RETRACTED ARTICLE: The autocrine glycosylated-GREM1 interacts with TGFB1 to suppress TGFβ/BMP/SMAD-mediated EMT partially by inhibiting MYL9 transactivation in urinary carcinoma

Abstract

Purpose

Urothelial carcinoma (UC) is a common disease in developed counties. This study aimed to identify autocrine roles and signaling pathways of gremlin 1, DAN family BMP antagonist (GREM1), which inhibits tumor growth and epithelial-mesenchymal transition (EMT) in UC.

Methods

Systematic in vitro and in vivo studies using genetic engineering, different urinary bladder urothelial carcinoma (UBUC)-derived cell lines, and mouse models were performed, respectively. Further, primary upper tract urothelial carcinoma (UTUC) and UBUC specimens were evaluated by immunohistochemistry.

Results

GREM1 protein levels conferred better disease-specific and metastasis-free survival rates and played an independent prognostic factor in UTUC and UBUC. Hypermethylation is the primary cause of low GREM1 levels. In different UBUC-derived cell lines, the autocrine/secreted and glycosylated GREM1 interacted with transforming growth factor beta 1 (TGFB1) and inhibited TGFβ/BMP/SMAD signaling and myosin light chain 9 (MYL9) transactivation, subsequently cell proliferation and epithelial-mesenchymal transition (EMT). Secreted and glycosylated GREM1 also suppressed tumor growth, metastasis, and MYL9 levels in the mouse model. Instead, cytosolic GREM1 promoted cell proliferation and EMT by activating the tumor necrosis factor (TNF)/AKT/nuclear factor kappa B (NFκB) axis.

Conclusions

Clinical associations, animal models, and in vitro indications provided solid evidence to show that the epithelial autocrine GREM1 is a novel tumor suppressor in UCs. The glycosylated-GREM1 hampered cell proliferation, migration, invasion, and in vitro angiogenesis through interaction with TGFB1 to inactivate TGFβ/BMP/SMAD-mediated EMT in an autocrine manner.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-023-00788-8.

Introduction

Urothelial carcinoma (UC), especially urinary bladder urothelial carcinoma (UBUC), is a common disease in developed counties [1]. Another UC, upper tract urothelial carcinoma (UTUC), accounts for only ~ 5–10%, predominantly due to regional environmental exposures [2]. Genetic and environmental components contribute to UC occurrences [3, 4]. In addition, clinicopathological features, including histological stage, grade, tumor size, and multiplicity, impact their advancements [5]. Around 75% of diagnosed UBUCs are early-stage and noninvasive. However, the prevalent and high recurrent properties of noninvasive UBUCs impose a substantial economic loading [6]. In addition, genomic heterogeneity detected in UBUC patients confounds the general remedies [7]. Therefore, one modern and prospective scheme to enhance the current consequences of treatments is to recognize crucial and involving molecules for targeted therapies.

Based on the Read per Kilobase Per Million (RPKM), the normal urinary bladder (58.60 ± 24.71) highly expresses Gremlin 1, DAN family BMP antagonist (GREM1) transcript, only next to the gall bladder (207.54 ± 99.52) and appendix (85.15 ± 8.29) among 27 normal tissues [8]. The GREM1 gene is conserved in vertebrates and belongs to the DAN (Differential screening-selected gene Aberrant in Neuroblastoma) gene family. As antagonists of the Bone Morphogenetic Protein (BMP), the DAN family proteins contain a highly conserved cysteine knot domain and the structure shared by the transforming growth factor beta (TGFβ) superfamily. This family comprises TGFβs, BMPs, nerve growth factors (NGFs), platelet-derived growth factors (PDGFs), and other secreted proteins [9]. Protein analysis identified that GREM1 is present in two primary forms: glycosylated and non-glycosylated [10]. In addition, UniProt [11] annotates one signaling peptide (#1–24), a predicted glycosylation site (#42), and a cysteine knot motif (#94–184) in the GREM1 protein.

Cytokines of the TGFβ family, including BMPs, are versatile secreted cytokines with highly context-dependent functions. BMPs induced their effects through interaction with type I and II cell surface receptors that activate these serine/threonine receptor kinases. Diverse secreted antagonists, including GREM1 that directly bind BMPs, firmly govern the interactions between BMPs and their receptors to restrain the associations with corresponding cognate BMP receptors [12]. GREM1 signaling occurs via diverse cell mechanisms [13]. Agitation in this pendulum implicates a range of developmental disorders, tissue homeostasis, and diseases, including fibrosis and cancers [14]. GREM1 preferentially binds BMP2 and BMP4 to inhibit canonical BMP signaling by preventing the phosphorylation of SMAD family member 1 (SMAD1)/5/8 [15]. Afterward, their downstream target genes, such as the inhibitor of DNA binding 1 (ID1), cannot be transactivated. A recent review suggested that BMPs and their antagonists have tumor-promoting and suppressive activities depending on cancer cell types and tumor stages [12]. However, renal cell and breast carcinomas showed hypermethylation on the GREM1 promoter [16–19]. Mesenchymal BMP signaling decreased by GREM1 regulates the progression of colorectal cancar [20]. The effects of autocrine GREM1 remained elucidative. Therefore, we aimed to study the expression levels, physiological functions, and underlying molecular mechanisms of the secreted/autocrine and glycosylated GREM1 in UC patients, mouse models, and distinct UBUC-derived cell lines to evaluate its therapeutic potential further.

Materials and methods

Cell culture, quantitative RT-PCR, protein concentration, and immunoblot analysis

Human UBUC-derived cell lines, including T24 (pTa, HTB-4™), RT4 (pT1, HTB-2™), UMUC3 (pT2-4, CRL-1749™), J82 (pT3, HTB-1™), TCCSUP (HTB-5™), SW780 (CRL-2169™), kidney epithelial Phoenix-AMPHO (CRL-3213™) and human umbilical vein endothelial cells (HUVEC, CRL-1730™) were obtained from the American Type Culture Collection (Manassas, VA, USA). We also attained human normal urothelial cells, SV-HUC-1 (#EP-CL-0222, Elabscience®, Houston, TX, USA), and another human UBUC-derived cell line, BFTC905 [pT4, #60,068, Bioresource Collection and Research Center, Hsinchu, Taiwan]. Supplementary Fig. S1 provides the genetic backgrounds of these cell lines [21]. The short tandem repeat DNA profiling (Mission Biotech, Taipei, Taiwan) authenticated these cell lines. Our previous studies [22] described cell culture conditions, quantitative RT-PCR, and immunoblot assays (Supplementary Methods). Protein concentrators (10,000 MWCO, #88,516, ThermoFisher Scientific, Waltham, MA, USA) concentrated GREM1 proteins in the conditioned media for the immunoblot and ELISA assays.

Patients, specimens, ELISA, and immunohistochemistry

The institutional review board of Chi-Mei Medical Center approved the retrospective retrieval of 340 UTUCs and 295 primary UBUCs with available tissue blocks who underwent surgery with curative intent between January 1996 and May 2004 [23], excluding those who underwent palliative resection. Our earlier publication [23] showed detailed information about patients. Two expert pathologists (CF Li and TJ Chen), blinded to clinicopathological details and patient outcomes, interpreted the immunoexpression. A subset of the UBUC cohort (n = 47) and non-UBUC controls (n = 70) provided the serum samples. Another 2 UBUC subsets (n = 25 and n = 30) were subjected to DNA extraction and bisulfite pyrosequencing on the GREM1 promoter. Human GREM1 ELISA Kit (#SEC128Hu, Cloud-Clone Corp., Houston, TX, USA) measured the GREM1 protein levels in media from cell culture, sera from UBUC patients, and normal controls (Supplementary Fig. S2). Immunohistochemistry probing anti-human GREM1 antibodies detected GREM1 protein levels in a tissue microarray (Supplementary Methods).

