Difference metabolites induced by Deoxynivalenol in 1 serum and urine of weaned rabbits detected using 2 LC-MS based metabolomics 3

39 Background: The main toxin effects of Deoxynivalenol (DON), which known as one 40 of the mycotoxins with the highest pollution rate, is the result of long-term accumulation, 41 and there are no obvious clinical signs at the early stage. Specific metabolites in blood 42 and urine can be used as biomarkers and become an important diagnostic indicator for 43 DON poisoning monitoring. At present, studies on the metabolic pathways and 44 characteristics of DON mainly focus on humans, pigs and poultry, but few study on 45 rabbits. This study aims to reveal the difference in DON-induced metabolites in the 46 serum and urine of weaned rabbits, so as to help find potential biomarkers and 47 understand the mechanism of DON in rabbits. 48 Methods: A total of 32 weaned rex rabbits were divided evenly into two groups, namely 49 the control group and DON group. Both groups of rabbits were fed with the basic diet. 50 Rabbits in DON group were intraperitoneally injected with DON at 1.5 mg/kg b.w. every 51 two days before feeding, while rabbits in control group were injected with saline at 1.5 52 mg/kg b.w. in the same way. After the 25-day trial, the serum and urine samples at 53 different experimental period were collected for LC/MS analysis. 54 Results: The results based on the LC-MS/MS method showed that DON can be 55 metabolized rapidly in blood, and urine is the main metabolism pathway for DON. The 56 data based on metabolomics illustrated that underlying biomarkers in serum were 57 mainly involved in Glycerophospholipid metabolism, Tryptophan metabolism and 58 Pentose and glucuronate interconversions, while those in urine samples involved in 59 Caffeine metabolism, Glycine, serine and threonine metabolism, and Terpenoid backbone biosynthesis. Correlation analysis suggested that DON can induce the changes in certain disease-related metabolites in serum and urine. Conclusions: The pathogenic mechanism of DON includes multiple levels, indicating 63 that DON poisoning is caused by multiple factors acting on multiple links.

individually placed in metabolic cages, were fed with the basic diet. Weaned rabbits in 132 the DON group were intraperitoneally injected with 1.5 mg/kg b.w. DON every two days before feeding in the morning, while the rabbits in the control group were injected with 134 the same amount of normal saline every two days in the same way. The experiment 135 lasted for 25 days, that is a total of 12 injections to rabbits by the end of the trail. On day 136 8, day 16 and 24 of experiment, 16 rabbits (8 rabbits each group) were selected for blood 137 collection by ear vein before injecting DON (0h), as well as 1h, 2h, 3h, 4h and 6h after 138 injecting DON, respectively. All collected blood coagulated naturally and centrifuged at 139 1500g for 10 min in order to acquire serum samples, which stored at −70°C for further 140 analysis. On day 8, day 16 and 24 of experiment, the urine samples of 24 rabbits at empty 141 stomach were collected, which centrifuged at 2,500 g for 10min and stored at -70°C until 142 sample procession, while the feces samples were collected and stored at -70°C after 143 drying at the same day. On day 25 of experiment, the serum and urine samples without 144 injecting DON were collected for LC/MS analysis. 145 Company (Guelph, Guelph, Canada). Sample preparation steps referred to the method 152 described by Brezina et al. (2014) [24]. Contents of DON and DOM-1 were detected via 153 LC-MS/MS following the protocol reported by Li et al. (2017) [14]. These detection 154 procedures were performed in the Institute of Quality Standards and Detection 155 Technology, Chinese Academy of Agricultural Sciences. 156

