ARTICLES

Feasibility of a High-Flux Anticancer Drug

Screen Using a Diverse Panel of Cultured
Human Tumor Cell Lines
Anne Monks * Dominic Scudiero, Philip Skehan, Robert Shoemaker,
Kenneth Paull, David Vistica, Curtis Hose, John Langley, Paul Cronise,
Anne Vaigro-Wolff, Marcia Gray-Goodrich, Hugh Campbell, Joseph
Mayo, Michael Boyd

subsequent in vivo xenograft testing of compounds of interest.

We describe here the development and implementation of a While the concept is simple, the technical challenges for im-
pilot-scale, in vitro, anticancer drug screen utilizing a panel plementation of such a screen are greatly magnified by the re-
of 60 human tumor cell lines organized into subpanels repre- quired scope of the program.
senting leukemia, melanoma, and cancers of the lung, colon, The goal of the program is to implement the full-scale screen
kidney, ovary, and central nervous system. The ultimate with a capacity for testing new substances at a rate of more than
goal of this disease-oriented screen is to facilitate the dis- 10 000 per year against a broadly representative panel of 100 or
covery of new compounds with potential cell line-specific more human tumor cell lines. In moving toward this goal, we in-
and/or subpanel-specific antitumor activity. In the current itiated the pilot-scale screen outlined here, incorporating ex-
screening protocol, each cell line is inoculated onto perience gained from previous laboratory-scale research and
microtiter plates, then preincubated for 24-28 hours. Sub- development efforts. With the pilot screen we have explored a
sequently, test agents are added in five 10-fold dilutions and variety of conceptual issues, as well as technical and managerial
the culture is incubated for an additional 48 hours. For each
issues critically involved in a scale-up to the full screen. This
test agent, a dose-response profile is generated. End-point
paper describes our experiences with the pilot screen and its
determinations of the cell viability or cell growth are per-
evolution through a variety of technical stages, including com-
formed by in situ fixation of cells, followed by staining with a
protein-binding dye, sulforhodamine B (SRB). The SRB parative evaluations of two types of tetrazolium assays (4,5) and
binds to the basic amino acids of cellular macromolecules; an assay using the protein-binding dye sulforhodamine B (SRB)
the solubilized stain is measured spectrophotometrically to (6), all of which are microculture assays to measure cell
determine relative cell growth or viability in treated and un- viability and cell growth.
treated cells. Following the pilot screening studies, a screen- The results from operation of the pilot screen have been
ing rate of 400 compounds per week has been consistently evaluated both continuously and retrospectively with a par-
achieved. [J Natl Cancer Inst 83:757-766,1991] ticular view toward determining the feasibility of implementa-
tion of the full-scale screen.

The National Cancer Institute is implementing a new inves-

tigational, in vitro, disease-oriented, drug-discovery screen. Received November 19, 1990; revised March 8, 1991; accepted March 20,
Other publications describe the overall concept and rationale for 1991.
Sponsored in part by the National Cancer Institute, National Institutes of
this screen (1-3) and various aspects of its technical evolution Health, Department of Health and Human Services, under contract N0ICO-
(4-7). The purpose of the screen is to provide for the initial 23910 with Program Resources, Inc.
evaluation of more than 10 000 new substances per year for A. Monks, D. Scudiero, C. Hose, J. Langley, P. Cronise, A. Vaigro-Wolff, M.
Gray-Goodrich, H. Campbell (Program Resources, Inc), P. Skehan, R.
cytotoxic and/or growth-inhibitory activity against a wide diver- Shoemaker, K. Paull, D. Vistica, J. Mayo, M. Boyd (Developmental
sity of tumor types and to allow the detection of possible tumor- Therapeutics Program, Division of Cancer Treatment), NCI-Frederick Cancer
Research and Development Center, Frederick, Md.
type-specific sensitivity. The data from such an in vitro primary
Correspondence to: Anne Monks, PhD, NCI-Frederick Cancer Research and
screen will allow selection of the most sensitive cell line(s) for Development Center, Bldg 432, Rm 232, Frederick, MD 21701 -1013.

