SP1 and STAT3 Functionally Synergize to Induce the RhoU Small GTPase and a Subclass of Non-canonical WNT Responsive Genes Correlating with Poor Prognosis in Breast Cancer

Breast cancer is a heterogeneous disease whose clinical management is very challenging. Although specific molecular features characterize breast cancer subtypes with different prognosis, the identification of specific markers predicting disease outcome within the single subtypes still lags behind. Both the non-canonical Wingless-type MMTV Integration site (WNT) and the Signal Transducer and Activator of Transcription (STAT)3 pathways are often constitutively activated in breast tumors, and both can induce the small GTPase Ras Homolog Family Member U RhoU. Here we show that RhoU transcription can be triggered by both canonical and non-canonical WNT ligands via the activation of c-JUN N-terminal kinase (JNK) and the recruitment of the Specificity Protein 1 (SP1) transcription factor to the RhoU promoter, identifying for the first time SP1 as a JNK-dependent mediator of WNT signaling. RhoU down-regulation by silencing or treatment with JNK, SP1 or STAT3 inhibitors leads to impaired migration and invasion in basal-like MDA-MB-231 and BT-549 cells, suggesting that STAT3 and SP1 can cooperate to induce high RhoU expression and enhance breast cancer cells migration. Moreover, in vivo concomitant binding of STAT3 and SP1 defines a subclass of genes belonging to the non-canonical WNT and the Interleukin (IL)-6/STAT3 pathways and contributing to breast cancer aggressiveness, suggesting the relevance of developing novel targeted therapies combining inhibitors of the STAT3 and WNT pathways or of their downstream mediators.


Introduction
Breast cancer (BC) is a heterogeneous disease that is routinely classified according to the immunohistochemical assessment of Estrogen and Progesterone Receptors (ER, PR), human epidermal growth factor receptor 2 (HER2) and Ki67 [1,2]. Alternatively, gene expression profiling according to the prediction analysis of microarray (PAM) 50 assay [3] can subdivide BC into intrinsic subtypes pathway. Similar to WNT1, both WNT4 and WNT5a could trigger RhoU mRNA induction acting via the JNK-dependent PCP pathway, since the effect was abolished by the JNK inhibitor SP600125 ( Figure 1A,B). Moreover, shRNA-mediated silencing of the two isoforms showed that JNK1, rather than JNK2, is involved ( Figure 1C and Figure S2A). In order to identify the transcription factors (TFs) involved, deletion fragments of the TATA-less RhoU promoter fused to a luciferase reporter gene were transiently transfected in MEF cells, followed by co-culturing with WNT-expressing or wild-type HEK-293 cells. All three WNT ligands could induce transcription of the RhoU promoter at similar levels, involving the same promoter region between positions −756 and −167 ( Figure 1D). As expected, the β-catenin-dependent control TOPflash was solely responsive to the canonical WNT1 and WNT4 ligands. Further dissection identified a minimal region required for both basal and WNT-inducible promoter activity between positions −366 and −234 ( Figure S1A), conferring however weaker induction as compared to the sequence extending to position −756. Sequence analysis did not reveal binding sites for transcription factors known to be involved in WNT/PCP-mediated transcription such as those belonging to the AP1 family [19]. Therefore, in order to identify WNT-responsive elements we subjected the −756 construct Linker Scanning mutagenesis. Surprisingly, none of the mutations tested could affect promoter inducibility ( Figure S1B), suggesting that WNT-mediated RhoU induction does not depend on a single responsive element but requires the cooperation of disseminated sites within the responsive region. involved ( Figure 1C and Figure S2A). In order to identify the transcription factors (TFs) involved, deletion fragments of the TATA-less RhoU promoter fused to a luciferase reporter gene were transiently transfected in MEF cells, followed by co-culturing with WNT-expressing or wild-type HEK-293 cells. All three WNT ligands could induce transcription of the RhoU promoter at similar levels, involving the same promoter region between positions −756 and −167 ( Figure 1D). As expected, the β-catenin-dependent control TOPflash was solely responsive to the canonical WNT1 and WNT4 ligands. Further dissection identified a minimal region required for both basal and WNTinducible promoter activity between positions −366 and −234 ( Figure S1A), conferring however weaker induction as compared to the sequence extending to position −756. Sequence analysis did not reveal binding sites for transcription factors known to be involved in WNT/PCP-mediated transcription such as those belonging to the AP1 family [19]. Therefore, in order to identify WNTresponsive elements we subjected the −756 construct Linker Scanning mutagenesis. Surprisingly, none of the mutations tested could affect promoter inducibility ( Figure S1B), suggesting that WNTmediated RhoU induction does not depend on a single responsive element but requires the cooperation of disseminated sites within the responsive region. RhoU promoter activity in mouse embryonic fibroblasts (MEF) cells transiently co-transfected with the indicated reporter constructs and a Secreted Embryonic Alkaline Phosphatase (SEAP)-expressing vector for normalization. TOPflash (TOP) and FOPflash (FOP) were used as positive and negative controls, respectively, for the β-catenin-dependent pathway. Transfected cells were stimulated or not with WNT ligands as described above. Data are mean ± SEM of the luciferase induction fold, relative to unstimulated condition. All WNT ligands significantly induced the activity of the −1765, −1300 and −756 RhoU promoter constructs (p < 0.001), but not of the −167 unresponsive control (ns). * p < 0.05, ** p < 0.01, *** p < 0.001; n = 3.   (JNKi, B), or upon knockdown of JNK 1 or JNK 2 by means of 2 independent shRNAs/each (C). Data are mean ± SEM of the relative expression levels, normalized to the 18S rRNA. n = 3.

