PRMT5 methylating SMAD4 activates TGF-β signaling and promotes colorectal cancer metastasis

Perturbations in transforming growth factor-β (TGF-β) signaling can lead to a plethora of diseases, including cancer. Mutations and posttranslational modifications (PTMs) of the partner of SMAD complexes contribute to the dysregulation of TGF-β signaling. Here, we reported a PTM of SMAD4, R361 methylation, that was critical for SMAD complexes formation and TGF-β signaling activation. Through mass spectrometric, co-immunoprecipitation (Co-IP) and immunofluorescent (IF) assays, we found that oncogene protein arginine methyltransferase 5 (PRMT5) interacted with SMAD4 under TGF-β1 treatment. Mechanically, PRMT5 triggered SMAD4 methylation at R361 and induced SMAD complexes formation and nuclear import. Furthermore, we emphasized that PRMT5 interacting and methylating SMAD4 was required for TGF-β1-induced epithelial-mesenchymal transition (EMT) and colorectal cancer (CRC) metastasis, and SMAD4 R361 mutation diminished PRMT5 and TGF-β1-induced metastasis. In addition, highly expressed PRMT5 or high level of SMAD4 R361 methylation indicated worse outcomes in clinical specimens analysis. Collectively, our study highlights the critical interaction of PRMT5 and SMAD4 and the roles of SMAD4 R361 methylation for controlling TGF-β signaling during metastasis. We provided a new insight for SMAD4 activation. And this study indicated that blocking PRMT5-SMAD4 signaling might be an effective targeting strategy in SMAD4 wild-type CRC.


INTRODUCTION
Transforming growth factor-β (TGF-β) signaling is a crucial enforcer of cell fate decisions during development and tissue homeostasis and plays a critical role in the occurrence and metastasis of colorectal cancer (CRC) [1][2][3][4]. Hence, perturbations in this pathway can lead to a plethora of developmental disorders and diseases, including cancer. Double faces of TGF-β signaling in cancer have been explored by multiple studies [1][2][3][4]. In the initial stage of CRC, TGF-β/SMADs signals inhibit tumorigenesis by regulating cell cycle arrest and apoptosis [1][2][3][4]. However, in the advanced stage, TGF-β signaling promotes the growth and metastasis of tumors by inducing immune escape, promoting vascular formation and inducing epithelial-mesenchymal transition (EMT) [1][2][3][4]. In canonical TGF-β signaling pathway, the active TGF-β dimer signals bring together type I and type II receptors [2]. The type II receptors can phosphorylate and activate the type I receptors, followed by propagating the signal through phosphorylating SMADs transcription factors. Once TGF-β signaling is activated, the receptor substrate SMADs (R-SMADs) form complexes with SMAD4 (Co-SMAD). Subsequently, the complexes shuttle to the nucleus, and initiate transcription of downstream molecules, triggering tumor metastasis and other processes [1][2][3][4][5].
Mutations and posttranslational modifications (PTMs) contribute to the formation, activation, and destruction of SMAD complexes [6,7]. SMAD4, as a major partner of SMAD complexes, is essential for TGF-β signaling pathway activation. Germline mutations in SMAD4 are the major cause of juvenile polyposis syndrome (JPS), an autosomal dominant predisposition to multiple gastrointestinal polyps and cancer [8,9]. Accumulation of evidence confirmed that the loss or mutation of SMAD4 in earlyonset CRC occurs at a frequency of about 30%, which suggested that SMAD4 acted as a tumor suppressor in colorectal carcinogenesis and was similar to the role of TGF-β signaling in tumor initiation [7,10]. However, most of CRC were SMAD4 wild-type, and TGF-β signaling was known as an important molecule to induce EMT in a SMAD4-dependent manner through the induction of translocation of SMAD2/3/4 complexes to the nucleus, leading to the expression of mesenchymal markers Snail, Slug, Twist and Zeb [1,2]. Thus, SMAD4 plays a role of a tumor metastasis inducer in cancer progression [1][2][3][4].
In this study, we reported that PRMT5 interacted with SMAD4 and triggered SMAD4 methylation at R361 under TGF-β1 treatment, which was the most frequent mutated site of SMAD4 in CRC and reported to show significant effects on SMAD complexes formation and nuclear accumulation [6,7,10,29,30]. We proved that SMAD4 R361 methylation was required for SMAD complexes formation. PRMT5 interacting and methylating SMAD4 was required for TGF-β-induced CRC metastasis. Furthermore, we explored that PRMT5 overexpression or a high level of SMAD4 R361 methylation indicated worse outcomes in clinical specimen analysis. Our study highlights the critical interaction of PRMT5 and SMAD4 and the roles of SMAD4 R361 methylation for controlling TGF-β signaling during metastasis in SMAD4 wild-type CRC.