Vol. 83, No. 11, June 5, 1991 ARTICLES 757

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 Materials and Methods tained. For each cell line, the seeding density per flask is deter-
mined for production of a healthy culture of 70%-9O% confluen-
Cell Lines cy after 7 days for continued routine maintenance or after 4 or 5
days for the microculture assay. These seeding densities are then
The human tumor cell lines currently used in the pilot screen
utilized twice a week to maintain sufficient cells for routine cul-
have been presented previously (7); details of their origin and
ture and for anticancer drug screening.
characterization are described elsewhere (Stinson SF, Alley
MC, Kopp WJ, et al: manuscript submitted). Parallel to the im- Preparation and Inoculation of Cells
plementation and operation of the pilot screen, a major effort
continues for the acquisition and evaluation of new cell lines for All of the adherent cell lines are detached from the culture
possible addition or substitution into the panel. Ultimately, the flasks by addition of 2-3 mL of 0.05% trypsin-EDTA (GIBCO
cell line panel may contain as many as twice the current number Laboratories, Grand Island, NY). Thereafter, trypsin is inac-
of lines, with each disease-related subpanel containing 10-15 tivated by addition of 10 mL of 5% serum-containing RPMI-
representative lines (3). However, for purposes of the pilot 1640 medium. Cells are separated into single-cell suspensions
screen, we have focused on a fixed set of 60 lines (Table 1). by a gentle pipetting action, then counted using trypan-blue ex-
Criteria for selection of a cell line for use in the interim panel clusion on a hemacytometer or by a Coulter counter, which is
were as follows: used when viability as determined by trypan-blue exclusion is
routinely greater than 97% and the cells maintain suspension as
(a) adaptability to growth on a single medium (RPMI-1640 single cells. After counting, dilutions are made to give the ap-
plus 5% fetal bovine serum and 2 nW glutamine); propriate cell densities for inoculation onto the microtiter plates.
(b) a negative test for mycoplasma and mouse antibody Cells are inoculated in a volume of 100 (iL per well at densities
production; between 5000 and 40 000 cells per well (Table 1); the basis for
(c) isoenzyme and karyotype profiles verifying the human selection of the particular inoculation densities for each cell line
origin of the cells; is described in the Results section. A 100-fiL aliquot of com-
(d) mass doubling time that allows for harvesting of ap- plete medium is added to cell-free wells. Cells from all cell lines
proximately 3 x 107 cells twice a week; and are counted, diluted, and inoculated onto microculture plates
(e) suitability for use with microculture assays. within a 4-hour period on 2 days each week. The microtiter
Once a line had been established as suitable, the number of plates containing the cells are preincubated for approximately
cells was massively expanded in a minimal number of passages, 24 hours at 37°C to allow stabilization prior to addition of drug.
and the cells were cryopreserved in a large repository of am-
Solubilization and Dilution of Samples
pules, each containing 1 x 106 cells, to provide a consistent,
long-term frozen stock for future use (4). Within the primary For initial screening of pure compounds, each agent is
drug evaluation laboratory, new cell line stock samples are routinely tested at five 10-fold dilutions, starting from a maxi-
thawed as each cell line culture used for screening approaches mum concentration of 10^ M. Alternatively, a maximum of 10"3
its 20th passage or if there is a noticeable change in growth, M may be selected if solubility permits, if the higher upper limit
morphology, or drug-response characteristics. Once the growth is desired, and if sufficient compound is available. Alternative
of the new stock is established at the second or third passage, initial concentrations may also be utilized for retesting, par-
the older passage line is replaced with the new stock for use in ticularly for potent cytotoxins, for compounds of limited
the screening laboratory. solubility, or for nonroutine detailed comparisons of selected
compounds when previous information is available. Crude ex-
Cell Line Maintenance
tracts of natural products are tested at five threefold dilutions,
Major considerations with respect to cell line maintenance in- starting at an upper limit of 250 |ig/mL regardless of solubility,
clude (a) the necessity of continually producing approximately 3 ie, paniculate matter may be present. All samples are initially
x 107 cells for biweekly inoculation for the pilot screen, (b) the solubilized in dimethyl sulfoxide (DMSO) or water at 400 times
potential for cross-contamination of cell lines, and (c) the pos- the desired final maximum test concentration. Drug stocks are
sibility of low-grade microbial contamination. Individual tech- not filtered or sterilized, but microbial contamination is
nicians are therefore assigned only six specific cell lines, which controlled by addition of gentamicin to the drug diluent. Multi-
they monitor continually for growth characteristics in tissue-cul- ple aliquots are stored frozen at -70°C to provide uniform
ture flasks and microculture plates, for behavior in the microcul- samples for initial tests as well as for retests, if required. Just
ture assays, and for microscopic appearance. Furthermore, the prior to preparation of the drug dilutions in cell-culture medium,
use of a limited number of passages from a frozen-stock vial these frozen concentrates are thawed at room temperature for 5
also helps prevent long-term cross-contamination of a cell line. minutes. The concentrates are then diluted with complete
Cells are grown and passaged in antibiotic-free growth medium medium containing 50 |ig/mL gentamicin to twice the desired
to ensure freedom from microbial contaminants. final concentrations.
Cells are maintained in multiple T150 tissue-culture flasks.
Drug Incubation
Cells for each inoculation day are maintained separately (no
common reagents) and passaged on separate days to prevent Immediately after preparation of these intermediate dilutions,
catastrophic loss of growing cell line stocks to microbial con- 100-(iL aliquots of each dilution are added to the appropriate
tamination. Additional backup flasks of cells are also main- microtiter plate wells according to the format in Fig 1, in which

