PMEPA1 induces EMT via a non‐canonical TGF‐β signalling in colorectal cancer

Abstract Prostate transmembrane protein androgen induced 1 (PMEPA1) has been reported to promote cancer progression. Metastasis is the main factor leading to cancer progression and poor prognosis, and at the beginning of metastasis, epithelial‐to‐mesenchymal transition (EMT) is a crucial activation. However, the relationship between PMEPA1 and EMT in colorectal cancer metastasis is still poorly understood. In this study, we first testified that PMEPA1 expresses higher in tumour than normal tissue in Gene Expression Omnibus database, in the Cancer Genome Atlas (TCGA) as well as in the clinical data we collected. Moreover, the higher expression was associated with poor prognosis. We furthermore demonstrated PMEPA1 promotes colorectal cancer metastasis and EMT in vivo and in vitro. We found that PMEPA1 activates the bone morphogenetic proteins (BMP) signalling of TGF‐β signalling resulting in promoting EMT and accelerating the proliferation and metastasis of colorectal cancer.

of TGFβ family to acquire a mesenchymal phenotype. 6,7 As for EMT, it is demonstrated that EMT is activated via the TGFβ/Smad-dependent pathway in several cell lines. After binding with TGF-β1, TGF-β1 receptor leads Smad2 and Smad3 to phosphorylate. The phosphorylated Smad2 and Smad3 then combine with Smad4 to form a complex, which is translocated to the nucleus and promote the transcription of target genes, for example, the EMT related genes. Meanwhile, the Smad2/3/4-independent pathway also activates EMT. 8 Prostate transmembrane protein, androgen induced 1, PMEPA1 (also called TMEPA1, STAG1, ERG1.2 or N4WBP4), is located at chromosome 20q13. PMEPA1 expresses at the membrane of the cell and some subcellular organelles, such as endoplasmic reticulum and Golgi apparatus. PMEPA1 contains a transmembrane domain at the N-terminus and two PY motifs (PPxY) which interact with WW domain of the E3 ubiquitin ligase Nedd4. PMEPA1 was firstly found in prostate cancer as an androgen-induced gene. 9 However, PMEPA1 has duplicate roles in prostate cancer. In androgen receptor-positive prostate cancer cell, PMEPA1 promotes the proliferation. But in the androgen receptor-negative prostate cancer cells, PMEPA1 inhibits the proliferation by suppressing Smad3/4-c-Myc-p21Cip1. 10 PMEPA1 also inhibits proliferation through the inhibition of NEDD and PTEN in prostate cancer. 11,12 Moreover, PMEPA1 inhibits the bone metastasis by blocking the TGF-β and androgen signalling in prostate cancer. 13 In ovarian cancer, PMEPA1 has been reported to promote apoptosis. 14 However, PMEPA1 promotes the cancer progression in other solid cancer, 15 such as lung cancer, 16,17 breast cancer, [18][19][20] gastric cancer 21 and ovarian cancer. 22 However, the anticancer role of PMEPA1 has been reported to promote apoptosis. 14 The difference might be resulted from tissue-specificity. Different types of cancers exhibit different and special characteristic. As for PMEPA1, some studies have shown PMEAP1 inhibited proliferation through androgen receptor, 11 however, androgen receptor and the related signalling pathways have been activated in prostate cancer, but maybe not in lung cancer or breast, or colorectal cancer.
And the different role of PMEPA1 can be explained by the relationship of PMEPA1 and TGF-β pathways. PMEPA1 is induced by the TGF-β signalling, but meanwhile, it inhibits the phosphorylation of Smad2 and Smad3 to antagonize TGF-β signalling. 23 Considering the negative loop of PMEPA1 and TGF-β signalling, the phenotype of PMEPA1 in specific cancer needs deeper investigation. PMEPA1 in CRC is associated with cancer poor prognosis.
Moreover, we built stable PMEPA1 overexpressed and PMEPA1knockdown CRC cell lines to demonstrate that PMEPA1 promotes the cancer cell proliferation via inhibiting G1/S cell cycle arrest and inducing EMT related tumour metastasis. Besides, we investigated the regulation of PMEPA1 on TGF-β signalling. PMEPA1 blocked the canonical TGF-β signalling via dephosphorylating of Smad2 and Smad3. Interestingly, the BMP signalling, a non-canonical TGF-β signalling also promoted EMT and the pathway was activated by PMEAP1 which was BMP-depended in CRC. These new findings have thrown light on the role of PMEPA1 in colorectal cancer.

