Critical role of toxicologic pathology in a short-term screen for carcinogenicity.

Carcinogenic potential of chemicals is currently evaluated using a two year bioassay in rodents. Numerous difficulties are known for this assay, most notably, the lack of information regarding detailed dose response and human relevance of any positive findings. A screen for carcinogenic activity has been proposed based on a 90 day screening assay. Chemicals are first evaluated for proliferative activity in various tissues. If negative, lack of carcinogenic activity can be concluded. If positive, additional evaluation for DNA reactivity, immunosuppression, and estrogenic activity are evaluated. If these are negative, additional efforts are made to determine specific modes of action in the animal model, with a detailed evaluation of the potential relevance to humans. Applications of this approach are presented for liver and urinary bladder. Toxicologic pathology is critical for all of these evaluations, including a detailed histopathologic evaluation of the 90 day assay, immunohistochemical analyses for labeling index, and involvement in a detailed mode of action analysis. Additionally, the toxicologic pathologist needs to be involved with molecular evaluations and evaluations of new molecularly developed animal models. The toxicologic pathologist is uniquely qualified to provide the expertise needed for these evaluations.

Introduction

The standard for evaluation of carcinogenic potential for a chemical is the two year bioassay in two rodent species, usually rats and mice[1]. Although there have been some minor refinements during the past four decades, this bioassay has remained essentially the same as that developed in the 1960’s as part of the National Cancer Institute Bioassay Program. Over the years, several difficulties have been identified regarding this assay, including its high cost and the length of time to perform and adequately evaluate it. Several criticisms have been raised including the use of high doses, usually the maximum tolerated dose (MTD) and fractions thereof, and the use of large numbers of animals. In addition, fundamentally the only information that is gained from this assay is whether or not the chemical increases the incidence of some type of tumor or tumors in rats and/or mice. Minimal dose response information is obtained since the doses used in these studies are in the range of only one order of magnitude, not the multiple orders of magnitude that are needed for a better assessment of the dose response, particularly down to doses relevant to human exposure. However, the most significant limitation of the two year bioassay is that it does not generate information about the relevance of the findings to human risk. Initially several chemicals were tested where a strong relationship between the findings in the animal test and human cancer development existed. These generally involved the potent, DNA-reactive (genotoxic) carcinogens such as aromatic amines, polycyclic aromatic hydrocarbons, nitrosamines, and aflatoxin[1]. However, over the past decades, numerous chemicals were tested in which either extremely high doses had to be used to produce tumors since the chemical was non-toxic, or the relevance of the tumors produced by the chemical were shown to be produced by a mechanism that was not relevant to humans. Such examples of non-relevant modes of action include the production of calcium phosphate-containing urinary precipitate by sodium saccharin and other sodium salts leading to the induction of bladder tumors in rats[2], [3], d-limonene and the induction of kidney tumors in male rats by binding to α2u-globulin[3], ethyl acrylate induction of forestomach tumors in rodents[4], and a variety of others. Determining the relevance of tumors in the bioassay to humans had to be determined in follow-up mechanistic research. The proposal is to utilize a short term screen to identify potential pre-neoplastic changes, and then evaluate the mechanistic processes involved to determine both a detailed dose response and relevance to humans. The two year bioassay is not necessary for evaluation of carcinogenic potential.

Carcinogenesis

Any studies performed in animal models, whether the two year bioassay or others, are based on two fundamental assumptions: 1) the effects produced at the doses used in the bioassay will also occur at doses to which humans are exposed (dose extrapolation); and 2) the effect that is produced in rodents will produce a similar effect in humans (species extrapolation)[1], [5]. As was stated by George Box more than five decades ago, “Models: all are wrong, some are useful”[6]. As scientists, it is incumbent on us to evaluate the results of the use of a given model as to how they might extrapolate to humans. Unfortunately, evaluation of the extrapolation to humans is not always investigated.

For cancer and non-cancer endpoints, the US EPA, Health Canada, and International Programme on Chemical Safety (IPCS) have evolved a framework for a transparent, structured, disciplined evaluation of mode of action data in animal studies, followed by a detailed evaluation of the human relevance of the mode of action[7],[8],[9],[10],[11],[12]. It is based on determination of the key events necessary for production of the adverse effect by the chemical in the animal model, and then an evaluation of those key events in humans qualitatively and quantitatively. The criteria for evaluating the various aspects of mode of action utilize a modification of the Bradford Hill criteria used in assessing causality in epidemiology studies[7]. These include temporality, dose response, strength of the evidence, consistency of the findings, and biological plausibility. In addition, alternative modes of action are evaluated and data gaps identified regarding a given mode of action for the induction of an adverse effect by a chemical. This framework continues to evolve, and now is incorporated into the adverse outcome pathway analysis that utilizes molecular biological investigations[13],[14],[15],[16]. It is a fundamental understanding of the key events in the process of producing the adverse effect that is essential for our understanding of the mode of action and for extrapolating between the animal model and the human. The key event is defined as a necessary step in the process, and all of the key events together are sufficient to explain the induction of the adverse effect.

Utilizing this framework can form the basis for a short term screen for carcinogenesis that provides a much more rational approach to this assessment than the current two year bioassay.

