Synthesis and Anti-Hepatocarcinoma Effect of Amino Acid Derivatives of Pyxinol and Ocotillol

Aiming at seeking an effective anti-hepatocarcinoma drug with low toxicity, a total of 24 amino acid derivatives (20 new along with 4 known derivatives) of two active ocotillol-type sapogenins (pyxinol and ocotillol) were synthesized. Both in vitro and in vivo anti-hepatocarcinoma effects of derivatives were evaluated. At first, the HepG2 human cancer cell was employed to evaluate the anti-cancer activity. Most of the derivatives showed obvious enhanced activity compared with pyxinol or ocotillol. Among them, compound 2e displayed the most excellent activity with an IC50 value of 11.26 ± 0.43 µM. Next, H22 hepatoma-bearing mice were used to further evaluate the anti-liver cancer activity of compound 2e. It was revealed that the growth of H22 transplanted tumor was significantly inhibited when treated with compound 2e or compound 2e combined with cyclophosphamide (CTX) (p < 0.05, p < 0.01), and the inhibition rates of tumor growth were 35.32% and 55.30%, respectively. More importantly, compound 2e caused limited damage to liver and kidney in contrast with CTX causing significant toxicity. Finally, the latent mechanism of compound 2e was explored by serum and liver metabolomics based on ultra-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) technology. A total of 21 potential metabolites involved in 8 pathways were identified. These results suggest that compound 2e is a promising agent for anti-hepato-carcinoma, and that it also could be used in combination with CTX to increase efficiency and to reduce toxicity.


Introduction
Liver cancer, the fourth major lethal malignancy worldwide, is generally divided into HCC (hepatocellular carcinoma), ICC (intrahepatic cholangiocarcinoma) and MHC (mixed hepato-cellular and cholangiocellular carcinoma) based on the pathological classification [1]. HCC, with 5-year overall survival, is only 10%, yet accounts for nearly 75% to 85% of liver cancer cases [2,3]. Despite the enormous efforts such as surgery, chemotherapy, radiation therapy, immunotherapy, and monoclonal antibody therapy to combat liver cancer, the disease burden imposed by it continues to increase [4]. Numerous scientific researchers devoted themselves to the modification or transformation of lead compounds with the purpose of obtaining derivative products with good properties [5][6][7][8]. Natural products and their structurally modified derivatives have been widely used as anti-tumor drugs due to the low toxicity and the ability of reversing multidrug resistance [9][10][11][12].

Chemistry
A total of 24 amino acid derivatives of pyxinol and ocotillol were synthesized as shown in Figure 2. Compounds 1a-1f and 3a-3f were separately synthesized by pyxinol and ocotillol in the presence of EDC {1-ethyl-3-(3-dimethylaminopropyl) carbodiimide} and DMAP (4-dimethylamino-pyridine) mediated esterifications of N-Boc protected amino acids in dry CH2Cl2 or THF (tetra-hydrofuran) at room temperature. The tert-butyl carbonate groups of them were removed in a mixed solvent of dry CH2Cl2 and TFA (trifluoroacetic acid) at room temperature to obtain compounds 2a-2f and 4a-4f. Silica gel column chromatography was used for purification of all the derivatives from the reaction mixture, and the structures of all products were confirmed by 1 H-NMR, 13 C-NMR and HR-MS. Taking compound 2e as an example, its molecular formula was established as C35H59NO5 by HR-ESI-MS. Compared with pyxinol (C30H52O4, the lead compound), compound 2e has an increased part (C5H7NO, equivalent to the deshydroxyproline). The 1 H-NMR (CDCl3, 600 MHz) spectrum of compound 2e showed eight methyl singlet signals at δH 1.25, 1.24, 1.08, 0.96, 0.88, 0.86, 0.84, 0.81; also showed the hydrogen singlet signal at δH 3.96, which was attached to the α-carbon of proline. The 13 C-NMR spectrum of compound 2e showed 35 carbon signals in total. A carbonyl carbon signal was also shown at δC 173.27, indicating the presence of an ester bond. The chemical shifts of C-3 moved to low-field region compared with pyxinol, demonstrating proline was conjugated at C-3 of pyxinol. Based on the above data, compound 2e was elucidated as (20S, 24R)-epoxy-3β-O-(l-prolyl)-dammarane-12β, 25-diol. Compounds 2a, 2c, 2d and 2f were prepared according to the published literature [31]. The other 20 derivatives are the new compounds.