| Public datasets analysis
CRC expression profiling studies including relevant clinical information were identified by searching the public datasets. Dataset GDS2947 included 32 prospectively collected adenomas with those of normal mucosa from the same individuals. 24 And the comparison between 40 paired colorectal adenoma and adjacent normal tissue samples were performed by dataset GSE31737. 25 Datasets with gene expression profile comparing CRC or colorectal adenoma to paired adjacent normal tissue were obtained from Dataset GSE32323 which contained 17 paired samples. 26 GSE41328 contained five colorectal adenocarcinomas and matched normal colonic tissues were analysed with Affymetrix HG-U133-Plus-2.0 microarrays. 27 GSE38832 includes survival information of 122 patients with CRC. 28 GSE17537 includes expression and clinical data for 55 patients with CRC. 29 CRC expression and copy number profiling study from TCGA dataset were used to analysis association and survival.

| Pathway enrichment analysis
The correlated genes with PMEPA1 were screened in TCGA database and GSE35834 dataset by Pearson product-moment correlation analysis. A threshold of P < 0.05 and odds ratio >0.3 were used to screen the gene with significant correlations. The GSEA tool (http://software.broadinstitute.org/gsea/msigdb/) was used for pathways enrichment.

| Clinical specimens
This study was approved by the ethics committee of Zhejiang University's School of Medicine and was carried out in accordance with the approved guidelines and regulations. One hundred and fiftyfive CRC patients were recruited from Sir Run Run Shaw Hospital of Zhejiang University. Pathologic diagnoses were evaluated by pathologists via biopsy reports. Patients with familial adenomatous polyposis, hereditary non-polyposis CRC and inflammatory bowel disease were excluded. All tissue samples were obtained from colorectal adenocarcinoma patients without any adjuvant treatment including radiotherapy or chemotherapy prior to surgery and diagnosis.

| Cell lines and cell culture
The human colon cancer cell lines SW620, HT29, HCT116, HCT 8 and HEK293t were purchased from the American Type Culture Collection (Manassas, VA, USA), and all the colon cancer cell above were cultured in RPMI 1640 medium (HyClone, Tauranga, New Zealand) with 10% foetal bovine serum (HyClone, Tauranga, New Zealand). HEK293T cell was cultured in DMEM high Glucose (HyClone, Tauranga, New Zealand) with foetal bovine serum (HyClone, Tauranga, New Zealand). All the cells are grown at 37℃ in an atmosphere of 95% and 5% CO 2 .

| DNA and siRNA constructs
The full CDS sequence of PMEPA1 was amplified and cloned into p3×FLAG-CMV-14 (Sigma). The pLKO.1 lentivirus vector was used to construct shRNA-PMEPA1 vector, and lentiviruses were co-transfected into HEK293T cells with the packaging plasmids pMD2.G and psPAX2. Lipofectamine 2000 (Invitrogen) was used in all transfection experiments according to its manufacture instructions. And the siRNA-Negative (UUCUCCGAACGUGUCACGUTT), siRNA-PMEPA1(GGAGCUGGAGUUUGUUCAGTT) and siRNA-smad1(CCAAUAGCAGUUACCCAAATT) were transfected by GenMute siRNA Transfection Reagent (SignaGen). DNA and selected in 10μg/ml G418 for 2 weeks. SW620 and HT29 PMEPA1 knocked-down cell lines were built by the lentivirus which is produced by the HEK293T transfected with the pLKO vector containing sh-PMEPA1 and selected in 1 μg/mL puromycin for 2 weeks. was added to each well. Subsequently, the cells were incubated for 3 hours at 37℃ and 5% CO 2 . The supernatants were removed to new 96-well plates and recorded the optical absorbance at 490 mn.

