The SUM149, HCC1937, MDA-MB-231 and MDA-MB-468 cells were used as models of BLBC; all are estrogen receptor negative, progesterone receptor negative and HER-2 negative 34. SUM149 cells were purchased from Asterand (Ann Arbor, MI, USA) and were cultured as previously described 1. MDA-MB-231 and MDA-MB-468 (both American Type Culture Collection, Manassas, VA, USA) cells were grown in DMEM (Gibco/Invitrogen, Burlington, ON, Canada) supplemented with 10% FBS and 100 units/ml penicillin/streptomycin. HCC1937 cells (kind donation from WD Foulkes, McGill University, QC, Canada) were cultured in RPMI-1640 media supplemented with 5% FBS, 10 mM HEPES, 4.5 g/l glucose (Sigma, Oakville, ON, Canada), 1 mM sodium pyruvate (Sigma) and 100 units/ml penicillin/streptomycin.

HTR-YB#5 (HTRY) are human mammary epithelial cells immortalized with HPV16, and express YB-1 if induced with tetracycline 35. These were maintained in the same media as SUM149 cells supplemented with 10 ng/ml epidermal growth factor (provided by author IMB). RSK1/RSK2 specific inhibitor SL0101 (Toronto Research Chemicals Inc., North York, ON, Canada) was dissolved in methanol 313637, and PD098059 (Cell Signaling Technologies, Danvers, MA, USA), phorbal 12-myristate 13-acetate (PMA) (Sigma) and epidermal growth factor (EGF) were dissolved in dimethylsulfoxide (DMSO).

SUM149 cells were seeded at a density of 4 × 10⁵ cells in a six-well plate. Cells were subsequently serum-starved for 24 hours prior to 6 hours treatment with vehicle, PD098059 (20 μM) or SL0101 (50 μM). Treated cells were stimulated with the following growth factors for 15 minutes before harvesting; 5% FBS/Ham's/F12 (serum stimulation), EGF (25 ng/ml) and PMA (50 ng/ml), lysed in egg lysis buffer (ELB) and subjected to western blot analysis 2. In all other experiments, HCC1937, MDA-MB-231 and HTRY cells were treated with 100 μM SL0101 and the SUM149 cells with 50 μM for 6 hours. The experiment was performed three times.

Protein was extracted from log-growing cells in ELB 2, supplemented with protease and phosphatase inhibitors, and quantified using the Bradford assay (Biorad, Hercules, CA, USA). Immunoblotting was performed as previously described 2. Specific proteins were detected using the following antibodies: EGFR, 1:1,000 (Stressgen, San Diego, CA, USA); ERK, 1:1,000 (p44/42 MAP kinase; Cell Signaling Technology, Danvers, MA, USA); RSK1, 1:1,000 (Santa Cruz Biotechnology, Santa Cruz, CA, USA); RSK2, 1:500 (Santa Cruz Biotechnology); YB-1, 1:2,000 (Abcam, Cambridge, MA, USA); P-ERK, 1:500 (Cell Signaling Technology); P-RSK^S380, 1:1,000 (Cell Signaling Technology); P-YB-1^S102, 1:1,500 (Cell Signaling Technology, Danvers, MA, USA); Vinculin, 1:1,000 (Upstate, Temecula, CA, USA); and Pan-actin, 1:1,000 (Cell Signaling Technology). Densitometry was performed where appropriate.

A synthetic peptidomimetic of the YB-1 S102 region was manufactured by Sigma with the sequence PRKYLR

S

The kinase assay reactions consisted of active protein kinase (250 ng/assay), substrate (optimized, YB-1 peptide or assay buffer) (40 μM), radiolabeled ³³P-ATP (50 μM in kinase assay buffer; 25 mM MOPS, 12.5 mM β-glycerol phosphate, 25 mM MgCl₂, 5 mM ethylene glycol tetraacetic acid (EGTA), 2 mM ethylenediamine tetraacetic acid, 0.25 mM dithiothreitol) to a final volume of 25 μl. Assays were performed at 30°C for 60 minutes, and then the reaction mixture was dotted on phosphocellulose P81 paper and the radioactivity measured. Activity greater than 5% of the optimized positive control is considered highly significant.

