MKNK1 is a YB-1 target gene responsible for imparting trastuzumab resistance and can be blocked by RSK inhibition

Trastuzumab (Herceptin) resistance is a major obstacle in the treatment of patients with HER2-positive breast cancers. We recently reported that the transcription factor Y-box binding protein-1 (YB-1) leads to acquisition of resistance to trastuzumab in a phosphorylation- dependent manner that relies on p90 ribosomal S6 kinase (RSK). To explore how this may occur we compared YB-1 target genes between trastuzumab-sensitive cells (BT474) and those with acquired resistance (HR5 and HR6) using genome-wide chromatin immunoprecipitation sequencing (ChIP-sequencing), which identified 1391 genes uniquely bound by YB-1 in the resistant cell lines. We then examined differences in protein expression and phosphor- ylation between these cell lines using the Kinexus Kinex antibody microarrays. Cross-referencing these two data sets identified the mitogen-activated protein kinase-inter- acting kinase (MNK) family as potentially being involved in acquired resistance downstream from YB-1. MNK1 and MNK2 were subsequently shown to be overexpressed in the resistant cell lines; however, only the former was a YB-1 target based on ChIP-PCR and small interfering RNA (siRNA) studies. Importantly, loss of MNK1 expression using siRNA enhanced sensitivity to trastuzu- mab. Further, MNK1 overexpression was sufficient to confer resistance to trastuzumab in cells that were previously sensitive. We then developed a de novo model of acquired resistance by exposing BT474 cells to trastuzumab for 60 days (BT474LT). Similar to the HR5/HR6 cells, the BT474LT cells had elevated MNK1 levels and were dependent on it for survival. In addition, we demonstrated that RSK phosphorylated MNK1, and that this phosphorylation was required for ability of MNK1 to mediate resistance to trastuzumab. Further- more, inhibition of RSK with the small molecule BI-D1870 repressed the MNK1-mediated trastuzumab resistance. In conclusion, this unbiased integrated approach identified MNK1 as a player in mediating trastuzumab resistance as a consequence of YB-1 activation, and demonstrated RSK inhibition as a means to overcome recalcitrance to trastuzumab.

Keywords: trastuzumab resistance; breast cancer; MNK; RSK; YB-1

Introduction

In approximately 25% of breast cancers, HER2 is amplified leading to abnormally high levels of the encoded protein (Slamon et al., 1987). Patients diag- nosed with HER2-positive breast tumors have poor prognosis, with considerably shortened disease-free survival (Slamon et al., 1987; Seshadri et al., 1993). Development of trastuzumab (Herceptin), a monoclonal antibody that targets the ectodomain of HER2, has been a significant advance in the care for these patients (Carter et al., 1992). Unfortunately, only 30% of them respond to single-agent trastuzumab treatment; thus intrinsic resistance is apparent (Vogel et al., 2002). Moreover, those that respond frequently experience acquired resistance within 1 year (Nahta and Esteva, 2006). An understanding of the mechanisms promoting trastuzumab resistance has the potential to lead to rational combinatorial therapies that could improve outcome for patients with HER2-positive breast cancers. Heterodimerization of HER2 with EGFR or HER3 (Motoyama et al., 2002; Diermeier et al., 2005), hyper- activation of phosphatidylinositol-3-kinase (PI3K) sig- naling through loss of PTEN or PIK3CA mutations (Nagata et al., 2004; Berns et al., 2007; Junttila et al., 2009), overexpression or activation of MET (Shattuck et al., 2008), increases in insulin-like growth factor-1 receptor (IGF-1R) signaling from receptor overexpres- sion or heterodimerization with HER2 (Lu et al., 2001; Camirand et al., 2002; Nahta et al., 2005), and activation of non-receptor tyrosine kinase c-SRC (Zhang et al., 2011) have all been implicated in decreased sensitivity to trastuzumab. In order to under- stand the mechanisms of trastuzumab resistance further, use of physiologically relevant models is essential. Ritter et al. (2007) established resistant cell lines from BT474 xenografts in mice that initially responded to trastuzu- mab but eventually recurred (HR5 and HR6 clones) (Ritter et al., 2007). This model is valuable as it recapitulates the scenario commonly seen in clinical development of trastuzumab resistance.