| Plate clone assay and soft agar clone assay
2×10 3 cells were plated in 6-well plates and fixed by 4% (w/v) paraformaldehyde and stained by 0.1% crystal violet after 2 weeks. Soft agar colony formation assay was carried out as described previously. 2×10 3 cells were plated on the 6-well plates coated soft agar.
Three weeks after seeding, the cell was fixed by 4% (w/v) paraformaldehyde and stained by 0.1% crystal violet.

| Cell migration and invasion assay
The migration and invasion capacity of cells were tested by transwell migration and transwell invasion assays. The cell was plated in the upper compartment chambers of 24-well plates equipped with cell culture inserts containing 8.0 μm pore size membrane (Costar Corp. Cambridge, MA, USA) with 1% FBS medium and the lower chamber was containing 10% FBS medium. Diluted extracellular matrix gel (BD Biosciences Bedford, MA, USA) was coated in the upper chamber for the invasion assays, but not migration assays. Moreover, 5 × 10 4 cells were incubated for migration assays, and for the invasion, assays 1 × 10 5 cells are required. The cells in the chamber were fixed in 4% paraformaldehyde and stained by 0.1% crystal violet 48 or 72 hours after incubating. Moreover, 30% acetic acid was used to wash the chamber, and the washing solution was recorded the optical absorbance at 570 mn.

| Western blot, immunohistochemistry and immunofluorescence
Western blot, immunohistochemistry and immunofluorescence were performed as previously. 30,31 And the related antibodies were GAPDH was used as a loading control for Western blots.

| RNA extraction and quantitative RT-PCR
Total RNA from the tissues were extracted using TRIZOL Reagent (Invitrogen). The RNA concentration was determined using UV spectrophotometry. cDNA was synthesized with PrimeScript ® RT regent kit (Takara Biotechnology, Dalian, China). RT-PCR was performed with Thunderbird SYBR Master Mix (Takara, Japan). The PCR was performed on a Real-time PCR Detection System (StepOnePlus, ABI) with the following cycles: 95°C for 1 minute, followed by 40 cycles of 95°C for 15 seconds, 60°C for 15 seconds and 72°C for 45 seconds to detect the target gene level and GAPDH gene levels.
GAPDH expression was used as an internal control. The 2 −ΔCT was calculated for every sample and normalized to GAPDH.

| Statistical analysis
The statistical package SPSS (version 20.0; IBM New York, NY, USA) was applied. Unpaired Student's t tests were used for normally distributed data and non-parametric Mann-Whitney U tests were used for non-normally distributed data to compare central tendencies. For results in CRC tissues, Relapse-free, metastasis-free or overall survival were compared between high and low PMEPA1 expression groups using median gene expression value as a bifurcating point. Correlations were analysed by the Spearman coefficient test. Significance was set at P < 0.05. Stata software was used to be a comprehensive evaluation that associated public datasets with clinical samples.

