Integrin-uPAR signaling leads to FRA-1 phosphorylation and enhanced breast cancer invasion

Background The Fos-related antigen 1 (FRA-1) transcription factor promotes tumor cell growth, invasion and metastasis. Phosphorylation of FRA-1 increases protein stability and function. We identify a novel signaling axis that leads to increased phosphorylation of FRA-1, increased extracellular matrix (ECM)-induced breast cancer cell invasion and is prognostic of poor outcome in patients with breast cancer. Methods While characterizing five breast cancer cell lines derived from primary human breast tumors, we identified BRC-31 as a novel basal-like cell model that expresses elevated FRA-1 levels. We interrogated the functional contribution of FRA-1 and an upstream signaling axis in breast cancer cell invasion. We extended this analysis to determine the prognostic significance of this signaling axis in samples derived from patients with breast cancer. Results BRC-31 cells display elevated focal adhesion kinase (FAK), SRC and extracellular signal-regulated (ERK2) phosphorylation relative to luminal breast cancer models. Inhibition of this signaling axis, with pharmacological inhibitors, reduces the phosphorylation and stabilization of FRA-1. Elevated integrin αVβ3 and uPAR expression in these cells suggested that integrin receptors might activate this FAK-SRC-ERK2 signaling. Transient knockdown of urokinase/plasminogen activator urokinase receptor (uPAR) in basal-like breast cancer cells grown on vitronectin reduces FRA-1 phosphorylation and stabilization; and uPAR and FRA-1 are required for vitronectin-induced cell invasion. In clinical samples, a molecular component signature consisting of vitronectin-uPAR-uPA-FRA-1 predicts poor overall survival in patients with breast cancer and correlates with an FRA-1 transcriptional signature. Conclusions We have identified a novel signaling axis that leads to phosphorylation and enhanced activity of FRA-1, a transcription factor that is emerging as an important modulator of breast cancer progression and metastasis. Electronic supplementary material The online version of this article (10.1186/s13058-018-0936-8) contains supplementary material, which is available to authorized users.


Background
The transcription factor Fos-related antigen 1 (FRA-1) influences tumor heterogeneity [1] and is an important driver of cancer cell stemness and resistance in breast cancer [2]. FRA-1 is a member of the AP-1 family of transcription factors that regulate cell proliferation, differentiation, apoptosis and other biological functions and is encoded by the fosl1 gene (reviewed in [3,4]). They function as heterodimers composed of one Fos (c-FOS, FOSB, FRA-1 or FRA-2) and one JUN (c-JUN, JUNB or JUND) family member. FRA-1 was originally shown to transform Rat1 fibroblasts [5] and has since been implicated in the invasiveness and progression of several cancers [6][7][8], with a prominent role in enhancing the malignant phenotypes of breast cancer cells [9][10][11][12]. FRA-1 is also a target of the mircoRNA miR34, which is frequently downregulated in metastatic breast cancer cell lines and primary breast tumors with lymph node metastases. Forced expression of miR34 impairs cellular invasion and the ability of breast cancer cells to metastasize [13].
In breast cancer, FRA-1 expression is associated with the transition from normal epithelium to hyperplasia/ ductal carcinoma in situ (DCIS) [14][15][16] and elevated FRA-1 correlates with increasing grade in invasive ductal carcinoma [2,16]. Correlation between FRA-1 expression and clinical outcomes is more controversial. One study failed to detect an association between FRA-1 protein expression and overall survival [16], while others identified positive correlation between FRA-1 gene expression and shorter time to distant metastasis [2,17,18]. A curated FRA-1 transcriptional signature, when applied to numerous gene expression data sets, showed positive correlation with shorter time to distant metastasis or relapse across breast cancer subtypes [9,10]. More recently, high FRA-1 expression was shown to be correlated with shorter overall survival and higher rates of lung metastases in patients with estrogen receptor (ER)positive disease but not ER-negative cancers [19].
Similar to other members in the Fos family of transcription factors [26,27], FRA-1 is phosphorylated. Two serine residues, S252 and S265, in the c-terminal DEST sequence are phosphorylated, leading to increased protein stabilization by protecting FRA-1 from proteosomal degradation [23,[28][29][30][31]. FRA-1 transcriptional activity is correlated with protein stability and phosphorylation status [32] and the c-terminal region of FRA-1 is required for its transforming activity [33]. Receptor tyrosine kinase signaling (including epidermal growth factor receptor (EGFR) and MET), via the ERK pathway, has been shown to mediate FRA-1 phosphorylation in numerous cancers [34,35]. FRA-1 can also be phosphorylated by protein kinase C (PKC)θ and PKCα on additional serine residues in the c-terminal DEST domain, which is thought to synergize with ERK-mediated phosphorylation to stabilize FRA-1 [2,11,36]. AKT signaling has also been shown to regulate the activity of AP-1 complexes, including FRA-1/c-JUN heterodimers [35].
Here, we demonstrate that engagement of the extracellular matrix protein vitronectin (VN), via the integrin and urokinase/plasminogen activator urokinase receptors (uPARs), leads to activation of SRC and mitogenactivated protein (MAP) kinase (MAPK) signaling and ultimately enhanced FRA-1 phosphorylation and the induction of breast cancer invasion.
The cDNA for fosl1 was purchased from GE Healthcare Bio-Sciences Company (Lafayette, CO, USA) and cloned into an expression vector to add an HA-tag to the N-terminus. Phospho-deficient and phosphomimetic versions were created using Quick-change mutagenesis (Agilent Technologies, Santa Clara, CA, USA) following the manufacturer's directions. Sequences for the oligonucleotides used to make these mutants are listed in Additional file 1: Table S1.

