Mutagenicity of carcinogenic heterocyclic amines in Salmonella typhimurium YG strains and transgenic rodents including gpt delta

Chemical carcinogens to humans have been usually identified by epidemiological studies on the relationships between occupational or environmental exposure to the agents and specific cancer induction. In contrast, carcinogenic heterocyclic amines were identified under the principle that mutagens in bacterial in the Ames test are possible human carcinogens. In the 1970s to 1990s, more than 10 heterocyclic amines were isolated from pyrolysates of amino acids, proteins, meat or fish as mutagens in the Ames test, and they were demonstrated as carcinogens in rodents. In the 1980s and 1990s, we have developed derivatives of the Ames tester strains that overexpressed acetyltransferase of Salmonella typhimurium. These strains such as Salmonella typhimurium YG1024 exhibited a high sensitivity to the mutagenicity of the carcinogenic heterocyclic amines. Because of the high sensitivity, YG1024 and other YG strains were used for various purposes, e.g., identification of novel heterocyclic amines, mechanisms of metabolic activation, comparison of mutagenic potencies of various heterocyclic amines, and the co-mutagenic effects. In the 1990s and 2000s, we developed transgenic mice and rats for the detection of mutagenicity of chemicals in vivo. The transgenics were generated by the introduction of reporter genes for mutations into fertilized eggs of mice and rats. We named the transgenics as gpt delta because the gpt gene of Escherichia coli was used for detection of point mutations such as base substitutions and frameshifts and the red/gam genes of λ phage were employed to detect deletion mutations. The transgenic rodents gpt delta and other transgenics with lacI or lacZ as reporter genes have been utilized for characterization of mutagenicity of heterocyclic amines in vivo. In this review, we summarized the in vitro mutagenicity of heterocyclic amines in Salmonella typhimurium YG strains and the in vivo mutagenicity in transgenic rodents. We discussed the relationships between in vitro and in vivo mutagenicity of the heterocyclic amines and their relations to the carcinogenicity. Supplementary Information The online version contains supplementary material available at 10.1186/s41021-021-00207-0.


Development of Salmonella typhimurium YG strains
In the 1970s, Dr. Bruce N. Ames, University of California, developed a bacterial mutagenicity test (Ames test) and reported that a high percentage of bacterial mutagens in the Ames test are rodent carcinogens [27,28]. The test is simple, rapid and economical; therefore, large number of environmental chemicals were tested for potential mutagenicity in the Ames tester strains. Typical tester strains of the Ames test are Salmonella typhimurium TA98 and TA100, which detect frameshift-type mutagens and base-substitution-type mutagens, respectively [19,20]. In the same era, Dr. Takashi Sugimura, National Cancer Center in Japan, was interested in the possibility that smoke from broiled fish might be mutagenic and carcinogenic. Dr. Sugimura and his collaborators examined this possibility using the Ames test and isolated many heterocyclic amines as mutagens from pyrolysates of amino acids, proteins, meat or fish as mutagens [6,29]. Similarly, Dr. Daisuke Yoshida, the Japan Tobacco & Salt Public Cooperation, isolated AαC and MeAαC from pyrolysis products of soybean globulin and Dr. James S. Felton, Lawrence Livermore National Laboratory, U.S.A., identified PhIP and the related chemicals from fried ground beef [9,14]. Heterocyclic amines require metabolic activation for mutagenesis and carcinogenesis. In general, they are first oxidized by CYP1A2 to N-hydroxy derivatives, which are further activated by O-acetyltransferase or sulfotransferase to the nitrenium ions, thereby inducing DNA adducts and mutations [30][31][32][33]. In the Ames test, these metabolic enzymes are provided as 9,000 x g supernatant of rat liver homogenates (S9) [34]. It must be pointed out, however, Salmonella typhimurium used in the Ames test has enzymes involved in metabolic activation. In fact, strain TA98/1,8-DNP 6 is significantly resistant to the mutagenicity and killing effects of aromatic amines and nitro aromatics, because this strain is devoid of acetyltransferase activity [21].