MCF-7 cells stably expressing Flag-YB-1 were serum starved for 16 hours prior to being lysed in radio immunoprecipitation assay (RIPA) buffer. As described above, 500 μg lysate was precleared with protein G agarose for 2 hours. YB-1 was then immunoprecipitated from the cells by overnight incubation at 4°C with 5 μg anti-Flag M2 antibody (Sigma) followed by 2 hours of incubation with protein G agarose. Complexes were then collected by centrifugation and washed firstly in Tris-buffered saline/1% NP40 and then once in modified wash buffer (100 mM Tris, pH 7.4, 50 mM NaCl, 1.5 mM MgCl₂, 1 mM ethylenediamine tetraacetic acid, 0.5% NP40). YB-1 was isolated from protein G through incubation in 0.1 M glycine, pH 3.5, for 5 min at room temperature. Kinase assays were performed for RSK1 as described above.

Log-growing SUM149 cells were lysed in RIPA buffer supplemented with protease inhibitors. Cell lysates were subjected to a Bradford assay for quantification and 500 μg protein was used in subsequent immunoprecipitations. For YB-1 pull-down, lysates were precleared with 60 μl PrecipHen beads (previously described 2) for 2 hours at 4°C with rotation, and the supernatants then incubated with IgY or chicken anti-YB-1 antibodies (5 μg) overnight at 4°C with rotation. Immunocomplexes were collected on PrecipHen beads after incubation at 4°C for 3 hours, by centrifugation. The beads were washed once with PBS/1% NP40, twice with wash buffer (100 mM Tris, pH 7.4, 100 mM NaCl, 1.5 mM MgCl₂, 1 mM ethylenediamine tetraacetic acid, 0.5% NP40) and the proteins eluted by boiling in 5× loading dye for 5 minutes.

Similarly, for total RSK1/RSK2 immunoprecipitations, lysates (500 μg) were precleared with 35 μl protein G agarose for 2 hours prior to incubation with either control IgG or RSK1 or RSK2 antibodies (5 μg) (Santa Cruz Biotechnology) for 16 hours at 4°C with rotation. Immunocomplexes were retrieved through the addition of protein G agarose for 2 hours. Immunoprecipitated proteins were resolved on acrylamide gels and immunoblotted as described above. Horseradish peroxidase protein A was used as the secondary antibody to avoid detection of denatured immunoglobulins (1:2,000; Amersham Biosciences, Piscataway, NJ, USA).

SUM149 cells (4 × 10⁵/well) were transfected with 20 nM siRNA (Qiagen, Mississauga, ON, Canada) using Hiperfect (Qiagen). The fast-forward protocol was followed as described by the manufacturer. RSK1 and RSK2 siRNA sequences were as previously described 38.

HCC1937 cells were seeded at a density of 4 × 10⁵ in a six-well plate 24 hours prior to transfection. Cells were transfected with 2 μg plasmid DNA using 10 μl Lipofectamine 2000/well as per instructions, and were lysed at 24 hours. Plasmid constructs for RSK overexpression studies were empty vectors (pRK7 and pKH3), pKH3-avRSK1, pKH3-mRSK2, pRK7-Myr-avRSK1 and pKH3-avRSK1(K112/464R) (kinase-dead) as previously described 39. The experiment was carried out three times.

Wild-type and RSK2^-/- mouse embryo fibroblasts (MEFs) were cultured as described previously 40, and were stimulated with EGF (10 ng/ml) and cell lysates collected at 5, 15, 30, 60 and 120 minutes (kind donation from Dr YY Cho, University of Minnesota, Austin, MN, USA). Two sets of samples were analyzed.