WNT-Mediated
(D) RhoU promoter activity in mouse embryonic fibroblasts (MEF) cells transiently co-transfected with the indicated reporter constructs and a Secreted Embryonic Alkaline Phosphatase (SEAP)-expressing vector for normalization. TOPflash (TOP) and FOPflash (FOP) were used as positive and negative controls, respectively, for the β-catenin-dependent pathway. Transfected cells were stimulated or not with WNT ligands as described above. Data are mean ± SEM of the luciferase induction fold, relative to unstimulated condition. All WNT ligands significantly induced the activity of the −1765, −1300 and −756 RhoU promoter constructs (p < 0.001), but not of the −167 unresponsive control (ns). * p < 0.05, ** p < 0.01, *** p < 0.001; n = 3.

WNT-Mediated Induction of the RhoU Promoter Rquires JNK-Dependent SP1 Recruitment
The failure to impair promoter inducibility with single mutations could be consistent with one or more TFs exhibiting diffused binding to the promoter due to high Total Binding Affinity (TBA), which considers the contribution of low-affinity binding sites to calculate the overall affinity of transcription factors for a chosen promoter region [25][26][27]. The TBA of the WNT-responsive RhoU promoter region for the TFs included in the JASPAR core vertebrata database was calculated (Table S1). Among high scoring TFs we focused our attention on SP1, known to play a major role in the transcription of GC-rich TATA-less promoters [28]. Although SP1 is mainly known as a basal activator of many housekeeping genes, it is also a target for JNK-mediated phosphorylation, which in turn can modulate its transcriptional and DNA binding activities at specific promoters [28][29][30][31]. We thus reasoned that WNT-activated JNK may mediate SP1 activation, enabling it to induce RhoU gene transcription via diffused binding to its promoter. SP1 involvement was confirmed by the observation that WNT-mediated RhoU mRNA induction was abolished by the SP1 inhibitor mithramycin A (Mit. A) (Figure 2A), which could also suppress the WNT1-mediated induction of the RhoU promoter but not of the TOPflash control ( Figure 2B). The failure to impair promoter inducibility with single mutations could be consistent with one or more TFs exhibiting diffused binding to the promoter due to high Total Binding Affinity (TBA), which considers the contribution of low-affinity binding sites to calculate the overall affinity of transcription factors for a chosen promoter region [25][26][27]. The TBA of the WNT-responsive RhoU promoter region for the TFs included in the JASPAR core vertebrata database was calculated (Table  S1). Among high scoring TFs we focused our attention on SP1, known to play a major role in the transcription of GC-rich TATA-less promoters [28]. Although SP1 is mainly known as a basal activator of many housekeeping genes, it is also a target for JNK-mediated phosphorylation, which in turn can modulate its transcriptional and DNA binding activities at specific promoters [28][29][30][31]. We thus reasoned that WNT-activated JNK may mediate SP1 activation, enabling it to induce RhoU gene transcription via diffused binding to its promoter. SP1 involvement was confirmed by the observation that WNT-mediated RhoU mRNA induction was abolished by the SP1 inhibitor mithramycin A (Mit. A) (Figure 2A), which could also suppress the WNT1-mediated induction of the RhoU promoter but not of the TOPflash control ( Figure 2B).  Accordingly, ChIP analysis revealed strong in vivo binding of SP1 to the RHOU WNT-responsive promoter region following treatment with all three WNT ligands tested ( Figure 2C,D). SP1 binding, undetectable in unstimulated cells, was abolished by both Mit.A treatment ( Figure 2C) and JNK inhibition ( Figure 2D). Thus, all tested WNT ligands can activate a non-canonical, JNK-mediated pathway that leads to SP1 recruitment to the RHOU promoter, driving its transcriptional activation. Accordingly, treatment with all three WNTs could increase the ratio of phosphorylated versus non-phosphorylated SP1 in a JNK-dependent manner, as shown by the relative increased intensity of the slower migrating band, corresponding to phosphorylated SP1 ( Figure 2E,F). As expected Mit.A, which binds to GC-rich regions inhibiting SP1 DNA binding activity, did not affect WNT-mediated SP1 phosphorylation.