Chemistry
A total of 24 amino acid derivatives of pyxinol and ocotillol were synthesized as shown in Figure 2. Compounds 1a-1f and 3a-3f were separately synthesized by pyxinol and ocotillol in the presence of EDC {1-ethyl-3-(3-dimethylaminopropyl) carbodiimide} and DMAP (4-dimethylamino-pyridine) mediated esterifications of N-Boc protected amino acids in dry CH 2 Cl 2 or THF (tetra-hydrofuran) at room temperature. The tert-butyl carbonate groups of them were removed in a mixed solvent of dry CH 2 Cl 2 and TFA (trifluoroacetic acid) at room temperature to obtain compounds 2a-2f and 4a-4f. Silica gel column chromatography was used for purification of all the derivatives from the reaction mixture, and the structures of all products were confirmed by 1 H-NMR, 13 C-NMR and HR-MS. Taking compound 2e as an example, its molecular formula was established as C 35 H 59 NO 5 by HR-ESI-MS. Compared with pyxinol (C 30 H 52 O 4 , the lead compound), compound 2e has an increased part (C 5 H 7 NO, equivalent to the deshydroxyproline). The 1 H-NMR (CDCl 3 , 600 MHz) spectrum of compound 2e showed eight methyl singlet signals at δ H 1.25, 1.24, 1.08, 0.96, 0.88, 0.86, 0.84, 0.81; also showed the hydrogen singlet signal at δ H 3.96, which was attached to the α-carbon of proline. The 13 C-NMR spectrum of compound 2e showed 35 carbon signals in total. A carbonyl carbon signal was also shown at δ C 173.27, indicating the presence of an ester bond. The chemical shifts of C-3 moved to low-field region compared with pyxinol, demonstrating proline was conjugated at C-3 of pyxinol. Based on the above data, compound 2e was elucidated as (20S, 24R)-epoxy-3β-O-(l-prolyl)-dammarane-12β, 25-diol. Compounds 2a, 2c, 2d and 2f were prepared according to the published literature [31]. The other 20 derivatives are the new compounds.

Cytotoxic Activity Determination on HepG2 Cells
All compounds including lead compounds (pyxinol and ocotillol) and derivatives were evaluated for their cytotoxic activity against the HepG2 cell line by using the CCK-8 assay method with cyclophosphamide (CTX) as the positive control. The results revealed that most derivatives, such as 1a, 1c, 1e, 2a-2f, 3c, 3e and 4a-4f, could enhance the antitumor activity of lead compounds. In particular, compound 2e exhibited the highest inhibitory effect on HepG2 cells in vitro with IC50 similar to the positive drug (Table 1). Therefore, compound 2e, the most potent growth inhibitor of the HepG2 cells, was selected for further study.

Cytotoxic Activity Determination on HepG2 Cells
All compounds including lead compounds (pyxinol and ocotillol) and derivatives were evaluated for their cytotoxic activity against the HepG2 cell line by using the CCK-8 assay method with cyclophosphamide (CTX) as the positive control. The results revealed that most derivatives, such as 1a, 1c, 1e, 2a-2f, 3c, 3e and 4a-4f, could enhance the antitumor activity of lead compounds. In particular, compound 2e exhibited the highest inhibitory effect on HepG2 cells in vitro with IC 50 similar to the positive drug (Table 1). Therefore, compound 2e, the most potent growth inhibitor of the HepG2 cells, was selected for further study.  Table 2. Tumor volume is shown in Figure 3. There is no significant difference in body weights of mice between the groups.  Tumors of the model group reached 10-18 mm in diameter on day 14. While CTX, compound 2e and compound 2e + CTX treatment resulted in delayed tumor growth compared with model group. It was showed that compound 2e groups (50, 100 mg/kg) could significantly reduce the tumor weight and volume compared to the model group (p < 0.05, p < 0.01). Treatment with compound 2e could decrease the tumor weight and increase the tumor inhibition rate (TIR) in a dose-dependent manner. Both CTX group and compound 2e + CTX group could significantly reduce tumor weights and volumes compared to the model group (p < 0.01). Compound 2e + CTX group displayed the highest TIR (55.30%).
The data also showed that the liver and kidney indexes were significantly increased in the model group compared with the normal group (p < 0.01). While the indexes of compound 2e-treated mice were basically consistent with the model group. However, the compound 2e + CTX group, which displayed similar indexes to the normal group, could significantly decreased the liver (p < 0.01) and kidney (p < 0.05) indexes compared with model group.

