RASSF10 Is a TGFβ-Target That Regulates ASPP2 and E-Cadherin Expression and Acts as Tumor Suppressor That Is Epigenetically Downregulated in Advanced Cancer

The Ras Association Domain Family (RASSF) encodes members of tumor suppressor genes which are frequently inactivated in human cancers. Here, the function and the regulation of RASSF10, that contains a RA (Ras-association) and two coiled domains, was investigated. We utilized mass spectrometry and immuno-precipitation to identify interaction partners of RASSF10. Additionally, we analyzed the up- and downstream pathways of RASSF10 that are involved in its tumor suppressive function. We report that RASSF10 binds ASPP1 (Apoptosis-stimulating protein of p53) and ASPP2 through its coiled-coils. Induction of RASSF10 leads to increased protein levels of ASPP2 and acts negatively on cell cycle progression. Interestingly, we found that RASSF10 is a target of the EMT (epithelial mesenchymal transition) driver TGFβ (Transforming growth factor beta) and that negatively associated genes of RASSF10 are significantly over-represented in an EMT gene set collection. We observed a positive correlation of RASSF10 expression and E-cadherin that prevents EMT. Depletion of RASSF10 by CRISPR/Cas9 technology induces the ability of lung cancer cells to proliferate and to invade an extracellular matrix after TGFβ treatment. Additionally, knockdown of RASSF10 or ASPP2 induced constitutive phosphorylation of SMAD2 (Smad family member 2). Moreover, we found that epigenetic reduction of RASSF10 levels correlates with tumor progression and poor survival in human cancers. Our study indicates that RASSF10 acts a TGFβ target gene and negatively regulates cell growth and invasion through ASPP2. This data suggests that epigenetic loss of RASSF10 contributes to tumorigenesis by promoting EMT induced by TGFβ.


Introduction
Cancer is caused by a multistep genetic and epigenetic transformation of normal cells into highly invasive and immortal tumor cells. Epithelial cells are immobile cells that are responsible for organ integrity and structure. A key event of the transformation of epithelial cells into invasive tumor cells is associated with increased motility and disruption of cell-adhesion referred to as epithelial RASSF10 expression was determined in confluent cells and in single cell suspension in A549. Therefore, confluent cells were trypsinized, resuspended in full medium, placed at 37 °C under constant slow rotation and RNA was isolated at indicated time points. (c) Verification of CRISPR/Cas9 RASSF10 knockout in A549 on genomic and expression level by PCR for three RASSF10 negative clones (∆RASSF10) and three wtRASSF10 clones. (expected RASSF10 sizes: genomic wtRASSF10 333 bp; RT wtRASSF10 244 bp, CRISPR/Cas9 induces deletion and shortening of PCR product). (d) Morphological differences observed between confluent ∆RASSF10 and wtRASSF10 clones. 100 µm standard white. (e) Morphological differences observed between confluent ∆RASSF10 and wtRASSF10 clones when wide scratch was placed in cell culture dish. 100 µm standard black. (f) Growth differences by flow cytometry of wtRASSF10 (wtRX) and ∆RASSF10 (∆RX) clones isolated expression was determined in confluent cells and in single cell suspension in A549. Therefore, confluent cells were trypsinized, resuspended in full medium, placed at 37 • C under constant slow rotation and RNA was isolated at indicated time points. (c) Verification of CRISPR/Cas9 RASSF10 knockout in A549 on genomic and expression level by PCR for three RASSF10 negative clones (∆RASSF10) and three wtRASSF10 clones. (expected RASSF10 sizes: genomic wtRASSF10 333 bp; RT wtRASSF10 244 bp, CRISPR/Cas9 induces deletion and shortening of PCR product). (d) Morphological differences observed between confluent ∆RASSF10 and wtRASSF10 clones. 100 µm standard white. (e) Morphological differences observed between confluent ∆RASSF10 and wtRASSF10 clones when wide scratch was placed in cell culture dish. 100 µm standard black. (f) Growth differences by flow cytometry of wtRASSF10 (wtRX) and ∆RASSF10 (∆RX) clones isolated during proliferation at 10% density and (g) by wound healing assay. 100 µm standard black or white. (h) RASSF10 (RX) re-expression effects on cell cycle progression negatively (HeLa negative for endogenous RASSF10). Table 1. GO (Gene ontology) term analysis of RASSF10-associated genes in cancer cell lines [28].

GO Term
GOPath p-Value  These observations indicate that RASSF10 expression is regulated by cellular contact, density and/or epithelial-mesenchymal transition. Therefore, we generated a RASSF10 knockout in the prototypic epithelial cell line A549 by CRISPR/Cas9 (Figure 1c; frameshift deletion mutation). Loss of RASSF10 was associated with detachment from monolayer ( Figure 1d). We observed that at the scratch site the monolayer RASSF10 wildtype (wt) cells adhered strongly to neighboring cells in contrast to RASSF10 knockout cell clones (Figure 1e). RASSF10 loss was also associated with induced proliferation (phase G1/G0 decrease) by flow cytometry and increased wound healing ability (Figure 1f,g). In contrast, RASSF10 re-expression in HeLa cells with epigenetically inactivated endogenous RASSF10 halted cell proliferation (G1/G0 induction) measured by flow cytometry (Figure 1h).

