SETD3 acts as a prognostic marker in breast cancer patients and modulates the viability and invasion of breast cancer cells

In several carcinomas, the SET Domain Containing 3, Actin Histidine Methyltransferase (SETD3) is associated with oncogenesis. However, there is little knowledge about the role of SETD3 in the progression and prognosis of breast cancer. In this study, we first analyzed the prognostic value of SETD3 in breast cancer patients using the database of the public Kaplan-Meier plotter. Moreover, in vitro assays were performed to assess the role of SETD3 in the viability and capacity of invasion of human breast cancer cell lines. We observed that the high expression of SETD3 was associated with better relapse-free survival (RFS) of the whole collective of 3,951 patients, of Estrogen Receptor-positive, and of Luminal A-type breast cancer patients. However, in patients lacking expression of estrogen-, progesterone- and HER2-receptor, and those affected by a p53-mutation, SETD3 was associated with poor RFS. In vitro analysis showed that SETD3 siRNA depletion affects the viability of triple-negative cells as well as the cytoskeletal function and capacity of invasion of highly invasive MDA-MB-231 cells. Interestingly, SETD3 regulates the expression of other genes associated with cancer such as β-actin, FOXM1, FBXW7, Fascin, eNOS, and MMP-2. Our study suggests that SETD3 expression can act as a subtype-specific biomarker for breast cancer progression and prognosis.

Breast cancer has a large global impact, accounting for 15% of cancer deaths in women worldwide 1 . Although newly introduced molecularly targeted therapies such as PARP, PI3K/AKT/mTOR and other kinase inhibitors as well as immunotherapy allow for better treatment of breast cancer, much remains to be understood 2 . Not only through the introduction of more specific morphologic subtypes, but also through a thorough molecular classification, breast cancer therapy has evolved towards targeted therapies with a more favorable side-effects profile 3,4 . This advance is largely exemplified by biomarkers such as estrogen receptor (ER) or human epidermal growth factor receptor 2 (HER2), which serve as decision support tools. Indeed, these biomarkers are excellent examples of functionally relevant prognostic markers that have been successfully utilized as therapeutic targets in breast cancer 5,6 .
Among potentially relevant novel biomarkers that can expand the diagnostic repertoire in malignant disease, factors modulating the cytoskeleton are prime candidates, as they may have a direct impact on cell motility and metastasis, which is the leading cause of death in cancer patients [7][8][9] . Novel evidence has highlighted the relevance for posttranslational methylation for cytoskeletal function 10 . Post-translational modifications such as methylation modulate the function of numerous eukaryotic proteins 11 . Most information in this field refers to the methylation of histones on arginine and lysine residues and helped to better understand the epigenetic control of gene expression 12 . However, it has been long known that non-histone proteins like beta-actin are methylated as well 13 . Methylation may, therefore, constitute a regulatory mechanism that does not only take place at the level of DNA expression. Recently, the SET Domain Containing 3, Actin Histidine Methyltransferase (SETD3) has been identified to be an actin-specific histidine N-methyltransferase 10,14 . SETD3 is a ubiquitously expressed protein, which is able to methylate the His73 position of actin in a physiologically relevant manner by using the co-factor S-adenosyl-methionine (SAM) as a methyl group donor 10 . Structural analysis has shown that actin methylation regulates the flexibility and stability of the interdomain of actin 15 . SETD3 does not affect the basal function of actin, as a knockout mouse model (SETD3 −/− ) is still viable. However, homozygous knockout females showed primary dystocia. These findings suggest that the phenotype of irregular uterine contractions may be caused by the lack of actin methylation 10 . Importantly, altered regulation of actin is an important step during pre-malignant stages, as actin is one of the key proteins driving cancer cell migration 16,17 . In order for cells to invade and metastasize they need to acquire the capacity to migrate 18 . Migratory single cells can show amoeboid or mesenchymal movement, whereas groups of cells can move in the form of collective migration. In both cases, the re-organization of actin is a critical step. 19 .