RESULTS
PRMT5 symmetrically dimethylates SMAD4 at R361 Posttranslational regulation of SMADs was already reported in our previous work [11] and some other research [31]. To better understand the regulations of the Co-SMADs molecule, SMAD4, we conducted a series of mass spectrometric analyses to detect PTMs of SMAD4 and its binding proteins. Firstly, SMAD4-Flag fusion protein was overexpressed in HEK293T cells and a Co-IP assay was performed to enrich the SMAD4-Flag fusion protein (Fig.  S1A). Remarkably, we found that SMAD4 could be methylated at R361 (Fig. 1A). SMAD4 mutation frequently occurred in multiple types of cancer, including CRC [6,10,32,33] (Fig. S1B). R361 was the most frequent mutational site of SMAD4 in CRC [6,7,10,29,33]. As shown in OncoKB database (https:// www.oncokb.org), SMAD4 R361 mutation was detected in over 10% of CRC patients (Fig. S1C). Notably, this site is highly conserved across species (Fig. 1B). SMAD4 R361 site was also reported to show significant effects on SMAD complexes formation and nuclear accumulation [7,30].
To determine the methyltransferase mediating SMAD4 R361 methylation, we conducted mass spectrometric analyses of SMAD4 binding proteins. Two protein arginine methyltransferases, PRMT1 and PRMT5, were clarified from the mass spectrometric analyses (Dataset 1). Co-IP assays were performed to verify that SMAD4 was able to bind with PRMT1 or PRMT5 (Fig. 1C). PRMT1 and PRMT5 mediated two different types of protein arginine methylation, asymmetrical (ADMA) and symmetrical dimethylation (SDMA) [20,21,23,34]. We then sought to determine which methyltransferase(s) mediated the SMAD4 R361 methylation by a quantified Co-IP (qCo-IP) assay. An equal amount of SMAD4-Flag WT or SMAD4 R361K-Flag mutant plasmid was transfected into HEK293T cells. Symmetrical and asymmetrical dimethylation antibodies were utilized to detect SDMA and ADMA levels of SMAD4. SDMA was downregulated in the R361K mutant, while ADMA was not significantly changed (Fig. 1D). It indicated that PRMT5, but not PRMT1, was the key enzyme that could symmetrically dimethylate SMAD4 at R361. To confirm our hypothesis, we tried to overexpress PRMT5 in HEK293T cells or treat HEK293T cells with the PRMT5-specific inhibitor GSK3326595, which was reported to be substrate competitive and was a potential drug in clinical trials for various malignant tumors. We verified the effect of GSK3326595 in HEK293T and human CRC cell lines SW48 and LoVo (Fig. S1D). qCo-IP assays revealed that SDMA level of SMAD4 was increased when PRMT5 was overexpressed, while it was decreased after GSK3326595 treatment (Figs. 1E, F, S1E). Co-IP assays confirmed the combination of SMAD4 and PRMT5 in human colorectal cancer cell lines SW48 and LoVo (Fig.  S2A). Through immunofluorescence (IF) assay, we observed that SMAD4 was co-localized with PRMT5 in HEK293T and SW48 and LoVo (Fig. S2B). To evaluate the level of SMAD4 R361 symmetrical dimethylation, we generated an antibody that could specifically recognize SMAD4 R361 SDMA (anti-SMAD4 R361me2s). A dot-blot assay of SMAD4 naked and SMAD4 R361(me)2(Symmetrical) peptide was performed to verify the potency and specificity (Fig.  1G). Most importantly, an in vitro methylation assay was performed to confirm the methylation of SMAD4 by PRMT5 (Fig.  1H). Taken together, our data showed that PRMT5 interacted with SMAD4 and mediated SMAD4 R361 symmetrical dimethylation.
PRMT5 promotes the SMAD complexes formation depending on methylation of SMAD4 at R361 SMAD4 plays a central role as it is the shared heterooligomerization partner of other SMADs. Hetero-trimerization of SMADs is mediated by the highly conserved C-terminal MH2 domains of SMAD2/3 and SMAD4. Interactions among MH2 domains bring the linker and MH1 domains of all subunits into proximity. Tumor-derived trimer interface mutations D351H, R361C, V730D, and D537E of SMAD4 MH2 domain disrupt both homo-and hetero-oligomerization of SMADs [7,29]. As SMAD4 R361 was reported to be important for the activation of TGF-β/ SMADs signaling, we sought to determine whether PRMT5 could regulate TGF-β/SMADs signaling via SMAD4 R361 methylation. Firstly, we employed a gene set enrichment analysis (GSEA) and the result showed a strong correlation between PRMT5 expression and TGF-β signaling activity (Fig. S2C). Gene expressing data of CRC samples downloaded from the TCGA database (cohort: TCGA Colon and Rectal Cancer (COADREAD)) was divided into groups according to whether SMAD4 was mutated. In samples with higher expression of PRMT5, GSEA revealed the enrichment of TGF-β signaling in the datasets of SMAD4 wild-type samples, but not in the datasets of the SMAD4 R361 mutants (Fig. S2D, E). These Fig. 1 PRMT5 methylates Smad4 at R361. A Secondary mass spectrogram of a peptide covering R361. M/z represents mass divided by charge number. The horizontal axis in a mass spectrum is expressed in units of m/z. B Structure of SMAD4 and R361 in multiple species. C Co-IP assays and followed immunoblot (IB) analysis performed in HEK293T cell line. D qCo-IP assays and followed IB analysis performed in HEK293T cell line. SMAD4 WT and R361K mutant plasmids were equally transfected into HEK293T for qCo-IP assays. A followed IB assay measured the SDMA and ADMA levels of pull-down proteins. E HEK293T cells were collected for qCo-IP to test the change of SDMA after PRMT5 overexpression. F HEK293T cells were collected for qCo-IP assays to test the change of SDMA upon GSK3326595 (60 nM, 4 h) treatment. Representative graph and statistical analysis were shown in (D-F) (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, t-test). G A dot-blot assay of SMAD4 naked and SMAD4 R361me2s peptide (0-1 μg) was conducted to verify the potency and specificity of anti-SMAD4 R361me2s. H An in vitro methylation assay was conducted to determine the methylation of SMAD4 mediated by PRMT5 at R361. The reaction product was collected for IB analysis.
Since R361 was a critical site for SMADs hetero-oligomerization [7,29], we suspected that R361 methylation might play a role in the formation of SMAD complexes. A qCo-IP assay showed that SDMA of SMAD4 increased within 5 min after TGF-β1 stimulation, while the binding of SMAD2/3 to SMAD4 appeared subsequently ( Fig. 2A). Moreover, both the qCo-IP assay and IF assay revealed A. Liu et al. that the binding of PRMT5 and SMAD4 was enhanced under TGF-β1 treatment ( Fig. 2A, B). To further explore the effect of PRMT5 on SMAD complexes formation, we carried out a series of qCo-IP assays. A SMAD4 R361K HEK293T cell line was generated by knockout SMAD4 (Fig. S2F) and stably expressing SMAD4 R361K mutant. Expectedly, under the treatment of TGF-β1, SMAD4 R361me2s level and SMAD2/3 binding to SMAD4 increased, but the upregulation was significantly impaired in the SMAD4 R361K mutant or under the treatment of GSK3326595 (Fig. 2C, D). To identify the detailed mechanism of PRMT5 mediating the combination of SMAD4 with SMAD2/3, we employed a qCo-IP assay to further explore whether the formation of SMAD complexes regulated by PRMT5 was dependent on SMAD4 R361 methylation. Our data showed that PRMT5 overexpression significantly enhanced the combination of SMAD2/3 with SMAD4 upon TGF-β1 stimulation in WT cells, but there was no obvious change in the R361K mutant (Fig. 2E, F). Meanwhile, in R361K mutant, GSK3326595 treatment showed little effect on SMAD complexes formation upon TGF-β1 stimulation (Fig. S2G). According to the above, PRMT5 could affect the interaction of SMAD4 between SMAD2/3 in a SMAD4 R361 methylation-dependent way.

