|Home | About | Journals | Submit | Contact Us | Français|
The efflux pump P-glycoprotein (ATP-binding cassette B1, multidrug resistance [MDR] 1, P-gp) has long been known to contribute to MDR against cancer chemotherapeutics. We describe the development of a dual-fluorescent cell line system to allow multiplexing of drug-sensitive and P-gp-mediated MDR cell lines. The parental OVCAR-8 human ovarian carcinoma cell line and the isogenic MDR NCI/ADR-RES subline, which stably expresses high levels of endogenous P-gp, were transfected to express the fluorescent proteins Discosoma sp. red fluorescent protein DsRed2 and enhanced green fluorescent protein, respectively. Co-culture conditions were defined, and fluorescent barcoding of each cell line allowed for the direct, simultaneous comparison of resistance to cytotoxic compounds in sensitive and MDR cell lines. We show that this assay system retains the phenotypes of the original lines and is suitable for multiplexing using confocal microscopy, flow cytometry, or laser scanning microplate cytometry in 1,536-well plates, enabling the high-throughput screening of large chemical libraries.
The ATP-binding cassette (ABC) family of membrane-spanning active efflux transporters are well known for their involvement in cancer chemotherapy multidrug resistance (MDR)1 and present a significant clinical challenge for drug design and development.2 This is typified by P-glycoprotein (P-gp) (ABCB1, MDR1), whose contribution to MDR has been characterized for a range of malignancies and correlates with poor clinical response to chemotherapy. A majority of anticancer drugs in current use are substrates for at least one ABC transporter.1 Endogenous expression of a number of ABC transporters is also responsible for drug absorption and distribution of small molecules, including at the blood–brain barrier where P-gp, ABCG2, and the MRP (ABCC) family collectively conspire to exclude a wide range of molecules from the brain, preventing their pharmacological utilization.3 ABCG2 has also been implicated as a marker of cancer stem cells,4 potentially explaining the intractability of a subpopulation of cancer cells such as occurs in the minimum residual disease phenomenon. We believe a strong understanding of the contributions of relevant transporters to absorption, distribution, and cellular accumulation of small molecules in both healthy and disease states will assist drug design and development as well as aid in the prediction of adverse drug effects through altered pharmacokinetics.
Despite the appreciation that ABC transporters play numerous clinical roles, many of the so-called “targeted therapies” entering the clinic are also substrates for ABC efflux transporters.5–8 Newer therapies may be more specific within the target class but suffer from the same resistance phenotype that plagues the oft-maligned “shotgun” cytotoxics they are intended to replace, and a need exists to elucidate the characteristics that render a compound susceptible to efflux by P-gp and other ABC transporters. Characterization of ABC transporter substrates appears mainly on a molecule-by-molecule basis, and no comprehensive screens have been reported; only about 300 P-gp substrates have been characterized to date.9–12 These likely represent the tip of the iceberg, as few small molecule libraries–some of which possess upwards of 100,000 compounds–have been screened for P-gp substrates. In silico approaches have approximated the fundamental chemical features of P-gp substrates ([N+O] ≥ 8, MW > 400, and pKa > 4),10 but these too are based on limited in vitro knowledge. We believe this paucity and fragmented reporting of experimental data must be addressed given the polyspecific nature of P-gp and the expectation that a significant portion of the pharmacopoeia are indeed substrates for ABC transporters.
Given the therapeutic potential of P-gp inhibitors to overcome MDR, several HTS approaches have been developed to identify new P-gp inhibitors,13–15 though published screening sets have largely been limited to small libraries used during assay validation, with one exception (~125,000 compounds, PubChem AID 1326). However, all such screens are alike in that they measure only P-gp inhibition and provide no information on substrate specificity. Inhibitor development thus far has been based on a limited number of lead structures with generally disappointing clinical efficacy,16,17 and the opportunity exists to identify new inhibitor classes with the potential to persevere where previous inhibitors have failed clinically. While our laboratory has recently reported a bioinformatics-guided approach to identify ABC transporter substrates by correlating drug response with transporter expression,18,19 a number of factors interfere with this predictive process, including the indeterminate and variable purity of compounds submitted to the National Cancer Institute (NCI) Developmental Therapeutics Program, the requirement for a range of gene expression across the NCI-60 cell line panel examined, the ability of mRNA levels to predict functional transporter activity in cells, and the relatively low number of statistically significant correlations. This approach unearthed agents that can selectively kill P-gp-expressing cells, a phenomenon termed “collateral sensitivity.” We believe this to be a promising pathway for resolving ABC transporter-mediated MDR, and while we are currently developing one such agent–the thiosemicarbazone NSC7330620–few “selective” chemotypes have thus far been identified, meriting further investigation through comprehensive HTS.