Histopathological Examination
Histological analyses of tumors, liver and kidney using H&E staining are shown in Figure 4.
The tumor cells in the model group showed aggressive growth in different forms and sizes. Mitotic activity and different stained tumor cells were also detected. The necrosis in the CTX group was more obvious than in the compound 2e high-dose group, but had extensive necrosis in the compound 2e + CTX group.
The liver and kidney cells in the model group and CTX group had many vacuoles and edema. This phenomenon had been alleviated in compound 2e high-dose group and in compound 2e + CTX group.  The tumor cells in the model group showed aggressive growth in different forms and sizes. Mitotic activity and different stained tumor cells were also detected. The necrosis in the CTX group was more obvious than in the compound 2e high-dose group, but had extensive necrosis in the compound 2e + CTX group.
The liver and kidney cells in the model group and CTX group had many vacuoles and edema. This phenomenon had been alleviated in compound 2e high-dose group and in compound 2e + CTX group.

Effects of Compound 2e Treatment on Cytokine Levels in Mice
The effects of compound 2e treatment on the expression of IL-2 (Interleukin-2), TNFα (Tumor necrosis factor-α) and VEGF (Vascular endothelial growth factor) were listed in Table 3. The level of TNF-α and IL-2 were greatly reduced after CTX treatment compared with the normal group (p < 0.01), which indicated that the immune system was suppressed by CTX. The level of TNF-α was dose-dependently increased in compound 2e treatment groups compared to the model group, and the highest level was observed in high-dose compound 2e group. These data indicated that compound 2e might enhance immune function in H22 tumor-bearing mice by increasing cytokine levels. The levels of IL-2 in compound 2e groups (moderate and high-dose) were significantly elevated compared to model group (p < 0.01), and they were also dose-dependently increased. These results showed that the proliferation and activation of T lymphocytes might be stimulated to cause tumor cell death. Meanwhile, the levels of VEGF expression in compound 2e treatment groups were all significantly decreased in a dose-dependent manner compared with the model group. CTX and compound 2e (moderate and high dose) could significantly inhibit the VEGF level in serum (p < 0.01). It was indicated that compound 2e might inhibit tumor growth by effectively reducing tumor angiogenesis. The effects of compound 2e treatment on the expression of IL-2 (Interleukin-2), TNF-α (Tumor necrosis factor-α) and VEGF (Vascular endothelial growth factor) were listed in Table 3. The level of TNF-α and IL-2 were greatly reduced after CTX treatment compared with the normal group (p < 0.01), which indicated that the immune system was suppressed by CTX. The level of TNF-α was dose-dependently increased in compound 2e treatment groups compared to the model group, and the highest level was observed in high-dose compound 2e group. These data indicated that compound 2e might enhance immune function in H22 tumor-bearing mice by increasing cytokine levels. The levels of IL-2 in compound 2e groups (moderate and high-dose) were significantly elevated compared to model group (p < 0.01), and they were also dose-dependently increased. These results showed that the proliferation and activation of T lymphocytes might be stimulated to cause tumor cell death. Meanwhile, the levels of VEGF expression in compound 2e treatment groups were all significantly decreased in a dose-dependent manner compared with the model group. CTX and compound 2e (moderate and high dose) could significantly inhibit the VEGF level in serum (p < 0.01). It was indicated that compound 2e might inhibit tumor growth by effectively reducing tumor angiogenesis.