SMAD4 R361 methylation is required for SMAD complexes nuclear translocation
For TGF-β signaling, SMADs form multimeric complexes and translocated into the nucleus to drive multiple tumor progression related genes expression under the TGF-β stimulation [7,30]. Since our data showed that PRMT5 could affect the interaction between SMAD4 and SMAD2/3 through a SMAD4 R361 methylation-dependent way, we speculated that PRMT5 could consequently promote SMAD complexes transporting into the nucleus. IF assays were carried out to observe SMAD2/3 nuclear accumulation. SMAD2 and SMAD3 were distinctly enriched in the nucleus after TGF-β1 stimulation, while SMAD complexes nuclear translocation was inhibited by GSK3326595 treatment or under SMAD4 R361K mutant condition (Fig. 3A, B). Cytoplasmic and nuclear proteins were isolated after stimulation of TGF-β1. Consistently, western blot assays showed that PRMT5 intervention or SMAD4 R361 mutation deprived SMADs of accumulation in the nucleus in response to TGF-β1 (Fig. 3C, D). Furthermore, overexpression of PRMT5 enhanced the nuclear aggregation of SMAD complexes in WT cells, but not in SMAD4 R361K mutant (Fig. 3E, F). Collectively, our data proved that PRMT5 promoted SMAD complexes formation and nuclear import, which were necessary for TGF-β signaling activation, through a SMAD4 R361 methylation-dependent way.

