IPCS INCHEM Home


        INTERNATIONAL PROGRAMME ON CHEMICAL SAFETY

        WORLD HEALTH ORGANIZATION



        SAFETY EVALUATION OF CERTAIN
        FOOD ADDITIVES AND CONTAMINANTS



        WHO FOOD ADDITIVES SERIES 40





        Prepared by:
          The forty-ninth meeting of the Joint FAO/WHO Expert
          Committee on Food Additives (JECFA)



        World Health Organization, Geneva 1998



    AFLATOXINS

    First draft prepared by
    S. Henry1, F.X. Bosch2, J.C. Bowers1, C.J. Portier3, 
    B.J. Petersen4 and L. Barraj4
    1 US Food and Drug Administration, Washington, DC
    2 Institut d'Oncologia, Unitat d'Epidemiologia, Hospitalet del
    Llobregat, Barcelona, Spain
    3 National Institute of Environmental Health Sciences, Research
    Triangle Park, NC, USA
    4Novigen Sciences Inc., Washington, DC, USA (authors of section 4)

    1.  Explanation
    2.  Biological data
        2.1 Biochemical aspects
            2.1.1   Metabolism of aflatoxins
        2.2 Toxicological studies
            2.2.1   Acute toxicity
            2.2.2   Special studies on reproductive toxicity
            2.2.3   Special studies on genotoxicity
            2.2.4   Special studies on immunosuppression
            2.2.5   Factors modifying carcinogenicity of aflatoxins
            2.2.6   Special studies on covalent binding of aflatoxin
                    residues with nucleic acids and proteins
            2.2.7   Special studies on glucose tolerance
            2.2.8   Special studies on effect of ammoniation of AFBI in
                    contaminated cottonseed
            2.2.9   Special studies on aflatoxin and hepatitis B virus
                    infection in woodchucks, ducks, ground squirrels and
                    tree shrews
            2.2.10  Observations in humans
                2.2.10.1    Biomarkers of aflatoxin exposure
                2.2.10.2    Mutations in p53 tumour-suppressor gene in
                            human hepatocellular carcinoma
                2.2.10.3    Epidemiology of primary liver cancer
            2.2.11  Summary of information on other aflatoxins
                2.2.11.1    Aflatoxin B2
                2.2.11.2    Aflatoxin G1 
                2.2.11.3    Aflatoxin G2
                2.2.11.4    Aflatoxin M1
    3.  Estimating carcinogenic risks from the intake of aflatoxins
        3.1 Information from various scientific disciplines and its
            contribution to aflatoxin carcinogenic risk
            3.1.1   Laboratory animal, mutagenicity and metabolic studies
            3.1.2   Studies on the p53 gene 
            3.1.3   Epidemiological studies
            3.1.4   Aflatoxin biomarker studies
        3.2 General modelling issues
            3.2.1   Choice of data
            3.2.2   Measure of exposure
            3.2.3   Measure of response
            3.2.4   Choice of mathematical model

        3.3 Potency estimates
            3.3.1   Potency estimates based upon epidemiological data
            3.3.2   Potency estimates not accounting for HBV infection
            3.3.3   Potency estimates accounting for HBV infection
            3.3.4   Potency estimates based on biomarker studies
            3.3.5   Potency estimates from test species
    4.  Aflatoxin dietary intake estimates
        4.1 Introduction
        4.2 Background
        4.3 Methods
            4.3.1   Period of intake of relevance
            4.3.2   Estimated levels of aflatoxins in foodstuff
            4.3.3   Estimated intakes
        4.4 Results
            4.4.1   Aflatoxin levels in foods: general
            4.4.2   Aflatoxin levels in foodstuffs: Occurrence data by
                    commodity
                4.4.2.1 Amount of commodity imported
                4.4.2.2 Accounting for the change in aflatoxin levels
                        during processing
            4.4.3   National estimates of aflatoxin intake
                4.4.3.1 Australia
                4.4.3.2 China
                4.4.3.3 European Union
                4.4.3.4 USA
                4.4.3.5 Zimbabwe
            4.4.4   Relative impact of establishing maximum limits on
                    estimate of intake
                4.4.4.1 Average aflatoxin concentrations using four
                        possible scenarios
                4.4.4.2 Intake of total aflatoxins using four scenarios
                4.4.4.3 Intake of aflatoxin b1 within four scenarios
            4.4.5   Summary
    5.  Comments and evaluation
        5.1 Aflatoxin potencies
        5.2 Population risks
        5.3.    Conclusions
    6.  References

     List of abbreviations

    AAT        alpha-1-antitrypsin
    ADA       aflatoxin-DNA adduct
    AF        aflatoxin (general)
    AF-alb    aflatoxin-albumin (adduct)
    AFB1      aflatoxin B1
    AFB2      aflatoxin B2
    AFG1      aflatoxin G1
    AFG2      aflatoxin G2
    AFL       aflatoxicol
    AFM1      aflatoxin M1
    AFP        alpha-fetoprotein
    AL         ad libitum
    ALT       alanine aminotransferase
    AM        alveolar macrophage
    APAT      ambient temperature ammoniation procedure
    BNF        beta-naphthoflavone
    CMI       cell-mediated immunity
    CR        calorically restricted
    CYP       cytochrome P450
    DHBV      duck hepatitis B virus
    DTH       delayed type hypersensitivity
    eAAIR     estimated age adjusted incidence rate
    EPHX      epoxide hydrolase
    GGT        gamma-glutamyltranspeptidase
    GHIS      Gambia hepatitis intervention trial
    GSHV      ground squirrel hepatitis virus
    GST       glutathione S-transferase
    GSTM1     glutathione S-transferase M1
    HBV       hepatitis B virus
    HC        high carbohydrate (diet)
    HCC       hepatocellular carcinoma
    HCV       hepatitis C virus
    HF        hypercaloric fat-containing (diet)
    HPHT      high temperature ammoniation procedure
    IC        isocaloric fat-containing (diet)
    I3C       indole-3-carbinol
    LC        liver cancer
    LDH       lactate dehydrogenase
    MDA       malonaldehyde
    OECD      Organisation for Economic Co-operation and Development
    Orm       matched odds ratio
    PCR       polymerase chain reaction
    PHC       primary hepatocellular carcinoma
    PLC       primary liver cancer
    ROS       reactive oxygen species
    SeY       selenium-enriched yeast extract
    WHV       woodchuck hepatitis virus

    1.  EXPLANATION

         Aflatoxins B1, B2, G1, and G2 are mycotoxins that may be
    produced by three moulds of the  Aspergillus species:  A. flavus, 
     A. parasiticus and  A. nomius, which contaminate plants and plant
    products. Aflatoxins M1 and M2, the hydroxylated metabolites of
    aflatoxin B1 and B2, may be found in milk or milk products obtained
    from livestock that has ingested contaminated feed. Of these
    six aflatoxins, aflatoxin B1 is the most frequent one present in
    contaminated samples and aflatoxins B2, G1 and G2 are generally not
    reported in the absence of aflatoxin B1. Most of the toxicological
    data relate to aflatoxin B1. Dietary intake of aflatoxins arises
    mainly from contamination of maize and groundnuts and their products.

         Aflatoxins were evaluated at the thirty-first meeting of the
    Committee (Annex 1, reference 77), at which time the Committee
    considered aflatoxin to be a potential human carcinogen. Sufficient
    information was not available to establish a figure for a tolerable
    level of intake. The Committee urged that the intake of dietary
    aflatoxin be reduced to the lowest practicable levels so as to reduce,
    as far as possible, the potential risk. A working group convened by
    the International Agency for Research on Cancer also concluded that
    naturally occurring aflatoxins are carcinogenic to humans1.

         At the forty-sixth meeting (Annex 1, reference 122), potency
    evaluations and population risk estimates were considered, and the
    Committee recommended that these analyses be completed and presented
    in an updated toxicological review.

         At its present meeting, the Committee reviewed a wide range of
    studies in both animals and humans that provided qualitative and
    quantitative information on the hepatocarcinogenicity of the
    aflatoxins. This monograph reviews the experimental evidence
    concerning the carcinogenicity of the aflatoxins, evaluates the
    potencies of these contaminants, links these potencies to intake
    estimates, and discusses the impact of hypothetical standards on
    sample populations and their overall risks.

         The scientific literature on aflatoxins in the past thirty years
    includes more than 3000 research articles. In 1971 aflatoxins were
    reviewed in Volume 1 of the International Agency for Research on
    Cancer (IARC) Monographs on the Evaluation of Carcinogenic Risk and
    again in Volume 56 of the IARC monographs in 1993. Aflatoxins were
    last reviewed by the JECFA in 1987. A key recent publication in the
    aflatoxin field was the review by Eaton & Groopman (1994). Eaton &

                   

    1 Some naturally occurring substances: food items and constituents,
    heterocyclic aromatic amines and mycotoxins. Lyon, International
    Agency for Research on Cancer, 1993 (IARC Monographs on the Evaluation
    of Carcinogenic Risks to Humans, Vol. 56): 245-395.

    Gallagher (1994) wrote a review of the mechanisms of aflatoxin
    carcinogenesis. This review for JECFA will focus on key reports that
    have appeared in the literature since the publication of the Eaton &
    Groopman review and the 1993 IARC review.

    2.  BIOLOGICAL DATA

    2.1  Biochemical aspects

    2.1.1  Metabolism of aflatoxins

    An excellent review on the cellular interactions and metabolism of
    aflatoxin has been produced by McLean & Dutton (1995). Gorelick (1990)
    compared metabolism of aflatoxin by different species. Guengerich 
     et al. (1996) discussed the involvement of cytochrome P450,
    glutathione-S-transferase and epoxide hydrolase in the metabolism of
    aflatoxin B1 (AFB1) and the relevance to risk of human liver cancer. A
    wide variety of vertebrates, invertebrates, plants, bacteria and fungi
    are sensitive to aflatoxins, but the range of sensitivity is wide for
    reasons not yet fully understood (Cullen & Newberne, 1994). Two
    important factors in species and strain variation of sensitivity are
    1) the proportion of AFB1 that is metabolized to the 8,9-epoxide,
    relative to other metabolites that are considerably less toxic, and 2)
    the relative activity of phase II metabolism, which forms non-toxic
    conjugates and inhibits cytotoxicity1. The 8,9-epoxide of AFB1 is
    short-lived but highly reactive, and is believed to be the principal
    mediator of cellular injury (McLean & Dutton, 1995).

         Formation of DNA adducts of AFB1-epoxide is well characterized
    (Cullen & Newberne, 1994). The primary site of adduct formation is the
    N7 position of the guanine nucleotide.

         It has been hypothesized that viral infection and associated
    liver injury alter expression of carcinogen-metabolizing enzymes.
    Kirby  et al. (1994) tested this hypothesis in a hepatitis B virus
    (HBV)-transgenic mouse model in which a synergistic interaction occurs
    between AFB1 and HBV in the induction of hepatocellular carcinoma
    (HCC). In this transgenic mouse lineage, overproduction of the HBV
    large envelope protein results in progressive liver cell injury,
    inflammation, and regenerative hyperplasia. Initially, two cytochrome
    P450s important in AFB1 metabolism in the mice were identified -
    CYP2a-5 and CYP3a, using specific antibodies and chemical inhibitors.
    The expression of these P450 isoenzymes and an alpha-class glutathione

                   

    1 Phase I enzymes of major importance to carcinogen metabolism are
    certain members of the superfamily (primarily within families 1-3) of
    CYPs. In general, P450 enzymes catalyse the formation of more polar,
    non-toxic products; however, bioactivation is sometimes a sequela. The
    phase II enzymes of primary importance are the GST, which catalyse
    conjugation of potentially toxic electrophiles to the tripeptide GSH,
    generally rendering them non-toxic.

    S-transferase (GST) isoenzyme, YaYa, was examined. Increased
    expression and altered distribution of CYP2a-5 were demonstrated, by
    immunohistochemical analysis, to be associated with the development of
    liver injury in mice and to increase with age between 1 and 12 months.
    CYP3a expression was also increased in HBV-transgenic mice, but the
    increase was not as clearly related to age. GST YaYa levels were the
    same in HBV-transgenic mice and their non-transgenic littermates of
    all ages.

         These results show that expression of specific cytochrome P450s
    is altered in association with over-expression of HBV large envelope
    protein and liver injury in this model. These findings may have
    general relevance to human HCC, which is associated with a diverse
    range of liver-damaging agents.

         Judah  et al. (1993) studied an aldehyde reductase in the rat,
    which, in contrast with fractions from control animals, catalysed the
    reduction of AFB1-dihydrodiol, in the dialdehyde form at physiological
    pH values, to AFB1-dialcohol. This aldehyde reductase was expressed in
    cytosolic fractions prepared from rat livers bearing pre-neoplastic
    lesions, or following treatment with the anti-oxidant ethoxyquin. This
    expression paralleled the development of resistance to the toxin. This
    enzymatic mechanism might also have relevance in terms of the
    development of resistance to other cytotoxic agents, the mechanism of
    which involves metabolism to a reactive aldehyde. The authors
    suggested that other systems, in particular human, be examined to
    determine if this enzyme activity is expressed, and if so in what
    circumstances, before its potential significance in the carcinogenic
    process can be evaluated. For example, do the livers of humans
    consuming diets contaminated with aflatoxins express such enzymes?

         The monkey CYP1A1 has been expressed in BALB 3T3 A31-1-1 cells
    and the expressed proteins were assayed for their capacity to activate
    AFB1 and benzo[a]pyrene (B[a]P (Itoh  et al., 1993). The transformed
    cells were approximately 5.4- to 4.7-fold more sensitive to AFB1 and
    B[a]P than the parental cells, respectively. The authors concluded
    that monkey CYP1A1-cDNA encoded a functional protein and that the
    expressed CYP1A1 enzyme is active in the activation of B[a]P as well
    as AFB1 to produce toxic metabolites.

