BioTechniques for Preclinical Development 44 ı No. ... University Health Network, Toronto, ON, Canada ... (D) A custom software program was created to convert the list output into a matrix containing the number of colonies per well on the.
Short Technical Reports Increased efficiency for performing colony formation assays in 96-well plates: novel applications to combination therapies and high-throughput screening David Katz1,2, Emma Ito1,2, Ken S. Lau1,3, Joseph D. Mocanu1,2, Carlo Bastianutto1,2, Aaron D. Schimmer1,2,4, and Fei-Fei Liu1,2,4 1University
of Toronto, 2Ontario Cancer Institute, University Health Network, 3Samuel Lunenfeld Research Institute, Mount Sinai Hospital, and 4Princess Margaret Hospital, University Health Network, Toronto, ON, Canada BioTechniques 44:Pix-Pxiv (February 2008) doi 10.2144/000112757
The colony formation assay (CFA) is the gold standard for measuring the effects of cytotoxic agents on cancer cells in vitro; however, in its traditional 6-well format, it is a time-consuming assay, particularly when evaluating combination therapies. In the interest of increased efficiency, the 6-well CFA was converted to a 96-well format using an automated colony counting algorithm. The 96-well CFA was validated using ionizing radiation therapy on the FaDu (human hypopharyngeal squamous cell) and A549 (human lung) cancer cell lines. Its ability to evaluate combination therapies was investigated by the generation of doseresponse curves for the combination of cisplatin and radiation therapy on FaDu and A549 cells. The 96-well CFA was then transferred to a robotic platform for evaluating its potential as a high-throughput screening (HTS) readout. The LOPAC1280 library was screened against FaDu cells, and eight putative hits were identified. Using the 96-well CFA to validate the eight putative chemicals, six of the eight were confirmed, resulting in a positive hit rate of 75%. These data indicate that the 96-well CFA can be adopted as an efficient alternative assay to the 6-well CFA in evaluating single and combination therapies in vitro, providing a possible readout that could be used on a HTS platform.
INTRODUCTION The colony formation assay (CFA) has been the gold standard for determining the effects of ionizing radiation therapy on in vitro cellular systems since first described by Puck and Marcus in 1956 (1). Despite being the gold standard, the CFA can be time consuming when counting the number of colonies manually under the microscope. For this reason, many different assays have been used in lieu of the CFA in order to assess the effects of cytotoxic agents on cancer cell growth in vitro. While many of these techniques are able to detect specific cellular processes such as apoptosis (2), proliferation (3), or senescence (4), the CFA is the only assay that monitors a cancer cell’s ability to produce a
viable colony after treatment. Unlike most other assays, the CFA is unbiased to the mode of cell death. It is able to detect the cytotoxic effect of an agent, regardless of mechanism, as long as the agent affects the cell’s reproductive ability to form progenies. Most of the advancements made in cancer therapy in recent years have resulted from the combination of previous individual modalities, such as radiation and chemotherapy. Chemotherapy with such agents as cisplatin, 5-fluorouracil, doxorubicin, temozolomide, or cetuximab have been combined with radiotherapy for the treatment of head and neck cancer (5), non-small cell lung carcinoma (6), glioblastoma (7), cervix (8), and bladder cancers (9), to name a few. Initially, discovery of such combina-
BioTechniques for Preclinical Development 44 ı No. 2 | 2008
tions was conducted in the laboratory using tissue culture as the primary testing platform. To assess novel potential combinations, the majority of experiments were performed in a traditional 6-well tissue culture plate CFA, with each plate representing a different combination of two potential treatments. Commonly, such experiments test up to six different doses of radiation (0–10 Gy) and up to seven different doses of drug (10). Using the traditional CFA, this would result in using 42 individual plates. Beyond the technical challenges of conducting an experiment with 42 individual plates, a significant amount of time would be required to manually count the colonies on all such plates. This underscores the need for a more modern approach to this assay. With recent improvements in fluorescent probes and highcontent microscopy, it is now possible to adapt the traditional method to a more efficient approach using a 96well plate and an automated colony counting algorithm. While this article describes this novel approach in evaluating combination therapies including radiation and chemical treatment, it can be extended to any combination of treatments including two different chemical treatments administered concurrently. Another area of advancement in cancer research is the use of highthroughput screening (HTS) for the identification of novel anticancer compounds. There are two basic approaches to HTS. The forward chemical biology approach identifies a phenotype of interest, after treatment with chemical compounds, and the mechanism is subsequently elucidated (11). The reverse chemical biology approach identifies a target, and hits are selected as compounds that modulate that specified molecule (11). Both forward and reverse approaches have yielded clinically useful anticancer drugs. The forward chemical biology approach is illustrated by the use of paclitaxel, which was shown to be effective against tumors long before it was identified to target microtubules (12). The reverse chemical biology approach has been highly successful in the identification www.biotechniques.com ı BioTechniques ı ix
Short Technical Reports
A
B
D
C
E
Figure 1. Representation of the 96-well colony formation assay (CFA). (A) Fluorescence image obtained using the INCell Analyzer 1000 (GE Healthcare, Buckinghamshire, England) at 4× objective of FaDu cells stained with Cell Tracker Orange CMTMR and Hoescht 33342 (both from Invitrogen, Carlsbad, CA, USA). (B) Colonies are defined using the Developer Toolbox software (GE Healthcare) as an adjacent set of cytoplasmic signals (blue outline) containing stained nuclei (green outline). (C) The readout from Developer Toolbox is an Excel worksheet that lists all colonies with their corresponding number of nuclei within a specific well. (D) A custom software program was created to convert the list output into a matrix containing the number of colonies per well on the original plate, filtered to include only colonies with ≥6 and ≤350 nuclei. (E) The matrix data are then plotted as a clonogenic survival curve.
