Mouse Skin Models for Carcinogenic Hazard Identification: Utilities and Challenges

Abstract

This report addresses 1) the predictability of mouse skin models for carcinogenic hazard identification, 2) the association between early changes in the skin and later tumorigenic responses, and 3) the relative sensitivity of three mouse models of skin tumorigenesis; i.e. the genetically-initiated Tg.AC and RasH2 lines and the SENCAR mouse model. All three mouse models responded similarly, with mild inflammation and epidermal hyperplasia, to several weeks of treatment with a topical agent. Based on our previous research experience, we hypothesized that inflammation, irritation, proliferation, and/or hyperplasia in the skin would precede and predict the appearance of tumors in these sensitive mouse skin models. Consistent with our hypothesis, the test agent caused a low but significant tumorigenic response in Tg.AC mice. We propose that inflammation, irritation, and hyperplasia are sensitive predictors of a later tumorigenic response in Tg.AC mice. Further studies are needed, however, to better determine the relative sensitivity of these 3 models to a wider variety of agents.

Introduction

Rodent models are commonly used as predictors of carcinogenic risk to humans. Animal models have also provided substantial mechanistic information applicable to human disease. In particular, the mouse skin model of carcinogenesis has revealed the multistage nature of carcinogenesis, which involves both genetic and nongenetic events. By separating the process of carcinogenesis into mechanistically and temporally distinct stages, the steps of initiation, tumor promotion (or premalignant progression), and malignant progression were identified (reviewed in (Yuspa, 1994)). These stages have subsequently been identified in colorectal and other human cancers as well (Goelz et al., 1985; Fearon et al., 1987; Law et al., 1988; Vogelstein et al., 1988).

Mouse models with enhanced susceptibility to skin tumorigenesis have recently gained interest for carcinogen identification. At issue in the scientific community today are questions about (1) how well these sensitive mouse skin models predict carcinogenic responses compared to 2-year rodent studies, (2) whether there are certain types of agents that cause false positive responses idiosyncratic to these models, (3) whether early changes in the skin can predict later tumor development, and (4) the relative sensitivity of several commonly used and highly sensitive lines. This article presents current thinking about these issues and a recent evaluation of the relative sensitivity of 3 mouse skin models using early markers that are known to be associated with tumorigenic response.

Modified Mouse Skin Models For Identification Of Carcinogenic Hazard

Mice historically have been used in safety evaluation studies and are recommended by regulatory agencies for this purpose. Several mouse skin models with enhanced sensitivity to carcinogens have been developed and proposed for carcinogenic hazard identification. Prominent among these models for carcinogen identification are three lines that are the subject of this review: the Tg.AC mouse, the RasH2 mouse, and the SENCAR mouse. Both Tg.AC and RasH2 models are genetically-initiated ras ^Ha transgenic lines. Tg.AC mice were developed following the introduction of a ζ-globin promoter-driven v-ras ^Ha transgene (Leder et al., 1990). One of several founders exhibited a tumorigenic phenotype following topical exposure to the tumor promoter 12-O-tetradecanoyl phorbol-13-acetate (TPA).

Because of the sensitivity and specificity of the response of Tg.AC mice to carcinogens, they have been frequently used for carcinogenic hazard identification (reviewed in Pritchard et al., 2003). The RasH2 mouse, also known as Tg.rasH2, was originally described by Saitoh et al. (1990). This line was created by insertion of a human c-ras ^Ha transgene driven by its own promoter. Hemizygous RasH2 mice respond with greater sensitivity to carcinogens than nontransgenic mice and are similarly recommended for genotoxic and nongenotoxic carcinogen identification (reviewed in Yamamoto et al., 1998; Usui et al., 2001; Morton et al., 2002)).

In contrast to the other 2 models, SENCAR mice were not genetically-engineered, rather the line was selected over 8 generations by Boutwell and colleagues for increased skin tumor multiplicity and decreased tumor latency in response to 7,12-dimethyl benz[a]anthracene (DMBA) and TPA treatment (reviewed in Slaga, 1986). SENCAR mice have an approximately 10–20 fold increase in sensitivity to DMBA initiation and 2–3-fold increase in sensitivity to TPA promotion compared to the parental CD-1 stock (Slaga, 1986). The SENCAR mouse is an accepted and commonly used short-term model system for evaluating the promoting or initiating activity of test items for two stage skin carcinogenesis in mice. These mice respond rapidly and sensitively with skin tumors following topical application of a single low dose of an initiating agent, typically a mutagenic carcinogen, followed by multiple applications of a tumor promoting agent. Tumor experiments in these 3 models are generally completed within 20–26 weeks, with benign squamous papillomas as the endpoint of interest. This sensitivity and quick response has made these models appealing for drug developers because of the time-consuming and costly nature of two year rodent bioassays.

