Transmembrane and Coiled-Coil Domain 1 Impairs the AKT Signaling Pathway in Urinary Bladder Urothelial Carcinoma: A Characterization of a Tumor Suppressor

Urinary bladder urothelial carcinoma (UBUC) is a common malignant disease, especially in developed countries (1). Both environmental and genetic factors affect UBUC development (2, 3). Clinicopathologic features, including histologic grade, stage, size, and multiplicity, are associated with its progression (4). Despite improvements in surgical techniques and multimodal therapy, 5-year survival rates for patients with muscle-invasive UBUC remain suboptimal. Almost 50% of patients eventually progress and develop systemic disease (5). Clinical and genetic heterogeneity observed in UBUC patients further complicates the use of general therapies (6). Cell-cycle dysregulation resulting in uncontrolled cell proliferation has been associated with UBUC development (7, 8). Thus, targeting a critical molecule for therapies is a rational approach for UBUC treatment (9).

To identify transcripts that are potentially involved in UBUC development, data mining on established expression profiles (GSE31684, n = 93) in the Gene Expression Omnibus database (GEO, NCBI) was performed. One differentially expressed transcript, transmembrane and coiled-coil domain 1 (TMCO1), was identified to be associated with growth factor activity (GO:0008083; molecular function) and was highly expressed in pTa (P = 0.0006) and pT1 (P = 0.0174), compared with pT2-T4 patients with UBUC. The human TMCO1 gene, mapped to human chromosome 1q24.1, encodes a transmembrane protein (10), is a member of the DUF841 superfamily of several eukaryotic proteins with unknown function or involvement in any biological process (11). The 3D structure of TMCO1 protein has not been resolved, yet its topology contains 2 transmembrane domains (#10-30; #91-111) and 1 intramembrane fragment (#138-154; ref. 12). Green fluorescent protein–tagged and Myc-tagged TMCO1 were found to be expressed in the endoplasmic reticulum (ER) and/or the Golgi apparatus of COS7 and HeLa cells (12, 13). TMCO1 mRNA is highly expressed in porcine heart, liver, and kidney (14). Data mining, and the fact that membrane proteins constitute the largest class of drug targets (15), prompted us to systematically analyze the relevance of TMCO1 immunoexpression and clinicopathologic features in UBUC patients and its biological significance in vitro and in vivo.

The procedure for data mining the GEO database to identify downregulated transcripts in UBUCs is described in the Supplementary Materials and Methods. For immunohistochemistry, the Institutional Review Board of Chi Mei Medical Center approved the retrospective retrieval of 295 primary UBUCs with available tissue blocks (IRB10207-001), which underwent surgical treatment with curative intent between January 1998 and May 2004. These patients received surgical resection with curative intent between 1998 and 2004, whereas those who underwent palliative resection were excluded. Patients with confirmed or suspected lymph node metastasis received regional lymph node dissection. Cisplatin-based postoperative adjuvant chemotherapy was performed in patients with pT3-pT4 status or nodal involvement. The histologic diagnosis of UBUCs was confirmed in all cases based on the latest World Health Organization classification. Histologic grading was assigned on the basis of Edmonson–Steiner criteria, whereas tumor staging was determined according to the 7th edition of the American Joint Committee on Cancer system. Medical charts were reviewed for each patient to ascertain the accuracy of other pertinent clinicopathologic data. Follow-up information was available in all cases with a median period of 42 months (ranging 3–176 months). To determine the clinical relevance of the TMCO1 transcript level, an independent cohort comprised of 15 pTa-T1 and 15 pT2-T4 tumors, and 9 nontumor urothelial samples were enrolled and evaluated by branched DNA assay. Immunohistochemical staining was performed on representative tissue sections cut from formalin-fixed, paraffin-embedded tissues at 3-μm thickness as in our previous study (16) with a few modifications (Supplementary Materials and Methods).

Human normal urothelial cells (HUC; #4320, ScienCell Research Laboratories) were obtained and cultured with the recommended medium (#4321, ScienCell Research) in poly-l-lysine–coated flasks (2 μg/cm²). The human UBUC-derived cell lines RT4 (Food Industry Research and Development Institute, Hsinchu, Taiwan), J82 (ATCC), T24 (ATCC), and BFTC905 and BFTC909 (kindly provided by Dr. C.C. Tzeng; ref. 17) were respectively maintained in McCoy's 2A, DMEM, DMEM, RPMI-1640, and DMEM supplemented with 10% (v/v) FBS (Biological Industries), appropriate nutrients, and antibiotics in a humidified incubator with 5% CO₂ at 37°C. All media were obtained from CORNING. RT4 (18) and BFTC905 (19) were characterized with wild-type TP53; however, J82 was an allele-specific mutation of TP53 (20). The pan-PH domain and leucine-rich repeat protein phosphatase (PHLPP) inhibitor, NSC117079 [1-amino-9,10-dioxo-4-(3-sulfamoylanilino)anthracene-2-sulfonic acid], was obtained from AOBIOUS (Gloucester). All cell lines were authenticated by short tandem repeat genotyping, periodically confirmed to be mycoplasma-free using PlasmoTest (Invivogen) in IMDM (Invitrogen) supplemented with 15% FBS, 100 U/mL.