Effects of Compound 2e Treatment on Hepatic and Renal Function
The results listed in Table 4 indicated that the serum AST (Aspartate aminotransferase), ALT (Alanine aminotransferase), CRE (Creatinine) and BUN (Blood urea nitrogen) levels dramatically increased in the model group compared with the normal group. The increased levels could be significantly re-regulated with the treatment of compound 2e (p < 0.01) or CTX (p < 0.01). The high dosage of compound 2e showed the best effect. Moreover, the combination group showed the lower values than those in CTX group, which indicated that compound 2e could increase CTX's efficiency and reduce its toxicity. These results implied that compound 2e did less harm to the liver and kidney of tumor-bearing mouse. Table 3. Effects of compound 2e on cytokine levels (mean ± SD, n = 6) in H22 tumor-bearing mice.

Validation of UPLC-QTOF-MS
UPLC-QTOF-MS system was used to perform the metabolomic study. The validation test was first conducted to monitor the durability and stability of system. Serum and liver quality control (QC) samples were run randomly covering the whole analysis process, respectively. The tests included: (1) five consecutive serum or liver QC samples were detected to evaluate the injection precision; (2) five parallel replicates of a serum or liver test sample were assayed to assess the reproducibility; (3) the post-preparation stability of the sample was estimated by detecting one serum or liver test sample that was placed in the autosampler at 10 • C for 0, 2, 8, 12, and 24 h. In the above test, the exact mass/retention time pairs of eight ions in serum or in liver QC samples both in positive and negative modes of ESI (electrospray ionization) were all monitored. The relative standard deviations (RSDs) of peak intensity or retention time were calculated and were listed in Table 5. The validation results indicated the chromatographic and spectrometric system were well-suited for the following metabolomics analysis. The serum and liver metabolic characteristics of normal group, model group and compound 2e (100 mg/kg) group in positive ESI (ESI+) and negative ESI (ESI-) modes were acquired. The representative BPI (base peak intensity) chromatograms are shown in Figure 5.  Model group-serum-ESI-Normal group-serum-ESI-

