It is often necessary to assess the potential cytotoxicity of a set of compounds on cultured cells. Here, we describe a strategy to reliably screen for toxic compounds in a 96-well format.
Cytotoxicity is a critical parameter that needs to be quantified when studying drugs that may have therapeutic benefits. Because of this, many drug screening assays utilize cytotoxicity as one of the critical characteristics to be profiled for individual compounds. Cells in culture are a useful model to assess cytotoxicity before proceeding to follow up on promising lead compounds in more costly and labor-intensive animal models. We describe a strategy to identify compounds that affect cell growth in a tdTomato expressing human neural stem cells (NSC) line. The strategy uses two complementary assays to assess cell number. One assay works via the reduction of 3-(4,5-dimethylthizol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) to formazan as a proxy for cell number and the other directly counts the tdTomato expressing NSCs. The two assays can be performed simultaneously in a single experiment and are not labor intensive, rapid, and inexpensive. The strategy described in this demonstration tested 57 compounds in an exploratory primary screen for toxicity in a 96-well plate format. Three of the hits were characterized further in a six-point dose response using the same assay set-up as the primary screen. In addition to providing excellent corroboration for toxicity, comparison of results from the two assays may be effective in identifying compounds affecting other aspects of cell growth.
One of the most important characteristics that needs to be determined for a chemical compound that has therapeutic potential is its toxicity to animal cells. This characteristic will determine whether a drug is a good candidate for more extensive study. In most instances, compounds with minimal toxicity are sought but there are situations in which a compound with the capacity to kill specific cell types is of interest, e.g., anti-tumorigenic drugs. Although whole animals are the best model systems to determine systemic toxicity, the cost and labor involved is prohibitive when more than a few compounds need to be tested. As such mammalian cell culture is generally used as the most efficient alternative1,2. Small to medium throughput drug screens are an important modality through which toxicity can be assessed in cell culture. These screens can be used to interrogate annotated libraries targeting individual signaling pathways. The general format of such a screen is to initially test all the compounds in the library at a single dose (generally 10 µM) in an exploratory primary toxicity screen, and then perform an in-depth secondary dose response screen to fully characterize the toxicity profile of hits from the primary screen. The methods to implement this strategy will be described here and provide a quick, efficient, and inexpensive way to identify and characterize toxic compounds.
Multiple methods have been developed to assess cytotoxicity of small compounds and nanomaterial in mammalian cells3,4. It should be noted that certain materials can interact with the assay providing misleading results, and such interactions should be tested when characterizing hits from toxicity screens4. Cytotoxicity assays include trypan blue exclusion5, lactate dehydrogenase (LDH) release assay6, Alamar blue assay7, calcien acetoxymethyl ester (AM)8, and the ATP assay9. All these assays measure various aspects of cell metabolism which can serve as a proxy for cell number. While all offer benefits, tetrazolium salt-based assays such as 3-(4,5-dimethylthizol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), 2,3-bis(2-methoxy-4-nitro-5-sulfopheny)-2H-tetrazolium-5-carboxyanilide inner salt (XTT)-1, and 4-(3-[4-Iodophenyl]-2-[4-nitrophenyl]-2H-5-tetrazolio)-1,3-benzene disulfonate (WST-1)10,11 provide good accuracy and ease of use at low cost. MTT, which will be used in this demonstration, is reduced to an insoluble formazan by a mitochondrial reductase and the rate of this conversion correlates strongly with cell number. This assay has been routinely utilized at both a small scale and for screening libraries with up to 2,000 compounds12. Direct counting of cells by a labeled marker offers another method to assess the cellular number, and unlike the MTT assay it can provide additional information about the dynamics of cellular growth. Several publicly available algorithms are available to perform automated cell count analyses and there are also proprietary algorithms that are part of software packages for imaging readers13,14. In this method description, a human neural stem cell (NSC) line that has been genetically edited to constitutively express tdTomato15 will serve as a test line to compare cellular viability results between an MTT assay and an automated cell counting assay in a screen assessing toxicity of 57 test compounds. Although the primary goal of this strategy was to identify and characterize toxic compounds, it has the additional benefit of potentially identifying growth inhibitory and growth enhancing compounds and thus provides an effective method for identifying drugs that can modulate cellular growth.
1. NSC culture
NOTE: Manipulation of a human NSC line will be described below but any cell line can be used for this protocol. All cell culture work is performed in a biological safety cabinet.
