The protocol describes a rapid, high-throughput, reliable, inexpensive, and unbiased assay for efficiently determining cellular viability. This assay is particularly useful when cells' mitochondria have been damaged, which interferes with other assays. The assay uses automated counting of cells stained with two nuclear dyes – Hoechst 33342 and propidium iodide.
The contribution of mitochondria to oncogenic transformation is a subject of wide interest and active study. As the field of cancer metabolism becomes more complex, the goal of targeting mitochondria using various compounds that inflict mitochondrial damage (so-called mitocans) is becoming quite popular. Unfortunately, many existing cytotoxicity assays, such as those based on tetrazolium salts or resazurin require functional mitochondrial enzymes for their performance. The damage inflicted by compounds that target mitochondria often compromises the accuracy of these assays. Here, we describe a modified protocol based on differential staining with two fluorescent dyes, one of which is cell-permeant (Hoechst 33342) and the other of which is not (propidium iodide). The difference in staining allows living and dead cells to be discriminated. The assay is amenable to automated microscopy and image analysis, which increases throughput and reduces bias. This also allows the assay to be used in high-throughput fashion using 96-well plates, making it a viable option for drug discovery efforts, particularly when the drugs in question have some level of mitotoxicity. Importantly, results obtained by Hoechst/PI staining assay show increased consistency, both with trypan blue exclusion results and between biological replicates when the assay is compared to other methods.
The first step to identifying effective cancer treatments is the selection of a robust, unbiased cytotoxicity assay that can be used to examine the effect of treatment. A common choice for low-throughput experiments is the exclusion of trypan blue dye from living cells. This method is favored because it allows a relatively unbiased method for quantifying cell survival. trypan blue passively diffuses into cells whose membranes are compromised, but it is effectively blocked from entering healthy cells1. The quotient of the living cells and the total cells represents the percent viability, which indicates the efficacy of the treatment. The most significant disadvantage of the trypan blue assay is that it is poorly suited for high-throughput methodologies. It has a relatively low signal-to-noise ratio and prolonged staining can result in artifacts due to the staining of viable cells. Consequently, trypan blue exclusion is typically, but not always2, relegated to manual counting. This makes it too slow and introduces the strong possibility of bias due to subjective judgment of the researcher (unless blinding or independent counts are used, which further reduce laboratory throughput). In general, the throughput of this assay is insufficient for modern drug discovery.
Viability assays, which generally have a much higher throughput, allow researchers to circumvent this limitation, but come with significant caveats (see Table 1). These methods generally fall into two groups. One group is comprised of colorimetric assays that are based on the function of cellular redox enzymes. Colorless or non-fluorescent substrates are converted into vibrant products that can be quantified using a spectrophotometer. Classic examples include tetrazolium salts (MTT, WST-1, XTT, etc.) and resazurin. This category also includes luminescent assays that utilize luciferin to evaluate ATP level. Assays of this type have the underlying limitation that they are measuring cellular metabolism, which is not cellular viability per se. It is quite common for cells to become quiescent under adverse conditions, but still retain the ability to divide3,4,5. For example, cancer stem cells are often relatively metabolically quiescent6,7,8,9, and are likely to be difficult to assay using these techniques. Effectiveness of treatments that damage mitochondrial function, such as most mitocans, is also likely to be significantly overestimated.
An alternative methodology leverages the chemical properties of various substances that allow them to either cross or not cross biological membranes. One example are nuclear stains such as SYTOX or propidium iodide (PI). This category also includes assays that are similar in concept but different in function, such as the lactate dehydrogenase (LDH) assay, which measures the release of LDH into the extracellular milieu as an indicator of cellular necrosis (Figure 1, Table 1). These assays are more capable of distinguishing between metabolically inactive and dead cells.
