Sister chromatid exchanges and micronuclei in peripheral lymphocytes of shoe factory workers exposed to solvents.

We examined sister chromatid exchanges (SCEs) and micronuclei (MN; cytokinesis-block method) in cultured peripheral lymphocytes from 52 female workers of two shoe factories and from 36 unexposed age- and sex-matched referents. The factory workers showed an elevated level of urinary hippuric acid, a biomarker of toluene exposure, and workplace air contained high concentrations of various organic solvents such as toluene, gasoline, acetone, and (in one of the plants only) ethylacetate and methylenediphenyl diisocyanate. The shoe factory workers showed a statistically significant higher frequency of micronucleated binucleate lymphocytes in comparison with the referents. This finding agreed with three preliminary MN determinations (each comprising 27-32 shoe workers and 16-20 controls) performed in one of the plants 2-5 years earlier. The shoe factory workers also had a lower average level of blood hemoglobin than the referents. In contrast, no difference was found between the groups in SCE analysis. Smokers showed significantly higher mean frequencies of SCEs per cell and high frequency cells (HFC) than nonsmokers. Aging was associated with increased MN rates and reduced cell proliferation. Polymorphism of the glutathione S-transferase M1 gene (GSTM1) did not affect the individual level of SCEs; but in smoking shoe workers an effect of the occupational exposure on the frequency of micronucleated cells could be seen only in GSTM1 null subjects. The low prevalence of the glutathione S-transferase T1 (GSTT1) null genotype precluded the evaluation of the influence of GSTT1 polymorphism. Our results show that the shoe factory workers have experienced genotoxic exposure, which is manifest as an increase in the frequency of MN, but not of SCEs, in peripheral lymphocytes. The exposures responsible for the MN induction could not be identified with certainty, but exposure to benzene in gasoline and methylenediphenyl diisocyanate may explain some of the findings.

During the last few years, genotoxicity biomarkers have received considerable interest as tools for detecting human genotoxic exposure and effects, especially in health surveillance programs dealing with occupational exposure to chemical carcinogens. Currently, only cytogenetic end points in peripheral blood lymphocytes allow a reasonable epidemiologic evaluation of cancer predictivity. The largest databases are available for chromosomal aberrations (CA); high CA level has been associated with increased cancer risk (1,2). Evaluations have also been performed for lymphocyte sister chromatid exchanges (SCEs) and micronuclei (MN), but thus far these biomarkers have not been shown to predict cancer (1,2). These findings may reflect the relatively small databases, the comparatively young cohorts, and variation in techniques, which have made it difficult to standardize individual SCE and MN values obtained from different sources. In a few years, accumulation of more uniform data will make it possible to reassess the value of both SCEs and MN. In the meantime, SCEs and MN are used as biomarkers of genotoxic exposure, as simpler alternatives to CA analysis. In the present study, we applied these methods to monitor solvent-exposed workers from two Bulgarian shoe factories.
People employed in the shoemaking industry are at an increased risk of leukemia and nasal cancer (3), and an excess of mortality due to other types of cancer has also been reported (4)(5)(6). Workers in shoe and boot factories are exposed to a mixture of organic solvents, among which toluene and acetone are usually the most common. Neither of these solvents is considered a genotoxin or a carcinogen; the weight of evidence from human in vivo studies suggests that exposure to toluene does not cause somatic cell genotoxic damage (7), although this view has been questioned by recent studies of rotogravure printers (8,9).
The glues and gasoline used in shoe manufacture may contain benzene, which could be responsible for some of the cancers found in shoe workers. Benzene is a wellknown clastogen that requires metabolic activation to be mutagenic. The genotoxic metabolites are also thought to play an important role in benzene myelotoxicity and leukemogenesis. The quinone metabolites of benzene can break chromosomes by inducing reactive oxygen species but may also act as aneuploidogens, causing microtubule disruption (10,11). Recent investigations have indicated that structural CA are increased in shoe factory workers exposed to benzene and toluene (12,13).
In a preliminary study of the genotoxic effects of occupational exposure to organic solvents, using the alkaline Comet assay, we did not detect any increase in DNA damage in cryopreserved peripheral blood mononuclear leukocytes from a group of women employed in the two Bulgarian shoe factories (14).
