Oncogene (2011) 30, 3513–3521

2011 Macmillan Publishers Limited All rights reserved 0950-9232/11

&

www.nature.com/onc

ORIGINAL ARTICLE

An siRNA screen identifies RSK1 as a key modulator of lung cancer metastasis

,

R Lara^{1 7}, FA Mauri², H Taylor³, R Derua⁴, A Shia³, C Gray⁵, A Nicols⁵, RJ Shiner², E Schofield⁶, PA Bates⁶, E Waelkens⁴, M Dallman³, J Lamb³, D Zicha⁵, J Downward⁷, MJ Seckl¹ and OE Pardo¹

2

¹Department of Oncology, Hammersmith Campus, Cyclotron Building, London, UK; Histopathology Imperial College London,

3

Hammersmith Campus, London, UK; Division of Cell and Molecular Biology, Department of Life Sciences, Faculty of Natural Sciences, Imperial College London, London, UK; ⁴Labo Proteı¨ne Fosforylatie en Proteomics, Katholieke Universiteit Leuven, Leuven,

• 5
• 6

Belgium; Light Microscopy Department, London Research Institute, London, UK; Biomolecular Modelling Laboratory, London Research Institute, London, UK and ⁷Signal Transduction Laboratory, Cancer Research UK, London Research Institute, London, UK

We performed a kinome-wide siRNA screen and identified 70 kinases altering cell migration in A549 lung cancer cells. In particular, ribosomal S6 kinase 1 (RSK1) silencing increased, whereas RSK2 and RSK4 downregulation inhibited cell motility. In a secondary collagenbased three-dimensional invasion screen, 38 of our hits cross-validated, including RSK1 and RSK4. In two further lung cancer cell lines, RSK1 but not RSK4 silencing showed identical modulation of cell motility. We therefore selected RSK1 for further investigation. Bioinformatic analysis followed by co-immunoprecipitation-based validation revealed that the actin regulators VASP and Mena interact with RSK1. Moreover, RSK1 phosphorylated VASP on T278, a site regulating its binding to actin. In addition, silencing of RSK1 enhanced the metastatic potential of these cells in vivo using a zebrafish model. Finally, we investigated the relevance of this finding in human lung cancer samples. In isogenically matched tissue, RSK1 was reduced in metastatic versus primary lung cancer lesions. Moreover, patients with RSK1- negative lung tumours showed increased number of metastases. Our results suggest that the findings of our high-throughput in vitro screen can reliably identify relevant clinical targets and as a proof of principle, RSK1 may provide a biomarker for metastasis in lung cancer patients.
Introduction

Lung cancer is the most common cancer killer with a 5-year survival rate o5%. Non-small cell lung cancer

(NSCLC) accounts for 80% of cases of which adenocarcinoma represents the majority. Most patients present with metastatic lesions and are incurable. Hence, a better understanding of the biological processes underlying lung cancer cell motility, invasion and metastasis is needed to improve patient survival.
The 90-kDa ribosomal S6 kinase (RSK) family is activated downstream of the Ras/MEK/Erk signalling pathway (Blenis et al., 1991). Four human isoforms (RSK1–4) exist (Anjum and Blenis 2008). RSKs are characterised by the existence of two kinase domains that come into close proximity following activating phosphorylation events. Their downstream substrates include a number of cytoplasmic and nuclear targets (CREB, Fos, Jun, TSC2 and filamin A (Anjum and Blenis 2008)) that explain their involvement in diverse cellular processes, such as cell proliferation and survival. Increased expression of RSKs was shown in breast (Smith et al., 2005) and prostate cancer (Clark et al., 2005), whereas RSK2 activity has been linked to cell transformation (Cho et al., 2007; Kang et al., 2007). RSKs have been shown to phosphorylate filamin A (Woo et al., 2004) and p27Kip (Larrea et al., 2009), with RSK1 being implicated in promoting melanoma cell migration in vitro (Larrea et al., 2009). However, the relevance of these findings to cancer metastasis in vivo or in patients has not been established.
Oncogene (2011) 30, 3513–3521; doi:10.1038/onc.2011.61; published online 21 March 2011

