Tumor Progression Is Mediated by Thymosin-β4 through a TGFβ/MRTF Signaling Axis

Dynamic cytoskeletal remodeling is critical to various morphogenetic and pathologic events, such as gastrulation, neural tube formation, tissue fibrosis, and tumor progression. TGFβ family proteins are well-known pivotal inducers of these events (1). For example, TGFβ induces the phenotypic transition of various types of cells into mesenchymal, myofibroblast, and smooth muscle–like cells, accompanied by dramatic cytoskeletal remodeling and increased cell motility (2, 3). Myocardin-related transcription factors (MRTFs; MKL1 and MKL2), or robust SRF transcriptional coactivators, have been reported to organize actin cytoskeletal rearrangement and regulate cell motility by controlling the expression of dozens of cytoskeletal/adhesion genes (4–6).

We have previously reported that MRTF activation is required for cytoskeletal rearrangement during the TGFβ1-induced epithelial–mesenchymal transition (EMT; ref. 7). MRTFs also induce cytoskeletal and extracellular matrix gene expression and are therefore critical for TGFβ-mediated myofibroblast activation in fibrosis (8). In short, MRTFs are central players in the TGFβ-induced phenotypic changes accompanied by cytoskeletal remodeling. However, the pathway from TGFβ signaling to MRTF activation is not thoroughly understood. MRTF activity is generally regulated by the binding of N-terminal RPEL motifs to three to five monomeric G-actin molecules, which prevents the importin-mediated nuclear translocation of MRTFs (4, 9, 10). Depletion of the cytosolic free G-actin pool leads to the nuclear accumulation and activation of MRTFs as transcriptional regulators.

Thymosin-β4 (Tβ4; TMSB4X) is a very abundant small (43-amino-acid) protein that regulates actin dynamics by binding monomeric G-actin via its WH2 domain with 1:1 stoichiometry, thus sequestering G-actin from spontaneous polymerization in the cytosol (11). In addition, Tβ4 has garnered considerable attention as a tumor promoter (12). Augmented Tβ4 expression is frequently observed during tumor progression and is associated with poor prognosis in various cancers. Although additional functional targets of Tβ4 have been reported (13, 14), the molecular mechanisms underlying the role of this protein in tumor progression remain largely unclear. Previously, we have identified Tβ4 as a regulator of MRTF activity: Tβ4-bound G-actin cannot interact with MRTFs (15). Specifically, the Tβ4 WH2 domain and MRTF RPEL motifs competitively interact with the same hydrophobic clef in the actin molecule (16, 17). Therefore, an increased Tβ4 level causes actin/MRTF complex dissociation and consequently, nuclear MRTF accumulation and activation. Here, we uncover the molecular mechanisms underlying the TGFβ/Tβ4/MRTF signaling pathway in EMT, metastasis, and tumor progression by demonstrating that TGFβ increases Tβ4 expression by activating Smad and T-cell factor/lymphoid enhancer factor (TCF7/LEF1) signaling. Subsequently, the elevated Tβ4 level induces changes in MRTF-regulated gene expression profiles.

All animal procedures were approved by the Animal Care Committee of Osaka University (Osaka, Japan), and all animal care and protocols were conducted according to the guidelines for animal experiments of the Osaka University School of Medicine.

For metastatic analyses, suspensions of wild-type B16F1, Tβ4 KO, and KO/MRTF cells (2 × 10⁴ cells/0.5 mL of HBSS) were intravenously injected into 8- to 10-week-old female C57BL/6j mice. Three weeks after the injection, the extracted lungs were fixed and bleached in Fekete's solution (60% ethanol, 3% formaldehyde, and 4% glacial acetic acid) to easily visualize the B16F1 tumor nodules. Clearly visible metastatic nodules on the lung surface were counted.

In addition, B16F1 and Tβ4 KO cells (4 × 10³ cells/0.1 mL of HBSS) were subcutaneously inoculated into mice, and resulting tumors were extracted 3 weeks after the inoculation. To inhibit TGFβ pathway activation, EW-7197 (AdooQ Bioscience; 5 mg/kg) was orally administered twice on days 2 and 1 before tumor extraction.

To investigate the in vivo effects of TGFβ1 on Tβ4 expression, TGFβ1 (PeproTech; 100 μg/kg) was intraperitoneally administered to 3-day-old female mice. Liver tissues were extracted 3 days after the administration.

