SNAI2 Attenuated the Stem-like Phenotype by Reducing the Expansion of EPCAMhigh Cells in Cervical Cancer Cells

SNAI2 (Snai2) is a zinc-finger transcriptional repressor that belongs to the Snail family. The accumulated evidence suggests that SNAI2 exhibits biphasic effects on regulating a stem-like phenotype in various types of cells, both normal and malignant. In this study, by exogenously expressing SNAI2 in SiHa cells, SNAI2 exhibited the capacity to inhibit a stem-like phenotype in cervical cancer cells. The SNAI2-overexpressing cells inhibited cell growth, tumorsphere formation, tumor growth, enhanced sensitivity to cisplatin, reduced stem cell-related factors’ expression, and lowered tumor initiating frequency. In addition, the EPCAMhigh cells sorted from SiHa cells exhibited an enhanced capacity to maintain a stem-like phenotype. Further study demonstrated that the trans-suppression of EPCAM expression by SNAI2 led to blockage of the nuclear translocation of β-catenin, as well as reduction in SOX2 and c-Myc expression in SiHa and HeLa cells, but induction in SNAI2 knockdown cells (CaSki), which would be responsible for the attenuation of the stem-like phenotype in cervical cancer cells mediated by SNAI2. All of these results demonstrated that SNAI2 could attenuate the stem-like phenotype in cervical cancer cells through the EPCAM/β-catenin axis.


Introduction
The Snail family is a highly conserved zinc-finger transcriptional family. SNAI2 (Snai2) is a member of this family, participating in numerous types of biological regulation processes, including epithelial-mesenchymal transition (EMT), cancer metastasis, and cellular reprogramming. More interestingly, SNAI2 has also been reported to play a specific role in the regulation of stem-cell-like properties in both normal and malignant cells, either by enhancing or attenuating the cell's self-renewal capacity. By binding to YAP/TAZ, Snail/SNAI2 controlled self-renewal and differentiation in skeletal stem cells [1] and mesenchymal stem cells [2]. In head and neck squamous cell carcinoma (HNSCC), by facilitating sphere formation, inducing stem-cell-related factor expression, and enhancing chemoresistance to cisplatin, SNAI2 enhanced the self-renewal capacity of cells [3]. When SNAI2 and SOX9 are co-expressed in human lung carcinoma, SNAI2 can protect SOX9 avoid from ubiquitin-mediated proteasomal degradation, which is necessary for SOX9 to promote cancer stem cells and maintain its stem-like phenotype [4]. The co-operation of SOX9 and SNAI2 can also convert luminal cells into mammary stem cells (MaSCs) with

