Suppressive effect of α-mangostin for cancer stem cells in colorectal cancer via the Notch pathway

Since colon cancer stem cells (CSCs) play an important role in chemoresistance and in tumor recurrence and metastasis, targeting of CSCs has emerged as a sophisticated strategy for cancer therapy. α-mangostin (αM) has been confirmed to have antiproliferative and apoptotic effects on cancer cells. This study aimed to evaluate the selective inhibition of αM on CSCs in colorectal cancer (CRC) and the suppressive effect on 5-fluorouracil (5-FU)-induced CSCs. The cell viability assay was performed to determine the optimal concentration of αM. A sphere forming assay and flow cytometry with CSC markers were carried out to evaluate the αM-mediated inhibition of CSCs. Western blot analysis and quantitative real-time PCR were performed to investigate the effects of αM on the Notch signaling pathway and colon CSCs. The in vivo anticancer efficacy of αM in combination with 5-FU was investigated using a xenograft mouse model. αM inhibited the cell viability and reduced the number of spheres in HT29 and SW620 cells. αM treatment decreased CSCs and suppressed the 5-FU-induced an increase in CSCs on flow cytometry. αM markedly suppressed Notch1, NICD1, and Hes1 in the Notch signaling pathway in a time- and dose-dependent manner. Moreover, αM attenuated CSC markers CD44 and CD133, in a manner similar to that upon DAPT treatment, in HT29 cells. In xenograft mice, the tumor and CSC makers were suppressed in the αM group and in the αM group with 5-FU treatment. This study shows that low-dose αM inhibits CSCs in CRC and suppresses 5-FU–induced augmentation of CSCs via the Notch signaling pathway.


Background
CRC is the second-most frequent cause of cancer-related deaths in United States and many other high-income countries [1,2]. While the best way to treat CRC is the complete surgical resection of the primary lesion, less than 25% of all patients are operable, and high percentage of patients may experience recurrence [3][4][5][6]. Patients with inoperable CRCs are usually treated with palliative chemotherapy, and a large number of patients have also required postsurgical chemotherapy for preventing tumor recurrence [7].
CSCs are small subset of the cancer cells with characteristics including proliferation, self-renewal, and asymmetric differentiation [8][9][10]. Conventional chemotherapeutic agents and radiotherapy may show therapeutic effects on rapidly growing tumors but cannot inhibit CSCs [11]. Previous studies reported that conventional chemotherapy can lead to an increase in colorectal CSCs [12,13]. CSCs are closely associated with chemoresistance, cancer metastasis, and recurrence after primary therapy [8,[14][15][16]. Therefore, targeting of CSCs has emerged as an important aspect of effective cancer treatment.
Recently, certain components from fruit and vegetables were identified to have a chemopreventive effect on cancers and anticancer properties [17]. Among them, mangostin (Garcinia mangostana), a tropical evergreen tree commonly found in Southeast Asia [18][19][20][21], has been used in the traditional treatment of skin infections and in wound-healing for a long time [22]. Among the various secondary metabolites of mangostin, xanthones and polyphenolic substances show a variety of physiological activities including anti-inflammatory, antibacterial, and anticancer effects [23]. α-mangostin (αM) is one of the main bioactive and most abundant xanthones extracted from mangostin [23]. To date, αM has been widely investigated as a chemotherapeutic and chemopreventive bioactive compound [24]. In addition, novel xanthone derivatives based on αM were synthesized and evaluated as anticancer agents [25]. Consequently, αM has been shown to be effective in various cancers, including CRC, pancreatic, prostate, oral squamous, and breast cancers [18,20,21,[26][27][28][29]. In this study, we aimed to evaluate whether αM can selectively inhibit CSCs in CRC and whether it can also suppress an increase in the number of CSCs in combination with conventional anticancer agents.

