Article Text

KITENIN recruits Dishevelled/PKCδ to form a functional complex and controls the migration and invasiveness of colorectal cancer cells
  1. D H Kho1,
  2. J A Bae1,
  3. J H Lee1,
  4. H J Cho1,
  5. S H Cho2,
  6. J H Lee3,
  7. Y-W Seo4,
  8. K Y Ahn1,
  9. I J Chung2,
  10. K K Kim1
  1. 1
    Medical Research Center for Gene Regulation, Chonnam National University Medical School, Kwangju, South Korea
  2. 2
    Department of Hematology-Oncology, Chonnam National University Medical School, Kwangju, South Korea
  3. 3
    Department of Pathology, Chonnam National University Medical School, Kwangju, South Korea
  4. 4
    Korea Basic Science Institute, Kwangju Center, Kwangju, South Korea
  1. Professor K K Kim, Department of Pharmacology, Chonnam National University Medical School, Hak-Dong 5, Dong-Ku, Kwangju 501-190, South Korea; kimkk{at}chonnam.ac.kr

Abstract

Background and aims: KITENIN was previously reported to promote metastasis in mouse colon tumour models; however, the signalling mechanism of KITENIN at the cellular level was unknown. Here the functional role of KITENIN with respect to colorectal cancer (CRC) cell invasion and its expression in CRC tissues were investigated.

Methods: The effect of KITENIN on cell motility was analysed in a migration and invasion assay upon its overexpression and knockdown. Immunoprecipitation was used to elucidate binding partners, and immunohistochemistry was used to study expression levels.

Results: KITENIN overexpression enhanced the migration of rat intestinal epithelial cells, whereas a loss of invasiveness was observed in CRC cells after KITENIN knockdown. Mechanically, KITENIN served as a scaffolding molecule that simultaneously recruited both Dishevelled (Dvl) and protein kinase Cδ (PKCδ) through the membrane-spanning C-terminal region to form a complex that stimulated extracellular signal-regulated kinase (ERK)/activating protein-1 (AP-1) via a PKCδ component but also organised the actin filament via a Dvl component. The KITENIN complex controlled the invasiveness of CRC cells aetiologically harbouring various mutations in APC, β-catenin or K-ras, in which AP-1 activation is redundant but the organisation of the actin filament is indispensable for cell motility. Clinically, KITENIN expression was significantly higher in colon cancer tissues from advanced stage (III, IV) than that of stage I CRC and also in corresponding metastatic tissues.

Conclusions: The functional KITENIN complex acts as an executor with regard to cell motility and thereby controls CRC cell invasion, which may contribute to promoting metastasis.

View Full Text

Statistics from Altmetric.com

Aberrant canonical Wnt signalling contributes to early progression in 90% of colorectal cancers (CRCs). This aberrant signalling derives from a loss-of-function mutation of adenomatous polyposis coli (APC) and sometimes of axin, and from a mutation of β-catenin that creates a constitutively stabilised form.1 Non-canonical Wnt signalling contains distinct pathways,2 3 one that antagonises the canonical Wnt pathway and one that activates the c-Jun N-terminal kinase (JNK)/activating protein-1 (AP-1) pathway, and has been implicated in malignant transformation and tumour progression, such as Wnt5a in melanoma4 and breast cancer.5 Dishevelled (Dvl) protein is considered to act at the intersection of Wnt signal traffic.6 Dvl organises pathway-specific subcellular signalling through distinct domains: the DIX domain is essential for the canonical Wnt pathway, whereas the C-terminal DEP domain is required for activation of JNK and membrane translocation of Dvl-1 upon Wnt stimulation.7 8 The specificity of Dvl transduction depends on its binding partners, sublocalisation and phosphorylation.9 10

The planar cell polarity (PCP) cascade has been implicated in cell motility, given that PCP signalling leads to regulation of the actin cytoskeleton and AP-1-dependent transcription.4 5 In the PCP pathway, Dvl controls cell protrusions during directed migration of individual cells by regulating Rho and Rac, which modulate the cytoskeleton. That PCP proteins themselves localise to particular membranes during PCP signalling suggests that polarising events occur at epithelial plasma membranes and that membrane localisation of Dvl might be connected with its participation in the direct regulation of cell motility.11 Genetic loss-of-function studies revealed that Vangl2 (trilobite), a tetraspanin protein, crucially regulates PCP in flies and convergent extension movements in vertebrates. Vangl1 has a different expression pattern from Vangl2 but has a similar function, judging from restoration of zebrafish trilobite mutants by Vangl1.12 13 Both Vangl1 and Vangl2 bind all three Dvl family members via the C-terminal half of each14; however, the meaning of their interaction in the PCP pathway is not fully understood at the cellular level.

Previously, we characterised a cDNA clone of Vangl1 that interacts specifically with the C-terminal domain of KAI1, a metastasis suppressor gene, and renamed it KAI1 C-terminal-interacting tetraspanin (KITENIN). We observed metastasis-suppressive and tumour-regressive effects of small interfering RNA (siRNA)-KITENIN when given intravenously in a mouse colon tumour model.15 16 Here, we address the downstream signalling mechanism of KITENIN, a metastasis-promoting protein, and focus on elucidating the meaning of the interaction of KITENIN with Dvl in terms of regulation of CRC cell invasion and whether KITENIN expression is upregulated in primary tumours and their corresponding metastatic tissues according to the stage of CRC.

