Nuclear expression of Glycogen synthase kinase-3β and lack of membranous β-catenin is correlated with poor survival in colon cancer


  • Tavga Salim,

    1. Cell and Experimental Pathology, Department of Laboratory Medicine, Lund University, Skåne University Hospital Malmö, Sweden
    Search for more papers by this author
  • Anita Sjölander,

    1. Cell and Experimental Pathology, Department of Laboratory Medicine, Lund University, Skåne University Hospital Malmö, Sweden
    Search for more papers by this author
    • Sjölander and sand-Dejmek contributed equally to this work.

  • Janna Sand-Dejmek

    Corresponding author
    1. Section of Surgery, Department of Clinical Sciences, Malmö, Lund University, Skåne University Hospital Malmö, Sweden
    • Cell and Experimental Pathology, Department of Laboratory Medicine, Lund University, Skåne University Hospital Malmö, Sweden
    Search for more papers by this author
    • Sjölander and sand-Dejmek contributed equally to this work.

Correspondence to: Dr. Janna Sand-Dejmek, Cell and Experimental Pathology, Clinical Research Center, Jan Waldenströms gata 35, 205 02 Malmö, Sweden, Tel: +4640391154, Fax: +4640391177, E-mail:


Dysregulation of Wnt/β-catenin signaling is a hallmark of colon cancer. Glycogen synthase kinase-3β (GSK-3β) can be a positive regulator of survival and proliferation of cultured colon cancer cell but its role in clinical colon cancer is unknown. Our objectives were to evaluate the role of GSK-3β in colon cancer. A tumor tissue microarray of primary colon cancers and metastases was used to evaluate expression and subcellular localization of GSK-3β and β-catenin. In total, 85 primary colon cancer samples were evaluated by immunohistochemistry. Immunoreactivity was correlated to known markers of adverse prognosis. Overall survival was the primary end-point. We found nuclear accumulation of GSK-3β in 39% (33/85) of evaluated tumors. Nuclear GSK-3β was significantly associated with shorter overall survival (p = 0.008), larger tumor size (p = 0.015), distant metastasis (p = 0.029) and loss of membranous β-catenin (p = 0.007). Loss of membranous β-catenin occurred in 37% (30/82) of the tumors and was associated with poor survival (p = 0.016). The combination of nuclear GSK-3β and lack of membrane β-catenin occurred in a total of 26% of the studied tumors (21/61) and was significantly and independently associated with poor prognosis. Our results suggest that nuclear expression of GSK-3β and loss of membrane β-catenin identify a subset of colon carcinomas with worse prognosis.

Colorectal cancer (CRC) is the third most common cancer worldwide, with over one million cases occurring every year. The mortality rate for CRC is approximately half of its global incidence. Five-year survival estimate for CRC exceeds 50%, but is highly variable depending on the stage of the disease.[1] The etiological factors and pathogenic mechanisms underlying the development of CRC are complex and heterogeneous and include inflammation, with inflammatory bowel disease being associated with an increased risk of CRC; dietary and life style factors, such as diets rich in red meat and unsaturated fat, excessive alcohol consumption and reduced physical activity.[2] Further, it is estimated that 15–30% of patients have a major hereditary component, only a quarter of which can be attributed to hereditary nonpolyposis colorectal cancer or familial adenomatous polyposis.[3]

Wnt signaling is one of the key signaling pathways controlling proliferation, differentiation and morphogenesis of cells during development. In all, >90% of CRCs have a mutation in a key regulatory factor of the Wnt/β-catenin pathway, most often in APC or β-catenin, resulting in activation of the pathway.[4] Wnt signaling is initiated by the binding of Wnt family members to a receptor complex consisting of the Frizzled family of transmembrane receptors, together with the coreceptors LRP5/6. Wnt signaling inactivates GSK-3β and prevents it from phosphorylating β-catenin, and thus stabilizing β-catenin in the cytoplasm. As β-catenin accumulates, it translocates into the nucleus where it binds to T-cell-specific transcription factor (TCF)/LEF and increases transcription of proto-oncogenes such as c-myc and cyclin-D1.[5] In addition to its role in embryogenesis and malignant transformation of cells, β-catenin exists at the cell membrane in a complex with E-cadherin and α-catenin and has a role in cell–cell adhesion and cell polarization, disruption of which results in dissociation of the cells and is an important step in invasion and metastasis.[4] The role of β-catenin in the development of CRC is well documented but as a single biomarker, β-catenin does not appear to have any prognostic value.[6]

