Metabolomics coupled with pathway analysis characterizes metabolic changes in response to BDE-3 induced reproductive toxicity in mice

Polybrominated diphenyl ethers (PBDEs) may affect male reproductive function. 4-bromodiphenyl ether (BDE-3), the photodegradation products of higher brominated PBDEs, is the most fundamental mono-BDE in environment but is less studied. The purpose of this study was to investigate the reproductive toxicity induced by BDE-3 and explore the mechanism by metabolomics approach. In this study, mice were treated intragastrically with BDE-3 for consecutive six weeks at the dosages of 0.0015, 1.5, 10 and 30 mg/kg. The reproductive toxicity was evaluated by sperm analysis and histopathology examinations. UPLC-Q-TOF/MS was applied to profile the metabolites of testis tissue, urine and serum samples in the control and BDE-3 treated mice. Results showed the sperm count was dose-dependently decreased and percentage of abnormal sperms increased by the treatment of BDE-3. Histopathology examination also revealed changes in seminiferous tubules and epididymides in BDE-3 treated mice. Metabolomics analysis revealed that different BDE-3 groups showed metabolic disturbances to varying degrees. We identified 76, 38 and 31 differential metabolites in testis tissue, urine and serum respectively. Pathway analysis revealed several pathways including Tyrosine metabolism, Purine metabolism and Riboflavin metabolism, which may give a possible explanation for the toxic mechanism of BDE-3. This study indicates that UHPLC-Q-TOFMS-based metabolomics approach provided a better understanding of PBDEs-induced toxicity dynamically.

may be more harmful to human health 14 . Mono-BDE may induce genetic recombination in mammalian cells 15 . Our previous studies in C. elegans also demonstrated BDE-3 could induce reproductive dysfunction and germ cell apoptosis by induction of ROS and DNA damage 16 . By far, however, few study in mono-BDE toxicity have been reported in rodents 17 .
Metabolomics protocols are used to comprehensively characterize the metabolite content of biological samples by exploiting cutting-edge analytical platforms. Liquid chromatography-mass spectrometry technology platform is the most commonly used analytical technique for metabolomics research in recent years.
Nowadays, non-targeted metabolomics has gained widely attention as a method of profiling endogenous metabolites because of its high sensitivity, high throughput, and unbiasedness. Therefore, Metabolomics techniques were gradually applied to many aspects, such as evaluating and monitoring drugs and even atmospheric pollutant toxicity 18 . The analysis also showed enormous potential for the detection of health effect of POPs, e.g. metabolomic-based profiling revealed plasma responses to dioxin-associated dietary contaminant exposure 19 ; identificated changed metabolites as potencial hepatotoxicity biomarker of Polychlorinated Biphenyls (PCBs) and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) 20 ; determinated toxic effects of PCBs and explained the mechanism of metabolic disturbance in obesity 21 ; revealed the mechanisms of complex pesticide mixtures 22 exposure increase oxidative stress and disturb energy metabolism.
In this study, we examined the effects of BDE-3 on the reproductive function in mice, with the lowest dose of 0.0015 mg/kg/day, which equal to the highest ingestion concentration of PBDEs in human (141 ng/kg/day) by the U.S. Environmental Protection Agency (EPA), 1.5 mg/kg/day, an equal dose to the lowest effective dose of its congener of BDE-47 17 , middle dose of 10 mg/kg/day and high dose of 30 mg/kg/day, respectively. We then performed metabolomics analyses for the testis, urine and serum samples after BDE-3 treatment using UPLC-Q-TOF/MS. The differential metabolites were identified by the pattern recognition using PLS-DA and univariate analysis for the control and BDE-3 treated mice. Our results may provide more evidence to better understand the mechanism of PBDEs-induced reproductive toxicity.

Result
Body, tissue weight and clinical observation. During the experimental period, animals were weighted once a week and no significant differences were observed in the body weight or clinical observation of the animals ( Fig. S1 and Table S1). Testis and epididymis weight of each animal was measured on the day of the anatomy. No statistical differences in the testis weight or epididymis weight were found between the solvent group and BDE-3 groups after 6 weeks of BDE-3 treatment (Fig. S2).
