The hemolytic behavior of OPD and OPD' in vivo
As it was shown in Figure 2 and Table 2, Hemolysis was found in each drug treatment group, and hematological indicators related to hemolysis, such as RBC, HGB, RET, and RET%, were significantly changed. Compared with the NC group, RBC and HGB decreased significantly, while RET and RET% were significantly increased, which means that RBCs were destroyed faster, and the immature RBCs (i.e. RET) were released to the peripheral blood increased. Indicated that the lifespan of RBC may be shortened. The results of the urine examination showed that the urobilinogen (URO) of each treatment group changed significantly, and compared with the NC group, they were all positive (degree 1+ to 2+. Under normal circumstances, there should be no red blood cells, urobilinogen and hemosiderin in the urine, which is represented by“-”. However, with the occurrence and aggravation of hemolysis, the number and concentration of RBC, URO, and hemosiderin will increase, expressed on a scale from 1+ to 4+.). The results of the urine occult blood tests were consistent with the results of the URO test. The Rous method  was used to determine the hemosiderin in urine, which is a commonly used clinical test to diagnose intravascular hemolysis. As shown in Table 2, obviously scattered or piles of blue glitter particles were observed in each treatment group (degree 1+ to 2+). These indicators confirm that OPD and OPD' can cause intravascular hemolysis  when used alone or in combination.
From the perspective of the chemical structure and physical and chemical properties of OPD and OPD', hemolysis caused by OPD' can be reasonably explained by its chemical structure, but OPD cannot [24-26]. In another article we published , relationship of hemolytic behavior in vitro of OPD and OPD' was studied. The results showed that OPD' has hemolytic effect in vitro, but OPD does not. In other words, OPD' hemolysis occurred both in vivo and in vitro, but the OPD only occurred in vivo. This indicates that the structure of OPD' is more likely to cause hemolysis, meanwhile, the inconsistency of OPD’s hemolytic behavior in vivo and in vitro may be closely related to the metabolic process of OPD after entering the body. This is also the problem that this article wants to reveal when conducting metabolomics and lipidomics research.
Global metabolic and lipidomic shifts induced by OPD and OPD'
The metabolic and lipidomic profiles were acquired under positive and negative ionization modes using UPLC-MS. A total of 10364 ions in ESI (+) and 9741 ions in ESI (-) were obtained by non-targeted metabolic profiling. In lipidomic profiling, 313 ions in ESI (+) and 260 ions in ESI (-) were obtained. PLS-DA score plots were applied to identify the differential metabolites or lipids among OPD and OPD' groups. Figure. 3 showed the clear segregation of OPD and OPD' from NC group, indicating that both OPD and OPD' induced obvious disturbance of inter-cellular metabolites and lipids.
Metabolic changes induced by OPD and OPD' in vivo
According to OPLS-DA model, Variable importance of projection (VIP) value was obtained, and VIP> 1 is the one of criterion for screening potential biomarkers. In each comparison group, the number of differential metabolites we found were 145 (NC vs OPD), 124 (NC vs OPD') and 84 (OPD vs OPD'), respectively. The volcano graph can be used to visualize the p value and Fold change value, which is helpful for screening different metabolites (Figure. S1). Top 20 differential metabolites were listed in Table 3 among different comparison group, and the distribution of fold changes of metabolites in different comparison group from the perspective of sub-class were shown in Figure. 4.
As it were shown in Table 3 and Figure. 4. Compared with the NC group, OPD and OPD' group had similar results in terms of the type, quantity and VIP value of the differential metabolites. In top 20 differential metabolites, LysoPC account for 45% in OPD and OPD' group, while the phospholipid differential metabolites account for 75%. Among them, the metabolites correlated to hemolysis were labeled by “b”. The abnormal increase of LysoPC and the disorder of phospholipid metabolism were consistent with the phenomena observed in in vivo hemolysis experiments.
When OPD was compared with OPD', the same differential metabolites based on comparison with NC group are neutralized, therefore, more details between OPD and OPD' in vivo metabolic processes were discovered. In its top 20 differential metabolites, LysoPC account for 20%, while cholic acid ranked first place（VIP value was 28.13, Fold change value was 1.72）. In another article we submitted, molecular modeling was adopted and screen out a protein called Q9NPD5 which can interact with both OPD and OPD'. This protein belongs to organic anion transporters (OATP) family and it was closely related to cholic acid metabolism , existing studies have confirmed that the down-regulation of OATP1A2 and OATP1B3 in particular leads to abnormal fetal bile acid metabolism between maternal and fetal fetuses . This may explain from another perspective how OPD and OPD' induce hemolysis. OATP1B3, which is highly expressed in the liver, is localized at the plasma membrane at the subcellular level. When OPD or OPD' enters the liver for metabolism, it binds to OATP1B3 , which disturbs the fluidity of the cell membrane and further induces hemolysis.