2. Treating cells with compounds
NOTE: The home-made library tested in this demonstration contains compounds that modulate wingless/integrated (Wnt), retinoic acid, transforming growth factor-beta (TGF-β), and sonic hedgehog signaling pathways as well as a variety of tyrosine kinases.
3. Imaging cells on a plate reader
4. Terminal MTT cytotoxicity assay
NOTE: Begin the MTT assay within two hours of completing tdTomato imaging.
5. Data analysis
The automated cell count data identified eleven compounds with less than 25% viability when normalized to the DMSO control while the MTT data identified these same compounds plus two additional ones (Table 1 and Table 2, shaded red). The two compounds found to be toxic only in the MTT assay (wells F3 and G10) had 31% and 39%, respectively, the number of tdTomato-positive cells as the control and by rank order were the next two most toxic compounds in this library after those deemed to be toxic. The standard deviation values for these two wells did not suggest that there was an outlier amongst the three plates that skewed the averages, and when examining the numbers for each of the three replicate plates neither compound fell below the 25% threshold on any of the plates (data not shown). Representative images of tdTomato fluorescence are shown from several wells in Figure 3. Examination of images of the two wells discordant for toxicity between the MTT and cell count assay revealed that the compounds in F3 (Figure 3B) and G10 (Figure 3C) were both toxic although in one of the three replicate plates there were a few residual live cells in well G10 (data not shown). It appears that in this instance the MTT assay was better able to score for cytotoxicity as sometimes the imager’s cell counting algorithm mistakenly counts dead/dying cells.
The MTT assay is designed for determining toxicity, but because a library may contain compounds that enhance and inhibit cell growth it would be informative to assess how well the assay quantifies both the potential growth inhibitory and proliferative effects of tested compounds. To do this a filter was used whereby compounds were classified as growth inhibitory if their normalized mean absorbances or cell counts were greater than 25% and less than two standard deviations below the control means on each of the three replicate plates (shaded yellow in Table 1 and Table 2). Eleven compounds met this criterion for the cell count assay and only two for the MTT assay with only one (E10) overlapping between the two assays although two of the eleven for the cell count assay were the ones previously mentioned to be toxic by the MTT assay (F3, G10).
Compounds whose normalized means were two standard deviations above control means on each of the three replicate plates were classified as growth enhancing (shaded green in Table 1 and Table 2). Only one compound fit this criterion for each assay and the compound did not overlap between the assays. Further examination of images of wells with discrepancies between the MTT and cell count assays indicated that in some instance wells in which MTT overestimated cell count relative to the tdTomato assay the cells appeared to be larger (Figure 3C), whereas those wells where MTT underestimated cell count relative to tdTomato the cells appeared to be smaller (Figure 3D). In summary, the tdTomato assay classified eleven compounds as toxic, eleven as growth inhibitory, and one as growth enhancing with thirty-four having no apparent effect on cell growth (Table 1). The MTT assay classified thirteen compounds as toxic, two as growth inhibitory, and one as growth enhancing with forty-one having no apparent effect on cell growth (Table 2).
A six-point dose response assay was conducted on three of the compounds identified as being toxic. These three compounds were the STAT3 inhibitors WP1066 (B5) and stattic (E4) and the epidermal growth factor receptor inhibitor tyrphostin 9 (E11). The doses were successive two-fold dilutions starting at a maximum concentration of 10 µM and going to a minimum concentration of 312.5 nM. The graph of the log of concentration versus normalized percentage of viable cells for both the cell count and MTT assays for one of these compounds (WP1066) is shown in Figure 4. The curve is relatively flat with no toxicity for the four lowest concentrations, falls rapidly at the 5 µM dose, and drops to nearly full toxicity at 10 µM. The lethal dose 50 (LD50) was calculated as 4.4 µM for the tdTomato assay and 6.0 µM for the MTT assay. The tdTomato and MTT LD50 values for the other two compounds were 3.4 µM and 4.7 µM, respectively, for static, and 0.8 µM and 1.6 µM, respectively, for tryphostin 9.
Figure 1: Plate map for master compound plate used in the primary toxicity screen. All the outer wells are shaded in grey indicating that they contained media without cells. DMSO controls (100%) are labeled in bold in wells B2, D6, and G11. All wells labeled Cmpd contained unique test compounds at 10 mM concentration in 100% DMSO. Please click here to view a larger version of this figure.