758 Journal of the National Cancer Institute

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 Sulforhodamine B Assay
A D1 5 D14 \ D n i g Blank
01, Dl, ">11 c« CB D2, « i 02, 02«
j (No catty The detailed methodology for the SRB assay is presented
B Qc Oc elsewhere (6). Briefly, adherent cell cultures are fixed in situ by
I Can Una
C adding 50 uX of cold 50% (wt/vol) trichloroacetic acid (TCA)
D
Oc «C
1 ' (final concentration, 10% TCA) and incubating for 60 minutes
Oc Oc Call Una
E 2 at 4°C. The supernatant is then discarded, and the plates are
Oc Oc
washed five times with deionized water and dried. One hundred
F Oc Qc Call Una
3 microliters of SRB solution (0.4% wt/vol in 1% acetic acid) is
a Qc Qc
added to each microtiter well, and the culture is incubated for 10
H Vonig Blank
1
/ (No call*) minutes at room temperature. Unbound SRB is removed by
washing five times with 1% acetic acid. Then the plates are air-
dried. Bound stain is solubilized with Tris buffer, and the optical
Fig 1. Microtiter plate format used for screening shows three cell lines and two
test agents (Dl and D2), each inoculated at five concentrations (Dl.-Dl, and densities are read on an automated spectrophotometric plate
D2.-D2,). One plate holds six dose-response experiments with quadruplicate reader at a single wavelength of 515 nm.
growth control (Gc) wells, four control background (C,) wells, and duplicate test
wells, for each dose-response set. Letters A-H and Nos. 1-12 represent the For suspensions of cell cultures (eg, leukemias), the method is
microtiter plate map. the same, except that at the end of the drug-incubation period,
the settled cells are fixed to the bottom of the microtiter well by
gently adding 50 ^L of 80% cold TCA (final concentration,
three cell lines are inoculated per plate. As the microtiter wells 16% TCA).
already contain the cells in 100 ^L of medium, the final drug
concentration tested is 50% of that in the intermediate dilutions. Data Calculations
Agents are then added immediately to the cultures in the Unprocessed optical density data from each microtiter plate
microtiter plates. During development of these procedures, drug are automatically transferred from the plate reader to a
incubation times were 1, 2, 3, 4, or 6 days at 37°C in an atmos- microcomputer, where the background optical density (OD)
phere of 5% CO2 and 100% relative humidity. The plates were measurements (ie, complete medium plus stain, minus cells) are
then assayed for cellular growth and viability by microculture subtracted from the appropriate control well values and where
assay—either by one of the two types of tetrazolium assay or by the appropriate drug-blank measurements (ie, complete medium
the SRB assay. In the current screening procedure, the cultures plus test compound dilution plus stain, minus cells) are sub-
are incubated with test agents for 2 days, and the end point is tracted from the appropriate test well values. The values for
measured by the SRB assay. mean ± SD of data from replicate wells are calculated. Data are
Microculture Tetrazolium Assay expressed in terms of %T/C [(OD of treated cells/OD of control
cells) x 100], as a measure of cell viability and survival in the
The MTT assay is based on metabolic reduction of 3-(4,5- presence of test materials. Calculations are also made for the
dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT). concentration of test agents giving a T/C value of 50%, or 50%
Details of this assay have been published previously (4), but cer- growth inhibition (ICJO), and a T/C value of 10%, or 90%
tain modifications were made for use in the pilot screen. A 50- growth inhibition (IQo).
(iL aliquot of MTT solution (1 mg/mL) in RPMI-1640 medium, With the SRB assay, a measure is also made of the cell
with no serum or glutamine, is added directly to all the ap- population density at time 0 (the time at which drugs are added)
propriate microtiter plate wells containing cells, complete from two extra reference plates of inoculated cells fixed with
growth medium, and test agents. The culture is then incubated TCA just prior to drug addition to the test plates. Thus, we have
for 4 hours to allow for MTT metabolism to formazan. After this three measurements: control optical density (C), test optical den-
time, the supernatant is aspirated and 150 (iL of DMSO is added sity (T), and optical density at time zero (To).
to dissolve the formazan. Plates are agitated on a plate shaker to Using these measurements, cellular responses can be calcu-
ensure a homogeneous solution, and the optical densities are lated for growth stimulation, for no drug effect, and for growth
read on an automated spectrophotometric plate reader (model inhibition. If T is greater than or equal to T^, the calculation is
312; Biotech Research Laboratories, Inc, Rockville, Md) at a 100 x [(T - T0)/(C - To)]. If T is less than T,,, cell killing has oc-
single wavelength of 550 nm. curred and can be calculated from 100 x [(T - To)/To] (8). Thus,
The XTT assay is based on the metabolic reduction of 2,3- for each drug-cell line combination, a dose-response curve is
bis[2-methoxy-4-nitro-5-sulfophenyl]-5-[(phenylamino)carbonyl]- generated and three levels of effect are calculated. Growth in-
2//-tetrazolium hydroxide (XTT). The detailed methodology is hibition of 50% (GIJO) is calculated from 100 x [(T - T0)/(C -
presented elsewhere (5). Briefly, 1 mg/mL XTT is mixed with To)] = 50, which is the drug concentration causing a 50% reduc-
phenazine methosulfate (7.65 |ig/mL) immediately prior to its tion in the net protein increase in control cells during the drug
addition to microtiter plates. Fifty microliters of the mixture is incubation. The drug concentration resulting in total growth in-
added to each well and incubated for 2-4 hours, depending on hibition (TGI) is calculated from T = To, where the amount of
the metabolizing characteristics of the individual cell lines. Each protein at the end of drug incubation is equal to the amount at
plate is agitated for 2-3 minutes on a plate shaker; then optical the beginning. The final calculation, LCJQ, is the concentration
densities are read on an spectrophotometric plate reader at a of drug causing a 50% reduction in the measured protein at the
single wavelength of 450 nm. end of the drug incubation, compared with that at the beginning,

Vol. 83, No. 11, June 5, 1991 ARTICLES 759

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 indicating a net loss of cells following drug treatment. LC^ is variation exceeding 25%, as determined from the CV of the
calculated from 100 x [(T - To)/To] = - 50. mean of duplicate dose-response optical densities; (c) mean
Values are calculated for each of these parameters if the levelbackground optical densities greater than 10% of control; (d)
of effect is reached, but if the effect is not reached or is ex- mean control optical densities exceeding 3 SD of the mean opti-
ceeded, then the value is expressed as greater or less than the cal densities for the six previous experiments for that cell line;
maximum or minimum drug concentration tested. and (e) a dose-response curve transecting the point of 50%
growth inhibition more than once (reversal).
Computer Support and Automation These codes constitute a "failure" of part or all of a dose-
The projected goal of the National Cancer Institute (NCI) in response determination and eliminate those data from further
vitro drug screen is to support the initial yearly evaluation of analysis. On a daily basis, data are copied to an on-site
10 000 or more new agents against as many as 100 or more cell microcomputer for backup and laboratory analysis, then
lines. This would result in the generation of more than 1 million downloaded nightly to a large central computer (VAX 8820) for
dose-response curves per year. Extensive computer support has future detailed analysis, comparisons, and generation of screen-
necessarily been developed for each step of the screening ing-data reports for suppliers of the samples tested.
process, from the shipping of an agent to the drug-preparation Procedure for Submitting New Cell Lines
laboratory to the analysis and return of screening data to a sup-
plier. Investigators who wish to submit cell lines for consideration
Agents to be screened are received by the drug-preparation of inclusion in the panel should contact:
laboratory, and the NCI drug identification code is entered into
the computerized data management and monitoring system. Mr. Richard Camalier, Biologist
After solubilization of the agents, details of the solubility and BTB, DTP, DCT, NCI-FCRDC
priority for testing are entered to update the database. In addi- Building 321, Room 7
tion, bar-coded labels containing all the appropriate information Fort Detrick, Frederick, MD 21702
are generated and attached to the storage vials containing the Telephone (301) 846-5065
concentrates, which are then placed in interim storage at -70°C. FAX (301) 846-5439
Screening-laboratory staff determine the available test
capacity for the following week and enter this information into Results
the mainframe database. Then, automated assignment of test
Choice of Assay for Cell Growth and Viability
agents to fill the scheduled requirements is performed according
to preassigned priority status. Alternatively, assignments or For the initial feasibility studies, the microculture tetrazolium
reassignments may be performed manually. assays (MTT and XTT) were utilized to determine cell growth
Immediately prior to the day of an assignment, information is and viability. During the evolution of the large-scale screen,
loaded via modem to microcomputers (Compaq 386-20 running however, certain technical limitations of these assays became
an SCO XENIX operating system) in the screening laboratories. especially apparent. As a result, we investigated alternative ap-
Labels are generated and printed by computer for each proaches, using protein and biomass stains. We found an assay
microculture plate. Each label contains an identification number using sulforhodamine B (SRB), a dye that binds to basic amino
that is specific to a particular cell line-drug-plate format. acids of cellular macromolecules, to be promising for purposes
After each technical task (ie, cell inoculation, addition of of this screen (6).
drug, or staining), the technicians confirm successful comple- The major technical disadvantage of the microculture
tion, using computer terminals available in each tissue-culture tetrazolium assays (MTT and XTT) is that the "moving-target"
laboratory. This procedure allows for cross-checking of each nature of the optical density, which is determined by formazan
step of the assay to ensure that the actual and assigned proce- production, is a function of time, in addition to cell number.
dures are identical. Once an identity number has been entered, With the requirements to screen at least 10 000 agents per year
the computer system tracks it through all the technical steps so and to compare each cell line against an agent in a single test,
that when the optical densities are read spectrophotometrically, 1000 or more microculture plates must be evaluated for cell
dose-response data can be calculated automatically and printed growth inhibition and viability on the same day. This presents a
within minutes of the plate reading. At this point, there is severe logistical challenge, especially at the stage of plate read-
provision for the manual elimination of suspicious data points ing, as there is a very narrow window of time within which the
by insertion of "reason codes" (eg, for microbial contamination, tetrazolium plates must be processed.
improper cell inoculation, or improper drug inoculation). These The MTT microculture assay is a reasonably simple, sensi-
procedures are designed to prevent entry into the mainframe tive, reproducible assay that exhibits good signal-to-noise ratios
database of dose-response data known to be flawed for specific (4). However, the MTT procedure requires aqueous aspiration
reasons. and DMSO-solubilization steps, making the procedure more
In addition to these provisions for manual deletions to ensure time-consuming and potentially error-prone. Moreover, the use
quality control, a series of automated quality control parameters of large quantities of DMSO is undesirable in a busy screening-
are performed by computer. These include "flags" for (a) varia- laboratory environment.
tion exceeding 20%, as determined from the coefficient of varia- The XTT microculture assay is technically very simple and
tion (CV) of the mean of control optical densities (n - 4); (b) can be applied directly to suspension or monolayer cultures.