Besides regulating actin, SETD3 shows various additional protein interactions, while its role in cell signaling remains unclear. In breast cancer, the forkhead box protein M1 (FoxM1) has tumor-promoting functions impacting patient prognosis 20 . FoxM1 is also methylated by SETD3, claiming a core position in oncogenesis 21 . It was demonstrated that SETD3 is a factor with relevance to cell cycle regulation. It has an expression peak in the S-phase and underlies a regulatory degradation by F-box/WD repeat-containing protein 7 beta isoform (FBXM7-beta) and glycogen synthase kinase 3 beta isoform (GSK3-beta). Inhibition or depletion of these two factors results in elevated SETD3 levels and promotes cell proliferation. This concurs with elevated SETD3 levels which have been found in human liver cancer 22 . Supporting findings suggest that SETD3 is required for p53 target recruitment and activation 23 . This identifies several relevant targets; the phenotypical and physical properties due to changes in beta-actin methylation (yet characterized by primary dystocia) and signaling properties of SETD3 that may affect oncogenesis 10,21 . Based on this evidence, we hypothesize that alterations in SETD3 expression may affect cytoskeletal function, cancer cell motility and clinical outcome in breast cancer patients. As there are multiple regulatory mechanisms in which SETD3 is involved, it is expected to define specific subgroups that exhibit a strong effect of SETD3 in terms of prognosis.
In this study, we first used the Kaplan Meier Plotter resource 24 incorporating the gene expression data of over five thousand breast cancer tumor samples, to correlate SETD3 expression to the survival of breast patients stratified by different clinical subtypes. Then, we employed SETD3 siRNA knockdown in vitro in a range of well-established human mammary carcinoma cell lines to analyze the role of SETD3 depletion in cytoskeleton-associated functions with relevance to malignant cell behavior. Our data assign a prognostic value to altered SETD3 expression in breast cancer that depends on p53 mutation status and provides proof for a role of SETD3 in the viability, motility, cytoskeletal organization, and invasiveness of triple-negative breast cancer cells.

Results
Patients with triple-negative and p53 mutated tumors expressing high levels of SETD3 have a poor outcome. To have an overview of SETD3 expression in different solid tumors, we analyzed its expression in samples from the TCGA, which demonstrated that SETD3 is mainly expressed in renal and thyroid cancer (Fig. 1). However, in cancers associated with women, we observed higher expression of SETD3 in breast cancer compared to cervical, endometrial and ovarian cancer (Fig. 1).
In order to establish whether SETD3 expression could have an impact on breast cancer prognosis, we investigated its relation to relapse-free survival (RFS). Gene expression data from tumor samples of 3,951 patients without any stratification and patients classified according to different molecular subtypes and receptor status (Table 1) were analyzed using the online tool Kaplan Meier Plotter 24 . Table 1 shows the HR (Hazard ratio) and P values of SETD3 in each classification. We observed that over all samples, high SETD3 levels were a protective factor for patient survival with an HR = 0.8; P = 4.3e-05 ( Fig. 2A). This applied to ER-positive tumors only (HR = 0.85; P = 0.012) (Fig. 2B) and to the intrinsic subtype Luminal A (HR = 0.75; P = 0.0049) (Table 1). However, in patients with PR-negative tumors, high SETD3 expression correlated with a worse survival (HR = 1.4; P = 0.026) (Fig. 2C), while in the other groups associated with hormone receptor or Her2 expression, we did not find a correlation (Table 1). Of note, in the clinically highly relevant subgroup of triple-negative tumors, the expression of SETD3 was associated with poor relapse-free survival (HR = 1.72; P = 0.031) (Fig. 2D).
We next determined the correlation of SETD3 expression with nodal status (Table 1). We found that the HR of the expression of SETD3 had a high score but that it was not significantly related to the survival of the patients ( Table 1). The treatment decision and the prognosis of a breast cancer patient depend on many factors including the grade of the tumor associated with LN status 25 . Like in our analysis regarding nodal status, we did not find a correlation between SETD3 expression and the grade of the tumor (Table 1).
Since SETD3 is a positive regulator of p53 23 , we studied its expression in our patient collective stratified by p53 status. In patients with tumors in which p53 is mutated, the high expression of SETD3 correlated with inferior RFS (HR = 2.1; P = 0.0024) (Fig. 3A, left panel), but we did not find a correlation in the wild type (WT) p53 group (HR = 0.95; P = 0.82) (Fig. 3A, right panel and Table 1).Taking a closer look at the expression of SETD3 and p53 in patients with ER-, and PR-positive or negative tumors and the Luminal A and B classification, we observed that in patients with ER-and PR-positive and p53 mutated tumors, a high expression of SETD3 was associated with worse RFS (HR = 2.93; P = 0.0014, and HR = 11.23; P = 0.0039, respectively) ( Fig. 3B,C, left panel). Also, in p53 mutated Luminal B tumors, but not in Luminal A tumors, SETD3 correlated with shorter survival (HR = 2.42; P = 0.041) (Fig. 3D, left panel and Table 1). Again, we did not find any correlation of SETD3 with survival in p53 WT and ER-, PR-negative and Luminal A or Luminal B tumors ( Fig. 3B-D, right panel and Table 1). Our analysis results suggest that the expression of SETD3 influences the clinical outcome according to the expression of hormone receptors and p53 status. We can also conclude that the function of SETD3 depends on the presence of mutations and the expression patterns of other genes in breast cancer.
