Hesperadin

Experiments in the EpiDerm 3D Skin In Vitro Model and Minipigs In Vivo Indicate Comparatively Lower In Vivo Skin Sensitivity of Topically Applied Aneugenic Compounds

Maik Schuler, Lindsay Tomlinson, Michael Homiski, Jennifer Cheung, Yutian Zhan, Stephanie Coffing, Maria Engel, Elizabeth Rubitski, Gary Seitis, Katherine Hales, Andrew Robertson, Saurabh Vispute, Jon Cook, Zaher Radi, and Brett Hollingshead
Pfizer Worldwide Research and Development, Groton, Connecticut 06340 and Cambridge, Massachusetts 02139, USA

ABSTRACT
Risk management of in vitro aneugens for topically applied compounds is not clearly defined because there is no validated methodology to accurately measure compound concentration in proliferating stratum basale keratinocytes of the skin. Here, we experimentally tested several known aneugens in the EpiDerm reconstructed human skin in vitro micronucleus assay and compared the results to flow cytometric mechanistic biomarkers (phospho-H3; MPM2, DNA content). We then evaluated similar biomarkers (Ki-67, nuclear area) using immunohistochemistry in skin sections of minipigs following topical exposure the potent aneugens, colchicine, and hesperadin. Data from the EpiDerm model showed positive micronucleus responses for all aneugens tested following topical or direct media dosing with similar sensitivity when adjusted for applied dose. Quantitative benchmark dose-response analysis exhibited increases in the mitotic index biomarkers phospho-H3 and MPM2 for tubulin binders and polyploidy for aurora kinase inhibitors are at least as sensitive as the micronucleus endpoint. By comparison, the aneugens tested did not induce histopathological changes, increases in Ki-67 immunolabeling or nuclear area in skin sections from the in vivo minipig study at doses in significant excess of those eliciting a responsein vitro. Results indicate the EpiDerm in vitro micronucleus assay is suitable for the hazard identification of aneugens. The lack of response in the minipig studies indicates that the barrier function of the minipig skin, which is comparable to human skin, protects from the effects of aneugens in vivo. These results provide a basis for conducting additional studies in the future to further refine this understanding.

INTRODUCTION
Genotoxicity testing of pharmaceuticals requires the assess- ment of aneuploidy as part of the test battery to determine a compound’s genotoxic potential (ICH Guideline, 2011).
Aneugens are typically detected in the in vitro micronucleus as- say or the in vitro test for structural chromosomal aberrations and by in vivo assessment of erythrocyte micronucleusfrequency in the blood or bone marrow. It is not uncommon to find in vitro aneugens in drug development programs, especially for selective kinase inhibitors that become non-selective when tested at higher concentrations than intended for clinical effi- cacy (Olaharski et al., 2009). Aneugens are expected to exert their aneugenic properties with a threshold mechanism both in vitro (Aardema et al., 1998; Bentley et al., 2000; Elhajouji et al., 1995; 1997) and in vivo (Asano et al., 2006; Cammerer et al., 2010; Marchetti et al., 1996). Typically, compounds with high systemic exposures are risk managed by calculating safety margins be- tween the concentration that causes aneuploidy in vivo and the clinical exposure. However, this approach is not straightforward for compounds that are intended for topical application, where it is technically challenging to accurately measure compound concentrations in the different layers of the skin (Herkenne et al., 2008), most importantly at the stratum basale where cell di- vision occurs and aneugen risk assessment is most relevant (Iizuka, 1994; Weinstein et al., 1984).
One of the other approaches for skin aneugenicity as part of a risk evaluation could be the measurement of aneuploidy in the form of micronuclei in the keratinocytes of a relevant ani- mal species or humans. These methods exist for rodent skin (Nishikawa et al., 1999, 2002, 2005) but it is not the ideal model because of major differences in both morphology and penetra- tion properties when compared with human skin (Todo, 2017). In addition, rodents have not been fully validated for aneugenic compounds and lack proliferation markers to ensure that proper concentration ranges are tested. No such system exists in more relevant animal species like the minipig, which is the preferred animal species in the safety testing of topical drug candidates because of the anatomic and physiologic similarity of skin to humans (Avon and Wood, 2005; Lavker et al., 1991; Vardaxis et al., 1997). A possible alternative is to use a weight of evidence approach by using the human EpiDerm reconstructed skin micronucleus (RSMN) assay that is currently undergoing validation efforts (Pfuhler et al., 2020).
Recently, adverse outcome pathways (AOPs) for aneugens have become available in the scientific literature (Lynch et al., 2019; Sasaki et al., 2020). Our best understanding to date on themechanisms and consequences of aneuploidy induction came from the most commonly observed class of aneugenic agents, tubulin binders, which are compounds that stabilize or destabi- lize microtubules (Lynch et al., 2019). The mechanism of action for the binding of tubulin leading to aneuploidy induction is presented in Figure 1 using the AOP framework.
Another important class of compound with high potential for aneuploidy induction by off-target effects are kinase inhibi- tors (Olaharski et al., 2009). Kinase targets have been pursued for multiple disease indications and aneuploidy induction in kinase programs that try to target the conserved ATP-binding pocket are very common (Klaeger et al., 2017). Many kinases like CDK1, polo-like kinase 1, aurora, and NEK kinases are essential for the proper timing and fidelity in mitosis (Kops et al., 2005). Some kinases involved in mitosis, especially aurora kinases, are com- monly identified as off-target hits in kinase programs at higher than therapeutic concentrations (Anastassiadis et al., 2011). Aurora kinases are a family of highly conserved serine/threo- nine kinases that are important for the accurate distribution of chromosomal material during mitosis and meiosis (Goldenson and Crispino, 2015). As aurora kinase inhibition produces cellu- lar effects that are usually phenotypically similar to the inhibi- tion of aurora kinase B (Ditchfield et al., 2003; Hauf et al., 2003), we present the mechanism of aneuploidy induction following aurora B kinase inhibition in the AOP diagram of Figure 1.
In order to address testing gaps in characterizing the aneu- genic risk as part of a weight of evidence strategy in the skin fol- lowing topical application, we explored several experimental approaches with known aneugens. First, we established the usefulness of the 72-h EpiDerm RSMN assay for the detection of aneugens with known mechanisms like tubulin binding and au- rora B kinase inhibition. We explored the impact of the barrier function of the stratum corneum in the EpiDerm model by com- paring the responses following topical and via media dosing. Second, using the AOP framework, we evaluated the usefulness of key event biomarkers for known aneugenic mechanisms (phosphorylated histone H3 at serine 10 or pH3, mitotic protein monoclonal 2 or MPM2, DNA content) using flow cytometry and compared them quantitatively to the micronucleus induction using benchmark dose-response modeling. Lastly, we used sim- ilar biomarkers that are suitable for in vivo studies, namely im- munohistochemistry (IHC) for Ki-67 and morphometric measurement of nuclear area in skin sections of minipigs treated topically with high doses of the tubulin binder colchi- cine (COL) and the aurora kinase inhibitor hesperadin (HES) to determine the in vivo translatability of these endpoints.

MATERIALS AND METHODS
EpiDerm In Vitro Studies
Test Chemicals
A list of test compounds used in the EpiDerm and minipig experiments can be found in Table 1. The positive control mito- mycin C (CAS 50-07-7) was purchased from Moltox (Boone, North Carolina). For all test compounds apart from the positive control, a fresh source was utilized to prepare dosing stocks just prior to the initial dosing of each experiment. The stocks werethen stored at —20◦C and utilized for all subsequent administra-tions within a single experiment.

Assay Procedures
For handling of EpiDerm tissues and procedure for the RSMN as- say, we followed the detailed protocol described by Dahl et al. (2011) with a few exceptions as outlined below. A general sche- matic of the treatment and harvesting schedule is presented in Figure 2.
Briefly, all EpiDerm kits (EPI-212-MNA; MatTek Corporation, Ashland, Massachusetts) were inspected for any defects such as air bubbles or blisters upon arrival. Each tissue was then trans- ferred aseptically to a 6-well plate containing 1 ml New Maintenance Medium (NMM, MatTek Corporation, Ashland, Massachusetts) and all plates were incubated at 37◦C in a hu-midified atmosphere of 5% CO2 for approximately 1 h. The tis-sues were subsequently re-fed with 1 ml each of fresh, warmed NMM. For experiments designated for micronucleus induction, the NMM was first supplemented with 3 lg/ml of cytochalasin B (CYB; ENZO Life Sciences, Inc., Farmingdale, New York). For those tissue models designated for the flow cytometric mecha- nistic endpoints, the NMM was not supplemented with CYB. For all tissues dosed topically, 10 ll of the test articles or controls were placed on the surface of the models ensuring that the en- tire surface was covered. For tissues in which the test article was added directly to the medium, 10 ml of a stock concentration test compound was added directly into 1 ml NMM, mixed and added to the tissue culture well. The models were then incu- bated under standard culture conditions. After 24 6 2 h, the testmedium was removed, the models were re-fed and dosed a sec- ond time using the same parameters. For the 72 h micronucleus and mechanistic flow endpoint, 24 6 2 h after the second dosing, the models were re-fed and dosed a third and final time using the same parameters. All tissues were harvested either at 48 h (6 3) or 72 h (6 3) after initial treatment to obtain single-cell sus- pensions from the basal layer according to Curren et al. (2006)with minor modifications. Tissues were submerged in Ca2þ-and Mg2þ-free Dulbecco’s phosphate-buffered saline (CMF- DPBS; MatTek Corporation, Ashland, Massachusetts) for 15 min and then transferred to an ethylenediaminetetraacetic acid so-
lution (EDTA; 1 mg/ml in CMF-DPBS; Thermo Fisher Scientific Corporation, Waltham, Massachusetts) for an additional 15 min at room temperature. Each tissue insert was then removed, blotted and placed in a new well containing 1 ml of pre-warmed (~37◦C) trypsin (0.25%)–EDTA (0.02%) solution (Sigma LifeScience, St. Louis, Missouri). An additional 0.5 ml of the trypsinsolution was added inside each insert and incubated at 37◦C for 15 min. After this time, each tissuewas carefully removed using fine forceps and transferred to a new well containing 1 ml offresh trypsin–EDTA solution. In order to maximize cell yield, the insert and supporting membrane of each model was rinsed me- ticulously using the trypsin–EDTA to collect any remaining basal cells. The tissue model was then gently agitated several times to release additional attached cells and then discarded. The single-cell suspension (~1.5 ml) was then transferred toconical tubes containing 1.0 ml each of warm McCoy’s 5A (GibcoLife Technologies Corporation, Grand Island, New York) with 10% FBS (Gibco Life Technologies Corporation, Grand Island, New York) to inactivate the trypsin.

