Address reprint requests to J. Ashby.
We are grateful to D. Paton for technical assistance,
to J.M. Mackay and D. Fox for the cytogenetic data on the oxazolinone,
to S.M. Galloway and M. Armstrong for the cytogenetic data on FITC, and
to A. Wait for determining the log P value for streptozotocin. M.D. Shelby
provided many helpful and definitive comments on early drafts of this paper.
Introduction
There are health consequences of both acute and chronic exposures of
the skin to reactive chemicals. Dermal exposure to some chemicals may cause
allergic sensitization such that subsequent skin contact will result in
erythema and/or edema. Other chemicals may induce cancer at the site of
their chronic, repeated application. Erythema and edema are temporally distinct
and have not hitherto been related, as evidenced by their independent assessments
when screening chemicals for toxicity. The chemical induction of skin cancer
is triggered by dermal exposure to either natural alkylating agents (electrophiles)
such as ß-propiolactone (see Table l), hydrolytically derived alkylating
agents such as N-methyl-N-nitroso-N´-nitro-guanidine
(MNNG), or agents such as benzo[a]pyrene (BaP) and 7,12-dimethylbenz[a]anthracene
(DMBA), which can be metabolized in the skin to an electrophilic epoxide
derivative. Covalent binding of such electrophiles to nuclear DNA can lead
to DNA mutations, and, under appropriate conditions, tumors. As expected,
such agents are mutagenic in assays such as the Salmonella test.
In contrast, although the induction of skin sensitization depends on interaction
of the sensitizing (allergenic) chemical with epidermal Langerhans cells
and the initiation of specific T-lymphocyte responses, the nature and site
of this interaction in the skin is not precisely defined.
Skin sensitization and skin carcinogenicity almost converged in the early
1960s. In 1963, Old et al. (1) reported the ability of BaP, DMBA,
and 3-methylcholanthrene to sensitize guinea pig skin, but not mouse skin.
These authors suggested that the selective immunogenic activity was associated
with the selective carcinogenicity of these agents to mouse skin but not
to guinea pig skin. In particular, Old et al. asserted that chemically modified
proteins triggered an immunological response in the guinea pig that prevented
carcinogenicity.
This possible role of protein interactions in carcinogenicity rapidly
faded when Brookes and Lawley (2) reported a correlation between
the extent of binding of polycyclic aromatic hydrocarbons such as BaP to
mouse skin DNA and their respective carcinogenic potencies to the skin.
The sensitizing activity of BaP in the guinea pig was confirmed in 1978
(3), but in 1987 Klemme et al. (4) demonstrated sensitizing
activity for both BaP and DMBA in the mouse, without making reference to
the earlier negative results of Old et al. In 1967, Stevens (5) reported
the results of a study in which the ability of 44 organic chemicals to sensitize
guinea pig skin was discussed. Three of the sensitizing chemicals listed
by Stevens, N-methyl-N-nitrosourea (MNU) (6), MNNG
(7), and ß-propiolactone (7), have since been reported
as skin carcinogens, but no attempt to correlate skin carcinogenicity and
skin sensitization had been reported. Likewise, more recently, Roberts and
Basketter (8) explored the structure-activity relationship between
the physicochemical properties of six long-chain sulfonic esters and their
sensitizing properties to mouse skin, but there was no discussion of the
possible skin carcinogenicity of these agents.
A major limitation of the earlier skin sensitization literature is the
subjective nature of the observations: edema and erythema were scored on
a scale from light pink to bright pink (5). However, the recently
developed local lymph node assay (9,10) has placed measurement of
sensitization on an objective footing (see below), and this led us to reevaluate
the possible relationship between the electrophilicity/ mutagenicity of
chemicals and their skin-sensitizing properties and skin carcinogenicity.
The mutagenicity of a chemical might provide an indication of its skin-sensitizing
potential, and this, in turn, might provide a rapid and convenient indication
of its carcinogenic potential to the skin--the definition of which is both
time and resource consuming.