Identification of the Differential Metabolites and Metabolic Pathways
In order to confirm whether the endogenous metabolites were different in serum or in liver between normal, model and compound 2e groups, principal component analysis (PCA), an unsupervised pattern recognition approach, was firstly performed in both ESI+ and ESI-modes ( Figure 6A-D). Both in serum and in liver PCA scores, each spot represented a sample; the tightly clustered QC spots indicated the satisfactory stability of system; three groups were separated indicating their being differential; compound 2e group was located between the normal group and the model group, indicating that compound 2e might regulate the metabolic disturbances in H22 tumor-bearing mice. ESI-modes ( Figure 6E-H). In the OPLS-DA score plots, each spot also represented a serum or a liver sample. Either in serum samples or in liver samples, the model group and compound 2e group were both separated with satisfactory R 2 and Q 2 parameters. Permutation test was then performed to validate the prediction ability and the reliability of OPLS-DA. Figure 7A-D showed the test results. All Q 2 -values (blue spots) to the left were lower than the original points to the right, indicating that the OPLS-DA displayed good prediction ability and reliability. To find the potential biomarkers that contributed to the differentiation between the model group and compound 2e group, S-plots ( Figure 7E-H) under OPLS-DA were generated to visualize the variables. Each spot in S-plots represents an endogenous metabolite in the model group and compound 2e group. The farther away the spot in the S-plots from the origin, the more significantly the metabolite contribute to the clustering of two groups. The metabolites with VIP (variable importance in the projection) >1.0 and p < 0.05 were considered as potential biomarkers. There were 21 robust endogenous metabolites were identified as the candidate biomarkers (listed in Table 6). These biomarkers were marked with measured mass in S-plots. And it was showed that these biomarkers were involved in eight metabolic pathways ( Table 7). The relationship between the main metabolites & its involved metabolisms and liver diseases is discussed as follows: Arachidonic acid metabolism (AM): (1) Arachidonic acid (AA), recognized as the main factor mediating repeated liver injury and compensatory proliferation [62,63], could suppress the growth of hepatic cells and increase cellular transglutaminase 2 transamidase   Aiming at obtaining the maximum separation between the model group and compound 2e group, orthogonal projections to latent structures discriminant analysis (OPLS-DA), a supervised method of pattern recognition, was then established in both ESI+ and ESI-modes ( Figure 6E-H). In the OPLS-DA score plots, each spot also represented a serum or a liver sample. Either in serum samples or in liver samples, the model group and compound 2e group were both separated with satisfactory R 2 and Q 2 parameters.
Permutation test was then performed to validate the prediction ability and the reliability of OPLS-DA. Figure 7A-D showed the test results. All Q 2 -values (blue spots) to the left were lower than the original points to the right, indicating that the OPLS-DA displayed good prediction ability and reliability. ESI-modes ( Figure 6E-H). In the OPLS-DA score plots, each spot also represented a serum or a liver sample. Either in serum samples or in liver samples, the model group and compound 2e group were both separated with satisfactory R 2 and Q 2 parameters. Permutation test was then performed to validate the prediction ability and the reliability of OPLS-DA. Figure 7A-D showed the test results. All Q 2 -values (blue spots) to the left were lower than the original points to the right, indicating that the OPLS-DA displayed good prediction ability and reliability. To find the potential biomarkers that contributed to the differentiation between the model group and compound 2e group, S-plots ( Figure 7E-H) under OPLS-DA were generated to visualize the variables. Each spot in S-plots represents an endogenous metabolite in the model group and compound 2e group. The farther away the spot in the S-plots from the origin, the more significantly the metabolite contribute to the clustering of two groups. The metabolites with VIP (variable importance in the projection) >1.0 and p < 0.05 were considered as potential biomarkers. There were 21 robust endogenous metabolites were identified as the candidate biomarkers (listed in Table 6). These biomarkers were marked with measured mass in S-plots. And it was showed that these biomarkers were involved in eight metabolic pathways ( Table 7). The relationship between the main metabolites & its involved metabolisms and liver diseases is discussed as follows: Arachidonic acid metabolism (AM): (1) Arachidonic acid (AA), recognized as the main factor mediating repeated liver injury and compensatory proliferation [62,63], could suppress the growth of hepatic cells and increase cellular transglutaminase 2 transamidase  To find the potential biomarkers that contributed to the differentiation between the model group and compound 2e group, S-plots ( Figure 7E-H) under OPLS-DA were generated to visualize the variables. Each spot in S-plots represents an endogenous metabolite in the model group and compound 2e group. The farther away the spot in the S-plots from the origin, the more significantly the metabolite contribute to the clustering of two groups. The metabolites with VIP (variable importance in the projection) >1.0 and p < 0.05 were considered as potential biomarkers. There were 21 robust endogenous metabolites were identified as the candidate biomarkers (listed in Table 6). These biomarkers were marked with measured mass in S-plots. And it was showed that these biomarkers were involved in eight metabolic pathways ( Table 7). The relationship between the main metabolites & its involved metabolisms and liver diseases is discussed as follows: Arachidonic acid metabolism (AM): (1) Arachidonic acid (AA), recognized as the main factor mediating repeated liver injury and compensatory proliferation [62,63], could suppress the growth of hepatic cells and increase cellular transglutaminase 2 transamidase activity. The elevated serum levels of AA had been observed in HCC Patients. (2) 8,9-epoxyeicosatrienoic acid (8,9-EET), a kind of epoxyeicosatrienoic acids (EETs), was closely related to hepatic tumor [64]. (3) 15(S)-hydroperoxy eicosatetraenoic acid (15(S)-HPETE) decreased production of CD31 and VEGF in endothelial cells and had an anti-angiogenic effect in adipose tissue [65,66]. (4) Leukotriene B4 (LTB4) played a role in acute and chronic liver injury. It is important in mediating the inflammatory response, and it is involved in pathogenesis of several autoimmune, inflammatory diseases as well as in tumor proliferation [67]. (5) Prostaglandin I2 (PGI2), an important downstream from AA, is also a mediator of tumor progression [68]. Exogenous PGI2 could decrease tumor cell proliferation. In this experiment, the elevated levels (AA, 8, 9-EET, 5-HPETE, LTB4) and decreased levels (15(S)-HPETE, PGI2) in the model group indicated the imbalance of arachidonic acid metabolism. While the above levels could be re-regulated by compound 2e treatment.
Retinol metabolism (RM): In liver tumors, retinyl ester levels were significantly decreased [73]. Hydroxyretinoic acid and all-trans-5, 6-Epoxyretinoic acid could inhibit the growth of breast cancer cell and rat rhabdomyosarcoma cell [74]. All-trans-retinoic acid, the active metabolite of vitamin A, could affect liver cancer stem cells. All the above metabolites decreased in the model group and could be up-regulated following compound 2e treatment.
Porphyrin and chlorophyll metabolism (PCM): The level of bilirubin was significantly elevated in the H22 liver tumor model group [75]. In our study, bilirubin was also markedly decreased in the model group, and was re-regulated following compound 2e treatment.
Tryptophan metabolism (TryM): L-tryptophan, an essential amino acid for human, could prevent hepatic fibrosis progression [76]. It was shown that L-tryptophan decreased in the model group and could be re-regulated by compound 2e.
Glycerophospholipid metabolism (GlyM): Altered glycerophospholipid metabolism (Lyso-PC a C18:1) has been reported in both blood and liver tissue samples from nonalcoholic fatty liver disease patients [77]. It had also been disturbed in the liver injury mouse model [78]. In our study, LysoPC (18:1(9Z)/0:0) increased in the model group and could be re-regulated by compound 2e.
α-Linolenic acid metabolism (ALAM): Alterations in phospholipid and fatty acid metabo lism may play key roles in hepatocarcinogenesis. It was reported that the tumor exhibited significantly lower level of α-linolenic acid than peritumoral liver tissue [79]. Stearidonic acid inhibited intestinal tumor development and cancer cell proliferation, and could be potentially chemo-preventive [80]. All above metabolites, decreased in the model group and could be re-regulated by compound 2e.
In addition, the predictive ROC curves, generated by using 21 candidate biomarkers, showed that the candidate biomarkers were potential diagnostic markers for liver tumor ( Figure 8A) and were contributed to compound 2e treatment ( Figure 8B). The area under curve (AUC) values and p-values of the biomarkers in ROC curves were listed in Table 8.      In order to characterize and visualize the biomarkers' relative abundance in three groups, the heatmap ( Figure 9A) was then generated with green color representing low abundance and red color representing high abundance. Moreover, the established metabolic network of the biomarkers was shown in Figure 9B.