Assay/dye | Type(s) of cell death detected | Necessary equipment | Key features |
MTT, CKK-8, Alamar Blue (resazurin) | Apoptosis/Necrosis | Spectrophotometer | Inexpensive, rapid; endpoint assay; dependent on enzymes' activity (exclusively mitochondrial in case of MTT) and does not discriminate between modes of cell death1,10 |
LDH release | Necrosis | Spectrophotometer | Rapid, independent of mitochondrial enzymes’ activity; expensive for high-throughput tests; detects necrotic cells with compromized plasma membrane11,12 |
Trypan Blue (TB) | Apoptosis/Necrosis | Microscope | Cell-impermeant; does not discriminate between modes of cell death; laborous and not suitable for high-throughput screening; more difficult to use with adherent cells; prone to subjective judgment of the user, but is considered the standard cell viability measurement method13 |
Acridine orange (AO) | Apoptosis/Necrosis/ | Fluorescence microscope | A nucleic acid dye with unique spectral properties, can distinguish between apoptosis and necrosis/necroptosis14 |
Necroptosis | |||
Hoechst 33342, DAPI | Apoptosis | Fluorescence microscope or flow cytometer | Cell-permeable; inappropriate on its own to monitor cell death; useful for co-staining; can be used to assess chromatin condensation and nuclei fragmentation in early apoptosis; can be paired with propidium iodide to distinguish apoptosis from necrosis15,16 |
Propidium Iodide (PI) | Late apoptosis/Necrosis | Fluorescence microscope or flow cytometer | Cell-impermeant intercalator; detects both late apoptosis and necrosis modes of cell death17. Toxic and permeable after long incubation times18 |
Table 1. List of cytotoxicity assays. Cytotoxicity assays, some of which were used in this study, listed along with the brief description of their key features.
Recent studies have demonstrated that mitochondrial metabolism is altered in some cancers19,20,21,22,23,24,25. For example, acute myeloid leukemias (AML) have been shown to upregulate their mitochondrial mass, mtDNA content, and mitochondrial respiration to meet their energy demands19,26,27. On the other hand, some solid tumors are characterized by mitochondrial dysfunction, or rather “metabolic reprogramming”, such as downregulation of mitochondrial proteins involved in OXPHOS or decreased mtDNA content, that has been associated with tumor invasiveness, metastatic potential and resistance to apoptosis-inducing drugs28,29. Furthermore, recently, there has been an increased interest in using mechanistically diverse compounds that affect mitochondrial function (generally called mitocans30), as potential therapies for particular cancers. These drugs target ATP synthesis, mitochondrial DNA, OXPHOS, and ROS production, as well as pro-apoptotic and anti-apoptotic proteins associated with mitochondria30,31. Several studies have shown that this approach has significant promise19,32,33,34. However, these metabolic deviations in cancer cell biology or mitochondria-targeting treatments may significantly affect conventional viability assays that are based on mitochondrial functionality.
Here, an optimized protocol for a differential nuclear staining assay is described. The protocol allows fast and accurate determination of cytotoxicity of mitocans or their combinations with other compounds. Hoechst 33342 is a cell-permeant nuclear dye that readily crosses cell membranes to stain DNA, allowing the total cell count to be obtained. By co-staining with PI, which only enters the nuclei of dead cells, the proportion of living (Hoechst only) and dead (stained with both) cells can be accurately determined. This protocol refines the published assay35 by adding a step for the optimization of the dye concentration (by cross-referencing results with orthogonal trypan blue method) and centrifugation of the plate prior to imaging. Since many cell lines are semi-adherent or suspended, centrifugation increases the proportion of cells that are imaged and strongly improves accuracy. The assay has several advantages, including that staining does not require removal of media or washing. The dye mixture is also inexpensive, easy to prepare, and compatible with multichannel/robotic pipetting systems.
After cells have been stained, they are imaged with an automated microscope. This has the added advantage of creating a permanent record of the images that can be re-analyzed later and the effects of particular compounds can be re-evaluated by visual inspection of captured images. Once images have been obtained, cells can be counted either manually or by using any of several software packages, including both free (e.g., ImageJ, CellProfiler, etc) and commercial software (e.g., Metamorph, Gen5, etc). Automated cell counting is generally preferable since properly developed automated cell counting pipelines are more accurate and less biased than manual counts. They also more effectively disregard cell debris or insoluble complexes. Development of these pipelines is generally straightforward and is simplified by the efficiency of the stains used. The output is quantitative since the actual number of dead cells is automatically calculated with respect to total cell number, and different thresholds can be applied to increase or decrease the stringency of detection35. For convenience, optimized parameters for counting cells using Gen5 v. 3.00 software compatible Cytation 5 Cell Imaging Multi-Mode Reader are included.