Figure 2: Plate map for the compound plate used in the dose response assay. All the outer wells are shaded in grey indicating that they contained only media without cells. The DMSO controls were housed in column 2. Triplicates of all compound/dose combinations are indicated. Please click here to view a larger version of this figure.
Figure 3: tdTomato fluorescent images of selected wells 72 hours post-treatment. (A) Well B2: DMSO control, (B) Well F3: cell count data suggested no toxicity but MTT data did, (C) Well E6: overestimated cell count by MTT relative to tdTomato count, (D) Well G6 underestimated cell count by MTT relative to tdTomato. All images were taken at 10x magnification using an RFP filter with 531/593 nm wavelength for excitation/emission. Please click here to view a larger version of this figure.
Figure 4: Curve for log concentration versus viability percent DMSO for the compound in well B5. The points on the curve represent the average normalized viable cells at six doses ± standard error of the mean for three biological replicates. Please click here to view a larger version of this figure.
Table 1: Means of tdTomato cell counts normalized to percentage DMSO control for three replicate plates ± standard deviation. Well shading indicates following: red, toxic compounds; yellow, potentially growth inhibitory; green, potentially growth enhancing. No shading indicates that compounds did not appear to affect cell growth.
Table 2: Means of MTT absorbance normalized to percentage DMSO control for three replicate plates ± standard deviation. Well shading indicates following: red, toxic compounds; yellow, potentially growth inhibitory; green, potentially growth enhancing. No shading indicates that compounds did not appear to affect cell growth.
Well | Compound | Notes |
B2 | DMSO | Negative control |
C2 | cAMP | Protein kinase A activator |
D2 | FRACTALKINE | Chemokine |
E2 | LDN212854 | Bone morphogenetic protein (BMP) receptor inhibitor |
F2 | AG370 | Platelet derived growth factor receptor (PDGFR) kinase inhibitor |
G2 | DAPT | Gamma-secretase inhibitor; neuronal differentiation positive control |
B3 | AY9944 | 7-dehydrocholestrol reductase inhibitor; hedgehog pathway inhibitor |
C3 | STA-21 | Signal transducer and activator of transcription 3 (STAT3) inhibitor |
D3 | GM-CSF | Granulkocyte-macrophage colony-stimulating factor; cytokine |
E3 | TNP470 | Methionine aminipeptidase-2 inhibitor |
F3 | BIO | Glycogen synthase kinase-3 inhibitor; WNT pathway activator |
G3 | CNTF | Ciliary neurotrophic factor; neuropeptide |
B4 | SANT | Smoothened receptor antagonist; hedgehog pathway inhibitor |
C4 | AG825 | ERBB2 inhibitor |
D4 | M-CSF | Macrophage colony-stimulating factor; cytokine |
E4 | STATTIC | Signal transducer and activator of transcription 3 (STAT3) inhibitor |
F4 | SC79 | AKT (protein kinase B) activator |
G4 | DMH1 | Bone morphogenetic protein (BMP) receptor inhibitor |
B5 | WP1066 | Signal transducer and activator of transcription 3 (STAT3) inhibitor |
C5 | INSULIN | |
D5 | IL-3 | Interleukin-3; cytokine |
E5 | AG494 | Epidermal growth factor receptor (EGFR) inhibitor |
F5 | LY294002 | Phosphoinosotide 3-kinase inhibitor |
G5 | IGF2 | insulin growth factor-2 |
B6 | SAG | Smoothened agonist; hedgehog pathway activator |
C6 | AG370 | Platelet derived growth factor receptor kinase inhibitor |
D6 | DMSO | Negative control |
E6 | EC23 | Retinoic acid receptor agonist |
F6 | TORIN2 | Mechanistic target of rapamycin (MTOR) inhibitor |
G6 | Y27362 | Rho-associated, coiled-coil containing protein kinase (ROCK) inhibitor |
B7 | CELECOXCIB | Cyclooxygenase-2 (COX-2) inhibitor |
C7 | SB525334 | Transforming growth factor beta-receptor (TGBFR) inhibitor |
D7 | DAPT | Gamma-secretase inhibitor; neuronal differentiation positive control |
E7 | CHIR99021 | Glycogen synthase kinase-3 inhibitor; WNT pathway activator |
F7 | LDN 193189 | Bone morphogenetic protein (BMP) receptor inhibitor |
G7 | TARAZOTINE | Retinoic acid receptor agonist |
B8 | AM580 | Retinoic acid receptor agonist |
C8 | DHBP | Calcium release inhibitor |
D8 | JSK | Nitric oxide donor |
E8 | DORSOMORPHIN | Bone morphogenetic protein (BMP) receptor inhibitor; 5' adenosine monophospate-activated protein kinase (AMPK) inhibitor |
F8 | IMATINIB | Tyrosine kinase inhibitor |