We recently used this model of acquired trastuzumab resistance and demonstrated that Y-box-binding protein-1 (YB-1) modulates resistance to trastuzumab in a phos- phorylation-dependent manner (Dhillon et al., 2010). YB- 1 is an oncogenic transcription/translation factor whose expression is associated with relapse and poor overall survival in various malignancies, including breast cancer (Wu et al., 2006; Habibi et al., 2008). The p90 ribosomal S6 kinase (RSK) (Stratford et al., 2008), and to a lesser extent AKT (Sutherland et al., 2005), can phosphorylate YB-1 on S102, promoting its nuclear translocation. Within the nucleus, YB-1 binds to inverted CCAAT boxes of genes (Didier et al., 1988) including MDR1 (Bargou et al., 1997), EGFR (Wu et al., 2006), PIK3CA (Astanehe et al., 2009), MET (Finkbeiner et al., 2009) and CD44 (To et al., 2010) to increase their transcription and thereby promote tumor progression. Elevated P-YB- 1S102 decreases the sensitivity of HER2-positive breast cancer cells to trastuzumab (Dhillon et al., 2010); however, the intracellular mechanisms underlying this YB-1-mediated trastuzumab resistance remain unclear. MKNK1 and MKNK2 code for mitogen-activated protein (MAP) kinase-interacting serine/threonine protein kinases MNK1 and MNK2, respectively (Mahalingam and Cooper, 2001). MNKs are members of MAP kinase-regulated kinases with activation being dependent on phosphorylation of their T197/T202 residues (Fukunaga and Hunter, 1997; Waskiewicz et al., 1997). MNKs are responsible for phosphorylation of the eukaryotic initiation factor, eIF4E, and as such have functionally been linked to translational regulation (Mahalingam and Cooper, 2001). The involvement of MNKs in tumor initiation and progression, particularly in epithelial tumors, remains largely unexplored. Mice reconstituted with hematopoietic stem cells expressing an activated MNK1 showed accelerated lymphomagen- esis, indicating that it can act as an oncogene (Wendel et al., 2007). Conversely, in a PTEN-negative lymphoma mouse model, tumorigenesis was suppressed by loss of MNK1/2 (Ueda et al., 2010). MNKs were shown to be more highly phosphorylated in HER2-positive breast cancer as compared with non-tumorigenic cells (Chrestensen et al., 2007), and their inhibition in prostate cancer cells repressed the translation of mRNAs required for cell-cycle progression (Bianchini et al., 2008). Recently, MNK1 was shown to be overexpressed in glioblastoma multiforme primary tumors, and its silencing reduced the proliferation of glioblastoma multiforme cell lines (Grzmil et al., 2011). These studies suggest a role for MNKs in cancer progression and emphasize the importance of under- standing the contribution of these emerging kinases to malignancy. In the present study, using the same model of acquired trastuzumab resistance described above, we undertook an unbiased approach combining genome- wide chromatin immunoprecipitation sequencing (ChIP-sequencing) and antibody microarrays to understand YB-1-mediated resistance at a deeper level. Using this integrated approach MKNK1 was identified as a YB-1 target that conveys resistance to trastuzumab treatment. Further, our data indicate that inhibiting the YB-1/ MNK1 pathway through inhibition of an upstream mediator RSK could be a useful strategy to overcome trastuzumab resistance in HER2-positive breast cancers.

Results

ChIP-sequencing identified unique YB-1 transcriptional target genes in trastuzumab-resistant cells

We recently reported that P-YB-1S102 levels are elevated in trastuzumab-resistant HR5 and HR6 cells relative to sensitive BT474 cells (Dhillon et al., 2010). Moreover, we demonstrated that YB-1 overexpression confers resistance in BT474 cells in a phospho-dependent manner (Dhillon et al., 2010). In the present study, we show that nuclear P-YB-1S102 levels decreased in BT474 cells after trastuzumab treatment, but remained elevated in HR5 and HR6 cells (Figure 1a). To determine how elevated nuclear YB-1 may mediate trastuzumab resis- tance, we turned to genome-wide ChIP-sequencing of YB-1 pulled-down DNA from BT474, HR5 and HR6 cells after 72 h of trastuzumab (20 mg/ml) treatment. The ChIP-sequencing approach not only identified candidate YB-1 targets at a genome-wide level, but the differential binding of YB-1 in trastuzumab-sensitive versus resis- tant cell lines provided clues as to what its targets responsible for mediating resistance might be. The Venn diagram (Figure 1b) indicates the number of genes bound by YB-1 in each of the cell lines. A detailed list of genes identified in all three cell lines, as well as the 1391 YB-1 target genes unique to resistant cell lines, is provided in Supplementary Table S1. A more focused list of YB-1 target genes unique to resistant cell lines is provided in Table 1. Of note, there are clusters of genes associated with drug resistance, proliferation, apoptosis, stem cells, chromatin remodeling and metastasis. Within this gene list MKNK1 piqued our interest, as it was one of the few YB-1 target genes that encodes for a protein with kinase activity, making it a potential drug candidate. Furthermore, pathway analyses based on the three gene lists (BT474, HR5 and HR6; Supplemen- tary Table S1) were generated by using the Ingenuity Pathway Analysis software to further define how YB-1 may mediate trastuzumab resistance. These analyses identified biological functions that were significant to the data set. More specifically, in HR5 and HR6 cells, genes related to inhibition of apoptosis were predomi- nant, including MKNK1 (Figure 1c). In conjunction, we used the Kinexus Kinex antibody microarray, which included antibodies to approximately 800 various signaling proteins, as an unbiased approach to profile changes in signal transduction in BT474, HR5 and HR6 cells after trastuzumab treatment (20 mg/ml for 72 h). Table 2 lists the proteins that showed at least a 40% change in expression or phosphorylation in both HR5 and HR6 compared with BT474 cells.

