Mechanisms of benzene-induced hematotoxicity and leukemogenicity: cDNA microarray analyses using mouse bone marrow tissue.

Although the mechanisms underlying benzene-induced toxicity and leukemogenicity are not yet fully understood, they are likely to be complicated by various pathways, including those of metabolism, growth factor regulation, oxidative stress, DNA damage, cell cycle regulation, and programmed cell death. With this as a background, we performed cDNA microarray analyses on mouse bone marrow tissue during and after a 2-week benzene exposure by inhalation. Our goal was to clarify the mechanisms underlying the hematotoxicity and leukemogenicity induced by benzene at the level of altered multigene expression. Because a few researchers have postulated that the cell cycle regulation mediated by p53 is a critical event for benzene-induced hematotoxicity, the present study was carried out using p53-knockout (KO) mice and C57BL/6 mice. On the basis of the results of large-scale gene expression studies, we conclude the following: (a) Benzene induces DNA damage in cells at any phase of the cell cycle through myeloperoxidase and in the redox cycle, resulting in p53 expression through Raf-1 and cyclin D-interacting myb-like protein 1. (b) For G1/S cell cycle arrest, the p53-mediated pathway through p21 is involved, as well as the pRb gene-mediated pathway. (c) Alteration of cyclin G1 and Wee-1 kinase genes may be related to the G2/M arrest induced by benzene exposure. (d) DNA repair genes such as Rad50 and Rad51 are markedly downregulated in p53-KO mice. (e) p53-mediated caspase 11 activation, aside from p53-mediated Bax gene induction, may be an important pathway for cellular apoptosis after benzene exposure. Our results strongly suggest that the dysfunction of the p53 gene, possibly caused by strong and repeated genetic and epigenetic effects of benzene on candidate leukemia cells, may induce fatal problems such as those of cell cycle checkpoint, apoptosis, and the DNA repair system, finally resulting in hemopoietic malignancies. Our cDNA microarray data provide valuable information for future investigations of the mechanisms underlying the toxicity and leukemogenicity of benzene.