Extracellular matrix (ECM) stimulation and gene expression analysis
For the siRNA-mediated knockdown of uPAR, 48 hours post-transfection with siRNA, cells were harvested with 2 mM NaEDTA in PBS, washed with serum-free medium and plated for 30 minutes on culture dishes that were left uncoated or coated with the appropriate ECM. For gene expression analysis, cells were grown on the indicated ECM-coated or uncoated dishes for 18 hours prior to RNA extraction. For the rescue of plaur/fosl1 knockdowns, cells were first transfected with siRNA then, 24 hours later, transfected with the indicated expression plasmid: 24 hours later, these cells were plated for 18 hours on VN-coated dishes prior to RNA extraction. RNA was extracted using RNeasy kits (Qiagen Inc, Toronto, ON Canada) according to the manufacturer's protocol and reverse transcribed using a high-capacity cDNA reverse transcription kit (Thermo-Scientific Inc, Waltham, MA, USA). Gene expression analysis was performed using the LightCycler 480 and associated software using Advanced Relative Gene Expression Analysis (Roche Diagnostics, Laval, Quebec, Canada). The sequences used for qPCR primers are listed in Additional file 1: Table S1.

Invasion assays
Cell invasion assays were performed as previously described [39] with the addition, where indicated, of vitronectin (24 ng/ml) within the Matrigel matrix alone or within both the Matrigel matrix and coated on the bottom surface of the Boyden chamber at 400 ng/cm 2 . In the latter assay, cells invade towards serum freemedium.

In vivo tumor growth and establishment of explant cultures
The fourth mammary fat pad of ten athymic nude mice was injected with 1 × 10 6 BRC cells (n = 10 mice/cohort). Tumor growth was monitored by weekly caliper measurements and volume calculated according to the formula:

ðπLxW^2Þ=6
where L refers to the length and W to the width of the tumor. After 9 weeks of growth, tumors were removed, digested with collagenase B and Dispase I (Roche Diagnostics, Laval, Quebec, Canada) and then incubated with cell growth medium to establish tumor explants.

