Inhaled crocidolite mutagenicity in lung DNA.

We used transgenic mice carrying the lacI reporter gene to study the mutagenesis potential of asbestos crocidolite. The animals were exposed by nose-only inhalation to an aerosol containing 5.75 mg/m(3) crocidolite dust for 6 hr/day and 5 consecutive days. After 1, 4, and 12 weeks, we examined four end points: the cytology of bronchoalveolar lavage, the lung load of crocidolite, the hydrophobic DNA adducts, and the mutations in the lacI reporter gene. Twelve weeks after exposure, nearly 10% of the inhaled fibers remained in the lung (227 +/- 103 ng/mg lung). There was evidence of a typical inflammatory response consisting of multinucleate macrophages at weeks 4 and 12, whereas immediately after the exposure, we observed numerous polymorphonuclear neutrophils. The mutant frequency significatively increased during the fourth week after the exposure: 13.5 [time] 10(-5) in the exposed group versus 6. 9 10(-5) in the control group. The induction factor, defined by the ratio of checked mutants of exposed mice to checked mutants of control mice, was 1.96. The mutation spectrum of control lung DNA and exposed lung DNA was similar, suggesting the possible involvement of a DNA repair decrease in crocidolite-treated animals. We used the (32)P-postlabeling method and did not detect any increase of either 5 mC or bulky adduct in treated mice. This is the first study that demonstrates asbestos mutagenicity in vivo after a nose-only inhalation.

Although the use of asbestos fibers has been banned in most industrialized countries, they are still a major environmental, occupational, and personal health concern. Asbestos fibers are considered tumorigenic and mutagenic for humans according to the U.S. Registry of Toxic Effects of Chemical Substances (1).
Physical dimension, biopersistence, surface reactivity, and fiber overload are all involved in the pathogenesis of asbestosis, bronchogenic carcinoma, and mesothelioma [reviewed by Kane (2)]. At the molecular level, oxidative stress and cytokine release by alveolar macrophages (AMs) and neutrophils are invoked in the pathogenesis of these illnesses (2).
Asbestos is thought to be a complete carcinogen (3), which means that it has the capability to initiate and promote tumors. Because asbestos fibers are not directly electrophilic, it has been proposed that asbestos is not a mutagenic compound per se (4). In fact, past in vitro studies have shown no mutagenic or weak mutagenic activity of asbestos samples [reviewed by Jaurand (5)]. Using the Salmonella typhimurium TA 102 oxidative stress sensitive strain, Faux et al. (6) showed mutagenicity induced by crocidolite, but not by chrysotile (chrysotile is the most-used asbestos worldwide). However, the mammalian cell paradigm found mutagenicity for both crocidolite and chrysotile at the S, locus (7) and for 50 pg/mL chrysotile at the HLA-A locus.
Nevertheless, no mutagenicity was manifested at the HPRTlocus (7)(8)(9) with crocidolite fibers. Furthermore, crocidolite induced a loss of heterozygosity in vitro at the HLA-A locus (10), as did chrysotile in Big Blue transgenic Rat 2?. cells (Stratagene, La Jolla, CA) at the X locus (11). Although many experimenits have been performed in vitro to demonstrate the mutagenicity of asbestos (clastogenic effects and mitotic abnormalities), there is a lack of in vivo studies to assert this important property. Thus we investigated more extensively the in vivo mutagenesis potential, if any, of asbestos fibers. To our knowledge, no mutagenesis studies have been conducted on whole animals. We used a transgenic mouse system carrying the lacI reporter gene and investigated the effect of crocidolite fibers in a short-term assay. The animals were exposed by nose-only inhalation for 5 days, then four end points were analyzed: the inflammatory response of the lung, the lung burden of fibers, the DNA adduct formation, and the mutation rate as well as the position and the nature of the mutation on the lacI reporter gene induced by crocidolite fibers.

Materials and Methods
Mice. We obtained transgenic male Lacl mice C57/B16 (lambda LIZ, BigBlue) from Stratagene. The mice were randomized and numbered by tattoo. The animals were housed in polycarbonate cages (1/cage) covered with spun-bonded polyester cage filters.
