Smad2Δexon3 and Smad3 have distinct properties in signal transmission leading to TGF-β–induced cell motility

In mammalian cells, Smad2 and Smad3, two receptor-regulated Smad proteins, play crucial roles in the signal transmission of transforming growth factor-β (TGF-β) and are involved in various cell regulatory processes, including epithelial–mesenchymal transition–associated cell responses, that is, cell morphological changes, E-cadherin downregulation, stress fiber formation, and cell motility enhancement. Smad2 contains an additional exon encoding 30 amino acid residues compared with Smad3, leading to distinct Smad2 and Smad3 functional properties. Intriguingly, Smad2 also has an alternatively spliced isoform termed Smad2Δexon3 (also known as Smad2β) lacking the additional exon and behaving similarly to Smad3. However, Smad2Δexon3 and Smad3 signaling properties have not yet been compared in detail. In this study, we reveal that Smad2Δexon3 rescues multiple TGF-β–induced in vitro cellular responses that would become defective upon SMAD3 KO but does not rescue cell motility enhancement. Using Smad2Δexon3/Smad3 chimeric proteins, we identified that residues Arg-104 and Asn-210 in Smad3, which are not conserved in Smad2Δexon3, are key for TGF-β–enhanced cell motility. Moreover, we discovered that Smad2Δexon3 fails to rescue the enhanced cell motility as it does not mediate TGF-β signals to downregulate transcription of ARHGAP24, a GTPase-activating protein that targets Rac1. This study reports for the first time distinct signaling properties of Smad2Δexon3 and Smad3.