Here we report results of lymphocyte SCE and MN analyses in a larger group of shoe workers from the same plants. We also present MN data from three other cross-sectional studies conducted 2-5 years earlier than the last sampling. In all cases, the concentrations of toluene, gasoline, acetone, ethylacetate, and methylenediphenyl diisocyanate (MDI) in workplace air and the concentration of hippuric acid (HA) in the urine were determined. To evaluate whether genetically determined individual variations in xenobiotic metabolizing capacity modified individual susceptibility to the possible genotoxic effects of the occupational exposure, we examined the subjects for their glutathione S-transferase M1 (GSTM1) and T1 (GSTT1) genotypes.

Materials and Methods
Subjects studied. We conducted four independent biomonitoring studies of Bulgarian shoe factory workers in 1992, 1993, 1995, and 1997. In the first three years, only workers from plant B (Gabrovo) were examined; the last survey also included subjects from plant A (Sofia). In 1997, we examined 33 workers from plant A and 19 from plant B. The control group consisted of 36 unexposed women from the administrative staff of these plants. The subjects represented about half of the staff in each factory and were chosen randomly. In general, the workers participated in the study only once, except for a few individuals who were included in two or three of the samplings. All subjects gave written informed consent and were interviewed for personal data, duration of employment, and smoking habits. The characteristics of the studied groups are summarized in Table 1. The subjects were all women, except eight control subjects and three exposed workers enrolled in 1992. In all cases, the numbers of smokers among the controls and the exposed subjects were adequately matched, except in the 1993 study, where the number of smokers was higher among the controls than among the exposed. In the shoe factory workers, the total duration of occupational exposure gradually increased during the follow-up period; the difference in the average time of employment in the plant between the first and the last sampling was 3 years.
Exposure measurements. We used special diffusive monitors (Hygitest Co., Sofia, Bulgaria) placed at breath level to obtain workplace vapor samples. The average concentrations of gasoline and methylenediphenyl diisocyanate (MDI) were measured by express linear-colorimetric method. We measured environmental levels of toluene, acetone, and ethylacetate with an infrared gas chromatography analyzer Miran 1B2 (Foxboro, MA, USA). We monitored these substances in workplace air according to the Bulgarian analytic protocols (15) and Bulgarian Health Ministry's Order (16). Because the administrative staff was expected to be exposed to very low environmental levels of organic solvents, compared with the exposed workers, and given the results of earlier studies (17)(18)(19), exposure values for control subjects included in this study were taken to be zero. The mean workplace levels (milligrams per cubic meter) of some organic solvents in the shoe factories are shown in Table 2, expressed as time-weighted average (8-hr TWA) values. It is evident that the shoe workers were exposed to a complex mixture of solvents. In most cases, the concentrations of these solvents clearly exceeded the limits recommended by European and American standards (20)(21)(22).
As a biomarker of toluene exposure, urine samples were collected in the end of the working day, also from the controls, and were analyzed for HA, according to a method previously described (23). As can be seen from Table 3, the shoe factory workers showed a marked increase in the mean HA levels, with a statistically significant difference from controls [p < 0.001; analysis of variance (ANOVA)] in each sampling. A significant relationship between HA concentrations in urine and toluene air levels was found in the entire population studied (β = 0.651, p < 0.001). The presence of phenol, the main urinary metabolite of benzene, was analyzed (24) in urine samples from 29 shoe factory workers and 16 controls included in the 1992 study. As there were no differences between these groups in urinary phenol levels (exposed: 0.12 ± 0.03 mmol/L; controls: 0.10 ± 0.02 mmol/L; detailed data not shown), the workers did not appear to be exposed to high levels of benzene. However, urinary phenol may not reliably indicate lower benzene exposures from gasoline, but these benzene levels may have genotoxic effects.
We analyzed the glues used in these plants in 1997 for benzene content, using a gas chromatograph with a mass-spectrometric detector (Hewlett-Packard 6890/5972A, Wilmington, DE, USA). We detected no benzene traces in these glues.