Methylation analysis on the GREM1 promoter

MethPrimer [24] identified a CpG-island-rich fragment (374 bp) in the GREM1 proximal promoter (-1180 to + 150; + 1: transcription start site) (Supplementary Fig. S3A). Due to the length limitation of pyrosequencing, we designed 4 primers (S1 to S4: #19-#52, Supplementary Fig. S3B) to examine 5 distinct UBUC-derived cell lines and one primer (S4: #41 to #52) to investigate the specimens from 25 UBUC patients (Supplementary Methods). Bisulfite pyrosequencing determined the methylation status of distinct cell lines and UBUC patients (Supplementary Fig. S3C, S3D).

Chemicals, expression plasmids, mutagenesis, and reagents

Unless otherwise stated, all chemicals were purchased from Merck KGaA (Darmstadt, Germany). The expression vector carrying GREM1 complete DNA with Myc-DDK-tag (#RC210835, pCMV-GREM1, Origene, Herford, Germany), pHAGE-TGFB1 (#116,799, Addgene, Watertown, MA, USA) were obtained and reconstructed into the pHTC HaloTag® CMV-neo Vector (pHaloTag). Another two plasmids, pGREM1(ΔSP)- (removal of the signal peptide, residues #1–24), and pGREM1(N42A)-HaloTag [mutation of a potential N-lined glycosylated residue #42, asparagine (N) into aniline (A)], were generated using the wild-type pGREM1(WT)-HaloTag plasmid as the template with the PCR-based technique, QuickChage II Site-directed mutagenesis kit (Agilent, Santa Clara, CA, USA) and InFusion® HD Cloning Kit (Takara, Mountain View, CA, USA) (Supplementary Methods). Twist Biosciences (South San Francisco, CA, USA) synthesized the pGL4.17-(CAGA)₁₂ (SRE: SMAD responsive element) plasmid. The pGL3-BRE-Luc plasmid (#45,126) was obtained from Addgene. Human recombinant BMP4 protein (rBMP4, #314-BP-050, R&D Systems, Inc. Minneapolis, MN, USA) was dissolved in sterile 4 mM HCl containing 0.1% BSA. Human recombinant GREM1 protein (active, #GTX53832-pro, GeneTex, Inc. Irvine, CA, USA) was prepared in PBS. Acti-Stain 488 Phalloidin (#PHDG1-A, Cytoskeleton, Inc., Denver, CO, USA) was used to determine cytoskeleton structure.

Stable overexpression and knockdown of specific genes in UBUC-derived cells

Cells after Transfection of pHaloTag, pGREM1- and/or pTGFB1-HaloTag plasmids were selected with 200 μg/mL of the Geneticin™ Selective Antibiotic (G418 sulfate, #10,131,035, TheroFisher, Waltham, MA, USA) for stable Overexpression. For in vitro studies and animal models, we performed the lentivirus-based gene delivery system for GREM1(WT), GREM1(ΔSP), GREM1(N42A), and myosin light chain 9 (MYL9) overexpression based on our previous study [23]. Likewise, cells were transduced with replication-incompetent virus particles carrying the small hairpin RNA (shRNA) targeting the GREM1 and MYL9 genes for the knockdown. Stable clones were selected with 4 μg/mL puromycin for 2 days and replaced with 2 μg/mL puromycin afterward. Supplementary Methods describes the preparation of replication-incompetent viruses.

In vitro assays and in vivo animal experiments

Flow cytometric, 5-bromo-2'-deoxyuridine (BrdU, #K306, Biovision, Milpitas, CA, USA) with Propidium Iodide (PI) staining, CytoSelect™ 96-well in vitro tumor sensitivity assay (soft agar colony formation) (#CAB-150, Cell Biolabs, Inc., San Diago, CA, USA) along with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), transwell migration, transwell invasion (ECM554, Merck, Darmstadt, Germany) and HUVEC tube formation assays determined the alterations of cell cycle distribution, cell proliferation, colony formation/anchorage-independent cell growth after overexpression or knockdown of a specific gene and/or chemical treatments in vitro, respectively, following our previous protocols [22, 23, 25] with some modifications (Supplementary Methods). The InnoZyme™ Assay Kit, Fluorogenic (CBA003, Merck) measured the matrix metalloproteinase 2 (MMP2)/MMP9 activities. Immunocytofluorescence examined the F-actin and cell structure using a microscope (Axioskop 2 Plus, Zeiss, Oberkochen, Germany) with 495/518 (excitation/emission). NOD/SCID mice (LASCO, Taiwan) were randomly grouped and subjected to subcutaneous and tail vein injections with UMUC3 cells, stably carrying GREM1(WT), GREM1(ΔSP), and GREM1(N42A) genes, respectively, to analyze the tumor size and metastatic lesions (Supplementary Methods).

Next-generation sequencing and analysis

Lysates from HaloTag- (control), GREM1(WT)- and GREM1(ΔSP)-overexpressed J82 cells; shLuc (control) and GREM1-knockdown BFTC905 cells were collected, subjected to RNA extraction with TRIzol (ThermoFisher). 28 s/18 s rRNA ratio was detected for sample quality control, followed by next-generation sequencing (NGS, Genomics, Taipei, Taiwan). The clusterProfiler software [26] and Gene Set Enrichment Analysis (GSEA) [27] determined whether a priori-defined set of genes showed statistical significance and concordant differences between each experimental group and its corresponding control.

Dual-luciferase reporter assay

To evaluate the effects of GREM1 levels on the SMAD- and BMP-responsive elements, 0.34 μg of pGL4.17-(CAGA)₁₂ and pGL3-BRE-Luc plasmids, respectively, were cotransfected with 0.07 μg Renilla luciferase plasmid (control of the transfection efficiency) in 12-well plates. Passive lysis buffer lyzed the cells after Transfection for 48 h. For the GREM1 promoter, Dual-Luciferase® Reporter Assay System (#E1910, Promega, Madison, WI, USA) and a GloMax® 20/20 Luminometer (Promega) analyzed the relative luciferase activities in GREM1-overexpressing J82 and GREM1-knockdown BFTC905 cells. Different MYL9 promoter regions were cloned as pGL4.17-A_{(-700 to +1)}, pGL4.17-B_{(-400 to +1)}, and pGL4.17-C_{(-320 to +1)} and subjected to dual-luciferase assays with or without GREM1-overexpression.

Coimmunoprecipitation and nuclear/cytosol fractionation

The Immunoprecipitation (IP) Kit (#K286, Biovision) and immunoblot assay detected the protein–protein interactions. The Nuclear/cytosol Fractionation Kit (K266, Biovision) isolated the nuclear component from the cytoplasmic fraction, followed by immunoblot assays. Supplementary Methods describe the detailed protocols.

Data mining and statistics

SPSS v.18.0 (IBM, Armonk, NY, USA) performed the statistical analysis. The protein levels of GREM1 and MYL9 determined by immunohistochemistry were dichotomized into high (≥ median H-score) and low (< media H-score). The Chi-square test evaluated the correlations between the protein levels and different parameters. Further, the log-rank test and Kaplan–Meier estimated and plotted disease-specific survival (DSS) and metastasis-free survival (MFS). Multiple Cox regression estimated the independence of every survival factor. We performed all in vitro experiments ≥ 3 repeats. The means ± SD describes the data with normal distribution. A two-tailed t-test examined significant differences between 2 groups. One way of variance analyzed differences among ≥ 3 groups, followed by the Scheffe multiple comparison test. Supplementary Methods narrate the procedure for data mining the Gene Expression Omnibus (GEO) database to identify upregulated transcripts in UBUCs. A p < 0.05 is considered statistically significant.