SNAI2 Inhibited the Stem-like Phenotype in Cervical Cancer Cells
In our previous study, SNAI2 exhibited a capacity to inhibit cell growth and tumor formation in cervical cancer cells [28,29]. In this study, the clonogenic formation assay showed that in both 2D (plate clone formation experiment, Figure 1A, p < 0.05) and 3D (soft agar assay, Figure 1B, p < 0.01) cell culture systems, the number and size of cell colonies were obviously decreased in SNAI2-overexpressing SiHa cells (SiHa-SNAI2), compared with control cells (SiHa-Vec). Moreover, SNAI2 was previously reported to play a mediator role in regulating the stem-like phenotype in several types of cancer [5,10]. To assess the effect of SNAI2 on the self-renewal ability of cervical cancer cells, SiHa-SNAI2 and SiHa-Vec cells were cultured in serum-free medium under optimal conditions for growing tumorspheres. As shown in Figure 1C, SiHa-SNAI2 cells generated fewer tumorspheres and fewer cell aggregates than the more conventional tumorspheres generated from SiHa-Vec cells. Moreover, tumorspheres were generated from SiHa-Vec cells with an efficiency of 12.13%, 8.725%, and 7.40% during three consecutive passages, whereas the tumorspheres growing tumorspheres. As shown in Figure 1C, SiHa-SNAI2 cells generated fewer tumorspheres and fewer cell aggregates than the more conventional tumorspheres generated from SiHa-Vec cells. Moreover, tumorspheres were generated from SiHa-Vec cells with an efficiency of 12.13%, 8.725%, and 7.40% during three consecutive passages, whereas the tumorspheres were generated from SiHa-SNAI2 cells with an efficiency of 3.10%, 2.60% and 2.45% during three consecutive passages ( Figure 1C). . The colony formation assay was used to detect the long-term cell survival and growth of SNAI2-overexpressed SiHa-SNAI2 cells and control cells (SiHa-Vec). (B). The soft agar assay was used to detect the long-term cell survival and growth in 3D culture of SiHa-SNAI2 and SiHa-Vec cells. (C). The number of tumorspheres cells formed by SiHa-SNAI2 and SiHa-Vec cells was measured from three consecutive passages; the representative photos are shown. (D). MTT assay was used to detect the cell viability of SiHa-SNAI2 and SiHa-Vec cells by treating with different concentrations of cisplatin (3,6,12,24,48, 96 µg/mL) for 24 h. (E). MTT assay was used to detect the cell viability of SiHa-SNAI2 and SiHa-Vec cells by treating with a constant dose of cisplatin (6 µg/mL) for 0, 24, 48, or 72 h. (F). The mRNA level of SNAI2, SOX2, KLF4, NANOG, Oct3/4, ALDH1, and ALDH2 in SNAI2-modified cells was detected by real-time quantitative PCR. The protein level of SNAI2, SOX2, KLF4, NANOG, Oct3/4, ALDH1, and ALDH2 in SNAI2-modified cells (G) or SNAI2-modified cells derived mouse xenografts (H) was detected by western blotting. * p < 0.05, ** p < 0.01 vs. control.
Resistance to the chemotherapeutics of CSCs is one of the manifestations of cell selfrenewal capacity [30]. Therefore, the effects of cisplatin in SNAI2-modified cells were . The colony formation assay was used to detect the long-term cell survival and growth of SNAI2-overexpressed SiHa-SNAI2 cells and control cells (SiHa-Vec). (B). The soft agar assay was used to detect the long-term cell survival and growth in 3D culture of SiHa-SNAI2 and SiHa-Vec cells. (C). The number of tumorspheres cells formed by SiHa-SNAI2 and SiHa-Vec cells was measured from three consecutive passages; the representative photos are shown. (D). MTT assay was used to detect the cell viability of SiHa-SNAI2 and SiHa-Vec cells by treating with different concentrations of cisplatin (3,6,12,24,48, 96 µg/mL) for 24 h. (E). MTT assay was used to detect the cell viability of SiHa-SNAI2 and SiHa-Vec cells by treating with a constant dose of cisplatin (6 µg/mL) for 0, 24, 48, or 72 h. (F). The mRNA level of SNAI2, SOX2, KLF4, NANOG, Oct3/4, ALDH1, and ALDH2 in SNAI2-modified cells was detected by real-time quantitative PCR. The protein level of SNAI2, SOX2, KLF4, NANOG, Oct3/4, ALDH1, and ALDH2 in SNAI2-modified cells (G) or SNAI2-modified cells derived mouse xenografts (H) was detected by western blotting. * p < 0.05, ** p < 0.01 vs. control.
Resistance to the chemotherapeutics of CSCs is one of the manifestations of cell selfrenewal capacity [30]. Therefore, the effects of cisplatin in SNAI2-modified cells were tested in this study. After treatment with cisplatin for 24 h at different concentrations, cell viability was detected by using an MTT assay. As shown in Figure 1D, SiHa-SNAI2 cells were more sensitive to cisplatin treatment than SiHa-Vec cells when cisplatin concentrations were ≥6 µg/mL. Moreover, SiHa-SNAI2 cells were also more sensitive to cisplatin treatment than SiHa-Vec cells when cisplatin concentration was at 6 µg/mL and cultured for 24, 48, and 72 h ( Figure 1E).