NMuMG, B16F0, and B16F1 cells were obtained from ATCC in 2006, ATCC in 2012, and RIKEN BRC in 2012, respectively. Cell authentication and mycoplasma testing were not performed by ourselves. All cells were cultured in DMEM (WAKO) supplemented with 10% FBS (Bovogen Biologicals) and maintained for no more than 10 passages. Cells were treated with 10 ng/mL of TGFβ1, 50 ng/mL of Wnt3a (PeproTech), 10 μmol/L RexSox (Cayman Chemical), 10 μmol/L SIS3 (Cayman Chemical), or 10 μmol/L FH535 (Focus Biomolecules). To observe the nuclear translocation of MRTF-A, 5 ng/mL of leptomycin B (Calbiochem) was added into the culture medium along with TGFβ1. Vectors were transfected into cells using Lipofectamine 3000 (Thermo Fisher Scientific) or ViaFect Transfection Reagent (Promega) according to the manufacturer's instructions. In the knockdown experiments, cells were transfected with predesigned siRNAs via Lipofectamine RNAiMAX (Thermo Fisher Scientific). The target sequences of the siRNAs used in this study are available in Supplementary Table S1. MISSION siRNA Universal Negative Control (Sigma-Aldrich) was used as a negative control. To establish stable Tβ4-overexpressing cell lines, NMuMG cells were cotransfected with a mammalian expression vector (pCAGGS) encoding the mouse Tmsb4x gene and Linear Hygromycin Markers (Clontech Laboratories). To establish Tβ4 KO cell lines using the CRISPR-Cas9 system, B16F1 cells were transfected with all-in-one Cas9/gRNA plasmid pSpCas9 BB-2A-GFP (PX458; Addgene; gRNA target sequence, ccatgtctgacaaacccgatatg). Gene knockout was validated by genome sequencing and Western blotting. KO/MRTF cell lines were isolated from Tβ4 KO2 cells transfected with pcDNA3.1-CA-MRTF-A (7).

NMuMG, B16F0, and B16F1 cells were cultured on coverslips, fixed with 4% paraformaldehyde in PBS for 15 minutes and incubated in blocking solution (0.1% Triton X-100, 0.2% BSA, and 10% normal goat serum in PBS) for 30 minutes. The cells were incubated in a primary antibody solution [1:50 to 1:100 dilutions in Can Get Signal immunostaining reagent (TOYOBO)] for 2 hours and in a secondary antibody solution [1:400 dilutions of Alexa 488- or 568-conjugated secondary antibodies (Thermo Fisher Scientific) in blocking solution] for 2 hours. Hoechst 33342 (Thermo Fisher Scientific) was added to the secondary antibody solution for nuclear visualization. The mounted cells were observed using an all-in-one fluorescence microscope (BZ-9000; Keyence). Anti-Tβ4 (EMD Millipore), anti-MKL1 (Santa Cruz Biotechnology), anti-GFP (Thermo Fisher Scientific) and anti-myc (Santa Cruz Biotechnology) were commercially purchased.

Cells were lysed in SDS sample buffer [2% SDS, 125 mmol/L diothiothreitol (DTT), 0.005% bromophenol blue, 10% glycerol and protease inhibitor cocktail (Nacalai Tesque), in 62.5 mmol/L Triton-HCl (pH 6.8)] and heated at 98°C for 5 minutes. Proteins were separated electrophoretically on 10% to 15% polyacrylamide gels and transferred to polyvinylidene difluoride (PVDF) membranes (EMD Millipore). The membranes were incubated in blocking solution [1% skim milk in TBS-T solution (0.1% Tween-20, and 137 mmol/L NaCl in 20 mmol/L Tris-HCl, pH7.5)] for 30 minutes, in a primary antibody solution [1:2,000–5,000 dilutions in Can Get Signal immunoreaction enhancer solution (TOYOBO)] for 2 hours, and in a secondary antibody solution [1:5,000 dilutions of horseradish peroxidase–linked secondary antibodies (GE Healthcare) in TBS-T] for 2 hours. To detect Tβ4 protein, the PVDF membranes were fixed with 1% paraformaldehyde for 30 minutes before blocking. Anti–β-actin (Sigma), anti-Tβ4 (EMD Millipore), anti-FLAG (Sigma), and anti-HA (Roche Life Science) antibodies were commercially purchased.

Total RNAs were extracted using RNAiso Plus (TAKARA BIO) and then reverse transcribed using a PrimeScript RT Reagent Kit with gDNA Eraser (TAKARA BIO). Real-time PCR was performed using the THUNDERBIRD SYBR qPCR Mix (TOYOBO) and a LightCycler Nano (Roche Life Science). Data were normalized to GAPDH, RPL13A, and 18S rRNA expression. The primer sequences used in this study are presented in Supplementary Table S1.