Results

The serum GREM1 level is proportional to its cytosolic counterpart, which confers an independent favorable prognostic factor in urothelial carcinoma

Quantitative RT-PCR and immunoblot assay identified that SV-HUC-1 and BFTC905 highly while J82, T24, UMUC3, RT4, and SW780 cells lowly expressed GREM1 mRNA and their corresponding protein levels (Fig. 1a, b). ELISA and immunoblot assay (using concentrated proteins) showed secreted GREM1 in the conditioned media (Fig. 1c, d). Across 5 cell lines: SVV-HUC-1, J82, BFTC905, T24, and UMUC3, the expression level of GREM1 protein examined by concentrated conditioned media/immunoblot analysis and ELISA was highly correlated (r = 0.980, p = 0.020, data not shown), suggesting, either way, can be used to measure the GREM1 concentration. Pearson correlation (r) between cytosolic GREM1 and secreted GREM1 was 0.974 (p = 0.026, data not shown), denoting that the cytosolic GREM1 level is highly proportional to the secreted GREM1 level. ELISA showed that the serum GREM1 levels in UBUC patients (n = 47) were lower (p < 0.01) than those non-tumor ones (n = 70) (Fig. 1e). High-stage tissues (T2-T4) expressed low GREM1 proteins than low-stage UTUC and UBUC (Ta-T1) and non-tumor urothelium (p < 0.001, Supplementary Table S1; Fig. 1f). Further, GREM1 protein level negatively correlated to the advanced primary tumor (p < 0.001; p < 0.001), nodal metastasis (p = 0.002; p = 0.001), histological grade (p = 0.008; p < 0.001), vascular invasion (p < 0.001; p < 0.001), and mitotic rate (p = 0.019; p < 0.001) in UTUC and UBUC, respectively (Supplementary Table S1). High GREM1 levels conferred superior disease-specific survival (p < 0.0001) and metastasis-free survivals survival (p < 0.0001) in UTUC and UBUC patients (Fig. 1g-h). Also, GREM1 protein levels remained a better and independent prognostic factor for DSS (p = 0.005, p = 0.035) and MFS (p = 0.008; p = 0.034) UTUC (Supplementary Table S2) and UBUC patients (Supplementary Table S3). Therefore, secreted GREM1 is proportional to its cytosolic form, and high GREM1 levels independently predicted better DSS and MFS in UTUC and UBUC patients.

Fig. 1.

High GREM1 protein levels confer better survival rates in upper urinary tract carcinoma (UTUC) and urinary bladder urothelial carcinoma (UBUC). RT-PCR, immunoblot, enzyme-linked immunosorbent assay (ELISA), and concentrated media protein/immunoblot assays showed GREM1 mRNA and their corresponding protein levels in distinct UBUC-derived cells compared to a normal bladder cell line, SV-HUC-1. (a, b) Cell lysates, (c) conditioned media subjected to protein concentration, and (d) conditioned media (ELISA). (e) GREM1 concentrations were ELISA-measured using sera from UBUC patients (n = 47) and normal (non-tumor) individuals. (f) Immunohistochemistry shows the GREM1 protein levels in the representative non-tumor urothelium, low-stage, and high-stage tissues in UTUC and UBUC patients. (g, h) Kaplan Meir plotted disease-specific survival and metastasis-free survivals in UTUC (n = 61) and UBUC (n = 52) patients. Results are expressed as the mean ± SD. For immunoblot, pan-actin and Ponceau S staining served as a loading control. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

Hypermethylation on the GREM1 promoter results in low GREM1 mRNA levels in UBUC-derived cells and specimens

Data mining on the Bladder Urothelial Carcinoma (BLCA)/The Cancer Genome Atlas (p < 0.000, TCGA) using UALCAN [28] identified GREM1 hypermethylation in the primary tumors compared to normal tissues derived from UBUC patients (Fig. 2a). The cBioportal [29] next showed a negative correlation between GREM1 mRNA levels and the promoter methylation state by using the Infinium HumanMethylation450 beadarray (p < 0.000, Fig. 2b). A DNA methylation inhibitor, 5-Aza-2’-deoxycytidine (5-aza: 5 or 10 μM), evaluated whether demethylation could upregulate GREM1 levels. Treatments with 5-aza for 24 h in 4 UBUC-derived and one UTUC cell line (BFTC909) increased their GREM1 mRNA and their corresponding protein levels (Fig. 2c). Bisulfite pyrosequencing further identified hypermethylation of GREM1 promoter in J82, T24, UMUC3, and BFTC909 cells while hypomethylation in BFTC905 cells (Fig. 2d). Bisulfite pyrosequencing (#47 to #52) on 25 UBUC specimens identified that patients with high GREM1 transcripts were hypomethylated (p < 0.001, Fig. 2e). Compared to non-muscle-invasive and adjacent non-tumor tissues, respectively, muscle-invasive (p < 0.001) and non-muscle-invasive (p = 0.004) UBUC displayed hypermethylation (Fig. 2f). These observations suggested that hypermethylation on the promoter caused low GREM1 transcription in distinct UBUC-derived cell lines and UBUC patients, with stepwise increases of GREM1 methylations during tumor progression. Low GREM1 mRNA indeed devotes to low GREM1 protein levels.

Fig. 2.

Hypermethylation at the GREM1 promoter in UBUC patients. (a) Data mining on The Cancer Genome Atlas Program (TCGA) database through the UALCAN platform showed hypermethylation at the proximal promoter of the GREM1 gene in patients with primary UBUC compared to normal ones (p < 0.000, BLCA: Urothelial Bladder Carcinoma) [30]. (b) The cBioportal project demonstrated the negative correlation between GREM1 mRNA levels and the GREM1 methylation status in a UBUC cohort from TCGA (n = 408, p < 0.001). (c) Quantitative RT-PCR identified GREM1 mRNA after 5-Azacytidine (5-aza) treatments for 24 h in 5 distinct UBUC-derived cell lines. Immunoblot analysis displayed the GREM1 protein levels after treatments with 5-aza (5 μM) for 24 h. (d) In 4 UBUC-derived (J82, BFTC905, T24 & UMUC3) and one UTUC-derived (BFTC909) cell line, bisulfite pyrosequencing identified the methylation status (%) of 34 CpGs (#19 to #52) in the GREM1 promoter. (e) Methylation status (%, CpG: #47 to #52) in the GREM1 promoter (Y-axis) and the GREM1 mRNA levels (X-axis) in UBUC patients. (f) The methylation status (%, CpG: #47 to #52) and UBUC patients from different tumor stages. NMIBC: nonmuscle invasive bladder cancer, MIBC: muscle-invasive bladder cancer. Statistical significance: **p < 0.01, **p < 0.001

Manipulation on the GREM1 level alters the cellular physiology and related protein levels

Exogenous expression of the GREM1 gene in J82 cells markedly upregulated the GREM1 mRNA level (p < 0.001, Fig. 3a), cytosolic (Fig. 3b), and secreted (Fig. 3c) GREM1 protein levels compared to the control. GREM1-overexpressing induced G₁ cell-cycle arrest (p < 0.001, Fig. 3d), suppressed cell proliferation in a time-dependent manner (p < 0.05, Fig. 3e), colony formation/anchorage-independent cell growth (p < 0.001, Fig. 3f), cell migration (p < 0.01), and cell invasion (p < 0.001) (Fig. 3g). Moreover, J82 cells treated with conditioned media from GREM1-overexpressed J82 cells inhibited HUVEC tube formation (p < 0.05, Fig. 3h). In parallel, GREM1-overexpressing notably altered the protein levels of several cell-cycle [tumor protein p53 (TP53), phospho-TP53(S15)[pTP53(S15)], cyclin-dependent kinase inhibitor 1 A (CDKN1A), CDKN1B, and cyclin A2 (CCNA2)] and epithelial-mesenchymal transition (EMT) regulators, cadherin 1 (CDH1), CDH2, fibronectin 1 (FN1), vimentin (VIM), and twist family bHLH transcription factor 1 (TWIST1)] (Fig. 3i), and MMP2/9 activity (p < 0.001) (Fig. 3j). Conversely, GREM1 knockdown in BFTC905 cells with 2 distinct shRNA clones showed opposite results (Fig. 3k-s). These aspects implied that the GREM1 level is crucial in maintaining normal cellular physiology and the expression levels of cell-cycle and EMT regulators.

Fig. 3.