The room temperature was 21 ± 1°C, and the pressure was 5 mm H20 above the atmospheric pressure. The humidity ranged from 40 to 60% and a fluorescent light was on for 12 hr/day. The animals were fed with pellet food and water ad libitum. Twomonth-old animals were exposed to the crocidolite aerosol; control animals were housed in nose-only exposure tubes without crocidolite during the experiment (6 hr/day for 5 days). We terminated the experiment 1, 4, or 12 weeks after the beginning of the exposure, depending on the animal group.
Exposure techniques. The crocidolite sample used in this study came from a batch (gift of R.E.G. Rendall, National Center for Occupational Health, Johannesburg, South Africa) that replaced the original Union Internationale Contre le Cancer (Geneva, Switzerland) sample when it ran out. The fiber size distribution of the gift batch is extensively described in a previous study (12 Al, 0.05; Cr, < 0.002; and Ni, 0.017. The fiber aerosol generator was previously described by Rihn et al. (12). In brief, the fibers were packed into a cylinder and pushed with a Teflon-coated piston on a steel brush. Clean air was provided by a six-bar compressor that delivered a 100-L/min air stream by an inverted cyclone device. The tangential position of the air inlet pipe gave the air flow a helicoidal movement from the top to the bottom of the cell to ensure the aerosol homogeneity. The aerosol concentrations were monitored online by photometry.
We used the chamber as previously described by Rihn et al. (12). Briefly, the control process included measurements of the air flow, the depression, the temperature, and the humidity. We transferred all output signals to a computer, which processed sensor information, calculated regulation loops, and dispatched analogical tensions to the regulation valves. A second computer was connected to the first and was used to monitor the inhalation chamber conditions (air flow and relative humidity), the aerosol concentration, and the adjustments of the aerosol generator parameters. The temperature of the atmosphere was maintained between 20 and 22°C, the humidity between 40 and 60%, and the air volume exchange (100 L) was 60 times/hr. To avoid contamination of ambient air by inhalation of chamber pollutants, pressures inside the inhalation chamber and the glove box surrounding the chamber were maintained below the room pressure at 5 and 2.5 mm H20, respectively. Animals were housed in transparent restraining tubes inserted in the inhalation chamber, a stainless concentric cylinder with a 48-tube capacity. The mice were continuously exposed nose only to the crocidolite aerosol (subjects) or to the ambient air (control mice). Small holes in each of the restraining tubes prevented hyperthermal effects. This flow-past design allowed each animal to breathe fresh crocidolite aerosol of well regulated particle concentration and size at all times; reinhalation of exhaled test atmosphere was minimized. The fraction of exhaled air in the chamber was < 1/1,000 on the basis of regular mouse breathing (25 mL/min/mouse) (13).
Atmosphere control. We determined the concentration of airborne dust by sampling on GF/C filters (Whatman International Ltd., Maidstone, UK) for 5-6 hr at a flow rate of 2 L/min. The sampling head was an open-faced Millipore cassette (Millipore, Bedford, MA). The weight of the filters before and after sampling was corrected according to the weight variations of three unexposed filters.
To determine the fiber number, we chose a short sampling period (5-10 sec) to avoid fiber amounts that were too high fiber and to avoid filter overload. We used 0.8-pm mixed ester cellulose membranes to sample the fibers, which were counted by phase-contrast microscopy (12). We assessed the fiber size distribution of either the aerosol or the bulk crocidolite sample by transmission electron microscopy (IOOCX; JEOL Ltd., Tokyo, Japan). For this measurement, the fibers were deposited on a Nucleopore (Millipore) filter covered with a thin layer of carbon.
To check the absence of contaminating metals produced by the generator, we also sampled dust on Whatman QM-A filters and analyzed by emission spectroscopy (ICAP 61E; Thermo Jarrell Ash, Franklin, MA).
