Critical role for transcriptional repressor Snail2 in transformation by oncogenic RAS in colorectal carcinoma cells

Abstract

Activating mutations in the KRAS gene are among the most prevalent genetic changes in human cancers. To identify synthetic lethal interactions in cancer cells harbouring mutant KRAS, we performed a large-scale screen in isogenic paired colon cancer cell lines that differ by a single allele of mutant KRAS using an inducible short hairpin RNA interference library. Snail2, a zinc finger transcriptional repressor encoded by the SNAI2 gene, was found to be selectively required for the long-term survival of cancer cells with mutant KRAS that have undergone epithelial–mesenchymal transition (EMT), a transdifferentiation event that is frequently seen in advanced tumours and is promoted by RAS activation. Snail2 expression is regulated by the RAS pathway and is required for EMT. Our findings support Snail2 as a possible target for the treatment of the broad spectrum of human cancers of epithelial origin with mutant RAS that have undergone EMT and are characterized by a high degree of chemoresistance and radioresistance.

Introduction

The concept of synthetic lethality was originally defined in fruit fly genetics (Dobzhansky, 1946) and was elaborated in a series of yeast genetic studies by Hartwell et al., (1997). Synthetic lethality occurs when alteration of a gene or treatment with a drug results in cell death only in the presence of another nonlethal genetic alteration, such as a cancer-associated mutation (Kaelin, 2005). Targeting a gene that is synthetic lethal to a cancer-specific mutation should kill only cancer cells and spare normal cells without such a mutation.

RAS-activating mutations, especially KRAS mutations, are one of the most prevalent genetic changes found in cancer, occurring in about 20% of human tumours (Downward, 2003; Karnoub and Weinberg, 2008). In these tumours, the activated RAS protein contributes significantly to several aspects of the malignant phenotype, including the deregulation of tumour cell growth, invasiveness and the ability to induce new blood-vessel formation, and the suppression of programmed cell death. Thus, identifying synthetic lethal genetic interactions in the context of mutant KRAS would provide additional drug targets for therapeutic exploration and also shed new light on RAS signalling pathways (Cully and Downward, 2008).

With the advent of RNA interference technology, it has become possible to systematically determine the functional consequence of gene suppression in cancer cell lines (Downward, 2004; Bernards et al., 2006; Iorns et al., 2007). Isogenic paired cancer cell lines that differ by a single oncogenic lesion can be used to identify potential targets for selectively killing tumour cells. In this study, we screened a small hairpin RNA (shRNA) library for those genes the inhibition of which shows synthetic lethality with the KRAS oncogene. Using paired colon cancer cell lines that differ in the expression of mutant KRAS, we identified a zinc finger transcriptional repressor, Snail2, which is selectively required for the survival of cancer cells with mutant KRAS. We further showed that Snail2 is regulated by the RAS pathway and is very important for the epithelial–mesenchymal transition (EMT) initiated in part by RAS pathway activation. Our findings also support Snail2 as a target for treatment of a broad spectrum of human cancers that have undergone EMT, associated at least in part with mutational activation of RAS.

Results

Identification of genes required for survival in cells with mutant KRAS

To identify those genes the targeting of which selectively kills cancer cells with an activating KRAS mutation, we performed large-scale loss-of-function RNA interference screens using a pair of human isogenic colon cancer cell lines containing a mutant KRAS allele (HCT-116 parental) or only wild-type (wt) KRAS (HKe-3 isogenic counterpart). HCT-116 cells carry an endogenous activating KRAS G13D point mutation required for maintaining their oncogenic state. Their isogenic counterpart, HKe-3, was created by genetic disruption of the activated KRAS allele and is impaired in both anchorage-independent growth and the ability to form tumours in mice (Shirasawa et al., 1993). We screened each cell line with a doxycycline-inducible retroviral shRNA library targeting 2500 human genes, including the majority of known protein kinases and cancer-related genes. The library was screened in six pools using a protocol described previously (Ngo et al., 2006; Shaffer et al., 2008). We analysed the change in the bar code abundance of each shRNA by microarray to identify those that are essential for cell survival and are thus depleted from the surviving cell population. We compared the lethality signature of HCT-116 and HKe-3 cells to identify those shRNAs showing selective depletion in the KRAS mutant, but not in KRAS wt cells. The strongest hit that we found to be selectively lost from HCT-116 cells relative to Hke-3 cells was Snail2, shRNAs targeting which were depleted from HCT-116 cultures by twofold (Figure 1a), but were not lost from HKe-3 cell cultures (P>0.05). Two out of three Snail2 shRNAs were selectively lost from HCT-116 cells. To expand our repertoire of shRNAs for functional validation studies, we also constructed a fourth Snail2 shRNA that effectively decreased the expression of its cognate mRNA (Gupta et al., 2005; Supplementary Figure S1a).

Figure 1.