Comparison of The Sensitivity and Predictive Ability of Sensitive Mouse Strains For Carcinogenic Hazard Identification

The National Toxicology Program (NTP) 2-year bioassays in rats and mice have been the standard approach for carcinogen identification for many years. More recently, genetically-engineered mouse models have been included as potential alternatives for the NTP bioassay. These models include the v-ras ^Ha transgenic Tg.AC mouse, the rasH2 mouse, and hemizygous p53 (Trp53+/−) mouse (reviewed in (Pritchard et al., 2003)). Perceived advantages of the transgenic models include practical considerations; such as more rapid and inexpensive testing, somewhat better overall predictability compared to the 2-year bioassay, and the opportunity to ascertain mechanistic information. Information from both the 2-year bioassay and transgenic models has been collected for many chemicals with widely divergent mechanisms of action, which allows an analysis of the suitability of these transgenic models for predicting carcinogenic potential in humans.

The accuracy of these models for carcinogen identification has been compared across a large, matched set of chemicals across various routes of administration (Table 1). As reviewed in Pritchard et al., the genetically engineered rasH2, hemizy-gous p53 (Trp53+/−), and Tg.AC models showed better predictability than the 2-year bioassay (Pritchard et al., 2003). For Tg.AC mice, the overall accuracy was 74% compared to the 69% for the 2-year, 2-species NTP bioassay. For RasH2 mice, overall accuracy was 81% (Pritchard et al., 2003). It was suggested that combining the 2-year rat bioassay with studies in transgenic models may be the most effective approach to identify compounds that may be a carcinogenic hazard. For example, Tg.AC and Trp53+/− models combined with the rat bioassay resulted in an overall accuracy of 84%.

The SENCAR mouse model has also been frequently used to identify carcinogens following dermal application, in particular for agents with tumor promoting activity on the skin. A review of the literature reveals a number of agents that have been tested in both SENCAR mice and the NTP 2-year bioassay (NTP, 1996 and Table 1). Most of these agents were tested because of their presumptive tumor promoting ability. In these studies, the SENCAR mice are first initiated with a mutagenic agent at low dose before chronic application of the test agent. Using these experimental conditions, SENCAR mice are a sensitive predictor of tumor promoting ability.

We wished to develop a scientific rationale to support the selection of the best model for a carcinogenesis study involving topical application of novel agents. Our review of the literature revealed very few agents utilized in all three of the Tg.AC, RasH2, and SENCAR models (Table 1). This was somewhat surprisingly because these models are frequently used to assess tumorigenesis in the skin. The archetypal tumor promoter TPA induces tumors in all three models; although TPA did not induce tumors in SENCAR mice unless it was administered following initiation with DMBA (Table 1). This is consistent with the lack of genetic initiation in the SEN-CAR mouse. Although a few papillomas arise in SENCAR mice treated over many weeks with TPA, a very strong tumor promoter, no dose-response relationship occurs in the absence of initiation. In contrast, genetically initiated Tg.AC and RasH2 mice do not require the application of an initiator for tumor development.

The sensitivity of Tg.AC mice is, however, strongly and positively correlated with the age of the mice at the start of treatment (Battalora et al., 2001). The mechanism of this effect is not known, although it may be related to an expansion of transgene expressing cells in the skin as the mice age. Older (21–32-week-old) Tg.AC mice do seem to be more sensitive to the tumor promoting effects of TPA when compared to SENCAR mice. They generally respond with greater tumor multiplicity (more than 50 papillomas per mouse compared to at most 22 papillomas/mouse in SENCAR) and shorter latency (only 3 weeks compared to 6 or more weeks in SEN-CAR), and greater sensitivity to a lesser cumulative dose of TPA (1 application is sufficient to cause papillomas in Tg.AC mice) (Slaga, 1986; Hansen et al., 1998; Battalora et al., 2001).