The long-term aim of this collaborative effort is to identify substrates, inhibitors, and selective agents against the ABC transporters P-gp (ABCB1), ABCG2, and ABCC1. Identifying ABC transporter substrates will prove useful for lead drug selection by other investigators (by discarding leads that are substrates) and allow for pharmacophoric modeling and characterization of molecular features across thousands of substrates, guiding the availability of lead drugs targeted to tissue expressing high levels of an ABC transporter.
We have developed a cell-based system using co-cultured cell lines capable of being optically discriminated via automated cell counting in a high-throughput system, using a characterized isogenic cell line pair with stable P-gp expression. This system allows sensitive and resistant cells to be treated with compounds under shared conditions in the same wells, improving sensitivity and accuracy of determinations of drug effects. We here report the development and validation of two fluorescent cell lines–the parental OVCAR-8 human ovarian carcinoma cell line expressing Discosoma sp. red (DsRed2) fluorescent protein and the isogenic MDR subline NCI/ADR-RES that stably expresses high levels of endogenous P-gp, expressing enhanced green fluorescent protein (EGFP) protein. These cell lines (OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP) express cytosolic red and green fluorescent proteins, respectively, yet retain the parental genotype and phenotype, rendering them a powerful tool for the study of P-gp-specific drug transport at the bench top. The dual-fluorescence system additionally provides a system suitable for the multiplexing of the two cells lines in a manner conducive to HTS of large chemical libraries using laser-scanning microplate cytometry.21,22
NSC73306 was initially obtained from the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, NCI.20 Previous studies have shown that the compound supplied was unchlorinated at m/z 327, despite the structure listed in the Developmental Therapeutics Program database. Accordingly, all experiments were conducted with this unchlorinated version of NSC73306. Doxorubicin, paclitaxel, mitoxantrone, cyclosporin A (CsA), calcein acetoxymethyl ester (CaAM), and dimethyl sulfoxide were purchased from Sigma-Aldrich (St. Louis, MO).
The fluorescent OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP lines were generated using the isogenic, ovarian carcinoma OVCAR-8 and NCI/ADR-RES cell line pair (NCI/ADR-RES was originally classified as a drug-selected variant of MCF7 and, incidentally, called MCF7/ADR, though subsequent bioinformatic analysis revealed it to be of OVCAR-8 origin23,24). Expression vectors encoding the fluorescent proteins DsRed2 and EGFP (pDsRed2-N1 and pEGFP-N1, respectively) were purchased from Clontech (Mountain View, CA) and transfected into the parental lines with Lipofectamine® 2000 (Invitogen Corp., Carlsbad, CA). Expression of the fluorescence plasmids was enforced by selection in G418 (Invitrogen). Stable transfectants were maintained in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal bovine serum (HyClone, Logan, UT), penicillin, streptomycin, l-glutamine (Invitrogen), and G418 (500 μg/ml and 200 μg/ml for OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP lines, respectively). Clones were preliminarily screened for plasmid expression by examining the intensity of green and red fluorescence (channels FL1 and FL2, respectively) by flow cytometry. The most intensely fluorescent clones were isolated and characterized to select those that behaved most like the untransfected, parental lines. The fluorescent lines reported here were the clones found to retain the properties of growth, ABC transporter expression, drug resistance, and efflux activity inherent in their nontransfected counterparts.
Growth curves were generated by seeding 5 × 104 cells on 35-mm-diameter plates and culturing in 2 ml of G418-free medium. A plate of each cell line was trypsinized every 24 h, and total cell counts were assessed by a Cellometer® Automatic Cell Counter (Nexcelcom, Lawrence, MA). Time points were measured in duplicate for each cell line, and total cell counts were generated until cessation of exponential growth at day 9. The effect of short-term G418 deprivation on retention of cell fluorescence intensity was qualitatively measured by confocal microscopy. OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells were sparsely plated onto glass coverslips in six-well culture dishes and cultured for 72 h in the absence and presence of G418 selection. Cells were then washed with phosphate-buffered saline (PBS), mounted onto glass slides without fixation, and immediately imaged at ×40 on a Zeiss (Jena, Germany) LSM 510 UV confocal microscope on their respective fluorescence channels at uniform thickness.