Manipulating the GREM1 gene expression altered cellular physiology and the expression levels of several cell cycle and EMT-related proteins in urinary bladder urothelial carcinoma (UBUC)-derived cells. (a-j) In J82 cells, stable Transfection of the pGREM1-HaloTag plasmid compared to the control (pHaloTag). (a) Quantitative RT-PCR displayed the GREM1 mRNA levels. Immunoblot analysis identified the GREM1 protein levels in cell lysates (b) and conditioned medium after concentration (c). (d) Flow cytometric with propidium iodide (PI) staining demonstrated the cell cycle distribution. (e-j) 5-Bromo-2’-deoxyuridine (BrdU), soft agar along with MTT, transwell migration, and transwell invasion, stained with crystal violet and read with OD₆₀₀, HUVEC tube formation using conditioned media (150 μL) from GREM1-overexpressing J82 cells, stained with Calcein AM, and quantified with the ImageJ software, immunoblot and MMP2/9 activity assays showed the cell proliferation (e) colony formation/anchorage-independent cell growth (f), cell migration, cell invasion (g), in vitro angiogenesis (h), and the changes of several cell-cycle and EMT-related proteins (i, j). The same experiments showed the phenotypes in GREM1-knockdown BFTC905 cells (k-r). The mean ± SD denotes the continuous data. For immunoblot assay, pan-actin, TUBA1B, or Ponceau S staining served as loading controls. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

The autocrine glycosylated-N42 GREM1 plays tumor-suppressive roles in vitro

Since GREM1 is a secreted protein, we next studied the secretory effects on cellular physiology in J82 and T24 cells Treatments with conditioned media collected from GREM1-overexpressed J82 cells (Fig. 4a) and human recombinant GREM1 protein (rGREM1, 3 μg) (Fig. 4b) inhibited cell migration (p < 0.05) and invasion (p < 0.05). Supplementary Fig. S4 identified the optimal anti-GREM1 antibody (3 μg to 100 ng/mL) to neutralize rGREM1. Therefore, treatment with anti-human GREM1 antibody in BFTC905 cells for 24 h enhanced cell migration (p < 0.05) and invasion (p < 0.05) compared to those treated with IgG (control) (Fig. 4c), suggesting that secreted-GREM1 suppressed in vitro metastatic phenotypes. Uniprot annotates that residue #42 of GREM1, asparagine (N), is a potential N-linked glycosylation site. Thus, we generated pGREM1(ΔSP)- and pGREM1(N42A)-HaloTag plasmids to delete the signal peptide and mutate N42 to A to abolish their secreted and glycosylated capacity, respectively. In J82 and T24 cells, quantitative RT-PCR (p < 0.001, Fig. 4d) and immunoblot analysis (Fig. 4e) displayed successful overexpression of GREM1(WT)-, GREM1(ΔSP)- and GREM1(N42A)-HaloTag mRNA and fusion proteins. However, the GREM1(ΔSP)-HaloTag transfectants are short of the secreted fusion protein (Fig. 4e). ELISA further identified high secreted GREM1-HaloTag protein levels in pGREM1(WT)- and pGREM1(N42A)-HaloTag transfectants while not in the pGREM1(ΔSP)-HaloTag transfectant (p < 0.001, Fig. 4f). For cellular phenotypes, GREM1(ΔSP) and pGREM1(N42A) restored GREM1(WT)-suppressed colony formation/anchorage-independent cell growth (p < 0.05, 9 d; Fig. 4g). Additionally, both GREM1(ΔSP) and GREM1(N42A) recovered GREM1(WT)-suppressed cell migration (p < 0.01), invasion (p < 0.001) (Fig. 4h), and HUVEC tube formation (p < 0.01, Fig. 4i). Another UBUC cell line, UMUC3, demonstrated comparable phenotypes, including cell proliferation, cell migration and invasion, and HUVEC tube formation upon the above manipulations (Supplementary Fig. S5). These facets suggested that secreted and N42-glycosylated GREM1 play tumor suppressor roles by inhibiting colony formation/anchorage-independent cell growth, EMT: migration and invasion, and in vitro angiogenesis in distinct UBUC-derived cell lines.

Fig. 4.

The autocrine glycosylated-GREM1 protein inhibited cell proliferation, migration, invasion, and HUVEC tube formation in UBUC-derived J82 and T24 cells. (a) Conditioned media was collected from J82 cells after stable Transfection of the pHaloTag and pGREM1-HaloTag plasmids, respectively. Treatments with the conditioned media (500 μL, upper chamber) for 24 h, transwell migration, and transwell invasion examined cell migration and invasion. (b) Treatments with PBS (control) and human recombinant GREM1 (rGREM1) dissolved in PBS (100 ng/mL) for 24 h, respectively, transwell migration and transwell invasion examined cell migration and invasion. (c) Treatments with IgG (control) and anti-human GREM1 antibody (αGREM1, 2 μg/500 μL, upper chamber), respectively, transwell migration and transwell invasion examined cell migration and invasion. We constructed two plasmids with deletion of the signal peptide (ΔSP) and mutation at residue #42 (from Asparagine to Alanine, N42A) of the GREM1 protein as pGREM1(ΔSP)-HaloTag and pGREM1(N42A)-HaloTag plasmids. Stable transfection of the pHaloTag (control), pGREM1(wild-type, WT)-, pGREM1(ΔSP)-, and pGREM1(N42A)-HaloTag plasmids into cells. (d) Quantitative RT-PCR showed transfection efficiencies. (e) Immunoblot analysis displayed cytosolic (cellular lysates) and secreted (concentrated conditioned media) GREM1 protein levels. (f) ELISA measured the GREM1 protein levels in the conditioned media. (g) Soft agar/MTT assays showed colony formation/anchorage-independent cell growth. (h) Transwell migration and transwell invasion assays evaluated cell migration and cell invasion. (i) J82 cells were stably transfected with the pHaloTag (control), pGREM1(WT)-, pGREM1(ΔSP)-HaloTag plasmids, respectively, and conditioned media were collected. Exactly 150 μL conditioned media were added into 96-well plates to observe the HUVEC (15,000 cells/cm²) tube formation/stained with calcein AM and quantified using the ImageJ software. Results are expressed as the mean ± SD. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

The autocrine glycosylated-GREM1 inhibits tumor growth and metastasis in vivo

We performed xenografting (n = 6, each group) in NOD/SCID mice with UMUC3 cells stably carrying genes that expressed mock (control), GREM1(WT), GREM1(ΔSP), and GREM1(N42A) proteins, respectively. Figure 5a shows that tumor growth of GREM1(WT) was the most slowly increased among all groups. After xenografting for 24 days, we sacrificed the mice and dissected the tumors. Likewise, tumors in the GREM1(WT) group were the most minor compared to those in other groups (Fig. 5b), suggesting that cytosolic and non-glycosylated GREM1 could not inhibit tumor size in vivo. In addition, H&E staining displayed xenografted tumors harboring GREM1(WT) with hypocellularity and more fibrosis (Fig. 5c). Immunohistochemistry probing anti-GREM1, -Ki-67, and -MYL9 antibodies demonstrated high GREM1 but low MYL9 protein levels in the GREM1(WT) group, while GREM1(ΔSP) and GREM1(N42A) restored MYL9 protein levels. In constrast, GREM1(WT) showed low Ki-67, whereas GREM1(ΔSP) and GREM1(N42A) rehabilitated Ki-67 levels (Fig. 5d). For tail-vein injection of UMUC3 cells carrying different genotypes (5 mice/group) in NOD/SCID mice, a Luminar II in vivo imaging system identified lung metastasis in GREM1(ΔSP) (3/5 mice) and GREM1(N42A) (1/5 mice) while not in the mock and GREM(WT) group (Fig. 5e). However, microscopic examination of tissue dissections detected numerous intravascular tumor emboli (arrow) without tumor growth within pulmonary veins in the mock group. Furthermore, those from the GREM1(WT) group were negative for pulmonary metastasis. Nevertheless, mice injected with tumor cells harboring GREM1(ΔSP) and GREM1(N42A) developed mass-forming lung metastasis (Fig. 5f). These phenomena implied that only GREM1(WT) inhibited tumor migration in vivo.

Fig. 5.