Wounding and chronic irritation are also well known tumor promoters (Hennings et al., 1970; Argyris, 1985). Both Tg.AC and DMBA-initiated SENCAR mice develop tumors following a single full-thickness incision (Leder, 1990; Di-Giovanni et al., 1993; Hansen et al., 1994a; Cannon et al., 1997), although tumor multiplicity (app. 5 pap/ms v. 0.76) and tumor incidence (100% v. 41%) were higher in Tg.AC compared to SENCAR mice. Among the chemicals agents of Table 1, di(2-ethylhexyl)phthalate was also tested in all 3 lines. Surprisingly, this agent was clearly positive only in the RasH2 mouse, consistent with the NTP bioassay result, and suggesting greater sensitivity of RasH2 compared to Tg.AC and SENCAR mice.

Tumorigenesis assays of benzene, positive in the Tg.AC and RasH2 mice but not in initiated SENCAR mice, also supports an increased sensitivity of the transgenic lines compared to SENCAR. Several mutagenic carcinogens such as DMBA, benzo[a]pyrene, and N-methyl-N-nitrosourea (MNU), which also serve as initiating agents in the two-stage model, and UV have been tested in SENCAR mice and at least one of the other models. These mutagens produce tumors in all of the lines tested. Comparison of the sensitivity of these three strains to additional agents would provide important information necessary to the selection of the best model for skin tumorigenesis in future studies.

CHALLENGES IN MOUSE MODEL SELECTION FOR CARCINOGENIC HAZARD IDENTIFICATION

There are similarities between human and mouse non-melanoma skin carcinogenesis, including the development of malignant squamous cell carcinoma from pre-existing benign lesions. However, there are significant differences between the mouse model and humans as well. For example, ras ^Ha mutations are much less frequent in human skin cancers. Additionally, basal cell carcinomas and actinic keratoses are much more common in humans while mice usually develop benign squamous papillomas. The stages of carcinogenesis are much more discrete in the mouse as well, with SENCAR mice an especially successful model for the detection of tumor promoters. The transgenic Tg.AC and RasH2 models respond to tumor promoters as well. Regulatory decisions with regard to nonmutagenic tumor promoters are not clearcut, however.

The ras ^Ha transgenic lines are generally not sensitive to non-carcinogens; including those rodent-specific carcinogens which were tested positive in the NTP bioassay, but found not to be carcinogenic in humans. The RasH2 model, in particular, showed nearly perfect discrimination for non-carcinogens (Pritchard et al., 2003). Tg.AC mice, however, have been suggested to respond aberrantly to some agents due to their heightened susceptibility to irritation- and inflammation-induced tumorigenesis, which has complicated interpretation of some Tg.AC carcinogenesis studies.

A better understanding of the mechanisms of tumorigenesis in Tg.AC mice may aid in the appropriate interpretation of results generated using this model. The ζ-globin-promoted v-ras ^Ha transgene is expressed in fetal hematopoietic tissue, but is largely silent after parturition (Leder et al., 1990). Tg.AC mice were initially observed to develop skin tumors following abrasion (Leder et al., 1990). Subsequent experiments demonstrated that Tg.AC mice rapidly develop benign squamous papillomas following topical application of a wide variety of carcinogens. While all Tg.AC tumors express the ras ^Ha transgene (Hansen et al., 1996), the skin of naïve, untreated Tg.AC mice does not detectably express v-ras ^Ha and is histologically normal (Hansen and Tennant, 1994a;Hansen et al., 1994b;Battalora, Spalding, Szczesniak, Cape, Morris, Trempus, Bortner, Lee, and Tennant, 2001). Papillomas arise from the clonal expansion of transgene-expressing follicular keratinocytes in focal hyperplasias (Hansen and Tennant, 1994b). The localization of this population is consistent with a stem cell origin (Hansen and Tennant, 1994b).