In the mid-1980s, we were interested in the metabolic activation mechanisms of chemical carcinogens and cloned the oat gene encoding bacterial O-acetyltransferase [16,22]. For this purpose, we constructed a gene library of Salmonella typhimurium strain TA1538 with a multicopy-number plasmid pBR322 and introduced the gene library into strain TA1538/1,8-DNP, which is the same as TA98/1,8-DNP 6 but lacks plasmid pKM101 (Fig. 1). We searched for colonies that could grow on plates without 2-nitrofluorene (2-NF) but could not grow on plates with 2-NF. The principle was that if a plasmid carrying the oat gene was introduced into the host strain TA1538/1,8-DNP, the transformants would not grow on plates with 2-NF but grow on plates without 2-NF because 2-NF requires activities of O-acetyltransferase for cytotoxicity and mutagenicity. Fortunately, we successfully isolated candidate colonies and confirmed that the sensitivity was maintained after the plasmids extracted from the candidate colonies were introduced to the fresh background of TA1538/1,8-DNP. Plasmid pYG121 and pYG122 were the first isolated plasmids that carried the oat gene (Table 1). We then constructed plasmid pYG213, a deletion derivative of pYG122, which contains a 1.35kb-DNA fragment of pYG122 including the oat gene. However, pYG213 has the ampicillin-resistance-gene and is incompatible with strains TA98 and TA100, both of which possess pKM101 that confers ampicillin resistance. Therefore, we subcloned the 1.35-kb DNA fragment into the ScaI site of pBR322 and generated pYG219. Subcloning into this site disrupted the ampicillin-resistance gene and permitted the selection of pYG219 in TA98 and TA100. The resulting strains were named as YG1024 and YG1029, respectively [16]. N-hydroxy-Glu-P-1 O-acetyltransferase activities of TA1538/1,8-DNP harboring pBR322, pYG122, pYG213 or pYG219 were 0, 28.0, 228 Fig. 1 Gene cloning of the oat gene encoding O-acetyltransferase in Salmonella typhimurium. The chromosome DNA of Salmonella typhimurium TA1538 was partially digested with Sau3A1 and ligated to BamH1-digested plasmid pBR322, thereby generating a plasmid library of TA1538. Then, the library DNA was introduced into Salmonella typhimurium TA1538/1,8-DNP and screened the colonies that could grow on plates without 2-NF but could not grow on plates with 2-NF or 54.6 nmol/min/mg protein, respectively [16,23]. Although strain YG1012, which is TA1538/1,8-DNP harboring pYG213, had the highest O-acetyltransferase activity, it exhibited lower sensitivity to the mutagenicity of 1-aminonaphthalene + S9, 1-nitropyrene and 1,8-dinitropyrene compared to YG1024 [23]. It suggests that these chemicals require the presence of pKM101 for maximum frameshift mutagenesis. Plasmid pKM101 carries the mucAB genes encoding DNA polymerase RI, an error-prone DNA polymerase involved in translesion DNA synthesis [35]. Owing to the possession of pKM101 and the wider range of sensitivity, strain YG1024 is more widely used for mutation assays than strain YG1012 [36]. However, YG1024 showed comparable or slightly lower sensitivity to 2-hydroxy-acetylaminofluorene, Glu-P-1 + S9 and 2-aminoanthracene +S9 compared to YG1012 [23]. It appears that these chemicals are not strongly dependent on the presence of pKM101 for maximum mutagenesis. Later, we constructed plasmid pYG233 carrying the oat gene and the cnr gene encoding classical nitroreductase [24] and introduced it to strains TA98 and TA100. The resulting strains YG1041 and YG1042, respectively, overexpressed both acetyltransferase and nitroreductase [26]. They were more sensitive to the mutagenicity of nitroaromatics such as 2-NF, 2,6-dinitrotoluene and 1-nitropyrene than YG1024 or YG1029. A possible problem with YG1041 and YG1042 is the extreme sensitivity to the killing effects of nitro, amino and hydroxyamino compounds. The number of revertants increased very sharply and decreased quickly with increasing doses. In addition, the number of spontaneous revertants per plate of YG1041 and YG1042 was higher than that of spontaneous revertants per plate of YG1024 and YG1029, respectively. The high number of spontaneous revertants obscures the weak mutagenicity of chemicals. Therefore, we recommend using these strains along with other strains such as YG1024 and YG1029 to avoid overlooking the mutagenic responses of test chemicals.

