Suppressor of cytokine signaling 1 modulates invasion and metastatic potential of colorectal cancer cells

Abstract

Suppressor of cytokine signaling (SOCS) 1 is an inducible negative regulator of cytokine signaling but its role in human cancer is not completely established. Here we report that, while SOCS1 is expressed in normal colonic epithelium and colon adenocarcinomas, its level decreases during progression of colon adenocarcinomas, the lowest level being found in the most aggressive stage and least differentiated carcinomas. Forced expression of SOCS1 in metastatic colorectal SW620 cells reverses many characteristics of Epithelial–Mesenchymal Transition (EMT), as highlighted by the disappearance of the transcription factor ZEB1 and the mesenchymal form of p120ctn and the re‐expression of E‐cadherin. Furthermore, miRNA profiling indicated that SOCS1 also up‐regulates the expression of the mir‐200 family of miRNAs, which can promote the mesenchymal–epithelial transition and reduce tumor cell migration. Accordingly, overexpression of SOCS1 induced cell morphology changes and dramatically reduced tumor cell invasion in vitro. When injected in nude mice, SOCS1‐expressing SW620 cells induced metastases in a smaller number of animals than parental SW620 cells, and did not generate any adrenal gland or bone metastasis. Overall, our results suggest that SOCS1 controls metastatic progression of colorectal tumors by preventing the mesenchymal–epithelial transition (MET), including E‐cadherin expression. This pathway may be associated with survival to colorectal cancer by reducing the capacity of generating metastases.

Abbreviations used in this paper

CRC

ColoRectal Carcinoma

EMT

Epithelial–Mesenchymal Transition

Ets

Erythroblastosis virus E26 Transforming Sequences

HES

Hemalum Eosin Safran

IHC

ImmunoHistoChemistry

MET

Mesenchymal–epithelial transition

SH2

Src Homology

Sp2

Specificity proteins 2

SOCS

Suppressor of Cytokine Signaling

STAT

Signal Transducers and Activators of Transcription

ZEB1

Zinc finger E‐box binding homeobox‐Box 1.

1. Introduction

The suppressors of cytokine signaling (SOCS) belong to a family of adaptor proteins that negatively regulate cellular signaling. The SOCS proteins contain a central SH2 domain that interacts with phosphorylated tyrosines (e.g. tyrosines residues located in the intracellular region of cytokine receptors, which become phosphorylated upon cell stimulation). In line with this notion, SOCS1 was originally described as an inhibitor of the signaling mediated by the cytoplasmic tyrosine kinases JAK (Janus kinase) (for reviews see Ilangumaran and Rottapel, 2003; Valentino and Pierre, 2006). In addition, the SOCS proteins contain a C‐terminal SOCS box that interacts with regulators of the protein ubiquitylation machinery, Elongin B/C and Cullin5 (Kamura et al., 2004; Mahrour et al., 2008). Accordingly, SOCS1 plays a critical role in the proteasomal degradation of its binding partners (Zhang et al., 1999). Through these two modes of action, SOCS1 participates in the regulation of multiple cell functions. The neonatal lethality observed in socs1 knockout mice (Alexander et al., 1999) highlights the physiological importance of this protein.

SOCS1 expression is tightly regulated by several mechanisms. SOCS1 is primarily regulated at the transcriptional level. Indeed, upon cytokine cell stimulation, active JAK kinases and downstream Signal Transducers and Activators of Transcription (STAT) proteins induce the expression of SOCS genes. Besides, the SOCS1 promoter is actively repressed by the nuclear proteins GFI‐1B (Jegalian and Wu, 2002), Ets‐1 (Travagli et al., 2004) and Sp2 (Letourneur et al., 2009). The SOCS1 gene can also be regulated by methylation of the CpG islands located in the translated exon 2 (Yoshikawa et al., 2001). Stabilization of SOCS1 proteins by proteasome inhibitors suggests that cells may also regulate SOCS1 levels through the proteasome pathway (Zhang et al., 1999).

Expression of SOCS1 is often de‐regulated in cancer cells. The SOCS1 gene is methylated in human hepatocellular carcinoma (Yoshida et al., 2004; Yoshikawa et al., 2001) or colorectal tumors (Fujitake et al., 2004; Lin et al., 2004). In colon cancer, SOCS1 gene methylation is mainly associated with the CpG island methylator phenotype (CIMP) subclass (Shen et al., 2007), and represents one of the nine high‐ranking CIMP‐predicting markers (Weisenberger et al., 2006). Besides, analysis of genetic variations in colorectal cancers indicated that single nucleotide polymorphisms (SNPs) in genes belonging to the JAK/STAT signaling pathway, including two in the SOCS1 gene, were associated with cancer survival (Slattery et al., 2013).

Beyond being mere diagnostic/prognostic markers, genetic and epigenetic events occurring at the SOCS1 locus may also have a functional relevance in cancers. On one hand, several studies have characterized SOCS1 as a tumor suppressor gene. socs1 ^−/− (Tg) mice, in which lethality conferred by the lack of SOCS1 is bypassed through expression of ectopic SOCS1 in T and B cells, spontaneously developed colorectal carcinomas at 6 months of age (Hanada et al., 2006). Of note, socs1 ^−/− Tg mice treated with anti‐IFN‐γ antibody did not develop such tumors, suggesting that chronic inflammation was a critical determinant for the development of these colorectal tumors. In vitro studies with human hepatocellular carcinoma‐derived cell lines indicated that restoration of SOCS1 expression leads to growth‐suppressing activity (Yoshikawa et al., 2001). Similarly, SOCS1 expression was found to be reduced in melanoma‐derived tissues that had metastasized to the brain, as compared to the corresponding melanoma tumors (Huang et al., 2008). Thus, these studies suggested that SOCS1 has an important, protective role in cancer development. On the other hand, SOCS1 has been also identified as a progression marker of human melanoma. Indeed, in human melanoma, Li et al. (2004) indicated that the level of SOCS1 protein correlates with tumor invasion and stage of diseases. Specifically, melanocytes of the normal skin or melanocytic nevi lack SOCS1 protein expression while melanoma cells express SOCS1 (Li et al., 2004). The later finding suggests that the exact role of SOCS1 in tumor progression is likely to depend on the type of cancer considered.