Thus, tumor development in Tg.AC skin is a focal response arising from transgene expressing cells in an otherwise transgene-negative tissue. Two possible scenarios could explain this phenotype. Firstly, the focal, transgene-expressing populations may result from the clonal expansion of a very small and, as yet, undetectable resident population of cells that constitutively expresses the transgene. Secondly, Tg.AC tumorigens may induce transgene expression in a few cells within the skin that then expand to form tumors. Unique among all skin tumor models, Tg.AC mice have an increased susceptibility to tumor formation with increased age (Battalora et al., 2001), which is consistent with the expansion of tumor precursor cells with increased age. However, transgene-expressing keratinocytes within untreated, nontumor-bearing Tg.AC skin have not been identified.

Transgenic mouse skin carcinogenicity models, which require 20–26 weeks of treatment, are relatively short and inexpensive when compared to the traditional 2-year, 2-species bioassay. However, these models also have significant limitations intrinsic to genetic background-associated effects; such as gene mutations that may not occur in the human cancer being modeled, tumor site specificity that is not reflective of the patterns of human cancer, and sensitivity to wounding or irritation of questionable applicability to human carcinogenesis.

It has been suggested that Tg.AC mice respond to some agents that are not human carcinogens because of either: (1) a hypersensitivity and predisposition to cutaneous irritation and inflammation in response to some agents, or (2) an increased susceptibility to irritation- and inflammation-induced tumorigenesis. We identified several relevant sets of results with regard to the role of inflammation in inducing papilloma formation in Tg.AC mice. Some agents produce tumors in Tg.AC mice without causing irritation or inflammation. Wyde et al. demonstrated a dose-response for cutaneous papillomas in Tg.AC mice following daily application of only nanogram quantities of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (Wyde et al., 2004). At this dose, the effect of TCDD is clearly a receptor-mediated response that did not produce significant inflammation or irritation. The tumorigenesis in Tg.AC mice induced by agents that cause sustained irritation of the skin should be considered mechanistically distinct from assays in which significant irritation does not occur.

Inflammation in response to chemical application in mouse skin may be induced by either of 2 distinct mechanisms. First, cytotoxicity-induced inflammation is caused by release of pro-inflammatory cytokines from injured keratinocytes. Second, cell-mediated immunity occurs in the case of the several hundred known contact sensitizers in which protein adducts trigger immune cell recruitment to the skin. These two mechanisms may have different implications for carcinogenic hazard. Examples of both mechanisms of inflammation exist among agents used in Tg.AC tumorigenesis assays. Albert et al. (1996) tested the hypothesis that cell-mediated immune response to contact sensitizers, such as 2,4-dinitro-1-fluorobenzene (DNFB), may be responsible for the tumorigenicity of these agents. In these experiments, the anti-inflammatory fluocinolone acetonide was not able to inhibit tumorigenesis induced by the contact sensitizer in Tg.AC mice because DNFB induced inflammation through cytotoxicity under the conditions tested (Albert et al., 1996). On the other hand, overexpression of glucocorticoid receptors was shown to inhibit TPA-induced tumorigenesis in Tg.AC mice; suggesting anti-inflammatory agents may suppress skin carcinogenesis in this model (Budunova, 2003).

Other investigations on the role of irritation and inflammation in tumorigenesis in Tg.AC mice have also produced contradictory results, necessitating further research (Albert et al., 1996; Murphy et al., 2003). The results of Tg.AC studies using two contact sensitizers, ethyl acrylate and tripropylene glycol diacrylate showed a lack of association between inflammation and tumorigenesis in Tg.AC mice. Although both agents cause inflammation, tripropylene glycol diacrylate is a tumorigen in Tg.AC mice, while ethyl acrylate is not (Table 1). Rotenone, acetic acid, and phenol also cause irritation but not papillomas in Tg.AC mice (Sistare et al., 2002). However, the irritating effects of phenol and acetic acid were not well-characterized. These results suggest that the induction of irritation and inflammation are not sufficient to induce tumors in Tg.AC mice, although the timing and extent of irritation caused by these agents could be critical in understanding the response of the mice.

The Tg.AC mouse responds with sensitivity to a wide variety of carcinogens (Spalding et al., 1993; Pritchard 2003). However, the sensitivity of this model to cutaneous injury may also be important to consider when evaluating results from the model. The sensitivity of Tg.AC mice to proliferative and pro-inflammatory stimuli is evidenced by the development of tumors following a single full-thickness incision in the skin (Cannon et al., 1997). Cutaneous wounding is also tumorigenic in SENCAR mice (Table 1), indicating wounding is a true tumor promoting stimulus.