Cytologic, histologic, and ultrastructural examinations. At the end of each time-point, we measured bronchoalveolar liquids by washing the lung 2 times with 500 pL phosphate buffered saline solution (pH 7.4). We determined the cell number after Trypan Blue (0.04% w/v) staining, and the nucleated cells were recorded in triplicate using a Thoma-Weiss counting cell (PolyLabs, Strasbourg, France). In addition, a portion of the bronchoalveolar liquid (100 pL) was spun for 5 min at 29g (Cytospin 2; Shandon, Pittsburgh, PA). The May Griinwald Giemsa staining was then performed and AMs, polymorphonucleated cells (PMC), and the lymphocytes (L) were counted. Our results are the mean of three independent measures. For histologic observations, we fixed lung specimens from control and treated animals in Bouin's solution or in 4% (v/v) aqueous formalin solution, then embedded in paraffin, sectioned, stained with hematoxylin/eosin, and evaluated by adapting the Wagner scale as described previously (12) with an Ortoplan microscope (Leitz, Wetzlar, Germany).
To evaluate the fiber burden in the lungs, we cleaned the frozen left lobe of four control and four treated animals in acetone to remove the fat, dried them to constant weight, and reduced them to ashes in a lowtemperature oven (150°C) for 2 hr. The ashes were gently homogenized in ultrapure water by handshaking, then filtered over a carbon-covered Nucleopore membrane (25 mm diameter, 0.4 pm pore size). We did not choose ultrasonic treatment because we previously demonstrated that this treatment can change the fiber size distribution (12). The retained particles were covered with a second layer of carbon. Under this prepared filter, five electron microscopy grids were deposited onto the filtering apparatus. Ten milliliters of chloroform was filtered through the grids to dissolve the Nucleopore filters and retain the fibers, if any, on the grids. We observed the fibers with the transmission electron microscope (Jeol 100CX) at 33,OOOx magnification. This technique allowed us to determine the quantity of fibers and the volume and the mass that could be calculated with the knowledge of the density (d = 3.35 g/cm3). Finally, we determined the ratio of the fiber mass to the lung weight (expressed in nanograms per milligram lung).
Lung DNA extraction, in vitro packaging, and mutation analysis. We prepared high-molecular-weight lung DNA from a small piece of lung (100 mg) according to the protocol developed by Stratagene (14). After homogenization (six strokes in a dounce) using Rnace-it buffer (Stratagene), the DNA samples were phenol/chloroform extracted, then we carefully ethanol precipitated the samples and resuspended them in 500 pL TE (Tris-HCl, 10 mmol, EDTA 1 mmol, pH = 8.0). Before packaging, the DNA was electrophoresed in 1.5% agarose gel to check the size and control the absence of a smear below 50 kB. We recovered the lambda-shuttle vector in viable phage by incubating the extracted DNA (approximately 50 pg) with the terminase and the phage proteins contained in the Transpack (Stratagene) as described by the manufacturer. We matched control and exposed animals for each packaging reaction. The average of plaques per packaging reaction was 38,000, and varied from 19,000 to 69,000. There was no statistical difference for transformation efficiency between control and exposed animals. The in vitro packaged phages were adsorbed on E. coli SCS-8 cells and screened for blue mutant plaques on 625 cm2 NYZ agar plates containing 1.5 mg X-gal/ml top agar. The plaque density varied from 8,000 to 10,000 per assay tray. We applied internal color sensitivity control (CMO, CM1, CM2, and CM3; Stratagene) for each plating. Incubation lasted between 16 and 20 hr and the blue mutants were screened using a red filter.
We confirmed the phenotype of each mutant plaque by phage replating at low density in the presence of X-gal. The phage was further isolated by coring the plaque and resuspending the phage in the buffer described in Stratagene's protocol (14). We extracted DNA and amplified it with the appropriate primers of the lad gene for further sequencing (forward direction 1781: GACACCATCGAATGGTGAAAAC; reverse direction 1780: CCGCTCA-CAATTCCACACAACAT). The lad gene of 60 phages from both control and exposed mice was entirely sequenced using the six primer sets recommended in the BigBlue protocol (14) to determine the distribution of the mutations. We determined the sequences using the Abi prism dye terminator cycle sequencing ready reaction kit (Perkin Elmer Biosystems, Foster City, CA) on a 373A sequencer (Perkin Elmer Biosystems).
