Momordicine-I, a Bitter Melon Bioactive Metabolite, Displays Anti-Tumor Activity in Head and Neck Cancer Involving c-Met and Downstream Signaling

Simple Summary The incidence of head and neck cancer (HNC), one of the most aggressive cancers, is increasing rapidly globally. Conventional and targeted therapies show limited success with several undesirable side effects. Thus, there is a critical clinical need to identify additional alternative therapeutic strategies for successfully managing the disease. Preclinical and clinical studies indicate the crucial roles of dietary phytochemicals to manage different cancers. We and others previously showed the potential anticancer effect of bitter melon extract (BME) to prevent various cancers, including HNC. In this study, we identified momordicine-I (M-I) as a bioactive component in the BME. Subsequent mechanistic study showed that M-I inhibited HNC cell (JHU022, JHU029, Cal27) proliferation involving c-Met and downstream signaling. In pre-clinical mouse models, M-I showed similar effectiveness to prevent HNC tumor growth in mice with no apparent toxic side effect, suggesting an additional option for HNC therapy. Abstract Head and neck cancer (HNC) is one of the most aggressive cancers, and treatments are quite challenging due to the difficulty in early diagnosis, lack of effective chemotherapeutic drugs, adverse side effects and therapy resistance. We identified momordicine-I (M-I), a bioactive secondary metabolite in bitter melon (Momordica charantia), by performing liquid chromatography-high resolution electrospray ionization mass spectrometry (LC-HRESIMS) analysis. M-I inhibited human HNC cell (JHU022, JHU029, Cal27) viability in a dose-dependent manner without an apparent toxic effect on normal oral keratinocytes. Mechanistic studies showed that M-I inhibited c-Met and its downstream signaling molecules c-Myc, survivin, and cyclin D1 through the inactivation of STAT3 in HNC cells. We further observed that M-I was non-toxic and stable in mouse (male C57Bl/6) blood, and a favorable pharmacokinetics profile was observed after IP administration. M-I treatment reduced HNC xenograft tumor growth in nude mice and inhibited c-Met and downstream signaling. Thus, M-I has potential therapeutic implications against HNC.


Introduction
Head and neck cancer (HNC) arises from the mucosal surfaces of the oral cavity, oropharynx, larynx, paranasal sinuses, nasal cavity, and salivary glands, and is considered the sixth most common cancer worldwide [1]. Major risk factors for HNC are habitual

Identification of Active Ingredients in Bitter Melon Extract (BME)
We and others have reported the potential anticancer effect of BME in several cancer models [11][12][13][14][15][16]. The biological activity of BME depends on its chemical constituents. Among several ingredients in bitter melon (like phenolic acids, flavonoids, essential oils, fatty acids, amino acids, lectins, sterols, saponin and proteins), cucurbitane type triterpenoids and cucurbitane type triterpene glycosides are a major chemical class in the family Cucurbitaceae and are suggested to be responsible for bitterness and much of the biological activities in the family [16]. We performed liquid chromatography coupled to high resolution electrospray ionization mass spectrometry (LC-HRESIMS) analysis to identify the cucurbitane type triterpenoids and triterpene glycosides present in the BME. The LC-HRESIMS data revealed the presence of a total of 28 secondary metabolites, of which 4 metabolites belonged to cucurbitane type triterpenoids and 20 belonged to cucurbitane triterpene glycosides in the extract (Table 1, Figure 1A). The BME also contained a cucurbitane triterpenoid at a retention time 21.64 min with a molecular formula of C 33 H 48 O 7 that was not reported previously. The molecular formula and MS/MS analysis indicated this compound to be a new derivative of 7,23-dihydroxy-3-O-malonylcucurbita-5,24-dien-19-al with an extra double bond. Additionally, the HRMS analysis indicated the presence of three monoterpenoid glycosides and one oleanane-type triterpene saponin. It worth noting that many of those metabolites have unknown biological functions, otherwise reported in Table 1. involving c-Met and downstream signaling in both in vitro and in vivo models. This is the first study describing the therapeutic potential of M-I for the regression of HNC tumors in a pre-clinical model.