Consistent with this result, agents that cause severe, chronic and overt irritation, inflammation, and increased proliferation are commonly tumorigenic in Tg.AC mice (Sistare et al., 2002). For example, resorcinol, which was negative in the NTP bioassay, induced papillomas in Tg.AC mice when given at a level that also caused inflammation in the skin (Sistare et al., 2002). Another example of irritation-associated tumorigen in Tg.AC mice that proved nontumorigenic in long-term rodent studies include tripropylene glycol diacrylate (Nylander-French et al., 1998).

How should this irritation-induced tumorigenesis be interpreted? Chronic irritation increases carcinogenesis in humans (reviewed in (Albert et al.1996)) and is considered by some to be a tumor promoting stimulus. However, these examples of irritation-associated tumorigenesis in Tg.AC mice may be at least partially due to the high doses used in rodent models and thus would not be expected to cause similar injury when appropriately applied to humans in lesser, non-toxic amounts. These issues have important regulatory implications. The lack of consensus on the link between tumor promotion and irritation clearly calls for additional study of this issue and is the focus of active research. To further investigate these issues, as well as the relative sensitivity of several mouse skin models to topically applied agents, several studies were initiated using a novel test agent.

A TEST CASE

Validating Early Biomarkers as Predictors of Tumorigenesis

The identification of biomarkers of exposure that can predict a tumorigenic response in mouse skin models would facilitate carcinogenic hazard identification greatly. Several short-term responses have been identified as being tightly associated with tumor promotion in mouse skin. These early changes are relevant for human carcinogenesis as well, since many human carcinogens; including ultraviolet irradiation, components of tobacco smoke, and contact sensitizers; have tumor-promoting ability. Tumor promoters almost without exception (Moser et al., 1992; Meyer et al., 1993), increase cell proliferation, induce inflammation, and cause hyperplasia. Inflammation, irritation, and hyperplasia are frequently associated with, and predictive of, tumor promoting activity in mouse models including the Tg.AC mouse (Slaga et al., 1996; Hanausek et al., 2004; Humble et al., 2005). Complete carcinogens, by definition able to both initiate and promote tumors, also cause increased proliferation, inflammation and hyperplasia within the skin. With this in mind, we hypothesized that cutaneous inflammation, irritation, proliferation, and epidermal hyperplasia could be used as short-term endpoints in skin-painting assays to predict the tumorigenic response in Tg.AC mice.

In order to be useful biomarkers of tumorigenic response, alteration in the marker must be measurable after short-term treatment. Thus, to test our hypothesis, Tg.AC mice were treated with a test agent (to be identified in this review as the “vehicle formulation”). The vehicle formulation consisted of a carbomer-based aqueous gel containing excipients that are generally regarded as safe (GRAS), including propylene glycol, nonionic surfactants, and paraben preservative. Mice were treated with the vehicle formulation for several weeks in a short-term experiment, and also for 26 weeks in a long-term tumor experiment. In the short-term experiment, groups of Tg.AC mice were treated with 0.1 mL of the test agent vehicle formulation, 5 μg TPA in 0.1 mL acetone as a positive control, or were untreated (negative control).

As expected, TPA-treated mouse skin was grossly irritated in all genotypes beginning 1 week after the start of treatment, as shown by desquamation, roughness, and erythema (Figure 1A). Significant edema also occurred in the TPA-treated mice by 2 weeks of treatment, and worsened over time (Figure 1B). Tumors appeared beginning at 5 weeks after the start of TPA treatment. As a result, the TPA-treated mice were euthanized at 6 weeks. In contrast to the results with TPA, the vehicle formulation-treated mouse skin sustained little change in appearance over the course of 10 weeks of treatment (Figure 1A). A significant increase in skin-fold thickness, consistent with the induction of edema, occurred in the vehicle formulation treated mice, although the response was less than that of TPA-treated mice (Figure 1B). As shown in Fig. 2A–B, the test agent vehicle induced a very slight but significant epidermal hyperplasia (determined using a Student’s t–test, where p = 0.05). Epidermal thickness was approximately doubled after test agent treatment when compared to the untreated mouse skin (28.8 ± 3.2 and 14.5 ± 0.8 μm, respectively).