PRMT5 promotes TGF-β-induced EMT and metastasis of CRC in vitro and in vivo
We compared the expression of PRMT5 between normal colon and tumor and in varying degrees of malignancy in GEO database. The results showed that PRMT5 was remarkably upregulated in CRC rather than in normal colon and tended to overexpress in the tissue of higher malignancy (Fig. S3A, B). Highly expressed PRMT5 predicted poor disease-specific survival and disease-free survival rates (Fig. S3C, D). The expression level of PRMT5 in various tumors and adjacent tissues in TCGA database were also been analyzed by using UALCAN web portal, data showed that PRMT5 expression was higher in most cancers than in adjacent tissue, including CRC (Fig. S3E).
In the advanced stage of cancer, TGF-β signaling pathway enhances tumor immunosuppression and facilitates tumor angiogenesis, invasion, and metastasis [3]. Since our previous data showed that SMAD4 R361me2s mediated by PRMT5 promoted the formation and nuclear accumulation of SMAD complexes, which was essential for their transcriptional activity [7], the effect of PRMT5 on TGF-β-mediated EMT turned out to be an interesting question. As the bioinformatics analysis showed that PRMT5 was overexpressed in CRC, we suspected that PRMT5 might affect the EMT of tumor cells through activating TGF-β signaling pathway. Firstly, we measured the mRNA levels of SNAIL and SLUG, which could be transcripted by SMAD complexes in response to TGF-β and induce EMT [1,2,35]. Human CRC cell lines SW48 and LoVo have mutant TGF-β receptors but were reported to respond to TGF-β1 [36]. Consistent with previous reports, the relative mRNA levels were remarkably increased after TGF-β1 treatment (Fig. 4A). The upregulation was impaired by GSK3326595 treatment (Fig.  4A). However, when PRMT5 was overexpressed, the transcription of SNAIL and SLUG improved more profoundly (Fig. 4B). The protein expression of Snail and Slug was also increased in responding to TGF-β1 stimulation, which was consistent with the changes in mRNA levels (Fig. 4C, D). We also found that N-cadherin was upregulated while E-cadherin was decreased (Figs. 4C, D and S4A, B). But, after GSK3326595 treatment, the change of the above proteins triggered by TGF-β1 was deprived (Figs. 4C and S4A). While the change was more obvious when PRMT5 was overexpressed (Figs. 4D and S4B). All these data indicated that PRMT5 played a role in TGF-β-mediated EMT.
Given that PRMT5-mediated SMAD4 R361me2s was necessary for TGF-β signaling pathway activation, we sought to determine if the effect of PRMT5 on TGF-β-induced EMT was dependent on SMAD4 R361me2s. Under TGF-β1 treatment, both mRNA and protein expression of indicated markers showed that EMT was impaired by SMAD4 R361K mutation, and could not be totally reversed by PRMT5 overexpression (Fig. 4E, F). Transwell assays showed that the migration ability appeared to be stronger when PRMT5 was overexpressed, while it was weakened by SMAD4 R361K mutation under TGF-β1 treatment. Although PRMT5 overexpression resulted in more cell migration in the R361K mutant, the extent of migration was not as much as in the WT cell line (Fig. 4G).
We further explored the role of SMAD4 R361me2s in CRC metastasis in vivo. Firstly, PRMT5 was stably knocked down in luciferase-expressing LoVo cell lines. shNC and shPRMT5 cells were inoculated into the spleen of BALB/c nude mice to evaluate metastatic ability. As the results show, the knockdown of PRMT5 remarkably prevented metastasis (Figs. S2H and 5A, B). IHC staining confirmed the level of PRMT5 and SMAD4 R361me2s in tumor tissue (Fig. 5C). In addition, compared to WT tumor cells, the R361K mutant showed less metastasis ability. PRMT5 overexpression in WT cell line dramatically increased metastasis, but not so much in R361K mutant (Fig. 5D-E). Moreover, since PRMT5 was found to play a role in CRC progression, whether PRMT5 inhibition Fig. 2 PRMT5 promotes the formation of SMAD complexes dependent on methylation of SMAD4 at R361. A Cells were starved in serumfree medium for 12 h. A qCo-IP assay measured the SDMA of SMAD4 and PRMT5, SMAD2/3 binding to SMAD4 for the indicated duration of TGF-β1 (5 ng/mL) treatment, followed by IB analysis. B Cells were starved in serum-free medium for 12 h. IF staining of SMAD4 and PRMT5 in HEK293T for the indicated duration of TGF-β1 (5 ng/mL) treatment. Representative images were obtained by a confocal scanning microscope. C HEK293T cells were treated with GSK3326595 (60 nM) for 4 h and then treated with TGF-β1 (5 ng/mL) for 20 min. The cell extracts were harvested for qCo-IP assay with anti-SMAD4, followed by IB analysis. D HEK293T WT or R361K cells were treated with TGF-β1 (5 ng/mL) for 20 min. The cell lysis was harvested for qCo-IP assay with anti-SMAD4, followed by IB analysis. E, F HEK293T WT or R361K cells were transfected with an equal amount of vector or HA-PRMT5. A qCo-IP assay followed by an IB analysis was carried out with or without TGF-β1 (5 ng/mL, 20 min) treatment. The qCo-IP assay and IB analysis was repeated three times for gray value statistics. Representative graph E and statistical analysis F (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, t-test).
was efficient against metastasis is worth investigating. As shown in Fig. 5D, E, GSK3326595 treatment could significantly inhibit liver metastasis in SMAD4 WT CRC. However, the efficacy was weaker towards R361K mutant cells (Fig. 5D, E). Taken together, PRMT5 could promote EMT and tumor metastasis partially depending on methylating SMAD4 at R361. Also, GSK3326595 showed the potential for CRC treatment, especially in the SMAD4 wild-type patients.
The level of SMAD4 R361me2s was a prognostic marker for CRC Since our data showed that SMAD4 R361me2s mediated by PRMT5 could promote CRC metastasis both in vitro and in vivo, we supposed that the level of SMAD4 R361me2s was promising to be a prognostic marker for CRC. Tumor specimens were collected from patients with CRC who underwent surgical resection. IHC staining showed that PRMT5 was significantly upregulated in most samples (Fig. 6A). The same sections were stained with previously generated anti-SMAD4 R361me2s. The intensities of SMAD4 R361me2s were increased in most of the samples, and it appeared to be related to the level of PRMT5 expression (Fig. 6A). To investigate the relationship between SMAD4 R361me2s and CRC outcomes, a tissue microarray was analyzed by calculating the IHC score of SMAD4 R361me2s staining (Fig. S3F). Excluding the specimens without tumors or adjacent glands, 6 out of 63 samples were not stained, the ratio was similar to the mutation frequency of SMAD4 R361 in CRC. And we did not include these 6 samples in the follow-up statistics. We found that the IHC score of SMAD4 R361me2s in tumor lesions was higher than in the adjacent tissue, which was consistent with our data before (Fig.  6B). Compared to patients with no metastasis detected at surgery, IHC scores of tumor tissue were higher in patients with lymph nodes or distant metastases (Fig. 6C). Also, the level of SMAD4 R361me2s was positively correlated with the number of metastatic lymph nodes (Fig.  6D). Moderately and poorly differentiated tumors seemed to show higher IHC scores compared with well-differentiated tumors, but there was no statistical significance (Fig. 6E). For patients without postoperative chemotherapy, higher levels of R361me2s (IHC score ≥ 6) indicated worse outcomes (Fig. 6F). However, there was no significant difference between the two groups if patients underwent postoperative chemotherapy were included (Fig. 6G). These results confirmed that SMAD4 R361me2s was able to be a prognostic marker to predict the risk of metastasis and the outcome.

