Benzene induces rapid leukemic transformation after prolonged hematotoxicity in a murine model

To the Editor:

Benzene is an important industrial chemical and ubiquitous environmental mutagen known to induce hematological malignancies particularly leukemia [1]. The development of benzene-induced leukemia takes an average of 11.4 years to occurs in workers chronically exposed to high levels of benzene [2]. Before the onset of the malignancy, benzene workers experience a strong and prolonged hematotoxicity characterized by significantly reduced white blood cell counts, and in severe forms, pancytopenia or anemia, and the later has been referred to as benzene poisoning (BP) [3]. These observations are consistent with in vitro findings that benzene inhibit hematopoietic growth and self-renewal through induction of DNA double strand breaks and adducts of bone marrow DNA [4]. Benzene-induced hematotoxicity is associated with future risk of developing haematological malignancy. While benzene exposure has an overall 7-fold risk for development of leukemia, BP is associated with a 71-fold risk for development of the acute myeloid leukemia (AML) or myelodysplastic syndromes (MDS) in human [2, 5]. Malignant transformation in these benzene-exposed workers can take place in a very short time, as evident by rapid 10-fold expansion of white blood cell (WBC) within 6-month in reported long-term exposure BP cases [3].

While these population-based cohort studies characterize key aspects of this unique entity of AML in humans, models are needed to recapitulate this distinct evolutionary trajectory of benzene-induced leukemia. Such model will be a powerful tool to explore the pathobiology and the underlying molecular mechanisms of benzene-induced carcinogenesis. The inhibitory effects of benzene on hematopoietic cells have been extensively characterized in vitro and in vivo [1]. However, hematotoxicity does not lead to leukemic transformation in these systems. In the efforts to develop cancer models, benzene treatment induced mostly lymphoma [6], a different type of blood cancer from what have been commonly observed in benzene-exposed human population. Benzene also has been shown to promote tumor development in transgenic mice carrying Trp53 or Ras mutation [6, 7]. The toxic effect on hematopoiesis, however, was not observed in these mice. Therefore, the continum of hematotoxicity and malignant transformation in the development of benzene-induced human leukemia has been difficult to model.

In this study, we aimed to develop a mouse model to recapitulate the two-stage (hematotoxicity and malignant transformation) process of benzene-induced AML. To observe leukemia development under benzene exposure, preleukemic bone marrow (BM) cells and spleen cells (CD45.2⁺) harboring Mll–Af9 oncogenic fusion gene were transplanted into recipient mice (CD45.1⁺) (Fig. 1a). After allowing for engraftment and reconstitution of BM, Mll–Af9 mice were exposed to benzene at 299 ppm (972 mg/m³) (n = 25) or air (n = 25) by inhalation for a lifetime exposure (8 h/d for 5 d/w). This level of benzene exposure concentration is sufficient to induce myeloid leukemia both in wild-type and transgenic mice [1, 6, 8]. (See methods). The effect of benzene on hematopoiesis was assessed by measuring the complete blood cell counts and preleukemic cells counts (CD45.2⁺ WBC). After maintaining a preleukemic condition for a period of 15 weeks (leukemia development stage), mice from both groups started to develop leukemia and all died of the disease within 28 weeks (Leukemic death stage) (Supplementary Fig. 1a). Although the survival rate was similar among the groups, benzene-exposed mice and air control mice exhibited a drastic difference in the growth trajectory of CD45.2⁺ preleukemic cells. CD45.2⁺ cells in the air control group experienced a steady 2.6-fold increases (2.4-fold increases for WBC) over an extended time of 15 weeks (Fig. 1b). Subsequently, mice underwent a 16-fold expansion (12-fold expansion for WBC) in the last 4 weeks towards impending death (Fig. 1c, Supplementary Fig. 1b). In sharp comparison, benzene exposure induced a 36–65% inhibition on the growth of CD45.2⁺ cells (36%-67% inhibition for WBC) over the 15-week leukemic development stage. Consistently, red blood cell (RBC) counts, hemoglobin (HGB), and Platelet (PLT) levels were significantly decreased in benzene-exposed mice compared to air control (Fig. 1d,e, Supplementary Fig. 1c). Remarkably, a 63-fold drastic expansion of CD45.2⁺ cells (57-fold expansion for WBC) were observed during the last 4 weeks towards impending death in the benzene group (n = 25) (Fig. 1c, Supplementary Fig. 1b). This rapid leukemic transformation following prolonged growth inhibition recapitulate the distinct evolutionary trajectory of benzene-induced leukemia in human.