Migration of MDA-MB-231 Cells Downstream of Constitutively Active WNT5a Requires both JNK and SP1 Activities and RhoU Expression
β-Catenin-independent WNT signaling is often aberrantly activated in both tumor and stromal cells [11,12,32,33], promoting JNK-dependent invasion [11,33]. To assess the functional role of RHOU in this context we analyzed the highly aggressive basal-like breast cancer cell line MDA-MB-231, whose motility and invasivity have been shown to require the expression of the non-canonical WNT5a and WNT5b ligands [11,33]. Indeed, as compared with the weakly motile MCF-7 cells, MDA-MB-231 cells express significantly higher levels of WNT5b and of the ROR1 co-receptor, implicated in the non-canonical WNT pathway [34] ( Figure 3A). Accordingly, RHOU expression was significantly higher in MDA-MB-231 cells, and fully dependent on both JNK and SP1 activities ( Figure 3A,B).
The motility of MDA-MB-231 cells in a wound healing assay could be significantly impaired both by shRNA-mediated silencing of RHOU ( Figure 3C,D and Figure S3) and by treatment with the JNK or SP1 inhibitors ( Figure 3E,F). Thus, SP1 activity is required to enhance cell migration downstream of the aberrant activation of the WNT/PCP/JNK pathway in basal-like cancer cells, at least partly via the induction of RHOU expression. whose motility and invasivity have been shown to require the expression of the non-canonical WNT5a and WNT5b ligands [11,33]. Indeed, as compared with the weakly motile MCF-7 cells, MDA-MB-231 cells express significantly higher levels of WNT5b and of the ROR1 co-receptor, implicated in the non-canonical WNT pathway [34] ( Figure 3A). Accordingly, RHOU expression was significantly higher in MDA-MB-231 cells, and fully dependent on both JNK and SP1 activities ( Figure 3A,B).  Migrating distance (D,F) was calculated as described in the Materials and Methods, and shown as mean ± SEM. n = 3. * p < 0.05, ** p < 0.01, *** p < 0.001 between the indicated groups.

RhoU is also Involved in Mediating Cell Migration Downstream of STAT3 Signaling
We recently identified RhoU as a STAT3 target gene downstream of gp130 cytokines, and reported a positive correlation between STAT3 activation and RHOU expression levels in a panel of human tumor cell lines, including MDA-MB-231 cells [19]. In order to investigate the role of RhoU in STAT3-induced migration, Stat3 −/− MEFs, which express very low RhoU levels and display impaired migration compared to their wild-type counterparts [19,35], were stably transfected with a doxycycline-inducible RHOU-GFP expression vector and treated with increasing doxycycline concentrations, obtaining dose-dependent RHOU induction ( Figure S2B). RhoU expression could significantly rescue the defective migration of Stat3 −/− MEFs in a wound-healing assay ( Figure 4A,B), demonstrating its involvement in STAT3-mediated migration. A and of the JNK and STAT3 inhibitors (JNKi, S3I), followed by RhoU mRNA quantification by RT-PCR (C), wound healing assay (quantified in D), and transwell invasion assay (E). The same inhibitors were used to treat BT549 cells, followed by RNA quantification (F) and transwell invasion assay (G). (E,G) Invasion was calculated as percentage of control cells (vehicle treated). Mean ± SD of three independent experiments. * p < 0.05, ** p < 0.01, *** p < 0.001 between the indicated groups.
We then explored potential synergic effects between the WNT and STAT3 pathways in MDA-MB-231 cells. Treatment with the STAT3 inhibitor S3I considerably reduced the levels of RHOU mRNA ( Figure 4C), leading to impaired cell migration and invasion at 24 h, in synergy with the JNK inhibitor (p < 0.01) ( Figure 4D,E and Figure S4A). In order to extend these observations, we performed the same invasion experiments on a second basal-like BC cell line, the BT-549 cells, obtaining similar results (Figure 4 F,G and Figure S4B). These data are in line with the idea that RHOU indeed represents a common mediator of STAT3 and non-canonical WNT/SP1 functions, under both physiological and pathological conditions.