RNA preparation and microarray analysis
Total RNA was extracted using TRIzol reagent (Gibco/ BRL, Life Technologies, Inc, Grand Island, NY, USA) according to the manufacturer's protocol. The quality of RNA was assessed using a 2100 Bioanalyzer with the RNA 6000 Nano LabChip kit (Agilent Technologies, Mississauga, ON, Canada) according to the manufacturer's protocol.
Microarray hybridization experiments were performed at McGill University and the Genome Quebec Innovation Center (Montreal, QC, Canada) using the HG-U133A GeneChip arrays. This chip allows the analysis of approximately 18,400 transcripts and variants, including 14,500 well-characterized human genes, composed of more than 22,000 probe sets. Protocols are available at the Affymetrix Web site (http://www.affymetrix.com/; Affymetrix, Santa Clara, CA, USA). Methods for labeling and hybridization of RNA were previously described [40].

Gene expression statistical analysis
Raw microarray expression data from Neve et al., 2006 was downloaded from (http://www.ebi.ac.uk/arrayexpress/experiments/E-TABM-157/) and combined with the data from the five BRC cell lines and together they were pre-processed and clustered using GeneSpring software (V7.3, Agilent Technologies). Pre-processing included first robust multiarray averaging (RMA) normalization, then genes with expression below 0.01 were forced to meet this threshold, per-chip normalization was performed to the 50 th percentile, and per-gene normalization to the median. Data are presented as a log ratio, log2. Clustering was performed using the 305 gene classifier as previously described [38] using average linkage with similar branches merged and bootstrapping. The microarray data for the BRC cell lines can be accessed through the Gene Expression Omnibus (GEO) repository [GEO:GSE69915].
Generation of the FRA-1 transcriptional activity signature A publically available dataset ([GSE:46440] [18]) was used to generate an FRA-1 transcriptional activity signature. Significance analysis of microarrays (SAM) [41] was used to identify genes with 1.24-fold increase in controltransfected BT549 cells relative to BT549 cells transfected with an siRNA pool targeting fosl1. To apply the gene signature, the average of the signature genes was first calculated from RMA-normalized gene expression data and subsequently mean-normalized to obtain a value between 0 (low) and 1 (high). The signature was then validated on human breast cancer cell lines with either high or low FRA-1 phosphorylation (Fig. 7b) and only the top 65 genes within this signature were used for survival analysis (Fig. 7c). The number of available patient samples for analysis of overall survival (OS), relapse-free survival (RFS) and distant metastasis-free survival (DMFS) analyses was 3955, 1747 and 1402, respectively ( Fig. 7f-h).

Breast cancer dataset analysis
To assess potential clinical associations between FRA-1 levels, FRA-1 transcriptional activity, and molecular signaling components capable of activating FRA-1, the kmplot.com dataset [42] was used. Patients were split into high and low expression groups based upon the top quartile of patients versus the remainder of patients. Biased arrays were excluded from the analysis. No other filters were applied to the patient population before Kaplan-Meier analysis.

Statistical analysis
GraphPad Prism 7.0 was used for statistical analysis. For sample sizes of three, the Shapiro-Wilk test was used to determine the normality of the data. For samples of eight or more the D'Agostino test was used to determine the normality of the data. Unless otherwise stated in the figure legends Student's t test was used determine statistical significance. Supplemental methods and figure legends can be found in Additional file 2: Document 1: Supplemental Figure legends and Methods.

