STON2 negatively modulates stem-like properties in ovarian cancer cells via DNMT1/MUC1 pathway

The human epithelial ovarian cancer cell line, Caov3, was obtained from the American Type Culture Collection (ATCC, Manassas, Virginia, USA). The human ovarian adenocarcinoma cell line 3AO was acquired from the Women’s Hospital, School of Medicine, Zhejiang University, where it was tested and authenticated. It was not cultured continuously for more than 3 months. Adherent Caov3 cells were cultured in their standard medium as recommended by ATCC. 3AO cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium (BI, Kibbutz Beit-Haemek, Israel), supplemented with 10% fetal bovine serum (FBS) (Invitrogen, New York, USA), maintained at 37 °C in 5% CO₂ and detached using trypsin/EDTA solution. The anchorage-independent spheroids were generated from Caov3 or 3AO cells after plating, 5 × 10⁴ cells per well, in ultra-low attachment 6-well culture plates (Corning, New York, USA) and culturing in a SFM composed of Dulbecco’s modified Eagle’s medium (DMEM)/F12 (BI, Kibbutz Beit-Haemek, Israel), 10 ng/mL basic fibroblast growth factor, 20 ng/mL epidermal growth factor (PeproTech, Rocky Hill, USA), 1 mg/mL insulin (Sigma-Aldrich, Burlington, MA, USA), 10 μL/mL B27 additive (Life Technologies, Carlsbad, California, USA), 100 U/mL penicillin, and 100 μg/mL streptomycin. Fresh medium was added after every two or three days and the cultures were maintained at 37 °C in 5% CO₂ for 7 days. During the passaging of spheroids, the cells were centrifuged and digested with trypsin, and cultured in SFM.

After plating 400 cells per well in ultra-low attachment 96-well culture plates (Corning, New York, USA), Caov3 or 3AO cells were cultured in a SFM composed of DMEM/F12 (BI, Kibbutz Beit-Haemek, Israel), 10 ng/mL basic fibroblast growth factor, 20 ng/mL epidermal growth factor (PeproTech, Rocky Hill, USA), 1 mg/mL insulin (Sigma-Aldrich, Burlington, MA, USA), 10 μL/mL B27 additive (Life Technologies, Carlsbad, California, USA), 100 U/mL penicillin, and 100 μg/mL streptomycin. Fresh medium was added after every two or three days and the cultures were maintained at 37 °C in 5% CO₂ for 7 days.

Parental and spheroids of 3AO cells (three independent samples in each group) were cracked in 200 μl lysis buffer, disrupted with agitation, boiled for 10 min, and centrifuged at 12000 rpm for 10 min. The supernatant was then collected and digested overnight at 37 °C with trypsin (Promega, Madison, Wisconsin, USA). The peptide of each sample was desalted on C18 cartridges (Sigma, Burlington, MA, USA) and then resuspended in 0.1% trifluoroacetic acid after centrifugation. MS analysis was performed on a Finnigan LTQ VELOS MS (ThermoFisher, Waltham, MA, USA). Each scan cycle included one full MS1 scan in profile mode and 20 MS2 scans in centroid mode. The peptides were investigated automatically according to the MS/MS spectra in the International Protein Index (IPI) human protein database using the Bioworks Browser software suite (Thermo, Waltham, MA, USA). The quantification of peptides was performed using the DeCyder differential analysis software (GE Healthcare, Munich, Germany).

FLAG-tagged full length open reading frames (ORFs) of human STON2 (NM_033104.3) were ligated into a pcDNA3.1 vector (BioVector, Beijing, China) after HindIII and XhoI digestion. HA-tagged human MUC1 (NM_001204286.1) was cloned into a pcDNA3.1 vector (BioVector, Beijing, China) after HindIII and EcoRI digestion. The sequences of all the constructs were verified by DNA sequencing. The primer sequences are listed in Additional file 1: Table S1. Transfections were performed using X-tremeGENE HP DNA transfection reagent (Roche, Basel, Switzerland) following the manufacturer’s protocols.

The short-hairpin RNAs (shRNAs) for STON2 and MUC1 oligonucleotides were cloned into hU6-MCS-CMV-puromycin lentivirus expression vectors between the AgeI and EcoRI sites and then transfected into adherent cells according to instructions. After 72 h of transfection, the cells were selected using 5 μg/ml puromycin for 4 days. All construct sequences were confirmed using DNA sequencing. The shRNA oligonucleotides are listed in Additional file 2: Table S2.