Identification of Active Ingredients in Bitter Melon Extract (BME)
We and others have reported the potential anticancer effect of BME in several cancer models [11][12][13][14][15][16]. The biological activity of BME depends on its chemical constituents. Among several ingredients in bitter melon (like phenolic acids, flavonoids, essential oils, fatty acids, amino acids, lectins, sterols, saponin and proteins), cucurbitane type triterpenoids and cucurbitane type triterpene glycosides are a major chemical class in the family Cucurbitaceae and are suggested to be responsible for bitterness and much of the biological activities in the family [16]. We performed liquid chromatography coupled to high resolution electrospray ionization mass spectrometry (LC-HRESIMS) analysis to identify the cucurbitane type triterpenoids and triterpene glycosides present in the BME. The LC-HRESIMS data revealed the presence of a total of 28 secondary metabolites, of which 4 metabolites belonged to cucurbitane type triterpenoids and 20 belonged to cucurbitane triterpene glycosides in the extract (Table 1, Figure 1A). The BME also contained a cucurbitane triterpenoid at a retention time 21.64 min with a molecular formula of C33H48O7 that was not reported previously. The molecular formula and MS/MS analysis indicated this compound to be a new derivative of 7,23-dihydroxy-3-O-malonylcucurbita-5,24-dien-19al with an extra double bond. Additionally, the HRMS analysis indicated the presence of three monoterpenoid glycosides and one oleanane-type triterpene saponin. It worth noting that many of those metabolites have unknown biological functions, otherwise reported in Table 1.

M-I Inhibits c-Met Signaling in HNC Cells
We observed that BME inhibited c-Met signaling and its downstream signaling molecules to prevent HNC growth [17]. To investigate the mechanistic effect of M-1, Cal27, JHU029 and JHU022 cells were treated with M-I for 48 hr along with vehicle control. BME was used in parallel as a control. A significant reduction of c-Met expression was observed in these cells following BME or M-I treatment ( Figure 2A). The receptor tyrosine kinase c-Met induces tumor development by activating multiple downstream molecules, including the oncogenic transcription factor signal transducer and activator of transcription 3 (STAT3) [22,23]. Upon activation of c-Met signaling, STAT3 is activated by phosphorylation at the Tyr-705 residue, dimerized and translocated to the nucleus for activation of several proliferation and survival related genes including c-Myc, survivin and cyclin D1 [23,24]. Activated STAT3 can also transcribe STAT3 gene as a positive feedback mechanism [25]. We observed a significant reduction in phospho-STAT3 (Tyr-705) expression following treatment with BME or M-I in Cal27 and JHU029 cells ( Figure 2B). A substantial reduction in total STAT3 level was also noted in M-I treated cells. Next, we examined whether the short term (15, 30, 60, 120, 240 min) BME or M-I treatment can inhibit p-STAT3 levels to ascertain whether initially M-1 inhibits p-STAT3 levels without affecting total STAT protein levels. Interestingly, we have observed that BME or M-I treatment on HNC cells inhibits pSTAT3 in 15 min without affecting total STAT3, however, after 240 min of treatment, total STAT3 was inhibited ( Figure 2C). This may be due to the feedback mechanism of pSTAT3 on STAT3 regulation as discussed earlier [25]. Further, depletion of c-Met by specific siRNA inhibits pSTAT3 ( Figure 2D). We also observed a significant inhibition of STAT3 downstream molecules; c-Myc, survivin and cyclin D1, in Cal27 and JHU029 cells following treatment with BME and M-I ( Figure 3). Taken together, our results suggested that M1 inhibits c-Met signaling in the prevention of HNC growth ( Figure 4).
Met by specific siRNA inhibits pSTAT3 ( Figure 2D). We also observed a significant inhibition of STAT3 downstream molecules; c-Myc, survivin and cyclin D1, in Cal27 and JHU029 cells following treatment with BME and M-I ( Figure 3). Taken together, our results suggested that M1 inhibits c-Met signaling in the prevention of HNC growth ( Figure  4).   Met by specific siRNA inhibits pSTAT3 ( Figure 2D). We also observed a significant inhibition of STAT3 downstream molecules; c-Myc, survivin and cyclin D1, in Cal27 and JHU029 cells following treatment with BME and M-I ( Figure 3). Taken together, our results suggested that M1 inhibits c-Met signaling in the prevention of HNC growth ( Figure  4).