DISCUSSION
PTMs on SMAD4, including phosphorylation, ubiquitylation, sumoylation and ADP-ribosylation, were previously reported. The function and stability of SMAD4 are extensively regulated [12-16, 31, 37-40]. Here, we found that SMAD4 could be modified by methylation, and we clarified the necessity of SMAD4 R361me2s in SMAD complexes formation and nuclear import. PRMT5 could promote EMT and CRC metastasis through methylating SMAD4 at R361 thus increasing TGF-β pathway activity. We investigated that SMAD4 R361me2s functioned as a factor in predicting the outcomes of patients with CRC and we emphasized that targeting PRMT5 was promising to treat CRC in patients without SMAD4 R361 mutation (Fig. 6H).
Our study showed that PRMT5 promoted SMAD4 R361 methylation, which was positively correlated with the activity of the TGF-β pathway. Structurally, the hydrogen bond formed at the R361 was essential for the formation of SMADs oligomers. Mutations at this residue disrupted the intricate hydrogen bond network at the interface [29]. SMAD complexes formation and nuclear import were intervened [7,30]. We inferred that the R361 methylation was involved in the interaction between SMADs, and mutation of R361 deprived its methylation state, which probably impaired the formation and nuclear entry of SMAD complexes, and then inhibited the normal activation TGF-β pathway. Our inference needs to be further verified by experiments. We clarified that PRMT5 promoted metastasis by methylating SMAD4 at R361 thus activating TGF-β pathway in the advanced stage of cancer. However, stable activation of the TGF-β signaling pathway is crucial for the regulation of cell cycle arrest and apoptosis in normal tissues and early-stage tumors, and the inactivation of TGF-β pathway will lead to carcinogenesis and tumor progression. SMAD4 R361 mutations are extremely common in colorectal cancer [3,6,7,10,29,30]. That means, in normal tissues and early-stage tumors, once R361 is mutated, loss of methylation may inhibit the normal activation of TGF-β pathway and promote carcinogenesis or tumor progression to some extent. Therefore, the effect of methylation at R361 site in carcinogenesis and early-stage tumors deserves further study.
In the current study, we found a decreased activity of TGF-β signals towards GSK3326595 treatment. As a PRMT5 inhibitor, GSK3326595 has entered phase I clinical trials (NCT02783300) for treating non-Hodgkin lymphoma and some solid tumors [20,[45][46][47]. A previous study showed that inhibition of PRMT5 initiated lymphocyte infiltration in the early stage of MYC-driven hepatocellular cancer, and the combination of GSK3326595 improved the efficacy of anti-PD1 therapy [48]. Our work revealed the positive relationship between PRMT5 and TGF-β pathway. In vitro and in vivo studies proved that PRMT5 could promote EMT partially depending on methylating SMAD4 at R361. And we clarified that GSK3326595 had the potential for CRC treatment, especially in the SMAD4 wild-type patients. In addition, our study also explored that the level of SMAD4 R361me2s was positively correlated with the number of metastatic lymph nodes and higher levels of R361me2s indicated worse outcomes, indicating SMAD4 R361me2s was able to be a prognostic marker to predict the risk of metastasis.