Small-interfering RNAs (siRNAs) against STON2 and DNMT1, along with RNAi negative controls, were purchased from Genepharma (Shanghai, China). Cells were transfected with the siRNAs (50 nM) using Lipofectamine™ RNAiMAX (ThermoFisher, Waltham, MA, USA) following the manufacturer’s protocols. The siRNA sequences are listed in Additional file 2: Table S2.

Total RNA was extracted using an RNA extraction kit (TaKaRa, Dalian, China) and reverse-transcribed into cDNA using the reverse transcription cDNA kit (TAKAR, Dalian, China). PCR reactions were performed using SYBR® Premix Ex Taq™ (TaKaRa, Dalian, China) and applied biosystems 7900HT fast real-time PCR system (Life Technologies, Carlsbad, California, USA). The relative mRNA expression was calculated using the 2^-∆∆Ct method and normalized to GAPDH expression. Primer sequences are shown in Additional file 1: Table S1.

Cells were lysed using a radioimmunoprecipitation assay (RIPA) lysis buffer (Beyotime, Shanghai, China) supplemented with PMSF inhibitor (Beyotime, Shanghai, China). Protein lysates were loaded and separated on a 10% sodium dodecyl sulfate polyacrylamide gel and transferred onto 0.22-μm polyvinylidene fluoride (PVDF) membranes (Bio-Rad, Hercules, California, USA). The membranes were then blocked with Tris buffered saline Tween 20 (TBST) containing 5% non-fat milk for 1 h, and probed with primary antibodies overnight at 4 °C. They were then washed thrice with TBST for 10 min each and probed with secondary antibodies for 1 h, followed by washing three times in TBST for 10 min per wash. The bands were visualized using an enhanced chemiluminescence (ECL) kit (ThermoFisher, Waltham, MA, USA) in Image quant LAS400 mini (GE Healthcare, Munich, Germany). Primary antibodies against STON2 (Santa, Dallas, Texas), MUC1 (ThermoFisher, Waltham, MA, USA), NANOG (Sigma, Burlington, MA, USA), c-MYC (CST, Danvers, MA, USA), DNMT1 (Abcam, Cambridge, MA, USA), DNMT3A (Abcam, Cambridge, MA, USA), DNMT3B (Abcam, Cambridge, MA, USA), E-cadherin (Abcam, Cambridge, MA, USA), N-cadherin (CST, Danvers, MA, USA), and fibronectin (Abcam, Cambridge, MA, USA) were used. GAPDH (Abcam, Cambridge, MA, USA) was the loading control.

Spheroids were centrifuged and digested with trypsin, after which 1 × 10⁵ cells were counted and resuspended in 100 μl PBS, stained with either anti-CD44-FITC (eBiosciences, Vienna, Austria), anti-CD24-APC (eBiosciences, Vienna, Austria) antibodies or isotype control antibodies (eBiosciences, Vienna, Austria), and incubated for 30 min at room temperature following the manufacturer’s protocol. They were then detected using a FC 500 series flow cytometer (Beckman Coulter, Brea, CA, USA), and the data was subsequently analyzed using the CXP 2.1 software.

3AO cells transfected with shSTON2 lentivirus or control shRNA were cultured in a SFM for 7 days. Total RNA was prepared from spheroids using an RNeasy mini kit (Qiagen, Dusseldorf, Germany) according to the manufacturer’s instructions. RNA quality was evaluated using a RNA Nano 6000 assay kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). High-quality RNA from shSTON2 and control groups was used for transcriptome sequencing (three independent samples for each group). The libraries were sequenced on an Illumina HiSeq platform using a 125 bp/150 bp paired-end sequencing strategy. The original image data generated by the sequencing machine was converted into sequence data via base calling using a TruSeq PE cluster kit v3-cBot-HS (Illumia, San Diego, CA, USA). These were then subjected to standard quality controls. Clean data was obtained by removing reads containing adapter, poly-N, and low quality reads from the raw data. Next, the Q20, Q30, and GC contents of the clean data were calculated. Paired-end clean reads were aligned to the reference genome using STAR. Read numbers mapped to each gene were counted by HTSeq v0.6.0. Differential expression analysis of two conditions was performed using the DESeq2 R package (1.10.1). The resulting P-values were adjusted using the benjamini and hochberg’s approach for controlling the false discovery rate (FDR). Genes with an adjusted P-value< 0.05, as identified by DESeq2, were considered to be differentially expressed.