Benzene is well documented as an environmental pollutant that can induce hematotoxicity and hemopoietic neoplasia in humans and mice (Aksoy et al. 1974(Aksoy et al. , 1976Cronkite et al. 1984Cronkite et al. , 1989Snyder et al. 1980;Vigliani and Forni 1976). To date, studies on benzene have focused on its metabolic pathways to determine the metabolites responsible for its hematotoxicity and leukemogenicity (Henderson 1996;Schlosser et al. 1989;Schrenk et al. 1996;Snyder and Hedli 1996). Benzene and its major metabolites are not mutagenic in the Ames Salmonella test (Dean 1985), but they do induce chromosomal aberration both in vitro and in vivo (Dean 1985;Wolman 1977;Yager et al. 1990). This is comparable to classic carcinogens that are generally being activated to a single carcinogenic metabolite having a mutagenic property. Benzene can be characterized further in terms of its multisite carcinogenicity (Huff et al. 1989;Maltoni et al. 1989). Mice exposed to benzene develop different types of tumor in various glandular tissues and organs, including the hemopoietic system, Zymbal gland, Harderian gland, preputial gland, mammary gland, ovary, and lung. Results of the study of Low et al. (1995) strongly suggest that the carcinogenicity of benzene on target organs depends on the ability of enzymes in the organs to metabolize benzene.
As postulated by several investigators, the metabolism of benzene to reactive metabolites by hepatic enzymes, mainly cytochrome P450-2E1 (CYP2E1), is a prerequisite to the cyto-and genotoxicities associated with benzene exposure (Gut et al. 1996;Snyder and Hedli 1996;Valentine et al. 1996). Primary benzene metabolites include phenol, hydroquinone, catechol, and trans-trans muconic acid (Ross 2000). The synergistic interactions between these phenolic metabolites exacerbate benzene toxicity (Chen and Eastmond 1995;Eastmond et al. 1987;Subrahmanyam et al. 1990). This mechanism of multimetabolite genotoxicity is another unique aspect of benzene that distinguishes it from other chemicals in terms of the mechanism of its toxicity and carcinogenicity. Benzene metabolites subsequently undergo secondary activation by myeloperoxidase (MPO) that is present at high levels in the bone marrow tissue. This results in the production of genotoxic quinones and reactive oxygen species, thereby inducing not only hemopoietic cellular damage (Farris et al. 1997;Kolachana et al. 1993; Lee and Garner 1991;Smith et al. 1989) but also the dysfunction of bone marrow stromal cells (Niculescu et al. 1995).
Exposure duration and dose are also important factors in determining benzeneinduced hematotoxicity and leukemogenicity (Cronkite et al. 1989;Snyder and Kalf 1994), which may be related to the limited capacity of enzymes for benzene metabolism and to the dynamic responses of hemopoietic microenvironmental conditions against the adverse effects of benzene.
Despite intensive studies over several decades, the mechanisms underlying benzene-induced hematotoxicity and leukemogenicity are still not fully understood. Nevertheless, previous studies strongly suggest that the toxic effects of benzene on bone marrow tissue can be realized through pathways such as those of metabolism (Snyder and Hedli 1996), growth factor regulation (Niculescu et al. 1995 production of oxidative stress (Laskin et al. 1996;Subrahmanyam et al. 1991), DNA damage and repair (Lee and Garner 1991), cell cycle regulation (Yoon et al. 2001b), and apoptosis Ross et al. 1996). These studies indicate that investigation of the roles of a few specific genes may not be sufficient to explain the complete molecular mechanism of benzene-induced hematotoxicity and leukemogenicity.
Bone marrow tissue, a major target organ of benzene, is an active hemopoietic system in which various counterbalanced genes are organized through their network interactions that maintain cellular-environmental homeostasis as well as protect cells from endogenous and exogenous hematotoxic effects such as benzene-induced effects. The dysregulation of such a multidimensional counterbalance, possibly induced by the genetic and epigenetic effects of benzene, may result in the altered expression of a number of genes associated with the mechanisms of benzene-induced hematotoxicity and leukemogenicity.
In this study we investigated the changes in DNA expression during and after benzene exposure (300 ppm) to probe further the molecular mechanisms underlying benzene toxicity. Because previous studies (Boley et al. 2002;Yoon et al. 2001b) demonstrated that the p53 tumor suppressor gene is important in cell cycle regulation associated with the mechanisms of benzene-induced toxicity, these analyses were carried out by cDNA microarray analyses in C57BL/6, wild-type (WT), and p53-knockout (KO) mice.

Materials and Methods Animals
Specific pathogen-free, 7-week-old, male C57BL/6 mice were purchased from Japan SLC (Hamamatsu, Japan) and quarantined for 1 week in 1.3-m 3 inhalation chambers (Shibata Scientific Technology Ltd., Tokyo, Japan) in ambient air. To obtain WT and p53-KO mice for use in this study, male and female heterozygous p53-KO C57BL/6 mice, originally bioengineered by Tsukada et al. (1993), were mated; the pups produced were then identified by polymerase chain reaction analysis of the DNA samples extracted from the tail of each mouse. The mice were grouped randomly into untreated control and benzene-exposed groups and maintained in stainless-steel wire cages inside inhalation chambers under a 12-hr light-dark cycle during the study. A basic pellet diet (CRF-1; Funabashi Farm, Funabashi, Japan) was provided ad libitum except during the daily 6-hr benzene inhalation period. Water was delivered by an automated tubing nozzle and provided ad libitum throughout the study.

Benzene Exposure
Benzene vapor was generated and its concentration was monitored as described elsewhere (Yoon et al. 2001b). Temperature and humidity inside the chambers were maintained automatically at 24 ± 1°C and 55 ± 10%, respectively. Mice were exposed to 300 ppm benzene for 6 hr/day, 5 days/week for 2 weeks; the sham control groups were maintained in the inhalation chambers in ambient air over the same period. Experimental schedules for sham and benzene-treated mice are shown in Figure 1. Immediately after the first 5 days of exposure (D5), the second 5 days of exposure in the second week (D12), and 3 days after D12 for recovery (D+3), the mice were sacrificed. D12 is also designated as the 2-week exposure. To investigate changes in gene expression, three C57BL/6 mice from each of the sham control and benzene-exposed groups were decapitated after euthanasia at 1 week (D5) and 2 weeks (D12), respectively, during a 2-week benzene exposure period and 3 days after benzene removal (D+3), and poly(A) + RNA extracted from each group was applied to Incyte gene expression microarray (GEM) assay (Incyte Pharmaceuticals, Inc., Palo Alto, CA, USA) (see "Microarray Preparation"). Our previous study (Yoon et al. 2001b) showed that mice are able to recover from benzeneinduced hematotoxicity 3 days after a 2-week benzene exposure. In studies using WT and p53-KO mice, two to four mice from each group and genotype were sacrificed immediately after the 2-week benzene exposure and applied to the Affymetrix system (Affymetrix, Inc., Santa Clara, CA, USA) (see "Microarray Preparation").