Snail2 is required for the survival of cells with mutant KRAS. (a) HCT-116 and HKe-3 cells were screened using a retrovirally delivered, doxycycline-inducible, shRNA library to identify genes required for cell survival. Depletion of cells bearing three Snail2-targeted shRNAs in shRNA-uninduced versus induced cells is plotted; error bars represent the standard deviation of triplicate measurements. (b) A Snail2 shRNA is more toxic to HCT-116 cells compared with HKe-3 and HKh-2 cells. A vector for inducible expression of Snail2 shRNA was introduced into cell lines and cell numbers were monitored at indicated days after doxycycline addition. Data were the ratio of live cell number in shRNA-induced versus uninduced cells. *P<0.05; error bars indicate s.d. (c) Snail2 knockdown causes strong inhibition effects on soft agar colony formation of HCT-116 cells. Data were the ratio of soft agar colonies in shRNA-induced versus uninduced cells. Error bars represent the s.d. of triplicates. (d) Snail2 knockdown is more toxic to SW48 KRAS G13D cells compared with SW48 wt cells. siRNAs against Snail2 were transfected into both cell lines and cell viability was measured. **P<0.01; error bars indicate s.d.

To confirm and extend the results from the bar-code screen, we inducibly expressed shRNAs targeting Snail2 in HCT-116, HKe-3 and HKh-2 cells. HKh-2 cells are another isogenic counterpart of HCT-116 cells, in which the activated KRAS allele was also disrupted using targeted homologous recombination (Shirasawa et al., 1993). In agreement with the primary screen, the knockdown of Snail2 killed HCT-116 cells much more than HKe-3 cells (P<0.05), and HKh-2 cells were also relatively resistance to Snail2 knockdown compared with HCT-116 cells (Figure 1b; P<0.05). In addition, we found that shRNA-mediated knockdown of Snail2 expression in HCT-116 cells by two different shRNAs also severely impaired colony formation in soft agar (Figure 1c), thus confirming the inhibition of Snail2 to be sufficient for suppressing the malignant phenotype of HCT-116 cells.

To further determine whether our finding of a correlation between mutant KRAS dependency and sensitivity to Snail2 knockdown is true in a different system, we investigated the effects of Snail2 suppression in another colon cancer cell pair. SW48 cells are wt for KRAS, whereas SW48 KRAS G13D cells were created by knock-in of an activating mutation into one KRAS allele (Di Nicolantonio et al., 2008). We found that Snail2 knockdown using two small interfering RNA (siRNA) oligos differentially killed SW48 KRAS G13D cells, but had a minimal effect on SW48 wt cells (Figure 1d, P<0.01), even though Snail2 knockdown effects were stronger in SW48 wt cells (Supplementary Figure S1b).

These experiments showed that Snail2 is preferentially required by cells that rely on mutant KRAS for their survival, but not otherwise by isogenic cells harbouring wt KRAS.

Snail2 and Snail1 expression levels are both regulated by RAS

Our observations suggested a causal relationship between the presence of a transforming RAS mutation and the requirement for Snail2. To confirm whether Snail2 expression is regulated by RAS activity, we first checked Snail2 mRNA and protein level in HCT-116/HKe-3/HKh2 cells and SW48 wt/SW48 KRAS G13D cells. We found that, in HCT-116 cells, Snail2 mRNA expression was around 1.5-fold that seen in HKe-3 cells (Figure 2a; P<0.05), whereas Snail2 expression in HKh-2 cells was very low (Figure 2a). Expression of Snail2 was also higher in SW48 KRAS G13D cells, around 2.5-fold higher than that in SW48 wt cells (Figure 2b). Similar protein expression pattern was also observed in these cells (Figure 2c). These data suggest that cells with mutant KRAS tend to have a higher expression of Snail2 compared with their isogenic counterparts with wt KRAS.

Figure 2.