The number of nucleated epidermal cell layers was significantly increased (3.2 ± 0.2 and 2.3 ± 0.1 cell layers, respectively). Epidermal keratin 6 expression; induced in the epidermis following trauma, hyperproliferative stimuli, or tumorigenesis; was also assessed in skin sections from test agent-treated and untreated mouse skin. Keratin 6 is normally expressed in the innermost layer of the outer root sheath of hair follicles but not normally found in the interfollicular epidermis (Wojcik et al., 2000).

As shown in Figure 2, follicular keratin 6 expression was detected following immunofluorescence analysis in untreated and vehicle formulation treated skin sections, serving as an internal positive control. As expected, epidermal keratin 6 was not detected in any sections from untreated mice (Figure 2C). Vehicle formulation-treated skin was also largely keratin 6 negative in the epidermis (Figure 2D). In this experiment, irritation and epidermal keratin 6 expression were less sensitive indicators of topical effects when compared to hyperplasia and inflammation. Our hypothesis predicts that agents causing cutaneous hyperplasia and inflammation in the short-term will be tumor promoters when applied in a long term experiment. According to this hypothesis, the very mild hyperplasia and inflammation detected after treatment with the vehicle formulation predict that this test agent will be produce a very weak tumorigenic response in Tg.AC mice.

Tg.AC Tumorigenic Response to a Novel Test Agent

To test our hypothesis and to examine the predictive ability of the above biomarkers for tumorigenesis, we assessed whether daily dermal administration of the test agent vehicle formulation plus a commonly used topical antibiotic active pharmaceutical ingredient (API) in this vehicle for 26 weeks, would increase the incidence of dermal tumors at the site of application in transgenic Tg.AC mice. The active pharmaceutical ingredient was shown to be non-carcinogenic in a 2-year dermal carcinogenicity study in CD-1 mice. Four groups of 20 Tg.AC transgenic mice were treated daily with 1 of 4 formulations containing active pharmaceutical ingredient (API) in 0×, 1×, 3× and 5× concentrations, respectively.

Additionally, a group consisting of the same number and strain of animals served as a shaved but untreated control group. A dose-dependent increase in papilloma formation was observed (Figure 3). Seven (35%) animals treated with the vehicle formulation (0× API) were positive after 26 weeks. Animals treated at 1×, 3×, and 5× API, showed an incidence of 50%, 65%, and 70%, respectively, by the end of the experiment (Figure 3B). In addition, the average number of papillomas was dose-dependent as well; i.e., 0.5, 0.5, 1.1, and 1.6 for 0×, 1×, 3×, and 5× API respectively (Figure 3A). No papilloma formation was observed in the untreated control mice. The weak tumorigenic activity of the vehicle formulation is consistent with the predicted response from the short-term study and proves our hypothesis correct for at least this agent.

A Comparison of the Sensitivity of Tg.AC, SENCAR, and RasH2 Mice

Information about the relative sensitivity of mouse skin tumorigenesis models to various carcinogens would be useful in the selection of the best model for carcinogen identification. As discussed in previous sections, however, few direct comparisons between the models have been made. With this in mind, further experiments were conducted to (1) assess if early histological and/or biological endpoints in the skin at SOA are also good indicators for skin tumorigenicity in the RasH2 and SENCAR models, and (2) to compare the sensitivity of Tg.AC, RasH2, and SENCAR mice. In order to compare the sensitivity of the RasH2 and SENCAR mouse models to Tg.AC mice, the short-term response of these mice to the test agents was assessed and compared to the response of Tg.AC mice. Groups of mice of each line were topically treated daily for several weeks with the vehicle formulation containing 0× or 1× API.

We determined how topical application affects parameters associated with skin tumorigenesis, such as inflammation, in Tg.AC, RasH2, and SENCAR mice. Cutaneous edema, measured using calipers to determine skin-fold thickness, was used as a surrogate to detect inflammation in the skin. Vehicle (0× API) and 1× API both increased skin-fold thickness in a dose-dependent manner, when compared to the untreated controls of each line (Figure 4).

