Allergic contact sensitizing chemicals as environmental carcinogens.

Chemicals that were bioassayed by the National Toxicology Program (NTP) and that also produce allergic dermatitis (ACD) in humans were evaluated for their tumorigenic characteristics. The impetus for the study was that most contact sensitizers, i.e., those that produce ACD, and genotoxic carcinogens are chemically similar in that they are electrophilic, thereby producing adducts on macromolecules including protein and DNA. This similarity in chemical behavior suggests that many contact sensitizers might be environmental carcinogens. All of the published NTP bioassays by early 1996 that had both genotoxicity and carcinogenicity studies were included in this analysis. The NTP chemicals had been chosen for bioassay without regard to their ability to produce ACD. Of the 209 chemicals that were bioassayed, there were 36 (17%) that were known to be human contact sensitizers; about half of these were positive on tumor bioassays. The contact sensitizers differed from the NTP sample as a whole by having a proportionately larger number of nongenotoxic chemicals by the Ames Salmonella assay, presumably because more of them were selected on the basis of widespread usage rather than structural resemblance to known carcinogens. Compared to the nongenotoxic chemicals, the genotoxics were stronger carcinogens in that they had a higher incidence of positive tumor bioassays, with twice the number of organs in which tumors were induced. The nongenotoxic chemicals had a preference for tumor induction in parenchymal tissues in contrast to epithelial tissues. The contact sensitizers showed essentially the same characteristics as the whole NTP sample when stratified according to genotoxicity. Judging by the chemicals that were chosen primarily for their widespread use rather than for their structural resemblance to carcinogens, the addition of a test for contact sensitization to the Ames test as a screening tool would increase the tumorigenic detection efficiency by about 40% because of the nongenotoxic tumorigens. A ballpark estimate suggests that there could be several thousand contact sensitizers for humans in commercial use that are rodent tumorigens.

.T..ox P ailaerac demtts(ACD)inhumns= weeeautdfrterrm gnccaatrs ICsThimtus for the S*tu Wa that' most contat se thj a prduce D,adenni cinogens are chemyll similr i that t t ding a C o mr molecules including protein and DNA. This ity n ceicl beihior t tt many contact sensitizers might be entcarcinogen A of the p li NW bioasa by early 1996 dtat had both genocity and cacinogenicit studies e included in alysis. The N chemicals had been chmen for bioass witout rd thdir ability ti produceACD. Ofthe 209 chemicals tdt were bioassayed, the were 36.(17%) thatw knwn t ihma t sensitizers; about half of these were positive on tumor bioassays. The  There is a possibility that many of the hundreds of chemicals that produce allergic contact dermatitis (ACD) in humans might be environmental carcinogens (1). Most contact sensitizers are electrophiles (2). They are Table 1 Gavage form on DNA as well as on protein, resulting in genetic damage and the initiation of the carcinogenic process. Genotoxic carcinogens have been shown to be contact sensitizers in the skin of the guinea pig (4) and in the mouse ear (5). A number of chemicals known to be contact sensitizers for humans (6) are found in the NIH list of carcinogens (7 these include nickel, chromium, benzidine, beryllium, cadmium, coal tar, epichlorohydrin, butyl hydroxyanisole, DDT, p-dichlorobenzene, dimethylaminoazobenzene, formaldehyde, hydrazine, lindane, 4,4'-methylenedianaline, thiourea, and toluene 2,4-diisocyanate. Hence, carcinogens can be contact sensitizers and contact sensitizers can be carcinogens. While contact sensitization is a skin response, it is clear from the above examples that cancer induction by agents that are contact sensitizers can occur internally by a variety of exposure routes not involving the skin. Given the possibility that contact sensitization might be a practical means of identifying previously unsuspected environmental carcinogens, we wondered what pattern of tumorigenic responses would be obtained if a sample of the many hundreds of known human contact sensitizers were subjected to conventional genotoxicity and rodent cancer bioassays. As an indirect approach to the question, we examined the NTP carcinogenesis and genotoxicity bioassay series for the presence of contact sensitizers and their tumorigenic effects.