To understand the basis for the shorter term approach to this risk assessment process, the fundamental basis for carcinogenesis needs to be understood[17], [18]. It has long been known that cancer is the result of numerous errors in the genome occurring in a single pluripotential, tissue stem cell. Cancer is a clonal disease, so all of the critical genetic errors must occur in a single cell. It is clear that more than one genetic abnormality must be present for cancer to develop, although the precise number is not usually known. Furthermore, these errors are fixed into place during DNA replication. Also, it has long been known that DNA replication, although incredibly precise, does not have 100% replication precision. Rather, rare mistakes occur every time DNA replicates. Thus, a chemical, or for that matter, any agent, can increase the risk of cancer only one of two fundamental ways[17],[18],[19],[20]: 1) it can increase the rate of DNA damage per cell division (DNA reactivity, genotoxicity); or 2) it can increase the number of cell divisions increasing the opportunity for spontaneous errors to occur (non-DNA reactive, increased proliferation) every time DNA replicates. DNA reactive carcinogens at high doses nearly always are toxic, in addition to being DNA reactive. Thus, at high doses, DNA reactive chemicals both increase proliferation and directly damage DNA.

Synergy Between DNA Reactivity and Cell Proliferation

When there is both an increase in DNA reactivity (genotoxicity) and increased cell proliferation, there is a strong synergism. In effect, the increased proliferation provides more numerous targets for the DNA reactive effect and increases the possibility of spontaneous errors. There are numerous examples of this synergy in human cancer risk, including some instances where the increased risk is in one agent, such as cigarette smoking, and in other circumstances the increase in DNA reactivity and increase in DNA replication are induced by different sources[18], [20]. An excellent example which provides a quantitative indication of this synergy is the interaction between aflatoxin (DNA reactive) and hepatitis B virus infection (cytotoxicity with consequent regenerative cell proliferation)[5], [20]. In parts of China where there is increased exposure to aflatoxin without hepatitis B virus infection, the increased risk of hepatocellular carcinoma is approximately three times. In areas where there is no increased aflatoxin exposure but increased hepatitis B virus infection, the increased risk is approximately ten to twelve times. In populations that have both an increased exposure to aflatoxin and have hepatitis B virus infection, the overall risk is approximately sixty-five times, clearly more than additive and more than multiplicative, a true synergy.

In animal experiments, there are numerous examples of interaction between increased cell proliferation and DNA reactivity. One example is the ED01 study involving more than 24,000 mice orally administered acetylaminofluorene (AAF) at relatively low doses, with a detection limit of an increased incidence of 1% instead of the usual 10% in two year bioassays[21]. A number of subgroups were also investigated, but the results clearly showed that the incidence of liver tumors was increased at all doses, even as low as 30 ppm. In contrast, an increase in urinary bladder tumors only occurred at doses of 60 ppm and above, despite a linear dose response for DNA adduct formation in both tissues. In this study, at the doses used, there was no effect of AAF on hepatocellular proliferation, whereas in the urinary bladder, urothelial cell proliferation was increased at doses of 60 ppm and above, the same doses at which there was an increase in detectable tumor incidence. It is apparent that at the lower doses, the presence of DNA adduct formation was not sufficient to increase the incidence of tumors in the urinary bladder to a detectable level (greater than 1%), but that the increase in DNA reactivity combined with increased cell proliferation at the dose of 60 ppm and above produced detectable tumor incidences.

In animal studies utilizing different stimuli for DNA reactivity and increased cell proliferation, an experiment involving the administration of sodium saccharin at 5% of the diet and N-[4-(5-nitro-2-furyl)-2-thiazolyl]-formamide (FANFT) at a dose of 0.005% of the diet, illustrates the interaction[22]. FANFT is a DNA reactive carcinogen specific for the urinary bladder in rats. Sodium saccharin increases cell proliferation and has no effect on DNA reactivity or genotoxicity in general. At these doses, in a standard two year bioassay, there was no increase in tumor incidences, but when they were administered together, there was a detectable incidence of tumors, approximately 30%.

In summary, there are only two fundamental ways by which chemicals can increase the risk of cancer, either DNA reactivity, which increases the rate of the number of errors in the DNA with each replication, or an increase in cell proliferation, which produces an increased number of replications in which spontaneous errors can occur during replication. The two can both occur, which results in synergy.

DNA Reactivity, Immunosuppression and Estrogenic Activity

For the past two decades, there has been an effort to utilize mode of action evaluations for risk assessment of chemicals and then determine if these modes of action have possible human relevance[7],[8],[9],[10],[11],[12],[13],[14],[15],[16], [23]. The criteria for evaluation of mode of action have been clearly delineated. The methods used for extrapolating to human relevance both qualitatively and quantitatively have also been delineated and continue to be refined and evolve.