Novel heterocyclic amines
Heterocyclic amines were initially isolated from the pyrolysates of food or food components. Later, they were isolated from various environmental sources such as river water [37], automobile exhaust particles [38], cigarette smoke [39], human excretion [40] and rainwater [41]. Appropriate devices and methods are required to efficiently collect environmental mutagens. In the case of river water, it is critical to effectively collect and concentrate the target molecules from a large volume of water samples because pollutants are present in only minute concentrations. Sakamoto and Hayatsu developed an effective method, i.e., the blue rayon hanging technique, in which blue rayon covalently bound to the blue pigment copper phthalocyanine is hung in the river to specifically adsorb polycyclic planar compounds including heterocyclic amines [42]. The blue rayon absorbing water pollutants, instead of a large volume of water samples, is transferred to laboratories for chemical analyses and mutagenicity assays. Kataoka et al. [43] isolated and identified IQ, Trp-P-1 and AαC in the Danube River in Vienna by the method. The river water samples exhibited higher mutagenicity in YG1024 than in TA98 in the presence of S9 activation, which suggested a significant contribution of the heterocyclic amines to the whole mutagenicity of the water samples ( Table 2). The source of the heterocyclic amines in the Danube River may be the emission and discharge from food processing, e.g., smoke sausage, and wood burning. The collection of mutagens in river water by the blue-rayon hanging technique and the subsequent mutagenicity assays with YG1024 were conducted in samples from the Chao Phraya River in Bangkok, Thailand, and the Sumida and Ara Rivers in Tokyo [44]. Similar methods were employed for samples from rivers in Boston, New York, Washington D.C. and Montreal in North America [45]. In the latter case, YG1041 and YG1024 were much more sensitive than TA98 in the presence of S9 plus an NADPH-generating system (S9 mix).
TA98, TA100, YG1024, YG1029 W PBTA-2 was a novel aromatic amine mutagen isolated from river water in Kyoto. The sensitivity was YG1024> > TA98. The mutagen may be produced from an azo dye in dyeing factories and treatment at sewage plants.

Metabolic activation of heterocyclic amines
Heterocyclic amines require metabolic activation for mutagenesis via CYP enzymes and either O-acetyltransferase or sulfotransferase. As expected, strain YG1024 overexpressing the acetyltransferase exhibited higher sensitivity, i.e., more induced revertants per nmol or μg of chemical, than strain TA98. In fact, YG1024 showed more than 10 times higher sensitivity than TA98 for Glu-P-1, IQ, MeIQ, MeIQx and PBTA-1 [36,46,57,62,70]. However, YG1024 exhibited similar or only slightly higher sensitivity to PhIP and Trp-P-2 than TA98, suggesting that these heterocyclic amines are not activated by acetyltransferase [46]. Consistent with this, Wu et al.
reported that CHO UV-5 cells expressing mouse CYP1A2 and human N-acetyltransferase did not exhibit any significant sensitivity or genotoxicity to PhIP [81]. Wu et al. reported in the next paper that CHO UV5 cells expressing mouse CYP1A2 and human aryl sulfotransferases, i.e., HAST1 or HAST3, exhibited higher sensitivity to the killing effects of PhIP than CHO UV5 cells expressing only mouse CYP1A2 [82]. Thus, N-hydroxy-PhIP may be activated by sulfotransferase rather than acetyltransferase. Knasmüller et al. examined the comparative mutagenicity of several heterocyclic amines with strain YG1024 and reported that IQ and MeIQ were the most potent mutagens followed by MeIQx and Trp-P-1 and PhIP was the weakest mutagen [51]. This order was basically the same when strain TA98 was used [6]. Part of the reason for the weak mutagenicity of PhIP in strains YG1024 and TA98 may be its low dependency on acetyltransferase for the metabolic activation.
The crystal structure of Salmonella acetyltransferase was determined at 2.8Å resolution, and it was revealed that a Cys-His-Asp catalytic triad is involved in the catalytic mechanism [83]. The critical Cys residue is conserved between the acetyltransferase of Salmonella typhimurium and mammalian acetyltransferases NAT1 and NAT2 [84]. Both acetyltransferases of Salmonella typhimurium and mammals catalyze N-acetylation (usually inactivation) and O-acetylation (usually activation) of heterocyclic amines and the N-hydroxy derivatives [85]. Mammalian NAT1 and NAT2 are polymorphic and epidemiological studies suggest that the polymorphisms modify the risk of developing various cancers such as urinary bladder, colorectal and breast cancers.