760 Journal of the National Cancer Institute

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 However, the XTT assay yields relatively poor signal-to-noise Table 1. Human tumor cell lines and inoculation densities used in the NCI
disease-oriented in vitro drug screen
ratios and requires addition to the assay medium of an electron-
coupling agent (phenazine methosulfate) representing yet Cell line Cells/well Cell line Cells/well Cell line Cells/well
another potential source of experimental variation. Furthermore,
with the XTT assay, there was occasional formation of a crystal- Lung cancer Colon cancer CNS cancer
line material in the culture wells, which strongly interfered with NC1-H23 20000 HT29 5000 SNB-19 15000
NCI-H226 20000 HCC-2998 10000 SNB-75 20 000
the assay. This phenomenon was subsequently related in part to NCI-H322M 20000 HCT-116 5000 SNB-78 20000
CO2 variations and buffering instability (9) and was one of the NCI-H460 5000 SW-620 10000 U251 7500
factors leading to development of a new CO2-independent cul- NCI-H522 15000 COLO 205 15000 SF-268 15 000
A549/ATCC 10000 DLD-1 5000 SF-295 10000
ture medium (70) to obviate the problem. EKVX 20000 HCT-15 10000 SF-539 15000
The SRB assay has a stable end point, since the drug-exposed HOP-18 20000 KM12 15000 XF498 20 000
cells are fixed with TCA at the end of the incubation period (6). HOP-62 15000 KM20L2 10000
HOP-92 20000 Ovarian cancer
Once the cells are fixed and stained, there is no critical time LXFL529 10000 Melanoma OVCAR-3 10000
restriction for subsequent determination of the SRB optical den- DMS 114 20000 LOX IMVI 5000 OVCAR^ 10000
sities, which change less than 2% over a 7-day period if DMS 273 5000 MALME-3M 20000 OVCAR-5 20 000
SK-MEL-2 20000 OVCAR-8 10000
evaporation is restricted (data not shown). Furthermore, the Renal cancer SK-MEL-5 10000 IGR-OV-1 10000
SRB assay provides excellent signal-to-noise ratios. The techni- UO-31 20 000 SK-MEL-28 10 000 SK-OV-3 20 000
cal manipulations required for staining are greater than for the SN12C 15000 M19-MEL 10000
A498 20000 UACC-62 10000 Leukemia
tetrazolium assays, requiring up to 10 washing steps and two CAKI-1 10000 UACC-257 20 000 CCRF-CEM 40000
drying steps. However, these steps can be automated and, under RXF 393 20 000 M14 15000 K-562 5000
the current protocol, can be readily scheduled around the ac- RXF631 10000 MOLT-4 30 000
ACHN 15000 HL-60 20 000
tivities of cell inoculation, addition of drug, and in situ staining 786-0 10000 RPMI-8226 20 000
of cells. TK-10 15000 SR 30 000
Comparisons of data from several hundred agents screened in
parallel by MTT and SRB assays indicated that, under the ex-
perimental conditions employed and within the limits of the data
To examine the difference between the results at 1,2, 3, and 4
analyses applied, the MTT and SRB assays generally yielded
days of incubation, the SRB assay was used to perform a dose-
quite similar results (77). Therefore, the SRB procedure was
response analysis on 20 standard drugs tested in six cell lines
subsequently selected for routine use in the pilot screen, prin-
(A2780, HT29, HOP-62, OVCAR-5, SN12KI, and SK-MEL-5).
cipally because of the technical advantages described above.
A2780 and SN12KI have since been deleted from the cell line
panel. The 20 agents were doxorubicin, amsacrine (AMSA),
Assay Parameters
bleomycin, cisplatin, dibromodideoxymannitol (mitobronitol),
gossypol, melphalan, mitomycin-C, retinoic acid, vinblastine,
Assay parameters refer here to the particular selections of (a)
actinomycin (dactinomycin), carmustine (BCNU), chromo-
initial cell densities for inoculation and (b) drug incubation
mycin, cordycepin (3'-deoxyadenosine), fluorouracil (5-FU),
times. To avoid biasing the screen excessively toward detection
homoharringtonine, methotrexate, phyllanthoside, tamoxifen,
of antiproliferative effects that might otherwise obscure interest-
and etoposide (VP-16).
ing patterns of differential cytotoxicity, we have focused our
developmental efforts on screening protocols incorporating rela- With the assays using 1-4 days of incubation, it is possible to
tively high initial inoculation densities and relatively short in- determine a signal (optical density) at the time drug is added
cubation times. Table 1 shows the specific cell inoculation (To). Using these values, calculations can be made for the con-
densities used for each cell line; optimal initial cell densities centration of drug causing 15% growth inhibition (GII3), 50%
were determined for each line as described in the following sec- growth inhibition (GIjo), total growth inhibition (TGI), and 50%
tion. cell kill (LCso). Fig 2 shows representative dose-response curves
As the cell line panel was expanded to include new lines, it comparing 1, 2, 3, or 4 days' exposure of six cell lines to
became apparent that many cell lines were subject to severe doxorubicin and BCNU. Table 2 shows a summary of the cel-
nutrient depletion by the end of a 6-day incubation period. This lular response parameters calculated as a function of drug in-
was indicated by extreme acidity of the growth medium and cubation time from such curves obtained for all 20 of the drugs
peeling of the cells from the microculture plate well surface. An used as standards. These data indicated that growth inhibition
additional consideration was the effect of incubation time on (GI,j and GI50) by the selected standard clinical agents could be
cell doubling times. An investigation into the growth rate of reliably determined at 2-4 days of drug exposure but that this
cells inoculated into microculture plates was undertaken. Six determination was much less reliable after only 1 day of drug
cell lines with different growth rates were selected and their exposure. The more demanding end points of TGI and LCJQ
growth in the described protocol was examined. During days 1 were less commonly achieved than the end points of GI,5 and
and 2, the doubling time increased as expected, but between GIso, but the 1- and 2-day incubations discriminated these
days 3 and 4, two of the six cell lines showed negative growth parameters less well than the longer incubation periods. The
rates, indicating that the populations were dying faster than they TGI parameter was achieved in five or more cell lines in the 1-
were growing (data not shown). or 2-day assay only with doxorubicin, AMSA, BCNU, gossypol,