GSK-3 is a multifunctional serine/threonine kinase involved in the regulation of cell fate, protein synthesis, glycogen metabolism, cell mobility, proliferation and survival.[7] There are two mammalian GSK-3 isoforms; GSK-3α and GSK-3β. Dysregulation of GSK-3β has been implicated in the development of a number of human diseases such as diabetes, cardiovascular disease, some neurodegenerative diseases and bipolar disorder but also in tumorigenesis and cancer progression.[7] Studies indicate that GSK-3β can act both as tumor suppressor and as tumor promotor.[7] The protein is constitutively active in resting cells and undergoes a rapid and transient inhibition in response to a number of external signals. GSK-3β activity is regulated by site-specific phosphorylation with full kinase activity requiring phosphorylation at tyrosine (Tyr216): On the other hand, phosphorylation at serine (Ser9) inhibits GSK-3β activity. One of the most known substrates of GSK-3β is β-catenin. In the absence of Wnt signaling, GSK-3β phosphorylates β-catenin, leading to its ubiquitin-mediated degradation.[7] In addition to having its effect in the cytosol, there is evidence that GSK-3β can enter the nucleus and form a complex with β-catenin, thereby lowering the levels of β-catenin/TCF-dependent transcription.[8]

GSK-3β is also proposed to be involved in cancer cell metastasis with inhibition of GSK-3β promoting epithelial–mesenchymal transition (EMT), a prerequisite for tumor cell invasion and dissemination. Contrastingly though, some studies suggest that GSK-3β can promote tumorigenesis and cancer development. In pancreatic cancer cells, nuclear localization of GSK-3β was associated with cell dedifferentiation and inhibition of GSK-3β impaired NF-κB-mediated pancreatic cancer cell survival and proliferation in tumor xenografts.[9] Nuclear accumulation of GSK-3β was found in bladder cancer samples and associated with impaired survival.[10] In colon cancer, the total amount of GSK-3β was found to be higher in tumors than in samples from normal colon.[11] Depletion of nuclear GSK-3β or pharmacological inhibition of its kinase activity impaired survival and proliferation of cultured colon cancer cells.[12]

The potential dual role for GSK-3β is especially interesting in CRC, given that the Wnt/β-catenin signaling pathway is deregulated in a majority of tumors. Little is known about GSK-3β with respect to the protein's subcellular localization in colon cancer, nor has its role in survival in this cancer type been studied. Therefore, we here investigated the expression and prognostic role of GSK-3β in clinical colon cancer samples. As GSK-3β is a major regulator of β-catenin levels in the cells, we also evaluated β-catenin expression and subcellular localization.

Material and Methods


Patients were selected retrospectively and all patients who underwent surgery for colon cancer at Malmö University Hospital during the selected time period (1990) were included. No stratification or matching was done. All tumors with available slides or paraffin blocks were histopathologically re-evaluated on hematoxylin and eosin (H&E)-stained slides. Clinical, treatment and histopathology data were retrieved from patient charts and pathology records. All patients had a pathologically confirmed diagnosis of adenocarcinoma before surgery. Tumors were staged according to the UICC TNM classification of malignant tumors.[13] Tumor grade was determined according to Jass et al.[14] as high, medium or low. Information on vital status and cause of death was obtained from the Swedish Cause of Death Registry until December 31, 2000. Follow-up started at the time of diagnosis and ended at death, emigration or December 31, 2000, whichever came first. Median follow-up time was 4.58 years (range, 0.08–10.67) for the full cohort (n = 89) and 8.78 years (range, 0.33–10.67) for patients alive (n = 31). The primary endpoint of the study was overall survival (OS). The study was approved by the Ethical Committee at Lund University.