SP1 and STAT3 In Vivo Binding Defines a Subclass of Genes Belonging to Non-Canonical WNT or IL-6
Pathway. A and of the JNK and STAT3 inhibitors (JNKi, S3I), followed by RhoU mRNA quantification by RT-PCR (C), wound healing assay (quantified in D), and transwell invasion assay (E). The same inhibitors were used to treat BT549 cells, followed by RNA quantification (F) and transwell invasion assay (G). (E,G) Invasion was calculated as percentage of control cells (vehicle treated). Mean ± SD of three independent experiments. * p < 0.05, ** p < 0.01, *** p < 0.001 between the indicated groups.
We then explored potential synergic effects between the WNT and STAT3 pathways in MDA-MB-231 cells. Treatment with the STAT3 inhibitor S3I considerably reduced the levels of RHOU mRNA ( Figure 4C), leading to impaired cell migration and invasion at 24 h, in synergy with the JNK inhibitor (p < 0.01) ( Figure 4D,E and Figure S4A). In order to extend these observations, we performed the same invasion experiments on a second basal-like BC cell line, the BT-549 cells, obtaining similar results ( Figure 4F,G and Figure S4B). These data are in line with the idea that RHOU indeed represents a common mediator of STAT3 and non-canonical WNT/SP1 functions, under both physiological and pathological conditions.

SP1 and STAT3 In Vivo Binding Defines a Subclass of Genes Belonging to Non-Canonical WNT or IL-6 Pathway
The TFs SP1 and STAT3, whose binding sites are often found close together in the regulatory regions of genes, are involved in several functions related to tumorigenesis including apoptosis, cell growth, angiogenesis, epithelial to mesenchymal transition (EMT) and invasion [36]. In order to investigate the STAT3 and SP1 regulatory network and to identify genes potentially regulated by both factors, we analyzed the Encyclopedia of DNA Elements (ENCODE) Transcription factor binding sites (TFBS) collapsed ChIP Sequencing track (wgEncodeRefTfbsClusteredV3) on H1-ESC, HeLa, MCF10A-Er-Src, K562, GM12878, HepG2 and HCT-116 cell lines, identifying~3600 genes displaying ChIP-Sequencing peaks for both STAT3 and SP1 (see Section Materials and Methods). While STAT3 binding could be detected on the RHOU promoter in MCF10A-Er-Src cells, the only mammary cell line available, which display active STAT3 [37], SP1 binding could not be detected in any of the cell lines analyzed. We reasoned that STAT3 and SP1 may coordinately mediate RHOU transcription in a cell type-specific manner, depending on the activation status of both factors. To expand our analysis, we took advantage of the "Integrative ChIP-Seq analysis of regulatory regions (ReMap 2018) [38], which combines Public and ENCODE datasets, where we could detect in vivo SP1 binding to the RHOU promoter in HEK293T and A549 cells. The latter are lung adenocarcinoma cells displaying non-canonical WNT constitutive activity [39], and express high RHOU mRNA levels as detected by RNA sequencing. By ChIP-qPCR experiments we could demonstrate SP1 binding to the RHOU promoter in the MDA-MB-231 cells line, abolished in the presence of the SP1 inhibitor Mit. A ( Figure 5A). Enrichment analysis of the genes presenting both STAT3 and SP1 binding in published gene sets involving the activation of either the STAT3 or the WNT pathways revealed a positive and significant enrichment (p < 0.05, Fisher exact test) for eight out of seventeen gene sets ( Figure 5B). The ChIP-Seq traces for two representative genes are shown in Figure 5C. Of note, a high percentage of the peaks is located between positions −1000 and +1000 with respect to the TSS of the corresponding gene(s) (~74% for SP1, 57% for STAT3), compatibly with a regulatory role of the binding. Importantly, the two top-ranking gene sets according to significance were linked to IL-6 (Broad Institute, Boston, MA, USA) and to non-canonical WNT [40] signaling pathways, respectively (IL-6,~37% overlap, 32 out of 87 genes; non-canonical WNT,~23% overlap, 113 out of 489 genes). The TFs SP1 and STAT3, whose binding sites are often found close together in the regulatory regions of genes, are involved in several functions related to tumorigenesis including apoptosis, cell growth, angiogenesis, epithelial to mesenchymal transition (EMT) and invasion [36]. In order to investigate the STAT3 and SP1 regulatory network and to identify genes potentially regulated by both factors, we analyzed the Encyclopedia of DNA Elements (ENCODE) Transcription factor binding sites (TFBS) collapsed ChIP Sequencing track (wgEncodeRefTfbsClusteredV3) on H1-ESC, HeLa, MCF10A-Er-Src, K562, GM12878, HepG2 and HCT-116 cell lines, identifying ~3600 genes displaying ChIP-Sequencing peaks for both STAT3 and SP1 (see Section Materials and Methods). While STAT3 binding could be detected on the RHOU promoter in MCF10A-Er-Src cells, the only mammary cell line available, which display active STAT3 [37], SP1 binding could not be detected in any of the cell lines analyzed. We reasoned that STAT3 and SP1 may coordinately mediate RHOU transcription in a cell type-specific manner, depending on the activation status of both factors. To expand our analysis, we took advantage of the "Integrative ChIP-Seq analysis of regulatory regions (ReMap 2018) [38], which combines Public and ENCODE datasets, where we could detect in vivo SP1 binding to the RHOU promoter in HEK293T and A549 cells. The latter are lung adenocarcinoma cells displaying non-canonical WNT constitutive activity [39], and express high RHOU mRNA levels as detected by RNA sequencing. By ChIP-qPCR experiments we could demonstrate SP1 binding to the RHOU promoter in the MDA-MB-231 cells line, abolished in the presence of the SP1 inhibitor Mit. A ( Figure 5A). Enrichment analysis of the genes presenting both STAT3 and SP1 binding in published gene sets involving the activation of either the STAT3 or the WNT pathways revealed a positive and significant enrichment (p < 0.05, Fisher exact test) for eight out of seventeen gene sets ( Figure 5B). The ChIP-Seq traces for two representative genes are shown in Figure 5C. Of note, a high percentage of the peaks is located between positions −1000 and +1000 with respect to the TSS of the corresponding gene(s) (~74% for SP1, 57% for STAT3), compatibly with a regulatory role of the binding. Importantly, the two top-ranking gene sets according to significance were linked to IL-6 (Broad Institute, Boston, MA, USA) and to non-canonical WNT [40] signaling pathways, respectively (IL-6, ~37% overlap, 32 out of 87 genes; non-canonical WNT, ~23% overlap, 113 out of 489 genes).