Bone Marrow Cell Collection for RNA Extraction
The mice from which bone marrow cells were collected for RNA extraction were carefully chosen on the basis of our evaluation of peripheral blood number and bone marrow cellularity using a blood cell counter (Sysmex M-2000; Sysmex Co., Tokyo, Japan) and our comparison of the values with those previously reported (Yoon et al. 2001b).
We harvested bone marrow cells from both femurs of individual mice of each group (Yoon et al. 2001b). Using a 27-gauge hypodermic needle, we flushed out bone marrow cells of the bone shafts with 2 mL Dulbecco's modified minimum essential medium without phenol red (Invitrogen Corp., Carlsbad, CA, USA). Single-cell suspensions were then prepared by repeatedly passing the harvested bone marrow cells through the needle. After the lysis of red blood cells, the bone marrow cells were immediately frozen in liquid nitrogen and stored at -80ºC until RNA extraction.

Preparation of Total RNA and Poly(A) + RNA
Total RNA was extracted from the collected bone marrow cells using ISO-GEN (Wako Chemical Co., Osaka, Japan) in accordance with the manufacturer's instructions. The total RNA yielded optical density (OD) ratios (OD 260/280) of 1.7-2.1; its purity was confirmed by gel  Benzene inhalation schedule diagram: exposure was for 6 hr/day from 10 A.M. to 4 P.M. On the days marked with an asterisk (*), both sham exposure mice were killed at 4 P.M., and exposed mice were killed immediately after exposure (Yoon et al. 2002). On D+3, both sham and recovery mice were killed at 4 P.M. D5 designates practically a 1-week exposure and D12, a 2-week exposure. D+3 is the group with a 3day recovery period after D12. chromatography, and its concentration was determined on the basis of its absorbance at 260 nm that was measured with a Beckman spectrophotometer (DU640; Beckman Coulter, Inc., Fullerton, CA, USA). Independent total RNA and poly(A) + RNA samples were separately extracted from three C57BL/6 mice and two to four WT and p53-KO mice; those samples from equivalent materials were analyzed using the Incyte GEM system and the Affymetrix system. We used the Affymetrix system to analyze further two separate RNA samples from benzeneexposed and sham-exposed WT mice at each time point, and two separate RNA samples from benzene-exposed and four separate samples from sham-exposed p53-KO mice. In addition, for further comparison, RNA samples from three mice in each of the four groups were pooled and processed with the Incyte GEM system. No duplicate or triplicate runs were performed using the Incyte GEM system. Poly(A) + RNA was prepared from the total RNA using Oligo (dT) Microbeads (Daiichi Co., Tokyo, Japan) in accordance with the manufacturer's instructions.