Edited by Phyllis Hanson
In mammalian cells, Smad2 and Smad3, two receptorregulated Smad proteins, play crucial roles in the signal transmission of transforming growth factor-β (TGF-β) and are involved in various cell regulatory processes, including epithelial-mesenchymal transition-associated cell responses, that is, cell morphological changes, E-cadherin downregulation, stress fiber formation, and cell motility enhancement. Smad2 contains an additional exon encoding 30 amino acid residues compared with Smad3, leading to distinct Smad2 and Smad3 functional properties. Intriguingly, Smad2 also has an alternatively spliced isoform termed Smad2Δexon3 (also known as Smad2β) lacking the additional exon and behaving similarly to Smad3. However, Smad2Δexon3 and Smad3 signaling properties have not yet been compared in detail. In this study, we reveal that Smad2Δexon3 rescues multiple TGFβ-induced in vitro cellular responses that would become defective upon SMAD3 KO but does not rescue cell motility enhancement. Using Smad2Δexon3/Smad3 chimeric proteins, we identified that residues Arg-104 and Asn-210 in Smad3, which are not conserved in Smad2Δexon3, are key for TGF-βenhanced cell motility. Moreover, we discovered that Smad2-Δexon3 fails to rescue the enhanced cell motility as it does not mediate TGF-β signals to downregulate transcription of ARHGAP24, a GTPase-activating protein that targets Rac1. This study reports for the first time distinct signaling properties of Smad2Δexon3 and Smad3.
Transforming growth factor-β (TGF-β) is a pleiotropic cytokine involved in the regulation of various cellular processes during embryogenesis as well as adult tissue homeostasis. TGF-β-derived signaling is transmitted through both the Smad and non-Smad signaling pathways. In the Smad signaling pathway, ligand stimulation induces the TGF-β type I receptor-mediated phosphorylation of Smad2 and Smad3, receptor-regulated Smads (R-Smads), at their C termini, leading to their heterotrimeric complex formation with Smad4, and translocation into the nucleus. The Smad complex subsequently upregulates or downregulates gene expression, through binding to genomic regulatory regions, in cooperation with Smad-binding transcription factors and coactivators/corepressors (1). Smad-binding transcription factors, collectively "Smad cofactors," are thought to assist the selective and stable binding of Smad proteins to the genomic DNA, thereby contributing to context-dependent gene expression in the target cells (2). Alternatively, R-Smads can affect gene expression via the repression or derepression of other transcription factors via physical interaction (3,4), where the DNA-binding activity of Smads could be dispensable.
Five mammalian R-Smad proteins have been identified so far, among which Smad2 and Smad3 transmit TGF-β-, activin-, and Nodal-related signaling, whereas Smad1, Smad5, and Smad8 transmit signaling from bone morphogenetic proteins (5). Although Smad2 and Smad3 share 92% sequence identity, they are not functionally redundant: they exhibit distinct binding partners and oligomeric states, nuclear import mechanisms, DNA-binding properties, as well as different target genes and spatiotemporal expression patterns (6,7). The pathophysiological importance of Smad2 and Smad3 is context dependent: in cells cultured in vitro, Smad3 appears to be crucial in several TGF-β-induced cell responses. In contrast, Smad2 is indispensable for in vivo embryonic development as its KO induces embryonic lethality in mice, whereas Smad3 KO mice are viable (6). Smad2-mediated transcriptional regulation has thus been well explored in the embryonic developmental context. In FoxH1-related developmental signaling, Smad2 is essential for mesendoderm gene induction, whereas Smad3 plays only a limited role in the process (8,9). In immune cell development, Smad2 regulates positively, whereas Smad3 regulates negatively Th17 differentiation (10)(11)(12). In skin squamous cell carcinoma and non-small cell lung cancer cells, Smad2 suppresses, whereas Smad3 promotes cancer formation, malignant progression, and metastasis (13)(14)(15)(16). In addition, Smad2 mutations are found in various cancers at a low frequency, whereas Smad3 mutations are rare and could be found only in colon cancers (17). Although Smad2 and Smad3 are independently recruited to often distinct genomic regions upon TGF-β stimulation in A549 lung cancer cells (16), the underlying mechanisms of the distinct Smad2 and Smad3 functions remain mostly unclear.
The most striking biochemical difference between Smad2 and Smad3 is the interaction of Smad3, but not full-length Smad2, with CAGA motif-containing DNA sequences upon C-terminal phosphorylation. Exon 3 of Smad2 is key to this process, encoding 30 amino acid residues located at the Nterminal side of the DNA-binding β-hairpin structure to interfere with Smad2 DNA binding (18,19). Moreover, exon 3 affects the cytoplasmic localization of Smad2 (20). Consistently, Smad2Δexon3 (Smad2β), an alternatively spliced Smad2 isoform (21), exhibits DNA-binding ability and behaves similarly to Smad3 in biochemical assays including CAGA-Luc reporter and electrophoretic motility shift assays (18,19). In Xenopus animal cap assay, where Smad2 and Smad3 exert distinct target gene induction, Xenopus Smad2Δexon3 activity was more similar to that of Xenopus Smad3 than that of Xenopus Smad2; Xenopus Smad2Δexon3 thus partially loses unique Xenopus Smad2 features (22). Smad2Δexon3 and Smad3 are reportedly functionally interexchangeable during early embryonic development in mice (23), although Smad2-Δexon3 is significantly less efficient compared with full-length Smad2 in the nodal stimulation-related mesendoderm gene induction of embryoid body cells (9).
In the present study, we examined if Smad2Δexon3 and Smad3 are functionally distinct during epithelialmesenchymal transition (EMT), a process involved in tumor progression (24). Using the CRISPR-Cas9 technology, we prepared SMAD2 exon 3/SMAD3-double KO A549 cells that exclusively express the Smad2Δexon3 isoform as R-Smad that transmits signals from TGF-β/activin/Nodal. Smad2Δexon3 could mediate multiple TGF-β-induced and EMT-associated cellular responses, including cell morphological changes, E-cadherin downregulation, stress fiber formation, although it failed to mediate cell motility enhancement. The signaling difference was attributed to Arg-104 in the β4 region of the MH1 domain and Asn-210 in the linker region of Smad3. Both amino acid residues are required for downregulating ARH-GAP24, encoding the GTPase-activating protein (GAP) Fil-GAP that targets Rac1. To the best of our knowledge, this study describes first the functional differences between Smad2Δexon3 and Smad3 either in vitro or in vivo.
We performed a DNA affinity precipitation assay using the Smad-binding 3xCAGA probe. Our results revealed that the probe precipitated Smad3 and Smad4 upon TGF-β stimulation in parental A549 cells, whereas no Smad4 precipitation could be detected in SMAD3-KO cells, which could be recovered in A549-S2E3/S3-KO cells expressing Smad2Δexon3 (Fig. 1C). In addition, the TGF-β-responsive CAGA 12 -MLP-Luc reporter, which was almost inactive in A549-S3-KO cells, was activated in A549-S2E3/S3-KO cells (Fig. 1D). These biochemical properties were consistent with a previous report on overexpressed Smad2Δexon3 (19).