Exogenous Expression of SNAI2 Attenuated the Tumorigenic Capacity of Cervical Cancer Cells In Vivo
The powerful ability to initiate tumors was thought to be an important capacity of CSCs. In our previous study, SNAI2 exhibited an inhibitory effect on tumor formation in cervical cancer in a tumor-bearing model in nude mice [29]. To further determine the potential function of SNAI2 during tumor initiation, SNAI2-modified SiHa cells and their control cells were subcutaneously injected into NOD/SCID mice by limiting dilutions to 10, 100, 1000, and 10,000 cells, and then the tumor volume, tumor initiating frequency, and incidence were monitored. As shown in Figure 2A-D, the NOD/SCID mice that were subcutaneously injected with SiHa-Vec or SiHa-SNAI2 cells formed xenograft tumors at each cell concentration (10, 100, 1000, and 10,000 cells). However, the xenograft tumors formed by SiHa-SNAI2 cells were much smaller and grew much slower than the xenograft tumors formed by SiHa-Vec cells. In addition, the average weight of the xenograft tumors derived from SiHa-SNAI2 cells at each cell dose (10, 100, 1000 and 10,000 cells) was much lighter than the xenograft tumors derived from control cells ( Figure 2E, p < 0.05).
Tumor latency was monitored after injection of SiHa-Vec and SiHa-SNAI2 cells into the NOD/SCID mice. As shown in Figure 2F, SiHa-SNAI2 cells exhibited a significantly longer tumor-free period than SiHa-Vec cells. In the SiHa-SNAI2 cell group, the shortest tumor-free period was 48 days, compared to 39 days in the SiHa-Vec cell group. The SiHa-SNAI2 cell group also exhibited a higher tumor-free rate (54.16% in SiHa-SNAI2 cells versus 4.17% in SiHa-Vec cells, p < 0.05) than the SiHa-Vec cell group.
The tumor incidence of SiHa-Vec and SiHa-SNAI2 cells is summarized in Figure 2G. The tumor-initiating frequency of SiHa-Vec was 1/286 at the lowest and 1/40.5 at the highest, compared to 1/7434 at the lowest (26.0-fold lower than SiHa-Vec cells, p < 0.05) and 1/1121.3 at the highest (27.7-fold lower than SiHa-Vec cells, p < 0.05) in SiHa-SNAI2. Therefore, the tumor formation assays revealed that the xenograft tumors derived from SiHa-SNAI2 cells grew much slower, and had a lower tumor initiating frequency, higher tumor-free rate, and longer tumor latency, compared with SiHa-Vec cells, suggesting that the tumorigenic capacity was attenuated by exogenous expression of SNAI2 in cervical cancer cells in vivo.

SiHa-EPCAM high Cells Exhibit an Enhanced Stem-Like Phenotype
In our previous study [28], SNAI2 exhibited the capacity to act as a transcriptional repressor of EPCAM in cervical cancer ( Figure 3A). Western blot ( Figure 3B) and immunocytochemical ( Figure 3C) results also verified the decreased protein levels of EPCAM in SNAI2-overexpressing SiHa cells. Moreover, flow cytometry results also revealed a lower