Keywords: ribosomal S6 kinase 1; metastasis; lung cancer; siRNA screen; VASP
The Ena/VASP family of actin-binding proteins is not known to mediate RSK1 effects, but is involved in various processes, including cell migration (Krause et al., 2003). Three human family members exist; Mena, VASP and EVL. Their overlapping function is modulated by both homo- (Zimmermann et al., 2002) or heterotetramerisation (Gertler et al., 1996) and phosphorylation events (Benz et al., 2009) that control their effect on actin polymerisation. Increased expression of Ena/VASP family members enhances cancer cell invasiveness in vitro and in vivo (Han et al., 2008;

Correspondence: Professor MJ Seckl and Dr OE Pardo, Department of Oncology, Cancer Medicine, Imperial College London, Hammersmith Campus, Cyclotron Building, Du Cane Road, London, W12 0NN, UK. E-mail: [email protected] or [email protected]

Received 5 July 2010; revised 29 November 2010; accepted 2 February 2011; published online 21 March 2011

RSK1 in lung cancer cell metastasis

R Lara et al

3514

Philippar et al., 2008) and correlates with tumour progression in several malignancies (Gurzu et al., 2008; Hu et al., 2008; Toyoda et al., 2009), including lung adenocarcinoma (Dertsiz et al., 2005).

779 Targets
Motility screen

1.8 1.6 1.4 1.2
1

RSK1
Scramble
160 Hits

Toxicity screen
ERBB2

Here, we performed an siRNA kinome library screen in A549 human lung adenocarcinoma cells and demonstrated that RSK1 silencing increased migration and invasion. This could potentially be explained by VASP phosphorylation on T278, a site regulating binding to actin (Blume et al., 2007). Knockdown of RSK1 enhanced the metastatic potential of A549 cells in vivo. Moreover, immunohistochemical staining revealed that RSK1 was downregulated in metastatic lung cancer samples compared with isogenically matched primary tumours. Thus, in lung cancer, our results suggest that RSK1 expression inhibits metastasis and might be a predictive biomarker for disease dissemination.

0.8 0.6 0.4 0.2
0

RSK2
RSK4
SRC MST1R
70 Hits
Publication

Individual Targets

RSK2
57 Hits
RSK1

• Sc
• RSK4

210-1 -2
210-1 -2
210-1 -2
210-1 -2

• -2 -1 0
• 1
• 2
• -2 -1 0
• 1
• 2
• -2 -1 0
• 1
• 2
• -2 -1
• 0
• 1
• 2

RSK1

**

• RSK2
• RSK4

• 8
• ***

2

• 2
• 7

654321
***
*
1.5

1
1.5

Results

*
**
*
*
*
*
**
**

P
1

Identification of novel regulators of cell motility in lung cancer cells

**
**

1
0.5
0.5

Using the well-characterised A549 NSCLC cell line, we established a high-throughput random-walk motility screen in which 60 cells per condition were automatically tracked using Metamorph software (Molecular Devices, Chicago, IL, USA). The resulting trajectory plots were analysed with a Mathematica notebook (Katso et al., 2006) and analysis of variance performed to assess the significance of the changes observed with Po0.05 used as a cutoff. Our

screen identified 160 proteins that when downregulated, significantly altered A549 cell migration (Figures 1a and b). To exclude cellular toxicity as a cause for decreased cell motility, we used data from our previously published toxicity screen (Swanton et al., 2007) performed with the same cells and RNAi library. Cell motility hits that showed 410% toxicity were discarded, thereby reducing the number of our candidates to 70 proteins (Figure 1a). Reassuringly, several of these were already known modulators of cell migration, including MAP4K4 (Collins et al., 2006), AKT2 (Irie et al., 2005), EphB6 (Fox and Kandpal, 2009) and CDC42 (Nobes and Hall, 1995). However, 57 of our hits were novel modulators of migration (Figure 1a and Supplementary Figure 1 for a full hit list). Four kinase families were found, with several members capable of altering migration as follows (Supplementary Figure 2): the ribosomal protein S6 kinase , casein kinase , CDC- like kinase and Arf-domain containing kinase families. To validate our hits, a second toxicity assay and SmartPool (Dharmacon, Lafayette, CO, USA) deconvolution were performed. We confirmed that changes in cell viability did not account for the observed motility effects (Supplementary Figure 2) and that 14 of the 17 targets validated with at least 2 of the 4 deconvoluted individual sequences (Supplementary Figure 2 and Figure 1d). These effects correlated with target downregulation at the protein and/or mRNA levels (Figure 1d and Supplementary Figure 3).