         The combined presence of CYP1A2 and 3A4, both of which oxidize
    AFB1 to the reactive AFB1-8,9-epoxide and to hydroxylated inactivation
    products aflatoxin M1 (AFM1) and aflatoxin Q1 (AFQ1), substantially
    complicates the kinetic analysis of AFB1 oxidation in human liver
    microsomes. Gallagher  et al. (1996) examined the reaction kinetics
    of AFB1 oxidation in human liver microsomes (N = 3) and in human
    CYP3A4 and CYP1A2 cDNA-expressed lymphoblastoid microsomes for the
    purpose of identifying the CYP isoform(s) responsible for AFB1
    oxidation at low substrate concentrations approaching those
    potentially encountered in the diet. CYP3A4 with AFB1 was found to
    have sigmoidal kinetics such that the rate of product formation fell
    off quickly as the substrate concentration was reduced. CYP1A2 obeyed
    Michaelis-Menten kinetics. Thus, at the low substrate concentrations

    that probably occur  in vivo, the formation of AFB oxide, as well as
    AFB1 clearance, were predicted to be dominated by CYP1A2. Even at a
    relatively higher substrate concentration, CYP1A2 formed approximately
    three times as much AFB-exo-epoxide and generated three times as much
    DNA binding as an equivalent amount of cDNA-expressed CYP3A4.

         The authors pointed out that because AFB1 is highly lipophilic,
    it is difficult to know how nominal concentrations of AFB in  in 
     vitro microsomal preparations relate to concentrations  in vivo. 
    The authors also discussed the work of Ueng  et al. (1995), who
    reported that CYP1A2 formed less AFB oxide than CYP3A4, using human
    CYP1A2 and 3A4 proteins that were expressed in a bacterial expression
    system. The discrepancy between the two studies, according to
    Gallagher  et al. (1996), may have been due to different sources of
    P450s used in the two experiments.

         Gallagher  et al. (1996) concluded that the dominant route for
     in vivo AFB1 activation at dietary concentrations obtained in human
    liver is primarily thorough CYP1A2. Evidence that both CYP1A2 and 3A4
    are involved in AFB1 metabolism  in vivo is substantiated by
    biomarker studies indicating the presence of AFM1 and AFQ1 in the
    urine of individuals exposed to dietary AFB1 (Ross  et al., 1992;
    Qian  et al., 1994). The ratio of activation:inactivation products
    catalysed by CYP1A2 (roughly 2.5:1, AFBO:AFM1) and CYP3A4 (1:10;
    AFBO:AFQ1) is likely to be a key determinant of the pathway and
    biological consequences of  in vivo AFB1 exposure. Unfortunately, the
    actual urinary and faecal levels of these two metabolites, (in
    particular, AFQ1 and possible secondary metabolites) following
    exposure to AFB1 are not known. Thus, the relative ratio of these two
    metabolites in individuals exposed to dietary AFB1, a key ratio, is
    also unknown.

         Sawada  et al. (1993) show that human placental microsomes
    activated AFB1, AFB1 showed relatively high mutagenic activity in the
    Ames test when incubated with human placental microsomes. Addition of
     alpha-naphthoflavone or aminoglutethimide, known inhibitors of
    cytochrome P450 1A and P450 19, respectively, into the test system
    partially inhibited the mutagen-producing activity.

         Induction of glutathione-S-transferase placental form (GST-P)
    positive hepatic foci has been examined by immunohistochemical
    analysis in young male Fischer rats 3 weeks after a single i.p.
    injection of AFB1 (Gopalan  et al., 1993). Pretreatment of rats with
    L-buthionine sulfoximine (BSO), a GSH depleter, at a dose of 4 mmol/kg
    bw 4 and 2 hours before 1.0 mg AFB1 treatment enhanced both the number
    of AFB1-induced hepatic foci and the area occupied by these foci by
    approximately 400 and 575% above their respective controls without
    affecting the mean diameter of these foci. Pretreatment of rats with
    0.1% phenobarbital (PB) in their drinking water for 1 week before AFB1
    (1 mg) treatment, inhibited AFB1-induced foci almost completely.
    However, the number of AFB1-induced foci in PB-treated rats was not
    significantly increased by BSO pretreatment.

         Fetal rat liver has been shown to possess substantial levels of
    glutathione-S-transferase (GST) activity toward AFB1-8,9-epoxide. The
    enzyme responsible for this activity was an alpha-class GST
    heterodimer comprising Yc1 and Yc2 subunits (Hayes  et al., 1994).
    The cDNAs encoding these polypeptides have been cloned and shown to
    share approximately 91% identity over 920 base pairs, extending from
    nucleotide -23 to the AATAAA polyadenylation signal. GST Yc2Yc2
    expressed in  Escherichia coli was found to exhibit 150-fold greater
    activity toward AFB1-8,9-epoxide than GST Yc1Yc1. Comparison between
    the structures of alpha-class GST suggested that tyrosine at residue
    108 and/or aspartate at residue 208 is responsible for the high AFB1
    detoxification capacity of Yc2. Immunoblotting and enzyme assays have
    shown that liver from adult female rats contains about 10-fold greater
    levels of Yc2 than is found in liver from adult male rats. This
    sex-specific expression of Yc2 in adult rat liver may contribute to
    the relative insensitivity of female rats to AFB1. Dietary
    administration of oltipraz, a synthetic antioxidant which protects
    against aflatoxin-hepatocarcinogenesis served as an inducer of GST
    Yc2.

         Gallagher & Eaton (1995) have investigated the biotransformation
    of AFB1 in hepatic microsomal and cytosolic fractions from channel
    catfish, an aquatic species shown to be refractory to AFB1 toxicity
    and reported to be resistant to AFB1 hepatocarcinogenesis, and in
    rainbow trout, a species sensitive to AFB1 toxicity and
    hepatocarcinogenesis. AFB1 was poorly oxidized by channel catfish
    microsomes, suggesting that the lack of microsomal AFB1 activation
    together with the rapid conversion of AFB1 to aflatoxicol (AFL)
    contributes to the apparent resistance of channel catfish to AFB1
    toxicity and hepatocarcinogenesis.

         Oltipraz is currently under evaluation as a possible
    chemopreventive agent in humans. Primiano  et al. (1995) investigated
    the chemopreventive efficacy achieved by administration of
    intermittent doses of oltipraz in rats. Fischer 344 rats were treated
    with oltipraz (0.5 mmol/kg, p.o.) once weekly, twice weekly, or daily
    over a 5-week period. After the first week, all rats were gavaged with
    20 µg/kg AFB1 for 28 consecutive days. Livers were analysed 2 months
    after the last AFB1 dose, and the volume of liver occupied by
    glutathione-S-transferase (GST)-P positive foci, a presumptive marker
    of neoplasia, was observed to be decreased by >95%, >97% or >99% in
    livers of rats receiving once-, twice-weekly or daily oltipraz
    treatments, respectively. The chemopreventive actions of oltipraz have
    been associated with increases in the levels of phase 2 detoxifying
    isozymes. Accordingly, GST conjugation activity measured with
    1-chloro-2,4-dinitrobenzene as a substrate increased 1.5, 1.8 or
    2.4-fold for the once-weekly, twice-weekly or daily treatments,
    respectively, throughout a 7-day period. The authors suggested that
    the protracted pharmacodynamic actions of oltipraz on enzyme induction
    may account from the marked reduction in the hepatic burden of
    AFB1-induced putative preneoplastic tumours after intermittent dosing.
    Consequently, scheduling of intermittent dosing protocols may sustain

    efficacy while improving drug tolerance and patient compliance over
    long-term treatments. These properties of oltipraz increase its
    attractiveness for clinical chemopreventive interventions, the authors
    emphasized.

         Langouet  et al. (1995) investigated metabolism of AFB1 in
    primary human hepatocytes with or without pretreatment by oltipraz.
    AFM1, glutathione conjugates of AFB1 oxides and unchanged AFB1 were
    quantified in cultures derived from eight human liver donors.
    Parenchymal cell obtained from the three GST M1-positive livers
    metabolized AFB1 to AFM1 and to AFB1 oxides derived from the isomeric
    exo and endo-8,9-oxides, whereas no AFB1 oxides were formed in the GST
    M1-null cells. Pretreatment of the cells with 3-methylcholanthrene or
    rifampicin, inducers of CYP1A2 and CYP3A4 respectively, caused a
    significant increase in AFB1 metabolism. Although oltipraz induced GST
    A2, and to a lesser extent GST A1 and GST M1, it decreased formation
    of AFM1 and AFB1 oxides, which involves CYP1A2 and CYP1A2. Inhibition
    by oltipraz of AFB1 metabolism through a reduction in CYP1A2 and
    CYP3A4 was also shown by decreased activity of their monooxygenase
    activities toward ethooxyresorufin and nifedipine, respectively. The
    significant inhibition by oltipraz of human recombinant yeast CYP1A2
    and CYP3A4 was also shown. These results demonstrated that AFB1 oxides
    can be formed by GST M1-positive human hepatocytes only, and suggested
    that chemoprotection with oltipraz is due to an inhibition of
    activation of AFB1 in addition to a GST-dependent inactivation of the
    carcinogenic  exo-epoxide.

         AFB1-induced carcinogenesis has been shown to be both inhibited
    and promoted by indole-3-carbinol (I3C), found in cruciferous
    vegetables. Stresser  et al. (1994a) examined the influence of
    dietary treatment with I3C and the well-known Ah receptor agonist
     beta-naphthoflavone (BNF) on the relative levels of different
    cytochrome P-450 (CYP) isoforms known to metabolize AFB1 in male
    Fischer 344 rats. After 7 days of feeding 0.3% I3C or 0.04% BNF alone
    or in combination, the relative levels of hepatic CYP1A1, 1A2, 2B1/2
    2C11 and 3A were assessed by laser densitometry of Western blots. Both
    diets containing I3C markedly increased band densities of CYP1A1, 1A2,
    and 3A1/2 with less effects on 2B1/2 and no effect on CYP2C11. BNF
    also strongly increased band densities of CYP2C11, but had no effect
    on CYP2C11 or 3A1/2, and repressed CYP2C11. In addition the  in 
     vitro hepatic microsomal metabolism of AFB1 was examined at 16, 124,
    and 512 TM substrate levels. The authors' results suggested that BNF
    inhibits AFB1 carcinogenesis, in part by enhancing net production of
    less toxic hydroxylated metabolites of AFB1, as a result of elevated
    levels of P450, and that I3C may share this mechanism. However, other
    mechanisms, such as direct inhibition of P450 bioactivation by I3C
    oligomers, or induction of phase II enzymes, also appeared to
    contribute.

         Stresser  et al. (1994a) also examined the influence of I3C and
    BNF on the AFB1 glutathione detoxication pathway and AFB1-DNA
    induction in rat liver. After 7 days of feeding approximately equally
    inhibitory doses of I3C (0.2%) or BNF (0.04%) alone or in combination,

    male Fischer 344 rats were administered [3H]AFB1 (0.5 mg/kg, 480
    TCi/kg) i.p. and killed 2 hours later. All three diets inhibited 
     in vivo AFB1-DNA adduction. Using an improved HPLC method for
    separation of the two isomeric forms of AFB1 8,9-epoxide-glutathione,
    both I3C diets were shown to induce GST activities strongly toward
    AFB1  exo-epoxide, whereas BNF alone induced activity weakly. Data
    suggest that enhanced detoxication of AFB1 via increased glutathione
    conjugation efficiency, as a result of elevated levels of the Yc2 GST
    subunit, is one mechanism that contributes to a protective effect of
    I3C against AFB1-induced preneoplastic lesions in the rat, and that
    this mechanism also participates to a lesser degree in protection by
    BNF.

         The role of reactive oxygen species (ROS) in AFB1-induced cell
    injury was investigated using cultured rat (male Fischer 344)
    hepatocytes (Shen  et al., 1995). Malonaldehyde (MDA) generation and
    lactate dehydrogenase (LDH) release were determined as indices of
    lipid peroxidation and cell injury, respectively. Exposure to AFB1 for
    up to 72 hours resulted in significantly elevated levels of LDH being
    released into the medium as well as the MDA generation in cultured
    hepatocytes. These effects were dose-dependent, indicating that AFB1
    was capable of inducing oxidative damages in the cell. Further, MDA
    generation and LDH release were effectively inhibited by the addition
    of the following: 1) superoxide dismutase (500 units/ml); 2) catalase
    (1500 units/ml); 3) 10 mM desferrioxamine (a specific iron chelator),
    or 4) 260 mM dimethyl sulfoxide (a hydroxyl radical scavenger). This
    evidence therefore suggests that ROS, such as superoxide radicals,
    hydroxyl radicals and hydrogen peroxides, are involved in AFB1-induced
    cell injury in cultured rat hepatocytes, the authors concluded.

         Kirby  et al. (1993) examined liver tissues from 20 liver cancer
    patients from Thailand, an area where the incidence of this tumour is
    high and where exposure to aflatoxin occurs. The expression of hepatic
    cytochrome P450s and GST was examined and this expression was compared
    to the  in vitro metabolism of AFB1. There was a >10-fold
    inter-individual variation in expression of the various P450s
    including CYP3A4 (57 fold), CYP2B6 (56-fold), and CYP2A6 (120 fold).
    Microsomal metabolism of AFB1 to AFB1 8,9-epoxide and AFQ1, the major
    metabolite produced, was statistically significantly correlated with
    CYP3A3/4 expression and, to a lesser extent, with CYP2B6 expression.
    There was a significantly reduced expression of major P450 proteins in
    microsomes from liver tumours compared to microsomes from the paired
    normal liver when analysed by Western immunoblot analysis.

         The immunoreactive expression of the major human classes of
    cytosolic GSTs  (alpha, mu and  pi) was also analysed in normal and
    tumorous liver tissue. The expression of GSTA  (alpha) and GSTM 
     (mu) class proteins was markedly decreased and GSTP  (pi) increased
    in the majority of tumour cytosols compared to normal liver. Cytosolic
    GST activity was significantly lower in liver tumours compared to
    normal liver. There was no detectable conjugation of AFB1 8,9-epoxide
    to glutathione by cytosol either from tumorous or normal liver. Thus,

    capacity of human cytosols to conjugate reactive AFB1 metabolites to
    GSH resembled AFB1-sensitive species such as rat, trout and duck
    rather than resistant species such as mouse and hamster. These data
    indicate a strong capacity of multiple forms of human hepatic P450s to
    metabolize AFB1 to both the reactive intermediate AFB1-8,9-epoxide and
    the detoxification product AFQ1. The authors suggested, that, in view
    of the lack of significant GST-mediated protection against AFB1 in
    human liver, variations in expression of hepatic P450, due either to
    genetic polymorphisms or to modulation by environmental factors, may
    be important determinants in the risk of liver cancer development in
    AFB1-exposed populations.