www.nature.com/scientificreports www.nature.com/scientificreports/ Expression of SETD3 in cell lines representing different molecular classifications of breast cancer. As a prerequisite for studying SETD3 function in vitro, we characterized its expression by quantitative  26 . Only MCF-7 cells showed a p53 WT status. SETD3 was expressed in all cells and no significant differences in the relative expression (represented by 2 -ΔCt ) were noted between the subtypes and within each subtype (Fig. 4A). Taking these results into account, we transfected only the Luminal cells T47D and MCF-7, and the basal cells MDA-MB-468 and MDA-MB-231 with a small interfering RNA (siRNA) to silence the expression of SETD3. Forty-eight hours after transfection, we observed that all cell lines downregulated the expression of SETD3 compared to control, almost 0.6 fold in T47D and 0.9 fold in the other cell lines (Fig. 4B). By performing western blot, we confirmed that in all cells SETD3 siRNA substantially decreased the levels of SETD3 protein (Fig. 4C), and was therefore found suitable to study the functional consequences of SETD3 depletion.

SETD3 differentially affects viability and invasiveness of mammary carcinoma cells in a subtype-specific manner.
After the initial characterisation of our experimental system, we aimed to uncover the possible participation of SETD3 in processes associated with tumor progression such as proliferation and invasion 27 . Following the downregulation of SETD3 expression, MTT assays revealed that the basal-type cell lines had the lowest viability: MDA-MB-231 showed a significant reduction by 50% compared to control siRNA-treated cells, and MDA-MB-468 displayed a reduction of viability by 35%. (Fig. 5A). In contrast, cell viability was not significantly altered in the Luminal cell lines (Fig. 5A). MDA-MB-231 is a highly aggressive breast cancer cell line according to its invasive and metastatic properties while MCF-7 is not 28 . Due to these properties, we used them as a model to study the role of SETD3 in invasion using the established Matrigel invasion chamber assay. As shown in Fig. 5B, SETD3 siRNA treated MDA-MB-231 cells reduced their invasive potential by 40% with respect to control. In contrast, SETD3 siRNA did not significantly reduce the invasion characteristics of MCF-7 cells. Because the reduction in SETD3 expression affected the invasion capacity only in MDA-MB-231 cells, we performed a collagen contraction assay in this cell line. The capacity to contract collagen I gels was diminished by SETD3 siRNA-treatment compared to control after 24 and 72 hrs in the triple-negative cell line (Fig. 5C), indicating a possible defect in cytoskeletal function. As SETD3 is involved in cytoskeletal rearrangement 14 , which is implicated in migration and invasion 17 , we performed immunofluorescence microscopy employing the fluorescently labeled actin-binding protein phalloidin to study the consequence of SETD3 knockdown on cell morphology. In agreement with the invasion and collagen contraction assays, both cell morphology and the structure of actin filaments were altered when SETD3 was knocked down (Fig. 5D). While the control cells had an elongated, mesenchymal-like morphology typical of the invasive cell line, a proportion of the SETD3-depleted cells showed a round morphology (Fig. 5D). Overall, these results demonstrate a role for SETD3 in cytoskeletal organization and function of aggressive breast cancer cells, which correlates with our observation that the expression of SETD3 in triple negative tumors was related to poor prognosis (Fig. 2D). SETD3 regulates the expression of different effector genes with relevance to cancer progression. It is known that SETD3 regulates the expression of genes related to normal physiology, development and cancer progression such as FOXM1 (Forkhead Box M1), ACTB (β-actin), ASMA (alpha-smooth muscle actin), ACTG (λ-actin), FSCN (Fascin Actin-Bundling Protein), FBXW7 (F-Box and WD Repeat Domain Containing 7) 10,22,29-32 but it is not yet known whether this regulation also occurs in breast cancer. To find out if SETD3 influences the expression of the genes listed above in our in vitro system, we transfected SETD3 siRNA into the basal cells MDA-MB-231 and MDA-MB-468, and Luminal MCF-7 and T47D cells. Our results show that in all SETD3 siRNA treated cell lines, some genes are significantly up-or down-regulated, however, this regulation occurred in a cell-type-specific manner, and was quantitatively modest in most cases (Fig. 6). In MDA-MB-231 cells, Gamma-actin was up-regulated 0.4 fold, while Fascin was 0.2 fold up-regulated, and FOXM1 was downregulated 0.1 fold (Fig. 6A). In MDA-MB-468 cells FOXM1 and Gamma-actin were downregulated 0.1 and 0.2 fold respectively while Fascin was up-regulated 0.3 fold (Fig. 6B). With respect to Luminal cells, MCF-7 FOXM1 was downregulated 0.5 fold and ASMA was upregulated 0.84 fold (Fig. 6C), and in T47D cells Fascin was upregulated 0.9 fold, whereas Gamma-actin was up-regulated 0.1 fold (Fig. 6D). Also, to analyze the gene expression of additional molecules regulated by SETD3 10,22,29-32 that could potentially explain the differences between the prognostic impact on diverse breast cancer subtypes, we also quantified MMP-2, KLC4, iNOS and eNOS mRNA levels in MCF-7 and MDA-MB-231 cells. We found that MMP-2 was up-regulated in the absence of SETD3 in both cell lines, however, this regulation was only significant in the luminal MCF-7 cell line (Fig. 6E,F). In addition, eNOS was up-regulated only in MCF-7 (Fig. 6E). According to our data, we can suggest that FOXM1 is the most homogeneously regulated gene affected by SETD3 regardless of the aggressiveness or classification of the cell lines and that numerous genes are affected by SETD3 knockdown in a cell-type-specific manner.