RASSF10 Is Induced by the EMT Driver TGFβ and RASSF10 Depletion Promotes TGFβ Induced Invasion
As epithelial-mesenchymal transition was found in RASSF10 associated genes, we questioned if RASSF10 was regulated by the EMT driver TGFβ, which is also a modulator of tumorigenesis [29]. We found, that RASSF10 expression and promoter activity is dynamically regulated by TGFβ (Figure 2a-e) and TGFβ induction of RASSF10 was further verified in different cell lines, that express RASSF10 (Figure 2g). Additionally, we observed that induction of RASSF10 by TGFβ vs. mock was highest under low cellular density (Figure 2f), which was accompanied by the phenotypical spindle like morphology and upregulation of SNAI2 [30]. We also observed that TGFβ treatment of A549 cells for two days led to an upregulation of RASSF10 and long-term TGFβ exposure (six days), when cells adopted an EMT program [31], caused an inhibition of RASSF10 expression (Figure 2h). Consistently, short-term TGFβ treatment led to cell cycle arrest at phase G1-G0 as measured by flow cytometry (Figure 2i) in accordance with earlier results [32]. We assumed that loss of the tumor-suppressor RASSF10 in cancer contributes to the transition of epithelial to mesenchymal cell phenotypes.
To analyze the effect of RASSF10 in TGFβ induced invasion, we performed extracellular matrix (ECM) transwell invasion assays in A549 cells (Table 2). In RASSF10 knockout cells (∆RASSF10) we found that TGFβ treatment promotes a significant induction of ECM invasion compared to mock treated cells (p-value ≤ 0.03). However, this induction was not significant (p-value ≥ 0.14) in the wtRASSF10 cells suggesting that depletion of RASSF10 enhances significantly the ability of A549 cells to invade an ECM.

RASSF10 Interacts and Stabilzes Apoptosis-Stimmlating of p53 Protein 2 (ASPP2)
For RASSF10 interactome analysis, we performed co-immuno-precipitations of overexpressed RASSF10 using the green fluorescent protein GFP-Trap system in biological triplicates ( Figure 4a) and analyzed the RASSF10 partners by mass spectrometry (ESI). Results were normalized to likewise treated controls for each experiment (#1-#3) and we precipitated putative binding partners in overlap of three experiments (Table 3). Results came down to the apoptosis-stimulating proteins of p53 (ASPP1 and ASPP2) with a reproducible interaction with RASSF10 ( Figure 4), in all three experiments under high stringency. It has been reported that ASPP2 controls epithelial plasticity and inhibits EMT through regulation of β-catenin and CDH1 [26].    analyzed using [28]).

RASSF10 Interacts and Stabilzes Apoptosis-Stimmlating of p53 Protein 2 (ASPP2)
For RASSF10 interactome analysis, we performed co-immuno-precipitations of overexpressed RASSF10 using the green fluorescent protein GFP-Trap system in biological triplicates ( Figure 4a) and analyzed the RASSF10 partners by mass spectrometry (ESI). Results were normalized to likewise treated controls for each experiment (#1-#3) and we precipitated putative binding partners in overlap

RASSF10 Stabilizes ASPP2 Protein through Its Coiled-Coils Domain
To characterize the ASPP1/2-RASSF10 interaction, we generated mutants of RASSF10 containing only its RA-domain (RASSF10D1-133) or its C-terminus with coiled-coils (RASSF10D237-508). The binding of ASPP to RASSF10/RASSF10-mutants was tested by co-precipitation ( Figure S3a). Precipitation of ASPP1/2 was almost lost with RASSF10D1-133. Subsequently, we tested the strength of the ASPP-RASSF10 interaction by using RASSF10-mutants as competitors ( Figure S3b). ASPP2 was co-precipitated with most of the domain mutants and competition with RASSF10D1-133 did not interfere with the binding of ASPP2 to wt-RASSF10 ( Figure S3b). However, wt-RASSF10 can compete with RASSF10D237-362 regarding the binding to ASPP2 ( Figure S3b). The vice versa experimental set up with GFP-trapped wt-RASSF10 that co-precipitates ASPP1/2, is not interfered with by the competitor RASSF10D237-362 ( Figure S3c). For further competition experiments, we used the known ASPP binding partners YAP, p53 and p65 [35]. The strength of the ASPP-RASSF10 complex was further emphasized by the fact that YAP, p53 and p65 overexpression failed to compete with the coprecipitation of ASPP with RASSF10 ( Figure S3d).

Verification of RASSF10 Interaction with ASPP1 and ASPP2
For confirmation we used co-precipitation by GFP-Trap and GST-glutathione pulldown for exogenous ASPP1/2-RASSF10 and endogenous ASPP1/2-RASSF10 interaction. RASSF10 was overexpressed and co-precipitated overexpressed ASPP1 and ASPP2 in Western blot ( Figure 4b) and also endogenous ASPP1/2 ( Figure 4c). RASSF10 co-localized with ASPP1 and ASPP2, whereas the latter was altered in its localization by RASSF10 ( Figure S2). The interaction was further confirmed by co-precipitation of exogenous RASSF10 with GFP-trapped ASPP2 ( Figure 4d). Interestingly, we observed that RASSF10 not only interacts with ASPP1/2 but also stimulates the endogenous levels of ASPP2 but not ASPP1 (Figure 4c; 5th panel lane 2). To test the interaction of ASPP2 with endogenous RASSF10 we used the RASSF10 expressing lung cancer cell line A549. We overexpressed and isolated ASPP2 by GFP-Trap in HEK, which was then incubated with A549 lysates or from HEK overexpressing Flag-RASSF10 ( Figure 4e). We were able to detect co-precipitated endogenous RASSF10 in A549 approx. at the height of Flag-RASSF10 at 70 kDa. We used CRISPR/Cas9 to generate three RASSF10 Cancers 2019, 11,1976 9 of 24 negative clones in A549 (∆RX-A,B,C), which we controlled by PCR and sequencing the deletion within the coding region of RASSF10 ( Figure 1). Using three A549 RASSF10-negative clones and three RASSF10-positive wt clones (wt-A,B,C), we could show that knockout of endogenous RASSF10 abolishes its co-precipitation with ASPP2 ( Figure 4f). Without prior co-precipitations/enrichment, we were unable to detect endogenous RASSF10. In our experiments, the RASSF10 'SDS PAGE-displayed MW' is 70 kDa vs. its estimated 57 kDa size. The RASSF10 displayed size was verified with detected endogenous RASSF10 vs. loss of RASSF10 upon CRISPR/Cas9 knockout (Figure 4f). This detected difference in size can be due to its specific amino acid composition, post-translational modification and/or SDS (sodium dodecyl sulfate) occupancy in page [33,34].