In addition to CYP enzymes, prostaglandin-H synthase activates several heterocyclic amines. This enzyme is an arachidonic acid-dependent peroxidase and is suggested to be involved in the metabolic activation of xenobiotics in extrahepatic tissues. Ram seminal vesicle microsomes, a rich source of prostaglandin-H synthase, activate IQ and MeIQ for mutagenesis [53,56]. The mutagenicity was more sensitively detected in YG strains overexpressing Salmonella acetyltransferase, i.e., YG1006 (TA1538/ 1,8-DNP with pYG121) and YG1024, than in TA98. The primary mutagenic metabolite of IQ by prostaglandin-H synthase is nitro-IQ, while N-hydroxy derivatives are the active metabolites of IQ and MeIQx by CYP enzymes [25,33,56,62]. Since nitro-IQ and N-hydroxy IQ are further activated by acetyltransferase, the same DNA adduct, i.e., C8-dG-IQ-adduct, is formed in DNA when YG1024 is treated with prostaglandin-H synthase-oxidized IQ or hepatocyte-exposed IQ [58].

Development of gpt delta transgenic rodents for mutagenicity assays in vivo
In the late 1980s and the early 1990s, two transgenic mouse models were developed with E. coli lacZ or lacI as reporter genes for mutations in vivo [87,88]. In these mouse models, i.e., Muta Mouse with lacZ and Big Blue Mouse with lacI, the λ phage DNAs with the reporter gene were integrated into the chromosome of all the cells of mice [17]. After the mice are treated with chemical agents, the phage is rescued as phage particles from the mouse genome of various organs and tissues by in vitro packaging reactions. The rescued phages are introduced into indicator E. coli strains to select mutant plaques by color selection, i.e., visual search of colorless plaques in Muta Mouse or blue color plaques in Big Blue Mouse in more than 100,000 background plaques. Transgenic mouse mutagenicity assays allow detection of mutations in any organs or tissues of mice including the liver, lung, bone marrow or testis. However, color selection is time-consuming and expensive because the visual search of plaques of different color is laborious and the chromogenic agent X-gal is expensive. To overcome this limitation, a positive selection with the cII gene of phage λ has been developed [89]. The cII gene encodes a repressor protein that controls the lysogenic and lytic cycle of λ. Mutations in the cII gene can be positively identified with an indicator E. coli strain deficient in Hfl protease. In the hflstrain, only λ phage with inactive cII can form plaques at 24 o C. In contrast, all the rescued λ phage can form plaques at 37 o C regardless of the status of cII. Thus, the mutant frequency (MF) can be calculated by dividing the number of plaques formed at 24 o C by the number of plaques formed at 37 o C and the dilution factor. The coding size of the cII gene is approximately 300 base pairs (bps), which are approximately 1/10 of lacZ and 1/3 of lacI. Thus, DNA sequencing analysis of the mutants is feasible. The cII selection detects point mutations, i.e., base substitutions and frameshifts, but not large deletions. In addition, the cII selection is applicable to both Muta Mouse and Big Blue Mouse. Later, Big Blue Rat was developed with the same λ phage DNA, i.e., λ LIZ DNA, with the lacI and cII genes [90].
In the mid-1990s, we developed another transgenic mouse model named gpt delta by introducing λEG10 DNA into fertilized eggs of C57BL/6J mice [91]. λEG10 DNA was integrated into a single site of the mouse chromosome 17 [92,93]. A feature of the transgenic mutation assay is the incorporation of two distinct selections for detecting different types of mutations, i.e., gpt selection for point mutations and Spiselection for deletions (Fig. 2) [17]. The gpt selection uses the gpt gene of E. coli as a reporter gene for mutations. The gpt gene is a bacterial counterpart of the human Hprt gene and encodes guanine phosphoribosyltransferase. When the gpt gene is inactivated by mutations, E. coli cells can survive on the plates containing 6-thioguanine (6-TG), whereas E. coli cells with the wild-type gpt gene cannot survive on the plates because they phosphoribosylate 6-TG to a toxic substance, i.e., 6-TGMP. Thus, the gpt selection is a positive selection. The coding size of the gpt gene is 456 bp, which is convenient for DNA sequencing analysis. The Spiselection positively detects deletion mutations in λ phage [94]. The selection name Spistands for "sensitive to P2 interference". This selection takes advantage of the restricted growth of the wild-type λ phage in P2 lysogen, which is E. coli cells carrying prophage P2 in the chromosome. Only mutant λ deficient in the functions of both the gam gene and the redBA genes can grow well in P2 lysogens and display the Spiphenotype. Because the gam gene and the redBA genes are located side by side in the λ genome, inactivation of both functions is most likely to be induced by deletions in the region. Because of the size limitation of the λ phage in in vitro packaging reactions, the size of deletions detectable by the selection is less than 10 kb. However, tandem array of multiple copies of 48-kb λEG10 DNA in the chromosome amounts to a potential target of more than 1 mega bps. Deletion mutations with a molecular size of more than 1 kb were detected by the Spiselection in various organs such as the liver, spleen, kidney or brain of the mice irradiated with heavy-ions, gamma-rays or X-rays [95][96][97]. Ultraviolet-B irradiation and treatment with mitomycin C also induced large deletions in the epidermis and bone marrow, respectively [98,99]. The molecular nature of the deletion mutations can be characterized by DNA sequencing of the mutated gam and redBA region [100]. Some of the Spilarge deletions have junctions of two broken ends overlapping with short homologous sequences, while others have flush ends. It suggests that non-homologous end-joining plays an essential role in the induction of deletion mutations. The Spiselection also detects -1 frameshifts in the gam gene that interfere with the start of translation of the downstream redBA genes [95]. The -1 frameshifts mostly occur in run sequences such as AAAAAA to AAAAA in the gam gene, and this type of mutation accounts for most of the spontaneous Spimutations.