Snail2 and Snail expression levels are regulated by the RAS pathway. (a) mRNA expression of Snail2 or Snail1 is higher in HCT-116 than in HKe-3 and HKh-2 cells. *P<0.05; **P<0.01; error bars indicate s.d. (b) mRNA expression of Snail2, but not of Snail1, is higher in SW48 KRAS G13D cells, compared with SW48 wt cells. *P<0.05; error bars indicate s.d. (c) Protein expression of Snail2 and Snail1 in HCT-116/HKe-3/HKh-2 cells, SW48 wt KRAS G13D cells and HKe-3 ER:HRAS V12 cells treated with or without OHT. (d) Snail2 mRNA is regulated by RAS pathway in a dose-dependent manner. HKe-3 ER:HRAS V12 cells were treated with different concentrations of OHT and Snail2 expression was measured by quantitative reverse transcriptase PCR after 3 days of OHT treatment, normalized to ethanol (EtOH)-treated cells. Error bars represent the s.d. of triplicates. (e) Snail2 mRNA is regulated by the RAS pathway in a time-dependent manner. HKe-3 ER:HRAS V12 cells were treated with 50 nm OHT, and Snail2 expression was measured by quantitative reverse transcriptase PCR (RT–PCR) at indicated time points after OHT treatment, normalized to EtOH-treated cells. Error bars represent the s.d. of triplicates. (f, g) Snail2 expression induced by RAS pathway activation is inhibited by LY294002 or UO126. HKe-3 ER:HRAS V12 cells were pretreated with LY294002 (16 μm) or UO126 (10 μm) for 30 min, then 50 nm OHT was introduced together with these inhibitors. After 6 h, Snail2 expression was measured by quantitative RT–PCR (f) or by western blots (g). Snail2 mRNA level was normalized to EtOH-treated cells. Error bars represent the s.d. of triplicates. LY, LY294002; UO, UO126. (h) Snail1 mRNA is regulated by RAS pathway in a dose-dependent manner. HKe-3 ER:HRAS V12 cells were treated with different concentrations of OHT, and Snail1 expression was measured by quantitative RT–PCR after 3 days of OHT treatment, normalized to EtOH-treated cells. Error bars represent the s.d. of triplicates. (i) Snail1 mRNA is regulated by RAS pathway in a time-dependent manner. HKe-3 ER:HRAS V12 cells were treated with 50 nm OHT and Snail1 expression was measured by quantitative RT–PCR at indicated time points after OHT treatment, normalized to EtOH-treated cells. Error bars represent the s.d. of triplicates. (j) Snail2 expression was reduced following KRAS knockdown. HCT-116 KRAS shRNA cells were treated with doxycycline. Snail2 or Snail1 expression was measured by quantitative RT–PCR at indicated time points and normalized to uninduced cells. Error bars represent the s.d.

We then introduced into HKe-3 cells a regulatable RAS construct made up of mutant HRAS fused to the oestrogen receptor (ER) ligand-binding domain that is conditionally responsive to 4-hydroxytamoxifen (OHT; Dajee et al., 2002). Addition of OHT acutely activates the RAS pathway in HKe-3 cells expressing ER:HRAS V12. As shown in Figures 2c-e, Snail2 protein and mRNA level were massively increased after activating the RAS pathway, with more than a 60-fold change in mRNA after 3 days of treatment with 50 nm OHT. Snail2 expression was upregulated as early as 6 h (Figure 2e). To determine whether the mitogen-activated protein kinase or phosphatidylinositol 3-kinase (PI3K) pathway is involved in RAS-mediated activation of Snail2 expression, HKe-3 ER:HRAS V12 cells were treated with OHT and inhibitors for these pathways. Treatment with the MEK inhibitor UO126 partially abolished Snail2 mRNA and protein upregulation because of activation of the RAS pathway, and the PI3K inhibitor LY294002 could completely abolish these effects (Figures 2f and g). Together, these results show that Snail2 mRNA expression is regulated by RAS, with requirement for both the PI3K and mitogen-activated protein kinase pathways.

Snail2 belongs to the Snail superfamily of zinc finger transcription repressors, along with Snail1 (Hemavathy et al., 2000; Katoh, 2005; Cobaleda et al., 2007; Peinado et al., 2007). Snail1 mRNA expression is also regulated by RAS activation, but to a much lesser extent than Snail2. Snail1 expression has previously been reported to be induced by RAS and transforming growth factor-β (TGF-β) signalling through pathways involving both MAP kinase and PI3-kinase activities (Peinado et al., 2003; Barbera et al., 2004). With the ER:HRAS V12 system in HKe-3 cells, we showed that the Snail1 mRNA level was increased around 2.5-fold with 50nM OHT after 3 days of induction (Figures 2h and i). Snail1 expression was also regulated by RAS both in a dose-dependent and time-dependent manner. Snail1 expression level was relatively higher in HCT-116 cells than in both HKe-3 and HKh-2 cells (Figures 2a and c), but expression of Snail1 was not increased in SW48 KRAS G13D cells (Figures 2b and c).

The regulation of Snail2 by RAS was further confirmed by knocking down of KRAS in HCT-116 cells. Following doxycycline treatments in HCT-116 KRAS shRNA cells, the expression of KRAS was decreased (Figure 4a). Meanwhile, Snail2 mRNA expression was markedly reduced on day 6 after induction of KRAS shRNA, whereas expression of Snail1 was not significantly altered (Figure 2j).

Figure 4.