Functional enrichment analysis for SETD3 related proteins. To find functional interaction networks of SETD3 and its potential effector genes such as FOXM1, FBXW7, ACTB (β-actin), Fascin, MMP-2, KLC4, eNOS, and iNOS, we used the online bioinformatic tool STRING 33 . The STRING analysis showed that SETD3 has close interactions only with FOXM1, while ACTB has interactions mainly with PFN1 and CFL1, but also has close www.nature.com/scientificreports www.nature.com/scientificreports/ interactions with Fascin 1-3 and is interconnected with NOS2 (iNOS) and NOS3 (eNOS). The rest of the molecules are also interconnected. FOXM1 is highly interconnected with FBXW7, AKT1, and MMP2. On the other hand, FBXW7 is associated with SKP1 and AKT1. Finally, KLC4 is not related to any of the molecules (Fig. 7A). Figure 7B and Table S1 show the molecular function, the cellular component and the biological process of the analyzed proteins. Interestingly, we identified enriched pathways linked to cancer such as VEGF, proteoglycans, Rap1, estrogen, and HIF-1, parasite and bacterial infection, in other pathological processes such as diabetes and atherosclerosis, and normal processes including arginine biosynthesis ( Fig. 7C and Table S2). Table S3 shows the 10 most significantly enriched terms based on PubMed co-citation analysis 33 , which also includes several terms linked to cancer.
The expression of FOXM1 has a negative impact on the survival of breast cancer patients and does not depend on nodal status and tumor grade. SETD3 regulates the expression of and has very close interactions with FOXM1 according to our STRING analysis, and simultaneous expression of SETD3 and FOXM1 was linked to tumorigenicity in an in vivo model of liver cancer 22 . Our data indicate that the expression of FOXM1 negatively influences the relapse-free survival of the whole study collective (HR = 1.67; P = <1E-16) as well as in ER-positive (HR = 1.81; P = <1E-16) (Supplementary Fig. 2A, left panel), PR-positive (HR = 2.66; P = 1.3e-07) and Her2-negative tumors (HR = 1.81; P = 1.1e-05) ( Table 2). When we analyzed the expression of FOXM1 in patients based on the intrinsic subtypes Luminal A and B, we observed a poor relapse-free survival with an HR = 1.81; P = 8.5e-13 in the Luminal A group (Supplementary Fig. 2A, right panel) and a HR = 1.36; P = 0.0017 in the Luminal B group (Table 2). Interestingly, the expression of FOXM1 correlated with reduced RFS of breast cancer patients regardless of their nodal status and tumor grade (except for Grade III) (LN-positive: HR = 1.74, P = 3.9e-08; LN-negative: HR = 1.7, P = 9.1e-10; Grade I: HR = 3.31, P = 3.7e-05; Grade II: HR = 1.55, P = 0.00034) ( Table 2). In contrast, in patients with estrogen receptor-negative tumors and in the molecular subtype Her2, higher FOXM1 was associated with better survival (HR = 0.75; P = 0.0087; HR = 0.63; P = 0.021, respectively). Although the correlation of this gene with p53 mutated tumors was negative, in p53 WT cancers, high FOXM1 expression correlated with worse RFS (HR = 1.66; P = 0.019) ( Table 2). Taking into account that SETD3 regulates the expression of FOXM1, we analyzed the co-expression of SETD3/FOXM1 only in ER-positive and Luminal A tumors where SETD3 expression correlated with better RFS (Fig. 2B), with the www.nature.com/scientificreports www.nature.com/scientificreports/ purpose of investigating if the interaction of both genes affects the survival of patients. We observed that the co-expression of both genes correlates with worse survival in ER-positive tumors (HR = 1.22; P = 0.028) while in the Luminal A subgroup, we did not find a significant correlation (HR = 1.02; P = 0.84) ( Supplementary Fig. 2B, left and right panel). We conclude that although we observe that SETD3 regulates FOXM1 in cell lines regardless www.nature.com/scientificreports www.nature.com/scientificreports/ of their classification, the expression of FOXM1 in tumors is sufficient to confer worse patient survival even when SETD3 (which is associated with better survival) is co-expressed. We suggest that within a tumor the regulation of FOXM1 may not only be carried out by SETD3 but also through other molecules.