The increase in skin-fold thickness was significantly different from the untreated control beginning 3 weeks from the start of treatment for both agents in Tg.AC and RasH2 mouse lines, and by 5 weeks in SENCAR mice. Although the extent of edema varied from week to week in all of the lines, the Tg.AC mice generally responded with more edema compared to the other strains. The RasH2 and SENCAR mice responded similarly to each other with lesser edema following treatment. These data reveal increased sensitivity of Tg.AC mice to inflammation caused by this compound when compared to both SENCAR and RasH2 mice.

The induction of epidermal hyperplasia following treatment was assessed in hematoxylin and eosin stained sections of skin removed from the treatment area after 6 (RasH2 and SENCAR) or 8 (Tg.AC) weeks of treatment. As shown in Figure 5, vehicle and 1× API induced a very mild hyperplasia in the epidermis of Tg.AC, RasH2 and SENCAR mice. Epidermal thickness was approximately doubled after 1×API treatment when compared to the untreated mouse skin in all 3 lines (Figure 6B). The number of nucleated epidermal cell layers was also significantly increased in 1× API-treated mice (Figure 6A). The formulation containing 1× API caused a more pronounced, but still mild, epidermal hyperplasia.

Interestingly, untreated RasH2 skin had greater epidermal thickness and number of nucleated cell layers when compared to Tg.AC and SENCAR mice (Fig. 6). Proliferation was also assessed in Tg.AC skin using Ki67 immunofluorescence. As shown in Fig. 7, the number of Ki67-positive keratinocytes was increased after treatment with both agents when compared to the untreated control animals. Although the mean Ki67 labeling was higher in 1× API treated skin compared to vehicle treated skin, the difference was not significant.

Similarly to inflammation and hyperplasia, chronic irritation is frequently associated with, and believed to be mechanistically linked to, skin tumorigenesis. Consistent with our previous results in the Tg.AC only study of the previous section, no irritation was induced by the Vehicle containing 0× and 1× API in Tg.AC mice (Figure 1A and data not shown). Similarly, neither treatment resulted in irritation in RasH2 and SENCAR animals by 6 weeks of treatment (data not shown). Thus, irritation proved a less sensitive marker for cutaneous response to these test agents when compared to inflammation and hyperplasia.

Implications for Model Selection and Early Biomarker Assessment

Our hypothesis predicted that tumorigenic agents cause inflammation, irritation, and/or hyperplasia in the skin of these 3 mouse models. The slight but significant cutaneous changes in the short-term studies were consistent with the very weak tumorigenic activity of the test agent vehicle and 1× API in Tg.AC mouse skin. In addition, our comparison of the effects of 2 agents produced qualitatively similar results in all 3 models. Thus, we predict that the test agents are weak tumor promoters in the RasH2 and SENCAR models as well, although confirmation of this prediction is beyond the scope of this review. Based on our limited studies, we further predict the three models are likely to respond to a given agent through similar mechanisms and with roughly similar tumorigenic responses.

Although further experimentation is clearly needed, our primary data and review of the literature suggest that Tg.AC, SENCAR and RasH2 models are likely to respond to topically applied agents in skin tumorigenesis studies with similar sensitivity. We suggest that even a very mild epidermal hyperplasia and inflammatory response after several weeks of treatment is causally associated with and highly predictive of a tumorigenic response in the mouse skin models. For all three lines, hyperplasia appears to be the most sensitive predictor of a cutaneous response. Therefore, careful analysis of short-term studies can be used to guide decision-making about whether a particular agent is worth the investment in a carcinogenesis study and the best choice of animal model for carcinogenic studies.

SUMMARY

In general, the Tg.AC, SENCAR and RasH2 models presented with a reporter phenotype, papillomas, upon carcinogen treatment and are fairly predictive of carcinogenic potential in humans. The molecular mechanisms of skin tumorigenesis, still under active investigation, are tightly associated with hyperproliferation, epidermal hyperplasia, and inflammation in these models. Our studies demonstrated that there are (1) early endpoints that may be used as predictors of dermal tumorigenesis in mouse skin models; and (2) relatively similar responses in three sensitive mouse models for skin carcinogenesis. We propose that assessment of cutaneous edema and epidermal hyperplasia after several weeks of treatment are robust predictors of a later tumorigenic response in mouse skin studies. Further investigation of the mechanism of induction of skin tumors in the mouse models and comparison between various sensitive strains will better elucidate the value of these models in predicting human carcinogenicity for future new chemical entities.