MATERIALS AND METHODS

Plasmids

cDNAs for deletion constructs of Flag- or haemagglutinin (HA)-tagged human KITENIN, HA-tagged mouse Dvl1, Flag-tagged mouse protein kinase Cδ (PKCδ), DN-PKCδ, DN-c-fos and AP-1-luc were generated by PCR-based methods and were confirmed by sequencing.

Cell culture

Human CRC cell lines (HT-29, DLD1, HCT116, SW480 and Caco2), rat intestinal epithelial (RIE) cells and 293T cells were obtained from the American Type Culture Collection and were grown as described.15 For the generation of stable cells, electroporation was performed according to the manufacturer’s instructions. Selected clones expressing each construct were obtained in the presence of antibiotics and were assayed. The sequences of all oligonucleotide primers and siRNA are given in table 1. The siRNA duplexes were prepared and transfected by using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, California, USA) according to the manufacturer’s instructions.

Table 1 Primer and siRNA-specific sequences

Luciferase assay

For the reporter assay, 293T cells were plated at a density of 4×104 cells/well 24 h before transfection, and were transfected with 50 ng of the reporters (AP-1-luc), 1 ng of phRL-CMV (cytomegalovirus) and effector constructs (50–200 ng) by using FuGENE6 (Roche, Indianapolis, Indiana, USA). After incubation for 24 or 48 h, luciferase activity was measured with a Dual-Luciferase Reporter Assay system (Promega, Madison, Wisconsin, USA) and normalised to that of Renilla luciferase. Each experiment was carried out in triplicate. Fold induction of reporters by effectors was calculated in comparison with the control vector.

Cell invasion assay

Cell invasion was measured by using the Transwell migration apparatus as described.15 Briefly, cultured cells pretreated with pharmacological inhibitor (for 1 h) or siRNA (for 48 h) were loaded into the top of a 24-well invasion chamber assay plate. Conditioned Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum (FCS) or phorbol-12-myristate-13-acetate (PMA; 100 nM) was added to the bottom chamber as a chemoattractant. After 20 h of incubation, the cells were stained and analysed.

Immunohistochemistry

The Ethics Committee of Chonnam National University Hospital approved our experimental protocols. Four and 80 colon cancer tissues were obtained for RNA preparations and immunohistochemistry, respectively, from surgically resected specimens at Chonnam National University Hospital, Kwangju, Korea. The tumours were histologically examined, and the pathological stage was estimated by the TNM score. Patients (stage I, n = 20; stage II, n = 20; stage III, n = 20; stage IV, n = 20) were randomly selected for this study according to TNM score. The tissue sections were deparaffinised, rehydrated, rinsed, stained with KITENIN, Dvl1, c-Jun and β-catenin, and examined as described.16

Assessment of KITENIN staining in CRC specimens

For semiquantification of immunohistochemical KITENIN expression, a score was attained by multiplying the intensity of the staining by the size of the area that stained positively. The intensity of cell staining appearing as a dark brown colour was graded according to the following scale: 0, no staining; 1+, mild staining; 2+, moderate staining; 3+, marked staining. The area of staining was evaluated on the following scale: 0, <5% of cells stained positive; 1+, 5–25% stained positive; 2+, 25–50% stained positive; 3+, >50% stained positive. The maximum combined score was 9, and the minimum was 0. Generally, specimens with scores >4 were considered as positive for a candidate protein expression, but this criterion was not used in this study.

RESULTS

KITENIN enhances migration and invasion of intestinal epithelial cells in an extracellular signal-regulated kinase (ERK)/AP-1-dependent manner

Previously, we found that KITENIN is implicated in cell adhesion and invasion in murine colon adenocarcinoma (CT-26) cells.15 To investigate the downstream signalling of KITENIN with respect to cell motility, we first obtained RIE cells stably overexpressing KITENIN (RIE/KITENIN) for use in a wound-healing assay. For most assays using stable cell lines, we used mixed polyclonal cells to exclude clonal variation. RIE/KITENIN cells showed enhanced migration compared with the vector control (fig 1A). To address which signalling pathways enabled RIE/KITENIN cells to migrate faster, we pretreated cells with PD98059 (a mitogen-activated protein kinase (MAPK) inhibitor) or wortmannin (a phosphatidylinositol 3-kinase (PI3 kinase) inhibitor) and assessed the migration rate in RIE/KITENIN cells. The MAPK and PI3 kinase signalling pathways are known to play a role in cell motility and invasion in CRCs.17 Both inhibitors reduced the migration rate in RIE/vector and RIE/KITENIN cells, but MAPK inhibitor only abolished the difference in the migration rate of RIE cells by KITENIN overexpression (fig 1A). Consistent with this findings, the level of p-ERK, a downstream signal of MAPK, was more elevated in RIE/KITENIN cells than in the control. Thus, the increased cell motility induced by KITENIN is dependent on MAPK/ERK signalling (fig. 1B).