High Expression of SP1 and STAT3 Dual Targets is Associated with Low Survival in Breast Cancer.
We then assessed the functional relevance of combined STAT3 and SP1 activity in breast cancer. To this end, we examined the METABRIC breast cancer dataset [41] for the expression of a signature composed of the genes belonging to the IL-6 or the WNT non-canonical pathways characterized by in vivo binding of both STAT3 and SP1, as identified above, denominated the SP1-S3 signature (Tables S2 and S3). All profiled tumors (n = 1982) where classified according to the cumulative expression of the SP1-S3 signature (SP1-S3 score, high and low). This score was significantly higher in basal-like tumors compared to other subtypes, in keeping with the known frequent activation of both STAT3 and WNT pathways in this subset ( Figure S5A). Patients with a high SP1-S3 score displayed significantly shorter overall survival as compared to patients scored as low ( Figure 5D). To exclude that this correlation was due to the contribution of basal-like tumors, which can predict shorter survival when considered as a whole versus all other subtypes (p < 0.005), we repeated the survival analysis excluding basal-like tumors. Even under these conditions, a high SP1-S3 score was still significantly predictive of poor prognosis, supporting the idea that SP1 and STAT3 can functionally cooperate to enhance tumor aggressiveness independently of the tumor subtype ( Figure  S5B). A high SP1-S3 score was also significantly positively associated with tumor grade, higher

High Expression of SP1 and STAT3 Dual Targets is Associated with Low Survival in Breast Cancer
We then assessed the functional relevance of combined STAT3 and SP1 activity in breast cancer. To this end, we examined the METABRIC breast cancer dataset [41] for the expression of a signature composed of the genes belonging to the IL-6 or the WNT non-canonical pathways characterized by in vivo binding of both STAT3 and SP1, as identified above, denominated the SP1-S3 signature (Tables S2 and S3). All profiled tumors (n = 1982) where classified according to the cumulative expression of the SP1-S3 signature (SP1-S3 score, high and low). This score was significantly higher in basal-like tumors compared to other subtypes, in keeping with the known frequent activation of both STAT3 and WNT pathways in this subset ( Figure S5A). Patients with a high SP1-S3 score displayed significantly shorter overall survival as compared to patients scored as low ( Figure 5D). To exclude that this correlation was due to the contribution of basal-like tumors, which can predict shorter survival when considered as a whole versus all other subtypes (p < 0.005), we repeated the survival analysis excluding basal-like tumors. Even under these conditions, a high SP1-S3 score was still significantly predictive of poor prognosis, supporting the idea that SP1 and STAT3 can functionally cooperate to enhance tumor aggressiveness independently of the tumor subtype ( Figure S5B). A high SP1-S3 score was also significantly positively associated with tumor grade, higher degree of lymph node positivity and chemotherapy treatment, and negatively correlated with age at diagnosis and ER positivity ( Figure 5E).