Here we addressed the role of SOCS1 in colorectal cancer (CRC). We report that while SOCS1 is well expressed in CRC, its expression level decreases with the aggressiveness of the tumors. By manipulating SOCS1 levels in colorectal tumor cells, we show that this protein controls the EMT process, and the tumor cell invasiveness and metastatic potential.

2. Material and methods

2.1. Plasmids

The pcDNA3‐myc‐mSocs1‐expressing plasmid is a generous gift from Yasukawa et al. (1999). Site‐specific mutations in this plasmid were performed using the quick change site‐directed mutagenesis kit (Stratagene) and the oligonucleotides described in Table S1. For the ΔSOCS‐box mutant, a stop codon was introduced at position 760 (NM_001271603.1), just before the SOCS‐box‐encoding sequence. Those plasmids were used for SW620 cell transfection. For retroviral infection of SW620, HT29 or HCT116 cells, the mutated SOCS1‐encoding sequences were subcloned into the pBabe vector in‐between the BamH1 and XhoI restriction sites. For each construct, the SOCS‐encoding region and the cloning junctions were verified by sequencing.

Silencing of ectopic SOCS1 expression was performed through lentiviral infection using a PLKO vector (TRCN0000067419) from OpenBioSystems, in which the targeting sequence of the short hairpin (mature antisense) was CGCATCCCTCTTAACCCGGTA. Silencing of CDH1 expression was performed using GIPZ lentiviral CDH1 ShRNA (CDH1 ShRNA‐1‐GFP: #V3LHS_346823: AAAATTTCCAATTTCATCG; CDH1 ShRNA‐2‐GFP: #V2LHS_14834: ATAATAAAGACACCAACAG) from OpenBioSystems. The pMD2.G (VSV‐G) and pCMVdeltaR8.74 plasmids, as well as the lentiviral construct carrying an expression cassette containing the firefly luciferase and the green fluorescent protein (GFP) genes separated by an internal ribosomal entry site (pMEGIX‐Luc plasmid), were kindly provided by Dr J.L. Villeval (UMR 1009; Villejuif, France) (Morgenstern and Land, 1990).

2.2. Viruses production and transduction of cells

Replication‐defective lentiviral particles were produced by transient transfection of 293T cells with the above‐mentioned plasmids, together with the pMD2.G (VSV‐G) and pCMVdeltaR8.74 plasmids, according to standard protocols (Naldini et al., 1996). The day before transduction, cells were plated at low cell density (30,000 cells/well) in a 24‐well dish. Viruses were added to cells for 24 h. To obtain stable cell lines in a polyclonal background, the GFP‐positive luciferase‐expressing cells (when applicable) were sorted for green fluorescence using a MoFlo cell sorter (Beckman–Coulter, Miami, FL) and maintained under selective pressure in the presence of puromycin (1 μg/mL). Retroviral infections were performed as previously described (Sirvent et al., 2007) and stable cell lines were obtained by puromycin selection as above.

2.3. Cell culture, transfections, growth and invasion assays

The human SW620 (ATCC^® CCL‐227), SW480 (ATCC^® CCL‐228), HT29 (ATCC^® HTB‐38) or HCT116 cells (ATCC^® CCL‐247) were cultured in Dulbecco's modified Eagle's medium supplemented with antibiotics (50 μg/mL penicillin, 50 μg/mL streptomycin), 1 mM sodium pyruvate, and 10% fetal calf serum. For stable transfection experiments, SW620 cells were transfected using JET‐PEI (Ozyme) with the pcDNA3‐myc‐mSocs1 plasmids and then cultured under 500 μg/mL G418 selective conditions for at least 1 month (Travagli et al., 2004). Two independent clones were selected and used for further experiments. For transient transfection, the p4xTSV luc (GAS‐reporter plasmid) (Moriggl et al., 1996) and the pSRα‐Renilla luc reporter (Seguin et al., 2009) were used as previously described (Travagli et al., 2004). For soft agar growth assays, 2 × 10⁴ cells were suspended in DMEM with 0.36% low melting agarose (Invitrogen) and 10% FCS and plated onto 1 mL/35 mm dish of DMEM with 0.6% agarose and 10% FCS. After 3 weeks, plates and individual clones were photographed and the number of colonies was determined in 3 independent experiments. For in vitro invasion assays, cells were seeded at 2 × 10⁵ per well in 24‐well matrigel precoated transwell filter plates (BD BioCoat Matrigel Invasion Chambers) in serum‐free medium. Serum containing medium (10%) was used as the chemo‐attractant in the lower chamber. After 24 h of incubation, cells that had invaded to the lower surface of the matrigel‐coated membrane were fixed with 100% methanol, stained with DAPI and whole‐well images were acquired using a Zeiss AxioImager Z1 fluorescence upright microscope (Zeiss, Jena, Germany). Images were processed using the Image J software. Data are expressed as mean ± SD of at least 3 independent experiments.

2.4. DNA isolation and methylation‐specific PCR

The methylation status of the SOCS1 gene was analyzed using the methyl detector kit (Active motif). DNA from SW480, SW620, HT29 or HCT116 colorectal cell lines was purified according to standard protocols. After bisulfite treatment of DNA, two nested PCR were performed for each sample. In both cases, the outer primers were used to amplify the exon 2 region of the SOCS1 gene (corresponding to the target region of methylation‐mediated silencing), whether methylated or not. In one case, the inner primers used were methylation‐specific primers that allowed amplification of a 170 bp‐long sequence, while in the other case the inner primers used hybridized only on unmethylated sequences and allowed amplification of a 190 bp‐long sequence. Conditions used for all these amplification reactions were: 5 min at 94 °C, followed by 35 cycles of 1 min at 94 °C, 1 min at 59 °C, 1 min at 72 °C, and final extension at 72 °C for 10 min. PCR fragments were then run on ethidium bromide‐containing agarose gels and analyzed under UV using a GelDoc system (BioRad, Hercules, CA).