Figure 1

KITENIN overexpression enhances cell migration via mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK)/activating protein-1 (AP-1) signalling in rat intestinal epithelial (RIE) cells. (A) RIE/KITENIN cells were subjected to a wound-healing assay (left). The effects of the signalling inhibitors PD98059 (50 μM) and wortmannin (0.1 μM) on cell migration are displayed as relative healing distances (right). Values are the mean (SE) for three independent experiments (n = 3). KIT·V5 represents the full-length KITENIN fused with the V5 epitope. (B) Phosphoprotein level of the MAPK and phosphatidylinositol 3-kinase pathways in RIE/KITENIN cells. (C) AP-1 reporter assay upon KITENIN overexpression in 293T cells. A-fos is dominant-negative c-Fos, an AP-1 transcriptional regulator. Psv40 and LacZ indicate SV40 promoter-luciferase as a negative control and cytomegalovirus β (CMVβ) as an effecter control, respectively (n = 3). (D) c-Jun and Fra-1 levels in RIE/KITENIN cells. (E) Analysis of the expression level of several AP-1 target genes by semiquantitative reverse transcription-PCR upon overexpression or downregulation of KITENIN. IB, immunoblotting; MMP, matrix metalloproteinase.

Downstream of the MAPK pathway, the AP-1 transcription factor, which is composed of c-Jun, c-Fos and Fra-1, regulates the expression of matrix metalloproteinases (MMPs), adhesion and cytoskeletal regulators, which control cell motility and invasion.18 Thus, we tested whether KITENIN activates AP-1 transcription. A reporter assay in 293T cells showed that KITENIN elevated AP-1 activity more than did the control and that its activation was suppressed by A-fos, a dominant-negative c-Fos that lacks the N-terminal phosphorylation sites (fig 1C). Consistent with the above results, the expression of c-Jun and Fra-1, which are autoregulated by AP-1 activation, was upregulated upon KITENIN overexpression (fig 1D).

We also investigated the expression of several target genes of AP-1 in KITENIN-overexpressing and KITENIN knockdown cells. Although the expression pattern was a little different depending on the cellular context, the target genes of AP-1 were elevated in KITENIN-overexpressing RIE and 293T cells, but were downregulated in CT-26 cells stably transfected with antisense KITENIN (fig 1E).

AP-1 activation by KITENIN is stimulated by a membrane-associated signal

To identify which region of KITENIN is responsible for AP-1 activation, we constructed several deletion mutants (fig 2A) and co-transfected these with the AP-1 reporter into 293T cells. The mutant with partial deletion of the N-terminal region (amino acids (aa) 79–524) or with deletion of the C-terminal PDZ-binding motif (ΔPDZ-bm) exhibited activation of AP-1 (fig 2B), indicating that KITENIN-induced AP-1 activity requires an intact membrane-spanning cytoplasmic region.

Figure 2

KITENIN activates mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK)/activating protein-1 (AP-1) signalling through interaction with Dishevelled (Dvl). (A) Schematic representation of KITENIN deletion mutants. ·V5 or ·flag indicate tag epitopes. (B) Reporter assay to identify which domain of KITENIN is responsible for AP-1 activation in 293T cells (n = 3). (C) Effect of KITENIN mutants (1–259, 1–465) on wild-type KITENIN-induced activation of AP-1 reporter and p-ERK (extracellular signal-regulated kinase) in 293T cells (n = 3). (D) AP-1 reporter activity and p-ERK in 293T cells transfected with each KITENIN mutant in which the second and third transmembrane domains were deleted but harboured the extracellular region (left side (L), right side (R), and left–right side fusion (LR)) (n = 3). (E) AP-1 reporter activity, p-ERK and p-JNK (c-Jun N-terminal kinase) in 293T cells overexpressing KITENIN under knockdown of all Dvl subtypes (n = 3). siRNA-con represents scrambled small interfering RNA. (F) Identification of the KITENIN region interacting with Dvl1 via immunoprecipitation (IP) in 293T cells, which were transiently transfected with KITENIN mutant constructs and full-length Dvl1. Arrows indicate the interacting band. HA, haemagglutinin; IB, immunoblotting; IgG, immunoglobulin G.

Next, we investigated the possibility that KITENIN works directly as a receptor for the extracellular ligand to activate AP-1. We examined whether the C-terminal deleted mutants (aa 1–259 and 1–465) that did not activate AP-1 suppressed wild-type KITENIN (KITENINwt)-induced AP-1 activation and phenotypically suppressed cell migration in a dominant-negative manner. Co-transfection of these mutants with KITENINwt did not reduce the AP-1 reporter activity of KITENINwt (fig 2C). Moreover, with respect to cell motility, CT-26 cells stably transfected with each mutant (aa 1–259 and 1–465) were shown to be similar to control cells (Supplementary fig 1A), which suggests that these mutants do not exhibit a dominant-negative effect on endogenous KITENIN. To refine this issue, we designed mutants of the extracellular region of KITENIN in which the second and third transmembrane domains were deleted but which contained the extracellular region (left side, right side and left–right side fusion) of the tetraspanin structure (fig 2A). Interestingly, these mutants activated the AP-1 luciferase and increased the level of p-ERK similar to KITENINwt (fig 2D). Meanwhile, to study the localisation of these mutants, green fluorescent protein (GFP) fusion constructs were prepared. Whereas the aa 238–524–GFP construct was displayed diffusely in a whole cell, a similar membrane localisation pattern was observed in 293T cells harbouring these GFP–mutant fusions (KIT-L, KIT-R and KIT-LR) or aa 1–465–GFP compared with that of KITENINwt–GFP (Supplementary fig 1B). However, we could not exclude the possibility that stimulation is triggered through the first or fourth transdomain, which is a source for a physical interaction with candidate proteins.