RNA was extracted from cultured cells (1 × 106 cells per extraction) using a RNeasy® mini kit (Qiagen, Inc., Valencia, CA) according to the manufacturer's instructions. Primers designed against human ABCB1, ABCC1, ABCG2, and plasma membrane Ca2+-ATPase 4 (PMCA4) cDNA sequences were generated and reverse-phase cartridge-purified by Lofstrand Labs Ltd. (Gaithersburg, MD). ABCB1 primers were 5′-GCCTGGCAGCTGGAAGACAAATAC (forward) and 5′-ATGGCCAAAATCACAAGGGTTAGC (reverse). ABCC1 primers were 5′-TGTGTGGGCAACTGCATCG (forward) and 5′-GTTGGTTTCCATTTCAGATGACATCCG (reverse). ABCG2 primers were 5′-CCGCGACAGCTTCCAATGACCT (forward) and 5′-GCCGAAGAGCTGCTGAGAACTGTA (reverse). ABC mRNA expression was normalized against expression of PMCA4, a previously characterized housekeeping gene exhibiting low variability among drug-sensitive and -resistant cancer cell lines.25 PMCA4 primers were 5′-ATCTGCATAGCTTACCGGGACT (forward) and 5′-TGCCAGCTTGTTTGCATTTGGCAATA (reverse). Real-time quantitative RT-PCR was performed on a LightCycler® II instrument (Roche Applied Science, Indianapolis, IN) using LightCycler RNA Master SYBR Green I master mix (Roche Applied Science). Reactions were set up according to the manufacturer's instructions; each reaction contained 300 ng of RNA. Crossing point (Cp) values were obtained from the LightCycler software (Roche Applied Science) using the second derivative maximum algorithm. Fold change in gene expression was calculated using the following equation:
Whole cell protein lysates were prepared by trypsinizing cultured cells (1 × 106 cells per extraction) and resuspending them in protease inhibitor cocktail containing buffer (10 mM Tris-HCl, 0.1% Triton X-100, 10 mM MgSO4, and 2 mM CaCl2), 2 mM dithiothreitol, 1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 20 μg/μL micrococcal endonuclease (DNase I), and 1% aprotinin. Cell suspensions were then incubated at room temperature for 5 min and snap-frozen in dry ice. Upon thawing, lysates were sonicated three times each for 1 min in a sonicator bath, and 5× sodium dodecyl sulfate buffer was added prior to loading. Equally loaded lysates were resolved on 3–8% gradient NuPAGE® Novex® Tris-acetate gels (Invitrogen) and transferred to nitrocellulose membranes using the iBlot gel transfer system (Invitrogen). Membranes were incubated overnight with either mouse monoclonal anti-P-gp C21926 primary antibody at 1:10,000 dilution or mouse anti-human glyceraldehyde-3-phosphate dehydrogenase at 1:5,000 dilution and subsequently incubated with horseradish peroxidase-conjugated goat anti-mouse immunoglobulin G (Cell Signaling Technology, Inc., Danvers, MA) at 1:10,000 dilution for 1 h. Proteins were visualized with ECL Western blotting detection reagent (GE Healthcare Bio-Sciences Corp., Piscataway, NJ) on Amersham Hyperfilm™ ECL film (GE Healthcare Bio-Sciences Corp.).
Cell survival was measured by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) viability assay. Cells were seeded in 100 μl of medium at a density of 5,000 cells per well in 96-well plates and incubated at 37°C in humidified 5% CO2 for 24 h. Serially diluted drugs were added in an additional 100 μl of medium to give the intended final concentrations. Cells were then incubated an additional 72 h, and the MTT assay was performed according to the manufacturer's instructions (Molecular Probes, Eugene, OR). Absorbance values were determined at 570 nm on a Spectra Max 250 spectrophotometer (Molecular Devices, Sunnyvale, CA). All MTT assays were performed three times in triplicate. The 50% inhibitory concentration (IC50) values were defined as the drug concentrations required to reduce cell viability to 50% of the untreated control well and are reported here as such. Resistance ratio (RR) values were reported as a measure of relative drug sensitivity and were defined for each drug as the IC50 of the high-P-gp-expressing line divided by the IC50 of the low-P-gp-expressing line.
Trypsinized cells were resuspended in fresh medium. Cells (5 × 105 per condition) were then centrifuged and brought up in Iscove's modified Dulbecco's medium (IMDM) (Invitrogen) containing 5% fetal bovine serum (HyClone). Cell-only controls were incubated for 10 min in drug-free IMDM. CaAM-treated cells were incubated for 10 min in IMDM containing 0.25 μM CaAM. CsA-treated cells were preincubated for 10 min in IMDM containing 5 μM CsA, spun down, aspirated, and incubated an additional 10 min in IMDM containing 5 μM CsA and 0.25 μM CaAM. All cells were then spun down, resuspended in 0.1% bovine serum albumin in PBS, and placed on ice prior to flow cytometry analysis. Green fluorescence intensity was measured using a FACSCalibur™ flow cytometer equipped with a 488 nm argon laser (Becton Dickinson Biosciences, San Jose, CA) for a total of 10,000 events per sample. Relative CaAM accumulation was calculated from geometric mean fluorescence intensities as determined by FlowJo flow cytometry analysis software (Tree Star, Inc., Ashland, OR).