Blood collection. Heparinized venous blood samples were drawn from each donor, and hemoglobin values were measured in Bulgaria by standard methods. In the 1992, 1993, and 1995 studies, the blood samples were processed in Bulgaria, and blood cultures were set up in a few hours after sampling. To carry out the 1997 study, the blood samples were sent to Barcelona and were put in culture within 48 hr after the collection. Briefly, the blood samples were centrifuged at 170 g for 5 min, the plasma was removed, and RPMI 1640 culture medium (GIBCO, Eragny, France) was added to each blood pellet to attain its initial volume.
Sister chromatid exchanges. We performed the SCE assay only for blood samples obtained in 1997, following standard methodology (25). Briefly, two whole-blood lymphocyte cultures were set up for each donor, and the cultures were incubated covered from light for 72 hr in the presence of 5-bromodeoxyuridine (BrdU; final concentration 8 µg/mL). The cultures were treated with colcemid for the last 2 hr, and the cells were then harvested, treated with hypotonic solution (0.075 M KCl) for 20 min, and fixed three times with methanol-glacial acetic acid. Microscopic slides were prepared from the fixed cells by air drying, and the slides were stained by the fluorescence-plus-Giemsa procedure (26 (Rochester, NY, USA), and Giemsa from Merck (Darmstadt, Germany). For each donor, we analyzed 50 metaphases (25/replicate) on coded slides to determine the frequency of SCEs per cell. The analysis was conducted by two scorers, one per replicate. We estimated the percentage of high frequency cells (HFC) for each individual using the pooled distribution of all SCE measurements. We defined HFC as cells with SCE number over the 95th percentile of the distribution of SCEs per cell in the whole population. We examined 100 metaphases to calculate the proliferation rate index (PRI) (27).
Micronuclei. Whole blood cultures for the cytokinesis-block MN test were established according to Surrallés et al. (28), with some minor adaptations to the specific conditions of the Sofia and Barcelona laboratories. In both cases, the cultures were incubated at 37ºC for 72 hr and, at 44 hr after initiation of the cultures, cytochalasin-B (Sigma Chemicals, St. Louis, MO, USA) was added at a concentration of 6 µg/mL. The cells were harvested, treated with a mild hypotonic solution (0.075 M KCl) for 2-3 min, and gently fixed three times using methanol/acetic acid (5:1) solution. Airdried preparations were made, and the slides were stained in a 10% (v/v) solution of Giemsa in phosphate buffer (pH 6.8) for 20 min.
We examined 1,000 binucleate cells for each subject on coded slides (the same scorer in Barcelona and Bulgaria), and the total numbers of MN and the frequency of binucleate cells with MN (BNMN) were scored. We scored 500 lymphocytes to evaluate the percentage of cells with 1, 2, 3, or 4 nuclei, and we calculated the cytokinesis-block proliferation index (CBPI) according to Surrallés et al. (29).
Genotype analysis. We used genomic DNA extracted from leukocytes by standard methods as a template in the GSTM1 and GSTT1 genotype analyses, as described elsewhere (30). Briefly, the GSTM1 and GSTT1 genotypes were determined simultaneously in a multiplex polymerase chain reaction (PCR) approach using β-globin specific primers in addition to the GSTM1 and GSTT1 specific primer pairs. The presence of the β-globin specific signal verified the proper functioning of the PCR reaction, whereas the absence of GSTM1 and GSTT1 specific amplification products revealed the corresponding null genotypes.
Statistical analysis. For the crosssectional studies performed in 1992-1997, all variables analyzed were compared with the normal distribution by the Kolmogorov-Smirnov test of goodness of fit. The distributions of our data did not significantly depart from normality (p > 0.05) in any case, so parametric tests were adequate for the statistical analysis.
For each independent study, the effects of various factors on HA, hemoglobin, BNMN, and SCEs were simultaneously assessed by the ANOVA. These factors were occupational exposure, smoking, and, if available, GST genotypes, and we considered the age of the subjects as a covariate. To determine the relationship between solvent exposure levels and both HA and hemoglobin data, we applied a multiple regression analysis in the whole population. To this effect, from those subjects sampled several times, only the latest results were included in the analysis. Because of the low prevalence of the GSTT1 null genotype, GSTT1 genotype was not included as an independent variable in the analysis. The results were analyzed using the CSS: STATISTICA/W (StatSoft, Tulsa, OK, USA) statistical package. Table 3, the hemoglobin values of shoe factory workers were reduced compared those of nonexposed women. This difference, apparently stemming from the occupational exposure, reached statistical significance (p < 0.001) in the ANOVA.