Pharmacokinetic and Toxicity Profile of M-I
To evaluate the pharmacokinetic (PK) profile of M-I, a single dose (20 mg/kg) was given to C57Bl/6 male mice by either intraperitoneal injection (IP group; n = 3) or oral gavage (oral group; n = 3). A dose of 20 mg/kg was chosen based on previous publications [26,27]. The average plasma concentration ± standard deviation at each time point is shown in Figure 5A and corresponding PK parameters are summarized in Table 2. M-I was rapidly absorbed with a maximum plasma concentration 1 h post-IP and PO dose. Cmax values were 18 µM and 0.5 µM after the single 20 mg/kg IP and PO dose, respectively. The observed elimination half-life was 0.9 h in the IP group and 2 hr in the PO group. The oral group had loose stool, starting four hours post-dose, which had not resolved by 8 h but all mice appeared normal after twenty-four hours. No adverse events were observed for the mice in the IP dosing group. We therefore selected IP administration of M-I for subsequence studies.

Pharmacokinetic and Toxicity Profile of M-I
To evaluate the pharmacokinetic (PK) profile of M-I, a single dose (20 mg/kg) was given to C57Bl/6 male mice by either intraperitoneal injection (IP group; n = 3) or oral ga-vage (oral group; n = 3). A dose of 20 mg/kg was chosen based on previous publications [26,27]. The average plasma concentration ± standard deviation at each time point is shown in Figure 5A and corresponding PK parameters are summarized in Table 2. M-I was rapidly absorbed with a maximum plasma concentration 1 h post-IP and PO dose. Cmax values were 18 µM and 0.5 µM after the single 20 mg/kg IP and PO dose, respec-tively. The observed elimination half-life was 0.9 h in the IP group and 2 h in the PO group. The oral group had loose stool, starting four hours post-dose, which had not re-solved by 8 h but all mice appeared normal after twenty-four hours. No adverse events were observed for the mice in the IP dosing group. We therefore selected IP administration of M-I for subsequence studies. Table 2. Pharmacokinetics parameters of momordicine I (M-I) administered either through intraperitoneally (IP) or orally (PO) in C57Bl/6 male mice (data provided as mean ± SD). 0.33 ± 0.08 6.84 ± 1.5

Pharmacokinetic Parameters
Abbreviations: T 1 2 : Elimination half-life; Tmax: time to reach maximum (peak) plasma concentration following drug administration; Cmax: maximum (peak) plasma drug concentration; AUClast: area under the plasma concentration-time curve from time zero to time of last measurable concentration; AUCINF_obs: area under the concentration-time curve extrapolated from zero up to infinity; AUC (%Extrap): area under the first moment of the plasma concentration-time curve extrapolated from time t to infinity as a percentage of total AUC; Cl_obs: apparent total body clearance of the drug from plasma. tively. The observed elimination half-life was 0.9 h in the IP group and 2 hr in the group. The oral group had loose stool, starting four hours post-dose, which had not solved by 8 h but all mice appeared normal after twenty-four hours. No adverse eve were observed for the mice in the IP dosing group. We therefore selected IP adminis tion of M-I for subsequence studies.  For toxicity analysis, mice (n = 3) were given 20 mg/ kg of M-I twice a day or 30 mg/kg IP dose of M-I once a day by IP injection for five days. In parallel, there were untreated control (n = 3) and BME treated group (n = 3) for comparison. Behavior and body weight of mice were monitored daily. The body weight in all the mice was stable, with no drastic changes seen in any of the mice tested. We collected blood on day six, and serum chemistries related to liver and kidney function were examined. The concentration of total bilirubin, alanine transaminase, aspartate transaminase, alkaline phosphatase, creatinine, urea, and glucose were comparable in M-I and BME treated groups with untreated control mice, indicating no toxic effect ( Figure 5B).