SUM149 cells were plated in six-well plates (4 × 10⁵ cells/well) and transfected with a luciferase construct containing the first 1 kb of the EGFR promoter (pER1) (kind gift from Alfred C. Johnson US National Cancer Institute, Bethesda, MD, USA – previously described in 141). Cells were transfected with a total of 1.5 μg DNA using Lipofectamine 2000 (Invitrogen). To account for transfection efficiency, cells were co-transfected with a renilla-expressing plasmid (pRL-TK, 10:1 luciferase:renilla; Promega). After 18 hours, cells were treated with vehicle or SL0101 (50 μM) for 6 hours prior to harvesting in 1 × passive lysis buffer (Promega). Luciferase activity was measured and normalized to the renilla reading from the same sample.

SUM149 cells (1 × 10⁷ cells) were treated with vehicle, PD098059 (20 μM) or SL0101 (50 μM) for 6 hours. Crosslinks were established between protein and DNA following 15 minutes of incubation with 1% formaldehyde. Cells were washed and collected by centrifugation. Chromatin immunoprecipitation with anti-P-YB-1 antibody (gift from Dr P Mertens, University Hospital RWTH – Aachen, Aachen, Germany) was carried out as described previously 12. The resulting DNA was amplified using the EGFR2a primers (previously described 12).

While we have previously established that AKT can interact and phosphorylate YB-1^S102 11, it is unclear whether other kinases are also able to perform this function. Serum-starved SUM149 cells were stimulated with 5% FBS/growth media, EGF or the tumour promoter PMA. All of these stimuli activated signaling through the MAP kinase/ERK pathway and led to the induction of P-YB-1^S102 (Figure 1a). The activation of the MAP kinase/RSK/YB-1 cascade was completely reversible by pretreating the cells with the MEK inhibitor PD098059 (Figure 1b). SUM149 cells secrete amphiregulin, resulting in activation of EGFR even in serum-free conditions 42. We therefore also treated the cells with the EGFR inhibitor Iressa. As expected, inhibiting EGFR signaling with Iressa decreased P-YB-1^S102 (Figure 1c). By screening a panel of BLBC cell lines, we noted that YB-1 was activated at varying levels but, interestingly, the level of phosphorylation did not always correlate with the expression of P-AKT^S473. RSK was activated in all cell lines including the MDA-MB-231 cells. These cells do not express P-AKT^S473; however, the level of P-YB-1^S102 is comparable with that of the SUM149 cells, which express activated AKT as well as P-RSK^S380 (Figure 1d). Similarly, in the immortalized normal breast cell line HTRY, P-RSK^S380 is also elevated along with P-YB-1^S102. These cells also do not express activated AKT (Figure 1e). It therefore appears that activation of the MAP kinase pathway can also lead to the induction of P-YB-1^S102. This is of particular importance in BLBC given the role of EGFR signaling in this particular type of breast cancer.

To further explore the role of the MAP kinase pathway in the phosphorylation of YB-1^S102 we next investigated the effect of modulating RSK, which lies downstream of ERK, either pharmacologically or genetically. Initially, using an in vitro kinase assay, we show that RSK1 and RSK2 are able to directly phosphorylate an YB-1 S102 peptide that mimics the region surrounding the S102 site (Table 1). The activity of RSK1 and RSK2 towards the YB-1 target peptide was 80% and 78% compared with the activity of these kinases towards the optimized positive control target, respectively (Table 1). Interestingly, this was greater than the activity of AKT1 towards the YB-1 target (7% of optimized control activity) (Table 1). The activity of AKT1, however, was still considered significant in this assay. Of note, the YB-1 target peptide was also phosphorylated by PKCα (Table 1). Weak RSK1 kinase activity was also detected when using flag-tagged YB-1 immunoprecipitated from stably expressing MCF-7 cells as a substrate (data not shown). In this case the salts required for the protein isolation compromised the level of activity.