RSMN Assay
Viable Cell Counts
Viable cell counts were generated by flow cytometry as a quality control measure to ensure an adequate number of cells were collected per tissue (≥50 000 viable cells). This was collected from 100 ml of each single cell suspension and transferred to designated wells of a 96-well plate. Eleven microliters of freshlyprepared staining solution (10 mg/ml propidium iodide in PBS; Invitrogen, Waltham, Massachusetts) containing 6 mm cell sort- ing set up beads for blue lasers (2 drops/5 ml PBS; Invitrogen, Waltham, Massachusetts) was added per well. Samples were mixed gently and incubated at room temperature in the dark for 15 min prior to flow cytometric analysis.
Flow cytometry was performed on a BD FACS Canto II equipped with an HTS plate loader using the FACS Diva software version 8.0.1. A dot plot of forward scatter (FSC; log height)versus side scatter (SSC; log area) area included all events detected by the flow cytometer with a FSC threshold value set at 5000. One region gate was placed around the cell population (all cells) and another region gate around counting bead population (beads). From the parent all cell population, a dot plot of propi- dium iodide (PI) area (log scale) versus FSC height (log scale) was created to differentiate viable (PI negative) cells from dead cells (PI positive). Ten thousand cell events were used as the stopping gate for flow cytometric data acquisition. Percent viability was calculated using the viable cell populations. Bead count was used to confirm cytometer volume and sampling consistency.

Cell fixation.
The remaining single cell suspension volume from each harvested tissue was taken from the collection tube anddispersed evenly within 8 wells (1 column) of a standard 96- well plate. When the transfer was completed, the plates were centrifuged at room temperature for 5 min at 100 × g. The su- pernatant was carefully aspirated using a BioTek plate washerleaving about 35 ml of residual solution covering the cell pel- lets. The plates were shaken gently using a plate shaker to loosen the cell pellets. Then 200 ml of warm (~37◦C) KCl solu- tion was added to each well using a multichannel micropipet- tor. After approximately 3 min, 30 ml of freshly prepared ice cold (4◦C) methanol/acetic acid (3:1) fixative was added toeach well and mixed up and down with the micropipettor.
The plates were then centrifuged at 100 × g for 5 min; aspi- rated; shaken and 250 ml of fresh fixative was added to each well. The plates were sealed with adhesive sealing film andplaced in a refrigerator at least overnight or over the weekend.
Slide preparation and acridine-orange staining.
On the day of slide preparation, the plates were centrifuged 100 × g for 5 min, aspi- rated and shaken gently to resuspend the cell pellet. The cellsfrom 4 wells were then combined into a single well and brought up to a volume of ~ 250 ml/well with fresh fixative. The plates were centrifuged; the supernatant aspirated; the cell pelletsloosened and 250 ml/well of fresh fixative/well was added. The plates were then centrifuged, aspirated, and shaken one last time, leaving ~35 ml/well of cell suspension. Typically, two slides per tissue were prepared from each EpiDerm model using a con-trolled environment metaphase spreading chamber (Monalisa: Oslermedicine System Compatible). The slides were dried out- side of the chamber for at least 1 h, after which they were stained in acridine-orange (40 mg/ml in CMF DPBS; Sigma- Aldrich, St. Louis, Missouri) for 1.5 min, rinsed with CMF-DPBS and submerged to de-stain in fresh buffer solution for at least 3 min and then allowed to dry. Prior to any fluorescence based microscopic analysis, a drop of CMF-DPBS was placed onto the slide and a coverslip was added.
Cytotoxicity determination.
For evaluation of cytotoxicity, at least 500 cells were scored per EpiDerm tissue to determine CBPI (cy- tokinesis block proliferation index), as measured by the percent- age of mononucleated, binucleated, and multinucleated cells. Relative CBPI was the basis for and the primary measure of test article-induced cytotoxicity utilized in this study. In order to get a better sense of the dose-response for aneugenic agents in the test system, we expanded the cytotoxicity limit of the assay from 45 6 5% of the vehicle control CBPI (OECD, 2016) to 60 6 5%. Percent binucleation of the cells was also calculated to ensurethat the individual untreated/solvent control tissues had ≥ 25% binucleation and that any treated tissue had ≥40% relative %binucleation. Solvent control tissues with less than 25% binu- cleation and test article-treated models with less than 40% rela- tive % binucleation were not further analyzed for micronucleus induction.

Micronucleus Evaluation
Using standard scoring criteria (Dahl et al., 2011), typically 2000 binucleated cells from 2 replicate slides were scored per tissue to determine the frequency of micronucleated cells (MCN) within the binucleated cell population. When possible, replicate slides were scored by two independent microscopists and the results were then pooled.

Mechanistic Biomarkers by Flow Cytometry
Fixation and Staining
Following the 48 and 72 h incubation with the test compound, each EpiDerm tissue was processed as described in the micro- nucleus section. Each cell pellet was resuspended in 200 ml of PBS and transferred to a 96-well plate. Each 96-well plate con- taining sample was centrifuged at 300 × g for 5 to 6 min and the
PBS aspirated off using a BioTek plate washer. After repeatingthe PBS wash, 150 ml of 0.5% paraformaldehyde (w/v) was added to the re-suspended cell pellet and immediately mixed. After a 15-min incubation period at room temperature, cells were pel- leted again at 300 × g for 5 to 6 min, the supernatant removedusing a BioTek plate washer and the cell pellet vortexed. Cellwere then fixed in 150 ml of ice-cold methanol and stored at—20◦C until labeled for flow cytometry.
For immunofluorescence staining, samples were removed from the freezer and centrifuged at 300 × g for 5 to 6 min. Methanol was aspirated off using a BioTek plate washer and samples were re-suspended in ~100 ml of 0.1% Tween/PBS solu- tion (Sigma) for 15 min at room temperature. Samples were cen-trifuged again; the supernatant was aspirated using a BioTek plate washer, and the cells were stained with 25 ml of antibody cocktail containing FITC conjugated p53 antibody (Product no. 645803; clone DO-7; 1:50; BioLegend, San Diego, California), PE- conjugated Rabbit Anti-H3 (pS10) antibody (Product no. 650807; clone D2C8 XP; 1:50, BioLegend; San Diego, California) or PE- conjugated Rabbit Anti-H3 (pS10) antibody (Product no. 5764; clone D2C8 XP; 1:50, Cell Signaling; Danvers, Massachusetts), Cy-5 conjugated Mouse Anti-phospho-Sr/Thr-Pro MPM2 anti- body (catalog no. 16-220; 1:50; EMD Millipore; St. Louis, Missouri) and V421-conjugated Mouse anti-cleaved PARP (Asp214) anti- body (clone F21-852; 1:50; BD Pharmingen; San Diego, California) for at least 90 min at room temperature. After the incubation,150 ml of PBS was added, 96-well plates were centrifuged 300 × gfor 5–6 min, the supernatant was removed using a BioTek plate washer, and the DNA was counterstained with 100 ml of a propi- dium iodide (PI)/RNase A stain (25 mg/ml PI diluted and 0.2 mg/ ml RNase A diluted into PBS) for 15 min at room temperature and then stored at 4◦C for 20–24 h prior to analysis by flowcytometry.

Flow Cytometric Analysis
Flow cytometry was performed on a BD FACS Canto II equipped with a 96-well plate loader and three lasers as excitation sour- ces using the FACS Diva software version 8.0.1. The MPM2 fluo- rescent signal (AF-647) was excited by the red 633 nm laser and collected in the far red fluorescence channel (660/20 nm), the p53 (AF-488) signal was excited by the blue 488 nm laser and the fluorescent signal was collected in the green fluorescence channel (530/30 nm), the pH3 fluorescence signal was excited by the blue 488 nm laser and collected in the red fluorescence channel (585/42), the cleaved PARP signal was excited by the 405 nm laser and collected in the blue fluorescence channel (450/50 nm) and PI was excited with a 488-nm laser and col- lected in the far-red fluorescence channel (670LP). Acquisition criteria were set to obtain 100 000 cycling cell events, if possible, or analyze 75 ml for each control and all treated samples on low speed (0.5–1.0 ml/s).