of the bcr-abl inhibitor, Imatinib (Novartis, Basel, Switzerland), used for treatment of chronic myelogenous leukemia (13), and the src-abl kinase inhibitor, Dasatinib (Bristol-Myers Squibb, New York, NY), used to treat imatinib-resistant chronic myelogenous leukemia (14). In the current study, we have adapted the 6-well CFA to a more efficient 96-well CFA that will allow for rapid analysis of combination therapies and opens the potential for high-throughput drug screening using the CFA as the readout.
x ı BioTechniques ı www.biotechniques.com
MATERIALS AND METHODS Cell Culture Human head and neck squamous cell carcinoma FaDu cells were cultured in MEM-F15 medium containing 10% fetal bovine serum (FBS), 1 mM pyruvate, 1.5 g/L sodium bicarbonate, 100 mg/L penicillin, and 100 mg/L streptomycin. Human lung adenocarcinoma A549 cells were cultured in RPMI media containing 10% FBS, 100 mg/L penicillin, and 100 mg/L streptomycin.
6-Well Colony Formation Assay Cells were trypsinized and plated in 6-well dishes at different densities depending on the potency of the treatments (from 50 to 104 cells/well). Cells were allowed to attach overnight and then exposed to radiation therapy (0–16 Gy) or chemical treatment at the corresponding dilution. Forty-eight hours after chemical treatment, the media was replaced with fresh media, and the plates were incubated at 37°C. Seven to eleven days later, the cells were fixed and stained with 10% methylene blue in 70% ethanol. The number of colonies, defined as >50 cells/colony
BioTechniques for Preclinical Development Vol. 44 ı No. 2 | 2008
Short Technical Reports
were counted, and the surviving fraction was calculated as the ratio of the number of colonies in the treated sample to the number of colonies in the untreated sample. Triplicate wells were set up for each condition. 96-Well Colony Formation Assay Cells were trypsinized and plated in 96-well plates at different densities depending on the stringency of the treatments (from 50 to 2500 cells/ well). The cells were allowed to attach overnight and then exposed to radiation therapy (0–16 Gy) or chemical treatment at the corresponding dilution. Radiation therapy was administered 24 h after addition of the chemical, and the media was replaced 48 h later. Six days after seeding, cells were fixed in 3.7% formaldehyde at room temperature for 15 min, followed by staining with 10 μM Hoescht 33342 and 10 μM Cell Tracker Orange CMTMR (both from Invitrogen, Carlsbad, CA, USA) in serum-free media, and incubated at 37°C for 30 min. After staining, the wells were scanned (four fields/well) at 4× objective using the INCell Analyzer 1000 (GE Healthcare, Buckinghamshire, England). The excitation and emission filters used for Hoescht 33342 and Cell Tracker Orange were 360 nm/460 nm and 535 nm/620 nm, respectively. Colonies were recognized by an algorithm setup on the Developer Toolbox software (GE Healthcare) using the overlay of the blue and red images. The algorithm’s output indicated the location of the colony within the plate and the number of blue nuclei contained within the colony. The number of colonies per well was then filtered by a custom computer program (www.uhnres. utoronto.ca/labs/liu/ACC) to include all colonies with ≥6 cells/colony and ≤350 cells/colony. The surviving fraction was then calculated by dividing the number of colonies by the number of cells seeded, multiplied by 2.217 (to account for the proportion of the well that was scanned), then divided by the surviving fraction of untreated cells. Three to eight replicates were set up per treatment.