SiHa-EPCAM high Cells Exhibit an Enhanced Stem-like Phenotype
In our previous study [28], SNAI2 exhibited the capacity to act as a transcriptional repressor of EPCAM in cervical cancer ( Figure 3A). Western blot ( Figure 3B) and immunocytochemical ( Figure 3C) results also verified the decreased protein levels of EPCAM in SNAI2-overexpressing SiHa cells. Moreover, flow cytometry results also revealed a lower level of EPCAM-positive ratio in SiHa-SNAI2 cells compared with control cells (44.3% in SiHa-SNAI2 cells versus 93.4% in SiHa-Vec cells, Figure 3D).  Figure 3D). As the advantageous effect of EPCAM on regulating self-renewal has previously been reported in various types of cancer cells, the potential role of EPCAM in regulating selfrenewal in SiHa cells was confirmed in this study. First, EPCAM high and EPCAM low SiHa cells were sorted by FACS from SiHa cells (fluorescence-activated cell sorting, Figure 3E) and then cultured for generating tumorspheres. As shown in Figure 4F, EPCAM high cells generated many more conventional tumorspheres than EPCAM low cells. EPCAM high cells generated tumorspheres with an efficiency of 15.53%, 14.25%, and 12.51% during three consecutive passages, whereas EPCAM low cells generated tumorspheres with an efficiency of 3.50%, 4.36%, and 3.12% during three consecutive passages ( Figure 3F).
To further determine the potential effect of EPCAM on regulating tumor initiation in cervical cancer cells, EPCAM high and EPCAM low cells were subcutaneously injected into NOD/SCID mice at limited dilutions with 10, 100, 1000, and 10,000 cells, respectively, and the tumor volume and incidence were monitored. As shown in Figure 3G, the NOD/SCID mice subcutaneously injected with EPCAM high and EPCAM low cells formed palpable tumors at each cell dose (10, 100, 1000, and 10,000 cells). However, the xenograft tumors formed by EPCAM high cells grew much faster and to much larger sizes than the xenograft tumors formed by EPCAM low cells. In addition, the average weight of the xenograft tumors derived from EPCAM high cells at each cell dose (10, 100, 1000, and 10,000 cells) were much heavier than the xenograft tumors which derived from EPCAM low cells ( Figure 3H, p < 0.05).
Moreover, as shown in Figure 3I, EPCAM high cells exhibited a significantly shorter tumor-free period than EPCAM low cells. A shorter tumor-free period was observed in EPCAM high cells compared to EPCAM low cells. In EPCAM high cells, the shortest tumorfree period was 29 days, compared to 41 days in EPCAM low cells. EPCAM high cells also exhibited a lower tumor-free rate (8.33%) compared to EPCAM high cells (41.67%, p < 0.05).
The tumor initiation frequency of EPCAM high and EPCAM low cells is summarized in Figure 3J. The tumor initiation frequency of EPCAM high cells was 1/25.9 at lower levels (44.9-fold lower than EPCAM low cells, p < 0.05) and 1/3.78 at higher levels (53.9-fold lower than EPCAM low cells, p < 0.05), and the tumor-initiating frequency of EPCAM low cells was 1/1162.4 at lower levels and 1/200.12 at higher levels. Therefore, the tumor formation assays from NOD/SCID mice revealed that EPCAM high cells grew much faster, had a higher tumor initiating frequency, lower tumor-free rate, and shorter tumor latency compared to EPCAM low cells, suggesting that EPCAM high cells have a potential ability to promote tumor initiation and growth in vivo. All of these results suggested that EPCAM high cervical SiHa cells exhibit enhanced self-renewal capacity compared with EPCAM low cells, and that EPCAM expression is advantageous for maintaining self-renewal capacity in SiHa cells.