1. Cytotoxicity Assay: Setup
2. Cytotoxicity Assay: staining with Hoechst 33342 and propidium Iodide
3. Cytotoxicity Assay: data acquisition
The aforementioned protocol has been developed using OCI-AML2 cells, which were taken as a representative acute myeloid leukemia cell line. AML is characterized by abnormal proliferation of undifferentiated and non-functional hematopoietic cells in the bone marrow26. Despite recent developments in AML targeted therapy, the standard of care has remained unchanged for several decades, and consists of induction therapy (typically comprised of three days of anthracycline, e.g., daunorubicin, idarubicin, or anthracenedione mitoxanthrone, and 7 days of cytarabine) followed by consolidation (typically comprised of rounds of cytarabine treatment followed by periods of recovery)36.
Following the protocol outlined in Figure 1A, OCI-AML2 cells were seeded in 96-well plates and treated with the mitochondrial uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP) or the glycolytic inhibitor 2-deoxy-D-glucose (2-DG) at a range of concentrations. Cells were treated for 24 h at 37 °C in serum-free RPMI-1640 media, and then viability was assessed using trypan blue (TB) exclusion or one of four viability assays (Hoechst/PI differential nuclear staining, LDH release assay, MTT, or alamarBlue). Representative images of cells stained with Hoechst/PI are shown in Figure 2A. Several important observations can be made immediately. First, the total number of cells (Hoechst staining) is reasonably high and is markedly greater than the number of dead cells (PI staining). This suggests that the media conditions are not triggering high rates of cell death. Second, since only cells being labeled with both Hoechst and PI are counted as dead (purple cells in the merged image), the probability of counting debris is very low. This image shows a good example of properly stained cells.
Figure 1. Experimental timeline and comparison of existing cytotoxicity assays. (A) Flowchart summarizing the timeline for the experimental procedure, e.g. Hoechst/PI staining. (B) Comparison of cytotoxicity assays, some of which were used in this study. Dye exclusion assays involve impermeant nuclear dyes that stain dead cells with compromised plasma membrane: TB – trypan blue, PI – propidium iodide, EtBr – ethidium bromide, and SYTOX. Second group of assays depend on cellular metabolism, for example, tetrazolium salts MTT, XTT, and CKK-8 (WST-8), resazurin-based reagent alamarBlue, etc. Please click here to view a larger version of this figure.
Figure 2. Comparison of a panel of cytotoxicity assays with trypan blue exclusion. (A) Representative images of total (Hoechst 33342) and dead (propidium iodide, PI) OCI-AML2 cells via Hoechst/PI staining. (B) Assessment of viability of OCI-AML2 cells using different methods after treatment with a gradient of CCCP (top) or 2-DG (bottom) concentrations. OCI-AML2 were treated with CCCP or 2-DG in serum-free RPMI-1640 for 24 h prior to determination of cell viability. Shown is mean, error bars represent SEM. C. Difference in cell viability between cytotoxicity assays in (B) vs. trypan blue staining (see Supplementary Tables S1-2 for the exact numbers and median difference). Stars indicate significant difference vs. trypan blue staining. ** p < 0.01, *** p < 0.001, ns – non-significant. Group comparison was done via t-test with correction for multiple hypothesis testing. Three independent biological replicates were performed. Please click here to view a larger version of this figure.
As we recently reported19, leukemias are very sensitive to mitotoxic treatments, which indicates that the cells already have underlying mitochondrial damage. On this basis, we predicted that the MTT and alamarBlue assays, which are based on mitochondrial enzyme activity, would inaccurately measure cellular viability. As expected, these assays (especially alamarBlue) showed significantly lower viability compared to trypan blue exclusion, see Figure 2B,C, Supplementary Tables S1-S2). This is consistent with the fact that doses of these compounds required to induce apoptosis or necrosis are higher than those needed to compromise mitochondrial function.
Amongst the assays that were tested, dual staining with Hoechst 33342 and PI had the best combination of robustness, sensitivity, and consistency with TB staining, with the smallest median deviation from TB exclusion method after CCCP or 2-DG treatment (Supplementary Tables S1-2). Interestingly, at higher doses of CCCP (50 and 100 µM) viability estimated with Hoechst/PI or TB was increased compared to lower doses (Figure 2B, top). This is likely due to the precipitation of CCCP at higher doses due to its hydrophobicity, reducing its effective concentration and impact on cells. Increasing the dose of CCCP further reduced Hoechst/PI-estimated mean viability of OCI-AML2 cells, however: 74% at 150 µM, 62% at 200 µM, 6% at 300 µM (data not shown).
