± SEM, and data from the Pfizer study are average cell data ± SEM. The 0 μM drug treatments are set to 0.2 nM (6-point curves) and 1 nM (12-point curves) to allow them to be plotted on a logarithmic scale.

first signal for 16% of the drugs. However, it occurred alone as the first signal for only one drug, Amiodarone. As in the Pfizer study, IC and PMP were the least sensitive parameters being the first signals for 6 and 10% of the tested compounds, respectively. This is, however, much lower than the 20 and 17% noted for the parameter and corresponding compounds, respectively, in the Pfizer study. To determine the correlation between the cytotoxicity assessment of the Pfizer and present studies, the lowest values where cytotoxicity was noted were plotted against each other on

As in the Pfizer study, CN was the first parameter to be affected, either alone or in a group of several other parameters. It was the first affected parameter for 71% of the tested compounds in this study, which is higher than the 64% noted for the same parameter in corresponding compounds in the Pfizer study. The second most sensitive parameter was NA, which was the first signal for 23% (31% in the Pfizer study) of the compounds followed by MMP, which was the first signal for 18% (16% in the Pfizer study) of the tested drugs. NI was not included as a parameter in the Pfizer study, but in the present study, it was the 818

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

Figure 6. Variation in patterns of the dose−response curves for six cytobiomarkers of toxicity of 21 drugs. As in Figure 5, the top six rows show dose− response curves for drug-induced toxicity for NA, NI, MMP, IC, PMP, and CN for seven drugs or chemicals, ryanodine (intracellular calcium channel agonist), 4-Cl-m-cresol (toxic chemical), folate (vitamin), chloroquine (antiparasitic), isoproterenol (β-adrenergic), novobiocin (antibiotic), and caffeine (stimulant). Similarly, the next six rows show the effects of chloropromazine (antipsychotic), diquat (herbicide), doxorubicin (antineoplastic), FCCP (mitochondrial uncoupler), ibuprofen (anti-inflammatory), niacin (vitamin), and pravastatin (statin). The last six rows show the effects of quinacrine (antiparasitic), stavudine (anti-AIDS), troglitazone (withdrawn due to hepatotoxicity), ethanol (alcohol), lactose (milk sugar), sodium hydroxide (base), and sodium chloride (table salt). At a glance, cell proliferation is seen to be by far the most sensitive biomarker of toxicity.

logarithmic axes (Figure 7). Samples with negative results in both studies were assigned a concentration of 100 μM, which was the highest tested concentration. A 0.1 μM concentration was assigned to both doxorubicin assessments in this study because

this was the lowest concentration tested in the Pfizer study, and a cytotoxic response was already found at this concentration. Correlation was determined by linear regression using leastsquares fit. 819

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Str+/Str+

Str+/Str+

Str+/Str+

Str+/Str+

Str+/Str+

Str+/Str+

Str+/Str+

Str+/Str+

+/+

amodiaquine

cyclosporin A

danazol

dantrolene

diclofenac

disulfiram

etoposide

imipramine

isoniazide

Str+/Str+

methyldopa

820

Str+/+

Str+/+

lovastatin

+/Str+

hydralazine

Str+/+

Str+/Str+

furosemide

leflunomide

Str+/Str+

furazolidone

ibuprofen

Str+/+

cyclophosphamide

erythromycin

+/+

chloropromazine

Str+/Str+

Str+/Str+

chloramphenicol

Str+/Str+

Str+/Str+

azathioprine

doxorubicin

Str+/+

sensitivity

doxorubicin

+/Str+

95%/100%

valporate

+/Str+

Str+/Str+

novobiocin

stavudine

Str+/Str+

nitrofurantoin

+/+

Str+/Str+

labetalol

niacin

false−/Str+

ketorolac

Str+/Str+

Str+/Str+

amiodarone

ketoconazole

Str+/Str+

result

acetaminophen

drug name

11/6

#/# #/#,C,P

8000/ 204/ 3.7/0.4

#/# #/A

/75

#/#

1667/2300

#/#,A

150/100

#/# P/M

117/

11

204

8000

150

350

0.4↑1.2↓ 1.2

0.015↑0.4↓/0.4

600

100

200

167

200

9

11

22

22

100

3.7

11

3.7

NI (I)

0.015↑0.4↓/1.4

#/#

A,#/#,C,P

5000/2300↑

1800/

#/# A,#/#,A

100/

/8000

#/#

#/#

#/#

600/400

/0.7↑

#/A M,P/#,P

/3800↑

500/330

67/6

/220

9/13↑50↓

56/

11/50

0.8↑7.4↓/0.5↑

P/#,A

I,M,#/all

#/A

/A,M,C,P

A,I,#/A

A/M

A,I/#

A/#,A

67/25

150↑/126↑

#/# I,P,#/#

/50

33/13

/13

11/25

11/25

1333/7800

NA (A)

C,#/#

A, M,#/#,A

I,M,#/M

A,I,#/C

I/#

#/#

first signal

Table 4. Cytotoxicity Potentialof Drugs withKnown Effects on Humansa IC (C)

33/50

/16

67/25

83/50

/25

11/13

33/13

11/13

/500

600/400

/50

/330

200/50

/220

11/3↑100↓

204/

8000/

/135↑

/570

50/25

39↑/50↑

0.02↑0.4↓11↑/1↑2↓

0.02↑0.4↓11↑/1↑2↓

33/25

3.7/

/13

204/

2667/

/270

/2300

/25

350/

1.2/0.1

0.4/0.1

33/13

Moderately HepatotoxicDrugs

/4000↑

67↑200↓/400

22/50

/7500

167/330

200/50↑

/220↑

28/25↑100↓ 83/25

1500↑/50↓

33/100

7.4/505

67/25

150/250

/100

33/50↑100↓

3.7/2↑

33/50

11/100

1333/4000

Severely HepatotoxicDrugs

MMP (M)

100/6

204/

8000/

556/2300

50/50

3.7/0.1

1.2/0.1

33/13

33/25

/4000

/1000

67/100

/50

4000/

/330

200/100

/220

250/50

100/50

22/4

22/25

150/126

/50

100/50

100/13

33/25

11/25

4000/

PMP (P)

1.2/1.6↑25↓

68/25

889/50

50/67

1667/1100

17/6

1.4↑13↓/1.6

100 μM/

95%/96%

sensitivity

flumazenil

Str+/Str+

Str+/Str+

tacrine

A/A

I/A,M,# #/#

67/36

90/15

#/C M/A,C,P,#

11/3

/320

11/13

NA (A)

C/C

/A

#/#

#/#

first signal

+/+

tamoxifen

sulfamethoxazole

Str+/+

Str+/Str+

quinine

simvastatin

Str+/Str+

quinidine

?/Str+

Str+/Str+

quinacrine

procainamide

Str+/Str+

Hi TI−/Hi TI−

pravastatin

result

paroxetine

drug name

Table 4. continued IC (C)

67/36

90/0.5

1.2/0.4

/6.3

25

1500/100

/50

11/

100/6

/50

11/6

4200/

67↑200↓/100

100/50

4.7/100

/0.4

3.7↑11↓/6

200/50

33/25

/100

33/

11/6

0.69↑6.1↓/0.2↑0.4↓3↑ /0.4

11/3↑6↓13↑50↓ 0.4/0.8

500/100

50/25

3.7↑33↓/26↑50↓

4200/

11/6

Drugs Toxic toOther Organs

67/

500/2

22/100↑

113/

Nontoxic Drugs

11/25

100/50↑100↓ 100/50

3.7/0.2↑0.9↓3↑50↓ /6

2.5↑200↓/70

270/30

3.7↑11↓/3

/630↑

11/13

Moderately HepatotoxicDrugs

MMP (M)

200/50

100/25

110/100

11/50

/13

33/1.6

500/

/50

3.7/13

2500

11/6

200/25

/100

200/50

3.7/

/100

340/

/50

11/25

/25

100/6

200/36

270/1

11/3

/630

3.7/13

PMP (P)