Sperm count, vitality and morphology. BDE-3 treated mice showed no change in sperm vitality compare with vehicle control group, (Fig. 1A). However, the sperm count was dose-dependently decreased by the treatment of BDE-3 (Fig. 1B). The average counts of sperm at 1.5, 10 and 30 mg/kg dose groups were 3.2, 2.7, 2.1 × 106/mL, and were significantly decreased by 36%, 47% and 57% compared to the vehicle control group, respectively. Moreover, as shown in Fig. 1C, in the sperm morphology analysis, the rate of tail folding sperm in BDE-3 treated mice were increased 1.4 to 2.2-fold higher compared to the vehicle group although without statistical difference. No significant differences were observed in the banana gate, enlarged-headed, amorphous, double-headed and double-tailed (Fig. S3).
Testis and epididymis histopathology. Compared with vehicle control group, minimal to mild germ cells decrease were found in seminiferous tubules in four males (4/6) of 30 mg/kg BDE-3 treated group as shown in Fig. 2A,B. Mature sperm decrease was revealed in epididymides in three males at 30 mg/kg BDE-3 treated group with 1 animal showed cellular debris in lumen and inflammatory cell infiltration in epididymal interstitium ( Fig. 2C-H). No microscopically change was found in 0.0015, 1.5 and 10 mg/kg treated mice.
Metabolomics profiling analysis. Metabolomic data of the testis, urine and serum samples were acquired using the abovementioned optimal conditions of UPLC-Q-TOF/MS. The representative total ion chromatograms (TICs) in both positive and negative ion modes of testis, urine and serum from the control and BDE-3 groups are shown in Fig. 3.
Initially, unsupervised PCA was used to observe the stability of the assay in the sequence between QC samples and other testis, urine and serum samples in different groups. Score plots from the PCA model have shown that the QC samples clustered together well in positive and negative mode, respectively, indicating the stability of the system was satisfying (Fig. S4).
In order to fully differentiate the testis, urine and serum metabolites between the control and the BDE-3 groups, the PLS-DA model was applied. PLS-DA is an efficient method to identify the ions that contribute to the clustering of samples. It also helps to eliminate non-correlated variations within the data set.
Score plots from the PLS-DA model have shown that the testicular samples from control group and dose groups clustered together respectively, and were clearly separated from each other (Fig. 4). Besides, the different BDE-3 groups exhibited an obvious separation trend, indicating that BDE-3 resulted in difference indeed among groups. The similar separation trend can also be observed in the score plots from the PLS-DA model of urine and serum samples (Figs S5 and S6).
In order to further differentiate the testicular metabolites, we applied PLS-DA model to characterize the differences between the control and each different BDE-3 groups. There was a distinct clustering between the control group and each BDE-3 groups at different dosages by UPLC-Q-TOF/MS method (Figs S7 and S8). The detailed model validation parameters (R2X, R2Y, and Q2) were shown on the Figs S7-S12. That showed good degree of fitting and predictive ability to screen the differential variables between groups.
The corresponding S-plot showed the contribution of different variables for the difference between the control and the BDE-3 groups. Each point in the S-plot represents an ion. Ions far away from the origin are significantly important to the differences between groups and have greater VIP values. Independent sample t-test was performed to assess the statistical significance. The important ions differentiating the BDE-3 groups from the control group at different dosages (VIP > 1, p < 0.05) were identified according our previous methods. Finally, 76 differentiated metabolites were identified in testis and listed in Table 1. Furthermore, we applied the PLS-DA model to characterize the differences among urine and serum samples in each group. Score plots from the PLS-DA model have shown that dose groups clustered together respectively, and were clearly separated from the control group (Figs S5 and S6). There was also a distinct separation of control group and each BDE-3 groups at different dosages by UPLC-Q-TOF/MS method (Figs S9-S12). We used T-Test method to screen the significantly changed metabolites between the control group and each BDE-3 groups at different dosages. Using the P < 0.05 and VIP > 1 as cut-off, 38 and 31 differentiated metabolites were identified in urine and serum respectively in positive mode and negative mode eventually (See Tables 2 and 3).