Study approval
This study was approved by Huazhong University of Science and Technology Ethics Committee. All animal experiments were approved by the Animal Care and Use Committee of Tongji Hospital.

MATERIALS AND METHODS Cell lines and cell culture
Human cell lines SW48, LoVo and HEK293T were purchased from American Type Culture Collection (ATCC). Cells were cultured in DMEM high glucose medium (Gibco, USA) containing 10% FBS (Thermo Fisher Scientific, USA). All of the cells were cultured at 37°C in a 5% CO 2 incubator. Fig. 3 The methylation of SMAD4 at R361 facilitates SMAD complexes' nuclear import. A, B WT, R361K mutant and HEK293T cells treated with or without GSK3326595 (60 nM, 4 h) were stimulated with TGF-β1 (5 ng/mL) for 1 h. An IF staining of SMAD2/3 and F-actin to observe the transporting of indicated proteins. Representative images were obtained by a confocal scanning microscope. C HEK293T cells were respectively treated with GSK3326595 (60 nM, 4 h) and TGF-β1 (5 ng/mL, 1 h). Cytoplasmic and nuclear proteins were separated. A western blot assay detected indicated proteins from each sample. D HEK293T WT and R361K mutant cells were respectively treated with TGF-β1 (5 ng/mL) for 1 h. Cytoplasmic and nuclear proteins were separated for IB analysis. E, F HEK293T WT and R361K mutant cells were transfected with vector or HA-PRMT5 plasmid. Cells were respectively treated with TGF-β1 (5 ng/mL) for 1 h. Cytoplasmic and nuclear proteins were separated for IB analysis. Representative graph (E) and statistical results (F) (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, t-test).

Western blot, dot-blot, and Co-IP assays
For western blot assays, cells were washed with PBS and lysed with NP40 lysis buffer containing protease inhibitor cocktail (MCE, HY-K0010, USA). Bicinchoninic Acid Assay (BCA) assay kit (Thermo Fisher Scientific, USA) was used to measure the protein concentration. Cytoplasmic and nuclear protein fractions were extracted by subcellular structure nuclear and cytoplasmic protein extraction kit (Boster, AR0106, China). Protein samples were electrophoretically separated in Tris-HCl sodium dodecyl sulfatepolyacrylamide gel (SDS-PAGE) and transferred to PVDF membrane. For the dot-blot assay, different quantified proteins were dropped on a PVDF membrane. The above antibodies and HRP-conjugated secondary antibodies were used to incubate with the membrane. The membrane bands were visualized with ECL reagents (Abbkine, China). Each experiment was repeated at least three times for gray value statistics.
For Co-IP assays, corresponding antibodies were incubated with protein A/G magnetic beads (MCE, HY-K0202, USA), and the cell lysate was added to the reaction system for a further 2 h at room temperature. Magnetic beads were washed 4 times with NP40 lysis buffer and then boiled in SDS-loading buffer for 10 min. Protein binding to the beads was analyzed by Western blot subsequently. For quantified Co-IP (qCo-IP) assays, equal numbers of cells were seeded in each group. The same volume of antibody and magnetic beads was used in each group.