7.4/13

3.7/3

110/3↑100↓

33/27

0.14↑3.7↓/0.14/0.4

18.5/3

1.2/1.6

167/

/6

3.7/6

1400/1250

3.7/6

67/50

500/13

7.4/50

33/6↑100↓

/100

113/

/100↑

100/25

/1100↑

3.7/0.2↑50↓

22/36

10/8

3.7/2

/630

100/100

0.4↑11↓/0.8↑25↓

CN (#)

7.4/3

3.7/3

4.7/3

33/0.4

1.1

0.69/0.1

0.4/0.8

167/100

50/6

3.7/6

1400/1250

3.7/6

22/25

500/2

7.4/50

3.7/6

/100

113/>100

/50

1.2/0.4

33/13

28.8/1100

0.4/0.2

2.5/36

10/0.5

1.2/0.4

/320

100/100

0.4/0.8

lowest cytotoxicconcn

145

0.11

11

0.09

0/0.4

0.45

46

0.1

0.48

42

0.01

0.1

50

0.0001

0.006

0.73

3.4

0.1

0.4

0.1

217

0.02

19

9

1

12

0.1

0.2

Cmax

n/a/n/a

34/27

0.4/0.3

367/4

1/2

4/2

500/60

8/13

33/30

370/600

220/250

10/0.04

74000/500000

617/1000

>137/137

33/>30

>1000/500

3/1

330/130

0.1/5

20/10

0.1/2

1.1/0.06

1.2/0.4

>42/27

1000/1000

2/4

TI

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

93%/97%

89%/89%

sensitivity

specificity

33/13↑

#/

#/

/#

100

100

NI (I)

IC (C)

33/

11/50

11/50 33/50

/50

Positive Controls:Chemicals Causing Toxicity

MMP (M)

100/50

100/100

PMP (P)

333/

1000 200/

3000/

Overall Sensitivityand Specificity

200/

333/

Previously NotTested 1000/

Negative Controls:Chemicals Nontoxic to Humans (at Concentrations up to 100 μM)

33/

33↑100↓/100

#/# #/A

A,M,#/

NA (A)

first signal

67/

111/

/5000

33/

3.7/25

3.7/13

CN (#)

67/

111/

>100/

>100/

>100/5000

33/

3.7/13

3.7/13

lowest cytotoxicconcn

4.2

n/a

n/a

n/a

n/a

n/a

n/a

Cmax

16/

n/a/n/a

n/a/n/a

n/a/n/a

n/a/n/a

n/a/n/a

n/a/n/a

TI

The 70 compoundstested in thisstudy are grouped according to their human hepatotoxicity. The overallresult of the HCA assay is listed in column 2 followed by the parameter(s)where toxicity was first noted (column 3). The lowest cytotoxic concentrationsfor individual assay parameters (NA (A), nuclear area; NI (I), nuclearintensity; MMP (M), mitochondrial membrane potential; IC (C), intracellularcalcium; PMP (P), plasma membrane potential; and CN (#), cell number)are listed in columns 4−9, and the lowest overall concentrationis indicated in column 10. The maximal plasma concentration associatedwith efficacy (Cmax) of the drug is indicatedin column 11, and the therapeutic index (TI, lowest cytotoxic concentration/Cmax) is indicated in column 12. Results aretabulated for both this study and compared with corresponding resultspresented by O’Brien et al.8 (currentstudy/O’Brien et al.8). The sensitivity/specificityis indicated at the bottom of each group, and the overall sensitivityand specificity are shown at the end of the table. Concentrationsare presented in μM except for sodium chloride (mM) (n/a, notapplicable; Str+, strong positive; +, positive; −, negative;Hi TI−, negative due to high TI; >100 μM, negativebecausetoxicity occurred higher than 100 μM; ?, no cytotoxic responsewas observed or the drug was not tested at a high enough concn toelicit a clear cytotoxic response).

Str+/

troglitazone

a

Str+/

4-chloro-m-cresol

+/

lactose

100%/100%

+/

ethanol

specificity

100%/75%

sensitivity

?/+

Str+/false −

ryanodine

sodium hydroxide

Str+/Str+

Str+/Str+

diquat

result

fccp

drug name

Table 4. continued

Chemical Research in Toxicology Article

822

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

when performed at different times and different sites (Figure 5). The lowest concentration of the drug found to be cytotoxic was the same or one point apart for the different sites. Furthermore, the data correlate well with previously published 6-point dose− response curves of doxorubicin,64 further confirming the assay’s reproducibility. After treatment with 20 μM (6-point curves) or 33 μM (12-point curves) doxorubicin, only a few cells were observed. In contrast, cells were affected much earlier in the Pfizer study, where no cells were detected at concentrations higher than 4 μM. A possible explanation for this is the hormetic responses for NA and MMP that can be observed in the 12-point curves and in Figure 5. By using 3-fold dilutions, a larger concentration range can be covered compared to the 2-fold dilutions used in the Pfizer study, where hormesis in response to doxorubicin was not detected. The 10-fold dilutions used in the 6-point curves allow coverage of even larger concentration ranges, but the drawback is a cruder estimate of the lowest cytotoxic concentration. It is also difficult to correctly distinguish whether a change in a tested parameter for one concentration should be attributed to a hormetic response or be rejected as an outlier. The dynamic ranges of the IC and PMP signals were found to be much higher (16- and 24-fold, respectively) in the Pfizer study compared to those in this study. Doxorubicin has been reported to intercalate with DNA at high concentrations.64,142 It is therefore likely that the observed decrease in NI results from doxorubicin competing for DNA binding with Hoechst (Figure 2B, left column, and Figure 5). Sixty-nine compounds were assessed as part of this study, 67 of which had previously been examined in the Pfizer study. In this study, the sensitivities of the assay were found to be 95, 95, 83, and 100% for the severely hepatotoxic drugs, moderately hepatotoxic drugs, drugs toxic to other organs, and positive controls, respectively. In the Pfizer study, these were found to be 100, 96, 92, and 75%. Differences between the two studies can be attributed to false negatives or uncertain results. Overall, the sensitivity and specificity of the assay in this study were 93 and 89% compared to 97 and 89% in the Pfizer study, respectively.8 The decrease in specificity with the current set of drugs can most likely be attributed to the small number of nontoxic drugs and negative controls, meaning a false result had greater impact in this analysis. Linear regression using least-squares fit was carried out to determine the correlation between the two studies. The slope of the line was found to be 0.96 with an R2 value of 0.79. On the basis of the results of the Pfizer study,8 various adaptations were made to the assay protocol to improve the precision as well as the user friendliness of the assay. In the Pfizer study, cell-to-cell variance was found to be the largest source of imprecision, and cell number is the parameter with the highest variance. This could be due to clumping of cells in the plate, cell passage number, cell viability, and variation in cell counts. It can be costly and time- and labor-intensive to maintain and prepare cells for use. Therefore, it was decided that HepG2 cells would be produced at a large scale and frozen in aliquots. This allowed for highly standardized, quality-controlled cell handling and precision over time. These cells were always available and easy to use with just a thaw-and-seed step, making this approach time and cost-effective. Cell number was as much as 50% lower in 1 of 5 plates seeded at sites 1 and 3, where dilution was based on new cell counts before seeding, but this problem did not occur at site 2 where cell counts from the cell factory were used (data not shown). To reduce the time required per plate and improve the scale-up potential, images were not acquired in a time series but

Figure 7. Scatter plot of the correlation between the current UCD and GE studies with the previous pfizer studies. The cytotoxic concentration was determined for 67 compounds, and the results were compared with those of the Pfizer study. Only 47 points are discriminable on the graph because 17 of the 67 points were identical. The slope of the line is 0.96 with an R2 of 0.79. Picotamide (filled circle in graph), a nontoxic drug with a false positive result in both studies, showed the worst correlation and was excluded as a statistical outlier from the correlation analysis.

The total number of comparable samples was 68, but four samples (ketorolac, procainamide, ryanodine, and sodium hydroxide) were excluded because of ambiquity where a positive result was expected, e.g., the toxic concentration could be determined in one study but not in the other. For the 64 included compounds, the slope was 0.96 (R2 = 0.79). The worst correlation was observed for picotamide, a nontoxic drug that elicited a false positive response in both studies. The negative results for sodium hydroxide and procainamide are regarded as uncertain because they were not tested to concentrations as high as in the Pfizer study. In the present study, sodium hydroxide was only tested to 100 μM, whereas the first cytotoxic signal in the Pfizer study was observed for CN at 5000 μM. Procainamide was tested up to 500 μM in the present study. In the Pfizer study, toxicity was noted at 320 μM for NA and at 630 μM for the other three parameters.