Discussion
BDE-3 is the most fundamental mono-BDE which could be degraded from PBDEs in environment, although its bioaccumulation properties and potential healthy risk have raised concern, the effect on the adult male reproductive system are currently unclear. To our knowledge, it is the first report on assessment of male reproductive toxicity induced by mono-BDE in rodent. In this study, we found that short-term exposure to low-dose BDE-3 may have adverse effects on spermatogenesis in adult male mice. The sperm count decreased after the treatment of BDE-3 for 6 weeks at a very low dose (0.0015 mg/kg/day, a dose based on environmental exposure), and at the dose of ≥1.5 mg/kg/day, the sperm count decreased with statistical significance, and microscopical changes of germ cell loss in seminiferous tubules and epididymides were clearly observed at the dose of 30 mg/kg. Our previous studies on C. Elegans revealed that BDE-3 could induce ROS and DNA damage in germ cell 16 , which might be involved in the male reproductive toxicity that observed in mice.
Metabolomics has been applied in toxicity studies of chemicals as a newly emerging technology, especially in risk assessment of low dose exposure of environmental pollutants [26][27][28][29] , it may lead to critical insights because of its high sensitivity. Although several studies on toxicological effects of PBDEs using metabolomics analysis have been reported [30][31][32] , studies in rodent are limited and few data of mono-BDE is available. In this study, to obtain a better understanding of the toxic mechanism of BDE-3, we applied the UPLC-Q-TOF/MS profile to the metabolites of testicular tissue, urine and serum in the control and BDE-3 treated mice at different dosages.
In our study, we analyzed the testis tissue, serum and urine of BDE-3 treated mice. The differential metabolites from biological samples at different levels were screened and analyzed. As mentioned above, testis is susceptible to BDE-3 induced toxicity, so the metabolic alterations of testis tissue could, to some extent, reflect BDE-3 induced reproductive toxicity. Based on the pathway enrichment analysis on differential metabolites screened in testis tissue, we could preliminarily reveal that BDE-3 induced toxicity disturbed organism metabolic regulation network. Meanwhile, the metabolic alterations would spread through blood circulation and urine directly or indirectly. So pathway enrichment analysis and correlation analysis on differential metabolites screened in serum and urine could further clarify and verify the BDE-3 induced metabolic alterations. Besides, body fluid sample like serum and urine could reflect the systematic metabolic state of the body, so the differential metabolites which were highly correlated with BDE-3 induced toxicity screened in body fluid were expected to be potential toxic biomarker, improving the efficiency to detect the toxicity of PBDEs. In total, 76 differential metabolites in testicular tissue, 38 differential metabolites in urine and 31 differential metabolites in serum were identified. According to the score plots from the PLS-DA model of tissue and urinary samples, we can observed that the different BDE-3 groups showed a consistent separation trend from the control group. While as for serum samples, it had a minor pullback in the last BDE-3 group (30 mg/kg/day), and some metabolites corresponded with this variation trend, such as Inosine, L-Phenylalanine and some lipids metabolites (Fig. S14). We thought the possible reason is that    the mice had developed resistance to the BDE-3 induced toxicity and self-healing capacity. Furthermore, the metabolic changes in urine was secondary to the changes in serum and target organs injury of BDE-3. That may be why the urine samples did not show a similar pullback change in the highest dose group.

Nucleotides metabolism.
Nucleotides play a key role as precursors of DNA and RNA, as activated intermediates in many biosynthetic processes, and as metabolic regulators. In the testis, nucleosides and nucleobases are important substrates of the salvage pathway for nucleotide biosynthesis 33 . We observed several purine and pyrimidine nucleobases metabolites in this study, such as Adenosine, Xanthine, Hypoxanthine and UDP-glucose, which significantly decreased in testicular tissue of BDE-treated mice compared with the control ones. Indicating that there will not be enough substrates for DNA and RNA biosynthesis.