Vol. 83, No. 11, June 5, 1991 ARTICLES 761

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 Doxorubicin BCNU

A27S0 HOP-82 A 2 790 ^^__^^ HOP-82

100 » —

50 -N
\
\ ^ ^ ~ ~ ~ * i
\
-50

100
1
As -ao

-100
0 » 10 10 ' 10 m
LJ
?,3

HT-M OVCAR-5

I 50
5o "
O
I -50

100 p 6 ]
I io ° io • io * 10 ' io to » io ' io io * io * to * to • io * io * vo- * i o - 1

SK-MEL-5 SN12KI ^~-+^ SK-MEL-5 ^-*^_ SN12KI

100

50

-50

2
- 100
1

Doxorubicin Concentration (M) BCNU Concentration (M)

Fig 2. Dose-response curves for six cell lines incubated with doxorubicin or BCNU for 1, 2, 3, or 4 days' continuous exposure.

Table 2. Cellular response as a function of drug-incubation time as determined

by SRB assay* .98 for all three of these lines and greater than .95 for all panel
lines. In addition, cell growth in the microculture plates was
Incubation Responders normalized characterized at a variety of cell densities (5000, 10 000, 20 000,
time % response! to 4-dav valuei 40 000, and 50 000 cells per well) over a 4-day period of
(days) GI|5 GIso TGI LC50 GI,5 GIso TGI LCso
growth. From these data, the specific inoculation density
1 89.1 68.9 45.4 21.0 96.8 77.9 59.8 35.1 selected for each cell line was that which produced an optical
2 91.5 83.1 51.7 36.4 99.5 94.0 68.1 60.9 density signal above the noise level of the assay and within the
3 91.7 85.8 69.2 52.5 99.7 97.1 91.2 87.8
4 92.0 88.4 75.9 59.8 100.0 100.0 100.0 100.0 linear range of the SRB signal (ie, within the range of 0.2-2.0
OD units), for both the To and control optical density (C) meas-
*GIis = 15% growth inhibition; GI50 = 50% growth inhibition; TGI = total urements.
growth inhibition; and LC50 = 50% cell kill. The To and control optical densities generated from SRB-
t% of total responses (6 cell lines x 20 drugs = 120responses)achieving each stained cells are a function of cell mass and growth rate. Thus,
calculated parameter.
JTotal responses achieving each calculated parameter normalized to the 4-day cells with a small mass, such as the leukemias CCRF-CEM and
incubation value, which is expressed as 100%. MOLT-4, are inoculated at the relatively high densities of 40
000 and 30 000 cells per well, respectively. Furthermore, cells
that divide relatively slowly (eg, XF 498 and HOP-18) are in-
homoharringtonine, phyllanthoside, tamoxifen, and vinblastine. oculated at 20 000 cells per well, while more rapidly dividing
The LC^ was achieved only with BCNU, gossypol, homohar- cells (eg, NCI-H460 and HCT-116) are inoculated at 5000 cells
ringtonine, and tamoxifen. per well.

Selection of Cell Lines and Inoculation Densities Quality Control and Reproducibility of Cell Line
Performance
Prior to inclusion of cell lines in the screening panel, their
growth and compatibility with the screening model were deter- Once a cell line has been included as part of the panel, all
mined. These determinations included comparison of the data generated from the line are stored in a database for various
linearity of the SRB-generated optical density signal with the analyses. Fig 4 shows typical examples of the control optical
cell number in the microculture assay. Fig 3 shows repre- densities measured for three cell lines in sequential experiments.
sentative linear signals from 5000-50 000 cells per well in the Within each experiment, the range of control optical densities
MALME-3M melanoma line and in the U251 and XF 498 from 40 microculture plates (mean of quadruplicate wells) is
central nervous system (CNS) tumor lines, measured 1 hour also included. Cell line SK-OV-3 (ovarian cancer) showed little
after inoculation. The correlation coefficient (r) was greater than variation of control optical density either within or between ex-

