The study was approved by the Ethical Review Board of the Graduate School of Medicine, Osaka University (approval no. 15234), and was performed in accordance with the committee guidelines and regulations. We examined 47 patients undergoing surgery for EC at Osaka University Hospital from 2011 to 2014. All patients provided written informed consent. The clinicopathological features of the enrolled patients were shown in Table I. Resected specimens were fixed in 10% formalin and processed for paraffin embedding. Specimens were stored at room temperature in a dark room, sectioned at 4 µm thickness, and subjected to immunohistochemical analysis.

The human EC cell lines HEC-1B and HEC108 were obtained from the Health Science Research Resources Bank of Osaka, Japan. Cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM)-High glucose supplemented with 10% FBS, penicillin (100 IU/ml), and streptomycin (100 µg/ml) and maintained at 37°C in 5% CO₂.

The antibodies against ALDH1A1 and β-actin were used as previously reported (13). The antibody against GPM6B (HPA002913, Sigma-Aldrich) was used for immunohistochemistry (dilution at 1:200). The antibody against GFP (#2956, Cell Signaling Technology) was used for immunoblotting (dilution at 1:500).

The plasmids Empty-EGFP (pRP-EGFP-CMV) and GPM6B [pRP-EGFP-CAG-FLAG/3×GGGGS/hGPM6B (NM_001001995.3)] were obtained from Vector Builder, Inc (Chicago, IL, USA).

The ALDEFLUOR kit (STEM CELL Technologies) was used according to the manufacturer's instructions. Cells were analyzed by FACS CantoII and AriaII flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). Data analysis was perfomed using Cell Quest software (BD Biosciences).

The ALDEFLUOR kit and PKH26 Red Fluorescent Cell Linker Kit (SIGMA-ALDRICH) were used according to the manufacturer's protocols. Fluorescence signals were visualized using fluorescence microscope (BZ-8000, KEYENCE, Osaka, Japan).

For generation of HEC-1B subline cells, original cells were treated with threefold diluted plasma-activated medium (PAM) which was culture medium irradiated by non-thermal plasma device that containing of gas, electrons, ions, radicals, and ultraviolet light as previously described (14). PAM induced apoptosis in a large number of cells due to production of reactive oxygen species (ROS) such as H2O2. Therefore, we selected the single cell survived from ROS using microscope. After the selected cell was cultured in 96-well plate, cells proliferated in 6-well plate. The same procedure was repeated once more to prepare HEC-1B subline cells.

Total RNA was extracted using the miRNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. cDNA libraries were constructed using the TruSeq Stranded mRNA Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer's protocol. Sequencing was undertaken on the Illumina HiSeq 2500 platform in 75-base single-end mode. Casava version 1.8.2 software (Illumina) was used for base calling. The sequenced reads were mapped to a human reference genome sequence (hg19) using TopHat version 2.0.13 (http://ccb.jhu.edu/software/tophat/index.shtml), Bowtie2 version 2.2.3 (http://bowtie–bio.sourceforge.net/bowtie2/index.shtml), and SAMtools version 0.1.19 (http://samtools.sourceforge.net/). The fragments per kilobase of exon per million mapped fragments values were calculated using Cuffnorm version. 2.2.1 (http://cole–trapnell–lab.github.io/cufflinks/) to identify upregulated (2.0-fold, P<0.05) and downregulated (−0.5-fold, P<0.05) genes.

GPM6B in HEC-1B was disrupted using the TrueGuide™ CRISPR/Cas9 system (Invitrogen, Carlsbad, CA, USA), in accordance with the manufacturer's instructions. The crRNA (A35509, CRISPR714997_CR, sequence: GUGUUGCUCAAGAAUCGCCA, target exon: exon2, Invitrogen) was annealed with TrueCut™ tracrRNA (Invitrogen). HEC-1B cells was co-transfected with the gRNA (crRNA:tracrRNA duplex) and TrueCut™ Cas9 Protein v2 (Invitrogen) using Lipofectamine™ CRISPRMAX™ Cas9 Transfection Reagent (Invitrogen). Single cell clones were then isolated by using limiting dilution cloning in 96-well plates. The positive clones were confirmed the absence of GPM6B by PCR of genomic sequence. Untransfected HEC-1B parent cells (WT) was used as a negative control.