NOD/SCID mice models showed that the autocrine glycosylated-N42 GREM1 is critical to inhibit tumor growth and metastasis. UMUC3 cells stably carrying control (mock), GREM1(WT), GREM1(ΔSP), and GREM1(N42A), respectively, were used for xenografting (5 × 10⁶; n = 6/group) (a-c) and tail-vein injection (2 × 10⁶; n = 5/group) (d, e). (a) A caliper and the formula [V (mm³) = π/6(width)² (mm²) × length (mm)] measured the tumor size every 3–4 days to day 24 after xenografting. (b) Tumors in each group after mice were sacrificed. (c) H&E staining showed the histology of xenograft dissections. (d) Immunohistochemistry on UMUC3-xenografts probing anti-GREM1, anti-Ki-67, and anti-MYL9 antibodies in different groups. (e) Bioluminescence images captured by an in vivo imaging system (VIIS) showed lung metastasis on day 34 after tail-vein injection with cells harboring GREM1(ΔSP) and GREM1(N42A) genes. (f) H&E staining showed the histology of lung dissections. Statistical significance: **p < 0.01, ***p < 0.001

Epithelial/autocrine GREM1 plays a tumor suppressor through binding TGFB1 and inhibition of the TGFβ/BMP/SMAD pathway in distinct UBUC-derived cell lines

Since GREM1 is an antagonist of BMPs belonging to the TGFβ signaling, which plays a vital protagonist in EMT [31], we further explored its roles in EMT. Phalloidin staining exhibited well-formed F-actins microfilament bundles in the cytoplasm of HaloTag-overexpressing (control) but not in GREM1-HaloTag-overexpressing J82 cells (Fig. 6a). Overexpression and knockdown of the GREM1 gene in J82 and BFTC905 cells, respectively, decreased (p < 0.001) and increased (p < 0.001) a direct transcriptional BMP/SMAD target gene, ID1 (Supplementary Fig. S6). Dual-luciferase assays further identified that GREM1-overexpressing inhibited the SRE (p < 0.001) and BMP responsive element (BRE) in J82 cells (p < 0.001, Fig. 6b) while GREM1-knockdown in BFTC905 displayed the opposite effects (p < 0.001; Fig. 6c), suggesting that GREM1 may regulate ETM through the TGFβ/BMP/SMAD pathways. Furthermore, GREM1-overexpressing downregulated TGFBR1, phosphorylated/activated SMAD2 at serine 465 and 467 [pSMAD2(S465/S467)], pSMAD3(S423/S425), and pSMAD1/5(S463/S465) protein levels in J82 and/or T24 cells, while GREM1-knockdown showed the opposite pattern in BFTC905 cells (Fig. 6d).

Fig. 6.

GREM1 restrains EMT in distinct UBUC-derived cells by inhibiting the TGFβ/BMP/SMAD signaling. (a) Phalloidin staining (F-actin) examined actin polymerization in control (pHaloTag) and GREM1-overexpressed J82 cells (pGREM1-HaloTag). (b, c) Transfection of pGL4.17-(CAGA)₁₂ (SMAD binding element: SBE) and pGL3-BRE-Luc (BMP responsive element), respectively, in stable GREM1-overexpressing J82 cells and GREM1-knockdown BFTC905 cells for 48 h, dual-luciferase assay identified the promoter activities. (d) Immunoblot analysis showed the expression levels of several TGFβ signaling-related proteins upon manipulating GREM1 levels in J82, T24 (overexpression), and BFTC905 (knockdown) cells. (e–h) Stable Transfection of the pHaloTag (control), pGREM1- and pTBFB1-HaloTag plasmids, respectively, in J82 cells, for in vitro functional assays. (e) Soft agar/MTT assay for anchorage-independent cell growth, (f) transwell migration and invasion, crystal violet staining, read with OD₆₀₀, (g) ImageJ measured the HUVEC tube formation upon treatment with conditioned media, stained with calcein AM, (h) immunoblot analysis examined EMT-related protein levels. (i-k) In vitro, functional assays in cells stably transfected with the pHaloTag (control), pGREM1-HaloTag, and/or treatments with HCl (control; 4 mM for rBMP4 reconstitution), rBMP4/HCl, pGREM1-HaloTag and rBMP4, and cotransfection of pGREM1- and pBMPR2-HaloTag. (i) Soft agar/MTT assay for anchorage-independent cell growth, (j) transwell migration and invasion, (k) ImageJ measured the HUVEC tube formation. (l) Reciprocal coimmunoprecipitation assays examined the effects of GREM1 overexpression on TGFBR1/TGFB1 interactions compared to the control (pHaloTag). By probing anti-TGFBR1 antibody (αTGFBR1, left panel) and anti-TGFB1 antibody (αTGFB1, right panel), respectively, and immunoblotted using anti-TGFB1 antibody and anti-TGFBR1 antibody. (m) Transfection of the pTGFB1-HaloTag and pGREM1-DDK in J82 cells and collected the conditioned media. The pull-down assay examined the conditioned media by probing anti-HaloTag and anti-DDK antibodies, respectively, and immunoblot with anti-DDK and anti-HaloTag antibodies. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

To decipher the role of TGFβ family members in GREM1-mediated tumor suppression, we subsequently stably Transfection of pHaloTag (control), pGREM1-HaloTag, and pTGFB1-HaloTag, respectively, in J82 and/or T24 cells for functional assays. GREM1-overexpressing suppressed colony formation/anchorage-independent cell growth (p < 0.001; Fig. 6e), cell migration (p < 0.05), invasion (p < 0.001) (Fig. 6f), and HUVEC tube formation after treatment with conditioned media from J82 cells (p < 0.001; Fig. 6g); downregulated pSMAD2(S465/S467), CDH2, VIM, and TWIST1 while upregulated CHD1 protein levels compared to the control (#2 vs. #1; Fig. 6h). Instead, TGFB1-overexpressing increased colony formation/anchorage-independent cell growth (p < 0.01), cell migration (p < 0.01) and invasion (p < 0.01), and upregulated pSMAD2(S465/S467), CDH2, VIM, and TWIST1 while downregulated CDH1 protein level compared to the control (#3 vs. #1; Fig. 6e-h). Overexpression of both GREM1 and TGFB1 genes inhibited TBFB1-induced colony formation/independent cell growth (p < 0.001), cell migration (p < 0.001) and invasion (p < 0.001); downregulated pSMAD2(S465/S467), CDH2, VIM, and TWIST1 while upregulated CDH1 protein levels compared to the TGFB1-overexpressing group (#4 vs. #3; Fig. 6e-h). These indications suggested that GREM1 alleviated EMT by inhibiting the TGFB1 activity in UBUC-derived cells.

In parallel, to elucidate the characters of BMP family members in GREM1-mediated tumor suppression, we stably transfection of pHaloTag, pGREM1-, and pBMPR2-HaloTag (BMPR2: BMP receptor type 2), respectively, without or with rBMP4 in J82 cells (Fig. 6i-k). Furthermore, a dose-course experiment identified that treatment with 50 ng/mL of rBMP4 notably upregulated the pSMAD1/5(S463/S465) protein level (Supplementary Fig. S7). Compared to the control (pHaloTag + HCl), GREM1-overexpressing (pGREM1 + HCl) inhibited colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.001), invasion (p < 0.01), and HUVEC tube formation (p < 0.001) (#2 vs. #1). On the contrary, treatment with rBMP4 (pHaloTag + rBMP4/HCl) increased colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.05), and cell invasion (p < 0.01), and HUVEC tube formation compared to the control (#3 vs. #1). GREM1-overexpressing and rBMP4 treatment (pGREM1 + rBMP4/HCl) inhibited rBMP4-induced colony formation/anchorage-independent cell growth (p < 0.01), cell migration (p < 0.01) and invasion (p < 0.001), and HUVEC tube formation (p < 0.001) (#4 vs. #3). Compared to the control, overexpression of a BMP4 receptor, BMPR2 (HCl + pBMPR2), increased colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.001) and invasion (p < 0.001), and HUVEC tube formation (p < 0.001) (#5 vs. #1). In addition, overexpression of both GREM1 and BMPR2 genes (pGREM1 + pBMPR2) augmented colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.001) and cell invasion (p < 0.001), and HUVEC tube formation (p < 0.01) compared to the GREM1-overexpressing group (#6 vs. #2), while impeded BRPR2-induced colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.01) and cell invasion (p < 0.01), and HUVEC tube formation (p < 0.001) (#6 vs. #5). These features suggested that GREM1 may also inhibit cell proliferation and EMT by inhibiting the BMP signaling.