Discussion
Molecular classification of tumors based on gene signatures correlated with clinical history may guide prognosis, therapeutic choices and, finally, the development of novel specific therapies [4]. Both the WNT and the STAT3 pathways are often altered in solid tumors including breast cancer, correlating with enhanced tumor cells migration, invasion and metastasis [10,42]. In an attempt to assess potential synergistic mechanisms between the two pathways, we focused on their common transcriptional target, the atypical small GTPase RhoU [19]. Our observation that RhoU induction via the JNK-dependent PCP pathway can be equally mediated by canonical and non-canonical WNT ligands suggests that this small GTPase may act as a common effector of both, expanding the scope for its potential relevance in mediating the pathogenic effects of disrupted WNT signaling. Moreover, we could determine that JNK1, rather than JNK2, is involved in WNT-mediated RhoU induction.
Small GTPases belonging to the Rho family are well known to mediate downstream effects of the JNK/PCP pathway [43] that, although best characterized for its transcription-independent functions, is also known to lead to the transcriptional modulation of several genes, mainly via JNK-dependent activation of members of the AP1 family of transcription factors [13,14]. The finding that RhoU is one of these transcriptional targets expands the repertoire of Rho GTPases downstream of the pathway. Moreover, our data identify SP1 as a novel transcriptional mediator of the JNK/PCP non canonical pathway, since its JNK-dependent recruitment to the RhoU promoter is essential for transcriptional induction downstream of both canonical and non-canonical WNT ligands. Although we cannot discriminate between direct or indirect JNK-mediated SP1 activation, several pieces of evidence implicate this factor in inducible transcription upon both JNK and MAPK-mediated phosphorylation [29][30][31].
Our data establish a clear functional correlation between constitutive activation of the PCP pathway, JNK activity, SP1-mediated RhoU transcription and tumor cell motility. Indeed, RhoU downregulation dramatically impairs migration and invasion of the basal-like MDA-MB-231 cells, which display constitutively high expression of the WNT5b ligand and of the ROR1 co-receptor, both implicated in non-canonical WNT pathways. Importantly, we show that also STAT3 converges on RhoU to support cell migration, since RhoU ectopic expression can partially rescue the defective migration of STAT3 null MEFs. Coordinated inhibition of both STAT3 and JNK cooperatively suppresses MDA-MB-231 cells migration and invasion, supporting the idea that the WNT/PCP/SP1 and STAT3 pathways can functionally synergize. The similar results achieved with BT-549 cells further strengthen our observations. Strikingly, the analysis of genes displaying in vivo binding for both SP1 and STAT3 detects a significant proportion of genes listed as non-canonical WNT or IL-6/STAT3 pathways, suggesting that the molecular mechanisms regulating RhoU transcription here identified may indeed extend to a subclass of WNT and IL-6/STAT3 responsive genes. The significant correlation of this gene signature with lower survival, high tumor grade, lymph node positivity, early age at diagnosis and chemotherapy treatment suggests that indeed the functional convergence of WNT/JNK/SP1 and STAT3 on downstream genes can confer aggressive features to breast tumors, independent of tumor subtype. Analysis of the genes of the signature reveals several interesting groups of genes, also exemplified by the results of the gene ontology analysis. Among the most enriched GO categories are "Chromatin Assembly", "Chromatin Organization" and "DNA Metabolic Process", including the genes encoding histone variants H2AX, essential for DNA repair and linked to chemoresistance in breast cancer [44] and H2AZ, whose high expression correlates with BC progression, lymph node metastasis and survival [45]. Other enriched categories are "Signal Transduction", "Cell Communication" and "Response to Stimulus", and include cytokine receptors involved in inflammation and immune response, integrins, important for migration and invasion, and several MAP Kinases. Indeed, the category "MAPK Cascade" is also significantly enriched. MAP kinase genes showing dual SP1 and STAT3 binding are, for example, those encoding for MEK1, the proto-oncogenic MAP3K8 and p38, all well-known players in Growth Factor and stress signaling linked to tumoral growth. Notably also JNK, the SP1 activator downstream of WNT, is a dual SP1-STAT3 target. Another significantly enriched category is "proteolysis", and indeed it includes a striking number of proteasomal subunits, whose levels and activity are increased in a high percentage of primary breast cancers [46]. Among the "Regulation of Gene Expression, Epigenetic" and the "Transcription, DNA-dependent" categories are included many genes encoding subunits of the Mediator complex, often observed to contribute to the progression of several types of cancer [47], and several well-known oncogenic transcription factors such as JUN and YAP. Finally, suggesting positive feed-forward mechanisms, SP1 itself as well as the WNT ligands WNT5b and WNT2b, the latter involved in both canonical and non-canonical signaling, are also dual SP1 and STAT3 targets.
Our data, shedding light on the molecular mechanisms regulating WNT/PCP-mediated transcriptional induction and pointing towards a cooperative co-regulation of many genes and gene families involved in tumor progression and aggressiveness by WNT/SP1 and STAT3, provide the rationale for efforts towards obtaining a coordinated inhibition of both pathways.