Preparation of serum and urine samples 157
All chemicals regeants used in the present study were analytically pure or 158 chromatographic grade. Methanol, chloroform, formic acid, water, acetonitrile were 159 purchased from CNW Company in German. l-2-chlorophenylalanine was purchased 160 from Hengchuang Biotechnology Company (Shanghai, China). In this study, 16 serum 161 and 16 urine samples from the two test groups were used for metabolomics analysis. 162 Firstly, 10 μL of 2-chloro-l-phenylalanine (0.3 mg/mL) was dissolved in methanol as 163 internal standard, then mixed accurately 100 μL serum samples and vortexed for 10 s. 164 Secondly, after 300 μL of pre-cooling mixed solution with methanol and acetonitrile (2:1, 165 v/v) was added and vortexed for 1 min, these mixtures were extracted using ultrasound 166 in ice water bath for 10 min. Thirdly, these mixed solution was centrifuged at 13,000 rpm 167 for 15 min after stood at −20°C for 30 min, then 200 μL extracted supernatant were 168 filtered with 0.22 μm pinhole filters and transferred to LC vial, which were stored at 169 -80℃ for future step.Lastly, all samples were mixed to be pooled sample at a certain ratio, 170 which acted as quality control samples (QC) sample and used to verify the results in present research. For urine samples, the treatment procedure was in keeping with the serum proceeding, except that 150 μL urine specimen was centrifuged at 13,000 rpm for 173 10 min firstly, and then 100 μL supernatant was accurately extracted and added into the 174 mix solution with internal standard. 175

LC-MS analysis 176
Metabolic profiling was monitored through an ACQUITY UPLC system (Waters 177 including 0.1% formic acid and B was acetonitrile with 0.1% formic acid, while for 182 negative ion mode, mobile phase A was water including 6.5 mM ammonium bicarbonate 183 and B was water and 95% methanol containing 6.5 mM ammonium bicarbonate. The 184 temperature of the column was 50℃, the injection volume was 5 μL and flow rate was at 185 0.35 mL/min.The elution gradient of HPLC and parameters of mass spectrum in this 186 experiment were listed in Table S1 and Table S2, respectively. Throughout the whole 187 analysis process, the QC samples were inserted at every 8 samples so as to evaluate the 188 repeatability. 189

Data processing 190
The raw data were transformed to mzML format via MSconventer, then peak extraction 191 was carried out by XCMS software. All data from positive and negative ion modes were combined into a data matrix, which contains all the information extracted from the raw 193 data that can be used for subsequent analysis. The data matrix was exported into the

Statistical analysis 211
The test data were analyzed by one-way ANOVA and t test, which expressed using mean 212 ± standard deviation (SD). p < 0.05 was considered statistically significant, and p < 0.01 213 was considered extremely significant.

Results and disscusion 215
DON concentration in rabbit serum samples of different experimental periods. 216 Studies have shown that the absorptivity of DON varied greatly among animals, for 217 example, the absorptivity of pigs, chickens, sheep and cattle was 82%, 19%, 9.9% and 1%, 218 respectively [10,25]. According to statistics, DON can reach the peak in blood after 15-30 219 min oral intake in pig, and reach the absorption peak within 3-4 h [26]. The results of this 220 study showed that the changes of DON concentration in serum of rabbits on day 8, 16 221 and 24 exhibited a highly consistent trend after DON was injected ( Fig.1). Namely, the 222 content of DON in serum before injection was consistent with that of the control group, 223 while the content of DON in serum increased rapidly after injection, reached the peak at 224 1h after injection, and then decreased rapidly. At 4 hours after injection, the concentration 225 of DON was close to the control group, and 6 hours after injection, the concentration was 226 almost the same as that of the control group. Moreover, with the increase of injection 227 days, the concentration of DON did not change significantly in the same period, 228 indicating that DON can be rapidly metabolized in the blood, with almost no cumulative 229 effect. This result suggested that, in the clinical monitoring of DON poisoning, the 230 collection of blood samples has a strong timeliness, that is, blood indicators within 1-2 h 231 after ingestion are more meaningful for detection. 232

Concentration of DON in ruine and feces samples 233
In this study, the content of DON in urine samples ( Fig.2A)  relatively lower compared with poultry or ruminants, thus pigs are more sensitive to 260 DON [29]. Previous experiments proved that the destruction induced by DON in the 261 anterior segment of the small intestine for rabbits more seriouser than to the posterior 262 segment, suggesting that part of DON in the posterior segment is converted to DOM-1 by 263 intestinal microflora, thus relieve the destruction to the ileum [14]. 264