         Liu  et al. (1991) evaluated the functional significance of the
    glutathione transferase (GST)  mu polymorphism by measuring its
    effect on AFB1-DNA adduct formation  in vitro. Human liver cytosols
    prepared from persons having low or high glutathione transferase
    toward  trans-stilbene oxide were incubated with human liver
    microsomes, calf thymus DNA, and AFB1. AFB1-DNA binding was inhibited
    to a greater extent in high conjugators than low conjugators; the
    correlation between AFB1-DNA adduct concentrations and GST  mu 
    activity was highly statistically significant. The authors suggested
    that GST  mu plays an important role in detoxifying DNA reactive
    metabolites of AFB1, and this enzyme may be a susceptibility marker
    for AFB1-related liver cancer.

         Heinonen  et al. (1996) studied the profile of AFB1 metabolism
    and the extent of AFB1 binding to cell macromolecules in human liver
    slices under experimental conditions that would allow direct
    comparison to similar end-points in the rat, a species sensitive to
    the carcinogenic actions of AFB1. Liver slices were prepared from
    three individual human liver samples with a Krumdieck tissue slicer
    and incubated with 0.5 µM [3H]AFB1 for 2 hours. Significant
    inter-individual variations were observed in the rates of oxidative
    metabolite formation and in specific binding to cell macromolecules.
    The rates of oxidative metabolism of AFB1 to AFQ1, AFP1 and AFM1 in
    the human liver samples were similar to those previously observed in
    rat liver slices. AFB1-GSH conjugate formation was not detected in any
    of the human liver samples, and yet specific binding of AFB1 to cell
    macromolecules was considerably lower in the human liver slices
    relative to that in rat liver slices. The authors postulated that
    these studies suggest that an as yet unidentified protective pathway
    may exit in human liver. These studies support the hypothesis that
    humans do not form as much aflatoxin B1-8,9-epoxide as the rat, but
    humans do not possess GST isozymes with high specific activity toward
    the epoxide. Significant interindividual differences in AFB1
    metabolism and binding between humans suggest the presence of genetic
    and/or environmental factors that may make some humans more or less
    susceptible to AFB1.

         Gallagher  et al. (1994) studied the metabolism of AFB1 in
    microsomes derived from human lymphoblastoid cell lines expressing
    transfected CYP1A2 or CYP3A4 (cytochrome P450) and in microsomes

    prepared from human liver donors (n=4). The authors summarized their
    findings as follows. Both CYP3A4 and CYP1A2 were involved in the
    activation of AFB1 to the AFB1-8,9-epoxide; 1A2 appeared to have a
    higher affinity for AFB1 and produced a higher ratio of activation
    (AFB1-8,9-epoxide) to detoxification (AFM1) products, relative to 3A4.
    3A4 may be expressed in human liver at a much higher level than 1A2,
    such that in some individuals, 3A4 may be the predominant source of
    AFB1-8,9-epoxide at low substrate concentrations, even though 3A4
    produces AFQ1 predominantly. Such differences in the apparent kinetics
    of these two P450s toward AFB1 indicate that the most important
    determinant of individual susceptibility to AFB may well be the level
    of expression of 1A2. Individuals with relatively high 1A2 expression
    may be at particular risk for AFB1-induced DNA damage, since human
    GSTs are relatively ineffective in detoxifying AFB1-8,9-epoxide.
    Inhibition of 1A2 may prove to be an effective means of
    chemointervention in AFB1-exposed populations. Of course,  in vivo 
    human toxicity is ultimately determined by a complex set of processes.

         In experiments by Ueng  et al. (1995), human cytochromes P450
    1A2 and 3A4 were expressed in  Escherichia coli, purified, and used
    in reconstituted oxidation systems. Relatively high catalytic
    activities were obtained with such a system for AFB1
    3 alpha-hydroxylation and 8,9-epoxidation. P450 3A4 was more active
    than P450 1A2 in forming genotoxic AFB1 oxidation products; P450 3A4
    formed AFQ1 and the  exo-8,9-epoxide; P450 1A2 formed AFB1, some
    AFQ1, and both the  exo- and  endo-8,9-epoxides. Plots of AFB1
    3 alpha-hydroxylation and 8,9-epoxidation  vs. AFB1 concentration
    were sigmoidal in both human liver microsomes and the reconstituted
    P450 3A4 system. The results were consistent, the authors
    hypothesized, with the view that P450 3A4 is a major human liver P450
    enzyme involved in AFB1 activation, although the  in vivo situation
    may be more complex due to the presence of the enzyme in the
    gastrointestinal tract.

    Guengerich  et al. (1996) have reviewed a series of studies to show
    the complexities encountered with metabolism of AFB1; the complexity
    demonstrates the difficulties in doing molecular epidemiology studies,
    even when a single chemical carcinogen has been identified. Figure 1
    shows these metabolism complexities. With all the enzymes,
    stereochemistry of the epoxide must be considered. In addition, the
    P450s both activate and detoxify AFB1, and the effect of inducing
    individual P450s is not easy to predict. P450 3A4 is expressed in the
    small intestine, the site of absorption of orally ingested AFB1, where
    the extent of detoxification is unknown. Even activation of AFB1 and
    DNA alkylation in the small intestine may be considered to be a
    detoxification process since the cells are sloughed rapidly, and
    cancers of the small intestine are very rare.

    FIGURE 1

    2.2  Toxicological studies

    2.2.1  Acute toxicity

         No additional acute toxicity studies have been reported in the
    literature since the review by Eaton & Groopman (1994).

    2.2.2  Special studies on reproductive toxicity

         Ankrah  et al. (1993) exposed ddy mice to AFB1 and AFG1 via
    their feed (4.8 ng AFG1, 0.8 ng AFB1 (or both) per kg bw per day while
     in utero. Levels of aflatoxin used were realistic relative to the
    level of human exposure currently seen in Ghanaian foods. Offspring of
    these animals (control and aflatoxin-fed) were continued on the
    respective diets received by the parent stock until sacrifice at six
    months of age. Blood obtained by cardiac puncture was used to
    determine haema-tological indices and the sera were used to determine
    glucose, triglyceride, total protein and albumin. AFG1 caused
    significant accumulation of only neutral fat in the liver, a slight
    rise in serum triglyceride and intensified hepatorenal inflammation,
    necrosis and bile duct proliferation. AFB1 caused the accumulation of
    both neutral fat and fatty acids in the liver, and was cytotoxic to
    the liver and kidney. Iron storage in the liver, haematological
    indices, serum total protein and albumin levels were not affected by
    the aflatoxins. At the level used, AFG1 was six times in excess of
    AFB1, but the latter was more severe in the observed hepatorenal
    effects.

         The authors pointed out that the mouse liver has been shown to
    metabolize aflatoxin in a manner similar in some ways to the human
    liver, although not all investigators would agree on this point. Hence
    they postulated that the action of aflatoxin on mouse organs may shed
    light on aflatoxin cytotoxicity in humans; results of this study are
    of particular relevance to population groups that ingest foods known
    to contain mainly AFG1 and to some extent AFB1.

    2.2.3  Special studies on genotoxicity

         AFB1 covalently binds to DNA and efficiently induces G to T
    trans-versions; codon 249, one site in p53, is a striking hot spot for
    AFB1 mutagenesis (Sengstag & Wurgler, 1994). Often, such mutations are
    followed by the loss of the second functional alleles of tumour
    suppressor genes, a phenomenon called loss of heterozygosity. To test
    whether mitotic recombination leading to loss of heterozygosity is
    induced by certain carcinogens, the authors genetically engineered a
     Saccharomyces cerevisiae tester strain so that it metabolized two
    important classes of carcinogens including AFB1. Human cDNAs coding
    for the cytochrome P450 (CYP) enzymes CYP1A1 or CYP1A2 in combination
    with NADPH-CYP oxidoreductase in a strain heterozygous for two
    mutations in the trp5 gene were inserted. AFB1, when activated
    intracellularly in microsomes isolated from the yeast strains
    containing either human CYP enzyme, significantly induced mitotic
    recombination. The authors concluded that activated AFB1 is a potent

    inducer of DNA recombination in  S. cerevisiae strains harbouring
    various heterologous xenobiotic-metabolizing systems.

         Young weanling Swiss albino mice were orally administered crude
    AFB1 in a dose mimicking human exposure, i.e., at 0.05 µg/kg bw per
    day for 14 weeks (Sinha & Dharmshila, 1994). Vitamin A (retinol) was
    orally administered along with the toxin at double (132 IU/kg bw per
    day) the human equivalent therapeutic dose. The authors concluded that
    vitamin A minimized the frequency of toxin-induced clastogeny in both
    mitotic and meiotic chromosomes. The decreases in sperm count, as well
    as increases in abnormality in the gross morphology of the sperm head,
    as observed upon toxin treatment, was ameliorated by the vitamin A.

         Marquez-Marquez  et al. (1993) evaluated the effects of an AFB1
    inactivating system with ammonia on the genotoxicity of AFB1 measuring
    micronucleus (MN) and sister chromatid exchange (SCE) analyses. Four
    groups of CD1 male mice were fed for 8 weeks with a special diet
    mainly composed of maize: 1) uncontaminated; 2) uncontaminated/
    inactivated; 3) contaminated/ inactivated; and 4) contaminated. The
    inactivating treatment was performed with ammonium hydroxide by
    homogeneously impregnating the grain and leaving it for 20 days in
    hermetically closed plastic bags and then heating in an oven for 24
    hours to eliminate the residual ammonia. AFB1 was quantified before
    and after inactivation. MN was evaluated at weekly intervals in
    peripheral blood; SCE was quantified in bone marrow cells at weeks 4
    and 8. The results showed that mice fed with AFB1 contaminated/
    inactivated maize had a 45% lower level of induced cytogenetic damage
    than those animals fed with AFB1 contaminated (but not inactivated)
    maize. A residual amount of AFB1 remaining after the inactivating
    treatment and the reconversion back to AFB1 in the organism may
    account for the remaining increased levels of SCE and MN.

         Marquez-Marquez  et al. (1995) evaluated the efficiency of the
    AFB1 inactivating system with ammonia, as described above, and using
    mice (male CD-1) and micronucleus (MN) and sister chromatid exchange
    (SCE) analysis. Apparently this study was the same as that published
    earlier.

         Occupational exposures to respirable dusts contaminated with the
    mycotoxin AFB1 have been associated with an increased incidence of
    upper airway tumours. To investigate this possible etiology Ball 
     et al. (1995) compared the abilities of tracheal and lung S9 from
    rabbit (male, New Zealand white), hamster (male Syrian Golden) and rat
    (male Sprague-Dawley) to activate AFB1 to mutagens in  Salmonella 
     typhimurium TA98. These species differ in airway morphology with
    respect to numbers of metabolically active non-ciliated tracheal
    epithelial cells. Tracheas from hamster and rabbit and lung from
    rabbit were active in converting AFB1 to bacterial mutagens. Tracheas
    from hamster were more efficient in activating AFB1 to mutagens than
    lung, while rabbit lung was over 4 times more active in converting
    AFB1 to mutagens than that from trachea. In all cases, AFB1 was more
    mutagenic than B[a]P. The relative capabilities of trachea to activate
    AFB1 were in agreement with the ability of cultured tracheas from

    these species to form AFB1-DNA adducts. These results demonstrate that
    AFB1 is activated more efficiently than B[a]P in the lung, and that
    the metabolic capabilities of airway epithelium to activate AFB1 are
    not predictable by airway morphology.

         A study by Shi  et al. (1995a) examined the effect of two
    selenium compounds, namely, sodium selenite and selenium-enriched
    yeast extract (SeY) on the cytotoxicity, DNA binding, and mutagenicity
    of AFB1 in cultured Chinese hamster ovary (CHO) cells. CHO cells,
    after treatment with 2 µg/ml selenite or 80 µg/ml SeY, exhibited
    increased resistance to AFB1-induced cell killing. At a concentration
    of 50 µg/ml AFB1, cell survival, measured by the clonogenicity assay,
    was increased by 21- and 10-fold in selenite- and SeY-treated cells,
    respectively. However, selenium treatment did not appear to affect
    AFB1-DNA binding. Similarly, no effect was observed on AFB1
    mutagenicity, as determined by the hypoxanthine-guanine phosphoribosyl
    transferase (HPRT) gene mutation assay. The results showed that
    selenium could effectively protect cells from AFB1 cytotoxicity in
    cultured cells, but had no effect on AFB1-DNA adduct formation or
    mutagenesis. The authors suggested that there are multiple pathways of
    AFB1 toxicity and that selenium can modulate AFB1-induced cell killing
    independent of its genotoxicity.

         Rats and mice differ markedly in sensitivity to AFB1
    hepatocarcinogenicity, the former being sensitive and the latter
    resistant. The purpose of this study was to determine whether the
    formation of AFB1-albumin (AF-alb) adducts was related to the
    induction of cytogenetic changes  in vivo as a step to understanding
    whether such markers of exposure may be informative with respect to
    genetic alterations important in the carcinogenic process (Anwar 
     et al., 1994). The comparison was made at two levels: between
    species and between individuals within a species. Animals (male Helwan
    Wistar albino rats and Swiss albino mice) were treated with single
    doses of different concentrations of AFB1 between 0.01 and 1.0 µg
    AFB1/g bw. The frequency of chromosomal aberrations and micronuclei in
    the bone marrow was measured and compared to the level of AFB1 bound
    covalently to albumin in the peripheral blood. Both chromosomal
    aberrations and micronuclei were significantly increased in treated
    rats compared to the control group at doses above 0.1 µg/g. In
    contrast, in mice, a slight increase in chromosome aberrations was
    seen in the highest dose group (1.0 µg/g), but no increase in
    micronuclei was observed at any of the doses. The level of chromosomal
    aberrations was about 10 times higher in rats than in mice at the
    highest dose of AFB1. AFB1-albumin increased linearly with dose of
    AFB1, and there were strong statistically significant correlations at
    the individual rat level with both chromosomal aberrations.