Equal numbers of fluorescent OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells were seeded on glass coverslips in six-well dishes in drug-free medium and incubated at 37°C for 24 h to allow cells to attach. Co-cultures were then dosed with concentrations of either paclitaxel (Taxol®; Sigma-Aldrich, St. Louis, MO) (0, 50, and 100 nM) or NSC73306 (0, 2, and 5 μM) and incubated for an additional 72 h prior to imaging. Coverslips were removed from medium, washed with 1× PBS, inverted onto Superfrost Plus glass slides (Thermo Fisher Scientific Inc., Waltham, MA), and sealed. Slides were then immediately imaged on a Zeiss LSM 510 UV confocal microscope on both green and red fluorescent channels at uniform thickness. Fields shown in images are at ×40 and are representative of the entire co-cultures.
Equal numbers of OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells were seeded on glass coverslips in six-well dishes in drug-free medium and incubated at 37°C for 48 h prior to dosing. Cultures were then incubated at 37°C in medium containing 40 μM mitoxantrone for 45 min, followed by drug-free medium for an additional 30 min. Inhibitor-treated cells were preincubated at 37°C in medium containing 5 μM CsA for 10 min. They were then incubated for 45 min in medium containing 5 μM CsA and 40 μM mitoxantrone, followed by medium containing only 5 μM CsA for an additional 30 min. Coverslips were removed from medium, washed with 1× PBS, inverted onto Superfrost Plus glass slides, and sealed. Slides were then immediately imaged on a Zeiss LSM 510 UV confocal microscope on green, red, and deep red fluorescent channels at uniform thickness. Merged images of red (DsRed2) and purple (mitoxantrone) fluorescence appear as magenta, while green (EGFP) and purple fluorescent images appear white upon merging. Fields shown in images are at ×40 and are representative of the entire co-cultures.
Cells were trypsinized and mixed at known concentrations (100% OVCAR-8-DsRed2, 75% OVCAR-8-DsRed2–25% NCI/ADR-RES-EGFP, 50% OVCAR-8-DsRed2–50% NCI/ADR-RES-EGFP, 25% OVCAR-8-DsRed2–75% NCI/ADR-RES-EGFP, and 100% NCI/ADR-RES-EGFP). Cells were then spun down, washed once with IMDM supplemented with 5% fetal bovine serum, spun again, and resuspended in 0.1% bovine serum albumin in PBS. Green and red fluorescence intensities were measured using a FACSCalibur flow cytometer equipped with a 488 nm argon laser for a total of 10,000 events per sample. Gating analysis was performed using FlowJo flow cytometry analysis software. Gates were set for the two fluorescent populations on a forward scatter (FSC) versus side scatter (SSC) density plot. These subpopulations were then backgated onto histograms of both green (FL1) and red (FL2) fluorescence to show that cell size and granularity are different for both fluorescent lines and that these parameters enable predictive discrimination by scatter, in addition to fluorescence. The percentages of fluorescent cells present within each transfected cell line were determined by performing gating analysis on the 100% OVCAR-8-DsRed2 and 100% NCI/ADR-RES-EGFP cell populations. Total cell counts of each line were determined by FSC and SSC plot data, and numbers of fluorescent cells were subsequently determined by plotting the FSC of each line against its respective fluorescent channel and autogating the densest cluster of cells using FlowJo flow cytometry analysis software.
Validation of red/green fluorescence discrimination by fluorescent cytometer (see Fig. 5A–G) was performed by plating OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells in mono- and co-culture (1,000 cells each per well) onto black 96-well clear-bottom plates (see Fig. 5A) and allowing them to attach for 24 h. Rows C and G were plated with red and green cells in monoculture (to define red and green populations), whereas rows D–F were plated with equal numbers of red and green cells in co-culture. Cells were then dosed with serially diluted Taxol (nine dose points starting with 500 nM, 1:2 dilutions) and incubated for an additional 48 h. Plates were then read on an Acumen eX3® microplate cytometer (TTP LabTech Ltd., Melbourn, UK) using a 488 nm laser, and fluorescent cells were counted on their respective channels. EGFP and DsRed2 fluorescence was detected using 500–530 nm and 565–605 nm bandpass filters, respectively. The Acumen software assigns a color to each well based on the range of cell counts observed on each plate (bright green corresponding to the highest value and black to the lowest). Red and green cells are optically “sorted” by the cytometer, assigned a false color based on their individual fluorescence characteristics, and counted accordingly.
To distinguish whole cells from lint and cellular debris, cell counts were restricted to objects measuring 10–100 μm in both width and depth (excluded objects are labeled as light blue in well images). Dead and apoptotic cells were observed to have significantly higher fluorescence intensities than viable cells, most likely because of decreased cell area (but unchanged levels of cytosolic fluorescent protein) upon rounding up. As such, nonviable cells were excluded from counts by gating out these outlying, highly fluorescent populations. Viability analysis using Sytox Orange® cell-impermeant vital dye (Invitrogen) gave a nearly identical approximation of nonviable cells (Appendix Fig. A1), suggesting gating by peak fluorescence intensity is an accurate method of measuring viability in the fluorescent lines. Cells exhibiting peak fluorescence intensities within these gates (green cells, 1,257.08–9,988.30 fluorescence units; red cells, 2,650.79–11,423.01 fluorescence units) were counted as “viable” (see Fig. 5D and E) and subsequently identified as either red or green based on their ratio of green to red fluorescence intensity (green cells, 0.57–3.00; red cells, −2.00 to 0.57); cells outside these thresholds were excluded. Cell counts, as well as individual red and green fluorescence intensities, were then exported and quantified.