In the early 2000s, Hayashi et al. introduced λEG10 DNA into fertilized eggs of Sprague-Dawley (SD) rats and established SD gpt delta rats [101]. λEG10 DNA was integrated into a single site of the chromosome 4 [93]. The SD gpt delta rats were crossed with Fischer 344 (F344) rats for 15 generations and established F344 gpt delta rats [102]. Unlike gpt delta mice, which have λEG10 DNA in both alleles of chromosome 17, gpt delta rats are heterozygous, where λEG10 is integrated into only one allele of chromosome 4. This is because homozygous rats are defective in tooth development and cannot survive after weaning. To overcome this limitation, a new homozygous gpt delta rat strain was established in the genetic background of Wistar Hannover [103]. In the new version of gpt delta rat, λEG10 was integrated into both alleles of chromosome 1 and exhibited a significantly higher packaging efficiency than the heterozygous gpt delta rats. The average of spontaneous gpt and Spi -MFs in the liver of heterozygous and new homozygous gpt delta rats are 4.4-6.5 x 10 -6 and 2.8-5.5 x 10 -6 , respectively, which are significantly lower than those of the lacI and cII genes [104]. The low frequencies of spontaneous MFs of gpt and Spiare similar to those of gpt delta mice. Transgenic mouse and rat mutation assays with gpt delta mouse/rat, Big Blue mouse/rat and Muta Mouse are recommended for regulatory genotoxicity assays in vivo in OECD Test Guideline 488 [105]. For the reason that rats are more frequently used for toxicological studies and cancer bioassays than mice, the transgenic rat mutation assays are expected to be combined with 28-day repeated-dose toxicity studies [106].

In vivo mutagenicity of heterocyclic amines in transgenic rodents
Organ specificity and gender difference PhIP is the most abundant mutagenic and carcinogenic heterocyclic amine produced in cooked meat and fish [14]. It induces colon and prostate cancers in male F344 rats and mammary cancer, but not colon cancer, in female rats [107,108]. Okonogi et al. [109] examined the mutagenicity of PhIP in the colon of male and female Big Blue rats and concluded that the MFs in the colon mucosa were enhanced by treatment with PhIP, but there were no gender differences in the MFs (Table 3). Masumura et al. [110] examined the organ specificity of Fig. 2 Protocols of gpt delta transgenic rodent mutation assays. Gpt delta mice or rats are exposed to chemicals by feeding, gavage or others. Then, the genomic DNA is extracted from various organs or tissues to recover λ phage EG10 particles by λ packaging reactions. Then, the rescued phages are introduced to indicator E. coli for gpt selection and for Spi − selection that detect point mutations and deletion mutations, respectively. DNA is extracted from 6-TG-resistant colonies or Spi − plaques for DNA sequencing      PhIP-induced mutations in male and female gpt delta mice and reported that the highest MF was observed in the colon, followed by the spleen and liver. There were no gender differences in the MFs in the colon and liver. Stuart et al. [111] also examined the organ specificity of PhIP in Big Blue rats and reported that the MF in the colon was higher than that in the cecum and also that no gender differences were observed in the MFs in the colon. IQ induces intestinal tumors and hepatocellular carcinoma but not in the kidney of rats [141][142][143]. Bol et al. [127] examined the mutagenicity of IQ in Big Blue rats and reported that the highest MF was observed in the liver, followed by the colon and kidney, a non-target organ. The higher MF in the liver than in the colon induced by IQ was also reported by Moller et al. [128]. MeIQ induces tumors in the Zymbal gland, oral cavity, colon, skin and mammary glands in F344 rats and tumors in the liver and forestomach of CDF1 mice [143]. Suzuki et al. [132] examined the mutagenicity of MeIQ in female Big Blue mice (C57BL/6N) and reported that the highest MF was in the colon, followed by the bone marrow, the liver and the forestomach. MeIQx induces liver tumors in CDF1 mice where the female mice are more susceptible than males, but does not induce tumors in the colon in both sexes [144]. Itoh et al. [135] examined the mutagenicity of MeIQx in Big Blue mice (C57BL/6) and reported that the MF in the liver was higher in female mice than in males. They also observed an increase in MFs in the