A dot plot of forward scatter (FSC area, linear scale) versusside scatter (SSC area, linear scale) included all events detected by the flow cytometer with a FSC threshold value set at 5000. To establish the parent cycling cell population, a dot plot of PI area (linear scale) versus PI width (linear scale) was created and a gating region was drawn around the cycling cell population to exclude debris, doublets, and/or aggregates from the analysis. A dot plot of PI area (linear scale) versus PE area (log scale; pH3) was created using the cycling cell population (Figure 8). A gating region was drawn around the pH3 population to differentiatethe cells in metaphase from the cycling cells in interphase. Increases ≥2-fold of the histone 3 positive cell population were considered indicative of increased mitotic activity. Similarly, adot plot of PI area (linear scale) versus AF488 area (log scale; MPM2) was created using the cycling cell population (Figure 8). A gating region was drawn around the MPM2 positive popula- tion to differentiate the cells in metaphase from the cycling cells in interphase. Increases ≥2-fold of the MPM2 positive cellpopulation were considered to be indicative of increased mitoticactivity.
Cleaved PARP and p53 were measured in all samples but data are not shown or discussed because they did not prove to be useful biomarkers. A dot plot of PI area (linear scale) versus blue violet (BV) 421 area (log scale; cleaved PARP) was created using the cycling cell population. Violet 450/50 voltage was ad- justed to place the negative population approximately in the second decade (102) on the cleaved PARP axis. A region gate was placed around the PARP positive cell population which spans from the above the S population (channel 50 K) to the far end of the DNA area axis. A dot plot of PI Area (linear scale) versus AF 488 Area (log scale; cleaved PARP) was created using the cycling cell population. Blue-530/30 voltage was adjusted to place the negative population approximately in the second decade (102) on the p53 axis. A region gate was placed around the entire cell population to measure the median channel fluorescence of the p53 signal. Negative and positive staining controls were utilized to ensure gate placement.
Minipig study with aneugens colchicine and hesperadin. A dermal tol- erability study was conducted with 12 to 16-week-old Go¨ ttingen minipigs at Charles River Laboratory, Spencerville, Ohio. Dosing formulations of COL (0.02–5 mg/ml) and HES (0.06 to 15 mg/ml) were prepared as solutions in a commonly used vehicle con- taining 15% w/w ethanol, 30% w/w polyethylene glycol 400, 45% w/w diethylene glycol monoethyl ether (transcutol), and 10% glycerin. Vehicle, COL, or HES formulations were administered to male minipigs twice daily, approximately 6 h apart, for three consecutive days at a dose volume of 0.01 ml/cm2. The applica- tion site for each test condition encompassed a skin area of 4cm2 (ie, 2 cm × 2 cm) in the dorsal region of the minipig, and 1male minipig served as a control, with 2 designated application sites; 1 untreated site and 1 vehicle site. COL or HES administra- tion groups included 3 separate males per test compound with each animal having a total of 6 application sites (1 vehicle site and 5 test article sites). Doses of 0, 0.0004, 0.0016, 0.0062, 0.025,and 0.10 mg/cm2/day or 0, 0.0012, 0.0046, 0.0188, 0.075, and0.3 mg/cm2/day were administered for COL or HES, respectively. There was a minimum of 8 cm between each application site to minimize the risk of cross-contamination. Animals were evalu- ated daily for clinical observations and dermal scoring. On day 4 of the study, representative sections of skin were collected from each application site post-mortem and fixed in 10% neutral buffered formalin.
The care and use of animals in this study was conducted in adherence to the guidelines of the U.S. National Research Council and approved by Testing Facility Institutional Animal Care and Use Committee (IACUC).
Staining and immunohistochemistry of skin sections. Formalin fixed skin sections from the in vivo minipig study were processed by routine histologic methods, embedded in paraffin, and sec- tioned at 5 mm. Sections were stained with hematoxylin and eo- sin (H&E) and examined by light microscopy. Light microscopic evaluation was performed by a board-certified veterinary pa- thologist for epidermal changes including evidence of cell pro- liferation and/or death/necrosis. Epidermal death/necrosis severity scores were semi-quantitatively graded as follows:0 ¼ absent, 1 ¼ minimal (rare dead cells in the epidermis); 2 ¼ mild (easily identified dead cells with occasional cell clus- ters); 3 ¼ moderate (prominent cell death with clusters of cells affected); 4 ¼ marked (confluent areas of cell death); 5 ¼ severe (>80% of tissue with confluent cell death).
For Ki-67 IHC, formalin-fixed, paraffin-embedded blocks of minipig skin samples were sectioned at 5 mm, mounted onpositively charged slides, and air dried prior to immunolabeling. The IHC staining was performed on a Leica Bond RX automated instrument (Leica Biosystems, Buffalo Grove, Illinois). Briefly, sections were pre-treated with Leica heat-induced Epitope Retrieval 2 (EDTA buffer) for 20 min, followed by both a 10-min endogenous peroxidase block and a 10-min protein block (Agilent, Santa Clara, California). A rabbit monoclonal antibody to Ki-67 (Product no. 275 R-16 clone SP6, Cell Marque, Rocklin, California) was applied at 1:100 for 40 min at room temperature, followed by detection and visualization with Leica Refine DAB (diaminobenzene) Polymer. A rabbit isotype control (Vector Laboratories, Burlingame, California) was run at the same con- centration on select slides and all IHC slides were counter- stained with hematoxylin and mounted with a permanent mounting medium.

Image Acquisition and Analysis
For morphometric analysis of Ki-67 IHC, sections of EpiDerm specimens and minipig skin on which IHC had been performed were scanned without manipulation on an Aperio AT2 digital brightfield whole slide scanner (Leica Biosystems), and image analysis algorithms were developed using the Visiopharm im- age analysis platform (Visiopharm, Denmark). Thresholds for signal intensity and size gating were established on control specimens and Ki-67 expression was quantified on images by measuring the numbers of DAB-positive and total (hematoxy- lin-stained) nuclei in each specimen and calculating the per- centage of total cells that were Ki-67 positive per section ofepidermis, reported as %Ki-67þ cells/epidermis.
Visiopharm newCAST 2D stereology software (Visiopharm) was used to determine the nuclear area per cell. After the region of interest was selected, random meander sampling was used to select random, unbiased fields for analysis. A point probe was applied to each field to randomly identify the nuclei to be measured, followed by a nucleator probe to measure the cross- sectional area of each selected nucleus.
Statistical methods and benchmark dose (BMD) analysis. Data for the creation of quality control charts were analyzed using the sta- tistical procedures available in Minitab (Minitab 18.1 Statistical Software. Minitab, Inc., State College, PA) (Lovell et al., 2018). Quality control charts are widely used to monitor the variability of samples and to show that their processes are stable and do not drift over time. Micronucleus frequencies from individual tissues of vehicle and positive controls (mitomycin C; 5 mg/ml) from the counting of 2000 binucleated cells were analyzed as % MNC and the CBPI was examined based on the calculated values from the counting of 500 cells. I-charts for individual tissue val-ues of % MNC and CBPI for control data, collected over a period of two years ðX¯ ) were created with decision lines for upper (UCL¼ X¯ þ 3 SD) and lower control limits (LCL ¼ —3 SD). In addition, a2-sample t test was used to compare the impact of the treat- ment condition (media vs topical) on the mean micronucleus frequencies of both the vehicle control and positive mitomycin C micronucleus frequencies. Lastly, the same analysis was used to compare the mean micronucleus frequencies for the vehicle control with the mean micronucleus frequencies for the mito- mycin C controls for each of the two treatment conditions.
Micronucleus data from the EpiDerm RSMN assay were ana- lyzed statistically using the SAS 9.4 program (SAS Institute Inc. Cary, North Carolina). The Cochran-Armitage binomial propor- tions trend test was performed to test for a trend in the increase of MNC across the dose groups. A positive trend was followed with a one-tailed Fisher’s exact test on the total number of MNCfrom each test article group compared with the total number of MNC from the negative controls. p-values of ≤.05 were consid- ered statistically significant. In addition, in order to be consid- ered a positive test, the frequency of MNC had to be greater than the upper reference interval of the current negative histor- ical control.
Benchmark dose (BMD) analysis was performed using the European Food Safety Authority (EFSA) web tool available at https://shiny-efsa.openanalytics.eu/app/bmd (accessed on 8/6/ 2020) and uses the R-package PROAST, version 69.0, for the un- derlying calculations. The BMD approach was used to derive points of departure (POD) for the mechanistic biomarkers %pH3 positive cells, %MPM2 positive cells, % 4N cells, and % 8N cells at 48 and 72 h. These PODs were then compared with the POD for micronucleus induction at 72 h. In addition, micronucleus frequencies following topical and media dosing were compared. Although a 5%–10% change in the mean response compared with the negative control is often used as a benchmark re- sponse (BMR), BMR values for genotoxicity endpoints have not been established (White et al., 2020) and EFSA notes that default BMRs might be modified based on statistical and biological con- siderations (EFSA Scientic Committee et al., 2017). Some recent publications indicate that the BMRs used for other toxicological endpoints are too low for genotoxicity endpoints and recom- mend to use values that are significantly higher (Slob, 2017; Zeller et al., 2017). A BMR of 50% change in mean response com- pared with the controls was selected based on the publication by Wheeldon et al. (2020) and Dertinger et al. (2019). The higher BMR value of 50% was chosen to reduce the confidence intervals for the BMD analysis (EFSA Scientic Committee et al., 2017). The higher BMR was expected to have little impact on the BMD val- ues when rank ordering compounds or comparing biomarker endpoints (Wills et al., 2016). This is deemed appropriate as this study does aim to compare in vitro endpoints rather than calcu- late a PoD value for determining a human equivalent dose for risk assessment.