The Hoechst/PI assay was also effective at determining cellular viability after treatment with other mitochondria-targeting molecules. These included rotenone, a poison that compromises Complex I of the mitochondrial electron transport chain37, and 3-bromopyruvate, a glycolytic inhibitor and alkylating agent that impairs mitochondrial respiration and mitochondrial metabolism38 (Figure 3A-D, Supplementary Tables S3-4).
Figure 3. Validation of Hoechst/PI cytotoxicity assay in leukemia cells. (A,B) OCI-AML2 (AML) cells were treated with different concentrations of rotenone (A) or 3-bromopyruvate, 3-BP (B) in serum-free RPMI-1640 media for 24 h, then viability was determined. Shown is mean with SEM. (C,D) Comparison of viability difference between Hoechst/PI assay vs. trypan blue staining (for the cells in A-B, see Supplementary Tables S3-4 for quantitation). Stars indicate statistical significance vs. trypan blue staining. ** p < 0.01, *** p < 0.001, ns – non-significant. Group comparison was done via t-test with correction for multiple hypothesis testing. E. MOLM-13 (AML), primary AML cells isolated from a patient, MOLT-4 (ALL), and K562 (CML) cells were treated with the indicated concentrations of rotenone in serum-free RPMI-1640 media for 24 h, prior to viability determination using Hoechst/PI staining. Shown is mean with SEM. Three independent biological replicates were performed. Please click here to view a larger version of this figure.
The Hoechst/PI cytotoxicity assay was further validated using a panel of leukemia cell lines representing multiple types of leukemia, as well as primary cells. They included MOLM-13 (an AML cell line), primary AML cells derived from a representative patient sample, MOLT-4 (an acute lymphoblastic leukemia cell line, ALL), and K562 (a chronic myelogenous leukemia cell line, CML) (Figure 3E, Supplementary Figure S1). Results of the Hoechst/PI assay showed that these cells had profound differences in rotenone sensitivity, ranging from very sensitive (MOLT-4) to resistant (K562) cells.
To demonstrate the robustness of the assay, OCI-AML2 cells were treated with a concentration gradient of 3-bromopyruvate, as described in the protocol above. Cell counts were collected for 6 wells at each concentration and are shown (Figure 4A,B). These counts were used to calculate viability, and showed that as few as 4 wells were sufficient to accurately capture assay outcome (Figure 4C). The resulting dose-response curve is shown (Figure 4D).
Figure 4. Reproducibility of Hoechst/PI staining. Viability of OCI-AML2 cells after 24 h of treatment with different concentrations of 3-bromopyruvate (3-BP). (A) Total number of cells counted via Hoechst 33342 staining, (B) Number of dead cells counted via propidium iodide staining. (C) Viability (%) calculated using (A-B). (D) Representative dose-response graph. Shown is mean with SEM. Three independent biological replicates were performed. Please click here to view a larger version of this figure.
Although the assay is relatively robust, care must still be taken. Improper performance of the assay may compromise its accuracy. Several compromised outcomes are shown for troubleshooting purposes (Figure 5). The first set of images shows the consequences of over-staining with PI, which may arise from the following sources: using the dye at too high concentration, staining for too long, or using too high LED intensity/integration time in the red channel (Figure 5A). These errors will generate artificially inflated numbers of dead cells. The second issue arises from neglecting the centrifugation step. Often, dead cells begin to lose their attachment from the surface of the dish. Consequently, they will be underrepresented during image acquisition, which generally occurs near the bottom of the well (Figure 5B). Finally, overexposure in the Hoechst channel artificially expands the size of the cells, significantly reducing the total counts per well and limiting the assay power (Figure 5C).
Figure 5. Comparison of optimal and sub-optimal assay parameters. Comparison of results acquired via optimized parameters (left) or sub-optimal parameters (right). (A) Healthy PBMCs were stained with propidium iodide at 1 µg/mL (left) or 5 µg/mL (right) for 15 min before imaging. (B) Images of OCI-AML2 cells taken with (left) or without (right) plate centrifugation. (C) Images of OCI-AML2 cells acquired with optimal (left) or excessive (right) integration time for the Hoechst channel. Please click here to view a larger version of this figure.