The Synthesis of Compounds 1a-1f and 3a-3f
At room temperature, EDC (3.15 mmol) and DMAP (1.05 mmol) in dry CH 2 Cl 2 (20 mL) were added to N-Boc protected amino acids (4.20 mmol) solution in a triangle flask with a magnetic stirrer. Pyxinol (2.10 mmol) was added 30 min later. The reaction mixtures were shaken for 12 h and progress was monitored by TLC (n-hexane/ethyl acetate 3:1). Then the organic solution was washed with saturated aqueous NaHCO 3 solution (3 × 10 mL), water (3 × 10 mL) and brine (3 × 10 mL), dried (MgSO 4 ), filtered and concentrated under reduced pressure to give the crude product. The crude products were chromatographed using silica gel and eluted with n-hexane and ethyl acetate (5:1) to obtain the pure products 1a-1f used as substrates for the Boc deprotection reaction. The ocotillol N-Boc protected amino acids derivatives 3a-3f using the same procedure described above for compounds 1a-1f, except all reactants were dissolved in THF.

Tumor-Bearing Mice Model and Drug Administration
H22 cells, resuspended in normal saline with final concentration of 5 × 10 6 cells/mL, were injected into the abdominal cavity of mice under aseptic conditions. After 1 week, ascites tumor cells were extracted and dispersed with physiological saline to a concentration of 1 × 10 7 cells/mL.
The diluted ascites tumor cell suspension was then injected (0.2 mL, 2 × 10 6 cells/mouse) subcutaneously into the right forelimb armpit of mice at day 1 to establish the tumorbearing mice model. The mice did not injected the tumor cells were used as the normal control.