         The level of AF-alb adducts was higher for a given dose in rats
    than mice, as has been seen for the level of liver DNA adducts in the
    two species. The metabolic basis of these differences has been
    investigated and has been shown to be associated with the expression
    of a specific glutathione-S-transferase isoenzyme in mice, which
    efficiently conjugates the AFB1-epoxide to glutathione.

         In rats, the level of AF-alb adducts was strongly correlated with
    the frequency of both micronuclei and chromosomal aberrations in the
    bone marrow. An increase in adduct levels was seen with exposures as
    low as a few ng AFB1/g bw, whereas the genetic alterations were only
    increased above control levels at doses around 0.1 µg/g.

         The authors suggested two considerations for interpretation of
    the present studies. First, the cells in which the micronuclei and
    chromosomal aberrations were examined are not the target cells for
    AFB1 hepatocarcinogenesis, and second, that this type of genetic
    marker is relatively non-specific. Thus, the genetic alterations being
    measured are not directly relevant to the carcinogenic process; this
    limitation may be overcome as sensitive molecular techniques are
    developed to measure mutation induced by aflatoxin in specific gene
    sequences in somatic cells (See Aguilar  et al., 1993). Recent
    studies by these authors (Wild  et al., in preparation) suggest that
    AF-alb adducts reflect the differing species sensitivity to AFB1
    carcinogenesis. This peripheral blood marker could be an indicator of
    risk of liver cancer development in addition to being a marker of
    exposure, the authors suggest, as has been further supported by the
    study of Ross  et al. (1992), in which the level of AFB1-N7 guanine
    adduct in the urine was related to the subsequent risk of developing
    hepatocellular carcinoma in a Chinese cohort. This study will be
    discussed in section 2.2.10.

    2.2.4  Special studies on immunosuppression

         The immunosuppressive potential of AFB1 was evaluated in growing
    rats (Raisuddin  et al., 1993). The weanling rats (species
    unspecified) were sub-chronically exposed to 60, 300 or 600 µg AFB1/kg
    bw for four weeks on alternate days by oral feeding. Various
    parameters of cell-mediated immunity (CMI) and humoral immunity were
    assessed in control and treated animals. CMI was evaluated by
    measuring delayed type of hypersensitivity (DTH) response and humoral
    immunity was measured by plaque forming (PFC) assay. The
    lympho-proliferative response assay for T- and B-cells was also
    performed. It was observed that AFB1 selectively suppressed
    cell-mediated immunity in growing rats. AFB1 suppressed CMI at the 300
    and 600 µg dose levels only as measured by DTH response assay. The
    authors concluded that continuous low level exposure of aflatoxin to
    the growing host may enhance its susceptibility to infection and
    tumorigenesis.

         Jakab  et al. (1994) conducted experiments to demonstrate the
    immunosuppressive effects of AFB1 ingestion, in this case respiratory
    tract exposure to AFB1. Rats (male Fischer 344) and mice (female
    Swiss) were exposed either by aerosol inhalation or intratrachael
    instillation to AFB1. Nose-only inhalation exposure of rats to AFB1
    aerosols suppressed alveolar macrophage (AM) phagocytosis at an
    estimated dose of 16.8 µg/kg with the effect of persisting for
    approximately 2 weeks. To determine whether another mode of
    respiratory tract exposure, intratrachael instillation, reflected
    inhalation exposure, animals were treated with increasing

    concentrations of AFB1, which also suppressed AM phagocytosis in a
    dose-related manner, albeit at doses at least an order of magnitude
    more than that obtained by aerosol inhalation. Intratrachael
    administration of AFB1 also suppressed the release of tumour necrosis
    factor-alpha from AMs and impaired systemic innate and acquired immune
    defences as shown, respectively, by suppression of peritoneal
    macrophage phagocytosis and the primary splenic antibody response. The
    authors concluded that experimental respiratory tract exposure to AFB1
    suppressed pulmonary and systemic host defences; they indicated that
    inhalation exposure to AFB1 is an occupational hazard where exposure
    to AFB1-laden dust is common, such as in grain dust.

    2.2.5  Factors modifying carcinogenicity of aflatoxins

         Young adult male Fischer rats maintained on a reduced calory diet
    (60% of  ad libitum food consumption) for 6 weeks showed a decrease
    in the binding of AFB1 to hepatic or renal nuclear DNA and a reduction
    of AFB-induced hepatocellular damage (Chou  et al., 1993). Repeated
    dosing of rats with AFB1 resulted in the inhibition of hepatic and
    renal DNA synthesis as measured by [3H]thymidine incorporation.
    However, the rate of DNA synthesis was greater in  ad libitum (AL)
    rats than in calorically restricted (CR) animals. Three days after
    AFB1 dosing, the rate of DNA synthesis had recovered to the control
    level. Cell cycle analyses measured by a flow cytometric method on
    kidney cells of both AL and CR rats showed that there were no
    significant changes in cell populations in the S phase between these
    two groups of rats. AFB1 inhibited the cell proliferation by 33% (on
    average). The restoration of the cell proliferation in kidney cells
    was found on the third day after AFB1 dosing. The rate of regenerative
    cell proliferation was found to be slightly greater in AL rats than in
    CR animals. The AFB1-induced regenerative DNA synthesis in both liver
    and kidney was retarded by CR.

         Youngman & Campbell (1992) demonstrated that with young Fischer
    344 rats the post-initiation development of AFB1-induced 
     gamma-glutamyltranspeptidase-positive (GGT+) hepatic foci was
    markedly inhibited by low protein feeding, even though the energy
    intake was greater. These investigators also studied this dietary
    effect upon the development of hepatic tumours and the correlation of
    foci development with tumour development. Following AFB1 dosing (15
    daily doses of 0.3 mg/kg each), animals were fed diets containing 6,
    14 or 22% casein (5.2, 12.2 or 19.1% protein) for 6, 12, 40, 58 or 100
    weeks. Foci at 12 weeks and tumours at 40, 58 and 100 weeks developed
    dose-dependently to protein intake. Foci development, tumour
    incidence, tumour size and the number of tumours per animal were
    markedly reduced, while the time to tumour emergence was increased
    with low-protein feeding. Non-hepatic tumour incidence also was lower
    in the animals fed the lowest protein diet. Foci development indices
    (foci number, per cent liver volume occupied) were highly correlated
    with tumour incidence at 58 and 100 weeks (r = 0.90-1.00). Tumour and
    foci inhibition occurred in spite of the greater energy intake.

         Previous results from a large ecological study in 65 rural
    counties in China suggested that primary liver cancer in humans
    primarily is associated with chronic HBV infection, coupled with
    nutritional factors (e.g., animal protein) that elevate plasma
    cholesterol level and encourage cancer growth (Campbell  et al., 
    1990). To test this hypothesis, the authors investigated the effect of
    dietary animal protein on tumour development in HBV transgenic mice.
    Male F2 offspring of 50-4 HBV transgenic mice were randomly assigned
    to 6, 14 and 22% dietary casein. Serum was collected from the retro-
    orbital vein and was analysed for the level of hepatitis B virus
    antigen (HBsAg), the products of the S-transgene. The increases from
    baseline in S-gene product observed for the normal protein animal
    (22%) at 3 months was inhibited in the mid- and low-protein animals by
    42% and 72%, respectively, with a highly significant dose-response
    relationship (P<0.001). Serum glutamic-pyruvic transaminase activity
    was not affected by diet treatment. The authors concluded that their
    results strongly suggest that dietary casein controls, in a
    dose-response manner, S-transgene expression in these experimental
    animals.

         Hasler  et al. (1994) fed Fischer 344 rats a low-fat high
    carbohydrate (HC) diet, an isocaloric fat-containing (IC) diet, a
    hypercaloric fat-containing (HF) diet or a commercial rodent chow. The
    effects of these diets were studied on the binding of AFB1 to
    exogenous DNA and on the activities of hepatic glutathione
    transferases (GSTs), cytochromes 2B1 and 1A1. Microsome-mediated
    binding of [3H]AFB1 to exogenous DNA was significantly lower in the
    HC rats than in the chow- and IC-fed rats. No significant differences
    were noted between HF and either HC or IC rats. There was no
    significant difference in hepatic GST activity of rats fed the
    different diets. The authors suggested that high carbohydrate/low fat
    diets may reduce microsome-mediated epoxidation of AFB1 to a larger
    extent than high-fat diets. In general, high-fat diets increased
    cytochrome 1A1 and 2B1 activities relative to chow and
    high-carbohydrate diet. This suggested greater detoxification of AFB1,
    thus reducing the amount of AFB1 available for hepatic macromolecular
    binding, the authors concluded.

         An excellent review by Massey  et al. (1995) covers the
    biochemical and molecular aspects of mammalian susceptibility to AFB1
    carcinogenicity. Important considerations include: 1) different
    mechanisms for bioactivation of AFB1 to its ultimate carcinogenic
    epoxide metabolite; 2) the balance between bioactivation to and
    detoxification of the epoxide; 3) the interaction of AFB1 epoxide with
    DNA and the mutational events leading to neoplastic transformation; 4)
    the role of cyto-toxicity in AFB1 carcinogenesis; 5) the significance
    of non-epoxide metabolites in toxicity; and 6) the contribution of
    mycotoxin-unrelated disease processes.


    2.2.6  Special studies on covalent binding of aflatoxin residues with
    nucleic acids and proteins

         Shi  et al. (1994) studied the effect of selenium on AFB1-DNA
    binding and adduct formation. Male Fischer 344 rats, fed with up to 8
    mg/litre sodium selenite in drinking-water for 8 weeks, were given a
    single i.p. dose of AFB1. The rats were killed 24 hours later and the
    amount of AFB1 bound to hepatic DNA and the amount of DNA adducts
    formed were determined. Selenium pretreatment resulted in a
    dose-dependent inhibition of AFB1-DNA binding as well as adduct
    formation. This was accompanied by an increase of reduced glutathione
    (GSH) in the liver of selenium-treated animals. These results
    suggested that selenium could effectively inhibit AFB1-induced DNA
    damage, which may be partially responsible for its anticarcinogenic
    effect against AFB1.

         Choy (1993) has reviewed the dose-response induction of DNA
    adducts by AFB1 and its implication to quantitative cancer risk
    assessment. Dose-response curves of DNA adduct formation after
    ingestion or injection treatments in the rat were reviewed; a linear
    dose-response relationship was observed in both injection and
    ingestion studies at low doses. The author concluded that this
    observation is consistent with the assumption of the linear
    dose-response risk assessment model for genotoxic agents and justifies
    the use of this model for quantitative cancer risk assessment for
    aflatoxins. The author also concluded that although AFB1-DNA adducts
    generated in rats, mice and humans reflect the "molecular dose" and
    DNA damage in the target organ, bypassing the need for interspecies
    pharmacokinetic dose adjustments, it is not possible to extrapolate
    from rodents to humans at this time because human DNA adduct data are
    incomplete.

    2.2.7  Special studies on glucose tolerance

         Glyoxalase-1 activity plays an important role in glucose
    metabolism and has been reported to be depressed in mice fed low
    levels of AFB1 (Ankrah, 1995). In the present study, glyoxalase-1
    activity, glucose tolerance and pancreatic beta cell sensitivity were
    examined in mice (male and female ddy) fed 0.045 ng AFB1 plus 0.450 ng
    ABG1/g feed prenatally and for 6 months after birth. After glucose
    challenge, the ratios between 0- and 2-hour serum glucose levels were
    significantly higher than controls, indicating an increase in
    tolerance of glucose in the aflatoxin-fed mice with lower glyoxalase-1
    activity. Pancreatic beta cell sensitivity to stimulation by
    tolbutamide was similar in both groups. However, liver malonic
    dialdehyde was significantly higher in the aflatoxin-fed mice,
    suggesting that the altered tolerance for glucose in the aflatoxin-fed
    mice might be a consequence of aflatoxin-mediated peroxidative action
    in the liver, the authors suggested.

    2.2.8  Special studies on effect of ammoniation of AFB1 in
    contaminated cottonseed

         The effectiveness of ammonia in inactivating aflatoxin in
    contaminated cottonseed was investigated (Bailey  et al., 1994). Two
    aflatoxin-contaminated cottonseed lots were treated separately using
    atmospheric pressure, ambient temperature ammoniation procedure (APAT)
    or a high pressure, high temperature ammoniation procedure (HPHT), and
    incorporated into dairy cow rations. Isocalorific diets containing 25%
    defatted, dried milk from cows fed aflatoxin-contaminated cottonseed
    without or with APAT or HPHT treatment, or an aflatoxin-free human
    grade commercial milk powder, were then fed for 12 months to rainbow
    trout  (Oncorhynchus mykiss). AFM1 concentrations in milk powders
    without and with seed treatment were: APAT, 85 and <0.05 µg/kg; HPHT,
    32 and <0.05 µg/kg. In the APAT experiment, trout consuming the diet
    containing milk from cows fed the aflatoxin-contaminated cottonseed
    had a 42% incidence of hepatic tumours; APAT cottonseed treatment
    reduced this to 2.5%. Positive controls were included to demonstrate
    trout responsiveness. AFB1 fed continuously for 12 months at 4 µg/kg
    resulted in a 34% tumour incidence, whereas positive controls fed 20
    µg AFB1/kg, 80 µg AFM1/kg, or 800 µg AFM1/kg for 2 weeks and killed 9
    months later had a 37, 5.7 and 50% incidence of tumours, respectively.

         The authors concluded that APAT ammonia treatment of
    aflatoxin-contaminated dairy cattle cottonseed feedstock abolished the
    detectable transfer of AFM1 or AFB1 into milk powder, and greatly
    reduced the carcinogenic risk posed by any carry-over of aflatoxins or
    their derivatives into milk.

         In addition, the results confirm AFM1 to be a lower level
    hepatocarcinogen in comparison with AFB1 in the trout carcinogenicity
    assay. In the separate HPHT experiment, no tumours were observed in
    the livers of trout fed diets containing milk from either the
    ammonia-treated or untreated source, or the control diet containing 8
    µg AFM1/kg. Positive controls fed 64 µg AFB1/kg for 2 weeks exhibited
    a 29% tumour incidence 12 months later. Thus in this experiment,
    neither AFM1 at 8 µg/kg nor any HPHT-derived aflatoxin derivatives
    that might have been carried over into milk represented a detectably
    carcinogenic hazard to trout, the authors conclude.