Smad2Δexon3 rescued cellular responses to TGF-β attenuated in SMAD3-KO A549 cells except for enhanced cell motility
Although Smad2Δexon3 restored CAGA 12 -MLP-Luc reporter activity in A549-S3-KO cells, it remained unclear if other cellular responses, attenuated by Smad3 deficiency, could be restored. Therefore, as a next step, we examined cellular responses induced by TGF-β in A549-S2E3/S3-KO cells. We could observe that Smad2Δexon3 expression restored cell morphology changes, E-cadherin downregulation, stress fiber formation, and TGF-β-induced cytostasis (Fig. 2, A-E). However, we discovered that TGF-β-enhanced cell motility, assessed by chamber migration and wound healing assays, was not restored (Fig. 2, F and G). These findings indicate that Smad2Δexon3 has distinct signaling properties from Smad3.
Lysine-114 substitution with arginine in the β4 region renders the Smad2Δexon3 MH1 domain the signal transmission ability leading to enhanced cell motility The MH1 domains of Smad2Δexon3 and Smad3 share 88% identity with an insertion of 10 additional amino acid residues in Smad2 (Fig. 4A). We recently reported that the β4 region in the Smad3 MH1 domain is crucial for cell motility-related signal transmission, via downregulating ARHGAP24 to prevent accelerated Rac1 inactivation (25). A divergent residue could be identified between Smad2Δexon3 (Lys-114) and Smad3 (Arg-104) in this region (Fig. 4A). Therefore, we constructed a Lys114Arg mutant of the S2Δ/3(Δ2-3-3) chimeric protein and introduced it into A549-S3-KO cells (Fig. 4B). As anticipated, TGF-β-enhanced cell motility, Rac1 activation, and downregulation of ARHGAP24 were successfully restored by this mutant (Fig. 4, C-F).
Knockdown of ARHGAP24 rescued the defect of TGF-βenhanced cell motility in A549-S2E3/S3-KO cells Finally, we examined how ARHGAP24 knockdown could affect TGF-β-enhanced cell motility. ARHGAP24 knockdown was verified by mRNA and protein expression analyses (Fig. 8A). Figure 3E shows that TGF-β failed to enhance cell motility in A549-S3-KO cells expressing S2Δ/3(3-2-3) or S2Δ/ 3(Δ2-3-3), whereas TGF-β successfully enhanced cell motility upon ARHGAP24 knockdown in these cells (Fig. 8B). These findings indicate that ARHGAP24 downregulation is the key biochemical event downstream of the Smad3 MH1 domain and linker region during TGF-β-mediated cell motility enhancement. However, these regions might induce other biochemical events leading to enhanced cell motility to some extent, as the ARHGAP24 knockdown-induced cell motility restoration was partial.