Quantitative RT-PCR assay was applied to quantify the expression levels of several transcripts using predesigned TaqMan assay reagents, including TMCO1 (Hs00976965_m1, 123 bp), tumor protein p53 (TP53; Hs01034249_m1, 108 bp), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Hs03929097_g1, 58 bp, internal control), along with a LightCycler (Roche Life Science) and ΔΔC_T calculation (21). Immunoblot analyses were performed as in our previous study (Supplementary Materials and Methods).

The pCMV6-TMCO1 (RC200219; NM_019026.4) plasmid was purchased from OriGene Technologies. The TMCO1 complete DNA (564 bp) was subcloned into a pHTC HaloTag CMV-neo vector (pHaloTag) using 5′-CTAGCTAGCATGAGCACTATGTTCGCGGA-3′ and 5′-CCGCTCGAGAGAGAACTTCCCAGAAGGAGGT-3′ primers with Nhe I and Xho I restriction sites (underlined) to generate the pTMCO1-HaloTag plasmid. The pcDNA3-PHLPP2 (#22403) and pHRIG-AKT1 (#53583) plasmids were obtained from Addgene (22, 23). All plasmids were sequence verified. Cells (5 × 10⁵) were transfected with 2.5 μg of a specific plasmid by mixing with 7.5 μL of PolyJet reagent (SignaGen Laboratories) in Opti-MEM (Life Technology). Transfectants were selected with medium containing 800 μg/mL of G418 (AMRESCO) for 7 days and maintained in medium with 400 μg/mL of G418 for subsequent experiments. The same protocol was used for transient transfections without selection by G418.

Small hairpin RNA interference (shRNAi) plasmids were inserted into the pLK0.1 vector downstream of the U6 promoter. Clones were obtained from the National RNAi Core Facility, Institute of Molecular Biology, Academia Sinica, Taipei, Taiwan. A total of 5 plasmids targeting TMCO1 gene were preliminarily screened. The TMCO1 mRNA levels could be effectively downregulated by only 2 clones. The plasmids shTMCO1#3 (TRCN0000062125: 5′-CCCTAATGGGAATGTTCAATT-3′) and shTMCO1#5 (TRCN0000062127: 5′-CATCGAAATCTGCTGGGAGAT-3′) were used for knockdown of the TMCO1 gene, and shLuc (TRCN0000072243:5′-CTTCGAAATGTCCGTTCGGTT-3)′ was used as a negative control clone. For stable shRNAi, lentiviral particles were produced. Experimental details are shown in Supplementary Materials and Methods.

Flow cytometric, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), bromodeoxyuridine (BrdUrd), and soft-agar assays were used to determine alternations of cell-cycle distribution, cell viability, cell proliferation, and colony formation/anchorage-independent cell growth after the overexpression and knockdown of the TMCO1 gene in vitro following our previous protocols (21). Cell migration and invasion were analyzed using the wound-healing assay and the QCM ECMatrix Cell Invasion Kit (ECM554, Millipore). For the above assays, details are described in Supplementary Materials and Methods.

To examine whether TMCO1 colocalizes or interacts with PHLPP1 or PHLPP2, immunocytochemistry and coimmunoprecipitation (co-IP) were performed using our previous procedures (24), and details are described in Supplementary Materials and Methods.

Three plasmids, pTMCO1(T33A)-, pTMCO1(S60A)-, and pTMCO1(S84A)-HaloTag, with mutations at residues #33, #60, or #84 from threonine/serine to alanine of the TMCO1 protein were constructed using the QuikChange Lightning Site-Directed Mutagenesis Kit (#210518, Agilent) and verified by DNA sequencing. The plasmid containing the wild-type TMCO1 gene, pTMCO1(WT)-HaloTag, served as the template for site-directed mutagenesis on residues T33, S60, and S84 using PCR-based technology. Details of primers sequences are listed in Supplementary Materials and Methods.

AKT activity was analyzed by using the KinaseSTAR Akt Activity Assay Kit (#K435-40; BioVision) according to the manufacturer's instruction. Briefly, BFTC905 cells (5 × 10⁵) were seeded and transfected overnight with pHaloTag, pcDNA3-HA-PHLPP2, and/or 3 TMCO1-mutated plasmids. For each assay, 2 μL of AKT-specific antibody was joined to 250 μg protein. Protein A-Sepharose slurry was applied to capture AKT-specific antibody, and 2 μL of the recombinant glycogen synthase kinase 3 alpha protein (GSK3A, substrate)/ATP mixture was added. Protein A-Sepharose were subsequently spun down, and the supernatant was collected and subjected to immunoblot analysis by probing anti-GSK3A and anti-pGSK3A(S21) antibody.