762 Journal of the National Cancer Institute

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 1.0-

0.8-
/
Optical Density

0.6-
/

0.4-
^ ^
3 0
2 a

0.2- 2 6 SF-539
2 «
2 2
0)
<D
2 0 »
! i .
o.o- * * * Q

C) 10 20 30 40 50
3
Cells/Well (xio- ) 5.
o
Fig 3. Three representative cell lines, showing linear relationship between in-
creasing cell number and optical density measured by the SRB assay. 0 =
3 0
MALME-3M, B = U251,and A = X F 4 9 8 . 2 8
2 6
2 I SNB19
periments. The data for the SF-539 (CNS tumor) line are from 2 2
2 0
two different cell inoculation densities; cultures in experiments S
1
1, 2, 3, and 5 were inoculated at 20 000 cells per well, and the ^ I
1
remainder were inoculated at 15 000 cells per well. The SNB-19
(CNS tumor) line initially showed more variation within and be- | l l < 'I., ,• i i
tween experiments than the SK-OV-3 line, but in the latter • ' • » I •
stages, the variation appeared to decrease, perhaps due to in-
creasing experience of the technicians handling the line. In
general, the variation for a cell line, within any experiment,
ranged from 0.1 to 0.4 OD units, but the interexperimental Sequential Experiments
variation occasionally exceeded 1.0 OD units.
In another analysis (data not shown), the variation in the con- Fig 4. Range of control optical densities measured for three cell lines (SK-OV-3,
trol optical density value with successive experiments was SF-539, and SNB-19) in each experiment and between sequential experiments,
ranked from the cell line with the most reproducible values to at the rate of two experiments per week.
the cell line with the least reproducible values. Examples of
human tumor cell lines with the most reproducible values are Reproducibility of Screening Assay
SK-OV-3 (OD = 0.95 ±0.15) and XF 498 (OD = 0.74 ± 0.13).
Examples of the cell lines with the least reproducible values are Fig 6 gives representative illustrations of the experimental
SF-268 (OD = 1.3 ± 0.52); NCI-H522 (OD = 0.91 ± 0.37); variation in data obtained with the three types of assay (MTT,
HOP-18 (OD = 0.82 ± 0.36); and SK-MEL-2 (OD = 0.99 ± XTT, and SRB) using a 6-day continuous incubation of mel-
0.51). phalan with the U251 CNS tumor line, which was independently
Since there was considerable variation in the resulting control grown, plated, and tested by 13 technicians. This experiment
optical density values from inoculation to inoculation with cer- was repeated for three cell lines, six agents, and two incubation
tain lines, it was important to determine whether the apparent periods. All three assays, but particularly MTT and SRB, gave
sensitivity of a cell line to a drug was significantly related to the good agreement in the dose-response profiles. Generally, the
different control optical densities. Fig 5 shows the relationship greatest variations were observed at concentrations with inter-
between optical density and ICso (/i>30) measured for mediate cytostatic activity (ie, 20%-75% T/C, rather than 100%
doxorubicin against three cell lines. The data for the KM 12 or0%-5%).
(colon tumor) line showed variation of more than 1 OD unit but Fig 7 shows examples of the experimental variation in
a consistent ICJO, with no relationship between optical density doxorubicin ICso values for sequential experiments with dif-
and ICJO. ' n o u r database on 64 cell lines, only five lines showed ferent cell lines. The IC50 values for most lines were highly con-
a significant relationship between optical density and ICso (r>.7; sistent. For example, reproducibility was good, showing only
P<.05); an example is the melanoma cell line UACC-62 (r = occasional discrepancies in the doxorubicin ICy, values for the
.77; P = .0001). A498 renal tumor line and the COLO 205 colon tumor line.

Vol. 83, No. 11, June 5, 1991 ARTICLES 763

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 KM12 SNB75 UACC62
2.5 2.5 2.5
A
B
2.0- A 2.0 2.0- AAA B

A •
AA A
A A
al Densi

1.5- C A
1.5- 15 A A
A A
B AA A A
A A B A AAA
AA A
AAA
1.0 A EB
BBA A
1.0 A A 1.0
AA C AA
* BA A B
AAA A A A
A A AA
Q. 0.5 0.5- A 0.5 A
A
O
0.0 0.0- 0.0

4
1 X 10" 1 x 10' 5 1x10"* 1 x 10 7 1 X 10"
4
1 x 10"5 1 x 10' 1 x 10"7 1 X 10"
4
1 x 10"5 1 x 10"6 1 x 10"7 1 X 10"8

Doxorubicin (M)

Fig 5. Correlation between control optical density and doxorubicin \CX (M) for three cell lines (KM 12, SNB-75, and UACC-62). A = one observation, B = two ob-
servations, C = three observations, E = five observations.

The cell lines showing the least consistency were NCI-H226

MTT XTT
P 120- 120- (mean ICjo = 6.52 x 10"7 M, 95% CL = 9.71 x 10"6 M to 4.38 x
£,00- i I
10""8 M)\ DLD-1 (mean IC50 = 7.36 x 10"7 M, 95% CL = 1.17 x
S 80- \ 80- 10"5 M to 4.64 x 10"8 M); and CCRF-SB (mean IC^ = 8.5 x 10""8
O 60- 60-
M, 95% CL = 1.68 x 10"6 M to 4.30 x 10"9 M). This continuous
o 40-
\
C \ evaluation and comparison of cell line performance over time
§ 20-
o
1 X 10*
V.
1 x 1 0 ' 1 x 1 0 * 1 x 1 0 * 1x10"*
20-

1 x 1 0 * 1 x 1 0 ' 1 x 1 0 * 1 x 1 0 *1 x
has proved to be a useful approach to identify intractable lines
for elimination from the panels or to identify technical problems
Cone(M) Cone(M) leading to irreproducible results for certain lines in the screening
assay. For example, erratic performance was detected with
T/C)

SRB
20 DLD-1, which was found to be due to cross-contamination of
S-100- DLD-1 with NCI-H460 in the laboratory. NCI-H460 is a cell
1 80 line identical to DLD-1 in appearance but more sensitive to
O 60
doxorubicin (mean IC^ = 5 x 10"8 M). In some instances, the
JT 40- \
source of erratic behavior of particular lines could not be
eo 20-
resolved. An example is CCRF-SB, which was dropped from
CL
1 i 10* 1x10* 1x10* 1x10* 1x
the panel because of highly variable responsiveness.
Cone(M)