Mass spectrometric analyses
The SMAD4-Flag fusion protein (expressed in HEK293T cells) was enriched by Co-IP assay. The protein sample was separated by SDSf-PAGE. After Coomassie blue staining, the gel bands of interest were excised from the gel, reduced with 5 mM of DTT, alkylated with 11 mM iodoacetamide, and digested in the gel with α-lysate in 50 mM ammonium bicarbonate overnight at 37°C. The next day, the sample was quenched by adding 10% TFA to adjust the pH to below 2. Then the sample was digested with sequencing-grade modified chymotrypsin in 50 mM ammonium bicarbonate at 25°C overnight. The peptides were extracted twice with 0.1% trifluoroacetic acid in 50% acetonitrile aqueous solution for 1 h and then dried in a speedVac. Peptides were redissolved in 25 μL 0.1% trifluoroacetic acid and 6 μL of extracted peptides were analyzed by Orbitrap Fusion Lumos mass spectrometer.
For LC-MS/MS analysis, the peptides were separated by a 40 min gradient elution at a flow rate of 0.30 µL/min with a Thermo-Dionex Ultimate 3000 HPLC system, which was directly interfaced with an LTQ Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, Germany).

Immunofluorescence assays
Cells were seeded on a cover slide in 24-well culture plates with different treatments. The cells on the cover slide were washed with pre-cool PBS and then fixed with 4% paraformaldehyde for 30 min at room temperature. Next, the non-specific antigens were blocked with 1% BSA-TBST solution for 2 h. Slides were incubated with the primary antibody overnight at 4°C. Then, the slides were embedded with fluorochrome-conjugated secondary antibodies for 2 h at room temperature in the dark and incubated with DAPI (Cat# 9542, Sigma, USA) for 10 min subsequently. We used a confocal scanning microscope (Olympus FLUOVIEW FV1000, Japan) or a fluorescence microscope (Nikon ECLIPSE Ti2, Japan) to obtain representative images. Tumor specimens, CRC tissue microarrays, and immunohistochemistry staining Clinical samples were obtained from CRC patients who underwent surgical resection at the department of gastrointestinal surgery, Wuhan Tongji Hospital, and informed consent was obtained. The human CRC tissue microarray of 80 primary tumors and adjacent normal tissues bought from Wuhan Baiqiandu Technology Co., Ltd. (Cat. #80A1) was used to evaluate the level of SMAD4 R361me2s. For IHC staining, lesions or adjacent tissues Fig. 4 PRMT5 promotes TGF-β-induced EMT and metastasis of CRC in vitro. A qPCR was conducted to measure the relative mRNA fold change of SNAIL and SLUG in SW48 and LoVo cells upon GSK3326595 (60 nM, 24 h) and TGF-β1 (5 ng/mL, 24 h) treatment (*p < 0.05, **p < 0.01, ***p < 0.001, t-test). B HA-PRMT5 plasmid or vector was transfected into SW48 and LoVo cells. qPCR was conducted to measure the relative mRNA change of SNAIL and SLUG in response to TGF-β1 (5 ng/mL, 24 h) stimulation (*p < 0.05, **p < 0.01, ***p < 0.001, t-test). C SW48 and LoVo cells were treated with GSK3326595 (60 nM, 36 h) and TGF-β1 (5 ng/mL, 36 h). Western blot assays detected indicated proteins. D HA-PRMT5 plasmid or vector was transfected into SW48 and LoVo cells, followed by TGF-β1 (5 ng/mL, 36 h) treatment. Western blot assays detected indicated proteins. E HA-PRMT5 plasmid or vector was transfected into SW48 and LoVo WT or R361K cells, followed by TGF-β1 (5 ng/mL, 24 h) treatment. qPCR measured the relative mRNA change of SNAIL and SLUG (*p < 0.05, **p < 0.01, ***p < 0.001, t-test). F HA-PRMT5 plasmid or vector was transfected into SW48 and LoVo WT or R361K cells followed by TGF-β1 (5 ng/mL, 36 h) treatment. Western blot assays detected indicated proteins. Each western blot assay was repeated 3 times. Representative graph was shown. G A transwell assay was performed to evaluate the migration of CRC in the R361K mutant or during PRMT5 overexpression (*p < 0.05, **p < 0.01, ***p < 0.001, t-test).

In vitro methylation assays
were fixed with formalin solution followed by paraffin embedded. Slides were dewaxed, rehydrated, and heated in sodium citrate buffer for antigen retrieval. Endogenous peroxidase was inhibited with 3% hydrogen peroxide and 0.1% sodium azide for 30 min at room temperature. Nonspecific epitopes were blocked with incubation in 5% BSA for 2 h at room temperature. Next, the slides were incubated with indicated primary (diluted anti-PRMT5 (1:150) and diluted anti-SMAD4 R361me2s (1:50)) antibodies at 4°C overnight and then with the secondary antibodies at room temperature for 1 h.