Cell Transfection and Transduction, Co-Culture and Luciferase Assays
Mouse RhoU promoter fragments were generated by PCR using a genomic BAC vector as a template. The primers used are reported in Table S4. PCR products were cloned into the promoterless luciferase reporter vector pGL3 basic (Promega, Madison, WI, USA). Linker Scanning mutagenesis constructs, carrying 10 bp mutations starting from the indicated position ( Figure S1), were generated from the RhoU −756 promoter fragment. The primers used are shown in Table S5. The Super8X-TOPflash (TOP) and Super8X-FOPflash (FOP) constructs [6] were used to monitor WNT pathway activation.
Transient transfections, co-cultures with WNT-expressing cells (24 h), luciferase and SEAP assays were performed essentially as previously described [19]. Briefly, transfections were performed with Lipofectamine TM (Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer's instructions. Each construct was co-transfected at a 10:1 ratio with a vector expressing secreted embryonic alkaline phosphatase (SEAP) as an internal control for transfection efficiency.

Chromatin Immunoprecipitation Assays
ChIP experiments were performed as previously described [19] on three independent biological samples, using the following antibodies: anti-SP1 sc-59X (Santa Cruz Biotechnology, Santa Cruz, CA, USA); anti-histone H3 06-755 (Merck Millipore, Darmstadt, Germany); IgG (ChIP-IT Control Kit-Mouse; Active Motif, La Hulpe, Belgium). Non-immunoprecipitated chromatin was used as Total Input control (T.I.). Immunoprecipitated chromatin was analyzed by quantitative real-time PCR using Platinum SYBR Green PCR SuperMix-UDG with ROX (Life Technologies, Carlsbad, CA, USA). Values were normalized to T.I. and to IgG signals. The β-globin promoter region was used as a negative control (Table 1).

Wound Healing Migration Assay
In vitro wound healing assays were performed essentially as described by Liang et al. [49], with some modifications. In brief, confluent cells in 60-mm dish were serum-starved O/N, followed by mitomycin A treatment (10 µg/mL) for 2 h. The monolayer was then scraped with a sterile pipette tip, washed with PBS and incubated in fresh medium containing 3% FBS. Phase-contrast images were taken at the indicated times (Axiovert 200M microscope, Carl Zeiss, Wetzlar, Germany) (10×). Wound's width was measured with the software ImageJ as the distance between the edges (National Institute of Health, Bethesda, MD, USA), and the migrating distance calculated as the difference between widths with respect to time 0, in three individual fields per experiment. As the edges were often irregular, the distance in each field was integrated from measures taken at three different position. Data are shown as mean ± SEM of three independent experiments.

Matrigel Invasion Assay
Tumor cell invasion assays were performed with 8-µm pores Transwell chambers, and membranes pre-coated with Matrigel (VWR International, Radnor, PA, USA) from Corning (New York, NY, USA). Briefly, cells were harvested and re-suspended in DMEM (MDA-MB-231 cells) or RPMI (BT-459 cells) medium without FBS, and 7.5 × 10 4 cells were seeded into the upper chamber of a 24-well plate. The lower chambers were filled with 10% FBS medium. After 24 h, invading cells were fixed with 4% Paraformaldehyde, stained with Crystal Violet and visualized under an inverted light microscope (Olympus, Tokyo, Japan). Invasion was quantified by Image J, and calculated as percentage of the control cells.

Protein Extracts and Immunoblotting
Whole protein extractions and immunoblotting were performed as described [19] using the following antibodies: anti-SP1 sc-59 (Santa Cruz Biotechnology); mouse monoclonal anti-GFP and mouse monoclonal anti-vinculin (home-made production).