Table 1. Confirmed Hits in the LOPAC1280 Library Compound
B-Score Value
Emetine Dihydrochloride Idarubicin Mitoxantrone Ouabain Vincristine Sulfate Vinblastine Sulfate Salt
-4.479 -4.781 -4.792 -4.648 -5.367 -5.754
Summary of confirmed hits in the LOPAC1280 library screened at 0.5 μM final concentration on FaDu cells using the 96-well colony formation assay (CFA) as the read-out.
A
B
Figure 2. Reproducibility of a 96-well colony formation assay (CFA) compared with a traditional 6-well CFA. (A) CFAs were performed on FaDu cells using the 96-well and the 6-well assays. (B) CFAs were performed on A549 cells using the 96-well and the 6-well assays. Each datum represents the mean ± sem from three independent experiments for 6-well data and two independent experiments for 96-well data.
High-throughput Screening FaDu cells were seeded in 96-well plates at a dilution of 250 cells/well in 100 μL media using a Biomek FX liquid handler (Beckman Coulter, Fullerton, CA, USA). After allowing the cells to attach overnight, the LOPAC1280 library of compounds (Sigma-Aldrich, St. Louis, MO, USA) was added to the cells at a final concentration of 0.5 μM. On each plate, column 1 was the vehicle control and column 12 was 0.5 μM cisplatin, providing the positive cytotoxic control. After 48 h, the chemical containing media was removed, and fresh media was replaced in all wells. After 72 h, the cells were dual-stained using Hoescht 33342 and Cell Tracker Orange CMTMR; the number of colonies per well was then determined, as described previously for the 96-well CFA. All reagents and media were added and removed using the robotic platform available at the Samuel Lunenfeld Research Institute Robotics Facility (Toronto, Canada).
BioTechniques for Preclinical Development 44 ı No. 2 | 2008
The B-score was then used to normalize the data with respect to systematic variation between plates using the HTS Corrector software (15; www.labunix. uqam.ca/∼makarenv/hts.html). Putative hits were then determined to be any compounds with a B-score lower than 3 standard deviations from the median B-score. The LOPAC1280 library was supplied by the Samuel Lunenfeld Research Institute and screened in two independent experiments; the first using plates 1 through 8, and the second using plates 9 through 16. The data were analyzed, and putative hits were selected separately. All putative hits were reordered from Sigma-Aldrich, and fresh batches were used for followup testing. RESULTS AND DISCUSSION The primary objective of this study was to automate the conduct of the CFA, thereby increasing efficiency in performing large-scale high-throughput www.biotechniques.com ı BioTechniques ı xi
Short Technical Reports
A
B
C
D
Figure 3. Dose response curves created using the 96-well colony formation assay (CFA). (A) Evaluation of the 96-well CFA on FaDu cells for the analysis of combination therapy of radiation therapy (0–6 Gy) and cisplatin (0–1 μM). (B) Evaluation of the 96-well CFA on A549 cells for the analysis of combination therapy of radiation therapy (0–6 Gy) and cisplatin (0–0.5 μM). (C) Survival curve created using the 96-well CFA on FaDu cells treated with ouabain (0–0.4 μM), an example of a confirmed hit from the high-throughput screening (HTS). (D) Survival curve created using the 96-well CFA on FaDu cells treated with mitoxantrone (0–5 nM), another example of a confirmed hit from the HTS. Each datum represents the mean ± sem from two independent experiments.