Knockdown of KRAS or Snail2 promotes reversal of EMT. (a) Microarray data show upregulation of epithelial markers (CDH1, OCLN, CLDN3) and downregulation of mesenchymal markers (VCAN, EFNB2) following KRAS knockdown in HCT-116 cells. (b) Immunofluorescence staining of E-cadherin in HCT-116 KRAS shRNA cells treated without or with doxycycline (25 ng/ml) for 1 week (upper panel). Lower panel: phase contrast pictures. (c, d) RAS pathway activation leads to increased Snail2 expression and reduced E-cadherin expression, and Snail2 knockdown promotes increased expression of E-cadherin. HKe-3 ER:HRAS V12 cells were transfected with siRNA against Snail2. After 24 h, cells were treated with 5 or 10 nm OHT for 48 h before being collected and measured by quantitative reverse transcriptase PCR, normalized to EtOH-treated cells. Error bars represent the s.d. of triplicates. (e–h) Snail2 knockdown enhances E-cadherin expression. (e) Immunofluorescence staining of E-cadherin in HCT-116 Snail2 shRNA cells treated without or with doxycycline (25 ng/ml) for 1 week (upper panel). Lower panel: phase contrast pictures. (f) Immunofluorescence staining of E-cadherin in HCT-116 cells transfected with siRNA against Snail2 or Snail1 for 72 h (upper panel). Lower panel: phase contrast pictures. (g) Western blot analysis of lysates from HCT-116/HKe-3/HKh-2 cells expressing an inducible Snail2 shRNA after doxycycline addition showing effects on E-cadherin. (h) Western blot analysis of lysates from SW48 wt/KRAS G13D cells transfected with RISC free or Snail2 siRNA, showing effects on E-cadherin.

Overall, these data suggest that the RAS pathway regulates Snail2 expression and also the expression of Snail1, although to a much lesser extent. Induction of Snail2 expression by RAS pathway activation can be blocked by inhibition of either the PI3K or mitogen-activated protein kinase pathway.

RAS pathway activation leads to EMT

As expression of activated mutant RAS in the HCT-116 cell system leads to elevated expression of Snail2 and Snail1, both very important mediators of the transdifferentiation process known as EMT (Hemavathy et al., 2000; Nieto, 2002), we next sought to determine whether these cells expressing mutant RAS have undergone EMT.

To test this hypothesis, we performed microarray analysis in HCT-116/HKe-3/HKh-2 cells. We found HCT-116 cells have increased levels of mesenchymal markers such as VCAN, EFNB2, and lower levels of epithelial markers such as CDH1, OCLN and CLDN3, compared with HKe-3 and HKh-2 cells (Figure 3a), suggesting that HCT-116 cells have undergone at least a partial EMT relative to their mutant KRAS-deleted derivatives. This point was also confirmed by E-cadherin staining in HCT-116/HKe-3/HKh-2 cells. The alternative EMT inducers, ZEB2, Twist and E47, do not show significant differences in expression under these different conditions (data not shown). As shown in Figure 3b, there was much less E-cadherin membrane staining in HCT-116 cells compared with HKe-3 and HKh-2 cells. This was also the case in SW48 KRAS G13D, relative to wt SW48 (Figure 3c). In HKe-3 ER:HRAS V12 cells, RAS pathway activation by adding OHT resulted in a marked reduction in E-cadherin and a scattered cellular phenotype (Figures 3c and d). These results were consistent with a recent microarray analysis in a human colon cancer cell system (Joyce et al., 2009), showing that human colon cancer cells with mutant RAS have a molecular signature for EMT that is reflected in many markers such as E-cadherin and Snail2.

Figure 3.

RAS pathway activation leads to EMT process. (a) Microarray data show higher levels of mesenchymal markers (ZEB1, VCAN, EFNB2), lower levels of epithelial markers (CDH1, OCLN, CLDN3) and KRAS expression in HCT-116 cells compared with HKe-3 and HKh-2 cells. (b) Immunofluorescence staining of E-cadherin in HCT-116/HKe-3/HKh-2 cells (upper panel). Lower panel: phase contrast pictures. (c) Western blot analysis of lysates from SW48 wt and SW48 KRAS G13D cells or from HKe-3 ER:HRAS V12 cells after OHT treatment, showing effects on E-cadherin, occludin, ZEB1 and activation status of ERK. p, phospho. (d) Immunofluorescence staining of HRAS and E-cadherin in HKe-3 ER:HRAS V12 cells treated with or without OHT. (e, f) E-cadherin is regulated by the RAS pathway in a dose-dependent manner. HKe-3 ER:HRAS V12 cells were treated with different concentrations of OHT, and E-cadherin expression was measured by quantitative RT–PCR (e) or western blot analysis (f) after 3 days of OHT treatment. Error bars represent the s.d. of triplicates. (g, h) E-cadherin is regulated by the RAS pathway in a time-dependent manner. HKe-3 ER:HRAS V12 cells were treated with OHT, and E-cadherin expression was measured by quantitative RT–PCR (g) or by western blot analysis (h) at indicated time points after OHT treatment. Error bars represent the s.d. of triplicates. In f and h, for western blot analysis, the data are representative of three independent experiments.

