*3.3. Issues Regarding the Toxicity of Alkenylbenzenes*

#### 3.3.1. The Genotoxic and Carcinogenic Potential of Alkenylbenzenes

Toxicity data regarding estragole, safrole, and methyleugenol show that the genotoxic and carcinogenic potential of these compounds is complex and may differ—at least in part—from that of other structurally related alkenylbenzenes, such as *trans*-anethole and eugenol. This may be based—amongst other things—on toxicokinetic differences of alkenylbenzenes, albeit having only slight structural differences, such as estragole vs. *trans*-anethole or methyleugenol vs. eugenol. In many cases, there are no adequate studies regarding carcinogenicity of different alkenylbenzenes, such as elemicin and apiol. In addition, studies investigating the genotoxic potential are often missing or the study design is not adequate to reliably address the genotoxic potential of that class of compounds. Therefore, additional studies are needed, especially those designed according to international guidelines and taking into account the alkenylbenzene specific bioactivation via SULTs to allow a comparative analyses and assessment of the (geno-)toxic potential of alkenylbenzenes in a conclusive manner.

#### 3.3.2. Weaknesses of Standard Genotoxicity Tests and Implications for Hazard Assessment

In order to identify the possible genotoxic activity of a given substance, genotoxicity studies are conducted and evaluated in several legal sectors. As a general rule, at least one mutagenicity test with bacteria and one cytogenicity test with mammalian cells is required [161–165]. Depending on the legal area, in vivo genotoxicity studies are either generally requested or may be subsequently required based on the findings of the in vitro tests.

There are many new or revised Organisation for Economic Co-operation and Development (OECD) test guidelines [166] for several genotoxic endpoints available. These test guidelines comprise the requirements for a reliable study design as well as an acceptable presentation of the study results.

Nevertheless, the tests described by OECD should be regarded as standard tests, which are generally suitable for identifying a possible genotoxic activity of a test substance, but need to be adapted to the individual case. As a prerequisite, information on metabolism as well as mechanistic understanding of the test compound is needed before the genotoxicity test is carried out in a modified way.

The weaknesses of the in vitro and in vivo standard test systems can manifest themselves in either false-positive or false-negative results. False test results are to be avoided in the regulatory process in order to prevent unnecessary animal studies, but also to enable protection of human health. Some pitfalls of standard genotoxicity studies—with special focus on the situation for alkenylbenzenes—will be discussed in the following section.

False-Negative Results

Depending on the existence of particular functional chemical groups in the molecular structure of the test substance, it might be possible that the test substance reacts with components of the metabolic activation system (i.e., proteins of the S9 mix) or the solvent (i.e., DMSO). As described by Nestmann et al. in 1985, DMSO can undergo chemical reactions with alkyl halides [167]. Consequently, the lowered effective concentration of the test substance reduces the sensitivity of the test system and can provoke false-negative results.

Another possibility for an artificial negative result in the Ames test can be extreme experimental conditions, such as drastic changes in pH leading to cytotoxicity. This could, ultimately, mask the quantitative formation and detection of revertant colonies.

Moreover, the solubility as well as the stability of the test substance plays an important role in the sensitivity of the test system. Substances with a short half-life could decay before passing through the bacterial cell wall. Thus, contact with the genetic material in the bacterium would be prevented. This circumstance is particularly critical as the test substance is per se reactive due to its low stability. Thus, nucleophilic reactions of the DNA with the test substance are considered likely.

For some substances, it has also been shown that they can be detected better in the Ames test with the pre-incubation method than with the standard plate incorporation assay. Among these are substances with special structural characteristics, such as short chain aliphatic nitrosamines, divalent metals, aldehydes, azo-dyes and diazo compounds, pyrollizidine alkaloids, allyl compounds, and nitro compounds [168]. In order to avoid false-negative results, the appropriate modification of the Ames test should be favored.

For other genotoxic endpoints, such as clastogenicity, false-negative results are also described in the literature. Substances producing crosslinks with DNA should be handled with care in Comet assays in which DNA strand breaks are detected. To DNA/DNAintra-strand and DNA/DNA-inter-strand-crosslinkers belong cisplatin and mytomycin C, respectively [169]. Both compounds impair the sensitivity of the Comet assay due to their potential to reduce DNA fragmentation. Consequently, the formation of comets is not adequately captured.

Many substances are not inherently genotoxic but require critical enzymatic steps to form reactive intermediates. The bioactivation process is often mediated by so-called phase I enzymes. On the contrary, phase II enzymes serve the purpose of making substances more inert, water-soluble and thus easier to excrete. Metabolic competence of a test system depends on various factors (e.g., cell type, S9 mix, co-factors). For instance, different cell types express different enzymes. To illustrate this, the metabolic competence of liver cells is greater than that of human lung fibroblasts (V79 cells). However, the latter are often used in common genotoxicity studies, and even the addition of S9 mix—as recommended by OECD test guidelines—cannot compensate for many phase II enzymes or their co-factors. The individual enzyme capacity and activity level within a given test system depends on many further factors. For instance, S9 mix can be obtained from the liver, but also from other organs, such as the lungs or kidney. Of note, the enzyme composition and level are also species-dependent and can be affected by the use of chemical inductors. Although S9 mix is often obtained from livers of arochlor-treated rats, the use of hamster S9 mix or even human S9 mix is advantageous in specific cases. S9 mix largely represents phase I enzymes (CYP). The addition of the appropriate co-factor (nicotinamide adenine dinucleotide phosphate (NADPH)-generating system) thus primarily boosts oxidative conversions. Other enzymes, such as the epoxide hydrolases, require no co-factors other than water, and are thus also addressed by using S9 mix. The situation is different for most phase II enzymes. These are either not expressed by the bacteria or mammalian cells, are not components of the standard S9 mix, or are less active due to a lacking co-factor. Even if a reactive metabolite is formed extracellularly by a metabolic activation system, such as S9 mix, it is questionable whether this metabolite can permeate the cellular barrier and reach the genetic material. All these aforementioned factors regarding biotransformation can contribute to false-negative results.