Animal experiments were approved (#10615) by Affidavit of reviewing of Animal Use Protocol, National Sun Yat-sen University. Cells were implanted into 40 NOD/SCID mice (8 for each group) by subcutaneous injection. BFTC905 cells (1.5 × 10⁷) stably overexpressing either pHaloTag, pTMCO1(WT)-, pTMCO1(T33A)-, TMCO1(S60A)-, or TMCO1(S84A)-HaloTag were resuspended in 100 μL of PBS, mixed with 100 μL of matrigel (BD Biosciences), and introduced into the right flank of 7-week-old, male mice. Tumor diameters were measured with a digital caliper every other day, and the tumor volume in mm³ was calculated as volume = π/6(width)² × length. Whole sections from formalin-fixed xenograft samples were analyzed by immunohistochemistry using pertinent antibodies described in Supplementary Materials and Methods.

All calculations were performed using SPSS 14.0 software. To determine the prognostic impact of selected transcripts identified in GSE31684, the deposited cases were subdivided into two clusters based on the expression level of each transcript, detected by a specific probe and computerized by k-means clustering (k = 2). The survival difference of the two clusters was next calculated by the log-rank analysis and plotted using the Kaplan–Meier method for overall survival. The association and comparison between various clinicopathologic factors and TMCO1 expression were assessed by the χ² test. The endpoint analyzed for survival analysis was disease-specific and metastasis-free survivals. The Student t test was used to examine the significance of differences in fold changes of mRNA levels, cells in different phases of the cell cycle, percentages of cell viability, proliferation, and anchorage-independent cell growth. Comparison of xenograft tumor sizes was performed by using one-way ANOVA. For other analyses, two-sided tests of significance were used, and a P value of < 0.05 was considered to be statistically significant.

From the transcriptomic profiles of 308 UBUCs deposited in the GEO database using Illumina HumanHT-12 V3.0 Expression BeadChip for analysis, GSE32894, the TMCO1 transcript, was found to be significantly downregulated in muscle-invasive compared with non–muscle-invasive UBUCs (Fig. 1A). In another dataset, GSE31684, containing 93 UBUC specimens, the downregulation of TMCO1 transcript predicted inferior overall survival was also identified (Fig. 1B and Supplementary Table S1); this result further suggested that the TMCO1 gene might function as a tumor suppressor in UBUCs. As shown in Fig. 1C, the TMCO1 mRNA level was highly expressed in nontumor (P = 0.001) and low-stage (pTa-T1; P = 0.004), compared with high-stage (pT2-T4) UBUC patients. High TMCO1 expression was also identified in nontumor urothelium and noninvasive urothelial carcinomas compared with muscle-invasive carcinomas (Fig. 1D). Correlations between TMCO1 expression and various clinicopathologic factors are listed in Supplementary Table S2. Univariate log-rank analysis identified that pT, nodal metastasis, histologic grade, vascular invasion, perineural invasion, mitotic rate, and TMCO1 immunostainings were significantly correlated with disease-specific and metastasis-free survivals in 295 UBUC patients (Table 1). Kaplan–Meier plots revealed that low TMCO1 protein levels predicted poor disease-specific survival (P = 0.0001; Fig. 1E) and metastasis-free survival (P < 0.0001; Fig. 1F). Multivariate analysis additionally demonstrated that pT, mitotic rate, and TMCO1 protein level significantly correlated to disease-specific survival; pT, nodal metastasis, and TMCO1 protein level considerably correlated with metastasis-free survival (Table 1). These results suggest that low TMCO1 protein level confers an independent prognostic indicator in UBUC patients. Array comparison genomic hybridization (aCGH) was performed, as analyzed (40 UBUCs) in our previous study (25), and showed frequent DNA copy-number gain at loci spanning TMCO1 gene at 1q24.1 (9/40, 22.5%), excluding the possibility of TMCO1 gene deletion in UBUC patients (Supplementary Fig. S1). Thus, downregulation of TMCO1 protein is an independent prognostic factor in UBUCs.