G8 | BMS 493 | inverse retinoic acid receptor agonist |
B9 | CYCLOPAMINE | Smoothened receptor antagonist; hedgehog pathway inhibitor |
C9 | SEMAGACESTAT | Gamma-secretase inhibitor |
D9 | BOSUTINIB | Tyrosine kinase inhibitor |
E9 | PURMORPHAMINE | Smoothened agonist; hedgehog pathway activator |
F9 | JAG | Jagged; Notch receptor agonist |
G9 | SB431542 | Transforming growth factor beta-receptor (TGBFR) inhibitor |
B10 | SC79 | AKT (protein kinase B) activator |
C10 | DANTROLENE | Ryanodine receptor antagonist |
D10 | TYRPHOSTIN46 | Epidermal growth factor receptor (EGFR) inhibitor |
E10 | AM80 | Retinoic acid receptor agonist |
F10 | IFN-Y | interferon-gamma; cytokine |
G10 | PQ401 | Insulin-like growth factor receptor (IGF1R) inhibitor |
B11 | DAPT | Gamma-secretase inhibitor; neuronal differentiation positive control |
C11 | A2M | Extracellular glycoprotein; protease inhibitor |
D11 | AG490 | Epidermal growth factor receptor (EGFR) inhibitor |
E11 | TYRPHOSTIN9 | Platelet derived growth factor receptor (PDGFR) kinase inhibitor |
F11 | BMP-2 | Bone morphogenetic protein-2 |
G11 | DMSO | Negative control |
Supplementary Table 1: List of primary screen compounds. Well location, name, and notes on each of the compounds that was used in the primary screen are provided.
The primary goal of this article was to describe a strategy that could efficiently and inexpensively identify compounds affecting cell growth in a low- to moderate-throughput screening. Two orthogonal techniques were utilized to assess cell number to increase confidence in the conclusions and offer additional insights that would not be available if only a single assay was used. One of the assays used a fluorescent cell imager to directly count tdTomato-positive cells and the second was dependent on the well-characterized ability of mitochondria to cleave MTT to formazans thus serving as a proxy for cell number10. A total of 57 test compounds were assessed in this demonstration although the MTT wing of the assay has been used for testing a library with as many as 2,000 compound14. The results of the screen pointed out how the two assays could reinforce one another in reaching certain conclusions with more confidence, and highlighted scenarios where the two assays were complementary providing additional information that would ordinarily require performing at least two separate experiments.
The most critical step in the protocol occurs just prior to plating the cells. Metabolic conditions in cell culture can become very volatile, particularly in the case of glutamine and glucose consumption, if cells are seeded at too high a density17,18. Under these conditions cell death will be due to factors inherent to the cell culture conditions and unrelated to the toxicity of tested compounds. The result will be an in increase in false positives for cytotoxic compounds as well as difficulty in reproducing results18. Success at this step requires knowing the appropriate cell density of the cell line being used, accurate determination of cell number before plating, and complete resuspension of cells to ensure homogeneous plating distribution within and across the wells of the 96-well plate. It is also important to visually confirm that cells are present at approximately the correct density 2-3 h after plating by looking at them under a microscope.
As far as the assays themselves, the most critical step for the MTT assay is ensuring that MTT is fully dissolved in the cell culture medium. Residual precipitates of MTT may by themselves result in acute cellular toxicity so it is important to completely dissolve MTT with vigorous vortexing. The most critical point for the cell counting assay is to establish the correct exposure time for imaging tdTomato. Exposure times that are too short can result in truly fluorescent cells going uncounted by the software, and exposure times that are too long can make the signal so strong that it blends neighboring cells together such that the software counts multiple cells as one cell because it is unable to resolve them14. Most software packages that come with imaging readers allow for a preview step showing which cells are being counted. It is important to run this preview step at several exposure times and pick the one that identifies fluorescent cells most accurately.