As shown in
In 1992-1995, we used the MN assay to assess whether the occupational exposure of the shoe factory workers could exert some genotoxic effects. We observed significant increases in the mean frequency of BNMN of the exposed groups each year in comparison with the respective control groups (Figure 1). Especially noteworthy was the constant increase in the mean frequency of BNMN in the exposed groups from 1992 (38.00, SD 20.18) to 1995 (55.2, SD 17.1), whereas control subjects consistently showed similar basal mean values (from 18.6, SD 6.3 to 22.0, SD 10.0).
In the last study, carried out in 1997 and involving both factories, the MN assay confirmed our previous positive findings. Table 4 shows the MN and SCE results, separately for nonsmokers and smokers, among the controls and exposed workers. In the multiple ANOVA, occupational exposure (plant A, plant B, and controls) constituted a statistically significant (p < 0.001) source of variation for the BNMN values. The post hoc least squares difference test showed that workers of plant B had a clearly higher BNMN frequency (p < 0.001) than controls or plant A workers, but also plant A workers differed significantly (p = 0.035) from the controls. We detected no effect on cell-cycle kinetics, measured as CBPI, as a consequence of the solvent exposure.
In the SCE assay (Table 4), we were not able to detect any genotoxic effect related to occupational exposure. We found no correlation between MN and SCEs. The occupational exposure was associated with a slight decrease in PRI (p = 0.048).
Age was a significant covariate (p = 0.006) in the multiple analysis of variance of the 1997 BNMN data, reflecting an

Articles • SCE and MN analyses in shoe factory workers
Environmental Health Perspectives • VOLUME 110 | NUMBER 4 | April 2002 401 Table 3. Urinary hippuric acid and blood hemoglobin levels in exposed shoe workers and control subjects (means ± SD).   age-dependent increase in BNMN frequency. Age appeared to be partially associated with a decreased cell proliferation, as indicated by its significant effect on PRI (p = 0.001); CBPI was not affected by age. In the same analysis, smoking did not significantly affect BNMN frequencies in the 1997 study. However, both the mean frequency of SCEs per cell (p = 0.006) and the mean number of HFC (p = 0.004) were increased by smoking. This smoking-related effect was stronger among the controls than among plant A workers (p < 0.05 in both cases; t-test), but it did not attain statistical significance among plant B workers. In agreement with reported frequencies for Caucasian populations, the prevalence of GSTM1 and GSTT1 null subjects in the 1997 study was 56.8% and 13.6%, respectively (Table 5). When the few GSTT1 null subjects were removed for a subsequent ANOVA (including occupational exposure, smoking, and GSTM1 genotype as independent factors and age as a covariate), neither the MN nor the SCE assay outcome was significantly affected by the GSTM1 genotype. However, considering plant A and B workers together, the post hoc least squares difference test indicated a higher BNMN frequency in the exposed than the controls in GSTM1 null subjects among both smokers (p = 0.006) and nonsmokers (p = 0.003), but in GSTM1 positive subjects only among nonsmokers (p = 0.017). Tables 6 and 7 show the average of BNMN, CBPI, SCEs, HFC, and PRI values for GSTM1 positive and GSTM1 null subjects among the different subgroups.

Discussion
The present study showed an increased frequency of MN in peripheral lymphocytes of shoe factory workers exposed to a mixture of organic solvents. In plant B, where lymphocyte MN frequencies were followed up over 5 years, all four samplings yielded the same result. Although the shoe factory workers did not show any effects in the frequency of SCEs, HFC, or DNA breakage, as measured earlier by the Comet assay (14), our findings indicate the presence of genotoxic exposure at the workplace. This would appear to act via an aneugenic or clastogenic mechanism that does not involve DNA breakage readily detectable by the Comet assay and that does not efficiently lead to SCE formation.