Screening for bulky adducts and determination ofm5dC level in DNA. We detected hydrophobic DNA adducts using the [32P]-postlabeling method, which had not previously been used, to check the presence of DNA adducts in asbestos studies. Briefly, the DNA was digested into 3'-phosphonucleosides by micrococcal nuclease and spleen phosphodiesterase. Adducts were privileged versus normal nucleotides by nuclease P1 enrichment. We postlabeled the adducts using [y 32P]-adenosine triphosphate (ATP) Volume 108, Number 4, April 2000 * Environmental Health Perspectives and T4 polynucleotide kinase, as described previously (15 Schleicher and Schiill, Dassel, Germany) and revealed using the solvents as previously mentioned (15). We used radioautography to detect the labeled nucleotides and we evaluated the radioactivity using a Bio-Imager Bas2000 analyzer (Fuji, Tokyo, Japan).

Results
Measurement of the fiber concentration in the chamber and in the lung ofthe mice. We measured the dust concentrations in the chamber by sampling onto glass filters 6 hr/day, as described in "Materials and Methods." The dust concentration mean was 5.75 ± 0.87 mg/m3. The average fiber density was 1,875 ± 1,216 fibers/cm3 (n = 24) when we considered only the fibers > 5 ,um in length and < 3 ,um in diameter. The fiber size distribution was similar to that of a previous study (12). The proportion of fiber that met Stanton's criteria (16) did not vary over the experiment time and remained < 3% of the overall fibers; thus the crocidolite aerosol used in our study was mainly composed of short fibers ( Table 1). The elementary fiber analysis either in the inhalation chamber or in the mice lungs did not show any difference as compared to the initial crocidolite sample (data not shown).
We checked the lung fiber content in four control and four exposed mice 1 week and 12 weeks after the intoxication. No fibers were detected in control mice after either 1 or 12 weeks. Indeed, the measured mass of fiber was < 1 ng/mg lung ( Table 2). In contrast, exposed animals had an average of 693 ± 496 ng fibers/mg lung corresponding to 1.63 X 106 (± 1.19 x 106) fibers/mg lung 1 week after the intoxication. Twelve weeks later, the lung content was still 227 ± 103 ng/mg corresponding to 0.62 x 106 (± 0.21 x 106) fibers/mg lung. When considering normal breathing of 25 mL/min air, on average, 17.1 and 6.6% of the total inhaled fibers (estimated approximately to 84.4 x 106) lasted 1 and 12 weeks, respectively, in the mice lung after the intoxication. The average of the length-on-diameter ratio was 19.15 and 19.88 at 1 and 12 weeks after the exposure, respectively (p > 0.05). Moreover, the fiber size distribution remained unchanged in the lung after 12 weeks and was similar to the fiber size distribution of the aerosol sample (Table 1). Cellular responses ofthe bronchoalveolar liquids. Table 3 shows the cellular response of the 1-, 4-, and 12-week series. One week Table 1. Size distribution of the fibers in the exposed animals and of the crocidolite aerosol. Sample L <5 pm 5   after the intoxication, the number of viable cells elevated dramatically because of a huge increase of PMCs and lymphocytes. However, the number of AMs with two nuclei was approximately 13 times higher in the treated group as compared to the control group. In addition, syncitia composed of three or more macrophages appeared in the exposed group. Four weeks after the intoxication, only AMs with two or more nuclei were increased. Twelve weeks after the intoxication, the number of AMs was still increased and the AM population was responsible for the increase in total viable cells. The averages of Wagner scores were 1.0 and 1.7 for control and exposed mice, respectively, indicating an alveolar inflammatory response but no fibrosis in the lung ( Table 2).
Mutant frequency and mutation spectrum determinations. More than one million genetically independent events were analyzed for each mouse set. We used approximately 55-60 transpack reactions for each time point. The mutant frequency significantly increased 4 weeks after the intoxication: 13.5 x 10-5 ± 1.7 x 10-5 in the exposed group versus 6.9 x 10-5 ± 1.1 x 10-5 in the control group ( Table 4). The induction factor, defined by the ratio of checked mutants to exposed mice versus checked mutants of control mice, was 1.96 (p < 0.05). In contrast, 1 and 12 weeks after the intoxication, the mutant frequencies of both mice groups were not significantly different.