Therapeutic Potential of M-I in HNC Xenograft Model
We investigated the therapeutic efficacy of M-I in HNC xenograft model. For this, we implanted JHU029 cells in flanks of nude mice. After the formation of a palpable tumor, mice were divided into three groups: untreated control (n = 5), BME treated group (n = 5), and M-I treated group (n = 5). In the BME group, mice were given 30% BME through drinking water as described previously [20]. The mice in M-I group received 30 mg/kg of M-I through IP, once a day till the end of the experiment. Body weight and tumor volume were measured. We observed a sudden drop in body weight (~10%) in three out of five mice in M-I group on Day 20; however, they gained back the weight ( Figure 6A). We also observed the formation of fluid in the tumors as reported previously [17,28]. Some of the tumors in the control group had an open wound; therefore we needed to sacrifice all the animals on day 32. The BME and M-I group mice displayed significantly reduced tumor volume ( Figure 6B,C). However, M-I showed a better effect in reducing tumor growth. We examined c-Met signaling in control and treated tumors. We observed a significant reduction in expression of c-Met and its downstream molecule c-Myc in BME and M-I treated groups compared to untreated control tumors ( Figure 6D). For further validation, we also examined the effect of M-I in Cal27 xenograft model. We observed increasing body weight in all the mice throughout the experiment, and ~50% reduction of tumor growth in BME or M-I treated xenograft tumor ( Figure 7A,B). We further observed a significant reduction in c-Met and c-Myc expression in BME and M-I treated groups compared to the control ( Figure 7C,D). Our results demonstrated that M-I has the potential as a therapeutic candidate for HNC treatment. For further validation, we also examined the effect of M-I in Cal27 xenograft model. We observed increasing body weight in all the mice throughout the experiment, and~50% reduction of tumor growth in BME or M-I treated xenograft tumor ( Figure 7A,B). We further observed a significant reduction in c-Met and c-Myc expression in BME and M-I treated groups compared to the control ( Figure 7C,D). Our results demonstrated that M-I has the potential as a therapeutic candidate for HNC treatment.
We observed increasing body weight in all the mice throughout the experiment, and ~ reduction of tumor growth in BME or M-I treated xenograft tumor ( Figure 7A,B). We ther observed a significant reduction in c-Met and c-Myc expression in BME and treated groups compared to the control ( Figure 7C,D). Our results demonstrated that has the potential as a therapeutic candidate for HNC treatment.