3D spheroid invasion assay
The 3D spheroid invasion assay was conducted with the aforementioned HT29 cells. HT29 cells were trypsinized, and 1 × 10 5 cells were resuspended in 5 mL DMEM with 20% methocel solution (methylcellulose, Sigma-Aldrich) and 1% Matrigel (Corning). Hanging drops (25 μl) were suspended on petri dishes (SPL), and cells were harvested after 2 d. Harvested spheroid cells were embedded in collagen gels (rat tail collagen, BD bioscience), which were polymerized at 37 °C. These spheroids were incubated for 5 d, and invasion ratios were calculated using ImageJ software (version 1.51 J8; National Institutes of Health, Bethesda, MD, USA).

Quantitative real-time PCR analysis
Total cellular RNA was extracted from HT29 and SW620 cells, by using Trizol reagent (Invitrogen) and the RNeasy Mini Kit (Invitrogen) in accordance with the manufacturer's protocol. The total RNA concentration was measured using a Nanodrop spectrophotometer (Nabi UV/ Vis Nano spectrophotometer, Microdigital, Gyeonggi, Republic of Korea) with an A 260/280 cut-off of approximately 2.0. Purified RNA (2 μg) was reverse-transcribed (with the Reverse Transcription Kit, Applied Biosystems, Framingham, MA, USA). Quantitative real-time PCR was performed with power SYBR Green master mix (Applied Biosystems) on Quant studio 3. The cycling conditions were as follows: denaturation for 2 min at 50 °C, 10 min at 95 °C, followed by 40 cycles at 95 °C for 5 s and 60 °C for 60 s, followed by dissociation for 15 s at 95 °C and annealing and extension at 60 °C for 20 s. The relative mRNA levels were normalized to those of β-actin mRNA using the 2 -ΔΔCt method. Primers for quantitative real-time PCR are listed in Supplementary Table S1.

Western blot assay
The Western blot assay was conducted to determine the expression levels of Notch1, NICD1, Hes1, Hey1, and β-actin, under 4 experimental conditions. Proteins were extracted from cells by using radioimmunoprecipitation assay (RIPA) lysis buffer (iNtRON Biotechnology, Gyeonggi, Republic of Korea). The concentration of the isolated proteins was determined using a bicinchoninic acid (BCA) protein assay (Thermo Scientific-Pierce, Waltham, MA, USA). Proteins (20 μg) were separated through 8, 10, and 12% SDS-PAGE (Hoefer, San Francisco, CA, USA) and transferred to polyvinylidene fluoride membranes (PVDF, Merck). The membranes were blocked using 3% bovine serum albumin (BSA, Sigma-Aldrich) for 30 min at room temperature (RT). Protein extracts were incubated with primary antibodies overnight at 4 °C and with secondary antibodies for 1 h at RT. Proteins were detected using the enhanced chemiluminescence (ECL) Western blotting Luminol reagent (Santa Cruz). Images were obtained using a Lumino image analyzer (LAS-4000 Mini, Fujifilm, Tokyo, Japan).

Flow cytometry analysis
For flow cytometry, cells were washed with PBS and incubated with Accutase (Gibco) for 10 min. After adding flow cytometry buffer (

Assessment of in vivo anticancer efficacy
Six-week-old male Balb/c athymic mice were purchased from Orient Bio (Seongnam, Republic of Korea) and acclimated for 1 week. All mouse experiments were conducted under approved guidelines of the Animal Care and Use Committee of Ewha Womans University (EUM17-0368). HT29 cells (1 × 10 6 cells) were suspended in DMEM with Matrigel matrix (1:1 ratio). The mixed cells were injected subcutaneously into the right rear flank of each mouse. After 11 days of injection, mice were divided into 4 treatment-based groups (5 mice per group): control, 5-FU only, αM only, and 5-FU and αM. 5-FU (30 mg/kg body weight) or/and αM (5 mg/kg) were administered intraperitoneally thrice a week for 18 d. Tumor volume was calculated (volume = length × width × width/2), and body weight was measured thrice a week. All mice were euthanized through CO 2 asphyxiation, and the weight and volume of the excised tumor were measured on day 29.