A DNA fragment from the promoter region of mouse Tmsb4x was isolated by PCR and inserted into the pGL3-Basic vector (Promega). Twelve copies of a synthetic TRE sequence (agATCAAAGggggta) were tandemly inserted into pGL3-γ-actin-TATA (7). The CArG-promoter and Acta2-promoter constructs have been described previously (18). These promoter-reporter constructs were introduced into NMuMG or B16F1 cells using pSV-βGal (Promega), and the resulting luciferase and β-galactosidase activities were measured using a Luciferase Assay System (Promega) and Galacto-Star (Roche Life Science), respectively. Luciferase activity levels were normalized to β-galactosidase activity levels.

NMuMG cells were treated with 10 ng/mL of TGFβ1 for 24 hours, followed by 1 μmol/L latrunculin A (Cayman Chemical) for 30 minutes. The cells were subsequently incubated with 2.5 mmol/L dithiobis (succinimidyl propionate; DSP; Thermo Fisher Scientific) in PBS for 1 hour, after which crosslinking was terminated by adding 20 mmol/L Tris-HCl (pH8.0). The resulting cells were lysed in DTT-free SDS-PAGE buffer, and the cross-linked proteins were separated and visualized by Western blotting.

Biotin-labeled DNA probes containing wild-type or mutated ST1 sites were synthesized by PCR (WT; gggagacaagcgagggAGACAAAGagggccgggt, mut SBE; gggagacaagcgagggAtACAAAGagggccgggt, mut TCF; gggagacaagcgagggAGACgggGagggccgggt) and subsequently bound to streptavidin on Dynabeads M-280 (Thermo Fisher Scientific). The recombinant proteins HA-Smad3, HA-Smad4, and FLAG-LEF1 were synthesized in HEK293T cells and incubated with the DNA-bound Dynabeads in 0.5% NP-40, 10% glycerol, and protease inhibitor cocktail in PBS for 3 hour. The resulting DNA–protein complexes were magnetically pulled down and using SDS sample buffer. The precipitated proteins were then separated and visualized by Western blotting.

Total RNAs were extracted from Tβ4 KO and parental B16F1 cells using RNAiso Plus and purified using NucleoSpin RNA Plus (Macherey-Nagel). Next, 100 ng of total RNA was amplified and labeled using a Low Input Quick Amp Labelling Kit, One-Color (Agilent Technologies). Cy3-labeled cRNA were hybridized on a SurePrint G3 Mouse Gene Expression 8 × 60K Microarray (Agilent Technologies). The hybridized images were scanned using a SureScan Microarray Scanner (Agilent Technologies) and analyzed using Feature Extraction Software (v12.0.3.1; Agilent Technologies). A Gene Ontology (biological process) analysis was performed using DAVID functional annotation tools.

The Cancer Genome Atlas (TCGA) mRNA expression and clinical datasets were downloaded from the cBioPortal for Cancer Genomics and included cutaneous melanoma (TCGA, provisional), breast invasive carcinoma (TCGA, provisional), renal clear cell carcinoma (TCGA, Nature 2013), hepatocellular carcinoma (TCGA, provisional), papillary thyroid carcinoma (TCGA, Cell 2014), lower grade glioma (TCGA, provisional), and glioblastoma (TCGA, Cell 2013). These datasets did not distinguish the expression of TMSB4X and TMSB4XP8, which have highly similar sequences; only six single-nucleotide mismatch sites and one four-nucleotide deletion site were detected in the 633-b TMSB4X mRNA. However, TMSB4XP8 is a pseudogene expressed at extremely low levels relative to TMSB4X, at least in cultured human cell lines.

Data from the in vitro experiments were subjected to a statistical analysis using Student two-tailed t test or the two-tailed Mann–Whitney U test. Correlation coefficients were calculated using Pearson correlation test or Spearman correlation test. All experiments were repeated independently at least three times. Data are presented as means ± SEM. Statistical significance is indicated by P values (*, P < 0.05 and **, P < 0.01). In the regression analysis, outliers were excluded using the Smirnov–Grubbs test, and differences between two groups were statistically evaluated using an analysis of covariance (ANCOVA). The Kaplan–Meier survival analysis was performed using JMP Pro 12 software, and survival data were statistically analyzed using the Wilcoxon rank-sum test and log-rank test.

All microarray data for this study have been deposited in the Gene Expression Omnibus (GEO) database under the accession number GSE99492. Previously published RNA-seq and clinical datasets that were reanalyzed in this study are available via the TCGA Data Portal (https://tcga-data.nci.nih.gov).