In the mammalian spermatogenic pathway, Sertoli cells can provide nutrients and metabolic precursors to spermatogenic cells located within the blood-testis barrier (BTB) 34 . Because large amounts of nucleotides are required for spermatogenesis, less substrates will decelerate spermatogenesis. Accordingly, BDE-3 might affect the transportation of nucleobases and nucleosides across the BTB, or disturb the uptake of substrates by Sertoli cells. On the other hand, we observed several purine nucleobases metabolites, such as Adenosine and Inosine, significantly increased in serum and urine of BDE-treated mice compared with the control ones, which was potentially related to the reduction of purine and pyrimidine nucleobases metabolites level in the testis. Adenosine is an important endogenous nucleoside. Extracellular adenosine has the potential to influence the target cell metabolism in many aspects 35 . Adenosine can be phosphorylated to AMP, which can further generate Inosine by adenosine deaminase and IMP by AMP deaminase. The previous study had determined that Adenosine is also an important signaling molecule which is released under inflammatory conditions and there are four distinct subtypes are known, termed A1, A2A, A2B and A3 36 . It usually shows its endocrine effects including insulin and glucagon secretion interference by A1 receptor in mice 37 . Besides, Hiroyasu Nakata presented the first purification of a peripheral A1 adenosine receptor of rat testis membranes 38 . In this study, we observed reduction of nucleobases and nucleosides level in testis and the accumulation in serum and urine accordingly. Less nucleobases and nucleosides in testis will affect spermatogenesis and the extra purine nucleobases metabolites in serum will disturb the internal environment homeostasis, which was consistent with previous studies, indicating that the BDE-3 possibly disturbed the above purine and pyrimidine metabolism, and induced reproductive toxicity.
Lipids metabolism. Lipids are abundant in testicles, which play an important role in membrane structure and function, energy storage and cell signaling 39 . Linoleic acid (LA) (18:2n-6) is the major Polyunsaturated fatty acid (PUFA) in vegetable oils and is a metabolic precursor to arachidonic acid (AA) (20:4n-6) 40 . A Previous report has indicated that LA family PUFA plays a criticle role in testicular function 41 . AA (20:4n-6) is a compound of important potent bioactivity, which mainly exists in phospholipid and LA mainly exists in phospholipid and triglyceride 41 . AA affects testis as a membrane component and providing energy storage, and a signal molecule in regulating steroidogenesis in Leydig cells as well 42 . In this study, decreased levels of LA and AA were found in BDE-3 exposed groups. Similarly, the increasing consumption of AA and LA level and corresponding ascending oxidative stress were found in testis after Bisphenol A exposure 43 . PUFA composition change observed in this metabolomics study was consistent with decreased spermatogenesis in this mice model. Therefore, testicular AA and LA alterations might be involved in BDE-3 testicular toxicity. Meanwhile, PUFA is very susceptible to peroxidation 44 . Thus, we speculated that the PUFA composition alterations indicated testicular oxidative stress.
We found significant increased GSH and L-carnitine levels in BDE-3 exposed group. The efficacy of l-carnitine and GSH as antioxidant substance has been confirmed both in humans and mice 45 . This kind of anti-oxidative composition might biologically adapt to oxidative stress 46 , thus the alterations was consistent with the hypothesis that testicular oxidative stress was increased in the BDE-3 exposed group. Oxidative stress could lead to both membrane lipid peroxidation and DNA fragmentation in testes 47 . It is reported that spermatogenesis 48 and Leydig cell steroidogenisis 49 are both vulnerable to oxidative stress 43 . There are clinical studies demonstrating that male infertility patients showed higher oxidative stress 50 . Additionally, oxidative stress and disturbed equilibrium of oxidant/antioxidant has been considered as a major mechanism of reproductive toxicity 51 . In present study, a possible explanation of decreased spermatogenesis of the mice model is oxidant/antioxidant imbalance, which has been well supported by population and animal studies 52,53 . Besides, in our study, a series of metabolites related to lipids metabolism such as lysophosphatidylcholine (LPC) and lysophosphatidyl ethanolamine (LysoPE) were detected, especially LPE significantly decreased in testicular tissue of the BDE-treated mice compared with the control group. LPE is an amphiphilic metabolite that is produced from membrane-phospholipids via the activation of phospholipase A2 (PLA2) 54 , and it also can synthesize PE under the action of LPCAT3 enzyme 55 . On the one hand, LysoPEs can stimulate invariant natural killer T cell activation through self-antigenicity, suggesting a possible role in innate immunity 56 , so we hypothesized that decreased LysoPEs in testicular tissue may cause its own immune disorder. On the other hand, it is essential that lysophospholipases (LysPLAs) and other phospholipases regulate phospholipid levels to maintain membrane homeostasis, flexibility and permeability 57,58 . This is critical to cell maintenance. In our study, the increase of various lipids metabolites potentially contributed to the BDE-induced toxicity.