On the basis of our two results showing the lack of a dominant-negative effect by the C-terminal deleted mutants and AP-1 activation despite changing the extracellular region of KITENIN, we proposed that KITENIN might not act as a receptor for a diffusible ligand and that KITENIN-induced AP-1 activation is stimulated by a membrane-associated signal through the membrane-spanning intact C-terminal cytoplasmic region harbouring aa 238–524.

KITENIN-induced AP-1 activation is mediated by Dvl interaction

To understand the junction between KITENIN and MAPK signalling, we considered the c-jun/AP-1 axis via interaction of vangl2 and Dvl during gastrulation.19 First, we tested whether Dvl is a downstream molecule responsible for the KITENIN/AP-1 axis. The experiments using siRNA targeting three Dvl isoforms (Dvls) showed that KITENIN/AP-1 reporter activity and the p-ERK level was suppressed by Dvls knockdown (fig 2E). Next, we used an immunoprecipitation (IP) assay to identify which region of KITENIN was responsible for binding to Dvl. Consistent with the region responsible for AP-1 activation, the membrane-spanning intact cytoplasmic region of KITENIN was required for the association with Dvl (fig 2F). Reciprocally, to confirm which region of Dvl interacted with KITENIN, we constructed various deletion mutants (Supplementary fig 1C). IP using lysates from 293T cells co-transfected with KITENIN and serial mutants of Dvl1 showed that aa 328–499 harbouring the DEP domain were responsible for the interaction with KITENIN (Supplementary fig 1D).

PKCδ is required for AP-1 activation by KITENIN, which recruits PKCδ/Dvl to form a functional tertiary complex

PKCδ is required for c-Jun activation through interaction with Dvl and translocation on the membrane in convergent extension movements.20 Moreover, PKCδ plays an oncogenic role through the Raf/MEK/MAPK signal cascade in some transformed cells.21 Because PKCδ may be one of the candidate kinases for KITENIN/Dvl-induced MAPK signalling, we constructed a dominant-negative mutant of PKCδ (PKCδ-DN) that harboured only a diacylglycerol (DAG)-binding domain (Supplementary fig 2A) and co-transfected this into 293T cells with KITENIN or a constitutively active form of Dvl1 (ΔPDZ-Dvl1, Supplementary fig 1E). Consequently, PKCδ-DN markedly suppressed the KITENIN- or ΔPDZ-Dvl1-induced AP-1 activation and the p-ERK level, indicating that PKCδ is a downstream effector of the KITENIN/Dvl complex (fig 3A).

Figure 3

Protein kinase Cδ (PKCδ) is recruited to a KITENIN/Dishevelled (Dvl) complex and induces mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK)/activating protein-1 (AP-1) activation. (A) The dominant-negative effects of PKCδ-ΔC (PKCδ-DN) on KITENIN- or ΔPDZ-Dvl1-induced p-ERK (extracellular signal-regulated kinase)/AP-1 luciferase activity (n = 3). (B) Immunoprecipitation (IP) to identify the Dvl domain interacting with PKCδ in 293T cells, which expressed both full-length Flag·PKCδ and haemagglutinin (HA)·Dvl1 (upper) or were co-transfected with deletion mutants of Dvl and Flag·PKCδ (lower). Arrows indicate the interacting band. (C) Immunoprecipitation to examine whether endogenous Dvls and PKCδ interact with KITENIN in 293T cells transfected with KITENIN·V5. (D) Immunoprecipitation to test whether KITENIN interacts with PKCδ under Dvls knockdown. (E) Immunoprecipitation to assess whether KITENIN interacts with Dvls under PKCδ knockdown. (F) The effect of KITENIN knockdown on the expression of AP-1 targets in Caco2 and DLD1 cells (n = 3). AP-1 reporter activity (left) and semiquantitative reverse transcriptase-PCR analysis (right) are shown. COX-2, cyclo-oxygenase-2; IB, immunoblotting; IgG, immunoglobulin G; MMP, matrix metalloproteinase; siRNA, small interfering RNA.