Overall, there are three types of mode of action that clearly are relevant to human cancer risk[1], [5]. These are DNA reactivity, immunosurveillance, and estrogenic stimulation of cell proliferation. These effects can readily be evaluated in short term screens rather than waiting for a two year bioassay to determine carcinogenicity. DNA reactivity can be evaluated utilizing structure activity relationship (SAR) computer programs, by assessment of mutagenicity in vitro (Ames assay), and more recently by the utilization of in vivo mutagenicity assays[24],[25],[26],[27],[28],[29],[30],[31],[32],[33]. If a chemical is DNA reactive, an evaluation utilizing human cells can be performed to demonstrate the relevance to humans, and evaluate metabolic pathways that are necessary to generate the reactive intermediate that will form DNA adducts. Demonstration of DNA reactivity results in presumptive evidence of carcinogenicity in humans. This is not true for other screening tests for genotoxicity, such as micronucleus formation, the Comet assay, and others, which involve indirect damage to DNA and can be greatly influenced by a number of factors, especially cytotoxicity, that do not portend an increase in cancer risk. Especially in in vitro studies, many of these genotoxicity assays have been shown to produce false positive results which cannot be extrapolated to the in vivo situation in animals, and certainly not to an increase in cancer risk either in animals or in humans.

Immunosuppression is another process that predicts an increased risk of carcinogenesis in humans, and this can be readily evaluated in a variety of both in vitro and in vivo assays[34]. Hematologic effects and effects on lymphoid organs such as lymph nodes, thymus, spleen, and bone marrow can be demonstrated morphologically in a 90 day study. More specific assays can also be used to determine certain aspects of the immune response.

Individuals that are immunosuppressed are at greatly increased risk of developing cancer[1], [20], [35]. This includes individuals that are born with inherited immunosuppressive disorders, transplant recipients treated with immunosuppressive therapy, patients with a variety of neoplastic and non-neoplastic disorders treated with drugs that are immunosuppressive, and individuals that have AIDS. However, the increased risk of tumors is not uniform for all types of tumors. In reality, the increased risk of cancer in immunosuppressed individuals is related to those malignancies that are produced by infectious diseases, such as B cell lymphoma (EBV), squamous cell carcinoma (HPV), hepatoma (HBV and HCV), and Kaposi’s sarcoma (HHV8, also known as Kaposi’s sarcoma virus). Thus, the immunosurveillance is for the infectious diseases, not the neoplasms themselves. The infectious diseases can lead to the production of certain malignancies. With respect to human risk, if a chemical is immunosuppressive at levels to which humans are exposed, it will lead to an increased risk of developing these infectious disease-related malignancies.

The relationship of estrogen to the production of certain malignancies, especially breast and endometrium, is well-documented [20], [36], [37]. Other tumors that might be related to estrogen include liver and ovary, although the relationship is not as strong as for breast and endometrium. Screening for an estrogenic effect can be accomplished in a variety of in vitro and in vivo assays, including estrogen receptor assays and bioassays which include examination of estrogen-dependent tissues such as ovary, uterus, cervix, vagina, and breast[1], [20], [36], [37]. Examination of these tissues is included in routine 90-day studies and could also be included in shorter term studies. Substances with estrogenic activity at doses to which humans are exposed, are likely to pose an increased risk for development of estrogen-related malignancies. The dose and potency of the estrogenic activity are particularly critical in evaluating human risk, but a two year bioassay is not necessary to determine these factors.

Short Term Screen for Carcinogenicity

Currently the standard process for evaluating substances for carcinogenic activity is to perform a dose range finding study, usually 90 days, in the test species, followed by a two year bioassay. If increased incidences of tumors occur in the two year bioassay, further investigative work is then conducted to determine the mode of action and whether the mode of action is relevant to humans, and to evaluate a more detailed dose response of the key events in the mode of action. Given the vast amount of knowledge that has been gained over the past five decades from the performance of these bioassays and the investigation of the modes of action of a wide variety of chemicals that have been tested, we now know the modes of action in animal models for most tumor types, and furthermore, we have a reasonable basis to judge which of these modes of action are relevant to humans. The proposal was first put forward nearly fifteen years ago to take advantage of this vast knowledge and use it in a more predictive and rational way to evaluate chemicals in short term screening assays instead of in a two year bioassay[1], [5]. The short term screening assays are focused on mode of action and human relevance. What really is needed is a determination as to whether the chemical poses a cancer risk to humans. not an investigation to determine if a chemical poses a cancer hazard in rats or mice. We will present how this process can be applied to an evaluation of liver carcinogenesis and urinary bladder carcinogenesis, and then present an overall approach to the predictive evaluation of cancer risk from chemicals. The role of toxicologic pathology is both essential and noteworthy.

Liver Carcinogenesis

Production of hepatocellular carcinomas (hepatomas) in rodent species was one of the first tumor types to be produced experimentally[5]. This was accomplished by Yoshida in 1932 and involved the administration of an azo dye for production of liver cancer. His seminal observation was followed by further investigations by Kinosita, published in 1937. Extensive investigations on azo dyes and other substances over the past eight decades have provided a good understanding of how chemicals produce liver cancer in rodents[5], [38]. Furthermore, we have also developed a basic understanding of what produces hepatocellular cancer in humans[39], [40]. For liver carcinogenesis, the modes of action can be divided into those that are DNA reactive and those that are non-DNA reactive as described above[5]. Non-DNA reactive chemicals that are carcinogenic for the liver can be divided into those that act through specific receptors versus those that do not. Furthermore, as for all non-DNA reactive modes of action, the increase in cell proliferation can either be due to an increase in cell births or a decrease in cell deaths, which results in an accumulation of more cells. Increased cell births can be produced by cytotoxicity and regeneration or by direct mitogenesis. Decreased cell deaths can be produced by either blocking apoptosis or blocking differentiation (a cell death process). Listed in Table 1 are the various modes of action that have been specifically identified in rodents for hepatocellular carcinogenesis. Examples are known for all of these, but we describe in detail a few of these to illustrate issues that arise with regard to mode of action and extrapolation to human relevance.

