Gambogic Acid Efficiently Kills Stem-Like Colorectal Cancer Cells by Upregulating ZFP36 Expression

Colorectal cancer (CRC) is one of the most common cancers worldwide [1]. While local resection is an effective approach for CRC therapy, distant metastases are still a major issue in its treatment [2, 3]. Therefore, the identification of novel and effective therapies against CRC is urgently needed.

Gambogic acid (GA, C₃₈H₄₄O₈) is a polyprenylated xanthone and one of the main active compounds of Gamboge hanburyi (a traditional Chinese medicine), which has been used for thousands of years for detoxification, maintenance of homeostasis, anti-inflammation, and as a parasiticide [4]. GA has been proven to have anti-tumor effects in many tumor types [5, 6]. Recent reports suggest that the molecular mechanisms underlying the anti-tumor effects of GA include inducing apoptosis and cell cycle arrest, as well as the inhibition of telomerase activity, angiogenesis, and the invasion and metastasis of tumor cells [7-11].

Cancer stem cells (CSCs) play an essential role in tumor initiation, metastasis, and relapse. CSCs are therefore considered to be the primary cause of tumor treatment failure, due to their resistance to conventional therapies [12-14]. Therefore, inhibiting the growth of tumoral CSCs is critically important for cancer treatment. In this study, we investigated the effects of GA on the growth of CRC CSCs, both in vitro and in vivo. Our study may provide a novel approach for the effective treatment of CRCs.

The human CRC cell lines HCT116 and SW480 were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) in a humidified incubator with 5% CO₂ at 37°C.

The tumor-killing activity of Gambogic acid (GA, Sigma) against CRC cells was assessed using a CCK8 assay. Briefly, HCT116 and SW480 cells were plated in 96-well plates at 5x10³ cells/well. GA was added after the CRC cells had adhered at the final concentrations (0, 0.125, 0.25, 0.5, 1, 2, or 3 µM) and were incubated for 24h, 48h, or 72h. The inhibition rate was calculated according to the following formula. Inhibition rate (100%) = 1- [(A value in experimental well) / (A value in target cell wells)] ×100%. After CRC cells were incubated with GA at different concentrations (0, 0.5, or 1 µM) for 24h, cells were washed with PBS, and used for the colony formation assay. The assay was performed in triplicate.

After CRC cells were treated with GA at different concentrations (0, 0.5, or 1µmol) for 24h, cancer cells were harvested and counted for the tumor spheroid formation assay. Next, 10, 000 CRC cells were plated in 6-well ultralow attachment plates (Corning, Corning, NY) in serum-free DMEM-F12 (Hyclone), supplemented with epidermal growth factor (20ng/mL, Peprotech), basic fibroblast growth factor (10ng/mL, Peprotech), and B27 (1: 50 dilution; BD). After one week of incubation, the tumor spheres were counted under a microscope. The experiment was performed in triplicate.

SW480 cells were co-cultured at different GA concentrations (0, 0.5, or 1µmol) for 24h, and CRC cells were then washed and digested with 0.25% trypsin, washed twice with PBS, and resuspended in RPMI-1640 medium (supplemented with 2% FBS) at a concentration of 5×105 cells/mL. The DNA binding Hoechst 33342 dye (Sigma, USA) was added and the cells were incubated at 37°C in a 5% CO₂ incubator for 90 min in the dark with periodic mixing. The cells were again washed twice with PBS, 1 mg/mL propidium iodide (Sigma, USA) was added, and the cells were maintained at 4°C in the dark prior to flow cytometry analysis (BD FACSAria). All samples were assayed in triplicate.

The CD133⁺/CD44⁺ cell ratio was analyzed by FACS, according to the manufacturer’s protocol (MACS Miltenyi Biotec). The cells were treated with GA, then washed and resuspended in PBS. Next, 10µl of the CD133/1 (AC133)-PE and 10µl CD44-FITC antibodies were added per 100 ml of suspension, and the suspension was mixed well and incubated for 20 minutes in the dark at 4°C. Finally, after two more washes, the cell pellet was resuspended in 400-500µl buffer for flow cytometry analysis. All samples were assayed in triplicate.