We determined the mutation spectrum 4 weeks after the exposure for 60 mutant plaques that were sequenced in both groups. Considering the 60 lad genes from mutant plaques sequenced for each group, there were no significant differences between control and exposed mice, as mentioned in Table 5.
We located the mutation positions for both groups (Figure 1). No difference appeared in the location between exposed and control mice (65 and 62 mutations determined, respectively). Similarly, there were no differences in either the 1-week or the 12-week groups (data not shown).
Bulky adducts and methylation status. The formation of bulky DNA adducts was ruled out by the [32p] postlabeling method (15). As shown in Figure 2A   46,100 ± 13,700* AM1, AM2, and AM3 are alveolar macrophage with one, two, and three or more nuclei, respectively. *Significant at p < 0.05 calculated using the Student's t-test (n = 9; three independent measures per animal).
Environmental Health Perspectives * Volume 108, Number 4, April 2000 was no difference in the DNA adduct profiles: both exhibited the same adducts, i.e., Icompounds generated by the endogenous metabolism. A slight decrease in the total amount of these I-compounds was detected: 160 versus 60 per 109 nucleotides for control and treated mice, respectively. In addition, the m5dC/dC ratios measured in lung DNA were 3.0 ± 0.7 (n = 16) in control mice versus 3.5 ± 0.4 (n = 16) for the exposed mice. There was no statistical difference in the m5dC/dC ratio (p > 0.01) because it was assessed by radioactivity counting of m5dC and dC spots on thin-layer chromatography by phosphorimaging.

Discussion
The cumulated dose of asbestos administrated in our study by inhalation was low (5.75 mg/m3 for 5 days) as compared to previous long-term assays on rodents (10-50 mg/m3 for 200-500 days) (17). Taking into account the average ventilation volume of a mouse, the total inhaled dose of inhaled dust was 0.26 mg/mouse, approximately corresponding to < 0.5% of the average inhaled dose in classical long-term assays. However, it is remarkable that 6.6% of the inhaled dose was still in the lung 12 weeks after the end of the intoxication. Contrary to other studies, there was no overload effect in our study because < 10% (_ 9 x 106) of the total inhaled fibers was present in the lung 4 weeks after the inhalation when the mutagenic effect was significant. The Wagner scores of both groups were similar, indicating that the crocidolite dose was far below the maximum tolerated dose and thus avoiding overload effects (18). Moreover, the length/diameter ratio and the size distribution of the fibers deposited in the lung were similar 1 and 12 weeks after a 5-day exposure (Table 1). This similarity shows that in this period of time all fibers, whatever the size, had the same probability to be cleared, which contrasts with long-term assays in which short fibers are removed from the lung more rapidly than long fibers (19). It would be interesting to study the mutagenic effect of an asbestos batch composed mainly of long fibers (> 20 pm). Coin et al. (19) showed that chrysotile long fibers, which are more carcinogenic (16), remained for at least 6 months in the lung after a short treatment comparable to the present study.
Asbestos fibers induce chromosomal changes in human cultured cell lines.
Crocidolite and chrysotile (2 mg/cm2) induced anaphase aberrations and abnormal mitoses in human immortalized mesothelial cell lines (20). In human lymphocytes, chromosome and chromatid breaks have also been reported, as well as polyploidy, provided that the dose was higher than 2 x 105 fibers/cm2 (approximately 5 fibers/cell) (5). This dose is higher as compared to the dose used in vivo in the present study. Considering a total lung area of 700 cm2 for a 3-month-old mouse (13) and a total number of alveolar type I and II cells of 26 x 106, the number of fibers per cell can be estimated 4 weeks after the intoxication as 0.35 fibers/cell. This value is not surprising because generally the doses needed to show in vitro effects are higher than those needed to show in vivo effects. After our study, we can conclude that mutations occur using a dose below the dose necessary for chromosome abnormalities. Asbestos fibers also induce binucleation, which is a consequence of mitotic abnormalities in pleural mesothelial cells and macrophages. In the current study, binucleated and multinucleated AMs, known as biomarkers of fiber burden, increase significantly after 4 and 12 weeks: these results were similar to those published previously (12).