| PMEPA1 expresses higher in tumour and associated poor prognosis
Our previous gene expression microarray and bioinformatics works have shown the PMEPA1 expresses higher in tumour cells and tumour budding cells than that in stroma cells. 32 Tumour budding, occurring at the invasive front of cancer has a metastatic and stem-cell-like feature indicating a poor prognosis. Tumour budding is partly responsible for cancer metastasis, and its initiation is based on the epithelial-mesenchymal transition (EMT) process. The expression of PMEPA1 was higher in tumour budding cells than tumour parenchyma cells and normal epithelial cells ( Figure 1A). We then confirmed that mRNA expression of PMEPA1 was higher in tumour than normal tissue in TCGA and GEO database GDS2947 n = 32 P = 0.001 ( Figure 1B), GSE31737 n = 40 P < 0.0001 ( Figure 1C), GSE41329 n = 10 P = 0.0039 ( Figure 1D), GSE32323 n = 17 P = 0.0002 ( Figure 1E) and TCGA n = 32 P < 0.001 ( Figure 1F), and all the data were from the paired samples. To identify the changes of PMEPA1 mRNA is related with copy number, we investigated the copy number of PMEPA1 in TCGA database, and found the mRNA level of PMEPA1 was increased in the copy number gained group ( Figure 1G). We then analysed the relationship between copy number and mRNA expression of PMEPA1, which showed there was a significantly positive correlation between copy number and mRNA level of PMEPA1 ( Figure 1H). These data indicated that mRNA level of PMEPA1 is higher in CRC tumour tissue than normal tissue.
To explore the relation between PMEPA1 and the prognosis, we obtained a validation cohort from the GEO databases.
F I G U R E 1 Prostate transmembrane protein androgen induced 1 (PMEPA1) expresses higher in tumour and related with poor prognosis (A) the expression of PMEPA1 in normal epithelium cells, tumour parenchyma cells and tumour budding cells in colorectal cancer. B, GSD2941 showed that PMEPA1 mRNA expressed higher in tumour tissue when compared with paired adjacent normal tissue (P = 0.0001). C, GES31737 showed that PMEPA1 mRNA expressed higher in tumour tissue when compared with paired adjacent normal tissue (P < 0.0001). D, GSE41328 showed that PMEPA1 mRNA expressed higher in tumour tissue when compared with paired adjacent normal tissue (P = 0.0039). E, GSE32323 showed that PMEPA1 mRNA expressed higher in tumour tissue when compared with paired adjacent normal tissue (P = 0.0002). F, TCGA data showed that PMEPA1 mRNA expressed higher in tumour tissue when compared with paired adjacent normal tissue (P = 0.0039). G, TCGA data showed that mRNA level of PMEPA1 is higher in copy number gained group (focal CNV values larger than 0.  Figure 1L) and GSE 17537 ( Figure 1M). PMEPA1 median centred ratio was related to recurrence in GSE38832, P = 0.03 ( Figure 1N).
We also testified the samples from Sir Run Run Shaw Hospital and found mRNA level of PMEPA1 expressed higher in the tumour than normal tissue and the higher expression is related to the poor prognosis ( Figure 1O,P).
Moreover, the results of TCGA database and Sir Run Run Shaw Hospital database were analysed by multivariate Cox proportional hazards regression model to find an independent prognostic value of PMEPA1 by adjusting location, differentiation, infiltrating depth, lymph node metastasis distant metastasis and TNM stage. After adjustment, PMEPA1 still showed a significant prognostic value.

| Knockdown of PMEPA1 inhibits proliferation and metastasis in colorectal cancer cells
To investigate the molecular role of PMEPA1 in colorectal cancer cells, first, we detected the mRNA and protein level in several colorectal cancer cell lines by RT-PCR and Western blot ( Figure S1A).
Subsequently, we selected PMEPA1 higher expressed cell lines, HT29 and SW620, for building the stable PMEPA1 knockdown cell lines.
We then tested mRNA and protein level of PMEPA1 by RT-PCR and Western blot, we investigated the influence on colorectal cancer cells with the comparison of relative control cell lines (Figure 2A). Compared with the Scramble-shRNA, PMEPA1-shRNA inhibited the proliferation and clones formation of HT29 and SW620 ( Figure 2B-D). To investigate the related mechanism of proliferation, cell cycle was analysed by flow cytometer, which showed PMEPA1 knockdown cells were arrested in the G1/S cell cycle ( Figure 2E). Moreover, down-regulation of PMEPA1 inhibited migration and invasion and reduced the capacity of wound healing ( Figure 2F,G). Considering EMT is a significant process of cell migration and invasion, we detected the proteins level of EMT markers in PMEPA1 knockdown cell lines and control cell lines. As the epithelium marker, E-cadherin was up-regulated; as the mesenchymal markers, MMP9 and Snail were down-regulated in the PMEPA1 knockdown cell lines, which indicates PMEPA1 knockdown inhibited EMT ( Figure 2H). Immunofluorescence assay also validated that down-regulated PMEAP1 increased expression of E-cadherin but decreased fibronectin, a mesenchymal marker ( Figure 2I). Taken together, the data shows PMEPA1 knockdown arrests cell at G1/S and inhibits CRC cell proliferation and PMEPA1 knockdown inhibits EMT and metastasis of CRC cells.