HCT116 and SW480 cells were labeled using CD133-PE and CD44-FITC antibodies, as described above. CD133⁺/CD44⁺ fractions and non-CD133⁺/CD44⁺ fractions were sorted for subsequent tumor-killing activity by fluorescence-activated cell sorting (FACS) (BD FACSAria). All samples were assayed in triplicate.

Total RNA was extracted from the colorectal cancer cells using the Trizol reagent (TaKaRa) and reversely transcribed into cDNA using the PrimeScript RT reagent Kit (TaKaRa), according to the manufacturer’s instructions. We analyzed mRNA expression with qRT-PCR on a Stratagene Mx3005P qRT-PCR system using the SYBR Green qRT-PCR master mix (TaKaRa) and GAPDH for normalization. All samples were normalized to the internal controls and the fold changes were calculated through relative quantification (2^-ΔΔCt). Three independent experiments were carried out.

Protein lysates were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to a polyvinylidene difluoride (PVDF) membrane. The blots were probed with the indicated primary antibodies (β-catenin, BD Biosciences, 1: 1000), (Vimentin, BD Biosciences, 1: 1000), (Bmi-1, Abcam, 1: 1000), (ALDH1, Abcam, 1: 1000), (Oct4, Abcam, 1: 1000), (Nanog, Abcam, 1: 1000), (ABCG2, Abcam, 1: 1000), (ZFP36, SANTA CRUZ, 1: 1000), (EGFR, Cell Signaling Technology, 1: 1000), (p-EGFR, Cell Signaling Technology, 1: 1000), (ERK1/2, ABclonal, 1: 1000), (p-ERK1/2, ABclonal, 1: 1000), followed by HRP (horseradish peroxidase)-labeled secondary antibody. The hybridization signal was detected using enhanced chemiluminescence (ECL) (Cat.No:KGP1122, KeyGEN BioTECH). GAPDH was used as the loading control.

After CRC cells were treated with different concentrations of GA for 24h, the cancer cells were collected and counted for the Transwell migration and Boyden invasion assays, as previously described [15]. Three independent experiments were carried out.

RNA from SW480 cells with and without GA treatment was extracted with a total RNA isolation kit (Agilent). The microarray expression profiles were collected using Affymetrix PrimeView^TM Human Gene Expression Array.

The pLV-shZFP36-EGFP plasmid was purchased from Gene Copoeia. The pLV-P_Nanog-GFP-T2A-Luc plasmid housed the reporter genes [i.e., GFP and luciferase (Luc)] under the control of Nanog promoter, the puromycin resistance gene under the control of the EF-1a promoter, and the lentiviral packaging plasmids psPAX2 and pMD2.G, all of which were kindly provided by Prof. Dong Xiao (Southern Medical University, China).

In order to generate stable cell lines, recombinant lentiviruses (LV-shZFP36-EGFP and LV-P_Nanog-GFP-T2A-Luc) were generated as previously described [16] and were used to infect HCT116 and SW480 cells. Next, stable infected cells were selected with puromycin at a 4μg/ml concentration to establish the puromycin-resistant stable reporter cell line (HCT116 and SW480 cells) carrying the P_Nanog-GFP-T2A-Luc transgene, followed by a GFP assay with an inverted fluorescence microscope (Nikon, Japan) and flow cytometry (BD FACSAria). GFP+ and GFP- fractions were then sorted for subsequent experiments by fluorescence-activated cell sorting (FACS) (BD FACSAria).