    2.2.9  Special studies on aflatoxin and hepatitis B virus infection in
    woodchucks, ducks, ground squirrels and tree shrews

         Interactive hepadnavirus and chemical hepatocarcinogenesis has
    been studied in woodchucks inoculated as newborns with woodchuck
    hepatitis virus (WHV), which is closely related to the human hepatitis
    B virus (Bannasch  et al., 1995). When the woodchucks reached 12
    months of age, AFB1 was administered in the diet at dose levels of 40
    µg/kg bw per day for 4 months and subsequently 20 µg/kg bw per day
    (5 days/week) for a lifetime. WHV DNA was demonstrated by Southern
    blot hybridization in the serum and by PCR in the serum and/or liver
    tissue. The histomorphology and cytomorphology of the liver were

    investigated by light and electron microscopy. WHV carriers with and
    without AFB1 treatment developed a high incidence of preneoplastic
    foci or altered hepatocytes, hepatocellular adenoma and hepatocellular
    carcinomas that appeared 6-26 months after the beginning of the
    combination experiment. Administration of AFB1 to WHV carriers
    resulted in a significantly earlier appearance of hepatocellular
    neoplasms and a higher incidence of hepatocellular carcinomas compared
    to WHV carriers not treated with AFB1. Neither hepatocellular adenomas
    nor carcinomas (but preneoplastic foci of altered hepatocytes) were
    detected in woodchucks receiving AFB1 alone, and no preneoplastic or
    neoplastic lesions were found in untreated controls.

         The authors pointed out that these results provide conclusive
    evidence of a synergistic hepatocarcinogenic effect of hepadnaviral
    infection and dietary AFB1. The striking similarities in altered
    cellular phenotypes of preneoplastic hepatic foci similarities in
    altered cellular phenotypes of preneoplastic hepatic foci emerging
    after both hepadnaviral infection and exposure to AFB1 suggested
    closely related underlying molecular mechanisms that may be mainly
    responsible for the synergistic hepatocarcinogenic effect of these
    oncogenic agents.

         In addition, the authors observed that the decisive role of the
    chronic WHV infection for hepatocarcinogenesis became particularly
    evident in those animals that seroconverted after 1 year and showed
    neither a chronic active hepatitis nor hepatocellular neoplasms, no
    matter when AFB1 was given. From this observation, the authors
    concluded that chronic hepatitis is not an absolutely necessary
    condition for the development of HCC in WHV carriers.

         To determine whether p53 mutations are common to HCCs of hosts
    infected with related hepadnaviruses with and without treatment with
    aflatoxin, Rivkina  et al. (1994) studied the occurrence of mutations
    in the p53 gene in HCCs of ground squirrels and woodchucks with a
    history of infection with ground squirrel hepatitis virus (GSHV) and
    woodchuck hepatitis virus (WHV). Sequencing of wild type p53 genes
    from ground squirrels and woodchucks revealed remarkable homology
    between the two species; using direct polymerase chain reaction
    sequencing, the investigators analysed the state of the p53 gene in 20
    HCCs from ground squirrels (2 uninfected, 7 with past and 11 with
    ongoing infection with GSHV) and in 11 HCCs from woodchucks
    persistently infected with WHV. Five GSHV carrier and two uninfected
    ground squirrels received i.p. administration of AFB1. Only one
    mutation - located in codon 176 of exon 5 - in the p53 gene of the
    tested animals was detected and that in a GSHV-positive ground
    squirrel treated with AFB1. The investigators suggested that in view
    of the considerably lower apparent rate of mutations in comparison to
    human HCCs, other etiological factors may be of greater significance
    in the development of HCC in ground squirrels and woodchucks.

         The unique mutation from G to T at the third base in codon 249
    observed in human hepatocellular carcinoma has been suggested to be

    linked to aflatoxin exposure. Imazeki  et al. (1995) studied six
    ducks with HCC, three of which were infected with duck hepatitis B
    virus and five of which were fed a diet containing AFB1 for 1-2 years.
    Liver tissues were analysed for the presence of point mutations at
    this codon of the p53 gene by polymerase chain reaction and direct
    nucleotide sequencing. None of the six ducks with HCC showed the
    change at this codon regardless of duck hepatitis B virus infection.
    Integration of duck hepatitis B virus DNA into the host genome was not
    observed in two ducks that were chronically infected with the virus
    and treated with AFB1. A third duck from Qitong Province in China,
    where HBV and AFB1 are risk factors for HCC in humans, did show viral
    integration. This suggested, in the opinion of the authors, that AFB1
    itself might not be involved in the unique mutation at codon 249 in
    hepatocarcinogenesis, or that other factors coincident with aflatoxin
    may be responsible for this unique mutation.

         Cova  et al. (1996) used a Pekin duck model to examine the
    effect of congenital duck hepatitis B virus (DHBV) infection and AFB1
    exposure in the induction and development of liver cancer. The study
    of the two major risk factors in the development of HCC, i.e.,
    persistent hepatitis virus infection and exposure to dietary
    aflatoxins, has been hampered by lack of an animal model, and these
    experiments were undertaken to this end. AFB1 was administered to
    groups of 13 DHBV infected or non-infected ducks at two doses (0.08
    and 0.02 mg/kg) by i.p. injection once a week from the third month
    posthatch until they were sacrificed 2.3 years later. Two control
    groups of ducks not treated with AFB1 (one of which was infected with
    DHBV) were observed for the same period. Higher mortality was observed
    in ducks infected with DHBV and treated with AFB1 compared to
    non-infected ducks treated with AFB1 and other control ducks. In the
    groups of non-infected ducks treated with high and low doses of AFB1,
    liver tumours developed in 3 out of 10 and 2 of 10 ducks,
    respectively. In infected ducks treated with the high dose, 3 of 6
    showed liver tumours; there were none with the low dose of AFB1. No
    liver tumours were observed in the two control groups. Ducks infected
    with DHBV and treated with AFB1 showed more pronounced periportal
    inflammatory change, fibrosis and focal necrosis compared to other
    groups. All DHBV carrier ducks showed persistent viraemia throughout
    the observation period. An increase of viral DNA titres in livers and
    sera of AFB1-treated animals compared to infected controls was
    frequently observed.

         No DHBV DNA integration into the host genome was observed,
    although in one hepatocellular carcinoma from an AFB1-treated duck, an
    accumulation of viral multimer DNA forms was detected. Unlike the
    situation observed for woodchuck and ground squirrel, HCC has rarely
    been associated with DHBV infection or integration of viral DNA in the
    duck. HCC has to date been reported only in Chinese ducks from
    Chi-tung County, not always associated with detectable virus, and with
    only a single reported case of integrated DHBV. Colonies of
    DHBV-infected ducks from other parts of the world do not develop HCC.
    Prevalence of liver tumours observed in Chi-tung County ducks

    reportedly correlated with the AFB1 food contamination and with the
    incidence of primary liver cancer in these areas.

         The authors observed a lower level of AFB1 binding to liver DNA
    and plasma protein in the DHBV-infected ducks compared to non-infected
    ducks after a single dose of AFB1; this finding appeared inconsistent
    with the hypothesis that DHBV infection could increase the metabolic
    activation of AFB1, as has been observed in woodchucks and in some
    human data. The investigators noted that their observations were made
    at a single dose at a single exposure and using one specific age of
    ducks; all of these factors could have influenced the AFB1-DNA adduct
    level.

         Yan  et al. (1996) reported the successful establishment of an
    animal model in tree shrews  (Tupaia belangeri chinensis) captured
    from the wild and experimentally infected with human hepatitis B
    virus. In animals exposed to AFB1 and infected with HBV, the incidence
    of HCC was significantly higher than in the animals solely infected
    with HBV or exposed to AFB1. AFB1-exposed animals received a total
    dosage of 15-16 mg/animal. No HCC or precancerous lesions were found
    in the controls that were neither HBV-infected nor AFB1-exposed. HBV
    DNA and the protein it encodes were detected in the cancer cells
    and/or the surrounding hepatocytes. Integration of HBV DNA into the
    host liver genome was found during hepatocarcinogenesis among the
    animals infected by HBV.

         The investigators pointed out that the cumulative dose of AFB
    used in their experiment was much lower than those (24-66 mg/animal)
    used in previous experiments on tree shrews where HCC was seen. This
    suggested that HBV infection might increase the hepatocarcinogenic
    effect of AFB1. The occurrence of precancerous GT foci in the tree
    shrews exposed only to AFB1 was much more frequent than in those
    infected by HBV alone. Among the animals exposed to the same dose of
    AFB1, the  gamma-glutamyltranspeptidase (GGT) foci were more numerous
    and larger in HBV-infected than in uninfected animals during the late
    state (after the 83rd week), but not at the early state. This suggests
    that, although both AFB1 and HBV may induce GGT foci and have a
    synergistic effect, the effect of HBV is weaker and slower than that
    of AFB1.

    2.2.10  Observations in humans

    2.2.10.1  Biomarkers of aflatoxin exposure

         A key issue in the use of aflatoxin biomarkers is whether the
    ratio of AFB1-albumin adduct to DNA adduct suggested in rodent
    experiments is the same in humans (Wild  et al., 1996). Direct
    evidence for this is not available due to the limitation of measuring
    DNA adducts in human liver. The human populations where AFB1 intake
    vs. AFB1-albumin adduct relationship has been examined are populations
    in which aflatoxin intake is relatively high. Examination of the
    relationship in low-exposure populations would be important to test
    whether the linear dose-response relationships seen in rats at

    exposures as low as 1 ng AFB1/kg bw are also observed in man. There
    are some data discussed in Wild  et al. (1996) indicating that the
    amounts of AFB1 intake bound to albumin are similar for rats and
    humans; assuming that the majority of AFB1-DNA adducts are formed in
    liver, then the initial ratio between the serum albumin and liver DNA
    adducts would be expected also to be similar in humans and Fischer
    rats. However, the capacity of human intestine to metabolize AFB1 must
    be further explored to clarify this point.

         It would appear that the AFB1-albumin adduct in peripheral blood
    is a reliable marker of AFB1-DNA adducts in the liver in rodents (Wild
     et al., 1996). Both of these parameters are at least qualitatively
    associated with species susceptibility to AFB1 hepatocarcinogenesis.
    Cross-species extrapolation to man suggests that the amount of
    AFB1-albumin formed for a given exposure more closely approximates
    that in the sensitive species rather than the resistant, and indicates
    that the Fischer rat may be a more appropriate model than the mouse
    for molecular dosimetry studies of AFB1 when, for example, validating
    approaches for chemoprevention studies.

         However, carcinogenesis is a multistep process; as pointed out in
    Wild  et al. (1996) AFB1-albumin adduct is acting as a surrogate
    marker only for one critical step, the formation of AFB1-DNA adducts
    in the target cell. The relationship between this marker and the
    genetic consequences of exposure as well as the quantitative
    association with HCC risk in man remain to be determined. In addition,
    HBV and possibly HCC infection, are major risk factors. The
    availability of more reliable markers of biologically effective dose
    of AFB1 should contribute to improving attempts to understand the
    mechanism of interaction between these two and other risk factors.

    2.2.10.2  Mutations in p53 tumour-suppressor gene in human
    hepatocellular carcinoma

         Molecular epidemiological studies have found that a G to T
    mis-sense mutation at the third base of codon 249 of the p53 gene,
    effecting an arginine to serine substitution, occurs in high frequency
    (up to 67%) in human liver tumours in regions with high risk of
    aflatoxin exposure, but not in regions of low aflatoxin exposure
    (Ozturk, 1991). Hsieh & Atkinson (1995) performed experiments to
    confirm this, using liver tissue from liver cancer patients in Taiwan
    and Japan. This was analysed for the presence of aflatoxin-DNA adducts
    (ADA) as a marker for aflatoxin exposure and an AGG to AGT
    transversion at codon 249 of the p53 gene. Ten per cent of samples
    containing ADA, indicating definite exposure of the subjects to
    aflatoxin, were found to harbour the codon 249 mutation, whereas 18%
    of the samples with no detectable adducts also contained the mutation.
    Since the presence of ADA in the liver tissue samples is an indication
    of definite recent exposure of the liver cancer patients to aflatoxin,
    these data indicated that the codon 249 mutation is not a high
    frequency event associated with recent aflatoxin exposure. If recent
    exposure to aflatoxin is indeed involved in the late stage
    hepatocarcinogenesis, these data suggested that it is through some

    mechanism other than codon 249 mutation. If either mutation at codon
    249 of the p53 gene or exposure to aflatoxin is involved in earlier
    stages of hepatocarcinogenesis, whether codon 249 of the p53 gene is a
    "hot spot" for aflatoxin attack could be shown by the present
    experiment, the authors concluded.

         The tumour suppressor p53 exerts important protective functions
    towards DNA-damaging agents (Gerbes & Caselmann, 1993). Its
    inactivation by allelic deletions or point mutations within the p53
    gene as well as complex formation of wildtype p53 with cellular or
    viral proteins is a common and crucial event in carcinogenesis.
    Mutations increase the half-life of the p53 protein allowing the
    immunohistochemical detection and anti-p53 antibody formation.
    Distinct  G to  T mutations in codon 249 leading to a substitution
    of the basic amino acid arginine by the neutral amino acid serine are
    responsible for the altered functionality of the mutation gene product
    and were originally identified in 8 of 16 Chinese and 5 of 10 African
    HCC patients, both groups living in regions with traditionally high
    exposure to mycotoxins. None of these mutations was detectable in 20
    patients with HCCs recently studied in the United Kingdom; only two of
    13 HCC DNAs from Germany displayed a C to T and a T to A transversion,
    respectively, in codons 257 or 273, but not in codon 249. An average
    p53 gene nutation rate of 25% is currently assumed for high-AFB1
    exposure regions; this is double the rate observed in low-AFB1
    exposure countries. The authors concluded that although many HCC
    patients displaying P53 mutations also suffer from HBV infection,
    which itself can lead to rearrangement of P53 coding regions or induce
    the synthesis of viral proteins possibly interacting with p53, the
    specific G to T transversion within codon 249 of the P53 gene seems to
    directly reflect the extent of AFB1 exposure and is not pathognomonic
    for all HCCs.