Discussion
Smad proteins play a crucial role in signal transduction downstream of cytokines of the TGF-β family. Among those, Smad2 and Smad3 principally mediate TGF-β-, activin-, and Nodal-related signaling to the nucleus. Although these two Smad proteins share 92% sequence similarity, their functions have been considered distinct (6,(26)(27)(28). TGF-β-induced growth inhibition of epithelial cells is dependent on Smad3 but not Smad2 (29). Breast cancer cells become more aggressive in the absence of Smad2 in part because Smad2 represses basal vascular endothelial growth factor A expression (15). Certain TGF-β-induced genes are both Smad2 and Smad3 dependent, whereas others solely depend either on Smasd2 or Smad3 (16,28). In certain cases, Smad2 and Smad3 counteract each other (8,(10)(11)(12)15). However, the differences in Smad2 and Smad3 signaling properties have not yet been well understood. In the present study, we described the distinct cell signaling properties of Smad2Δexon3, an alternatively spliced Smad2 isoform, and Smad3.

Function of the Smad2Δexon3 isoform
Although Smad2Δexon3 is a minor Smad2 isoform in mouse tissues (9), its signaling properties during embryonic development have already been examined (23). Gene-engineered mice exclusively expressing Smad2Δexon3 as R-Smad downstream of TGF-β/activin/Nodal (Smad2 Δexon3 :Smad3 null homozygous mice) exhibit no apparent phenotypes except for osteoarthritis, observed in Smad3 null mice. In addition, Smad2 null homozygous mice expressing Smad2Δexon3 are viable, whereas those expressing the full-length Smad2 are embryonic lethal (23). Therefore, Smad2Δexon3-mediated signaling is necessary and sufficient for embryonic development.
In contrast, Smad2Δexon3 function in adult tissue homeostasis remains elusive. In several healthy and cancerous human cell lines, Smad2Δexon3 represents only a small fraction of the total Smad2 protein (19). To date, several Smad3-specific binding partners have been identified (6,30). However, certain Smad3-specific binding partners reportedly interact with Smad2Δexon3 (6). Therefore, it was anticipated that certain defects caused by the loss of Smad3 could be restored by Smad2Δexon3. Consistently, we found that Smad2Δexon3 rescued multiple TGF-β-induced EMTrelated cellular phenotypes but not enhanced cell motility. Together with the previous finding that Smad2Δexon3 is sufficient for embryonic development (23), TGF-β/activin/ Nodal-induced cell motility might be dispensable for morphogenesis.