Microarray Preparation
All procedures such as experimental design, array design, sampling, hybridization, signal measurements, and normalization control were performed according to the MIAME (minimum information about a microarray experiment) guidelines (Brazma et al. 2001).
Probe array washing, staining, and antibody amplification. After hybridization, washing and staining were performed with a GeneChip fluidic station (Affymetrix) using appropriate antibody amplification washing and staining protocols.
Probe array scanning. The phycoerythrin-stained array was performed with a confocal scanner (Agilent Affymetrix GeneArray scanner), processed into digital image files, and analyzed using the Affymetrix analysis software Microarray Suite (MAS, version 4.0). Data normalization. GeneSpring software (Silicon Genetics, Redwood City, CA, USA) was used to normalize the data. The 50th percentile of all measurements was used as a positive control for the sample; each measurement for each gene was divided by this synthetic positive control, assuming that this was at least 10. The bottom 10th percentile was used as a test for correcting background subtraction. This was never less than the negative values of the synthetic positive control. The measurement for each gene in each sample was divided by the corresponding mean of the sham controls, assuming that the cutoff value is more than 0.01.
Incyte GEM system. Fluorescence labeling of probe for GEM system. For comparison of the array data obtained using the Affymetrix system, the samples were simultaneously sent to the Incyte GEM system to analyze the time course of gene expression changes after benzene inhalation and its cessation. Poly(A) + RNA (200 ng) from each sample was sent to Incyte Co Ltd. (MouseUniGEM: GEM-5200; Fremont, CA, USA) via GEM custom screening services (Kurabo Co Ltd., Osaka, Japan). Briefly, the samples were incubated for 2 hr at 37°C with 200 U M-MLV reverse transcriptase (Life Technologies, Gaithersburg, MD, USA), 4 mM DTT, 1 U RNase inhibitor (Ambion, Austin, TX, USA), 0.5-mM dNTPs, and 2 µg 5´Cy3 or Cy5labeled 9-mers (Operon Technologies Inc., Alameda, CA, USA) in 25-µL volume with an enzyme buffer supplied by the manufacturer, and then reverse-transcribed to cDNA. The reaction was terminated by heating at 85°C for 5 min. The paired reaction mixtures were combined and purified with a TE-30 column (Clonetech, Palo Alto, CA, USA), diluted to 90 µL with distilled water, and precipitated with 2 µL of 1 g/mL glycogen, 60 µL of 5 M ammonium acetate, and 300 µL ethanol. After centrifugation, the supernatant was decanted and the pellet was resuspended in 24 µL hybridization buffer, 5× sodium chlorine-sodium citrate buffer, 0.2% sodium dodecyl sulfate, and 1 mM DTT.
Hybridization. The probe solutions were thoroughly resuspended by incubating them at 65°C for 5 min, with mixing. The probe was applied to the array and covered with a 22-mm 2 glass cover slip and placed in a sealed chamber to prevent evaporation. After hybridization at 60°C for 6.5 hr, the slides were consecutively washed 3 times in a washing buffer of decreasing ionic strength.
The GEM system scanning. After hybridization, the GEM was scanned at 10-µm resolution to detect Cy3 and Cy5 fluorescence. Both Cy3 and Cy5 channels were scanned simultaneously with independent lasers. The emitted fluorescent light was optically filtered before photo-multiplier tubes translated the photons into an analog electrical signal, which was further processed into a 16-bit digital signal. This provided electronic images of both Cy3 and Cy5 with a 65,536-color resolution. A 16-color log scale was used for visual representation.
Normalization and ratio determination. Incyte GEM Tool software (Incyte) was Toxicogenomics | Mechanisms of benzene-induced hematologic toxicity Environmental Health Perspectives • VOLUME 111 | NUMBER 11 | August 2003 used for image analysis. A grid-and-region detection algorithm was used to determine the elements. The area surrounding each element image was used to calculate the local background and was subtracted from the total element signal. Backgroundsubtracted element signals were used to calculate the Cy3:Cy5 ratio. The average of the resulting total Cy3 and Cy5 signals gives a ratio that is used to balance or normalize the signals.