Taxol concentration–response curves (CRCs) were generated by plating OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells in both monoculture (4,000 cells per well) and co-culture (1,000 cells each per well) in 96-well plates and allowing them to attach for 24 h. Cells were then dosed with serially diluted taxol (nine dose points, 1:2 dilutions) and incubated for an additional 48 h. OVCAR-8-DsRed2 cells were dosed with a maximum Taxol concentration of 500 nM, while NCI/ADR-RES-EGFP cells, because of their inherent levels of drug resistance, were dosed with a maximum concentration of 50 μM. Co-cultures were dosed with both ranges of Taxol concentration to achieve full CRCs for both lines. Plates were then read on an Acumen eX3 microplate cytometer, and counts were quantified as described above (using identical gates). Cytotoxicity was validated by MTT assay (as described earlier) using fluorescent cells grown separately and dosed under identical conditions.
To determine what effect, if any, our co-culture conditions had on drug toxicity estimations at the 1,536-well scale, Taxol CRCs were generated by plating OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells in both mono- (200 cells per well) and co-culture (200 cells each per well) in 1,536-well plates and allowing them to attach for 2 h. Cells were then dosed with serially diluted Taxol (16 dose points, 1:2 dilutions) and incubated for an additional 46 h. Cell dispensing was performed using the solenoid-based dispenser (Flying Reagent Dispenser [FRD]27) from Aurora BioSciences (San Diego, CA). The assay plates used were 1,536-well Aurora Lobase black clear-bottom plates (catalog number 2932628). Compounds were added using a Kalypsys pin-tool equipped with 1,536 10-nl slotted pins (VPN Scientific, San Diego29); 23 nl of a dimethyl sulfoxide solution was transferred to the assay plates by pin-tool. OVCAR-8-DsRed2 cells were dosed with a maximum taxol concentration of 657.25 nM, while NCI/ADR-RES-EGFP cells were dosed with a maximum concentration of 65.73 μM. Co-cultures were dosed with an intermediate range of Taxol (maximum concentration, 6.57 μM) to achieve partial CRCs for both lines. Plates were then read on an Acumen eX3 microplate cytometer, and counts were quantified as described above.
In developing the two fluorescent cell lines amenable to high-throughput assay development, it was essential to ensure that the final transfected lines retained the characteristics and ABC transporter expression of their parental lines. Following transfection and selection in G418, both OVCAR-8-DsRed2 and the MDR NCI/ADR-RES-EGFP cells demonstrated strong fluorescence when grown in G418-free medium for 96 h (Fig. 1A), and this expression persisted after 4 weeks (data not shown). The growth dynamics of both cell lines were found to closely match their nontransfected parental cells over the first 6 days (Fig. 1B), although the NCI/ADR-RES-EGFP line was observed to grow at a lower rate than its parental counterpart beyond 6 days. However, as most applicable cell-based assays utilize timelines of 72 h or less (<48 h for most high-throughput applications), the lines are suitable for the desired dual-color fluorescence-based assays. Morphologically, the NCI/ADR-RES-EGFP line exhibits a somewhat larger cell size than its drug-sensitive counterpart; however, this trend is seen in the parental cell line pair as well (Appendix Fig. A2)–though to a less pronounced degree–and may contribute to the differential long-term growth rates.
The expression patterns of MDR-implicated ABC transporters were measured in parental and transfected cell lines, by both RT-PCR (Fig. 1C) and Western blot analysis (Fig. 1D). Fluorescent protein expression did not greatly affect the expression of ABCB1 transcript; the fluorescent pair retained the large difference in ABCB1 expression, with NCI/ADR-RES-EGFP cells expressing over 1,000-fold greater levels of ABCB1 transcript than the OVCAR-8-DsRed2 cells (Fig. 1C). The mRNA transcript expression of the ABC transporters ABCC1 (MRP1) and ABCG2 was also measured to confirm that the transfection and clone selection process did not result in changes between the lines that would interfere with the ABCB1-mediated drug resistance phenotype (Fig. 1C). Indeed, ABCG2 transcript was undetectable in both parental and transfected cell lines (normalized to 1 in Fig. 2A), and ABCC1 was detectable at very low levels, being expressed only 2.65-fold less in NCI/ADR-RES-EGFP cells relative to OVCAR-8-DsRed2. Western blotting confirmed that parental levels of B1 transporter protein in these cells were retained (Fig. 1D).