Figure 1 ChIP-sequencing identified genome-wide YB-1 targets. (a) The nuclear and cytoplasmic proteins of untreated cells (BT474, HR5 and HR6) and those treated with trastuzumab for 72 h were analyzed by immunoblotting. (b) A Venn diagram of the YB-1 target genes identified by ChIP-sequencing in BT474, HR5 and HR6 cells after trastuzumab (20 mg/ml) treatment for 72 h. (c) Ingenuity Pathway Analysis of YB-1-bound target genes in BT474, HR5 and HR6 cells identified the biological functions that are most significant to the data set. Genes involved in inhibition of cell death are listed, and those shared between the HR5 and HR6 cells, including MKNK1, are shown in bold.

MNK family of proteins was elevated in the resistant cell lines to the greatest degree. Cross-referencing the Kinex antibody microarray data with the ChIP- sequencing data obtained under the same conditions (20 mg/ml trastuzumab for 72 h) identified MKNK1 as a YB-1 target enriched in trastuzumab-resistant cell lines. Therefore, we chose to pursue MNK1 as a potential lead to understanding the mechanisms of trastuzumab resistance downstream from nuclear YB-1.

P-YB-1S102 induces MNK1 expression by direct promoter occupancy

In agreement with the Kinex antibody microarray, immunoblotting verified higher levels of MNK1 and MNK2 in the HR5 and HR6 resistant cell lines as compared with BT474 cells (Figure 2a). In response to trastuzumab treatment, MNK1 levels decreased in BT474, but remained elevated in HR5 and HR6 cells
(Figure 2a). This paralleled the phosphorylation pattern of YB-1 and its upstream kinase RSK (Figure 2a). On the contrary, MNK2 levels did not decrease in BT474 cells after trastuzumab treatment (Figure 2a).

ChIP–PCR was used to validate the interaction of YB-1 with the MKNK1 promoter identified by ChIP- sequencing (Figure 2b). The first 1 kb of the MKNK1 promoter was analyzed for putative YB-1-binding sites. Four sets of oligonucleotides (ChIP1, 2, 3 and 4) were designed around the six potential YB-1-binding sites on the MKNK1 promoter (Figure 2b). YB-1 bound to region ChIP2 (Figure 2b) but not ChIP1, 3 and 4 (Supplementary Figure S1). Moreover, YB-1 binding to the MKNK1 promoter was confirmed in a subsample of the DNA from cell lines treated with trastuzumab submitted for ChIP-sequencing by performing ChIP– PCR for it along with known YB-1 target genes CD44, EGFR and MET (Figure 2c). There was enriched YB-1 binding to MNKN1, CD44, EGFR and MET promoters in HR5 and HR6 cells as compared with BT474 cells treated with trastuzumab (Figure 2c).

The evidence that MKNK1 was a YB-1 target gene was further confirmed by showing that silencing YB-1 reduced MNK1 transcript (Figure 2d) and protein levels (Figure 2e and Supplementary Figure S2). This was further accompanied by induction of apoptosis as evidenced by elevated phosphorylation of gH2A.XS139 and cleavage of PARP (Figure 2e). Conversely, BT474 cells that stably express either wild-type YB-1 or a phospho-mimic mutant YB-1 (YB-1 S102D) had higher MNK1 transcript (Figure 2f) and protein (Figure 2g) levels compared with the empty vector (EV) control and the inactive mutant YB-1 (YB-1 S102A) cells. We also questioned whether YB-1 might elevate MNK1 in non- tumorigenic cells. Expressing YB-1 under a tetracycline- inducible system in immortalized human mammary epithelial cells, a system previously described by us (Astanehe et al., 2009; Davies et al., 2011), increased MNK1 transcript levels (Supplementary Figure S3). Together, these studies suggest that P-YB-1S102 induces MNK1 through transcriptional control. By contrast, YB-1 did not bind to the MKNK2 promoter based on ChIP-sequencing (Supplementary Table S1), and silen- cing YB-1 did not decrease MNK2 transcript or protein levels (data not shown).