We also analyzed the impact of FBXW7 and Fascin on patient survival, as regulation of these genes has been linked to SETD3 22   www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ (HR = 1.6; P = 0.034) (Table S4). High Fascin levels correlated with poor survival in Her-2 negative and Grade III tumors with a HR = 1.37; P = 0.018; HR = 1.34; P = 0.0091, respectively (Table S5). The other analyzed groups had no association with RFS.
Our investigation suggests that high tumoral levels of SETD3 are associated with a better outcome of breast cancer patients especially in ER-positive tumors. Moreover, in PR-negative and in triple-negative tumors a high expression of SETD3 results in reduced survival. Interestingly, in the presence of p53 mutations, the higher expression of SETD3 correlates with worse relapse-free survival even in ER-positive tumors. In addition, the HR of PR-negative tumors is worse when p53 is mutated and shows a high expression of SETD3. Finally, we observed that SETD3 regulates the invasive potential and the cytoskeleton of highly aggressive mammary cancer cells.

Materials and Methods
Kaplan-Meier plots and TCGA expression data. To assess the prognostic value of SETD3 and further functionally linked genes, the online tool kmplot.com 24 , which allows a meta-analysis of gene expression in relation to breast cancer patient survival, was employed. Gene expression data was obtained through microarray analysis of widely used arrays of the GEO database and converted into Kaplan-Meier plots 34 . This study includes relapse-free survival (n = 3,951) and overall survival (n = 1,402). The package "survival" was used in the R programming environment to plot Kaplan-Meier survival curves and compute the number-at-risk 24 . To distinguish between high and low expression, the median was selected as cut off-value to reduce the impact of outliers and produce equal numbers in both categories to only show strong correlations. Also, the JetSet probe set was selected to acquire unambiguous expression estimates 35 and redundant samples were removed to enhance the quality of the sample. Patients were stratified by ER status, HER2 status, lymph node status, p53 status, molecular classification, Pietenpol subtype, and grading. We analyzed the expression of SETD3, FOXM1, FBXW7, and Fascin, and visualized the correlation to survival by drawing Kaplan-Meier survival plots. The Affymetrix IDs are 212465_at (SETD3), 202580_x_at (FOXM1), 218751_s_at (FBXW7), and 201564_s_at (FSCN1). The shown hazard ratios are not inverted (HR < 1 favorable). Gene expression data as plotted in Fig. 1 were derived from The Pathology Atlas section 34 of the Human Protein Atlas (www.proteinatlas.org/) 35   Western blot. Total cell lysates were prepared with protein Sample Buffer Laemmli 2×. 15 µl of protein per lane were separated on 10% SDS gels, and electro-transferred to nitrocellulose membranes (Amersham, Pharmacia Biotech, Piscataway, NJ, USA, cat. No. 10600008). Following 1 hour of blocking with 5% (w/v) skimmed milk powder in 1X TBST buffer, detection was performed with rabbit polyclonal primary antibody against SETD3 (1:1000, Sigma-Aldrich, cat no. HPA003639), and goat anti-rabbit IgG, H & L Chain Specific Peroxidase Conjugate (Sigma-Aldrich, cat. No. 401353). Then, membranes were subjected to a chemiluminescence reaction and signal quantification with NIH ImageJ software, normalizing the values of SETD3 to tubulin (1:10000, Sigma cat no. T5168; as a loading control). For tubulin detection, membranes were stripped with 0.2 M glycine buffer (pH 2.5), washed, re-incubated with primary antibody for tubulin as a housekeeping protein and subjected to the procedure described above. Collagen contraction assay. SETD3-and control siRNA-transfected MDA-MB-231 cells were harvested and re-suspended in DMEM media at 3 × 10 6 cells/ml. Cell-embedded collagen gels were prepared on ice by neutralizing and diluting Type-I collagen solution (4.88 mg/ml; Corning, cat. No. 8274002, NY, USA) with