HepGentox: a novel promising HepG2 reportergene-assay for the detection of genotoxic substances in complex mixtures

Cell line

HepG2 (ATCC HB-8065, CVCL_0027) cells were stably transfected with a p53 reporter construct using the PiggyBac transposon system (Wilson, Coates & George, 2007). For this, a pGVL8 backbone was used (Mertl et al., 2019), with a six times multimerized p53 binding site from GADD45 (sense: GAACATGTCTAAGCATGCTG) (Hollander et al., 1993). The development of the HepGentox cell line was based on previous reporter optimizations for different signaling pathways (Mertl et al., 2019; manuscript in preparation: Steurer, 2020). A six times multimerized p53 binding site was introduced upstream of an Nluc reporter gene. We chose the short-lived NlucPAU (NanoLuc containing mRNA and protein destabilizing sequences Steurer et al., 2018) to reduce the background signal (= no accumulation) and obtain high induction rates after a short incubation time at lower cytotoxic side effects. The construct was stably integrated into HepG2 cells and one clone was selected as the HepGentox cell line.

The cells were cultivated in DMEM, substituted with 10% FBS and 1% Pen/Strep at 37 °C and 5% CO₂. Individual clones were raised and tested for their performance. Through initial experiments with a luciferase assay the maximum induction of several clones were tested with a selection of genotoxic substances. In a next step, promising clones were screened concerning their LEC values and the most suitable clone was selected. Cells were frozen at a passage of three and used up to a maximum of 12 passages. The clones were selected by adding puromycin to ensure the stability of the cell line and the inserted construct. Further, induction levels, response of the negative and positive controls and background results were closely monitored throughout the course of this study to test for the cell line’s stability. For testing of pure substances, the cells were seeded at a concentration of 2 × 10⁴ cells/well in a 96 well plate with 100 µL of cell suspension per well. After 24 h, the cells were treated with the genotoxic substance and the following day the cell response was measured. For substance treatment, DMSO was used as a solvent vehicle and applied at a maximum of 1% in DMEM, supplemented with 5% FBS. A maximum substance concentration of 1 mM in the well was chosen. If cytotoxic effects or precipitation/insolubility was observed, the concentrations were altered accordingly.

Optimisation experiments

For optimisation experiments, the cells were treated with the pure substances 4NQO at a top concentration of 0.63 µM and BαP at 10 µM solved in DMSO. As a vehicle control 1% DMSO was used and the DMSO concentration was steady over the whole plate. To determine the optimal cell concentration, the cells were cultivated as described above and 100 µL of a cell suspension was seeded in 96 wells plate with 1 × 10⁴, 2 × 10⁴, 4 × 10⁴, 6 × 10⁴, 8 × 10⁴ and 1 × 10⁵ cells/well. The cells were treated with 4NQO and BαP and incubated for 24 h before measurement. For incubation time experiments, the cells were seeded at 2 × 10⁴ cells/well in a 96 well plate and incubated with 4NQO and BαP for 2, 6, 24, 48 or 72 h until measurement. To determine the optimal FBS concentration, the cells were seeded at 2 × 10⁴ cells/well in a 96 well plate and treated with 4NQO or BαP solved in DMEM supplemented with 5%, 10% and 15% FBS for each plate and measured after 24 h of incubation. For DMSO experiments, 2 × 10⁴ cells/well were seeded in a 96 well plate and treated with 4NQO and BαP solved in DMEM. Over half a plate, a DMSO concentration of 0.25, 0.5, 1.0, 1.5 and 2.0% was ensured and the vehicle control was adjusted accordingly. Measurements were done after 24 h of incubation.

Measurement

Viability was determined using a resazurin assay as described previously (Riegel et al., 2017) prior to luciferase measurement with a multiplate reader Infinite^®200 Pro (Tecan, CH). single NanoLuc measurement was performed as described in Steurer et al. (2018) using a Luminoskan™ Microplate Luminometer (Thermo Fisher, Waltham, MA, USA). For viability measurement, resazurin was diluted in 1xPBS and added in the wells to a final concentration of 5 µM in the plate. The plates were further incubated with the resazurin for 1 h, before measurement with an Infinite^®200 Pro (Tecan, CH) multiplate reader at excitation wavelength 540 nm and emission wavelength 590 nm. For viability, a threshold of 70% was used. For evaluation, a threshold of 1.7 was applied, which was determined through statistical analysis of blank values (= vehicle control with 1% DMSO) by addition of three times the standard deviation. In these experiments a fold induction of 0.7 for the vehicle control was found with a standard deviation of 0.312. This data was obtained from two individual blank experiments (192 wells in total) and the background data of 113 experiments (12 wells each). A fold induction of a substance or sample above the threshold of 1.7 was considered as positive.