4. DISCUSSION 4.1. Validation of the HCS Assay across Sites and Technologies. The assay described herein involves the high content analysis of toxicity in human HepG2 hepatoma hepatocytes. This assay has previously been validated8 and offers a number of advantages over traditional in vitro cytotoxicity assays, which are specific but not sensitive, usually detecting only extreme toxicity. In contrast, this high content analysis detects prelethal toxicity by measuring the effect of the drug of interest after a three day incubation period and by measuring multiple parameters to obtain a large amount of informative data. The nonlethal effects of a drug are key in determining its ultimate safety. The potential of this assay to improve safety screening has been established with toxicity prediction at 93% sensitivity and 97% specificity.8 A further requirement for its widespread use is its reproducibility between different technologies and locations. Therefore, an aim of this study was to select a subset of the drugs originally tested in the Pfizer study and compare the results between three different sites. Imprecision in the data can be due to various sources and can affect the accuracy of the results. This research demonstrated that the imprecision for the measured parameters was, for the most part, consistent from location to location, serving as confirmation of the transferable nature of the assay. The doxorubicin dose− response curves demonstrate the reproducibility of the assay 823

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

Increased ROS production and altered mitochondrial membrane potential were also found in the cells exposed to steatotic drugs.115 Various attempts have been made to adapt the HepG2 cell line to make it a better model for cytotoxicology assays. Adenovirusmediated expression of the drug-metabolizing enzymes cytochrome P450s increased their catalytic activity in HepG2 cells, which are normally low.148,149 Exposure of HepG2 cells to compounds that are normally metabolized by the P450 system induce this system and metabolic competence.8,136 HepaRG cells have also been suggested as an alternative model to HepG2.105,115,116,150 They were derived from a human hepatocellular carcinoma and maintain cytochrome P450 levels seen in vivo.150 At present, the amount of value added by using HepaRG cells in this cytotoxicity assay would not warrant switching from HepG2 cells that are widely used, easily available commercially, and probably the most well-characterized of all human hepatocyte lines. Primary human hepatocytes are also a model system for experiments on liver toxicity. However, their use is hampered by a limited supply and heterogeneity in enzyme levels from donor-to-donor149 and at present is not a practical alternative to use the HepG2 cell line for the HCA cytotoxicity assay described herein. Despite the limitations of the HepG2 cell line, it is currently the most suitable system in which to test for hepatotoxicity using this assay. The use of HCA means that multiple parameters can easily be measured, and the assay is compatible with the testing of a large number of compounds in a short time, allowing for highthroughput. The original study presenting this novel cell-based cytotoxicity assay used a Kinetic-Scan HCS Reader, whereas the present study utilized the more modern IN Cell Analyzer 1000 and demonstrates the transferability of the assay across technologies. This research has also shown that the assay is comparable when performed by different personnel in varied locations. The adaptation of the assay by other researchers has shown that additional or alternative biomarkers can be included, depending on the specifics of the compounds being tested. We conclude that there is an overwhelming amount of evidence that has accumulated over the last 10 years to support the validity of the multiparametric HCA cytotoxicity assay for effective use in predictive toxicology originally reported in 2006.8 Moreover, this study reproduced a similarly high level of concordance in excess of 90% for HCA assay results with human toxicity, and this accuracy held across sites and HCA technologies. Similar assays have been applied across a wide variety of toxicants, toxicities, cell types, and disciplines. Review herein of the wider literature on cytotoxicity, since the first assay was reported 100 years ago, supports the selection of key cytobiomarkers along a final common pathway of cell injury, including cell proliferation, mitochondrial activity, apoptosis, lysosomal mass, oxidative stress, and cell membrane permeability. HCS studies without inclusion of multiple key cytobiomarkers or without testing to sufficiently high concentration or duration have not been as successful at predicting human toxicity potential. A protocol is demonstrated for operational validation of the assay within laboratories to demonstrate proficiency and quality management.

only at 1 h after dye addition. This also reduced the amount of time required for analysis and data storage space. In the last ten years, HCA has moved from being a novel technology to one that is used as the standard in a range of areas of research including cell signaling, oncology, and in vitro toxicology.143 Improvement over more traditional techniques has been demonstrated with the increased information obtained resulting in more accurate results compared to those from ELISA, for example.144 In the area of cytotoxicology, O’Brien et al.8 demonstrated marked improvement in sensitivity and specificity of the assay used in this research compared to those of seven more traditional toxicity assays. Since then, numerous studies (see Table 2) have carried out multiparameter cytotoxicity assays using HepG2 cells and HCA similar to the original assay format8 and demonstrated its utility. Tolosa et al.111 tested a similar assay, adding a measurement of oxidative stress, and found a specificity of 92% and sensitivity of 94%. However, Xu et al.95 found their HCA assay in primary human hepatocyes95 had 50−60% sensitivity. This lower predictivity may be attributed to the use of only 24 h exposure of cells to drugs as 72 h exposure has been found to give higher sensitivity in time-course studies;8,103 cell proliferation, which is the most sensitive parameter for the detection of cytotoxicity, and cell membrane permeability were not measured. Persson et al.103 used a similar assay with 100 compounds with varying degrees of hepatotoxicity and found a sensitivity of 50−60%. However, this lower estimate was attributable to their testing drugs only up to an exposure of 100 μM and not to a therapeutic index of 100 (i.e., 100 × Cmax). If only the drugs tested to this exposure were considered in their study, then the sensitivity and specificity would be similar to those of the original HCA assay. Nevertheless, even with the limitations of the Xu and Persson study designs, the accuracy of their performance of the HCA cytotoxicity assay was much higher than for historical assays. The study performed herein confirms the high sensitivity and specificity previously reported8 even with major differences in time, geography, laboratory, staff, and technology. As the liver is the primary organ involved in drug metabolism, and excretion assessment of hepatotoxicity is critical for determining the safety of a drug, in vitro assays using the liver cell line HepG2 therefore have the potential to accurately predict the in vivo hepatotoxicity of a drug. However, there has been some debate over the metabolic competence of HepG2 cells and if it is sufficient to mimic in vivo liver metabolism.145 Several studies have shown8,96,111 though that their HCA assays using HepG2 have identified most bioactivatable hepatotoxins (including aflatoxin B1, acetaminophen, clozapine) suggesting these cells do not restrict the analysis of these compounds. Recently, it was shown that HepG2 cells have low but inducible enzyme activities of cytochrome P450 (CYP) enzymes 1A1, 1A2, 2B6, and 3A4, which are responsible for the metabolism of most but not all drugs.136 This may explain the need for multiple days of incubation of HepG2 cells with most drugs to best detect their toxicities.8 Recently, HepG2 cells have also shown themselves to be a useful model in which to test for steatosis.100,105,146 Hepatocytes and HepG2 cells were demonstrated to have similar levels of lipid accumulation in response to treatment with steatosis inducers.147 Donato et al.110 developed an HCS assay for steatosis that included the dye BODIPY493/503 to monitor the lipid content of the cells. The dye showed significant increases in fat deposits in HepG2 cells in response to 16 drugs known to cause liver steatosis but not the negative controls, six nonsteatotic compounds.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; phone: +353 1716 6048. ORCID