Results
Novel breast cancer lines derived from primary human breast tumors are representative of the intrinsic subtypes We have examined a set of breast cancer cell lines that have been isolated directly from primary tumors of breast cancer patients [37]. We first characterized these explants by gene expression analysis and performed unsupervised clustering using a 305 gene signature [38] to classify human breast cancer cell lines as either luminal, basal A or basal B. Four of the five primary breast cancer explants (BRC-17, 32, 36 and 196) clustered with other luminal cell populations, whereas one explant (BRC-31) clustered closely with basal B cell lines (Fig. 1a). We specifically examined the expression of genes within the Affymetrix dataset that are characteristic of each intrinsic subtype including the estrogen receptor (esr1) and progesterone receptor (pgr) (luminal subtype), the ErbB2 receptor tyrosine kinase (erbb2) (human epidermal growth factor receptor 2 (HER2) + subtype) or the epidermal growth factor receptor (egrf ) (basal subtype). Luminal breast cancer cells (BRC-17, 32, 36 and 196) expressed higher levels of the esr1 and erbb2 relative to the BRC-31 cell population (Fig. 1b). Conversely, the BRC-31 explant exhibited high expression of egfr relative to the other explant populations, which is characteristic of 50% of basal breast tumors [43] (Fig. 1b). To further validate the basal nature of BRC-31 cells, we examined the expression of epithelial (E-cadherin (cdh1), cytokeratin-8 (krt8)) and mesenchymal (N-cadherin (cdh2), fibronectin (fn1) and vimentin (vim)) markers. BRC-31 breast cancer cells expressed high levels of cdh2, fn1 and vim and low levels of cdh1 and krt-8 compared to BRC-17, 32, 36 and 196 cells (Fig. 1c). These data reinforce the classification of BRC-17, 32, 36 and 196 breast cancer cells as representative of the luminal subtype and the BRC-31 cell population as representative of the basal subtype.
When injected into the mammary fat pads of athymic mice, all five human explant cell lines were tumorigenic; however, only two cells lines (BRC-31 and BRC-36) demonstrated a reproducible ability to establish primary tumors (>60% incidence) and maintain sustained growth (Fig. 1d, e). No metastatic lesions were observed in lung tissue collected from tumor-bearing mice (data not shown). To verify phenotypic stability of these tumors following growth in the mammary fat pad of mice, we established explants from tumor-bearing mice (explant A or B). Immunoblot analyses revealed that BRC-31 breast cancer cells retained expression of EGFR, Ncadherin, fibronectin (FN) and vimentin at higher levels relative to BRC-32, 36 and 196 breast cancer cells (Fig. 1f ). Conversely, BRC-31 cells exhibited low levels of Her2, estrogen receptor-α (ERα), E-cadherin and cytokeratin-8 relative to BRC-32, 36 and 196 cells (Fig. 1f ). These data confirm the basal (BRC-31) and luminal nature (BRC-17, 32, 36 and 196) of these model systems and validate the Affymetrix gene expression profiles ( Fig. 1a-c). While the incidence of tumor formation was similar, the BRC-36 cell line, but not the BRC-31 cell line, exhibited enhanced primary tumor growth in mice implanted with estrogen pellets (data not shown). Thus, we have characterized five novel breast cancer cell lines, four that represent ER-positive luminal breast tumors and one that represents basal breast tumors.