MNK1 mediates trastuzumab resistance

We previously reported that YB-1 overexpression in BT474 cells induced trastuzumab insensitivity (Dhillon et al., 2010). Since we identified MNK1 as a downstream target of YB-1, we questioned whether MNK1 influ- enced trastuzumab sensitivity. When MNK1 or MNK2 was silenced, cells underwent apoptosis only when the former was inhibited, and the double gene knockdown did not show an additive effect (Supplementary Figure S4). Similar results were obtained using a second set of small interfering RNA (siRNA) to MNK1 and MNK2 (data not shown). Given these data, and because only MKNK1 was a YB-1 target, we focused on the role of MNK1 in trastuzumab resistance.

Silencing MNK1 in BT474, HR5 and HR6 cells decreased viability (Figure 3a) and induced apoptosis (Figure 3b). In BT474 cells, siMNK1 or trastuzumab treatment alone decreased viability by approximately 40% (Figure 3a). More importantly, however, silencing MNK1 in combination with trastuzumab reduced the viability of BT474 cells by more than 80%, suggesting that MNK1 inhibition could improve the effect of trastuzumab clinically (Figure 3a). In HR5 and HR6 cells, trastuzumab treatment alone decreased survival by approximately 15%, whereas silencing MNK1 in combination with trastuzumab reduced viability by nearly 60% (Figure 3a). Further, loss of MNK1 overcame trastuzumab resistance in HR5 and HR6 cells as apoptosis was induced (Figure 3b). Similarly, silencing MNK1 in another trastuzumab-sensitive cell line, AU565, increased drug sensitivity (Supplementary Figure S5a). Because some patients have tumors that are intrinsically resistant to trastuzumab, we also asked whether inhibition of MNK1 might overcome insensi- tivity in this situation. Silencing MNK1 in intrinsically resistant, MDA-MB-453 and JIMT-1 cells decreased viability in the presence of trastuzumab by as much as 70% (Supplementary Figure S5b). Therefore, loss of MNK1 decreased the viability of both acquired (HR5 and HR6) and intrinsic (MDA-MB-453 and JIMT-1) trastuzumab-resistant cells.

Figure 2 MNK1 levels are regulated through transcriptional induction by YB-1. (a) BT474, HR5 and HR6 cells treated with trastuzumab for 72 h or left untreated were analyzed by immunoblotting. (b) The first 1-kb region of the MKNK1 promoter with putative YB-1-binding sites. Genomic coordinates based on assembly version hg18 (Ensembl v53 Mar 2006) are shown. ChIP–PCR demonstrates YB-1 binding in region ChIP2, which includes sites 2 and 3. Densitometric measurements of the bands are indicated as percentage comparison to the BT474 sample and indicate enhanced YB-1 binding to the MKNK1 promoter in the HR5 and HR6 cells as compared with the BT474 cells. (c) ChIP–PCR of BT474, HR5 and HR6 cells treated with trastuzumab (20 mg/ml) for 72 h verified enriched binding of YB-1 to MKNK1, CD44, EGFR and MET promoters in the trastuzumab-resistant cell lines as compared with the sensitive cells. (d, e) Cells transfected with a scrambled control or a YB-1 siRNA for 96 h were analyzed by real-time quantitative PCR and immunoblotting, respectively, for changes in MNK1 expression. (e) Evidence that loss of YB-1 led to cell death was assessed by blotting for P-gH2A.XS139 and PARP cleavage. (f, g) BT474 cells stably expressing EV, wild-type YB-1, YB-1 S102D (phospho-mimic mutant) or YB-1 S102A (inactive mutant) were analyzed by real-time quantitative PCR and immunoblotting. (g) The immunoblots were also probed with a YB-1 antibody to verify expression of the recombinant protein, and with a P-YB-1S102 antibody to show that wild-type YB-1 recombinant protein was phosphorylated. *P-value o0.05 compared with the control.

To further support the involvement of MNK1 in mediating resistance to trastuzumab, we overexpressed MNK1 in BT474 cells. After trastuzumab treatment, there were approximately 40% more viable cells in the MNK1-overexpressing population as compared with EV control cells (Figure 3c). Moreover, MNK1-over-expressing BT474 cells showed a response to trastuzu- mab comparable to resistant HR5 and HR6 cells (Figures 3a and c). Similar results were obtained using an alternate MNK1 plasmid, whereas MNK2 over- expression had no effect on trastuzumab response (data not shown). In addition, immunoblots confirmed that the MNK1-overexpressing BT474 cells did not undergo apoptosis as readily as control cells in response to drug treatment (Figure 3d and Supplementary Figure S6). Therefore, forced overexpression of MNK1 in BT474 cells rendered them less sensitive to trastuzumab.
As described above, HR5 and HR6 cells were derived in vivo by exposing mice harboring BT474 tumors to trastuzumab until resistant clones arose (Ritter et al., 2007). This model of acquired resistance led us to discover a possible role for MNK1 in mediating trastuzumab resistance. To take this further, we created a de novo model of acquired trastuzumab resistance by exposing BT474 cells to the drug for 60 days in culture. Using this approach, we created a resistant variant of