Genomic DNA was extracted using a QIAamp DNA mini kit (Qiagen, Dusseldorf, Germany). One microgram of genomic DNA was bisulfite-modified using an EpiTect bisulfite kit (Qiagen, Dusseldorf, Germany) following the manufacturer’s protocol. The modified DNA was used as a template for subsequent PCR. Primers for PCR amplification and sequencing were designed using the pyromark assay design 2.0 software (Qiagen, Dusseldorf, Germany). A Hotstar Taq DNA polymerase PCR kit (Qiagen, Dusseldorf, Germany) was used for the experiments under the following conditions: 95 °C 3 min; 40 cycles of 94 °C 30 s, 52 °C 30 s, 72 °C for 1 min; 72 °C for 7 min. The acquisitions were assessed using gel electrophoresis and pyrosequenced with pyromark Q96 ID (Qiagen, Dusseldorf, Germany). The primers for pyrosequencing are shown in Additional file 1: Table S1.

With the permission of the patients and the ethical committee of the Women’s Hospital, School of Medicine, Zhejiang University, 165 formalin-fixed and paraffin-embedded tissue samples were obtained from patients that had been diagnosed with ovarian serous, mucinous, endometrioid, clear-cell and mixed carcinomas from January 2002 to December 2009. These were used for immunohistochemical analyses. All patients underwent primary surgery, followed by paclitaxel-based chemotherapy. The deadline of the follow-up was July 30, 2017. As 20 patients failed to re-contact, only the remaining 145 patients were used for the calculation of progression-free survival (PFS) and overall survival (OS). All pathological diagnosis was reconfirmed blindly by an expert pathologist. The section slides were counter-stained with haematoxylin, dehydrated and mounted. The details were described as previously [25]. Anti-STON2 (Abcam, Cambridge, MA, USA) (1:200) and anti-MUC1 (Abcam, Cambridge, MA, USA) (1:500) were used for IHC. Five microscope fields were picked for evaluating the semiquantification of STON2 and MUC1 staining. The intensity of immunostaining was graded as follows: 1+, weak; 2+, moderate; 3+, strong or 4+, very strong (Additional file 3: Figure S1). The area of positive cancer cells in was categorized as follows: 1+, 0 to 25%; 2+, 25 to 50%; 3+, 50 to 75% or 4+, 75 to 100%. The score for each microscopic field was obtained by multiplying the two parts of score, The sum was acquired by adding the scores of the five microscopic fields. The sum from 0 to 42 was assigned as “low expression”, from 43 to 80 was assigned as “high expression” [26].

All animal experiments were performed in accordance with Animal Research Reporting In Vivo Experiments (ARRIVE) guidelines for the use of laboratory animals and were approved by the ethics committee of Zhejiang University. NOD/SCID mice (female, 4–6-week-old) were purchased from the Chinese Academy of Medical Sciences (Beijing, China) and randomly assigned to each treatment group. For experiments, 1 × 10⁵ or 1 × 10⁴ 3AO spheroids transfected with shSTON2 or shNC were resuspended in 100 μl PBS and injected into the right flank of mice, which were then monitored weekly for 4 weeks. The tumor volumes were calculated according to the formula V = a × b² × 0.5, where a is the largest diameter and b is the smallest diameter of the tumor.

Differences in clinicopathological characteristics were assessed using the chi-squared test. Cumulative survival probabilities were calculated using the Kaplan-Meier method. Survival rates were compared using the log-rank test. Two-tailed Student’s t-tests were used to perform a statistical comparison between two groups unless otherwise indicated. Statistical tests were performed using the SPSS software, version 20.0 (SPSS Inc.) or with GraphPad Prism 6.0 (GraphPad Software, Inc.). The level of statistical significance was set at *P < 0.05, **P < 0.01, ***P < 0.001.

The ovarian cancer cells were cultured in suspension culture conditions to enrich the spheroids [9]. As shown in Fig. 1a, spheroids with considerably increased CD44⁺CD24⁻ subpopulation are characterized as CSCs and referred to as OCSCs for the subsequent experiments. Using proteomics, we analyzed the differential protein expression between parental cells and spheroids from 3AO cells. The 2-dimensional image generated by LC-MS/MS analysis (Fig. 1b) revealed that 685 peptides and 187 nonredundant peptides had been differentially expressed between parental cells and spheroids (FDR < 0.05, Additional file 4: Table S3). The heat map of Fig. 1c shows 74 of the most differentially expressed proteins in DeCyder differential analysis (FDR < 0.05, fold change ≥2 or ≤ 0.5). Notably, qPCR revealed a lesser STON2 expression in the spheroids than in the parental cells (Fig. 1d). Western blotting further validated the significantly lower expression of STON2 protein in the spheroids than in parental 3AO and Caov3 cells (Fig. 1e).