As with any method there are certain limitations to these assays. To gain higher throughput, the primary toxicity assay uses only a single treatment/single dose paradigm which can come at the cost of more false positives and false negatives. Additionally, although several earlier studies have shown that MTT correlates very well with cell number when using either a colony forming assay19 or a thymidine incorporation asssay20, treatment with certain compounds can either enhance or inhibit mitochondrial activity in such a way that the results of the MTT assay no longer correlate with cell number21. Results from this demonstration indicate that while MTT is excellent at identifying toxic compounds, its ability to identify compounds that either inhibit or enhance proliferation is limited perhaps because such compounds alter mitochondrial activity in a manner that it correlates less well with cell number. There are also some limitations to the tdTomato counting assay. An obvious limitation is the need to have a cell line stably expressing a fluorescent protein. Recent advances in genome manipulation have made it much easier to develop such lines but the work required to generate them may be beyond the capabilities of some labs. From a technical standpoint, the biggest issues with any cell counting assay that uses image analysis is the inability of these assays to distinguish between cells that are clustered together resulting in an undercount14, therefore, proper plating is critical for accurate results. Another potential problem is the counting of dead or dying cells that fluoresce brightly. One way to avoid this problem is to wash cells with PBS before counting to remove these background cells. This may not be convenient for certain less well adhering cells lines as live cells may detach upon washing. An alternative solution to this problem is to utilize the flexibility inherent in many analysis programs to customize the parameters for cell identification within a narrow range so that only live, fluorescent cells are counted.
The strategy described in this article provides a powerful way to efficiently screen up to several hundred compounds. The MTT assay readout is the physiological result of mitochondrial activity and can have cell line or compound-specific effects that can produce inaccurate results4,21. By combining it with a cell counting assay using a fluorescent reporter, these limitations can be greatly mitigated. As shown, comparing the results of both assays can result in close to 100% accuracy in identifying toxic compounds. A previous study has shown that in an HEK293T line stably expressing tdTomato, there is high IC50 correlation between MTT and tdTomato for a library of toxic compounds22. Although this study did not run secondary confirmations on enough hit compounds to perform a similar concordance analysis, the calculated LD50 values for the three compounds that were tested were similar.
In addition to their ability to reinforce conclusions about toxicity, the two assays can complement one another when addressing the potential growth inhibitory and proliferative effects of test compounds. For several test compounds the data between the two assays diverged substantially. When examining images of tdTomato fluorescence for some of these compounds, there were noticeable morphological changes between treatment and control. This suggests that the divergence in the normalized values between the two assays may be based upon physiological changes that differentially affect the MTT readout and the tdTomato cell count. The ability to acquire such data with a single experiment greatly increases the robustness of this strategy making it more generally applicable. As such, it has the capacity not only to identify toxic compounds with high accuracy but to point out compounds with more subtle effects on cell growth/physiology that can be more extensively characterized.
The authors have nothing to disclose.
This work was supported by the NINDS Intramural Research Program.
B-27 (50X) | ThermoFisher Scientific | 17504001 | Neural stem cell medium component. |
BenchTop pipettor | Sorenson Bioscience | 73990 | Provides ability to pipette compound library into a 96-well plate in one shot. |
BioLite 96 well multidish | Thermo Scientific | 130188 | Any 96 well cell culture plate will work. We use these in our work. |
Cell culture microscope | Nikon | Eclipse TS100 | Visual inspection of cells to ensure proper density. |
Cytation 5/ Imaging reader | BioTek | CYT3MFV | Used for cell imaging and absorbance readings. |
DMSO | Fisher Scientific | 610420010 | Solvent for compounds used in screen. Dissolves MTT precipitates to facilitate absorbance measurements. |
FGF-basic | Peprotech | 100-18B | Neural stem cell medium component. |
GelTrex | ThermoFisher Scientific | A1413202 | Neural stem cell basement membrane matrix. Allows cells to attach to cell culture plates. |
Gen5 3.04 | BioTek | Analysis software to determine cell counts for tdTomato expressing cells. | |
Glutamine | ThermoFisher Scientific | 25030081 | Neural stem cell medium component. |
Microtest U-Bottom | Becton Dickinson | 3077 | Storage of compound libraries. |
MTT | ThermoFisher Scientific | M6494 | Active assay reagent to determine cellular viability. |
Multichannel pippette | Rainin | E8-1200 | Column-by-column addition of cell culture medium, MTT, or DMSO. |
Neurobasal medium | ThermoFisher Scientific | 21103049 | Neural stem cell base medium. |
RFP filter cube | BioTek | 1225103 | Filter in Cytation 5 used to image tdTomato expressing cells. |
TrypLE | ThermoFisher Scientific | 12605036 | Cell dissociation reagent. |