To confirm retention of the MDR phenotype, drug sensitivities to the P-gp substrate doxorubicin and the MDR1-selective agent NSC73306 were assessed (Fig. 2). NSC73306 demonstrated MDR1 selectivity against both parental (RR = 0.57) and transfected (RR = 0.36) NCI/ADR-RES cell lines, and likewise, resistance to doxorubicin was retained (parental RR = 66.43, transfected RR = 61.23). The distinct fluorescence emission spectra of the OVCAR-8-DsRed2/NCI/ADR-RES-EGFP pair afford optical discrimination by confocal microscopy of the two lines when grown in co-culture (Fig. 2B and C), allowing for the simultaneous analysis of both lines exposed to shared conditions during the course of an assay. As such, drug toxicity phenotypes can be dynamically observed in this co-culture system by confocal microscopy. The P-gp substrate Taxol demonstrated dramatic dose-dependent toxicity towards the red OVCAR-8-DsRed2 cells, while the MDR NCI/ADR-RES-EGFP cells persisted in 100 nM Taxol (Fig. 2B). In an inverse fashion, NSC73306 selectively reduced the population of green P-gp-expressing cells relative to the red “drug-sensitive” cells (Fig. 2C).
P-gp-mediated efflux was confirmed by assessing accumulation of CaAM, a fluorescent P-gp substrate (Fig. 3). Accumulation of CaAM was virtually unaffected (3.3% increase) by addition of the P-gp inhibitor CsA in OVCAR-8-DsRed2 cells (Fig. 3A). In contrast, the P-gp-expressing NCI/ADR-RES-EGFP line displayed a 334% increase in CaAM accumulation upon addition of inhibitor (Fig. 3B). While CaAM (excitation, 492 nm; emission, 513 nm) fluoresces in the same region as EGFP, accumulation differences (observed as shifts in fluorescence intensity) can still be readily measured because of the higher fluorescence intensity of CaAM relative to the “background” fluorescence conferred by the NCI/ADR-RES-EGFP cells.
Confocal microscopy was employed to observe live cell accumulation and P-gp-mediated efflux of mitoxantrone in the co-culture system (Fig. 3C and D). When cells were co-cultured with the fluorescent P-gp substrate mitoxantrone, strong accumulation was observed only in the OVCAR-8-DsRed2 line (Fig. 3C, bottom right, shows mitoxantrone/red cell merge as magenta). Co-incubation with the P-gp inhibitor CsA (10 μM) resulted in strong mitoxantrone accumulation in both the OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells (Fig. 3D, bottom right, shows mitoxantrone/green cell merge as white), indicating that the effect of P-gp inhibition can be observed directly in a mixed live-cell system by detecting the accumulation of a fluorescent substrate by virtue of fluorophore intensity.
The cells are readily detected by flow cytometry without the use of cell dyes or fluorophore-tagged antibodies (Fig. 4); the NCI/ADR-RES-EGFP cells can be detected in the FL1 (green) channel (Fig. 4A), whereas the OVCAR-8-DsRed2 cells can be detected in the FL2 (red) channel (Fig. 4B). Mixed populations of red and green cells can be simultaneously quantified; OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells mixed at known proportions yield accurate histogram populations whose red and green events appear proportional to the starting ratio of cells. Figure 4A also reveals that optical spillover from the red cells into the green channel can be observed as a lower intensity population in the FL1 histogram. An unanticipated benefit of this emission spillover is that heterogeneous mixtures of the two cell lines can be simultaneously quantified in the FL1 channel, presenting opportunities for flow-based MDR assays that can be coupled with substrates, antibodies against a range of markers, or viability stains. The FSC and SSC scatter histograms show the cell lines cluster into two distinct populations, possibly because of subtle morphological differences between the lines (Fig. 4C). These gated populations appear identical to the original ungated populations on both FL1 (Fig. 4D) and FL2 (Fig. 4E) channels when overlaid, suggesting that the size and granularity differences between the lines can additionally be used to distinguish between them in co-culture.
The homogeneous populations of both lines were also examined to determine what percentage of the total cell populations was fluorescent. Comparison between scatter and fluorescence channel data determined that 98.67% and 98.84% of OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cell populations, respectively, were fluorescent in their respective channels (Appendix Fig. A3A–C). Additionally, representative confocal images of fluorescent co-cultures stained with 4′,6-diamidino-2-phenylindole nuclear stain (Appendix Fig. A3D) demonstrate highly fluorescent red and green cell populations with minimal presence of nonviable or viable, nonfluorescent cells.