         Yap  et al. (1993) analysed 24 HCC liver biopsy samples from
    patients in Durban, South Africa, for p53 mutations and HBV infection.
    One patient was negative for HBV (Hbsag, anti-HBcAb, anti-HBsAb) and
    possessed the p53 249 mutation (which results in an arginine to serine
    substitution). The authors suggested that HBV infection or integration
    increases the likelihood of, but is not essential for, this p53
    "hot-spot" mutation in HCC. AFB1 or other as yet unidentified
    environmental carcinogens and cofactors are implicated; the mechanisms
    by which cells exposed to these agents acquire such a specific
    mutation and then expand clonally remains to be elucidated.

         The subject of the mutation at codon 249 of the p53 tumour
    suppressor gene has continued to be the subject of much research.
    Fifty-eight per cent of HCCs from Quidong, China, contain this
    mutation which is rarely seen in HCCs from Western countries (Aguilar
     et al., 1994). The population of Qidong is exposed to high levels of
    AFB1 and this toxin has been shown to induce the same mutation in
    cultured human HCC cells. To investigate the role of AFB1 and of these
    p53 mutations in hepatocarcinogenesis, normal liver samples from the
    USA (5), Thailand (3), and Qidong (14) (where AFB1 exposures are
    negligible, low, and high, respectively), were examined for p53

    mutations. The frequency of the AGG to AGT mutation at codon 249
    paralleled the level of AFB1 exposure, which supports the hypothesis
    that this toxin has a causative - and probably early - role in
    hepatocarcinogenesis. However, a role for other carcinogens cannot be
    ruled out, the authors point out; bulky heterocyclic amines in cooked
    foods and oxidants released by inflammatory leukocytes possess the
    same specificity for G to T transversion and HBV infection is
    associated with inflammation. All of the liver samples from Qidong and
    Thailand were from HBV-infected individuals.

         The presence of elevated frequencies of codon 249 AGT mutations
    in the non-malignant tissue of HCC patents from Qidong suggested that
    the mutagenic event occurred early in hepatocarcinogenesis. In
    contrast, p53 mutations in HCCs from geographic areas with low
    exposure to AFB1 could be late events. For example, p53 mutations have
    been observed more frequently in large tumours and in advanced grades
    of malignancy in HCCs from Japan. In other organs, such as the colon
    and the bladder, p53 mutations are thought to occur late in
    tumorigenesis. However, the methods used in previous work may not have
    been sensitive enough to detect mutations at early stages of
    tumorigenesis.

         Fujimoto  et al. (1994) tested the hypothesis that exposure to
    AFB1 alone or coincident with other environmental carcinogens might be
    associated with allelic losses occurring during development of human
    hepatocarcinogenesis (HCC) in China. The HCCs were obtained from two
    different areas in China: Qidong, where exposure to HBV and AFB1 is
    high; and Beijing, where exposure to HBV is high, but that to AFB1 is
    low. Tumours were analysed for mutations in the p53 gene and loss of
    heterozygosity for the p53, Rb and APC genes and at marker loci on
    chromosomes 4, 13 and 16. The data indicated that mutation and/or loss
    of heterozygosity in the p53 gene, independent of the 249 mutation,
    played a critical role in the development of HBV-associated HCCs in
    China. The authors postulated that different mechanisms appeared to be
    responsible for the development of HCC in Beijing and may have
    resulted from exposure to unknown environmental carcinogens or a
    different subtype of HBV. Also, the results demonstrated that multiple
    alterations in DNA located on different chromosomes may be involved in
    the development of HCC.

         Additional support for the etiological role of AFB1 in
    hepatocarcinogenesis in regions of the world with AFB1-contaminated
    food has come from the studies of Aguilar  et al. (1993). These
    investigators studied the mutagenesis of codons 247-250 of p53 by rat
    liver microsome-activated AFB1 in human HCC cells HepG2 by restriction
    fragment length polymorphism/polymerase chain reaction genotypic
    analysis. AFB1 preferentially induced the transversion of G to T in
    the third position of codon 249, and also induced G to T and C to A
    transversions into adjacent codons, albeit at lower frequencies. Since
    the latter mutations are not observed in HCC, the investigators
    concluded that both mutability on the DNA level and altered function
    of the mutant serine 249 p53 protein are responsible for the observed
    mutational hot spot in p53 HCC from AFB1-contaminated areas. The fact

    that this mutation is only rarely found in HCC from low AFB1 regions
    indicates that it is not a prerequisite for hepatocarcinogenesis;
    perhaps HBV and the mutant serine 240 p53 protein play a synergistic
    role.

         In a later study, Aguilar  et al. (1995) examined normal liver
    samples from the USA, Thailand and Qidong, where AFB1 exposures are
    negligible, low and high, respectively, for p53 mutations. The
    frequency of the AGG to AGT mutation at codon 249 paralleled the level
    of AFB1 exposure, which, according to the authors, provides additional
    support for the hypothesis that this toxin has a causative and
    probably early role in hepatocarcinogenesis.

         Hulla  et al. (1993) analysed the p53 gene at the site
    corresponding to codon 249 of the human gene in AFB-induced
    preneoplastic hepatic nodules from rats. No mutations were detected in
    the tissues examined. Thus, at least in the rat, the authors suggested
    that AFB exposure alone may not be sufficient for the specificity of
    p53 mutations observed in HCC. The selective mutations have been
    identified only in populations at risk for hepatitis B; it is possible
    that both AFB1 and chronic hepatitis are essential for mutation at
    codon 249 in the human p53 gene.

         In another study in rats, Liu  et al. (1996) looked at the
    effects of AFB1 on the p53 locus at the preneoplastic stage of rat
    liver oncogenesis. Male Wistar rats received a single dose of 1.5 mg
    AFB1/kg bw by a gastric tube. Liver biopsies over a period of one year
    were examined for aberrations of the p53 gene together with the
    expression of placental GST, a marker for preneoplasia.
    Immunohistochemistry, Western blot, polymerase chain
    reaction-single-strand conformation polymorphism and DNA sequencing
    techniques were used. AFB1 induction resulted in GST overexpression,
    forming GST-positive multi-foci and nodules of hepatocytes but no
    aberrations in the p53 expression and the microstructure of exons 5-8
    of the p53 gene. Thus, the authors concluded that p53 mutations might
    not occur at this early stage of AFB1-induced hepatocarcinogenesis.

         Shi  et al. (1995b) characterized p53 mutations in 44
    hepatocellular carcinomas from Chinese patients residing in a
    high-incidence area. In contrast to HCCs from other high HCC incidence
    areas with endemic aflatoxin exposures, in which codon 249 is a
    mutational hotspot, no mutations were observed at codon 249. The
    authors concluded that risk factors other that dietary exposure to
    aflatoxin may contribute to the high HCC incidence in Singapore.

         Liang (1995) recently reviewed the relationship of p53 proteins
    and AFB1. He pointed out firstly that the murine mutant p53 gene
    p53Ser249 appears to have a hepatocyte-specific phenotype, which
    suggests that this gene may interact with cellular factors(s) in a
    liver-specific manner to alter the growth property of hepatocytes. It
    is not known if the human form of p53Ser249 exhibits the same
    properties. Secondly, cooperative interaction of this p53 mutation and
    viral-induced cellular changes are probably involved in the

    transformation of hepatocytes in situations where aflatoxin exposure
    and hepatitis viral infection are evident. Recent studies of
    non-aflatoxin-associated HCC showed that p53 mutations are not as
    common as other human malignancies. This difference could be explained
    by the relatively low proliferation rate of hepatocytes as compared
    with other epithelial cells, such as colonic mucosa and mammary gland.
    Because p53 plays a critical role in "damage control" of proliferating
    cells and in regulation of abnormal proliferation, it is reasonable to
    speculate that p53 mutations may play a lesser role in
    hepatocarcinogenesis. However, dysregulated p53 function may still be
    an important step in this process, in view of the recent observation
    that HBX protein encoded by HBV appears to interact with p53 and
    inhibit its function. 

         Thirdly, it has not been possible to induce the same p53 mutation
    with aflatoxin exposure in a murine model, which casts a shadow of
    doubt on the applicability of studies in the murine model to human
    hepatocarcinogenesis. Liang (1995) recommended using human p53 genes
    to perform parallel experiments.

         Harris (1995) has also reviewed this subject. He reiterated that
    in high-incidence liver cancer areas such as China and Mozambique, the
    high frequency of G:C to T:A transversions in human hepatocellular
    carcinomas in this region could be due to the high mutability of the
    third base of codon 249 by AFB1 or a selective growth advantage of
    hepatocyte clones carrying this specific p53 mutant in liver
    chronically infected with HBV. The third base of codon 240 in a human
    liver cell line exposed to AFB1 has been shown to be preferentially
    mutated, and transfected 240Serp53 mutant enhances the growth rate of
    the p53 null hepatocellular carcinoma cell line Hep3B.

         The hypothesis that some of the mutations observed in the p53
    tumour-suppressor gene may be specific markers of exposure to
    aflatoxin may represent a real breakthrough in the field of liver
    cancer epidemiology. In particular, the confirmation of the
    specificity of the p53/aflatoxin association could be useful in
    assessing and quantifying the responsibility of aflatoxin as an
    independent cause of liver cancer and in evaluating the likely
    interactions with the hepatitis viruses in humans. A word of caution
    should be raised regarding the interpretation of the early studies
    because of: 1) the small sample size and limited methodology as to the
    criteria of specimen inclusion; 2) inadequate adjustment of the
    correlations for exposures to other viral and non-viral risk factors
    at the individual level; 3) limited information on the sensitivity and
    specificity of the proposed genetic markers; in particular, some
    animal data and cell system data are inconsistent in showing a
    specific association between p53 codon 249 mutations and previous
    exposure to aflatoxin; and 4) insufficient knowledge of the additional
    genetic changes in p53 and other genes (i.e., N-ras, C-myc, c-fos,
    alpha-TGF) associated with liver cancer development.

    2.2.10.3  Epidemiology of primary liver cancer

     (a) Descriptive epidemiology.

         Liver cancer is a disease prevalent in some of the developing
    parts of the world. It is frequent in China, South East Asia and
    subsaharan Africa. In some of these regions, like the Qidong area in
    Southern China, liver cancer is the major cause of death to cancer
    among men. It is relatively common in Japan and in the countries in
    the Mediterranean basin and it is rare in the Americas and Northern
    Europe. Pockets of high risk populations have been described in the
    Amazonian basin, among Eskimos and in special populations like the
    renal transplant patients. The incidence of liver cancer is
    consistently higher in men than in women with a sex ratio ranging from
    2 to 3 in most countries. Within countries, further variation in
    incidence rates is observed across cancer registries, men showing
    greater variation than women. Worldwide, the incidence of liver cancer
    in men and women shows a strong correlation.

         Migration from high risk areas to lower risk areas tends to
    reduce the risk to the levels of the host country, and this is
    observable within first and second generations.

     (b) Etiology

         The etiology of primary liver cancer is nowadays largely
    understood. Table 1 summarizes the range and the point estimates of
    the attributable fractions in two different settings, the low-risk
    areas in Europe and the USA and the high-risk areas in Africa and
    Asia.

         In both scenarios, viral infections to hepatitis B or C virus are
    associated with liver cancer in a range from 65% to 100% of cases. In
    low-risk countries HBV predominates and the other relevant factors are
    alcohol, tobacco and oral contraceptives. In high-risk areas HBV
    predominates and aflatoxins play a role, although quantification has
    been difficult.

         The evidence points to a synergistic interaction between HBV and
    AF in the etiology of liver cancer and some debate exists as to the
    independency of AF as an etiologic agent in humans.

         It is noteworthy that the large majority of the available
    epidemiological studies including data on aflatoxin exposure are based
    on high-risk countries where both HBV and AF are highly prevalent.
    Since the nature of the interaction at low levels of exposure is
    unknown, extrapolation of results from available studies to other
    settings is questionable.

    In addition to these established factors, studies have identified
    other factors that may modulate the incidence of the disease. Risk
    factors identified are the use of contaminated drinking-water, liver
    flukes and severe malnourishment.  Protection from liver cancer has



        Table 1.  Causal factors of liver cancer and estimates of the attributable fractions

                                                                                          
    Factor                        Low-risk countries               High-risk countries
                              Japan Europe and the USA               Africa and Asia
                         Estimate   Range    Estimate   Range      Estimate      Range
                                                                                          

    Hepatitis B            <15%     4-50%       20%     18-44%       60%         40-90%
    Hepatitis C3            60%     12-64%      50%     40-80%      <10%         NE
    Aflatoxin             limited exposure     limited exposure     important exposure1
    Alcohol                <15%4               <20%     11-30%5      NE
    Tobacco                <12%4                40%     38-51%5      NE
    Oral contraceptive    10-50%2               NE                   NE
    Other                  <5%                                      <5%
                                                                                          

    1  Attributable risk not quantified. One study suggested attributable fraction close 
    to 50%.
    2  Restricted to liver cancer in women. Likely to increase in future generations. 
    Uncertain if hepatitis infections (notably HCV) are necessary co-factors.
    3  Not including double infections with HBV and HCV. Very few studies available using 
    second-generation assays.
    4  Estimates for the USA
    5  Estimates from three studies of LC in men

    NE non evaluated.
    Note: attributable fractions do not necessarily add to 100% due to multiple exposures and 
    possible interactions between risk factors.

    Adapted from CDC, 1989; Bosch & Muñoz, 1991; Thomas, 1991; Tanaka et al., 1993; IARC, 
    1994; Bosch, 1995
    


    been reported in the case of diets rich in retinol and protein.
    Associations have been reported between liver cancer and blood
    testosterone levels, HLA types, and predisposition due to
    polimorphisms in some of the SGT and CYT metabolic regulatory genes.

     (c) Vaccination against HBV as a preventive measure against liver
    cancer

        In 1983, the World Health Organization proposed as a medium-term
    objective trials of immunization against hepatitis B to prevent liver
    cancer. Since then more than 70 countries have introduced HBV
    vaccination into their routine vaccination schemes. A recently
    published study in Taiwan (Chang  et al., 1997) has described the
    rigorous application of universal immunization against hepatitis B and
    the prevention of the carrier state in children; these data provide
    further evidence of a direct causal relationship between HBV and liver
    cancer.