S9 experiments

For metabolization experiments, 1254 aroclor induced S9 rat liver extract was used (Moltox, NC, USA) and cofactors nicotinamide adenine dinucleotide phosphate (NADPH), Glucose-6-Phosphate (G6P) and MgCl₂ were purchased from Carl Roth (Karlsruhe, GER) and Glucose-6-Phosphate-Dehydrogenase (G6P-DH) from Sigma Aldrich (US). Two different S9 protocols were followed, with different S9 composition depending on the incubation time with the S9 mixture. Final concentration of the compounds in the wells were: 5 mM MgCl₂, 3 mM G6P, 0.2 mM NADPH, 0.3 units/mL G6P-DH and 330 µg/mL (3 h protocol) or 10 µg/mL (24 h protocol according to Mollergues et al. (2016)) of S9 liver extract. Cells were either treated for 3 h with a higher concentrated S9 mix, then washed with Dulbecco’s phosphate buffered saline (DPBS) and further incubated with DMEM containing 5% FBS and 1% DMSO for another 21 h. Alternatively, treatment was done for 24 h with a lower concentrated mix, without a change of medium. Luciferase and resazurin measurements were conducted the same way as without S9 addition. The 1254 aroclor induced S9 rat liver extract was used simultaneously in the same laboratory for the Ames MPF™ assay to prove its functionality.

Complex mixtures

For testing of complex mixtures, the cells were cultivated as described above and treated with 1% of an FCM sample migrate solved in DMSO. The FCM migrate was produced through migration and concentration of polyethylene, following the protocol by Rainer et al. (2019). Upon addition to the HepGentox, the sample was spiked with 4NQO or BαP in a range where a positive response was expected. The spikes were solved in DMEM with additional 1% DMSO, therefore the DMSO concentration remained at 1% over the whole plate.

Results – assay optimization –cell number and incubation effects

A low cell number is leading to a higher amount of substance per cell. To observe if this can be directly translated into a lower LEC value in the assay we tested 10,000 to 100,000 cells per well in a 96 well plate. The results in Figs. 1A and 1B clearly show, that a low cell number led to a LEC value of 0.16 µM for 4NQO and 0.63 µM for BαP, compared to the highest cell concentration of 100,000 cells per well, with four times higher LEC values of 0.63 µM and 2.5 µM, respectively. This was the case for both substances; which may or may not need metabolic activation. Of course, a higher amount of substance per cell might also result in greater cytotoxicity, therefore viability was closely observed in parallel. A threshold of 70% was taken as a limit for the viability. For 4NQO, this limit was reached earlier with lower cell concentrations (2 to 4-fold compared to higher cell concentrations). However, for BαP, the viability was stable through all concentrations (Figs. S1A and S1B). A concentration of 2 × 10⁴ cells/well was chosen as optimum, as here the LEC value was low at 0.31 µM for 4NQO and 0.63 µM for BαP. Further, the viability was considered to be reasonably stable at higher concentrations of genotoxic substances as it remained above the 70% threshold.

Genotoxic substances have very heterogeneous chemical properties and therefore cover a wide variety of modes of action (MoA). Further, the MoA together with differences in the kinetics of the cellular uptake greatly influences the kinetics of the induced DNA damage and the cellular response. To analyze the influence of the incubation time on the resulting LEC values, the HepGentox cells were tested after 2, 6, 24, 48 and 72 h treatment with the model substances 4NQO or BαP (Figs. 1C and 1D). The experiment clearly showed that substances, which have a genotoxic effect independent of a metabolic activation system, such as 4NQO affected the cells shortly after substance treatment, as a signal could already be seen after 6 h with a LEC value of 0.08 µM and after 24 h with a LEC of 0.16 µM (Fig. 1C). However, at later time points induction above the threshold was no longer observed. Contrary, for BαP a signal was observed only after 24 h with a LEC of 0.31 µM or more (Fig. 1D). Further, viability dropped at higher 4NQO concentrations with increasing time at 0.31 µM below the 70% threshold, which was also observed for BαP at a concentration of 0.63 µM after 48 and 72 h of incubation (Fig. S1). This leads to the conclusion that an incubation time of 24 h is the most reasonable, since only at this measurement point both substances, which act genotoxic with and without a metabolic activation system, could be detected.

Results –assay optimization –serum and DMSO effects

Supplementation of cell culture media with serum is known to greatly benefit the cell viability (OECD, 2018). However, binding of genotoxic substances to serum proteins might negatively influence the LEC values, as this leads to a reduction of free available compounds (Craig & Kunin, 1976). To analyze the influence of the presence of serum proteins on the toxicological and analytical sensitivity of the HepGentox assay, 0.16 µM of 4NQO and 0.63 µM of BαP were tested in the presence of different serum concentrations, namely 5, 10 and 15% FBS. Preliminary experiments (data not shown) found these concentrations to be suitable, since lower amounts of FBS led to a decrease in viability or a reduction in proliferation in the control culture. Therefore 5% FBS was used as a minimum level. As shown in Figs. 2A and 2B between the various FBS concentrations, no apparent differences could be found for the LEC values with 4NQO. However, the induction of 5% FBS was elevated by a factor of 2.5 to 3.5 for BαP compared to the other concentrations and was therefore chosen as optimum.