2.5. Immunoblotting and immunofluorescence analysis

For immunoblotting, cells were lysed in a buffer containing 20 mM Tris–HCl, pH 7.5, 1% Triton X‐100, 150 mM NaCl, 1 mM PMSF, 10 mM NEM, 10 μg/mL aprotinin, 5 μg/mL leupeptin. Cellular extracts were separated through 8% SDS‐PAGE and processed for Western Blotting as described (Travagli et al., 2004). For immunofluorescence analysis, cells were seeded on poly‐Lysine (Sigma) coated cover slides for 24 h. Cells were then rinsed in PBS prior to fixation in 4% PFA, permeabilization in PBS–Triton X‐100 0.1%, saturation in PBS–BSA 5% and incubation with required antibodies. For labeling procedure, primary antibodies were visualized by incubation with Rhodamine Red™‐X Goat or Alexa Fluor^® 488 Goat conjugated anti species immunoglobulin G, as indicated. Cover slides were mounted onto slides with Dako (Dakocytomation) supplemented with 1% DAPI (to stain nuclei). Cells were examined under a Zeiss AxioImager Z1 fluorescence upright microscope (Zeiss, Jena, Germany) equipped with apotome and 63 X/1.40 oil M27 objective. Images were processed using the Image J software.

2.6. Immunohistochemical analyses

For human tissues: a total of 60 formalin‐fixed, paraffin‐embedded colon tissue samples (Colorectal cancer‐metastasis‐normal [CDA2] human tissue array) was obtained from Super Bio Chips. This array included 40 cores of colorectal cancer of different stages, 10 cores of metastatic lesions and 9 cores of normal colon tissue. Sections of paraffin‐embedded tissue specimens were stained with an anti‐SOCS1 antibody (1:40 dilution). Tissue sections were immunostained with nonspecific IgG and used as negative controls (not shown). A score of 0–3 that reflected the amount of SOCS1 protein detected by IHC was assigned to each colon cancer tissue sample. Samples with a score of 0 or 1 were categorized in the “low expression” group, while samples with a score of 2 or 3 were listed in the “high expression” group. Two investigators, while blinded to the clinical data, independently scored SOCS1 level in each sample. For xenograft histology: 4% PFA paraffin‐embedded, 4 μm‐sections of subcutaneous tumors or metastases were analyzed by standard immunohistochemical methods with specific antibodies or by HES (Hematoxylin–Eosin–Safran).

2.7. Human TissueScan Colon Cancer cDNA array

We used the human ‘TissueScan Colon Cancer cDNA Array I’ for in‐house analysis of SOCS1 mRNA expression by q‐PCR in 48 tissues covering different stages of colorectal cancers (Origene, HCRT101). The real‐time PCR reactions were carried out using an ABI PRISM 7000 detection system in the presence of 5% Me₂SO, as previously described (Travagli et al., 2004), using a SYBR green‐based protocol for the SOCS1 specific primers (SOCS1, Table S1) or a TaqMan probe‐based protocol for the 18S (Hs9999901_s1; Applied Biosystems). SOCS1 expression levels were normalized to those of the 18S using the ΔΔCt method. Data in Figure 1A are expressed as the log2 ratios between tumors and control tissues.

Figure 1.

SOCS1 expression inversely correlates with stages of human colorectal carcinomas. A. Transcript expression of SOCS1 in TissueScan Colon Cancer cDNA Array I. SOCS1 expression levels are expressed as the log2 ratios between tumors and control tissues. *P < 0.05, **P < 0.01, unpaired 2‐tailed Student's test, using Prism software. B. Paraffin‐embedded sections of colonic tumors samples were examined for SOCS1 expression by immunohistochemistry using a polyclonal rabbit anti‐SOCS1 antibody. Expression of SOCS1 was quantified as described in Materials and methods. The relationship between the SOCS1 immunoreactivity scores and the tumor stages is shown. C: Representative, high magnification pictures of SOCS1 immunostaining of different UICC stages in well‐(upper and middle) or moderately‐(lower) differentiated tissue sections are shown.

2.8. Xenograft assays in mice

Six‐week‐old female‐Hsd:athymic Nude‐nu/nu mice (Preclinical Evaluation Platform, Institut Gustave Roussy) were used for all the in vivo experiments. Procedures involving animals and their care were conducted in conformity with national and international laws and policies. For in vivo tumor growth assays, mice were subcutaneously injected in the right flank with 0.1 mL of a 50% Matrigel (BD Biosciences) solution diluted in complete media and containing 2 × 10⁶ cells of the various clones, as previously described (Marsaud et al., 2010). Tumor volumes based on caliper measurements were calculated by the ellipsoidal formula [1/2 (length² × width)]. For metastases formation assays, luc‐positive cells were injected either into the left ventricle of anesthetized Hsd:athymic Nude‐nu/nu mice (1 × 10⁶/0.1 mL in PBS) or in the tail vein (5 × 10⁶/0.2 mL in PBS). Mice were imaged for luciferase activity immediately after cell injection to control for successful xenograft, and once a week to monitor appearance of metastases. For bioluminescent imaging, mice were anesthetized and injected in peritoneum with 10 mL/kg of d‐luciferin (15 mg/mL in PBS). Imaging was completed 5 min after d‐luciferin injection, using a Xenogen IVIS system coupled to Living Image acquisition and analysis software. At the end of the experiment (week 5), mice were sacrificed and subjected to dissection for anatomical analyses of metastases.

2.9. Microarray experiments and bioinformatic analyses

Details are available in the Supplementary Materials and Methods file.