NaOH, nuclease-free water, and 10X PBS into 2.5 mg/ml then cell suspension was immediately mixed at 2:8 ratio to collagen solution. The cell collagen mixture (500 µL/well) was immediately pipetted in a 24-well plate and placed in a culture incubator to polymerize at 37 °C for 60 minutes, then cell-seeded collagen gels were carefully detached from the well to remain free-floating in the culture medium. Subsequently, 1 mL of DMEM media with FCS was added to each well and incubated at 37 °C, 7.5% CO 2 humidified atmosphere. The medium was changed every two days up to the decided time points. The collagen gel contraction progress was monitored by measuring the surface area of the gels every day 36 . The 24-well plate and cell-embedded gel were photographed after 24 hours and 72 hours using a digital camera with a resolution of 2160 × 1080 pixels and an Axiovert 100 microscope (Carl Zeiss, Jena, Germany). The digital images of the surface were analyzed and calculated using the NIH ImageJ program (U.S. National Institutes of Health, Bethesda, Maryland, USA). The contracted area was presented as a percentage of the initial surface area. Protein interaction network analysis. STRING v11 (http://string-db.org/) 33 was used to generate in silico protein interaction networks for the gene products that we analyzed in KMplot, SETD3, FOXM1, FBXW7, ACTB, FSCN1, MMP-2, KLC4, iNOS, and eNOS. All interactions are predicted with a high confidence threshold of 0.700, and all active predictive methods were allowed. For the enrichment analysis, STRING implements well-known classification systems such as Gene Ontology (GO) and KEGG (Kyoto Encyclopedia of Genes and Genomes) 33 .

Statistical analyses.
For survival analysis, in the R statistical environment, we utilized the Kaplan-Meier-Plotter database via the statistical package "survival" to calculate Kaplan-Meier survival curves and the number-at-risk. Furthermore, the hazard ratio (and 95% confidence intervals) and log-rank P were calculated for each gene 24 . For experimental assays, statistical analysis was performed with GraphPad Prism 4.02 (GraphPad Software, La Jolla, CA). GraphPad Prism 4.02 was used to perform two-tailed t-tests, one-way ANOVA with Dunn´s posttest, or nonparametric Friedman's test with Dunn's posttest, where appropriate. Significant P-values are indicated as follows: ≤0.05 by one asterisk *, ≤0.01 by two asterisks ** and ≤0.001 by three asterisks ***.

Discussion
One of the most studied epigenetic modifications is the methylation of DNA. Under physiological conditions, it contributes to the regulation of the expression of genes that must be expressed in a defined physiological state, while silencing those that should not 37 . Since DNA modification is a pivotal player in transcriptional regulation, defects in this mechanism may lead to tumorigenesis and cancer progression 38 . Already in pre-invasive (2020) 10:2262 | https://doi.org/10.1038/s41598-020-59057-5 www.nature.com/scientificreports www.nature.com/scientificreports/ stages of cancer, hypermethylation leads to the abnormal silencing of some genes such as p16 and glutathione S-transferase-π1 (GSTP1), allowing neoplastic cells to progress towards more aggressive stages 39 . Notably, the methylation of actin plays an important role in the remodeling of the cytoskeleton which occurs during development and steady-state physiology, and in processes associated with cancer like invasion and metastasis, promoted by epithelial to mesenchymal transition (EMT) 17,40,41 . Also, the methylation of actin on His73 by SETD3, a novel methyltransferase, prevents primary dystocia 10 . Interestingly, the human methyltransferasome includes 208 known members, of which 22 SETD proteins have been associated with different diseases including cancer in both humans and animal models 42,43 . Previously, the expression of SETD3 has been associated with oncogenesis of lymphoma and liver cancer 44 . In breast cancer, however, our study indicates a bivalent role of SETD3.