| PMEPA1 promotes proliferation and metastasis in colorectal cancer cells
As shown in Figure S1A  the expression of MMP9 and Snail ( Figure 3H). Immunofluorescence assay also validated that PMEPA1 overexpression decreased the expression of E-cadherin, and increased fibronectin in HCT116 cells lines ( Figure 3I). In order to confirm that the changed phenotypes are related to PMEPA1, we transfected siRNA-PMEPA1 and Negative control (siRNA-NC) into PMEPA1-overexpressed cell lines. As shown in Figure 3B, PMEPA1-siRNA inhibited the proliferation compared with siRNA-NC in PMEPA1 stable expressing HCT116 and HCT8 cell lines.
Transwell migration and invasion assays also indicated PMEPA-siRNA inhibited migration and invasion in the PMEPA-overexpressed cell lines ( Figure 3F). As a result, PMEPA1 promotes proliferation via arresting G1/S cell cycle and enhances migration and invasion via EMT in vitro. group is larger than the control group. The weight of xenografts with PMEPA1 overexpressed were higher than the control xenografts which also confirms the proliferation induction of PMEPA1 ( Figure 4A). Furthermore, we inoculated control and PMEPA1-knockdown SW620 cell into male BALB/c nude mice, the continuous tumour volume measurement and the weight of the tumour indicated PMEPA1 knockdown reduces the proliferation in vivo ( Figure 4B).

| PMEPA1 promotes proliferation and metastasis in vivo
We then intravenously injected luciferase-labelled control and PMEPA1 knockdown SW620 cells into NOD/SCID mice and monitored metastasis with bioluminescent imaging. After 40 days, the whole-body luminescence signals in the PMEPA1 knockdown group were ~10-fold lower than the control group ( Figure 4C,D). Moreover, the number of metastatic foci in the lung decreased significantly in the mice which were injected with PMEPA1 knockdown cells ( Figure 4E). The results above indicate that PMEPA1 knockdown decreases the metastasis in mice.

Expression of E-cadherin was increased and the expression of
Vimentin was decreased in the PMEPA1-knockdown xenografts ( Figure 4F). However, in the PMEPA1-overexpressed xenograft, expression of E-cadherin was decreased and expression of MMP9 was increased ( Figure 4I).
We then detected the mRNA levels of PMEPA1, E-cadherin, MMP9, Twist and Zeb1 of the xenograft. E-cadherin was decreased and MMP9, Twist and Zeb1 were increased in the PMEPA1-knockdown xenografts ( Figure 4G). And E-cadherin was increased and MMP9, Twist and Zeb1 were decreased in PMEPA1-overexpressed F I G U R E 2 Knockdown PMEPA1 inhibits proliferation and metastasis in colorectal cancer cells. A, The efficiency of PMEPA1 knocked down has been testified by qRT-PCR and Western blot in HT29 and SW620. (B-D) The proliferation of HT29 and SW620 cells with PMEPA1 down-regulation was detected by CCK8 assay, plate clone assay and soft agar clone assay. E, Flow Cytometer detected the cell cycle and the proportion of each cell cycle. F, Migration and invasion assay were used for HT29 and SW620 cells with PMEPA1 down-regulation. The chambers were washed by 30% acetic and absorbance of washing solution was recorded at 570 mn for the quantification of the relative migration and invasion cells. G, Wound healing assay was used for the HT29 and SW620 cells with PMEPA1 down-regulation. And the length of the wound has been measured by Image J. H, Western blot detection of E-cadherin, MMP9, snail and PMEPA1 in HT29 and SW620 with PMEPA1 down-regulation. I, Immunofluorescence assay for expression of E-cadherin and Fibronectin in PMEPA1 down-regulated SW620 cells.The designations for levels of significance were used within this figure: *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant xenografts ( Figure 4J). In the PMEPA1-knockdown group, protein levels protein E-cadherin was increased and the expression of Twist and Vimentin were decreased ( Figure 4H). And protein levels of Ecadherin were decreased in the PMEPA1-overexpressed xenografts; and the protein levels of MMP9 were decreased in the PMEAPA1knockdown group ( Figure 4K).
Moreover, we performed a correlation analysis between the expression of PMEPA1 and EMT markers. The results indicated a strong positive correlation between PMEPA1 and CHD2, Twist, FN1.