Peter James O’Brien: 0000-0003-4290-3585 824

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology

(6) Walum, E. (2014) Scandinavian Society for Cell Toxicology − thirty years of scientific pioneering. Basic Clin. Pharmacol. Toxicol. 115, 88−92. (7) O’Brien, P. J., and Haskins, J. R. (2007) In vitro cytotoxicity assessment. In High Content Screening. A Powerful Approach to Systems Cell Biology and Drug Discovery (Taylor, D. L., Haskins, J. R., and Giuliano, K. A., Eds.), pp 415−425, Humana Press, Totowa, NJ. (8) O’Brien, P. J., Irwin, W., Diaz, D., Howard-Cofield, E., Krejsa, C. M., Slaughter, M. R., Gao, B., Kaludercic, N., Angeline, A., Bernardi, P., Brain, P., and Hougham, C. (2006) High concordance of drug-induced human hepatotoxicity with in vitro cytotoxicity measured in a novel cellbased model using high content screening. Arch. Toxicol. 80, 580−604. (9) Olson, H., Betton, G., Stritar, J., and Robinson, D. (1998) The predictivity of the toxicity of pharmaceuticals in humans from animal data–an interim assessment. Toxicol. Lett. 102−103, 535−538. (10) O’Brien, P. J. (2008) High content analysis of sublethal cytotoxicity in human HepG2 hepatocytes for assessing potential and mechanism for chemical- and drug-induced human toxicity. In High Content Screening: Science, Techniques and Applications (Haney, S. A., Ed.), pp 293−315, John Wiley & Sons, Inc., Hoboken, NJ. (11) Combes, R. D. (2009) Some thoughts on the use of replacement alternatives for toxicity testing and risk assessment. Arch. Toxicol. 83, 199−201. (12) Clinical and Laboratory Standards Institute (2005). User Verification of Performance for Precision and Trueness; Approved Guideline, 2nd ed, CLSI document EP15-A2, Clinical and Laboratory Standards Institute, Wayne, PA. (13) Pappenheimer, A. M. (1917) Experimental studies upon lymphocytes. I. The reactions of lymphocytes under various experimental conditions. J. Exp. Med. 25, 633−650. (14) Martin, A. J. (1992) Cytotoxicity testing in vitro: investigation of 5 miniaturized, colorimetric assays. Ph.D. Thesis, Dublin City University. (15) Weimar, V. (1962) Effect of amino acid, purine, and pyrimidine analogues on activation of corneal stromal cells to take up neutral red. Invest. Opthamol. 1, 226−232. (16) Eagle, H., and Foley, G. E. (1956) The cytotoxic action of carcinolytic agents in tissue culture. Am. J. Med. 21, 739−774. (17) Eagle, H., and Foley, G. E. (1958) Cytotoxicity in human cell cultures as a primary screen for the detection of anti-tumor agents. Cancer Res. 18, 1017−1025. (18) Oyama, V. I., and Eagle, H. (1956) Measurement of cell growth in tissue culture with a phenol reagent (Folin-Ciocalteau). Exp. Biol. Med. (London, U. K.) 91, 305−307. (19) Smith, C. G., Lummis, W. L., and Grady, J. E. (1959) An improved tissue culture assay. I. Methodology and cytotoxicity of anti-tumor agents. Cancer Res. 19, 843−846. (20) Smith, C. G., Lummis, W. L., and Grady, J. E. (1959) An improved tissue culture assay. II. Cytotoxicity studies with antibiotics, chemicals, and solvent. Cancer Res. 19, 847−852. (21) Grady, J. E., Lummis, W. L., and Smith, C. G. (1960) An improved tissue culture assay. III. Alternate methods for measuring cell growth. Cancer Res. 20, 1114−1117. (22) McDonald, G. O., Stroud, A. N., Brues, A. M., and Cole, W. H. (1963) In vivo and in vitro assay for drug effect on cancer cells. Ann. Surg. 157, 785−796. (23) Porstmann, T., Ternynck, T., and Avrameas, S. (1985) Quantitation of 5-bromo-2−deoxyuridine incorporation into DNA: an enzyme immunoassay for the assessment of the lymphoid cell proliferative response. J. Immunol. Methods 82, 169−179. (24) Gentry, M. K., and Dalrymple, J. M. (1980) Quantitative microtiter cytotoxicity assay for Shigella toxin. J. Clin. Microbiol. 12, 361−366. (25) Rage, R., Mitchen, J., and Wilding, G. (1990) DNA fluorometric assay in 96-well tissue culture plates using Hoechst 33258 after cell lysis by freezing in distilled water. Anal. Biochem. 191, 31−34. (26) Skehan, P., Storeng, R. l, Scudiero, D., Monks, A., McMahon, J., Vistica, D., Warren, J. T., Bokesch, H., Kenney, S., and Boyd, M. R. (1990) New colorimetric cytotoxicity assay for anticancer-drug screening. J. Natl. Cancer Inst. 82, 1107−1112.

Funding

The authors gratefully acknowledge the financial support of this work by (a) Food Health of Ireland, UCD, Enterprise Ireland Grant Number CC20080001, (b) Advanced Diagnostics Laboratory (ADL), Park West Enterprise Centre, Lavery Ave., Park West, Dublin 12, and (c) GE Healthcare, The Maynard Centre, Forest Farm Industrial Estate, Longwood Dr, Cardiff, Wales CF14 7YT, UK. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors gratefully acknowledge the technical and scientific support of Dr. Suzanne Hancock who provided the data from GE Healthcare.



ABBREVIATIONS APAF, apoptotic protease-activating factor; ATP, adenosine triphosphate; BAK, Bcl-2 killer; BAX, Bcl-2-associated X protein; Bcl-2, B-cell lymphoma 2; BH3-only, Bcl-2 sequence homology only in protein domain 3; BID, BH3-interacting domain death agonist; BIK, Bcl-2 interacting killer; BIM, BH3-only Bcl-2interacting mediator of cell death; CAD, caspase-activated deoxyribonuclease; CADS, cationic amphiphilic drugs; Cmax, maximum drug concentration in serum with a therapeutic dose; CN, cell number; DiOC6, 3,3′-dihexyloxacarbocyanine iodide; HCA, high content analysis; HCS, high content screening; FCCP, carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone; IAP, inhibitors-of-apoptosis proteins; IC, intracellular calcium; JC-1, 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazlocarbocyanine iodide; LDH, lactate dehydrogenase; MMP, mitochondrial membrane potential; MOMP, mitochondrial outer membrane permeabilization; MPTP, mitochondrial permeability-transition pore; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; NA, nuclear area; NADH, reduced nicotinamide adenine dinucleotide; NI, nuclear intensity; NRTI, nucleoside reverse transcriptase inhibitors; PIP2, phosphatidyl 4,5-bisphosphate; PS, phosphatidyl serine; SMAC-DIABLO, second mitochondria-derived activator of caspase direct inhibitor of apoptosis-binding protein with low pI; TI, therapeutic index; TMRE, tetramethylrhodamine ethyl ester; TMRM, tetramethylrhodamine methyl ester; XTT, 2,3-bis(2-methoxy-4-nitro-5sulfophenyl)-2H-tetrazolium-5-carboxanilide



REFERENCES

(1) O’Brien, P. J. (2009) Discovery toxicology screening: predictive, in vitro cytotoxicity. In Hit and Lead Profiling: Identification and Optimization of Drug-like Molecules (Faller, B., and Urban, L., Eds.), pp 325−344, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim. (2) Fung, M., Thornton, A., Mybeck, K., Wu, J. H.-h., Hornbuckle, K., and Muniz, E. (2001) Evaluation of the characteristics of safety withdrawal of prescription drugs from worldwide pharmaceutical markets1960− 1999. Drug Inf. J. 35, 293−317. (3) Dykens, J. A., and Will, Y. (2007) The significance of mitochondrial toxicity testing in drug development. Drug Discovery Today 12, 777− 785. (4) Segall, M. D., and Barber, C. (2014) Addressing toxicity risk when designing and selecting compounds in early drug discovery. Drug Discovery Today 19, 688−693. (5) Olson, H., Betton, G., Robinson, D., Thomas, K., Monro, A., Kolaja, G., Lilly, P., Sanders, J., Sipes, G., Bracken, W., Dorato, M., Van Deun, K., Smith, P., Berger, B., and Heller, A. (2000) Concordance of the toxicity of pharmaceuticals in humans and in animals. Regul. Toxicol. Pharmacol. 32, 56−67. 825