Next, to confirm the domain of PKCδ needed to induce AP-1 activity, we designed serial mutants (Supplementary fig 2A) including one lacking aa 1–145 (PKCδ-CA) on the basis of the DR144/145A substitution mutant, which escapes intramolecular suppression and is constitutively active in murine fibroblasts.22 As expected, AP-1 activity was increased by PKCδ-CA in a p-ERK-dependent fashion, but not by PKCδwt under the same conditions (Supplementary fig 2B). This functional difference between the two constructs implies that PKCδwt is in an inactive state caused by intramolecular suppression, which may occur at the region encompassing the N-terminal 145 aa. We then examined whether endogenous DAG is required for PKCδ-CA-induced p-ERK/AP-1 activity and found that PKCδ-DN suppressed PKCδ-CA-induced AP-1 activity (Supplementary fig 2C). Moreover, we observed that PKCδ-CA-induced p-ERK/AP-1 activity was not reduced by siRNA-KITENIN or siRNA-Dvls, supporting that PKCδ-CA already has the structural changes needed, as a downstream effector of KITENIN/Dvl (Supplementary fig. 2D). Consequently, we confirmed that PKCδ-induced AP-1 activation requires binding of DAG and structural changes of PKCδ to overcome intramolecular suppression.23

PMA is well characterised as being a functional analogue of DAG for activating the PKC family,21 and we tested whether PKCδ translocates to the membrane upon PMA stimulation in CRC cells just like in various other cells by using PKCδ–GFP.23 Observation of PKCδ translocation on the membrane (Supplementary fig 2E) prompted us to assess the possibility that PKCδ interacts with KITENIN. On the basis of AP-1 activation by ΔPDZ-Dvl1 in the absence of KITENIN (Supplementary fig 1F), we first tested whether PKCδ associates with Dvl rather than KITENIN. Intriguingly, IP with PKCδwt (fig 3B) and PKCδ-DN (data not shown) showed that PKCδ interacts with the DEP region of Dvl1, which is also necessary for AP-1 activation. These results indicate that Dvl-induced AP-1 activation involves direct PKCδ interaction. However, concerning PKCδ’s various roles,24 we did not exclude other mechanisms by which PKCδ escapes intramolecular suppression of its activation.

Of note, Dvl could interact with KITENIN (fig 2F) and PKCδ (fig 3B), which led us to investigate the formation of a tertiary complex among them. Importantly, IP with 293T cells overexpressing KITENIN revealed that KITENIN recruits endogenous PKCδ and Dvl homologues (Dvl1, 2, 3) (fig 3C). However, it raised the question of how each PKCδ or Dvl associates with KITENIN. IP experiments with a Dvls or PKCδ knockdown strategy showed that PKCδ associates with KITENIN in the presence of Dvls (fig 3D), but Dvls interact with KITENIN in the absence of PKCδ (fig 3E). Collectively, KITENIN-induced ERK/AP-1 activation involves direct interaction with Dvls and PKCδ through the formation of a tertiary functional complex. Based on our observation that KITENIN does not have a receptor function in cell motility, we propose that KITENIN acts as a scaffolding protein for Dvl/PKCδ interaction and that overexpressed KITENIN could provide the increased chance for Dvl/PKCδ interaction to make a functional complex.

KITENIN is redundant in AP-1 activation but is crucial for invasiveness of CRC cells

Given that accumulation of genetic alterations, including mutation in K-ras, p53 and TGFβ, contributes to CRC progression,25 the KITENIN/AP-1 axis would be predicted to extend to various CRC cells. In particular, concerning the Raf/ERK/AP-1 activation by the K-ras oncogenic mutation,26 Caco2 (K-raswt, APCmut) and DLD1 (K-rasmut, p53mut) cells were selected for KITENIN knockdown. The AP-1 reporter and expression of AP-1 target genes were downregulated in Caco2 cells but not in DLD1 cells under KITENIN knockdown (fig 3F). Similar results were also obtained in HCT116 and SW480 cells harbouring the K-ras mutation (data not shown). These data suggest that the KITENIN/AP-1 axis might work in CRC cells not harbouring the K-ras mutation.

Because PMA promotes cancer progression via a PKC/ERK/AP-1 activation21 and PKCδ is a component of the KITENIN complex, we speculated whether PMA is an upstream signal for formation of a functional KITENIN complex and then examined whether PMA-induced AP-1 activity was blocked by the absence of KITENIN or Dvl. As expected, the p-ERK level was induced in a time-dependent manner, and c-Jun and Fra-1 (Fosl1) expression was also elevated by PMA (fig 4A). However, PMA-induced AP-1 activation was not affected by the knockdown of KITENIN or Dvl (fig 4A, upper panel), and PMA overcame the downregulated c-Jun and Fosl1 levels in Caco2 cells under KITENIN or Dvls knockdown (fig 4A, lower panel). Thus, at this time, it is unclear whether PMA acts as an upstream signal to the functional KITENIN complex as the result of interference by other PKC family members for p-ERK/AP-1 activation.