Difference with trypan blue method, % | ||||
CCCP, µM | Hoechst/PI | LDH assay | MTT | Alamar Blue |
3.1 | -1.19 | 33.00 | -31.05 | -59.04 |
6.3 | 6.99 | 41.90 | -11.03 | -59.34 |
12.5 | 0.71 | 41.60 | 4.31 | -52.52 |
25 | 30.73 | 66.33 | -3.68 | -51.85 |
50 | 47.38 | 50.71 | -17.71 | -60.27 |
100 | 13.81 | 17.12 | -45.31 | -67.25 |
Median | 10.40 | 41.75 | -14.37 | -59.19 |
Supplementary Table S1. Viability differences for a panel of assays in OCI-AML2 cells after CCCP treatment. Assessment of viability was performed after 24 h of treatment with different concentrations of CCCP. Median viability differences between tested assays (Hoechst/PI, LDH, MTT, or alamarBlue) and trypan blue staining with automated counting was calculated based on viability differences at each drug concentration.
Difference with trypan blue method, % | ||||
2-DG, mM | Hoechst/PI | LDH assay | MTT | Alamar Blue |
3.1 | 16.16 | 20.77 | 5.82 | -20.27 |
6.3 | 22.93 | 31.96 | -11.13 | -31.11 |
12.5 | 30.85 | 45.07 | -34.27 | -36.29 |
25 | 13.81 | 42.39 | -55.79 | -52.29 |
50 | 16.96 | 81.12 | -29.38 | -47.68 |
100 | 7.79 | 149.16 | -19.37 | -28.67 |
Median | 16.56 | 43.73 | -24.38 | -33.70 |
Supplementary Table S2. Viability differences for a panel of assays in OCI-AML2 cells after 2-DG treatment. Assessment of viability was performed after 24 h of treatment with different concentrations of 2-DG. Median viability differences between tested assays (Hoechst/PI, LDH, MTT, or alamarBlue) and trypan blue staining with automated counting were calculated based on viability differences at each drug concentration.
Rotenone, µM | Viability difference, % (Hoechst/PI-trypan blue) |
10 | 4.45 |
25 | 7.61 |
50 | 17.70 |
100 | 7.74 |
150 | 56.38 |
200 | 16.57 |
Median | 12.15 |
Supplementary Table S3. Viability difference between Hoechst/PI and TB exclusion method in OCI-AML2 cells after rotenone treatment. Assessment of viability was performed after 24 h of treatment with different concentrations of rotenone. Median viability difference between Hoechst/PI and trypan blue staining with automated counting was calculated based on viability differences at each drug concentration.
3-BP, µM | Viability difference, % (Hoechst/PI-trypan blue) |
5 | 5.73 |
10 | 9.00 |
25 | 7.72 |
50 | -1.90 |
100 | -28.23 |
200 | -20.77 |
Median | 1.91 |
Supplementary Table S4. Viability difference between Hoechst/PI and TB exclusion method in OCI-AML2 cells after 3-bromopyruvate (3-BP) treatment. Assessment of viability was performed after 24 h of treatment with different concentrations of 3-BP. Median viability difference between Hoechst/PI and trypan blue staining with automated counting was calculated based on viability differences at each drug concentration.
Supplementary Figure S1. Representative images of cells stained with Hoechst/PI, with or without rotenone treatment at indicated concentrations in serum-free RPMI-1640 media for 24 h: A-B. AML: MOLM-13 cell line (A), and primary AML cells (B) derived from a representative patient sample. C. ALL: MOLT-4 cell line. D. CML: K562 cell line. Please click here to download this file.
Although the protocol for Hoechst/PI cytotoxicity assay is robust and requires comparatively little hands-on time, there are several experimental details that are very important to ensure accurate results. First, it is essential to make sure that the concentration of DMSO remains below 0.5% (v/v). It is generally agreed that exposure to even low doses of DMSO can substantially alter the morphology and attachment of cells and significantly delay cell cycle progression39,40.
