Redirecting RNA splicing by SMAD3 turns TGF-β into a tumor promoter

ABSTRACT

Transforming growth factor β (TGF-β) is a well-known growth inhibitor of normal epithelial cells, but it is also secreted by solid tumors to promote cancer progression. Our recent discovery of SMAD3-PCBP1 complex with direct RNA-binding properties has shed light on how this conversion is implemented by controlling pre-mRNA splicing patterns.

Transforming growth factor β (TGF-β) is well known to be a potent growth inhibitor and tumor suppressor; however, at the time when TGF-β was discovered, it was first identified as an activity that transformed normal fibroblasts to form colonies in soft-agar in cooperation with epidermal growth factor (EGF) or TGF-α, both of which are ligands of EGF receptors (EGFR).¹ Fast forward 35 y to present time, it has been appreciated that the two seemingly opposing functions of TGF-β are cell-contextual dependent: in normal epithelial, endothelial, and hematopoietic cells. TGF-β strongly inhibits growth, but this inhibition is lost in late stages of carcinomas with elevated levels of signaling from the mitogen-activated protein kinase (MAPK) pathway sustained by oncogenic mutations.^2,3 Multiple mechanisms by which TGF-β inhibits cell growth have been unraveled, chief among them are transcriptional induction of cyclin dependent kinase (CDK) inhibitors and suppression of proto-oncogene MYC and apoptosis inducer B-cell lymphoma 2 (BCL2). On the other hand, advanced tumors are frequently found with high levels of TGF-β that promotes tumor progression by inducing epithelial to mesenchymal transition (EMT) and invasion.^4,5 Several key transcriptional targets of TGF-β including Snail Family Zinc Finger 1 (SNAl1), Zinc finger E-box binding homeobox 1 (ZEB1), high mobility group AT-Hook 2 (HMGA2), and forkhead box A1 (FOXA1) have been shown to play important roles in TGF-β-induced EMT, while other nontranscriptional mechanisms have also been reported. Despite a vast body of the literature on the nuts and bolts of how TGF-β signals, the mechanism that underpins the switch of TGF-β from a growth inhibitor to a tumor promoter still remains elusive. In searching for proteins that confer regulation of the key TGF-β pathway transcription effector SMAD3 via phosphorylation of threonine 179 (T179) in the linker region, we identified an RNA-binding protein poly(RC) binding protein 1 (PCBP1, also known as hnRNP E1), and serendipitously discovered that by partnering with PCBP1, SMAD3 is brought onto the pre-mRNA of a cancer stem cell marker gene CD44 to regulate its alternative splicing.⁶ In addition to CD44, our global RNA-seq study revealed a plethora of cancers genes whose splicing patterns are altered by the SMAD3-PCBP1 interaction in favor of tumor progression. These findings let us to propose that regulation of alternative splicing by the concerted action of receptor-activated SMAD3 and PCBP1 is a key mechanism that propels TGF-β to a tumor promoter (Fig. 1).

Figure 1.

Oncogenic stress and TGF-β signaling stimulate SMAD3 to regulate alternative splicing independent of transcription. In normal epithelial cells, TGF-β binds to its type II and type I receptors and induces c-terminal phosphorylation of SMAD3, which enables SMAD3 to form a complex with SMAD4 and accumulate in the nucleus to regulate transcription of many genes, including those functioning in growth suppression, e.g., CDK inhibitor 1A (CDKN1A, also known as p21), CDK inhibitor 2B (CDKN2B, also known as p15), proto-oncogene MYC, and apoptosis inducer BCL2. In cancer cells, oncogenic stimulation, e.g., epidermal growth factor receptor (EGFR) amplification, RAS activation, or phosphatase and tensin homology (PTEN) mutation, results in threonine 179 (T179) phosphorylation of SMAD3 as well as PCBP1 phosphorylation. In the presence of TGF-β, although it is unable to form complex with SMAD4 or bind DNA, T179 and c-terminal phosphorylated SMAD3 associates with poly(RC) binding protein 1 (PCBP1) in the nucleus, binds pre-mRNA, and inhibits recruiting U2 snRNP auxiliary factor 2 (U2AF2) and spliceosomal assembly on suboptimal exons, thus affecting alternative splicing of CD44 and other EMT-related genes, such as platelet-derived growth factor subunit A (PDGFA), mitogen-activated protein kinase kinase kinase 7 (MAP3K7), STE20-like kinase (SLK), and kinesin-related protein 13 (KIF13).