Discussion
In this study, we identified M-I as a bioactive secondary metabolite using LC-HRESIMS analysis of the BME. We observed that M-I inhibits c-Met signaling in HNC cell (JHU022, JHU029, Cal27) lines. We further demonstrated a significant regression of tumor growth in HNC xenograft models following daily administration of M-I with no toxicity, suggesting the therapeutic efficacy of BME and M-I by targeting c-Met signaling.
Bitter melon plant and its fruit contain many phytochemicals, and among those cucurbitane type triterpenoids and cucurbitane type triterpene glycosides are the major chemical constituents [16]. The cucurbitane-type triterpenoids and cucurbitane-type triterpene glycosides are suggested to be responsible for bitterness and confer much of the biological activities of the plant [16,29]. M-I belongs to the class cucurbitane-type triterpene. This secondary metabolite was first identified and characterized in leaves and vines of Momordica charantia L. [30]. In our study, we identified M-I in the water extract of the fruit. Although the compound was identified in 1984, the biological function of M-I was not well evaluated. M-I was recently reported to have inhibitory effects on high-glucose-induced cell proliferation and collagen synthesis in rat cardiac fibroblasts [31] and stimulate insulin secretion in vitro [32], but to the best of our knowledge, the anticancer effect of M-I was not reported. We initially screened a few metabolites from bitter melon and got a significant inhibition with M-I with IC 50 of less than 8 µg/mL in Cal27 and JHU029 cells. We observed that M-I inhibited HNC growth and c-Met signaling. Aberrant activation of c-Met signaling through overexpression of c-Met and its downstream molecules c-Myc, cyclin D1, and survivin were observed predominantly in HNC [3][4][5], which were reduced following M-I treatment. Increased c-Met signaling is associated with HNC progression and metastasis, and c-Met signalling inhibition by neutralizing antibody inhibited tumor growth and its metastatic potential [3,5].M-I may have other targets to inhibit cell proliferation which was not investigated in this study. We observed a significant reduction of tumor growth in two HNC xenograft models following BME treatment with no toxicity. Further, treatment of M-I (30 mg/kg/mouse) once a day worked similarly to twice a day in our pilot experiment (20 mg/kg/mouse).
HNC therapy often has limited success. Resistance to approved anti-EGFR therapy sometimes makes the treatment management difficult [3,5]. Thus, targeting c-Met is suggested to be a promising alternative strategy. Many drugs that target c-Met signaling have been developed, showing promising results in preclinical and clinical studies. However, these drugs manifested several adverse side effects [3,5,33]. c-Met inhibitor GEN-203 and compound 8 showed liver and bone marrow toxicity in mice and myocardial degeneration in rats. Foretinib caused fatigue, hypertension, and gastrointestinal toxicities. Golvatinib caused supraventricular tachycardia, convulsion, and pulmonary embolism. Tivantinib (ARQ197) showed adverse events, including leukopenia, anemia, and neutropenia in clinical trial. Thus, identifying a new natural and non-toxic c-Met signaling inhibitor would have a high impact on HNC treatment, and M-I might be one of the major contributors with biological activity in BME. However, the functional mechanism of M-I mediated c-Met signaling inhibition remains to be further elucidated. Effect on M-I in the presence of an intact immune system would need further investigation in future. Bitter melon extract (BME) was prepared from the Chinese variety of young bitter melons (raw and green) as described previously [17]. Briefly, BME was prepared by aqueous extraction from whole fruit without seeds using a household juicer at room temperature with subsequent centrifugation at 15000× g at 4 • C for 30 min. BME was stored at −80 • C for further analysis. Cal27 and JHU029 cells were treated with 2% BME and JHU022 cells were treated with 3% BME as described previously [17,21]. The momordicine-I (>98% pure) was purchased from Chemfaces (Cat. No.: CFN92076; Hubei, China). The powder was dissolved in DMSO and added to the cells at different concentrations. Based on cytotoxicity data, Cal27 and JHU029 cells were treated with 10 µg/mL and JHU022 cells were treated with 20 µg/mL dose of M-I and incubated for 48 hr for further experiments. All the experiments were done at least in triplicate.

Cytotoxicity Assay
Cal27, JHU029, JHU022 and NOK were seeded in 96 well-plate (5000 cells/well) and cells were treated with different concentration of M-I for 48 h. There were untreated control and DMSO treated vehicle control group for comparison. Cytotoxicity assay was performed using Cell Counting Kit-8 (Dojindo Molecular Technology, Rockville, MD, USA) according to manufacturer instruction.

Liquid Chromatography High Resolution Electrospray Ionization Mass Spectrometry (LC-HRESIMS)
HRESIMS analysis of the BME was done using a LTQ Orbitrap spectrometer coupled to an HPLC system (PDA detector, PDA autosampler, and pump, ThermoFisher Scientific, Inchinnan, Renfrew PA4 9R, UK). The following conditions were used: capillary voltage of 45 V, capillary temperature of 260 • C, auxiliary gas flow rate of 10−20 arbitrary units, sheath gas flow rate of 40−50 arbitrary units, spray voltage of 4.5 kV, and mass range of 100−2000 amu (maximal resolution of 30,000). For LC-HRESIMS, a Sunfire C 18 analytical HPLC column (5 µm, 4.6 mm × 150 mm) was used with a mobile phase of 0 to 100% MeOH over 20 min followed by 100% MeOH over 5 min at a flow rate of 1 mL min −1 .