Fig. 1. Benzene induces rapid leukemic transformation after prolonged hematotoxicity in transplant Mll–Af9 murine model.

a Outline of experiments. Transplant Mll–Af9 mice were obtained by transplanting bone marrow (BM) and spleen cells from an 8-weeks-old Mll–Af9 knock-in mouse (preleukemia state, CD45.2⁺) and equal BM cells from an 8-weeks-old B6.SJL mouse (normal, CD45.1⁺) into B6.SJL recipient mice. After 3 weeks of reconstitution, these mice were randomly divided into benzene exposure group (Target concentration: 1000 mg/m³ or 308 ppm. Conversion factor: 20 °C, 101 kPa 1 ppm = 3.25 mg/m³). (n = 25) and air control group (0 mg/m³, n = 25) for lifetime exposure. The actual concentration in the chambers is 972 mg/m³ or 299 ppm. See methods. Peripheral blood (PB) parameters and the number of preleukemic (CD45.2⁺) cells in PB were analyzed once a week. b Growth pattern of white blood cells (WBCs) and preleukemic cells during the whole 28 weeks of exposure. Preleukemic cells in PB were identified by flow cytometry analysis (FACS). The count of WBCs or preleukemic cells was significantly lower during the first 20 weeks in benzene-exposed mice than in air control mice (P < 0.05). c Growth pattern of preleukemic cells prior to the time of death. The fold-change of preleukemic cells number in benzene-exposed mice (63-fold) and air control mice (16-fold) during the last 4 weeks before death are marked on the graph. Fold-change = Y/X–1. The Y represents the number of preleukemic cells in mice at 4 weeks prior to death and the X represents the number of preleukemic cells in mice at the stage of impending death. d, e Red blood cell (RBC) counts and Platelet (PLT) counts in the PB of Mll–Af9 mice during the whole 28 weeks of exposure. Data are presented as the median with interquartile range (b, c) or the mean with SEM (d, e). P-values in b and c were obtained using the Wilcoxon rank-sum test. P-values in d and e were obtained using the unpaired Student’s t-test. Asterisk Means that the difference between benzene group and air control group was significant (P < 0.05).

We went on to use this model to explore the pathobiology of benzene-induced AML. We carried out RNA-sequencing on sorted CD45.2⁺ bone marrow cells from AML mice to examine potential functional difference between the benzene group (n = 4) and air control group (n = 4). Although AML-associated pathways were significantly upregulated in both two groups (Supplementary Fig. 2a), we found a significant difference in the transcriptome profile between benzene-exposed mice and control mice (Fig. 2a). Transcriptome of benzene group preferentially enriched for “stem or progenitor cell”, “metastasis/invasion”, “mixed-lineage leukemia (MLL)” and “proliferation”, as shown by Gene Set Enrichment Analysis (GSEA) (P < 0.01, FDR < 0.05) (Fig. 2b). Consistent with the notion that benzene and its metabolites induces DNA double strand breaks and DNA repair, five top upstream regulators (P < 1 × 10⁻³) predicted by Ingenuity Pathway Analysis (IPA) were all related to DNA damage and mutagenesis in benzene group, including activation of Trp53, Crebbp, and Dnase2a, as well as inhibition of Cgas and Ifna (Fig. 2c, Supplementary Table 1). Interestingly, Trp53 and Crebbp were previously found to positively regulate DNA damage response [9]. Dnase2a is key to degrade damaged double-stranded DNA [10], and Cgas is an inhibitor of DNA repair [11]. Together, these data suggested that AML cells from benzene-exposed mice exhibited increased activity of DNA damage response, as well as an enhanced function in stem cell activity, malignant proliferation, and invasion.