The underlying mechanisms of TGF-β-enhanced cell motility through Smad3
TGF-β is a representative extracellular EMT inducer. We previously reported that signals leading to cell motility are Smad2Δexon3/Smad3 chimeras and E-cadherin were determined by immunoblotting with the indicated antibodies; α-tubulin was used as a loading control. C, chamber migration assay. TGF-β1 stimulation for 12 h. D, wound healing assay. TGF-β1 stimulation for 24 h. Quantification is shown in the bottom. E, Rac1 activation assay. The amount of active GTP-loaded Rac1 was determined by using a glutathione-S-transferase (GST) pull-down assay. Rac1 was detected by immunoblotting. One representative result from three independent experiments is shown. Others are presented in Fig. S2. Quantification is shown on the right. F, TGF-β-induced ARHGAP24 downregulation. Cells were stimulated with TGF-β1 for 8 h on collagen-coated plates and subjected to quantitative real-time PCR. The scale bars represent 200 μm (D). Error bars represent the SD (n = 3 for E and F and n = 5 for C and D). The p values were determined by Student's t test (C, D, and F) or Dunnet's multiple comparison test (E). *p < 0.01. One representative result from two independent experiments is shown (C, D, and F). TGF-β, transforming growth factor-β.
Signaling from the Smad2Δexon3 isoform    (25,31,32). TGF-β-induced cell motility is thought to contribute to the invasive and metastatic properties of malignant tumor cells. Consistently, blocking Smad signaling reportedly attenuates the metastasis of breast cancer cells without affecting primary tumor growth (33).
Thus far, several Smad-binding transcription factors (Smad cofactors), such as JunB (34), Olig1 (31), and Sox4 (35), have been reportedly involved in TGF-β-induced cell motility or invasive properties. In addition, Smad3 linker phosphorylation reportedly enhanced cell motility via a Pin1-dependent mechanism in PC-3 human prostate cancer cells and normal immunoblotting with the indicated antibodies. B, the expression of E-cadherin was determined by immunoblotting with anti-E-cadherin; α-tubulin was used as a loading control. C, chamber migration assay. TGF-β1 stimulation for 12 h. Quantification is shown in the bottom. D, formation of actin stress fibers. F-actin was stained using rhodamine-phalloidin. E, cells were stimulated with TGF-β1 for 8 h on collagen-coated plates and subjected to quantitative realtime PCR. The scale bars represent 100 μm (C) and 25 μm (D). Error bars represent SD (n = 3 for D and n = 4 for C). The p values were determined by Student's t test. *p < 0.01. One representative result from two independent experiments is shown (C and D). F, schematic illustration of differences among Smad2, Smad2Δexon3, and Smad3 in TGF-β signaling. Smad2 and Smad3 differ in their activities in TGF-β-induced cellular responses. Smad2Δexon3 transmits Smad3-dependent cellular responses other than cell motility. Arg-104 and Asn-210 in Smad3 contribute to TGF-β-enhanced cell motility through the downregulation of ARHGAP24. TGF-β, transforming growth factor-β. The p values were determined by Student's t test. *p < 0.01. C, Rac1 activation assay. The amount of active GTP-loaded Rac1 was determined by using a glutathione-S-transferase (GST) pull-down assay. Rac1 was detected by immunoblotting. One representative result from two independent experiments is shown (A-C). TGF-β, transforming growth factor-β. murine mammary gland epithelial cells (31,36). However, the downstream events leading to enhanced cell motility remained poorly understood.
We recently reported that the Smad3 β4 region in the MH1 domain is involved in transmitting signals triggering cell motility. The signal through the β4 region leads to ARHGAP24 (encoding Fil-GAP, a Rac1-targeting GAP) downregulation, a key event during TGF-β-enhanced cell motility, although this is not the sole Smad3-mediated signaling output (25). In this study, we found that the Smad3 linker region is also required for downregulating ARHGAP24.
Any potential cooperativity between the β4 and the linker regions in downregulating ARHGAP24 remains to be elucidated. Alternatively, these two regions might independently affect ARHGAP24 downregulation. Intriguingly, the responsible region in the linker (Asn-210-Asn-218 in Smad3 and Thr-252-Ser-260 in Smad2) is closely located or overlapped with the linker phosphorylation sites (37). In the C-terminal half of the linker region, Smad3 displays three phosphorylation sites (Ser-204, Ser-208, and Ser-213), whereas Smad2 exhibits six phosphorylation sites (Ser-240, Ser-245, Ser-250, Thr-252, Ser-255, and Ser-260). Among these Smad2 sites, four (Ser-240, Ser-245, Ser-250, and Ser-255) are shared with Smad3, whereas Thr-252 and Ser-260 are not conserved between Smad2 and Smad3 (Fig. 5A). Thr-252 and Ser-260 are phosphorylated by Araf kinase (38) and calmodulin-dependent kinase II (39), respectively. A possibility is that Thr-252 (Thr-222 in Smad2Δexon3) phosphorylation is inhibitory to the linkermediated signaling through attenuating or facilitating the physical interaction with certain binding partners. However, this possibility could be excluded because when Asn-210 in Smad3 was mutated to Ala, TGF-β-enhanced cell motility was attenuated (Fig. S3). Therefore, Asn-210 is likely to play an enhancing role in signal transmission. We previously found that the presence of the linker region affects the MH2 domain-mediated interaction with Olig1 (31). Therefore, the linker sequence might affect the MH2 domain-mediated physical interaction. Arg-104-and Asn-210-dependent Smad3 binding partners, which are likely to be involved in enhanced cell motility, remain to be identified.