Signaling via ERK2 leads to constitutive FRA-1 phosphorylation in basal B breast tumor cell lines
We next characterized the signaling pathways that were active in the BRC series of cell lines and derived tumor explants. We assessed the activation of the PI3-kinase pathway by examining AKT phosphorylation and the MAPK pathway by detecting ERK1/ERK2 phosphorylation. While the degree of AKT expression and phosphorylation was not noticeably different in the BRC cell lines and explants, we noted that the pattern of ERK1/2 phosphorylation, and to some extent expression of total ERK1/2, was clearly divergent between basal and luminal BRC cell lines. Specifically, we observed that the BRC-31 cell line and both tumor explants exhibited prominent p42 ERK2 expression and phosphorylation relative to the luminal BRC breast cancer cells (Fig. 2a). Previous observations have linked ERK2-dependent signaling to an epithelial-to-mesenchymal transition (EMT) that relied on phosphorylation and stabilization of FRA-1, a component of the AP-1 transcription factor family [31,44]. Interestingly, the Fos-like antigen 1 (fosl1) gene, which encodes FRA-1, is overexpressed in basal breast cancer cell lines when compared to luminal breast cancer cells, with an overabundance specifically in basal B breast cancer cells (Additional file 3: Figure S1A).
The observation that BRC-31 breast cancer cells exhibit preferential ERK2 activation and express markers of an EMT transition (E-cadherin and cytokeratin-8 low; N-cadherin, FN and vimentin high) prompted us to examine the phosphorylation status of FRA-1. In the BRC panel, FRA-1 expression was uniquely elevated in the BRC-31 cells (Fig. 2b). FRA-1 is phosphorylated on serine residues 252 and 265, which leads to protein stabilization [23,[28][29][30][31]. In a panel of established human breast cancer cell lines (Fig. 2b) representing the luminal-like and basal-like subtypes we observed that, similar to previous reports [18], FRA-1 expression and phosphorylation of serine-265 is elevated in basal breast cancer cell lines (Fig. 2b). While ERK1 and ERK2 expression (protein, Fig. 2c; RNA, Additional file 3: Figure S1B Given the elevated expression of EGFR in these cells (Fig. 1f), we anticipated that ERK activation occurred downstream of EGFR. While stimulation of BRC-31 cells with EGF led to an increase in pFRA-1/FRA-1 levels, inhibition of the EGFR with small molecule kinase inhibitors (AG1478 or gefitinib) did not alter ERK phosphorylation or basal pFRA1 status in BRC-31 cells (Additional file 4: Figure S2A, B). To identify the signaling pathways responsible for basal pFRA-1 levels in BRC-31 cells, we extended our characterization of the BRC panel of cell lines and examined FAK and SRC phosphorylation. We observed elevated phosphorylation on the FAK auto-phosphorylation site Y397 and on sites that are phosphorylated by SRC (Y576 and Y925) specifically in the basal BRC-31 cell line (Fig. 2d). Consistent with increased SRC activity we also observed increased SRC phosphorylation on Y416 in the BRC-31 cell line. This observation is of interest considering recent data implicating SRC in the phosphorylation of FRA-1 [45]. To determine if the SRC family kinase (SFK)rapidly accelerated fibrosarcoma (RAF)-MEK pathway led to FRA-1 phosphorylation we treated cells with multiple small molecule kinase inhibitors that individually target each component of this pathway to identify optimal inhibitor concentrations, thus reducing the possibility of offtarget effects. Inhibiting the activity of each protein kinase in this pathway reduced the level of FRA-1 phosphorylation relative to vehicle (Fig. 2e, f). These observations link SFK-RAF-MEK signaling activity to S265 phosphorylation on FRA-1.