We also found that, following immunoprecipitation of endogenous YB-1 from log-growing cells, RSK1 is present in the complex (Figure 2a, left). Similarly, by performing the reverse experiment, immunoprecipitation of RSK1, YB-1 was detected (Figure 2a, right). We were unable to determine an interaction of RSK2 with YB-1 due to a lack of suitable antibody for this application. Since RSK2 could not be detected following YB-1 immunoprecipitation, we believe the interaction between the two proteins maybe weaker. This prompted us to investigate the consequence of inhibiting RSK1 or RSK2 on YB-1 phosphorylation. Following suppression of RSK1 expression with siRNA for 72 hours, the level of P-YB-1^S102 was greatly reduced in SUM149 cells (Figure 2b). The loss of RSK2 also resulted in a decrease in YB-1 phosphorylation, although to a lesser degree than that by RSK1. Simultaneous knockdown of RSK1 and RSK2 produced an effect on the level of P-YB-1^S102 greater than either gene knockdown alone (Figure 2b). The levels of total YB-1 and actin remained unchanged (Figure 2b). In a complementary study, introducing exogenous RSK1, RSK2 or a constitutively active RSK1 (myr-RSK) for 24 hours induced P-YB-1^S102 in HCC1937 cells (Figure 2c) compared with cells transfected with the empty vectors pKH3 and pRK7 (myr-RSK empty vector). The kinase-dead RSK1 mutant, however, was unable to phosphorylate YB-1 at S102 (Figure 2c). Taking an alternative genetic approach, we turned to using MEFs that have a homozygous deletion for RSK2 40. Loss of RSK2 prevented the induction of P-YB-1^S102 following EGF stimulation in a time-dependent manner, as compared with the wild-type MEFs (Figure 2d). ERK and RSK were still phosphorylated in response to EGF in the RSK2^-/- MEFs (data not shown). Interestingly, the YB-1 downstream target gene EGFR could be induced in the wild-type cells after 120 minutes; however, this was not the case in the RSK2^-/- cells (relative intensity of EGFR expression compared with wild-type cells at each time point given under blot) (Figure 2d).

We then used the RSK1/RSK2 specific inhibitor SL0101 31 to confirm these findings. SL0101 was used at concentrations in line with previous studies in MCF-7 cells 31. Following treatment of SUM149 cells with SL0101 (50 μM) for between 6 and 16 hours, we observed a reduction in P-YB-1^S102 at all time points whilst the YB-1 level remained constant (Figure 3a). This finding was confirmed in the HCC1937, MDA-MB-231 and HTRY cells treated for 6 hours with SL0101 (100 μM) (Figure 3b,c). Likewise, pretreating SUM149 cells with SL0101 prevented the stimulation of P-YB-1^S102 by serum, EGF or PMA after 6 hours compared with cells treated with the vehicle (methanol) control (Figure 3d). P-RSK^S380 is phosphorylated by the C-terminal kinase, and SL0101 inhibits the N-terminal kinase activity. One therefore cannot measure the effect of SL0101 by studying P-RSK^S380.

We have previously established the importance of phosphorylation of YB-1^S102 for its transcriptional activity in breast cancer 2, and in particular the regulation of EGFR in BLBC. Firstly, we performed a reporter assay using a 1 kb EGFR–luciferase construct that contains an YB-1 binding site at -968 base pairs 1. Knocking down YB-1 with siRNA or inhibiting signaling with PD098059 decreased the EGFR promoter activity by ~80% (P < 0.001) (Figure 4a), while inhibition further downstream with the RSK inhibitor SL0101 decreased EGFR reporter activity by 30% (P = 0.02) (Figure 4a). Consistent with this observation, PD098059 and SL0101 prevented P-YB-1(S102) from binding to the EGFR promoter based on chromatin immunoprecipitation (Figure 4b).

Inhibition of RSK2 by siRNA in SUM149 cells (Figure 4c) led to a decrease in EGFR expression. This downregulation was mirrored in HTRY and MDA-MB-231 cells following treatment with SL0101 (Figure 4d); densitometric analysis for MDA-MB-231 gave a 35% decrease. We thereby conclude that there is a feed-forward signaling pathway in BLBC where EGF binds to the EGFR, which in turn leads to activation of the MAP kinase/RSK pathway resulting in phosphorylation of YB-1 at S102. Activated AKT and PKCα also have the ability to activate YB-1. Following this, P-YB-1^S102 binds to and transactivates the EGFR gene, further fueling the growth potential of BLBC (Figure 4e).