Tumor tissue microarray and immunohistochemistry

Cases without available paraffin blocks or with an insufficient amount of tumor material were excluded. Out of a total number of 89, 85 patients (95.5%) were suitable for tumor tissue microarray (TMA) construction. To confirm the diagnosis and histological grading, the archival formalin-fixed, paraffin-embedded (FFPE) samples were cut into 1 µm sections, dried, deparaffinized, rehydrated and stained with H&E. Two to three 1.5-mm tissue cores from each donor block were placed in a new paraffin block by using an automated Beecher Micro-Arrayer (Beecher Instruments, Sun Prarie, WI, USA). For the samples from distant metastases, whole sections of the tumors were evaluated by H&E to verify the pathological diagnosis. For TMAs as well as larger tumor sections, immunohistochemical stainings were performed as described previously.[15] Briefly, the paraffin array blocks were cut into 1 µm sections, which were pretreated as described previously and then placed in citrate buffer and heated for 2 × 10 min in a microwave oven. All immunohistochemical procedures were performed using a Dako automatic slide stainer (Dako, Glostrup, Denmark) according to the manufacturer's instructions. After immunostaining, all slides were counterstained with H&E. Immunoreactivity for GSK-3β and β-catenin was assessed by two investigators (Salim T and Sand-Dejmek J). Disagreement between the observers was <10%, and those cases were reviewed until an agreement was reached. Tumors were graded according to intensity and subcellular localization (membrane, cytoplasm and nucleus). Tumors were evaluated for intensity of staining (0 = negative, 1 = weak, 2 = intermediate and 3 = strong), percentage of staining cells (1 = 0–5%, 2 = 6–25%, 3 = 26–75% and 4 = 76–100%) and to subcellular localization (membrane, cytoplasm and nuclear). Tumors were considered positive for nuclear GSK-3β staining if more than 5% of cells exhibited nuclear expression and staining intensity was moderate or high. Normal colon exhibited weak to moderate cytoplasmic GSK-3β expression (Fig. 1a). In normal tissue, nuclear GSK-3β was expressed in <5% of cells and the staining intensity was weak (Fig. 1a). For membranous β-catenin, tumors were considered positive if >50% of the cells exhibited membranous expression of the protein and negative if the expression was below 50%. In reality, however, staining for membranous β-catenin was very homogenous, with a majority of tumors being either strongly positive for membranous β-catenin with close to 100% of the cells expressing β-catenin at the membrane, or completely negative, with <5% of cells exhibiting immunoreactivity for β-catenin at the cell membrane.

Figure 1.

(a–g) Subcellular localization of GSK-3β and β-catenin determined by immunoreactivity in representative sections of invasive colon carcinomas. (a) Moderate cytoplasmic expression of GSK-3β in normal colon mucosa. (b)Tumor with cytoplasmic expression of GSK-3β, (c) tumor with moderate homogenous nuclear expression of GSK-3β. (d) Tumor with strong heterogeneous nuclear expression of GSK-3β. (e) Tumor with membranous β-catenin expression. (f) Tumor with membranous β-catenin expression and cytoplasmic β-catenin expression, and (g) tumor with weak cytoplasmic expression of β-catenin (microscopy images; left hand panels: 10× magnification, right-hand panels: 40×). (h–j) Nuclear GSK-3β and lack of membrane β-catenin is associated with poor survival. Kaplan–Meier survival curves according to (h) nuclear GSK-3β (n = 82), (i) membranous β-catenin (n = 82), and (j) nuclear GSK-3β in combination with lack of membrane β-catenin (n = 82). OS is shown in months and differences between groups were assessed using log-rank testing.

A mouse monoclonal antibody against β-catenin (dilution 1:1000, BD Transduction Laboratories) and a rabbit polyclonal antibody against GSK-3β (dilution 1:50, Cell Signaling Technology, Danvers, MA, USA) were used. For Ki67, nuclear immunoreactivity was evaluated by counting Ki67-positive cells in a number of fields of view and estimating the percentage of positive cells in the tumor as a whole. For statistical evaluation, tumors were classified as low proliferation = 0–29% Ki67-positive nuclei, and high proliferation = >29% Ki67-positive nuclei.