| PMEPA1 has dual role in TGF-β signalling
To explore the mechanism that PMEPA1 promotes EMT and metastasis. We analysed the public data from TCGA and GSE35834, and screened the gene expression profiles associated with F I G U R E 3 Prostate transmembrane protein androgen induced 1 (PMEPA1) promotes proliferation and metastasis in colorectal cancer cell lines (A) The efficiency of overexpressed PMEPA1 has been testified by qRT-PCR and Western blot in HCT116 and HCT8. And the efficiency of siRNA-PMEPA1 has been testified by qRT-PCR. B, The proliferation of HCT116 and HCT8 cells with PMEPA1 overexpressed and PMEPA1-overexpressed+siRNA-PMEPA1 was detected by CCK8 assay. The * markers in purple colours indicate the statistical analysis is between PMEPA1+NC and PMEPA1+siRNA; the * markers in red colours indicate the statistical analysis is between EV and PMEPA1. (C,D) The proliferation of HCT116 and HCT8 cells with PMEPA1 overexpressed was detected by plate clone assay and soft agar clone assay. E, Flow Cytometer detected the cell cycle and the proportion of each cell cycle. F, Migration and invasion assay have been used for HCT116 and HCT8 cells with PMEPA1 overexpressed and PMEPA1-overexpressed+siRNA-PMEPA1.The chambers were washed by 30% acetic and the absorbance of washing solution was recorded at 570 mn for the quantification of the relative migration and invasion cells. G, Wound healing assay has been used for the HCT116 and HCT8 cells with PMEPA1 overexpressed. And the length of the wound has been measured by Image J. The designations for levels of significance were used within this figure: *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. As we know, the role of PMEPA1 is complicated in TGF-β signalling. We then explored the relationship between PMEPA1 and TGF-Β signalling in CRC. We then tested phosphorylation of smad1/5, smad2 and smad3 after 3, 6 and 9 hours treated with 5 ng/mL TGF-β to find the optimal point of time when phosphorylation was activated ( Figure 5A). As shown in Figure 5B Figure S1D). Taken together, PMEPA1 inhibits the canonical TGF-β signalling, but activates the BMP signalling.
To explore the role of BMP signalling in EMT induced by PMEPA1, we first verified whether PMEPA1 interacts with Smad1, a significant protein in BMP signalling. As shown in Figure S1C, the co-immunoprecipitation assay indicated no interaction between and Smad1. Moreover, to explore the role of BMP signalling pathway, we used siRNA-smad1 to decrease the expression of smad1 in the PMEPA1-overexpressed cells ( Figure S1E,F). And the decreased mRNA level of E-cadherin caused by overexpressed PMEAP1 has been rescued by the siRNA-smad1 ( Figure S1G). As shown in Figure   S1H, the decreased Smad1 inhibited the cell mobility induced by the PMEPA1. Taken together, PMEPA1 promotes migration via BMP signalling pathways.
F I G U R E 4 Prostate transmembrane protein androgen induced 1 (PMEPA1) promotes proliferation and metastasis in vivo (A) PMEPA1overexpressed HCT8 cells and the control HCT8 cells (EV) have been subcutaneously inoculated into immune-deficient mice. And the volume of xenografts was recorded after injection. After 30 d, the weight of xenografts has been photographed and measured after killing the mice. B, PMEPA1-knockdown SW620 cells and control SW620 cells have been subcutaneously inoculated into immune-deficient mice. And the volume of xenografts was recorded after injection. After 26 d, the