RASSF10 Stabilizes ASPP2 Protein through Its Coiled-Coils Domain
To characterize the ASPP1/2-RASSF10 interaction, we generated mutants of RASSF10 containing only its RA-domain (RASSF10D1-133) or its C-terminus with coiled-coils (RASSF10D237-508). The binding of ASPP to RASSF10/RASSF10-mutants was tested by co-precipitation ( Figure S3a). Precipitation of ASPP1/2 was almost lost with RASSF10D1-133. Subsequently, we tested the strength of the ASPP-RASSF10 interaction by using RASSF10-mutants as competitors ( Figure S3b). ASPP2 was co-precipitated with most of the domain mutants and competition with RASSF10D1-133 did not interfere with the binding of ASPP2 to wt-RASSF10 ( Figure S3b). However, wt-RASSF10 can compete with RASSF10D237-362 regarding the binding to ASPP2 ( Figure S3b). The vice versa experimental set up with GFP-trapped wt-RASSF10 that co-precipitates ASPP1/2, is not interfered with by the competitor RASSF10D237-362 ( Figure S3c). For further competition experiments, we used the known ASPP binding partners YAP, p53 and p65 [35]. The strength of the ASPP-RASSF10 complex was further emphasized by the fact that YAP, p53 and p65 overexpression failed to compete with the co-precipitation of ASPP with RASSF10 ( Figure S3d).
Based on our initial observation that RASSF10 overexpression led to the stabilization of endogenous ASPP2, but not ASPP1 (Figure 4c), we verified that RASSF10 overexpression increases endogenous ASPP2 levels irrespective of the cell line or vectors used (Figure 5a).
We created a RASSF10-inducible cell line (TetOn-TREx-System, Thermo Fisher Scientific) in HEK293 using doxycycline. Induction of RASSF10 strongly increased endogenous ASPP2 protein levels ( Figure 5b), which was not due to increased ASPP2 RNA expression ( Figure 5c). Increased endogenous ASPP2 levels could also be shown by IF upon RASSF10 overexpression ( Figure S2b). Next, we questioned which RASSF10 domain facilitates the ASPP2 stabilization. We used RASSF10-domain mutants (RASSF10D1-133, RASSF10D237-362, RASSF10D237-508) and deletion mutants (RASSF10∆RA, RASSF10∆M, RASSF10∆CT, RASSF10∆C1, RASSF10∆C2 and RASSF10∆C1+C2) (Figure 5d,e). The RA-domain or coils alone did not stabilize ASPP2 levels ( Figure 5d). Deletion of RA-domain of RASSF10 or one coil retained the ability to stabilize ASPP2 (Figure 5e). In contrast, deleting the middle region of RASSF10 (∆M) or the C-terminus (∆CT) led to the loss of ASPP2 stabilization. Deletion of both coils almost completely abolishes ASPP2 stabilization (Figure 5e lower panel). Here we show that RASSF10 stabilizes ASPP2 through its coiled-coil domains. To understand the aberrant mechanism of the RASSF10-ASPP2 pathway which is relevant for human carcinogenesis, we investigated the genetic and epigenetic alteration of these genes in different tumor entities. Based on our initial observation that RASSF10 overexpression led to the stabilization of endogenous ASPP2, but not ASPP1 (Figure 4c), we verified that RASSF10 overexpression increases endogenous ASPP2 levels irrespective of the cell line or vectors used (Figure 5a). We created a RASSF10-inducible cell line (TetOn-TREx-System, Thermo Fisher Scientific) in HEK293 using doxycycline. Induction of RASSF10 strongly increased endogenous ASPP2 protein levels (Figure 5b), which was not due to increased ASPP2 RNA expression ( Figure 5c). Increased endogenous ASPP2 levels could also be shown by IF upon RASSF10 overexpression ( Figure S2b). Next, we questioned which RASSF10 domain facilitates the ASPP2 stabilization. We used RASSF10domain mutants (RASSF10D1-133, RASSF10D237-362, RASSF10D237-508) and deletion mutants (RASSF10∆RA, RASSF10∆M, RASSF10∆CT, RASSF10∆C1, RASSF10∆C2 and RASSF10∆C1+C2) (Figure 5d,e). The RA-domain or coils alone did not stabilize ASPP2 levels (Figure 5d). Deletion of RA-domain of RASSF10 or one coil retained the ability to stabilize ASPP2 (Figure 5e). In contrast, deleting the middle region of RASSF10 (∆M) or the C-terminus (∆CT) led to the loss of ASPP2 stabilization. Deletion of both coils almost completely abolishes ASPP2 stabilization (Figure 5e lower panel). Here we show that RASSF10 stabilizes ASPP2 through its coiled-coil domains. To understand the aberrant mechanism of the RASSF10-ASPP2 pathway which is relevant for human carcinogenesis, we investigated the genetic and epigenetic alteration of these genes in different tumor entities.