Statistical analysis

OS was the primary endpoint. Comparisons of clinical data and tumor characteristics according to GSK-3β and β-catenin status were done by -test. Kaplan–Meier estimates were used to illustrate survival according to GSK-3β and β-catenin expression and the log-rank test to assess for equality of survival curves. Hazard ratios (HRs) were estimated using Cox proportional hazards model for OS in uni- and multivariate analyses. The power to detect a HR of 2.0 at the level of significance 0.05 was 77%. Calculations were performed using SPSS version 19.0 (SPSS, IBM, Armonk, NY, USA) All p-values corresponded to two-sided tests and values of ≤0.05 were considered significant.

Cell culture

Human colorectal adenocarcinoma cell lines SW480 (DMS ACC 313) and SW620 (ATCC CCL-227) were grown in RPMI medium (SW480) or Leibovitz's L-15 medium with 2 mM l-glutamine (SW620) supplemented with 10% fetal bovine serum, 100 U/mL penicillin and 100 µg/mL streptomycin at 37°C in a humidified atmosphere containing 5% CO2. Cells were regularly tested to ensure the absence of mycoplasma. For experiments, cells were used at a confluence of around 70%.

Cell fractionation, gel electrophoresis and immunoblotting

For the preparation of membrane and cytosolic fractions, SW480 and SW620 cells were washed with ice cold PBS (1×), covered with buffer A[16] and kept on ice for 30 min whereafter the cells were homogenized in a Dounce homogenizer and centrifuged for 10 min at 500g. The supernatant was centrifuged for 10 min at 10,000g and the resulting supernatant was separated into plasma membrane and cytosol by centrifugation at 200,000g for 1 hr. For the preparation of nuclear extracts, the Nuclear Extraction Kit from Millipore was used according to the manufacturer's instructions (Millipore, Billerica, MA, USA). Cellular fraction samples were solubilized by boiling in sample buffer and DTT for 10 min. The samples were run on a Mini-PROTEAN TGX Precast gel and were subjected to electrophoresis whereafter the separated proteins were transferred to polyvinylidene difluoride membranes, blocked with 3% BSA in 1× PBS for 1 hr at room temperature followed by overnight incubation with primary antibodies at 4°C. GSK-3β expression was detected by probing with the mouse monoclonal antibody at 1:1000 dilution. For β-catenin, the monoclonal antibody was used at a mouse monoclonal antibody (BD Transduction Laboratories, Franklin Lakes, NJ, USA) was used at a 1:1000 dilution. Secondary, horseradish peroxidase (HRP)-coupled, antibodies were from Dako. Immunoreactive proteins were detected using the Immobilon Western Chemiluminescent HRP Substrate (Millipore).


For immunofluorescence analysis, cells grown on 22-mm glass cover slips were fixed with 3.7% paraformaldehyde and permeabilized with 0.1% Triton X-100. Fixed cells were incubated anti-GSK-3β (1:500) or anti-β-catenin (1:500) antibodies for 1 hr at 37°C, and subsequently incubated for 45 min at 37°C with goat antirabbit and antimouse secondary antibodies conjugated to Alexa Fluor 488 or Alexa Fluor 594 fluorochromes (Molecular Probes, Eugene, OR, USA), respectively, at a 1:1000 dilution. DNA was visualized by 4′,6-diamidino-2-phenylindole (DAPI) staining (Sigma-Aldrich, Copenhagen, Denmark). Fluorochromes were visualized with a Nikon microscope and imaged with NIS-Elements AR software (Nikon, Tokyo, Japan).