Extracellular matrix components engage integrin receptors for FRA-1 activation
Given that EGFR inhibitors failed to diminish pFRA-1 levels, we reasoned that additional upstream receptors were responsible for SFK activation and ultimately FRA-1 phosphorylation. SFKs can be activated downstream of integrin engagement; thus, we assessed the expression levels of several integrin members. Interestingly, α v , α 5 , β 1 and β 3 integrin subunits were uniquely upregulated in BRC-31 breast cancer cells relative to the luminal BRC cell lines (Fig. 3a). These observations argue that integrin receptormediated FAK and SRC activation may represent a new mechanism leading to FRA-1 phosphorylation. Distinct integrin receptors bind to specific components of the ECM. For example, α 5 β 1 receptors bind FN, α V β 3 integrin receptors bind VN and α 2 β 1 integrin receptors bind to laminin (LN). To determine which of these integrin heterodimers was responsible for increased FRA-1 phosphorylation, BRC-31 basal breast cancer cells were plated on VN, FN or LN. Only VN led to a significant increase in FRA-1 phosphorylation in BRC-31 breast cancer cells (Fig. 3b, c), which occurred prior to significant cell spreading on the ECM (Additional file 5: Figure S3A, B). To determine if VN-induced FRA-1 phosphorylation was mediated through ERK1/2, cells were transfected with siRNAs directed against either mapk3 (ERK1), mapk1 (ERK2) or scrambled control (Scrambled) and plated on VN-coated dishes (Fig. 3d). Only knockdown of ERK2, but not ERK1, diminished VN-induced FRA-1 phosphorylation (Fig. 3d).
Using available gene expression data from BT549 breast cancer cells transfected with siRNAs targeting fosl1 [18] we generated an FRA-1 transcriptional signature composed of genes that are positively regulated by FRA-1 (Table 1). Using four of the top regulated genes from this list, we assessed whether activation of FRA-1 phosphorylation correlated with increased FRA-1 transcriptional activity. Consistent with previous reports linking FRA-1 transcriptional activity with phosphorylation [32], only cells plated on VN, but not FN or LN, exhibited a significant increase in FRA-1 regulated transcriptional targets (Fig. 3e).
Urokinase plasminogen activator receptor (uPAR) is a known regulator of VN signaling in conjunction with integrin receptors [46,47] and is a known transcriptional target of FRA-1 [48]. We speculated that uPAR could engage with specific integrin receptors to induce FRA-1 phosphorylation, which in turn maintains uPAR expression. Of the integrin receptors expressed in BRC-31 cells, α 5, β 1 and β 3 were also elevated in established basal breast cancer cells (Fig. 4a). There was also strong correlation between high levels of plaur (uPAR) expression and the basal subtype (Fig. 4a, b and Additional file 6: Figure S4). We asked whether reduction of uPAR expression in basal-like breast cancer cells would affect ECM-induced signaling and FRA-1 phosphorylation. When BT549, BRC-31 or HCC1143 cells were plated on plastic, knockdown of uPAR expression using siRNA-mediated approaches did not alter FRA-1 phosphorylation when compared with control-transfected cells (Fig. 4c, d). When plated on VN, BT549, BRC-31 and HCC1143 cells harboring control siR-NAs, displayed an increase in FRA-1 phosphorylation. In contrast, when breast cancer cells with diminished uPAR expression were plated on increasing concentrations of VN, we observed a decrease in FRA-1 phosphorylation (Fig. 4c, d). FAK activation, measured by Y925 phosphorylation, was diminished with uPAR knockdown (Fig. 4c); however, we only detected a modest decrease in ERK phosphorylation despite the requirement for ERK2 expression for the phosphorylation of FRA-1 in these cells (Fig. 3c). Taken together, the data argue that VN engages an integrin/uPAR complex to induce downstream SRC/ FAK signaling that ultimately leads to FRA-1 phosphorylation.
FRA-1 increases the invasive properties of breast cancer cells [12]. We postulated that mixing VN with Matrigel would further enhance BRC-31 invasion, as this would lead to increased integrin engagement and enhanced FRA-1 activity. To test this, we coated the bottom surface of the Boyden chamber with VN, mixed VN together with Matrigel in the chamber, and allowed the cells to invade towards serum-free medium. In the absence of VN, BRC-31 and HCC1143 cells did not invade through the Matrigel (data not shown), consistent with our previous observation that LN (the major component of Matrigel) failed to stimulate FRA-1 phosphorylation. In the presence of VN, BRC-31 and HCC1143 cells invaded through the Matrigel/VN mix (Fig. 5a, b, Scr); however, siRNA-mediated knockdown (KD) of FRA-1 (Fig. 5c, d) suppressed VN-induced BRC-31 and HCC1143 cellular invasion (Fig. 5a, b, fosl1 KD). Consistent with a role for uPAR in stimulating the engagement of VN with integrins, siRNA-mediated knockdown of uPAR (Fig. 5c, d) reduced the invasive properties of BRC-31 and HCC1143 cells (Fig. 5a, b, plaur KD). Knockdown of FRA-1 or uPAR had no effect on cell proliferation over the duration of this assay (Additional file 7: Figure  S5A-D). These data demonstrate that uPAR and FRA-1 are required for VN-induced cellular invasion.