The TMCO1 mRNA levels were highly expressed in HUC, RT4, and J82 cells and low in BFTC905 cells (Fig. 2A). In addition, TMCO1 protein levels were high in RT4 and J82 cells and low in BFTC905 (Fig. 2B). Therefore, BFTC905 cells were used for overexpression, and RT4 and J82 cells were used for knockdown of the TMCO1 gene for functional studies in vitro. No methylation in the TMCO1 promoter region was found in RT4, T24, J82 cell lines, nontumor urothelium (n = 8), and UBUCs with high (n = 7) and low (n = 7) TMCO1 protein levels (Supplementary Materials and Methods; Supplementary Table S3). Exogenous expression of the TMCO1 gene in BFTC905 cells resulted in stable expression of the TMCO1-HaloTag fusion protein (Fig. 2C); this stably expressed protein-induced G₁ cell-cycle arrest (P < 0.05) and decreased the number of cells in S phase (P < 0.01; Fig. 2D) as well as suppressed cell viability (P < 0.001; Fig. 2E), cell proliferation (P < 0.001; Fig. 2F), and colony formation (Fig. 2G)/anchorage-independent cell growth (P < 0.001; Fig. 2H). Alternately, stable knockdown of the TMCO1 gene in RT4 cells inhibited TMCO1 mRNA (P < 0.001) and protein levels, induced cell-cycle progression to S phase (P < 0.001), and increased cell viability (P < 0.01), cell proliferation (P < 0.01), and colony formation/anchorage-independent cell growth (P < 0.001; Fig. 2I–N). These findings suggest that TMCO1 functions as a tumor suppressor by regulating cell-cycle progression in vitro.

Stable exogenous expression of the TMCO1 gene in BFTC905 cells notably upregulated the protein levels of the TMCO1-HaloTag, RB transcriptional corepressor 1 (RB1), TP53, pTP53(S15), cyclin-dependent kinase inhibitor 1A (CDKN1A), and CDKN1B; however, TMCO1 stable expression also downregulated cyclin D1 (CCND1), cyclin-dependent kinase 4 (CDK4), CCNE1, and CDK2 protein levels (Fig. 2O), as well as reduced the following ratios: pCDKN1A(T145) (inactive form)/CDKN1A and pCDKN1B(T157)(inactive)/CDKN1B (0.26 and 0.25; Supplementary Fig. S2A). A declined pCDKN1A(T145)/CDKN1A or pCDKN1B(T157)/CDKN1B ratio indicates an increase in the corresponding active form. Upregulation of nuclear TP53 and CDKN1A, rather than the cytosolic forms, accounted for the total TP53 and CDKN1A levels (Fig. 2P). However, TP53 mRNA levels remained unchanged (NS; Supplementary Fig. S2B), suggesting that TMCO1 might stabilize TP53 at the protein level. Accordingly, the expression levels of MDM2, an E3 ubiquitin ligase of TP53, and its upstream regulators, AKT and pAKT1(S473) (26), were next examined. Substantial downregulation of pAKT1(S473), MDM2, and pMDM2(S166) proteins was found in TMCO1-overexpressed BFTC905 (Fig. 2Q).

Stable TMCO1-knockdown RT4 cells exhibited an opposite protein expression pattern compared with TMCO1-overexpressed BFTC905 cells (Fig. 2R). The ratios of pCDKN1A(T145) (inactive)/CDKN1A (shTMCO1#3: 0.70; shTMCO1#5: 0.89) and pCDKN1B(T157) (inactive)/CDKN1B (shTMCO1#3: 0.92; shTMCO1#5:0.92) were similar to the control (shLuc; Supplementary Fig. S2C). Nuclear TP53 and CDKN1A as well as nuclear and cytosolic CDKN1B were downregulated (Fig. 2S), yet TP53 mRNA levels were upregulated (P < 0.05; Supplementary Fig. S2D). Further, treatment with a proteasome inhibitor, MG132, increased the abundance of the TP53 protein in the shLuc group compared with the DMSO control. MG132 further restored shTMCO1-suppressed TP53 protein levels in RT4 cells (Fig. 2T), reinforcing the observation that TMCO1 stabilizes TP53 at the protein level. Stable knockdown of the TMCO1 gene in J82 cells showed similar results to TMCO1-knockdown RT4 cells, except the levels of phospho/inactive CDKN1A (shTMCO1#3: 1.89; shTMCO1#5: 2.42) and CDKN1B (shTMCO1#3: 1.76; shTMCO1#5: 1.67) were much higher than the shLuc control, implying that TMCO1 predominantly inhibits phospho/inactive CDKN1A and CDKN1B in J82 cells (Supplementary Fig. S3A–S3G). Moreover, 5 cyclin-dependent kinase inhibitors, CDKN1C, CDKN2A, CDKN2B, CDKN2C, and CDKN2D, were not consistently upregulated or downregulated after overexpression and knockdown of the TMCO1 gene in BFTC905 and RT4 cells, respectively (Supplementary Fig. S4).