Wound-healing assay

SW480 and SW620 cells were grown to confluence in cell-culture dishes. A wound was inflicted on the monolayer using a pipette tip. Cells were serum starved for 2 hr and subsequently incubated in medium containing 20% fetal bovine serum. The cells were allowed to migrate for 24 hr at 37°C. Pictures were taken at 0, 18 and 24 hr with a Nikon phase contrast (DS-Fi1) microscope using a 10× objective and NIS-Elements Basic Research software. The area of the wound was measured both at the beginning and after 18 and 24 hr with Adobe Photoshop CS4 software.[17]


Expression of GSK-3β and β-catenin in colon tumors

In our study, evaluation of GSK-3β and β-catenin expression was possible for 85 and 82 cases, respectively. GSK-3β was expressed in the cytoplasm of all tumors with no tumors exhibiting membranous staining. Thirty-three (39%) tumors showed moderate or strong nuclear staining for GSK-3β. In the remaining 52 (61%) of tumors, GSK-3β was absent from or very weakly expressed in the nucleus. For β-catenin, cytoplasmic, membranous and/or nuclear immunereactivity was found. Membrane β-catenin was present in 39% of tumors (32/82). Nuclear β-catenin was found in 45% (37/82) of tumors. Figure 1 shows representative images of normal colon mucosa (Fig.1a) and tumors with different subcellular expressions of GSK-3β (Figs. 1b1d) and β-catenin (Figs. 1e−1g). For 20 patients, a corresponding sample from normal colon mucosa was included in the TMA. Normal colon mucosa samples expressed various levels of GSK-3β in the cytoplasm but none exhibited any nuclear expression of the protein. Figure 1a shows cytoplasmic expression of GSK-3β in normal colon mucosa (Fig. 1a). Normal colon mucosa did not express nuclear GSK-3β staining.

Association of nuclear GSK-3β and membranous β-catenin expressions with other clinico-pathological variables

Nuclear expression of GSK-3β was significantly correlated with the absence of membranous β-catenin (p = 0.007), larger tumor size (p = 0.015) and distant metastasis at the time of primary surgery (p = 0.029) (Supporting Information Table 1). Given the strong correlation between nuclear GSK-3β and lack of β-catenin at the cell membrane, we decided to compare patients with tumors with no membranous β-catenin in combination with nuclear GSK-3β with tumors not exhibiting this protein combination. Out of the 82 patients who were evaluated for both GSK-3β and β-catenin, the combination of nuclear GSK-3β and lack of β-catenin was found in 26% (21/82). When examining this subgroup, we found that those patients were significantly younger (average age at diagnosis 68.9 vs. 74.1 years, p = 0.031). Moreover, metastatic lymph node involvement was more common in this subset (p = 0.009). As for nuclear GSK-3β and membrane β-catenin individually, the presence of the combination was associated with larger tumor size and with distant metastasis at the time of diagnosis. No association was found with proliferation as estimated by Ki67 or with histological tumor grade (Table 1).

Table 1. Clinical and pathological variables in patients with primary tumors expressing the combination of nuclear GSK-3β and no membrane β-catenin versus tumors with no nuclear GSK-3β and/or expression of membranous β-catenin
 Nuclear GSK-3β and membrane β-cateninAll other tumorsp-Value
Characteristic(n = 21)(n = 61) 
Age at diagnosis (years)68.974.10.031
Tumor size   
T10 (0%)3 (5%) 
T20 (0%)8 (13%) 
T313 (86%)48 (79%) 
T48 (14%)2 (3%)<0.001
Missing0 (0%)0 (0%) 
Lymph node metastasis   
N06 (28%)40 (66%) 
N110 (48%)18 (29%) 
N23 (14%)1 (2%) 
N32 (10%)2 (3%)0.009
Missing0 (0%)0 (0%) 
Distant metastases   
M012 (57%)53 (87%) 
M19 (43%)8 (13%)0.006
Missing0 (0%)0 (0%) 
Male10 (53%)28 (47%) 
Female11 (47%)33 (53%)0.146
Missing0 (0%)0 (0%) 
Histological grade   
10 (0%)3 (5%) 
215 (71%)45 (74%) 
36 (29%)13 (21%)0.497
Missing0 (0%)0 (0%) 
Low7 (50%)19 (46%) 
High7 (50%)22 (54%)0.528
Missing7 (33%)20 (33%) 