Statistical analysis
Data are expressed as mean ± standard error of the mean (SEM) or mean ± standard deviation (SD) values. All analyses were performed using Graph Pad Prism 8.0 software (Graph Pad Software, La Jolla, CA, USA) and SPSS software (version 22.0, Chicago, IL, USA). A P value of < 0.05 was considered significant. Statistical significance was determined using the Mann-Whitney U test for nonparametric data and a two-tailed Student t test for parametric data.

Inhibitory effect of low-dose αM on colosphere formation
The number of spheres from SW620 cells decreased after the treatment with αM in a dose-dependent manner (Fig. 1D). Compared to the control group, 1.25 μM (P < 0.01) and 2.5 μM (P < 0.001) αM significantly decreased sphere formation in SW620 cells. In HT29 cells, the number of spheres were significantly decreased upon treatment with 0.25 μM (P < 0.001) and 0.5 μM (P < 0.001) αM (Fig. 1E). Sphere formation was not observed for SW620 cells treated with 5 and 10 μM αM and HT29 cells treated with 1 and 2 μM αM. These results indicate that sphere formation was suppressed with low-dose αM in both SW620 and HT29 cells.

Reduction of CSCs and 5-FU-induced CSCs upon treatment with low-dose αM
To evaluate the inhibitory effect of αM on CSCs and 5-FU-induced increase in CSCs, the expression levels of CD133 and CD44, which are well-known as CSC markers, were monitored after treating HT29 cells with αM with or without 5-FU for 72 h (Fig. 2A). The proportion of CD133 + CD44 + cells significantly decreased upon treatment with 0.5 μM (control: 31.48% vs αM: 25.86%; P < 0.01) and 1.0 μM αM (control: 31.48% vs αM: 23.94%; P < 0.001). The proportion of CD133 + CD44 + cells increased to 56.72% upon treatment with 2 μM 5-FU and decreased to 46.89% or 40.23% upon treatment with 0.5 μM or 1.0 μM of αM, respectively. The number of spheres from SW620 cells decreased after the treatment with 5-FU with or without αM ( Supplementary Fig. S2A). These results suggest that αM selectively inhibits CSCs and the 5-FU-induced increase in CSCs. In addition, a 3D spheroid invasion assay was conducted to analyze the effect of αM on cancer cell invasion. As shown in Fig. 2B, αM significantly inhibited cancer cell invasion compared to the control (54.77% vs 100%, respectively; P < 0.05).

Inhibition of CSCs via the NOTCH-HES1 pathway upon treatment with low-dose αM
Notch signaling, a highly conserved pathway, is reportedly involved in the self-renewal of CSCs and contributes to cancer metastasis (Gu et al., [30]; Pannuti et al., [31]). In colospheres of HT29 cells, treatment with 0.25 and 0.50 μM αM downregulated Notch1, Hes1, and Hey1 ( Fig. 3A) and significantly attenuated Hes1 mRNA levels (Fig. 3B). Notch signaling proteins including Notch1, Hes1, and Hey1 were downregulated after αM treatment in HT29 and SW620 cells ( Furthermore, other signaling pathways related to CSCs, except for Notch signaling, were analyzed, which showed that they were not suppressed in a dose-dependent manner (Supplementary Fig. S4). These results suggest that Notch signaling is related to the CSCsuppressive effect of αM.