colon, a non-target organ for carcinogenesis, where no obvious differences in MFs between male and female were observed. Mutagenicity in the colon of mice has also been reported in male gpt delta mice fed a diet containing MeIQx [136]. APNH is formed from aniline and norharman in vitro and in vivo and induces liver and colon cancers in F344 rats [145]. The in vivo mutagenicity of APNH was examined in male gpt delta mice fed a diet containing 10 or 20 ppm of APNH for 12 weeks [139]. The MF was higher in the liver than in the colon, and the MF in the liver of the mice at 20 ppm was almost equivalent to that of the liver in the same mice fed a diet containing 300 ppm MeIQx for 12 weeks [136]. ABAQ is a heterocyclic amine formed from glucose and L-tryptophan via the Maillard reaction. ABAQ has a tumor initiating-activity in the colon of mice [146]. The in vivo mutagenicity of ABAQ was examined in male gpt delta mice treated by gavage for 3 weeks at 25 or 50 mg/kg [140]. The MFs in the liver increased in a dose-dependent manner, and no MF was enhanced by the treatments in the kidney. AαC is the second most abundant heterocyclic amine in very well-done meat and fish [147]. It induces cancers in the liver and blood vessels of CDF1 mice [148]. The in vivo mutagenicity of AαC was examined in F1 (C57BL/6 x SWR) mice with lacI as a reporter gene [112]. AαC enhanced MFs in the colon but not in the small intestine.

Mutation spectrum
Mutagens induce specific types of sequence changes in the genome, such as T to C mutations by ethyl nitrosourea, G to T mutations by benzo[a]pyrene and CC to TT by ultraviolet light irradiation. DNA sequence changes associated with mutagenic treatments are called the "mutation spectrum". In particular, specific sequence changes in cancer cells are called "mutational signatures," which are important clues for investigating the causes of human cancer [149,150]. PhIP induces colon cancer in male F344 rats where the adenomatous polyposis coli (Apc) gene is mutated by a guanine deletion at a 5'-GGGA-3' [151]. Okonogi et al. [109] examined the mutation spectrum in the colon of Big Blue rats fed a diet containing PhIP and reported that one bp deletion was the most frequent mutation, including a guanine deletion at 5'-GGGA-3' in male and female rats. Okochi et al. [116] investigated the mutation spectrum of mammary glands in female F1 (Big Blue rat x SD) rats  [112]. [115] administered 10 gavages of PhIP and concluded that G: C to T:A transversions were the most frequent mutations, followed by G:C deletions including G:C deletions at a 5'-GGGA-3'. Stuart et al. [119] examined the mutation spectrum in the prostate of Big Blue rats fed a diet containing PhIP and concluded that the predominant mutations were G:C to T:A transversions and deletions of G:C bp. In mice, Lynch et al. [114] treated Muta Mouse with PhIP and examined the mutation spectrum in the intestine. Approximately 40% of the total mutations were G:C to T:A transversions and 20% were G:C deletions, which were similar to those observed in the Hprt and DHFR genes in hamster and human cells in vitro. Okonogi et al. [115] examined the mutation spectrum of PhIP in the colon of Big Blue mice and reported that approximately half of the mutations were G: C to T:A transversions, in particular in runs of guanine, and approximately 1/4 of the total mutations were G:C deletions. In the colon, the rate of G:C to T:A transversions is significantly higher in mice than in rats [109]. Masumura et al. [117] treated male gpt delta mice with PhIP and reported that G:C to T:A transversions and G: C deletions in particular in 5'-TTTTTTG-3' to 5'-TTTTTT-3' were predominant mutations in the colon detected by gpt and Spiselections, respectively. Overall, it seems that PhIP induces G:C to T:A transversions and G: C deletions and that the transversions are more frequently induced in mice than in rats. IQ predominantly induces G:C to T:A transversions in the liver of gpt delta rats and also in the liver and colon of Big Blue rats [127,130]. G:C to T:A was also induced by MeIQ in the liver, bone marrow and colon of female Big Blue mice [115,133], APNH in the liver and colon of male gpt delta mice [139] and AαC in the colon of Big Blue mice [115]. Mutational hot spots for G:C to T: A transversions by PhIP, MeIQ and AαC are in runs of guanine, at 5'-GC-3' and in 5'-CGT-3', respectively [115].