Prognostic role of nuclear GSK-3β and membranous β-catenin expression

For survival analyses, dichotomized variables defined as low or absent staining versus moderate or strong staining were used for nuclear GSK-3β and for membranous β-catenin. As shown in Figure 1, nuclear expression of GSK-3β protein was associated with shorter OS by log-rank test (p = 0.008). For membranous β-catenin, the effect was the opposite, with lack of β-catenin expression at the membrane being associated with improved survival (p = 0.016). A univariate analysis according to the Cox proportional hazard regression model confirmed these results. Nuclear expression of GSK-3β in the primary tumor was associated with an increased risk of death (HR: 2.108; 95% confidence interval [CI]: 1.192–3.727). Membranous expression of β-catenin on the other hand was associated with lower risk (HR: 0.483; 95% CI: 0.263–0.889). As expected, larger tumor size, lymph node involvement and distant metastasis at the time of primary surgery were also associated with poorer survival.

Lack of membranous β-catenin in combination with nuclear GSK-3β expression is associated with poor prognosis

We found a strong correlation between nuclear expression of GSK-3β and lack of membranous β-catenin (p = 0.007). Therefore, we next investigated the effect of this protein combination on survival. In a univariate analysis, the risk of death was significantly increased (HR: 3.290; CI: 1.759–6.154) (Table 2 and Fig. 1). A multivariate analysis confirmed that nuclear GSK-3β in combination with lack of membrane β-catenin in the primary tumor is an independent factor associated with poor prognosis in colon cancer (HR: 1.989; CI: 1.016–3.894) (Table 2).

Table 2. Overall survival according to GSK-3β and β-catenin subcellular localizationa
 Univariate analysisMultivariate analysisb
VariableHR95% CIpHR95% CIp-Value
  1. a

    Cox uni- and multivariate analysis of overall survival in patients with tumors expressing nuclear GSK-3β but no membranous β-catenin compared to the whole cohort.

  2. b

    Multivariate analysis performed only for variables significant in the univariate analysis.

Nuclear GSK-3β + no membrane β-catenin3.291.759–6.154<0.0011.9891.016–3.8940.045
Yes versus no      
High versus low      
High versus low      
Male versus Female      
Tumor size4.9021.188–20.2260.0281.7730.409–7.6840.444
≤T2 versus >T2      
Lymph node status3.8382.095–7.031<0.0012.0490.959–4.3760.064
≥N1 versus N0      
Distant metastasis7.7413.899–15.367<0.0014.9142.239–10.781<0.001
M1 versus M0      

Differences in subcellular localization of GSK-3β and β-catenin in primary tumors versus metastases

In clinical samples, expressions of certain proteins have been shown to vary between primary tumors and metastases. An immunohistochemical study of primary colorectal tumors and the corresponding liver metastases demonstrated loss of membranous β-catenin expression in 26% of primary tumors and 60% of liver metastases. No correlation was found between β-catenin expression in primary tumors and metastases from the same patient.[18] Our aim was to investigate the expression of GSK-3β and β-catenin in metastasis from colon cancer. First, we evaluated the expression of GSK-3β and β-catenin in cultured colon cancer cells. To this end, we used the SW480 and SW620 cells, two cell lines derived from a primary Duke B colon cancer and a lymph node metastasis from the same patient. Although the SW480 cell line was established from the primary tumor in the colon, the SW620 cell line was isolated a year later, from a metastatic lymph node in the same patient when he experienced a massive intra-abdominal tumor recurrence.[19] Recently, SW480 and SW620 cells were analyzed using the iTRAQ approach and β-catenin levels were found to be decreased in the metastatic cells; the authors did, however, not further investigate cellular sublocalization of the protein.[20] We investigated untreated cells for their expression and localization of GSK-3β and β-catenin by immunofluorescence and found that SW480 colon cancer cells expressed GSK-3β in the cytosol as well as in the nucleus, whereas the metastasis-derived SW620 cells expressed GSK-3β in the cytosol and at the nuclear membrane (Fig. 2a). For β-catenin, SW480 colon cancer cells express the protein in the nucleus and the cytosol and at the cell membrane. In the metastatic cells, nuclear expression predominated with lower membrane as well as cytosolic expression (Fig. 2b). To confirm our results, we also evaluated the expression of the above-mentioned proteins by Western blot. To this end, nuclear, membrane and cytosolic fractions were isolated from SW480 and SW620 cells. Protein levels of GSK-3β did not differ between primary tumor cells and metastases but the primary tumor cells exhibit significantly higher levels of inactive, phosphorylated GSK-3β in the cytosol as well as in the nucleus (Fig. 2c). For β-catenin, both membranous cytosolic and nuclear protein levels were higher in primary tumor cells compared to the metastatic ones (Fig. 2d). This is in contrast to some previous studies where no differences between subcellular localization of β-catenin in primary tumors and metastases were found.[21] Expression of E-cadherin at the membrane correlated well with that of β-catenin, with metastatic cells expressing very low levels of E-cadherin, confirming, as expected, the predominantly mesenchymal phenotype of those cells.