The animal experiments were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of Guangzhou Medical University. The animal protocol was approved by the Ethics Committee for Animal Experiments of Guangzhou Medical University. Nude mice were purchased from the Medical Laboratory Animal Center of Guangdong Province, and were maintained in microisolator cages under aseptic conditions. HCT116 cells (1x10⁶ cells) harboring the P_Nanog-GFP-T2A-Luc transgene were resuspended in a mix of PBS and BD Matrigel (BD Biosciences) (1: 1), and then subcutaneously injected into the right dorsal thigh of each mouse. One week after HCT116 cell implantation, the mice were treated daily with GA (1mg/kg) or PBS, for the untreated control, via intraperitoneal injection. Tumor growth was determined by caliper measurement or in vivo bioluminescence (BLI) (see below). Tumor growth was monitored with calipers and the volume was calculated using the formula, V = (L×l²) /2, where L is the length and l is the diameter width of the tumor. Twenty days after cancer cell implantation, mice were sacrificed, tumors were dissected, weighed, and fixed overnight in 4% paraformaldehyde, dehydrated, paraffin-embedded, and sectioned. All surgery was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize the suffering of animals.

The protocols for whole-animal bioluminescence imaging to noninvasively detect Luc activity using the Xenogen IVIS Lumina Ⅱ Imaging System (Xenogen Corp., Alameda, CA, USA) have been previously described [17].

For histology analysis, tumor tissues were fixed with 4% paraformaldehyde (PFA) in PBS, embedded in paraffin, cut into 5mm thick sections, and then deparaffinized, followed by hematoxylin and eosin staining (H&E staining) according to standard procedures. After deparaffinization and rehydration, the paraffin-embedded sections were exposed to high pressure for 2 min for antigenic retrieval. The slides were then incubated overnight at 4°C with the indicated primary antibodies (Nanog, Abcam, 1: 50), (Ki67, Abcam, 1: 50), (ZFP36, SANTA CRUZ, 1: 50), (p-EGFR, Cell Signaling Technology, 1: 50), (ERK1/2, ABclonal, 1: 50).

Data are presented as mean ± SD. Statistical analysis was performed using the SPSS 13.0 software package and Graphpad 5.0. An independent-sample t-test was used for comparisons of 2 independent groups. One-Way ANOVA was used to compare multiple groups. The mixed model analysis of variance (Two-Way ANOVA) was employed to assess the GA cytotoxic activity curves in vivo. Statistical significance was assessed as *P < 0.05 and **P < 0.01.

The anti-tumor activity of GA against the CRC HCT116 and SW480 cell lines was first evaluated using the CCK8 assay. As shown in Fig. 1A and B, GA inhibited the growth of HCT116 and SW480 cells in a dose-dependent manner. The colony formation assay also showed that GA treatment resulted in fewer colonies in a concentration dependent manner compared with the control group (Fig. 1C). Taken together, these results suggest that GA significantly inhibited the growth of CRC cells.

Previous studies have reported that cancer stem cells can form tumor spheres in serum-free, non-adherent conditions in vitro, and that they can even become enriched in this condition [18]. Therefore, we examined the ability of HCT116 and SW480 cells to form tumor spheres after treatment with different concentration of GA by using an in vitro tumor sphere formation assay. The results showed that there was a dramatic decrease in the number of tumor spheres in GA treated CRC cells compared to the control group (Fig. 2A).

Side populations (SPs) and CD133⁺CD44⁺ cells have also been reported to exhibit cancer stem cell characteristics [19-21]. We further examined the effect of GA treatment on the percentages of stem cell-like SP cells in SW480 cells, and found that GA treatment sharply decreased the size of the SP cells in the SW480 cell line (from 3.6% to 0.8%) (Fig. 2C). As shown in Fig. 2B, GA treatment dramatically reduced the percentages of CD133⁺CD44⁺ cells in the HCT116 cell culture. Together, our results indicate that GA treatment can remarkably reduce the cancer stem-like cell population of CRC cells.