Table 1.

Modes of Action for Hepatocellular Carcinogenesis

Chloroform has been demonstrated repeatedly to produce increased incidences of tumors of the liver and kidney in rats and mice[5], [9], [38], [41], [42]. The mode of action has been delineated and involves metabolic activation to reactive products (phosgene) by cytochrome CYP2E1. This leads to the induction of cytotoxicity in hepatocytes and renal tubule cells, with consequent regenerative proliferation. If this continues for prolonged periods of time, it ultimately leads to an increased incidence of tumors of the liver and kidney.

Exposure to chloroform also occurs in humans. Several decades ago, chloroform was used at relatively high doses as an anesthetic. Occasionally, because controlling exposure was difficult, some patients developed toxicity to the anesthetic chloroform. The toxicity involved liver, and infrequently kidney, similar to what is seen in rats and mice. Furthermore, metabolic activation of chloroform to phosgene occurs in humans by the same cytochrome as in rodents. Thus, there is the potential for chloroform to theoretically increase the incidence of liver or kidney tumors in humans, since the precursor key events are known to occur in humans. However, there are two aspects that preclude the development of tumors in humans that need to be taken in to account in a risk assessment. First, in humans, the exposure is limited, especially if there is any evidence of toxicity, and second, chloroform is no longer used as an anesthetic since there are many better drugs available for this purpose. In the rodent, extensive investigations have demonstrated that the tumorigenic effect only occurs when the dose is sufficient to produce cytotoxicity in the target tissues, liver and kidney. Thus, there is clear evidence for a non-linear, threshold dose response. If the exposure is too low to produce cytotoxicity, there is no increased risk for development of tumors. Humans today are exposed to chloroform at extremely low levels as a by-product of chlorination of drinking water. The exposure in humans is several orders of magnitude below that which is known to produce toxicity in humans. Therefore, based on both time of exposure and dose of exposure, humans are not exposed to sufficient amounts of chloroform in the drinking water or other environmental sources to produce sustained cytotoxicity. Since there is no risk for increased cell proliferation, there is no increased cancer risk in humans.

Of great importance in investigating liver and kidney tumors produced by chloroform is a detailed understanding of the metabolism and identification of the cytotoxicity and the consequent increased proliferation that occurs in response to the treatment with chloroform. This can be performed in short term studies and involves comprehensive investigations of histopathology, serum enzyme markers for liver toxicity, and serum markers for kidney toxicity. Urinary markers of kidney toxicity also could be evaluated. Very sensitive and rapid techniques are now available to measure the markers for cytotoxicity in both of these tissues. Evidence of increased proliferation can be determined by observation of hyperplasia by routine histopathology or utilization of markers for cell proliferation such as bromodeoxyuridine (BrdU)[43], [44], proliferating cell nuclear antigen (PCNA)[45], and Ki-67[46]. Advantages and disadvantages of these labeling indices for evaluating cell proliferation have previously been described in detail[47], [48].

Another example of chemicals shown to be liver carcinogens in rodents are those known to induce cytochrome P450’s, such as phenobarbital and pyrethroid insecticides[49], [50]. These chemicals are non-genotoxic. Following extensive investigations over the past 40 years, the mode of action as to how these chemicals produce liver tumors in rodents has been delineated and includes activation of the CAR receptor and induction of P450 isozymes, with consequent increased hepatocellular proliferation and ultimately tumor formation[5], [38], [51],[52],[53],[54]. There are several associated events in the process, including centrilobular hypertrophy and increased liver weight. Humans have the same receptor (CAR) as the rodent, although there appears to be fewer receptors per cell in the human than in the rodent. In addition, activation of the CAR receptor in humans leads to induction of the same P450 isozymes as in the rodent. However, in mice and rats, CAR activation results in an increase in hepatocellular proliferation which is a required key event for the production of tumors. In human hepatocytes, there is no proliferative response. This has been demonstrated in vitro in studies involving human liver cells[55], and also in vivo utilizing chimeric mice with transplanted human hepatocytes[56]. The human hepatocytes respond to CAR activators such as phenobarbital or permethrin with the expected metabolic responses including induction of P450 isozymes, but there is no increase in human hepatocyte proliferation. Since increased proliferation is a required key event in the mode of action for the production of liver tumors, it is clear that liver tumors will not be produced by this mode of action in humans. Additional evidence supporting a lack of effect in humans by CAR activators are extensive epidemiology investigations of phenobarbital and pyrethroid insecticides[50], [57], [58]. This is particularly relevant for phenobarbital, since patients receive phenobarbital as treatment for epilepsy for many years. The dose administered to humans is similar to the dose to which the rodents are exposed, and the epidemiologic evidence consistently demonstrates the lack of a carcinogenic effect of these agents.