BT474 cells (BT474LT). BT474LT cells had increased levels of MNK1 (Figure 3e). Interestingly, similar to HR5 and HR6 cells, BT474LT cells also had higher P- RSK and P-YB-1S102 levels compared with BT474 cells (Figure 3e). Given this, we silenced MNK1 in BT474LT cells using siRNA and demonstrated that the cells regained sensitivity to trastuzumab (Figure 3f). Whereas trastuzumab treatment alone decreased the viability of BT474LT cells by approximately 10%, silencing MNK1 in combination with trastuzumab reduced cell survival by more than 70% (Figure 3f). Collectively, these studies indicate that MNK1 promotes a trastuzumab- resistant phenotype, and that loss of MNK1 enhances the effect of the drug on cells.

Inhibition of RSK blocks MNK1-mediated trastuzumab resistance

We demonstrated that MNK1 is a YB-1 target and its silencing enhances sensitivity of cells to trastuzumab. However, there being no specific inhibitors of MNK1 available, we sought to identify other strategies to inhibit MNK1. RSK is the major kinase responsible for phosphorylating YB-1 at S102 (Stratford et al., 2008). Therefore, we addressed whether its inhibition would affect MNK1 levels. We silenced RSK1 and RSK2, the main isoforms expressed in breast cancer (Smith et al., 2005), in BT474, HR5 and HR6 cells using siRNA for 72 h. Similar to knocking down YB-1, silencing its upstream kinase, RSK, decreased MNK1 transcript (Figure 4a) and protein levels (Figure 4b). Loss of RSK also induced apoptosis (Figure 4b). In addition, whereas RSK1/2 knockdown reduced MNK1 levels in BT474 EV control cells, this effect was rescued with YB-1 S102D (Figure 4c, solid arrow). Therefore, the decrease in MNK1 subsequent to RSK knockdown is YB-1- mediated.

Unexpectedly we noticed that, although YB-1 S102D rescued MNK1 levels after RSK1/2 knockdown, the P-MNK1T197/T202 levels decreased (Figure 4c, broken arrow), suggesting that RSK might selectively regulate MNK1 phosphorylation. We showed by in vitro kinase assays that active RSK1 phosphorylated MNK1 (Figure 4d). An interaction between RSK1 and MNK1 was further demonstrated in BT474 cells expressing a glutathione-S-transferase (GST)-tagged MNK1 plasmid (Figure 4e). Similarly, co-immunoprecipitation studies using RSK1 or MNK1 antibodies revealed that endogenous RSK1 and MNK1 proteins form a complex (Figure 4f). In order to assess the intracellular role of RSK on phosphorylation of MNK1 and not the long- term effect on its total levels by YB-1 transcriptional control, we inhibited RSK activity using a specific small- molecule inhibitor (BI-D1870) (Nguyen, 2008) in a short-term pulse assay. BT474, HR5 and HR6 cells were serum-starved overnight; pre-treated with BI-D1870 (10 mM) for 1 h; and subsequently stimulated with epidermal growth factor (EGF) for 15 min. This short- term BI-D1870 treatment markedly reduced P- MNK1T197/T202 levels (Figure 5a). Further, we established that RSK activity is required for the ability of MNK1 to mediate resistance (Figure 5b). MNK1 overexpression decreased the sensitivity of BT474 cells to trastuzumab; however, BI-D1870 treatment for 72 h re-sensitized these cells to trastuzumab as demonstrated by decreased cell
Figure 4 RSK1 phosphorylates MNK1. (a, b) BT474, HR5 and HR6 cells transfected with a scrambled control, or an RSK1 and RSK2 siRNA, for 96 h were analyzed by real-time quantitative PCR and immunoblotting. (c) BT474 cells stably expressing an EV or YB-1 S102D transfected with scrambled control, or an RSK1 and RSK2 siRNA, for 96 h were analyzed by immunoblotting. The expression of YB-1S102D rescued MNK1 (solid arrow), but not P-MNK1T197/T202 (broken arrow), levels. (d) An autoradiograph (top) and a Coomassie blue-stained gel (bottom) of in vitro kinase assay samples. The P-RSK1 band in lanes 2 and 5 of the autoradiograph demonstrates RSK1 autophosphorylation, whereas the P-MNK1 band in lane-5 indicates phosphorylation by active RSK1. (e) GST- tagged proteins from BT474 cells transfected with GST-tagged EV, MNK1 or MNK2 plasmids were purified and analyzed by immunoblotting. Probing immunoblots using an RSK antibody demonstrated an interaction with MNK1 but not with MNK2. YB-1 antibody was used as negative control. The MNK1 and P-MNK1T197/T202 antibodies were used to confirm transfection of GST-tagged plasmids. The P-MNK1T197/T202 antibody detects both P-MNK1 and P-MNK2 proteins. (f) BT474 extracts used for co- immunoprecipitation using RSK1 and MNK1 antibodies were analyzed by immunoblotting. A 5-mg weight of input protein was
loaded onto the gel. A YB-1 antibody was used as control. *P-value o0.05 compared with the control.