To explore the function of STON2 in ovarian cancer, a shRNA was used to suppress STON2 expression in 3AO and Caov3 cells cultured in SFM for 7 days. STON2 protein levels were successfully inhibited by specific shRNA (Fig. 2a). We observed that STON2 knockdown up-regulated the expression of stem cell related-markers such as NANOG and c-MYC [27] (Fig. 2a) and that spheroid numbers were significantly higher in OCSCs than in shNC cells (Fig. 2b). Flow cytometry analysis showed that the proportion of CD44⁺CD24⁻ cells had also increased (Fig. 2c). We quantified both the protein and mRNA levels of EMT-related key factors, including E-cadherin, N-cadherin, and fibronectin, all of which revealed significant differences in expressions between shSTON2 and shNC cells, except the N-cadherin mRNA expression in Caov3 line (Fig. 2d, e).

Next, we considered whether STON2 overexpression negatively affects OCSCs. 3AO and Caov3 cells were transfected with STON2 plasmids and cultured in SFM for 7 days. Cells transfected with the empty vectors acted as controls. Analysis of the expression of CSC-related proteins showed that their expression levels were reduced in cells showing STON2 overexpression (Fig. 3a). STON2 overexpression also significantly inhibited spheroid number (Figs. 3b) and the proportion of cells with the CD44⁺CD24⁻phenotype in the 3AO cell line, but not in the Caov3 cell line (Fig. 3c). It also altered the expression of EMT-related key factors (Fig. 3d), as compared to those of the control groups.

We then investigated the effect of STON2 knockdown on ovarian tumorigenicity. We performed xenograft assays using different gradients of CSC subpopulations enriched from 3AO-shNC or shSTON2 cells. As shown in Fig. 2f, by 4 weeks after injection, 1 × 10⁴ shSTON2 cells had efficiently generated large tumors in all mice (n = 5), but equal numbers of shNC cells had produced only three tumors among five injected NOD/SCID mice. These data indicated that STON2 knockdown enhanced the CSC subpopulation-induced tumorigenicity in vivo. In addition, the tumor diameters of mice injected with 1 × 10⁵ cells were measured weekly for 4 weeks after implantation. As shown in Fig. 2g, the volume of tumors in the shSTON2 groups was larger than those in the shNC groups.

As STON2 modulates stem-like properties in ovarian cancer cells, we attempted to determine the downstream targets of STON2. Total RNA was extracted from spheroid cells treated with shNC or shSTON2, and the differential mRNA expression was detected using RNA-seq (three independent samples in each group). DESeq2 software analysis showed that a series of genes was differentially expressed (< 0.05-Padj) (Fig. 4a). We focused on MUC1, which was shown to be particularly up-regulated (− 0.0000446-Padj) and has been demonstrated to be related to stemness in breast cancer [28] and pancreatic cancer [29].

We determined MUC1 expression in 3AO and Caov3 cells using qPCR and western blotting. Results showed that the expression levels of MUC1 mRNA and protein were elevated after STON2 knockdown (Fig. 4b, c), whereas, STON2 expression remained unaltered upon MUC1 knockdown or over-expression (Fig. 4d, e). This suggests that MUC1 acts downstream of STON2 in ovarian cancer cells.

We also analyzed the expression of MUC1 in parental as well as in cancer cell derived stem cells. Western blotting showed a higher expression of MUC1 associated with a lower expression of STON2 in spheroids, as compared to those in parental cells (Fig. 4f). We also employed a highly aggressive ovarian cancer cell line Skov3 [30] to examine MUC1 expression. As expected, a much higher expression of MUC1 was associated with a lower expression of STON2 in the Skov3 cells, as compared to those in Caov3 cells (Fig. 4g).

Further random examinations of specimen sections from the paraffin-embedded blocks of ovarian cancer patients confirmed that STON2 and MUC1 expression correlated negatively (Fig. 4h).

Previous studies have shown that MUC1 expression is regulated by DNA methylation [24, 31]. Hence, we treated 3AO and Caov3 cells with different concentrations of DNA methylation inhibitor azacitidine (Aza) for 48 h. An Aza dosage-dependent increase in MUC1 expression was observed (Fig. 5a). We then transfected a STON2-specific siRNA or control siRNA in 3AO and Caov3 cells and incubated them for 24 h. This was followed by treatment with Aza or control for 48 h. Results showed that Aza was able to partially block MUC1 up-regulation post-STON2 knockdown (Fig. 5b), suggesting that STON2 knockdown-induced MUC1 up-regulation might occur via a reduction in MUC1 promoter methylation. We further identified one CpG-rich region in the MUC1 promoter (Fig. 5c) and designed primers to analyze the CpG-rich region using a PyroMark assay design. 3AO and Caov3 cells were treated with Aza for 24 h after transfection with the STON2-specific siRNA or control siRNA. The DNA methylation levels of the specific CpG nucleotides were then detected using pyrosequencing. Results showed that the extent of DNA methylation at multiple sites in the siSTON2 group was less than that of the control group (Fig. 5d and Additional file 5: Figure S2).