Absence of EPCAM Inhibited the Nuclear Translocation of β-Catenin in SNAI2-Overexpressing Cervical Cancer Cells
In order to explore whether the alteration of EPCAM under SNAI2-overexpressing cells was involved in SNAI2 mediating the stemness of cervical cancer cells, EPCAM high and EPCAM low cells were sorted by FACS from SiHa-SNAI2 cells ( Figure 4A), and then cultured to generate tumorspheres. As shown in Figure 4B, EPCAM high cells from SiHa-SNAI2 cells generated many more tumorspheres than EPCAM low cells from SiHa-SNAI2 cells. SiHa-SNAI2-EPCAM high cells generated tumorspheres with an efficiency of 6.13%, 7.75%, and 6.91% during three consecutive passages, whereas SiHa-SNAI2-EPCAM low cells generated tumorspheres with an efficiency of 3.25%, 2.66%, and 2.59% during three consecutive passages ( Figure 4B). Moreover, an EPCAM recombinant plasmid was used to rescue the expression of EPCAM in SiHa-SNAI2 cells. As shown in Figure 4C, SiHa-SNAI2-EPCAM cells generated many more tumorspheres than SiHa-SNAI2-Vec cells. SiHa-SNAI2-EPCAM generated tumorspheres with an efficiency of 7.53%, 6.25%, and 4.15% during three consecutive passages, whereas the SiHa-SNAI2-Vec cells generated tumorspheres was with an efficiency of 2.65%, 1.86%, and 1.79% during three consecutive passages ( Figure 4C). These results suggest that EpCMA is involved in the SNAI2-mediated stem-like phenotype in cervical cancer cells.
The Wnt/β-catenin signaling pathway plays a critical role in promoting cancer progression and maintaining cell stemness, not only in embryonic stem cells but also in many types of cancer stem cells. The interaction between EPCAM and the Wnt/β-catenin signaling pathway for maintaining cell stemness was reported in a previous study [31]. Our previous study demonstrated that SNAI2 attenuated the activity of the Wnt/β-catenin signaling pathway in cervical cancer cells [29]. Therefore, the potential interactions between SNAI2, EPCAM, and the Wnt/β-catenin signaling pathway were investigated in this study. First, the protein levels of β-catenin and the Wnt/β-catenin signaling pathway target genes c-Myc and SOX2 were detected in SNAI2-modified cervical cancer cells. As shown in Figure 4D,E, the protein levels of β-catenin, c-Myc, and SOX2 were decreased in SNAI2-overexpressing SiHa and HeLa cells compared with the control cells (SiHa-Vec and HeLa-Vec cells, p < 0.05). Conversely, the protein levels of β-catenin, c-Myc, and SOX2 were increased in SNAI2-knockdown CaSki-shSNAI2 cells compared with control cells (CaSki-shCtr cells, p < 0.05).
EPCAM comprises an extracellular domain (EpEX) and an intracellular domain (EpICD) [11]. EpICD becomes part of a large nuclear complex containing the transcriptional regulator β-catenin, which is an important component of the Wnt/β-catenin signaling pathway, and enhances the nuclear translocation of β-catenin [32]. Therefore, the protein level of β-catenin in the nucleus was detected in this study. As shown in Figure 4D, the protein level of β-catenin in the nucleus was decreased in SNAI2-overexpressing SiHa and HeLa cells compared with the control cells (SiHa-Vec and HeLa-Vec cells, p < 0.05). Conversely, the protein level of β-catenin was increased in SNAI2-knockdown CaSki-shSNAI2 cells compared with control cells (CaSki-shCtr cells, p < 0.05).
To further confirm that the reduction in c-Myc and SOX2 in SNAI2-overexpressing cells could be attributed to the attenuated activity of the Wnt/β-catenin signaling pathway, the reduction in β-catenin was recovered by transiently transfecting a recombinant βcatenin plasmid. As shown in Figure 4E, the expressions of β-catenin, c-Myc, and SOX2 were recovered by transfection with pIRES2-AcGFP-β-catenin into SiHa-SNAI2 and HeLa-SNAI2 cells. Conversely, XAV939, which is an inhibitor of the Wnt/β-catenin signaling pathway, was used to block the stimulated β-catenin expression in CaSki-shSNAI2 cells. As shown in Figure 4F, after treatment with XAV939, the expressions of β-catenin, c-Myc, and SOX2 were reduced in CaSki-shSNAI2 cells. These results demonstrated that the alteration of c-Myc and SOX2 in SNAI2-modified cells could be attributed to changes in the Wnt/β-catenin signaling pathway. Furthermore, when the protein level of EPCAM was recovered in SNAI2-overexpressed cells, β-catenin, c-Myc, SOX2, and nuclear β-catenin expression were also recovered in SNAI2-overexpressed SiHa and HeLa cells ( Figure 4G,H, p < 0.05). Similarly, the same phenomenon was observed in SiHa-Vec and HeLa-Vec cells by transiently transfecting the EPCAM recombinant plasmid ( Figure 4G,H, p < 0.05). All of these results suggested that SNAI2 blocked the nuclear translocation of β-catenin by inhibiting EPCAM expression, further suppressing the expression of c-Myc and SOX2, and ultimately attenuating the stem-like phenotype in cervical cancer cells.
Moreover, the protein levels of SNAI2, EPCAM, β-catenin, c-Myc, and SOX2 were also detected in EPCAM high -and EPCAM low -cell-derived xenografted tumor tissues. As shown in Figure 5B,D, higher expressions of EPCAM, β-catenin, β-catenin (in the nucleus), c-Myc, and SOX2 were observed in the EPCAM high cells derived xenografted tumor tissues than in the EPCAM low cell-derived xenografted tumor tissues (p < 0.05). In particular, an increased protein level of SNAI2 was observed in EPCAM high -cell-derived xenografted tumor tissues than in the EPCAM low -cell-derived xenografted tumor tissues (p < 0.05). This could be attributed to the activation of the Wnt/β-catenin signaling pathway under higher EPCAM expression. As reported in a previous study, activation of the Wnt/β-catenin signaling pathway enhanced SNAI2 expression in various cancers [33].
Additionally, the SNAI2 and EPCAM protein levels in human CESC specimens were detected using the TMA (tissue microarray) approach. The selected CESC TMA included 13 adenocarcinoma of the uterine cervix specimens (AUC), and 13 squamous carcinomas of the cervix specimens (SCC). The relationship between EPCAM and SNAI2 in these CESC specimens was examined via Pearson correlation analysis by performing immunohistochemistry (IHC). Representative images of high and low staining of SNAI2 and EPCAM are shown in Figure 5E. A negative correlation between the expression of EPCAM and SNAI2 was observed in all 26 CESC specimens ( Figure 5E, r = −0.7393, p = 0.0008) and SCC (n = 13, Figure 5E, r = −0.7564, p = 0.0028) specimens, respectively, which was consistent with our previous study and the data from the GEPIA online database [29]. Furthermore, the IRSs (immunoreactive scores) of SNAI2 and EPCAM in AUC and SCC specimens are shown in Figure 5G. The mean IRS value for SNAI2 in SCC specimens (n = 13, 7.462 ± 4.115) was much higher than that in AUC specimens (n = 13, 1.385 ± 1.445, p < 0.05). The mean IRS for EPCAM in SCC specimens (n = 13, 6.000 ± 3.266) was much lower than that in AUC specimens (n = 13, 9.769 ± 3.270, p < 0.05). Together, these results suggest that SNAI2 is negatively correlated with EPCAM in human CESC patient specimens. detected in xenografted tumor samples formed by SiHa-Vec and SiHa-SNAI2 cells. As shown in Figure 5A,C, SNAI2 exhibited a much higher protein level in SiHa-SNAI2 cellderived xenografted tumor tissues than in SiHa-Vec cell-derived xenografted tumor tissues (p < 0.05). Additionally, lower EPCAM, β-catenin, β-catenin (in the nucleus), c-Myc, and SOX2 protein levels were seen in SiHa-SNAI2-cell-derived xenografted tumor tissues compared to SiHa-Vec-cell-derived xenografted tumor tissues (p < 0.05).