The NTP series encompasses hundreds of chemicals that have been subjected to highly standardized and carefully monitored bioassays. The chemicals were recommended for study mainly by federal health and environmental agencies without regard to their ability to produce ACD. We examined several hundred NTP bioassays and found that they included a number of chemicals, both genotoxic and nongenotoxic, that are known to be human contact sensitizers. The essential finding was that about half of the contact sensitizers were positive in the cancer bioassay, supporting the possibility that contact sensitizers might be an important group of environmental carcinogens.

Results
The NTP technical reports were the only source of bioassay data used in this analysis, and only those chemicals that had both genotoxicity and carcinogenesis bioassays were used. The lists of chemicals were provided by the NTP on the basis of what had been completed and published as technical reports by May 1996; the chemicals were stratified according to the results of the Ames Salmonella assay (SAL) for mutagenicity and the Chinese hamster ovary cell (CHO) assay for chromosomal damage. The Ames assay was selected because it is regarded as the most predictive index of carcinogenicity (8).
There were other tests of mutagenicity, e.g., on mammalian cells, that were not used consistently in the bioassay series and were not included in this analysis. The CHO assay was chosen on the possibility that contact sensitizers, by being interactive with protein, might cause genotoxicity by indirect mechanisms. Some of the chemicals had more repetitions of given genotoxicity tests than others. Chemicals were identified as contact sensitizers by the use of a list of 384 such agents in one dermatology text (6) and confirmed by another (9), with the addition of parathion. There is no formally recognized registry of contact sensitizers with standardized inclusion criteria.
Tables 1-4 list the chemicals according to the four combinations of the two genotoxicity indices: SAL+,CHO+; SAL+,CHO-; SAL-,CHO+; and SAL-,CHO-. These tables indicate which chemicals are contact sensitizers, the rationale for their being selected for bioassay, and the route of exposure. About 90% of the routes of exposure were via the gastrointestinal tract in the total NTP sample: feed (56%), gavage (33%), and drinking water (2%). The other routes of exposure included intraperitoneal injection (1%), dermal application (2%), and inhalation (6%). There were no differences in the routes of exposure according to genotoxicity or contact sensitization. Tumorigenicity is also indicated in Tables 1-4. Each chemical was effectively bioassayed four times with 50 animals each, i.e., in male and female rats and mice. We analyzed the data on the basis of the combined number of organs in the four test groups that had a statistically significant increase in tumor formation; the different tumor types within any given organ were lumped together, e.g., bile duct and hepatocellular tumors in the liver. The tumor Environmental Health Perspectives * Volume 105, Number 9, September 1997 Feed Abbreviations: SAL, Ames Salmonella mutagenesis assay; CHO, Chinese hamster ovary cell assay for chromosomal abnormalities; CAS, Chemical Abstracts Registry number; CS, contact sensitization. aRFB-primary rationale for the bioassay as follows: 1, widespread use; 2, structural relation with known carcinogen or mutagen; 3, inadequate previous study; 4, representative of a structural class; 5, drug with prolonged exposure; 6, potential use as an antineoplastic drug; 7, preexisting evidence of carcinogenicity; and 8, potential warfare agent. "Tumor score is the number of different organs that showed tumor induction in mice and rats of both sexes; parentheses indicate the number of tumor sites that were obtained only on a single species; se-tumor responses indicate those that were classified as some evidence in contrast to the stronger designation of clear evidence.
response to each chemical in Tables 1-4 was the sum of the number of organs that had statistically significant tumor induction across all four test groups; this is called the tumor score. Thus, if the liver were the only organ with statistically significant tumor induction by a given chemical, and this occurred in male and female rats and mice, the tumor score for that chemical would be 4. If four different organs in female mice were involved in tumor induction and none of the three other groups showed a tumor response, the tumor score for that chemical would also be 4; this illustration is an extreme example of a non-uniform response, which in fact did not occur. The tumor score for each chemical was a measure of the pervasiveness of tumor induction with respect to sex, organ, and species. It was not a measure of potency in the sense of the daily dosage required to induce a given level of tumorigenic response. A zero score, of course, indicated a nontumorigenic response.