Figure 4

The functional KITENIN/Dishevelled (Dvl)/protein kinase Cδ (PKCδ) complex regulates cell motility for invasiveness of colorectal cancer (CRC) cells. (A) Effect of KITENIN or Dvls knockdown on p-ERK (extracellular signal-regulated kinase) level in Caco2 cells, which were treated with small interfering RNA (siRNA) for 48 h and harvested after phorbol-12-myristate-13-acetate (PMA) treatment (100 nM) for 15 or 30 min. Immunoblotting (IB) (upper) and semiquantitative reverse transcription-PCR analysis (lower) are shown. (B) Invasion assay of Caco2 cells under KITENIN or Dvls knockdown. Caco2 cells transfected with siRNA were sequentially subjected to invasion assay with fetal calf serum (FCS; 10%) or PMA (100 nM) as a chemoattractant. Stained invading cells were counted and are represented as a bar graph among groups (mean (SE) n = 3). (C) Immunoprecipitation (IP) for elucidating the formation of an endogenous KITENIN complex upon PMA stimulation in Caco2 cells using anti-KITENIN antibody (left) or anti-Dvls antibody (Ab) under KITENIN knockdown (right). (D) AP-1 reporter activity in Caco2 cells stably expressing ΔPDZ-Dvl1, PKCδ-CA or K-ras (V12) upon KITENIN knockdown. (E) Invasion assay in stable Caco2 cells treated with siRNA (n = 3). HA, haemagglutinin.

Phenotypically, we examined whether each component of the functional KITENIN complex contributed to cell motility and invasion of Caco2 cells not harbouring the K-ras mutation, just like CT-26 cells.15 Similarly, we observed that knockdown of KITENIN or Dvls dramatically blocked cell invasion (fig 4B, left). Given that PMA overcame ERK/AP-1 activity in Caco2 cells under KITENIN knockdown (fig 4A), invasiveness under KITENIN or Dvls knockdown would be expected to be restored by PMA. To this end, Caco2 cells were pretreated with siRNAs and studied in the invasion assay with addition of PMA. However, PMA did not restore the invasive capacity of the cells targeted with siRNA-KITENIN or siRNA-Dvls (fig 4B, right). Moreover, we observed that the effect of PKCδ knockdown on epithelial and CRC cell invasion was similar to that of KITENIN or Dvls knockdown (Supplementary fig 3A,B), implying the role of each component comprising the KITENIN complex in cell invasion.

Here, we noticed that cell invasion with PMA addition was approximately four times that of control untreated Caco2 cells and was suppressed by knockdown of KITENIN, Dvls or PKCδ. This finding indicated that each component of the KITENIN complex was required individually as a downstream effector in cell invasion upon PMA stimuli. To gain further details, we addressed whether a KITENIN/Dvl/PKCδ complex would be required for PMA-induced CRC cell invasion. Interestingly, IP showed that Dvls and PKCδ, which interact with KITENIN, were increased by PMA stimuli (fig 4C, left). Moreover, Dvls did not interact with PKCδ under KITENIN knockdown, thus supporting the scaffolding role of KITENIN (fig 4C, right). Therefore, PMA-induced elevation of CRC cell invasion seems to require the functional KITENIN complex, which supports our proposition that KITENIN-induced AP-1 activation might be mediated by a membrane-associated signal.

Suppressed CRC cell invasion under KITENIN knockdown is partially restored by ΔPDZ-Dvl1

To refine further the downstream events of KITENIN in CRC cell invasiveness, we approached genetically whether ΔPDZ-Dvl1 and PKCδ-CA, which constitutively activate ERK/AP-1, restore the loss of invasion in KITENIN knockdown CRC cells. Moreover, K-ras, which activates ERK/AP-1 and drives cell motility through divergent downstream pathways,26 was also tested. First, Caco2 cells stably transfected with ΔPDZ-Dvl1, PKCδ-CA or oncogenic K-ras (V12), respectively, were obtained as polyclonal cell lines. Sequentially, the invasion assay was performed in these cells under KITENIN knockdown. Notably, cell lines harbouring each constitutively active form exhibited more invasiveness, approximately twofold that of control cells, supporting that MAPK/AP-1 signalling contributes to cell invasiveness.18 Interestingly, even though AP-1 activity was restored altogether in these stable cell lines (fig 4D), only ΔPDZ-Dvl1 partially restored the invasiveness in the absence of KITENIN (fig 4E). A similar result was also observed when ΔDIX-Dvl1 was transiently transfected (data not shown). Unexpectedly, K-ras did not restore the invasiveness under KITENIN knockdown (fig 4E), which suggests that a functional KITENIN complex is also required as a downstream effector for invasion triggered by K-ras. Collectively, the restoration of invasiveness by constitutive Dvl under KITENIN knockdown supports phenotypically that Dvl is a downstream effector of KITENIN. The failure of restoration by PKCδ-CA indicates that PKCδ-CA did not affect Dvl activation. Thus, considering that loss of function of KITENIN epistatically masks the impact of AP-1 activity on cell invasiveness, this genetic approach is consistent with observation of stimulation experiments using PMA. Therefore, we suggest that a functional KITENIN/Dvl/PKCδ complex is another mechanism in physiologically modulating cell invasiveness, independently of the AP-1 axis.