The plasmid Empty-GFP was transfected into HEC-1B and HEC108 cells using Lipofectamine 3000 reagent (Thermo Fisher Scientific). The plasmid GPM6B-GFP was transfected into HEC-1B (GPM6B-KO) and HEC108 cells. Green population was sorted with Cell Sorter SH800ZDP (SONY, Tokyo, Japan). When colonies formed after passage, we picked up different colonies and named OE1 and OE2. Untransfected HEC108 parent cells (WT) was used as a negative control. HEC1B and HEC-108 transfected with empty vector (EV) was used as another negative control.

The RT-qPCR was performed with StepOnePlus™ Real-Time PCR instrument (Applied Biosystems, Foster City, CA) using Taqman probe/primer sets specific for human GPM6B (Hs01041077_m1). GAPDH was used as a reference for gene amplification (Applied Biosystems).

Cells were rinsed three times with PBS and lysed in 500 µl of lysis buffer [1×SSC (418 µl), 1M Tris-HCl (pH 7.5) (5 µl), 0.5M EDTA (pH 8.0) (1 µl), 10% SDS (50 µl), 20 mg/ml Proteinase K (FUJIFILM) (25 µl), 10 mg/ml RNase A (Invitrogen) (1 µl)]. The samples were mixed by vortexing and centrifugation at 15,310 g for 10 min at room temperature. After centrifugation at high speed, the upper phase was carefully removed and transferred to a new tube. A mixture of Phenol: Chloroform: Isoamyl Alcohol (Nacalai) was added in equal volumes to samples and the samples were mixed gently and the aqueous layer was transferred into a new tube. After ethanol precipitation, precipitated DNA was dissolved in 30 µl of sterile 1×TE buffer (pH 8.0). Extracted genomic DNA was amplified with KOD FX Neo (TOYOBO). The following primers which include protospacer adjacent motif were used to amplify DNA:

(Forward) 5′-CCGTGGCGATTCTTGAGCAAC-3′

(Reverse) 5′-ATGCCCTGGGATCTGCTCTTC-3′

The following primers were used to check the disrupted alleles:

Human GPM6B:

Exon 2:

(Forward) 5′-ACTGCTCTGCCATTCACTACCCTTCCAG-3′

(Reverse) 5′-ACGCACCACCACGCCCAGCTAAATTTTT-3′.

PCR was done on the T100 Thermal Cycler (Bio-Rad, USA). The PCR amplification consisted of an initial denaturation for 2 min at 94°C, followed by 5 cycles of denaturation (10 sec, 98°C) and extension (30 sec, 74°C), 5 cycles of denaturation (10 sec, 98°C) and extension (30 sec, 72°C), 5 cycles of denaturation (10 sec, 98°C) and extension (30 sec, 70°C), 30 cycles of denaturation (10 sec, 98°C) and extension (30 sec, 68°C). The final extension step was carried out at 68°C for 7 min. Its analyzing was performed using 1.5% agarose gel electrophoresis and visualized using gel documentation (AE-9000 E-Graph, ATTO, Japan).

Immunohistochemical staining was conducted by the Dako Autostainer Link 48 + (Dako/Agilent Technologies, Inc.) according to the manufacturer's instructions. Primary antibodies are incubated for 30 min at room temperature.

Studies were performed as previously reported (13). LAS-4000 Image Analyzer (GE Healthcare, Chicago, IL, USA) or ChemiDoc Touch (Bio-Rad) were used for the detection of antibody reaction. The expression of β-Actin was used as a loading control.

Statistical analyses were performed using JMP Pro 14 software (SAS Institute). In vitro experiments were performed at least two times. The data are presented as means ± standard error of the mean of independent experiments. The significance of the differences was determined using Mann Whitney U test. The log-rank test was used for survival analysis. Kaplan-Meier survival plots were made by using GraphPad Prism 9. P<0.05 was considered to indicate a statistically significant difference.