Given that the fluorescent cell line pairs replicate their nontransfected parental lines, we finally wanted to ensure that our co-culture system was compatible with HTS applications. The Acumen eX3 microplate cytometer provides the ability to enumerate fluorescent characteristics of individual cells present in microtiter plates (96-, 384-, and 1,536-well formats),22 providing a system where cytometric analysis can be performed using three excitation lasers (405/488/633 nm) and four photomultiplier tube detectors. This provides an optimal balance between resolution and microplate scan time. At the National Institutes of Health Chemical Genomics Center, two Acumen eX3 cytometers have been integrated into a Kalypsys automated screening system where chemical libraries can be tested with throughputs as fast as five samples per second.30 In addition to determining object dimensions, the cytometer measures fluorescence characteristics of individual cells that, when coupled with its ability to define cell populations based on any number of measurable characteristics, provide detailed population statistics for each scanned well. To validate the dual-color cell pair on this system, cells were co-cultured in black clear-bottom 96-well plates, and the epifluorescence was collected from below with a size threshold of 10–100 μm (width and depth). The cytometer was able to identify individual cells on the well bottom and accurately determine object (cell) fluorescence (Fig. 5A–C and G). Examples of plate “heat maps” denoting relative cell counts on a single plate on both red (OVCAR-8-DsRed2, Fig. 5B) and green (NCI/ADR-RES-EGFP, Fig. 5C) channels demonstrate the cell quantification process and show that cell counts are consistent among red and green cells grown in both mono- and co-culture conditions.
Traditional experimental design for drug activity assays employs viability stains to exclude dead or apoptotic cells that may retain fluorescence. When counting cells by microplate cytometry, however, a subpopulation of cells with brighter fluorescence intensity (found to consist of rounded apoptotic and dead cells [Fig. 5D and E, denoted by*]) was observed in drug-treated wells. This brighter pool of cells, whose increased fluorescence intensity was likely due to the high amount of fluorescent protein present in the comparatively smaller area of the apoptotic cells, was “gated” out during cell counting (Fig. 5D and E) and resulted in excellent consensus with viability measurements using fluorescent cell viability dyes (Appendix Fig. A1). Cell populations within the gating range were found to decrease as Taxol concentration increased, while, conversely, the number of highly fluorescent “apoptotic” cells increased considerably. CRCs with Taxol against both individual and co-cultured cells were determined by counting total cells with the microplate cytometer and comparing to an MTT assay conducted under identical conditions (Fig. 5G). Separate color and co-culture conditions as measured by cytometry correlated closely, with Taxol RRs of 1,085 and 829, respectively, both of which approximated the MTT-derived MDR ratio of 1,018 (overlaid in Fig. 5G). Thus the OVCAR-8-DsRed2-NCI/ADR-RES-EGFP co-culture system measures cytotoxicity as accurately as the standard MTT assay performed in the laboratory, while simplifying the assay dynamics and increasing throughput.
To validate the assay for quantitative HTS we further tested the performance of the assay system in 1,536-well microtiter plates. Assay factors were optimized in terms of plating density, total assay time, and dosing timeline to maximize toxicity effects and minimize complications common to the 1,536-well scale, such as evaporation and dispenser variation (data not shown). The final optimized 1,536-well protocol is listed in Table 1. Not only did both lines remain intensely fluorescent and easily distinguishable at the 1,536-scale, but Taxol CRCs of OVCAR-8-DsRed2 and NCI/ADR-RES-EGFP cells grown in both mono- and co-culture on 1,536-well plates displayed the expected shift in IC50 (RR = 1,453) irrespective of culture conditions (overlaid in Fig. 5H), comparable to those determined by Acumen analysis at the 96-well scale (Fig. 5G). Additionally, Taxol CRCs demonstrated the co-culture assay to be reproducible and precise, with Z′-factors31 of 0.50 ± 0.15 and 0.52 ± 0.06 for red and green fluorescent cells, respectively, grown in co-culture. Thus miniaturization of the co-culture system was found to have a negligible effect on the discrimination and toxicity estimation of both red and green cell lines.
The two fluorescent cell lines generated here–a red fluorescent drug-sensitive cell line and a green fluorescent MDR cell line expressing high levels of P-gp–were shown to express high levels of DsRed2 and EGFP, respectively, and to maintain expression in the absence of selection pressure for long periods (Fig. 1). Furthermore, the two cell lines can be grown in co-culture and easily visually discriminated from one another by confocal microscopy (Figs. 2 and 3),3), flow cytometry (Figs. 3 and 4),4), or laser scanning microplate cytometry (Fig. 5). Expression of P-gp at the mRNA transcript and protein levels was unaffected relative to the parental lines (Fig. 1), and the drug response phenotypes to doxorubicin and NSC73306 were retained (Fig. 2), irrespective of the observed morphological differences common to many such drug-sensitive/drug-resistant cell pairs. Efflux assays using the substrates CaAM and mitoxantrone (Fig. 3) confirmed that P-gp is functional in NCI/ADR-RES-EGFP cells and unaffected by the transfection process.