Statins are a third example of a chemical that produces high incidences of liver tumors in rats and mice in the standard two year bioassay[59], [60]. However, the metabolic response to statins is different in rodents than it is in humans. Subsequent extensive epidemiologic investigations involving hundreds of thousands of patients on statins clearly demonstrated that there is no increased or decreased risk of liver cancer or any other types of cancer in patients receiving statins[61],[62],[63],[64],[65].

To screen for potential carcinogens in rodent species utilizing a short term assay, there must be evidence that the short term screening assay will provide markers for detection of potential carcinogens. Several years ago, investigators at the National Toxicology Program reviewed the tumor bioassays of more than 500 chemicals and correlated the eventual appearance of liver tumors in rats and/or mice with findings in the 90 day dose range finding study[66]. They demonstrated that all of the potential liver carcinogens showed one or more of the following markers in the 90 day study: hepatocellular necrosis, hepatocellular hypertrophy, hepatocellular cytomegaly, or increased liver weight. Obviously, these are non-specific, and many chemicals which produced these effects, particularly increased liver weight, in a 90 day study did not produce liver tumors in the full two year bioassay. However, the importance for a short term screening assay is that there be no false negatives, and that was found to be the case. All eventual hepatocellular carcinogens produced one or more of these effects in the short term assay.

The proposal is to utilize short term screens to detect potential carcinogens, in this case for the liver, and then to do more detailed mode of action analyses and dose response investigations to provide a more rational risk assessment for humans[5]. In Table 1, the modes of action are identified for hepatocellular carcinogenesis, but highlighted in bold print are those that are actually relevant to humans. Genotoxicity, a known mode of action for certain liver carcinogens such as aflatoxin, is relevant to humans. In addition, it is known that estrogen is related to the rare development of hepatocellular carcinomas in humans, evolving in those circumstances from adenomas. In humans, the most common modes of action for induction of hepatocellular carcinomas involve cytotoxicity, inflammation, and regenerative hepatocellular proliferation, whether due to inherited disorders, iron overload, or viral infection. The modes of action that are not highlighted in this table, such as CAR activation, PPARα activation, or exposure to statins, are not relevant to humans.

Based on our knowledge of hepatocellular carcinogenesis in rodents and in humans, a short term screen is readily achievable (Fig. 1). First, the chemical would be administered for 90 days, with evaluation of the liver for the four markers identified by the NTP (see above) in their screen for rodent hepatocarcinogens. If negative, we can be assured that there is no potential risk for development of liver cancer in humans. If it is positive, then evaluation of the mode of action is required to determine whether the chemical produces the effects on the liver by a mode of action that is relevant to human carcinogenesis. Methods that can be used for mode of action analysis are indicated in Fig. 1.

Fig. 1.

Proposed sequence of evaluation to screen for potential hepatocarcinogens. The initial screen involves the four criteria proposed by Allen et al.[66]. If one or more of these four signals are detected in the 13 wk. screening assay, follow-up mechanistic evaluations are performed to determine the MOA(s) and provide the basis for an assessment of human cancer risk. Modified from Cohen, SM, Toxicol Pathol, 38: 487–501, 2010[5], with permission from Toxicologic Pathology.

Keep in mind that it is critically important in these short term screenings that they be used to identify potential toxicities that could occur in humans, not simply as dose range finding studies for longer term bioassays. Liver toxicity is a common problem with various pharmaceuticals in development, so it is critical in these animal studies that potential for liver toxicity in general be identified. This is irrespective of any potential for carcinogenesis. Thus, the short term screening is critical not only for identifying potential carcinogens, but more importantly, identifying potential toxicities in humans. If the mode of action is determined to be relevant to humans, a more detailed dose response can then be investigated to determine whether the exposures in humans put them at risk for development of tumors. If the mode of action is not relevant to humans, such as for CAR activators, it is sufficient to conclude that these chemicals will not be carcinogenic for the liver in humans. In addition, if the dose response for a chemical such as chloroform shows that the exposure to humans is significantly below that which will produce the effect on the liver that is necessary for the eventual development of hepatocellular tumors, it can be concluded with confidence that there is no increased risk of carcinogenesis in humans.

All of this information can provide a more rational, mode of action-based risk assessment than is provided by proceeding to the two year bioassay. For such a proposal to actually be implemented, however, requires not only acceptance by regulatory authorities of the science behind the proposal, but would also require extensive changes in the processes by which chemicals are labeled, particularly in the pharmaceutical industry. A positive finding for one of the short term screening markers cannot be used as the basis for listing on the label that it is a potential carcinogen. If it is not relevant to humans either based on mode of action or on dose, the chemical should not be labeled as a potential carcinogen to the public.