Results of cDNA Microarray Analyses and Their Implications
In this study we investigated the changes in gene expression during and after benzene exposure (300 ppm). As previous studies (Yoon et al. 2001b) demonstrated that the p53 tumor suppressor gene plays an important role in a cell response to benzene toxicity, analyses were performed using WT and p53-KO mice. In the sections that follow, we compare the gene expression profile obtained from WT mice using the Incyte GEM system with that obtained using the Affymetrix system, which in turn are compared with those of previous reports (Boley et al. 2002;Ho and Witz 1997;Schattenberg et al. 1994, Zhang et al. 2002. In addition we also describe particular genes related to p53-KO mice, such as cell cycle-regulating genes, apoptosisrelated genes, and DNA repair-related genes. All gene names, abbreviations, and accession numbers from MAS 4.0 are equivalent to those of GenBank (http:// w w w . n c b i . n l m . n i h / g o v / G e n b a n k / index.html). Figure 2 shows differences in the expression patterns of specific genes between, during, and after exposure of the WT mice to 300 ppm benzene for 2 weeks, determined using the Incyte GEM system (see Figure 1 for experimental schedule). Figure 2A shows the genes upregulated during benzene exposure (D5, D12), and then downregulated afterward (D+3), as represented by the MPO gene. Figure 2B shows the genes that had been continuously upregulated after benzene exposure, i.e., p53-binding protein 1 (53BP1), adenosine triphosphate (ATP)binding cassette (ABC) transporter, and N-acetylglucosamine-6-O-sulfotransferase. Figure 2C shows the genes that had continuously been somewhat upregulated after benzene exposure, e.g., murine cathelin-like protein (MCLP), cell division cycle 2 (cdc2), and lipocalin 2. The expression patterns of MPO in Figure 2A may be induced by benzene metabolism during benzene exposure. This induction ceases after inhalation (Schattenberg et al. 1994), whereas 53BP1, a DNA damage-responsive gene (Ward et al. 2003), and ABC transporter, a detoxifying drug-transporter (Ambudkar and Gottesman 1998), in Figure 2B show prolonged expressions after benzene exposure. MCLP (Gombert et al. 2003) and lipocalin 2 (Jessen and Stevens 2002) function as marker genes for differentiation. The genes listed in Figure 2C, including cdc2, may be upregulated for the proliferation of bone marrow cells during the recovery phase. A particular expression change in the aryl hydrocarbon receptor (AhR) was observed for which a mechanism could not be specified (data not shown). As we previously observed, sensitivity to benzene toxicity is innate in AhR-KO mice, implying that AhR transmits this sensitivity to benzene toxicity (Yoon et al. 2002).

Gene Expression Profile of Wild-Type Mice after Benzene Exposure
The results of cDNA microarray analysis showed a broad consensus that the p53 tumor suppressor gene is central to the mechanism of benzene action, by strictly regulating specific genes involved in the pathways of cell cycle arrest, apoptosis, and DNA repair. Such close association of the p53 gene with the benzene toxicity mechanism raises the question: What would happen in mice whose p53 gene is knocked out after benzene exposure? Thus, the cDNA microarray data obtained from the WT and p53-KO mice were applied to the Affymetrix system and analyzed using GeneSpring software, as described in "Materials and Methods." The results are shown in Table 1. This table shows that the expression profiles of the many genes involved in benzene metabolism, cell cycle or cell proliferation, and hemopoiesis in WT mice were generally consistent with the cDNA microarray data of C57BL/6 mice described in Table 2.

Characteristics of Gene Expression Profile of p53-KO Mice after Benzene Exposure
Mice lacking the p53 gene and WT mice generally had similar expression patterns of the genes involved in benzene metabolism (CYP2E1 and MPO; Bernauer et al. 1999Bernauer et al. , 2000Schattenberg et al. 1994;Yoon et al. 2001b) and hemopoiesis, suggesting that p53-KO mice are also affected to a similar extent by benzene exposure. This is consistent with the high frequency of micronuclei observed in benzene-exposed p53-deficient mice (Healy et al. 2001) (Table 1, Table 3A; p53-independent, benzene-induced gene expression level increase or decrease.). Figure 3 shows scatterplots representing the expression levels of genes in the bone marrow cells of the benzene-exposed WT ( Figure 3A) and p53-KO mice ( Figure  3B) relative to the expression levels of the genes in those of the corresponding sham-control mice. To elucidate and visualize the difference in gene expression level between the WT and p53-KO mice, clustering analysis was performed ( Figure  4). The genes expressed include cell cycle/proliferation-associated genes. Table 3B lists the genes with a p53dependent, benzene-induced decrease (e.g., G protein-coupled receptor [GPCR]) or increase (e.g., caspase-11) in expression level in the WT mice. In the Toxicogenomics | Yoon et al. 1414 VOLUME 111 | NUMBER 11 | August 2003 • Environmental Health Perspectives  The studies involved two to four animals; data were obtained from the use of the Affymetrix gene chips. Mice were killed on day 12, immediately after benzene exposure (see Figure 1, "Experimental Schedule"). b Information for GenBank (http://www.ncbi.nlm.nih.gov/Genbank/index.html). c No data available for p53-KO mice.
p53-KO mice, these genes did not change their expression level with benzene exposure. Table 3C shows that some changes in gene expression were undetectable because of the function of the p53 gene, which can be "visualized" in the p53-KO microarray ( Figure 4). Namely, data from toxicogenomics studies of specific gene KO mice could possibly disclose homeostatic balances undetectable in conventional WT mice.