In order to separately evaluate the tumor-killing activity of GA against CD133⁺CD44⁺cells (putative CSCs) and non-CD133⁺CD44⁺cells, we sorted CD133⁺CD44⁺ and non-CD133⁺CD44⁺cells by flow cytometry. Our results showed that the antitumor activity of GA remained equally effective against both CD133⁺CD44⁺ and non-CD133⁺CD44⁺ cells (Fig. 2D), The IC50 between CD133⁺CD44⁺ and non-CD133⁺CD44⁺ cells is not statistically significant, while the IC50 of CD133⁺CD44⁺ cells treated with 5-FU was increased compared to non-CD133⁺CD44⁺ cells (Fig. 2E). This finding illustrates that the tumor-killing activity of GA was equally intense against putative CSCs and non-CSCs of CRC cells, and GA was more effective than 5-FU chemotherapy.

CSC-related genes (i.e., Nanog, Oct4, Sox2, ALDH1, and Bmi-1) have frequently been used to identify CSC populations from clinical samples and various cancer cell lines. In order to evaluate the effect of GA treatment on stemness-related gene expression in CRC cells, representative stem cell markers were analyzed with both qRT-PCR and western blot. Fig. 3A demonstrates a dramatically decreased expression of stemness-associated genes (i.e., Nanog, Oct4, Sox2, ALDH1, CD133, and Bmi-1) observed at the transcriptional level in the GA group compared with the control. The decreased expression of stemness-related genes was further verified at the protein level (Fig. 3B), further illustrating that GA treatment significantly reduced stemness-related gene expression in CRC cells.

EMT has been shown to endow cancer cells with CSC characteristics [22]. Thus, we reexamined molecular EMT markers in GA-treated CRC cells. As expected, GA-treated CRC cells exhibited the molecular characterization of mesenchyma with a reduced expression of E-cadherin and the enhanced expressions of N-cadherin, vimentin, and fibronectin at the transcriptional level (Fig. 4A). The decreased expression of β-catenin and vimentin were further verified at the protein level (Fig. 4B). In order to further evaluate whether GA inhibits CRC migration and invasion, we tested the migratory and invasive abilities of GA-treated CRC cells using Transwell migration and Boyden invasion assays. We found that CRC cells treated with GA displayed less migratory and invasive activities than did the control group (Fig. 4C). Our data demonstrate that the decreased migratory and invasive ability of GA-treated CRC cells is coincident with the EMT phenotype.

To identify which downstream genes are affected by GA, an mRNA microarray was performed for SW480 cells before and after GA treatment. After hierarchical clustering (Fig. 5A), the top 50 upregulated (>2-fold) and 37 downregulated (>1.5-fold) genes were correlated with different signaling pathways (Table 1). Notably, GA significantly increased the expression of a panel of genes in the apoptosis signaling pathway, many of which have already been studied by other researchers. Based on gene function and existing research results, we chose ZFP36 as a candidate gene, since it has been associated with tumorigenesis [23], EMT [24], and enhanced cancer invasiveness and metastasis [25]. Furthermore, qRT–PCR verified that partially upregulated and downregulated genes were also affected by GA, whereas ZFP36, as a tumor suppressor gene, increased by 6-8 fold in SW480 cells (Fig. 5B). In addition, western blotting results also showed that ZFP36 was upregulated in CRC cells after GA treatment (Fig. 5C).

In order to ascertain whether GA treatment decreased the stem cell-like population and stemness-related gene expression of CRC cells as mediated by ZFP36, we established stable CRC cells (transfected with shZFP36 or shSCR plasmids) (Fig. 5D-F) and performed loss-of-function experiments. We first examined the role of ZFP36 in the maintenance of CRC stemness by ZFP36 knockdown. As shown in Fig. 5, the loss of ZFP36 function through its exogenous expression enhanced the expression of stemness markers (Bmi-1, ALDH1, and ABCG2) (Fig. 5G), and increased the SP cell ratio, the number of colonies, and the number of tumor spheres (Fig. 5H-J). Subsequently, we investigated whether shZFP36 overexpression would rescue the GA treatment-mediated inhibition of stemness gene expression, and reverse the GA-induced drop in the percentages of stem cell-like SP cells, the number of colonies, and the number of tumor spheres in CRC cells. We found that shZFP36 did, indeed, reverse the GA induced downregulation of stemness gene expression (Fig. 5G) and rescued the GA treatment-induced decline in the percentages of stem cell-like SP cells, colony number, and tumor sphere formation ability in SW480 CRC cells (Fig. 5H-J). Taken together, these results demonstrate that ZFP36 is involved in the GA-induced decrease in stemness-related gene expression and the stem cell-like population of CRC cells.