According to the current paradigm of TGF-β signaling, SMAD3 is activated by phosphorylation at the carboxyl terminus SSXS motif by the ligand-bound receptor complex consisting of type I and type II receptors (TGFβRI and TGFβRII).⁷ Following this step, phosphorylated SMAD3 becomes accumulated in the nucleus in association with SMAD4 to regulate transcription in conjunction with a growing list of transcriptional co-activators, co-repressors, and chromatin modifiers. The linker region of SMAD3 is also heavily phosphorylated at multiple sites by MAPKs and CDKs through SMAD-independent conduits of noncanonical TGF-β signaling, or through crosstalks with the EGF, hepatocyte (HGF), fibroblast growth factor, and RAS-activated signaling pathways.⁸ Phosphorylation in the linker region generally suppresses transcriptional activities of SMAD3, which can result from the engagement of SMAD-specific E3 ubiquitin protein ligase 2 (SMURF2) or Neural Precursor Cell Expressed, Developmentally DownRegulated 4-Like (NEDD4L) to disrupt SMAD3-SMAD4 complex or prevent DNA binding due to steric hindrance imposed by monoubiquitin modification of SMAD3 in recognition of the phosphorylated T179.⁹ Since tumor cells generally bear the hallmark of elevated RAS-MAPK and CDK activities, phosphorylation at the SMAD3 linker region through crosstalks with the proliferative kinases offers a way to evade growth inhibition of TGF-β by shutting down its transcriptional programs. Underscoring the tumor-suppressing role of TGF-β, mutations of TGFβRI, TGFβRII, and SMAD4 are frequently identified in gastrointestinal, pancreatic, breast, head and neck, and cervical cancers.^2,3,10 Paradoxically, SMAD3 mutations are rarely seen in tumors, despite its close structural and functional similarity to SMAD4. Moreover, interrogation of pancreatic and colon cancer genomics has found that SMAD4 mutations usually occur during late stages of tumorigenesis and are prognostic indicators of metastasis, advanced diseases, and reduced survival.¹⁰ Clearly, SMAD3 and SMAD4 play very different roles in mediating TGF-β functions in cancer.

At the outset of our study, we noticed that many different cancer cell lines exhibited elevated basal levels of phosphorylation at T179 of SMAD3 compared to normal mammary epithelial MCF10A cells.⁶ We also found that a nonphosphorytable T179V mutant of SMAD3 while still retaining its transcriptional activity was nevertheless inept in restoring the mesenchymal gene expression, TGF-β-mediated invasion, and formation of lung metastasis to SMAD3-depleted cells. These experimental clues suggested that phosphorylation at T179 could contribute to the contextual cue of cancer progression and prompted us to seek nontranscriptional mechanism of SMAD3 in mediating TGF-β-induced mesenchymal gene expression and cell invasion and metastasis. Against this backdrop, PCBP1 was identified as a SMAD3-binding partner.⁶ The notable features of this new interaction include phosphorylation-induced binding and colocalization of both proteins in the nuclear antigen serine and arginine-rich splicing factor 2 (SRSF2, also known as SC35) positive speckles that are intimately involved in RNA splicing. Modeling on a heavily spliced CD44 gene, we demonstrated that SMAD3 indeed participates directly in the splicing of CD44 mRNA isoforms that are preferentially expressed in association with EMT. We showed that the SMAD3-PCBP1 complex binds directly to the variable region of CD44 pre-mRNA, thus denying the access of U2 snRNP auxiliary factor 2 (U2AF2), which is required for the assembly of spliceosome on the inherently weak variable exons of CD44. Since the interaction between SMAD3 and PCBP1 is mediated by phosphorylated T179, which can be phosphorylated either through noncanonical TGF-β signaling or responding to activated EGFR-RAS-MAPK pathways, our observations not only breaks a conceptual ground in revealing the splicing function of SMAD3, the dual requirement of TGF-β and EGFR activation for launching the splicing regulation by SMAD3-PCBP1 also cast an insight into how TGF-β becomes a tumor promoter in late stages of cancer cells where oncogenic mutations sustain elevated MAPK pathways. Moreover, since disruption of the SMAD3-SMAD4 complex would inevitably free up SMAD3 for participating in nontranscriptional, splicing control, our findings may in fact also offer an explanation to why SMAD4 mutations can be protumorigenic in certain instances.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

Research in Y.E. Zhang's lab is supported by intramural program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research.