Cell Cycle-Regulating Genes in p53-KO Mice and Wild-Type Mice
Cyclin genes were generally activated in p53-KO mice after benzene exposure, whereas cell cycle-regulating genes including the G2/M arrest-related gene cyclin G1 (Kimura et al. 2001) were upregulated in WT mice. These findings indicate that the hemopoietic cell cycle is still functional in p53-KO mice during benzene exposure, whereas in WT mice it is arrested because of alterations in the expression of cell-cycle checkpoint genes, particularly the p53 gene (Yoon et al. 2001b). Some upstream genes encoding p53, such as Dmp1 and Raf-1 of the p53-KO mice, compared with those of the corresponding experimental groups of the WT mice, were upregulated to a similar extent or were more strongly enhanced in their expression. This is another indication of the role of the p53-mediated pathway in the mechanism of benzene toxicity associated with cell cycle regulation. Such information could be important in helping investigators to understand yet unknown mechanisms of chemical toxicity.
It is important to note that such a conclusion possibly can be drawn by carefully and simultaneously screening different expression patterns of many genes with interrelated functions, including genes showing small changes in expression levels (about 1.5-to 2-fold). The investigation of the expression levels of a limited number of genes generally may not provide insight into the main mechanism of chemical toxicity or clues to the particular role of each of the investigated genes in this mechanism. Toxicogenomics may have a strong advantage from this point of view (Inoue 2003).

Apoptosis-Related Genes in p53-KO Mice and Wild-Type Mice
The microarray analysis results of the p53-KO mice reminded us of the importance of the p53 gene in the mechanism of benzene toxicity. The genes activated by the p53 gene, including p21, caspase 11 (Choi et al. 2001;Kang et al. 2000), and cyclin G1 (Kimura et al. 2001), were distinctly upregulated in the benzene-exposed WT mice (Table 1). It is interesting that caspase 11 instead of caspase 9 was highly expressed after benzene exposure. This suggests that the p53-mediated activation of caspase 11 is an important signaling pathway of apoptosis of bone marrow cells Toxicogenomics | Yoon et al. 1416 VOLUME 111 | NUMBER 11 | August 2003 • Environmental Health Perspectives Table 3. Differences in alteration of gene expression between WT and p53-KO mice after benzene exposure.

Expression of DNA Repair-Related Genes in p53 Gene Network
Despite the possible damage to the DNA of the bone marrow cells of a p53-KO mouse, the DNA repair system is not likely to be functioning efficiently in the p53-KO mice, as DNA repair-related genes that were actively functioning in the benzene-exposed WT mice were not activated but rather suppressed in the p53-KO mice. In association with cell proliferation and apoptosis, high expression levels of the tuberous sclerosis gene (Tsc-2), a tumor suppressor gene encoding tuberin, and metallothionein 1 gene were noted in the WT mice (Table 1), raising the possibility that these genes are regulated by the p53 gene. The association of metallothionein with p53 transcriptional activity has Figure 3. Scatterplots representing expression of genes in the bone marrow cells of benzene-exposed WT (A) and p53-KO mice (B) relative to expression of the genes in those of the corresponding sham-control mice, obtained using the Affymetrix system; x-axis and y-axis, respectively, indicate fluorescent signal intensity in the sham-control and benzene-exposed groups.  . Clustering diagram of gene trees focused on particular genes of interest in the Affymetrix system. GeneSpring software was used to normalize the data. Clustering of microarray data revealed the standardized expression intensities of relevant genes shown by colors; from low expression level to high expression level, blue < yellow < orange < red. WT mice with or without benzene exposure, as well as p53-KO mice with or without benzene exposure are compared. A cluster of genes in the box (top) was found to consist of genes, the functions of which are related to cellular proliferation (bottom), and expressed only when p53 is knocked out, after benzene exposure. The average expression levels were obtained from two to four independent RNA samples from the mice. Genes in the box ( recently been postulated in an in vitro system in which metallothionein acts as a potent chelator to remove zinc from p53, thereby modulating p53 transcriptional activity (Meplan et al. 2000). The Tsc-2 gene has recently been reported to regulate the insulin-signaling pathway mediated by protein kinase B (PKB/Akt) for cell growth (Gao and Pan 2001;Potter et al. 2002). It is noteworthy that Tsc-2 is a target gene of 2,3,5-tris (glutathion-S-y) hydroquinone, a metabolite of hydroquinone for renal cell transformation (Lau et al. 2001). The high expression level of the mph1/rae28 gene in the WT mice with severely suppressed bone marrow cellularity is noteworthy with respect to the maintenance of the activity of hemopoietic stem cells (Ohta et al. 2002). Furthermore, the Wnt-1 signaling pathway is also likely to be activated after benzene exposure, followed by the aberrant expressions of downstream genes such as WISP1 and WISP2 (Table 1). As the Wnt-1 signaling pathway was reported to regulate the proliferation and survival of various types of cell including B lymphocytes (Reya et al. 2000), the activation of both mph1/rae28 and Wnt-1 genes may be associated with the rapid recovery of suppressed bone marrow cellularity after cessation of benzene exposure.