This point was further strengthened using the ER:HRAS V12 system in HKe-3 cells. E-cadherin is an important component of the epithelial phenotype and downregulation of E-cadherin is of direct relevance to EMT (Thiery and Sleeman, 2006). Throughout the course of activated RAS induction in these cells (confirmed by increased phosphorylation of ERK and AKT, Figures 3f and h), E-cadherin mRNA expression and protein level were both strongly decreased in a dose-dependent and time-dependent manner (Figures 3e-h). After 3 days of 50 nm OHT treatment, the mRNA level of E-cadherin was reduced to only 20% compared with mock treatment, whereas the protein level was reduced to less than half.

This was also confirmed by knocking down KRAS in HCT-116 cells. After doxycycline treatment of HCT-116 KRAS shRNA cells to inhibit KRAS expression, microarray data showed that epithelial markers (such as CDH1, OCLN and CLDN3) were upregulated, whereas some mesenchymal markers (such as VCAN and EFNB2, although not ZEB1) were reduced (Figure 4a), suggesting that HCT-116 cells underwent mesenchymal–epithelial transition after KRAS knockdown. This was further supported by E-cadherin staining in HCT-116 KRAS shRNA cells treated with or without doxycycline (Figure 4b).

As RAS pathway activation leads to massive upregulation of Snail2 expression with simultaneous reduction of E-cadherin levels (Figures 2 and 3), we attempted to knock down Snail2 expression in HKe-3 ER:RAS cells treated with OHT. We found that even at very low concentrations of OHT, Snail2 mRNA level was still strongly elevated, whereas E-cadherin mRNA expression was reduced to half (Figures 4c and d). These data further confirmed that RAS pathway activation increases Snail2 expression and leads to EMT. The Snail2 siRNA oligo caused a significant decrease in Snail2 mRNA level (Figure 4c), and Snail2 knockdown allowed upregulation of E-cadherin mRNA, despite activated mutant RAS expression (Figure 4d). Both HCT-116 and SW48 KRAS G13D cells express mutant RAS and have less E-cadherin compared with their isogenic counterparts with wt RAS. Knockdown of Snail2 expression caused a return of E-cadherin protein expression in both HCT116 and SW48 KRAS G13D cells (Figures 4e-h), whereas knockdown of Snail1 expression in HCT-116 cells did not have much effect on E-cadherin expression (Figure 4f). Snail2 knockdown had no effect on KRAS expression or activation, or the phosphorylation state of ERK or Akt in HCT-116 or SW48 cell systems (data not shown).

Our above data confirm the hypothesis that some carcinoma cell lines with mutant RAS have undergone at least a partial EMT, and knockdown of Snail2 expression reverses the mesenchymal status.

Cells that have undergone EMT require continued Snail2 and Snail1 expression

Using the Oncomine research online tool (www.oncomine.com; Supplementary Figure S2), we found an inverse correlation between Snail2 expression and E-cadherin expression in many types of cancer. In breast cancer cell lines, BT-549 and Hs-578-T cells have undergone EMT as shown by reduced expression of E-cadherin and higher abundance of Snail2 expression compared with MCF7 and T47D cells (Figures 5a-c), consistent with previous reports (Hajra et al., 2002; Come et al., 2006). Of these lines, only Hs-578-T cells carry an activating mutation in a RAS oncogene (HRAS). Expression of Snail2, rather than that of Snail1, was strongly correlated with the reduced expression of E-cadherin (Figures 5a-c), implying that Snail2 is more likely to be a significant in vivo repressor of E-cadherin transcription in breast cancer.

Figure 5.

Cells that have undergone EMT have high Snail2 and Snail1 expression and are sensitive to their knockdown. (a) mRNA expression of E-cadherin, Snail1 and Snail2 in four breast cancer cell lines. Error bars indicate s.d. (b) Protein expression of Snail1 and Snail2 in four breast cancer cell lines. (c) Immunofluorescence staining of E-cadherin in four breast cancer cell lines (upper panel). Lower panel: phase contrast pictures. (d, e) Snail2 or Snail1 knockdown is more toxic to BT-549 and Hs-578-T cells than to MCF-7 and T47D cells. siRNAs against Snail2 or Snail1 were transfected into these cell lines and cell viability (d) or apoptosis (e) was measured 4 days after transfection. Error bars indicate s.d. (f) Expression of E-cadherin, Snail1 and Snail2 in SW480 and SW620 cells. Error bars indicate s.d. (g) Snail2 or Snail1 knockdown is more toxic to SW620 cells. siRNAs against Snail2 or Snail1 were transfected into both lines and cell viability was measured. *P<0.05; **P<0.01; error bars indicate s.d. (h, i) TGF-β induces EMT in H358 cells and increases the expression of Snail2 and Snail1. Expression of E-cadherin, Snail1 and Snail2 was measured after TGF-β treatment at indicated time points (h). Error bars indicate s.d. Western blot analysis with the indicated antibodies of lysates collected from H358 cells treated with or without TGF-β at indicated time points. (j) Snail2 or Snail1 knockdown is more toxic to H358 cells treated with TGF-β. H358 cells were treated with hTGF-β for 10 days, then transfected with siRNAs against Snail2 or Snail1 for 3 days. Cell viability was measured thereafter. *P<0.05; **P<0.01; error bars indicate s.d.