EGFR signaling regulates cell proliferation, differentiation, and tumor cell survival through the activation of EGFR's intrinsic kinases, leading to the subsequent activation of several downstream signaling pathways, including the extracellular related kinase (ERK), phosphoinositide 3-kinase (PI3K), and AKT pathways [26-28]. Furthermore, EGFR signaling is also required for self-renewal of rat embryonic stem cells and cancer stem cells [29-31]. Thus, we performed a western blot analysis to investigate whether GA suppressed the EGFR/ERK signaling pathway in colorectal cancer cells. We observed that GA inhibited the phosphorylation of EGFR and ERK, but that the total levels of EGFR and ERK were unchanged (Fig. 6A), and those of the ZFP36 mRNA and protein were enhanced (Fig. 6A-B). Previous studies have revealed that ZFP36 is a downstream gene of the EGFR/ERK signaling pathway [32]. Subsequently, we investigated the relationship between the EGFR/ERK signaling pathway and ZFP36. When CRC cells were treated with EGFR inhibitor PD153035, we found that the level of ERK phosphorylation was inhibited and ZFP36 protein levels increased (Fig. 6C). Likewise, ZFP36 protein also increased when the ERK inhibitor U0126 was used (Fig. 6D). Taken together, these results indicate that GA could enhance the expression of ZFP36 by inhibiting the EGFR/ERK signaling pathway.

The development of in vitro methods for optical imaging and non-invasive in vivo detection of putative CSCs is of great importance. Fluorescence and bioluminescence imaging may be the best approach for a given experiment [33-35], both of which have their own merits. Nanog has frequently been used as a CSC-associated marker for the identification of the CSC population from the clinical samples and cell lines of many tumors, including CRC [36-38]. Thus, based on a promoter-reporter gene strategy, we constructed the pLV-P_Nanog-GFP-T2A-Luc lentivirus vector carrying GFP and Luc reporter genes under the control of the human Nanog promoter (Fig. 7A), which allowed us to visualize CSCs both in vitro and in vivo.

In order to optically visualize CSCs, HCT116 and SW480 cells were infected with the lentivirus harboring P_Nanog-GFP-T2A-Luc (Fig. 7A). Ten days after the infection of CRC cells with LV-P_Nanog-GFP-T2A-Luc, we found that GFP was highly expressed in a small percentage of stably infected HCT116 and SW480 cells (Fig. 7B). A fluorescence activated cell-sorting analysis showed that 4.0% of cells were GFP+ in the HCT116 cell line and that 2.1% were GFP+ in the SW480 cell line (Fig. 7C). Importantly, the Luc signal was strongly correlated with HCT116 cell numbers (Fig. 7D-E) (R²=0.9922).

Next, GFP-positive (GFP+) and GFP-negative (GFP-) cells were sorted, and their respective gene expressions were analyzed, while the specific assays mentioned below were performed (Fig. 7F). Fig. 7F shows the significantly increased expression of CSC-related genes (Bmi-1, Oct4, and ABCG2) observed in sorted GFP+ cells, as compared with GFP- cells, indicating that GFP+ cells might have stem-cell–like characteristics. Previous studies have demonstrated that cancer stem cells can form tumor spheres in vitro in a non-attached culture condition. As shown in Fig. 7G, GFP+ cells formed more spheres than GFP- cells. In addition, the colony formation assay indicated that GFP+ cells have greater colony numbers compared to GFP-cells (Fig. 7H). GFP+ cancer cells exhibit the characteristics of CSCs.