RASSF10 Is Inactivated across Human Cancers and a Valuable Cancer Biomarker
Tumor-suppressor gene inactivation can occur by loss of function mutation or promoter methylation [36,37]. We only found minimal genomic alterations of RASSF10 (2.4%) and ASPP2

RASSF10 Is Inactivated across Human Cancers and a Valuable Cancer Biomarker
Tumor-suppressor gene inactivation can occur by loss of function mutation or promoter methylation [36,37]. We only found minimal genomic alterations of RASSF10 (2.4%) and ASPP2 (4.7%) compared to 62% of TP53 mutations in cancer cell lines (n = 881; Broad CCLE/Cancer Cell Line Encyclopedia; analyzed using [38]). Similar mutation frequencies were found in primary tumors. TP53 is heavily mutated in various cancers especially in lung (>80%), head and neck (>70%), colorectal (>50%), breast (>30%), kidney chromophobe (30%) and below 5% in kidney clear cell carcinoma. For ASPP2 genetic mutations remain at very low levels with <1% in lung adenocarcinoma and kidney clear cell carcinoma and <2% in bladder cancer. For RASSF10 there are no genetic alterations across primary cancers (TCGA/The Cancer Genome Atlas; analyzed using [39]). In summary, we are not convinced that mutations of RASSF10 and ASPP2 are likely to contribute to tumorigenesis, whereas occurrence of TP53 mutations in cancer is consistent with literature [40].

RASSF10 Is Frequently Hypermethylated in Human Cancers
RASSF10 and ASPP2 contain large CpG islands within their promoter regions ( Figure 6, Figure 7 and Figure S4). In normal tissues, these both genes with CpG rich promoters are expressed (Figure 7 and Figure S4). We questioned whether the newly identified complex partners RASSF10-ASPP2 are regulated by promoter hypermethylation in cancer. We used the combined bisulfite restriction analysis COBRA methylation analysis, based on bisulfite conversion of unmethylated cytosins, for various cell lines: lung cancer cell lines: HTB171, HTB175, H64, H1672, CRL-2062 (small cell lung cancer); H358, A427, A549, H322 (non-small cell lung cancer); SK-MES-1 (squamous cell carcinoma); head and neck cancer Hep2, liver carcinoma HepG2, embryonal kidney HEK, cervix carcinoma HeLa and human fibroblasts HF55. We found that six out of 14 immortal cell lines showed methylation of RASSF10 whereas HF55 was unmethylated. However, ASPP2 was unmethylated in these cancer cell lines (Figure 6b). Figure S4). In normal tissues, these both genes with CpG rich promoters are expressed ( Figure  7 and Figure S4). We questioned whether the newly identified complex partners RASSF10-ASPP2 are regulated by promoter hypermethylation in cancer. We used the combined bisulfite restriction analysis COBRA methylation analysis, based on bisulfite conversion of unmethylated cytosins, for various cell lines: lung cancer cell lines: HTB171, HTB175, H64, H1672, CRL-2062 (small cell lung cancer); H358, A427, A549, H322 (non-small cell lung cancer); SK-MES-1 (squamous cell carcinoma); head and neck cancer Hep2, liver carcinoma HepG2, embryonal kidney HEK, cervix carcinoma HeLa and human fibroblasts HF55. We found that six out of 14 immortal cell lines showed methylation of RASSF10 whereas HF55 was unmethylated. However, ASPP2 was unmethylated in these cancer cell lines (Figure 6b).  Genome wide analysis using publicly available datasets (NCBI-GEO; Gene Expression omnibus) verified our findings (Figure 7 and Figure S4). RASSF10 is significantly hypermethylated in tumor tissues and cancer cell lines (p = 1.5 × 10 −29 ; Figure 7c) but ASPP2 is unmethylated (Supplementary Figure S4c). Increased methylation of RASSF10 was revealed in primary pancreatic adenocarcinoma and invasive breast carcinoma compared to normal tissue (p = 2.0 × 10 −9 ; Figure 7d). Methylation of RASSF10 occurs across the complete CpG island in tumor tissues and methylation is further increased in cancer cell lines, whereas normal tissues are unmethylated (Figure 7f). In contrast, ASPP2 remains unmethylated in cancer across its CpG island ( Figure S4e). RASSF10 expression is significantly decreased in cancer cell lines vs. normal tissues (p = 4.1 × 10 −62 ; Figure 7b), but not ASPP2 ( Figure S4b). As a causal correlation RASSF10 promoter methylation significantly correlates with decreased expression in cancer cell lines (p = 1.3 × 10 −38 ; Figure 7e), but not ASPP2 ( Figure S4d). RASSF10 expression levels vary in cancer cell lines, but not the expression levels of ASPP2 and expression of RASSF10 and ASPP2 do not correlate (Figure 7g). in cancer cell lines, whereas normal tissues are unmethylated (Figure 7f). In contrast, ASPP2 remains unmethylated in cancer across its CpG island ( Figure S4e). RASSF10 expression is significantly decreased in cancer cell lines vs. normal tissues (p = 4.1 × 10 −62 ; Figure 7b), but not ASPP2 ( Figure S4b). As a causal correlation RASSF10 promoter methylation significantly correlates with decreased expression in cancer cell lines (p = 1.3 × 10 −38 ; Figure 7e), but not ASPP2 ( Figure S4d). RASSF10 expression levels vary in cancer cell lines, but not the expression levels of ASPP2 and expression of RASSF10 and ASPP2 do not correlate (Figure 7g).  (d) Increased RASSF10 methylation in primary pancreatic adenocarcinoma and invasive breast carcinoma compared to normal tissue (cg05817758, data Lokk vs. TCGA vs. TCGA, Anova one way) (e) Loss of RASSF10 expression correlates with its promoter hypermethylation (cg05817758 and 11742468_at, log 2 ps, data Esteller vs. Garnett). (f) Increased RASSF10 promoter methylation (yellow unmethylated; blue methylated) is observed from normal to tumor tissues and cancer cell lines. Methylation is depicted relative to CpG island/shores and for all RASSF10 (cg) reporters from 450K array. (g) RASSF10 expression plotted relative to ASPP2 expression across cancer cell lines (CCLE Cancer Cell Line Encyclopedia, log2, data Broad, n = 917).