The first in vitro gene mutation assay (i.e., the Salmonella typhimurium assay) was not relevant for testing the mutagenicity of crocidolite. An improvement of the Ames test using the TA102 strain showed a mutagenic potential of 50 pg/cm2 crocidolite because of its sensitivity to oxidative damage (6). The hypoxanthine guanine phosphoribosyl transferase locus (HPRI) has been investigated in rat liver, Chinese ovary, or human-hamster fibroblasts, but a mutagenic effect of crocidolite could not be demonstrated for any of the the cell systems used [reviewed by Jaurand (4)]. Therefore, it appears that the BigBlue transgenic system allows a more realistic picture of the genotoxic effects of an asbestos exposure in vivo.
In such studies, a nonessential and silent gene, the so-called reporter gene, was used for mutation studies. With this latter system, each point mutation is detected in contrast to assays based on the HPRT locus, which are generally phenotypic end points. Recently, however, Park and Aust (21) improved the HPRT locus assay by using hprtand gpt+ G 12 cells where the transgenic gpt gene is located on an autosomal chromosome, making this assay useful for asbestos mutagenicity testing. The improved assay showed a 2-fold increase of the mutant frequency with a 6-pg/mL dose of crocidolite.
Recent studies using transgenic rodents exposed to benzo[a]pyrene compared the mutant frequencies in both hprt and lad genes. The lad genes were 40 times more sensitive to a 150-mg/kg benzo[a]pyrene treatment in splenic T cells (22). Such disparity could be due to the phenotypic expression of the concerned gene. In fact, lad does not require in vivo expression, whereas the resistant 6-thioguanine phenotype (hprf) is required in T cells. Other factors could be the different rate of benzo[a]pyrene adduct formation and repair processes in both genes. The lad is a silent gene in the mouse; it is possible that the repair, before translation, is involved in a more efficient manner in the hprt gene. Therefore, it appears that a nonexpressed ladtransgene is more appropriate to study the genotoxicity of compounds or particles such as asbestos than other end points based on the loss of heterozygosity of the HLA locus or hprfphenotyping.
In our study, we sequenced the lad gene using the six primer sets in > 60 mutant colonies isolated from exposed and control mice that represented > 576,000 sequenced bases. We assessed the nature and the position of each mutation. The mutation frequency of control mice in our study varied from 6.9 x 10-5 to 8.7 x 10-5. These values are slightly higher than the spontaneous mutant frequency determined in lung by de Boer et al. (23). This variation could be due to differences in the experimental conditions (e.g., the age of the mouse, the plating density, or DNA extraction). Table 5 and Figure 1 show that the mutation spectrum is similar in both groups, suggesting that the increased number of mutations could be due to an oxidative stress directed randomly against all nucleotides rather than to a direct and specific mutagenic effect of fibers, as is seen for initiating compounds such as o-anisidine and benzo[a]pyrene. The reactive oxygen species (ROS) could be produced by the macrophages and neutrophils burst, as assumed by Hei et al. (24), or by extracellular compounds due to fiber surface reactivity (25)(26)(27). ROS induce mostly deletions and insertions (28); Lezon-Geyda et al. (11) used an in vitro assay based on the same principle (10). Moreover, the BigBlue test is unable to detect large insertions and deletions of DNA because these phenomena change the size of the X phage genome, which cannot be packaged. However an in vitro adaptation of the BigBlue system, using cultured Rat 2 X fibroblasts, allowed Lezon-Geyda et al. (11) to demonstrate a 2.5-fold increase of the mitotic recombination after a 6-gg/mL treatment of crocidolite, thus making this test useful for insertion/deletion studies. Although the possibility of mutations through large insertions and deletions cannot be excluded in our test conditions, and because we found a clear increase of the mutant frequency, the exact mutation frequency due to asbestos is probably underestimated. Position on the faclgene Figure 1. Mutation position on the lacl gene. Sixty-five control and 62 treated mutant plaques were sequenced for both mice groups 4 weeks after exposure. For insertion and deletion, only the first modified base was marked.