To examine whether GFP+ cells are efficiently migratory and highly invasive, we used Transwell migration and matrigel-coated Boyden chamber assays, respectively. Our data revealed that GFP+ cells illustrated a dramatic increase in motile capacity and invasive ability compared with GFP- cells (Fig. 7I). The epithelial-mesenchymal transition (EMT) has been shown to endow cancer cells with a strong invasive ability. Thus, we next used qRT-PCR to test for EMT-related markers in both GFP+ and GFP- cells. As expected, GFP+ cells exhibited the molecular characterization of mesenchyma with reduced E-cadherin expression and increased N-cadherin, Vimentin, and Fibronectin expression (Fig. 7J), demonstrating that the increased migratory and invasive capacity of GFP+ cancer cells is coincident with the EMT phenotype.

Our data explored the ways in which GFP+ cells harboring the P_Nanog-GFP-T2A-Luc transgene exhibited the characteristics of CSCs. Therefore, the newly established CRC cell line harboring the P_Nanog-GFP-T2A-Luc transgene allows for the better visualization and monitoring of a small population of putative CSCs both in vitro and in vivo.

As shown by various efficient in vitro methods, GA exhibited a strong ability to kill putative CSCs of CRC. We next evaluated the antitumor activity of GA against putative CSCs of CRC in nude mice by in vivo optical imaging. As mentioned above, the established HCT116 cells carrying GFP and the Luc reporter gene under the control of the Nanog promoter allowed us to monitor a rare population of putative CSCs in human tumor xenograft-bearing mice using in vivo bioluminescence imaging.

First, nude mice were subcutaneously implanted with HCT116 cells (1x10⁶) harboring the P_Nanog-GFP-T2A-Luc transgene. Starting one week after HCT116 cell implantation, GA (1mg/kg) was intraperitoneally injected into the mice each day. A significant reduction in tumor growth was observed in the GA-treated mice compared with the controls (Fig. 8A, B) As early as 4 days after GA treatment, the growth of the transplanted tumors between GA-treated mice and controls became statistically significant (P <0.05 or 0.01, Fig. 8B), respectively. On day 20, all tumors were excised and weighed, in order to further illustrate the strong anti-tumor activity of GA against HCT116 cells (Fig. 8D, E). The results of the ensuing immunohistochemical analysis revealed that the number of hyperproliferative Ki67-positive tumor cells in the GA-treated group decreased significantly (Fig. 8F).

The above-mentioned data from these conventional methods exhibited a significant delay in the growth of subcutaneous tumor xenografts observed in GA-treated mice (Fig. 8B, D, E, F). The established subcutaneous tumor xenografts, expressing both GFP and the Luc reporter gene under the control of the Nanog promoter, allowed us to detect putative CSCs in human tumor xenograft-bearing mice using bioluminescence imaging. Therefore, after the sizes of the subcutaneous tumors were determined by caliper measurement, tumor bearing mice were imaged at days 8, 12, 16, and 20 by in vivo bioluminescence imaging to detect the changes in the number of putative CSCs. As shown in Fig. 8A, C, after GA treatment, a significant bioluminescence signal reduction was found in GA-treated mice compared with untreated controls and, as early as 12 days post-treatment, the signal intensities between GA treated mice and the controls became statistically significant (P <0.05 or 0.01). We further detected the percentage of Nanog-positive cells using a Nanog antibody-based staining method. As shown in Fig. 8F, significant decreases in the percentage of Nanog-positive cells were found in GA treated mice compared with controls, suggesting a general downward trend in the number of putative CSCs within the tumor xenografts. These findings from both in vivo bioluminescence imaging and Nanog antibody-based staining clearly demonstrate that GA treatment leads to the remarkable reduction in the number of putative CSCs within tumor xenografts, demonstrating that GA can efficiently kill putative CSCs in vivo.