weight of xenografts has been photographed and measured after killing the mice. C, Representative images of luciferase signals and (D) quantification of photon flux for metastasis by tail-vein injection of SW620-SC and SW620-shRNA PMEPA1 cells in immune-deficient mice. E, H&E staining for pulmonary metastatic foci from SW620-SC and SW620-shRNA PMEPA1cells. And the foci in lungs of SW620-SC and SW620-shRNA PMEPA1 groups were counted. F, Immunohistochemistry detection for PMEPA1, E-cadherin, Vimentin in SW620-SC and SW620-shRNA PMEPA1 xenografts. G, qRTPCR dectection for PMEPA1, E-cadherin, MMP9, Twist and Zeb1vimentin in SW620-SC and SW620-shRNA PMEPA1 xenografts. H, Western blot detection for E-cadherin, vimentin, PMEPA1 and Twist in SW620-SC and SW620-shRNA PMEPA1 xenografts. I, Immunohistochemistry detection for PMEPA1, E-cadherin, MMP9in HCT8-EV and HCT8-PMEPA1 xenografts. J, qRT-PCR dectection for PMEPA1, E-cadherin, MMP9, Twist and Zeb1vimentin in HCT8-EV and HCT8-PMEPA1 xenografts. (K) Western blot detection for E-cadherin, vimentin and PMEPA1 in HCT8-EV and HCT8-PMEPA1 xenografts. The designations for levels of significance were used within this figure: *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant has been reported to promote cancer metastasis. In breast cancer, PMEPA1 reduces PTEN and promotes non-canonical PI3K/ Akt signalling to promote cancer progression. 33 In prostate cancer, PMEPA1 inhibits Smad3/4-cmyc-p21Cip1 signalling pathway to promote prostate cancer cell proliferation. 10 In lung cancer, PMEPA1 regulates ROS and IRS-1 signalling and induces EMT to promote metastasis. 35 Together, PMEPA1 could be considered as a versatile molecule, which functions the complex roles in different types of cancer.
As we know, TGF-β is a classical inducer of EMT to drive tumour cell migration, invasion, and metastasis in many carcinomas. 36 As a TGF-β-inducible gene, PMEPA1 regulated multiple biological process in several types of cancer. 16,20,37 Some reports showed PMEPA1 accelerates the metastasis and EMT through TGF-β signalling. 35 Nevertheless, we demonstrated PMEPA1 promoted migration and invasion, and induced EMT in CRC by activating the BMP signalling pathway with phosphorylation of Smad1 and Smad5. Previous reports have indicated the role of BMP signalling pathways in EMT. In colorectal cancer, BMP-2, an inducer of BMP signalling, induces EMT and drug resistance. 38 BMP-2 also induces BMP signalling and EMT in pancreatic cancer. 39 The agonists of BMP-2 and BMP-7 also block the BMP signal, and inhibit EMT and invasion in melanoma cells. 40 PMEPA1, which activates BMP signalling pathway like BMP-2 and BMP7, might have the potential to be a novel target for molecular agonists design. However, the detailed mechanism of PMEPA1 activating BMP signalling remained to further study.
In conclusion, we revealed that PMEPA1 promoted EMT-mediated metastasis through activating TGF-β non-canonical signalling cascade.
Although the conclusion needs more clinical studies to valid, PMEPA1 might has the potential to serve as a meaningful biomarker for high-risk CRC or to serve as a therapeutic target for intervene colorectal cancer.

ACK N OWLED G EM ENTS
This work was supported by grants from the National Natural

CO N FLI C T O F I NTE R E S T
We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled.

AUTH O R S' CO NTR I B UTI O N
Xue Wang analysed the data has shown in Figure 1 and Tables 1, 2