We next inspected whether GREM1 inhibited TGFβ/BMP/SMAD signaling by interacting with TGFB1 and its receptor, TGFBR1. In J82 cell lysates, reciprocal coimmunoprecipitation by probing anti-TGBFR1 and TGFB1 antibodies, and immunoblot with anti-TBFB1 and anti-TGFBR1 antibodies, respectively, were performed. Figure 6l showed that TGFB1 profoundly interacted with TGFBR1 (pHaloTag), while GREM1-overexpression reduced TGFB1-TGFBR1 interaction. In addition, pull-down assays in concentrated conditioned media from TGFB1-HaloTag- and GREM1-DDK-overexpressing J82 cells identified the interaction between TBFB1 and GREM1 (Fig. 6m). These aspects denoted that GREM1 binds TGFB1 to inhibit TGFB1-TGFBR1 interaction.

The epithelial/autocrine GREM1 inhibits TGFβ/BMP/SMAD-mediated EMT partially by transrepression of MYL9 in UC

As the TGFβ pathway leads to the activation of EMT through several different molecules, we next examined whether any novel effector was TGFβ/BMP/SMAD-mediated. Among the differentially expressed transcripts in GREM1-overexpressed J82 and GREM1-knockdown BFTC905 cells, we identified 8 overlapped transcripts within the gene set: HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION (Molecular Signatures Database, GSEA) [32] in both cell lines (Supplementary Fig. S8). Based on UniProt annotation [11], MYL9 is the only one with subcellular localization on the cytoskeleton among 8 candidates. Since quasi-mesenchymal cells exhibit front-rear polarity and extensively rearrange the actin stress fibers that form a critical part of their cytoskeleton [33], we spotlighted how GREM1 regulates the MYL9 and, after that, EMT.

Quantitative RT-PCR confirmed that GREM1-overexpressing J82 cells and GREM1–knockdown BFTC905 cells respectively downregulated and upregulated MYL9 mRNA levels (Fig. 7A). Datamining on the GSE31684 (n = 93, GEO, NCBI, Bethesda, USA) showed that high-stage (T2 and above) UBUC expressed higher MYL9 mRNA levels than low-stage patients (Ta and T1, p < 0.001 Supplementary Fig. S9). Immunohistochemistry further identified that high MYL9 protein levels contributed to low disease-specific survival (p < 0.0001; p = 0.0002) and metastasis-free survival (p < 0.0001; p < 0.0001) in UTUC (Fig. 7b) and UBUC (Fig. 7c) patients, respectively. Quantitative RT-PCR and immunoblot analysis examined the endogenous MYL9 mRNA and their protein levels in the normal SV-HUC-1 and 10 distinct UBUC- and UTUC-derived cell lines (Supplementary Fig. S10). Accordingly, we used BFTC905 and UMCU3 cells for MYL9 overexpression and J82 and T24 cells for MYL9 knockdown in further functional assays. Stable overexpression of the MYL9 gene upregulated its mRNA and protein levels (Fig. 7d). MYL9-overexpression also enhanced cell proliferation in time-dependent manners (p < 0.05, Fig. 7e), cell migration (p < 0.01) and invasion (p < 0.01) (Fig. 7f), and HUVEC tube formation (p < 0.01, Fig. 7g). Instead, MYL9-knockdown in J82 and T24 cells showed the opposite results (Fig. 7h-k).

Fig. 7.

GREM1 inhibited MYL9 transcription and EMT in distinct UBUC-derived cells. (a) Quantitative RT-PCR showed that GREM1-overexpressing decreased while GREM1-knockdown increased MYL9 mRNA levels in J82 and BFTC905 cells, respectively. (b, c) Immunohistochemistry on tissue specimens indicated that patients with low MYL9 protein levels have better disease-specific and metastasis-free survival rates in UTUC and UBUC. (d-g) Transduction of viruses packaged by cotransfection of the pLenti-MYL9, psPAX2, and pMD2.G plasmids in Phoenix-AMPHO cells, MYL9 was stably overexpressed in BFTC905 and UMUC3 cells. (d) Quantitative RT-PCR and immunoblot indicated that MYL9-overexpressing notably upregulated MYL9 mRNA and their corresponding protein levels. (e) BrdU assay showed that MYL9-overexpressing increased cell proliferation in time-dependent manners. (f) Transwell migration and transwell invasion indicated that MYL9-overexpressing increased cell migration and invasion. (g) ImageJ measured the HUVEC tube formation after treatment with conditioned media collected from MYL9-overexpressed cell cultures. (h–k) Stable knockdown of the MYL9 gene with 2 shRNA clones showed the opposite results compared to MYL9-overexpressed cells. The eukaryotic promoter database (EPD) identified several potential SMAD2:SMAD3:SMAD4 responsive elements in human MYL9 promoter regions conserved in mouse Myl9. (l) We cloned the human MYL9 promoter regions -700 to + 1, -400 to + 1, and -320 to + 1 into pGL4.17, as pGL4.17A, pGL4.17-B, and pGL4.17-C. (m) Dual luciferase assays identified that GREM1-overexpressing (pGREM1-HaloTag) impeded pGL4.17-C activities compared to the controls (pHaloTag) in UMUC3 and J82 cells. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

JASPAR in the Eukaryotic Promoter Database [34] identified 6 (human) to 10 (mouse) potential SMAD2:SMAD3:SMAD4 responsive elements in the promoter regions (-900 to + 1) of the MYL9 gene conserved across humans and mice (Supplementary Fig. S11). We cloned the MYL9 proximal promoter into the pGL4.17 reporter as the pGL4.17-A_{(-700 to +1)}, pGL4.17-B_{(-400 to +1)}, and pGL4.17-C_{(-320 to +1)} (Fig. 7l). Transfection of the pGL4.17-C in GREM1-overexpressing UMUC3 and J82 cells decreased the promoter activity compared to the HaloTag-overexpressing cells (p < 0.05, Fig. 7m). All the above suggested that GREM1 inhibited tumorigenesis at least partially by suppressing MYL9 transactivation.

Cytosolic GREM1 promotes tumor aggression by augmenting TNF/AKT/NFκB signaling in vitro

Next-generation sequencing examined the abundance of mRNA from GREM1(ΔSP)-overexpressing (signal peptide-deleted GREM1) J82 cells compared to the GREM1(WT)-overexpressing group. GSEA analysis highlighted the hallmarks: 'tumor necrosis factor A (TNFA)_signaling', 'inflammatory_response', and 'inteferron_gamma', which were positive-enriched (adjusted p < 0.0001) (Fig. 8a), implying that non-secreted GREM1 may play oncogenic roles through regulation of these pathways, distinct from the secreted form. Furthermore, in J82 cells, cytosolic GREM1(ΔSP)-overexpressing decreased NFKB inhibitor alpha (NFKBIA), AKT serine/threonine kinase 1 (AKT1), and RELA proto-oncogene (RELA) while increased active/phosphorylated pNFKBIA(S32/S36), pAKT1(S473), and nuclear pRELA(S536) protein levels compared to GREM1(WT)-overexpressing cells (Fig. 8b).

Fig. 8.