        The immunization programme against hepatitis B in Taiwan, an area
    of hyperendemic infection and moderate to high aflatoxin exposure,
    reduced the rate of HBV carriage in six-year-old children from about
    10% in the period from 1981 to 1986 to between 0.9 and 0.8% in the
    period from 1990 to 1994. The drop in the rate of carriage occurred as
    the proportion of infants immunized against hepatitis B increased from
    15% of children born to mothers at high risk during the earlier period
    to 84-94% of all newborn infants during the later period. This
    significant reduction in the prevalence of hepatitis B surface antigen
    was accompanied by a decline in the average annual incidence of
    hepatocellular carcinoma in children 6 to 14 years of age, from 0.7
    per 100 000 between 1981 and 1986 to 0.57 between 1986 and 1989 and
    0.36 between 1990 and 1994. The incidence of hepatocellular carcinoma
    in children 6 to 19 years old fell even more dramatically, from 0.52
    among those born between 1974 and 1984 to 0.13 among those born
    between 1984 and 1986. As the investigators pointed out, since the
    incidence of hepatocellular carcinoma in Taiwan peaks in the sixth
    decade of life, it may take 40 years or longer to see an overall
    decrease in the rate of hepatocellular carcinoma as a result of the
    vaccination programme.

        The Committee noted that studies like this one need to be observed
    carefully in coming years for the light they may shed on the
    relationship between aflatoxin, HBV and liver cancer.

     (d)    Effects of exposure to aflatoxins

        Ahmed  et al.(1995) undertook two prospective studies to
    determine a possible relationship between perinatal aflatoxin exposure
    and neonatal jaundice. First, cord blood samples from 37 neonates who
    subsequently developed jaundice and from 40 non-jaundiced (control)
    babies were analysed for six major aflatoxins and aflatoxicol.
    Peripheral blood samples of both groups were also analysed postnatally
    for aflatoxins. In a second study, serum aflatoxin levels of 64
    jaundiced neonates admitted from outside the hospital were compared

    with levels in 60 non-jaundiced control babies. Aflatoxins were
    detected in 14 (38%) cord blood samples of jaundiced neonates and in
    nine (23%) of the controls. The mean cord aflatoxin concentration was
    highest in jaundiced neonates with septicaemia, but the difference was
    not statistically significant. The frequency of detection of
    aflatoxins in peripheral blood was not significantly different in
    jaundiced and non-jaundiced babies. Aflatoxins were detected in the
    blood of over 50% of neonates with jaundice of unknown etiology. There
    was no correlation between severity of hyperbilirubinaemia and serum
    aflatoxin levels. Further studies are needed to determine the extent
    of pre-and postnatal exposure to aflatoxin in Nigerian infants and the
    effects of such exposure on fetal and neonatal health, the authors
    concluded.

        In October 1988, 13 Chinese children died of acute hepatic
    encephalopathy in the northwestern state of Perak in peninsular
    Malaysia (Lye  et al., 1995). Symptoms included vomiting,
    haematemesis, seizures, diarrhoea, fever and abdominal pain. All had
    liver dysfunction with increased aspartate aminotransferase and
    alanine aminotransferase levels greater than 100 IU/litre. The
    acuteness of the illness differed from previously reported outbreaks
    described in Kenya, India and Thailand; median incubation period for
    this outbreak was 8 hours, whereas the exposure was over a period of
    days to weeks of consumption of highly contaminated food such as maize
    in outbreaks in Kenya and India. Epidemiological investigations
    determined that the children had eaten a Chinese noodle,  loh see 
     fun, hours before they died. The attack rates among those who had
    eaten the noodles were significantly higher than those who had not 
    (P < 0.0001). The cases were geographically scattered in six towns in
    two districts along the route of distribution of the noodle supplied
    by one factory in Kampar town. Aflatoxins were confirmed in the
    postmortem samples from patients, but the noodles or their ingredients
    were not analysed for aflatoxins. The authors questioned the etiology
    of the outbreak.

        Ibeh  et al. (1994) examined the relationship between aflatoxin
    levels in serum of infertile men in comparison with random controls
    from the community. The subjects were 100 adult males, yielding 50
    semen samples, from men attending infertility clinics at a university
    teaching hospital and 50 normal men in the same community. The staple
    foods of the men were assayed for aflatoxin content. Aflatoxin was
    found in 20 semen samples from the infertile group (40 %) with a mean
    concentration of 1.7 µg/ml and four samples from the fertile group
    (8%) with a mean concentration of 1.0 µg/ml. Infertile men showed a
    higher percentage of spermatozoa abnormality (50%) than the fertile
    men (10-15 %).

        In a parallel experiment, adult male rats were given an aflatoxin-
    contaminated diet (8.5 µg purified AFB1/g of feed) for 14 days while
    10 control age-matched rats were fed a normal aflatoxin-free diet
    during the same period. Seven days after the withdrawal of aflatoxin
    from the diet of test rats, two rats were randomly selected from the
    test and control groups, killed and their semen harvested from the

    epididymis and vasa deferentia and analysed. The process was repeated
    at weekly intervals until four rats were left in the test and control
    groups. Thereafter, four fertile adult female rats were introduced to
    mate with test and control rats, and rats were observed for 90 days
    with an adequate, non-aflatoxin-contaminated diet. Results
    showed that rats exposed to dietary aflatoxin experienced changes in
    spermatozoal profile which differed in a statistically significant
    manner from the control rats; test rats showed depression in the
    motility, viability and number of sperm cells which resembled features
    seen in semen of infertile men exposed to aflatoxin. The four test
    rats who were mated in the conclusion of the study were unable to
    effect conception of fertile female rats, while the four control rats
    were able to do so.

        The authors hypothesized that aflatoxin may affect the
    reproductive system by its toxic effect on the liver, leading to the
    desquamation of the membranes of hepatocytes, the mitochondria, the
    cytosol and the endoplasmic reticulum. This cellular damage could
    include inhibition of enzyme synthesis and/or enzyme activities or
    inhibition of lipid metabolism or fatty acid synthesis, which may
    derail the capacity of the hepatocytes to handle the conversion of
    intermediate biomolecules, such as precursor molecules for hormones,
    e.g., testosterone and progesterone. Depression or absence of normal
    hormone levels could cause a wide range of degenerative changes in
    sexual organs. Aflatoxin may also affect the male reproductive system
    by causing lysis of sperm cells as a result of constant reversible
    reaction with the mycotoxin, binding of the toxin to free and/or bound
    amino acids in the seminal fluid, depressing the motility of
    spermatozoa and the formation of aflatoxin adducts with nucleic acids,
    giving a risk of mutations of the spermatogonia.

        El-Nazami  et al. (1995) examined the exposure of infants to
    aflatoxin M1 (AFM1) and of lactating mothers to AFB1, using AFM1 in
    breast milk as a biomarker for exposure to AFB1. Prevalence of AFM1 in
    breast milk samples from 73 women from Victoria, Australia
    (low-exposure area) and 11 women from Thailand (high-exposure area)
    was also compared. Assays were done by both HPLC and by ELISA. AFM1
    was detected in 11 samples from Victoria and five samples from
    Thailand at median concentrations of 0.071 mg/ml (range 0.028 - 1.031)
    and 0.664 ng/ml (range 0.039 - 1.736). Levels of AFM1 were
    significantly higher in milk samples from Thailand than in milk
    samples from Victoria.

        Ankrah  et al. (1994) attempted to ascertain if the presumed
    intake of dietary aflatoxins (AFB1 and AFG1) has adverse effect on the
    liver; aflatoxins were measured in serum, urine and faecal specimens
    obtained from a group of 40 apparently healthy adults (11 females and
    29 males) from the Greater Accra region of Ghana. Liver status of the
    subjects was monitored with serum  alpha-fetoprotein (AFP),
    alpha-1-antitrypsin (AAT) and direct: total bilirubin ratio. Aliquots
    of serum were tested for HBsAg. AFG1, AFB1, AFQ1, and AFM1 were
    detected in one or more of the body specimens in 35% of the subjects
    (AFB1+ group). Sixty-five per cent of the subjects had only AFG1 in

    their body specimens (AFB-group). Serum levels of AFP (greater than 20
    ng/ml), AAT (greater than 170 ng/dl) and direct: total bilirubin ratio
    (greater than 0.5), which indicated absence of predisposition to liver
    cancer in all the subjects but were suggestive of liver inflammation,
    were noted in both the AFB+ and AFB1-subjects. None of the subjects
    had malaria or hepatitis B virus infection. The authors suggested that
    the pattern of distribution of the aflatoxins in the subjects
    indicates that the suspected liver inflammation may involve other
    factors and may not be only due to present intake levels of
    aflatoxins.

     (e) Epidemiological studies on dietary aflatoxins and liver cancer

        A number of important epidemiological studies have been published
    since the Committee's last review of aflatoxins at its thirty-first
    meeting (Annex 1, reference 77).

        Yeh  et al. (1989) examined the roles of the hepatitis B virus
    and AFB1 in the development of primary hepatocellular carcinoma (PHC)
    in a cohort of 7917 men aged 25 to 64 years old in southern Guangxi,
    China, where the incidence of PHC is among the highest in the world.
    After accumulating 30 188 person-years of observation, 149 deaths were
    observed, 76 (51%) of which were due to PHC. Ninety-one per cent (69
    of 76) were HBsAg+ at enrollment into the study in contrast to 23% of
    all members of the cohort. Three of the four patients who died of
    liver cirrhosis were also HBsAg+ at enrollment. There was no
    association between HBsAg positivity and other causes of death. Within
    the cohort, there was a 3.5-fold difference in PHC mortality by place
    of residence.

        To estimate AFB1 exposure, between 1978 and 1984, the Fusui Liver
    Cancer Institute regularly sampled and tested staple foods consumed in
    the counties of southern Guangxi for contamination by AFB1. Twice a
    year, samples of raw foods were collected from all over the region and
    analysed for AFB1 content by TLC. An estimated mean level was computed
    for each commune as follows. The yearly amount consumed of a given raw
    foodstuff was multiplied by the average AFB1 content as determined
    from tested samples of raw foodstuffs. These cross-product terms were
    then summed over all staple foods, and the resultant figure was
    divided by the total population to obtain an estimated intake per
    person per year. These population-based levels of AFB1 were correlated
    with mortality rates of PHC among members of the cohort by the
    communes from which the subjects were derived.

        When estimated AFB1 levels in the subpopulations were plotted
    against the corresponding mortality rates of PHC, a positive and
    almost perfectly linear relationship was observed. On the other hand
    the prevalence of HBsAg was very high and homogeneous across the study
    areas (range 21.6%-24.7%) and therefore, no significant association
    was observed when the prevalence of HBsAg positivity in the
    subpopulations was compared with their corresponding rates of PHC
    mortality. The authors conclude that despite the "crudeness" of their
    exposure estimate, (i.e., population-based instead of personal

    exposure assessments), it is reasonable to conclude that AFB1 seems to
    play a role in the unusually high rates of PHC in southern Guangxi. 

        The population prevalence of HBsAg is extraordinarily high in this
    study population, almost one in every four adult men being a
    positive carrier of HBV. Primary infection occurs very early in this
    high-risk population, possibly through vertical transmission from
    carrier mothers to infants during the perinatal period, based on a
    survey of serum HBsAg in children ages 1 to 9 years in a county
    adjacent to Fusui. Even though most cases of liver cancer in this
    study did not have histopathological confirmation, the authors
    indicate that probably all were PHCs.

        The Yeh  et al. (1989) report is an early important study showing
    that, in a region where HBV is highly prevalent and PLC is common, the
    HBV carriers are at very high risk. It further indicates that in an
    area of high AFB1 exposure, the PLC mortality rates are higher than in
    areas of lower AFB1 exposure. This study provides the basic
    information for most potency estimates. Most of the early correlation
    studies (with or without HBV consideration) are consistent with the
    basic conclusion of the study of Yeh  et al. (1989), but other
    studies are not (Campbell  et al., 1990; Hsing  et al., 1991).
    However, the study has the general limitations of correlation studies
    in which: i) exposure to AFB1 is estimated from raw foodstuffs
    available to populations and attributed to individuals; ii) the
    correlation between PLC and AFB1 was not adjusted for any of the
    possible confounders such as HCV, alcohol, tobacco or nutritional
    status as shown in Taiwan by Yu  et al. (1995); iii) HBV exposure may
    have been underestimated due to lack of use of PCR methodology; iv)
    HBsAg prevalence was measured in a 25% sample of the cohort and
    attributed to the region.

        Campbell  et al. (1990) conducted a comprehensive cross-sectional
    survey in the People's Republic of China of possible risk factors for
    primary liver cancer (PLC) to include 48 survey sites, an
    approximately 600-fold aflatoxin exposure range, a 39-fold range of
    HCC mortality rates, a 28-fold range of hepatitis B virus surface
    antigen (HBsAg+) carrier prevalence, and estimation of exposures for a
    large number of other nutritional, dietary and life-style features
    (Campbell  et al., 1990). PLC mortality was unrelated to aflatoxin
    intake, but was positively correlated with HBsAg+ prevalence, plasma
    cholesterol, frequency of liquor consumption, and mean daily intake of
    cadmium from foods of plant origin. Multiple regression analysis for
    various combinations of risk factors showed that aflatoxin exposure
    consistently remained unassociated with PLC mortality regardless of
    variable adjustment. In contrast, associations of PLC mortality with
    HBsAg+, plasma cholesterol, and cadmium intake remained, regardless of
    model specifications, while the association with liquor consumption
    was markedly attenuated (nonsignificant) with adjustment for plasma
    cholesterol.