• Sc
• P
• 1
• 2
• 3
• 4
• Sc
• 2
• 3
• 4
• Sc
• P
• 1
• 2
• 3
• 4

RT

• RSK1
• RSK2
• RSK4

• GAPDH
• GAPDH
• GAPDH

• Figure
• 1

RSK family members regulate cell migration.
(a) Overview of the migration screen. (b) Distribution of the mean speed ratio to scramble for individual targets. (c) Trajectory plots of representative fields of view (n ¼ 15) following transfection

• with RSK1, 2,
• 4
• or non-targeting (Sc) SmartPool siRNAs.

(d) Representative validation experiment for RSK1, 2 and 4. P; pool, 1–4; deconvoluted single oligonucleotides. The average random migration speed distribution of 60 cells per condition is represented as a box plot. Box; 75% and side bar; 12.5% of the cell population. Transverse bar; median speed. Analysis of variance with Sc as reference: *Po0.05, **Po0.005, ***Po0.001. Repre-

sentative reverse transcriptase–PCR (RT) for the target mRNAs with GAPDH as a loading control.

As metastasis involves both enhanced motility and invasiveness, we next performed a secondary threedimensional high-throughput collagen-based invasion screen on our 70 motility hits. Thirty-eight of these cross-validated in the two screens (Supplementary Figure 4). Figure 2a shows side views of representative stacks in which knockdown of 4.1B was used as a positive control (Cavanna et al., 2007). We decided to focus on the RSK family as a paradigm to validate the importance of our screen and because of our previous interest in S6 kinases (Pardo et al., 2001, 2006). In keeping with our migration data, silencing of RSK1 enhanced, whereas RSK4 knockdown suppressed invasion (Figures 2a and b). However, unlike the motility screen, RSK2 silencing failed to suppress invasion. Hence, silencing of RSK1 and RSK4 alter invasion in a similar way to their effects on migration in A549 cells.

Oncogene

RSK1 in lung cancer cell metastasis

R Lara et al

3515

We next investigated whether the effects of RSK1 and
RSK4 silencing on motility could be generalised to other NSCLC cell lines. Figure 2c demonstrates that RSK1 but not RSK4 validated in both NME35 and H23 NSCLC cells. Consequently, we focused on RSK1 and further validated this hit in our three cell lines using four independent siRNA sequences from an alternative provider (Figure 2d). All four sequences were able to enhance the migration in the three NSCLC cell lines examined. In contrast, overexpression of wild-type RSK1 in A549 cells inhibited cell migration (Supplementary Figures 9A–C). This effect was even more pronounced upon overexpression of a kinase-active mutant of this protein (Supplementary Figure 9C). Hence, overexpression and silencing of RSK1 have opposing effects on A549 cell migration, and the role of this kinase on cell motility might be mediated through its kinase activity.

siRNA:
Sc

4.1B RSK1 RSK2 RSK4

RSK1 binds and phosphorylates VASP

*

To identity new interacting partners of RSK1 that might explain the effects of this protein on cell migration and invasion, we performed bioinformatic analysis. A human protein–protein interaction database was assembled that integrated two previously published collections containing both experimentally verified or in-silico predicted interactions (Jonsson and Bates 2006; Linding et al., 2007). Subsequent analysis revealed several known modulators of cell migration (for example, VASP(Krause et al., 2003), VIM (Ivaska et al., 2007)) that might interact with RSK1 (Figure 3a). The changes seen in the actin cytoskeleton upon downregulation of RSK1 (Figure 3b) suggested a potential link with modulators of actin dynamics. As VASP is known to mediate such effects (Trichet et al., 2008), we tested whether the predicted interaction between VASP and RSK1 could be validated. Figure 3c (upper panel) shows that immunoprecipitation of RSK1 pulled down endogenous VASP in A549 cells. Similarly, RSK1 was detected in VASP immunoprecipitates (Figure 3c, lower panel). As VASP is known to heterotetramerize with Mena (Gertler et al., 1996), another protein regulating actin dynamics (Kwiatkowski et al., 2003), we next tested the potential for Mena to associate with RSK1. Co-immunoprecipitation experiments confirmed that RSK1 associates with Mena in A549 cells (Figure 3d).
To investigate the functional relevance of the RSK1/
VASP interaction and determine whether these proteins directly interact, we performed an in vitro kinase assay using purified proteins. Figure 3e shows that RSK1