Urinary Bladder Carcinogenesis

A similar approach can be utilized for the urinary bladder. Urinary bladder carcinogenesis has been extensively investigated for the past eight decades, but traces a lineage of scientific investigations back to the observation by Rehn in 1895 when he observed that individuals working in the aniline dye industry had a markedly increased risk of developing bladder cancer[67],[68],[69]. It was eventually demonstrated that the increased risk was due to exposure to aromatic amines, which are metabolically activated by N-hydroxylation and subsequent esterification to reactive electrophiles that form DNA adducts and mutation. Several specific chemicals have been identified as human bladder carcinogens, including various aromatic amines, phosphoramide mustards, and arsenic[67],[68],[69],[70]. Arsenic, in contrast to the others, is a non-genotoxic carcinogen which produces bladder and other types of cancer in humans by a process involving cytotoxicity and regenerative proliferation[70].

Screening for potential human bladder carcinogens would involve a similar process (Table 2) as was just illustrated for the liver. Markers for potential urothelial carcinogenesis in rats or mice include evidence of increased cell proliferation[67]. It is usually adequate to use routine light microscopic examination of hematoxylin and eosin stained slides for the presence of hyperplasia. However, a more sensitive evaluation would be determination of a proliferation labeling index, such as BrdU, Ki-67, or PCNA. All potential urothelial carcinogens in rodents (and in humans) show evidence of increased proliferation in a standard 90 day study, whether the ultimate mode of action is due to DNA reactivity (e.g. aromatic amines) or due to increased cell proliferation without genotoxicity (e.g. inorganic arsenic). Again, we will cite some examples to illustrate the process.

Table 2.

Modes of Action for Urothelial Carcinogenesis

Sodium saccharin produces an increased incidence of urinary bladder tumors in rats, particularly males, when administered beginning before weaning and continuing for the full two year bioassay[2], [6], [7], [71], [72]. Administration beginning after weaning does not produce an incidence of tumors adequate for detection in a standard bioassay involving 50–60 animals per group. However, in the so-called two generation study, the incidence is increased sufficiently to be detected in such studies. The dose required to produce any effects by administration of sodium saccharin in the diet are enormous, ≥ 25,000 ppm. Nevertheless, hyperplasia is present in the urothelium after 4 to 13 weeks of administration, even if it begins after weaning, but most importantly, it turns out that the mode of action is not relevant to humans. The mode of action involves administration of the sodium salt at high doses in the diet which produces an alteration of the urinary milieu that results in formation of calcium phosphate-containing precipitate. This precipitate acts as a cytotoxic agent for the urothelium with consequent regenerative proliferation and ultimately the development of tumors. Various aspects of this mode of action have been extensively investigated and demonstrated in several laboratories. Since the mode of action was clearly not relevant to humans, the International Agency for Research in Cancer (IARC) downgraded its classification of saccharin from 2B to 3 in 1999[2], [71], and the National Toxicology Program removed it from its List of Carcinogens in 2000[72].

Sodium saccharin-induced urinary bladder carcinogenesis is specific for the rat [2], [3], [67], [71], [72]. When administered to mice at high doses, even up to 100,000 ppm of the diet, sodium saccharin produced no effect on the urothelium. The reason for this rat specificity is because mice do not form the precipitate in response to administration of sodium saccharin. In the mouse, the urinary concentrations of calcium, phosphate, and magnesium are approximately 5–10 times lower than in the rat, and are inadequate for formation of the precipitate. Without the precipitate there is no cytotoxicity, no regenerative proliferation, and no formation of tumors. In humans, there is adequate urinary calcium concentration, similar to the rat, but other components of the urine that are necessary for the formation of the precipitate do not occur in humans, including protein concentration, overall density (osmolality), and possibly other aspects such as citrate concentration (which chelates calcium and magnesium).

In addition, it became apparent based on extensive investigations from the laboratories of Dr. Nobuyuki Ito and Shoji Fukushima, that it was not the saccharin itself that was producing the effect but sodium salts in general[2], [3], [67], [72]. It was demonstrated that the sodium salts of ascorbate, bicarbonate, glutamate, citrate, chloride, and others produce similar effects in the rat. For these substances, the assumption of species extrapolation and dose extrapolation are both incorrect. It is a high dose phenomenon only, and it is only in the rat.

A related example of urinary bladder carcinogenesis in rodents involves formation of calculi. Calculi can be produced by a variety of substances, such as melamine, uracil, and many others[3], [9], [73]. Clearly, the effect only occurs at high doses, doses sufficiently high to produce precipitation of the material in the urine to form calculi. Calculi, in contrast to the calcium phosphate-containing precipitate produced by sodium salts, produce much greater urothelial damage and much higher incidences of tumors. The effect is a high dose phenomenon only. However, a number of substances that produce calculi in rats and mice resulting in an increased incidence of urinary bladder tumors can also produce calculi in humans[67], [73], [74]. This was clearly demonstrated in China where infants ingested formula contaminated with melamine[75]. Several infants developed urinary tract calculi and six died. Clearly toxicity can occur, so the mode of action for calculi-induced urinary bladder tumors is potentially relevant to humans. However, there are several bases to conclude that the potential for carcinogenesis in humans is significantly lower than in rodents and probably non-existent[3], [9], [73]. This is predominantly due to the anatomic differences between humans and rodents. In rodents, calculi can accumulate in the bladder and remain for the lifetime of the animal. In humans, calculi will produce obstruction, which produces pain and usually hematuria, either of which will lead the affected individual to a physician for removal of the stone. Thus, although stones can form in humans, they will be present for only a brief period of time. There are some individuals that can carry calculi for longer periods of time such as individuals with urinary bladder diverticuli or patients with neurogenic bladder. However, when there are long standing calculi in these individuals, there is also bacterial infection. Bacterial cystitis is a known risk factor for the development of bladder tumors, so it is unlikely that the calculi pose any additional risk.