Other metabolic pathway. Riboflavin (vitamin B-2) belongs to micronutrients that play important roles in carbohydrate energy metabolism 59 . There are studies that suggest that the supplement of Ribovlavin, can significantly reduce the embryolethal action of methophenazine 60 . On the contrary, other studies suggest a syndrome, which included various limb, brain, orofacial, gastrointestinal, and miscellaneous malformations, can be induced by an intense riboflavin deficiency 61 . Besides, cytoflavin is proven to reduce the manifestations of mitochondrial dysfunction and the cell damage degree because of the antioxidant and membrane-protective properties 62 . In this study, the reduction of riboflavin in the BDE-3 groups, played an important role in the BDE-3 induced reproductive toxicity.
Several amino acids and their metabolites significantly changed in both blood and urine of the BDE-treated mice. They are mainly involved in arginine and proline metabolism, valine, leucine and isoleucine degradation, cysteine and methionine metabolism, phenylalanine metabolism, tyrosine metabolism, tryptophan metabolism, glycine, serine and threonine metabolism, D-Glutamine and D-glutamate metabolism, cysteine and methionine metabolism, especially valine, leucine and isoleucine biosynthesis. Arginine, one of the most metabolically versatile amino acids, is a semiessential or conditionally essential amino acid in humans and serves as a precursor for the synthesis of urea, nitric oxide, polyamines, proline, glutamate, creatine, and agmatine. Arginine is metabolized through a complex and highly regulated set of pathways 63 . Besides, the present studies found that the arginine-rich dipeptides, in particular Proline-Arginine (PR), are potently neurotoxic. The Proline-Arginine dipeptide (PR), is potently neurotoxic when expressed in vivo and neurons with nuclear PR aggregates have a much higher risk to undergo degeneration 64 . In this study, L-Proline is increased in serum of BDE-treated mice. Beta-Hydroxyarginine and creatine were decreased in urine. The BDE-induced toxicity was possibly related to the disturbed arginine and proline metabolism.
In addition, we observed that L-Tyrosine, L-Leucine and L-Isoleucine, significantly decreased in the urine, while they increased in the blood, in the BDE-treated mice. These results suggested that the thiamine metabolism, valine, leucine and isoleucine degradation, and especially the polarity and clearance of the above metabolites are disturbed during the formation of HLP-induced toxicity, thus resulting in the abnormalities of body's reproductive system metabolism. In this study, we examined the effects of BDE-3 on the reproductive function in mice, with the lowest dose of 0.0015 mg/kg/day, which equal to the highest ingestion concentration of PBDEs in human (141 ng/kg/day) by the U.S. Environmental Protection Agency (EPA), 1.5 mg/kg/day, an equal dose to the lowest effective dose of its congener of BDE-471, middle dose of 10 mg/kg/day and high dose of 30 mg/kg/day, respectively.

Chemicals and reagents.