experiments. This task was addressed by using high-content microscopy, with differentially staining fluorescent dyes for morphologic distinction of the cell’s nucleus versus cytoplasm (Figure 1A). This allowed the determination of not only the number of colonies, but also the size of each colony. This additional level of assessment thereby allowed the filtering of data by removing any colonies that resulted from a radiationinduced giant cell, or a small aggregate of cells that would not score as a legitimate colony. The INCell Analyzer 1000 provided the necessary flexibility that could automate both image acquisition and analysis. Upon acquiring the images and using the Developer Toolbox software, the images were analyzed using an algorithm that first identified a colony based on adjacent signals from the cytoplasm stained with Cell Tracker Orange CMTMR. Once the colony was defined, the number of cells within this colony was determined by counting the number of Hoescht 33342 stained nuclei (Figure 1B). After analysis of all the images on a single xii ı BioTechniques ı www.biotechniques.com
plate, a list of the well location of each colony, and its corresponding number of nuclei, was produced (Figure 1C). A customized program then converted this list into a matrix representing the number of colonies per well (Figure 1D). In this process, a filter was also included to define the minimum and maximum number of nuclei necessary in order to qualify as a colony. The minimum number of cells per colony is determined by assuming a least growth scenario, wherein cells do not start dividing until the chemical has been removed, allowing for 72 h of growth. Given the doubling time of the cell line being investigated, one could then determine the number of divisions a single cell will undergo, thereby providing an estimate for the minimum number of cells in a colony. In the case of FaDu cells, with a doubling time of ∼22 h (16), approximately three doublings should have occurred within 72 h, thus a colony should constitute at least eight cells. For practical purposes, the threshold was set between 6 to 350
nuclei/colony. Once the number of colonies per well had been determined, the surviving fraction was then calculated in the same manner as the traditional CFA (Figure 1E). Historically, the CFA was first described as a method for assessing the effects of radiation (1). Since that time, there have been many modifications to this assay with the use of fluorescent dyes and smaller formats (17); however it has yet to be clearly illustrated whether these modified protocols are able to recapitulate the traditional gold-standard assay. In order to clearly validate the previously described protocol, the first set of 96-well experiments was conducted using radiation therapy on the FaDu cells, since they produced well-defined single colonies, facilitating miniaturization of the CFA. At a dilution of 100 cells/well in the control (untreated) sample, the colonies remained discrete and separable by the image analysis software after 7 days growth. This was achieved in the untreated samples and could be extended out to 2500 cells/well in samples exposed to 16 Gy. Survival curves were then generated using the traditional 6-well CFA format and compared with that of the 96-well CFA (Figure 2A). The two curves overlap almost completely, indicating that the 96-well CFA could recapitulate the 6-well CFA. The survival curves generated with the 96-well format however could only be extended to 8 Gy radiation therapy, since beyond this dose, too many radiation-induced giant cells would confound the analysis. To determine if this method could be applicable to other cell lines, the experiments were repeated using the A549 cell line (Figure 2B). Once again, the 96-well data set was able to reproduce the data from the 6-well assay, indicating that the 96-well format could recapitulate the traditional assay in the context of radiation-treated colony forming cell lines. The subsequent studies proceeded to evaluate combination treatments, a major objective in improving cancer therapy. Hence, FaDu cells were treated with radiation therapy combined with cisplatin, a common clinical regimen for treatment of head and neck squamous cell carcinoma patients. Again, the 96-
BioTechniques for Preclinical Development Vol. 44 ı No. 2 | 2008
Short Technical Reports
well assay appeared to be an excellent representation, as the radiation therapy alone curve replicated that of the 6-well assay results (Figure 3A). Similarly, the cisplatin data agree with previously published reports whereby ∼50% survival level was observed when FaDu cells were treated with 0.5 μM cisplatin (18). The corresponding experiments were performed with the A549 cells, with similar results obtained (Figure 3B). From the time of staining of the plates, the data were obtained within 24 h using automated counting. Comparison of estimated time required for these assays indicated time savings at almost every step of the protocol with the 96-well CFA, but the most substantial gain was observed at the colony counting step (see Supplementary Table S1 available online at www.BioTechniques.com). Given that the approximate reagent and consumable costs of both the 6-well and the 96-well assays are similar, this demonstrates the substantial gain in efficiency of generating and analyzing such volumes of data, particularly for the determination of multiple permutations of combinatorial therapies. From an analytical perspective, many different statistical methods could be used to test for additive, synergistic, or subadditive interactions of combined therapies. These methods include isobologram analyses (19), mean inactivation dose (20), or the median effect principle (21), all of which require the generation of data using a broad range of concentrations of different treatments. Such experimental data are time consuming to generate using the traditional 6-well format, which is easily overcome by the efficiency of the 96-well CFA, enabling the collection of masses of data required for such analyses in a much shorter time frame. The final question we sought to address using the 96-well CFA was its utilization for HTS of chemical libraries. Our laboratory has previously identified novel anticancer activities of existing antimicrobial compounds, such as alexidine dihydrochloride (22) and benzethonium chloride (23). However, the read-out for these screens was the 3-(4,5-dimethylthiazol-2yl)-5-(3-carboxymethoxyphenyl)-2-
(4-sulfophenyl)-2H-tetrazolium salt (MTS) assay, which is a measure of mitochondrial enzymatic activity. There are other readouts available for determining cytotoxicity (24), or other specific cellular events (25,26), however these assays do not measure reproductive potential. Hence, the 96well CFA was adopted for the HTS, anticipating the identification of novel and more potent anticancer agents. As an initial proof of concept experiment, the protocol for performing the assay was transferred to a robotic platform. The LOPAC1280 library was screened at 0.5 μM final concentration of all compounds for their ability to inhibit clonogenic growth, as measured by the number of colonies per well normalized to untreated samples. To correct for systematic variation, the data were normalized using the B-score, a statistical analysis that accounts for variability across rows, columns, and plates (27). After normalization by the B-score, only compounds whose score was