Pharmacokinetic and Toxicity Study
The pharmacokinetic (PK) study of M-I was performed in male C57Bl/6 mice. A sin-gle dose of M-I (formulated at 2 mg/mL dissolved in 5% DMSO/95% of a 30% w:v Captisol solution) was administered either by intraperitoneal injection (IP group, n = 3) or oral ga-vage (PO group, n = 3). After dosing, 20 µL blood samples were collected into heparin pre-coated tubes at 5 min, 15 min, 30 min, 1 h, 2 h, 4 h, 6 h, 8 h and 24 h. Samples were centrifuged, and the plasma collected. M-I concentration was determined using liq-uid chromatography-mass spectrometry/mass spectrometry (LC-MS/MS) as described earlier [34].
For toxicity study, C57Bl/6 male mice were received either 30% BME through drinking water or 20 mg/kg of M-I twice a day or 30 mg/kg of M-I once a day by IP injection for 5 days. There were three mice in each group, including untreated control. Blood was collected on day 6, and serum was prepared. Serum parameters related to hepatotoxicity (total bilirubin, alanine transaminase, aspartate transaminase, and alkaline phosphatase) and nephrotoxicity (urea and creatinine) were performed.

Tumorigenicity Assay
JHU029 or Cal27 cells (1.5 × 10 6 ) containing 40% Matrigel were injected subcutaneously into the flank of BALB/c athymic nude mice (7-8 weeks old). When the palpable tumor was developed (>60 mm 3 ), mice were divided into three groups randomly, 5 mice in each group. The control group was without any treatment. The BME group received 30% (v/v) BME through drinking water. M-I group received 30 mg/kg dose of M-I once in day, every day. Body weight was monitored, and tumor size was measured using a slide caliper and volume was calculated using the formula 1 2 L × W 2 . After animal sacrifice tumors were dissected out and snap frozen in liquid nitrogen for further analysis. All the animal experiments were carried out in accordance NIH guidelines, following a protocol approved (1017) by the Institutional Animal Care and Use Committee (IACUC) of Saint Louis University.

Statistical Analysis
The results are presented as means ± standard deviations. Data were analyzed by Student's t-test. p value of <0.05 was considered statistically significant. All experiments were repeated at least three times except animal experiments, and representative data are shown.

Conclusions
In conclusion, this is the first report showing M-I, a secondary metabolite from bitter melon, inhibited tumor growth in HNC xenograft models with no apparent toxicity. Mechanistic data demonstrated that M-I impairs c-Met signaling in HNC (JHU022, JHU029, Cal27) cells, which is schematically illustrated in Figure 4. Thus, M-I could be used as an additional chemotherapeutic agent or its structural motif to be developed against HNC.

Data availability statement:
The data generated for this study is included in this manuscript.

Acknowledgments:
We thank Bahaa Eldien Elgendy for the discussion and Hiroshi Nikanishi for initial involvement in the project. We like to thank Mike Cameron at the Scripps Institute at Florida for PK/PD study.

Conflicts of Interest:
The authors declare that they have no conflict of interest.

HNC
head and neck cancer BME bitter melon extract M-I momordicine-I IUPAC international union of pure and applied chemistry MAP30 Momordica antiviral protein, 30 kD NOK normal oral keratinocyte IC 50 half maximal inhibitory concentration c-Met mesenchymal-epithelial transition factor c-Myc Myc proto-oncogene protein STAT3 signal transducer and activator of transcription 3 EGFR epidermal growth factor receptor PD-1 programmed cell death protein 1 LC-HRESIMS liquid chromatography coupled to high resolution electrospray ionization mass spectrometry PK pharmacokinetics T 1 2 elimination half-life Tmax time to reach maximum (peak) plasma concentration following drug administration Cmax maximum (peak) plasma drug concentration AUClast area under the plasma concentration-time curve from time zero to time of last measurable concentration AUCINF_obs area under the concentration-time curve extrapolated from zero up to infinity AUC (%Extrap) area under the first moment of the plasma concentration-time curve extrapolated from time t to infinity as a percentage of total AUC Cl_obs apparent total body clearance of the drug from plasma