We have previously reported on the pivotal functions of MRTFs in TGFβ-induced EMT (7), an important event in the acquisition of invasive and metastatic abilities (19). In murine mammary gland NMuMG cells, TGFβ1 increased the activity of a synthetic reporter construct containing CArG boxes, or MRTF/SRF complex-binding cis-acting elements (Fig. 1A). Furthermore, TGFβ1 increased the promoter activity of Acta2, a major MRTF target in EMT, in a CArG box–dependent manner (Fig. 1B). The latter activity was completely suppressed by the MRTF inhibitor CCG-1423 (Fig. 1B). We also identified Tβ4 as a downstream target of TGFβ1 during EMT (Fig. 1C), and its expression exhibited a similar increase over time as that of MRTF/SRF activation after TGFβ1 stimulation (Fig. 1A and D). In contrast, the expression of thymosin-β10 (Tmsb10), a thymosin-β family protein with high similarity to Tβ4, was significantly but only slightly increased after TGFβ1 stimulation (Fig. 1D).

We have previously reported that increased Tβ4 level expression enhances MRTF activity through Tβ4/G-actin complex formation (15). Tβ4, a small (5 kDa) protein, forms a roughly 45-kDa complex with monomeric actin, as visualized by chemical crosslinking with DSP (Fig. 1E). TGFβ1 stimulation increased complex formation, whereas latrunculin A (LatA) prevented complex formation by binding to and occupying the Tβ4-binding sites of actin molecules (20). Tβ4 knockdown suppressed the nuclear accumulation of MRTF-A and activation of MRTFs/SRF after TGFβ stimulation (Fig. 1F and G). Furthermore, ectopic Tβ4 overexpression robustly enhanced MRTF-A nuclear accumulation and MRTFs/SRF activity even in the absence of TGFβ1 (Fig. 1H and I). In summary, elevated Tβ4 competitively prevents MRTF/G-actin complex formation, leading to nuclear MRTF accumulation and subsequent MRTFs/SRF signaling activation during TGFβ-induced EMT.

Previously, we have shown that MRTFs regulate the expression of various EMT-related genes during TGFβ1-induced EMT in canine normal kidney epithelial MDCK cell lines (7). We confirmed that MRTFs also controlled EMT-related gene expression in NMuMG cells (Supplementary Fig. S1A), whereas Tβ4 knockdown significantly abolished these transcriptional changes (Fig. 1J). Using stably Tβ4-overexpressing NMuMG cell lines, we demonstrated a morphologic change from an epithelial to a mesenchymal cell shape and altered EMT-related gene expression levels after EMT, even in the absence of TGFβ1 (Supplementary Fig. S1B–S1D). Thus, Tβ4 upregulation is critical for TGFβ-induced EMT.

As observed in NMuMG cells, TGFβ1 stimulation upregulated Tβ4 expression in various mouse and human cells (Fig. 2A) and increased Tβ4 expression, accompanied by Acta2 upregulation, in vivo (Fig. 2B and C). These data suggest that TGFβ signaling universally controls Tβ4 expression. Knockdown of Smad4, a canonical downstream effector of the TGFβ cascade, suppressed TGFβ1-induced Tβ4 expression (Fig. 2D). Tβ4 induction was also severely suppressed by treatment with the TGFβ type I receptor inhibitor RepSox and the Smad3-specific inhibitor SIS3 (Fig. 2E), indicating involvement of the canonical TGFβ/Smad pathway in Tβ4 induction.

For a detailed investigation of the mechanism underlying Tβ4 transcription regulation, a DNA fragment upstream of exon 1 of the mouse Tβ4 gene was subjected to a promoter analysis. TGFβ1 stimulation equivalently activated reporter constructs containing the 3,145-b or 523-b fragment, but not the 254-b fragment (Fig. 2F). From a Tβ4 promoter region sequence comparison across multiple mammals, we identified evolutionally conserved Smad-binding elements (SBE: cAGACa) between −254 and −523, together with other predicted cis-elements (Fig. 2G; Supplementary Fig. S2). The human and mouse genomes harbor two copies of highly conserved DNA sequences (ST1 and ST2: AGACAAAG) containing a partially overlapping SBE and TCF/LEF transcriptional response element (TRE: ACAAAG). Mutation of the ST1 and ST2 sequences in the 523-b construct completely abolished TGFβ responsiveness (Fig. 2H), suggesting the involvement of Smad and/or TCF/LEF. In a DNA–protein binding assay, both Smad3/4 and LEF1 proteins bound to a short DNA fragment containing the ST1 site, but not to fragments containing a mutated SBE or TRE (Fig. 2I). The ectopic expression of constitutively active Smad3/4 and LEF1 synergistically activated the Tβ4 promoter construct, whereas mutations of the ST1 and ST2 sites notably abolished these activations (Fig. 2J), suggesting that both Smad and TCF/LEF directly control Tβ4 expression via the ST1 and ST2 sites.