In the present study, we found that the expression of SETD3 was associated with better relapse-free survival in the patients in general without any classification, and in patients with ER-positive and Luminal A-type tumors. In contrast, in PR-negative and triple-negative tumors (ER-, PR-and HER2-negative), SETD3 is associated with poor RFS. Most importantly, in breast cancer patients with mutated p53, the expression of SETD3 is linked to worse prognosis even in patients with ER-positive tumors. Although it has been demonstrated that SETD3 is required for p53 target gene recruitment and the induction of apoptosis in response to DNA damage in cancer colon cells 23 , in our study we observed that downregulation of SETD3 is associated with reduced viability of aggressive breast cancer cells. In agreement with that, in hepatocellular carcinoma cells, the knockdown of SETD3 markedly reduced cell proliferation 22 and in consequence resulted in the formation of small tumors in vivo. Interestingly, in the same study, the expression of SETD3 was higher in specimens of liver cancer patients with grade III 22 . We found this information important because in breast cancer as well as other types of cancer the grade is associated with metastasis to lymph nodes 45 . However, while we obtained a high HR score regarding the expression of SETD3 and grade 3, we found no significant difference (Table 1). It seems that SETD3 has different roles in cancer depending on other mutations and the stage of tumor progression 22,23,32 . This is not surprising since it has been reported that other molecules have dual roles and can act as anti-tumoral and/or pro-tumoral factors. For example, the bone morphogenetic proteins (BMPs) that are part of the transforming growth factor-β (TGF-β) superfamily have been related to the induction of angiogenesis, EMT, cancer stem cells (CSCs) and metastasis in aggressive breast, skin, and prostate cancer cells lines and tissues 46 . On the other hand, BMP10 which is part of the BMP family suppresses the aggressiveness of prostate cancer cells by the induction of apoptosis 47 and in patients with low expression of BMPRIB exhibited poor prognosis, especially in the luminal subtype 48 . Also, the same group reported that patients that were treated with taxane and anthracycline regimens and have high expression of BMPRIB have a favorable prognosis 48 . Another clear example is TGF-β, which has a tumor-suppressive effect in non-tumorous cells and pre-malignant stages, while it facilitates tumor metastasis in late-stage cancer cells 49 .
An important result of our in vitro analysis is that SETD3 siRNA treated MDA-MB-231 cells, representing the basal subtype, were 35% less invasive than controls, whereas no significant impact on invasiveness was noted for SETD3 knockdown in the luminal cell line MCF-7 as the MDA-MB-231 cell line displays a highly invasive phenotype 50 , this could be explained by the already described effects of SETD3 on actin polymerization and cell metabolism 10 . Apparently the cytoskeleton of the highly metastatic basal MDA-MB-231 cells was affected in a stronger manner as the less dedifferentiated luminal cell line MCF-7. The cell-type-specific alterations in cytoskeletal gene expression (Fascin, ACTG2, ASMA) may be indicative of a differential effect on the cytoskeleton (Fig. 6). Indeed, numerous studies have demonstrated that the actin cytoskeleton of basal and luminal breast cancer cells shows structural and functional differences that result e.g. in an increased actin response in the interaction of breast cancer cells with natural killer cells 51 , in differential nuclear shape dynamics 52 , in the suppression of invasion 53 , and in cell fusion and the formation of cancer hybrid cells 54 . Other SETD proteins such as SETD5 have been associated with the invasion and migration of non-small cell lung cancer cells 55 . Interestingly, the authors found that SETD5 regulates the expression of master EMT transcription factors like Snail and ZO-1 55,56 . They also found that higher SETD5 expression was significantly correlated with advanced TNM stage, lymph node metastasis, and overall survival rate. The invasion process is characterized by a reorganization of the actin cytoskeleton, and the formation of membrane protrusions is a requirement for invasive growth 57 . Our morphological analysis indicates that MDA-MB-231 SETD3 depleted cells undergo changes in their cytoskeleton arrangement and lose their large membrane protrusions. This might be the result of a higher frequency of actin polymerization and destabilization of the actin cytoskeleton through hypomethylation 10 and could explain their less invasive capacity. Our finding of a substantially reduced capability of SETD3-depleted MDA-MB-231 cells to contract collagen gels supports this observation at the functional level, as it is indicative of impaired cytoskeletal function.
According to our and other studies, it seems that SETD3 significantly regulates mechanisms associated with actin regulation, however, the protein network in which SETD3 is involved is unknown. Cohn et al. demonstrated that SETD3 can interact with 172 novel proteins such as transcription factors of which FoxM1 was one of them 21 . Also, SETD3 methylates FoxM1 and negatively regulates the expression of Vascular Endothelial Growth Factor (VEGF) under normoxia conditions 21 , a known factor related to angiogenesis and tumor progression 58 . On the other hand, FoxM1 has an important role in cancer because its overexpression has been found to be related to poor prognosis in patients 29 . In our study, we found that the expression of FoxM1 also correlated with worse RFS in breast cancer patients (Table 2). When we analyzed the expression of both genes at the same time in the patients classified as ER-positive and Luminal A, we observed worse survival. However, the real-time gene expression analysis revealed that SETD3 knockdown influences the expression of FoxM1 which was downregulated (Fig. 6). Furthermore, the STRING analysis showed that SETD3 has a very close interaction with FoxM1 (Fig. 7). Our study could, therefore, confirm the observation of Cohn et al. that SETD3 regulates the expression of FoxM1 21 . Nevertheless, the discrepancy that we observed in our Kaplan-Meier analysis where the SETD3 expression correlates with a good prognosis while FOXM1 with a bad prognosis in patients with ER-positive tumors could be due to the following reasons: 1) the expression of FOXM1 (up-or down-regulation) is not only regulated www.nature.com/scientificreports www.nature.com/scientificreports/ by SETD3 but is also regulated by a wide range of factors such as c-Myc, H-Ras, N-Ras, EGFR (epidermal growth factor receptor), ErbB1, HER2/Neu (ErbB2), Notch1, STAT3 (signal transducer and activator of transcription 3), KLF4, GATA3, and others 59 . 2) Interestingly, it has been described that ERα activates FOXM1 transcription through an estrogen-response element (ERE) located within the proximal promoter region 59,60 . This could explain why even in ER-positive tumors the expression of FOMX1 correlates with a poor prognosis. On the other hand, Carr and collaborators observed that the expression of FoxM1 is highest in luminal progenitor cells and decreases upon differentiation. Thus, FOXM1 regulates the differentiation of the luminal cells in the breast and that later could lead to more aggressive characteristics in neoplastic cells 61 .