A default set of models were fitted. The Akaike information criterion (AIC) was calculated for each model to estimate the quality of each model relative to each other. The models with the lowest AIC (AIC ≤ AICmin þ 2) were selected for model aver-aging and were weighted according to their AIC values.
Confidence intervals for the BMD were based on 200 bootstrap data sets. A 90% confidence interval around the BMD was esti- mated, the lower bound of the prediction is reported as BMDL and the upper bound as BMDU.

RESULTS
Reproducibility and Historical Control Values
Following the recommendation of the in vitro micronucleus technical guideline TG487 (OECD, 2016), historical control ranges and distributions for both the vehicle control acetone and positive control mitomycin C (5 mg/ml) using the direct me- dia and topical treatment methods were established. Control Charts (ie, I charts) of individual tissues were created for the percent micronucleated cells (% MNC) and percent binucleated cells (% BNC) for both the vehicle and positive controls sepa- rated by media and topical treatments (Figure 3). The resulting control charts indicate that both the % MNC and % BNC for the negative controls were generally within the upper and lower control ranges except for a single data point for micronucleus induction for the topical treatment. The average micronucleus frequency for the vehicle controls in our laboratory was0.32 6 0.11% MNC with 95% confidence limits of 0.08 to 0.55% for direct in medium and 0.33 6 0.12% MNC with 95% confidence limits of 0.11 to 0.56% for the topical treatments. The average bi- nucleate frequency was 53.2 6 6.5% BNC with a 95% confidence interval of 43.9 to 62.5% for the direct in media and 52.3 6 6.3% BNC with a 95% confidence interval of 41.8%–62.7% for the topi- cal treatment. In contrast, the variability of % MNC and % BNC for the positive control appeared to be slightly higher with one individual tissue from both the in medium and topical treat- ments being higher than the upper control limit. The average micronucleus frequency was 2.4 6 0.98% MNC with 95% confi- dence limits of 1.3% and 3.6% for medium exposed and1.9 6 0.72% MNC with 95% confidence limits of 0.8% and 3.0% forthe topical exposures. The frequency of binucleates was 37 6 8.3% BNC with a 95% confidence interval of 17.7%–48.9% forthe medium exposed and 37% 6 8.3% BNC with a 95% confi- dence interval of 25.9%–52.8% for the topical exposure.
The route of exposure (topical vs media) did not result in significant differences for the mean % MNC for the vehicle(p ¼ 0.646; 2-sample t test) or the positive control mitomycin C (p ¼ .192; 2-sample t test). However, there was a statisticallysignificant difference (vehicle control vs mitomycin C positive control) for the mean % MNC for both the in media (p < .01; 2-sample t test) and topical (p < .01; 2-sample I test) treatments. Micronucleus Induction With Tubulin Binders The sensitivity of the EpiDerm RSMN assay was evaluated for the known tubulin de-polymerizing compounds COL and noco- dazole (NOC). The impact of the barrier function of the stratum corneum of the EpiDerm system was investigated by applying the test compound in acetone topically or into the maintenance media. The results from this comparison are shown in Figures 4and 5. Both tubulin binders produced dose-related statistically significant trends (p ≤ .05; Cochran Armitage trend test) for the increase of % MNC under all test conditions that were outside ofthe historical negative control confidence interval.COL, following direct application to the media, did not in- duce micronuclei up to 0.3 mg/ml but concentration between 0.4 and 1.2 mg/ml produced statistically significant increases in % MNC (p ≤ 0.05; 1-tailed Fisher’s exact test; Figs. 4 and 5). Similarly, COL applied topically did not induce micronuclei upto 0.25 mg/ml but produced statistically significant increases in micronuclei between 0.47 and 1.7 mg/ml (p ≤ .05; 1-tailed Fisher’sexact test). Higher concentrations were not evaluated for either dosing method because of excessive cytotoxicity. None of the treatments produced dose-related changes in viability. POD de- termination for micronucleus induction using BMD modeling (Figure 5) with COL determined BMDL0.5 to BMDU0.5 ranges of 0.26–0.33 mg/ml following media dosing while the range was 0.16–0.35 for topical dosing indicated that no significant differ- ence between the two dosing methods exist, when adjusted for total dose. Similar results were seen for the less potent aneugen NOC (Figs. 4C and 4D). No micronucleus induction was seen follow- ing direct application to the media up to 1.5 mg/ml. Concentrations between 2.2 and 5.2 mg/ml produced statisticallysignificant increases in micronuclei (p ≤ .05; 1-tailed Fisher’s ex-act test). Topically applied NOC did not induce micronuclei up to 0.9 mg/ml but produced statistically significant increases in micronuclei between 1.5 and 5.2 mg/ml (p ≤ .05; 1-tailed Fisher’sexact test) with 8 mg/ml being too cytotoxic for evaluation. TheNOC treatments did not produce dose-related changes in viabil- ity when compared with the negative controls. The PODdetermination for micronucleus induction using BMD response modeling (Figure 5) produced BMDL0.5 to BMDU0.5 ranges of 0.92–2.1 mg/ml for media dosing while the range was 0.74–1.1 for topical dosing. These results indicate that both dosing methods produced similar POD metrics. Micronucleus induction with aurora kinase inhibitors. A similar ap- proach was taken for the known aurora kinase inhibitors AMG900, barasertib (BAR) and HES, which are usually character- ized by steep dose-response relationships for micronucleus in- duction in 2D test systems. Similar to the studies with tubulin binders, the test compounds were applied topically in acetone or directly into the maintenance media (Figs. 6 and 7). As expected, all three aurora kinase inhibitors produced dose-related statistically significant trends (p ≤ .05; Cochran Armitagetrend test) for the increase of % MNC under all test conditions that were outside of the historical negative control confidence interval. AMG900 was the most potent inducer of the three aurora ki- nase inhibitors in the EpiDerm RSMN assay (Figs. 6A and 6B). No micronucleus induction was seen following direct application to the media up to 0.3 mg/ml. Concentrations at 0.6 and 2.5 mg/ml produced statistically significant increases in micronuclei (p ≤ .05; 1-tailed Fisher’s exact test) with 2.5 mg/ml being exces- sively cytotoxic. Likewise, topically applied AMG900 did not in- duce micronuclei up to 0.3 mg/ml but produced statistically significant increases in micronuclei at 0.63 and 2.5 mg/ml (p ≤ .05; 1-tailed Fisher’s exact test) with ≥10 mg/ml again being excessively cytotoxic. The AMG900 treatments did not produce dose-related changes in viability when compared with the neg- ative controls. The POD determination for micronucleus induc- tion using BMD response modeling (Figure 7) produced BMDL0.5 to BMDU0.5 ranges of 0.14–0.56 mg/ml for media dosing while the range was 0.02–0.48 for topical dosing. The results from this BMD analysis suggest that the dosing route has no significant impact on the micronucleus induction with AMG900. BAR was the weakest inducer of the three aurora kinase inhibitors in the EpiDerm RSMN assay (Figs. 6C and 6D). No in- duction of micronuclei was observed after application to the media up to 2.8 mg/ml. Concentrations from 4.0 to 12.2 mg/mlproduced statistically significant increases in micronuclei (p ≤ .05; 1-tailed Fisher’s exact test) with 12.4 mg/ml exceeding the recommended cytotoxicity limits for the assay. Topicallydosed BAR did not induce micronuclei up to 7.3 mg/ml but pro- duced statistically significant increases in micronuclei between11.3 and 41 mg/ml (p ≤ .05; 1-tailed Fisher’s exact test) with con- centrations ≥17.3 mg/ml again being excessively toxic. The BAR treatments did not produce dose-related change in viabilitywhen compared with the negative controls. The POD determi- nation for micronucleus induction using BMD response model- ing (Figure 7) produced BMDL0.5 to BMDU0.5 ranges of 2.46–3.97 mg/ml for direct in media dosing while the range was 2.84–9.42 for topical dosing. The results from this BMD analysis implies that the dosing route has no significant impact on the micronucleus induction with BAR. HES was the last of the three aurora kinase inhibitors in our EpiDerm RSMN studies tested (Figs. 6D and 6E). No induction of micronuclei was observed after both via media and topical ap- plication up to 1.1 mg/ml. Concentrations from 1.7 to 10 mg/ml produced statistically significant increases in micronuclei (p ≤ .05; 1-tailed Fisher’s exact test) with 10 mg/ml exceeding therecommended cytotoxicity limits for the assay under both treat-ment conditions. The BAR treatments did not produce dose- related change in viability when compared with the negative controls. The POD determination for micronucleus induction using BMD response modeling (Figure 7) produced BMDL0.5 toBMDU0.5 ranges of 1.06–1.54 mg/ml for media dosing while the range was 0.314–1.14 for topical dosing. The results from this BMD analysis implies that the dosing route has no significant impact on the micronucleus induction with BAR when dose is considered. Key event biomarkers evaluated by flow cytometry and quanti- tative relationship to in vitro micronucleus induction for tubulin binders. Several