To evaluate whether TMCO1-suppressed cell proliferation was AKT signaling–dependent, a constitutively active AKT1 plasmid, pHRIG-AKT1 (23), was cotransfected with the pTMCO1-HaloTag plasmid into BFTC905 cells. Exogenous constitutive expression of the active AKT1 gene markedly downregulated TMCO1-induced levels of RB1, TP53, pTP53(S15), CDKN1A, and CDKN1B, whereas it upregulated the TMCO1-suppressed protein levels of pCDKN1A(T145), pCDKN1B(T157), CCND1, CDK4, CCNE1, CDK2, exogenous AKT1 and pAKT1(S473), MDM2, and pMDM2(S166) (Fig. 2U) as well as phospho/inactive CDKN1A and CDKN1B (Supplementary Fig. S2A). Constitutively active AKT1 overexpression reinstated TMCO1-inhibited cell proliferation (P < 0.001; Fig. 2V). Therefore, TMCO1 suppresses cell proliferation by downregulating pAKT1(S473).

Stable overexpression of the TMCO1 gene in BFTC905 cells suppressed cell migration and invasion (P < 0.001; Fig. 3A and B), with marked downregulation of CD44 and VIM protein levels (Fig. 3C). On the other hand, stable knockdown of the TMCO1 gene in J82 cells enhanced cell migration and invasion (P < 0.001; Fig. 3D and E), with notable upregulation of CD44 and VIM protein levels (Fig. 3F). Stable knockdown of the TMCO1 gene in RT4 cells exhibited similar phenotypes to J82 cells (Supplementary Fig. S5). Accordingly, TMCO1 downregulation induces cell migration and invasion in vitro.

Recently, a family of protein phosphatases (PHLPPs: PHLPP1 and PHLPP2) were discovered and found to directly dephosphorylate and inactivate AKT, thus introducing a new negative regulator of the PI3K oncogenic pathway (22). Individually, PHLPPs dephosphorylate the S473 residue (22), and protein phosphatase 2 catalytic subunit alpha (PPP2CA) dephosphorylates the T308 residue (27) of AKT1. Because the TMCO1 protein contributes to the deactivation of AKT1, we first hypothesized that TMCO1 might regulate PHLPP1 or PHLPP2 protein expression. However, PHLPP1 or PHLPP2 protein level was not altered after TMCO1 overexpression in BFTC905, RT4, or J82 cells (Fig. 4A). A dose-course experiment was conducted in BFTC905 cells, and 15 μmol/L was found to be effective for a pan-PHLPP inhibitor, NSC117079 (Supplementary Fig. S6). In TMCO1-overexpressed BFTC905 cells, treatments with NSC117079 upregulated pAKT1(S473) compared with DMSO/pHaloTag (control) and DMSO/pTMCO1-HaloTag transfectants. Moreover, compared with NSC117079/pHaloTag, pAKT1(S473) was downregulated in NSC117079/pTMCO1-HaloTag transfectants (Fig. 4B). Hence, PHLPPs are indeed involved in TMCO1-mediated dephosphorylation of pAKT1(S473). Confocal immunocytochemistry insinuated that TMCO1 were predominately colocalized with PHLPP2 protein (Fig. 4C), whereas only somewhat colocalized with PHLPP1 (Supplementary Fig. S7) in BTFC905 and RT4 cells. Co-IPs additionally demonstrated that PHLPP2, rather than PHLPP1, interacted with TMCO1 in RT4 and J82 cells (Fig. 4D). These results suggested that TMCO1 recruits PHLPP2 to dephosphorylate pAKT1(S473) in UBUC-derived cells. Exogenous expression of both TMCO1 and PHLPP2 remarkably downregulated pAKT1(S473) protein level (Fig. 4E) and AKT activity (Fig. 4F) in BFTC905 cells compared with overexpression of either TMCO1 or PHLPP2 alone, indicating that TMCO1 recruits PHLPP2 to deactivate pAKT1(S473).

The TMHMM Server (28) predicted that residues #32 to #89 of the TMCO1 protein might reside in the cytoplasmic region (Fig. 5A). We therefore constructed 3 plasmids containing TMCO1 with T33A, S60A, and S84A mutations to evaluate whether these potential phosphorylation sites are critical for TMCO1 function. As shown in Fig. 5B, overexpression of the TMCO1(WT) or each mutant in BFTC905 cells notably upregulated TMCO1(WT)-, TMCO1(T33A)-, TMCO1(S60A)-, TMCO1(S84A)-HaloTag, pAKT1(S473), MDM2, and pMDM2(S166), and downregulated TP53, pTP53(S15), and CDKN1A protein levels. Overexpression of TMCO1(S60A), but not other mutants, released TMCO1(WT)-induced G₁ cell-cycle arrest (Fig. 5C), and increased AKT1 activity (Fig. 5D) compared with TMCO1(WT) overexpression. Neither WT nor any other mutant affected the TP53 mRNA level (NS; (Supplementary Fig. S2E). Accordingly, S60 in the TMCO1 protein was found to be critical for the expression and stability of pAKT1(S473), MDM2, pMDM2(S166), TP53, pTP53(S15), and CDKN1A proteins, and AKT activity.