To investigate whether cells that have undergone EMT are more sensitive to Snail2 or Snail1 knockdown, we introduced siRNAs against Snail2 or Snail1 into these four breast cancer cells. SiRNA-induced knockdown of Snail1 or Snail2 (Supplementary Figures S1g and h) significantly killed BT-549 and Hs-578-T cells. Cell viability was reduced to around 50% in both cell lines, with two independent siRNA oligos against Snail1 or Snail2 (Figure 5d). We also observed more apoptosis after Snail1 or Snail2 knockdown in both cell lines, as shown by caspase-3 assays (Figure 5e). In contrast, there were no significant effects of Snail1 or Snail2 knockdown on cell viability or apoptosis in MCF7 or T47D cells. These data confirmed that breast cancer cells that have undergone some degree of EMT are more sensitive to Snail2 or Snail1 knockdown than primarily epithelial breast cancer cells.

This conclusion was further strengthened by the study of a pair of KRAS mutant colon cancer cell lines, SW480 and SW620, derived from the same patient; SW480 was from the primary tumour site and SW620 was from a metastatic site. The metastatic tumour line SW620 has a more mesenchymal phenotype as shown by a lower E-cadherin expression and a higher level of Snail2 and Snail1 than the primary tumour line SW480 (Figure 5f). SW620 cells were more sensitive to Snail1 or Snail2 knockdown than SW480 cells (Figure 5g).

To show a causal relationship between Snail2 or Snail1 dependency and the mesenchymal phenotype, we tested the possibility that induction of EMT could induce Snail1 or Snail2 dependency. Exposure of KRAS mutant H358 cells to TGF-β1, a promoter of EMT, over 2 weeks resulted in a marked reduction of E-cadherin (at both mRNA and protein level), coupled with an increased expression of mesenchymal markers ZEB1 and vimentin (Figures 5h and i), and a scattered cellular phenotype (data not shown). These cells are typical in showing that RAS mutation alone is not sufficient to induce EMT, but that it requires coordinate activation of other pathways, such as TGF-β signalling. We observed a marked increase in Snail2 expression after TGF-β1 treatment in H358 cells and also a moderate increase in Snail1 expression (Figure 5h), suggesting potential roles of Snail2 and/or Snail1 in TGF-β1-induced EMT in H358 cells. Whereas parental H358 cells seemed resistant to Snail2 or Snail1 knockdown, mesenchymal H358 human TGF-β cells responded to the siRNAs against Snail2 or Snail1 with reduction in viability (Figure 5j; P< 0.05).

Our above data indicate that cells that have undergone EMT in part because of RAS mutation or other pathways have relatively high Snail2 or Snail1 expression and are sensitive to their knockdown.

Synergistic loss of viability due to Snail2 or Snail1 knockdown and DNA-damaging drug treatment

We next tried to investigate whether there are synergistic effects between Snail2 or Snail1 knockdown and treatment with the DNA-damaging anthracycline drug, doxorubicin.

As shown in Figures 6a and b, breast cancer cell lines BT-549 and Hs-578-T, which have a mesenchymal phenotype, were relatively resistant to doxorubicin treatment, whereas two other breast cancer cell lines with an epithelial phenotype, MCF-7 and T47D, were sensitive. Cell viability of MCF-7 and T47D was reduced to a minimum of around 50% at the indicated concentrations of doxorubicin (Figure 6a), and apoptosis was induced in these two lines (Figure 6b). There were no significant effects after doxorubicin treatment in BT-549 and Hs-578-T cells.

Figure 6.

Synergistic effects between Snail2 or Snail1 knockdown and doxorubicin treatment. (a, b) BT-549 and Hs-578-T cells are relatively resistant to doxorubicin treatments. Cell lines were treated with doxorubicin at indicated concentrations for 72 h and subjected to cell viability (a) and apoptosis assays (b). Error bars represent the s.d. of triplicate measurements. (c, d) Knockdown of Snail2 or Snail1 sensitizes BT-549 cells to doxorubicin treatments. BT-549 cells were treated 24 h after being transfected with siRNAs against Snail2 or Snail1 with the indicated concentration of doxorubicin for 48 h and subjected to cell viability (c) and apoptosis assays (d). Error bars represent the s.d. of triplicate measurements. (e, f) Knockdown of Snail2 or Snail1 sensitizes Hs-578-T cells to doxorubicin treatments. Hs-578-T cells were treated for 24 h after being transfected with siRNAs against Snail2 or Snail1 with the indicated concentration of doxorubicin for 48 h and subjected to cell viability (e) and apoptosis assays (f). Error bars represent the s.d. of triplicate measurements.