mechanistic key event biomarkers expected to respond following the treatment with aneugenic tubulin bind- ers and aurora kinase inhibitors were evaluated by flow cytome- try (Figs. 9 and 11). Historical vehicle control backgroundfrequencies for mitotic cells and polyploidy were generally low. The 48 h timepoint experiments showed 0.57 6 0.36% pH3 and0.58 6 0.39% MPM2 positive cells, while 0.36 6 0.12% polyploid cells were seen. At the 72 h timepoint, somewhat lower frequen- cies of each % pH3 (0.32 6 0.20%), % MPM2 (0.29 6 0.18%) and polyploid cells (0.26 6 0.16%) were seen. For the positive histori- cal control data for COL (0.89 mg/ml) at the 48 h timepoint higher% pH3 (4.33 6 2.26%; p ≤ .01, 2-sample t test) and MPM2(4.17 6 2.03%; p ≤ .01, 2-sample t test) frequencies when com- pared with the vehicle control occurred, while the frequency of polyploid cells (0.52 6 0.23%; p ¼ .015) remained largelyunchanged. Frequencies of % pH3 (1.48 6 0.92%; p ≤ .01, 2-sample t test) and % MPM2 (1.42 6 0.92; p ≤ .01, 2-sample t test), while higher than in the vehicle control, where somewhat lower than at the 48 h timepoint and the % polyploidy cells (0.30 6 0.15; p ¼ .494) was comparable to the negative control. The percentage of cleaved PARP positive cells in the negativecontrol at both the 48 h (0.10 6 0.17%) and 72 h (0.09 6 0.19%) timepoint were low. The percentage of cleaved PARP positive cells were significantly higher at both the 48 (1.25 6 0.54%;p ≤ 0.01, 2-sample t test) and 72 h timepoints (0.52 6 0.33%;p ≤ 0.01, 2-sample t test) in the COL treated samples. The mean channel fluorescence for p53 in the vehicle control was278.1 6 41.5 RLUs at 48 h and 280.9 6 34.8 at the 72 h time point. COL treated EpiDerm tissues showed higher p53 MCF values at both the 48 (394 6 29.5; p ≤ .01, 2-sample t test) and 72 h(409.1 6 75.6 RLUs; p ≤ 0.01, 2-sample t test) timepoints. For boththe negative control and COL, no statistically significant differ- ence was seen between the mitotic index markers % pH3 and MPM2 in a 2-sample t test. For tubulin binders, the expectation was that pH3 and MPM2 would increase because of SAC activated mitotic arrest. Model flow plots from the flow cytometric measurements for key event biomarkers in isolated keratinocytes from the Epiderm model are shown in Figure 8 and results for the two model tubu- lin binders COL and NOC are displayed in Figure 9. For clarity, only the endpoints for mitotic index markers % pH3 and % MPM2 as well as the DNA content markers % 4N (ie, cells in G2M or polyploid G1 cells) and % 8N (ie, polyploid G2M cells) are shown. Overall increases of %PARP and p53 were weak and are not presented. After the completion of a dose-range finder, COL was tested over a dose range of 0.01–1.68 mg/ml for the 48 h treatments and0.01 to 3.2 mg/ml for the 72 h treatments (Figs. 9A and 9B). COL induced dose-related increases in % pH3, % MPM2, and % 4N cells at both treatment times with the 48 h timepoint producing a higher frequency of cells arrested in mitosis than the 72 h timepoint but similar frequencies of % 4N cells observed. Only a very small increase in % 8N were detected at either timepoint. When comparing the POD derived by BMD analysis at a BMR of 50% change in mean response compared with the controls (Figure 10), both mitotic index markers % pH3 and % MPM2 aboth timepoints showed equal sensitivity compared with the % MNC at 72 h. On the other hand, the 90% confidence interval for the POD of % 4N cells was slightly higher than % MNC and be- cause of the weak response, the confidence interval for 8N cells is too wide to provide a useful comparison to % MNC. Following an initial dose-range finder, NOC was tested from1.5 to 8.0 mg/ml in the 48 h treatment and 0.8 to 8 mg/ml in the 72 h treatment (Figs. 9C and 9D). NOC produced dose-related increases of % pH3, % MPM2 and % 4N cells at both timepoints with much higher frequencies seen at the 48 h when compared with the 72 h timepoint. Similar increases in % of 4N cells were seen at both timepoints and only a slight increase in % 8N cells was seen at 48 h but not at 72 h under either treatment condi- tion. Subsequent BMD analysis showed that there was good agreement for the POD for % pH3 and % MPM2 at the 72 h time- point but to a lesser degree at the 48 h timepoint (Figure 10). The POD for % 4N cells was also slightly higher than for micronu- cleus and polyploidy was not useful because of the wide confi- dence intervals for the POD prediction. Overall, it appears that for tubulin binders, the two mitoticndex markers pH3 and MPM2 are more predictive biomarkers than the DNA content markers. Key event biomarkers evaluated by flow cytometry and quanti- tative relationship to in vitro micronucleus induction for aurora kinase inhibitors. For aurora kinase inhibitors, it was antici- pated that they would induce large increases of % 4N and % 8N cells because of their ability to inhibit cytokinesis and decrease in % pH3, which is one the phosphorylation targets for aurora B kinase. Results for the 3 model aurora kinase inhibitors AMG900, BAR, and HES are displayed in Figure 11. For clarity, only the endpoints for mitotic index markers % pH3 and % MPM2 as well as the DNA content markers % 4N and % 8N are shown. Because increases of % PARP were generally weak and p53 was used as a confirmatory cellular stress marker, they are not presented. All 3 aurora kinase inhibitors had consistent bio- marker patterns with dose-related increases in % 4N and % 8N cells as well as decreases in % pH3 and % MPM2 cells. As the fre- quency of mitotic cells is generally low, although qualitatively a decreased signal is apparent, it is difficult to quantitatively characterize a decrease in mitotic cells in the EpiDerm model. AMG900 was tested over a dose range of 0.003–40 mg/ml. Dose-related increases were seen at 3 mg/ml and higher for both% 4N and % 8N cell population at both timepoints, while the same concentrations produced decreases in % pH3 and % MPM2 (Figs. 11A and 11B). When comparing the POD derived by BMD analysis at a BMR of 50% change in mean response compared with the controls (Figure 12), the % 4N cells at both timepoints showed equal sensitivity to the % MNC at 72 h. The PODs for the% 8N cells had overlapping confidence intervals with the % MNC but the confidence interval was too wide to make a conclu- sive decision. The PODs for the 2 mitotic index markers, % pH3 and % MPM2, appeared to be higher and therefore less sensitive than the % MNC. BAR was tested from 1.8 to 26.7 mg/ml and produced a similarpattern of response with higher frequencies in % 4N and % 8N seen at doses ≥7.3 mg/ml with concurrent decreases in the fre- quency of % pH3 and MPM2 (Figs. 11B and 11C). In the case ofBAR, the PODs for the increase in % 4N and % 8N cellpopulations were in line with the POD for % MNC. The % pH3 and % MPM2 had overlapping confidence intervals with the % MNC but the usefulness was limited because of the wide confi- dence intervals. Lastly, HES was evaluated from 0.21 to 10 mg/ml with a simi- lar response pattern seen as with the other 2 aurora kinase inhibitors Figs. 11D and 11E). Doses around 5 mg/ml produced both an increase in % 4N and % 8N cells and a decrease in mi- totic cells as indicated by lower frequencies of % pH3 and % MPM2. The confidence interval for the POD for the % MNC in- duced by HES are somewhat wide but both the % 4N and % 8N PODs appear to be in good agreement while the PODs for the mi- totic index markers % pH3 and % MPM2 are too wide to be of value. Minipig study evaluating key event biomarkers following topical treat- ment with colchicine and hesperadin. In order to characterize the EpiDerm in vitro to in vivo translation capabilities a 3-day inves- tigative study in the minipig, which is used based on its similar- ity in skin morphology compared with human, was designed to evaluate the key event biomarkers described previously. In this study COL or HES were applied topically to male minipigs twicedaily for 3 consecutive days at doses exceeding (up to ~2000×)those resulting in positive micronucleus response in the EpiDerm model. An earlier study had shown that the frequency of mitotic cells in the stratum basale is quite low as shown by the low frequency of pH3 positive cells; therefore, more suitable bio- markers for IHC and image analysis were utilized. The mitotic index biomarkers pH3 and MPM2 were replaced by Ki67, which detects cycling cells and is expressed with increasing levels dur- ing S-phase with a maximum in M-phase. As an increase of both mitotic cells and 4N cells was seen with tubulin binders in the EpiDerm system and Ki67 is expressed in a higher number of cells, Ki67 is expected to be a suitable surrogate key event bio- marker. Additionally, a surrogate endpoint used for DNA content by fluorescence methods used in the EpiDerm system was the measurement of nuclear area by nonfluorescent image analysis techniques. H&E stained skin sections were reviewed by a board- certified veterinary pathologist for histopathological changes. With microscopic evaluation of H&E-stained sections of skin, all minipigs had similar variability in epidermal thickness and there was no morphological evidence of test article-related pro- liferation or cell death/necrosis. There were no COL or HES- related microscopic findings and examples of vehicle control and H&E skin sections treated with the highest doses of COL (5 mg/ml) and