Summary
As described above, the results of our cDNA microarray suggest that p53-KO mice are not resistant to benzene-induced toxic effects. These results were comparable with the dynamic protective responses of C57BL/6, WT mice at the gene functional level. On the basis of these observations, the effects of benzene on the bone marrow cells of p53-KO mice can be summarized as follows: a) cellular damage due to benzene metabolites and oxidative stress, b) dysfunction of the machinery of cell cycle arrest for repairing damaged DNA, resulting in continuous cycling of damaged cells even without undergoing repair, c) inhibition of apoptosis by both disruption of p53-dependent proapoptotic signaling and activation of survival genes, and d) failure of activating DNA repair genes. Such phenomena may lead to the increase in cell mutation frequencies at the candidate DNA locus, for instance, the hprt locus, responsible for benzene carcinogenesis, resulting in the development of hemopoietic malignancies. This hypothesis is based on multigene expression profiles that reasonably explain the high incidence and early onset of hemopoietic neoplasia, which were clearly observed in the p53 hetero-and homozygous KO mice chronically exposed to a critical dose of benzene for leukemogenicity tests (Kawasaki et al. Unpublished observation). We also noted that the genes involved in fatty acid β oxidation such as the acyl-Co-A thioesterase gene and those encoding adipose fatty acid-binding proteins, which are commonly induced by peroxisome proliferators such as diethylhexylphthalate and clofibrate (Bartosiewicz et al. 2001), were also upregulated in the WT mice exposed to benzene (Table 1). A possible signaling pathway induced by benzene exposure is shown by a schematic in Figure 5. The present study using p53-KO mice elucidated the role of the p53 gene not only in during benzene exposure, but also in the recovery state, and the gene expression profiling from p53-KO mice visualizes such oscillatory changes hidden behind the homeostatic balance organized by the p53 gene in WT mice.
In conclusion, The cDNA microarray system used in this study revealed the mechanism of benzene toxicity by showing the altered expression of a number of benzene-affected genes including physiologic and toxicologic gene repertoires. Our data will provide valuable targets for the future investigation of the mechanism of benzeneinduced toxicity and leukemogenicity.  Figure 5. Mechanism of benzene toxicity at the molecular level based on the altered multigene expression profiles after benzene exposure. Benzene toxicity is dependent on AhR. CYP2E1 oxidizes benzene and produces quinones. Metabolites of benzene generated by cytochrome P450 monooxygenase and/or MPO trigger redox cycling consistent with the upregulation of genes for Adh4, Cox, glutathione S-transferase 1 (GST1) and metallothionein. Induction of p53 after DNA damage is the central toxicologic event, which readily induces the expression of p21, GADD45, and cyclin G, on one hand, the production of Bax, caspase 11, and ADP ribosylation factor-like protein 1 (ARL-1), and subsequently, the expression of Wee1 kinase. Induction of p21, in conjunction with Rb, results in the G1/S cell cycle arrest, whereas cyclin G and elongation factor 1 delta induces G2/M cell cycle arrest. Whether these cells in the cell cycle arrest further go into apoptosis or DNA repair depends on the participation of other cell signaling genes such as bcl-2 and Rad50, Rad51, and Rad54. However, cells may mutate because of the possible failure of DNA repair.