To validate the tumor cell plasticity, we conducted Aldefluor assay using EC cell line HEC-1B. After culturing ALDH1A1-low cells, the cell distribution was analyzed by the assay. ALDH1A1-low population spontaneously yielded ALDH1A1-high population (Fig. 1A). Next, we stained ALDH1A1-low cells red using the PKH26 Red Fluorescent Cell Linker Kit and conducted Aldefluor assay. The percentage of cells that turned yellow in the total cell number was defined as Plasticity index (Fig. 1B). Furthermore, we found that the mixture of ALDH1A1-high population sometimes accelerated the plasticity index and observed changes in tumor cells with low ALDH1A1 activity under a fluorescence microscope (Fig. 1C and D). Many color changes were observed in the adhesive areas between tumor cells (Fig. 1D).

We speculated that physical contact between cells contributed to the activation of ALDH1A1. First, we generate some sublines of HEC-1B cells using PAM and co-cultured ALDH1A1-high cells and ALDH1A1-low cells by a method of culturing two types of cells in the same dish and a method of culturing the two types of cells so that they do not contact with each other via transwell (Permeable Supports 3.0 µm Polycarbonate membrane, Corning) (Fig. 2A). In subline A, the former method had a higher plasticity index than the latter method, however, there was no difference in subline B (Fig. 2B). And then, we extracted RNAs of subline A and B. We compared the RNA expression of both cells and focused on GPM6B as the candidate mediating tumor cell plasticity (Fig. 2C; Tables II and SI). Next, we performed immunohistochemistry analysis of ALDH1A1 and GPM6B in clinical samples of EC tissues and found that GPM6B tended to express in the border of ALDH1A1 expressing tumor cells and non-expressing tumor cells (Fig. 2D).

To examine the significance of GPM6B in EC, we disrupted the GPM6B gene in HEC-1B cells, using the CRISPR/Cas9 system and successfully established GPM6B-knockout (KO) HEC-1B cells (Fig. 3A and B). Furthermore, we showed the transfection efficiency (Fig. S1A), and we constructed HEC-1B cells (GPM6B-KO) stably expressing GPM6B (Figs. 3C and S1B). GPM6B knockout in HEC-1B cells resulted in a decreased level of ALDH1A1 expression and GPM6B expressing HEC-1B cells (GPM6B-KO) resulted in an increased level of ALDH1A1 (Fig. 3D and Table SII). Similarly, we constructed GPM6B-expressing HEC108 cells (OE1 and OE2) (Figs. 3E and S1B). GPM6B expressing HEC108 cells resulted in an increased level of ALDH1A1 (Fig. 3F).

Due to a limited number of enrolled cases in this study, we examined the effect of high GPM6B on prognosis with three kinds of publicly available datasets; gene expression profiling interactive analysis (GEPIA)2, human protein atlas, and the cancer genome atlas (TCGA). In GEPIA2 of uterine corpus endometrial carcinoma, high GPM6B expression was marginally correlated with poor overall survival of the patients (Fig. 4A), in which group cutoff was set as quartile (http://gepia2.cancer-pku.cn/#survival). The data of human protein atlas revealed high GPM6B expression was correlated with poor overall survival of endometrial carcinoma (https://www.proteinatlas.org/ENSG00000046653–GPM6B/pathology/endometrial+cancer), in which 432 cases of GPM6B-high and 109 cases of GPM6B-low were enrolled and the best expression cut-off score 1.94 was applied (Fig. 4B). The median follow-up time of human protein atlas was 2.5 years and the p-value was 0.037. Lastly, we examined TCGA database, in which the enrolled endometrial carcinoma cases were divided into recently published classification (15); POLE type (ultramutated) (POLE), microsatellite instability (MSI), copy number low (CN-low), and copy number high (CN-high). Sixteen cases of POLE, 65 cases of MSI, 87 cases of CN-low and 58 cases of CN-high were examined in TCGA database. When cutoff was set as quartile, patients with high GPM6B expression had significantly shorter OS compared with those with low GPM6B expression in CN-high group but not in other groups (Fig. 4C).