3. Theory

Lack of SOCS1 expression due to methylation of the SOCS1 gene is usually considered as a mere biomarker of the CIMP+ phenotype of colorectal cancer. Our study demonstrates that SOCS1 has a functional role in modulating the epithelial–mesenchymal transition as well as the invasive properties and metastatic potential of CRC cells.

4. Results

4.1. Association of SOCS1 with clinicopathological features of colon cancer

We performed q‐PCR to assess SOCS1 expression in a panel of 45 clinical colon cancer specimens representing different UICC stages of tumor progression. Aggressiveness of colon tumors correlated with a progressive reduction in SOCS1 mRNA expression. Indeed, tumors at stages II–IV expressed significantly lower levels of SOCS1 mRNA than stage I tumors (Figure 1A). To confirm that this trend in SOCS1 mRNA expression translated into differences at the protein level, we assessed SOCS1 expression by immunohistochemistry in primary and metastatic human CRC tumor biopsies from 50 individual patients using Tissue MicroArrays. No significant correlation between SOCS1 expression and gender and age was observed (Table S1). In contrast, SOCS1 protein levels decreased with the stage of UICC classification. While 37.5% of stage IV CRC lack detectable expression of SOCS1, high levels of SOCS1 were observed in 92% of stage II CRC (Figure 1B and Table S1). Moreover, SOCS1 immunoreactivity seemed reduced in metastases when compared to the primary tumor of the same patient (Table S1). Interestingly, we noted that SOCS1 expression levels were high in all of the well‐differentiated adenocarcinomas (N = 15), whereas only 11 out of 18 moderately‐differentiated tumors showed a strong signal (Figure 1C and Table S1), regardless of the UICC stage. This correlation suggested that the decrease in SOCS1 levels may be functionally implicated in the mechanisms controlling the structural integrity of the epithelium, in addition to be a molecular marker of CRC tumor progression.

4.2. SOCS1 expression promotes a mesenchymal–epithelial change in cell morphology, and increases cell invasion

To test whether SOCS1 participates in the mechanisms that control epithelial features of colon cells, we ectopically expressed SOCS1 in the SW620 colon cell line (SW620^SOCS). This CIMP+ cell line (Ahmed et al., 2013) is derived from a lymph node metastasis of a patient bearing a poorly differentiated, Dukes C, colorectal cancer. These cells have undergone the epithelial–mesenchymal transition (EMT) (Leibovitz et al., 1976) and do not express SOCS1 due to DNA methylation of the corresponding gene (Figure S1). We first confirmed that ectopic SOCS1 was active in these cells, as it strongly reduced IFN‐γ‐induced STAT1 phosphorylation and GAS promoter activity (Figure S2). Overexpressed SOCS1 did not change the cell proliferation rate indicating that SOCS1 does not regulate growth of these CRC cells (not shown). In contrast, SOCS1 induced a dramatic morphological change: SW620^SOCS displayed a clear epithelial morphology while controls cells [i.e. the parental, untransfected SW620 cells (“SW620”) or the SW620 cells transfected with the pcDNA3 control plasmid (“SW620 control”)] displayed a rounded, fibroblastic mesenchymal morphology (Figure 2A). The fact that this modification was accompanied by an increase in cell–cell contacts in SW620^SOCS cells prompted us to investigate the localization of E‐cadherin, a protein that displays a central role in organizing the cell–cell adhesion complexes typical of epithelial cells. While E‐cadherin membrane staining was absent in control cells (Figure 2B left hand side), it was readily detectable at sites of SW620^SOCS cell–cell contacts (Figure 2B right hand side). Collectively, these observations suggested that SOCS1 induced a morphological change that resembles a mesenchymal–epithelial transition (MET).

Figure 2.

Consequences of ectopic SOCS1 expression in the metastatic SW620 cells. A: Representative phase‐contrast microscopy images of parental, untransfected SW620 cells (“SW620”) or SW620 cells transfected with the pCDNA3 control plasmid (“SW620 control”) or SW620 cells transfected with the pCDNA3‐myc‐mSOCS1 plasmid (“SW620^SOCS”, clones #1 or #2) growing in monolayer cultures. B: Immunofluorescence microscopy images showing localizations of myc‐tagged SOCS1 (upper panels), E‐cadherin (middle panels) and both signals merged with DAPI‐labeled DNA (lower panels). C: Soft agar plates were photographed at 100× magnification. A representative colony is shown for each cell lines. D: Cell invasion assay was performed using 24‐well transwells coated with matrigel. 24 h after plating, cells that had invaded the coated membrane were fixed and counted. ***P < 0.001, unpaired 2‐tailed Student's test. SOCS1 levels in the different cell lines are shown.

Forced expression of SOCS1 did not modify anchorage‐independent growth ability of SW620 cells in soft agar as revealed by the number of colonies. However, cell colonies exhibited obvious morphological differences that were consistent with a specific function of SOCS1 in the control of a MET‐dependent process. Indeed, control SW620 cells aggregated only very loosely and cells were readily detectable at the periphery of the colonies, while SW620^SOCS cells formed tight, densely packed multicellular aggregates where single cells could not be discriminated (Figure 2C). This result suggested that SOCS1 diminish the ability of SW620 cells to scatter from each other. To directly test this hypothesis, we performed a Boyden Chamber cell invasion assay using matrigel matrix as the basement membrane. We observed a significant decrease in the invasive capacity of SW620^SOCS cells, as compared to that of control cells (Figure 2D). Efficient shRNA‐mediated silencing of ectopic SOCS1 (ShRNA‐2) restored cell behaviors similar to those of control cells, formally demonstrating that SOCS1 was, indeed, responsible for the decrease in invasion properties of SW620^SOCS cells. By retroviral infections, we also expressed SOCS1 in other CIMP+ colorectal cancer cell lines (Ahmed et al., 2013), in which the endogenous SOCS1 gene is inactivated by methylation (Figure S1). SOCS1 overexpression in HCT116 or HT29 cells also strongly diminished cell invasion in vitro (Figure 3D). Thus, SOCS1 effect on cell invasion is not restricted to SW620 cells and can be extended to other colon cancer cell lines.