Phosphorylation of FRA-1 is required for transcriptional activity
To determine if phosphorylation was required for FRA-1 activity, we knocked down the expression of both uPAR and FRA-1 with siRNA in BRC-31 cells (Fig. 6a) resulting in reduced expression of FRA-1 transcriptional targets relative to a non-targeting control (Fig. 6b), confirming the dependence, in part, on FRA-1 for the transcription of these targets. To rescue knockdown of endogenous FRA-1, we overexpressed either wild-type FRA-1, a phospho-deficient FRA-1 mutant (S252AS265A) or a phospho-mimetic FRA-1 mutant (S252DS265D) (Fig. 6a). Expression of wild-type FRA-1 or the phospho-mimetic mutant of FRA-1 restored expression of FRA-1 regulated transcriptional targets (Fig. 6b). In contrast, BRC-31 cells expressing the phospho-deficient mutant of FRA-1 displayed similar levels of these FRA-1 transcriptional targets to BRC-31 cells harboring the vector control (Fig. 6b). We analyzed representative breast cancer cells lines for expression of uPA, the ligand for uPAR, and observed that uPA mRNA expression and secretion was highest in basal breast cancer cells (Additional file 8: Figure S6A, B). We next wished to determine if this signaling axis was also upregulated in patients with breast cancer. As a first step, we examined tumor lysates from five breast cancer patient-derived xenografts (PDX) for the expression of uPAR, uPA and phosphorylated FRA-1 [49,50]. Two of five PDX samples had elevated uPAR expression and detectable FRA-1 phosphorylation (Additional file 8: Figure S6C). The PDX sample showing the highest phospho-FRA-1 levels exhibited the highest uPA expression in tumor lysates (Additional file 8: Figure S6D). We next assessed clinical correlation between fosl1 expression, FRA-1 activity and breast cancer patient outcomes. Kaplan-Meier analysis of a human breast cancer dataset revealed that fosl1 gene expression alone did not was not significantly prognostic of overall patient survival (Fig. 7a). To validate the robustness of the FRA-1 transcriptional signature (generated from available gene expression data [18] (Table 1)) as a surrogate readout of FRA-1 activity, we used it to segregate breast cancer cells in which FRA-1 phosphorylation status was previously established (Fig. 2b). Importantly, breast cancer cells characterized as high for the FRA-1 transcriptional signature were the same ones that displayed high FRA-1 phosphorylation (Fig. 7b) and the signature was prognostic of poorer overall survival in patients with breast cancer (Fig. 7c). Interestingly, we noted that the correlation between fosl1 mRNA expression and presence of the FRA-1 expression transcriptional signature, while significant, was not very strong (Fig. 7d) consistent with fosl1 expression alone not being prognostic in this dataset (Fig. 7a). This suggests FRA-1 expression alone does not translate to FRA-1 transcriptional activity. We speculated that breast tumors characterized by elevated expression for molecular components of the uPA/uPAR/ VN/FRA-1 signaling axis might possess elevated FRA-1 transcriptional activity. Indeed, a molecular components signature (plau/plaur/vtn/fosl1) correlated well with FRA-1 transcriptional activity (Fig. 7e). Kaplan-Meier analysis of the same breast cancer dataset revealed that this molecular component signature was also prognostic in predicting poorer overall survival, recurrence-free survival and distant metastasis-free survival (Fig. 7f, g, h,  respectively).