Discussion
SNAI2 is well known as an EMT regulator in various carcinomas. The accumulated evidence suggests that SNAI2 exhibits biphasic effects on regulating, attenuating, or enhancing the stem-like phenotype. In our previous study, SNAI2 exhibited the capacity to inhibit cell proliferation, and promote cell motility and distant metastasis by trans-suppressing EPCAM expression in cervical cancer [28]. In addition, the results of TMA (tissue microarray) also confirmed the negative correlation of SNAI2 and EPCAM protein level in cervical cancer tissues. In this study, SNAI2 exhibited the capacity to inhibit a stem-like phenotype in cervical cancer. As shown in Figure 1, by exogenously expressing SNAI2 in SiHa cells, SiHa-SNAI2 cells exhibited inhibited cell growth, decreased tumorsphere formation, enhanced sensitivity to cisplatin, reduced stem-cell-related factor expression, a slower tumor growth rate, and a lower tumor initiating frequency, indicating that SNAI2 plays an inhibitory role in regulating a stem-like phenotype in cervical cancer cells.
EPCAM is considered a multifunctional transmembrane protein involved in the regulation of cell stemness and is required for stem cell survival or proliferation, and the inhibitory role of SNAI2 on trans-suppressing EPCAM expression in cervical cancer was reported in our previous study [28]. Thus, in cervical cancer, there is a potential connection between SNAI2 trans-suppression of EPCAM expression and SNAI2-mediated inhibition of the stem-like phenotype. Although the advantageous effect of EPCAM on regulating cell stemness has been reported in various cancers, it has been rarely studied in cervical cancer. As shown in Figure 3, the EPCAM high cells sorted from SiHa cells generated many more conventional tumorspheres, grew much faster, and had a higher tumor initiating frequency, shorter tumor latency, and lower tumor-free rate. These results indicated that EPCAM expression facilitated the maintenance of cell renewal in SiHa cells. Moreover, both the EPCAM high cells sorted from SiHa-SNAI2 cells, or transiently transfected with an EPCAM recombinant plasmid to rescue EPCAM expression in SiHa-SNAI2 cells, exhibited enhanced cell-renewal ability, suggesting that alteration of EpCMA is involved in the SNAI2-mediated stem-like phenotype in cervical cancer cells.
EPCAM comprises an extracellular domain (EpEX) and an intracellular domain (EpICD) [11]. EpICD becomes part of a nuclear complex containing β-catenin, and thus is involved in the nuclear translocation of β-catenin [32]. In this study, SNAI2 overexpression inhibited the expression of EPCAM, β-catenin, and nuclear β-catenin in SiHa and HeLa cells ( Figure 4). Conversely, when EPCAM expression was rescued in SNAI2-overexpressing cells, the protein levels of EPCAM, β-catenin, and nuclear β-catenin were recovered in SNAI2-overexpressing cells. In general, blockade of the nuclear translocation of β-catenin attenuated the activity of the Wnt/β-catenin signaling pathway [34]. These findings are in accordance with our previous study, which demonstrated that SNAI2 attenuated the activity of the Wnt/β-catenin signaling pathway in cervical cancer cells [29]. All of these results suggested that SNAI2 could inhibit the nuclear translocation of β-catenin by transsuppressing EPCAM expression, further attenuating the activity of the Wnt/β-catenin signaling pathway.
The Wnt/β-catenin signaling pathway plays a critical role in promoting cancer progression and maintaining cell stemness, not only in embryonic stem cells but also in many types of cancer stem cells. SOX2 and c-Myc are considered target genes of the Wnt/β-catenin signaling pathway, and both are also necessary for maintaining cell stemness [35][36][37][38]. In this study, a reduction in SOX2 and c-Myc expression was observed in SNAI2-overexpressing SiHa and HeLa cells. Conversely, when the expression of β-catenin or EPCAM was rescued in SNAI2-overexpred cells, the protein levels of SOX2 and c-Myc were recovered in SNAI2-overexpressing cells. Additionally, the negative correlation between SNAI2 and EPCAM, β-catenin, SOX2, and c-Myc expression in xenograft tissues from NOD/SCID mice (derived from SiHa-Vec and SiHa-SNAI2 cells) was confirmed in this study.
In conclusion, this study reveals a negative relationship between SNAI2 and EPCAM in cervical cancer. Moreover, this study suggested that blocking the nuclear translocation of β-catenin and further reducing SOX2 and c-Myc expression, which contribute to the trans-suppression of EPCAM expression by SNAI2, are responsible for attenuation of the stem-like phenotype in cervical cancer cells mediated by SNAI2 ( Figure 6).