TCF/LEF has been reported as a noncanonical downstream effector of TGFβ (21–23). A synthetic reporter construct containing 12 tandemly arranged TRE sequences was designed to monitor TCF/LEF activity and was predictably activated by Wnt3a, a typical β-catenin/TCF/LEF pathway activator, and constitutively active LEF1 (Fig. 2K). TGFβ1 actually activated this construct, which was severely suppressed by the β-catenin/TCF inhibitor FH535, indicating the ability of TGFβ1 to activate TCF/LEF signaling. Furthermore, ST1/ST2 mutation abolished the synergistic activating effects of TGFβ1 and Wnt3a on the Tβ4 promoter construct (Fig. 2L). FH535 also inhibited the TGFβ1-induced upregulation of endogenous Tβ4, whereas upregulation of the Serpine1 gene, a direct downstream target of TGFβ/Smad signaling, was inhibited by SIS3 but not FH535 (Fig. 2E). These findings indicate that Smad and TCF/LEF cooperatively regulate Tβ4 transcription downstream of TGFβ signaling via the ST1 and ST2 elements.

Clark and colleagues have previously reported enhanced Tβ4 expression in highly metastatic B16 murine melanoma cell variants (F1, F2, and F3) subcloned by repeated in vivo metastatic selection from poorly metastatic B16F0 cells (24). We observed considerably higher basal expression of Tβ4 in B16F1 cells relative to B16F0 cells (Fig. 3A). Moreover, TGFβ1 more strongly enhanced Tβ4 expression in B16F1 cells relative to B16F0 cells, despite comparable levels of Serpine1 induction (Fig. 3A). In B16F1 cells, Tβ4 protein was strongly expressed and formed a complex with β-actin, especially after TGFβ1 stimulation; in contrast, Tβ4 was rather weakly expressed in B16F0 cells (Fig. 3B). Accordingly, CArG-box promoter analyses showed that TGFβ1 significantly activated MRTF/SRF only in B16F1 cells, a process attenuated by Tβ4 knockdown (Fig. 3C and D); thus, the high Tβ4 level promotes the activation of MRTFs in B16F1 cells. Furthermore, TGFβ1 increased the nuclear translocation of MRTFs in B16F1 cells, which was also attenuated by Tβ4 knockdown (Fig. 3E). In B16F0 cells, ectopic Tβ4 expression remarkably induced nuclear MRTF-A accumulation and MRTF/SRF activation (Fig. 3F and G).

As in NMuMG cells, RepSox, SIS3, and FH535 all inhibited TGFβ1-induced Tβ4 upregulation in B16F1 cells (Fig. 3H), whereas SIS3 but not FH535 inhibited Serpine1 induction (Fig. 3I). Treatment with these inhibitors and Smad4 knockdown also reduced basal Tβ4 expression (Fig. 3J and K), suggesting that autocrine TGFβ signaling contributes to the high basal Tβ4 expression in B16F1 cells. The expression levels of Smad and TCF/LEF pathways signaling components were compared between the two melanoma variants to elucidate the cause for the differences in TGFβ responsiveness (Fig. 3L). Notably, Tgfr1, Tgfr2, Smad3, Tcf7, and Lef1 expression was significantly stronger in B16F1 cells. Moreover, Tcf7 and LEF1 knockdown attenuated TGFβ1-induced Tβ4 expression (Fig. 3M), suggesting that in addition to the canonical Smad pathway, TCF/LEF activation is important for TGFβ-induced Tβ4 expression.