HES (15 mg/ml) are shown in the upper panel of Figure 13. Skin sections from untreated, vehicle control, COL (0.02– 5 mg/ml) and HES (0.06–15 mg/ml) from 3 animals each were stained for Ki67 (Figure 13, lower panel) and the percentage of Ki67 positive cells quantified. No dose-related increases or decreases of Ki67 positive cells were seen with any of the treat- ments (Figure 14). Lastly, H&E-stained skin sections from the same animals were evaluated for changes in the nuclear area using stereology. No treatment-related changes in nuclear area were detected with any of the treatments. Based on the parameters evaluated, there were no treatment-related changes in the skin of minipigs treated with high doses of the model aneugens COL and HES that would sug- gest aneuploidy induction in vivo. Measurement of COL and HES in blood from treated minipigs were below the limit of quantifi- cation confirming that the topically administered doses did not result in systemic exposure. DISCUSSION The studies presented here further confirm the suitability of the RSMN assay in the EpiDerm model for the in vitro assessment oftopically applied aneugenic compounds. When adjusted by ap- plied dose, however there is no significant difference whether the dose is applied topically or has direct access to the cells of interest when dosed via the media. Using the principles of AOPs and BMD analysis, the results indicate that mechanistic keyevent biomarkers can be quantitatively linked to micronucleus induction in the RSMN assay using the EpiDerm tissue system. Based on the limited number of compounds tested in this study, a significant limitation of the EpiDerm appears to be the ab- sence of barrier integrity to limit compound absorption similarto what would be expected in vivo based on the similar sensitiv- ity regardless of method of dosing (ie, topical compared with medium-based dosing results in similar results). This is further supported by the study that applied the knowledge gained from the key event biomarkers studies in the EpiDerm model to a topical minipig study that did not show evidence of micronu- cleus activity even with significantly higher applied doses of aneugens. These disparate results highlight the known impor- tance of the barrier function of skin that likely does not allow adequate HES or COL to reach the stratum basale and is possibly influenced by lower duration of exposure due to skin flux and/ or compound metabolism. As was reported by others, our studies show the robustness of the RSMN assay for testing compounds following topical ap- plication (Aardema et al., 2010, 2013; Chapman et al., 2014; Curren et al., 2006; Hu et al., 2009; Mun et al., 2009; Roy et al., 2016; Yuki et al., 2013). The EpiDerm model consistently showed good and consistent proliferation rates as exemplified by the % binu- cleated cells being consistently above 50%. However, the mean background frequency of % MNC of 0.33% in our studies is higher than what is reported in the literature with most other studies reporting the mean vehicle control frequency of 0.05%s– 0.1% MNC (Chapman et al., 2014; Curren et al., 2006; Hu et al., 2009; Roy et al., 2016). A possible explanation for this discrep- ancy is that a modified harvesting procedure employing climate-controlled slide dropping chamber improved the spreading of the binucleated keratinocytes, which could make the detection of micronuclei easier. The higher background fre- quencies of MNC had no impact on the sensitivity for the detec- tion of micronuclei induced by aneugens in this study as all compounds tested produced robust responses in the assay. The EpiDerm RSMN assay was shown to sensitively respondto two classical tubulin binders and three model aurora kinase inhibitors by two different modes of exposure. MNC could be in- duced in a dose-dependent manner either by adding genotoxi- cants directly to the media exposing the dividing keratinocytes from the stratum basale or by application to the apex of stratum corneum. The only other published data on the micronucleus in- duction with a known aneugen in the EpiDerm RSMN assay is vinblastine sulfate (Aardema et al., 2010; Curren et al., 2006; Roy et al., 2016). The dose-response for the aneugens tested is steep and dose selection is crucial to identify appropriate concentra- tion for micronucleus evaluation. Testing of all compounds, regardless of class of compound, no significant difference in the PODs was observed for either the direct in medium or topical dosing method. A similar obser- vation has been made by Chapman et al. (2014) for methyl meth- anesulfonate and the authors speculated that this might be related to the reduced barrier function, which has been de- scribed in greater detail in publications comparing human skin with reconstructed human skin models (Dreher et al., 2002; Schafer-Korting et al., 2008; Schreiber et al., 2005). These studies generally indicate that the rank order of skin penetration for hy-drophilic and amphipathic compounds in human skin < pigskin < rat skin < EpiDerm, whereas usually little difference isseen for lipophilic compounds. Acetone is often used in pene- tration studies for barrier disruption and can significantly alter the absorption of hydrophilic and amphipathic compounds, which is attributed to transepidermal water loss (Benfeldt, 1999; Cross and Roberts, 2000; Tsai et al., 2001). The impact of the sol- vent acetone on the penetration of the test compound can also not be underestimated because medicines designed for thera- peutic treatments in humans by topical use are not typically formulated in organic solvents such as acetone and barrierdisruption of the skin is not desired. If the result of these studies are the consequence of incomplete or disrupted barrier func- tion, the risk of dermal aneuploidy to humans exposed topically to aneugens is probably much lower than the in vitro micronu- cleus induction in the EpiDerm RSMN assay would suggest and caution is recommended when applying these studies to quan- tify risk to humans, as this model is currently most appropriate for hazard identification purposes only given the limited ability to contextualize in vitro potency data to in vivo context. This cau- tionary approach guidance is further supported by the lack ofresponse in the minipig with high doses (up to ~2000 timeshigher compared with EpiDerm concentrations resulting in pos- itive response) of COL and HES and is consistent with the rank order of skin penetration from the literature. One of the goals of these experiments was to evaluate the use of mechanistic key event biomarkers indicative of the mechanism of action of aneugens that could be used in the skin and serve as surrogate biomarkers for the induction of aneu- ploidy for topically applied compounds. Ultimately, biomarkers that could be applied to dermal toxicity studies in minipigs and by using IHC techniques on standard skin sections would be highly desirable. In order to quickly process a large number of samples and based on extensive experience with flow cytomet- ric assays, biomarkers that have been previously shown to be predictive of aneugenic mechanisms in 2D cell culture systems were applied to the EpiDerm model (Bernacki et al., 2019a,b; Bryce et al., 2017, 2018; Dertinger et al., 2019). In a proof-of- concept study in human lymphoblastoid TK6 cells with 27 refer- ence aneugens, which were mostly tubulin binders and kinase inhibitors that have the potential for aurora B kinase inhibition, Bernacki et al. (2019a) showed that mechanistic biomarkers can sensitively predict these classes of aneugens. The approach taken in this study is further supported by the recent efforts to generate AOPs for different genotoxic molecular initiating events (MIEs), which includes AOPs for tubulin binding and au- rora kinase inhibition leading to aneuploidy induction (Lynch et al., 2019; Sasaki et al., 2020). The experiments presented in this study are the first time that this approach using the AOPs and flow cytometric measurements of key event biomarkers have been applied successfully to a reconstructed 3D tissue sys- tem. An increase in the mitotic index caused by activation of the spindle assembly checkpoint due to the presence of kineto- chore that are not properly attached to the mitotic spindle is as sensitive as micronucleus induction and predictive for tubulin binders. This phenomenon is not surprising and has been previ- ously shown in 2D cell culture systems (Muehlbauer et al., 2008). In contrast, aurora B kinase inhibitors were mainly character- ized by increases in 4N and 8N (polyploid) cells, which are the consequence of the inhibition of cytokinesis through the inhibi- tion of aurora kinase B (Bavetsias and Linardopoulos, 2015; Chieffi, 2018; Wu et al., 2013). The quantitative relationships be- tween micronucleus induction and increases in the mitotic in- dex for tubulin binders or polyploidy induction for aurora kinase inhibitors indicate that these quantitative biomarkers could potentially be surrogates to assess potential aneugenicity of test compounds in vivo and might be able to define safety margins for human clinical trials. After 3 days of twice daily topical administration to minipigswith the potent model aneugens, COL and HES, it was unex- pected that no histopathological changes, Ki67 alterations or changes in the nuclear area were detected. The study was designed, using high doses up to ~2000 