When analyzing the genotoxicity of complex mixtures, the application of a maximum amount of sample is of interest to increase the substance concentration in the assay. Unfortunately, most samples of complex mixtures are not aqueous, but solved in organic solvents not tolerated well by mammalian cell culture cells such as DMSO. For mammalian cells, the DMSO compatibility usually ranges around 0.5 to 2%, greatly limiting the sample application (Timm et al., 2013). To determine the DMSO tolerance in the HepGentox assay the cells were treated either with 0.16 µM 4NQO or 0.31 µM BαP dissolved in 0.25, 0.50, 0.75, 1.00, 1.50 or 2.00% DMSO. Figs. 2C and 2D show that upon increasing concentration of DMSO with 4NQO a quenching of the signal was observed by 50% from the highest induction at 0.25% DMSO to the lowest signal at 2% DMSO, therefore possibly leading to higher LEC values. The same was observed with BαP, where the signal was reduced by 75% from its highest peak at 0.25% DMSO to its lowest at 2% DMSO. Contrary, the viability was not reduced at any tested concentration. At a DMSO concentration of 0.25% the highest induction levels could be observed. Nevertheless in regards of the research question, this concentration is not ideal for sample testing. Due to the fact, that this leads to a higher sample dilution and therefore indirectly increasing the LEC values when a sample is added. In terms of correlating sample input, viability and quenching effect, 1% DMSO was chosen as assay condition. This is a holistic approach so that the results of the determined LEC values can be directly compared to the sample testing.

Results –assay optimization –external metabolizing system

Many genotoxic substances need metabolic activation, which is normally achieved via the application of S9 rat liver extract in in vitro assays. The use of S9 does not only raise ethical questions, but is also expensive and due to cytotoxicity and variation of substrate quality its use is discussed (Jacobs et al., 2013). Further, more sample volume and laboratory time is necessary, as testing has to be done with and without the addition of S9, since it possesses both activating and detoxifying abilities, which could lead to false negative results. In this study, two different S9 protocols (incubation for 3 h with 330 µg/mL and 24 h with 10 µg/mL S9) as proposed by Mollergues et al. (2016) were tested, as well as the ability of the HepGentox cell line to metabolize the substances without S9 addition. Results were evaluated for LEC values, as well as for viability (Table 1 and and Figs. S2 and S3). The results showed that HepGentox cells tolerate both S9 treatments well, as the viability was hardly compromised (Fig. S3).

Concerning the LEC values, the 3 h protocol was more promising than the 24 h protocol without S9, since the LEC values were improved for aflatoxin B1 by a factor of two. For cyclophosphamide, (negative after 24 h to 625 µM with the 3 h protocol) the viability was hardly affected. However, for other substances there were no improvements or positive signals. It can be seen that substances needing a metabolizing system, show a response within the same order of magnitude (e.g., aflatoxin B1 with a LEC of 0.63 µM without S9 and 0.31 µM after 3 h with S9, ENU with a LEC of 625 µM for both with/without S9) or better (e.g., BαP with a LEC of 0.63 µM without S9 and 1.25 µM after 3 h with S9) LEC value. Further, the metabolizing activity does not compromise its ability to detect substances that might be negative with S9, such as cisplatin. However, it has to be noted that the substance cyclophosphamide would not have been detected without the addition of S9. Since the assay was developed to detect possible genotoxic substances at low concentration, it was considered as negligible that cyclophosphamide could not be detected without S9, as the LEC value was very high with 625 µM and close to the testing threshold of 1 mM.

Results –pure substances testing

For pure substances testing, a pool of known genotoxic and non-genotoxic substances was chosen from the updated ECVAM list (Kirkland et al., 2016) and some genotoxins of interest were added as well (e.g., 4NQO, actinomycin C). Overall, 16 known genotoxins, 11 known non-genotoxins and 7 non-genotoxins that tend to give positive results in in vitro tests (false positives) were tested. Substances were analyzed up to a top concentration of 1 mM or until the viability dropped below 70%. The maximum concentration of 1 mM was used to prevent the rise of false positive substances, as was proposed by Kirkland et al. (2007). Upon precipitation, insolubility of the stock or cytotoxic effects, a lower concentration was chosen. A threshold of 1.7 fold induction compared to the blank was used, which was calculated from a broad series of negative controls adding three times the standard deviation. For negative substances, a positive control of 2 µM 4NQO and a vehicle control of 1% DMSO was used. The maximum fold induction over the concentration range is given in Tables 2 and 3 as maximum IF. This is the ratio of the mean Nluc response compared to the background signal. The assay proved to have sufficient maximum inductions compared to the background, proving that a genotoxic response leads to a consistent increase in signal intensity in the HepGentox making the assay robust in its response. Overall, a toxicological sensitivity of 87.5% (14 out of 16) and a specificity of 94% (17 out of 18) was achieved as can be seen in Tables 2 and 3. This is within the range of current reporter gene assays dealing with genotoxicity, such as the BlueScreen™ HC with 80% sensitivity and 100% specificity (Hughes et al., 2012) and the p53 CALUX^® with 82% and 90% (Van der Linden et al., 2014).