In addition, inhibition of RSK with BI-D1870, particularly in combination with trastuzumab, de- creased viability by up to 90% (Figure 5c and Supple- mentary Figures S7b and c) and induced apoptosis (Figure 5d) in the BT474, HR5 and HR6 cells. Similar results were obtained by silencing RSK1/2 using siRNA (Supplementary Figure S7d). These results suggest that RSK inhibition may be an effective approach to overcome MNK1-mediated trastuzumab resistance.

Discussion

We recently reported that YB-1 mediates trastuzumab resistance in a phosphorylation-dependent manner (Dhillon et al., 2010). In the present study, we sought to identify downstream targets of YB-1 that promote resistance to trastuzumab. Through an integrated approach using genome-wide ChIP-sequencing and antibody microarrays, we identified MKNK1 to be a YB-1 target that is overexpressed in trastuzumab- resistant cell lines. A few recent studies have suggested involvement of MNKs in tumorigenesis (Chrestensen et al., 2007; Wendel et al., 2007; Bianchini et al., 2008; Ueda et al., 2010; Grzmil et al., 2011). Our study is the first to identify a role for MNK1 in drug resistance. The majority of the roles of MNKs in cells have been attributed to their effect on eIF4E phosphorylation and therefore translational regulation (Mahalingam and Cooper, 2001). In BT474, HR5 and HR6 cells, double knockdown of MNK1 and MNK2 inhibited eIF4E phosphorylation; however, P-eIF4ES209 levels were maintained when only MNK1 or MNK2 was silenced (data not shown). We demonstrated that MNK1, but not MNK2, is a mediator of decreased response to trastuzumab therapy. While silencing MNK1 promoted cell death, double knockdown of MNK1 and MNK2 in these cells did not further decrease cell survival, and we were unable to show a role for MNK2 in trastuzumab response. Therefore, we suspect that the role of MNK1 in mediating resistance to trastuzumab is independent of eIF4E phosphorylation. In the recent years, other substrates of MNKs have been identified such as heterogeneous nuclear ribonucleoprotein A1 (hnRNP A1), sprouty (SPRY2), cytoplasmic phospholipase A2 (cPLA2) and polypyrimidine-tract-binding protein- associated splicing factor (PSF) (Hefner et al., 2000; Buxade et al., 2005; DaSilva et al., 2006; Guil et al., 2006; Buxade et al., 2008). Most of these substrates have been implicated in inflammatory response enhancing the production of pro-inflammatory cytokines such as tumor necrosis factor-a or eicosanoids (Hefner et al., 2000; Buxade et al., 2005). More work is required to identify other downstream targets of MNKs, and to understand the unique roles of MNK1 and MNK2 in mediating tumor growth and drug resistance.

Figure 5 RSK1 phosphorylation of MNK1 is required for its ability to mediate resistance. (a) Serum-starved (SS) BT474, HR5 and HR6 cells were pre-treated with BI-D1870 (10 mM) or dimethyl sulfoxide (DMSO) for 1 h, stimulated with EGF for 15 min or not, and analyzed by immunoblotting for changes in P-MNK1T197/T202. (b) BT474 cells transfected with an EV or MNK1 plasmids were treated with BI-D1870 (10 mM) or DMSO either in combination or without trastuzumab for 72 h, and cell viability was assessed by MTT assays. (c, d) BT474, HR5 and HR6 cells were treated with DMSO control, BI-D1870 (10 mM), trastuzumab (20 mg/ml), or a combination of BI-D1870 and trastuzumab for 72 h. (c) Cell viability and (d) apoptosis were analyzed by MTT assays and immunoblotting for P-gH2A.XS139, respectively. *P-value o0.05 compared with each treatment alone.