Methods
All chemicals and reagents are commercially available, were of the highest
available purity, and were used as received (10). The biological
techniques we used are well established. We determined skin-sensitizing
activity by exposing the dorsum of both ears of groups of four CBA/Ca mice
to the test chemical in olive oil/acetone (4:1) on 3 successive days. Five
days after the initiation of exposure, we injected mice with tritiated thymidine
(3HTdR; 30 µCi/animal), and 5 hr later we measured lymphocyte
proliferation in the pooled auricular lymph nodes local to the treated ears
in terms of incorporated radioactivity.
Activity in the local lymph node assay is expressed as the ratio of test
to control incorporation of 3HTdR (T/C) (10). In practice,
chemicals which under these conditions elicit a 3-fold or greater increase
in 3HTdR incorporation relative to vehicle-treated controls are
considered to have skin-sensitizing potential (10). We determined
mutagenicity to Salmonella typhi-murium in the absence of auxiliary
metabolism using strain TA98 for the aromatic alkylating agents, strain
TA1535 for the simple alkylating agents, and strain TA98 (+S9 mix) for BaP,
DMBA, 2-acetylaminofluorene (2-AAF), and 2-nitrofluorene (2-NF) (11).
In two instances the clastogenicity of established sensitizing agents was
measured in cultured mammalian cells. The genotoxicity data for these two
chemicals are referred to in the text but will be published elsewhere (Ashby,
in preparation). We used a colorimetric assay for alkylating activity employing
the colorless reagent p-nitrobenzylpyridine (NBP), which yields a
deep-blue product upon alkylation of its pyridine nitrogen atom (12).
Results and Discussion
To initiate either carcinogenesis or skin sensitization, a biologically
significant level of DNA or protein adduction, respectively, must be achieved
in the appropriate epidermal target cells within the initial few hours of
exposure. This suggests a possible overlap of the structure-activity relationships
for mutagenesis, skin sensitization, and skin carcinogenesis. For these
initially nonspecific interactions with protein and DNA in the skin to occur,
three conditions must be met. First, the chemical, or a skin-derived metabolite,
must be electrophilic and therefore capable of covalent reaction with nucleophilic
sites on proteins and DNA. Second, a sufficient quantity of the agent must
penetrate the lipophilic stratum corneum of the skin. The partition coefficient
of the chemical will clearly influence penetration; agents with low coefficients
(i.e., high water solubility) are relatively disadvantaged. Third, once
in the epidermis, the agent must be sufficiently reactive to produce a biologically
significant level of protein/DNA alkylation before it is dispersed from
the site of application and/or metabolically detoxified. These physicochemical
factors are subsumed in the relative alkylation index (RAI) suggested by
Roberts and Basketter (8) for alkyl sulfonates.
Preliminary experiments established that the reference mouse skin carcinogens
and bacterial mutagens MNU (6), N-ethyl-N-nitrosourea
(ENU) (13), MNNG (7), N-ethyl-N-nitroso-N´-nitroguanidine
(ENNG) (14), ß-propiolactone (7), BaP (7), and
DMBA gave a strong positive response in the lymph node assay (Table 1).
The response observed for MNNG was the strongest yet recorded in this assay
and was induced at dose levels similar to those used in the skin-painting
cancer bioassay. The activity of ß-propiolactone depends on the electrophilicity
of its strained lactone ring, which adds further structural diversity to
this set of carcinogens. Similar strong sensitizing responses were seen
for the potent mutagens 1-chloromethyl-pyrene (1-CMP) and 2-chloromethylfluorene
(2-CMF), two aromatic chloromethyl compounds designed to model the ultimate
carcinogenic electrophiles formed metabolically from polycyclic carcinogens
such as BaP and DMBA (15). Comple-menting these strong sensitizing
activities was the inactivity observed for 2-chloro-ethanol (2-CE), a weak
bacterial mutagen and probably the most well-defined mouse skin noncarcinogen
yet described (16). The inactivity of 2-CE as a sensitizing agent
was established under both the present standard conditions of the test and
when applied under the conditions used in the skin-painting cancer bioassay
(a 1.6% solution in 70% ethanol). These data therefore indicate that genotoxic
carcinogens (DNA reactive/electrophilic) can also elicit immunogenic effects
via their inevitable parallel reactions with proteins. The one Salmonella
mutagen (2-CE) that was unable to express its intrinsic genotoxicity as
carcinogenicity to mouse skin was also inactive as a sensitizing agent to
mouse skin.