Cytosolic GREM1 plays oncogenic roles by activating the AKT-NFκB axis. (a) J82 cells were stably transfected with the pHaloTag (control) and pGREM1(ΔSP)-HaloTag plasmid, respectively, and cell lysates were subjected to RNA extraction and next-generation sequencing. Gene Set Enrichment Analysis (GSEA) highlighted the 'TNFA signaling via NFκB' as the top one HALLMARK [adjusted p < 0.0001, Enrichment Score (ES) = 0.61] compared to the control. (b) J82 cells were stably transfected with pHaloTag (control), pGREM1- and pGREM1(ΔSP)-HaloTag (deletion of the signal peptide) plasmid, respectively. Immunoblot analysis showed several protein levels related to the NFκB signaling pathway in total cell lysates and the nuclear fractionation from the pGREM1(ΔSP)-HaloTag compared to pGREM1-HaloTag and pHaloTag transfectants. (c-f) J82 and/or T24 cells were transfected with pGREM1-, pGREM1(ΔSP)-HaloTag and/or treatments with an inhibitor of IKBKB, IMD0354. Soft agar/MTT, transwell migration, transwell invasion, HUVEC tube formation with ImageJ, and immunoblot assays demonstrated the colony formation and anchorage-independent cell growth (c), cell migration and cell invasion (d), in vitro angiogenesis, and (f) levels of AKT-NFκB axis-related proteins. (g-j) Cells were treated with a PI3K inhibitor, LY294002, and transfected with the pGREM1(WT)- and pGREM1(ΔSP)-HaloTag plasmid, respectively. Soft agar/MTT assay, transwell migration and transwell invasion, HUVEC tube formation with ImageJ, and immunoblot examined colony formation and anchorage-independent cell growth (g), cell migration and invasion (h), and HUVEC tube formation (i), and the expression levels of AKT-NFκB axis-related phosphorylated/active proteins [phospho-AKT1(S473): pAKT1(S473) and pRELA(S536)] (j). For immunoblot analysis, pan-actin and PARP1 were total and nuclear controls, respectively. Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001

To elucidate whether cytosolic GREM1 promotes tumorigenesis in vitro mediated by TNF/Nuclear Factor kappa B (NFκB) signaling, we treated with an Inhibitor of nuclear factor kappa B kinase subunit beta (IKBKB, aka IKKβ) inhibitor, IMD0354, in GREM1(WT)- and GREM1(ΔSP)-overexpressing J82 and/or T24 cells (Fig. 8c-f). Compared to the GREM1(WT) group, cytosolic GREM1(ΔSP) enhanced colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.01) and invasion (p < 0.01), HUVEC tube formation (p < 0.001); upregulated active/phosphorylated pNFKBIA(S32/S36) and nuclear pRELA(S536) protein levels (#2 vs. #1). GREM1(ΔSP)-overexpressing combined with IMD0354 treatment suppressed GREM1(ΔSP)-induced colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.001) and cell invasion (p < 0.001), HUVEC tube formation (p < 0.001) and downregulated pNFKBIA(S32/S36), and nuclear pRELA(S536) protein levels (#3 vs. #2), suggesting that cytosolic GREM1-induced oncogenic characters by promoting the TNF/NFκB pathway.

We further treated with an inhibitor, LY294002, in GREM1(WT)- and GREM1(ΔSP)-overexpressing J82 and/or T24 cells, respectively, to evaluate whether cytosolic GREM1 plays oncogenic roles was TNF/AKT-interceded (Fig. 8g-j). Compared to the GREM1(WT)-overexpressing cells, cytosolic GREM1(ΔSP)-overexpressing enhanced colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.05) and cell invasion (p < 0.01), HUVEC tube formation (p < 0.001) and upregulated active/phosphorylated pAKT1(S473) and nuclear pRELA(S536) levels (#2 vs. #1). GREM1(ΔSP)-overexpressing combined with LY294002 treatment suppressed GREM1(ΔSP)-induced colony formation/anchorage-independent cell growth (p < 0.001), cell migration (p < 0.001) and cell invasion (p < 0.001), HUVEC tube formation (p < 0.001), and downregulated pAKT1(S473) and nuclear pRELA(S536) protein levels. These notations suggested cytosolic GREM1-induced oncogenic characters in vitro by promoting the TNF/AKT/NFκB pathway (Fig. 9).

Fig. 9.

Summary of this study

Discussion

In the present study, we identified that the autocrine glycosylated-GREM1 is a tumor suppressor in urothelial carcinoma. Previous studies showed that a secreted form of GREM1 is 28 kDa and represents the glycosylated form [35]. Our experiments in mouse models and UBUC-derived cells reinforced the clinical correlations. Promoter hypermethylation is the primary cause of low GREM1 protein levels in UBUC patients and UC-derived cell lines, similar to earlier reports regarding several tumor suppressors, including mutL homolog 1 (MLH1), O-6-methylguanine-DNA methyltransferase (MGMT), CDKN2A, CDKN2B, death-associated protein kinase 1 (DAPK1), and CDH1. So far, associations between epithelial GREM1 levels and tumorigenesis are controversial [12]. Former studies used immunohistochemistry, immunoblot analysis, RNA in situ hybridization, and quantitative RT-PCR to evaluate GREM1 protein and GREM1 mRNA levels and correlate them to distinct phenotypes. However, neither cytosolic GREM1 nor GREM1 mRNA is the functional form, leading to inconsistent interpretations. Our genetic engineering to remove the signal peptide and mutation on the potential glycosylation site of the GREM1 protein in cellular and animal models revealed the autocrine and tumor-suppressive roles of epithelial GREM1.

For the first time, we reported that autocrine GREM1 interacts with TGFB1, thus interrupting TGFB1/TGFBR1 binding, SMAD2/3 activation, and downstream pathways. The conservations of 3D [36] and 2D [37] structures between BMPs and TGFBs support our identification. Therefore, interactions among members of the TGFβ superfamily have expanded further. Traditional targets of GREM1 are BMPs and growth differentiation factors (GDFs) interacting with BMPRs and activin A receptors [38]. Interactions between GREM1 and BMP2/4/7 and inhibition of the BMP signaling are well-documented. Recent studies showed that BMP4 induced M2 macrophage polarization and favored tumor progression in bladder cancer [39]. Activated BMPR/TGFBR initiates downstream signaling by phosphorylating the intracellular mediators, receptor-regulated SMADs (R-SMADs: SMAD1/5/8 and SMAD2/3), to promote their binding to co-SMAD (SMAD4) and nuclear translocation as a heteromer to transactivate BMP and TGFB target genes, such as ID1 and MYL9, respectively, in this study, to regulate cell fate and differentiation [40]. In the mouse model, Grem1 is the principal BMP antagonist required for early limb outgrowth and patterning [41]. BMP, but not TGFβ, strongly activates the mouse Id1 promoter in a Smad-dependent manner [42]. The ID1 is a helix-loop-helix (HLH) protein, lacking a basic DNA binding domain yet forming heterodimers with basic HLHs (bHLHs), thus inhibiting their transcriptional activities. ID1 impacts cellular growth, senescence, differentiation, apoptosis, angiogenesis, and neoplastic transformation [43]. Besides, the overexpression of ID1 in several types of tumors promotes their progression through various pathways [44], reinforcing our findings.

Our study pointed out that epithelial and autocrine GREM1 inhibited TGFβ-mediated EMT in cellular and animal models. Notably, TGFβ induces EMT in both epithelial and endothelial cells and results in disease progression in cancer and fibrosis [31]. As cell migration requires losing cell–cell contacts and gaining fibroblastic properties, EMT boosts cell migration and invasion. The vertebrate genome contains more than 30 pleiotropic ligands that belong to the TGFβ superfamily, including TGFBs, BMPs, GDFs, activins, nodal growth differentiation factor (NODAL), and anti-Mullerian hormone (AMH) [45]. TGFβ involve numerous cell processes, comprising cytostasis, EMT (cell migration and invasion), extracellular matrix remodeling, and immune suppression [46]. Once EMT induction, the activated SMADs mediate transcriptional regulation through transcription repressors, snail family transcriptional repressor 1 (SNAI1), zinc finger E-box binding homeobox 1 (ZEB1), and HLHs, including ID1 [31]. Recent studies on mice showed that Grem1 inactivation in established pancreatic ductal adenocarcinoma (PDAC) caused the direct conversion of epithelial into mesenchymal PDAC cells within days, suggesting constant Grem1 activity is essential to maintain the epithelial PDAC subpopulations [47]. All the above studies and our manipulations on the GREM1 levels in distinct UBUC-derived cells showed positive correlations between GREM1 and CDH1 and negative correlations between GREM1 and FN1, CDH2, VIM, and TWIST protein levels, supporting that epithelial GREM1 suppresses EMT through inhibiting of the TGFβ/BMP/SMAD pathway.