Second, the staining should be performed as soon as possible after treatment. Since there are no washing steps, the compound remains in the wells and can continue to inflict damage on the cells. In comparison to the earlier protocol35, the cells were stained for only 15 minutes. This limits the possibility that viable cells could be inadvertently stained with propidium iodide. For this reason, we suggest staggering treatments and staining if more than two plates will be treated. This allows for imaging time between plates and improves reproducibility.
Third, appropriate dye concentration is dependent on cell type. It is important to empirically determine the optimized propidium iodide concentration, beginning with untreated cells, using trypan blue staining as an orthogonal reference.
Finally, the centrifugation step immediately before visualization is also critical. Based on this protocol, a range of 500-4,000 cells is recorded (depending on seeding density). This is a notable improvement over the 100-400 cells per well used previously35. Considering the variation in the population of cancer cells (especially primary cells), having more cells for analysis is important, and may allow more robust data analysis.
Due to the relative simplicity of the method, a wide variety of modifications can easily be performed. For example, many other cell-impermeant dyes are available from a number of vendors and can be substituted for propidium iodide. While this substitution has relatively little value on its own, the increased palette of colors means that more complex experiments can be performed. For example, the addition of a third color will allow differential survival to be assessed between subpopulations of cells, as is commonly performed with flow cytometry.
A more substantial modification of the protocol is foregoing cell visualization and using a spectrophotometer instead. This approach has the advantage of utilizing a nearly ubiquitous piece of equipment (a standard spectrophotometer with a microplate reader) and is the fastest method to acquire data. However, this method is far less accurate. Fluorescence intensity is representative of cell staining, but stochastic variations in staining intensity, combined with the relatively smaller sample area, introduce significant variability. While further optimization (such as the inclusion of a fixative, additional washing steps, or a larger number of samples) may address these problems, fluorescence microscopy is generally a superior approach.
In conclusion, the modified Hoechst/PI protocol is a fast, accurate, inexpensive, high-throughput cytotoxicity assay that is independent of mitochondrial function. This assay has substantial utility for efficiently screening compounds or compound combinations that target mitochondria.
The authors have nothing to disclose.
NVK, a CPRIT scholar in Cancer Research, thanks the Cancer Prevention and Research Institute of Texas (CPRIT) for their generous support, CPRIT grant RR150044. This work was also supported by the Welch Foundation Research Grant C-1930, and by the National Institutes of Health R35 GM129294 awarded to NVK. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
2-Deoxy-D-glucose/2-DG | Chem-Impex | 50-519-067 | |
3-bromo-pyruvate | Alfa Aesar | 1113-59-3 | |
96-Well plates | Greiner Bio-One | 655090 | Black or clear flat-bottomed 96-well plates |
Alamar blue HS cell viability reagent (100mL) | Thermo Fisher | A50101 | |
Countess II automated cell counter | Thermo Fisher | ||
Cytation 5 Cell Imaging Multi-Mode Reader | BioTek | ||
Hoechst 33342 | Thermo Fisher | 62249 | 20 mM solution; final concentration 1:1,000 |
HyClone fetal bovine serum | GE Healthcare | #25-514 | |
m-chlorophenylhydrazone/CCCP | Sigma Aldrich | C2759 | |
PBS tablets | Thermo Fisher | BP2944100 | 1 tablet + 200 mL of sterile water = 1x PBS solution |
Penicillin-Streptomycin-Glutamine (100X) | Gibco | 10378016 | |
Pierce LDH assay kit | Thermo Fisher | 50-103-5952 | |
Propidium Iodide | Thermo Fisher | 50-596-072 | Dry powder; stock 1 mg/mL in PBS; final concentration 5 µg/mL (leukemia cells), 1 µg/mL (normal PBMCs) |
Rotenone | Ark Pharm | AK115691 | |
RPMI-1640 Medium With L-glutamine and sodium bicarbonate, liquid, sterile-filtered, suitable for cell culture |
Sigma Aldrich | R8758-500ML | |
Thiazolyl blue tetrazolium bromide | ACROS Organics | AC158990010 | |
Trypan blue stain (0.4%) | Gibco | 15250-061 | |
Cell lines | |||
K562 | ATCC | CCL-243 | CML cell line |
MOLM-13 | ATCC | AML cell line |
|
MOLT-4 | ATCC | CRL-1582 | ALL cell line |
OCI-AML2 | ATCC | AML cell line |