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology (27) Crouch, S. P. M., Kozlowski, R., Slater, K. J., and Fletcher, J. (1993) The use of ATP bioluminescence as a measure of cell proliferation and cytotoxicity. J. Immunol. Methods 160, 81−88. (28) Medzihradsky, F., and Marks, M. J. (1975) Measures of viability in isolated cells. Biochem. Med. 13, 164−177. (29) Brunner, K. T., Mauel, J., Cerotttini, J.-C., and Chapuis, B. (1968) Quantitative assay of the lytic action of immune lymphoid cells on 51Crlabelled allogeneic target cells in vitro; inhibition by isoantibody and by drugs. Immunology 14, 181−196. (30) Korzeniewski, C., and Callewaert, D. M. (1983) An enzymerelease assay for natural cytotoxicity. J. Immunol. Methods 64, 313−320. (31) Bermudez, V. M., Miller, R. B., Rosendal, S., Fernando, M. A., Johnson, W. H., and O’Brien, P. J. (1992) Measurement of the cytotoxic effects of different strains of Mycoplasma equigenitalium on the equine uterine tube using a calmodulin assay. Can. J. Vet. Res. 56, 331−338. (32) Smits, B., Rosendal, S., Ruhnke, H. L., Plante, C., O’Brien, P. J., and Miller, R. B. (1994) Effects of Ureaplasma diversum on bovine oviductal explants: quantitative measurement using a calmodulin assay. Can. J. Vet. Res. 58, 114−121. (33) Mosmann, T. (1983) Rapid colorimetric assay for cellular growth and survival: application to proliferation and cytotoxicity assays. J. Immunol. Methods 65, 55−63. (34) Roehm, N. W., Rodgers, G. H., Hatfield, S. M., and Glasebrook, A. L. (1991) An improved colorimetric assay for cell proliferation and viability utilizing the tetrazolium salt XTT. J. Immunol. Methods 142, 257−265. (35) Buttke, T. M., McCubrey, J. A., and Owen, T. C. (1993) Use of an aqueous soluble tetrazolium/formazan assay to measure viability and proliferation of lymphokine-dependent cell lines. J. Immunol. Methods 157, 233−240. (36) Ishiyama, M., Tominaga, H., Shiga, M., Sasamoto, K., Ohkura, Y., and Ueno, K. (1996) A combined assay of cell viability and in vitro cytotoxicity with a highly water-soluble tetrazolium salt, neutral red and crystal violet. Biol. Pharm. Bull. 19, 1518−1520. (37) Berridge, M. V., Tan, A. S., McCoy, K. D., and Wang, R. (1996) The biochemical and cellular basis of cell proliferation assays that use tetrazolium salts. Biochemica 96, 14−19. (38) Page, B., Page, M., and Noel, C. (1993) A new fluorometric assay for cytotoxicity measurements in vitro. Int. J. Oncol. 3, 473−476. (39) Slaughter, M. R., Bugelski, P. J., and O’Brien, P. J. (1999) Evaluation of Alamar Blue reduction for the in vitro assay of hepatocyte toxicity. Toxicol. In Vitro 13, 567−569. (40) Slaughter, M. R., Thakkar, H., and O’Brien, P. J. (2002) Effect of diquat on the antioxidant system and cell growth in human neuroblastoma cells. Toxicol. Appl. Pharmacol. 178, 63−70. (41) Bernal, S. D., et al. (1982) Monitoring the effect of anti-cancer drugs on L1210 cells by a mitochondrial probe, rhodamine-123. Int. J. Cancer 30, 219−224. (42) Bellomo, G., Jewell, S. A., Thor, H., and Orrenius, S. (1982) Regulation of intracellular calcium compartmentation: Studies with isolated hepatocytes and t-butyl hydroperoxide. Proc. Natl. Acad. Sci. U. S. A. 79, 6842−6846. (43) Tsien, R. Y., Pozzan, T., and Rink, T. J. (1982) Calcium homeostasis in intact lymphocytes: cytoplasmic free calcium monitored with a new intracellularly trapped fluorescent indicator. J. Cell Biol. 94, 325−334. (44) Brattin, W. J., Pencil, S. D., Waller, R. L., Glende, E. A., Jr., and Recknagel, R. O. (1984) Assessment of the role of calcium ion in halocarbon hepatotoxicity. Environ. Health Perspect. 57, 321−323. (45) O’Brien, P. J., Kalow, B. I., Brown, B. D., Lumsden, J. H., and Jacobs, R. M. (1989) Porcine malignant hyperthermia susceptibility: halothane-induced increase in cytoplasmic free calcium in lymphocytes. Am. J. Vet. Res. 50, 131−135. (46) Onal, M., Ovet, G., and Onal, O. (2015) Review of apoptosis. M. O. J. Immunol. 3, 00073−77. (47) Elmore, S. (2007) Apoptosis: a review of programmed cell death. Toxicol. Pathol. 35, 495−516. (48) Afanas’ev, V. N., Korol, B. A., Mantsygin, Y. A., Nelipovich, P. A., Pechatnikov, V. A., and Umansky, S. R. (1986) Flow cytometry and

biochemical analysis of DNA degradation characteristic of two types of cell death. FEBS Lett. 194, 347−350. (49) Compton, M. M., Haskill, J. S., and Cidlowski, J. A. (1988) Analysis of glucocorticoid actions on rat thymocyte deoxyribonucleic acid by fluorescence-activated flow cytometry. Endocrinology 122, 2158−2164. (50) Ellwart, J. W., and Dörmer, P. (1990) Vitality measurement using spectrum shift in Hoechst 33342 stained cells. Cytometry 11, 239−243. (51) Nicoletti, I., Migliorati, G., Pagliacci, M. C., Grignani, F., and Riccardi, C. (1991) A rapid and simple method for measuring thymocyte apoptosis by propidium iodide staining and flow cytometry. J. Immunol. Methods 139, 271−279. (52) Schmid, I., Krall, W. J., Uittenbogaart, C. H., Braun, J., and Giorgi, J. V. (1992) Dead cell discrimination with 7-amino-actinomycin D in combination with dual color immunofluorescence in single laser flow cytometry. Cytometry 13, 204−208. (53) Telford, W. G., King, L. E., and Fraker, P. J. (1992) Comparative evaluation of several DNA binding dyes in the detection of apoptosisassociated chromatin degradation by flow cytometry. Cytometry 13, 137−143. (54) Becker, B., Clapper, J., Harkins, K. R., and Olson, J. A. (1994) In situ screening assay for cell viability using a dimeric cyanine nucleic acid stain. Anal. Biochem. 221, 78−84. (55) Lee-MacAry, A. E., Ross, E. L., Davies, D., Laylor, R., Honeychurch, J., Glennie, M. J., Snary, D., and Wilkinson, R. W. (2001) Development of a novel flow cytometric cell-mediated cytotoxicity assay using the fluorophores PKH-26 and TO-PRO-3 iodide. J. Immunol. Methods 252, 83−92. (56) Ashley, N., Harris, D., and Poulton, J. (2005) Detection of mitochondrial DNA depletion in living human cells using PicoGreen staining. Exp. Cell Res. 303, 432−446. (57) Vermes, I., Haanen, C., Steffens-Nakken, H., and Reutellingsperger, C. (1995) A novel assay for apoptosis flow cytometric detection of phosphatidylserine expression on early apoptotic cells using fluorescein labelled annexin V. J. Immunol. Methods 184, 39−51. (58) Koopman, G., Reutelingsperger, C. P., Kuijten, G. A., Keehnen, R. M., Pals, S. T., and van Oers, M. H. (1994) Annexin V for flow cytometric detection of phosphatidylserine expression on B cells undergoing apoptosis. Blood 84, 1415−1420. (59) Kamada, S., Kikkawa, U., Tsujimoto, Y., and Hunter, T. (2005) Nuclear translocation of caspase-3 is dependent on its proteolytic activation and recognition of a substrate-like protein(s). J. Biol. Chem. 280, 857−860. (60) Kerr, J. F. R., Wyllie, A. H., and Currie, A. R. (1972) Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br. J. Cancer 26, 239−57. (61) Darzynkiewicz, Z., Bruno, S., Del Bino, G. L., Gorczyca, W., Hotz, M. A., Lassota, P., and Traganos, F. (1992) Features of apoptotic cells measured by flow cytometry. Cytometry 13, 795−808. (62) Ormerod, M. G., Paul, F., Cheetham, M., and Sun, X. M. (1995) Discrimination of apoptotic thymocytes by forward light scatter. Cytometry 21, 300−304. (63) Arends, M. J., Morris, R. G., and Wyllie, A. H. (1990) Apoptosis. The role of the endonuclease. Am. J. Pathol. 136, 593−608. (64) Lieggi, N. T., Edvardsson, A., and O’Brien, P. J. (2010) Translation of novel anti-cancer cytotoxicity biomarkers detected with high content analysis from an in vitro predictive model to an in vivo cell model. Toxicol. In Vitro 24, 2063−2071. (65) Domingos, M. C., Davies, A. M., and O’Brien, P. J. (2014) Application of high content analysis in clinical cytology for translational safety biomarkers of drug-induced toxicity for lymphoma chemotherapy. Basic Clin. Pharmacol. Toxicol. 115, 145−153. (66) Voskoboinik, I., Whisstock, J. C., and Trapani, J. A. (2015) Perforin and granzymes: function, dysfunction and human pathology. Nat. Rev. Immunol. 15, 388−400. (67) Tait, S. W. G., and Green, D. R. (2010) Mitochondria and cell death: outer membrane permeabilization and beyond. Nat. Rev. Mol. Cell Biol. 11, 621−32. 826