The terminology used in the technical reports changed over the years. The earlier reports used the terms positive or suggestive evidence, either of which we considered to be a tumorigenic response. The later reports used the terminology for the tumor bioassays as clear evidence, some evidence, equivocal evidence, or no evidence. We considered either clear or some evidence as an indication of tumorigenicity; the less decisive some evidence responses are indicated in Tables 1-4. The rationale for doing the bioassay was obtained from each technical report. The primary rationales are as follows: 1) widespread usage; 2) structural resemblance to a known carcinogen or mutagen; 3) an inadequate previous study; 4) representative of a structural class that had not been adequately studied; 5) a drug whose use is prolonged; 6) an antineoplastic drug; 7) preexisting evidence ofcarcinogenicity; and 8) potential use as a warfare agent. The first two rationales included 82% of the chemicals. Table 5 summarizes the data from the entire NTP bioassay sample with respect to genotoxicity, contact sensitization, and tumorigenicity, including the proportion of some evidence tumor responses and the primary rationale for the decision to test the chemical. All of the differences, except where indicated, were statistically significant in a two-tailed test at p = <0.05. There was a total of 209 chemicals in the analysis, including the known contact sensitizers and those that were not known to be such. The majority of the chemicals were SALP, i.e., 122 (58%) compared to 87 (42%) that were SAL-. In both the SAL+ and SALgroups, the majority of chemicals were CHO+, 71% and 74%, respectively.
The majority of the 209 tested chemicals were tumorigens (64%). Somewhat more of the SAL+ chemicals were tumorigenic (74%), compared to the SALchemicals (51%). The 90 SAL+ tumorigens were more decisively tumorigenic in the sense that only 12% were in the some evidence category, compared to 34% for the 44 SALtumorigens.
Somewhat more of the 209 chemicals were selected for bioassay because of structural resemblance to carcinogens or mutagens (Rationale 2; 47%), compared to those that were tested solely because of widespread use (Rationale 1; 35%). This was not statistically significant at p = 0.05. Comparing Rationales 1 and 2, the SAP group was heavily weighted (2.5:1) toward chemicals with structural resemblance to carcinogens (Rationale 2), while the SALchemicals were heavily weighted (3.2:1) toward those that were selected because of widespread exposure (Rationale 1).
Of the 209 chemicals, there were 36 (17%) that were contact sensitizers for humans. Table 6 summarizes the data from the 36 contact sensitizing chemicals for the same parameters as those shown in Table 5 for the total sample of 209 chemicals. About two-thirds (63%) of the contact sensitizers were in the SAL-category. This differs from the bioassay population as a whole, in which 42% were SAL-. The same high proportion of both SAL-and SAL+ chemicals (80%) were CHO+ as in the entire bioassay sample (72%). As with the total sample of bioassay chemicals, the majority of the SAL+ sensitizers (55%) were tested because of structural resemblance to carcinogens and mutagens (Rationale 2), and most (80%) of the SALsensitizers were bioassayed because of widespread usage (Rationale 1). About half (54%) of the contact sensitizers were tumorigenic; a larger proportion of the SAL+ contact sensitizers were tumorigenic (77%) compared to the SAL sensitizers (41%). A larger proportion (44%) of the SAL-tumorigens were in the some evidence category compared with 20% in the SAL+ sensitizer carcinogen group. Thus the contact sensitizers differed from the total sample of bioassay chemicals in only two respects: the rationale for selection was different and there was a greater proportion of SAL-chemicals among the contact sensitizers. Table 7 shows the frequency distribution of the tumor scores, i.e., the number of organs in which tumors were induced, as described above according to the Ames assay, for the total NTP sample and for contact sensitizers alone, without regard to the route of administration. The frequency distributions were biased toward the low tumor scores and were analyzed as log distributions. In the total NTP sample, the SAL+ chemicals had about twice the tumor score as the SALchemicals, with a geometric mean and geometric standard deviation of 2.1 ± 2.3 and 0.9 ± 2.0, respectively. This was statistically, significant (p20.0001). The SAL+ contact sensitizers also had a higher geometric mean tumor score than the SAL-contact sensitizers, 1.3 ± 1.9 compared to 0.8 ± 2.1, but the difference was not statistically significant (p = 0.3). There were no blockbuster chemicals in any group except the SAL+, CHO+ group, in which 1,2-dibromoethane, 2,3-dibromo-1propanol, glycidyl, and 1,2,3-trichloropropane had tumor scores of 15, 23, 25, and 19, respectively. There were a few SAL+ contact sensitizers that had substantial tumor scores such as coumarin (9), and 2,4-toluene diisocyanate (9). Among the SALcontact sensitizers, only N-methylolacrylamide and 2-mercaptobenzothiazole had substantial tumor scores of 7 and 6, respectively.