In the tumor microenvironment, TGFβ is supplied to cancer cells in an autocrine/paracrine manner, which exacerbates tumor progression and metastasis (25). When B16F1 cells were subcutaneously inoculated into mice, Tβ4 expression levels increased considerably in the resulting tumors, although this was significantly attenuated by treatment with an orally bioavailable TGFβ receptor inhibitor, EW7197 (Fig. 4A). To investigate the role of Tβ4 in B16F1 metastasis, we isolated Tβ4-knockout B16F1 cell lines (KO1 and KO2) using the CRISPR-Cas9 system (Fig. 4B) and performed a microarray analysis. Downregulated genes in Tβ4-KO cells included several TGFβ1-inducible genes (Fig. 4C and D; Supplementary Table S2); furthermore, the expression of these genes was largely reduced by CCG1423 treatment (Fig. 4C–E). A Gene Ontology (GO) analysis of downregulated genes in KO cells revealed significant enrichment for terms related to circadian rhythm, cell proliferation, cell migration, and cytoskeleton/adhesion (Fig. 4F; Supplementary Table S3), which were also reportedly enriched among the GO terms of MRTF-regulated genes (6). Therefore, Tβ4 almost certainly regulates MRTF activity downstream of TGFβ signaling in B16F1 cells.

Among the Tβ4-regulated genes, we focused on Itgb1, Mmp14, Myh9, Myl9, Tpm3, and Wisp1, because expression of these genes reportedly depends on MRTF activity and is strongly associated with tumor metastasis (Fig. 4G; refs. 6, 26–31). We confirmed the reduced expressions of these genes in Tβ4-KO tumors formed by subcutaneous inoculation into mice (Supplementary Fig. S3). Transient Tβ4 knockdown, as well as SRF knockdown and double MRTF-A/B knockdown, also reduced their expressions in B16F1 cells (Fig. 4H). Meanwhile, transient overexpression of Tβ4 or MRTF-A partially restored the expression of these Tβ4-regulated genes in Tβ4-KO2 cells (Fig. 4I). These findings suggest that Tβ4/MRTFs regulate the expression of tumor-associated genes in B16F1 cells.

In humans, enhanced Tβ4 expression has been correlated with metastasis and poor prognosis (12), although it's functional roles in tumor progression remain largely unknown. We validated the importance of the TGFβ1/Tβ4/MRTFs signaling cascade in tumor progression using human cancer genomic datasets from The Cancer Genome Atlas. We observed significant positive correlations between the expression of TGFB1 and TMSB4X (encodes human Tβ4 protein) in cutaneous melanoma, breast invasive carcinoma, renal clear cell carcinoma, hepatocellular carcinoma, papillary thyroid carcinoma, lower grade glioma, and glioblastoma (Fig. 5A). As a dataset control, we confirmed that the TGFB1 expression levels positively correlated with those of SERPINE1 but not SRF, the expression of which is independent of TGFβ signaling (Supplementary Fig. S4A).

To clarify whether TMSB4X expression occurs downstream of TGFB1 in this context, the expression data were grouped by TGFβ receptor 1 (TGFBR1) levels (Fig. 5B). Low-TGFBR1–expressing melanomas, thyroid carcinomas, and gliomas exhibited significant low TMSB4X to TGFB1 expression ratios, suggesting that TGFB1/TGFBR1 signaling increases TMSB4X expression. SERPINE1 expression levels also depended on TGFBR1 expression (Supplementary Fig. S4B). In B16F1 cells, TGFβ/TCF/LEF signaling affected Tβ4 induction (Fig. 3H and M). To verify the contribution of this pathway in human cancers, the expression data were grouped according to TCF7 or LEF1 expression. Low-TCF7–expressing melanomas and thyroid carcinomas and low-LEF1–expressing gliomas had significant low TMSB4X/TGFB1 expression ratios (Fig. 5C and D), supporting our hypothesis that TGFβ signaling induces Tβ4 expression via TGFBR/SMAD and TCF/LEF signaling in human cancer tissues.

Next, we evaluated the correlation between Tβ4 and Tβ4-regulated tumor-associated genes (ITGB1, MMP14, MYH9, MYL9, TPM3, WISP1) in human cancers. In many cases, TMSB4X expression correlated positively with the tumor-associated gene expression, although the correlation between TMSB4X and MYH9 was relatively weak (Fig. 5A). When the expression data were grouped by MKL1 (MRTF-A) and MKL2 (MRTF-B) expression levels, MKL1/2-low groups exhibited extremely low tumor-associated gene to TMSB4X expression ratios, compared with MKL1/2-high groups (Fig. 6A). Thus, enhanced Tβ4 expression most likely promoted the increased expression of tumor-associated genes in human cancer tissues by activating MRTFs.