DNMT1, a key member of the DNA methyltransferase family [32], is known to maintain methylation patterns and suppress transcription [33]. Our results showed that STON2 knockdown decreased DNMT1 but increased MUC1 expression (Fig. 5e). We also observed that STON2 knockdown did not alter DNMT1 mRNA levels but changed DNMT1 protein levels (Fig. 5e, f). Growing evidence suggests that the stability of DNMT1 is regulated by post-translational modifications such as phosphorylation, acetylation, methylation, and ubiquitination [34‐36]. Using specific siRNAs, we confirmed that the STON2 knockdown-induced reduction in DNMT1 could be partially rescued by the proteasome inhibitor MG132, and that a time-dependent increase in MUC1 expression was observed (Fig. 5g). In addition, both the mRNA expression and protein level of MUC1 were increased by DNMT1 knockdown (Fig. 5h, i). Correspondingly, the MUC1 promoter DNA methylation levels of specific CpG nucleotides at multiple sites in the siDNMT1 group were lower than those in the control group (Additional file 6: Figure S3). These results suggest that, for cancer stem cells, STON2 functions at the post-transcriptional level via the DNMT1/MUC1 pathway.

Finally, we constructed shRNA to suppress MUC1 in 3AO and Caov3 cells cultured in SFM. We analyzed the relative expression levels of CSC-related proteins, NANOG and c-MYC. The results showed that MUC1 knockdown significantly reduced the expression level of NANOG in Caov3 (p < 0.05), but not in 3AO (Fig. 6a). The reduction of c-MYC expression was not significant in 3AO or Caov3 cells (Fig. 6a). MUC1 knockdown also significantly reduced CD44⁺CD24⁻phenotypes (Fig. 6b). Conversely, MUC1 overexpression significantly up-regulated both NANOG (p < 0.05) and c-MYC (p < 0.01) (Fig. 6c) and increased the proportion of cells with the CD44⁺CD24⁻ phenotype (Fig. 6d). Furthermore, we observed that the STON2 overexpression-induced reduction of sphere-forming ability was partially reversed by MUC1 overexpression (Fig. 6e).

Using MUC1 knockdown or overexpression in the 3AO cell line, we detected the expression levels of EMT-related key factors by western blotting. As expected, the results showed that the expression of E-cadherin increased, while N-cadherin and fibronectin decreased in shMUC1 cells, as compared to those of the shNC cells (Fig. 6f). MUC1 overexpression up-regulated the expression of N-cadherin and fibronectin, but did not significantly change E-cadherin, as compared to those of the control groups (Fig. 6f). However, we failed to co-express the shSTON2 and shMUC1 in 3AO cells. Based on previous results that the EMT-related protein expression was regulated by shSTON2 (Fig. 2d), it is conceivable that MUC1 participates in STON2 induced changes in EMT-related markers.

We firstly investigated the correlation between STON2 expression and clinical parameters based on immunohistological examination of tissue samples from patients. Kaplan–Meier survival analysis failed to reveal any correlation of STON2 expression with progression-free survival (PFS) or overall survival (OS) in our patient cohort (Additional file 7: Figure S5A). However, significantly lower STON2 immunoreactivity was detected in poorly differentiated ovarian cancer (P = 0.000, Additional file 8: Table S4). We also analyzed the correlation of MUC1 expression with clinicopathological characteristics, and prognosis of ovarian cancer. Results showed that significantly higher expression of MUC1 was associated with FIGO stage (P = 0.004), tumor grade (P = 0.01), serum CA125 (P = 0.005), and PFS (P = 0.006), OS (P = 0.019) (Additional file 9: Table S5 and Fig. 7a).

IHC analysis of the co-expression relationships between STON2 and MUC1 revealed that the prognostic value of STON2 expression was highly dependent on MUC1 expression, in which the high expression of STON2 twinned with a low expression of MUC1 was associated with a favorable prognosis (Fig. 7b). A trend was also revealed in that patients expressing high STON2 and low MUC1 were associated with a best survival of all the remaining patients (PFS, P = 0.095, and OS, P = 0.06). However, with a low expression of STON2, there was no prognostic value of MUC1 expression (Additional file 7: Figure S5B). Taken together, these observations provide evidence that STON2 acts as a negative modulator via the MUC1-mediated pathway in ovarian cancer.