Fig. 2. Transcriptome and genome analysis of AML cells among benzene group and air control group.

a Heatmap showing changes of all expressed genes across benzene group (AML, n = 4), air control group (AML, n = 4), and preleukemia group (n = 2). The colors reflect scaled values representing the degree of expression from low to high as blue to red, respectively. b GSEA analysis was performed for benzene group (vs preleukemia) and air control group (vs preleukemia), respectively. Presentative upregulated terms that specifically enriched in one of two groups are listed (P < 0.01; FDR < 0.05; Molecular Signatures Database, C2). c Ingenuity Pathway Analysis (IPA) was performed to identify the affected upstream regulators in both groups with differential expressed genes (threshold for differentially expressed genes: P < 0.01, FDR < 0.05, ∣Fold-change ∣ > 2; threshold for affected upstream regulators: ∣z-score ∣ > 2, P < 0.01). Top five regulators that affected in benzene group but not in control group were selected to build network with their differentially expressed target genes using Cytoscape (purple and blue ellipses represent differential expressed targets associated with DNA damage response and other pathways, respectively; red and green rectangles represent activated and inhibited regulators specifically affected by benzene exposure, respectively). d Contribution of known COSMIC signatures to somatic SNVs identified in each mouse. Given the mutational profiles and 30 reference COSMIC signatures, the weighted contributions of each reference signature for each mouse were iteratively inferred by deconstructSigs. Among contributing signatures found in four mice, signature 3 is associated with abnormal DSB repair and signature 9 is thought to result from error-prone polymerase η-associated mutagenesis. The underlying mechanisms behind other signatures are detailed at https://cancer.sanger.ac.uk/cosmic/signatures/SBS/. e Circos plots showing the distribution of structural variations within each mouse. The middle circle: deletions (blue) and insertions (red). The inner circle: interchromosomal (seagreen) or intrachromosomal (darkmagenta) translocations. f Clonal structure of each AML mouse analyzed by sciClone with all neutral copy SNVs and small indels. For the clonal structure graphics of each mouse, the upper and lower graph represent the kernel density plots of VAFs and the distribution of tumor coverage(depth) along the VAFs, respectively. Mutation clusters of different colors determined by model fit represent distinct clones. Six and nine clones were found in Benzene AML_1 mouse and Benzene AML_2 mouse. Two and Three clones were found in Air control AML_1 and Air control AML_2.

To further determine the genomic basis for the observed transcriptomic changes, we carried out whole-genome sequencing (WGS) at a median depth of 36.88 × (35.96~37.07×) on benzene-exposed mice (AML stage, n = 2) and air control mice (AML stage, n = 2). Using Mll–Af9 preleukemic cells as baseline control, we identified point mutations as well as structural changes in benzene-exposed and air control mice, respectively. Although the total number of genomic alterations had no significant difference between the two groups (Supplementary Fig. 2b), functional related changes including nonsynonymous, splicing, stopgain, stoploss, frameshift SNVs, and indels were significantly higher in benzene-exposed mice than air control mice (Supplementary Fig. 2c). Furthermore, we identified more AML-associated mutations in benzene-exposed mice including those known to be essential for the self-renewal potential of leukemic stem cells (e.g., Kmt2c, Bcor, and Nras. Supplementary Table 2). These data were supportive of the transcriptional changes for an enhanced stem cell function and proliferative capability of benzene AMLs.