CSC-inhibitory effect on αM in an in vivo xenograft mouse model
To evaluate the inhibitory effect of αM on CRC CSCs in vivo, its antitumor efficacy was assessed using a xenograft mouse model. 5-FU and/or αM were administered intraperitoneally to mice from days 11-28 (Fig. 4A). Among 4 groups including the control, 5-FU only, αM only, and 5-FU + αM, no significant differences were observed in body weight (Fig. 4B). Regarding tumor volume, tumors in the 5-FU + αM group were significantly smaller than those in the 5-FU only treatment group, and tumors in the αM group were significantly smaller than those in the control group (Fig. 4C). On day 29, the weight of the excised tumor of the 5-FU + αM group was significantly lower than that in the 5-FU only group (5-FU: 0.3 g vs 5-FU + αM: 0.14 g; P < 0.01), while no difference was observed between the control and αM only groups (Fig. 4D).
Regarding the CSC population in the excised tumors, treatment with αM only decreased the proportion of showing the protein levels of Notch1, NICD1, Hes1, Hey1, and β-actin in HT29 cells treated with or without 2 μM 5-FU and 1.0 μM αM. F Expression of CD44 and CD133 (CSC markers) was analyzed with or without 5-FU treatment through flow cytometry, using αM or DAPT. HT29 cells were treated with 1.0 μM αM and DAPT with or without 5-FU for 72 h for 11 times. The proportion of CD133 + CD44 + cells was significantly decreased with both αM and DAPT. In addition, the 5-FU-induced increase in CD133 + CD44 + cells was significantly attenuated by αM and DAPT treatment. N = 11 Data are expressed as mean ± SEM values; * P < 0.05, ** P < 0.01, *** P < 0.001 CD133 + CD44 + cells compared to control (17.74% vs 11.72%, respectively; P < 0.01), and 5-FU + αM treatment attenuated these cells compared to the 5-FU only group (13.35 vs 8.39%, respectively; P < 0.001) (Fig. 4E). αM inhibits the notch signal pathway, leading the CSCs inhibition consequently (Fig. 5). Overall, our results show that αM not only inhibits CSCs but also exerts synergistic therapeutic effects in combination with 5-FU.