We mainly found that cytosolic/non-secreted GREM1 played oncogenic roles in the xenograft and metastatic mice. In vitro evidence in this study also revealed that cytosolic GREM1 enhanced cell proliferation, migration, invasion, and HUVEC tube formation in distinct UBUC-derived cell lines by activating the TNF/AKT/NFκB pathway. TNF is one of the most intensively studied immune system cytokines, which involves cellular communication, differentiation, survival, and death [48]. Through binding to TNF receptor superfamily member 1B (TNFRSF1B), this complex activates the TNF receptor-associated factor 1 (TRAF1)/2/3 and phosphoinositide 3-kinase (PI3K). Subsequently, TRAF1/2/3 phosphorylates phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), AKT, and NFKB inhibitor alpha (NFKBIA/IκBα). Beta-transducin repeat containing E3 ubiquitin protein (BTRC) later targets NFKBIA for ubiquitination and degradation [49]. The dimer of NFKB1 (p50) and RELA (p65) is the most abundant form of the NFκB, which is a ubiquitous, inducible transcription factor involved in various physiological processes. In non-stimulated cells, NFκB retains a dormant form in the cytoplasm by inhibitory molecules such as NFKBIA [50]. Once activated, NFκB is released and translocated into the nucleus. It transactivates many genes, including TNF [51], in response to infections, inflammation, and other stressful situations [52], thus forming a positive feedback loop. Due to TNF being a cytokine, cytosolic GREM1 must initiate the TNF signaling in the cytoplasm; perhaps through protein–protein interactions, cytosolic GREM1 triggers AKT and NFKBIA phosphorylation, resulting in NFκB dissociation and activation.

We additionally highlighted the glycosylation on GREN1 residue #42 (asparagine) is conclusive for its functionality as a tumor suppressor. Even loss-of-function mutation GREM1(N42A) remains its secretion character, whereas it cannot inhibit tumorigenesis in vivo and in vitro. Glycosylation of specific proteins in cells modulates inflammatory responses, viral immune escape, tumor metastasis, and apoptosis [53]. Furthermore, N-glycosylation promotes the formation of secondary structural components required for protein folding, resists proteolysis, and modifies its half-life/stability [54]. Accordingly, loss of GREM1(N42) glycosylation might hurdle the interaction with its cognate ligands, which in turn is unable to inhibit TGFβ/BMP/SMAD signaling and related phenotypes. One similar example in the same pathway is BMPR2. Site-directed mutagenesis revealed unique glycosylation (N110) on the BMPR2 protein near its ligand-binding hydrophobic patch. Furthermore, patients with heritable pulmonary arterial hypertension showed a mutation at this glycosylation position [55]. In parallel to these disclosures, our studies strongly advocated that only secreted and glycosylated GREM1 encompasses tumor suppressor roles.

We moreover identified MYL9 as a novel TGFβ/BMP/SMAD-transactivated target, and clinical and in vitro evidence suggest that high MYL9 levels promote UC progression. Myosin is a structural component of muscle, comprising 2 heavy and 4 light chains. As a myosin light chain, MYL9, regulates smooth muscle and nonmuscle contraction by modulating the ATPase activity of myosin heads. Myosin light chain kinase activates MYL9 and implicates cytokinesis and cell locomotion [56], potentially contributing to cancer advancements. Indeed, recent studies showed the emerging roles of MYL9 on esophageal squamous cell carcinoma [57], pancreatic ductal adenocarcinoma [58], epithelial ovarian cancer [59], and glioblastoma [60] consisting to our observations. However, unlike the GREM1, MYL9 protein level did not independently denote disease-specific and metastasis-free survival rates, suggesting the existence of other involving machinery downstream of the GREM1 gene.

Conclusions

Overall, our study demonstrates that low GREM1 levels confer an independent and worse prognostic factor in UTUC and UBUC. Promoter hypermethylation in the GREM1 gene hampers its transcription and translation in UC-derived cells and UBUC patients. Especially, secreted and glycosylated GREM1 plays autocrine tumor suppressor roles in inhibiting cell proliferation, colony formation/anchorage-independent cell growth, cell migration, cell invasion, and in vitro angiogenesis through suppression of the GREM1/TGFB/SMAD signaling, partially via transrepression of the MYL9. The autocrine glycosylated-GREM1 also inhibits tumor growth and metastasis in mouse models. Manipulating the GREM1 levels in distinct UBUC-derived cell lines alters the expression levels of several proteins involving cell cycle regulation and EMT. Additionally, autocrine GREM1 interacts with TGFB1 to prevent its interaction with TGFBR1 and subsequently deactivates SMAD signaling. Our studies highlight the tumor-suppressive roles of the autocrine glycosylated-GREM1 in UC.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

5-aza

5-Aza-2’-deoxycytidine

AKT1

AKT serine/threonine kinase 1

AMH

Anti-Mullerian hormone

bHLHs

Basic HLHs

BLCA

Bladder urothelial carcinoma

BMP

Bone morphogenetic protein

BMPR2

Bone morphogenetic protein receptor type 2

BrdU

5-Bromo-2'-deoxyuridine

BRE

BMP responsive element

BTRC

Beta-transducin repeat containing E3 ubiquitin protein

CDH1

Cadherin 1

CDKN1A

Cyclin-dependent kinase inhibitor 1 A

DAPK1

Death-associated protein kinase 1

DSS

Disease-specific survival

EMT

Epithelial-mesenchymal transition

FN1

Fibronectin 1

GDF

Growth differentiation factors

GEO

Gene Expression Omnibus

GREM1

Gremlin 1, DAN family BMP antagonist

GSEA

Gene Set Enrichment Analysis

HUVEC

Human umbilical vein endothelial cells

ID1

Inhibitor of DNA binding 1

IKBKB

Nuclear factor kappa B kinase subunit beta

MFS

Metastasis-free survival

MGMT

O-6-methylguanine-DNA methyltransferase

MLH1

MutL homolog 1

MMP

Matrix metalloproteinase

MTT

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

MYL9

Myosin light chain 9

NFκB

Nuclear factor kappa B

NFKBIA

NFKB inhibitor alpha

NGF

Nerve growth factor

NODAL

Nodal growth differentiation factor

PDAC

Pancreatic ductal adenocarcinoma

PDGF

Platelet-derived growth factor

PI3K

Phosphoinositide 3-kinase

PIK3CA

Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha

RELA

RELA proto-oncogene

RPKM

Read per kilobase per million

shRNA

Small hairpin RNA

SNAI1

Snail family transcriptional repressor 1

SRE

SMAD responsive element

TCGA

The Cancer Genome Atlas

TGFβ

Transforming growth factor beta

TGFBR

Transforming growth factor beta receptor

TNF

Tumor necrosis factor

TNFRSF1B

TNF receptor superfamily member 1B

TP53

Tumor protein p53

TRAF1

TNF receptor-associated factor 1

TWIST1

Twist family bHLH transcription factor 1

UBUC

Urinary bladder urothelial carcinoma

UC

Urothelial carcinoma

UTUC

Upper tract urothelial carcinoma

VIM

Vimentin

ZEB1

Zinc finger E-box binding homeobox 1

Authors' contribution

CFL, WJW, YLS conceived the concepts; CFL, CTP, YLS designed the experiments; TCC, CTP, HYH, PV, RJW, BWY, LRC, and MSS performed the experiments; CTP and CFL analyzed the data. CFL, CTP and YLS wrote the manuscript. All authors read and approved the final manuscript.

Funding

The National Science and Technology Council (MOST-107–2314-B-110-MY3), NSYSU-KMU Joint Research Project (#NSYSUKMU 109-I008), Kaohsiung Armed Forces General Hospital (KAFGH-A-108021 & -109038), and Kaohsiung Medical University Research Center (KMU-TC111A02-0) supported this research.

Data availability

The raw data used to support the conclusions of this article will be made available by the corresponding author, without undue reservation to any qualified researcher.

Declarations

Human data and samples were obtained with informed consent, and the use was approved (IRB10207-001) by the Institutional Review Board of the Chi Mei Medical Center, Tainan, Taiwan. Animal treatments (#10626) were performed according to the Institutional Animal Care and Use Committee of National Sun Yat-sen University (NSYSU) Protocol, and NSYSU approved all protocols. This study does not contain any studies involving human participants performed by any of the authors.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Conflict of interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.