Cell Lines and Cell Culture
The cervical cancer cell lines that were used in this study were procured from ATCC (Manassas, VA, American Type Culture Collection), including HeLa, CaSki, and SiHa.

Colony Formation Assay
The colony formed assay was used to evaluate the capacity of long-term cell proliferation and survival in 2D culture. In brief, SiHa-GFP and SiHa-SNAI2 cells were seeded as the cell density of 500 cells/well into 6-well plates, respectively. Then, the cells were in culture for 2 weeks. At the end of this experiment, the cells were washed by using PBS, fixed with paraformaldehyde (4%, for 15 min), then stained using crystal violet (0.5%, w/v, Sigma-Aldrich, St. Louis, MO, USA). The ImageJ (version 1.48 v) software was used to count the number of colonies [39].

Soft Agar Assay
The soft agar assay was used to evaluate the capacity of long-term cell proliferation and survival in 3D culture. The culture solution that contained 0.5% low-melting agarose was plated in the lower-layer of the six-well plates, then cultured at 37 • C for 4-5 h. The upper-layer with low-melting agarose to a final concentration of 0.3% was mixed with 2000 cells/well and added to 500µL of the culture medium, then cultured for 2-4 weeks at 37 • C. The ImageJ (version 1.48v) software was used to count the number of colonies.

Real-Time PCR Analysis
The experimental procedure for extracting the total RNA was followed as described in our previously study [29]. Additionally, the experimental procedures for RT-PCR were implemented as per the description of the reagent supplies' manual and a previous study [40]. The ∆∆Ct method was used to analyze the results, and the house-keeping gene was GAPDH in this study. The primer sequences which performed real-time PCR in this study were as follows:

Tumorsphere Culture
The cells were cultured in stem cell media; this medium was constituted with the DMEM/F12 basal media, and added with hEGF (EGF, 20 ng/mL), N2 (Invitrogen), bFGF (20 ng/mL, PeproTech Inc., Rocky Hill, NJ, USA), and B27 (Invitrogen) supplements. The cells were plated as 1 cell (per well) in 96-well plates and cultured in this stem cell medium for the tumorsphere formation assay. After 2-3 weeks, the number of tumorspheres were counted. The tumorspheres were harvested and disaggregated with 0.25% trypsin/EDTA, then plated as 1 cell (per well) in 96-well plates, and cultured in this stem cell medium for serial tumorsphere formation assays.