Kaplan–Meier survival analyses of renal clear cell carcinoma, glioblastoma, breast invasive carcinoma, and lower grade glioma revealed that strong TMSB4X expression correlated significantly with a poor prognosis (TMSB4X-high groups, red vs. TMSB4X-moderate/low groups, yellow; Fig. 6B–I; Supplementary Fig. S5A–S5H). To validate the involvement of MRTFs/SRF signaling, TMSB4X-high groups were further subdivided by SRF expression level because the MKL1 level directly affected survival rates (Supplementary Fig. S5I). The TMSB4X-high/SRF-high groups (green) exhibited significantly shorter mean survival times (MST) than the TMSB4X-high/SRF-low groups (blue), particularly for renal cell carcinoma and glioblastoma where MSTs of TMSB4X-high/SRF-low groups were statistically comparable with those of TMSB4X-moderate/low groups [73.168 vs. 90.384; P = 0.2157 (Wilcoxon rank sum test) in renal cell carcinoma and 13.3 vs. 14.0; P = 0.7537 in glioblastoma; Fig. 6C and G]. We confirmed that SRF levels did not directly correlate with prognosis (Fig. 6D and H) and that Tβ4 expression levels were not lower in TMSB4X-high/SRF-low groups than in TMSB4X-high/SRF-high groups (Fig. 6E and I). These data suggest that Tβ4 expression affects human cancer prognosis via MRTFs/SRF signaling.

To further investigate the importance of Tβ4/MRTF signaling in tumor progression, we performed experimental metastasis assay using Tβ4 KO cells. In mice, intravenously injected parental B16F1 cells exhibited pulmonary metastasis and formed numerous pulmonary nodules within 3 weeks postinjection, whereas Tβ4 deletion led to significant decreases in metastatic potential and secondary tumor formation (Fig. 7A and B). We further isolated KO/MRTF clones from Tβ4 KO2 cells transfected with a constitutively active MRTF-A expression vector, because the expression levels of MRTFs-target genes were largely reduced in Tβ4 KO cells (Fig. 4C and E). KO/MRTF cells exhibited restored expressions of the Tβ4/MRTF–dependent tumor-associated genes and enhanced metastatic activity in mice (Figs. 7A–C), supporting the idea that MRTF activation is critical for the Tβ4-contributed tumor progression and metastasis.

Mounting evidence suggests that MRTFs promote tumor progression by regulating cytoskeletal component gene expression (26, 31–33). Medjkane and colleagues clearly demonstrated that MRTFs regulate the motility, invasiveness, and lung metastatic capacities of human breast carcinoma MDA-MB-231 and B16 melanoma cells (31). Furthermore, MRTF knockdown severely suppressed the metastatic potential of B16F2 cells, whereas constitutively active MRTF-A expression enhanced the lung metastatic capacity of B16F0 cells. Medjkane and colleagues identified MYH9 and MYL9, the expression of which is required for tumor invasion and metastasis, as critical MRTF targets. Similarly, we observed that MYL9 expression was suppressed in Tβ4 KO cells but robustly induced by ectopic Tβ4 and MRTF-A expression (Fig. 4) and identified significant positive correlations between TMSB4X and MYL9 expression levels in all examined human cancers (Fig. 5A). Medjkane and colleagues also identified additional eight common MRTF target genes between the two studied cell lines (31), of which six (PDE1C, VGLL3, TPM1, HELLS, DENND2A, and PHLDB2) were confirmed to be downregulated in our Tβ4 KO cells (Supplementary Table S2). Many downregulated genes in Tβ4 KO cells overlapped with those in CCG1423-treated cells (256/681 genes; 37.6%, Fig. 4D), and their expression levels were significantly positively correlated (Pearson r = 0.6238, P < 0.001; Fig. 4E), indicating a similar mode of regulation (e.g., MRTFs/SRF).

We demonstrated that increased Tβ4 expression is critical for TGFβ-induced MRTF activation. MRTF activity is generally regulated by Rho signaling, which in some cells is directly stimulated by TGFβ stimulation within 30 minutes (34, 35). In NMuMG and MDCK (8) cells, however, 16 to 24 hours were required to fully activate MRTFs after TGFβ1 stimulation (Fig. 1A), thus supporting our findings regarding the involvement of transcriptional regulation. Previously, several Rho signaling factors, including RhoB, RhoC, GEF-H1, and Net-1, have been reported to be transcriptionally upregulated by TGFβ stimulation (36–39). Our microarray analysis confirmed the weak upregulation of these factors along with Tβ4 (Tmsb4x: 4.52-fold) in B16F1 cells after TGFβ1treatment, excluding Net1 [Rhob: 1.88-fold, Rhoc: 1.60-fold, Arhgef2 (GEF-H1): 1.65-fold, Net1: 0.99-fold]. Furthermore, some Rho signaling factors are transcriptionally regulated by MRTFs, which likely forms a positive feedback loop (31). Although our knockdown/knockout experiments demonstrated a pivotal role for Tβ4 in the TGFβ/MRTF pathway, other factors might contribute to this complex system.