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology (68) Delbridge, A. R. D., Grabow, S., Strasser, A., and Vaux, D. L. (2016) Thirty years of BCL-2: translating cell death discoveries into novel cancer therapies. Nat. Rev. Cancer 16, 99−109. (69) Ichim, G., and Tait, S. W. G. (2016) A fate worse than death: apoptosis as an oncogenic process. Nat. Rev. Cancer 16, 539. (70) Novoderezhkina, E. A., Zhivotovsky, B. D., and Gogvadze, V. G. (2016) Induction of unspecific permeabilization of mitochondrial membrane and its role in cell death. Mol. Biol. 50, 43−58. (71) Biasutto, L., Azzolini, M., Szabò, I., and Zoratti, M. (2016) The mitochondrial permeability transition pore in AD 2016: An update. Biochim. Biophys. Acta, Mol. Cell Res. 1863, 2515. (72) Jenssen, H.-L., Redmann, K., and Mix, E. (1986) Flow cytometric estimation of transmembrane potential of macrophages - a comparison with microelectrode measurements. Cytometry 7, 339−346. (73) Cossarizza, A., and Baccaranicontri, M. (1993) A new method for the cytofluorometric analysis of mitochondrial membrane potential using the J-aggregate forming lipophilic cation 5, 5′, 6, 6′-tetrachloro-1, 1′, 3′-tetraethylbenzimidazolcarbocyanine iodide (JC-1). Biochem. Biophys. Res. Commun. 197, 40−45. (74) Floryk, D., and Houstek, J. (1999) Tetramethyl Rhodamine methyl ester (TMRM) is suitable for cytofluorometric measurements of mitochondrial membrane potential in cells treated with digitonin. Biosci. Rep. 19, 27−34. (75) Perry, S. W., Norman, J. P., Barbieri, J., Brown, E. B., and Gelbard, H. A. (2011) Mitochondrial membrane potential probes and the proton gradient: a practical usage guide. BioTechniques 50, 98−115. (76) Fitzhugh, O. G., Nelson, A. A., and Holland, O. L. (1948) The chronic oral toxicity of chloroquine. J. Pharmacol. Exp. Ther. 93, 147− 152. (77) Halliwell, W. H. (1997) Cationic amphiphilic drug-induced phospholipidosis. Toxicol. Pathol. 25, 53−60. (78) Shayman, J. A., and Abe, A. (2013) Drug induced phospholipidosis: an acquired lysosomal storage disorder. Biochim. Biophys. Acta, Mol. Cell Biol. Lipids 1831, 602−611. (79) Osman, S., Taylor, K. A., Allcock, N., Rainbow, R. D., and MahautSmith, M. P. (2016) Detachment of surface membrane invagination systems by cationic amphiphilic drug. Sci. Rep. 6, 18536. (80) Piccoli, E., Nadai, M., Caretta, C. M., Bergonzini, V., Del Vecchio, C., Ha, H. R., Bigler, L., Dal Zoppo, D., Faggin, E., Pettenazzo, A., Orlando, R., Salata, C., Calistri, A., Palù, G., and Baritussio, A. (2011) Amiodarone impairs trafficking through late endosomes inducing a Niemann-Pick C-like phenotype. Biochem. Pharmacol. 82, 1234−1249. (81) Funk, R. S., and Krise, J. P. (2012) Cationic amphiphilic drugs cause a marked expansion of apparent lysosomal volume: Implications for an intracellular distribution-based drug interaction. Mol. Pharmaceutics 9, 1384−1395. (82) Aki, T., Nara, A., and Uemura, K. (2012) Cytoplasmic vacuolization during exposure to drugs and other substances. Cell Biol. Toxicol. 28, 125−131. (83) Belkin, M., Hardy, W. G., Orr, H. C., and Lachman, A. B. (1962) Induction in vitro by autonomic drugs of cytoplasmic vacuoles in ascites tumor cells. J. Natl. Cancer Inst. 28, 187−201. (84) Bento, C. F., Renna, M., Ghislat, G., Puri, C., Ashkenazi, A., Vicinanza, M., Menzies, F. M., and Rubinsztein, D. C. (2016) Mammalian autophagy: how does it work? Annu. Rev. Biochem. 85, 685−713. (85) Hong, M., Wang, N., and Feng, Y. (2015) The role of cell autophagy in cancer and its application in drug discovery. In Biochemistry, Genetics and Molecular Biology: Cell Death - Autophagy, Apoptosis and Necrosis (Ntuli, T., Ed.), InTech, DOI: 10.5772/61780. Available from: http://www.intechopen.com/books/cell-deathautophagy-apoptosis-and-necrosis/the-role-of-cell-autophagy-incancer-and-its-application-in-drug-discovery. (86) Gewirtz, D. A. (2014) The four faces of autophagy: implications for cancer therapy. Cancer Res. 74, 647−651. (87) Sharma, K., Le, N., Alotaibi, M., and Gewirtz, D. A. (2014) Cytotoxic autophagy in cancer therapy. Int. J. Mol. Sci. 15, 10034−10051. (88) O’Brien, P. J. (2014) High content analysis in toxicology: screening substances for human toxicity potential, elucidating