Validation of the utility of these cell lines was followed by experiments to confirm that co-cultures of the two fluorescent lines were amenable to the high-throughput microplate cytometer platform used to optically count and bin cells. The Acumen reader was able to readily discriminate between the two cell lines based on their optical properties, count cell numbers in individual wells across 96-well and 1,536-well plates, and enumerate viable cells based on gated cell fluorescence intensity, producing comparable CRCs and RRs from both separate and co-cultures at well densities as high as 1,536 wells (Fig. 5 and Table 1).32
In considering the appropriate cell line pair to develop for this MDR system, it was desirable that the cell line pair was isogenic and that the P-gp-expressing cell line was drug-selected rather than transfected, with stable P-gp expression in the absence of continual drug selection, given the scale of cell line maintenance required. The OVCAR-8 ovarian carcinoma cell line and its drug-selected isogenic subline NCI/ADR-RES satisfy these requirements and provide the added advantage of being included in the NCI-60 cell line panel, for which a great deal of gene expression, drug sensitivity, and bioinformatics information is available,33,34 allowing cross-correlation of activity profiles from our high-throughput drug discovery screen with the performance of the NCI drug screening program. While other P-gp assays have employed fluorescent membrane-permeant cell tracker stains for short-term studies,35 we have found that some such stains are in fact substrates for P-gp (data not shown). These stains could serve as competitive substrates, reducing fluorescence and interfering with the cellular accumulation and efflux of drugs introduced to cells, compromising the sensitivity of both cytotoxicity and P-gp inhibition assays. Thus we were prompted to develop a more permanent fluorescence system.
EGFP (emission, 509 nm) and DsRed2 (emission, 583 nm) fluorescent proteins were selected to ensure a spectral window was available for the inclusion of a third fluorophore to allow the incorporation of a P-gp inhibition screen into the assay (Fig. 6) using the principles demonstrated by confocal microscopy (Fig. 3C and D). An appropriate fluorophore would be required to fluoresce at a wavelength ≥650 nm and possess a large Stokes shift (given excitation at 488 nm), such as 7-aminoactinomycin D (emission 650 nm),36 LDS 751 (emission 712 nm),37,38 and mitoxantrone (684 nm).39 When using a fluorescence-based reporter, it is important to consider that compounds within libraries may themselves fluoresce and interfere with assay performance. Simeonov et al.40 have recently shown that more than 5% of samples from a 70,000 compound library (available within PubChem, AIDs 587, 588, and 590–594) showed fluorescence similar to a common ultraviolet-active probe, and these compounds have been annotated, enabling cross-correlation and ensuring false-positive hits are minimized in our own screen.
In addition to the intended use of these cell lines, we have identified here a number of cell culture-based applications that enhance current assay systems and extend beyond the HTS realm, particularly through the ability to test sensitive and MDR cells in co-culture. A majority of the assays used to probe P-gp substrate and inhibitor specificity are cell-based and often optically quantified by fluorescence detection, rendering this assay system “backward compatible” with standard applications, while providing the unique ability to recover mixed populations by flow cytometry from culture or animal models. Real-time accumulation and efflux of fluorescent substrates in co-culture using confocal microscopy is also available using these cultures, and the spectral window available (including those at other excitation wavelengths) allows for the incorporation of multiple fluorophores.
This new method for color discrimination in co-culture provides an exquisitely controlled system for probing MDR, ideally leading to a greater understanding of its underlying mechanisms through the development of pharmacophores, structure–activity relationships, and the chemical homology of HTS hits with known mechanisms/targets in other screens. This system's utility is not limited to the study of P-gp-mediated drug resistance; the field of drug transporters is one that frequently utilizes such isogenic cell line pairs and so is naturally conducive for application of a similar co-culture system. Facilitated uptake drug transporters, such as the solute carrier group of transport proteins, as well as any of the other ABC efflux transporters could be easily adapted to profile their respective drug transport activities. Though transport-mediated shifts in IC50 values are theoretically linked to the fold difference in transporter expression between two such lines, reduced substrate accumulation (and hence, reduced toxicity) has been observed in virtually all isogenic sensitive/MDR pairs (which themselves have widely varying degrees of transporter expression), suggesting that this HTS approach would remain applicable for cell pairs with less marked differences in expression or for those expressing less efficacious transporters. Indeed, any cell line pair exhibiting differential expression of a specific gene target (selected or transfected) could be similarly labeled and, in turn, discriminated in a HTS co-culture system to unearth small molecules whose toxicities are potentiated or antagonized by expression of the target. As such, it is hoped that the generalities of this new approach will be applicable for probing a range of phenotypes.
The authors gratefully acknowledge the assistance of Susan H. Garfield and Poonam Mannan of the National Cancer Institute Confocal Microscopy Core Facility, Drs. Natasha Thorne and Noel Southall of the National Institutes of Health Chemical Genomics Center, and George Leiman for his editorial assistance. This research was supported by the Intramural Research Program of the National Institutes of Health. K.R.B. was supported by a Center for Cancer Research/NCI training fellowship.
No competing financial interests exist.