We demonstrated that MNK1 overexpression decreased the sensitivity of HER2-positive breast cancer cells to trastuzumab. Further, silencing MNK1 in trastuzumab-resistant cell lines enhanced sensitivity to trastuzumab and promoted cell death. However, silencing MNK1 did not sensitize the HR5 and HR6 cells to the level of the parental BT474 cells, suggesting coexistence of other resistance pathways. We demonstrated in this study that MNK1 levels are regulated by P-YB-1S102, which is downstream from RSK signaling. In addition, we demonstrated that RSK directly phosphorylates MNK1. Interestingly, inhibition of RSK in trastuzumab-resistant HR5 and HR6 cells by either siRNA or the small- molecule inhibitor BI-D1870 resulted in a comparable decrease in viability and cell death between parental BT474, and resistant HR5 and HR6 cells, after trastuzu- mab treatment. This suggests that inhibition of RSK may be a more effective therapeutic approach to direct inhibition of MNK1, because not only will it decrease total and phosphorylated MNK1 levels, but will also diminish other targets of YB-1 that are involved in mediating trastuzumab resistance.

Our comprehensive genome-wide ChIP-sequencing identified a number of downstream YB-1 targets that may be responsible for trastuzumab resistance. For example, the biological function identified by Ingenuity Pathway Analysis to be significant to the ChIP- sequencing data set was cell death. Products of many of the anti-apoptotic YB-1 target genes have previously been implicated in drug resistance. For example, inhibition of inhibitor of apoptosis-family members cIAP1 (BIRC2) and cIAP2 (BIRC3) increases apoptosis in response to trastuzumab, lapatinib or gefitinib in HER2-overexpressing cells (Foster et al., 2009), whereas inhibition of MCL-1 sensitizes cells to both lapatinib and trastuzumab (Henson et al., 2006; Martin et al., 2008). Moreover, TNFSF18 and TCF7L1 are associated with tamoxifen and erlotinib resistance, respectively (Treeck et al., 2004; Halatsch et al., 2009). Therefore, upregulation of these anti-apoptotic factors by YB-1 likely has a role in promoting resistance to trastuzumab. Furthermore, ChIP-sequencing identified CD44, EGFR and MET as YB-1 targets more highly expressed in resistant cells after trastuzumab treatment. Increased expression of EGFR (Motoyama et al., 2002), MET (Shattuck et al., 2008) and more recently CD44 (Dhillon et al., 2010) has been implicated in resistance to anti- HER2 therapy. Silencing YB-1 decreases MNK1, EGFR (Wu et al., 2006), MET (Finkbeiner et al., 2009) and CD44 (To et al., 2010). Therefore, inhibiting YB-1 function will likely increase sensitivity to trastu- zumab by reducing the expression of MNK1, EGFR,MET, CD44, the anti-apoptotic proteins mentioned above, and likely other mediators of resistance to the drug downstream from YB-1. For example, c-SRC was recently shown to promote trastuzumab resistance (Zhang et al., 2011). Our ChIP-sequencing analysis identified the non-receptor tyrosine kinase SRC-family member, YES, as a YB-1 target in the trastuzumab- resistant HR5 and HR6 cells. Additionally, ChIP- sequencing identified b-arrestin and EphA1 as YB-1 targets in the resistant cell lines. b-Arrestin forms a complex with YES leading to its activation (Imamura et al., 2001). Activated YES then interacts with HER2 to modulate EphA activity (Zhuang et al., 2010). EphA has been shown to mediate trastuzumab resistance by amplifying signaling through both the PI3K and MAP kinase pathways (Zhuang et al., 2010). Therefore, YB-1- mediated regulation of YES, b-arrestin and EphA can also contribute to trastuzumab resistance.

Figure 6 A summary illustration of RSK/YB-1/MNK1 signaling. Activated RSK phosphorylates YB-1 on S102, promoting its nuclear translocation, where it binds to MKNK1 to enhance its transcription. In addition, activated RSK phosphorylates MNK1, which is required for the ability of MNK1 to mediate resistance. Further, within the nucleus, YB-1 can bind to other targets responsible for mediating trastuzumab resistance, including EGFR, CD44, MET and anti-apoptotic genes. This RSK/YB-1/MNK1 signaling pathway can be inhibited using a small-molecule inhibitor (BI-D1870) to RSK.