subcellular mechanisms, and in vivo use as translational safety biomarkers. Basic Clin. Pharmacol. Toxicol. 115, 4−17. (89) O’Brien, P. J., Slaughter, M. R., Swain, A., Birmingham, J. M., Greenhill, R. M., Elcock, F., and Bugelski, P. J. (2000) Repeated acetaminophen dosing in rats: adaptation of hepatic antioxidant system. Hum. Exp. Toxicol. 19, 277−283. (90) Bondesson, I., Ekwall, B., Hellberg, S., Romert, L., Stenberg, K., and Walum, E. (1989) MEIC − a new international multicenter project to evaluate the relevance to human toxicity of in vitro cytotoxicity tests. Cell Biol. Toxicol. 5, 331−347. (91) Ekwall, B. (1999) Overview of the final MEIC results II. The in vivo/in vitro evaluation, including a selection of a practical battery of cell tests for prediction of acute lethal blood concentrations in humans. Toxicol. In Vitro 13, 665−73. (92) O’Brien, P. J., Slaughter, M. R., Biagini, C., Diaz, D., Gao, B., Irwin, W., Krejsa, C., Hougham, C., Abraham, V., and Haskins, J. (2003) Predicting drug-induced human hepatotoxicity with in vitro cytotoxicity assays. In Proceedings Toxicology for the Pharma Industry, 9th-10th Dec 2003, London, UK. (93) Xu, J. J., Diaz, D., and O’Brien, P. J. (2004) Applications of cytotoxicity assays and pre-lethal mechanistic assays for assessment of human hepatotoxicity potential. Chem.-Biol. Interact. 150, 115−128. (94) Diaz, D., and O’Brien, P. J. (2006) Identifying human toxicity potential. Eur. Pharmaceut. Rew. 4, 1−8. (95) Xu, J. J., Henstock, P. V., Dunn, M. C., Smith, A. R., Chabot, J. R., and de Graaf, D. (2008) Cellular imaging predictions of clinical druginduced liver injury. Toxicol. Sci. 105, 97−105. (96) Davila, J. C., Xu, J. J., Hoffmaster, K. A., O’Brien, P. J., and Storm, S. C. (2007) Chapter 1. Current in vitro models to study drug-induced liver injury. In Hepatotoxicity: From Genomics to In Vitro and In Vivo Models (Sahu, S., ed.) J. Wiley, pp 3−55. (97) Morelli, J. K., Buehrle, M., Pognan, R., Barone, L. R., Fieles, W., and Ciaccio, P. J. (2006) Validation of an in vitro screen for phospholipidosis using a high-content biology platform. Cell Biol. Toxicol. 22, 15−27. (98) Ainscow, E. K., Pilling, J. E., Brown, N. M., Orme, A. T., Sullivan, M., Hargreaves, A. C., Cooke, E. L., Sullivan, E., Carlsson, S., and Andersson, T. B. (2008) Investigations into the liver effects of ximelagatran using high content screening of primary human hepatocyte cultures. Expert Opin. Drug Saf. 7, 351−365. (99) Garside, H., Marcoe, K. F., Chesnut-Speelman, J., Foster, A. J., Muthas, D., Kenna, J. G., Warrior, U., Bowes, J., and Baumgartner, J. (2014) Evaluation of the use of imaging parameters for the detection of compound-induced hepatotoxicity in 384-well cultures of HepG2 cells and cryopreserved primary human hepatocytes. Toxicol. In Vitro 28, 171−181. (100) Pradip, A., Steel, D., Jacobsson, S., Holmgren, G., IngelmanSundberg, M., Sartipy, P., Björquist, P., Johansson, I., and Edsbagge, J. (2016) High content analysis of human pluripotent stem cell derived hepatocytes reveals drug induced steatosis and phospholipidosis. Stem Cells Int. 2016, 2475631. (101) Abraham, V. C., Towne, D. L., Waring, J. F., Warrior, U., and Burns, D. J. (2008) Application of a high-content multiparameter cytotoxicity assay to prioritize compounds based on toxicity potential in humans. J. Biomol. Screening 13, 527−537. (102) Garippa, R. J., and Hoffman, A. F. (2009) High-content screening with emphasis on cytotoxicity and cell health measurements. Drugs Pharmaceutical Services 196, 382−398. (103) Persson, M., Løye, A. F., Mow, T., and Hornberg, J. J. (2013) A high content screening assay to predict human drug-induced liver injury during drug discovery. J. Pharmacol. Toxicol. Methods 68, 302−313. (104) Persson, M., Løye, A. F., Jacquet, M., Mow, N. S., Thougaard, A. V., Mow, T., and Hornberg, J. J. (2014) High-content analysis/screening for predictive toxicology: application to hepatotoxicity and genotoxicity. Basic Clin. Pharmacol. Toxicol. 115, 18−23. (105) Saito, J., Okamura, A., Takeuchi, K., Hanioka, K., Okada, A., and Ohata, T. (2016) High content analysis assay for prediction of human hepatotoxicity in HepaRG and HepG2 cells. Toxicol. In Vitro 33, 63−70. 827

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology (106) GE Healthcare Life Sciences. (2011) Toxicity Applications Manual. Reference Protocols for Use with IN Cell Analyzer. (107) Sirenko, O., Hesley, J., Rusyn, I., and Cromwell, E. F. (2014) High-content assays for hepatotoxicity using induced pluripotent stem cell-derived cells. Assay Drug Dev. Technol. 12, 43−54. (108) Bauch, C., Bevan, S., Woodhouse, H., Dilworth, C., and Walker, P. (2015) Predicting in vivo phospholipidosis-inducing potential of drugs by a combined high content screening and in silico modelling approach. Toxicol. In Vitro 29, 621−630. (109) Talbert, D. R., Doherty, K. R., Trusk, P. B., Moran, D. M., Shell, S. A., and Bacus, S. (2015) A multi-parameter in vitro screen in human stem cell-derived cardiomyocytes identifies ponatinib-induced structural and functional cardiac toxicity. Toxicol. Sci. 143, 147−155. (110) Donato, M. T., Tolosa, L., Jimenez, N., Castell, J. V., and GomezLechon, M. J. (2012) High-content imaging technology for the evaluation of drug-induced steatosis using a multiparametric cell-based assay. J. Biomol. Screening 17, 394−400. (111) Tolosa, L., Pinto, S., Donato, M. T., Lahoz, A., Castell, J. V., O’Connor, J. E., and Gómez-Lechón, M. J. (2012) Development of a multiparametric cell-based protocol to screen and classify the hepatotoxicity potential of drugs. Toxicol. Sci. 127 (1), 187−198. (112) Martin, H. L., Adams, M., Higgins, J., Bond, J., Morrison, E. E., Bell, S. M., Warriner, S., Nelson, A., and Tomlinson, D. C. (2014) Highcontent, high-throughput screening for the identification of cytotoxic compounds based on cell morphology and cell proliferation markers. PLoS One 9, e88338. (113) Tolosa, L., Carmona, A., Castell, J. V., Gómez-Lechón, M. J., and Donato, M. T. (2015) High-content screening of drug-induced mitochondrial impairment in hepatic cells: effects of statins. Arch. Toxicol. 89, 1847−1860. (114) Grimm, F. A., Iwata, Y., Sirenko, O., Bittner, M., and Rusyn, I. (2015) High-content assay multiplexing for toxicity screening in induced pluripotent stem cell-derived cardiomyocytes and hepatocytes assay. Assay Drug Dev. Technol. 13, 529−534. (115) Tolosa, L., Gómez-Lechón, M. J., Jiménez, N., Hervás, D., Jover, R., and Donato, M. T. (2016) Advantageous use of HepaRG cells for the screening and mechanistic study of drug-induced steatosis. Toxicol. Appl. Pharmacol. 302, 1−9. (116) Mennecozzi, M., Landesmann, B., Harris, G. A., Liska, R., and Whelan, M. (2012) Hepatotoxicity screening taking a mode-of-action approach using HepRG cells and HCA. ALTEX Proc. 1/12 Proc. WC8, 193−204. (117) Plymale, D. R., Haskins, J. R., and De La Iglesia, F. A. (1999) Monitoring simultaneous subcellular events in vitro by means of coherent multiprobe fluorescence. Nat. Med. 5, 351−355. (118) Walsh, E. G., Maher, S., Devocelle, M., O’Brien, P. J., Baird, A. J., and Brayden, D. J. (2011) High content analysis to determine cytotoxicity of the antimicrobial peptide, melittin and selected structural analogs. Peptides 32, 1764−1773. (119) Zhang, T. T., Stilwell, J. L., Gerion, D., Ding, L. H., Elboudwarej, O., Cooke, P. A., Gray, J. W., Alivisatos, A. P., and Chen, F. F. (2006) Cellular effect of high doses of silica-coated quantum dot profiled with high throughput gene expression analysis and high content cellomics measurements. Nano Lett. 6, 800−808. (120) Jan, E., Byrne, S. J., Cuddihy, M., Davies, A. M., Volkov, Y., Gun’ko, Y. K., and Kotov, N. A. (2008) High-content screening as a universal tool for fingerprinting of cytotoxicity of nanoparticles. ACS Nano 2, 928−938. (121) Anguissola, S., Garry, D., Salvati, G., O’Brien, P. J., and Dawson, K. A. (2014) High content analysis provides mechanistic insights on the pathways of toxicity induced by amine-modified polystyrene nanoparticles. PLoS One 9 (9), e108025. (122) George, S., Pokhrel, S., Xia, T., Gilbert, B., Ji, Z., Schowalter, M., Rosenauer, A., Damoiseaux, R., Bradley, K. A., Mädler, L., and Nel, A. E. (2010) Use of a rapid cytotoxicity screening approach to engineer a safer zinc oxide nanoparticle through iron doping. ACS Nano 26, 15−29. (123) Ramery, E., and O’Brien, P. J. (2014) Evaluation of the cytotoxicity of organic dust components on THP1 monocytes-derived macrophages using high content analysis. Environ. Toxicol. 29, 310−319.