3) The expression of FOXM1 was associated with processes related to aggressiveness like EMT, CSCs, drug resistance and is a master regulator of metastasis, which is the ultimate cause of patient mortality 59,62-64 . All of the above could explain why although FOXM1 is regulated by SETD3 (which is associated with a good prognosis), it was associated with worse survival in patients.
Another gene that could be regulated by SETD3 is Fascin. We observed the downregulation of Fascin in SETD3 depleted-cells that suggest a negative regulatory mechanism of Fascin directly through the expression of SETD3. In a meta-analysis study, Fascin expression was associated with an increased risk of mortality in the breast, colorectal and gastric cancers 65,66 . Interestingly, Heinz and colleagues demonstrated that the expression of Fascin was up-regulated in highly aggressive MDA-MB-231 breast cancer cells and metastasis was also increased by an actin bundling-independent mechanism 67 . Another study observed that the downregulation of Fascin alters cell morphology and reduces cell migration and invasion, while the overexpression of this gene and increases the expression of genes associated with metastasis such as matrix metalloproteases (MMP)-2 and -9 and urokinase-type plasminogen activator (uPA) 68 . Finally, microRNA-mediated targeting of Fascin is part of the anti-invasive mechanism of the tumor suppressor miR-145 in breast cancer 69 . In contrast to our observation that Fascin only has clinical implications in Her2-negative and grade 3 tumors (Table S5), another study revealed that Fascin expression is associated with poor survival in ER-negative tumors 70 . Methodological differences may account for this discrepancy, as the aforementioned study analyzed tissue microarrays, whereas our study relied on transcriptomic analyses. Moreover, the patient collective of our study was 19-fold larger than the collective of Yoder et al. 70 .
Our qPCR analysis revealed that SETD3 siRNA treatment in luminal MCF-7 cells was associated with an upregulation of three pathogenesis factors that were not significantly affected in basal MDA-MB-231 cells, ASMA, MMP2, and eNOS. ASMA is an actin isoform that is upregulated in the tumor and promotes in cancer cells a metastatic activity associated to EMT 71,72 , MMP2 is a metalloproteinase that promotes metastatic growth by degrading basement membrane constituents 73,74 , whereas eNOS affects inflammation, angiogenesis, apoptosis, cell cycle, invasion, and metastasis 75,76 . Therefore, conversely, high expression of SETD3 in ER-positive tumors may result in a survival benefit due to a downregulation of these factors that is worth addressing in future studies.
In summary, our results suggest that the expression of SETD3 influences the survival of breast cancer patients. Particularly in triple-negative and p53 mutated tumors, SETD3 emerges as an important target for the treatment of BC patients. On the other hand, SETD3 expression was related to good prognosis in ER-positive tumors which suggests that its prognostic value and pathophysiological role depend on the expression of hormone receptors and p53. Additionally, elucidating its role in invasion and metastasis as well as in the regulation of the expression of genes such as FOXM1, MMP2, eNOS, ASMA, and Fascin could point out new therapeutic strategies for more specific treatments. Importantly, our results indicate SETD3 among the new potential targets in TP53 tumors as to date no targeted therapy is available for this important sub-cohort of patients. While a differential effect on the actin cytoskeleton and invasiveness of basal and luminal model cell lines have emerged as an underlying mechanism linked to metastatic behavior, the study of additional mechanisms by which SETD3 acts as a marker of good prognosis in receptor-positive tumors and as a marker for poor prognosis in triple-negative tumors will help us to better understand the role of SETD3 in tumor progression.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information Files.