To determine the effects of Snail2 or Snail1 knockdown in combination with doxorubicin treatment, siRNAs against Snail2 or Snail1 were transfected into BT-549 or Hs-578-T cells and, 24 h later, cells were treated with doxorubicin. Without Snail2 or Snail1 knockdown, no significant killing effects were observed in BT-549 cells after doxorubicin treatment at 90 or 180 nm (Figures 6c and d). We found that transfections of two independent siRNA oligos against Snail2 or Snail1, in combination with doxorubicin treatment, led to further reduced cell viability and more induction of apoptosis (Figures 6c and d). Similar results were observed in Hs-578-T cells (Figures 6e and f).

Thus, breast tumour cell lines that have undergone EMT and are of mesenchymal phenotype are relatively resistance to doxorubicin treatment; however, knockdown of Snail2 or Snail1 significantly sensitizes them to treatment with this drug, to a level comparable with that seen for breast cancer cell lines with a more epithelial phenotype.

Discussion

Functional RNA interference cell-based screens for genes that are differentially required, dependent on a particular genotype or phenotype, have been performed recently by several groups (Ngo et al., 2006; Shaffer et al., 2008; Luo et al., 2009; Scholl et al., 2009). The primary screen in this study was similar in design to that of Luo et al., using a pooled shRNA library and an isogenic colon cancer cell line pair differing only in the presence or absence of an activated mutant KRAS allele. shRNAs that were selectively toxic to KRAS mutant cells were identified by comparative hybridization on custom microarrays. However, for reasons that are not immediately obvious, the hits identified in our screen and that of Luo et al. did not overlap.

The strongest hit that we identified in this screen, as selectively required for the survival of mutant KRAS, but not KRAS wt, colon cancer cells, was the zinc finger transcriptional repressor Snail2, which is encoded by the SNAI2 gene. Snail2 belongs to the Snail superfamily of zinc finger transcriptional repressors. It is an important regulator of EMT and has a significant role in the malignant progression of cancer (Cobaleda et al., 2007; Peinado et al., 2007; Alves et al., 2009). Snail2-overexpressing mice develop mesenchymal tumours, whereas Snail2-deficient mice are resistant to BCR-ABL-induced leukaemogenesis (Perez-Mancera et al., 2005). Snail2 overexpression has been described in a wide spectrum of human cancers including breast cancer, colorectal cancer, gastric carcinoma, oesophageal carcinoma and other tumour types and seems to be important for tumour progression toward invasion and metastasis, in most cases involving E-cadherin downregulation (Cobaleda et al., 2007; Peinado et al., 2007; Alves et al., 2009). Snail2 expression has been reported to be important for the metastatic spread of RAS mutant melanoma cells in a mouse model (Gupta et al., 2005) and to be an independent prognostic marker for poor survival in colorectal cancer patients (Shioiri et al., 2006).

Our observations indicate a relationship between RAS signalling and Snail2 expression, with a downstream effect on the EMT state of the cell. Previous studies have shown that activated RAS can increase Snail2 expression in rat intestinal epithelial cells (Schmidt et al., 2005) and that Snail2 expression can be induced by multiple RAS effector pathways, by Raf and ERK (Conacci-Sorrell et al., 2003; Wang et al., 2007) and also by AKT (Saegusa et al., 2009). Snail2 knockdown does have a relatively modest effect on E-cadherin mRNA, even in cells without RAS activation, suggesting that there is at least some activity of Snail2 independent of elevated RAS signalling. It is possible that this could be due to basal, growth factor-stimulated RAS signalling, or signalling from other pathways. We find that expression of the related transcriptional regulator Snail1 is also induced by RAS, but to a much lesser extent compared with Snail2.

Although we find that RAS activation can clearly induce Snail2 expression in a given cell system, there is no evidence of a general correlation between RAS mutational status and Snail2 expression levels in tumours and cancer cell lines. This is likely due to the complexity of the regulation of Snail2 expression, with several pathways contributing to its control, including TGF-β, Wnt and p53 (Wu et al., 2005; Cobaleda et al., 2007; Peinado et al., 2007; Wang et al., 2009). It is well established that activation of RAS in epithelial cells can cooperate with other pathways, such as TGF-β, to lead to EMT (Oft et al., 1996; Lehmann et al., 2000; Huber et al., 2005; Larue and Bellacosa, 2005; Thiery and Sleeman, 2006). We find here that Snail2 is an important factor in the ability of RAS pathway activation in promoting EMT, with knockdown of Snail2 expression at least partially reversing the mesenchymal phenotype.