Mouse xenograft models were further applied to evaluate the effect of TMCO1 in vivo. In NOD/SCID mice, xenografts of BFTC905 cells with TMCO1(WT) (P = 0.007) and TMCO1(S84A) (P = 0.039) overexpression showed smaller tumors compared with the control (HaloTag) after mice were sacrificed. When compared with the TMCO1(WT) group, the TMCO1(T33A) (P = 0.036) and TMCO1(S60A) (P = 0.006) groups exhibited larger tumors. The TMCO1(S60A) group possessed greater tumor mass than TMCO1(S84A) (P = 0.035; Fig. 5E; Supplementary Fig. S8). Stable TMCO1(WT) overexpression induced a large area of necrosis and a much lower percentage of cancer components. Ki-67 was highly expressed in pTMCO1(T33A)- and pTMCO1(S60A)-HaloTag–expressing xenografts compared with pHaloTag xenografts (Fig. 5F), suggesting that TMCO1(S60A) and TMCO1(T33A) mutants particularly impair the growth-inhibitory property of the TMCO1 protein in vivo. In HTB-33 epithelial cells (cervix, derived from metastatic site, omentum), TMCO1 functions as a tumor suppressor in vitro and in vivo (Supplementary Fig. S9), thus reinforcing our findings.

In this study, we found that a high TMCO1 protein level could be an independent prognostic factor for disease-specific and metastasis-free survivals in a subset of UBUC patients. In addition, low TMCO1 protein levels correlated with aggressive tumor behaviors. We further identified that some patients (most were non–muscle-invasive UBUCs, data not shown) with low TMCO1 protein levels showed long-term survival. Recent genomic and transcriptomic experiments indicated that non–muscle-invasive UBUCs and carcinoma in situ/high-grade muscle-invasive UBUCs have distinct mutation and gene expression profile (29). Low invasive properties might weaken the tumor-suppressor role of TMCO1 in these patients. In vitro and xenograft mice models supported TMCO1 as a tumor-suppressor gene in vivo. Few studies have focused on understanding the biological functions of TMCO1. Using genome-wide mappings followed by candidate gene sequencing, homozygosity for a 2-bp deletion in the TMCO1 gene in patients with a syndrome characterized by craniofacial dysmorphism, skeletal anomalies, and mental retardation was identified (30). Another genome-wide association investigation in several cohorts detected a single-nucleotide polymorphism (rs4656461), locating approximately 6.5 Kb downstream of the TMCO1 gene, which was associated with advanced primary open-angle glaucoma (POAG) and less-severe POAG (31). TMCO1 was recently reported to provide a protective mechanism to prevent overfilling of ER stores with Ca²⁺ ions (12). Therefore, this is the first study to describe that loss of TMCO1 expression contributes to tumorigenesis. Immunohistochemical analysis indicates that the TMCO1 protein was highly expressed in the cytoplasm and cell membrane of nontumor urothelium and noninvasive urothelial carcinomas. Low TMCO1 protein levels can be traced back to low TMCO1 mRNA; however, predesigned assays for quantification of CpG methylation spanning 8 CpG islands by pyrosequencing did not detect methylated sites in 4 distinct UBUC-derived cell lines and UBUCs with low TMCO1 protein levels. Thus, the possibility that methylation in the promoter region caused low TMCO1 transcription was excluded. In addition, no evidence of the TMCO1 gene deletion was found from our previous aCGH data (25). Accordingly, the regulation of TMCO1 mRNA and subsequent protein levels might be attributed to the activities of its transcriptional factors or other epigenetic modifications except methylation.