As indicated above, the characterization of tumor responses as tumor scores was a measure of the pervasiveness of the tumorigenic action across species, organs, and sex. Potency was characterized, with respect to genotoxicity, on the basis of the dosage (milligrams per kilogram per day) estimated to produce a 50% tumor incidence (TD50) using published values (10) for many of the same chemicals that were included in this analysis but necessarily excluding those which were nontumorigenic. There were no statistically significant differences according to genotoxicity; this could hardly be otherwise because the range of potencies was enormous, as reflected by the extremes where the difference was a factor of 109.
There appeared to be a difference, according to genotoxicity, in the organs in which tumors were induced in both the total NTP sample and the contact sensitizers. The organs were stratified according to   aRFB-primary rationale for the bioassay as follows: 1, widespread use; 2, structural relation with known carcinogen or mutagen; 3, inadequate previous study; 4, representative of a structural class; 5, drug with prolonged exposure; 6, potential use as an antineoplastic drug; 7, preexisting evidence of carcinogenicity; and 8, potential warfare agent. bTumor score is the number of different organs that showed tumor induction in mice and rats of both sexes; parentheses indicate the number of tumor sites that were obtained only on a single species; setumor responses indicate those that were classified as some evidence in contrast to the stronger designation of clear evidence. subtotals are less than the aggregate totals because 14% of the tumors occurred in small numbers in miscellaneous organs and were not included in the analysis. This is true also for the SAL-chemicals in the total NTP sample and the contact sensitizers of both genotoxicity types. There was a predilection of SAL-chemicals for tumor induction in parenchymal tissues; i.e., 76% of tumors in the SALgroup were in parenchymal organs compared to 15% in epithelial organs; this was statistically significant at p<O.05. In contrast, there was a roughly equal distribution of epithelial and parenchymal organ sites among the SAL+ chemicals. Similarly, among the epithelial organ sites of tumor formation, there was approximately a threefold larger proportion of SAL+ chemicals (49%) compared to SAL chemicals (15%), whereas the opposite relationship was true for the parenchymal organs (37% and 76%, respectively). These comparisons were also statistically significant at p.O.O5. Roughly the same pattern was observed with the contact sensitizers as with the NTP sample as a whole, but without statistical significance. The much larger aggregate tumor score for the SAL+ chemicals (389), compared to the SAL-chemicals (1 15) was due to the larger number of SAL+ chemicals, the higher proportion of tumorresponding animals, and the larger number of organs in which tumors were induced.
The relative inability of SAL-chemicals to induce tumors in epithelial tissues raises the question of whether this might be due to differences in background tumor rates; i.e., if nongenotoxic chemicals are only tumor promoters and tumor promoters act by accelerating tumorigenic processes that are under way, they would be expected to have a predilection for organ sites with relatively high background tumor rates. There was a correlation between tumor induction and background tumor rates in the sense that, when a chemical produced a larger 47-> Abbreviations: SAL, Ames Salmonella mutagenesis assay; CHO, Chinese hamster ovary cell assay for chromosomal abnormalties. Each vertical arrow refers to the percentage of the number at the top of the same column. Each horizontal arrow refers to the percentage of the number in the right hand box in the same row. aPercentage of tumorigens that show some evidence.
Volume 105, Number 9, September 1997 * Environmental Health Perspectives whether the tumors arose partially or entirely from epithelial lining or surface (mucosal) cells or mainly from parenchymal cells. The organs where tumors derived, at least partly, from lining epithelium included the mucosa of the bladder, the bronchial and nasal mucosae, the lining of the gastrointestinal tract from the mouth through the intestines, the skin and adnexal structures, the sebaceous gland of the ear canal (Zymbal gland), and the epithelia of the breast and clitoral and preputial glands. The tumor response in a given organ in one species compared to another, the background tumor rate was generally higher in the organ of the species with the higher tumor response. Table 9 compares the background tumor rates in various organs of rats and mice (11) with the tumor scores for individual organs for the 209 chemicals in the total NTP samples. Only those organs that had a minimum tumor score of 5 are included in the table. For example, in Table 9, the tumor scores for the rat and mouse livers were 45 and 99, respectively, and the corresponding background rates for males and females combined were 4 and 62%, respectively, showing a concordance between the species predominance of liver tumor induction and background tumor rates. Of the 20 organ comparisons, 15 (75%) showed a concordance between the higher species response and the higher background tumor incidence. While there is an association of background tumor rates and species responses, some of the organs that had epithelial tumors also had high background tumor rates that were not enhanced by SALchemicals, such as the lung in mice and the mammary gland in rats. Hence the explanation for the differences in organ patterns of tumor formation according to genotoxicity is not likely to be simple.