No-observed effect level (NOEL) of in vivo mutagenesis
Toxicological assays, including in vivo mutagenicity assays of chemicals, are conducted at high doses, i.e., the maximum tolerable doses (MTDs), which are often 1,000 or 10,000 times higher than the human exposure levels in daily life. Therefore, it is unclear whether the toxicity or mutagenicity observed at high doses can also be observed at low doses where humans are actually exposed to the chemical [152]. Lynch et al. [113] examined the mutagenicity of PhIP in Muta mice treated by oral gavage at doses of 0.2, 2 and 20 mg/kg for 4 days and reported that PhIP was mutagenic only at a dose of 20 mg/kg in the large intestine and liver. No mutagenicity was observed in the kidney, even at 20 mg/kg. They suggested that 2 mg/kg may be a potential threshold dose for PhIP-induced mutagenesis. They argued, however, that the dose may be a detection limit instead of a threshold because of the high spontaneous MFs in the liver of Muta mice. Gi et al. [131] examined the mutagenicity of IQ in male F344 gpt delta rats fed diets at doses of 0.1, 1, 10 or 100 ppm for 4 weeks and reported that gpt MFs were significantly enhanced over the control level at doses of 10 and 100 ppm but not at 0.1 and 1 ppm in the liver. They reported, however, that the frequencies of G:C to T:A transversions were significantly enhanced over the control level at a dose of 1 ppm in addition to 10 and 100 ppm and that the increase in the frequencies was dose-dependent. It suggests that DNA sequencing analysis may enhance the sensitivity of mutation detection, thereby lowering the no-observed-effect level (NOEL). Masumura et al. [136] examined the mutagenicity of MeIQx in male gpt delta mice fed diets containing MeIQx at doses of 3, 30 or 300 ppm for 12 weeks. The MFs in the liver significantly increased at doses of 30 and 300 ppm but not at 3 ppm. The frequency of G:C to T:A did not significantly increase at 3 ppm, either. In this case, DNA sequencing analysis did not affect the NOEL. Hoshi et al. [137] examined the mutagenicity of MeIQx in male F344 Big Blue rats fed diets at doses of 0.01, 0.1, 1, 10 or 100 ppm for 16 weeks and reported that the MFs significantly increased at 10 and 100 ppm in the liver. In addition, they examined glutathione S-transferase placental form (GST-P)-positive foci in the liver, which is a marker for hepatocarcinogenesis. The number of GST-P-positive foci significantly increased beyond the number of the control group only at a dose of 100 ppm. They suggested that the NOEL for in vivo mutagenesis was lower than that for carcinogenesis.

Implication of in vitro and in vivo mutation assays
The discovery of carcinogenic heterocyclic amines is one of the most fruitful scientific achievements enabled by the Ames test. Before this test was developed, the identification of chemical carcinogens solely depends on timeconsuming animal tests. Multiple validation studies with more than 2,000 chemicals revealed that approximately 70-90% of chemical carcinogens are positive in the Ames test [153]. Therefore, this test is adopted in OECD test guideline 471 [154] and is widely used to eliminate potential carcinogens from pre-marketing chemicals developing for pharmaceuticals, pesticides, food additives and others. Owing to the power of the Ames test, it was initially expected that strong mutagens in the Ames test might be strong carcinogens in rodents. However, studies with a large database indicated that the potency in the Ames test does not quantitatively correlate with that in rodent carcinogenicity assays [155]. The lack of quantitative relationships between mutagenesis in bacteria and carcinogenesis in rodents may not be very surprising when considering the complex process of carcinogenesis such as mutation or initiation, promotion and progression. Since in vivo mutagenesis is much simpler than carcinogenesis, it was expected that the potency of the Ames test might correlate with that in transgenic mutation assays quantitatively. Although the mutagenic potency (revertants per μg) of MeIQ in strain TA98 is more than 300 times higher than that of PhIP [6], the MF of MeIQ in the colon of Big Blue mice fed a diet containing 300 ppm for 90 days is similar to that of PhIP in the mice fed a diet containing 400 ppm for 90 days [156]. It seems, therefore, that the mutagenic potency of the Ames test does not quantitatively correlate with the potency in in vivo mutation assays. It is also pointed out that the potency of the Ames test does not quantitatively correlate with that in in vitro mammalian cell assays for gene mutation and chromosome aberrations [153]. Despite the lack of quantitative correlations, the power of the Ames test to qualitatively predict potential carcinogens is outstanding, as evidenced by the successful discovery of carcinogenic heterocyclic amines.