It is also possible that asbestos impaired the repair potential of the cells, as shown by Okayasu et al. (29). An 8-pg/mL dose of chrysotile decreased the cell survival and the double strand repair efficiency of double strand repair deficiency cells (xrs-5) as compared to wild-type cells (Chinese hamster ovary cell). The repair potential of DNA was not verified in the present study. Regardless of the mechanism involved, one can hypothesize that the mutant frequency must have increased because of oxidative stress above the repair ability of lung DNA.
McGavran et al. (30) showed that a short-term chrysotile exposure (4 mg/m3) induced cellular proliferation, as assessed by in vivo [3H]-thymidine incorporation, in epithelial as well as interstitial lung cells. Adamson (31) reported similar findings in lung cells after a single intratracheal administration of 0.1 mg crocidolite. Adamson (31) showed a proliferation of mesothelial and fibroblast cells that peaked at 2% labeled nuclei and lasted for 3 weeks. Such an effect was also reported on mesothelial cells in Fischer 344 rats exposed to 8 mg/m3 crocidolite for 20 days (32). Using our experimental design (a moderate dose of 0.26 mg crocidolite for 5 days and a physiologic route for the toxicant), one can hypothesize that a proliferation occurred in lung. Mitogenic factors released in the milieu (33,34) by the oxidative stress may fix the mutations shown 4 weeks after the crocidolite exposure. The decrease of the mutant frequency in the exposed mice of the 12-week series could be explained by epithelial regeneration and restored DNA repair after the acute exposure period. The induced cell death could also explain the normalization of the mutant frequency seen after 12 weeks. Indeed, as demonstrated by Broaddus et al. (25), a fiber dose of 5 pg/mL induced apoptosis on mesothelial cells in a ROS-induction manner because the apoptosis process was inhibited by extracellular catalase and superoxide dismutase enzymes.
Adachi et al. (35) reported an increase of dOH8dG after an in vitro coincubation of deionized crocidolite and DNA. Furthermore, Takeuchi and Morimoto (36) showed an increase of this modified nucleotide in HL60 human cells treated with crocidolite and phorbol myristate acetate, a strong mitogenic factor as well as an inducer of ROS. However, they measured the total cell DNA, which could have been contaminated by mitochondrial DNA. dOH8dG is an early indicator of mitochondrial DNA injury (37). Future studies should determine the dOH8dG level in both nuclear and mitochondrial DNA. In our study, the mutation spectra of both groups of animals showed no statistical difference in the involvement of CpG sites (Table 5). Because mitochondrial DNA represents < 1% of total DNA, our results were not able to reveal a specific increase of this kind of mutation on dG nucleosides therein.
We did not find an excess of adducts in the present study in exposed mice. This is not surprising because the [32P]-postlabeling method detects only the bulky adducts. Inversely, the number of bulky adducts per genome seems more important in the control mice as compared to exposed animals: a similar finding was already noted by Randerath et al. (38), who found a gradual depletion of Icompounds during carcinogenesis in rodents. Howden and Faux (39) reported the presence of DNA adducts by using a fluorescence technique and demonstrated the role of lipid peroxidation products as malondialdehyde. This in vitro paradigm, however, was far from the physiological conditions. Moreover, the [32P]-postlabeling method is more sensitive and detected only I-compounds using an in vivo paradigm in both exposed and control animals.

Conclusions
We demonstrated a significant increase of the mutant frequency (approximately twice that of the control) of lung DNA for the first time in crocidolite-exposed mice as compared to nonexposed mice. Our study is the first attempt to quantify and qualify the mutagenic potential in vivo with respect to a physiologically based and controlled mode of exposure (nose-only inhalation) without overload and with the respect to AM clearance of fibers. Our study shows that mutagenesis is an early event in fiber intoxication, which could be evoked in cell transformation due to asbestos fibers. In addition, a low dose of fibers, as compared to previous studies, is sufficient to observe mutagenesis in situ in lung DNA. The mice paradigm developed in this study should now be experienced with man-made mineral fiber with (e.g., ceramic fibers) and without (e.g., glass fiber) known carcinogenic potential in rodents. It would also be useful to measure the mutant frequency in vivo using asbestos of various sizes to correlate, if possible, the mutagenic and the carcinogenic potential of fibers in vivo.