        The authors commented on the lack of an association between
    aflatoxin exposure and PLC mortality in this study, in view of the
    findings of most previous investigations. The absence of an
    aflatoxin-PLC association was consistent with a similar lack of
    association of PLC mortality with the consumption of the two foods
    most commonly contaminated with aflatoxin i.e. maize and mouldy
    groundnuts. In contrast to the lack of an association with aflatoxin,
    PLC mortality was highly correlated with HBsAg+ prevalence and not
    with past HBV infection, as assessed by the prevalence of antibody to
    the HBV core protein. In this study, the association of plasma
    cholesterol with PLC mortality was even more consistent than the
    association of PLC mortality with HBsAg+ prevalence. Mortalities from
    colon cancer, rectal cancer, lung cancer, leukemia, brain cancer and
    total aggregate cancer are also known to be associated with plasma
    cholesterol. This association was even more surprising in China where
    plasma cholesterol in this cohort ranged up to about 190 mg/dl, which
    is near the low end of the range for comparable Western subjects (Chen
     et al., 1990).

        The authors offered several explanations for the lack of an
    association between aflatoxin intake and PLC mortality, which
    contrasts with the finding of other previous studies. First, Chinese
    people might respond differently to aflatoxin, perhaps because of
    unique genetic or environmental characteristics. This is unlikely
    given the previously shown positive association between aflatoxin
    intake and PLC mortality in Chinese subjects (Yeh  et al., 1985; Yeh
     et al., 1989) and because major ethnic differences in risk for other
    cancers are greatly reduced or eliminated after migration to new
    environments.

        A second argument could be that the lack of an effect in this
    study may have been because measurement of aflatoxin exposure during
    the survey period was not representative of past intakes when the
    cancers were forming. However, a similar limitation existed for all
    other Chinese studies; this study is more reliable, in the opinion of
    the authors, because it is based on urinary aflatoxin metabolite
    excretion which directly represents and integrates over a day or so
    actual consumption. In addition, aflatoxin contamination rates in a
    county in the Guangxi Autonomous Region were relatively stable during
    the years 1972-1983.

        A third line of reasoning suggests that aflatoxin may not be a
    significant human carcinogen, in the opinion of the authors. The
    present study has greater statistical power and more comprehensive
    range, diversity and inclusiveness of risk factors than other previous
    studies. Humans may also be resistant to aflatoxin carcinogenesis, a
    finding which is supported by  in vitro aflatoxin studies on species
    of varying resistance (Booth  et al., 1981). Humans may also be more
    refractory when consuming lower protein diets; whereas acute toxicity
    of aflatoxin is increased in protein-malnourished children (Hendrickse
     et al., 1982).

        The authors pointed out that data from animal studies have shown
    that when animals were fed either lower levels of animal protein (5-
    or the same level (20%) of plant protein after completion of aflatoxin
    dosing, development or preneoplastic lesions and tumours was markedly
    inhibited (Appleton & Campbell, 1983; Schulsinger  et al., 1989).
    Protein in the Chinese diet is primarily of plant or fish origin, as
    compared to protein in the USA diet which is primarily of animal
    origin (Food and Nutrition Board, 1989; Chen  et al., 1990). 

        The authors continued with a critique of previous aflatoxin
    epidemiology studies and offered the following model to explain the
    etiology of PLC. The vast majority of individuals who are susceptible
    to PLC are those who are persistently infected with HBV. Within this
    HBsAg+ population, additional risk is contributed chiefly by
    nutritional and dietary practices that enhance liver cell
    proliferation, such as diets containing significant amounts of animal
    protein. Aflatoxin may act as a carcinogenic initiator, but
    contributes only a very small proportion of the initiating activity
    routinely exposing the liver. Therefore, HBsAg+ is a necessary but
    insufficient cause of PLC, aflatoxin is an unnecessary and
    insufficient cause, and sustained nourishment causing liver cell
    proliferation (and elevated plasma cholesterol) is a necessary and
    insufficient cause for HBsAg negative carriers, but a necessary and
    sufficient cause for HBsAg positive carriers. Why is PLC so much more
    common in undernourished and impoverished societies? The authors
    concluded that PLC is more common because HBsAg+ carriers are more
    common.

        In evaluating the significance of this study by Campbell  et 
     al. (1990), a number of issues, both statistical and
    non-statistical, should be considered. For example, PLC rates were
    determined for the years 1973-1975 and the biochemical analyses
    (covariate ascertainment) was conducted in 1983. With regard to the
    statistical analysis presented in the paper, there is some indication
    that the sample data do not adequately satisfy the normality
    assumptions upon which the univariate correlation and multiple
    regression analyses are based. Finally, the urinary aflatoxin
    measurements were of total aflatoxin metabolites, which have been
    shown not to correlate well with levels of AFB1 consumed (Wild 
     et al., 1992; Groopman  et al., 1993).

     (f) What can we learn from epidemiological studies that considered
     HBV, HCV and AFB in relation to liver cancer?

        Viral hepatitis is a major worldwide public health problem. It is
    estimated that over 300 million individuals are chronically infected
    with HBV and perhaps 100 million with HCV. Chronic infection with
    either virus has been linked to cirrhosis and liver cancer. HBV is
    prevalent in the developing parts of the world, and HCV is emerging as
    a major cause of hepatocellular cancer in Japan and western societies
    (Table 1).

        Tests to detect HBV markers have increased in sensitivity, largely
    due to the use of the polymerase chain reaction (PCR) to amplify HBV
    DNA in serum and liver tissues. HBV infection has been shown to
    persist in the serum (49.7%) or in the liver cancer tissue (24.9%) in
    a number of patients with liver cancer that are at the same time HBSAg
    negative (Bosch & Muñoz, 1991). This pattern has been documented in
    cases from areas at low and high risk for HBV infection.

        Although the significance of detecting low levels of HBV DNA in a
    patient with liver cancer is not fully understood, from an
    epidemiological viewpoint, these subjects could easily be classified
    as persistently exposed to HBV and grouped with the HBsAg carriers in
    computing risk estimates. Although data are sparse on the prevalence
    of equivalent markers in the general population, it is likely that
    among controls the prevalence of PCR-detected HBV DNA in the absence
    of any other HBV marker is extremely low. If this is the case, the
    case control studies that use PCR would increase the Risk Ratio
    estimates for HBV as well as the estimates of the Attributable
    Fraction. There are no case control studies that have used PCR methods
    (in serum or liver tissue) to detect HBV exposure.

        Ramesh & Panda (1993) have questioned the hypothesis that HBV
    causes chronic liver disease and liver cell carcinoma in
    HBsAg-positive individuals only. The presence of HBV in patients with
    HCC who are seropositive for the envelope antigen (HBsAg) is well
    established. Epidemiological studies have shown a small percentage of
    patients with HCC with past HBV infection, positive for anti-HBsAb or
    HBcAb. However, the role of HBV in HCC cases who seroconverted from
    HBsAg to HBsAb is unclear. The authors described a study of 36 HCC
    cases where four cases negative for HBsAg and with underlying
    cirrhosis were found. Biopsy tissue was investigated by polymerase
    chain reaction; all four samples tested were positive for a portion of
    the surface region (nucleotide position: 636-735), but were negative
    for the "X" and the "C" regions of HBV genome. Since hepatitis C virus
    (HCV) has been associated with HCC, the authors tested serum samples
    of the four cases for anti-HCV; one out of four was positive for
    anti-HCV. The authors concluded that these observations indicate that
    parts of the HBV genome can persist in liver cells of individuals who
    have recovered from clinical illness and seroconverted to HBsAg
    positive. However, the significance of these sub-genomic fragments in
    the development of HCC is not clear.

        The identification of HCV in the last decade has been a major step
    forward in the understanding of the origins of liver cancer and in the
    quantification of the proportion of cases related directly to viral
    infections (IARC, 1994). Epidemiological studies are largely
    consistent in showing a strong association between carriers of
    anti-HCV and liver cancer. The specific potential to induce PLC by
    each of the HVC types and variants of types as well as the impact of
    other factors from the host and the environment still requires further
    research. Likewise, few studies are available exploring the role of
    aflatoxin in the presence of HBV and HCV. High estimates of the
    Relative Risk for carriers of anti-HCV have also been reported in

    areas of HBV endemicity. The risk linked to HCV is independent of HBV
    and persons who are carriers of both HBsAg and anti-HCV are at a very
    high risk of developing liver cancer. HCV is likely to be the major
    cause of liver cancer in countries at low/intermediate risk like the
    USA and Europe.

     (g) Epidemiological studies including aflatoxins in countries where 
     the risk of liver cancer is low

        In countries where liver cancer is rare and aflatoxin exposure is
    low, most etiological studies on liver cancer have not considered
    aflatoxins as a risk factor. The populations with higher exposures are
    the workers occupationally exposed to grain dust in the animal feed
    processing plants. In studies conducted in the Nordic countries in
    Europe, Sweden, Denmark, the Netherlands and in the USA, aflatoxins
    have been isolated from dust samples and an excess of mortality of
    several such cohorts has been documented for liver cancer (risk, 2.4
    times the expected rates) liver and biliary tract cancer (risk, 2.5
    times the expected rates), lung cancer and lymphomas (risk, 1.5-3
    times the expected rates (Hayes  et al., 1984; Alavanja  et al., 
    1987; Olsen  et al., 1988). It should be noted that some of these
    studies did not evaluate other relevant exposures such as hepatitis
    infections and alcohol.

        It is of interest that few studies are available on liver cancer
    in Latin America. In this extensive region of the world, agricultural
    products are prone to mould growth, and consumption of maize is part
    of the staple food in many countries. Yet liver cancer is rare in
    these populations as is HBV infection. If that is the case, Latin
    America would be potentially a very informative field to investigate
    the occurrence of liver cancer in populations exposed to AF as the
    central risk factor.

     (h) Epidemiological studies that used biomarkers of exposure to 
     aflatoxins including studies on genetic susceptibility to aflatoxins

        Biomarkers have been developed and are being introduced in
    epidemiological studies with the purpose of increasing the accuracy of
    the assessment of exposure to aflatoxins. Various biomarkers have been
    developed, including urinary total aflatoxins, aflatoxin adducts in
    urine, aflatoxin albumin adduct in serum, aflatoxin adducts in liver
    cancer tissue and more recently p53 specific mutations in liver cancer
    specimens. Other studies are investigating genetic polymorphisms in
    some key genes involved in the metabolism of aflatoxin that may
    introduce some variability in the response to aflatoxin. (Groopman 
     et al., 1994; Wild  et al., 1996; IARC, 1997).

        The importance of the major aflatoxin-nucleic acid adduct,
    AFB-N7-guanine, in urine as a biomarker was enhanced by the finding
    that this metabolite is excreted exclusively in urine of exposed rats,
    thus simplifying pharmacokinetic considerations. The aflatoxin-albumin
    adduct in serum has also been examined as a biomarker of exposure;
    because of the longer half-life  in vivo of albumin compared to the

    urinary AFB-N7-guanine, the serum albumin adduct can integrate
    exposures over longer time periods. Data from human exposure studies
    have shown that the excretion of the urinary aflatoxin nucleic acid
    adduct and formation of the serum albumin adduct are highly
    correlated. In the rat, validation studies for the dose-dependent
    excretion of urinary aflatoxin biomarkers were conducted in rats
    following a single exposure to AFB1; excellent linear correspondence
    between oral AFB1 dose and excretion of AFB-N7-guanine in urine was
    shown (Scholl  et al., 1995).

        Aflatoxin metabolites in urine or adducts in serum can be a useful
    tool to evaluate exposure but with the currently available methods
    remain relatively short-term exposure markers; biomarkers are of much
    less use in predicting long-term or lifetime human exposure. As such,
    they reflect poorly the natural pattern of exposure to aflatoxin
    (i.e., seasonality, manual sorting of foodstuffs, age at exposure,
    etc.); therefore, it is not surprising that studies conducted using
    aflatoxin biomarkers as markers of exposure show conflicting results.

        At present, it is not fully understood how the functional status
    of the liver or the coexistence of other risk factors for liver cancer
    may affect the different biomarkers that are being proposed for
    epidemiological studies. Few studies have described the natural
    history of these markers in patients with chronic liver disease
    including chronic hepatitis and liver cirrhosis, conditions that
    usually precede liver cancer by months or years (Wild  et al., 1993;
    Wang  et al., 1996a). In this circumstance, the interpretation of the
    findings is complicated since the aflatoxin biomarker may be
    confounded by the presence of some risk factors, (e.g., HBV, HCV,
    alcohol), the presence of some protective factors (e.g., retinol in
    the diet) or by the presence of liver disease (i.e. chronic active
    hepatitis B infection).

        The mutations in the p53 gene claimed to be specific markers of
    exposure to aflatoxin are being actively investigated to confirm the
    strength and the specificity of the association in human populations
    (Harris, 1995). Recent data have been summarized earlier in this
    paper.

        The best evidence of an interaction between HBV and aflatoxin in
    the causation of human liver cancer is the cohort study in Shanghai
    (Ross  et al., 1992, Qian  et al., 1994; Yuan  et al., 1995). This
    is an ongoing prospective study of 18 244 middle-aged men in Shanghai,
    China. Assays for urinary AFB1, its metabolites AFP1 and AFM1 and DNA
    adducts have been performed to assess the relationship between
    aflatoxin exposure and liver cancer. After 35 299 person-years of
    follow-up, 22 cases of liver cancer had been identified. For each
    case, 5 or 10 controls were randomly selected from cohort members
    without liver cancer on the date the disorder was diagnosed in the
    case and matched to within 1 year of age, within 1 month for sample
    collection, and for neighbourhood of residence. Each subject provided
    a blood sample and a urine sample. A positive result was defined as
    the presence of at least 1 ng of an individual aflatoxin compound in

    the sample. Hepatitis B surface antigen was measured by a standard
    radioimmunoassay method.

        Subjects with liver cancer were significantly more likely than
    were controls to have detectable concentrations of any of the
    aflatoxin compounds; the strongest association was for AFP1.
    Positivity for HBsAg was strongly associated with risk of liver
    cancer. The authors concluded that their results are based on too few
    cases to give a reliable estimate of attributable risk, but they
    estimated that up to 50% of cases of liver cancer in Shanghai may be
    due to aflatoxin exposure.

        In further follow-up of the Shanghai study, Qian  et al. (1994)
    reported on 70 000 person-years of follow-up and 55 cases of HCC.
    Levels of urinary AFB1 and the oxidative metabolites, including the
    major aflatoxin nucleic acid adduct, aflatoxin-N7-guanine, were
    determined for 50 of the 55 identified cases of HCC; 267 controls were
    matched against the 50 cases as above. A nested case-control analysis
    showed highly significant association between the