(124) Rawlinson, L. A., O’Brien, P. J., and Brayden, D. J. (2010) High content analysis of cytotoxic effects of pDMAEMA on human intestinal epithelial and monocyte cultures. J. Controlled Release 146, 84−92. (125) Mundy, W. R., Radio, N. M., and Freudenrich, T. M. (2010) Neuronal models for evaluation of proliferation in vitro using high content screening. Toxicology 270, 121−130. (126) Wilson, M. S., Graham, J. R., and Ball, A. J. (2014) Multiparametric High Content Analysis for assessment of neurotoxicity in differentiated neuronal cell lines and human embryonic stem cellderived neurons. NeuroToxicology 42, 33−48. (127) Shah, I., Setzer, R. W., Jack, J., Houck, K. A., Judson, R. S., Knudsen, T. B., Liu, J., Martin, M. T., Reif, D. M., Richard, A. M., Thomas, R. S., Crofton, K. M., Dix, D. J., and Kavlock, R. J. (2016) Using ToxCast data to reconstruct dynamic cell state trajectories and estimate toxicological points of departure. Environ. Health Perspect., DOI: 10.1289/ehp.1409029. (128) Wilson, J., Berntsen, H. F., Zimmer, K. E., Frizzell, C., Verhaegen, S., Ropstad, E., and Connolly, L. (2016) Effects of defined mixtures of persistent organic pollutants (POPs) on multiple cellular responses in the human hepatocarcinoma cell line, HepG2, using high content analysis screening. Toxicol. Appl. Pharmacol. 294, 21−31. (129) Mohan, H. M., Collins, D., Maher, S., Walsh, E. W., Winter, D. C., O’Brien, P. J., Brayden, D. J., and Baird, A. W. (2012) The mycotoxin patulin increases colonic epithelial permeability in vitro. Food Chem. Toxicol. 50, 4097−4102. (130) Clarke, R., Connolly, L., Frizzell, C., and Elliott, C. T. (2015) High content analysis: A sensitive tool to detect and quantify the cytotoxic, synergistic and antagonistic effects of chemical contaminants in foods. Toxicol. Lett. 233, 278−286. (131) Papakonstantinou, S., and O’Brien, P. J. (2014) High content imaging for the morphometric diagnosis and immunophenotypic prognosis of canine lymphomas. Cytometry, Part B 86B, 373−382. (132) O’Brien, P. J., and Domingos, M. C. (2010) Use of high content analysis in toxicologic clinical pathology for identification and monitoring of translational safety biomarkers. Inter. Drug Discovery 5 (1), 50−55. (133) Phillips, G. W., Irwin, W. A., Howard-Cofield, E. J., Randle, L. E., Abraham, V., Haskins, J., and O’Brien, P. J. (2005) Incorporation of an oxidative stress biomarker into high content screening (HCS) for human toxicity potential. 11th Ann. Meeting Soc. Biomol. Screen. (SBS), P05020. (134) Schoonen, W. G. E. J., Westerink, W. M. A., de Roos, J. A. D. M., and Debiton, E. (2005) Cytotoxic effects of 110 reference compounds on HepG2 cells and for 60 compounds on HeLa, ECC-1 and CHO cells. I Mechanistic assays on ROS, glutathione depletion and calcein uptake. Toxicol. In Vitro 19, 505−516. (135) Schoonen, W. G. E. J., Westerink, W. M. A., de Roos, J. A. D. M., and Debiton, E. (2005) Cytotoxic effects of 110 reference compounds on HepG2 cells and for 60 compounds on HeLa, ECC-1 and CHO cells. II Mechanistic assays on NAD (P) H, ATP and DNA contents. Toxicol. In Vitro 19, 491−503. (136) Westerink, W. M. A., and Schoonen, W. G. E. J. (2007) Cytochrome P450 enzyme levels in HepG2 cells and cryopreserved primary human hepatocytes and their induction in HepG2 cells. Toxicol. In Vitro 21, 1581−1591. (137) Tomida, T., Okamura, H., Satsukawa, M., Yokoi, T., and Konno, Y. (2015) Multiparametric assay using HepaRG cells for predicting drug-induced liver injury. Toxicol. Lett. 236, 16−24. (138) Haskins, J. R., Rowse, P., Rahbari, R., and de La Iglesia, F. A. (2001) Thiazolidinedione toxicity to isolated hepatocytes revealed by coherent multiprobe fluorescence microscopy and correlated with multiparameter flow cytometry of peripheral leukocytes. Arch. Toxicol. 75, 425−438. (139) Rasola, A., and Geuna, M. (2001) A flow cytometry assay simultaneously detects independent apoptotic parameters. Cytometry 45, 151−157. (140) Taylor, D. L., DeBiasio, R., LaRocca, G., Pane, D., Post, P., Kolega, J., Giuliano, K., Burton, K., Gough, B., Dow, A., Yu, J., 828

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829

Article

Chemical Research in Toxicology Waggoner, J. S., and Farkas, D. L. (1994) Potential of machine-vision light microscopy in toxicologic pathology. Toxicol. Pathol. 22, 145−159. (141) Perlman, A. E., Slack, M. D., Feng, Y., Mitchison, T. J., Wu, L. F., and Altschuler, S. J. (2004) Multidimensional drug profiling by automated microscopy. Science 306, 1194−1198. (142) Hajian, R., Shams, N., and Mohagheghian, M. (2009) Study on the interaction between doxorubicin and deoxyribonucleic acid with the use of methylene blue as a probe. J. Braz. Chem. Soc. 20, 1399−1405. (143) Zock, J. M. (2009) Applications of high content screening in life science research. Comb. Chem. High Throughput Screening 12, 870−876. (144) Gasparri, F., Mariani, M., Sola, F., and Galvani, A. (2004) Quantification of the proliferation index of human dermal fibroblast cultures with the ArrayScan high-content screening reader. J. Biomol. Screening 9, 232−243. (145) Van Summeren, A., Renes, J., Mariman, E. C. M., Kleinjans, J. C. S., and van Delft, J. H. M. (2011) Response to pathophysiological relevance of proteomics investigations of drug-Induced hepatotoxicity in HepG2 Cells. Toxicol. Sci. 121, 431−433. (146) McMillian, M. K., Grant, E. R., Zhong, Z., Parker, J. B., Li, L., Zivin, R. A., Burczynski, M. E., and Johnson, M. D. (2001) Nile Red binding to HepG2 cells: an improved assay for in vitro studies of hepatosteatosis. In Vitro Mol. Toxicol. 14, 177−190. (147) Donato, M. T., Martinez-Romero, A., Jimenez, N., Negro, A., Herrera, G., Castell, J. V., O’Connor, J. E., and Gomez-Lechon, M. J. (2009) Cytometric analysis for drug-induced steatosis in HepG2 cells. Chem.-Biol. Interact. 181, 417−423. (148) Tolosa, L., Donato, M. T., Perez-Cataldo, G., Castell, J. V., and Gomez-Lechon, M. J. (2012) Upgrading cytochrome P450 activity in HepG2 cells co-transfected with adenoviral vectors for drug hepatotoxicity assessment. Toxicol. In Vitro 26, 1272−1277. (149) Aoyama, K., Yoshinari, K., Kim, H. J., Nagata, K., and Yamazoe, Y. (2009) Simultaneous expression of plural forms of human cytochrome P450 at desired ratios in HepG2 cells: adenovirus-mediated tool for cytochrome P450 reconstitution. Drug Metab. Pharmacokinet. 24, 209−217. (150) Aninat, C., Piton, A., Glaise, D., Le Charpentier, T., Langouet, S., Morel, F., Guguen-Guillouzo, C., and Guillouzo, A. (2005) Expression of cytochromes P450, conjugating enzymes and nuclear receptors in human hepatoma HepaRG cells. Drug Metab. Dispos. 34, 75−83.

829

DOI: 10.1021/acs.chemrestox.6b00403 Chem. Res. Toxicol. 2017, 30, 804−829