KITENIN also contributes to the invasion of CRC cells harbouring the K-ras mutation

Because ectopic K-ras (V12) expression could not restore the invasiveness of Caco2 cells under KITENIN knockdown, we further assessed whether the invasion of CRC cells harbouring the K-ras mutation was affected by knockdown of KITENIN, Dvls or PKCδ. Similarly, siRNA-transfected DLD1 (fig 5A) and HCT116 (β-cateninmut, K-rasmut) (Supplementary fig 3C) cells exhibited a loss of invasion. Moreover, PMA did not affect the decreased cell invasion by knockdown of KITENIN, Dvls or PKCδ, indicating that each component of the functional KITENIN complex is also indispensable for invasion of CRC cells harbouring the K-ras mutation.

Figure 5

Each component of the functional KITENIN complex modulates the actin cytoskeleton and KITENIN is up-regulated in colorectal cancer (CRC) cells. (A) The effect of the knockdown of KITENIN, Dishevelled (Dvl) or protein kinase Cδ (PKCδ) on the invasion of DLD1 cells harbouring the K-ras mutation (n = 3). (B) The F-actin configuration using a confocal microscope in DLD1 cells upon knockdown of KITENIN, Dvl or PKCδ. Scale bar: 2 μm. (C) Protein level of KITENIN and nuclear β-catenin in various CRC cells and non-transformed cells (293T, rat intestinal epithelial (RIE)). (D) Semiquantitative reverse transcription-PCR analysis of KITENIN and several activating protein-1 (AP-1) targets in four human colon cancer tissues that contain a tumour region (T) and their corresponding adjacent mucosa (Adj). (E) Schematic illustrating the network of a KITENIN complex with regard to its relevant component and downstream event, which modulates the invasiveness of CRC cells. COX-2, cyclo-oxygenase-2; DAG, diacylglycerol; FCS, fetal calf serum; MMP, matrix metelloproteinase; PMA, phorbol-12-myristate-13-acetate; Rb, retinoblasoma protein; siRNA, small interfering RNA.

Each component of the KITENIN complex organises the actin filament system in CRC cells

Dvl regulates cell motility by dynamically modulating the actin cytoskeleton through Daam/RhoA and Rac during convergent extension.11 Because we observed distorted actin filaments in CT-26 cells with KITENIN ablation,16 we wanted to test whether the organisation of the actin cytoskeleton participated as another route of modulating cell invasion by a KITENIN complex, independently of the AP-1 axis. We treated DLD1 (fig 5B) and Caco2 (data not shown) cells with siRNA-KITENIN, siRNA-Dvl or siRNA-PKCδ for 48 h and observed the F-actin configuration by confocal microscopy. In scrambled siRNA-treated DLD1 cells, the actin filament bundles were well organised in a radial pattern and showed a continuous filamentous shape (fig 5B, upper). However, the arrangement of filamentous actin was disrupted in all experimental siRNA-treated DLD1 cells (fig 5B, middle and lower), indicating that all components of the KITENIN complex participate in organising the actin filament system. Although it is not fully understood how the KITENIN complex dynamically organises the actin cytoskeleton in cell motility, we suggest that the functional KITENIN complex controls the invasiveness of human CRC cells aetiologically harbouring various mutations in APC, β-catenin or K-ras by organising the actin cytoskeleton system.

The KITENIN/AP-1 axis is upregulated in human colon cancer tissues and metastatic tissues

Because the expression level of KITENIN rather than that of Dvl or PKCδ determines the formation of the functional KITENIN complex and thereby increases AP-1 activity and invasiveness, we first examined the expression level of KITENIN in various CRC cells. We found a higher level of endogenous KITENIN and nuclear β-catenin in human CRC cells than in RIE cells (fig 5C). Similarly, in sporadic colon cancer tissues from four patients, we also found more elevated KITENIN and several AP-1 targets in the tumour mucosa than in adjacent normal mucosa (fig 5D).

Next, using immunohistochemistry, we examined the expression patterns of KITENIN, c-Jun and β-catenin in serial colon cancer tissue sections from stage I to IV CRC patients, but also in metastatic tissues from stage III and IV CRC patients. Compared with those from stage I (fig 6A) or stage II CRC (fig 6B), KITENIN, c-Jun and β-catenin were highly expressed in the colon tissues from stage III (fig 6C) and IV CRC (fig 6E), but also in a metastatic lymph node (stage III, fig 6D) and a liver nodule (stage IV, fig 6F). Moreover, in the marginal portion of the metastatic lymph node (fig 6G, upper) and liver nodule (fig 6G, lower), KITENIN-overexpressing cells roughly overlapped with nuclear β-catenin, indicating that the expression level of KITENIN is associated with local invasive function of CRC cells.

Figure 6

KITENIN is upregulated in advanced colon adenocarcinomas and their corresponding metastatic tissues. The expression of KITENIN, c-Jun and β-catenin was examined by immunohistochemistry (dark brown colour) in serially sectioned human colon tissues from stage I (A), II (B), III (C) to IV (E) colorectal cancer (CRC) patients and in metastatic tissues from stage III (D) and IV (F) CRC patients. A high-power view (A–F, inset) marked by a circle in (A–F) shows increased KITENIN expression in stage III and IV CRC tissue and in a metastatic lymph node (stage III) and liver tissue (stage IV) compared with stage I CRC tissue. Scale bar: 200 μm in (A–F); 50 μm in (A–F, inset). (G) KITENIN-overexpressing cells roughly overlapped with nuclear β-catenin in the marginal portion of metastatic tissues from advanced CRC stages. A high-power field view of KITENIN and β-catenin expression in the marginal portion of a metastatic lymph node (upper) from stage III and a metastatic liver nodule (lower) from stage IV CRC marked by a square in (D and F). Scale bar: 100 μm.