To evaluate the frequency of occurrence of contact sensitizers among chemicals in general, we searched the 10,000 entries in the Merck Index (12) for contact sensitizers. We made four passes through the chemicals by taking every hundredth chemical and comparing it to the list of contact sensitizers used above, starting with 1, 101, 1001, 9901; 20, 120, 1020, 9920; 50, 150, 1050, 9950; and 70, 170, 1070, 9970. In each of the four passes of about 100 chemicals each, we identified 4,4,2, and 2 contact sensitizers, respectively, for an average frequency of 3% [standard deviation (SD) = 1%].   bincludes mammary, clitoral, and preputial glands. clncludes thyroid, adrenal, and pituitary. dThe aggregate percentage is less than 100% because miscellaneous organs in which small numbers of tumors occurred were not included in the analysis.

Discussion
Summarizing the above findings, it can be said that of the 209 bioassayed chemicals included in this analysis, 36 were known to be human contact sensitizers (17%). The contact sensitizers differed from the NTP bioassay group as a whole in that a larger proportion were nongenotoxic, which was probably related to the fact that fewer of them were selected for bioassay on the basis of their structural resemblance to known carcinogens, in contrast to widespread usage. When sorted for genotoxicity, the Environmental Health Perspectives * Volume 105, Number 9, September 1997 contact sensitizers had about the same characteristics as the total population of bioassayed chemicals in the following respects: 1) roughly the same proportion were tumorigenic; 2) proportionately more of the SAL+ chemicals were tumorigenic than those that were SAL-; 3) the SALtumorigens had a greater proportion of less decisive (some evidence) tumor responses than the SAP tumorigens; 4) the tumor scores were greater in the SAL+ chemicals than in the SAL-chemicals; and 5) the SAL-chemicals had proportionately more parenchymal tumors (renal, liver, and endocrine tumors and leukemia) than the SAL+ chemicals, whereas the latter produced relatively more epithelial tumors [bladder, gastrointestinal, skin and zymbal gland, breast, clitoral and preputial glands, and respiratory tumors (nose and lung)]. Most of the chemicals were CHO+ regardless of whether they were SAL+ or SALor whether they were carcinogens. This may reflect the possibility that the underlying reason why all of the chemicals were chosen for bioassay, regardless of the specific rationale, was because they were highly reactive as intermediates in chemical synthesis or as reactive finished products; perhaps highly reactive chemicals tend to react with protein. This might mean that many of the NTP bioassay chemicals are contact sensitizers. If so, a comparison of the characteristics of the contact sensitizers to the total group of bioassay chemicals has the limitation that some substantial fraction of the chemicals that were not known to be contact sensitizers may actually be so.
The concentration of contact sensitizers in the SAL-CHO+ category is unexplained. The status of a chemical as CHO+ or CHO-did not consistently affect tumorigenicity. The SAL-CHO-group was anomalous because of its high proportion of tumorigens.
The similarity of the characteristics of the contact sensitizers with those of the entire NTP sample raises the issue of whether contact sensitization might be unrelated to tumorigenicity, i.e., selecting for contact sensitizers constitutes a random sample of bioassay chemicals. However, contact sensitizers were not a random sample because of the differences in the rationale for selection; the properties of the contact sensitizers and the total NTP sample only matched when the two groups were compared on the basis of mutagenicity.