An example of chemically induced urinary bladder cytotoxicity is provided by the natural substance pulegone[76]. Pulegone produced a slightly increased risk of bladder tumors in female rats in the NTP two year bioassay. In short term studies, it produced an increase in hyperplasia and labeling index, due to cytotoxicity of the urothelium with regenerative proliferation. There is no formation of urinary solids, so the toxicity must be produced by the chemical and/or one of its metabolites. The metabolism of pulegone is well-known, and the major urinary metabolites are piperitone and piperitenone. The hepatotoxic metabolite menthofuran also is excreted in the urine but only to a limited extent. By evaluating the urinary concentration of metabolites of the animals administered pulegone and comparing that to the IC50s of the chemicals determined in in vitro investigations with urothelial cell lines, it was shown (Table 3) that pipertenone is likely the major contributor to the cytotoxicity, although piperitone and pulegone itself may also contribute to the cytotoxicity. Menthofuran was present at too low of a concentration to produce cytotoxicity, and menthone was not detected in the urine. The mode of action for pulegone-induced urothelial tumors includes metabolism of pulegone, with excretion and concentration of the metabolites in the urine, which produce urothelial cytotoxicity, consequent regenerative proliferation, and ultimately a low incidence of urothelial tumors. Similar to chloroform for liver and kidney, the dose response requires sufficient chemical to be administered to produce a concentration of metabolites in the urine that will produce cytotoxicity. This does not occur in humans, since the exposure required to produce urinary levels of cytotoxic metabolites are substantially higher than what humans could tolerate.

Table 3.

Comparison of In Vivo Urinary Concentrations after Pulegone Administration and In Vitro Cytotoxicity

Similar to what was described above for the liver, a process could be outlined for evaluating the mode of action and potential relevance of rodent bladder carcinogens to humans (Fig. 2). To begin with, the 90 day screen would involve detection of increased proliferation by routine analysis of hematoxylin and eosin stained slides and by the use of a more sensitive proliferation marker such as BrdU, Ki-67, or PCNA labeling index. Similar to the liver, if one of these markers is not present in the 90 day study, there will be no development of bladder tumors in the two year bioassay. If one of these markers is positive, an investigation of the mode of action and dose response can be performed. The modes of action for urinary bladder carcinogens are listed in Table 2 and include those known for urinary bladder carcinogenesis in humans. Evaluations that can be used to investigate the mode of action and human relevance are urinalysis for solids, culture for infection, and light and scanning electron microscopy for evidence of cytotoxicity. There is clear evidence for DNA reactive carcinogenesis, such as by aromatic amines, and cytotoxicity with consequent regenerative proliferation induced carcinogenesis, such as by Schistosomiasis or inorganic arsenic, as being relevant in humans[68], [69]. Whether or not direct mitogenesis occurs in the bladder is yet to be demonstrated, although nicotine is a possibility in both rodents and humans[77].

Fig. 2.

Proposed sequence of evaluation to screen for potential urothelial carcinogens. The initial screen involves evaluation for hyperplasia or increased labeling index for cell proliferation. If positive, DNA reactivity is evaluated. If that is positive, evaluation for metabolism in humans is assessed. If not DNA reactive, evaluation of the urothelium by light microscopy and possibly scanning electron microscopy is performed. If cytotoxicity, an evaluation of urine for solids is performed along with an assessment of potential cytotoxicity of the chemical and metabolites and corresponding urinary concentrations after administration of carcinogenic dose. If no cytotoxicity, mitogenic activity is presumed.

Short Term Screen for Carcinogenic Potential

Although a detailed description of the process for the liver and urinary bladder is presented, a similar process could be performed for other tissues[1], [5]. However, other types of screening assays may be required for evaluations in other tissues. Most importantly, it is critical to remember that the primary purpose of the 90 day study is not to evaluate carcinogenesis, it is to evaluate potential toxicity, irrespective of the potential for carcinogenicity.

For some tissues, the rodent model can be a reasonable surrogate for potential toxicity and carcinogenicity in humans, such as for the liver and urinary bladder, and also for the kidney, glandular stomach and large intestine. However, there are several tissues in rodents that are not predictive of carcinogenicity in humans[1], [5]. There are tissues present in rodents that do not occur in humans, including Zymbal’s gland, the Harderian gland, and forestomach. Evaluation of these tissues for potential carcinogenicity is irrelevant to humans. Other tissues can show evidence of toxicity, but are not predictive of carcinogenicity in humans. These include the endocrine organs (thyroid, adrenal cortex, adrenal medulla, anterior pituitary, posterior pituitary, parathyroid, gastrointestinal endocrine cells, pancreatic islet cells) and also the different organs sensitive to endocrine stimulation (ovary, testes, endometrium, prostate). Detection of toxicity in these tissues is readily apparent in the 90 day study based on histopathology, and if there are specific concerns, detailed biochemical evaluations can be performed. Nevertheless, these will not provide information regarding potential risk to humans regarding carcinogenesis.