Figure 3.

The SOCS1‐dependent inhibition of CRC cells invasion requires both its SH2 and SOCS‐box domains. A: Schematic drawings of SOCS1 showing the SH2, NLS and SOCS‐box domains and of the two mutants used in this study: the R 104 K mutant, or the mutant deleted of its SOCS box (Δ SOCS Box). B: Cell invasion assays were performed as described in Figure 2, using the SW620 (B.1), HCT116 (B.2) or HT29 (B.3) cell lines. *P < 0.05, **P < 0.01, unpaired 2‐tailed Student's test. SOCS1 levels in the various cell lines are shown. C: E‐cadherin localization was analyzed by immunofluorescence in the SW620 cell populations expressing either the wild‐type or the mutated forms of SOCS1.

4.3. Both the SH2 and SOCS‐box domains are required for SOCS1‐induced effects on colon cancer cell morphology and invasion potential

In order to get insights into the molecular mechanisms underlying the SOCS1‐induced inhibitory effects, we performed a structure–function analysis. An SH2 domain and a C‐terminal SOCS‐box domain structurally characterize the SOCS1 protein. Using site‐directed mutagenesis, we generated two previously described SOCS1 mutants (Nicholson et al., 1999) in which one or the other domain was non‐functional (Figure 3A) and expressed them in the SW620, HT29 or HCT116 cell lines by retroviral infection. This procedure allowed stable expression of SOCS1 mutants in a polyclonal background. Mutation of the SH2 domain (SOCS1 R/K) impairs SOCS1 ability to block STAT‐mediated cytokine signaling ((Nicholson et al., 1999) and Figure S2D). The SOCS1 mutant that lacks the SOCS box (SOCS1Δ SOCS box) is no longer able to induce degradation of its binding partners (Frantsve et al., 2001), but conserves its ability to inhibit cytokine signaling (Figure S2D and (Subramanian et al., 2005)). We found that, in contrast to the wild‐type protein, SOCS1 R/K and SOCS1Δ SOCS failed to affect cell invasion (Figure 3B) and to promote E‐cadherin membrane recruitment (as illustrated in Figure 3C using the SW620 cell line). We concluded that the SH2 and SOCS‐box domains are both required for the capacity of SOCS1 to mediate E‐cadherin membrane localization. The critical role of the SH2 domain strongly suggests the existence of a pTyr‐dependent signaling pathway involved in the capacity of SOCS1 to induce a MET. The critical role of the SOCS box known to promote E3 ubiquitin ligase activity also argues for a proteasomal pathway involved in this SOCS1‐dependent process.

4.4. Expression profiling of genes or miRNAs regulated by SOCS1

To investigate the molecular mechanisms underlying the SOCS1‐induced changes in colon cancer cell morphology and invasion properties, we analyzed the transcriptome of SW620 cells or SOCS1‐expressing SW620 cells through microarrays. Stringent data analysis revealed differential expression of 889 genes between SOCS1‐expressing cells (average between SW620^SOCS#1/SW620^SOCS#2) and the parental cell line (Table S3), as well as 96 other entities (LincRNA, pseudogenes, etc) without official gene symbols. In order to better understand the genomic response of SOCS1 in SW620 cells, the 889 differentially expressed genes were analyzed using the Ingenuity Pathway Analysis (IPA) tool. Functional analysis of the differentially expressed genes revealed a significant association with two diseases, referred to in the Ingenuity database as “Cancer” and “Gastrointestinal Diseases”. Thus, this first level of analysis indicated that these SOCS1‐induced variations in gene expression were highly relevant to the biological context addressed through our study. The top cellular functions that were found associated with our list of 889 differentially expressed genes were “Cellular Movement” (205 genes), “Cellular Growth and Proliferation” (266 genes), “Cell‐to‐cell Signaling and Interaction” (155 genes) and “Cell Morphology” (173 genes). Thus, with the exception of “Cell Growth and Proliferation”, this second level of analysis pointed out to cellular functions that were found regulated by SOCS1 in SW620 cells. Moreover, the search for overlaps between our list of SOCS1‐regulated genes and gene sets available from the MsigDB (Molecular Signatures Database; Broad Institute) (Subramanian et al., 2005) (category C2‐CGP: curated gene sets) revealed a significant overlap between the SOCS1‐regulated genes and the list of genes differentially expressed between the metastasis‐derived SW620 cells and the primary tumor‐derived SW480 cells from the same patient. Specifically, genes down‐regulated by SOCS1 in SW620 cells overlapped with genes expressed at higher levels in SW620 than in SW480 cells. Conversely, genes up‐regulated by SOCS1 in SW620 cells overlapped with genes expressed at lower levels in SW620 than in SW480 cells. Thus, forced expression of SOCS1 in the metastatic SW620 cells induced gene expression changes consistent with a partial reversion toward the transcriptome of primary tumor‐derived SW480 cells. To seek for effector molecules involved in SOCS1 biological activity, we used the “upstream regulators” application of the Ingenuity software. We found 34 regulatory molecules that were predicted to be activated or inhibited by SOCS1. Although systematic analysis of the activation state of these 34 molecules was beyond the scope of our study, we noted that one of them was also present in our list of differentially‐regulated genes, namely ZEB1, a transcriptional repressor that has been described as a crucial EMT inducer (Eger et al., 2005; Spaderna et al., 2008). Accordingly, ZEB1 protein level was dramatically reduced in the SW620^SOCS cells, as compared to SW620 cells (Figure 4A). This observation raised the hypothesis that ZEB1 may be an important target of SOCS1 in CRC cells that, given its known ability to act as a transcriptional repressor, may in turn control several of the differentially‐regulated genes we observed. Search for overlaps between our list of SOCS1‐regulated genes and gene sets available from the MsigDB (category C3‐TFT: motif gene sets of transcription factor targets) indicated that the list of SOCS1‐regulated genes was enriched in genes containing ZEB1 binding sites in their promoter region.103 SOCS1‐regulated genes displayed ZEB1 binding sites in their promoter region and/or were previously described as regulated by ZEB1 (Table S3). Those included CDH1 (encoding E‐cadherin), JUP (γ‐catenin), FN1 (fibronectin‐1), KRT18 (keratin 18) and ITGB4 (integrin beta 4). Expression of those potential ZEB1 target genes was therefore investigated in more detail (see below).