Discussion
Fig 6. Percent variation (± SD) in dose-response curve for mclphalan incubated
with U251 cells for 6 days. Cells were grown, plated, and tested by 13 tech-
nicians using the MTT, XTT, or SRB assay. Cone = concentration.
From the pilot-screen results described here, we reached a
consensus that it was feasible to operate the full screen on the
Certain lines gave somewhat less consistent IQo values. For desired scale, using the SRB assay, the selected assay
example, the data for the K-562 leukemia line suggested initial parameters, and the quality control and other operational proce-
technical handling problems resulting in poor reproducibility. dures described. During November and December of 1989, after
As the technicians gained experience with the particular lines, a series of reviews of the developmental status and the pilot-
however, the reproducibility of measurements of doxorubicin screening operations, a similar consensus was reached by
IC^ values were markedly improved. A few other lines, such as several groups, including an Ad Hoc Expert Advisory Commit-
SN12KI, showed greater variation overall in the measured ICM tee, the National Cancer Advisory Board, and the Division of
values. Cancer Treatment's Board of Scientific Counselors (12,13). In
The reproducibility of IQo values for all cell lines in the panel addition, the specific recommendation was made that the screen
was ranked. As representative examples, the cell lines showing had reached a sufficient level of refinement that it should be
the greatest consistency were A498 [mean ICjo = 8.70 x 1(T7 M, placed immediately into operational status, using the current 60-
95% confidence limits (CL) = 1.33 x 10"6 M to 5.17 x 10~7 M]\ cell line panel. We have adopted that recommendation and do
COLO 205 (mean IC^ = 5.1 x 10"7 M, 95% CL = 8.1 x 10"7 M not anticipate further additions or substitutions to the present
to 3.21 x 10"7 Af); and SK-MEL-2 (mean IC,,, = 7.56 x 10~7 M, cell line panel in the immediate future. The screening of a large
95%CL= 1.21 x XffM to 4.73 x 10"7 M). backlog of submitted compounds that have accrued at NCI

764 Journal of the National Cancer Institute

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 8 A498 screening laboratory employs 44 technical staff supervised by
1 x 1 0" - two PhD scientist/managers. The maximum screening rate that
7
1 x 10 -
can currently be accomplished is approximately 400 compounds
per 5-day work week, or 20 000 compounds per year. Based on
1 x 10* - current expenditures, the estimated cost to screen one compound
5 is $200-$3O0.
1 x 10- -
The possibility for further increases in screening capacity,
1 x 10- H within current resources, also appears feasible as more of the
-7. COLO-205 routine procedures are automated. Moreover, future expansions
1 x 10
of the cell line panel could be easily incorporated, albeit with a
•-. 1 x 1 0 - 6 - concomitant impact on screening capacity. For example, if the
appropriate cell lines were available to expand the panel, screen-
1 x 10"5"1
ing operations could be performed at about half the present rate
o (eg, 10 000 compounds per year) on a panel about twice the
1 x 10-* H
present size (eg, 120 lines). Further acquisition and development
1 X10" 8 " K562
n of suitable lines for future addition or replacement of existing
2 -7. lines in the panel continues.
o 1 x 10
x The question of panel and subpanel size is an important one
o •€_
Q 1 x 10
that needs further resolution. Statistical considerations suggest
1 x 10
•5. that each of the disease-related subpanels should contain at least
10 representative cell lines to allow reliable determinations of
x10 - 4 . subpanel-specific activity (Rubinstein L, et al: unpublished
1 X 10" 8 -
SN12-KI data). Nevertheless, any future decision about further expan-
sions or other substantial modifications of the cell line panel
1 x10"7- will await sufficient experience and evaluation of the present
1 x 10" 6 "
60-line panel.
The screening of each compound generates a voluminous
1 x 10" 5 - amount of data, which present unique challenges for display and
analysis. The further development and refinement of data
1 x 10" 4 -!
management and analysis support technology for the screening
Sequential Experiments program continue to be areas of intense developmental effort.
Currently, suppliers of compounds will receive the screening
Fig 7. Representative variations in IC^, values for doxorubicin in four cell lines data in a variety of investigational formats {14). A tabular
(A498, COLO 205, K-562, and SN12KI) over 4 months of sequential testing at presentation of all of the mean optical density values will be
the rate of two experiments per week, in the screening protocol. provided, in addition to the entire series of dose-response
curves, plotted for each disease-related subpanel. Also included
will be a series of mean graph displays of relative cellular
during the developmental and pilot phases of the new screen is responsiveness with respect to three different mean response
now under way. Selected acquisition of new compounds from parameters—GI^, TGI, and LCJO. The derivation of the mean
suppliers worldwide for screening by the NCI program has graph display and examples of some unique applications, such
resumed. as the COMPARE pattern-recognition algorithm, have been
Currently, we are testing samples at the rate of 400 per week described elsewhere {14,15). Finally, the data report package
against the 60-line panel, using five sample dilutions, with each will also include a new data display and analysis procedure
cell line-dilution combination being performed in duplicate called the "dose-response matrix." In this analysis, the relative
wells. Appropriate measures of control (nondrug) well values cellular responses are evaluated and compared over the entire
and To values are also obtained. The testing of each compound dose-response range rather than just the midpoints. Thus, a com-
requires the equivalent often 96-well microtiter plates. prehensive, detailed package of data is returned to the suppliers
The minimum amount of compound required for screening following testing of their compounds in this screen.
can be readily estimated by submitters; each full test requires
approximately 40 mL (0.04 L) of cell-culture medium contain- References
ing two times the highest test concentration desired. For ex-
(/) BOYD MR: National Cancer Institute drug discovery and development. In
ample, for a routine upper test concentration limit of 10"4 mol/L, Accomplishments in Oncology. Cancer Therapy: Where Do We Go From
the minimal amount (grams) of compound can be calculated by: Here? (Frei EJ, Freireich EJ, eds), vol 1, No. 1. Philadelphia: Lippincott,
1986, pp 68-76
molecular weight x IQT* x 0.04 x 2.
(2) BOYD MR, SHOEMAKER RH, MCLEMORE TL, ET AL: New drug develop-
The fully operational screening laboratory and the laboratory ment. In Thoracic Oncology (Roth JA, Ruckdeschel JC, Weisenburger TH,
for sample preparation together occupy 8051 sq ft of space, con- eds), chap 49. Philadelphia: Saunders, 1989, pp 711-721
(5) BOYD MR: Status of the National Cancer Institute preclinical antitumor
taining 50 laminar-flow, cell-culture hoods; 12 automated drug discovery screen: Implications for selection of new agents for clinical
spectrophotometric plate readers; and 20 microcomputers. The trial. In Cancer Principles and Practice of Oncology Update (DeVita VT,