The skin-sensitizing and carcinogenic activities of BaP and DMBA depend
on their metabolism in the skin to electrophilic epoxides, and it therefore
became of interest to study the sensitizing activities of the two rodent
liver carcinogens 2-AAF and 2-NF, which are known to require complex gut/hepatic/biliary
metabolism to electrophiles by enzymes, some of which are not adequately
represented in the skin. The inactivity of 2-AAF and 2-NF as sensitizing
agents at the application concentrations studied was in contrast to the
strong sensitizing activity observed for the direct-acting (S9 independent)
mutagen 2-(N-acetyl-N-acetoxy)aminofluorene (2-AAAF), a model
of the ultimate electrophile formed metabolically from both 2-AAF and 2-NF
(Table 1). These data establish that some genotoxins are prevented from
eliciting biological responses in the skin due to a failure of appropriate
metabolism. It is possible that weak sensitizing activity would be observed
for 2-AAF and 2-NF at elevated dose levels, but the intention here was to
conduct comparative studies. The similar sensitizing activity of the two
2-fluorenyl derivatives, 2-AAAF and 2-CMF (Table 1), is probably related
to the similar and bulky fluorenyl adducts they produce on proteins. From
these data, we conclude that probably all genotoxic (DNA reactive) carcinogens
have intrinsic skin-sensitizing potential, but this potential is only realized
when the carcinogens are appropriately absorbed across the stratum corneum
and, if necessary, when their metabolic conversion to an electrophile is
achieved in the skin. These data also indicate that 2-AAAF, 2-CMF, and 1-CMP
have carcinogenic potential to the skin.
The 13 aliphatic alkylating agents shown in the middle panel of Figure
l display mutagenicity to Salmonella over a 107 dose range.
Although each of these agents was already known to be mutagenic, they have
not hitherto been tested concomitantly, and consequently the wide differences
in their mutagenic potencies and minimum effective dose levels have not
been known. It is probable that this dispersion of active dose ranges is
influenced both by the different intrinsic reactivities of the agents and
by the extent of formation and repair of O6-alkyl guanidine adducts
on DNA. The dose-response envelopes shown in Figure l are for mutagenicity;
the shading indicates that those agents also act as skin-sensitizing agents
(see also Table 1).
Given the preponderance of immunogenic chemicals shown in Figure 1, mutagens
that were inactive as sensitizing agents obviously become of interest. The
human antitumor agent semustine (MeCCNU) is active in slowing tumor growth
by virtue of its ability to cross-link DNA, a property dependent on its
hydrolysis to the bifunctional alkylating agent ClCH2CH2N+.
By cross-linking DNA, cell replication is impeded (17). The most
likely explanation for the inactivity of MeCCNU as a sensitizing agent is
therefore that it inhibits the necessary proliferation (9) of T-lymphocytes
in draining lymph nodes. Consistent with this, application of MeCCNU 1 hr
after exposure to MNNG led to a 75% attenuation of the concomitantly observed
local lymph node assay response to MNNG (Table 2).
Figure 1. Relative mutagencity
to Salmonella of 20 agents tested for skin sensitization. The mutagenicity
dose-response envelopes are shaded for skin-sensitizing agents (see Table
1).