RASSF10 Inactivation Correlates with Clinical Diagnosis and Prognosis of Cancer
At last, we tested the ability of RASSF10 to serve as a prognostic and diagnostic biomarker in independent data sets for human neoplasia across various primary samples. In thymoma RASSF10 expression was reduced with progressed tumor type and correlated with reduced survival (Figure 8b; Table S2). In lymphoma reduced RASSF10 expression correlated with reduced survival (mantle cell) or an earlier death of patients (B-cell) (Figure 8c,d; Table S2). In breast cancer we can show that high RASSF10 methylation is associated with reduced RASSF10 expression, we observed that loss of RASSF10 expression correlates with progressed breast cancer grade and reduced overall survival of breast cancer patients (Figure 8e-g, Table S2). Similarly, RASSF10 expression was reduced in metastatic melanoma vs. non-metastatic melanoma types and in colon carcinoma vs. adenocarcinoma (Figure 8h,i, respectively). Reduced RASSF10 levels further reveal a reduced survival of colon cancer (Figure 8j Table S2 summarizes that low RASSF10 expression is associated with reduced 5 year survival rates of cancers: kidney cancer, 24% (papillary, significant) and 12% (clear cell, significant); head and neck cancer, 18% (significant); lymphoma, 22% (mantle cell, significant) and 5% (B-cell); thymoma, 8%; liver cancer, 34% (significant); lung cancer, 18% (significant); gastric cancer, 7%; and breast cancer, 21% (significant). In summary, we observed that the levels of RASSF10 expression/methylation are suitable for prognosis and diagnosis of various cancer types in humans.
high RASSF10 methylation is associated with reduced RASSF10 expression, we observed that loss of RASSF10 expression correlates with progressed breast cancer grade and reduced overall survival of breast cancer patients (Figure 8e-g, Table S2). Similarly, RASSF10 expression was reduced in metastatic melanoma vs. non-metastatic melanoma types and in colon carcinoma vs. adenocarcinoma (Figure 8h und 8i, respectively). Reduced RASSF10 levels further reveal a reduced survival of colon cancer (Figure 8j), head and neck cancer patients (Figure 8k), as well as liver cancer (Figure 8l), lung cancer ( Figure 8m) and gastric cancer patients (Figure 8n). Supplementary table S2 summarizes that low RASSF10 expression is associated with reduced 5 year survival rates of cancers: kidney cancer, 24% (papillary, significant) and 12% (clear cell, significant); head and neck cancer, 18% (significant); lymphoma, 22% (mantle cell, significant) and 5% (B-cell); thymoma, 8%; liver cancer, 34% (significant); lung cancer, 18% (significant); gastric cancer, 7%; and breast cancer, 21% (significant). In summary, we observed that the levels of RASSF10 expression/methylation are suitable for prognosis and diagnosis of various cancer types in humans.  (j) Relapse free survival was positively associated with RASSF10 expression in colon carcinoma (data SieberSmith). In (k) head and neck cancer, (l) liver cancer, (m) lung cancer and (n) gastric cancer reduced RASSF10 expression is associated with reduced survival (KM plotter).

RASSF10 or ASPP2 Depletion Induces Activation of the TGFβ Signaling Pathway
To analyze functional effects of RASSF10 and ASPP2 on TGFβ mediated signaling, we performed siRNA mediated knock-down (KD) and analyzed the molecular consequences on SMAD2 phosphorylation which is activated through the canonical TGFβ pathway (Figure 9). Knockdown of RASSF10 induced constitutive phosphorylation of SMAD2 at Ser465/467 and this constitutive activation of SMAD2 was also observed for the ASPP2 knockdown (Figure 9a). Interestingly, we found an enhanced phosphorylation of SMAD2 after TGFβ treatment and knockdown of ASPP2. Moreover, increased levels of ß-Catenin that represents another TGFβ target and a classical regulator of EMT were detected (Figure 9a). However, YAP1 levels were rather unaffected after RASSF10 or ASPP2 knockdown. Subsequently, we have analyzed the expression levels of CTGF that is a downstream effector of TGFβ signaling involved in EMT [41]. We revealed that overexpression of RASSF10 and ASPP2 reduces levels of CTGF in HeLa cells (Figure 9b). These data indicated that loss of RASSF10 or ASPP2 induced constitutive activation of TGFβ signaling and EMT.
(n) gastric cancer reduced RASSF10 expression is associated with reduced survival (KM plotter).

RASSF10 or ASPP2 Depletion Induces Activation of the TGFβ Signaling Pathway
To analyze functional effects of RASSF10 and ASPP2 on TGFβ mediated signaling, we performed siRNA mediated knock-down (KD) and analyzed the molecular consequences on SMAD2 phosphorylation which is activated through the canonical TGFβ pathway (Figure 9). Knockdown of RASSF10 induced constitutive phosphorylation of SMAD2 at Ser465/467 and this constitutive activation of SMAD2 was also observed for the ASPP2 knockdown (Figure 9a). Interestingly, we found an enhanced phosphorylation of SMAD2 after TGFβ treatment and knockdown of ASPP2. Moreover, increased levels of ß-Catenin that represents another TGFβ target and a classical regulator of EMT were detected (Figure 9a). However, YAP1 levels were rather unaffected after RASSF10 or ASPP2 knockdown. Subsequently, we have analyzed the expression levels of CTGF that is a downstream effector of TGFβ signaling involved in EMT [41]. We revealed that overexpression of RASSF10 and ASPP2 reduces levels of CTGF in HeLa cells (Figure 9b). These data indicated that loss of RASSF10 or ASPP2 induced constitutive activation of TGFβ signaling and EMT.