times those in theEpiDerm, to achieve similar or greater concentrations of COLand HES at the site of aneugenic concern (ie, stratum basale ofthe epidermis) even if skin absorption was limited to <1%. As there is no expected anomaly in sensitivity of skin keratinocytes of minipigs to the biological effects of COL and HES, it is possiblethat despite the approximately 2000-fold higher doses than EpiDerm system, skin concentrations of both compounds in the stratum basale of the minipigs were not adequate to induce a biomarker response. The differences in response between the EpiDerm system and the minipigs could also be the influenced by the differences in the vehicles used. It is possible that the ac- etone solvent used in the EpiDerm system could, if used in vivo, lead to higher concentrations of the test compounds in the stra- tum basale. However, the use of acetone in the minipig study was avoided to minimize the potential confounding contribu- tion of solvent-mediated skin irritation. Alternatively, it is also possible that the duration of the study was insufficient to detect changes in the skin and longer duration studies could be needed, but by comparison, when COL is given to minipigs byoral dosing or intravenous infusion, early onset (ie, <72 h postdose) multi-organ failure and mortality and/or euthanasia is ob- served at a dose as low as 0.25 mg/kg (Eddleston et al., 2018) demonstrating that minipigs are sensitive to rapid onset toxic- ities resulting from tubulin disruption at low doses. It should also be mentioned that based on the low cumulativebodyweight-based dose applied (ie, <0.25 mg/kg), plasma expo-sures would not achieve levels expected to result in systemic toxicities, which does not equate to a lack of exposure in the stratum basale layer of the epidermis. Additionally, COL has been tested in human clinical trials using up to 1% or 10 mg/ml, concentrations greatly exceeding those resulting in positive MN response in the EpiDerm model, for actinic keratoses and recal- citrant psoriasis for up to 12 weeks with no changes in the skin or systemic toxicity (Akar et al., 2001; Faghihi et al., 2016; Kaidbey et al., 1975). This further indicates that COL, even at high doses, is unlikely to be aneugenic in human skin presum- ably due to its barrier function preventing sufficient exposure to the stratum basale. In conclusion, the experimental approaches presented here could help in identifying potential risk management approaches as part of a comprehensive weight of evidence strategy for topically applied drugs with aneugenic potential. The studies demonstrate that the EpiDerm RSMN is sensitive for the hazard identification of aneugens that exert their effect by tubulin binding and aurora kinase inhibition. For the first time, we applied flow cytometric mechanistic biomarkers and quantitative benchmark dose-response modeling to the EpiDerm model demonstrating that these biomarkers are at least as sensitive as micronucleus induction and could poten- tially serve as surrogate biomarkers in toxicology studies in vivo. The 3-day minipig study with high doses of model aneugens indicates that the skin of minipigs is less sensitive to the aneu- genic effects of the model aneugens tested in this study than reconstructed skin models. The lack of response in the skin of minipigs, which has similar morphology compared with human skin, provides evidence that topical aneugen potency in vivo is significantly lower than observed using in vitro systems. These data are a starting point for longer-term in vivo studies and along with improved exposure measurements in skin and pro- vide expanded capabilities to evaluate the aneugenic risk at the epidermal site of interest, the stratum basale. REFERENCES Aardema, M. J., Albertini, S., Arni, P., Henderson, L. M., Kirsch, V. M., Mackay, J. M., Sarrif, A. M., Stringer, D. A., and Taalman, R.D. (1998). Aneuploidy: A report of an ECETOC task force.Mutat. Res. 410, 3–79. Aardema, M. J., Barnett, B. B., Mun, G. C., Dahl, E. L., Curren, R. D., Hewitt, N. J., and Pfuhler, S. (2013). Evaluation of chemicals requiring metabolic activation in the EpiDerm 3D human reconstructed skin micronucleus (RSMN) assay. Mutat. Res. Genet. Toxicol. Environ. Mutagen. 750, 40–49. Aardema, M. J., Barnett, B. C., Khambatta, Z., Reisinger, K., Ouedraogo-Arras, G., Faquet, B., Ginestet, A.-C., Mun, G. C., Dahl, E. L., Hewitt, N. J., et al. (2010). International prevalida- tion studies of the EpiDerm 3D human reconstructed skin micronucleus (RSMN) assay: Transferability and reproduc- ibility. Mutat. Res. 701, 123–131. Akar, A., Bulent Tastan, H., Erbil, H., Arca, E., Kurumlu, Z., and Gur, A. R. (2001). Efficacy and safety assessment of 0.5% and 1% colchicine cream in the treatment of actinic keratoses. J. Dermatolog. Treat. 12, 199–203. Anastassiadis, T., Deacon, S. W., Devarajan, K., Ma, H., andPeterson, J. R. (2011). Comprehensive assay of kinase cata- lytic activity reveals features of kinase inhibitor selectivity. Nat. Biotechnol. 29, 1039–1045. Asano, N., Torous, D. K., Tometsko, C. R., Dertinger, S. D., Morita, T., and Hayashi, M. (2006). Practical threshold for micro- nucleated reticulocyte induction observed for low doses of mitomycin C, Ara-C and colchicine. Mutagenesis 21, 15–20. Avon, S. L., and Wood, R. E. (2005). Porcine skin as an in-vivomodel for ageing of human bite marks. J. Forensic Odontostomatol. 23, 30–39. Bavetsias, V., and Linardopoulos, S. (2015). Aurora KinaseInhibitors: Current Status and Outlook. Front. Oncol. 5, 278. Benfeldt, E. (1999). In vivo microdialysis for the investigation of drug levels in the dermis and the effect of barrier perturba- tion on cutaneous drug penetration. Studies in hairless rats and human subjects. Acta Derm. Venereol. Suppl. 206, 1–59. Bentley, K. S., Kirkland, D., Murphy, M., and Marshall, R. (2000). Evaluation of thresholds for benomyl- and carbendazim- induced aneuploidy in cultured human lymphocytes using fluorescence in situ hybridization. Mutat. Res. 464, 41–51. Bernacki, D. T., Bryce, S. M., Bemis, J. C., and Dertinger, S. D.(2019a). Aneugen molecular mechanism assay: Proof-of- concept with 27 reference chemicals. Toxicol. Sci. 170, 382–393. Bernacki, D. T., Bryce, S. M., Bemis, J. C., Dertinger, S. D., Witt, K.L., and Smith-Roe, S. L. (2019b). Evidence for an aneugenic mechanism of action for micronucleus induction by black cohosh extract. Environ. Mol. Mutagen. 60, 845–856. Bryce, S. M., Bernacki, D. T., Bemis, J. C., Spellman, R. A., Engel, M.E., Schuler, M., Lorge, E., Heikkinen, P. T., Hemmann, U., Thybaud, V., et al. (2017). Interlaboratory evaluation of a mul- tiplexed high information content in vitro genotoxicity as- say. Environ. Mol. Mutagen. 58, 146–161. Bryce, S. M., Bernacki, D. T., Smith-Roe, S. L., Witt, K. L., Bemis, J. C., and Dertinger, S. D. (2018). Investigating the generalizabil- ity of the multiflow (R) DNA damage assay and several com- panion machine learning models with a set of 103 diverse test chemicals. Toxicol. Sci. 162, 146–166. Cammerer, Z., Schumacher, M. M., Kirsch-Volders, M., Suter, W., and Elhajouji, A. (2010). Flow cytometry peripheral blood mi- cronucleus test in vivo: Determination of potential thresh- olds for aneuploidy induced by spindle poisons. Environ. Mol. Mutagen. 51, 278–284. Chapman, K. E., Thomas, A. D., Wills, J. W., Pfuhler, S., Doak, S. H., and Jenkins, G. J. S. (2014). Automation and validation of micronucleus detection in the 3D EpiDermVR human recon- structed skin assay and correlation with 2D dose responses. Mutagenesis 29, 165–175. Chieffi, P. (2018). Aurora B: A new promising therapeutic target in cancer. Intract. Rare Dis. Res. 7, 141–144. Cross, S. E., and Roberts, M. S. (2000). The effect of occlusion on epidermal penetration of parabens from a commercial al- lergy test ointment, acetone and ethanol vehicles. J. Invest. Dermatol. 115, 914–918. Curren, R. D., Mun, G. C., Gibson, D. P., and Aardema, M. J. (2006). Development of a method for assessing micronucleus induc- tion in a 3D human skin model (EpiDerm). Mutat. Res. 607, 192–204. Dahl, E. L., Curren, R., Barnett, B. C., Khambatta, Z., Reisinger, K., Ouedraogo, G., Faquet, B., Ginestet, A.-C., Mun, G., Hewitt, N. J., et al. (2011). The reconstructed skin micronucleus assay (RSMN) in EpiDermVR : Detailed protocol and harmonized scor- ing atlas. Mutat. Res. 720, 42–52. Dertinger, S. D., Kraynak, A. R., Wheeldon, R. P., Bernacki, D. T., Bryce, S. M., Hall, N., Bemis, J. C., Galloway, S. M., Escobar, P. A., and Johnson, G. E. (2019). Predictions of genotoxic poten- tial, mode of action, molecular targets, and potency via a tiered multiflow(R) assay data analysis strategy. Environ. Mol. Mutagen. 60, 513–533. Ditchfield, C., Johnson, V. L., Tighe, A., Ellston, R., Haworth, C., Johnson, T., Mortlock, A., Keen, N., and Taylor, S. S. (2003). Aurora B couples chromosome alignment with anaphase by targeting BubR1, Mad2, and Cenp-E to kinetochores. J. Cell Biol. 161, 267–280. Dreher, F., Fouchard, F., Patouillet, C., Andrian, M., Simonnet, J. T., and Benech-Kieffer, F. (2002). Comparison of cutaneous bioavailability of cosmetic preparations containing caffeine or alpha-tocopherol applied on human skin models or hu- man skin ex vivo at finite doses. Skin Pharmacol. Appl. Skin Physiol. 15, 40–58. Eddleston, M., Fabresse, N., Thompson, A., Al Abdulla, I., Gregson, R., King, T., Astier, A., Baud, F. J., Clutton, R. E., and Alvarez, J. C. (2018). Anti-colchicine Fab fragments prevent lethal colchicine toxicity in a porcine model: A pharmacoki- netic and clinical study. Clin. Toxicol. 