Discussion
This study shows that αM has an inhibitory effect on CRC CSCs and attenuates a 5-FU-induced increase in CSCs. The effects of αM on apoptosis and cell cycle arrest through several signaling pathways in CRC have been widely studied [19,[32][33][34][35]. Several studies have reported that αM arrests the cell cycle by regulating cyclins and p27 in DLD-1 cells [33,34]. Furthermore, αM induces apoptosis through the extrinsic and intrinsic pathways in COLO 205 cells [32,35]. Moreover, αM induces apoptosis via the mitochondrial signaling pathway, which is regulated by MAPK, ERK, and Akt [19,33]. In particular, apoptotic signals induced by the expression of proapoptotic proteins p21 and Bax owing to ERK activation, are relevant to the NF-κB pathway [19,33].
Furthermore, αM is reported to have potential anticancer and antiproliferative effects on cervical and pancreatic CSCs [36,37]. αM can inhibit CSC-like spheroids in human breast cancer cells, resulting in a significant reduction in the adherence and migration of cancer cells [38]. Combinatorial treatment with αM and cisplatin reportedly enhanced the therapeutic effects of cisplatin on cervical cancer and attenuated the chemoresistance of cancer cells to cisplatin by inducing apoptosis in CSC-like cervical cancer cells [36]. Combinatorial treatment with αM and doxorubicin reduces cell viability and decreases the retinaldehyde-dependent isoenzymes of aldehyde dehydrogenase (RALDH), a CSC marker, in spheroids of human breast cancer cell line MCF-7 [39]. Combinatorial treatment with αM and chemotherapeutic agents can help overcome chemoresistance through the suppression of CSCs.
The Notch signaling pathway, a highly conserved cellular signaling pathway, plays an important role in proliferation, stem cell maintenance, cell fate specification, differentiation, and homeostasis in multicellular organisms [40][41][42]. Some studies have reported that the Notch signaling is one of the most critical pathways in cancer metastasis [43]. Notch signaling can induce colon adenoma together with Wnt signaling and is necessary to eradicate drug-resistant CRC CSCs [44,45]. Notch1 expression is dysregulated in the initiation step of CRC [46][47][48], positively predicts poor overall survival [49], and is upregulated in advanced tumors [50]. The inhibition of this pathway may enhance therapeutic efficacy in the curative treatment of cancer by eradicating CSCs [43,51].
We hypothesized that αM selectively suppresses CSCs in CRC, and the combined use of αM and current anticancer agents including 5-FU exerts synergistic effects on CRC. The CSC-inhibitory effect was observed upon treatment with low-dose αM without concerns of cytotoxicity. Our results show the differences in αM concentrations and the sphere forming ability between the HT29 and SW620 cells. It should be noted that these differences originated from the different properties of each cell line. HT29 cells derived from primary cancer and SW620 cells were derived from lymph node metastasis [52]. Further, HT29 cells harbored V600E BRAF, P449T PIK3CA, R273H, and TP53 mutations, whereas SW620 cells harbored G12V KRAS, R273H, P309S, and TP53 mutations [52]. Even though both were CRC cell lines, the differences in experimental conditions were inevitable owing to intrinsic properties. In our study, the CSC proportion was around 10% for SW620 cells and 30% for HT29 cells. Accordingly, compared to HT29 cells, SW620 cells were inhibited with a higher concentration of αM in sphereforming assay. This pattern was also observed in cell viability assay. Other CRC cells with a lower CSC proportion were SW480, DLD-1, and HCT116 cells compared to HT29 cells. In cell viability assay, the inhibitory effect of αM was not concentration-dependent in HCT116, DLD-1, and SW480 cells. In addition, cancer cell viability was suppressed by a higher dose of αM in HCT116, DLD-1, and SW480 cells compared to HT29 cells. These results support our hypothesis that αM may selectively inhibit CSCs in CRC.
In both HT29 spheres and the cell line, Notch, Hes1, and Hey1 were downregulated after αM treatment. In addition, the CSC proportion decreased upon treatment with both αM and DAPT. The RNA and protein (See figure on next page.) Fig. 4 αM shows an inhibitory effect on CSCs in an in vivo xenograft mouse model. A Schematic representation of the experimental design. B Body weight of mice was not significantly different between the control group and the other groups during the experiment (Student's t test). C Tumor volume in the HT29 xenograft mice treated with each agent. The tumor volume of the control group was larger than that in the αM group. Tumor volume of the 5-FU + αM group was significantly larger than that in the 5-FU only group. D Tumors from each group were weighed immediately after resection. The tumor weight in the 5-FU + αM group was significantly lower than that in the 5-FU group. E CD44/CD133 double-positive tumors in the αM group were significantly fewer than those in the αM group. These proportions were significantly lower in the 5-FU + αM group than in the 5-FU group. All Data is N = 5. Data are expressed as mean ± SEM values; * P < 0.05, ** P < 0.01, *** P < 0. levels of Notch1, NICD1, Hes1, and Hey1 were inhibited by αM in a concentration-dependent manner. Therefore, αM may inhibit the Notch signaling pathway at the transcriptional level. Of note, the mechanism underlying the CSC-inhibitory effects of αM in CRC are associated with Notch signaling. Moreover, the attenuation of the 5-FU-induced increase in CSCs by αM suggests that αM has the potential to suppress chemoresistance in CSCs. The 5-FU-induced increase in CSCs was suppressed with αM treatment. However, the CSC suppressive effect of αM was higher with αM treatment only than with 5-FU + αM treatment. Based on the results, when considering the clinical use of αM in CRC chemotherapy, the use of αM before conventional chemotherapeutic agents could have a greater therapeutic effect compared to the effect of simultaneous treatment.

Conclusions
In conclusion, our results show that αM suppresses CSCs and inhibits the 5-FU-induced increase in CRC CSCs via Notch signaling. In particular, the fact that therapeutic efficacy is improved only with low-dose αM provides a strong advantage for clinical use. αM might be a promising adjunctive agent with conventional anticancer agents to improve treatment efficacy among patients with CRC.