The failure of streptozotocin to elicit a sensitizing response is initially
surprising because this nitrosourea is a rodent carcinogen and mouse somatic-cell
mutagen when administered by intraperitoneal injection (18). The
critical structural difference between streptozotocin and MNU is the sugar
substituent on the former. The measured log P value for streptozotocin was
-2.0 ± 0.3 (shake flask technique) compared to MNU's value of approximately
0, and this 100-fold enhancement of relative water solubility probably reduces
the bioavailability of streptozotocin by attenuating its translocation across
the lipophilic stratum corneum. In fact, streptozotocin was not completely
soluble in the standard acetone/olive oil vehicle; consequently, at the
highest three dose levels shown in Table 1, acetone/water (4:1) was used
as vehicle, but even in this vehicle complete solution was not achieved
at the highest dose level. All of the other agents in this study have log
P values in the range 0-6, so a critical log P value of around -2 probably
provides one key parameter for determining skin toxicities. It is interesting
to note that exposure by intraperitoneal injection is not impeded by the
protective lipid barrier of skin, thus explaining the genotoxic activities
of streptozotocin in mice exposed by the parenteral route. [For a review
of the rodent genotoxicity of streptozotocin, see Liegibel et al. (18)].
Dimethylsulfate (DMS), diethylsulfate (DES), and the two sulfonate esters
[methyl methanesulfonate (MMS) and ethyl methanesulphonate (EMS)] are interesting
because they demonstrate a wide range of reactivities as alkylating agents.
Ehrenberg et al. (19) have measured the hydrolytic half-lives of
three of these compounds and found DMS to have a t1/2
of 0.l hr and EMS and MMS to have t1/2 values of 4 hr
and 5 hr, respectively. These data suggest that the relatively high dose
levels of DMS required to elicit mutagenicity to Salmonella may be
due to its hydrolysis before reaction with DNA, while the similar high dose
levels required for EMS and MMS may be due to their low intrinsic reactivity.
Likewise, although each of these alkylating agents gives a blue color
in the NBP test, for DMS the color can be abolished by preincubation for
2 min with the nucleophile piperidine, while the other three compounds retain
alkylating activity after 20 min of such preincubation (data not shown).
It is relevant that the weak Salmonella mutagen 2-CE was of such
low intrinsic alkylating activity that it required warming with the NBP
reagent in order to develop a blue color, consistent with its noncarcinogenicity
and nonsensitizing activity. The skin carcinogenicity of EMS and MMS is
unknown, but the present data indicate EMS to be noncarcinogenic and MMS
to have very low carcinogenicity to the skin. A limited skin-painting carcinogenicity
bioassay of DMS revealed a negative response (20), but the present
sensitization data indicate a carcinogenic potential that may be realized
in an adequate bioassay such as that used by the National Toxicology Program
for the evaluation of 2-CE (16). Diethysulfate and N-propyl-N-nitroso-N'-nitroguanidine
are of unknown carcinogenicity to the skin, but based on the present sensitization
data (Table l), we suggest that they have such potential.
Given this correlation between the skin carcinogenicity of a genotoxin
and its activity as a skin-sensitizing agent, the role of genotoxicity data
in the prediction of skin sensitization becomes of interest. When screening
chemicals for genotoxicity, it is usual to conduct an assay for the induction
of chromosomal aberrations (clastogenesis) on agents found to be nonmutagenic
to Salmonella. The reason for this assay is that some chemicals are
mammalian cell mutagens despite their inability to mutate prokaryotic DNA
(21,22). This in turn raises the question of whether a chemical that
is nonmutagenic to Salmonella but which is a mammalian-cell clastogen
would have skin-sensitizing potential.
To address this possibility, we evaluated the genotoxicity of two established
sensitizing agents active in the local lymph node assay (23) but
whose chemical structures (22) indicated they were unlikely to be
mutagenic to Salmonella: the oxazolinone 4-ethoxymethylene-2-phenyl-2-oxazolidin-5-one
(EPO; I, Fig. 2) and the fluorochrome fluorescein isiothiocynate (FITC;
II, Fig. 2). Both chemicals were reproducibly nonmutagenic to Salmonella
but reproducibly clastogenic to cultured mammalian cells. The Salmonella
assays were conducted using strains TA1535, 1537, 1538, 98, and 100 (11)
in the presence and absence of induced rat-liver S9 mix to a high dose level
of 1 mg/plate (FITC) or 5 mg/plate (EPO). The oxazolinone EPO was reproducibly
clastogenic to cultured human lymphocytes (>50 µg/ml) when treated
in the absence of S9 mix for 24 hr (Mackay and Fox, unpublished data). Fluoroscein
isothiocyanate was re-producibly clastogenic to cultured Chinese hamster
ovary cells (>400 µg/ml) when treated in the absence of S9 mix
for 3 hr and harvested at 20 hr (Galloway and Armstrong, unpublished data).