The connection between tumorigens and contact sensitizers relates to the ability of both classes of chemicals to interact with protein. The contact sensitizers react with the amino acid side chains of proteins by a variety of mechanisms: two electron reactions, including nucleophilic and electrophilic substitutions and additions, as well as single electron (free radical) additions (2). Genotoxic tumorigens, as electrophiles, undergo the same kinds of interactions with proteins as contact sensitizers, but in the case of genotoxic tumorigens, the electrophilic interactions extend to DNA. There are subtle differences in chemical reactivity that depend on the strength of the electrophilic and nucleophilic centers, as well as the stearic properties of these chemicals that affect the amount and location of the adducts on protein and DNA. Also, there are important biological factors that have an effect on the outcome of the chemical reactions of carcinogens and contact sensitizers. When an electrophile, a hapten in the context of ACD, is adducted onto the amino acid of a protein, it may become an antigen if the resultant complex has a structure that is capable of sensitizing T lymphocytes; this is a biological response that can have a great deal of individual variability among humans. When an electrophile is adducted onto DNA, it may result in an oncogenic mutation as a consequence of mistakes in DNA repair processes. Hence, even though there is a commonality in the chemical behavior of contact sensitizers and genotoxic tumorigens as electrophiles, there are strong chemical and biological factors that can modulate their effects. However, there is sufficient overlap between the actions of contact sensitizers and tumorigens to raise the possibility that when a chemical is identified as a contact sensitizer, it has a significant likelihood of being a tumorigen.
ACD, caused by SALcontact sensitizers, might be a useful biomarker for identifying nongenotoxic tumor promoters. SAL' contact sensitizers would be expected to be both tumor promoters and initiators. A possible common denominator between contact sensitization and tumor promotion might be the induction of inflammation. ACD is an inflammatory reaction caused by a complex immune response set off by the chemical interaction of the sensitizer, the haptene, with skin proteins. The resultant complex, the antigen, ultimately sensitizes T lymphocytes, which are drawn to the site of skin exposure where they induce a Type IV cell-mediated inflammatory reaction. Contact sensitization also occurs in animals, and the inflammatory reaction is indistinguishable from inflammation produced by skin tumor promoters. Inflammation is a characteristic of many tumor promoters (13,14). Mechanistically, the connection between inflammation and tumor promotion is thought to involve the production of free radicals (15,16). Some tumor promoters are contact sensitizers, including benzoyl peroxide (17), phorbol Volume 105, Number 9, September 1997 * Environmental Health Perspectives myristate acetate (1), and 2,4-dinitro-1-fluorobenzene (1) for mouse skin and butyl hydroxytoluene for rat liver (18). The ability of tumorigens to induce inflammation may be the basis for the promoting component of their action; whether this is due to an immune or irritant reaction is not known (19).
The action of contact sensitizing tumorigens may be stimulated by the induction of humoral immunity. For example, mice injected with rabbit benzo(a)pyrene (BaP) antibodies show a much enhanced skin tumor response to BaP applied dermally (20). Moreover, there is evidence that transplanted malignant cells grow faster in a host that had been previously immunized against them (21).
We examined the issue of whether contact sensitizers are tumor promoters as a consequence of their immunological effects in a previous study that used a well-known contact sensitizer and tumor promoter (22,23), 1-fluoro 2,4-dinitrobenzene (DNFB), in Tg.AC mouse skin. This transgenic mouse is tumor initiated by the presence of a mutated ras oncogene and is therefore a model for skin tumor promotion (24). The skin application of DNFB produced the typical promotional response of squamous papillomas, with an associated dermal inflammatory reaction. However, the immune-depressing corticosteroid fluocinolone acetonide applied locally did not reduce either the inflammation or the tumor response, suggesting that some component of cytotoxicity, not immunogenicity, caused both the inflammation and the tumor promotion, namely, that contact sensitization and tumor promotion may be independent responses to electrophilic chemicals (1). Figure 1 illustrates some of the interrelationships that might exist between electrophilicity, macromolecular adduction, and various biological responses. Two linkages of electrophilicity are shown: DNA and protein adduction. DNA adduction is linked to mutagenicity and thus to tumor initiation and progression. Protein adduction is shown to have a four-way linkage to cell-mediated immunity (contact sensitization), humoral immunity, cytotoxicity, and indirectly to mutagenicity (e.g., possibly by interference with DNA repair and chromosomal segregation). The immune and cytotoxic responses are linked with inflammation and thus with tumor promotion. Promotion of atherosclerotic lesions is included on the basis that polycyclic aromatic hydrocarbon carcinogens like benzo(a)pyrene and 7,12 dimethylbenz(a)anthracene promote the development of aortic atheromata in chickens (25); this finding was prompted by the evidence that aortic atheromatous plaques in humans have a tumorlike monoclonal origin (26). The evidence for the existence of some of these linkages is modest. However, Figure 1 does illustrate that the contact sensitizing and carcinogenic properties of electrophilic chemicals could be independent expressions of their common underlying chemical (electrophilic) reactions, i.e., that contact sensitizers are not tumorigens on the basis of their immunogenic effects but rather because of some aspect of their toxicity. It would be useful to know whether the immune response of contact sensitizers enhances tumorigenicity; if so, immunogenically responsive people might be at increased tumorigenic risk.