Discussion
In our study, we wanted to understand the mechanism of regulation of RASSF10 and its contribution to inhibition of growth, migration and invasion as a tumor suppressor ( Figure 1 and Table 2). TGFβ is well known for its inhibition of epithelial cell proliferation, however, during tumor progression cells evade the antitumoral TGFβ effect and TGFβ becomes oncogenic in late stage tumors by activation of EMT [32,42,43]. TGFβ signaling transmits an extracellular signal into a cellular/nuclear signal, with phosphorylation and nuclear translocation of SMADs followed by altered gene transcription (also SMAD independent effect) [44,45]. Here, we report that the RASSF10

Discussion
In our study, we wanted to understand the mechanism of regulation of RASSF10 and its contribution to inhibition of growth, migration and invasion as a tumor suppressor ( Figure 1 and Table 2). TGFβ is well known for its inhibition of epithelial cell proliferation, however, during tumor progression cells evade the antitumoral TGFβ effect and TGFβ becomes oncogenic in late stage tumors by activation of EMT [32,42,43]. TGFβ signaling transmits an extracellular signal into a cellular/nuclear signal, with phosphorylation and nuclear translocation of SMADs followed by altered gene transcription (also SMAD independent effect) [44,45]. Here, we report that the RASSF10 tumor-suppressor is upregulated by TGFβ stimulation in epithelial cells (as well as by cellular density) and associated with a G1 cell cycle arrest restricting cell growth ( Figure 2). However, the RASSF10 promoter is not a directed SMAD target, since ChIP-sequencing (chromatin immune-precipitation) data indicate that SMAD2 and SMAD3 are not detected at the proximal RASSF10 locus in several cell lines (Supplementary Figure S5). This suggests that TGFβ induced RASSF10 expression is a delayed gene response involved in downregulation of overshooting TGFβ signaling and regulation of the dual nature of TGFβ [46]. Loss of RASSF10 altered the TGFβ gene expression profile and induced expression of COL5A1, MMP2, SNAI2 and SPOCK1 (Figure 3a) and constitutive SMAD2 phosphorylation (Figure 9a), suggesting that RASSF10 plays a suppressive role in TGFβ signaling and EMT. SNAI2 and SPOCK1 are both factors that are known to trigger TGFβ induced EMT [4,47]. Additionally, we observed that depletion of RASSF10 significantly promotes TGFβ induced invasion of A549 cells in an extracellular matrix ( Table 2). We presume that, during early TGFβ signaling expression of RASSF10 acts as a transmitter of antiproliferative TGFβ effects. Interestingly, we observed that RASSF10 induced E-cadherin (CDH1) levels and loss of RASSF10 expression reduced CDH1 levels ( Figure 3). CDH1 is essential for the maintenance and homeostasis of polarized epithelial monolayers [48] and its transcription is repressed through the TGFβ induced expression of SNAI2 [3]. We revealed, that RASSF10 expression positively correlated with the expression of CDH1 (Figure 3g) and GO terms associated with the plasma membrane, cell surface receptor signaling and epithelial/epidermis cell differentiation ( Table 1). The role of RASSF10 was further supported by finding its interaction partner ASPP2 (Table 3 and Figure 4) that also controls epithelial plasticity through regulation of CDH1 [26]. The strength of the ASPP-RASSF10 interaction was further emphasized by the fact that known ASPP2 binding factor YAP1 [49], p53 and p65 overexpression [50] failed to compete with the co-precipitation of ASPP with RASSF10 ( Figure S3d). There are several data supporting the ASPP2 role at the cell periphery through the cell polarity complex PAR [51][52][53]. Our findings regarding the cellular role and localization of RASSF10-ASPP2 are further supported by interactome analysis, in which ASPP2 and RASSF10 are placed together in a large interaction module that provides links to cell polarity [54]. Additionally, ASPP2 forms an apical-lateral polarity complex at the level of tight junctions in polarized epithelial cells, acting as a scaffold for, e.g., protein phosphatase 1 (PP1) and junctional YAP [55] and ASPP2 is also said to regulate epithelial plasticity through CDH1 and β-Catenin regulation [26]. Our data indicate that knockdown of ASPP2 induces β-Catenin levels and phospho-SMAD2 under TGFβ treatment (Figure 9a). Previously it has been reported that ASPP2 suppresses TGFβ induced EMT by inhibiting Smad7 degradation [27]. SMAD7 prevents SMAD2/3 phosphorylation and is an important antagonist of TGFβ signaling [56]. In summary, we presume that RASSF10 induces ASPP2 and ASPP2 inhibits degradation of SMAD7 that counteracts TGFβ signaling and EMT. Moreover, RASSF10 and ASPP2 overexpression reduced levels of the matricellular protein CTGF (CCN2). CTGF has an important growth promoting role in cancer and is also involved in angiogenesis, cell adhesion and migration [57]. We found that knockdown and knockout of RASSF10 increased mitosis and increased cell proliferation (Figure 1) and significantly promotes TGFβ induced ECM invasion (Table 2), as well as RASSF10 re-expression halted proliferation (Figure 1) [19,22]. Additionally, RASSF10 was found at centrosomes/microtubules during mitosis [58]. In accordance ASPP2 was said to be involved in centrosome linker assembly at the end of mitosis [59]. As RASSF10 and ASPP2 [26,[51][52][53][54][55] are both linked to the cell polarity network, their presence would negatively regulate the mitotic/proliferative potential. Loss of the RASSF10-ASPP2 complex would lead to disturbance of cell polarity and thereby would interfere with the coordination of mitosis, as it is known that the positioning of the spindle apparatus is coordinated with polarity signals at the cell cortex [60]. The disruption of cell polarity itself is regarded as a central hallmark of cancer [61].
In our study we found a significant epigenetic silencing of RASSF10, but not ASPP2 in different tumor entities (Figures 6 and 7). Promoter hypermethylation of tumor suppressor genes is an established mechanism of their silencing in carcinogenesis [62]. RASSF10 methylation increased from primary tumors to cancer cell lines (Figure 6), consistent with the progressive hypermethylation of tumor-suppressors during tumorigenesis [63,64]. In our previous work, we reported that the pharmacological inhibition (e.g., 5-Aza-2'-deoxycytidine) of DNA methylation restored RASSF10 expression in different cancer entities [17,19,22,65]. We observed that methylation levels of RASSF10 correlated significantly with its reduced expression in cancer (Figure 7e). Our clinical data set revealed that hypermethylation of the CpG island of RASSF10 is a common and general event in human tumorigenesis (Figures 7 and 8). We could broadly show that in independent data sets loss of RASSF10 correlated not only with reduced patient survival rates in various tumor types (kidney cancer, thymoma, lymphoma, breast cancer, colon carcinoma, head and neck cancer, liver cancer, lung cancer and gastric cancer), but also with tumor stage/grade (thymoma, breast cancer) and tumor types (kidney cancer, melanoma and colon carcinoma) (Figure 8 and Table S2). Further analyses will become possible, when newer data sets are available in which RASSF10 is now commonly integrated. Our present comprehensive work is the finalization of our previous research in smaller data sets in which we have shown that the promoter of RASSF10 is methylated in patient tumors samples of the adrenal gland [21], head and neck [19], sarcoma [19], pancreas carcinoma [19] and Merkel cell carcinoma [20]. We showed the epigenetic inactivation of RASSF10 in thyroid cancer [17], lung cancer [19], skin cancer [65], breast cancer [22] and showed that RASSF10 inhibited growth of breast cancer [22], pancreas carcinoma and sarcoma cell lines [19]. Here, we now confirmed the prognostic and diagnostic value of RASSF10 across tumor types ( Figure 8). As RASSF10 DNA promoter methylation correlates with its RNA expression across cancers (Figure 7e), we are confident that measuring RASSF10 methylation represents its expression profile. It is obvious to measure methylated tumor DNA instead of RNA due to its superior stability in cells and body fluids (circulating DNA). Furthermore, RASSF10 RNA levels vary in tissues, and determining the according threshold level of inactivation would have to be determined for each tissue or cancer type. However, in DNA, we observed that already low levels of methylation inactivated RASSF10 and therefore a common threshold level could abrogate RASSF10 expression irrespective of the tissue. Tumor DNA could be obtained from tumor resections/biopsies or also non invasive by liquid biopsies [66], where circulating tumor cells (CTC) or circulating tumor DNA (ctDNA) are present in blood/body fluids or even from exhaled breath condensates (EBC) [67]. It was reported that in ex-smokers hypermethylation of RASSF1A could be measured in EBC [68] and EBCs are in clinical trials [69]. Liquid biopsies are FDA approved for lung cancer EGFR mutation tests as companion diagnostic [70]. Bisulfite treatment of DNA still is the gold standard to analyze DNA methylation [71], high-throughput bisulfite conversion is available [72] and digital droplet PCR (ddPCR) even amplifies low levels of nucleic acid in disproportionate sample/target combinations [73]. RASSF10 methylation analysis should be evaluated for its integration in present cancer screens.