Lipidomic changes induced by OPD and OPD' in vivo
A total of 82 lipids were identified as differential lipids. In each comparison group, the number of differential metabolites we found were 37 (NC vs OPD), 31 (NC vs OPD') and 14 (OPD vs OPD'), respectively. As were shown in Table 4 and Figure. 5. Glycerophospholipids (GPs) and glycerolipids (GLs) was significantly perturbed. Among GPs, phosphatidylcholines (PC) changes accounted for the largest proportion, which was 48.65% (NC vs OPD), 54.05% (NC vs OPD'), and 64.29% (OPD vs OPD') respectively. This was consistent with the changing trend of different metabolites in metabolomics. Among GLs, compared with the NC group, when treated with OPD, there is a significant difference in the change of GLs compared with OPD'. However, when OPD compared with OPD', changes of GLs was similar. In the three comparison groups, there was another pattern that makes people confused. Lysophosphatidylcholine (LPC), the main component that causes hemolysis, has been down-regulated when treated with OPD and OPD', this was not consistent with the observed hemolytic behavior.
Metabolic and Lipidomic pathway analysis
In order to further explore the hemolysis mechanism of OPD and OPD', the different metabolites in the plasma of rats in the OPD and OPD' group were substituted into the progqenesis QI data processing platform for metabolic path analysis, as shown in Figure. 6. The results showed that the top-20 metabolic pathways disturbed by OPD or OPD' were similar. Since the pathway of lipid metabolites is insufficient in the KEGG website, not all differential metabolites can be reflected in KEGG pathway among the screened differential metabolites, this is also a challenge for current pathway analysis. However, the trend of the Figure.6 is consistent with the results of the differential metabolites which were shown in Table 3, its predictions were still very informative.
Compared with NC group, choline metabolism in cancer and glycerophospholipid metabolism pathways were the most significantly enriched and largest number of differential metabolites gathered pathways in OPD and OPD' group, from a super class perspective, these two metabolic pathways were related to lipids and lipid-like molecules. As we all know, the main body of the cell membrane is the phospholipid bilayer and the fundamental cause of hemolysis of red blood cells is that the membrane was damaged. Therefore, we can speculate that the hemolysis caused by OPD and OPD' in vivo was related to the interference of glycerophospholipid metabolism. In addition， the pathways that were significantly affected also include pyrimidine metabolism, biosynthesis of amino acids, phenylalanine metabolism and bile secretion, these pathways not only have a certain relationship with the metabolism of glycerophospholipids, but may also participate in the metabolism and activation of OPD in vivo, which may result in the occurrence of OPD hemolysis.
Abnormal phospholipid metabolism may play an important role in hemolysis induced by OPD and OPD'
Published literature  shows that OPD can regulate the metabolism of glucose and lipids and improve the metabolic syndrome by interfering with the types and proportions of gut microbiota when taken orally. However, oral administration is very different from intravenous administration. Judging from the changes in the metabolic and lipid profile of different groups, the disorder of phospholipid metabolism dominates the causes of OPD and OPD' induced hemolysis [32, 33]. Statistics of all differential metabolites from two levels of super class and class were used to analyze the proportion of lipids and glycerophospholipids in the metabolic profile. In the NC vs OPD group, the ratio of super class level, i.e. lipids and lipid-like molecules, to all differential metabolites was 67/145, the ratio of class level, i.e. glycerophospholipids, to super class level’s metabolites was 42/67. In the other two groups, these two ratios were 60/124 (to all differential metabolites), 39/60 (to super class level’s metabolites) in NC vs OPD' group and 52/84 (to all differential metabolites), 10/52 (to super class level’s metabolites) in OPD vs OPD' group. Further analysis of sub-class metabolites at the class level, we found that at the sub-class level, the proportion of LysoPC was 18/42, 13/39 and 7/10 respectively. These ratios have significant statistical differences in comparison with other categories of differential metabolites in their corresponding levels.
Phospholipid metabolism disorder play an important role in hemolysis. The metabolome and lipidome profiles show that the main cause of hemolysis is related to direct attack on red blood cell membrane. The metabolism of glycerophospholipid to lysophospholipid requires two steps of catalysis by phospholipase A2 and phospholipase B2, and the resulting lysophospholipid contains a hydrophobic hydrocarbon chain and a polar phosphate group. It was a very strong surfactant and has strong ability to destroy cell membranes . On the other hand, in addition to direct attacks on cell membranes, a significant increase in red blood cell apoptosis by inducing RBCs oxidative stress is another cause of hemolysis induced by phospholipid metabolism disorders . Such as lipid peroxidation  and mitochondrial dysfunctions . Although in the omics data, we have clearly observed phospholipid metabolism disorders, but how OPD and OPD' exactly participated in this process remains to be further explored.