Figure 4.

SOCS1‐induced modulation of the expression of EMT markers. A: The level of SOCS1 is shown in the first lane. Mesenchymal markers [ZEB1, mesenchymal isoform of p120ctn (p120ctn 1), Fibronectin, Vimentin] and epithelial markers [E‐cadherin, Keratin 18 (K18), epithelial isoforms of p120ctn (p120ctn all)], were analyzed by Western Blotting. Levels of δ‐catenin, γ‐catenin and ITGB4 are also shown. Hsc70 was used as protein loading control. Immunofluorescence analyses of expression and localization of myc‐tagged SOCS1 and of (B) p120ctn isoform 1 (6H11 antibodies) or (C) p120ctn all isoforms in the SW620control and SW620SOCS cell lines are shown.

The fact that SOCS1 exerted only a modest effect on ZEB1 mRNA level (−2.49 fold), while it drastically decreased ZEB1 protein level, was puzzling. This apparent discrepancy suggested that, in addition to the capacity of SOCS1 to regulate transcription of the ZEB1 gene, another mechanism may be at play. Since ZEB1 can also be regulated by miRNAs (Bracken et al., 2008), we searched for SOCS1‐regulated miRNA expression in SW620 cells. miRNA profiling revealed 6 miRNAs that were regulated by SOCS1. We observed a down‐regulation of miR‐192 in SOCS1‐expressing cells, and an up‐regulation of the five members of the miR‐200 family (Table 1). The miR‐200a, ‐200b and ‐429 family members are located contiguously in a cluster on chromosome 1 and share a common promoter, while the other family members, miR‐200c and 141 are clustered on chromosome 12 (Bracken et al., 2008). Interrogation of MsigDB (category C3‐MIR: gene sets of miRNA targets) indicated statistically significant enrichments between the list of SOCS1‐regulated genes and the gene sets corresponding to targets of miR‐200a, ‐200b, ‐200c, ‐429 or ‐141, with 49 genes in the overlap (Table S3). Further analysis indicated that among the highest scored target genes for these miRNAs were ZEB1 and ZEB2 (http://www.microrna.org). In addition, mir‐200c was shown to decrease ZEB1 expression level (Hurteau et al., 2007). Thus, the remarkable decrease in ZEB1 protein level in SOCS1‐expressing CRC cells, in which the ZEB1 mRNA was only modestly modulated, may result from a miRNA‐mediated impairment of ZEB1 mRNA translation. Reciprocally, ZEB1 is also known to directly repress transcription of miR‐141 and ‐200c (Burk et al., 2008). Consistent with other studies (Bracken et al., 2008; Burk et al., 2008), these data suggest the existence of an amplifying feedback loop where decrease in ZEB1 protein level would alleviate the repression of miRNA expression (i.e. lead to an increase in miRNA expression), which in turn would decrease translation of the ZEB1 mRNA and ultimately result in even less ZEB1 protein expression. Whether SOCS1 expression triggers this ZEB1/miRNA feedback loop by directly targeting ZEB1 and/or miRNA expression remains to be determined.

Table 1.

SOCS1 regulates miRNA expression in SW620 cells.

	SW620 (T) intensity	SW620SOCS (S) intensity	Fold change (S/T)
Up‐regulated miRNAs
hsa‐miR‐141	52.29	405.60	7.75
hsa‐miR‐200c	45.74	106.52	2.32
hsa‐miR‐200b	515.42	1198.06	2.32
hsa‐miR‐200a	69.43	148.24	2.13
hsa‐miR‐429	92.18	240.28	2.60
Down‐regulated miRNAs
hsa‐miR‐192	197.74	84.06	−2.35

4.5. Differential expression, at the protein level, of selected genes known to be involved in EMT and/or cell in cell adhesion

By Western Blot analysis, we monitored the protein expression of selected differentially‐regulated genes that, given their known cellular functions, were attractive candidates for participating in the SOCS1‐induced phenotype observed in CRC cells. Protein levels of ZEB1 (ZEB1), fibronectin‐1 (FN1), the mesenchymal form of p120ctn (p120ctn 1) and δ‐catenin (CTNND2, a potential target of SOCS1‐regulated miRNA) drastically dropped in SOCS1‐expressing cells (Figure 4A). We also observed a strong increase in the levels of E‐cadherin (CDH1) and to a certain extent of the epithelial forms of p120ctn and of the potential ZEB1 targets Keratin 18 (KRT18), Integrin beta 4 (ITGB4), γ‐catenin (JUP) (Figure 4A). The observed variations in E‐cadherin, ZEB1 and fibronectin‐1 are consistent with the capacity of SOCS1 to reverse EMT in CRC cells. Nevertheless, the lack of variation in Vimentin indicates that such reversion is only partial. As shown in Figure 4B, the decrease in the level of the mesenchymal form of p120ctn (p120ctn 1) upon SOCS1 expression was confirmed by immunofluorescence experiments. We also noted that the epithelial forms of p120ctn (p120ctn 1), whose level slightly increased upon SOCS1 expression, translocated from the cytoplasm to the cell cortex (Figure 4C). Membrane localization of E‐cadherin is regulated through its interaction with either δ‐catenin (Yang et al., 2010) or the p120ctn 1, which destabilize E‐cadherin, or the epithelial isoforms of p120ctn, which stabilizes E‐cadherin (Anastasiadis and Reynolds, 2000; Davis et al., 2003; Ireton et al., 2002). Even though levels of epithelial isoforms of p120ctn only modestly increased in SOCS1‐expressing cells, the disappearance of δ‐catenin and of the mesenchymal p120ctn isoform (p120ctn 1) would be sufficient to tip the balance toward E‐cadherin stability and its localization to the cell cortex. These molecular events are thus likely to account for E‐cadherin cortical localization in SOCS1‐expressing cells.

4.6. E‐cadherin plays a central role in the SOCS1‐induced mesenchymal–epithelial transition in CRC cells

To assess the role of E‐cadherin in the SOCS1‐induced phenotype observed in CRC cells, we performed shRNA‐mediated silencing of E‐cadherin in SW620^SOCS cells. FACS analyses indicated that CDH1 ShRNA‐1‐GFP (Figure 5A right), but not CDH1 ShRNA‐2‐GFP (Figure 5A left), was effective in inhibiting the expression of E‐cadherin in SW620^SOCS#2 cells. We found that depletion of E‐cadherin decreased cell–cell contacts and reversed the epithelial morphology that was induced by SOCS1 (Figure 5B, right). More importantly, E‐cadherin silencing restored invasiveness of SW620^SOCS cells (Figure 5C), demonstrating that E‐cadherin plays a central role in the SOCS1‐induced MET in CRC cells.

Figure 5.

E‐cadherin is involved in the SOCS1‐induced changes in cell morphology or cell invasion potential. A: Flow cytometry analysis of E‐cadherin expression at the surface of SW620SOCS#2 cells (red curve) and of SW620SOCS#2 cells after lentiviral transduction of CDH1 ShRNA‐2‐GFP (green curve; A) or of CDH1 ShRNA‐1‐GFP (blue curve; A). The gray curves display the signals obtained with isotype control antibody. B: Changes in cell morphology in the SW620SOCS#2 cells expressing either CDH1 ShRNA‐2‐GFP or CDH1 ShRNA‐1‐GFP. C: Cell invasion assays were performed as described in Figure 2, using SW620control (grey), SW620SOCS#2 (red), or SW620SOCS#2 cells expressing either CDH1 ShRNA‐1‐GFP (blue) or CDH1 ShRNA‐2‐GFP (green). Results are expressed as percentages of cell invasion in these populations relative to the SW620control cells.

4.7. SOCS1 expression in SW620 cells prevents metastases in nude mice

Having established that SOCS1 expression decreases with CRC aggressiveness, and that SOCS1 engages regulation of multi‐gene programs leading to a partial reversion of the EMT in CRC cells, we investigated whether SOCS1 is sufficient to regulate the differentiation status of tumors in xenograft assays, and/or to affect the metastatic potential of such tumoral cells. Subcutaneously xenografted animals were analyzed for tumor growth over time. No significant modification of tumor growth was observed with the cells expressing SOCS1 as compared to the parental cells (not shown). Pathology analysis of HES‐stained, SW620^control cells‐derived tumor sections revealed a heterogenous phenotype, with poorly differentiated cells as well as large carcinoma cells (Figure 6A). In the SOCS1‐expressing tumors, the histological sections showed a similar pattern except for the presence of glandular‐like structures. Interestingly, these partially differentiated structures showed high level of E‐cadherin (Figure 6A), in agreement with what we had observed in cell culture (Figure 2B). We finally assessed the effect of SOCS1 expression on the metastatic potential of CRC cells in nude mice. For this purpose, SW620^control or SW620^SOCS cells expressing luciferase reporter gene were generated through lentiviral transduction, and were injected in mice, either intravenously or intracardiacally. During the following weeks, on a regular basis, mice were injected with d‐luciferin and examined for luminescence using an in vivo imaging system. This technique allowed monitoring the number and localization of metastases. While control cells generated metastases in all engrafted animals, only a few metastases‐bearing animals was obtained with the SOCS1 positive clone (Table 2). Whereas the percentage of animals bearing muscle metastasis was equivalent, a striking difference was obtained on adrenal gland and bone metastasis: while 33 and 55% of animals injected with controls cells were scored positive for adrenal gland and bone metastasis respectively, no metastatic nodules could be detected at these locations with SOCS1‐expressing cells (Table 2). This result is independent of the route of cells injection as intracardiac or intravenous cell injections led to similar results. Histological sections of muscle (6B, left, upper), adrenal gland (6B left lower) and bones (6B right) are shown in Figure 6. We thus concluded that SOCS1 expression was sufficient to decrease the metastatic potential of CRC cells in xenograft models. In particular, SOCS1‐expressing cells were no longer capable of generating adrenal gland or bone metastases.

Figure 6.

In vivo consequences of SOCS1 expression in the SW620 cells. A: Representative HES and immunohistochemical staining of flank tumor xenografts samples with anti‐myc (to visualize SOCS1) and anti‐E‐cadherin antibodies (magnification ×100). B: Spontaneous metastases were observed in several organs with the SW620control‐luc+ cells, at week 5 after intracardiac injection, as illustrated after HES staining in the muscle (upper, left), or after immunostaining for MHC class1 in the adrenal gland (lower, left) or the bones (lumbar vertebra or femur) (right panels). Injection of SW620SOCS#2‐luc+ cells only led to metastases in the muscle (upper, right). Magnifications: HES × 100. B, MHC1 either ×20 or ×100 as indicated. AC: adrenal cortex; B: Bone; BM: bone marrow; M‐A: medulo‐adrenal; MF: muscle fibers; Meta: metastasis; V: vertebra. The experiment was repeated three times.

Table 2.

Summary of tissue localization of metastases in nude mice.

Cells	Route of injection	Number of mice with metastases in : (%)
		Adrenal gland	Bone	Muscle
SW620control	i. v.	1/7 (14.3%)	4/7 (57.2%)	1/7 (14.3%)
	i. c.	3/9 (33.3%)	5/9 (55.5%)	2/9 (22%)
SW620SOCS	i. v.	None	None	2/15 (13.3%)
	i. c.	None	None	3/12 (25%)

5. Conclusions

Our findings that levels of SOCS1 were lower in metastases than in primary tumors in CRC patients, that forced SOCS1 expression in metastatic SW620 cells induced gene expression changes consistent with a partial reversion toward the transcriptome of primary tumor‐derived SW480 cells, and that SOCS1 decreased the metastatic potential of CRC cells in xenograft models, collectively support the notion that SOCS1 is a critical suppressor of metastasis in human CRC pathology.

6. Discussion

During tumor progression, epithelial cells are converted into mesenchymal cells by EMT, which is characterized by the loss of E‐cadherin, the gain of mesenchymal markers such as ZEB1, Fibronectin or Vimentin, and the increase in cell invasion and migration. This phenotype is associated with a loss of cell polarity, low cell–cell interactions and anchorage‐independent growth (Voulgari and Pintzas, 2009). EMT gives to epithelial cells an advantage to escape from the primary tumor site and facilitates distant implantation of metastases (Thiery et al., 2009). EMT can be controlled by intrinsic oncogene activation such as KRAS mutation (Edme et al., 2002). It can also be triggered by external stimuli that emanate from extracellular matrix or immune cells, such as secreted soluble growth factors or cytokines (Thiery et al., 2009). Recent studies have shown that EMT may be particularly relevant for tumor progression in human colon cancer (Loboda et al., 2011). Therefore, understanding mechanisms that could prevent EMT or induce its reversion through MET is of critical importance. A number of groups has reported experimental conditions that could reverse EMT in colorectal cancer. Fibroblasts growth factor receptor 4 (FGFR4) is overexpressed in CRC and its silencing decreased the invasive capabilities of CRC cells and prevented them from forming tumors in nude mice. This effect was accompanied by a decrease in SNAIL and TWIST gene expression levels and an increase in E‐cadherin, leading to a more epithelial phenotype (Pelaez‐Garcia et al., 2013). Similarly, down‐regulation of the basic helix‐loop‐helix transcription factor AP4, a c‐myc target gene, in colorectal cancer resulted in mesenchymal–epithelial transition, and inhibited invasion and migration (Jackstadt et al., 2013). In this example, the downstream effectors were the SNAIL transcriptional repressors. It has also been reported that Claudin‐1, a member of a family of proteins that are the principal constituents of the tight junctions, was highly up‐regulated in CRC. Its expression correlated with colon cancer progression and metastasis.

Forced expression of Claudin‐1 was accompanied by an increase in ZEB1, which in turn reduced E‐cadherin expression. Inhibition of Claudin‐1 expression caused re‐expression of E‐cadherin and loss of invasive properties. This effect was associated with loss of the mesenchymal Vimentin (Dhawan et al., 2005).

In this report we provide evidence for an important role for SOCS1 in CRC tumor progression and metastatic potential. Specifically, our data support a model where SOCS1 controls tumor cell adhesion and motility by modulating EMT. SOCS1‐induced down‐regulation of ZEB1 and/or up‐regulation of miRNA or the mir‐200 family may play an important role in this process. Indeed, it has been reported that ectopic expression of miRNA of the miR‐200 family, including miR‐200c that inhibits the expression of ZEB1, can promote a MET and reduce tumor cell migration (Hur et al., 2013; de Krijger et al., 2011). The SOCS1‐induced mesenchymal–epithelial reversion appears to be only partial, as expression of the mesenchymal marker Vimentin was unchanged. EMT is not an all or nothing phenomenon and intermediate stages may exist (Huang et al., 2013). Of interest we found that SNAIL, SLUG and TWIST were not affected by SOCS1 (data not shown). ZEB, SNAIL and SLUG regulate distinct, yet overlapping, target genes that cooperate to induce a full EMT program. Our finding that SOCS1 regulates ZEB1 independently of the other relevant transcription factors may explain why SOCS1 only induced a partial EMT reversion.

In addition to the phenotypic changes induced in cultured cells, we report that SOCS1 metastatic abilities of CRC cells with a profound effect on their capacities to colonize bones and adrenal glands. Interactions of cancer cells with the tumor microenvironment play important roles in metastatic progression. Colonization is also promoted by the production of several cytokines secreted by cancer cells themselves. In particular, chemokines signaling dictates organ‐selective cancer metastasis in immunodeficient mice (Cambien et al., 2009), which may be targeted by SOCS1 in this in vivo model.

In summary, our results show that SOCS1 expression modulates invasive properties of cancer colon cells in vitro, and limits the occurrence of metastases of colon tumors in a mouse model, especially metastases to adrenal glands and bones. Finally, our work provides a list of target genes that may be worth investigating, together with SOCS1, to help in stratifying colon cancer patients and that may be useful in defining prognostic features.

Contributors

J. P.: conception, design of the study and results analysis. M. D., C. N., and M. L., acquisition of data, experimental procedure. M. P., and J‐M. R.: animal experimentation and management. V. L.: responsibility of the microarrays studies. P. D., and M. D.: computational analysis and interpretation of data. J. P., M. D., S. R. and J.B.: manuscript writing.

Funding

Supported by INSERM, and by La Ligue Nationale contre le Cancer (Equipe labellisée to J. B. and to S. R.)

Conflicts of interest

None.

Supporting information

The following are the supplementary data related to this article:

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data 1.

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2014.03.014.

David Muriel, Naudin Cécile, Letourneur Martine, Polrot Mélanie, Renoir Jack-Michel, Lazar Vladimir, Dessen Philippe, Roche Serge, Bertoglio Jacques and Pierre Josiane, (2014), Suppressor of cytokine signaling 1 modulates invasion and metastatic potential of colorectal cancer cells, Molecular Oncology, 8, doi: 10.1016/j.molonc.2014.03.014.