Figure 2.

Subcellular expression of GSK-3β, serine-9 phosphorylated GSK-3β and β-catenin in SW480 and SW620 cells assessed by immunofluorescence and Western blotting. Immunofluorescence for (a) GSK-3β and (b) β-catenin, in SW480 primary colon tumor cells and SW620 distant metastasis cells from the same patient. Western blots for (c) GSK-3β and serine-9 phosphorylated GSK-3β in cytoplasmic and nuclear fractions, and for (d) E-cadherin and β-catenin in membrane fractions and β-catenin in cytoplasmic and nuclear fractions of untreated SW480 and SW620 cells. β-actin and lamin B or alpha-tubulin was used as loading control.

To evaluate whether those results could be translated into a clinical setting, we retrieved FFPE samples from distant metastases from patients who already had their primary tumors evaluated for the expression of GSK-3β and β-catenin. A total of 22% (18/85) patients had distant metastases already at the time of primary surgery. For seven patients, FFPE samples from distant lung or liver metastases were available. Tumor samples were stained for GSK-3β and β-catenin and the subcellular protein expressions were evaluated and compared to those of the corresponding primary tumor. As shown in Figure 3, expression and localization of GSK-3β and β-catenin in metastases did not always correspond to that of the primary tumor (Fig. 3a). For GSK-3β, nuclear expression was lost in two cases, whereas for membrane β-catenin there was a trend toward re-expression of the protein in metastases (Fig. 3b). The apparent discrepancies between primary tumor/matched metastasis in cell lines and clinical samples could be owing to the fact that the cell line metastasis investigated is derived from a lymph node, whereas the clinical metastases samples were distant liver metastases. In contrast to metastases located in the liver, lymph node metastases represent a “in transit” tumor with a retained propensity for migratory behavior. Hence, it is not unlikely that SW620 cells represent tumor cells that have underwent EMT, whereas the clinical liver metastases have reverted back to an epithelial phenotype, and hence their re-expression of epithelial markers.

Figure 3.

Graphs showing expression of (a) nuclear GSK-3β, and (b) membrane β-catenin, in seven primary tumors and corresponding distant metastases from the same patient. Subcellular expression of the proteins was evaluated by immunohistochemical staining and scored by two independent observers.

Differences in migration rates between primary tumor cells and metastases

In order for a tumor cell to become metastatic, it needs to acquire the ability to invade. The capacity to migrate is required for this initial step toward the establishment of metastatic growth.[22] We evaluated the migratory capacity of SW480 and SW620 cells and found that the metastatic cell line exhibited a significantly faster migratory rate (Fig. 4).

Figure 4.