Cisplatin Resistance and MTT Assay
For cisplatin sensitivity assays, 10 4 cells per well were plated in 96-well plates and the various concentrations of cisplatin were added (the concentrations of cisplatin were 0, 3, 6, 12, 24, 48, or 96 µg/mL) the next day. Then, the cells were exposed for 24 h, and the viability of such cells was measured by using MTT assays. Additionally, in a separate experiment, the cells were cultured at 6 µg/mL for 24, 48, and 72 h as a constant concentration of cisplatin, and then the MTT assays was used to detect the cell viability. The experimental procedure for MTT was followed as described in our previous study [28].

Tumor Xenograft Assay
The SiHa-Vec, SiHa-SNAI2, EPCAM high , and EPCAM low cells were resuspended in 200 µL culture solution, which was mixed with DMEM/F12 basal media and Matrigel (#356234, BD, Biosciences) at 1:1, respectively. Then, it was injected subcutaneously into the female NOD/SCID mice (6 to 8-weeks old): left (SiHa-Vec or EPCAM low cells) and right (SiHa-SNAI2 or EPCAM high cells). All of the mice were fed in the SFP room, and the mice were observed every 3 days, and the length (a) and the width (b) of the tumors were recorded. The volume of the tumors was calculated as V = abb/2.
At the end of this experiment, all of the tumor samples were collected, weighed, and immunostained. The limiting-dilution analysis was used to evaluate the upper and lower limits of tumor formation, which represented the frequency of tumorigenic cells [41].

Plasmid and Transfection
The pIRES2-EPCAM [28] and pIRES2-AcGFP-β-catenin [42] plasmids were kept in our laboratory. In this study, the Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) was used to transiently transfect the plasmids in the cells that were used in our study. Additionally, the protocol was performed as per the description of the reagent supplies manual. Then, the proteins of these cells were collected and used for Western blotting.

EPCAM high Cells and EPCAM low Cells Isolation and Flow Cytometry Analysis
The antibody of EPCAM (1:100, sc-25308, Santa Cruz, CA, USA) in accordance with manufacturer's instructions were used for flow cytometry and FACS to measure the expression of EPCAM in human cervical cells and xenografts, and sort EPCAM high cells and EPCAM low cells. Flow cytometry was performed with a FACS Calibu, and FACS was performed with a BD FACSAria II cell sorter (Becton Dickinson). The data were analyzed by FlowJo 7.6 software (Tree Star Inc., Ashland, OR, USA).

Immunocytochemistry, Immunohistochemistry (IHC) and Tissue Microarray
The experimental procedures for immunocytochemistry and immunohistochemistry that were used in this study were followed as described in our previous study [28]. The cervical cancer tissue microarray (Cat No. F261301, Bioaitech, Xi'an, China) was used to analyze the correlation between the expression of EPCAM (1:100 dilution) and SNAI2 (1:50 dilution). Then, the IHC score of EPCAM and SNAI2 was analyzed by the Pearson correlation analysis (all of the data have been tested for normality of distribution using GraphPad Prism 9.0).

Statistical Analysis
SPSS (version 19.0, Inc., Chicago, IL, USA) and GraPhpad Prism 9.0 (GraphPad Software, LLC, USA) software was used for statistical analysis of the data in this study. Continuous data were compared using Student's t-tests, and data are expressed as mean ± SD. Pearson correlation analysis was used to analyze the correlation among protein expression. Survival free interval curves were calculated according to the Kaplan-Meier method, and analysis was performed using the logrank test.
Author Contributions: N.C. designed the research. Q.C., Q.F. and Y.Z. performed the animal study. X.L., N.Z., Q.F., Y.Z., Q.C., X.Y. and Z.W. performed research, analyzed data, and developed software necessary to perform and record the work. N.C. and H.L. wrote the paper. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by grants to Nan Cui from the National Natural Science Foundation of China (No. 81903042). This work was supported by grants to Hongbao Li from the National Natural Science Foundation of China (No. 82170443). This work was supported by grants to Zhiqiang Wang from Gansu Provincial Science and Technology Department Grant (No. 20JR5RA138).

Institutional Review Board Statement:
The animal experiments that were performed in this study were according to the guiding idea by the Medical School of Xi'an Jiaotong University (G-222). The Ethics Committee of the Animal Care and Use Committee of the Medical School of Xi'an Jiaotong Universit ratified this study.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.