To determine the mechanism by which GA killed putative CSCs of CRC through the inhibition of the EGFR/ERK/ZFP36 pathway in vivo, an IHC analysis was performed in the tumor tissues of the two experimental groups. The expression of p-EGFR, p-ERK was significantly reduced in the GA group, while ZFP36 was significantly enhanced compared with the control (Fig. 8F), in support of the in vitro results.

Recent studies have demonstrated the preclinical antitumor activity of GA against multiple types of tumors, including colorectal cancer [39, 40]. However, whether or not it inhibits the growth of tumoral CSCs has not yet been elucidated. Intensive research should be done to explore the inhibitory activity of GA against CSCs of various cancers, since CSCs are associated with tumor therapy failure. Our study showed the intense tumor killing activity of GA against CSCs of CRC both in vitro (tumor sphere formation assay, detection of SP and CD133⁺CD44⁺ cells, antitumor activity of GA against CD133⁺CD44⁺ CSCs, and detection of stem cells markers) and in vivo (optical imaging and quantifying putative CSCs within tumor bearing mice using in vivo bioluminescence imaging).

We further explored the mechanisms underlying the antitumor effects of GA on CSCs of CRC. We identified ZFP36 as a potential downstream gene affected by GA using an mRNA microarray. ZFP36 was originally described as an active regulator of the inflammatory response, but recent reports show that its expression is commonly deficient in a number of cancer cell types when compared to normal cells [41]. Moreover, ZFP36 expression seems to counteract the progression of malignancy [42, 43]. In this study, our results showed that ZFP36 was upregulated in CRC cells after GA treatment and was involved in the maintenance of CRC cell stemness through the suppression of ZFP36 expression by shRNA.

Next, we explored whether or not the GA treatment decrease in the stem cell-like population and stemness-related gene expression of CRC cells was mediated by ZFP36. ZFP36 has been reported to be involved in the regulation of proliferation and EMT of many tumors via different signaling pathways [24, 44, 45]. There are no publications, however, on whether ZFP36 regulates the stemness of colorectal cancer stem cells. In the present study, we first examined the role of ZFP36 in the maintenance of CRC cell stemness through the exogenous shZFP36 expression using various methods (tumor sphere formation assay detection of SP cell percentage, and detection of stem cells markers). In addition, we found that shZFP36 reversed the GA-induced downregulation of stemness gene expression and rescued the GA-induced reduction in colony and tumor sphere formation and the percentages of stem cell-like SP cells in CRC. GA treatment decreased the stem cell-like population and stemness related gene expression of CRC cells by upregulating ZFP36. Previous studies have revealed that ZFP36 is a downstream gene in the EGFR and ERK signaling pathway [32], which is also required for the self-renewal of cancer stem cells in addition to regulating proliferation and survival [31]. KEGG pathway analyses and related studies have reported that GA can downregulate the EGFR and ERK signaling pathway [46]. Furthermore, our results revealed that GA could increase the expression of ZFP36 by inhibiting the EGFR/ERK signaling pathway. As a traditional Chinese medicine extract, we showed that GA exerted a multi-target inhibitory effect, and might have the potential for further therapeutic use in colorectal cancer.

Increasing evidence has revealed that CSCs are responsible for tumorigenicity, progression, therapeutic resistance, and tumor recurrence. Thus, the development of methods for in vitro optical imaging and the non-invasive in vivo detection of putative CSCs is of great importance [34, 35]. In our previous nasopharyngeal carcinoma and liver cancer studies [47, 48], we developed a new methodology for both the in vitro and in vivo visualization of putative CSCs based on a lentiviral “CSC detector” vector encoding the GFP and Luc proteins controlled by the human Nanog promoter, which has never previously been reported. In this study, in order to visualize the CSCs of CRC cells, HCT116 and SW480 cells were infected with the lentiviruses carrying PNanog-GFP-T2A-Luc. We then sorted GFP+ cancer cells (putative CSCs) from CRC cells carrying the P_Nanog-GFP-T2A-Luc transgene and exhibiting the characteristics of CSCs. Based on these results, the noninvasive in vivo bioluminescence imaging, via the detection of Luc expression, allowed us to monitor the remarkable reduction in the number of putative CSCs within tumor bearing mice after GA treatment. Therefore, the newly established CRC cell line harboring GFP and Luc reporter genes under the control of the Nanog promoter allowed us to visualize putative CSCs within tumor xenograft-bearing mice, which has never been reported in colorectal cancer.