Transgenic rodent mutation assays have enabled us to analyze chemical-induced mutations in various organs and tissues at the sequence level. Therefore, it would be interesting to examine whether we can predict target organs and sensitive gender for carcinogenesis based on the high MFs in specific organs and gender of rats and mice. Thus, the MFs were compared between the target organs and non-target organs for carcinogenesis, and the gender specificity in mice and rats was examined. However, the MFs in various lobes of the prostate were almost equally sensitive to the mutagenicity of PhIP, while the ventral prostate was the only target for cancer induction in rats [108,125]. MeIQ induces much higher MF in the colon than in the liver, but the cancer incidence is higher in the liver than in the colon in mice [132,157]. PhIP induces mutations in the colon of male and female rats, while colon cancer is induced only in males [107][108][109]111]. MeIQx induces mutations in the colon of male and female mice, but it does not induce tumors in the colon [135,136,144]. These results indicate that target organs or tissues for carcinogenesis do not necessarily exhibit higher MFs compared to other organs or tissues, and also that mutations can be induced regardless of the gender specificity for carcinogenesis. In other words, the organs or tissues that are positive in the transgenic mutation assays are not necessarily carcinogenic targets. It appears, however, that tumors are induced in organs where mutations are induced when the carcinogens are genotoxic. Therefore, the transgenic mutation assays are employed to distinguish genotoxic carcinogens from non-genotoxic carcinogens [158]. The results of the transgenic mutation assays reflect in vivo metabolism and mammalian DNA repair, while the results of the Ames test reflect in vitro metabolism of S9 and bacterial DNA repair. Hence, the in vivo mutation assays may be useful to narrow down genotoxic carcinogens from chemicals that are positive in the Ames test. In fact, International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) M7 for regulation of mutagenic impurities in pharmaceuticals recommends conducting in vivo mutation assays when the chemical is mutagenic in the Ames test [159]. Research on carcinogenic heterocyclic amines has provided valuable lessons on the effectiveness and limitations of in vivo transgenic mutation assays.
Since carcinogenic heterocyclic amines are produced by cooking, a question is whether they induce cancers in humans. If so, the extent to which they impose cancer risks on the general population? The exposure levels of heterocyclic amines are reported to be less than 500 ng per person per day [6]. In general, genotoxic carcinogens are regulated under the policy that they have no threshold or safe doses [152,159,160]. Therefore, there is carcinogenic risk to people who take carcinogenic heterocyclic amines. However, humans have various protective mechanisms against mutagenic substances such as detoxification, DNA repair, error-free translesion synthesis and apoptosis [161]. It is expected, therefore, that low-dose exposure to mutagenic carcinogens may be negligible due to these mechanisms. In addition, people are constantly exposed to endogenous mutagens such as reactive oxygen species. Thus, mutagenic risk is inevitable in humans. European Food Safety Authority (EFSA) and World Health Organization (WHO) propose 150 ng per person per day as a sufficient protective threshold of toxicological concern (TTC) for DNA-reactive genotoxic chemicals [162,163]. Several studies with transgenic rodents exposed to low levels of carcinogenic heterocyclic amines have suggested the presence of NOEL [113,131,136,137]. Although TTC is a concept that was developed to prioritize chemicals that require further toxicological evaluation and NOEL does not mean the absolute safe level, there may be certain exposure levels for genotoxic carcinogens, which do not increase excess lifetime cancer risk substantially. However, humans are exposed to multiple chemicals. Therefore, the combined risk should be evaluated. It has been reported that six carcinogenic heterocyclic amines, each of whose doses was below non-detectable levels by the Ames test, became mutagenic when they were combined [164]. In addition, chemicals may exhibit co-mutagenic effects and produce mutagenic substances when more than one non-mutagenic substance is combined [165]. Risk assessment of multiple exposures to DNA reactive mutagenic carcinogens at low levels may be a challenge that research on carcinogenic heterocyclic amines has proposed us.