Wound-healing assay comparing migration rates in SW480 and SW620 cells. A scratch was made in a confluent layer of cells. (a) Images of wound closure at 0 and 24 hr in SW480 and SW620 cells. (b) Graph showing the differential migration capacity of the two cell lines. [Color figure can be viewed in the online issue, which is available at]


The aim of our study was to examine the role of nuclear expression of GSK-3β in colon cancer. Inactivation of the APC gene or activating mutations of β-catenin is reported in virtually all patients presenting with CRC and is believed to be the critical initiating step in malignant transformation.[4, 23] Interestingly, however, although most colon cancers have constitutively activating mutations of the Wnt pathway, such tumors often still reveal a certain degree of regulation of the pathway.[23] In addition to its role in targeting β-catenin for degradation and hence counteracting tumor progression induced by increase transcription of TCF/LEF target genes, GSK-3β has been demonstrated to have an oncogenic role in colon cancer.[9, 24] GSK-3β expression levels and kinase activities were markedly and significantly increased in colorectal adenocarcinomas and correlated with increased expression of β-catenin in the nucleus and cytoplasm of colorectal tumors.[11] Downregulation of GSK-3β by siRNA induced tumor cell apoptosis.[12] Moreover, experimental studies suggest that inhibition of GSK-3β activity enhances the β-catenin/E-cadherin-mediated adhesion, presumably through increased expression of β-catenin at the membrane.[25] On the other hand, nuclear GSK-3β has been shown to participate in the degradation of β-catenin and hence, in all logic, rather should have a tumor suppressive role.[8] In cells, β-catenin has both a signaling function, conferred by a soluble cytoplasmic pool that is unstable in the absence of a Wnt signal, and an adhesion function based on a cadherin-bound, stable pool of β-catenin at the membrane. EMT of tumor cells results in the redistribution of β-catenin from the membrane to the cytoplasm.[26] In line with this, loss of membrane β-catenin has been shown to correlate with poor prognosis in CRC.[27] Our results corroborate those results. The strong association with nuclear GSK-3β was, however, somewhat unexpected, given that GSK-3β is known to negatively regulate β-catenin levels in colon cancer and that an experimental study showed that GSK-3β can enter the nucleus and form complexes with β-catenin, leading to decreased transcription of TCF/LEF target genes.[8] A recent publication, however, demonstrated increased levels of functional GSK-3β in CRC, suggesting that GSK-3β can have a tumor-promoting function.[11] In bladder cancer, nuclear accumulation of GSK-3β was associated with metastasis and worse outcome.[10] It is possible that GSK-3β promotes tumor cell survival as inhibition of GSK-3β activity leads to apoptosis and when GSK-3β was depleted through a siRNA approach, cell viability was reduced[10]. It is not unlikely that nuclear translocation of GSK-3β has a similar function in colon cancer.

Our data indicate that although nuclear GSK-3β is expressed in primary tumors as well as metastases, a larger proportion of the protein is phosphorylated and hence inactive in the primary tumors. Assuming that the function of GSK-3β in the nucleus is equal to that of its cytoplasmic counterpart, this finding offers a clue to the mechanism behind an increased propensity for metastases in nuclear GSK-3β expressing tumors. Moreover, we found that tumor cells with high expression of active GSK-3β in the nucleus in combination with low levels of membranous β-catenin migrated significantly faster compared to their low-active GSK-3β/high-membrane β-catenin counterparts.

Our patient data suggest that analysis of the combination of nuclear GSK-3β expression and lack of membrane β-catenin expression in the primary tumor may be useful to predict prognosis in colon cancer. We performed our study on a small number of patients but nonetheless found that loss of β-catenin expression at the membrane together with nuclear expression of GSK-3β was independently associated with poor prognosis in colon cancer.

Taken together, our results indicate that nuclear GSK-3β could be a potential new target in the development of therapies for CRC. Hence, we believe that our results are very promising and will aim at validating our findings in a larger study.


The authors thank Elise Nilsson for excellent technical assistance. The study was supported by grants from Malmö University Hospital Cancer Foundation, Percy Falk Foundation (J.S.D.) and from Swedish Cancer Foundation, Swedish Research Council, Gunnar Nilsson's Cancer Foundation, Skåne University Hospital Research Foundations, and by Governmental Funding of Clinical Research within the national health services (A.S.). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.