CSCs are the root cause for cancer treatment failure. Nevertheless, the development of new therapeutic strategies targeting CSCs is currently hindered by the lack of special markers for their identification. Moreover, one of the key goals in cancer research over the past decade has been to develop therapeutic strategies to eliminate CSCs and to cure the cancer with little to no damage to normal tissues, but a major barrier to this goal lies in the identification of the key mechanisms that distinguish CSCs from normal endogenous tissue stem cells.

Gambogic acid (GA) is expected to be a potential new antitumor drug, but its poor aqueous solubility and accompanying toxic side effects limit its bioavailability in clinical applications [49, 50]. Fortunately, several strategies have been adopted to improve the anti-tumor efficacy of both soluble and insoluble drugs. These strategies include the use of tumor cell-derived microparticles (T-MPs) and biomimetic nanoparticle-based therapies to improve the antitumor efficacy of drugs with relatively low toxicity [51, 52]. Drug-packaging MPs have shown promising anti-tumor activity [51]. They possess multiple advantages in cancer therapy, including their (i) efficient packaging of both soluble and insoluble drugs; (ii) ability to package large amounts of drug molecules; and (iii) safety and nontoxicity. In addition, the preparation of tumor cell-derived, drug-packaging MPs is simple, further facilitating their applicability. These features make drug-packaging MPs highly efficacious in targeting tumor cells, especially for cancer stem cells [51].

RBCm (red blood-cell membrane)-GA/PLGA (poly lactic-co-glycolic acid) NPs (nanoparticles) has been used as a biomimetic nanocarrier and united the customizability and flexibility of synthetic copolymer PLGA with the functionality and complexity of natural liposome RBCm [52]. The loading of the potential antitumor drug GA bestowed a new antitumor efficacy on the nanocarrier. The combination of RBCm and PLGA has been shown to enhance the stability and biocompatibility of a composite carrier compared with bare nanoerythrosomes and a polymeric carrier [52]. Previous research has demonstrated that the synthetic RBC-NP platform enhances the anti-tumoral efficiency of GA [52]. It is necessary, therefore, to explore whether or not GA-loaded RBC NPs kill stem-like cancer cells of colorectal cancer more efficiently than does GA alone.

In conclusion, our in vitro and in vivo findings here demonstrate the intense tumor killing activity of GA against putative CSCs of CRC through the inactivation of the EGFR/ERK/ZFP36 signaling pathway. Further research is required to identify the role of the EGFR/ERK/ZFP36 signaling pathway in the therapeutic use of GA and the maintenance of the stemness of CRC cells. We will carry out further preclinical and clinical investigations on the prospective potential of targeting putative CSCs of CRC with GA-packaging MPs and GA-loaded RBC NPs.

This work was supported by the National Natural Science Foundation of China (Grant No. 81602598, to F. Wei), the Science and Technology Planning Project of Guangdong Province of China (Grant No. 2017A020215009, to F. Wei), and the Fundamental Research Funds for the Central Universities (Grant No. 2017MS125, to F. Wei), Guangzhou Science Technology and Innovation Commission (2014Y2-00137, to J. Cao).

All experiments of this study were authorized by the Ethics Committee of Guangzhou First People’s Hospital (K2016-103-01).

The authors declare that they have no competing interests.

F. Wei, T. Zhang, Z. Yang, J.-C. Wei contributed equally to this work.