In interpreting these findings, one must consider that although there are several examples of MN induction in human binucleated lymphocytes by known clastogenic exposures in vivo, evidence that the in vivo effects of an aneugen can be expressed in the first in vitro division of a binucleated cell is scanty (31,32). The binucleated cells that are exclusively analyzed in the cytokinesis-block MN assay have, by definition, completed their first in vitro mitosis, and most MN observed in such cells have been formed in the culture. Thus, an in vivo MN inducer would have to be able to initiate its effect in the resting G 0 lymphocyte. Further studies of the shoe factory workers using CA analysis or an MN assay incorporating pancentromeric fluorescence in situ hybridization or antikinetochore antibodies to identify the contents of the MN will be useful in identifying the nature of the exposure that produced the excess MN.
Although the glues used in the factories did not contain measurable amounts of benzene, there probably was benzene in the gasoline used as solvent. The air concentrations of gasoline were high, and benzene may well be the principal agent responsible for the observed genotoxic effects. In factory B, the levels of MDI (measured as diisocyanate) were very high and could explain some of the genotoxic effects. MDI is highly reactive and induced CA in human lymphocytes in vitro (33). MN frequencies were clearly higher in plant B than in plant A, and this could be caused by MDI, since diisocyanate was not found at all in plant A. Even in plant B, diisocyanate was not discovered during the first three sample collections. Thus, although it may not be possible to point out a single genotoxic agent at the workplace, benzene in gasoline and MDI may explain some of the findings. A possible indication of the heavy organic solvent exposure of the workers was the consistently lowered blood hemoglobin level, which we also observed in the previous study (14). Anemia and other hematologic abnormalities have been associated with organic solvent exposure in Bulgarian petroleum workers (34).
Several investigations have suggested an increased sensitivity of the GSTM1 null genotype to genotoxicity induced by various exposures, including tobacco smoke (35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47). Such an association was not, however, unequivocally demonstrated in most of the few studies where SCEs or MN were used as biomarkers (48)(49)(50)(51)(52)(53)(54)(55)(56)(57)(58) Table 5. Distribution of GSTM1 and GSTT1 genotypes among exposed shoe factory workers and control subjects in 1997. genotype, because in smokers an occupational exposure effect could be seen only among GSTM1 null subjects. Although this finding is based on a few individuals, it could indicate, for example, that smoking induces GSTM1 activity in GSTM1-proficient individuals, which would protect them from the genotoxic effects of the occupational exposure. Smoking is not known to markedly induce GSTM1 in humans in vivo. At present, the possible effect of GSTM1 genotype on MN formation is not well understood. Our previous studies suggested that the baseline level of MN is decreased in GSTM1 null subjects (55). In one investigation, the GSTM1 null genotype was described to increase SCEs in smokers, although the difference to GSTM1 positive smokers was quite small (0.23 SCEs/cell) and was statistically significant only in heavy smokers (0.53 SCEs/cell) (48). An increase of this magnitude may have been undetected in the present study. An effect of GSTM1 polymorphism on the baseline level of SCEs in nonsmokers was reported by Cheng et al. (59,60), but this finding has not been reported in other studies (52)(53)(54)57), and it was not supported by our data.

Controls
In contrast, there is some evidence that the GSTT1 null genotype is associated with an increased baseline frequency of SCEs (61)(62)(63)(64) and possibly also CA (54,65). This finding does not appear to be explained by smoking or other known exposures. Unfortunately, we could not properly evaluate the effects of GSTT1 polymorphisms because of the low frequency of the GSTT1 null genotype.
In conclusion, our results indicate that the shoe factory workers studied have experienced genotoxic exposure, which is manifested as an increase in the frequency of MN, but not of SCEs, in peripheral lymphocytes. The workers also showed a reduction in blood hemoglobin values. Although high air concentrations of several organic solvents were found at the factories, the exposures responsible for the MN induction could not be identified with certainty. Exposure to benzene in gasoline and MDI may explain some of the findings. Table 6. Mean (± SD) frequency of BNMN and CBPI in shoe factory workers and unexposed controls, according to smoking and GSTM1 genotype (last sampling).