The TGFβ/Smad and Wnt/TCF/LEF pathways commonly cross-talk during developmental processes and cancer progression. Many human tumors overexpress TGFβ and Wnt, which are important effectors of invasive properties, mainly via EMT induction (40). During morphogenesis, these signals cooperatively regulate gene expression; for example, the promoter regions of the homeobox genes Xtwn and Msx2 contain both SBE and TRE, and TGFβ and Wnt synergistically induce the expression of both genes (21, 41). Here, we demonstrated that Tβ4 expression is cooperatively regulated by Smad and TCF/LEF signals in NMuMG and B16F1 cells and some human cancer tissues (Figs. 2, 3, and 5). However, the expression of Serpine1, a typical transcriptional target of TGFβ/Smad, was controlled in a Smad-dependent but TCF/LEF–independent manner (Figs. 2E and 3I), indicating that the latter is not necessarily required for TGFβ/SMAD pathway activity. In the Tβ4 gene promoter region, we identified a novel TGFβ-responsive cis-acting element, AGACAAAG, which is conserved across several mammals and contains partially overlapping SBE and TRE sequences and interacts with both Smad3/4 and LEF1. Around ST1/2, we also identified conserved predicted binding elements for the SP-1, AP-1, and ETS transcription factors, which are frequently identified in functional Smad2/3 and Smad4-binding regions (42, 43). Although TCF/LEF is not a canonical downstream target of TGFβ, we observed TGFβ1-induced TCF/LEF activation in NMuMG cells (Fig. 2K). Previous studies of other cell types have also reported direct/indirect activation of the β-catenin/TCF pathway by TGFβ (21–23), and Smad3 and Smad4 reportedly interact with TCF/LEF (21, 44). These findings demonstrate that TGFβ signaling activates Smad and TCF/LEF and promotes their recruitment to ST1/2 elements individually or as a complex.

The multifunctional cytokine TGFβ is a potent inducer of EMT and fibrosis and both a promoter and suppressor of tumor progression (25). Within fibrotic tissues and tumors, myofibroblasts, tumor cells and immune cells secrete TGFβ, which acts in an autocrine/paracrine manner to provoke EMT, angiogenesis, phenotypic fibroblast and macrophage transition, and tumor immunosurveillance escape. Recent research has focused on tumor-associated macrophages (TAM) and cancer-associated fibroblasts (CAF), which secrete TGFβ within the tumor microenvironment and promote metastatic potential (45, 46). Our results suggest that TGFβ might direct cancer cells to increase Tβ4 expression within the tumor microenvironment (Figs. 4A and 5A), raising a possibility that TAMs and CAFs facilitate Tβ4 expression and MRTF activation to promote metastasis in the tumor microenvironment.

Taken together, our findings point to a principal role for Tβ4 in the TGFβ/MRTF pathway and provide the first demonstration of the molecular mechanism underlying the transcriptional regulation of Tβ4 genes by TGFβ signaling via a novel cis-acting element. Notably, in cancer cells, elevated Tβ4 contributed to tumor progression and metastasis by inducing tumor-associated genes, such as MYL9, via MRTF/SRF signaling. Although further studies are needed to understand the MRTF-independent functions of Tβ4, our study provides new and important insights into the roles of Tβ4 in EMT and tumor progression. We suggest MRTF signaling as a potential therapeutic target in tumors with enhanced Tβ4 expression.

T. Morita reports receiving a commercial research grant from Mitsubishi Tanabe Pharma Corporation and Project MEET, Osaka University Graduate School of Medicine. No potential conflicts of interest were disclosed by the other author.

Conception and design: T. Morita

Development of methodology: K. Hayashi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Morita,

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Morita,

Writing, review, and/or revision of the manuscript: T. Morita,

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Hayashi

Study supervision: T. Morita, K. Hayashi

This work was supported by JSPS KAKENHI grant number 15K07076 (to T. Morita) and 16K08142 (to K. Hayashi), the Takeda Science Foundation (to T. Morita), Mitsubishi Tanabe Pharma Corporation and Project MEET, Osaka University Graduate School of Medicine (to T. Morita).

The authors thank Prof. Yukio Kawahara and all the members in the laboratory for discussion and technical support. We thank all the staff in the Center for Medical Research and Education of Osaka University Graduate School of Medicine for their technical assistance. We also thank Enago (www.enago.jp) for the English language review.

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.