Possible electrophilic sites within the structures of EPO and FITC that
could be responsible for their clastogenicity and sensitizing activities
are indicated in Figure 2, but why such reactive centers should fail to
cause mutations in bacteria remains obscure. The concept that the Michael
center in EPO and the NCS group in FITC are reactive exclusively to proteins
is an initially appealing explanation for these selective mutagenic effects,
but it remains speculative.
Figure 2. Chemical
structure of 4-ethoxymethylene-2-phenyl-2-oxazolidin-5-one (I) and fluorescein
isothiocyanate (II). The arrows indicate the site of possible electrophilicity
on each molecule. Fluorescein isothiocyanate is sometimes drawn in the ring-closed
spiro-lactone form.
The above findings suggest that the mutagenicity of a chemical, as opposed
to just its activity in the Salmonella assay, indicates carcinogenic
and sensitizing potential of the chemical to the skin. Nonetheless, subtle
differences in the structure-activity relationships for these three toxicities
will exist. For example, all of the methylating agents studied (MNU, MNNG,
MMS, and DMS) are more potent sensitizing agents (Table 1) than are their
ethylating analogues (ENU, ENNG, EMS, and DES, respectively). This suggests
that methylated proteins have higher immunogenicity, per se, compared to
ethylated proteins. In contrast, although MNU and MNNG are more potent Salmonella
mutagens than their respective ethyl analogues (ENU and ENNG), EMS and DES
are clearly more potent mutagens than their methyl analogues MMS and DMS.
These selective mutagenic effects are related to differences in the nature
of the electrophilicity of the alkyl nitrosoure-as/guanidines as compared
to the alkyl sulfonates/sulfates. These chemical-class differences lead
to different adduction profiles on DNA, which in turn can lead to different
and organism-specific DNA-repair/ mutagenic consequences (24). Similarly,
differences in the mitogenic properties of a mutagen can lead to differences
in the progression of initiated cells to carcinogenesis (25). Subtle
differences are to be expected between the structure-activity relationships
of mutagenesis, skin sensitization, and dermal carcinogenicity, despite
an underlying coherence. This overall coherence is illustrated by the facts
that the most potent skin-sensitizing agent described here (MNNG) is also
the most potent of the present mutagens, and the weakest sensitizing agent
(MMS) is also the weakest mutagen among those that are skin-sensitizing
agents (Table 1; Fig. 1).
We therefore suggest that adequate evaluation of chemicals for genotoxicity
(21) will also yield information on skin-sensitizing potential. In
the case of metabolism-dependent genotoxins, the likelihood of the appropriate
biotransformation occurring in the skin will have to be considered. Further,
we suggest that the activity of a genotoxin as a skin-sensitizing agent
provides strong indications of its potential skin carcinogenicity. It must
be borne in mind, however, that relatively high dose levels are used when
assessing the skin-sensitizing potential of a chemical because this may
reflect the conditions of accidental human exposure (10). Chronic
exposure of the skin to high, acute dose levels may not always be possible.
Consequently, consideration of acute versus chronic allowable dose levels
must accompany extrapolation of skin sensitization data to predict skin
carcinogenic potential. Nevertheless, the dose levels used in this study
were generally within the ranges used to establish the potency of the such
skin carcinogens.
As it took more than a decade to qualify the initial observation that
carcinogens are mutagens (7), it would be unwise to transmute the
present correlations into inviolable relationships. Nonetheless, it is clear
that a common structure-activity relationship underpins genotoxicity, skin
sensitization, and skin carcinogenesis and that in the early stages of chemically
induced skin carcinogenesis, adduct-related immunogenic as well as mutagenic
activity is occurring. Such knowledge should aid and simplify the hazard
assessment of chemicals.