False-negative results also play an important role in animal experiments. The test for micronuclei formation in bone marrow or peripheral blood is generally only considered reliable and valid if the test substance (or its metabolites) is systemically available. To ensure this, the corresponding OECD test guideline 474 [170] recommends using a reduction in the polychromatic erythrocytes/normochromatic erythrocytes (PCE/NCE) ratio as a surrogate for bioavailability. If the ratio is not decreased compared to the PCE/NCE ratio of the control animals, systemic availability of the test substance is difficult to prove without further information (e.g., clinical signs of animals, detection of the test substance in blood plasma). However, study evaluators are often faced with the problem that many test substances are not cytotoxic in the bone marrow. Consequently, they do not modify the PCE/NCE ratio. In this respect, the test substance could be systemically available without a change in the PCE/NCE ratio.

#### False-Positive Results

False positive results can often be explained by the presence of extreme conditions in the culture medium. For example, strong fluctuations in pH or osmolality can lead to cytotoxicity and eventually to artificial positive test results.

Another example for false-positive results might be the enzyme equipment of the test system. The classical Salmonella strains applied in the Ames test express nitroreductases. These enzymes allow for azo- and nitroreduction. However, these enzymes are not present in mammals. Owing to this uncertainty, an extrapolation to the human situations might be difficult. Theoretically, metabolic activation of nitro compounds can be mediated by intestinal bacteria in humans. To clarify this, an absorption test should be carried out in these cases. If the test substance (or its metabolites) is completely absorbed, the risk of intestinal bioactivation is low, as direct contact with the intestinal bacteria is unlikely [169].

#### Possible Ways to Optimize Standard Genotoxicity Tests

One option to mimic possible bioactivation and detoxification steps—taking into account phase I and also phase II enzymes—is the use of bacteria, mammalian cells, and animals that have been genetically modified to artificially express certain enzymes. In this way, bacteria and cells with murine enzymes could serve the purpose of studying possible biotransformation processes in mice. Toxicological endpoints could be the bacterial reverse mutation assay, but also the hypoxanthine-guanine phosphoribosyltransferase (HPRT) assay or the mouse lymphoma assay. Apart from that, indicator assays such as DNA adduct formation could also give indications about the genotoxic activity of a test compound. Furthermore, genotoxicity studies applying bacteria, mammalian cells, or even animals expressing the corresponding murine enzymes might be of particular importance if, for example, tumor formation was observed in mice and the underlying mechanism has to be clarified. Finally, experiments utilizing human enzymes could help to better extrapolate findings from animal studies to the human situation.

In order to select a suitable genotoxicity model, information on the species-dependent metabolism as well as a hypothesis for the underlying genotoxic mode of action is indispensable. A number of in silico tools have been established for the identification of possible structural alerts for genotoxicity. Many of them are well-trained and generate reliable predictions about metabolism. These prediction programs can, therefore, help to identify critical metabolites that might mediate genotoxic events.

Of note, in silico programs should be selected with care. If a chemical structure is too dissimilar for what the in silico model is trained for, a reliable prediction for metabolism or genotoxicity can be challenging. For this reason, a prerequisite for conducting valid and robust predictions is the availability of appropriate representative training data within the in silico model [171].

Optimization of Standard Genotoxicity Tests Using the Example of the Alkenylbenzene Methyleugenol

The dominant metabolic pathway relevant for genotoxicity is conversion of methyleugenol to 1- -hydroxymethyleugenol via CYP enzymes [61]. After sulfo-conjugation of the allylic hydroxyl group by SULTs, electrophilic esters are formed, which can be attacked by nucleophilic structures in the cell (e.g., DNA or proteins). This critical bioactivation step has been described not only for methyleugenol, but also for other alkenylbenzenes, such as safrole and estragole [79,80,102,120,172–177]. If DNA adducts are not error-free repaired by the cell's repair system, they can manifest as mutations. This is particularly concerning if proto-oncogenes or tumor suppressor genes are affected, as cancer development might be triggered.

The genotoxic and mutagenic activity of methyleugenol has been tested in numerous standard in vitro tests. In the bacterial reverse mutation test with conventional bacterial strains, methyleugenol was not mutagenic [113,178,179]. This finding did not change with the addition of an exogenous activation system (S9 mix). The main reason for this observation is that SULTs are not considered in conventional genotoxicity studies. Whereas the use of an S9 mix can increase the metabolic competence for phase I enzymes (if appropriate co-factors are added) many phase II enzymes—such as SULTs—remain unconsidered.

In contrast, mutagenic findings have been observed in bacteria being genetically modified for the expression of murine and human SULTs [77]. Likewise, DNA adduct formation was higher in bacteria, cells, and animals expressing murine or human SULTs in comparison to the wild-type [77,180]. This illustrates that standard genotoxicity tests should be optimized and adapted to the relevant question. However, when quantitatively comparing DNA adduct levels and the mutagenic potential of alkenylbenzenes between conventional, murine, and humanized test systems, attention should be paid to how much SULT (and which form) is expressed in which bacterial strain and animal model [77,180]. Furthermore, the SULT status in mice may also vary between tissues [181].