To assess quantitatively the immunohistochemical KITENIN expression in the colon tissues among the CRC stages, a score was calculated by using the intensity of the staining and the size of the positively stained area. The KITENIN expression score was significantly greater in colon tissues from stage III (6.0±0.47), p<0.001) and stage IV (6.10±0.51, p<0.001) than in stage I CRC (3.95±0.32) (fig 7). However, there was no difference in KITENIN expression score between stage I and stage II (3.9±0.35), or between stage III and stage IV. These results indicate the existence of a positive correlation between the expression level of KITENIN and advanced stages of human CRC. Also, KITENIN expression in metastatic lymph nodes (6.0±0.45) and liver tissues (6.15±0.55) was comparable with that of colon tissues in stage IV CRC.

Figure 7

The positive correlation between the expression level of KITENIN and stage of human colorectal cancer (CRC). To assess the immunohistochemical level of KITENIN, a score was calculated as described in the Materials and methods section. LN, lymph node; NS, not significant.

DISCUSSION

Previously, we reported that KITENIN-overexpressing CT-26 mouse colon cancer cells show increased tumourigenicity and early hepatic metastasis in vivo.15 In the present study, we investigated the metastasis-promoting nature of KITENIN at the cellular level. There are two pieces of evidence for the importance of KITENIN in promoting CRC metastasis.

The first is that the functional KITENIN complex controls the invasion of human CRC cells as one of the modifiers for regulation of invasion. KITENIN is a scaffolding molecule for Dvl/PKCδ interaction at the membrane and is the upstream limiting protein that specifies Dvl to participate in the PCP pathway. Thus, formation of the functional KITENIN complex leads to PKCδ-mediated MAPK/AP-1 transcriptional activation and to Dvl-mediated organisation of the actin cytoskeleton system (fig 5E). These events increase the migration and invasion of CRC cells, which aetiologically have various mutations containing APC, β-catenin or K-ras.

Of note, KITENIN knockdown distorts the actin arrangement and suppresses the invasion of CRC cells, which is partially rescued by constitutively active Dvl. Because silencing Dvl2 in postdevelopmental endothelial cells results in aberrant cell membrane activity and actin disorganisation,27 and Dvl regulates convergent extension movement by activating the Rho family of GTPases,11 we suggest that formation of the functional KITENIN complex generates constitutively active Dvl, which regulates CRC cell motility. However, we cannot rule out the possibility that KITENIN may have another route in modulating CRC cell invasion independently of Dvl, in which vangl-associated proteins in the conserved PCP pathway, such as prickle, scribble and MAGI-3, might be involved.13

The second piece of evidence is that KITENIN expression is higher in cancer tissues of an advanced stage (III and IV) than in stage I or II CRC and also in metastatic tissues. The tumours diagnosed at an advanced CRC stage with regional (III) or distant liver metastasis (IV) showed elevated KITENIN expression, suggesting that KITENIN may predispose to the metastatic behaviour of CRC tumours. Because non-physiologically overexpressed KITENIN in 293T and RIE cells induced AP-1 activation and enhanced cell motility via the formation of a KITENIN/Dvl/PKCδ complex, this implies that a high level of expression of KITENIN in CRC cells provides an opportunity to form more KITENIN complex and subsequently further enhances the cell invasion, thereby leading to distant metastasis.

Dvl plays a dual role in the canonical β-catenin/transcription factor transcriptional pathway and the PCP pathway.9 Although the relationship between an aberrant β-catenin pathway, which is downstream of Dvl, and colorectal adenoma is well established,28 29 it remains unclear whether Dvl contributes to colorectal carcinogenesis through the PCP pathway. In this study, we found that Dvl participates in the regulation of CRC cell invasion as a component of the functional KITENIN complex in the PCP pathway. Because AP-1 is a critical regulator of gene expression that defines the invasive phenotype of cancer,18 and Dvl, one of the AP-1 modifiers, is overexpressed in human breast and lung cancers,30 31 we propose here the participation of Dvl via the PCP pathway in promoting CRC progression and in the invasion of various cancer cells.

Acknowledgments

We thank Song Eun Lee for assisting with immunohistochemistry. This work was supported by the Korea Science & Engineering Foundation through the Medical Research Center for Gene Regulation (R13-2002-013-04001-0) at Chonnam National University, and partly by a grant (0520280-1) from the National R&D Program for Cancer Control, Ministry of Health & Welfare, Republic of Korea.

REFERENCES

View Abstract

Supplementary materials

Footnotes

  • Competing interests: None.

  • Ethics approval: The Ethics Committee of Chonnam National University Hospital approved the experimental protocols.

  • ▸ Additional figures and methods are published online only at http://gut.bmj.com/content/vol58/issue4

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Digest
    Robin Spiller Magnus Simren