Carcinogen-induced deletion of proteins, vital to growth control, was a prominent theory of cancer that stemmed from early work on protein binding in the liver by p-aminoazobenzene, a chemical that, interestingly, is a contact sensitizer (27). This theory was overshadowed by the discovery of carcinogen-induced genotoxicity and was relegated to an epigenetic role (3). Perhaps the epigenetic role of electrophileinduced protein damage is promotion of genetic damage.
A combination of the Ames assay for mutagenicity and an assay for contact sensitization, as a screening tool, might be an improvement in the detection efficiency for environmental tumorigens over that from the Ames assay alone because it would help to uncover rodent tumorigens among the SAL-chemicals. This is seen in an analysis of the Rationale 1 chemicals (widespread use without structural resemblance to known carcinogens), the group that is probably the most relevant to the generality of chemicals in the environment. Of the 73 Rationale 1 chemicals, 23 were SAL' (32%); of these, 15 were tumorigens on bioassay (65%) with an overall tumorigenic detection efficiency of 21% (15/73). Of the remaining 50 chemicals that were SAL-, 16 (32%) were contact sensitizers; of these, 7 (44%) were tumorigenic on bioassay for an overall detection efficiency of 14% (7/50). Thus, in this illustration, if a contact sensitization bioassay followed the Ames test, the yield of rodent tumorigens among the 73 Category I chemicals would rise from 15 to 22, a 43% increase because of the inclusion of SAL-chemicals. An evaluation of false negatives cannot be made because the unknown proportion of chemicals that are not recognized as contact sensitizers, but might be if they were tested. There would be little or no advantage to combining the Ames assay with an assay for contact sensitization when the Ames test is positive because there is a statistically insignificant increase in the efficiency of detecting tumorigens; i.e., 77% compared to 65%.
The 3% of Merck Index chemicals that are contact sensitizers contrasts with the 17% of chemicals in the NTP bioassay series. The Merck chemicals were selected because they were biologically related, and the NTP chemicals were mostly reactive industrial intermediates or compounds in commercial use because of their pesticidal toxicity. The proportion of the 60,000 chemicals in commercial use (28) that are contact sensitizers ranges from 3% to 17%.
An average of 10%, for purposes of estimation, yields perhaps 6,000 contact sensitizers in commercial use. If, as in this analysis of the NTP experience, about half of the contact sensitizers are tumorigenic in the rodent bioassay, there would be about 3,000 chemicals in commerce that would have to be taken seriously as possible tumorigens.

Reviews * Albert
Furthermore, it is likely that only a fraction of contact sensitizers have been identified because it is notoriously difficult to specify a causal agent in most cases of ACD. If contact sensitizers induce tumors because of protein interaction, independent of immunogenicity, then the contact sensitization per se might underestimate tumorigenicity because only some protein adducts induce an immunogenic response.
The finding that contact sensitization might be a useful marker for nongenotoxic tumorigens is necessarily preliminary, because the identification of contact sensitizers in the NTP bioassay series was a matter of happenstance. A more systematic evaluation of the immune response to these NTP chemicals might provide a better definition of the utility of contact sensitization as a screen for potential nongenotoxic environmental tumorigens. An improved understanding of the nature of the macromolecular interactions of contact sensitizers might elucidate why some of these electrophiles are protein interactive but nongenotoxic; also, a better understanding of the chemical mechanisms might lead to an in vitro test that is more relevant to tumorigenesis than contact sensitization.