56, 773–781. EFSA Scientic Committee, Hardy, A., Benford, D., Halldorsson, T., Jeger, M. J., Knutsen, K. H., More, S., Mortensen, A., Naegeli, H., Noteborn, H., et al. (2017). Update: Use of the benchmark dose approach in risk assessment. EFSA J. 15, e04658. Elhajouji, A., Tibaldi, F., and Kirsch-Volders, M. (1997). Indication for thresholds of chromosome non-disjunction versuschromosome lagging induced by spindle inhibitors in vitro in human lymphocytes. Mutagenesis 12, 133–140. Elhajouji, A., Van Hummelen, P., and Kirsch-Volders, M. (1995). Indications for a threshold of chemically-induced aneu- ploidy in vitro in human lymphocytes. Environ. Mol. Mutagen. 26, 292–304. Faghihi, G., Elahipoor, A., Iraji, F., Behfar, S., and Abtahi-Naeini,B. (2016). Topical colchicine gel versus diclofenac sodium gel for the treatment of actinic keratoses: A randomized, double- blind study. Adv. Med. 2016, 1–6. Goldenson, B., and Crispino, J. D. (2015). The aurora kinases in cell cycle and leukemia. Oncogene 34, 537–545. Hauf, S., Cole, R. W., LaTerra, S., Zimmer, C., Schnapp, G., Walter, R., Heckel, A., van Meel, J., Rieder, C. L., and Peters, J. M. (2003). The small molecule Hesperadin reveals a role for Aurora B in correcting kinetochore-microtubule attachment and in maintaining the spindle assembly checkpoint. J. Cell Biol. 161, 281–294. Herkenne, C., Alberti, I., Naik, A., Kalia, Y. N., Mathy, F. X., Preat, V., and Guy, R. H. (2008). In vivo methods for the assessment of topical drug bioavailability. Pharm. Res. 25, 87–103. Hu, T., Kaluzhny, Y., Mun, G. C., Barnett, B., Karetsky, V., Wilt, N., Klausner, M., Curren, R. D., and Aardema, M. J. (2009). Intralaboratory and interlaboratory evaluation of the EpiDerm 3D human reconstructed skin micronucleus (RSMN) assay. Mutat. Res. 673, 100–108. ICH Guideline. (2011). Guidance on genotoxicity testing and data interpretation for pharmaceuticals intended for human use S2 (R1). In International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use. Cited May 2020. Available at: https://database.ich.org/sites/ default/files/S2%28R1%29%20Guideline.pdf. Iizuka, H. (1994). Epidermal turnover time. J. Dermatol. Sci. 8, 215–217. Kaidbey, K. H., Petrozzi, J. W., and Kligman, A. M. (1975). Topical colchicine therapy for recalcitrant psoriasis. Arch. Dermatol. 111, 33–36. Klaeger, S., Heinzlmeir, S., Wilhelm, M., Polzer, H., Vick, B., Koenig, P. A., Reinecke, M., Ruprecht, B., Petzoldt, S., Meng, C., et al. (2017). The target landscape of clinical kinase drugs. Science 358, eaan4368. Kops, G. J., Weaver, B. A., and Cleveland, D. W. (2005). On the road to cancer: Aneuploidy and the mitotic checkpoint. Nat. Rev. Cancer 5, 773–785. Lavker, R. M., Dong, G., Zheng, P. S., and Murphy, G. F. (1991). Hairless micropig skin. A novel model for studies of cutane- ous biology. Am. J. Pathol. 138, 687–697. Lovell, D. P., Fellows, M., Marchetti, F., Christiansen, J., Elhajouji, A., Hashimoto, K., Kasamoto, S., Li, Y., Masayasu, O., Moore, M. M. , et al.. (2018). Analysis of negative historical control group data from the in vitro micronucleus assay using TK6 cells.Mutat. Res. Genet. Toxicol. Environ. Mutagen.825, 40–50. Lynch, A. M., Eastmond, D., Elhajouji, A., Froetschl, R., Kirsch- Volders, M., Marchetti, F., Masumura, K., Pacchierotti, F., Schuler, M., and Tweats, D. (2019). Targets and mechanisms of chemically induced aneuploidy. Part 1 of the report of the 2017 IWGT workgroup on assessing the risk of aneugens for carcinogenesis and hereditary diseases. Mutat. Res. 847, 403025. Marchetti, F., Mailhes, J. B., Bairnsfather, L., Nandy, I., and London, S. N. (1996). Dose-response study and threshold esti- mation of griseofulvin-induced aneuploidy during female mouse meiosis I and II. Mutagenesis 11, 195–200. Muehlbauer, P. A., Spellman, R. A., Gunther, W. C., Sanok, K. E., Wiersch, C. J., O’Lone, S. D., Dobo, K. L., and Schuler, M. J. (2008). Improving dose selection and identification of aneu- gens in the in vitro chromosome aberration test by integra- tion of flow cytometry-based methods. Environ. Mol. Mutagen. 49, 318–327. Mun, G. C., Aardema, M. J., Hu, T., Barnett, B., Kaluzhny, Y., Klausner, M., Karetsky, V., Dahl, E. L., and Curren, R. D. (2009). Further development of the EpiDerm 3D reconstructed hu- man skin micronucleus (RSMN) assay. Mutat. Res. 673, 92–99. Nishikawa, T., Haresaku, M., Adachi, K., Masuda, M., and Hayashi, M. (1999). Study of a rat skin in vivo micronucleus test: Data generated by mitomycin C and methyl methane- sulfonate. Mutat. Res. 444, 159–166. Nishikawa, T., Haresaku, M., Fukushima, A., Nakamura, T., Adachi, K., Masuda, M., and Hayashi, M. (2002). Further eval- uation of an in vivo micronucleus test on rat and mouse skin: Results with five skin carcinogens. Mutat. Res. 513, 93–102. Nishikawa, T., Nakamura, T., Fukushima, A., and Takagi, Y. (2005). Further evaluation of the skin micronucleus test: Results obtained using 10 polycyclic aromatic hydrocarbons. Mutat. Res. 588, 58–63. OECD (2016). Test No. 487: In Vitro Mammalian Cell Micronucleus Test. Cited Dec 2020. Available at: https://database.ich.org/ sites/default/files/S2%28R1%29%20Guideline.pdf. Olaharski, A. J., Gonzaludo, N., Bitter, H., Goldstein, D., Kirchner, S., Uppal, H., and Kolaja, K. (2009). Identification of a kinase profile that predicts chromosome damage induced by small molecule kinase inhibitors. PLoS Computat. Biol. 5, e1000446. Pfuhler, S., van Benthem, J., Curren, R., Doak, S. H., Dusinska, M., Hayashi, M., Heflich, R. H., Kidd, D., Kirkland, D., Luan, Y., et al. (2020). Use of in vitro 3D tissue models in genotoxicity testing: Strategic fit, validation status and way forward. Report of the working group from the 7(th) International Workshop on Genotoxicity Testing (IWGT). Mutat. Res. 850– 851, 503135. Roy, S., Kulkarni, R., Hewitt, N. J., and Aardema, M. J. (2016). The EpiDerm 3D human reconstructed skin micronucleus (RSMN) assay: Historical control data and proof of principle studies for mechanistic assay adaptations. Mutat. Res. Genet. Toxicol. Environ. Mutagen. 805, 25–37 Sasaki, J. C., Allemang, A., Bryce, S. M., Custer, L., Dearfield, K. L., Dietz, Y., Elhajouji, A., Escobar, P. A., Fornace, A. J., Jr., Froetschl, R., et al. (2020). Application of the adverse outcome pathway framework to genotoxic modes of action. Environ. Mol. Mutagen. 61, 114–134. Schafer-Korting, M., Bock, U., Diembeck, W., Dusing, H. J., Gamer, A., Haltner-Ukomadu, E., Hoffmann, C., Kaca, M., Kamp, H., Kersen, S., et al. (2008). The use of reconstructed humanepidermis for skin absorption testing: Results of the valida- tion study. Altern. Lab. Anim. 36, 161–187. Schreiber, S., Mahmoud, A., Vuia, A., Rubbelke, M. K., Schmidt, E., Schaller, M., Kandarova, H., Haberland, A., Schafer, U. F., Bock, U., et al. (2005). Reconstructed epidermis versus human and animal skin in skin absorption studies. Toxicol. In Vitro 19, 813–822. Slob, W. (2017). A general theory of effect size, and its conse- quences for defining the benchmark response (BMR) for con- tinuous endpoints. Crit. Rev. Toxicol. 47, 342–351. Todo, H. (2017). Transdermal permeation of drugs in various ani- mal species. Pharmaceutics 9, 33. Tsai, J. C., Sheu, H. M., Hung, P. L., and Cheng, C. L. (2001). Effect of barrier disruption by acetone treatment on the permeabil- ity of compounds with various lipophilicities: Implications for the permeability of compromised skin. J. Pharm. Sci. 90, 1242–1254. Vardaxis, N. J., Brans, T. A., Boon, M. E., Kreis, R. W., and Marres,L. M. (1997). Confocal laser scanning microscopy of porcine skin: Implications for human wound healing studies. J. Anat. 190, 601–611. Weinstein, G. D., McCullough, J. L., and Ross, P. (1984). Cell prolif- eration in normal epidermis. J. Invest. Dermatol. 82, 623–628. Wheeldon, R. P., Bernacki, D. T., Dertinger, S. D., Bryce, S. M., Bemis, J. C., and Johnson, G. E. (2020). Benchmark dose analy- sis of DNA damage biomarker responses provides compound potency and adverse outcome pathway information for the topoisomerase II inhibitor class of compounds. Environ. Mol. Mutagen. 61, 396–407. White, P. A., Long, A. S., and Johnson, G. E. (2020). Quantitative interpretation of genetic toxicity dose-response data for risk assessment and regulatory decision-making: Current status and emerging priorities. Environ. Mol. Mutagen. 61, 66–83. Wills, J. W., Johnson, G. E., Doak, S. H., Soeteman-Hernandez, L. G., Slob, W., and White, P. A. (2016). Empirical analysis of BMD metrics in genetic toxicology part I: In vitro analyses to provide robust potency rankings and support MOA determi- nations. Mutagenesis 31, 255–263. Wu, Q., Zhang, W., Mu, T., Song, T., and Li, D. (2013). Aurora B ki- nase is required for Hesperadin cytokinesis through effecting spindle structure. Cell Biol. Int. 37, 436–442.
Yuki, K., Ikeda, N., Nishiyama, N., and Kasamatsu, T. (2013). The reconstructed skin micronucleus assay in EpiDerm: Reduction of false-positive results – A mechanistic study with epigallocatechin gallate. Mutat. Res. Genet. Toxicol. Environ. Mutagen. 757, 148–157.
Zeller, A., Duran-Pacheco, G., and Guerard, M. (2017). An ap- praisal of critical effect sizes for the benchmark dose ap- proach to assess dose-response relationships in genetic toxicology. Arch. Toxicol. 91, 3799–3807.