Based on the above discussion, an overall proposal for utilizing the 90 day screening assay for evaluating carcinogenic risk to humans in illustrated in Fig. 3[1], [5]. If a substance is DNA reactive, then evaluation of the metabolism in humans and potential for forming DNA adducts can be readily evaluated. If DNA adduct formation does occur, one must presume that the substance is going to be carcinogenic in humans. If the substance is not DNA reactive, the next step in the process is to evaluate the substance for immunosuppressive activity and estrogenic activity. This can be readily incorporated into the 90 day bioassay utilizing various receptor assays and immunologic assays. If the substance has either immunosuppressive or estrogenic activity, potential human carcinogenic risk must be assumed, and a dose response and a specific risk assessment will need to be performed with regard to cancer risk. Benefit and risk needs to be weighed, such as in the use of pharmaceuticals for specific disease treatments. If the substance is not DNA reactive, not immunosuppressive, and does not have estrogenic activity that will produce a biologic response in vivo at exposure levels for humans, the next step in the process is evaluation of the potential for inducing a proliferative response in the various tissues. Screens for this can be developed for all of the tissues that need to be evaluated, taking into consideration the various discussions above. If the substance shows no evidence of producing an increased proliferative response in any tissue in the 90 day study, it can be presumed that the substance does not carry a cancer risk for humans. If the substance does produce a proliferative response, evaluation of the mode of action for producing this response and evaluation of the dose response are then investigated to determine human relevance of the mode of action and the dose. Based on these evaluations, one can develop a strong, rationally based risk assessment for the potential of development of cancer in humans, without any need for performing a two year bioassay. In the next several years, various molecular analyses are likely to be developed that can aid in this screening process and in identifying mode of action[78],[79],[80],[81],[82],[83],[84],[85]. These could supplement or replace current procedures.

Fig. 3.

A proposed guide for evaluating the potential carcinogenicity of chemicals. Each box poses an evaluation to be performed. If the sequence results ultimately in a No that is in a circle, there is no (or negligible) carcinogenic risk in humans. If the sequence results ultimately in a Yes that is in a triangle, it poses a presumptive human carcinogenic risk. From Cohen, SM, Toxicol Sci, 80: 225–229, 2004[1], with permission from Toxicological Sciences.

Role of Toxicologic Pathology

As is evident in the above discussion, there is a critical role for toxicologic pathology in these assessments, both with respect to identifying potential toxicities and identifying potential carcinogenic risk. The toxicologic pathologist is involved not only with an evaluation for overall toxicity, but with an evaluation of potential immunologic and endocrine effects and the proliferative effects that are expected for a potential carcinogen. The pathologist would obviously be involved in the detailed investigations into dose response and mode of action analysis. Utilization of routine histopathology is only one part of the overall assessment that can be provided by the toxicologic pathologist. In addition, there are a wide variety of tools available that can be utilized to investigate mode of action, such as immunohistochemistry and molecular analyses. Collaboration with other investigators with expertise in biochemistry, metabolism, and molecular biology is essential. In addition, a pathologist is required for active participation in the design of studies and for selection of tissues to be utilized for these additional investigations.

A toxicologic pathologist also should be actively involved in basic research investigations. During the past two decades there has been an explosion of development of a variety of knock-out and knock-in animal models for investigation of various basic biologic mechanisms and toxicologic investigations. It is critical that pathologists be involved in the evaluation of the phenotype of these models, since individuals without that expertise can frequently misinterpret the morphologic findings. Examples of such misinterpretations are rampant in the literature. As an example, in an investigation to evaluate the potential for troglitazone to produce intestinal tumors in mice, the investigators interpreted the lesions as invasive intestinal adenocarcinomas. However, careful evaluation of the photographs provided in the publication clearly demonstrated that the lesions were actually ulcers with regenerative proliferation, not malignancies[86]. We implore pathologists to become more actively involved in these basic investigations in addition to their role in overall toxicology programs.

In summary, the toxicologic pathologist is critical for the evaluation of potential carcinogenicity and overall toxicity of chemicals. This involves performance of short term assays (≤ 90 days) for toxicity and also provides information for the potential for carcinogenicity. The pathologist needs to be involved with mechanistic evaluation of the toxic endpoints to determine the human relevance and the possible dose response if mode of action for the substance’s toxicity is relevant to humans. The pathologist is critical for the evaluation of potential toxicity. In addition, evaluation for the rates of apoptosis and cell proliferation utilizing immunohistochemical assays are particularly useful for both screening purposes and for assessment of mode of action. The pathologist is critical for the selection of specific tissues for molecular analyses, and is the individual best trained for the utilization of immunohistochemical and in situ hybridization markers and for the evaluation of these markers in tissues. Identifying the specific cell types involved is essential. In addition, the pathologist should be involved in the selection of tissues for molecular analyses including microdissection.

Toxicologic pathology has continued to expand its armamentarium of tools that are available for investigations in rodents, but also for the important scientific basis for eventually determining the relevance of the mode of action to humans and a detailed dose response.