Fifteen-weeks-old male mice were randomly allocated to 5 groups (n = 6 per group) and treated intragastrically with BDE-3 dissolved in corn oil 6 days/week for consecutive six weeks at different dosages of 0.0015, 1.5, 10 and 30 mg/kg. The vehicle control group received corn oil at the same manner. Mice were forced into metabolismcage for 24 hours once every two weeks to facilitate the collection of urine, the collections were immediately centrifuged at 4000 rpm for 5 min, and the supernatant was stored at −80 °C until being analyzed. Twenty-four hours after the last treatment, the mice were anesthetized by chloral hydrate. The blood samples were collected via eyeball and clotted at 4 °C for 40-60 minutes and then centrifuged at 4000 rpm for 5 minutes to obtain serum. The serum samples were stored at −80 °C until analysis. After blood collection, the epididymis and testis were dissected and weighed (wet weight). We collected the urine at different time points, but to match our serum samples (collected at 6th week), we did metabolomics profiling analysis only using the urine which were collected at 6th week. The remaining urine samples might be used for some deeper researches in the future. Sperm count, vitality and morphology analysis. For sperm count analysis, one epididymis from each individual was put in a 1.5 mL tube with 37 °C pre-warmed M199 (containing 0.5% Fetal Bovine Serum) and cut into pieces for the preparation of sperm suspension. A drop of sperm suspension was put on a counting plate to measure the sperm count and sperm vitality. For sperm vitality examination, at least 200 sperm were checked and divided into 4 categories: I-fast straight forward movement, II-slow forward movement or rotation, III-in situ fibrillation or spinning, IV-no movement. Rate of sperm vitality % = (I + II + III)/(I + II + III + IV) × 100%. For morphological examination, 1-2 drops of sperm suspension were smeared on a slide. After drying, the slides were fixed in methanol for 5 minutes and stained with eosin for about 60 minutes before examined. At least 500 intact sperm per individual were examined under the microscope to calculate the ratio of abnormality. The abnormality was divided for 6 types: banana gate, enlarged-headed, amorphous, tail fold, double-headed and double-tailed. Rate of sperm abnormality (%) = total deformed sperm/total observed sperm × 100%.
Histopathology. One testis from each animal was fixed in modified Davidson solution and one epididymis of each animal was fixed in 10% neutral formalin fixative. After being well fixed, the tissues were sampled, dehydrated in automatic tissue hydroextractor (Pathcentre, Thremo) and embedded in paraffin embedding machine (Shandon Histocentre 2, Thremo) and then cut into 3 µm paraffin section in microtome (Shandon Finesse E+, Thremo). The sections were stained with hematoxylin and eosin. All tissues were analysised by manual check and scanned by slice scanning system (Scan ScopeXT, Aperio).

UHPLC-Q-TOF/MS metabolomics analysis.
The frozen urine and serum samples were thawed at room temperature. 300 μL methanol solution that contained 5 μg/mL 2-Chloro-L-phenylalanine as the internal standard was added into 100 μL urine or serum samples. The mixed samples were homogenized by vortex for 5 min and then were placed at room temperature for 10 min and centrifuged at 13,000 rpm, 4 °C for 15 min. 150 μL of supernatant were added into the sample vials.
The frozen testicular tissue samples were thawed at room temperature. Then they were weighed, and homogenized in 900 μl 80% methanol solution that contained 5 μg/mL 2-chloro-l-phenylalanine as an internal standard. Then the mixture were centrifuged at 13,000 rpm, 4 •C for 15 min. 150 μl of supernatant were added into the sample vials.
A quality control sample (QC) was prepared by mixing the same volume from all supernatant samples (10 μL from each sample). The QC sample was run six times prior to the start of the analytical run to "condition" the system and analyzed after every 9 samples to check the system stability 65 .