Cell Culture and Treatment of Cell Lines
Cell lines were grown in appropriate medium (DMEM/Dulbecco's Modified Eagle's Medium, RPMI/Roswell Park Memorial Institute) supplemented with 10% FCS/fetal calf serum, 1% Penicillin/Streptomycin under cell culture conditions (37 • C, 5% CO 2 ). Cell lines were transfected for indicated time points using Polyethylenimmin (PEI, 4,9mM, Sigma, St. Louis, MO, USA) for HEK/HeLa, Turbofect (Thermo Fisher Scientific) or X-tremeGENE HP (Roche, Mannheim, Germany) for A549 according to manufacturer's protocol. Doxycycline (Dox, Thermo Fisher Scientific) was dissolved in water and used for RASSF10 induction in HEK-cells within the TetOn-TREx-system at working concentration of 2 µg/mL for 48 h [74]. SiRNAs (RASSF10 and ASPP2) were purchased from Dharmacon (Lafayette, CO, USA) On-target plus siRNA pool of four and was transfected using Lipofectamine RNAiMAX (Thermo Fisher Scientific). TGFβ (pH3 Citrate buffer) was used at 10 ng/ml. For flow cytometry assay overnight ethanol fixed cells were RNase A (50 µg/mL, 30 min, 37 • C) treated, stained with 50 µg/mL propidium iodide prior to measurement of DNA content in FACSCantoII (BD Biosciences). FACSDiva Software (BD Biosciences) was used for measurement and gating to distinguish cell cycle phases.