12
4
8

3
8

n.s.
2
4

4
1
*
10
10
0

+EGF
*
+EGF
6
+EGF

• H23
• NME35

1.8

1.6 1.4 1.2
1
*
5

432
0.8

• A549
• H23
• NME35

** *

Figure 2 Validation of RSK1 in additional cell lines and by invasion assay. (a) Side views of representative confocal stacks of A549 cells invading upward into a collagen matrix. Arrows point to the bottom of the wells. (b) Quantification of cell invasion in A549 treated with RSK1, 2, 4 or Sc siRNAs. The combined invasion distribution for 27 fields of view per condition is represented as a box-plot. Box; 75% and side bar; 12.5% of the cell population. Transverse bar; median. Analysis of variance with epidermal growth factor-treated Sc as reference: *Po0.05, **Po0.005. (c) RSK1

regulates cell migration in NME35 and H23 cell lines. (d) Validation of the effects of RSK1 silencing using four alternate siRNA oligonucleotides in A549, NME35 and H23 cells. (c, d) Random migration speed distributions are shown as in Figure 1.

**

• **
• **

5
*
4

321
*
4

321
**
*

• *
• 4

32
*
*
**
*
*

• P
• 1
• 2
• 3
• 4
• P
• 1
• 2
• 3
• 4
• P 1 2 3 4

• RSK1
• RSK1
• RSK1

Oncogene

RSK1 in lung cancer cell metastasis

R Lara et al

3516

polymerisation (Blume et al., et al., 2007). We therefore tested whether RSK1 might modulate phosphorylation of T278 in intact cells. Figure 3f (left panel) demonstrates that RSK1 silencing prevented phosphorylation of VASP on T278 in A549 cells growing in serum. Knockdown of VASP confirmed that the T278 phosphosite detected was specific for VASP. We also examined the phosphorylation of S157 and S239, two other sites involved in the binding of VASP to actin. Unlike T278, basal phosphorylation of these sites was low and was unaffected by RSK1 downregulation, but as expected, could be induced by forskolin (Figure 3f, right panel). Thus, RSK1 can directly associate and selectively phosphorylate VASP on T278 in A549 NSCLC cells.

WB:

• Ly
• B
• IP

RSK1

VASP S6
-+-
++-
--+
+-+
RSK1 VASP

As VASP T278 phosphorylation is known to correlate with an inhibitory effect on lamellipodia formation and cell migration, one might predict that silencing of VASP could reverse the promigratory effects of RSK1 knockdown. In addition, because VASP heterotetramerises with Mena we investigated the potential role of Mena in controlling RSK1 effects. Figure 4a shows that knockdown of either VASP or Mena decreased A549 cell migration. Moreover, knockdown of VASP or Mena partially reversed, whereas combined downregulation completely prevented the effects of RSK1 silencing. Furthermore, although RSK1 silencing induced a 10-fold increase in lamellipodia, combined VASP and Mena knockdown dramatically reduced this effect (Figures 4b and c). We next investigated the effect of downregulating VASP or Mena on the proinvasive phenotype of RSK1-silenced A549 cells. VASP, but not Mena silencing, slightly decreased the ability of RSK1 siRNAs to promote A549 invasiveness (Supplementary Figure 8). Taken together, these data suggest that VASP and/or Mena mediate the effects of RSK1 on A549 cell migration. However, it seems likely that in addition to VASP, one or more other molecules are involved in RSK1-mediated invasion.

VASP
RSK1

VASP
S6

WB:

• Ly
• B
• IP

siRNA siRNA
WB: