ERK2-Dependent Phosphorylation of CSN6 Is Critical in Colorectal Cancer Development

SUMMARY

Biomarkers for predicting prognosis are critical to treating colorectal cancer (CRC) patients. We found that CSN6, a subunit of COP9 signalosome, is overexpressed in CRC samples and that CSN6 overexpression is correlated with poor patient survival. Mechanistic studies revealed that CSN6 is deregulated by EGFR signaling, in which ERK2 binds directly to CSN6 Leu163/Val165 and phosphorylates CSN6 at Ser148. Furthermore, CSN6 regulated β-Trcp and stabilizes β-catenin expression by blocking the ubiquitin-proteasome pathway, thereby promoting CRC development. High CSN6 expression was positively correlated with ERK2 activation and β-catenin overexpression in CRC samples, and inhibiting CSN6 stability with cetuximab reduced colon cancer growth. Taken together, our study’s findings indicate that the deregulation of β-catenin by ERK2-activated CSN6 is important for CRC development.

INTRODUCTION

Colorectal cancer (CRC) is a leading cause of cancer mortality. The primary treatment for CRC patients is surgery, and conventional cytotoxic chemotherapy and/or targeted therapies are routinely used to treat patients who are at high risk of developing recurrent or metastatic disease (Laurent-Puig et al., 2009). The molecular alterations in CRC have been studied extensively. Specific molecular marker analyses (e.g., microsatellite instability, CpG island methylator phenotype, PIK3CA, RAS and BRAF mutations) are becoming routine clinical practice in determining colorectal cancer treatment (Cancer Genome Atlas, 2012; Karapetis et al., 2008; Ogino et al., 2014; Walther et al., 2009). For example, RAS and BRAF mutations have impact on the clinical management of colorectal cancer (Amado et al., 2008; Bardelli and Siena, 2010; Douillard et al., 2013; Pietrantonio et al., 2015). However, a more detailed picture of the pathways deregulated in CRC has yet to emerge. Defining these molecular alterations can help guide treatment and improve clinical care (Siena et al., 2009).

The COP9 signalosome (CSN) is a multiprotein complex involved in protein degradation, transcriptional activation (Seeger et al., 1998; Wei and Deng, 1999), signal transduction (Chen et al., 2012; Xue et al., 2012; Zhang et al., 2008), and tumorigenesis (Chen et al., 2014; Richardson and Zundel, 2005; Zhao et al., 2014; Zhao et al., 2013; Zhao et al., 2011). However, the detailed biological functions of the CSN’s subunits have not been well characterized. Several studies have indicated that mammalian CSN subunits are involved in developmental processes; examples include targeted disruptions of mammalian Csn2, Csn3, Csn5, and Csn8 that resulted in defective embryo development (Lykke-Andersen et al., 2003; Menon et al., 2007; Tomoda et al., 2004; Yan et al., 2003). The role of CSN subunits in cancer biology is emerging (Lee et al., 2011). Our own work has shed some light on the function of the CSN6 subunit specifically. In a previous study, we performed targeted disruption of the Csn6 gene in mice and found that Csn6-/- mice developed until 7.5 days post-coitus but not beyond this time (Zhao et al., 2011). In that same study, Csn6+/- mouse tumor experiments showed that Csn6 haplo-insufficiency helps impede the development of cancer (Zhao et al., 2011), suggesting that CSN6 signaling regulation is critical for tumor development. However, the mechanism and biological consequence of CSN6 overexpression in cancer remains unclear. In this study, we characterized the mechanism of CSN6 overexpression during the colorectal cancer tumorigenesis.

RESULTS

CSN6 Is Overexpressed in and Is a Biomarker for Colon Cancer

To identify potential gene deregulation correlated with the outcomes of CRC patients, we analyzed the gene expression profiles of snap-frozen CRC tissues resected from 20 CRC patients. A comparison of gene expression levels changed in CRC patients with recurrence-free survival (RFS) ≤3 years and those with RFS >3 years, indicated that CSN6 is one of the top genes deregulated in CRC (Figure S1A; Table S1).

Microarray analysis showed that CSN6 expression in CRC samples from patients with a recurrence-free survival (RFS) duration ≤3 years was significantly higher than that in CRC samples from patients with an RFS duration >3 years (p=0.004) (Figures 1A and 1B). Using the BioGPS Gene Expression Atlas, we found that CSN6, but not CSN5, was highly expressed in colon cancer (Figures S1B and S1C). Furthermore, an analysis of mRNA expression data from The Cancer Genome Atlas (TCGA) revealed that the level of CSN6 expression was higher in cancer tissue than in normal tissue and was positively correlated with advanced disease (Figure 1C), whereas the level of CSN5 expression in normal colon tissue did not differ significantly from that in CRC (Figure S1D). Validating these findings, quantitative reverse-transcription polymerase chain reaction (qRT-PCR) analysis of CSN6 expression in 33 paired samples of colon cancer tissue and adjacent normal mucosa revealed that CSN6 expression in cancer tissue was significantly higher than that in normal tissue (Figure 1D). Finally, an immunohistochemistry tissue microarray revealed that CSN6 expression in CRC was higher than that in normal tissue in 67 (82.7%) of 81 paired samples (Figures 1E and 1F). Western blot analysis of endogenous CSN6 expression in one colon cell line (HCEC) and four CRC cell lines (DLD-1, HCT116, SW480, and AKP) revealed that CSN6 is overexpressed in the CRC cell lines but not in the normal colon cells (Figure S1E). Therefore, CSN6 can be identified as a biomarker overexpressed in human colon cancer samples.

Figure 1. CSN6 Is Overexpressed and Associated with Poor Survival in Colorectal Cancer.

(A) Twenty CRC samples were collected and subjected to gene expression microarray. The expression of all subunits of COP9 signalosome is presented as a heat map. RFS, recurrence-free survival.

(B) Relative expression of CSN6 in 20 CRC patients. The bounds of the boxes denote interquartile ranges; solid lines within the boxes denote medians; and whiskers denote the 5th and 95th percentiles. ^★★p<0.01.

(C) CSN6 gene mRNA levels were assessed from The Cancer Genome Atlas (TCGA). Amplification of the CSN6 gene was associated with advanced disease stage. The bounds of the boxes denote interquartile ranges; solid lines within the boxes denote medians; whiskers denote the 5th and 95th percentiles. ^★★p<0.01; ^★★★p<0.001.

(D) Waterfall plot of relative CSN6 mRNA levels of 33 paired samples of CRC and normal tissue measured using qRT-PCR.

(E) Representative IHC staining for CSN6 in human colon cancer (right) and adjacent normal colon tissue (left). CSN6 expression in 81 paired samples of CRC and normal colon tissue was assessed by immunohistochemistry. Scale bars represent 50 μm.

(F) A box plot illustrates the relative expression of CSN6 in 81 paired samples of normal tissue and CRC. The horizontal lines represent the medians, the boxes represent the interquartile range, and the whiskers represent the 5th and 95th percentiles. ^★★★p<0.001, Student t-test.

(G and H) Kaplan-Meier survival curves of overall survival duration based on CSN6 expression in the CRC tissues of the testing and validation cohorts. The receiver operating characteristic curve was used to define the cutoff, and log-rank analysis was used to test for significance. See also Figure S1 and Table S1.

CSN6 Overexpression is Correlated with Poor Prognosis in CRC Patients

Because CSN6 expression is significantly increased in CRC, we sought to determine whether CSN6 expression is associated with the prognosis and/or clinical characteristics of CRC patients. The testing cohort included paired samples of normal colon and CRC from 81 patients. The validation cohort included CRC samples from 305 patients. CSN6 expression level was determined using immunohistochemistry tissue microarray. High CSN6 expression was positively correlated with elder patients in the testing cohort and with advanced clinical stage in the validation cohort (p<0.05, Table 1). Kaplan-Meier analysis showed that high levels of CSN6 expression were correlated with poor overall survival in both cohorts (Figures 1G and 1H). Further, multivariate Cox regression analysis revealed CSN6 expression to be an independent prognostic factor for poor survival (Table 2). In another analysis, Kaplan-Meier estimate shows significantly (p=0.006) shorter overall survival in patients with high CSN6 gene expression in the Colorectal Adenocarcinoma (Cancer Genome Atlas, 2012)(TCGA, Nature 2012) data (Figure S1F).

Table 1.

Correlation between expression of CSN6 and clinicopathological features of colorectal cancer patients

Variable	Testing cohort			Validation cohort
	Low CSN6	High CSN6	p valuea	Low CSN6	High CSN6	p valuea
Gender			0.350			0.568
Male	22 (71.0)	30 (60.0)		78 (54.5)	83 (51.2)
Female	9 (29.0)	20 (40.0)		65 (45.5)	79 (48.8)
Median age			0.010			0.301
<59 years	19 (61.3)	15 (30.0)		72 (50.3)	71 (43.8)
≥59 years	12 (38.7)	35 (70.0)		71 (49.7)	91 (56.2)
Histological grade			0.860			0.588
G1	4 (12.9)	7 (14.0)		8 (5.6)	13 (8.0)
G2	21 (67.7)	31 (62.0)		115 (80.4)	123 (75.9)
G3	6 (19.4)	12 (24.0)		20 (14.0)	26 (16.0)
pT status			0.509			0.161
T1	1 (3.2)	2 (4.0)		7 (4.9)	2 (1.2)
T2	4 (12.9)	7 (14.0)		21 (14.7)	17 (10.5)
T3	23 (74.2)	30 (60.0)		112 (78.3)	140 (86.4)
T4	3 (9.7)	11 (22.0)		3 (2.1)	3 (1.9)
pN status			1.000			0.289
N0	15 (48.4)	24 (48.0)		85 (59.4)	106 (65.4)
N1	16 (51.6)	26 (52.0)		58 (40.6)	56 (34.6)
pM status			0.518			0.038
M0	28 (90.3)	42 (84.0)		134 (93.7)	140 (86.4)
M1	3 (9.7)	8 (16.0)		9 (6.3)	22 (13.6)
Clinical stage			0.819			0.028
I	4 (12.9)	7 (14.0)		21 (14.7)	13 (8.0)
II	11 (35.5)	14 (28.0)		59 (41.3)	78 (48.1)
III	13 (41.9)	21 (42.0)		54 (37.8)	49 (30.2)
IV	3 (9.7)	8 (16.0)		9 (6.3)	22 (13.6)
Chemotherapy(%)			1.000			0.795
No	17 (54.8)	27 (54.0)		107 (74.8)	119 (73.5)
Yes	14 (45.2)	23 (46.0)		36 (25.2)	43 (26.5)

NOTE: All data are no. of patients (%).

^ap values were calculated in SPSS16.0 using a chi-square test. p values <0.05 were considered to indicate statistical significance.

Table 2.

Univariate and multivariate analysis of different prognostic parameters for colorectal cancer patients in the testing cohort and validation cohort

Variable	Univariate analysis		Multivariate analysis
	HR (95% CI)a	p valueb	HR (95% CI)a	p valueb
Testing cohort (n=81)
Gender (male versus female)	1.4 (0.7-2.9)	0.361	0.6 (0.2 -1.7)	0.384
Age (<59 years versus ≥59 years)	1.8 (0.8-4.0)	0.133	1.9 (0.8-4.5)	0.148
Chemotherapy (no versus yes)	1.7 (0.8-3.4)	0.167	1.7 (0.8 -3.8)	0.188
Histological grade (G1 or G2 versus G3)	2.1 (1.0-4.4)	0.060	1.7 (0.7-4.0)	0.250
pT status (T1 or T2 versus T3 or T4)	6.1 (0.8-44.8)	0.076	2.5 (0.3-20.0)	0.391
pN status (N0 versus N1)	3.0 (1.4-6.8)	0.007	2.6 (1.1-6.2)	0.034
pM status (M0 versus M1)	6.8 (3.1-14.9)	<0.001	8.0 (3.0-21.6)	<0.001
CSN6 expression (low versus high)	3.4 (1.3-9.0)	0.012	3.7 (1.3-10.6)	0.015
Validation cohort (n=305)
Gender (male versus female)	1.2 (0.8-1.8)	0.466	1.6 (1.0-2.4)	0.048
Age (<59 years versus ≥59 years)	1.4 (0.9-2.1)	0.152	1.4 (0.9-2.2)	0.100
Chemotherapy (no versus yes)	1.1 (0.65-1.7)	0.837	1.1 (0.6-1.8)	0.795
Histological grade (G1 or G2 versus G3)	1.7 (1.0-2.8)	0.048	1.6 (0.9-2.7)	0.096
pT status (T1 or T2 versus T3 or T4)	1.5 (0.8-3.0)	0.208	0.9 (0.5-1.8)	0.852
pN status (N0 versus N1)	1.7 (1.1-2.7)	0.010	2.2 (1.5-3.5)	<0.001
pM status (M0 versus M1)	7.2 (4.5-11.6)	<0.001	7.9 (4.6-13.4)	<0.001
CSN6 expression (low versus high)	2.7 (1.7-4.5)	<0.001	2.4 (1.5-4.1)	0.001

^aHazard ratios (HRs) and 95% confidence intervals (CIs) were calculated using univariate or multivariate Cox proportional hazards regression in SPSS 16.

^bp values were calculated using univariate or multivariate Cox proportional hazards regression in SPSS 16.0. p values <0.05 were considered to indicate statistical significance.

EGF-ERK Signaling Increases CSN6 Stability

We performed Gene Set Enrichment Analysis (GSEA) to identify association between expression of CSN6 and signaling pathways in two sets of colon cancer GSE14333 (n=290) and GSE2109 (n=373). Result showed that gene sets of EGFR and KRAS were enriched in samples with high expression of CSN6 (Figure S2A and S2B). Because EGFR signaling is highly activated in colon cancer, we sought to determine whether EGFR activation regulates CSN6 expression. Immunoblot analysis showed that EGF treatment increased the steady-state expression of CSN6 in DLD-1 colon cancer cells, whereas inhibition of MEK/ERK by the MEK1 inhibitor PD98059 diminished the EGF-induced increase of CSN6 expression (Figure 2A). CSN6 mRNA levels did not increase significantly in response to EGF (Figure S2C), suggesting that EGF regulates CSN6 posttranslationally. In addition, the expression of a constitutively active MEK1 (MEK1^CA) increased CSN6 expression in DLD-1 cells (Figure S2D).

Figure 2. The EGF-ERK Axis Regulates CSN6 Stability.

(A) DLD-1 cells were treated with or without PD98059 (20 μM) for 1 hr and then treated with EGF (100 ng/ml) for 30 min. Equal amounts of cell lysates were immunoblotted with the indicated antibodies.

(B) DLD-1 cell lysates were immunoprecipitated with an anti-CSN6 antibody and immunoblotted with the indicated antibodies.

(C) DLD-1 cells were treated with or without EGF (100 ng/ml). MG132 was added to the cells 6 h before they were harvested with a guanidine-HCl–containing buffer. The cell lysates were pulled down with nickel beads and immunoblotted with an anti-CSN6 antibody.

(D) DLD-1 cells were transfected with vectors expressing the indicated myc-tagged ERK and MEK1^CA proteins. MG132 was added to the cells 6 hr before they were harvested with a guanidine-HCl–containing buffer. The cell lysates were pulled down with nickel beads and immunoblotted with an anti-CSN6 antibody.

(E) DLD-1 cells were treated with PD98059 (10 μM) and then treated with cycloheximide (CHX; 100 μg/ml) for the indicated times. Cell lysates were immunoblotted with the indicated antibodies.

(F) DLD-1 cells were starved overnight, treated with EGF (100 ng/ml), and then treated with CHX (100 μg/ml) for the indicated times. Cell lysates were immunoblotted with the indicated antibodies.

(G) Equal amounts of cell lysates from DLD-1 cells transfected with myc-ERK2, myc-ERK2-D318/321N (DN), or myc-ERK2-T159/160E (TE) were immunoprecipitated with an anti-myc antibody and then immunoblotted with the indicated antibodies.

(H) The consensus sequence for the ERK2 substrate binding sites is highlighted. Sequences for CSN6 and other known ERK2 substrates are shown for comparison.

(I) Equal amounts of cell lysates from DLD-1 cells transfected with WT Flag-CSN6 or Flag-CSN6 L163A/V165A (LVAA) were immunoprecipitated with an anti-Flag antibody and immunoblotted with indicated antibodies.

(J) Structural model of the CSN6-ERK2 complex. The loop11 and α4 helix of CSN6 (residues 159-165, D-motif) fit into the major docking groove of ERK2.

(K) The best complex HADDOCK (High Ambiguity Driven DOCKing) model displays a ribbon presentation of the CSN6-ERK2 complex structure. The interacting residues involved in the interface between CSN6 and ERK2 are indicated.

See also Figure S2.

A coimmunoprecipitation assay revealed that ERK1/2 binds to CSN6 (Figure 2B). Furthermore, we found that EGF treatment or ERK2 activation decreases the ubiquitination level of CSN6 (Figures 2C and 2D). Congruently, PD98059 increased the turnover rate of CSN6 in the presence of the de novo protein synthesis inhibitor cycloheximide (Figure 2E), whereas EGF, which induced ERK activation, reduced CSN6 turnover (Figure 2F). Taken together, these results indicate that EGFR-ERK signaling helps to stabilize CSN6.

Model of CSN6-ERK Interaction

ERK cascades are critical to the regulation of signals initiated by growth factors and other stimuli, including EGF (Bardwell, 2006). MAP kinases (MAPKs) bind to their substrates through a docking groove comprised of an acidic common docking (CD) domain and glutamic acid–aspartic acid pocket (Lu and Xu, 2006). Immunoblotting of the immunoprecipitated myc-ERK2 proteins with an anti-CSN6 antibody showed that compared with wild type (WT) ERK2, ERK2 with mutations in the CD domain (D318/321N) or glutamic acid–aspartic acid pocket (T159/160E) had reduced binding to endogenous CSN6 (Figure 2G). Also, many MAPK substrates contain docking motifs (D-sites), which consist of a cluster of basic residues, a short spacer, and a hydrophobic (φ)-X-hydrophobic (φ) motif (K/R -X_1–6 - φ-X -φ) (Bardwell et al., 2009; Bardwell, 2006; Whisenant et al., 2010). An analysis of the CSN6 amino acid sequence identified the consensus D-site as 159-KHTDLPVS-166 (Figure 2H; Figure S2E). When its D-site was mutated to alanine (LV→AA), CSN6 lost its ability to bind to MAPK (Figure 2I), suggesting that the ERK docking groove binds to a D site at L163/V165 in CSN6.

To gain insight into the structural basis of CSN6’s recognition by MAPK, we used HADDOCK to generate a model of the CSN6-MAPK complex (de Vries et al., 2010; Kelley and Sternberg, 2009). Structure modeling using CSN6’s structure (Zhang et al., 2012) and the known MAPK docking groove (Tanoue et al., 2000) (CD groove, residues E81, D318, and D321) indicated close contact between the loop11 and α4 helix of CSN6 (residues 159-165, D-site) and MAPK docking groove (Figure 2J). Basically, MAPK docking grooves are adjacent to each other, and their side chains are exposed at the surface of MAPK. These negatively charged residues are critical to MAPK’s interaction with the D-sites of its substrates (Tanoue et al., 2000). The D-site of CSN6 (loop11 and α 4 helix of CSN6, residues 159-165) docks to the major docking groove of MAPK (Figure 2J). The highly positively charged residues in CSN6’s D-site bind to the negatively charged residues in MAPK’s CD groove. The docking model indicates that the electrostatic interactions are made between the charged residues K97, K102, H160, and K159 of CSN6 and the charged residues E81, D162, D318, and D321 of MAPK (Figure 2K). The hydrophobic interactions occur between the residues Y106, M157, T158, T161, and V165 of CSN6 and the residues Y128, T159, T160, and C161 of MAPK (Figure 2K). The model reveals that CSN6 D-site interacts with the MAPK docking groove.

The EGF-ERK2 Axis Phosphorylates CSN6 on S148 and Enhances CSN6 Stabilization

Sequence analysis of CSN6 using the NetPhos algorithm (www.cbs.dtu.dk/services/NetPhos) and GPS2.1 (Gps.Biocuckoo.org) revealed that CSN6 contains an ERK consensus phosphorylation motif (Ser-Pro) at the S148 residues (Figure 3A). Treatment with calf intestinal alkaline phosphatase (CIP) reduced the EGF-induced steady-state expression of CSN6, indicating that CSN6 phosphorylation is required in this process (Figure 3B) and presumably acts through ERK.

Figure 3. ERK2 Phosphorylates CSN6 S148.

(A) ERK consensus phosphorylation motif of CSN6 (serine 148).

(B) The lysates from DLD-1 cells that were serum-starved overnight and then stimulated with EGF for 1 hr were treated with calf intestinal phosphatase (CIP) and subjected to immunoblotting.

(C) Cell lysates from DLD-1 cells expressing Flag-tagged WT CSN6, CSN6 S148A, or CSN6 S148D that had been starved overnight and then treated with or without EGF (100 ng/ml) for 1 hr were subjected to immunoblotting.

(D) Flag-CSN6 was immunoprecipitated with an anti-Flag antibody from DLD-1 cells that had been pretreated with or without PD98059 for 1 hr and then treated with EGF (100 ng/ml) for 30 min. Cell lysates were subjected to immunoblotting with an anti-phosphoserine antibody.

(E) In vitro kinase assays were conducted by incubating recombinant active ERK2 with GST-CSN6 and mutant GST-CSN6-S148A proteins. Cell lysates were subjected to immunoblotting with an anti-phosphoserine antibody.

(F) DLD-1 cells were transfected with the indicated plasmids and treated with CHX (100 μg/ml) for the indicated times. Cell lysates were immunoblotted with the indicated antibodies. The density of Flag-CSN6 was measured and the integrated optical density (IOD) was measured. The turnover of Flag-CSN6 is indicated graphically.

(G) DLD-1 cells were transfected with Flag-CSN6(WT), Flag-CSN6(S148A), or Flag-CSN6(S148D) and treated with EGF (100 ng/ml). MG132 was added to the cells 6 h before they were harvested with guanidine-HCl–containing buffer. The cell lysates were pulled down with nickel beads and immunoblotted with an anti-Flag antibody.

(H) DLD-1 cells were transfected with the indicated plasmids. Cell proliferation rates were measured by MTT assay. The results were measured at an optical density of 570 nm. Data are means ± standard deviation (SD).

(I-K) DLD-1 cells were transfected with Flag-CSN6(WT), Flag-CSN6(S148A), and Flag-CSN6(S148D). Cell motility was analyzed by wound-healing assay, migration assay, and invasion assay. The scale bar represents 200 μm. Data represent three independent experiments. Data are means ±SD. ^★★p<0.01.

To determine the way in which the EGF-induced stabilization of CSN6 involves phosphorylation, we constructed S148A and S148D phosphorylation site mutants of CSN6 and assessed their steady-state expression in the presence of EGF. The CSN6 S148A mutant did not respond to EGF-induced upregulation, whereas the WT CSN6 was upregulated by EGF. Even in the absence of EGF, the phosphorylation-mimetic CSN6 S148D mutant had a higher level of expression than the WT CSN6 or the CSN6 S148A mutant did (Figure 3C).

To confirm the existence of an ERK-specific phosphorylation site in CSN6 in vivo, we used an anti–phospho-serine antibody to detect the increase of serine-phosphorylated CSN6 in EGF-treated cells (Figure 3D). MEK1 inhibition with PD98059 treatment abolished the serine phosphorylation of CSN6 (Figure 3D). Compared with WT CSN6, the CSN6 S148A mutant had significantly lower serine phosphorylation levels in vivo (Figure 3D). Accordingly, CSN6 S148A was resistant to phosphorylation by recombinant ERK, whereas WT CSN6 was highly phosphorylated by ERK, in an in vitro kinase assay (Figure 3E). We then found that the CSN6 S148A mutant had a faster turnover rate than WT CSN6 did. As expected, the turnover rate of the CSN6 S148D mutant was slower than that of WT CSN6 (Figure 3F). We assessed the ubiquitination of CSN6 in the presence of EGF and found that EGF reduced the ubiquitination level of WT CSN6 but not that of the CSN6 S148A mutant (Figure 3G). Even in the absence of EGF, the phosphorylation-mimetic CSN6 S148D mutant had a lower ubiquitination level than WT CSN6 or the CSN6 S148A mutant did (Figure 3G). Together, these results indicate that EGF/ERK mediated CSN6 phosphorylation and diminished CSN6 ubiquitination and turnover.

To determine the biological significance of CSN6 phosphorylation, we transiently transfected DLD-1 cells with WT CSN6 or CSN6S 148A plasmid and measured their growth. A 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay demonstrated that cells transfected with WT CSN6 had significant growth, whereas cells transfected with CSN6 S148A lost growth capability (Figure 3H). Also, WT CSN6 increased cell migration and invasion in a wound-healing assay (p<0.001, Figure 3I) and in transwell migration/invasion assays (Figures 3J and 3K), whereas CSN6 S148A lost such capability, demonstrating that the CSN6-promoted migration and invasion of colon cancer cells requires a phosphorylation event.

CSN6 Positively Regulates β-Catenin Protein Stability and Facilitates the Transcriptional Activity of β-Catenin

Because β-catenin plays an important role in CRC and can be upregulated by EGF activation (Yang et al., 2011), we assessed the expression of β-catenin and CSN6 under the condition of EGF-ERK activation to further investigate the role of EGF-mediated CSN6 phosphorylation/stabilization in regulating tumorigenicity. CSN6 and β-catenin expression were both upregulated in DLD-1 and SW480 colon cancer cell lines after EGF treatment (Figure 4A). CSN6 increased the steady-state expression of β-catenin in a dose-dependent manner while downregulating β-Trcp (Figure 4B). In line with this finding, western blotting revealed that knock down of CSN6 by shRNA decreased the expression of β-catenin and increased the expression of β-Trcp (Figure 4C).

Figure 4. CSN6 Increases β-Catenin Expression by Regulating Ubiquitination.

(A) DLD-1 or SW480 cells were treated with EGF for the indicated times. Cell lysates were immunoblotted with the indicated antibodies.

(B) DLD-1 cells were transfected with increasing amounts of CSN6 plasmids. Cell lysates were immunoblotted with the indicated antibodies.

(C) DLD-1 cells were treated with shCSN6. Equal amounts of cell lysates were immunoblotted with the indicated antibodies.

(D) DLD-1 cells were treated with shCSN6. qRT-PCR analysis was performed to measure the mRNA levels of β-catenin–induced target genes. Data are means ± SD.

(E) DLD-1 cells with control vector, WT-CSN6, or CSN6-S148A were transfected with either TOP-FLASH or FOP-FLASH (control vector). The relative luciferase activity levels were normalized to those of control cells and the Renilla control plasmid. Data represent three independent experiments and are means ± SD. ^★★p<0.01.

(F) DLD-1 cell lysates were immunoprecipitated with an anti–β-Trcp antibody and immunoblotted with the indicated antibodies.

(G) Cell lysates were prepared from DLD-1 cells expressing Flag-tagged WT CSN6 that had been treated with or without MG132 for 8 hr. Equal amounts of cell lysates were immunoblotted with the indicated antibodies.

(H) DLD-1 cells were transfected with the indicated plasmids and treated with CHX (100 μg/ml) for the indicated times. Cell lysates were immunoblotted with the indicated antibodies.

(I) DLD-1 cells were transfected with the indicated plasmids. MG132 was added to the cells 6 hr before they were harvested with guanidine-HCl–containing buffer. The cell lysates were pulled down with nickel beads and immunoblotted with an anti-Flag antibody.

(J) A heat map produced from transcriptomic analyses of CSN6 and β-catenin–induced target genes in 373 colon cancer patients. The human CRC patient dataset was obtained from Oncomine and the Gene Expression Omnibus (cohort GSE-2109).

See also Figure S3.

The CSN6 knockdown–mediated β-catenin destabilization translated into β-catenin’s reduced transcriptional activity, as evidenced by qRT-PCR findings showing diminished expression of β-catenin target genes, including myc, cyclin D, and vascular endothelial growth factor (Figure 4D). Overexpression of CSN6 could increase the expression of β-catenin target genes (Figure S3A). The TOP-FLASH TCF/LEF-1 luciferase reporter gene analyses showed that Flag-CSN6 significantly increased β-catenin transactivation but the activity of Flag-CSN6-S148A was compromised (Figure 4E). Thus, CSN6 is involved in the β-catenin pathway.

CSN6 Increases β-Catenin Expression by Hindering the Ubiquitin-Mediated Degradation of β-Catenin

Because the stability of β-catenin is regulated by β-Trcp–mediated ubiquitination/degradation (Liu et al., 1999; Winston et al., 1999), we hypothesized that CSN6 negatively impacts the β-Trcp protein to stabilize β-catenin. We found that CSN6 associated with β-Trcp in vivo (Figure 4F), suggesting that CSN6 interacts with F-box protein β-Trcp and regulates it, thereby stabilizing β-catenin.

qRT-PCR analysis revealed that the mRNA levels of β-Trcp and β-catenin were not affected by CSN6 overexpression (Figure S3B), and western blotting revealed that CSN6-mediated β-Trcp downregulation could be rescued by the proteasome inhibitor MG132 (Figure 4G). These results suggest that CSN6 regulates β-Trcp and β-catenin at the posttranscriptional level. CSN6 overexpression reduced the turnover rate of β-catenin and increased the turnover rate of β-Trcp (Figure 4H). Furthermore, CSN6 increased the ubiquitination level of β-Trcp and reduced the ubiquitination level of β-catenin (Figure 4I). Significantly, we searched the Oncomine database for the expression profiles of CSN6 and β-catenin target genes and found that transcriptomic analysis of 373 colon cancer samples showed a positive correlation between CSN6 expression and the expression of β-catenin target genes such as MYC, MET, and JUN (Figure 4J; Figure S3C).

The EGF–ERK–CSN6–β-Catenin Axis Promotes Cell Proliferation and Tumor Growth

The EGF/ERK pathway is often deregulated in CRC, and the monoclonal antibody cetuximab, an EGFR inhibitor, is used clinically to correct this deregulation (Allegra et al., 2009; Prewett et al., 2002). To investigate the role of the EGF–ERK–CSN6–β-catenin axis in colon cancer tumorigenesis, we injected DLD-1 cells into athymic nude mice. Treatment with cetuximab inhibited tumor growth (Figures 5A and 5B) led to reduced ERK1/2 phosphorylation and CSN6 expression in tumors (Figure 5C). qRT-PCR analysis showed that β-catenin target genes were also downregulated after cetuximab treatment (Figure S4A). Furthermore, we subcutaneously injected AKP colon cancer cells and CSN6-depleted AKP colon cancer cells into athymic nude mice. In the xenografted colon cancer samples, CSN6 depletion inhibited tumor growth by increasing β-Trcp while downregulating β-catenin (Figures 5D–5F; Figure S4B). Immunohistochemical staining revealed that the expressions of CSN6 and β-catenin were decreased in shCSN6 tumor samples (Figure 5G). β-catenin target genes in tumor samples were also downregulated after CSN6 knockdown (Figure 5H).

Figure 5. The pERK–CSN6–β-Catenin Axis Promotes Colon Cancer Growth in Mice and Cancer Patients.

(A and B) DLD-1 cell were subcutaneously injected into nude mice (n=4). Mice were treated with or without cetuximab. Tumor growth curves and representative tumor sizes are shown. Data are the means ± SD.

(C) The tumors in (A) were isolated at the end of the assay, and tissue lysates were immunoblotted with the indicated antibodies.

(D and E) shCSN6-AKP colon cancer cells and control stable AKP colon cancer cells were subcutaneously injected into nude mice (n=10). Tumor growth curves are shown. The tumors were isolated at the end of the assay, and the tumor weight of each group was measured. Representative tumor sizes are shown. Data are means ± SD.

(F) Tumors in (D) were isolated at the end of the assay, and tissue lysates were immunoblotted with the indicated antibodies.

(G) Representative IHC staining for p-ERK, CSN6, and β-catenin in tumor tissues from mice. Scale bars represent 50 μm.

(H) RNA was extracted from tumor tissues. qRT-PCR analysis was performed to measure the mRNA levels of β-catenin–induced target genes. Data are means ± SDs.

(I) Representative IHC staining for p-ERK, CSN6, and β-catenin in serial sections of colon cancer samples from patients. Case 1 is representative of a patient with CSN6-overexpressing colon cancer. Case 2 is representative of a patient with non–CSN6-overexpressing colon cancer. The samples were derived from 305 colon cancer cases. Scale bars represent 50 μm.

(J) Model of ERK-induced CSN6 modulating β-catenin stability and tumorigenesis. Upon phosphorylation of S148 by the EGFR-ERK axis, CSN6 is stabilized, which in turn downregulates β-Trcp to preserve β-catenin, thereby facilitating tumorigenesis.

See also Figure S4 and Table S2.

To determine the clinical relevance, we subjected a tissue microarray containing 305 human CRC specimens to immunohistochemical staining for p-ERK, CSN6, and β-catenin (Figure 5I). The CRC samples had high CSN6 expression, which correlated with high ERK activation and high β-catenin expression (representative case 1). Accordingly, CRC samples with low CSN6 expression had low β-catenin expression and reduced ERK activation (representative case 2). The p-ERK, CSN6, and β-catenin expression levels were correlated with each other (Table S2). These results strongly suggest that the p-ERK–CSN6–β-catenin axis is deregulated during the development of human CRC (Figure 5J).

DISCUSSION

In the present study, we found that CSN6 is a prognosis-associated marker that is deregulated in colon cancer. CSN6 expression is elevated in colon cancer and leads to worse RFS, suggesting that abnormal CSN6 expression causes treatment resistance. In two cohorts of colon cancer patients, high levels of CSN6 expression resulted in poor overall survival, indicating that CSN6 has a critical role in colon cancer development. CSN6 is a critical ubiquitination regulator involved in cell cycle regulation (Choi et al., 2011; Xue et al., 2012) and p53 signaling pathway (Zhao et al., 2011), but its role in colon cancer remains unclear. Noticeably, CSN6 is the major CSN subunit highly deregulated in colon cancer, while CSN5, another subunit frequently deregulated in cancer (Tomoda et al., 1999; Zhao et al., 2014), is not particularly overexpressed in colon cancer. In particular, we found that CSN6 has unprecedented biological activity in downregulating β-Trcp, thereby stabilizing β-catenin; this confirms CSN6’s role in promoting colon cancer development by influencing a major molecular defect in the disease.

ERK kinases bind to their substrates through a docking groove, and these substrates usually contain a D-site that interacts with ERK. In the present study, replacing some conserved residues on the D- site compromised CSN6’s capacity to interact with ERK, suggesting that these residues are involved in ERK-CSN6 protein-protein interaction. This interaction leads to a phosphorylation event. Our data show that S148 of CSN6 has an ERK phosphorylation site. This site is located in the end portion of CSN6’s MPN (Mpr1p and Pad1p N-terminal) domain, which is found in the N-terminus of yeast Mpr1 and Pad1 proteins (Penney et al., 1998; Rinaldi et al., 1995; Wei and Deng, 1999). Although the biological function of the MPN domain remains not well determined, the domain consists of polar residues that resemble the active site residues of metalloproteases (Maytal-Kivity et al., 2002) and is involved in a proteasome-associated deneddylation activity (Lyapina et al., 2001). Whether this phosphorylation event participates in any activity of the MPN domain is unknown; regardless, this phosphorylation is critical to maintaining the stability of CSN6. We previously showed that S60 of CSN6 has an Akt phosphorylation site (Xue et al., 2012) and that Akt increases the steady-state expression of CSN6 (Xue et al., 2012). Together, these results demonstrated that CSN6 is regulated by two major signaling pathways (i.e., the ERK and Akt signaling pathways).

Our studies also showed that CSN6 can function in connection with β-Trcp in a circuit, resulting in the negative regulation of β-catenin protein. We found that CSN6 also associates with β-Trcp and facilitates its degradation. The steady-state protein level of the F-box protein β-Trcp is important in determining the ubiquitin ligase activity of the whole SCF^β-Trcp. Cullin-RING ligase (CRL) substrate adaptors, such as F-box proteins, tend to have autocatalytic degradation in vivo (Wee et al., 2005; Zhou et al., 2003). Thus, keeping the F-box proteins stable has a positive impact on the CRL, thereby facilitating the degradation of CRL targets. Strikingly, our studies showed that CSN6 expression decreases the steady-state expression of β-Trcp. Because Cullin neddylation, a process controlled by CSN, decreases the steady-state expression levels of F-box proteins (Wee et al., 2005), we rationalized that CSN6 may be involved in Cullin neddylation, thereby increasing the autocatalytic degradation of β-Trcp. Indeed, we recently showed that CSN6 is competing with CSN5 for Cullin binding through its MPN domain (Chen et al., 2014). Structurally, CSN6 and CSN5 form a heterodimer through MPN domain, and the dimer is topologically knotted to engage in Cullin deneddylation (Lingaraju et al., 2014). Reducing CSN6 expression diminishes its competition with CSN5 in terms of Cullin binding, thereby allowing CSN5 to mediate the deneddylation of Cullin efficiently as observed in Csn6^+/– extracts (Chen et al., 2014). On the other hand, it is conceivable that in CSN6-overexpressing cancers abundant CSN6 binds more amounts of Cullin and thus reduces the accessibility of Cullin to CSN5 for deneddylation, thereby preserving Cullin neddylation. More Cullin neddylation increases the autocatalytic degradation of F-box protein such as Fbxw7 (Chen et al., 2014). Here, we provided another example of CSN6-preserved Cullin neddylation in affecting another F-box protein–– β-Trcp and showed how CSN6 can destabilize β-Trcp and subsequently increase the stability of β-catenin.

In our study, cetuximab inhibited tumor growth in xenograft colon cancer mice and prolonged their survival. Tumor sample analysis indicated that cetuximab ablated the CSN6 and β-catenin expression levels as well as ERK activation. Congruently, CSN6 knockdown xenograft tumors had elevated β-Trcp and diminished β-catenin expression and a slower growth rate. Thus, these studies recapitulated the relationships among EGF, ERK, CSN6, and β-catenin in vivo. Activating mutations in KRAS, BRAF, and/or PIK3CA are well-known carcinogenic mechanisms and may predict response to treatment of CRC (Douillard et al., 2013; Mao et al., 2012; Nishihara et al., 2013; Siena et al., 2009). It is known that human colon cancers bearing mutated KRAS do not respond to cetuximab well, whereas those bearing WT KRAS do respond to cetuximab (Karapetis et al., 2008). Our previous data show that AKT signaling pathway stabilizes CSN6 (Xue et al., 2012), and here we show that RAS-MAPK pathway has a positive impact on CSN6 stability. Because KRAS is an upstream regulator of these two pathways, KRAS mutations will continue to activate both RAS-MAPK and AKT pathways to stabilize CSN6, which in turn activates β-catenin activity to cause pathogenesis in colon cancer. The worse prognosis with EGFR antibody therapy (cetuximab) in colon cancer cells/tumors harboring KRAS mutation can, at least in part, be due to CSN6-β-catenin stabilization. CSN6 stabilization and its downstream impact on β-catenin activation may lead to cetuximab resistance. Thus CSN6’s expression level might be used as a predictor for drug selection.

We also found that CSN6 overexpression is quite common in colorectal cancer. Our cohort studies in Figure 1 indicate that CSN6 is overexpressed in about 35-80 % of colorectal cancers depending from mRNA expression or tissue microarray studies. The deregulation rate is comparable to other known markers. For example, Douillard et al. show that RAS mutations (any KRAS or NRAS mutations in exon 2, 3, or 4) in CRC could be about 52 % (Douillard et al., 2013), whereas BRAF V660E mutation is found in 20% of colonic tumors (Farina-Sarasqueta et al., 2010; Pietrantonio et al., 2015). CSN6 overexpression was positively correlated with β-catenin protein expression and ERK activation in human colon cancer samples. These findings demonstrate that CSN6 overexpression can at least partially account for β-catenin overexpression in colon cancer and provide important insights into the mechanisms underlying EGFR/ERK pathway deregulation in colon cancer. That CSN6 acts as a positive regulator of β-catenin raises the possibility that inhibiting the CSN6 signaling axis is an efficient therapeutic approach in β-catenin-deregulated colon cancer.

EXPERIMENTAL PROCEDURES

Detailed descriptions of the following procedures are provided in the Supplemental Experimental Procedures.

Patients and Tissue Samples

We obtained snap-frozen tissue samples from 20 colorectal cancer (CRC) patients with stage III disease who had undergone curative resection between 2007 and 2010 at the Sixth Affiliated Hospital of Sun Yat-sen University. All patients received adjuvant chemotherapy with modified FOLFOX6 (folinic acid, fluorouracil, and oxaliplatin). Of the 20 patients, 7 had disease relapse within 3 years after surgery.

Thirty-three fresh frozen paired samples of primary CRC and adjacent normal colon tissue were collected from the Department of Surgery at the Sixth Affiliated Hospital of Sun Yat-sen University. All patients had stage II or stage III disease at the time of specimen collection.

We also obtained paraffin-embedded samples of primary colorectal adenocarcinomas from two independent CRC patient cohorts: 1) 81 patients from the Sixth Affiliated Hospital of Sun Yat-sen University (the testing cohort) and 2) 305 patients from the First Affiliated Hospital of Sun Yat-sen University (the validation cohort). All samples were collected with the patients’ written informed consent and approval from each study center’s Institutional Review Board.

Microarray Analysis

RNA samples from 20 frozen CRC patient specimens were isolated using the RiboPure Kit (Ambion, Austin, TX). Gene expression profiles were determined using Affymetrix HG-U133 Plus 2.0 GeneChips (Affymetrix Inc., Santa Clara, CA) according to manufacturer’s instructions. Heat maps were generated using the Java Treeview software program (http://jtreeview.sourceforge.net; used under general public license).

Data Mining

CRC data sets were downloaded from the publicly available GEO databases (290 patients from GEO14333 and 373 patients from GEO2109). GSEA was performed by the JAVA program (http://www.broadinstitute.org/gsea) using MSigDB C6.oncogenic signatures gene set collection and visualized in the Enrichment Map software. β-catenin target genes were selected according to the Wnt home page (http://www.stanford.edu/group/nusselab/cgi-bin/wnt/). The mRNA expression profiles of the CSN6, CSN5, and β-catenin target genes in colon cancer tissue were obtained by analyzing the Oncomine Cancer Microarray database (http://www.oncomine.org). The genes’ expression was analyzed and visualized using Cluster and Treeview software (Eisen, MB) and was presented as heat map as previously described (Fuentes-Mattei et al., 2014).

Cell Lines

HCT116, DLD-1, and SW480 cells were obtained from ATCC (Manassas, VA). The HCEC cell line was kindly provided by Dr. Shay (Roig et al., 2010). The AKP cell line was kindly provided by Dr. Hung (Hung et al., 2010). Cells were maintained in Dulbecco’s modified Eagle’s medium/nutrient F12 media (in-house supplier) supplemented with 10% (v/v) fetal bovine serum.

Transfection and Generation of Stable Transfectants

Standard protocols were used; details are given in the Supplemental Experimental Procedures.

Quantitative Polymerase Chain Reaction

TRIzol (Invitrogen) was used to extract total RNAs from cells or tissues. The iScript cDNA Synthesis Kit (BioRad, #170-8891) was used to produce cDNA from 1 μg of RNA. Quantitative reverse-transcription polymerase chain reaction (qRT-PCR) was performed using a 7500 Real-Time PCR System (Applied Biosystems), and SYBR Green PCR Master Mix (BioRad). Primers for qRT-PCR of CSN6 and other genes are listed in TableS3. GAPDH was used for normalization.

Western Blot Analysis and Immunoprecipitation

For western blot analysis, we used antibodies against Flag (M2 monoclonal antibody, Sigma, #F3165), extracellular signal-regulated kinase (ERK)1/2 (Millipore, 06-182), p-ERK1/2 (Cell Signaling, #4370), p-serine (Cell Signaling, #2325), β-catenin (BD, #610153), CSN6 (Biomol International, #PW8295), β-Trcp (Cell Signaling, #4394), actin (Sigma, A2066), and Myc (sc-40, Santa Cruz Biotechnology). Experimental details and primer sequences are given in the Supplemental Experimental Procedures.

Ubiquitination Assay and Turnover Assay

Standard protocols were used as previously described(Choi et al., 2015; Gully et al., 2012); details are given in the Supplemental Experimental Procedures.

Cell Proliferation Assay, Wound Healing Assay, and Migration and Invasion Assay

Standard protocols were used; details are given in the Supplemental Experimental Procedures.

Luciferase Reporter Gene Assay

The transcriptional activation of β-catenin in DLD-1 cells was measured as described previously (Fang et al., 2007). Luciferase activity was assessed with the dual luciferase assay system (Promega, USA) according to the manufacturer’s instructions.

Xenograft Study

All animal work was conducted in accordance with American Association for Laboratory Animal Science regulations and approval of The University of Texas MD Anderson Cancer Center Institutional Animal Care & Use Committee. Male BALB/c nude mice, 4-6 weeks old, were obtained from The University of Texas MD Anderson Cancer Center. Mice were randomly divided into experimental groups. Mice were injected subcutaneously with 4×10⁶ DLD1 cells in a 50-μl volume using a 27-gauge needle. Five days after implantation, the mice were assigned to the PBS-only or PBS plus cetuximab group. The mice were treated three times per week with intraperitoneal injections of 1 mg of cetuximab in PBS or PBS alone.

For the shCSN6 study, mice were injected subcutaneously with 4×10⁶ AKP or shCSN6-AKP cells in a 50-μl volume using a 27-gauge needle. Tumor diameters were serially measured with calipers three times each week, and tumor volumes were calculated using the formula V=(L×W²)/2, where V=volume (in mm³), L=length (in mm), and W=width (in mm). Total RNA was extracted for qRT-PCR analysis. Sections of the harvested tissues were immunostained with the indicated antibodies.

Statistical Analysis

Chi square tests and one-way ANOVA were used to assess differences in clinical variables between the CRC cohorts. Kaplan-Meier survival analyses were used to compare survival times among CRC patients based on CSN6 expression; the log-rank test was used to generate p values. Cox proportional hazards regression analyses were used to assess the effect of clinical variables on patient survival. Univariate and multivariate analysis were used to assess the influence of clinical variables on survival. The p values and hazard ratios are indicated. Differences between groups were evaluated using a 2-tailed t test or a Mann-Whitney rank-sum test. All statistical analyses were performed using SPSS 16.0. Paired samples were compared using a paired t test.

Significance.

CSN6 is a biomarker that is elevated in colon cancer and leads to worse recurrence-free survival. CSN6, which is regulated through EGFR/ERK signaling, has unprecedented biological activity in downregulating β-Trcp, an E3 ligase for β-catenin, thereby stabilizing β-catenin; this confirms CSN6’s role in promoting colon cancer development. Our findings elucidate a mechanism by which β-catenin is regulated through the EGFR-ERK2-CSN6 axis during colorectal cancer development.

Highlights.

1. CSN6, a biomarker overexpressed in CRC, deregulated by EGFR/ERK signaling.

2. ERK2 binds directly to CSN6 and phosphorylates CSN6 at Ser148 to stabilize CSN6.

3. CSN6-mediated β-catenin stabilization involves β-Trcp downregulation.

4. Inhibition of ERK2 by cetuximab can inhibit CSN6 expression and tumor growth.