Future perspectives
We previously reported that TGF-β-induced actin stress fiber formation and cell morphological changes are also dependent on Smad3 (25). We have previously attempted to identify the region in Smad3 responsible for the signaling inducing these effects through approaches using Smad1/ Smad3 chimeric proteins, although we could not obtain satisfying results (25). However, we recently found that these responses are dependent on Snail, a core EMT-transcription factor (K. Miyazawa, unpublished observations). Snail can be induced by both TGF-β-and bone morphogenetic proteinrelated signaling (40), the latter of which use Smad1, Smad5, and Smad8 as downstream effectors. Therefore, approaches using Smad1/3 chimeric proteins might not be appropriate for identifying the region in Smad3 required for actin stress fiber formation and cell morphological changes. Intriguingly, Snail knockdown also attenuated TGF-β-enhanced cell motility (K. Miyazawa, unpublished observations). We previously reported that Smad3 exhibits a function other than downregulating ARHGAP24 in enhanced cell motility (25). Therefore, the additional Smad3-mediated signaling might be associated with the Snail pathway, which appears to be common with Smad2Δexon3. Future investigation should focus on unveiling these unsolved questions.

Experimental procedures
Cell lines and CRISPR-Cas9 system-mediated genome editing A549 cells were obtained from the American Type Culture Collection (41), cloned, and used as parental cells, authenticated by short tandem repeat analysis (4). The establishment of SMAD3 KO A549 (A549-S3-KO) cells was described previously (4). The KO of SMAD2 exon 3 was conducted using Double Nickase Plasmid (catalog no.: sc-400475-NIC-2; Santa Cruz Biotechnology). Deletion/disruption of the target exon was confirmed by sequencing (Fig. S1). The KO of exon 3 resulted in the dominant expression of Smad2Δexon3, an alternatively spliced isoform (Fig. 1A). Cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, supplemented with 50 units/ml of penicillin and 50 μg/ml of streptomycin, at 37 C under a 5% CO 2 atmosphere.

Biochemical assays
Cell lysis and immunoblotting were performed as described previously using semidry transfer (31) except for Figure 1, in which wet tank transfer was used to detect lowly expressed endogenous Smad3. Immunofluorescent E-cadherin detection was described previously (42). Actin stress fiber formation was detected by rhodamine-conjugated phalloidin staining (Cytoskelton, Inc) as described previously (43). Rac1 activation, chamber migration, wound healing, and cell proliferation assays were performed as described previously (25,32). The luciferase reporter assay, DNA affinity precipitation assay using 3xCAGA probe, and quantitative real-time PCR were performed as described previously (44). Table S1 shows the primer sequences for quantitative real-time PCR.

Conventional RT-PCR
To detect Smad2 isoform expressions, RT-PCR analysis was performed. Total RNAs were isolated by using ISOGEN (NIPPON GENE). First-strand cDNA synthesis was performed using the PrimeScript first Strand cDNA Synthesis Kit and random hexamers (Takara Bio). The sequences of the applied primers were 5 0 -TTTTCCTAGCGTGGCTTG (forward) and 5 0 -CTGGTGTCTCAACTCTCTGA (reverse) (19).

Lentivirus infection
The lentiviral vectors encoding either Smad3 or Smad2/3 chimeras were generated by the Gateway technology (Invitrogen). Lentivirus particles were produced as previously described (41) and then used to infect A549-S3-KO cells. Cells infected with Smad3 or Smad2/3 chimeras were cloned by limiting dilution in 96-well plates. Multiple clones with equivalent infected protein expression levels were used for the experiments. For experiments in Figure 7, pools of infected cells were used.

Statistical analysis
A two-sided Student's t test or Dunnet's multiple comparison test was used to determine the significant differences among the experimental groups. Probability values below 0.01 were considered significant; *p < 0.01.

Data availability
Data are contained within the article and its supporting information.