Vol. 83, No. 11, June 5, 1991 ARTICLES 765

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018
 Jr, Hellman S, Rosenberg SA, eds), vol 3, No. 10. Philadelphia: Lippincott, (10) VISTICA DT, SCUDIERO DA, SKEHAN P, ET AL: New carbon dioxide-inde-
1989, pp 1-12 pendent basal growth medium for culture of diverse tumor and nontumor
(4) ALLEY MC, SCUDIERO DA, MONKS A, ET AL: Feasibility of drug screening cells of human and nonhuman origin. J Nat] Cancer Inst 82:1055-1061, 1990
with panels of human tumor cell lines using a microculture tetrazolium (//) RUBINSTEIN LV, PAULL KD, SHOEMAKER RH, ET AL: Comparison of in
assay. Cancer Res 48:589-601, 1988 vitro anticancer drug screening data generated with a tetrazolium assay
(5) SCUDIERO DA, SHOEMAKER RH, PAUU. KD, ET AL: Evaluation of a soluble versus a protein assay against a diverse panel of human tumor cell lines. J
tetrazolium/formazan assay for cell growth and drug sensitivity in culture Natl Cancer Inst 82:1113-1118,1990
using human and other tumor cell lines. Cancer Res 48:4827^*833, 1988 (12) AD HOC REVIEW COMMITTEE: Proceedings for National Cancer Institute In
(6) SKEHAN P, STORENG R, SCUCHERO DA, ET AL: New colorimetric Vitro/In Vivo Disease-Oriented Screening Project. Bethesda, Md: NIH,
cytotoxicity assay for anticancer-drug screening. J Natl Cancer Inst Nov 13-15,1989
82:1107-1112,1990 (13) Reviewers Report Progress in New Drug Prescreen System Development.
(7) STINSON SF, ALLEY MC, KENNEY S, ET AL: Morphologic characterization Cancer Lett 15:1-5, 1989
of human cell carcinoma cell lines. Presented at the annual meeting of the (14) BOYD MR, PAULL KD, RUBENSTEIN LR: Data display and analysis
American Association for Cancer Research, San Francisco, Calif, May 24- strategies for the NCI disease-oriented in vitro antitumor drug screen. In
27, 1989 Proceedings of the 22nd Annual Detroit Oncology Symposium on Antican-
(8) SKEHAN P: On the normality of growth dynamics of neoplasms in vivo: cer Drug Discovery and Development (Valeriote FA, Corbett T, Baker L,
Data base analysis. Growth 50:496-515, 1986 eds). Detroit, April 25-27, 1990
(9) VisncA DT, SKEHAN P, SCUDIERO DA, ET AL: Tetrazolium-based assays (15) PAULL KD, SHOEMAKER RH, HODES L, ET AL: Display and analysis of pat-
for cellular viability: A critical examination of parameters which affect for- terns of differential activity of drugs against human tumor cell lines. J Natl
mazan production. Cancer Res. In press Cancer Inst 81:1088-1092, 1989

Effects of a Low-Fat Diet on Levels of

Oxidative Damage to DNA in Human
Peripheral Nucleated Blood Cells
Zora Djuric * Lance K. Heilbrun, Bruce A. Reading, Allison Boomer,
Frederick A. Valeriote, Silvana Martino

low-fat diet. The diet of each woman is being closely monitored,

Fat in the diet has been associated with increased breast can- and potential markers of fat intake and cancer risk are being es-
cer risk. In this study, blood samples were obtained from 21 tablished.
women at high risk for breast cancer who had been random- One intermediate marker of breast cancer risk may be oxida-
ly assigned to either a nonintervention diet or a low-fat diet. tive damage to the DNA. Such damage may play a role in tumor
Oxidative damage was examined in the DNA from nucleated promotion (J) and can be elicited by numerous endogenous
peripheral blood cells. The levels of oxidized thymine, processes, including lipid peroxidation (4). When DNA is ex-
specifically 5-hydroxymethyluracil, were threefold higher in posed to oxidants, thymine residues appear to be highly suscep-
the nonintervention diet group than in the low-fat diet tible to oxidation (5,6). Of the thymine oxidation products,
group. Without regard to diet arm, there also was a sig-
nificant linear relationship between daily total fat intake and
5-hydroxymethyluracil level. These results suggest that
oxidative damage to DNA may be a marker of dietary fat in- Received October 1, 1990; revised February 28, 1991; accepted March 6,
take. In addition, oxidative DNA damage may be a mech- 1991.
anistic link between fat in the diet and cancer risk, since Supported by the Wayne State University Ben Kastle Trust for Cancer Re-
search and by Public Health Service grant CA-22453 from the National Cancer
such damage is associated with the process of tumor promo- Institute, National Institutes of Health, Department of Health and Human Ser-
tion. [J Natl Cancer Inst 83:766-769,1991] vices. The GC/MS data were obtained at the Michigan State University Mass
Spectrometry Facility (East Lansing, Mich), which is supported in part by grant
RR-00480 from the Division of Research Resources, National Cancer Institute,
National Institutes of Health.
Epidemiological studies have suggested an association be- Presented in part at the 5th Biennial Meeting of the Society for Free-Radical
tween increased fat intake and breast cancer risk {12). In an ef- Research, Pasadena, Calif, November 14-18, 1990.
fort to modulate breast cancer risk in women at high risk for Department of Internal Medicine, Wayne State University, Detroit, Mich.
We thank Maria Alonso for blood sample collection.
breast cancer, we initiated an intervention study that randomly 'Correspondence to: Zora Djuric, PhD, Oncology Division, Wayne State
assigned high-risk women to either a nonintervention diet or a University, PO Box 02188, Detroit, MI 48201.

766 Journal of the National Cancer Institute

Downloaded from https://academic.oup.com/jnci/article-abstract/83/11/757/889634
by University of Durham user
on 24 April 2018