In a next step, the newly developed HepGentox assay was compared to the LEC values found for other commonly used mammalian assays, for genotoxicity testing, as can be seen in Table 4. The micronucleus has been recommended as part of a test battery for genotoxicity testing in regulatory guidelines (EFSA, 2011; ICH, 2012) and has been approved and standardized by the OECD (OECD, 2014b; OECD, 2014a). The comet test is included here as well, which is also an in vitro assay used for the detection of DNA breaks and damages, especially for clastogenic substances (Pfuhler & Wolf, 1996). For the comet assay, an OECD guideline exists only for the in vivo method (OECD, 2014b), but it can also be used for in vitro testing for genotoxicity. Since for the micronucleus and the comet different cell lines can be used not all substance LEC values could be found for HepG2 cells. Therefore, the used cell line for the LEC result is given in Table 4. Finally, the Ames test is also shown in Table 4, which is an assay used for the detection of direct DNA-reactive substances and especially for mutagens. The Ames test is widely applied and recommended by regulatory guidelines and standardized by the OECD (OECD, 1997). The results obtained for the HepGentox were based on the results in Table 2 and the LEC values for the micronucleus, the comet assay and the Ames test were taken from a literature survey. Comparing the results to several assays is challenging, as there is limited data for several substances and assays. We decided to compare the assays in groups and only for the data where a literature result was available for the assay group. Out of the 15 substances in Table 4, the HepGentox proved to have lower LEC values for 26% (4 out of 15) when looking at the micronucleus and the comet assay. Specifically, for cisplatin the HepGentox was 500 times more sensitive than the comet or the micronucleus tests. For 20% of the substances, higher LEC values were observed with the HepGentox by a factor of two to ten and for 54% the assay was within the range of the others. When comparing the HepGentox to the Ames test in Table 4 it can be seen that the mammalian assay only led to lower LEC values for the substances 7,12-DMBA and etoposide. For the other substances, the Ames test had superior LEC values, which was already observed in a literature survey by Pinter et al. (2020). When looking at the reporter gene assays in Table S1, the BlueScreen™ HC and the p53 CALUX^®, we found that the HepGentox had lower LEC values for 38% (5 out of 13) of the substances. For other substances, it performed in an equal concentration range detecting 31% (4 out of 13) with a similar LEC when compared to both assays, but 31% had a higher LEC than the BlueScreen^TM HC or the p53 CALUX^®. To sum up it can be seen that by comparing the HepGentox to the other genotoxicity assays, it can be found that all of these assays have their advantages and disadvantages when it comes to the analytical sensitivity of the assay, namely the LEC value. However, the HepGentox is the only assay, which has been specifically designed and evaluated for the application of complex mixtures. This makes it an interesting assay, compared to previous test systems exclusively designed for pure substances testing and could be incorporated into a comprehensive test battery together with chemical analysis and other in vitro bioassays, such as the Ames test.

Results –assay application –complex mixtures and cytotoxicity

The presence of a complex mixture matrix was an important aspect during development and validation of the assay, since the test application should include the analysis of complex mixtures. For these mixtures, so called matrix effects are crucial as they can strongly affect the outcome of a result and its reliability (Schilter et al., 2019). To determine the ability of the assay to detect genotoxic substances in the presence of a complex matrix, spiking experiments were conducted. For this, FCM polyethylene extracts solved in DMSO were used, to simulate the presence of a complex matrix and were spiked with 4NQO and BαP in different concentrations. Further, the viability was regarded more closely, since a cytotoxic effect of substances present in the matrix or mixture might mask a genotoxic effect. The results in Fig. 3A and 3B show that the presence of a complex mixture matrix did have an effect, as the induction for each sample differed slightly. However, the LEC was not affected for BαP and slight alterations were found for 4NQO, as the LEC varied by a factor of two, which is considered to be within the range of biological variation within the assay. Further. when compared to the signals observed for the pure substance without samples, no remarkable deviation could be observed. Moreover, the matrix did not interfere negatively with the cell’s viability. This leads to the conclusion that the presence of a complex mixture matrix is not likely to have any adverse effects regarding the detection of genotoxic substances.