Generation of Stable RASSF10 Inducible HEK-Cells
HEK-cells stably expressing the Tet Repressor under Blasticidine (5 µg/mL, Roth, Karlsruhe, Germany) were obtained from Thermo Fisher Scientific as part of the TetOn-TREx-System and were used as control cell line. These cells were transfected with the cloned RASSF10-pcDNA4ToMyc for stable insertion of Dox-inducible-RASSF10. Cells were selected for RASSF10 using Zeocin (500 µg/mL, Thermo Fisher Scientific). Three clones with stable Dox-inducible-RASSF10 were used in our experiments.

DNA Isolation and Methylation Analysis
The promoter region of RASSF10 and ASPP2 was analyzed by CpG plot (EMBL-EBI, Hinxton, United Kingdom) and both show large CpG islands. Primers for bisulfite treated DNA were designed to bind only fully converted DNA and amplify promoter region (Table S3). The precise promoter region was chosen for CpG content and presence of according restriction enzymes for COBRA analysis ( Figure 6). DNA was isolated after proteinase K (Thermo Fisher Scientific) digest and extracted either with phenol/chloroform or by QIAamp DNA extraction kit (Qiagen, Hilden, Germany), and concentrations were determined by UV-photospectrometery. For COBRA methylation analysis 2 µg genomic DNA was bisulfite treated (5 mM hydroquinone, 1.65 M sodium metabisulfite and pH 5.5 with 0.025M NaOH) and incubated over night at 50 • C. DNA was purified using MSB Spin PCRapace (STRATEC Molecular, Birkenfeld, Germany), eluted in 50 µL H 2 O and followed by 10 min incubation with 5 µL 3 M NaOH at 37 • C. DNA was then precipitated with 100% ethanol and 7.5 M ammonium acetate and resolved in 1 × TE buffer. The subsequent PCR product (COBRA primers) was digested with 0.5 µl of TaqI (Thermo Fisher Scientific) 1 h at 65 • C and resolved on 2% TBE gel together with mock control and DNA ladder. The COBRA product for RASSF10 is 167 bp (TaqI site at 66 bp) for semi-nested PCR and the product for ASPP2 is 117 bp (TaqI site at 50) for nested PCR. In vitro methylation of genomic DNA was performed using CpG Methyltransferase M.SssI (NEB, Frankfurt, Germany) according to manufacturer's protocol. For Primers see Supplementary Table S3. For further details on COBRA analysis see Richter et al. [22].

CRISPR/Cas9
To genomically delete RASSF10 we performed CRISPR/Cas9 targeted knockout. CRISPR/Cas9 vectors were obtained from Lienhard Schmitz (Giessen, Germany) and RASSF10 targeting oligos/guideRNAs were generated according to protocol [77]. RASSF10 knockout oligos (two combinations) were created to delete a region 3 prime of the translational start site, creating a frameshift and RASSF10 loss in p × 549 with wt Cas9. We transfected A549-cells with the CRISPR/Cas9 RASSF10 oligos and selected for positive clones by puromycin (1 µg/mL) for three days. Clones were expanded and the knockout was verified by PCR based amplification of the RASSF10 genomic region and RNA showing the deletion and was further verified by Western blot.

Invasion Assay
For the evaluation of the invasion capacity of A549 wildtype and RASSF10 deletion clones transwell extracellular matrix (ECM) invasion assays were performed. In 24-well plate a volume of 750 µL of DMEM with 10% FCS and 5 ng/mL TGFβ (or mock) was distributed into the wells. Filter inserts (8.0 µm pore size, Falcon cell culture insert, transparent PET membrane, Corning, Kaiserslautern, Germany) were coated with 100 µL ECM (0.25 mg/mL Matrigel Matrix, Corning). 5 × 10 4 cells were seeded in 200 µL DMEM with 5 ng/mL TGFβ (or mock) and incubated at 37 • C with 5% CO 2 for 16 h. Cells were fixed with 3.7% formaldehyde and 100% methanol for 2 min. Finally, cells were stained with 10% Giemsa for 15 min, washed with PBS and ECM was removed with a cotton swab. Bottom side of membranes were analyzed by microscopy (Motic AE21, Wetzlar, Germany) and all invasive cells were counted.

Conclusions
Cancer incidences are still growing and thereby the need for precision medicine and novel targeted therapies is rising. Our work directly serves this demand to identify for novel cancer targets. We are now showing that RASSF10 expression inhibits signs of EMT in cancer cell lines, but we are also reporting its mode of action as a tumor suppressor. RASSF10 is regulated by TGFβ signaling and promoter hypermethylation and RASSF10 itself modulated TGFβ signaling through its regulation of epithelial cadherin. Depletion of RASSF10 promotes TGFβ induced ECM invasion and constitutive SMAD2 phosphorylation. We identified ASPP2 (apoptosis-stimulating protein of p53 2) as the main binding target of RASSF10 by interactome analysis, which depended on the RASSF10 coiled-coils domain. We show that RASSF10 induces the protein levels of ASPP2 that is also an inhibitor of EMT. In human cancers, RASSF10, but not ASPP2, is epigenetically inactivated by promoter hypermethylation. In independent datasets we could validate, that loss of RASSF10 expression clinically correlated with decreased survival and with progressed disease state of cancer patients. Thus, loss of RASSF10 expression by promoter hypermethylation can serve as a diagnostic and prognostic cancer marker.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A. Additional Information on Datasets and Software Used in this Study
The information is listed in order of its appearance. Analysis performed using the R2 [28] Genomics Analysis and Visualization Platform.   [79][80][81][82]. Thymoma, Head-neck squamous cell carcinoma; mRNA Affymetrix ID 238755_at: Liver cancer, lung cancer, gastric cancer, Split patients by: Auto select best cutoff, Follow up threshold: 60 months. Analysis performed using MethSurv [83].