The data presented in this study highlight the RSK1/ YB-1/MNK1 network as a novel mechanism for trastuzumab resistance (Figure 6) and open prospects for therapeutic interventions against these targets in patients with HER2-positive breast cancers. Silencing MNK1 induced apoptosis of trastuzumab-resistant cells particularly in combination with trastuzumab; therefore one might consider MNK1 as a therapeutic target. Interestingly, in mammalian models it has been shown that MNKs are not necessary for normal growth and development as double knockout (MNK1/2) mice were viable with no apparent abnormality (Ueda et al., 2004). This suggests MNKs as effective targets for cancer therapy as their loss may selectively block the growth of cancer cells that depend on them, with minimal effect on normal cells. CGP57380 was developed to target MNKs; however, it was shown to be non-specific as it also inhibits MKK1, CK1 and BRSK2 with similar potency (Bain et al., 2007). Based on the data presented here, an alternative approach to decreasing MNK1 levels and overcoming trastuzumab resistance would be loss of YB-1. Small-molecule inhibitors of YB-1 are not currently available, although research in siRNA-based cancer therapeutics is advancing (Judge et al., 2009); thus siYB-1 could be a future therapeutic strategy. Our results indicate RSK inhibition as an alternate approach to block MNK1-mediated trastuzumab resistance. We have previously demonstrated that inhibition of RSK suppresses YB-1 phosphorylation (Stratford et al., 2008; Dhillon et al., 2010). Decreasing YB-1 phosphorylation will diminish its nuclear translocation and decrease the transcription of its downstream targets (Sutherland et al., 2005), including MKNK1. Further, inhibition of RSK will decrease MNK1 phosphorylation, which is required for its ability to promote resistance. There are currently three selective inhibitors of RSK available: BI-D1870, SL0101 and fmk (Nguyen, 2008). We demonstrated that inhibiting RSK with an siRNA or BI-D1870 induced apoptosis in trastuzumab-resistant cell lines. Furthermore, BI-D1870 treatment repressed MNK1-mediated trastuzumab resistance. Therefore, inhibition of RSK activity provides an alternate approach to block MNK1-mediated trastuzumab resis- tance and to inhibit phosphorylation, and thereby nuclear localization of YB-1. Collectively, the data presented here indicate that inhibiting RSK will induce cell death in trastuzumab-resistant cells by (1) decreas- ing MNK1 levels through suppression of YB-1 phos- phorylation and nuclear localization; (2) reducing MNK1 phosphorylation; and (3) decreasing other YB-1 targets that have a role in mediating trastuzumab resistance, including CD44, EGFR, MET and anti- apoptotic proteins (Figure 6). In addition, RSK is phosphorylated downstream from receptor and non- receptor tyrosine kinases, and is classically placed in the MAP kinase signaling pathway. We recently showed that RSK phosphorylation could also be PI3K-mediated (Astanehe et al., 2009). Therefore, RSK is a particularly attractive target as it acts as a hub for signals emitted through both the MAP kinase and PI3K networks, and can therefore have an effect on trastuzumab-resistant tumors with multiple altered signaling pathways. In conclusion, our findings suggest that interfering with the RSK1/YB-1/MNK1 network may be beneficial to combat resistance in patients with HER2-positive breast cancers.

Kinexus Kinex antibody microarray

Cells treated with 20 mg/ml trastuzumab (BC Cancer Agency Pharmacy, Vancouver, BC, Canada) for 72 h were suspended in lysis buffer provided by Kinexus Bioinformatics Corpora- tion (Vancouver, BC, Canada). Fluorescent labeling, hybridi- zation onto KAM-1.2 microarray with 800 antibodies, as well as scanning, imaging and quantitative analysis of the enhanced chemiluminescence signal of the detected proteins, were performed by Kinexus (www.kinexus.ca).

siRNA/plasmid transfections

Cells were transfected with 20 nM siRNA (Qiagen) using Lipofectamine RNAiMAX (Invitrogen). A 2-mg weight of MNK1-pEBG6P (GST-tagged) or MNK1-pCS3-MT (pro- vided by Dr Christopher Proud) was transfected into cells using Lipofectamine 2000 (Invitrogen). Detailed treatment protocols are described in the Supplementary information.

Viability assays

MTT at 5 mg/ml was diluted in growth media, added to cells for 1 h at 37 1C, precipitate was dissolved with dimethyl sulfoxide and absorption values were obtained at 570 nm using an EnSpire 2300 multilabel plate reader (PerkinElmer, Waltham, MA, USA).

In vitro kinase assays

The kinase assays were performed by SignalChem (www.signalchem.com) (Vancouver, BC, Canada) using RSK1 and MNK1 targets that were cloned, expressed, purified and characterized by them using proprietary methods.

Co-immunoprecipitation

Immunoprecipitations were performed as described previously (Stratford et al., 2008), with the following modifications: Cells were lysed in TBS-T (1% triton X100) with protease and phosphatase inhibitors. Equal protein amounts were used in immunoprecipitations using IgG, RSK1 (Santa Cruz Biotech- nology) or MNK1 (Cell Signaling Technology) antibodies. For GST purification, after cell lysis, equal protein amounts were added to glutathione agarose (Sigma, Oakville, ON, Canada) and incubated at 4 1C with rotation for 1 h. The resins were washed with TBS-T and eluted in buffer (0.5 M Tris–HCl (pH 9.5), 0.25 M reduced glutathione) with rotation at 4 1C for 15 min.

Statistical analysis

Results are reported as mean±s.d. of at least three experi- ments. Significance was examined by paired Student’s t-test or analysis of variance, and P-values o0.05 were considered significant.