Multivariate analysis of LC-MS data 265
Compared with the traditional HPLC-MS method, the LC-MS adopted in this study has 266 higher peak volume, separation degree and sensitivity, so it is more suitable for the 267 analysis of metabonomics for serum and urine samples of weaned rabbits. Test data of 268 serum and urine were uploaded severally into SIMCA for multivariate analysis in 269 present research. PCA was applied to analyze the entire allocation between samples and 270 stability of the whole analytic processing [18]. PCA score plots of total serum and urine 271 between two groups illustrated that the clustering of endogenous metabolites in the 272 DON group significantly changed compared with the control group ( Fig.3A and Fig.3B). 273 In addition, it can be seen from Table 2 that both R 2 X of serum (R 2 X = 0.543) and urine 274 (R 2 X = 0.662) were greater than 0.5, indicating that the model was reliable. OPLS-DA of 275 serum and urine samples were established respectively to further validate separation of 276 the metabolic profiles between two groups. As showed in OPLS-DA score plots (Fig.3C  277 and Fig.3D), either serum or urine samples between two groups were completely 278 separated, further indicating the metabolic characteristics in serum and urine were greater than 0.5, indicating that the OPLS-DA model was well established. Meanwhile, 282 the results from 200 response permutation testing (RPT), which were used to examine the 283 quality of the model in order to prevent the model from overfitting, indicated that the 284 OPLS-DA models were well established ( Fig.3E and Fig.3F). In brief, the multivariate 285 analysis of LC-MS data proved that the data quality was reliable and the models were 286 well established, as well as there were a significant difference between two groups, 287 indicating the metabolic characteristics in serum and urine were significantly affected by 288 DON exposure. 289

Screening of differential metabolites 290
The differential metabolites between groups were screened on the basis of the 291 combination of multi-dimensional analysis and single-dimensional analysis. According 292 to VIP≥1 coupled with p < 0.05 as the screening criteria in this study, total of 179 serum 293 features and 526 urine features (Table S3) were screened as discriminated variables. In Notes: N means the number of samples analyzed. R 2 X and R 2 Y refers to the cumulative interpretation rate of the model in the X-axis and Y-axis when modeling with the multivariate statistical analysis, or can be understood as the square of the percentage of the original data information retained at the X-axis or Y-axis direction, respectively. Q 2 means the parameters of the response ranking test, applying for measure whether the model is over-fitting. metabolites were upregulated and 128 were up-regulated in serum, while 327 296 metabolites were upregulated and 199 were down-regulated in urine samples. As a result, 297 26 metabolites in serum, as well as 36 metabolites in urine respectively were screened as 298 potential biomarkers based on the principle of p values from small to large (Table 3 and 299 Table 4). Furthermore, all the differential metabolites in serum and urine of rabbits were 300 analyzed applying VENN method, and the results showed that there are 15 differential 301 metabolites shared by serum and urine (Fig.4). As displayed in Table 5  To better understand the effect of identified metabolites in serum and urine samples of 336 rabbits of different groups, the enrichment of metabolic pathway was analyzed 337 separately. Firstly, all the metabolites in each group, with FC < 0.67 or FC > 1.5, were 338 selected and putted into MetaboAnalyst, which as a platform of comprehensive analysis 339 for quantitative metabonomics data based on web. Secondly, the metabolites were 340 matched according to the KEGG metabolic pathway database. Finally, the visualized 341 results of metabolic pathway analysis are illustrated in Figure 5. As shown in Fig.5 interconversions, and Ether lipid metabolism, while their pathway impact were 22%, 14%, 345 Deficiency and Sarcosinemia were correlated to the multiple diseases. The results of this 380 study indicated that DON addition would cause changes of some metabolites related to 381 specific diseases in the serum and urine of weaned rabbits. It is of great concern that the 382 first related disease in both of serum and urine samples were schizophrenia, and 383 multiple identified metabolites were associated with neurological diseases. 384 In recent years, numerous researches found that there exists a complex neuroendocrine 385 network which connect the brain and gastrointestinal tract, so it is called "brain-gut axis" tract, abnormal signal transduction is associated with numerous diseases, such as 388 inflammatory, functional gastrointestinal diseases and feeding disorders, and so on [32]. 389 DON can act on the central nervous system through quickly the blood-brain barrier, 390 cause changes in certain chemicals, and then lead to eating less or refusing food [33,34].