Plasma metabolomics study reveals the critical metabolic signatures for benzene-induced hematotoxicity

Metabolomics has been used to explore the molecular mechanism and screen biomarkers. However, the critical metabolic signatures associated with benzene-induced hematotoxicity remain elusive. Here, we performed a plasma metabolomics study in 86 benzene-exposed workers and 76 healthy controls, followed by a validation analysis in mice, to investigate the dynamical change of the metabolic profile. We found that 8 fatty acids were significantly altered in both benzene-exposed worker and benzene-exposed animal models. These metabolites were significantly associated with S-phenylmercapturic acid and WBC, and they mediated the benzene-induced WBC decline. Furthermore, in vivo results confirm that fatty acid levels were dynamically altered, characterized by a decrease at 15 days and then sharp increases at 30 and 45 days. Following these identified fatty acids, the potential metabolic pathways were investigated. Fatty acids, as precursors for fatty acid oxidation, may disturb the balance of fatty acid biosynthesis and degradation. Our results reveal that fatty acid metabolism was strongly reprogrammed after benzene exposure. This abnormal change of fatty acids might be the key metabolic signature associated with benzene-induced hematotoxicity.

redisclosed with 160 μL 15 mM ammonium acetate solution and transferred to 350 μL 96-well plate. 23 It was oscillated at 650 rpm for 10min at 10°C, then centrifuged at 4000 g for 10 min at 4°C, 24 transferred 100 μL supernatant to another 96-well plate, and waited for detection. The parameters 25 and methods of the instrument are set as shown in Table S5. There is reagent blank, system blank 26 and system balance biological sample before and after sample analysis of each batch. The addition 27 of these quality controls also monitors possible contamination and data quality during the analysis. 28 In order to eliminate errors caused by the sequence of analysis process, samples to be tested were 29 randomly measured according to group information. QC samples and blank samples are interspersed 30 in the whole sample for testing. The raw data generated by UPLC-MS/MS will adopt QuanMET 31 software (v1.0, Metabo-profile, Shanghai, China) perform peak integration, correction and 32 quantitative analysis for each metabolite. 33

Plasma metabolomics analysis 34
Plasma sample used to assess individual metabolite including amino acids, organic acids, amines, 35 fatty acids, carbohydrates, and bile acids which performed on an ultra-performance liquid 36 chromatography coupled to tandem mass spectrometry (UPLC-MS/MS) system. Briefly, samples 37 were thawed on ice-bath to diminish sample degradation. 25 μL of plasma was added to a 96-well 38 plate and the plate was transferred to the Biomek 4000 workstation (Biomek 4000, USA). Ice cold 39 methanol with internal standards was automatically added to each sample and vortexed for 5 min. 40 Then the plate was centrifuged at 4000 g for 30 min (Allegra X-15R，USA). 30 μL of supernatant 41 was transferred to a clean 96-well plate, and 20μL of freshly prepared derivative reagents was added 42 to each well. The plate was sealed and carried out at 30°C for 60 min. After derivatization, 350 μL 43 of ice-cold 50% methanol solution was added to dilute the sample and the plate was stored at -20°C 44 4 for 20 min and followed by 4000 g centrifugation at 4°C for 30 min. 135 μL of supernatant was 45 transferred to a new 96-well plate with 15 μL internal standards in each well. Serial dilutions of 46 derivatized stock standards were added to the left wells. Finally, the plate was sealed for UPLC-47 MS/MS analysis. 48 An ultra-performance liquid chromatography coupled to tandem mass spectrometry system 49 (ACQUITY UPLC-Xevo TQ-S, USA) was used to measure the metabolites. The instrumental 50 parameters of the analysis were set as shows in Table S5. A standard calibration solution with more 51 than 300 standards at 7 different concentration levels were analyzed to construct the calibration 52 curve. Peak annotation and quantitation were conducted by TargetLynx application manager (Waters 53 Corp., Milford, MA, United States). Internal standards were added to the test samples in order to 54 monitor analytical variations during the entire sample preparation and analysis processes. 55

Basis for selection of exposure time and dose in animal experiments 56
The hematopoietic system is a recognized target of benzene exposure. Long-term chronic 57 benzene exposure could inhibit the renewal and differentiation of BM HSCs, ultimately resulting in 58 altered hematology (1). To investigate the changes in metabolite profiles at different stages of 59 benzene exposure-induced hematotoxicity, mice were treated with subcutaneous injections of 125 60 mg/kg of benzene for 15, 30 and 45 days. Respiratory inhalation and dermal contact are the two 61 main pathways of benzene exposure in humans (2). Therefore, dynamic inhalation and subcutaneous 62 injection are widely used in animal experiments. In the pre-experiments of our group, it was found 63 that the ratio of mixed gas and pure gas in the dynamic inhalation device was unstable at low-level 64 benzene, which might lead to the system error of benzene exposure concentration. In addition, real-65 time benzene exposure concentrations in the contaminated room were monitored by activated 66 5 carbon adsorption-CS2 desorption combined with gas chromatography, which is a complex and 67 time-consuming procedure 68 (http://www.nhc.gov.cn/wjw/pyl/201712/0c849c68aa9b48549056712aea6f97b9.shtml). Compared 69 to dynamic inhalation, subcutaneous injection is convenient and allows the injection of accurate 70 doses by calculation to mice. In addition, the choice of dose is critical to the success of a mouse 71 model of benzene exposure-induced hematotoxicity. Previous literature on benzene exposure in 72 mice indicated that subcutaneous administration of 150 mg/kg benzene for 30 days had been 73 acknowledged by numerous researchers due to its stable and reproducible hematotoxicity (3,4). 74 Therefore, our current experimental dose was close to the previous studies, and a mouse model of 75 benzene-induced hematotoxicity had been successfully constructed at this dose. In traditional 76 toxicology, about 1/10 of the mouse life span (1-2 years) is considered a subchronic toxicity cycle. 77 Therefore, the 30-day exposure period was widely used in in vivo studies in benzene exposure-78 induced hematotoxicity (4,5). However, it is not clear whether there are dynamic changes in 79 hematotoxicity and metabolic disorders due to benzene exposure during different exposure periods. 80 Taken together, subcutaneous injections of 125 mg/kg benzene for 15, 30 and 45 days were selected 81 in our subsequent animal experiments. 82 83  Table S1.

Effect indicators
Gender Control group (n=76) Benzene-exposed group (n=86) p<0.05 indicates that the difference is statistically significant; q-value is adjusted p-value using FDR. Bold: q-value < 0.25.  Mann-Whitney-Wilcoxon (U-test) was performed to compare plasma differential metabolites levels between two group.
p < 0.05 indicates that the difference is statistically significant. q value is adjusted p-value using FDR. Bold: q value < 0.25.
Fold change (FC): the ratio of Exposeure/Control.  Linear regression models of urinary benzenes metabolites with whole blood cell and liver function indexes after adjusting for age, BMI, smoking and alcohol consumption.

Table S3. Associations of urinary benzenes metabolites with whole blood cell and liver function indexes with adjusting age, smoking, drinking, and BMI in subjects
*p < 0.05; ** p < 0.01; q value is adjusted p-value using FDR. Bold: q value < 0.25.