We further identified that TMCO1 inhibits cell-cycle progression accompanied with alterations of pAKT1(S473), MDM2, pMDM2(S166), TP53, pTP53(S15) and nuclear TP53, and CDKN1A protein levels in distinct UBUC-derived cell lines. Based on the Cancer Genome Atlas project, a few pathways were consistently dysregulated in UBUCs, including TP53 and RB1 tumor suppressors, receptor tyrosine kinase/related RAS viral (r-ras) oncogene homolog 2 (RRAS2), and the PI3K/AKT/mechanistic target of rapamycin pathways that affect cell proliferation and survival (32, 33). Along with our findings in this study, constitutive overexpression of active AKT1 reset the expression levels of TMCO1-altered cell-cycle regulators and cell proliferation. Thus, we hypothesized that TMCO1 might deactivate AKT and its downstream signaling pathways. TP53 mRNA levels were inconsistently altered in TMCO1-overexpressed and -knockdown cells, reinforcing the finding that TMCO1 stabilizes TP53 at the protein level. There are three isoforms of AKTs. These isoforms are encoded by different genes but share a conserved domain structure consisting of an N-terminal pleckstrin homology domain, a kinase domain, and a C-terminal regulatory domain containing a hydrophobic motif. AKT1 is ubiquitously expressed, AKT2 is primarily expressed in insulin-responsive tissues, and AKT3 is highly expressed in brain and testes (34). Former studies in various epithelial cell lines revealed that PHLPP1 binds and dephosphorylates AKT2 and AKT3, but not AKT1, whereas PHLPP2 binds and dephosphorylates AKT1 and AKT3, but not AKT2 (22). Unfortunately, exogenous expression and knockdown of the TMCO1 gene in 3 distinct UBUC-derived cell lines were not able to change the expression levels of PHLPP1 or PHLPP2 protein in our study. Because TMCO1 is a transmembrane protein, its upregulation or downregulation might alter the interactions with other proteins, including membranous and nonmembranous proteins, in UBUC-derived cells. Meanwhile, we found that a pan PHLPP inhibitor, NSC117079, notably augments endogenous as well as TMCO1-suppressed pAKT1(S473) levels in BFTC905 cells, implicating PHLPPs in TMCO1-mediated pAKT1(S473) deactivation. Our immunocytochemistry and co-IP data additionally signified that TMCO1 and PHLPP2 proteins colocalize and interact in vivo. Further, simultaneous overexpression of the TMCO1 and PHLPP2 genes synergistically decreased pAKT1(S473) level and AKT1 activity compared with overexpression of either the TMCO1 or PHLPP2 gene alone. Together, these data support the concept that TMCO1 recruits PHLPP2 to dephosphorylate pAKT1(S473). In a well-characterized mechanism, AKT downregulates TP53 protein level by enhancing MDM2-mediated targeting of TP53 degradation (35). Under nonstress conditions, MDM2, an E3 ligase of TP53, binds to TP53 and monoubiquitinates it prior to exporting TP53 to the cytoplasm, where it is polyubiquitinated. AKT-dependent phosphorylation of MDM2 on S166 facilitates this export (36). Therefore, TMCO1-suppressed pMDM2(S166) radically facilitated the stability of the TP53 protein. Moreover, phosphorylation on S15 of the TP53 protein, i.e., pTP53(S15), is necessary to mediate TP53-dependent transcription (especially CDKN1A) and growth arrest (37). In parallel to this scenario, a novel tumor suppressor, TMCO1, is uncovered and found to recruit PHLPP2 to dephosphorylate pAKT1(S473), downregulate MDM2 and phospho/active pMDM2(S166), upregulate TP53 and phospho/active/nuclear pTP53(S15), and inhibit cell proliferation.

In addition to TP53, we also found that TMCO1 upregulates nuclear CDKN1A, and downregulates phospho/inactive CDKN1A and CDKN1B in BFTC905 and J82, but not RT4 cells. Indeed, it has long been known that the AKT oncogenic kinase functionally phosphorylates [pCDKN1A(T145); pCDKN1B(T157), (S187), and (T198)] and inactivates the nuclear CDKN1A and CDKN1B by causing cytoplasmic mislocalization (38). In addition, AKT-dependent phosphorylation of CDKN1A on T145 prevents the formation of a complex between CDKN1A and proliferating cell nuclear antigen, decreases binding of CDKN1A to CDK2, and promotes the assembly of CCND1/CDK4 complex (39, 40). Similarly, upon AKT activation, the appearance of pCDKN1B(T157/T198) precedes CDKN1B–CCND1–CDK4 assembly in early G₁ (41), thereby promoting cell-cycle progression. In the cell cycle, the key regulators of the G₁–S transition are cyclin D-CDK4/6 and cyclin E-CDK2. The activities of these complexes are regulated by the TP53 checkpoint, the RB1 tumor suppressor, the INK4 family of proteins (CDKN2A/p16, CDKN2B/p15, CDKN2C/p18, CDKN2D/p19) and the Cip1/Kip1 family (CDKN1A and CDKN1B; 42). Our overexpression and knockdown experiments suggest that the TMCO1 protein specifically upregulated and downregulated Cip1/Kip1, but not INK4 proteins. TMCO1-suppressed cell proliferation might be partially CDKN1A- and/or CDKN1B-dependent, because TMCO1 positively regulated nuclear/active CDKN1A and CDKN1B protein levels in vitro. Knockdown of the TMCO1 gene was not able to upregulate phospho/inactive CDKN1A and CDKN1B, implying that AKT1 might not be the only kinase to phosphorylate CDKN1A and CDKN1B. Although CDKN1A can be induced by both TP53-dependent and -independent mechanisms, our findings are compatible with either pathway.