RNA Extraction, Retrotranscription and Quantitative Real-Time PCR (qRT-PCR)
These procedures were previously described [19]. qRT-PCR reactions were performed with the ABI Prism 7300 real-time PCR System (Applied Biosystems, Carlsbad, CA, USA) using Platinum Quantitative or SYBR Green PCR SuperMix-UDG with ROX (Life Technologies). Primers and probes used are reported in Supplementary Table S6. The 18S rRNA pre-developed TaqMan assay (Applied Biosystems) was used as an internal control.

Statistical Analysis
Results were assessed for statistical significance by a standard two-tailed Student's t test, using the GraphPad Prism v.5 software (GraphPad Software Inc., San Diego, CA, USA). Statistically significant p values: p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), with the exception of experiments of Figure 4E,G, where a two-way ANOVA test was performed.

ChIP-Seq Data Analysis
ChIP-Seq data were downloaded from the UCSC browser [50]. Peaks were filtered according to the absolute distance (5 Kb) from the closest gene. Genes displaying binding sites for both TFs were selected.

Gene set Data
Genes related to WNT and STAT3 signaling were obtained either from published manually curated gene sets [39,[51][52][53], or from the "Molecular Signatures Database" (IL6, MsigDB C2) [54] (http: //www.broadinstitute.org/gsea/msigdb/index.jsp). The SP1-S3 signature was defined as follows: genes defined by the ChIP-Seq data analysis were assessed for their enrichment with the above gene sets (Fisher exact test, p < 0.05). A combined gene signature was obtained from the intersection between the two top-ranking signatures and the ChIP-Seq derived signature, obtaining 144 genes, listed in Table S2).

Survival Analysis
Time-to-event analyses were performed using the Kaplan-Meier method (R, survival package). Patients were subdivided in two groups according to the median value of the SP1-S3 score, defined as the sum of SP1-S3 gene expression for each patient. Unstratified log-rank statistics were used to evaluate the effect of the change in SP1-S3 score on overall survival. 4.9.6. Univariate Analysis All univariate analyses were performed using the glm R function to assess the relationship between the SP1-S3 score and clinical features using the univariate Gaussian distribution. Logistic regression was applied for binary variables (https://www.R-project.org).

Conclusions
In conclusion, we demonstrated that JNK1 can mediate the functions of the WNT/PCP pathways by regulating transcription of the RhoU gene and possibly of other targets via the transcription factor SP1. Importantly, our data also show the cooperative regulation of RhoU, and of RhoU-mediated triple negative breast cancer cells migration and invasion, by WNT/JNK1/SP1 and STAT3. These data suggest that a family of SP1/STAT3 target genes could be responsible of mediating WNT/PCP-STAT3 convergent functions. Identifying them and assessing their relevance in different cancer subtypes where the two pathways are known to play a key pathogenic role could in turn support the pre-clinical experimentation of combinations of STAT3 and JNK1 inhibitors in triple-negative breast cancer. Several STAT3 and JNK inhibitors are under clinical trial, and could be tested with this purpose.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/11/1/101/s1, Figure S1: RhoU promoter analysis in MEF cells, Figure S2: Downregulation of Jnk1 or Jnk2 mRNAs by the indicated shRNAs as measured by qRT-PCR in MEF cells, Figure S3: Manipulation of RhoU levels in MEF and MDA-MB-231 cells, Figure S4: Migration and invasion assays upon treatment with various inhibitors, Figure S5: SP1 and STAT3 can functionally cooperate to enhance tumor aggressiveness independently of the tumor subtype. Table S1: Total Binding Affinity of JASPAR Transcription Factors (TF) with the mouse RhoU promoter, Table S2: List of genes included in the SP1-STAT3 signature, Table S3: GO of SP1-S3 signature, Table S4: Primers used to generate the RhoU promoter fusions, Table S5: Primers used to generate the Linker Scanning constructs, Table S6: Primers used for qRT-PCR. Funding: This work was supported by the Italian Cancer Research Association (AIRC IG13009 and IG16930 to V.P.); the Italian Ministry of University and Research (MIUR PRIN 2012 9JLHSY to V.P., MIUR PRIN 2015 YYLBTR to P.P.); the Ateneo/Compagnia di San Paolo (02_2012_0062 to V.P. and 02_2012_0094 to P.P.), and the Truus and Gerrit van Riemsdijk Foundation, Liechtenstein, donation to V.P. Em.M. was the recipient of a PhD fellowship from Fondazione CRT.