Transwell migration assays
Transwell migration assays were carried out using 24-well 8 μm pore size Transwell chambers (Corning Incorporated, costar 3422, USA). Colorectal cancer cells (2 × 10 5 ) in a 200 μL volume of serum-free medium were seeded into the upper chambers and cultured at 37°C for 16 h for in vitro migration assay. The up surfaces of the chamber were wiped with cotton swabs. Cells across pores were fixed with 4% paraformaldehyde and then stained with 1% crystal violet solution. The migrating cell numbers were counted in 3 randomly selected microscope fields (200×).

Quantitative real-time PCR (qPCR) reactions
The total RNA of tumor cells was isolated with RNAiso reagent (Takara, Japan) and then reversely transcripted into cDNA with cDNA Synthesis Kit (gDNA digester plus) (Yeasen, 11121ES60, China). Reactions were carried out using Hieff UNICON Universal Blue qPCR SYBR Green Master Mix (Yeason, 11184ES03, China) on QuantStudio 3 System (AppliedBiosystem Thermo Fisher Scientific, USA). The results were calculated using the Comparative Ct method with GAPDH as a housekeeping control.
Nude mice CRC liver metastasis model and imaging BALB/c nude mice (6 w, female) were bought from Gempharmatech Co., Ltd and were kept in the Laboratory Animal Center of Wuhan Tongji Hospital. All mouse experiments were performed following a protocol approved by our institutional Animal Care and Use Committee. Mice were randomly assigned to groups (5 mice per group). After administering appropriate anesthesia, 2 × 10 6 cells of each group suspended in 125 μL serum-free media were injected carefully into spleens. After 6 weeks, D-Luciferin, Potassium Salt D (150 mg/kg) was intraperitoneally injected, and imaging analysis was performed 10 min after injection. The mice were sacrificed and dissected subsequently. Then livers were removed, and tumors were stripped and weighed.

Statistical analysis
Data were statistically analyzed using one-way ANOVA or two-tailed Student's t-tests for comparison between groups. Clinicopathological parameters associated with the level of R361me2s were assessed by Chisquare tests and Fisher's exact tests. A Spearman test was used to analyze the correlation between gene expressions. Survival curves were plotted by the Kaplan-Meier method. All statistical analysis was performed with GraphPad Prism 8.0 (GraphPad Software) and SPSS 22.0 (IBM). Variance was similar between the groups that were statistically compared. All data represent the mean ± SEM. Statistical significance was set at p < 0.05. Fig. 5 Targeting the methylation of SMAD4 at R361 inhibits CRC metastasis in vivo. A, B 2 × 10 6 luciferase-expressing LoVo cells of different genotypes, shNC, PRMT5 sh#1 and sh#2, were injected intrasplenic to BALB/c nude mice. After 6 weeks, bioluminescence was detected, then livers were dissected, and metastatic tumors were weighed. Representative images (A) and quantitative analysis of metastatic tumor weight (B) (*p < 0.05, **p < 0.01, t-test). C IHC staining of PRMT5 and SMAD4 R361me2s in tumor tissue. D, E 2 × 10 6 luciferase-expressing cells of different genotypes, WT, R361K mutant, PRMT5 overexpressing and PRMT5 overexpressing R361K mutant, were injected intrasplenic into BALB/c nude mice. After 3 weeks of tumor cell injection, half of the mice injected with WT and R361K cells were treated with GSK3326595 (40 mg/kg) every 3 days. After another 3 weeks, bioluminescence was detected, then livers were dissected, and metastatic tumors were weighed. Representative images (D) and quantitative analysis of metastatic tumor weight (E) (*p < 0.05, **p < 0.01, t-test). Fig. 6 The level of SMAD4 R361me2s can be used as a prognostic indicator for colorectal cancer. A HE staining, anti-PRMT5 and anti-SMAD4 R361me2s IHC staining of tumor specimens from 6 patients who underwent surgery. B IHC score of tumor and adjacent tissue (*p < 0.05, **p < 0.01, ***p < 0.001, t-test). C IHC score of samples from patients with or without metastasis detected (*p < 0.05, **p < 0.01, ***p < 0.001, t-test). D The number of positive lymph nodes in different levels of SMAD4 R361me2s (*p < 0.05, **p < 0.01, ***p < 0.001, χ 2 -test). E IHC scores of tumors with different degrees of differentiation, the p value was calculated via χ 2 -test. F Kaplan-Meier survival analysis comparing high and low levels of SMAD4 R361me2s in patients without postoperative chemotherapy (p-value was calculated via a log-rank test). G Kaplan-Meier survival analysis comparing overall survival between patients with high and low levels of SMAD4 R361me2s (p-value was calculated via a log-rank test). Patients who underwent postoperative chemotherapy were included. H A schematic diagram of PRMT5 regulating TGF-β signaling pathway and colorectal cancer metastasis through methylating SMAD4 at R361.