Among three mutants disrupting potential phosphorylation sites in the cytoplasmic region, we revealed that overexpression of the TMCO1(S60A), but not TMCO1(T33A) or TMCO1(S84A), protein reverses TMCO1-mediated AKT activity to 85%, pAKT1(S473), pMDM2(S166), pTP53(S15), CDKN1A protein levels, and cell-cycle arrest, suggesting that S60 is a critical residue in maintaining functional TMCO1-AKT-TP53 regulation in vitro. To our surprise, xenografts showed that both TMCO1(T33A) and TMCO1(S60A) expressed high Ki-67 label and exhibited larger tumors compared with TMCO1(WT), indicating that TMCO1(T33A) might induce tumor growth via an AKT/MDM2/TP53/CDKN1A-independent pathway. Similar to TMCO1(WT), overexpression of the TMCO1(T33A), TMCO1(S60A), and TMCO1(S84A) did not change the TP53 mRNA level, strengthening the finding that TMCO1 affects downstream tumor suppressors at the posttranscriptional and/or posttranslational level. Undeniably, transmembrane proteins constitute approximately 20% to 30% of fully sequenced proteome, and they are crucial for a wide variety of cellular functions (28). Protein phosphorylation is the most important, well-studied posttranslation modification in eukaryotes and is involved in the regulation of several cellular processes, such as cell growth and differentiation, signal transduction, and apoptosis (43–45). Phosphorylation usually occurs at S, T, tyrosine (Y), and histidine (H) residues in eukaryotic proteins; approximately 30% to 50% of proteins are presumed to be phosphorylated at some point (46). In transmembrane proteins, phosphorylation sites are located at the cytoplasmic region (47). In nocodazole-induced mitotic arrested HeLa cells, phosphorylation on the intracytoplasmic S60 of the TMCO1 protein was identified (48), supporting our observations.

In addition to cell-cycle arrest, exogenous expression and knockdown of the TMCO1 gene additionally inhibited and enhanced cell migration and invasion in vitro. The CD44 protein has several important physiologic functions in cell–cell and cell–matrix interactions including proliferation, adhesion, migration, hematopoiesis, lymphocyte activation, homing, and extravasation (49). Epithelial cell migration was recently reported to require the interaction between the VIM and keratin intermediated filaments (50). The downregulation and upregulation of CD44 and VIM protein levels, respectively, after overexpression and knockdown of the TMCO1 gene was also reported, thus strengthening the tumor-suppressor roles of TMCO1 in cell migration in vitro.

Overall, we demonstrate low TMCO1 protein levels in a subset of UBUCs with aggressive behaviors. In distinct UBUC-derived cell lines, TMCO1 expression inhibited cell proliferation by modulating the protein levels of pAKT1(S473), MDM2, pMDM2(S166), TP53, pTP53(S15), nuclear/active CDKN1A and CDKN1B, CD44, and VIM in combination with decreasing cell viability, proliferation, colony formation/anchorage-independent cell growth, cell migration, and invasion. In addition, TMCO1 was found to recruit PHLPP2 to dephosphorylate pAKT1(S473) and reduce AKT activity, and the intramembrane S60 residue of the TMCO1 protein was found to play a crucial role in this AKT-dependent pathway. Clinical associations, in vitro indications, and xenografts serve robust evidence that the TMCO1 gene is a novel tumor suppressor in UBUCs. Downregulation of the TMCO1 protein can be an adverse prognostic factor for inferior outcomes in UBUC patients.

No potential conflicts of interest were disclosed.

Conception and design: C.-F. Li, Y.-L. Shiue

Development of methodology: C.-F. Li, Y.-L. Shiue

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.-R. Wu, T.-C. Chan, L.-R. Chen, W.-J. Wu, B.-W. Yeh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-F. Li, T.-C. Chan, L.-R. Chen, B.-W. Yeh, Y.-L. Shiue

Writing, review, and/or revision of the manuscript: C.-F. Li, L.-R. Chen, Y.-L. Shiue

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.-F. Li, Y.-H. Wang, W.-J. Wu, B.-W. Yeh, S.-S. Liang, Y.-L. Shiue

Study supervision: C.-F. Li, Y.-L. Shiue

C.-F. Li was funded by Ministry of Health and Welfare (MOHW104-TDU-M-212-133004) and Ministry of Science and Technology (MOST104-2314-B-037-050-MY3). W.-J. Wu was supported by Kaohsiung Medical University “Aim for the Top Universities” (KMU-TP105G00, KMU-TP105G01, KMU-TP105G02), Center for Infectious Disease and Cancer Research (KMU-TP105E24), Kaohsiung Medical University Research Foundation (KMUOR105), and NSYSU-KMU Joint Research Project (NSYSUKMU 106-P008). It is also supported by Kaohsiung Medical University Hospital (KMUH104-4R44, KMUH105-5R47) and Ministry of Science and Technology (MOST104-2314-B-037-050-MY3). Both C.-F. Li and W.-J. Wu were supported by the health and welfare surcharge on tobacco products, Ministry of Health and Welfare (MOHW106-TDU-B-212-144007). Y.-L. Shiue was funded by (MOST105-2314-B-110 -002 -MY2) and the Taiwan Protein Project (MOST105-0210-01-12-01 and MOST106-0210-01-15-04).

The authors are grateful to the Biobank at Chi Mei Medical Center.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.