Discussion
Phosphorylation provides a rapid mechanism to regulate FRA-1 transcriptional activity depending on the microenvironment encountered by the cell. Here we have described a vitronectin-stimulated pathway that leads to increased FRA-1 phosphorylation, transcriptional activity and invasion in basal-like breast cancer cells. Vitronectin is a soluble ECM protein that is found in the circulation in small vessels surrounding tumor cells [54]. FRA-1 regulates the expression of multiple proteins involved in cell migration and invasion and it can also promote cell migration by suppressing RhoA activity; however, the exact mechanism has not yet been identified [55,56].
The glycosyl phosphatidylinositol anchored membrane protein uPAR has been implicated in vitronectinstimulated tumor cell migration and invasion [46,47,57]. As a membrane-associated protein, uPAR requires other membrane receptors including G protein-coupled receptors (GPCRs), certain growth factor receptors and integrin complexes to facilitate intracellular signaling (recently reviewed [58]). The β 1 -integrin and β 3 -integrin subunits are frequently reported as important signaling partners for uPAR [59]. Both uPAR and β3-containing integrin receptors can bind vitronectin, with uPAR recognizing the SMB domain and the β3 integrin subunit the Arg-Gly-Asp sequence of vitronectin; however, it is unclear whether a ternary complex indeed forms between uPAR-VN-β 3 -integrin [60]. Integrin signaling may be enhanced by uPAR through a direct conformational change in the β 3 -integrin subunit [61] or uPAR may alter the membrane surface leading to integrin activation [62]. It was recently reported that cell spreading induced by uPAR-vitronectin is not dependent on β 3 -integrin signaling, but requires nonligand dependent activation of β 1 -containing integrin receptors, which is mediated through changes in membrane tension [62]. It is noteworthy that only vitronectin, and not fibronectin or laminin, was able to strongly stimulate FRA-1 phosphorylation in BRC-31 cells. It remains to be determined what additional cellular components are required to initiate this signaling axis.
Binding of urokinase plasminogen activator (uPA) to uPAR can enhance the binding of vitronectin to uPAR due to the fact that uPA and vitronectin utilize mutually exclusive binding sites to simultaneously bind uPAR [63,64]. Upon uPAR binding, uPA is activated and cleaves the zymogen plasminogen into the active protease, plasmin, ultimately leading to the degradation of ECM components [65]. uPAR is also a substrate for uPA and the presence of soluble uPAR (suPAR) fragments in the circulation of preoperative patients with breast cancer is indicative of poor prognosis [66]. As our experimental system removes any secreted protein, we suspect that uPA is not required for  Table 1). Patients are divided into high (red) or low (black) transcriptional activity. vitronectin-stimulated FRA-1 phosphorylation; however, it remains to be determined whether uPA is required for the invasive phenotype. The controversial relationship between FRA-1 expression and clinical outcome in patients with breast cancer may be, in part, due to the fact that expression levels of FRA-1 may not correlate with phosphorylation status and transcriptional activity. To address this possibility, an "'FRA-1 classifier" was constructed by identifying genes from an "'FRA-1 transcriptome" (differentially expressed genes in MDA-MB-231 LM2 cells harboring short hairpin RNAs (shRNAs) to FRA-1 versus vector controls cells) that also had prognostic significance in publicly available datasets [10]. The Desmet et al. FRA-1 classifier showed significant prognostic ability to identify distant metastasis across all subtypes with the exception of Her2+/ER-breast cancers. This curated FRA-1 classifier provides a useful readout of FRA-1 activity; however, the subset of genes regulated by FRA-1 likely differ significantly depending cellular context, epigenetic variation and microenvironment. Similar to the datasets examined by Desmet et al., our Kaplan-Meier analysis using fosl1 expression alone did not have prognostic significance in the independent breast cancer dataset utilized in this study; however, a gene signature containing the molecular components of the novel signaling pathway we have delineated in the present study (plau/ plaur/vtn/fosl1) was able to identify patients with breast cancer with poor overall survival, recurrence-free survival and distant metastasis-free survival. We speculate that tumors with these signaling components would possess elevated FRA-1 activity and therefore be more aggressive in nature. Until now, FRA-1 phosphorylation has been described downstream of receptor tyrosine kinase signaling, via ERK2 [23,[29][30][31], and via members of the PKC family [2,11]. Our data uncover a new signaling pathway, downstream of vitronectin engagement of integrin complexes that is augmented through the uPA/ uPAR axis, which ultimately engages SRC/RAF/MEK to mediate FRA-1 phosphorylation.
Many components of the vitronectin-uPAR-integrin signaling axis are transcriptionally regulated by FRA-1 [10,21], suggesting the existence of a positive feedback loop that further enhances FRA-1 activity. Strategies to suppress the pro-metastatic effects of FRA-1 include targeting downstream transcriptional targets and suppressing their activity [9,10,25]. However, given that numerous transcriptional targets likely contribute to the observed FRA-1 effects on breast cancer invasion and metastasis, such an approach may prove ineffective. Here we have demonstrated that vitronectin, and not laminin or fibronectin, stimulates FRA-1 phosphorylation via an uPARdependent process, suggesting that targeting this specific upstream axis could prove efficacious [67]. Application of the recently characterized small molecule inhibitors of uPAR-integrin association [68] and antibodies [69] may be beneficial in suppressing this signaling axis and reducing tumor cell invasion and metastases.

Conclusions
We have identified a vitronectin stimulated signaling axis that leads to phosphorylation and stabilization of FRA-1, which is associated with increased transcriptional activity and breast cancer invasion. Notably, components of this signaling axis, along with a transcriptional signature of FRA-1 activity, are associated with poor clinical outcomes in patients with breast cancer. These data highlight FRA-1 as a transcription factor important for promoting breast cancer progression and metastasis.