Considering benzene can directly cause DNA adducts and double strand breaks (DSBs), as well as to induce chromosomal aberrations in cultured cell lines [4], we further looked for potential genomic features that may specifically derived from benzene exposure in AML mice. As base substitution pattern of point mutations, termed “mutational signatures”, are often indicative of mutational processes (e.g., cellular defects or mutagen exposures), we examined the genomes for the presence of 30 COSMIC mutational signatures. This analysis showed that, comparing to air control, benzene-exposed mice had a significant higher proportion of signature 3 (Fig. 2d), which is known to be associated with abnormal DSB repair [12]. Notably, in line with the notion that genomes with DSBs are prone for structural alterations [13], benzene-exposed mice were characterized with extensive inter- or intrachromosomal translocations, and deletions (Fig. 2e). The analyses of base substitution signature and structural alterations suggested that benzene may impact on the AML genome through induction of DSBs and subsequent abnormal repair. As the AMLs in benzene and control group were developed through very distinct growth pattern of CD45.2⁺ preleukemic cells, we further carried out clonal structure analysis to infer the evolutionary process of the observed genomic alterations. The genome of each of air control AML mice was represented by a simple clonal structure with one to two subclones consisting of multiple mutations, suggesting that a subset of Mll–Af9 preleukemic cells acquired additional mutations and proceeded to malignant transformation. Notably, each of the benzene-induced AML genomes consisted multiple (5–8) subclones (Fig. 2f). This increased complexity of clonal structure suggested that benzene-induced-genomic mutations provided the basis for a genetically diversified pool of Mll–Af9 cells, and multiple subsets of preleukemic cells with unique mutations may be further selected by persistent benzene exposure to eventually transformed into AML. Together, these data indicated that benzene AML genomes bear mutational imprints associated with DSBs, a characteristic of benzene-induced DNA damage. Benzene exposure may reshape the genomic landscape and clonal structure during the development of leukemia.

In summary, our study employed a preleukemia mouse transplantation system to model benzene-induced AML. Benzene exposure resulted in a drastic difference in cellular dynamics and genome evolution of the disease. This model recapitulate the prolonged inhibitory phase followed by rapid onset of AML observed in benzene-exposed workers. Based on previous studies [1, 6, 8], a high level of approximate 300 ppm of benzene exposure is required to induce myeloid leukemia in mice (C3H/He: 8.7%; CBA/Ca: 19.3%; Trp53-deficient C3H/He: 37.5%). In order to explore the potential tumorigenic responses, an exposure concentration at the aforementioned level was selected in this mouse model study, although such level of exposure do not occur in industrial workers anymore [1]. MLL translocations are frequently occurred in pediatric, particular infant leukemia, as well as therapy-related leukemia induced by topoisomerase II inhibitors [14]. Considering children with low concentration of outdoor benzene exposure have been shown to have a significantly increased risk of AML [15], this Mll–Af9 mouse model may have relevance to specific subgroup of pediatric or therapy-related MLL-rearranged AML who had previous benzene exposure. Through combined transcriptome and genome analysis, we also showed that benzene AMLs had increased capability of malignant proliferation and self-renewal, which may be resulted from benzene-induced DSBs and genomic reconfiguration characterized by prevalence of leukemia-associated point mutations, structural alterations, as well as increased levels of clonal complexity. It is worthy to note that the current model used a preleukemia system involving a common AML oncogenic fusion gene Mll–Af9. Such preleukemic condition is present in general population. Hematological cancer-associated chromosomal abnomalities have been found in roughly 2–3% of elderly individuals without a history of blood-related disorders [16]. Around 10% of healthy elderly people (older than 65 years) have clonal hematopoiesis with somatic mutations in known leukemia driver genes DNMT3A, TET2, and ASXL1 [17]. Furthermore, around 5% of children are born with the common leukemia fusion gene, such as TEL-AML1 and AML1-ETO [18]. Benzene poisoning workers are likely to harbor oncogenic alterations. Our study may serve as a model to explore the pathobiology and to test therapeutic agent for benzene-induced AML derived from those preleukemic conditions.