The appearance of Snail2 as the strongest hit in our KRAS synthetic lethal screen and its critical role in EMT, in which RAS has a part, suggests that the lethality associated with inhibition of Snail2 in HCT-116 cells, but not in their mutant KRAS-deleted derivatives, may be linked to forcing a reversal of EMT on cells that are at least part of the way toward transdifferentiation from epithelial to mesenchymal phenotypes. Targeting Snail2 is selectively damaging to RAS mutant cells with a more mesenchymal, rather than epithelial, phenotype. It has recently been found that the requirement of RAS mutant cancer cell lines for continued expression of the RAS oncogene, so-called oncogene addiction, is restricted to cells with an epithelial phenotype (Singh et al., 2009). By contrast, mesenchymal cells with mutant RAS do not show RAS oncogene addiction, although this may reflect a broader resistance to induction of cell death associated with the mesenchymal state; for example, the sensitivity of non-small-cell lung cancer cell lines that are wt for EGF receptor to EGF receptor inhibitors correlates with their differentiation state, with mesenchymal cell lines being resistant (Thomson et al., 2005).

Snail1 knockdown by siRNA in breast carcinoma cells has a somewhat greater effect on the induction of cell death than does the knockdown of Snail2. In contrast, in the colon cancer line SW620 and in the lung cancer line H358, Snail2 knockdown has the same effect as Snail1 knockdown on induction of cell death, possibly reflecting a tissue-type difference. It is certainly established that Snail1 provides an important part of the signal leading to the establishment of the mesenchymal phenotype. However, expression of Snail2 is markedly more responsive to input from RAS signalling pathways, suggesting that it may be more important than Snail1 in the context of RAS mutation, as in colon and lung cancers with a mesenchymal phenotype, than in tumour types in which RAS mutation is rare, such as breast.

The significance of finding that more mesenchymal RAS mutant tumour cells are dependent on continued Snail2 expression lies in the fact that it is precisely these cells that are most resistant to existing therapies. Their lack of RAS oncogene addiction suggests that they would not respond to the RAS pathway inhibitors currently in clinical trials, such as RAF, MEK, PI3-kinase and Akt inhibitors. In addition to causing loss of viability in RAS mutant mesenchymal cells, Snail2 knockdown increases the sensitivity of these cells to cytotoxic drugs. Snail2 inhibition has previously been shown to sensitize cells to DNA-damaging agents, in part by directing the p53 response away from apoptosis (Inoue et al., 2002; Perez-Losada et al., 2003; Kajita et al., 2004; Wu et al., 2005; Vannini et al., 2007; Vitali et al., 2008; Kurrey et al., 2009). Although the function of Snail2 as a transcriptional regulator does not make it an attractive target for pharmacological targeting, the data presented here suggest that its inhibition could be a good way of targeting some of the most intractable tumour types, those with a mutant RAS oncogene and a mesenchymal phenotype. The relatively mild phenotype of Snail2 knockout mice (Jiang et al., 1998; Inoue et al., 2002) suggests that on-target inhibition of this factor may also be relatively free from unwanted side effects.

Materials and methods

Molecular biology

For cellular growth curves, cells were seeded in six-well plates and cell number was measured using a Vi-CELL Cell Viability Analyzer (Beckman Coulter, Brea, CA, USA). For cell viability assays, cells were seeded in 96-well plates and cell number was measured using CellTiter-Blue (Promega, Madison, WI, USA). Apoptosis was measured using the Apo-ONE Homogeneous Caspase-3/7 Assay (Promega). Caspase-3/7 activity was normalized to relative cell number using the CellTiter-Blue assay. For anchorage-independent colony assays, cells were seeded in each well of a six-well plate in media containing 0.35% low-melting point agarose and colonies were counted 2 weeks later with Giemsa stain.

Quantitative reverse transcription–PCR

Real-time reverse transcriptase–PCR was performed using gene-specific primers (QuantiTect Primer Assays) for Snail2, Snail1 or GAPDH with the FastLane Cell SYBR Green Kit (Qiagen, Hilden, Germany). Relative transcript levels of target genes were normalized to GAPDH mRNA levels.

Immunoblot analysis

For western blot analysis, cells were lysed in NuPage LDS sample buffer (Invitrogen, Carlsbad, CA, USA). Proteins were resolved by electrophoresis on NuPage 4–12% Bis-Tris gels (Invitrogen) and immunoblots were developed using the ECL Western Blotting Detection Reagent (Amersham, Munich, Germany). The following antibodies were used: anti-E-cadherin, anti-ZEB1 and anti-KRAS, anti-lamin B (Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-occludin (Invitrogen), anti-GAPDH (Abcam, Cambridge, UK), anti-Snail1, anti-Snail2, anti-phospho (Ser473)-AKT, anti-phospho (Thr202/Tyr204)-ERK and ERK (Cell Signaling Technology, Danvers, MA, USA).

Immunofluorescence

Cells were fixed in 4% phosphate-buffered saline-paraformaldehyde for 15 min, incubated in 0.2% Triton-X-100 for 5 min, then in 0.2% fish skin gelatine in phosphate-buffered saline for 10 min and stained for 1 h with antibodies against E-cadherin (1:100; Santa Cruz Biotechnology) or HRAS (1:50; Calbiochem, San Diego, CA, USA). Cells were then stained with the secondary antibody, followed by visualization under a fluorescence microscope.

Further materials and methods are described in the Supplementary Material.