UHPLC-Q-TOF/MS analysis was performed using an Agilent 1290 Infinity LC system coupled to an Agilent 6538 Accurate-Mass Quadrupole Time-of-Flight (Q-TOF) mass spectrometer (Agilent, USA). Chromatographic separations were performed at 25 °C using a Waters XSelect ® HSS T3 analytical column (2.1 mm × 100 mm, 2.5 μm, Waters, Milford, MA). The flow rate was 0.4 mL/min and the injection volume was 3 μL. The mobile phase consisted of 0.1% formic acid (A) and ACN modified with 0.1% formic acid (B). The gradient elution conditions for urinary samples were 5% B at 0-1 min, 5-20% B at 1-6 min, 20-50% B at 6-9 min, 50-95% B at 9-13 min, 95% B at 13-15 min and followed by re-equilibrated step of 5 min. As for serum samples, the following gradient program was used: 5%B at 0-2 min, 5-95%B at 2-17 min, 95%B at 17-19 min. The gradient elution conditions for tissue samples were 5% B at 0-2 min, 5-95% B at 2-13 min, 95% B at 13-15 min. The capillary voltage was 4 kV in positive mode and 3.5 kV in negative mode, the drying gas flow was 11 L/min, and the gas temperature was 350 °C. The nebulizer pressure was set at 45 psig. The fragmentor voltage was set at 120 V and skimmer voltage was set at 60 V. The reference ions were 121.0508 and 922.0098 in positive mode, 119.03632 and 966.000725 in negative mode. Data were collected in a centroid mode and the mass range was set at m/z 50-1100 using an extended dynamic range. MS/MS analysis was carried out to study the structure of the potential biomarkers and the collision energy was range from 10 to 40 eV. Data processing and statistical analysis. In the analysis of body weight, testis and epididymis weight, sperm count, vitality and morphology, the Dunnett-t test was used to determine significant differences between BDE-3 exposure groups and the vehicle control group after the one-way ANOVA was performed. A value of P < 0.05 was considered significant. SPSS version 17.0 (IBM, USA) was used for statistical analysis.
The raw UHPLC-MS data were transformed into a common data file format (.mzdata) using MassHunter Qualitative software. The interferences of the isotopes were excluded and the absolute peak height was set at 300 counts. In the R software platform, the XCMS program was applied for peak extraction, peak alignment, and automatic integration 66 . The 80% rule was used to filter the ions and the remaining ions were then normalized to internal standard peak area. The Retention time (Rt)-m/z pairs, observations, and relative ion intensities were imported into SIMCA-P 11.0 software package (Umetrics, Umea, Sweden) for the multivariate statistical analysis. Prior to multivariate analysis, the resultant data matrices were mean-centered and scaled to Pareto variance. Unsupervised principal component analysis (PCA) and partial least squares discriminate analysis (PLS-DA) were applied to observe the stability of the assay in the sequence, the separating trends, and further identify the metabolite candidates contributing to the clustering of samples. We applied variable importance plot (VIP) representing the confirmation of the importance or power of the selected candidates to select metabolite candidates with the threshold value of 1. In addition, the model of PLS-DA was evaluated according to the cross-validation of R2, Q2 value and permutation test. An independent sample t-test was performed for the statistical analysis using SPSS version 17.0 (IBM, USA) and p < 0.05 was considered statistically significant.
Pathway enrichment analysis. We applied the Pathway Analysis module on Metaboanalyst website (http://www.metaboanalyst.ca/). We firstly import the differential metabolites of testis tissue, serum, and urine to this website, matching the HMDB, PUBCHEM and KEGG database (The mass error tolerance was "<15ppm"), and the metabolites were categorized based on the link from KEGG. Select mouse as species to do pathway enrichment analysis. The results were shown on

Conclusion
In this study, we successfully observed the reproductive toxicity induced by BDE-3 in mice via clinical observation, Sperm count and vitality, Sperm morphology and Histopathology examination. A new scientific metabolomic method based on UPLC-Q-TOF/MS was applied to profile the differential metabolites in testis, urine and serum in the control and BDE-3 treated mice at different dosages. In total, 76 differential metabolites in testicular tissue, 38 differential metabolites in urine and 31 differential metabolites in serum were identified. Pathway analysis revealed that several pathways were potentially related to BDE-3 induced toxicity including Arginine and proline metabolism, Valine, leucine and isoleucine degradation, Phenylalanine metabolism, Cysteine and methionine metabolism, Tyrosine metabolism, Purine metabolism, Pentose and glucuronate interconversions, Tryptophan metabolism, Biotin metabolism, Oxidative phosphorylation, Riboflavin metabolism, and Glycerophospholipid metabolism. These findings demonstrated the BDE-3 induced reproductive toxicity in mice. Our study may be helpful to arouse our awareness of the toxicity of PBDEs and to give a possible explanation for the mechanism, which indicates that UHPLC-Q-TOFMS-based metabolomics approach allowed a better understanding of PBDEs-induced toxicity dynamically. Data Availability. The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.