Antitumor Activity Evaluation
Before administration every day, the tumor size and body weight were measured. When the tumor was larger than 20 mm in diameter, animals were euthanized according to the IACUC proposals.
24 h after the last administration, the whole blood was collected from the orbit. Serum was obtained from whole blood by centrifugation (3500 rpm, 15 min, 4 • C) and stored at −20 • C. The levels of IL-2, TNF-α, VEGF, AST, ALT, BUN and CRE in serum were determined in accordance with the procedures described in commercial kit instructions.
Mice were then killed by cervical dislocation after blood collected. The tumor, liver and kidney were rapidly separated, weighted and dissected for histopathological examination.
The tumor inhibition rate (TIR) was calculated as follows: TIR (%) = (W m − W t )/W m × 100% (W m : tumor weight of model group, W t : tumor weight of treatment group).
The liver and kidney indexes were calculated as follows: organ index (mg/g) =average weight of organ/average body weight.
The tissue of tumor, liver and kidney from each mouse was quickly fixed in 10% neutral-buffered formalin for histopathology. The H&E staining of H22 solid tumor, liver and kidney tissue were embedded in paraffin and sectioned into sections with a thickness of 5 µm which were conducted following the manufacturer's instructions in respective kit. Tumor, liver and kidney tissue sections were observed by light microscopy (Leica DM750, Solms, Germany) and recorded by photography (×400).

Statistical Analysis
The data were expressed as mean ± SD. All statistical analyses were performed using the SPSS 16.0 software (SPSS, USA). For statistical comparison of values, a Student's t-test was used, and p < 0.05 was considered statistically significant. One-way ANOVA, followed by t-test, was used to compare differences among groups.

UPLC-QTOF-MS Conditions
Waters UPLC system with electrospray ionization (ESI) interface combined with Xevo G2-XS QTOF mass spectrometer was applied for determination and analysis.
MS conditions were as follows: desolvation and source temperatures being and 150 • C, respectively; cone and desolvation gas flows being 50 L/h and 800 L/h, respectively; cone voltage setting at 40 V; capillary voltage being at 2.2 kV (ESI-) and 2.6 kV (ESI+); MS E centroid mode with low energy of 6 V and high energy of 20~40 V; sodium formate being used to calibrate the mass spectrometer in the range of 50 to 1200 Da; external reference leucine enkephalin (m/z 556.2771 and 554.2615 in ESI+ and ESI-modes) injected at a flow of 10 µL/min. QC sample was injected randomly throughout the whole worklist for 5 times. The volume injections of QC and test samples were all 5 µL per run. MassLynx V4.1 workstation (Waters, Manchester, UK) was used to record data.

Metabolomics Study
Firstly, MarkerLynx XS software (Version 4.1, Waters Co., Milford, MA, USA) was applied to analyze data with the optimized parameters (mass range 50~1200 Da, mass tolerance 0.10 with window 0.10, retention time 2~28 min with window 0.20, noise elimination level 6 and marker intensity threshold 2000 counts). Then the results of exact mass/retention time pairs and the corresponding intensities were shown in Extended Statistics (XS) Viewer.
Secondly, SIMCA-P software (Version 14.1, Umetric, Umea, Sweden) was used to perform the multivariate analysis including principal component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA) [34]. S-plots were then created to explore the remarkable potential biomarkers with VIP (variable importance in the projection) value above 1.0 and p-value below 0.05. In addition, the permutation test with R 2 /Q 2 values indicating statistical significance was performed to provide a reference distribution. The predictive ROC (receiver operating characteristic) curves were generated using the metabolites with AUC (area under curve) >0. 8  Finally, MetaboAnalyst 4.0 software was used to filter out the most vital potential metabolic pathways (impact-value threshold above 0.10) by analyzing the confirmed distinct metabolites.

Conclusions
In this paper, a total of 24 amino acid derivatives, including 20 new along with 4 known compounds of pyxinol and ocotillol were synthesized and evaluated in vitro and in vivo for the anti-hepatocarcinoma effect. Most of the amino acid derivatives showed obvious enhanced activity compared with pyxinol or ocotillol. Compound 2e displayed excellent activity in HepG2 human cancer cell and in H22 tumor-bearing mice. It was also revealed that compound 2e combined with cyclophosphamide (CTX) had the best antitumor activity and the lowest toxicity in mice. A total of 21 potential metabolites involved in 8 metabolic pathways were identified in anti-hepatocarcinoma effect of compound 2e. These results suggest that compound 2e is a promising agent for anti-hepatocarcinoma with low toxicity, and that it also could increase CTX's efficiency and reduce its toxicity.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.